XI INTERNATIONAL SCIENTIFIC CONGRESS INNOVATIONS 2025 PDF Free Download
1 / 138/138
100%
XI INTERNATIONAL SCIENTIFIC CONGRESS
INNOVATIONS 2025
23-26.06.2025 VARNA
ISSN 2603-3771 (Online)
ISSN 2603-3763 (Print)
SCIENTIFIC-TECHNICAL UNION OF MECHANICAL ENGINEERING - INDUSTRY 4.0
BULGARIA
INNOVATIONS
2025
XI INTERNATIONAL SCIENTIFIC CONGRESS
INNOVATIONS
PROCEEDINGS
“ THEORETICAL PROBLEMS IN INNOVATIONS. INNOVATION
POLICY AND INNOVATION MANAGEMENT
”
“INNOVATIVE SOLUTIONS”
23 – 26 JUNE, 2025,
VARNA, BULGARIA
ISSN 2603-3763 (Print)
ISSN 2603-3771 (Online)
Year IX
Volume 1(9)
JUNE 2025
INTERNATIONAL EDITORIAL BOARD
CHAIRMAN:
Prof. Dimitar Damianov
Bulgaria
MEMBERS:
Aliaksandr Ilyushchanka, Cor. Member
Belarus
Assoc. Prof. PhD Anna Guzanova
Slovakia
Prof. Askar Kamerbaev
Kazakhstan
Prof. Atanas Kochov
North Macedonia
Prof. Azem Kyçyku
Kosovo
Assoc.Prof. Claudia Barile
Italy
Prof. David Gurgenidze
Georgia
Assoc. Prof. Dr. Despo Ktoridou
Cyprus
Assoc. Prof. Emilia Abadjieva
Bulgaria
Prof. Egils Dzelzitis
Latvia
Prof. Feruza Zakirova
Uzbekistan
Prof. Galina Alekseevna
Russia
Prof. Günay Anlas
Tukey
Prof. Igor Kuzo
Ukraine
Prof. Dr. Eng. Ivan Kuric
Slovakia
Prof. Ivan Lamin
Russia
Assoc.Prof. Jelena Jovanovic
Montenegro
Assoc.Prof. Jurate Cerneviciute
Lithuania
Prof. Marcin Golabczak
Poland
Prof. Predrag Dašić
Serbia
Prof. Raul Turmanidze
Georgia
Assoc. Prof. Sanjin Troha
Croatia
Prof. Silvio Macuta
Romania
Prof. Sreten Savićević
Montenegro
Prof. Stojance Nusev
North Macedonia
Prof. Uwe Füssel
Germany
Prof. Vahram Abrahamyan
Armenia
Assoc.Prof.Volodymyr Zhavoronkov
Ukraine
Prof. Wolfgang Drechsler
Estonia
Prof. Zvonimir Guzović
Croatia
C O N T E N T S
INNOVATION POLICY AND INNOVATION MANAGEMENT
Innovation Marketing and Strategic Marketing
Naqib Daneshjo, Renáta Ševčíková , Dominika Popovičová , Filip Ţiak ..................................................................................... 5
Application of Cybersecurity Awareness Training and End-User Behavioural Protection Simulation Platforms into
Higher Education
Dimitar Tashkov, Angel Valchev, Veselina Gagamova, Elitsa Pavlova, Violeta Vasileva .......................................................... 8
Киберсигурност по веригите за доставки на софтуер
Велиян Димитров, Георги Димитров, Александър Кирков, Радостин Николов, Николай Керин .................................... 11
Immersive Technologies as an Educational Innovation for Training the Next Generation of Professionals
Kaloyan Dimitrov ........................................................................................................................................................................ 15
Regarding cybersecurity in Bulgarian educational institutions at the K–12 level
Marieta Hristova, Diana Netova, Nikolay Netov ......................................................................................................................... 18
INNOVATIVE SOLUTIONS
The Evolution, Current Impact and Future of Artificial Intelligence in Medicine
Martin Čillík, Ivan Kuric, Miriam Kuricová ................................................................................................................................ 21
Photophysical properties of some phthalocyanine derivatives using ultrafast spectroscopy
Ionut Radu Tigoianu,Anton Airinei, Carmen Gherasim, Natalia Simionescu, Tamara Potlog, Victor Suman,
Ion Lungu, Giulio Cerullo, Stefano Dal Conte, Edoardo Carraro ............................................................................................... 27
Simulation of solar-coal hybrid power plant based on the Rankine cycle
Paweł Józef Madejski, Isyna Izzal Muna ..................................................................................................................................... 31
Comparative Study of Bayesian-Optimized 1-D CNN, Bi-LSTM and MLP for Bearing Fault Classification from Raw
Vibration Signals
Paweł Knap, Urszula Jachymczyk ............................................................................................................................................... 35
Correlation-Based Sensor Pruning and Malfunction Detection in Multi-Sensor Condition Monitoring
Urszula Jachymczyk, Paweł Knap ............................................................................................................................................... 39
Control System Concept for an Omnidirectional Mobile Platform: Modeling and Design Aspects
Patryk Bałazy, Julia Jeleńska ....................................................................................................................................................... 43
Engineering tool integration for complex system simulation and optimization
Szymon Podlasek, Urszula Jachymczyk ...................................................................................................................................... 48
Prototype of a Wireless MEMS-Based Sensor Node within a Wireless Sensor Network Concept
Julia Jeleńska ................................................................................................................................................................................ 52
Study of the effect of thermomechanical treatment, including preliminary heat treatment and radial shear rolling, on
changes of brass mechanical properties
Abdrakhman Naizabekov, Evgeniy Panin, Sergey Lezhnev, Pavel Tsyba .................................................................................. 55
New technology for ferrous metals bar scrap recycling to obtain a reinforced screw reinforcement profile
Sergey Lezhnev, Evgeniy Panin, Elena Shyraeva, Maxim Bogachev ......................................................................................... 58
Innovative technical solution for emergency repair of a pressure tunnel water supply pipeline during its air passage
over a river bed
Valeriy Naidenov ......................................................................................................................................................................... 62
Stress intensity factor in rods under tension with twin semielliptical cracks
Pejo Konjatić, Ana Konjatić, Mato Jakus, Ţeljko Blavicki ......................................................................................................... 66
Nanosized BaTiO3 powder prepared via mechanochemical activation
Mihaela Aleksandrova, Bojidar Jivov, Vladimir Petkov ............................................................................................................. 70
Research of the resistance to intergranular corrosion of 316L steel samples created on the method “Metal Injection
Molding” technology
Kalin Anastasov ........................................................................................................................................................................... 73
On application of generative artificial intelligence in higher education: insights from bulgarian students
Teodora Varbanova, Diana Netova, Nikolay Netov, Kamen Spasov .......................................................................................... 77
Systematic Approach to Design Space Exploration of Pulley Supports Using Generative Design
Tomislav Solar, Ivan Grgić, Mirko Karakašić, Ţeljko Ivandić ................................................................................................... 81
Effect of organosilicate application on thermo-pressure bonding of metals and composites with thermoplastic matrix
Anna Guzanová, Dagmar Draganovská, Nikita Veligotskyi ....................................................................................................... 87
Detection of fibre continuity in joints of metallic and composite thin-walled materials .formed by thermal drilling
technology by computed tomography
Anna Guzanová, Nikita Veligotskyi, Teodor Tóth ...................................................................................................................... 91
Statistical analysis of modification of joints of metals and composites by thermal drilling
Anna Guzanová, Nikita Veligotskyi, Gabriela Iţaríková ............................................................................................................ 95
Effect of thermal drilling strategy on the geometrical characteristics of metal – composite joints
Anna Guzanová, Nikita Veligotskyi ............................................................................................................................................ 99
Optimization of femtosecond laser parameters on surface morphology of lithium disilicate glass ceramic
Andreja Carek, Ljerka Slokar Benić, Hrvoje Skenderović, Lorna Martić ................................................................................. 103
Electrical and hydrodynamic conditions in electrolyte-plasma treatment of internal surfaces
Aleksandr Korolyov, Vyacheslav Tomilo, Vladimir Niss ......................................................................................................... 107
Investigation of the local permeability of filter materials by methods with the opposite direction of local air flows
Aliaksandr Ph. Ilyushchanka, Iryna M. Charniak, Aliaksei R. Kusin, Anastasia A. Astapenko, Ruslan A. Kusin .................. 112
Recent Innovations in Metal Processing with the Use of Lasers 2025
Nikolaos Papageorgiou .............................................................................................................................................................. 116
Антиброкариани на четириъгълник
Станислав Стефанов, Хаим Хаимов ....................................................................................................................................... 119
Други свойства на антиброкарианите
Станислав Стефанов, Хаим Хаимов ....................................................................................................................................... 127
Synthesis of carbon adsorbents for adsorption of chlorhexidine gluconate from aqueous solution
I.Stoycheva, B. Petrova, B. Tsyntsarski, A. Kosateva, M. Argirova, G. Tirolski, N. Petrov, P. Dolashka, M. Kalapsazova ... 135
Study of the porous properties of carbon adsorbents obtained from waste materials
B. Petrova, I. Stoycheva, B. Tsyntsarski, A. Kosateva, N. Petrov, P. Dolashka ....................................................................... 136
Synthesis and characterization of novel bio-char adsorbents
B. Tsyntsarski, I.Stoycheva, B. Petrova, A. Kosateva, N. Petrov, P. Dolashka, Teodor Sandu, Andrei Sarbu ........................ 137
Innovation Marketing and Strategic Marketing
Naqib Daneshjo1, Renáta Ševčíková2 , Dominika Popovičová3 , Filip Žiak4
University of Economics in Bratislava, Slovakia1, 2, 3, 4
naqib.daneshjo@euba.sk1, renata.sevcikova@euba.sk2, dominika.popovicova@euba.sk3, filip.ziak@euba.sk4
Abstract: This article examines the role of innovative and strategic marketing in enhancing the competitiveness of businesses in an
environment of rapid technological change. Particular attention is paid to a strategic approach to marketing planning, which involves a
thorough analysis of consumer needs and market assumptions. The article emphasizes the need for continuous adaptation of product
portfolios to changing market conditions and increasing fragmentation of customer segments. The concept of strategic innovation marketing
is introduced as a key factor for successful product positioning, while its implementation in business strategies is considered essential for
effective management of innovation processes. The final findings of the paper suggest that strategic innovation marketing not only promotes
customer retention, but also enables the transformation of business models and supports the long-term growth of firms in an environment of
rapidly evolving technologies.
Keywords: INNOVATION MARKETING, STRATEGIC MARKETING, INNOVATIONS, MARKETING MANAGEMENT
1. Introduction
In economic terms, innovation means the development and
implementation of ideas and technologies that lead to higher quality
goods and services or to their more efficient production or
provision. A typical example of innovation is the development of
steam engine technology in the 18th century, with the beginning of
the use of steam engines leading to the start of mass production.
More recently, information technology has changed the way
businesses produce their goods and services, opening up new
markets and introducing new business models. [1]
Product quality is clearly a determining factor for a company to
position itself on the market and win the largest number of
customers, which automatically implies a growth in market share
and room for making the expected profit. Marketing is a stabilising
factor in management, which enables enterprises to orient
themselves securely in the market. It is one of the means by which
an enterprise can successfully establish itself in the market.
Innovative marketing is a very important tool for an enterprise. In a
competitive environment, innovation of existing products and the
development of new products is a necessity in order to retain
customers and market share. Enterprises that do not spend on new
product development or invest in the innovation of machinery and
equipment and existing products expose themselves to the risk of
losing regular and potential customers and, inevitably, of losing
their market share. An enterprise should strive to obtain good
quality and reliable information on its own activities as well as on
the activities of the external environment, which provide a good
starting point for the enterprise to make the right decisions and to
operate successfully on the market. [2]
Marketing innovations are aimed at better meeting the needs of
consumers, opening new sales markets to increase sales volumes,
which are key to the successful development of the organization. In
order to keep up with the rapidly changing needs of the market and
to make the most of the opportunities that open up in the external
environment, organisations need to continuously work on new
products, technologies and relationships with the outside world. The
key to solving these problems is innovative marketing activity,
which in modern times has become the core of corporate
competitive strategies. [3]
The production of goods, the provision of services, the
provision of information and their compulsory sale in the respective
markets are the basis of the cultural and economic life of the people.
Services, goods and information generate utility, which economists
in their own circles call utility, which enables the consumer to
satisfy a certain desire. The commercial tone of the relationship
between producers and buyers of products is set by four types of
basic instruments: form, time, place and ownership. The
commercialisation of innovations requires the use of marketing
innovations. Therefore, issues related to the development of
marketing innovations and the creation of mechanisms for the
implementation of marketing innovations in order to increase the
potential and competitiveness of domestic enterprises are nowadays
particularly important in the post-industrial stage of the country's
economic development. [3, 4]
Strategic marketing involves the acquisition of theoretical
knowledge by undergraduates and practical skills in the
development of marketing strategies for manufacturers, taking into
account the opportunities and threats existing in the external
environment, as well as the available skills, capabilities and
resources that contribute to the creation of a unique market offering
designed for the most attractive market segments. [3] Innovation is
becoming an integral part of the modern economy. The process of
economic and political transformations taking place in the country
has led to the need for significant changes in the activities of
economic actors. Existing organisational structures, interests and,
with them, behavioural mechanisms and ways of decision-making
are changing. The conditions of modernity require increased
attention to the management of innovative activities that ensure the
efficient development of production. Innovative activity is the most
progressive form of entrepreneurship in market relations. The
market creates real opportunities for the development of scientific
and technological progress. [5]
2. The functioning of strategic innovation
marketing
The logic of development of an innovative company leads to a
shift of the focus from operational tactical planning to the strategic
level, to the level of forming a new type of management -
innovative marketing. Innovative marketing in the modern concept
represents the unity of strategies, business philosophy, functions
and proce-dures of company management. In industrialised
countries, the marketing concept of company development has been
important for decades. The concept of innovative marketing is the
basis for market research and the search for a competitive strategy
for an enterprise. The innovative marketing complex includes the
development of an innovative strategy, market analysis and
operational marketing. Figure 1 shows that all phases of innovation
marketing are related to strategic or operational components. [2]
Strategic innovation marketing is defined by market
segmentation and product positioning. The key point of the
marketing strategy is the study and forecasting of the demand for
the new product, based on a thorough study of consumer perception
of the innovation. During the strategic research, innovation project
leaders should define what products will be offered, in what quality
and to which consumers. [7] Strategic marketing therefore focuses
on close contact between the company's marketing and sociological
services staff and the potential consumer, through various tools such
as questionnaires, telephone surveys, representative samples, etc. A
general economic analysis is usually carried out in the initial market
research phase. This type of analysis is closely linked to the study
of the external environment of the enterprise and allows for the
examination of macroeconomic factors related to the demand for
INNOVATIONS 2025
5
innovation, including population, growth rates, per capita income
and consumption, the consumer price index, the consumer basket
and the inflation rate. This includes the study of the legal conditions
as well as the practice of legislation concerning the import and
export of like products, quotas, restrictions on standards,
obligations, taxes, subsidies, etc. At the same time, it is necessary to
analyse the level of national production of such products, the
presence or possibility of imports, the level of exports, data on the
production of import substituting products and innovation. [3, 5]
Fig. 1 Innovative marketing complex [1]
Fig. 2 Model of Formation of the Company Innovation Strategy for
Activation of Market Activity [8]
3. Functions of innovation marketing
Innovative marketing involves the use of creative approaches in
all areas of business, focusing on the constant search for ideas, their
implementation in order to improve business technology and create
competitive products. Modern researchers emphasize the possible
areas of innovative marketing within both traditional thinking and
non-standard practices. [1]
If the former focuses on finding ideas and creating products
within the goals, target markets, and opportunities defined by the
enterprise, then the latter involves search processes that are not
constrained; innovative ideas are superior to the enterprise's goals.
The second direction identifies the spin-off of innovation
departments in large companies and the emergence of venture
capital firms focused on the implementation of venture projects and
the creation of fundamentally new products and technologies. [9]
F. Kotler and F. Trias de Bez, based on the type of thinking,
proposed to distinguish the concepts of vertical and lateral
marketing in the innovation process. Vertical marketing is based on
logic and consistency of thinking. The concept of lateral thinking
was introduced by Edward de Bono and defined as "a set of
processes designed to use information in a way that generates
creative ideas through the ingenious restructuring of concepts
accumulated in memory". Characteristics of innovative marketing:
[10]
1. The strategic focus on finding and satisfying new needs
suggests that innovation marketing is used not only at the
"output" but also at the "input" of innovation management.
2. The organization and management of the innovative
activity of the enterprise is carried out through the prism of
interaction with the market, which includes the use of
network theory and the study of modern forms of relations
in the innovation market.
3. The object of investigation and the product on the market
is not a finished product, but an idea, which determines the
use of methods for the exploitation and evaluation of
intellectual property. Thus, the goal of innovation
marketing can be defined as the creation and
implementation of an innovation strategy for the
organization's activities, which includes increasing its
competitiveness.
Table 1. Functions of innovative marketing [6]
Feature
Table of Contents
Marketing
research
Research of innovation processes of
external environment and internal
innovation potential, research of innovation
potential of competitors, research of
potential markets of finished products,
research of consumers of innovations,
research of marketing mix possibilities in
different stages of innovation process.
Planning policy
on innovation,
trade and
assortment policy
Development of new products,
development of measures to modify
existing products (improvement of quality
characteristics and competitiveness),
development of the product range structure,
etc.
Sales and
distribution
Creation of a sales network, determination
of the sales structure at different stages of
the innovation process, control of physical
flows in the distribution system.
Promotion
The development of advertising policy at
different stages of the innovation process,
the development of the trademark, the
image of innovation, the development of
measures to change the image, the
formation of demand for innovation.
Pricing
Forecasting new product prices, developing
pricing strategies, analysing prices of
substitute products and similar products.
Marketing
Management
Organization of the marketing management
structure at different stages of the
innovation process, optimization of
managerial decisions in the system of
marketing activities, marketing audit.
Innovative marketing involves the use of creative approaches in
all areas of the enterprise, focusing on the constant search for ideas,
their implementation in order to improve enterprise technology and
create competitive products. Modern researchers emphasize the
possible areas of innovative marketing within traditional thinking
and non-standard (combinatorial) thinking. If the former focuses on
the search for ideas and the creation of products within the
framework of enterprise-defined goals, target markets and
opportunities, then the latter assumes unrestricted processes of
searching for innovative ideas and their primacy over the objectives
of enterprises. The second direction determines the allocation of
innovative units in large enterprises and the emergence of venture
capital firms. According to the theory of innovative marketing, the
INNOVATIONS 2025
6
process of perception of a new product consists of the following
stages: [6, 7]
1. Primary awareness. The consumer learns about the
innovation but does not have enough information.
2. Product recognition. The consumer already has some
information, is interested in the novelty; may seek further
information about the new product.
3. New product identification. The consumer compares the
novelty with his needs.
4. Evaluation of the possibilities of using the innovation.
The consumer decides to test the innovation.
5. Consumer endorsement of the innovation to obtain
information about the innovation and the possibility of
acquiring it.
6. The decision to acquire or invest in the creation of an
innovation.
The logic of development of an innovative company leads to a
shift of the focus from operational tactical planning to the strategic
level, to the level of forming a new type of management -
innovative marketing. Innovative marketing in the modern concept
represents the unity of strategies, business philosophy, functions
and procedures of company management. In industrialised
countries, the marketing concept of company development has been
at the forefront for decades. The concept of innovative marketing is
the basis for market research and the search for a competitive
strategy for the enterprise. The innovative marketing complex
includes the development of an innovative strategy, market analysis
and operational marketing. [7]
The main objective of strategic innovation marketing is to
develop a strategy for bringing innovations to the market. The basis
of strategic marketing research is therefore the analysis of the
market situation with the subsequent development of market
segments, the organisation and shaping of demand and the
modelling of purchasing behaviour. [4]
Strategic innovative marketing is determined by market
segmentation, product positioning. The key point of the marketing
strategy is the study and forecasting of the demand for the new
product, based on a thorough study of consumer perception of the
innovation. In the course of the strategic research, the leader of the
innovation project must determine: what products, in what quality
and for which consumers will be offered. Strategic marketing
therefore focuses on close contact between the marketing and
sociological services staff of the company and the consumer
(through questionnaires, telephone surveys, representative samples,
etc.). [8]
4. Conclusion
The rapid evolution of technology requires significant changes
in the product portfolio and continuous market analysis. Due to the
fragmentation of mature markets and the emergence of customer
groups with specific requirements, a market segmentation strategy
is only possible on the basis of detailed information. Strategic
marketing comprises two parts: strategic analysis and strategic
decisions. Conceptually, the concepts of brand and strategy are
intrinsically linked. A brand is essentially a strategic decision that is
followed across the corporate structure to achieve key objectives.
When developing a brand strategy, it must be taken into account
that the market structure is always dynamic, not static. If the brand
strategy does not take into account strategic adaptation to market
demands and has a strong connection in relation to the brand's
history and standard, then the company may find itself in a situation
where the brand is overwhelmed by the market and is outperformed
by other brands.
The process of innovation and strategic marketing is indeed a
source of information for the management of innovation activities
and innovative solutions that can translate the value of innovations
from product innovations conditioned by technological innovations,
innovations in the organizational structure of the enterprise and in
synergy with marketing innovations into changes in quantitative
performance indicators and performance in monetary and non-
monetary indicators.
5. References
1. I. V. Afonin. Innovation management: study guide. (2005).
2. K. Anderson. Customer-oriented management. (2003).
3. K. M. Christensen. Solving the problems of innovation in
business. How to create a growing business and
successfully support its growth. (2004).
4. E. P. Golubkov. Innovative marketing as a tool for moving
the Russian economy to a new path of development.
(2010).
5. Y. H. Gordon. Marketing partnerships. (2001).
6. M. Kováč, D. Sabadka, L. Kováčová. Fundamentals of
small business. (2003).
7. E. I. Krylov. Analysis of the effectiveness of investment
and innovation activities of the enterprise: Proc.
Allowance. (2003).
8. V. Melas. The importance of innovation in the 21st
Century: The importance of innovation in the 21st century.
(2016).
9. A. Hanlon. How to create a marketing plan in 2025.
(2025). https://www.smartinsights.com/marketing-
planning/create-a-marketing-plan/how-to-create-a-
marketing-plan/
10. P. Kotler, F. Trias de Bes. Inovativní marketing. (2004).
This work has been supported by the Scientific Grant Agency of
the Ministry of Education of the Slovak Republic VEGA
1/0064/23.
INNOVATIONS 2025
7
Application of Cybersecurity Awareness Training and End-User Behavioural Protection
Simulation Platforms into Higher Education
Dimitar Tashkov1, Angel Valchev1, Veselina Gagamova1,*, Elitsa Pavlova 2, Violeta Vasileva3
Rakovski National Defence College Sofia, Bulgaria1
University of National and World Economy, Sofia, Bulgaria2
Future Innovation Labs, Sofia, Bulgaria3
v.gagamova@rndc.bg
Abstract: As cyber threats continue to grow in scale and complexity, the demand for robust cybersecurity education has become increasingly
urgent. This paper examines the application of cybersecurity awareness training and end-user behavioral protection simulation platforms
within higher education settings. It analyses the challenges universities face in delivering both theoretical knowledge and hands-on
experience in cybersecurity and highlights the pedagogical value of integrating simulation-based tools into undergraduate and postgraduate
curricula. The study also underscores the importance of collaboration between academia, industry, and government to develop strategies for
cultivating cyber competencies. By leveraging interactive platforms, institutions can better prepare students to recognize, respond to, and
mitigate cyber risks in real-world environments.
Keywords: CYBERSECURITY, AWARENESS TRAINING, SIMULATION PLATFORMS, HIGHER SCHOOLS. END-USER BEHAVIORAL
PROTECTION
1. Introduction
Universities in Europe are at different stages of their digital
transformation in relation to the available infrastructure and
resources at their disposal. Digitalization includes changing
curricula, changing organizational structures, creating interactive
content [17, 20]. Higher education policies do not directly affect
innovation, but they are related to the drivers of innovation, says a
study conducted in EU universities [16]. Improving university
he
European Union Agency for Cybersecurity, is key for all stages of
training [10, 17, 18].
Most of these findings lead to the conclusion that there is a need
for a clearer definition of the knowledge and skills that a student
should possess. The Cybersecurity4Europe Centre conducted a
study on whether the EU has good facilities for training and practice
for undergraduate and postgraduate students. The report states that
higher education curricula should encourage the use of cyber
training grounds. It draws attention to the fact that the content of the
curriculum should be enriched with topics on organizational or
human aspects of cybersecurity.
.
2. Comparative Analysis of Platforms with Cloud
Services Used in Cyber Security Training
A comparative analysis of the most prominent types of
cybersecurity training platforms, examines their features,
pedagogical models, user engagement, and practical applicability.
By evaluating platforms such as e-learning modules, cyber ranges,
Capture the Flag (CTF) competitions, and hybrid learning
environments, the analysis aims to provide a comprehensive
overview that supports informed decision-making in cybersecurity
skills development.
Table 1 provides a comparative analysis of the types of
platforms according to various criteria: interactivity, practical focus,
theoretical content, necessary resources, accessibility, audience,
price.
Table 1: Comparative analysis of types of cybersecurity training platforms
Criteria
Cyber Range
MOOC
LMS
Gamific
ation /
CTF
Interactivity
Very high
Medium
Medium
High
Practical
focus
Extremely high
Low
Medium
Very
high
Theoretical
content
Limited
Very high
High
Limited
Resources
Very high
Low
Moderate
Moder
ate
Accessibility
Limited (paid)
Very
high
High
Mediu
m to
high
Audience
Advanced /
Professional
Beginners
/ Self-
employed
Students
/
Teachers
Interm
ediate
/
Advan
ced
Price
High
Low to
Medium
Included
in
academic
licenses
Often
free
Source: Gamification in workforce training [2]
The comparison between platforms shows that there is no
universal solution. The main challenges described in the study
differentiation in terms of opportunities for working with digital
technologies, a lack of strategic vision regarding cybersecurity
training, achieving a balance between increasing societal demands
and expectations for higher education [12]. The lack of practical
experience of students leads to a mismatch between what businesses
are looking for in a job candidate and the skills that candidates
actually possess after graduation.
and the content should be updated regularly, covering emerging
security threats and security controls that are implemented to
is security training and
why is [14]. That is why universities should
implement a hybrid model, combining MOOCs (Massive Open
Online Courses), Cyber Range, LMS (Learning Management
System) and gamification. It is necessary to provide modern
infrastructure, as well as an extensive portfolio of tools and
educational resources to help increase competencies and develop a
stimulating environment for learning and research.
The transformation of cybersecurity education in higher
education can be made possible by:
- Investments in cloud laboratories and digital resources;
- introduction of laboratories with virtual environments;
- use of open resources and open-source tools;
- expansion of partnerships with industry;
- promotion of certification among students and teachers;
- creation of adapted courses based on Microsoft Learn
with teaching in Bulgarian and their integration into existing
disciplines.
- CTF competitions.
INNOVATIONS 2025
8
3. Cybersecurity Awareness Training and End-
User Behavioral Protection Platforms
Cybersecurity training platforms are a critical component of an
and safeguards against human-factor vulnerabilities. These
platforms typically employ a multimodal approach to training,
including interactive simulations, knowledge assessments, phishing
simulations, and behavioural analytics, to cultivate security
awareness among employees [18, 19, 24]. Recent research shows
that organizations that implement comprehensive security
awareness training experience a 70% reduction in successful
phishing attacks and a significant reduction in security incidents due
to human error. The effectiveness of these platforms is highly
correlated with the frequency of training, the relevance of the
content to specific organizational threats, and the extent to which
the training materials adapt to the changing cybersecurity
landscape. Challenges in implementing them include measuring
long-term behaviour change, addressing training fatigue, and
quantifying the return on security investments.
In the context of cybersecurity awareness training, several
leading platforms stand out, combining different approaches to
increase engagement and reduce human risks. NINJIO offers
training through engaging animated stories inspired by real-life
events, ensuring high user engagement [1]. It has a gamified leader
board that encourages engagement. Key features include a private
hosting portal, interactive questionnaires, multilingualism, etc.
KnowBe4 stands out with its extensive library of training and
automated phishing simulations, as well as risk analyses for
employees [11]. Key advantages of the platform are multilingual
support for the administrative console and end-user localization
options. There are additional customization features that allow for
gamification, a library of up-to-date content, statistics, and graphs
for training. Cofense specializes in phishing simulations and
reporting tools that enable employees to actively participate in
threat detection [4]. The platform provides personalized training
that allows employees to follow courses at their own pace and
covers best practices for remote work, such as credential theft and
social engineering. CybSafe uses behavioral science to measure and
improve security awareness, focusing on behavioral change rather
than just knowledge transfer [7]. Key features include personalized
and data-driven content, as well as certified training.
Proofpoint offers a holistic approach to training with adaptive
programs based on user risk and vulnerability assessments [9]. The
training includes interactive and game-based training content,
phishing simulations, and a user knowledge assessment tool that
covers critical security issues. Living Security creates engaging
learning experiences through game-based scenarios and team
competitions [13]. The admin dashboard allows management to
measure employee performance and progress, while Infosec IQ
provides customized training programs with a focus on regulatory
compliance across industries.
While all of these solutions share a common goal of reducing
risk by raising awareness, they differ in their approach: from
interactive video and gamification to behavioral analysis and attack
simulations, allowing them to be tailored to the audience, training
objectives, and organizational context.
Taking a deep approach to cybersecurity awareness training
leads to a change in employee attitudes and beliefs. The more
informed employees are, the more likely they are to succeed in
protecting information assets. With the help of predictive analytics,
high-risk positions can be easily identified and monitored according
to specific markers. Training programs are only effective when they
are conducted repeatedly and integrate new knowledge. A key
challenge for higher education institutions is the effective
integration of technology into curricula. Microsoft technologies are
widely used in cybersecurity curricula, but it requires continuous
updates to the curriculum and ensuring access for students from
different backgrounds [21, 22, 23]. Limited budgets are a challenge,
which is why many universities use only the basic levels of security
features included in Office 365 A1/A3 licenses. Licensing tools
such as Defender for Endpoint, Microsoft Sentinel is too expensive
for small public universities. This leads to misconfigured security
policies in M365 and Azure. Security gaps can arise when, for
example, a university deploys Microsoft Defender but fails to
enable threat analysis due to a lack of know-how, or security
policies implemented through Microsoft Intune do not reach all
devices in use. Key messages that ENISA is sending to
organizations in this regard include more awareness and education
about security at all levels, risk management, and collaboration with
academia [8].
Microsoft also offers specific educational tools aimed at
improving learning. Some European universities use Microsoft
HoloLens, a mixed reality headset, for hands-on learning through
3D models, virtual labs and simulations. The GitHub platform
provides open source resources, tools and labs for learning and
development. Microsoft Learn contains free modules focused on the
basics of security, network protection, including topics such as Zero
Trust, identity management, attack protection, etc. Microsoft Learn
courses offer certifications for SC-900 (fundamentals of cloud
security, identities, data protection), Azure Security Engineer
Associate (Microsoft Defender, policy management, cloud
protection) AZ-500 and MS-500 which could be used as a
supplement to the training courses. Another option is to include
teachers in Microsoft Certified Trainer programs and access to
training materials and licenses.
4. Conclusion
The development of digital platforms has drastically changed
cybersecurity training in recent years. Combining different
approaches according to educational goals and audience allows for
effective, motivating and sustainable development of key skills in
this critically important area.
The analysis of the four main types of cybersecurity training
platforms Cyber Range, MOOC, LMS and Gamification/CTF
shows that each of them has specific advantages and limitations [5,
6, 15]. Cyber Range platforms provide the highest degree of
practical training in a controlled and realistic environment, but
require significant resources for implementation and maintenance
[6]. MOOC courses provide accessibility and a theoretical basis
suitable for mass learning, but do not offer real simulations and
personalized feedback. LMS systems are ideal for university
organization and assessment, allowing the integration of diverse
content, but rely on external solutions for interactivity and practical
value. Gamification and CTF platforms motivate learners through a
game approach and a competitive element, building practical skills
and critical thinking, but are more suitable for more advanced levels
[2, 3, 25]. On this basis, it is recommended to use a hybrid model
that combines the strengths of different platforms according to the
course objectives, the level of the students, and the available
infrastructure.
Acknowledgements
This report was written thanks to funding from the Ministry of
Education and Science in implementation of the National Strategy
for Research Development 2017-2030 under the National Scientific
Programme "Security and Defence", adopted by Decision of the
Council of Ministers No. 731 of 21 October 2021.
INNOVATIONS 2025
9
5. References
1. All-In-One Cybersecurity Awareness Training Solution |
(Accessed: 16 October 2024).
2. et al.
Improving empl -efficacy and information security and
Journal of Business Research, 179, p.
114685. Available at:
https://doi.org/10.1016/j.jbusres.2024.114685.
3.
education in the EU: Will the new approach produce the required
Education and Information Technologies,
27(3), pp. 30113036. Available at: https://doi.org/10.1007/s10639-
021-10704-y.
4. Cofense (2024) Phishing Protection Solutions, Cofense.
Available at: https://cofense.com/ (Accessed: 16 October 2024).
5. Coursera | Degrees, Certificates, & Free Online Courses (no
date) Coursera. Available at: https://www.coursera.org/ (Accessed:
30 April 2025).
6. Cyber Range & Capacity Building in Cybersecurity (2025)
CYBERWISER.eu. Available at: https://cyberwiser.eu/.
7. CybSafe (2024) The Human Risk Management Platform,
CybSafe. Available at: https://www.cybsafe.com/ (Accessed: 16
October 2024).
8. ENISA Europa (2019) Good Practice Guide on Training
Methodologies, ENISA Europa. Available at:
https://www.enisa.europa.eu/publications/good-practice-guide-on-
training-methodologies.
9. Enterprise Cybersecurity Solutions, Services & Training |
Proofpoint US (2024) Proofpoint. Available at:
https://www.proofpoint.com/us (Accessed: 16 October 2024).
10. isaca.org (2024) The Path to Improved Cybersecurity Culture,
isaca. Available at: https://www.isaca.org/resources/news-and-
trends/isaca-now-blog/2018/the-path-to-improved-cybersecurity-
culture.
11. Knowbe4 (2024) Security Awareness Training, Knowbe4.
Available at: https://www.knowbe4.com (Accessed: 16 October
2024).
12. Link Springer (2022) A University Landscape for the Digital
World, Link Springer. Available at:
https://link.springer.com/chapter/10.1007%2F978-3-030-44897-
4_1.
13. Living Security (2024) Human Risk Management Solution,
Living Security. Available at: https://www.livingsecurity.com
(Accessed: 16 October 2024).
14. Mimecast (2022) What is Security Awareness Training and
Why is it Important?, Mimecast. Available at:
https://www.mimecast.com/content/what-is-security-awareness-
training/.
15. (PDF) A gamification framework to enhance students’
intrinsic motivation on MOOC (no date). Available at:
https://www.researchgate.net/publication/331686982_A_gamificati
on_framework_to_enhance_students'_intrinsic_motivation_on_MO
OC?_tp=eyJjb250ZXh0Ijp7ImZpcnN0UGFnZSI6InB1YmxpY2F0a
W9uIiwicGFnZSI6Il9kaXJlY3QifX0 (Accessed: 2 May 2025).
16. Roger G. Baldwin (2021) Technology in Higher Education,
Education.stateuniversity.com. Available at:
https://education.stateuniversity.com/pages/2496/Technology-in-
Education-HIGHER-EDUCATION.html.
17. Ivanova, Yoana, Optimizing Simulation Models of an
Artificial Neural Network for Digital Recognition, Sofia:
Publication of Union of Scientists in Bulgaria: International Journal
(vol. 13), ISSN 1313-8251, pp. 59-70 Web of Science IJITS has ISI
Impact Factor (IIF) = 1,449
18. Ivanova, Yoana, Applications of Simulation Modeling and
Computer Visualizations for Studying Structured Crystals for
Implementation in Technical Devices, 20th EAI International
Conference, CSECS 2024, Sofia, Bulgaria, June 2830, 2024,
https://csecs-conf.eai-conferences.org/2024/program-at-glance
19. Ivanova, Yoana, Applications of Simulation Modelling
Method in Prevention of Jamming Attacks, 2022 IEEE International
Conference on Information Technologies, Proceedings of the 36-th
International Conference on Information Technologies, 15-16
September, Sofia, 2022, ISSN 1314-1023, pp. 10, 55,
http://infotech-bg.com/proceedings
20. Ivanova, Yoana, Adaptive Digitalization Methods and Digital
Transformation Trends for Security, Sofia: Publication of Union of
Scientists in Bulgaria: International Journal on Information
1313-8251, pp. 51-62.
21. Krumova, Kristina. Analiz na aspektite na operativna
savmestimost na KIS v koalitsionna sreda. In Sofia; 2022. p. 329
338.
22. Krumova, Kristina. Kiberefektite v operatsii proektsia na sila
v kiberprostranstvoto. In Sofia; Nauchna konferentsia s
Bulgaria, 2-
nline) 978-619-7711-52-3, p.346-356.
23. Krumova Kristina, Analiz na kontseptsiyata za
multidomeynovi operatsii i integrirane na novovaznikvashtata ideya
za "boen oblak", AzBuki, Profesionalno obrazovanie, knizhka 2,
ISSN 1314-555X (Print), ISSN 1314-8567 (Online), Sofia, 2023,
str. 179-185
24. Kostadinov Ch., Kiberataka etapi, vidove, tendentsii v
protivodeystvieto. In 2022. p. 2108. Sbornik dokladi ot Godishna
nauchna konferentsia
-619-7478-95-2, pp.
210-218.
25. Kostadinov Ch., (2022). Industria 4.0 i predizvikatelstvata pred
osiguryavaneto na sigurnost i zashtita v kiberprostranstvoto.
Avtomatizatsia na diskretnoto proizvodstvo, Broy 4, Izdatelstvo na
TU-Sofia, ISSN: 2682-9584, Sofia, pp. 184-189.
INNOVATIONS 2025
10
Киберсигурност по веригите за доставки на софтуер
Велиян Димитров1, Георги Димитров1,2, Александър Кирков1, Радостин Николов1, Николай Керин3
1Университет по библиотекознание и информационни технологии
2Институт по роботика, БАН, 3Глобал Сат ЕООД
Резюме: Целта на статията е да предложи разширена визия за проблемите по сигурността на веригите за доставки при произ-
водството на софтуер. Предложен е анализ на предизвикателствата пред софистичността на кибератаките през веригите за
доставки на софтуер, която заплашва всеки цифров артефакт в киберфизичния свят. Представена е визия за обхвата на пробле-
ма със сигурността на софтуерните продукти по веригите за доставки. Изследвани са случаи с последствия от атаки по вери-
гите за доставки. Систематизирани са слабостите в настоящото състояние на регулаторната база за сигурност на софтуер по
веригите за доставки.
Ключови думи: вериги, доставки, киберсигурност, софтуер, атаки, кибер, сигурност
Увод
Индустрията за информационни и комуникационни техно-
логии осъзна въздействието на веригите за доставки върху
проблемите на киберсигурността. Обществото на изследовате-
лите предупреждаваше за този вид опасности отдавна. Пейза-
жът на заплахите се усложнява и институциите вече вземат
мерки за регулиране и контрол на сигурността по веригите за
доставки. Атаките през веригите за доставки са ефективни
защото чрез тях се преодоляват традиционните защити, които
са основани на конвенционални инструменти. При тези трудни
за откриване атаки, организациите неволно канят противника в
архитектурите на съществени приложения, използвайки непро-
верени технологични компоненти. Това ефективно води до
самокомпрометиране. Скрити в доставения продукт зловредни
компоненти са откривани в хардуерни и софтуерни продукти,
които са произведени от трети страни и пристигат по веригите
за доставки. Настоящото изследване е фокусирано върху проб-
лемите, които са свързани с опасностите при производството
на софтуерни продукти и през цикъла на тяхната експлоатация.
Атаките през веригите за доставки заплашват всеки цифров
артефакт в киберфизичния свят. Последствията може да бъдат
свързани с въздействия върху бизнеса и производството, кри-
тична инфраструктура, технологиите в здравеопазването и др.
Този подход от страна на зловредните актьори показа ограни-
ченията на традиционните методи за сигурност и необходи-
мостта от нови технически и регулаторни мерки.
Методология
Приложен е евристичен способ основан на емпиричен опит
в проблемната област, проучване на публикациите за инциден-
ти по веригите за доставки и систематизиране на дефицитите в
регулаторните мерки.
Анализът ни на множество случаи на инциденти по кибер-
сигурността показва обхват на участващите елементи от раз-
лични трети страни, които излизат извън схващанията, на кои-
то са основани известните регулаторни мерки и изследваните
научни публикации. На Фиг. 1 е представена общата картина
на проблема, а поясненията и мотивите за нашето схващане са
както следва.
Модулите са елементите, които участват в конструкцията
на софтуерния продукт и се доставят от репозитари с отворен
код или се произвеждат от подизпълнители. Те може да са
програмни библиотеки, класове, схеми с XML или JSON, маси-
ви с данни и др. Архитектурните компоненти са операционните
системи, машините за бази данни, уеб сървърите, конфигура-
цията на мрежовата среда. Услугите от външни доставчици
може да са достъп до чужди бази данни, ftp сървери, ботове,
платформи за изкуствен интелект, софтуерите с генеричното
наименование EDR (End Point Detection and Response), довери-
телната верига за криптографски ключове, телеконферентни
платформи, виртуални частни мрежи [1].
Връзките с външни информационни системи включват вся-
какви програмни интерфейси (Application Programming
Interface, API), през които софтуерните приложения взаимо-
действат с външни системи.
Инфраструктурните мрежови услуги са тези, които са ин-
тентни за интернет и без тях в общия случай е невъзможно
съществуването на информационните системи. Такива са точ-
ното време (Network Time Protocal, NTP), имената на домейни-
те (Domain Name Service, DNS), протоколите за маршрутизира-
не например Routing Information Protocol (RIP), Interior Gateway
Protocol (IGRP), Open Shortest Path First (OSPF). Тези услуги
често са намесени в изтичане на данни и отвличане на трафик.
Фиг. 1. Екосистема на софтуерно приложение
Услуги от външни доставчици се прилагат при информаци-
онни системи с висока степен на сложност. Когато контейнери-
те S3 се разпадат и впоследствие са изоставени конфигурация-
та позволява злонамерените лица да ги пререгистрират за себе
си. Клас грешки, известен под името „S3 bucket takeover“. Ов-
ладяването на контейнери от втори ред чрез неработещи връзки
е известен проблем, но не е засегнат в обхвата на сигурността
по веригите за доставки [2].
Крайните потребители са замесени в големите атаки по ве-
ригите за доставки по много начини. Заразена платформа на
краен потребител може да стане плацдарм за атаки от типа
странично преместване към приложения, с които потребителят
взаимодейства. Включително с интерфейси за банкиране, паза-
руване и други.
Общи и отраслови регулации
Настоящата точка съдържа преглед на общите и отраслови
регулации на киберсигурността по веригите за доставки, които
излязоха през последните години.
В САЩ Белият дом издаде Изпълнителна заповед за подоб-
ряване на киберсигурността на нацията (EO 14028). Тя устано-
вява изисквания за сигурност по веригата за доставки на соф-
туер на федералното правителство. Те включват систематични
прегледи, подобрения в процесите и стандарти за сигурност за
доставчиците, за разработчиците на софтуер и клиентите, които
закупуват софтуер за федералните структури.
INNOVATIONS 2025
11
ESF (Enduring Security Framework) е междусекторна работ-
на група, която действа под егидата на Консултативния съвет
за партньорство в областта на критичната инфраструктура
раздели CIPAC (Critical Infrastructure Partnership Advisory
Council ) с цел да се справи със заплахите за сигурността на
националните системи за сигурност на САЩ. Издава ръководс-
тво с предложени практики за разработчици, доставчици и
заинтересовани трети страни от страна на клиентите, за да се
гарантира по-сигурна верига за доставки на софтуер. То е орга-
низирано в три части: Част 1 се фокусира върху разработчици-
те на софтуер; Част 2 се фокусира върху доставчиците на соф-
туер; и Част 3 се фокусира върху клиентите на софтуер. Пред-
ложените практики може да се прилагат по веригата за достав-
ки на софтуер през времето на фазите на внедряване и експлоа-
тация. Отговорностите на доставчиците включват гарантиране
на целостта и сигурността на софтуера чрез договорни спора-
зумения, издаване на софтуерни актуализации, уведомления за
открити проблеми по сигурността и смекчаване на уязвимости.
Документът представя норми, включващи планиране на изиск-
ванията за сигурност, проектиране на софтуерна архитектура
от гледна точка на сигурността, добавяне на функции за сигур-
ност и съпровод на сигурността на софтуера и основната инф-
раструктура [3].
Рисковите профили на доставчиците се променят въз осно-
ва на нова информация, насоки от националните органи, разуз-
наване за заплахи и атаки. Тези информационни потоци са
спонсорирани и регулирани от държавите, които са достигнали
до разбирането за важността на киберсигурността по веригите
за доставки. Всяка организация може да ги използва в контекс-
та на собствените си доставчици. Стандартът за най-добри
практики за управление на взаимоотношенията с доставчиците
е ISO/IEC 27002:2022, глави 5.19–5.23. Той изисква собствени-
кът на активите да определи правила за доставчиците и достав-
чиците на услуги, които защитават информационните плат-
форми на собственика на активите и определя изисквания за
продуктите.
Изискванията за сигурност са специфични в зависимост от
предназначението на даден продукт и съответната организаци-
онна потребност [4]. Пример за добра отраслова практика за
изисквания за сигурност към придобиваните продукти и услуги
в енергийния сектор с бялата книга на Bundesverband der
Energie und Wasserwirtschaft. В нея са определени изискванията
за продукти, използвани в критична инфраструктура. Те са
извлечени от изискванията на ISO/IEC27002:2022 и ISO/IEC
27019:2017 (контроли, специфични за енергийната област).
DORA коментира проблема в раздели 76, 80, 81, 82 [5].
Софтуерен списък с материали (Software Bill of Materials,
SBOM). Подробностите за интегриран компонент на трета
страна трябва да бъдат докладвани в SBOM за разработен про-
дукт, за да се валидират одобрените компоненти и да се иден-
тифицира наличието на уязвими компоненти. Няколко специ-
фикации определят формата на SBOM: 1. Проектите на Linux
Foundation „Обмен на данни за софтуерни пакети (SPDX)“. 2.
OWASP „CycloneDX“. 3. NIST „Етикети за идентификация на
софтуер (SWID)“.
Препоръчителните смекчаващи мерки за елементите от
SBOM трябва да се актуализират при необходимост. Всички
несъответствия трябва да бъдат докладвани на доставчика. За
тази цел, към софтуера, предоставен от третата страна, трябва
да се прилагат инструменти за анализ на състава на софтуера
(Software Component Analysis, SCA). Инструментът за двоично
сканиране SCA може да идентифицира съдържанието на край-
ните продукти от софтуера на третата страна. Документът
„Минимални елементи за софтуерен списък с материали“ на
NTIA (National Telecommunications and Information Administra-
tion) в т. 2.4 Подобряване на средата за изграждане очертава
два вида среди за изграждане – индивидуална среда за разра-
ботчици и производствена среда [3].
Инструменти за контрол
CycloneDX е олекотен стандарт за SBOM, предназначен за
използване в контекста на сигурността на приложенията и
анализа на компонентите на веригата за доставки. Той се фоку-
сира върху точна и възпроизводима идентификация на компо-
нентите във веригата за доставки на софтуер. CycloneDX под-
държа набор от типове съдържание, което позволява цялостно
покритие на софтуерните компоненти.
Software Package Data Exchange (SPDX) предоставя унифи-
циран начин за документиране на компонентите в софтуерен
пакет. Целта му е да стандартизира начина, по който софтуер-
ните компоненти се идентифицират, атрибутират и одитират.
Улеснява споделянето на данни между различни платформи и
инструменти. Широко възприет поради адаптивността му.
Поддържа цялостна документация на компонентите, включи-
телно лицензиране и уязвимости в сигурността, свързани с
всеки компонент. Това гарантира разбиране и управление на
софтуерните лицензи и рисковете за сигурността.
Етикетите за идентификация на софтуер (Software
Identification Tag, SWID) предлагат метод за уникално иденти-
фициране на софтуерни продукти и компоненти. Те предоста-
вят жизненоважни метаданни за продукта, които могат да се
използват за проследяване и управление на софтуерни активи.
Таговете със SWID осигуряват автоматизиран механизъм за
идентификация, който е в съответствие с ISO/IEC 19770-2 [6].
Улесняват спазването на лицензионните условия и регулатор-
ните изисквания и осигуряват точна и актуална инвентаризация
на софтуера. Това подобрява мерките за сигурност, позволява
бързо идентифициране и реагиране на уязвимости или несъот-
ветствия в софтуера. Предотвратява прилагането на остарели
модули например библиотеки, криптиращи алгоритми и др.
Инциденти по веригите за доставки
Представеният преглед на значимите инциденти по вериги-
те за доставки съдържа подбрани случаи, които представят
различни области на въздействие.
Компанията SolarWinds стана причина за инцидент в ки-
берсигурността, разкрит в края на 2020 г. Софтуерът им за
управление на мрежи беше използван за атаки от институцио-
нално спонсорирана хакерска група. Засегнати са правителст-
вени агенции и частни компании, което засили фокуса върху
сигурността на веригата за доставки на софтуер [7], [8].
При атака по веригите за доставки са компрометирани 36
разширения за браузъра Google Chrome. Изследователите са
идентифицирали разширения за Chrome, използвани общо от
2,6 милиона души, в които хакер е инжектирал зловреден соф-
туер за кражба на данни [9]. Едно от разширенията е създадено
от стартъпа за киберсигурност Cyberhaven. Предназначено е да
защитава корпоративните данни от вътрешни заплахи.
3CX Desktop App е корпоративен софтуер, който осигурява
комуникации за своите потребители, включително чат, видео
разговори и гласови повиквания. В края на март 2023 г., след
компрометиране на веригата за доставки на софтуер се разп-
ространи зловреден софтуер чрез троянска версия на легитим-
ния софтуер на 3CX, която беше достъпна за изтегляне от тех-
ния уебсайт. Засегнатият софтуер беше 3CX DesktopApp
18.12.416 и по-стара версия, който съдържаше зловреден код.
Разследването на Mandiant Consulting по веригата за доставки
на 3CX разкри първоначалния вектор на проникване: софтуе-
рен пакет, заразен със зловреден софтуер, разпространяван чрез
по-ранно компрометиране на веригата за доставки на софтуер,
което е започнало с подправен инсталатор за X_TRADER,
софтуерен пакет, предоставен от Trading Technologies. Mandiant
Consulting идентифицира инсталатор с име на файл
X_TRADER_r7.17.90p608.exe, който е довел до внедряването
на злонамерената задна врата VEILEDSIGNAL. Въпреки че
платформата X_TRADER е била спряна от производство през
INNOVATIONS 2025
12
2020 г., тя все още е била достъпна за изтегляне от легитимния
уебсайт на Trading Technologies през 2022 г.
Разширенията за браузъри са редовна цел за хакерите, кои-
то се стремят да ги подкопаят и да получат достъп до браузъ-
рите на жертвите и данните включително пароли и бисквитки
за сесия. Компрометирано разширение за Chrome е проследено
до стартиращата компания за киберсигурност Cyberhaven,
която предлага едноименното разширение за Chrome. То е
предназначено да защитава корпоративните данни от вътрешни
заплахи, включително случайно излагане.
Cyberhaven съобщи на клиентите си, че „нападателят е из-
ползвал достъпа, получен при тази атака, за да публикува зло-
намерено разширение за Chrome (версия 24.10.4) в уеб магази-
на на Chrome на 25.12.2024 г. Атаката „е засегнала само маши-
ни, работещи с браузъри, базирани на Chrome, които са били
актуализирани чрез уеб магазина на Google Chrome“. Инфор-
мация може да е била открадната от всяка система, използваща
уязвимото разширение, но само ако машините са били онлайн
между 1:32 ч. UTC на 25 декември и 2:50 ч, се казва в съобще-
нието [10].
ReversingLabs представят демонстрации на рисковете при
използване на GoToMeeting и BlueJeans Meet за видеоконфе-
рентна връзка, на които разчитат организации от всякакъв
мащаб за глобална комуникация [11]. Друга демонстрация
показва опасностите от прилагане на корпоративни VPN мре-
жи. Те са примери за разпределени системи, които може да
показват подозрителна функционалност, съдържат уязвимости
и остарели компоненти или представляват риск за данните,
които са предназначени да защитават [12].
През март 2023 г. изследователи от Unit 42 откриха шест
злонамерени пакета в мениджъра на пакети Python Package
Index (PyPI). Списък с библиотеки за Python, които се използ-
ват за зловредни цели е представен на сайта на SANS [13].
Множество подправени през 2025 г. библиотеки със злонаме-
рен код за Python и RubyGems са коментирани в [14] и [15].
Предназначението им е да откраднат идентификационни данни
за облачни услуги, социални мрежи, лични данни и информа-
ция за проследяване на крипто портфейли, съдържание на
съобщения, прикачени файлове, идентификационни данни от
прокси и дори токени, които могат да бъдат използвани за
отвличане на Telegram ботове [16].
Предизвикателства
Атаките по веригите за доставки са много по-сложни от
настоящите схващания, при които се пропускат характеристи-
ките на важни компоненти от Фиг 1. Следващият систематизи-
ран списък съдържа емпирично синтезирани предизвикателст-
вата пред регулирането на сигурността по веригите за достав-
ки.
Първите четири имат организационен характер, а оста-
налите технологичен.
Навлизащите авангардни технологии за интелигентни неща
(Smart Things, Smart city), ИИ (Изкуствен Интелект), DLT
(Distribured Ledger Technology). Създаването и функциониране-
то на платформите за изкуствен интелект са базирани на комп-
лексна мрежа от доставчици, а криптовалутите функционират в
сложна разпределена среда, която има неясна идеология за
свързване с конвенционалните инфраструктурни и интеграци-
онни услуги.
Корпоративните сливания, придобивания и разделяне водят
до нарушаване на приемствеността в екипите за разработване и
тези по сигурността. Тогава липсата на документация и проце-
дури става предпоставка за интегриране на непроверени ИТ
инфраструктури.
Публичните корпоративни репозитари за приложения и
програмни модули се замърсяват със зловредни компоненти.
Пропуските и ограниченията в данните за CVE се дължат
на множество фактори, които са интентни за слабостите в си-
гурността на продуктите. Поради което задължителните обяв-
ление към засегнатите се издават със закъснения.
Атакуващите през веригата за доставки на софтуер подме-
нят софтуера преди внедряването му или вграждат зловреден
софтуер в софтуерните актуализации. Тестовете за проникване
не са предназначени да откриват вграден зловреден код или
софтуерни промени.
На практика мерките за сигурност по веригите за доставки
се отнасят към седмо ниво по OSI. Това изключва познатите
проблеми при останалите нива. Хардуерът от първо и второ
нинво по OSI се доставя с отворени портове, заразява се със
задни врати, съществуват атаки през второ и трето ниво по OSI.
Изводи
Считаме, че осъзнаването на предизвикателствата по темата
сигурност по веригите за доставки от страна на регулаторните
органи, индустрията за ИТ и изследователите на този етап е
недостатъчно. Част от регулаторните мерки по темата имат
формален характер и неясен резултат. Същевременно са про-
пуснати лесни мерки, които да бъдат ефективни например
показаните в [2].
Анализът на прегледаните регулаторни мерки и изследва-
ните инциденти от изложеното дотук ни доведе до следните
изводи, които имат обобщаващ, мета характер.
Знанието в областта на киберсигурността е фрагментирано
и с неустановен терминологичен апарат. Идеите за сигурност
на софтуера по веригите за доставки са откъснати от концеп-
циите за сигурност в дълбочина, нулево доверие, проактивни
мерки, разузнаване на заплахите и “перлите в короната” [17].
Разследванията са затруднени поради възможности за
сложна оперативна игра. Злонамерен, но легитимен актьор
може да присъства дълго време в даден софтуерен проект и да
изгради доверие. След което спокойно да инкорпорира задна
врата в отделен софтуерен модул само за специален клиент.
Или да напусне екипа по разработването и да остави в продук-
ционната среда активен агент, който да го прави.
Пропуска се контролът върху крайните потребители с тех-
ните крайни устройства. Те може да са носители на стари вер-
сии на операционни системи, приложения с уязвимости насле-
дени по веригата за доставки, От тях може да се активират
атаки от типа странично преместване.
Доверието към доставеният хардуер за производствените
среди на софтуерните разработчици остава извън обхвата на
регулациите. На практика дори един лаптоп съдържа компо-
ненти от множество производители от различни държави.
Регулаторните мерки не обхващат инфраструктурни услуги
и архитектурни интеграционни компоненти, каквито са VPN,
телеконферентните платформи, облачните услуги, изоставени
сайтове и домейни, които се наследяват от зловредни актьори.
Възможностите за атаки благодарение на изоставени домейни
са демонстрирани в [18].
Приключването на жизнения цикъл на софтуерните проек-
ти се неглижира. Пример са облачните услуги с изоставени
контейнери в Amazon с имена на домейни от [19].
Не са изяснени статута и задълженията на електронните ма-
газини за приложения и репозитарите за програмни модули за
саниране на средата и нулево доверие към предлаганите про-
дукти, част от идеята за проактивна сигурност [20].
Препоръки
Организацията, която приема продукта, трябва да следи на-
личните механизми за докладване на CVE и каналите за под-
INNOVATIONS 2025
13
дръжка от трети страни, за да определи дали уязвимостите,
идентифицирани в рамките на приет компонент на трета стра-
на, могат да повлияят на продуктите и да предприеме подхо-
дящи действия за разрешаване или смекчаване на уязвимостта.
Договорът с третата страна трябва да разрешава бъдещи
уязвимости. Собственикът на компонента на третата страна
трябва също да уведоми продуктовата организация за наличие-
то на уязвимост, риска, свързан с нея, и времева рамка за това
кога уязвимостта ще бъде отстранена и предоставена.
Добрите практики се предоставят и могат да бъдат прила-
гани организации заявители, основни и важни субекти съгласно
директивата NIS2, или от съответните им изпълнители и дос-
тавчици. Диобрите практики обхващат пет области: стратеги-
чески корпоративен подход, управление на риска във веригата
за доставки, управление на взаимоотношенията с доставчиците,
справяне с уязвимостите, качество на продуктите и практиките
за изпълнителите и доставчиците на услуги. Докладът [4] зак-
лючава следното. Съществува объркване по отношение на
терминологията около веригата за доставки на ИКТ. Взаимо-
действието между директивата NIS2 и предложението за закон
за киберустойчивост или друго законодателство, секторно или
не, което предвижда изисквания за киберсигурност за продукти
и услуги, следва да бъде допълнително проучено [4].
Считаме, че нормалите и инструментите от третата и чет-
въртата част на настоящата работа са неприложими при конст-
руирането и експлоатацията на платформите за изкуствен ин-
телект. За тази област са необходими специфични мерки.
Изграждането и осигуряването на верига за доставки е съв-
местно усилие. Управлението на бизнеса включва външни
партньорства и различни заинтересовани страни. Тази взаимос-
вързана екосистема формира верига за доставки, която включва
доставчици на софтуерни продукти, доставчици на услуги,
партньори и клиенти, които използват информационните сис-
теми на организацията по веригата за пласмент. Осъзнаването
на мястото на киберсигурността в тези взаимоотношения от
ръководствата на организациите гарантира, че мрежата на
веригата за доставки функционира хармонично и защитава
бизнеса.
Литературни източници
[1] C. O’Gara, “Cybersecurity Ratings of Remote Meeting
Apps: Zoom, Teams, Skype, and More.” Accessed: May 23,
2025. [Online]. Available:
https://www.secureworld.io/industry-news/cybersecurity-
and-privacy-ratings-of-remote-meeting-apps-like-zoom-
teams-and-skype
[2] “8 Million Requests Later, We Made The SolarWinds
Supply Chain Attack Look Amateur,” watchTowr Labs.
Accessed: Jun. 01, 2025. [Online]. Available:
https://labs.watchtowr.com/8-million-requests-later-we-
made-the-solarwinds-supply-chain-attack-look-amateur/
[3] Enduring Security Framework, “ECURING THE
SOFTWARE SUPPLY CHAIN RECOMMENDED
PRACTICES GUIDE FOR DEVELOPERS.” Aug. 2022.
[Online]. Available:
https://www.cisa.gov/sites/default/files/publications/ESF_SE
CURING_THE_SOFTWARE_SUPPLY_CHAIN_DEVELO
PERS.PDF
[4] European Union Agency for Cybersecurity., Good practices
for supply chain cybersecurity. LU: Publications Office,
2023. Accessed: May 22, 2025. [Online]. Available:
https://data.europa.eu/doi/10.2824/805268
[5] “REGULATION (EU) 2022/2554 OF THE EUROPEAN
PARLIAMENT AND OF THE COUNCIL of 14 December
2022 on digital operational resilience for the financial sector
and amending Regulations (EC) No 1060/2009, (EU) No
648/2012, (EU) No 600/2014, (EU) No 909/2014 and (EU)
2016/1011.” Official Journal of the European Union, Dec.
27, 2022. [Online]. Available: https://eur-
lex.europa.eu/legal-
content/EN/TXT/PDF/?uri=CELEX:32022R2554&from=FR
[6] “SBOM Tools: The Basics and 5 Free Tools to Get You
Started,” Aqua. Accessed: May 31, 2025. [Online].
Available: https://www.aquasec.com/cloud-native-
academy/supply-chain-security/sbom-tools/
[7] Gia Anisa and Fitria Widianingsih, “SolarWinds Attack:
Stages, Implications, and Mitigation Strategies in the Cyber
Age,” ENIGMA, vol. 2, no. 1, pp. 47–52, Oct. 2024, doi:
10.62123/enigma.v2i1.31.
[8] S. Peisert et al., “Perspectives on the SolarWinds incident,”
IEEE Security & Privacy, vol. 19, no. 2, pp. 7–13, Mar.
2021, doi: 10.1109/msec.2021.3051235.
[9] M. J. S. January 3 and 2025, “36 Chrome Extensions
Compromised in Supply Chain Attack.” Accessed: May 21,
2025. [Online]. Available:
https://www.inforisktoday.com/36-chrome-extensions-
compromised-in-supply-chain-attack-a-27207
[10] M. J. S. December 30 and 2024, “Hackers Launch Supply
Chain Attack Against Chrome Extensions.” Accessed: May
21, 2025. [Online]. Available:
https://www.inforisktoday.com/hackers-launch-supply-
chain-attack-against-chrome-extensions-a-27173
[11] T. Stahl, “Software Package Deconstruction: Video
Conferencing Software,” ReversingLabs. Accessed: Jun. 09,
2025. [Online]. Available:
https://www.reversinglabs.com/software-package-
deconstruction-series/video-conferencing-software-
deconstruction
[12] T. Stahl, “Software Package Deconstruction: Enterprise VPN
Comparison,” ReversingLabs. Accessed: Jun. 09, 2025.
[Online]. Available:
https://www.reversinglabs.com/software-package-
deconstruction-series/vpn-deconstruction
[13] S. I. S. Center, “Python Libraries Used for Malicious
Purposes,” SANS Internet Storm Center. Accessed: Jun. 09,
2025. [Online]. Available: https://isc.sans.edu/diary/31248
[14] “Multiple Malicious Packages Discovered on PyPI, npm, and
RubyGems,” Cloudsmith. Accessed: Jun. 09, 2025. [Online].
Available: https://cloudsmith.com/blog/multiple-malicious-
packages-discovered-on-pypi-npm-and-rubygems
[15] “Malicious RubyGems pose as Fastlane to steal Telegram
API data,” BleepingComputer. Accessed: Jun. 09, 2025.
[Online]. Available:
https://www.bleepingcomputer.com/news/security/malicious
-rubygems-pose-as-fastlane-to-steal-telegram-api-data/
[16] S. B. Hai, “Six Malicious Python Packages in the PyPI
Targeting Windows Users,” Unit 42. Accessed: Jun. 09,
2025. [Online]. Available:
https://unit42.paloaltonetworks.com/malicious-packages-in-
pypi/
[17] R. Gangupantulu, T. Cody, A. Rahman, C. Redino, R. Clark,
and P. Park, “Crown Jewels Analysis using Reinforcement
Learning with Attack Graphs,” Aug. 20, 2021, arXiv:
arXiv:2108.09358. doi: 10.48550/arXiv.2108.09358.
[18] “Backdooring Your Backdoors - Another $20 Domain, More
Governments,” watchTowr Labs. Accessed: Jun. 01, 2025.
[Online]. Available: https://labs.watchtowr.com/more-
governments-backdoors-in-your-backdoors/
[19] “8 Million Requests Later, We Made The SolarWinds
Supply Chain Attack Look Amateur,” watchTowr Labs.
Accessed: Jun. 09, 2025. [Online]. Available:
https://labs.watchtowr.com/8-million-requests-later-we-
made-the-solarwinds-supply-chain-attack-look-amateur/
[20] “Considering cyber pollution control.” Accessed: May 19,
2025. [Online]. Available:
https://www.arcticsecurity.com/resources/considering-cyber-
pollution-control
INNOVATIONS 2025
14
Immersive Technologies as an Educational Innovation for Training the Next Generation of
Professionals
Kaloyan Dimitrov
University of National and World Economy - Sofia, Bulgaria
kdimittrov@unwe.bg
Abstract: Immersive technologies, including virtual, augmented, and mixed reality, are transforming contemporary education by facilitating
interactive learning environments. This paper examines the potential impact of these technologies on the training of professionals in different
fields. The study identifies key tasks, including an analysis of the educational approach of the technology and an assessment of the
applicability of these tools in different educational contexts and domains. The results of the study indicate an enhancement in educational
outcomes, as evidenced by an increase in opportunities and benefits for learners. The paper also discusses the importance of VR, AR, and
MR in preparing future professionals who will work in highly dynamic, technology-driven work environments. The research posits that
immersive technologies are an indispensable component of contemporary educational approaches to teaching and learning, providing future
professionals with the competencies required in evolving, technology-driven industries.
Keywords: IMMERSIVE TECHNOLOGIES, VIRTUAL REALITY (VR), AUGMENTED REALITY (AR), MIXED REALITY (MR),
ENGAGEMENT, MOTIVATION, IMMERSIVE LEARNING, IMMERSIVE TEACHING
1. Introduction
The rapid development of digital tools and technology's
increasing role in all aspects of society have made changes to
educational approaches necessary. Traditional teaching methods
often fail to meet the needs of today's learners, particularly in
subjects that require practical experience, problem-solving, and
critical thinking skills. In response, immersive technologies - virtual
reality (VR), augmented reality (AR) and mixed reality (MR) - have
emerged as promising educational tools. They enable interactive,
experiential learning by immersing students in simulated
environments, thereby increasing engagement and bridging the gap
between theoretical knowledge and practical application.
Immersive technologies are important to modern education
because they meet the demands of dynamic, technology-driven
industries. The job market increasingly values individuals who
possess practical skills, adaptability, and critical thinking abilities.
In this context, educational institutions must adopt approaches that
effectively prepare learners for professional challenges. Immersive
technologies provide a realistic, interactive environment that allows
learners to safely practice their skills and develop competencies
through experience. This makes them important for the future of
vocational training [1].
The pedagogical benefits of immersive technologies remain
understudied. On the one hand, it has been noted that immersive
environments promote active learning, improve student engagement
and retention, and facilitate a better understanding of complex
concepts. Virtual reality, for example, enhances learning by
providing a more hands-on approach to education. Augmented
reality, on the other hand, enhances collaborative learning by
enabling real-time interaction with digital content. However, the
implementation of immersive tools faces challenges such as cost,
technological barriers, and the need to train educators. Despite these
challenges, research shows that using immersive technologies can
significantly improve problem-solving, motivation, and critical
thinking skills, particularly in STEM, medical, and technical
education.
The impetus for conducting this research stems from the
objective of examining and examining the potential of immersive
technologies as an innovative educational approach for training
future professionals. This objective is pursued through the
description of existing immersive technologies and the theoretical
justification of their use. The assessment results in the identification
of strategies to address the evolving demands of preparing today's
students to become the workforce. To that end, the following
research tasks have been identified as necessary to achieve the main
objective:
• To examine and describe existing educational immersive
technologies,
• To explore and analyze the applicability of VR, AR and MR in
different educational contexts and domains;
• To evaluate the impact of immersive learning on students'
academic achievement, motivation and future readiness;
• To analyze the educational potential of immersive
technologies to promote student engagement, critical thinking and
practical skills to use immersive technologies in education.
2. The importance of immersive technologies in
modern education and vocational training
The rapid development of digital technologies has precipitated a
paradigm shift in the manner in which knowledge is acquired and
applied. Consequently, educational institutions are compelled to
adopt innovative approaches to meet the needs of today's learners.
Immersive technologies, including VR, AR and MR, have emerged
as novel pedagogical tools, offering interactive, experience-based
learning environments that facilitate student engagement. The
efficacy of these tools in facilitating the transition between
theoretical concepts and practical applications has been
demonstrated, particularly within professions that demand a high
degree of manual dexterity and technical expertise, such as
medicine, engineering, and related technical domains.
In the contemporary job market, experiential learning, problem-
solving skills, and critical thinking are highly prized, thus rendering
immersive technologies a pivotal component in the preparation of
future professionals. Through the emulation of real-world scenarios,
these technologies empower learners to apply their knowledge in a
secure and controlled environment, enhancing learning outcomes
and mitigating the risks associated with conventional training in
high-stakes domains. Moreover, immersive learning has been
demonstrated to enhance student motivation and learning
engagement by addressing the challenges associated with student
retention and participation in traditional classrooms [2].
The significance of these technologies extends to the realms of
lifelong learning and professional development, where continuous
upskilling is imperative in rapidly evolving industries. As the
demand for professionals proficient in advanced cyphoric
technologies increases, the integration of immersive technologies
into education and training interventions is imperative to ensure that
learners acquire the necessary competencies to thrive in dynamic
and technological workplaces. Therefore, immersive technologies
are not only relevant but also essential for modernizing education
and ensuring learners' futures in changing professional
environments [3].
In general, immersive technologies are proving to be effective
tools in educational and training settings. These technologies
INNOVATIONS 2025
15
facilitate immersive, interactive experiences, enabling future
professionals to hone their skills in authentic real-world settings.
The evident advantages of immersive learning in the professional
training context can be categorized across various domains of
human activity, including [4]:
• Increased engagement and knowledge retention - immersive
tools stimulate multiple senses, which promotes deeper
engagement; studies show that knowledge retention increases when
learners actively participate rather than passively absorb content);
• Experiential learning - immersive environments offer hands-
on experience through virtual simulations; medical students, for
example, can perform surgeries in VR, gaining confidence without
real-life risks);
• Safe environment for making mistakes - learners can practice
and make mistakes without real-world consequences; for example,
pilots use flight simulators to train in difficult conditions without
endangering their lives;
• Develop and practice complex skills - fields such as
engineering and healthcare require mastery of complex tasks;
immersive technologies allow learners to practice repeatedly,
honing motor skills and critical thinking);
• Global and remote access to learning - using immersive
platforms, learners from anywhere can collaborate in the same
virtual environment, facilitating distance learning programs.
Immersive technologies have the potential to transform
education and vocational training, allowing learners to engage in
learning in safe, realistic, and risk-free environments using
immersive simulations of real work environments. The utilization of
a virtual reality module in an educational setting facilitates the
acquisition of practical skills in a risk-free, engaging, and
demanding environment, thereby enhancing students' readiness for
the professional realm.
It is imperative to acknowledge the accelerated advancements in
technology, which are progressively broadening the scope of virtual
reality's integration into contemporary learning methodologies. This
proliferation of possibilities renders the technology potentially
obsolete even before its formal incorporation into the educational
and training curricula of future professionals [5].
3. A brief overview of virtual, augmented and
mixed reality tools
Immersive technologies, including VR, AR and MR, have
emerged as innovative tools to enhance educational experiences by
integrating digital content with learning environments. The
utilization of these instruments fosters distinctive functionalities,
thereby facilitating interactive and experiential learning
methodologies [6, 7]. These instruments empower learners to
engage with academic materials in a more profound manner:
• Virtual reality has emerged as a significant technological
advancement in recent years, with applications ranging from
entertainment to education and beyond. VR engenders a sense of
total immersion in a digital environment, effectively isolating users
from the physical world. Through the use of virtual reality headsets
and motion-tracking devices, learners are able to explore simulated
spaces, perform virtual experiments, or participate in interactive
scenarios. The utilization of VR technology has gained significant
traction across diverse disciplines, including medicine, engineering,
and architecture. This technological advancement facilitates the
execution of procedures and operations in a controlled and risk-free
environment, thereby promoting skill development and training for
practitioners in these fields.
• Augmented Reality is a technology that overlays digital
content onto the physical world using smartphones, tablets or
augmented reality glasses. In contrast to the immersive experience
offered by VR, AR enables learners to maintain a connection with
their physical environment while engaging with virtual content or
information in real-time. The technology has applications in science
education, language learning, and fieldwork, as it provides
contextual information or enhances physical models with digital
elements.
• Mixed Reality combines the virtual and physical worlds so that
digital elements not only appear in the real environment, but can
also interact with it. The utilization of MR headsets enables learners
to interact with virtual objects in a manner that is indistinguishable
from real-world interactions, rendering it an optimal medium for
technical training and collaborative endeavors. The utility of MR is
particularly evident in fields such as engineering and product
design, where learners require the ability to visualize complex
processes in an interactive environment.
Nowadays, the most popular VR Heаdsets are shown in the
figure below:
1.
2.
3.
4.
Fig. 1 1)Meta Quest 3; 2) Lenovo ThinkRealityVRX; 3) HTC VIVE Focus 3;
4) PICO 4 Enterprise.
4. A discussion of potential applications and
benefits of educational immersive technologies
A growing body of research has identified a correlation between
certain applications and the subsequent benefits associated with
immersive learning [8, 9, 10]:
• Efficacy: The efficacy of hands-on, immersive training is well-
documented. This training method has been shown to increase
confidence in on-the-job performance, leading to improved skills,
knowledge retention, and behavior transfer. These devices have
been shown to enhance the retention of material.
• Efficiency: The efficiency of the system is such that a training
period that previously required a full day can now be completed in
as little as 15 minutes, without compromising the desired results.
Immersive training has been demonstrated to enhance focus,
accelerate skill acquisition, and enable multiple individuals to learn
simultaneously in an interactive environment.
• Cost-effectiveness: The issue of cost-effectiveness is
paramount in the context of traditional training. This training
method entails various expenses, including travel costs, organizer
salaries, expert fees, and room and equipment rental. Immersive
training eliminates the need for these by offering reusable modules
and increasing productivity.
• Standardization: The standardization is often compromised
due to variations in training programs, which can be attributed to
differences in instructor location and expertise. These
inconsistencies can lead to variations in the quality and
effectiveness of training programs. Immersive training provides a
standardized, repeatable experience that can be easily scaled across
the organization. The translation of modules into different
languages serves to ensure uniformity across regions and cultures.
• Measurement: The measurability of outcomes is a critical
component of immersive technology training. This training utilizes
xAPI data tracking to assess learner readiness, ensuring the
accuracy and reliability of the measurement process. The process of
INNOVATIONS 2025
16
personalization and performance assessment entails the utilization
of metrics such as decision-making, task completion time, and
performance deviations. Indeed, this combination serves to maintain
the engagement of the learning process.
• Security and safety: In environments characterized by elevated
risk levels, errors can have deleterious consequences that may result
in severe injury or death. Immersive training provides a risk-free
environment in which learners can safely participate in complex
scenarios and learn from their mistakes.
Immersive technologies have been shown to complement each
other, with each offering unique benefits for different educational
needs. The utilization of these tools fosters engagement, facilitates
hands-on skill development, and encourages problem-solving, all of
which are paramount for a new generation of professionals [11]. As
these instruments become increasingly available, their incorporation
into educational settings, such as classrooms, laboratories, and
professional development programs, will undergo significant
expansion. This integration will precipitate a fundamental shift in
the manner by which knowledge is transmitted, acquired, and
transferred.
Table 1: Examples of applications in various fields.
Field
Description
Medicine
Virtual reality has been incorporated into surgical training
regimens, enabling learners to execute intricate medical
procedures on virtual patients. Augmented reality (AR)
applications offer the capability to overlay anatomical
structures onto a three-dimensional model of the human
body.
Military and
Defence
Virtual simulations are utilized to equip military personnel
with the skills necessary to respond effectively to combat
scenarios and disaster situations by enhancing their
decision-making abilities under high-pressure
circumstances.
Aviation
Pilots and crew members are subjected to an intensive
training regimen that utilizes a variety of simulators to
instruct them in both routine procedures and emergency
responses.
Architecture
and design
Augmented and virtual reality tools facilitate the
visualization of three-dimensional structures, enabling
learners to assess their designs in real time.
Engineering
The utilization of virtual, augmented and mixed reality
facilitates the design and evaluation of prototypes through
the employment of holographic instruments.
Business skills
and leadership
Immersive simulations have been shown to facilitate the
development of leadership skills by subjecting participants
to high-pressure business scenarios, thereby enhancing
their problem-solving abilities.
Foreign
Language
The integration of VR and AR technologies with artificial
intelligence (AI)-powered language learning tools has the
potential to offer an even more immersive and interactive
learning experience.
Source: created by the author
5. Conclusion
The past decade has seen a particular move towards the
integration of various new technologies in education and training,
including immersive technologies. The integration of these
technologies into the educational environment has been shown to
enhance teaching and experiential learning, promote engagement,
and link theoretical knowledge with practical skills. The present
study is an ongoing investigation and examination of immersive
technologies as an innovative educational approach. This
investigation has yielded several key conclusions related to the
tasks outlined in this article. These conclusions suggest that
immersive technologies have the potential to create numerous
opportunities and benefits for education and training systems.
In the context of technological development, these tools offer
new educational opportunities and benefits for learners by
promoting interactive and experiential learning for better theoretical
knowledge and its practical application. Therefore, immersive
technologies have their significant potential and role in the
development of educational approaches, respectively in quality
improvement.
The analysis of previous state-of-the-art experimental studies
contributed to drawing relevant conclusions about the
appropriateness and high effectiveness of using educational
immersive technologies for teaching university students. This
finding indicates that immersive environments can effectively
enhance students' motivation, problem-solving abilities, and critical
thinking skills. Nevertheless, in order to obtain more objective
conclusions, additional field studies and observations are necessary
to directly address these issues.
In conclusion, it can be posited that immersive technologies
represent an essential component of modern educational
approaches, providing future professionals with the competencies
needed in evolving, technology-driven industries. The examples
presented in this study lay the foundation for further research and
projects at both the theoretical and practical level. Subsequent
researchers may wish to direct their attention to the exploration of
factors that enhance learner motivation through interactive learning.
6. Acknowledgements
This publication contains the results of a study funded by the
Scientific Research Fund of the Republic of Bulgaria under
administrative contract KP 06-H65/5 of 12.12.2022.
7. References
1. Abdraimova, E.T., Zh, S.A. and Zhorabaev, K.B. Immersive
technologies in a higher school in the modern digital reality.
Евразийский союз ученых, (10-1 (79)), pp.49-52. (2020)
2. Blyth, C., Immersive technologies and language learning.
Foreign Language Annals, 51(1), pp.225-232. (2018)
3. Parong, J. and Mayer, R.E., Learning science in immersive
virtual reality. Journal of educational psychology, 110(6), p.785.
(2018)
4. Huang, K.T., Ball, C., Francis, J., Ratan, R., Boumis, J. and
Fordham, J., Augmented versus virtual reality in education: An
exploratory study examining science knowledge retention when
using augmented reality/virtual reality mobile applications.
Cyberpsychology, Behavior, and Social Networking, 22(2), pp.105-
110. (2019)
5. Pellas, Nikolaos, Stylianos Mystakidis, and Ioannis Kazanidis.
Immersive Virtual Reality in K-12 and Higher Education: A
systematic review of the last decade scientific literature. Virtual
Reality, 25(3), 835-861. (2021)
6. Dimitrov, K., The Need of Immersive Technology in Digital
Higher Education Ecosystem. Innovative Information Technologies
for Economy Digitalization (IITED), 1, pp.37-43. (2024)
7. Schaffernak, H., Moesl, B., Vorraber, W. and Koglbauer, I.V.,
Potential augmented reality application areas for pilot education: An
exploratory study. Education Sciences, 10(4), p.86. (2020)
8. Stepan, K., Zeiger, J., Hanchuk, S., Del Signore, A.,
Shrivastava, R., Govindaraj, S. and Iloreta, A., Immersive virtual
reality as a teaching tool for neuroanatomy. In International forum
of allergy & rhinology, 7(10), pp. 1006-1013. (2017)
9. Tang, Y.M., Chau, K.Y., Kwok, A.P.K., Zhu, T. and Ma, X., A
systematic review of immersive technology applications for medical
practice and education-trends, application areas, recipients, teaching
contents, evaluation methods, and performance. Educational
Research Review, 35, p.100429. (2022)
10. Daher, S., Clark, A. and Barmaki, R., 2022. Immersive
Technologies in Healthcare. Frontiers in Virtual Reality, 3,
p.962950.
11. Facchino, A.P., Marchetti, D., Colasanti, M., Fontanesi, L. and
Verrocchio, M.C., 2025, February. The use of serious games for
psychological education and training: a systematic review. In
Frontiers in Education (Vol. 10, p. 1511729). Frontiers Media SA.
INNOVATIONS 2025
17
Regarding cybersecurity in Bulgarian educational institutions at the K–12 level
Marieta Hristova1, Diana Netova1, Nikolay Netov1
Sofia University "St. Kliment Ohridski"1
marietaivanova@feb.uni-sofia.bg, dianaht@feb.uni-sofia.bg, nnetoff@feb.uni-sofia.bg
Abstract: Cybersecurity studies increasingly prioritize empirical methodologies to understand and alleviate security risks arising from
human behavior, organizational practices, and the advancement of technologies used to perform cyberattacks. This research paper explores
how the school management and key educators in Bulgarian K-12 schools comprehend cybersecurity within the academic framework of the
Bulgarian education system. Our survey of 927 Bulgarian K–12 educational institutions revealed that a very small number of them assess
their level of cybersecurity as extremely low. Respondents see cybersecurity training as essential for ensuring information security within
their institutions. Our findings indicate that only 25% of K-12 employees are able to incorporate this training into their qualification
programs.
KEYWORDS: HUMAN FACTOR IN SECURITY, K-12 EDUCATION, BULGARIA
1. Introduction
Cybersecurity establishes policies, procedures, and technical
methods to protect, detect, correct, and defend against harm, illegal
use or modification, or exploitation of information and
communication systems and the data they contain. The rapid speed
of technical progress and innovation, along with the rapidly
developing nature of cyber threats, exacerbates the situation, [1].
The European Union underlines the significance of the science
and research sector, categorizing it as a vital infrastructure sector
subject to distinct cybersecurity rules. The NIS 2 Directive (EU
2022/2555) establishes a revised framework for cybersecurity inside
the European Union, superseding the original NIS Directive (2016).
It aims to improve cybersecurity within the European Union by
creating a uniform high standard of security for network and
information systems. While K–12 level educational institutions are
not covered by NIS 2 Directive, their cybersecurity must not be
overlooked. The rapid growth of digital educational infrastructures,
combined with the rise of cloud computing, is introducing new risks
that established security frameworks are struggling to manage.
A recent article from 2024 delineates and elucidates the primary
elements of cybersecurity protection in Bulgaria. The rules for
adopting and implementing effective measures to avoid cybercrime
are substantiated based on the ideas of "information security,"
"cyber security," and "cyber resilience." An analysis has been
conducted on the principal administrative documents and legal
instruments pertaining to national cyber security protection. A legal
analysis was performed for the purpose of examining potential
vulnerabilities, dangers, and risks to cybersecurity in Bulgaria, [2].
A study conducted in Bulgaria presents findings from a survey
of Cybersecurity Experts aimed at collecting and synthesizing
extensive information about the challenges of ensuring
cybersecurity in the country, with particular emphasis on the critical
role of human factors (HF) in this field. The survey was executed as
a component of a project funded by the Bulgarian Institute of Public
Administration in 2019. The survey inquiries relate to three specific
sectors: public administration, academia, and the business sector,
each of which supports the functioning of e-government in
Bulgaria. The study results corroborated the prevailing consensus
among experts that the human element may represent the 'weakest
link' in cybersecurity as a socio-technical system. Therefore, experts
believe that the majority of cybersecurity breaches are attributable
to human error or other HF vulnerabilities. The predominant issues
associated with successful assaults are spam and ransomware.
Nonetheless, authors observed a propensity among SMEs to regard
HF as a potentially robust method for identifying and alleviating
cyber hazards. Consequently, they deem it essential to focus on the
significance of the human element in the realm of cybersecurity in
Bulgaria. Simultaneously, authors highlighted the competency
level and inadequate capacity (knowledge and skills) of IT
personnel as the primary issue. The research indicates that, despite
the acknowledged significance of HF in cybersecurity, the training
remains inadequate, [3].
2. The Research Backgrounds
The educational institutions at the K–12 level are frequently
targeted by cyber threats. K-12 schools and districts are confronting
numerous formidable risks as contemporary cyber-attacks get
increasingly sophisticated and astute. A suitable starting point for
our analysis can be found in [4]. An analysis of the Cybersecurity
and Infrastructure Security Agency of the U.S. Department of
Homeland Security indicates that K–12 organizations require
simplicity, prioritization, and resources tailored to their unique
needs. [4] aims to advance the response to this call by offering clear
recommendations and resources to assist K–12 organizations in
effectively mitigating their evolving cybersecurity risks. The
report's key findings and recommendations are summarized as
follows:
Table 1: Key findings and recommendations, Source[2]
FINDING
RECOMMENDATION
1
With finite resources,
K–12 institutions can
take a small number
of steps to
significantly reduce
cybersecurity risk.
Invest in the most impactful security measures
and build a mature cybersecurity plan by
taking these three steps:
• Implement highest priority security controls.
• Prioritize further near-term investments in
alignment with the full list of CISA’s Cross-
Sector Cybersecurity Performance Goals
(CPGs).
• Over the long-term, develop a unique
cybersecurity plan that leverages the NIST
Cybersecurity Framework (CSF).
2
Many school districts
struggle with
insufficient IT
resources and
cybersecurity
capacity.
Recognize and actively address resource
constraints: Work with the state planning
committee to leverage the State and Local
Cybersecurity Grant Program (SLCGP). •
Utilize free or low-cost services to make near-
term improvements in resource-constrained
environments. • Expect and call for technology
providers to enable strong security controls by
default for no additional charge. • Minimize
the burden of security by migrating IT services
to more secure cloud versions.
3
No K–12 entity can
singlehandedly
identify and
prioritize emerging
threats,
vulnerabilities, and
risks.
Focus on collaboration and information
sharing:
• Join relevant collaboration groups, such as
MS-ISAC and K12 SIX.
• Work with other information-sharing
organizations, such as fusion centers, state
school safety centers, other state and regional
agencies, and associations.
• Build a strong and enduring relationship with
CISA and FBI regional cybersecurity
personnel.
Alongside cybersecurity issues, the proliferation of digital
content and the ever-improving access to it have heightened pupils'
vulnerability to a broader spectrum of online dangers and unsuitable
material. Consequently, the necessity for cybersecurity and school
child cybersecurity has become imperative. Given the evolving
digital landscape, it is crucial to implement safeguards to protect
kids from online risks, including cyberbullying, improper content,
sexting, sextortion, and online predation. In this context [5]
INNOVATIONS 2025
18
proposes a novel education-specific K-12 Cyber Protection
Framework (CPF) that provides industry-led cybersecurity and
cybersecurity standards.
Other current thorough research analyzes the fundamental
principles and procedures involved in the formulation and execution
of information security management policies inside educational
institutions, with a particular focus on "Dr. Ivan Bogorov"
Vocational High School of Economics in Varna. The development
of an information security policy within the framework of rules and
international standards is discussed. The research presents an
overview of the primary steps for implementing information
security policies in educational settings, specifically within "Dr.
Ivan Bogorov "Vocational High School of Economics, [6].
An intriguing study was undertaken from April 2024 to
September 2024 with the participation of 56 Bulgarian citizens
employed in higher education. The study aimed to evaluate the
state of information security in higher education institutions by
collecting the views, attitudes, and perceptions of professionals in
the field. The queries were classified according to three
fundamental principles: demographic characteristics, occupational
features, and individual viewpoints. The main finding is that
respondents feel secure regarding the performance of their work
duties and the protection of personal data in an information security
environment, [7].
3. Data and Methodology
In order to have a better knowledge of how to enhance security
and make schools more resistant to cyber-attacks, the first logical
step is to investigate the comprehension of individuals in
educational institutions with regards to cybersecurity.
The survey was conducted online from March 31, 2024, to May
3, 2025, involving a poll of 927 school principals, teaching staff,
administrative and other staff from Bulgarian K-12 institutions.
This study aimed to compare individual differences, evaluate the
availability of systematic cybersecurity training, analyze decision-
making styles, and assess behavioral intentions related to
cybersecurity for the protection of educational and administrative
platforms, software, hardware, and computer networks, along with
proactive awareness and knowledge enhancement. The survey was
designed and distributed via Microsoft Forms within the internal
Microsoft 365 environment of the Bulgarian Ministry of Education.
The survey consisted of 22 questions designed to assess
individuals' understanding of cybersecurity within their K-12
educational institutions and to gather their viewpoints on the impact
of cybersecurity risks on their daily tasks. No incentives were given
out, and participation was completely voluntary and anonymous.
Table 2, Table 3 and Table 4 illustrate the structure of the
schools involved in the study.
Table 2: Distribution of survey-participating schools by type and kind of
ownership.
type and kind of ownership
In numbers
percentage of
respondents
State
317
34.20%
Municipal
592
63.86%
Private
4
0.43%
(blank)
14.00
1.51%
Table 3: Distribution of survey-participating schools by category.
School category
In numbers
percentage of
respondents
Gymnasiums
14
2%
Primary school
31
3%
Integrated school
20
2%
Basic school
321
35%
Profiled gymnasiums
114
12%
Vocational gymnasiums
22
2%
Secondary school
400
43%
(blank)
5
1%
Table 4: Distribution of survey-participating schools by size.
School size
In numbers
percentage of
respondents
With up to 50 students
23
2%
Between 50 and 100 students
98
11%
Between 100 and 200 students
134
14%
Between 200 and 500 students
329
35%
Over 500
332
36%
(blank)
11
1%
This structure of the schools involved in the study gives us
reason to believe that the results obtained from our survey are
representative only of the different types and sizes of public K-12
schools. Only four private schools participated in the study,
rendering the data unrepresentative of this group of K-12
institutions.
Table 5 illustrates the distribution of survey participants by
position.
Table 5: Distribution of survey participants by position.
Participants’ position
In numbers
percentage of
respondents
Teaching staff
750
80.91%
Director
103
11.11%
Deputy director
43
4.64%
Administrative and other staff
27
2.91%
(blank)
4
0.43%
The distribution of survey participants by their positions
suggests that the results are indicative of the many roles within K-
12 institutions.
Table 6 illustrates the distribution of survey participants by age.
Table 6: Distribution of survey participants by years old.
Participants’ age (years old)
In numbers
percentage of
respondents
Up to 25 years old
23
2.48%
Between 24 and 40 years old
210
22.65%
Between 40 and 60 years old
611
65.91%
Over 60
79
8.52%
(blank)
4
0.43%
The age distribution of survey participants corroborates the
trend of an aging teaching workforce in Bulgarian K-12 institutions.
4. Key Results
Based on the assessments of the K–12 educational institutions
surveyed, only 3.34% indicate that they have faced cybersecurity
incidents in the past five years. The results are presented in Table 7.
Table 7: Has your school had any cybersecurity incidents in the past five
years?
Answer
In numbers
percentage of
respondents
Yes
31
3.34%
No
884
95.36%
(blank)
12
1.29%
For comparison, according to data from the USA, [8], 82% of
K-12 organizations experienced cyber incidents, nearly 14,000
security events were observed, over 9,300 are confirmed incidents.
Cyber threat actors target human behavior 45% more often than
technical vulnerabilities. We believe that the significant difference
in reported cybersecurity incidents between our data and the US
data comes from the absence of established and easy-to-follow
procedures for reporting every cybersecurity incident in K–12
educational institutions. In many cases, prevalent cybersecurity
incidents like spam, fishing and ransomware often go unreported.
INNOVATIONS 2025
19
This observation is corroborated by our survey data, which
reveals a significant lack of personnel accountable for cybersecurity
in K-12 schools. Only 5.39% of the respondents who participated in
the survey indicated that their school had personnel accountable for
cybersecurity, (See Table 8).
Table 8: Do you think it is necessary to have a separate position for
cybersecurity activities in your school?
Answer
In numbers
percentage of
respondents
There is a position in the school
that handles cybersecurity
activities.
50
5.39%
Yes, it is important to have such
a position.
235
25.35%
I can't say.
361
38.94%
Not really
215
23.19%
Definitely NOT.
50
5.39%
(blank)
16
1.73%
At the same time, the majority of respondents, (48.87%), are
unable to assess the state of cybersecurity at their institution, (See
Table 9). The positive aspect was that a very small percentage of
answers identified significant cybersecurity deficiencies in their K-
12 institutions.
Table 9: What do you thinking about the level of cybersecurity at your
school?
Answer
In numbers
percentage of
respondents
Very low
16
1.73%
Relatively low
100
10.79%
Can't tell
453
48.87%
Relatively high
293
31.61%
Very high
61
6.58%
(blank)
4
0.43%
A considerable number of respondents in our poll indicated
substantial resources and personnel challenges. The most commonly
identified essential components of cybersecurity in K-12 schools are
cybersecurity training (staff training, selected by 24.77% of
respondents, student training, selected by 20.43% of respondents,
parent training, selected by 15.61% of respondents), and
technological tools and software, selected by 26% of respondents,
(See Table 10). At the same time, а very small percentage of
respondents ( 8.00%) refer to the presence of regulations governing
cybersecurity as a significant factor.
Table 10: In your experience, what are the most crucial parts of
cybersecurity in your school?
Answer
In numbers
percentage of
respondents*
Technical tools and software for
cybersecurity
480
26.02%
Staff training
457
24.77%
Student training
377
20.43%
Parent training
288
15.61%
Regulations
148
8.02%
Other
95
5.15%
(blank)
37
3.99%
* The cumulative percentages surpass 100% as participants
were permitted to select several answers.
Despite the recognition of cybersecurity training as a critical
element, only 35% responded positively to the question, "Have you
participated in any cybersecurity-related training in the past five
years?". Additionally, merely 12.73% of respondents confirmed that
their K-12 institutions provide annual cybersecurity training, and
11.76% of respondents confirmed that their K-12 institutions
provide a rather general cybersecurity training, (See Table 11).
Table 11: Do you expect that your qualification plans will include
cybersecurity and computer security training?
Answer
In numbers
percentage of
respondents*
Not planned
143
15.43%
Somewhat unlikely
127
13.70%
Can't tell
368
39.70%
Somewhat likely
109
11.76%
Annually
118
12.73%
(blank)
62
6.69%
5. Conclusions
Our survey of 927 Bulgarian K–12 educational institutions
indicates that a unsignificant number of workers employed within,
define the level of cybersecurity in their institutions as very low.
They believe that cybersecurity training, technological instruments,
and software are essential components for maintaining information
security within their institutions. Simultaneously, our research
indicates that fewer than ¼ of those employed in K-12 institutions
can rely on such training as an integral component of their
qualification plans. We deem it necessary to do further our
investigation into the remarkably low number of cybersecurity
events observed in our study.
References
[
1]
R. Kaur, D. Gabrijelčič, T. Klobučar, "Artificial intelligence
for cybersecurity: Literature review and future research
directions," Information Fusion, vol. 97, p. 101804, (2023).
[
2]
R. Yanev, "Cyber security protection in Bulgaria," Yearbook
- Higher School of Security and Economics, no. 21, pp. 47-55,
(2025).
[
3]
Y. Yanakiev, D. Polimirova, "Exploring the Role of the
Human Factor in Cybersecurity: Results from an Expert Survey
in Bulgaria," Information & Security: An International Journal,
vol. 44, pp. 39-50, (2020).
[
4]
"Protecting our future: partnering to safeguard k–12
organizations from cybersecurity threat," Cybersecurity and
infrastructure security agency, U.S. department of homeland
security, (2023).
[
5]
M. Kamaludeen., S. Ismaeel., S. Asiri., T. Allen , C. Scarfo,
"A Framework for Cyber Protection (FCP) in K-12 Education
Sector," in IET Conference Proceedings, (2020).
[
6]
D. Vasilev, "Information security management in
educational institutions: policies, procedures and good
practices," in Conferences of the department Informatics, Varna,
(2024).
[
7]
E. Angelova, "Information Security in Higher Education
Institutions," in Knowledge, Science, Innovation, Technology,
(2024).
[
8]
Center for Internet Security, Inc.®(CIS) and Multi-State
Information Sharing andAnalysis Center® (MS-ISAC®), "An
18-Month, Retrospective Study of Cyber Threat Trends and
Defensive Impact in K-12 Education," Center for Internet
Security, in partnership with Consortium for School
Networking, (2025).
INNOVATIONS 2025
20
The Evolution, Current Impact and Future of Artificial Intelligence in Medicine
Martin Čillík1,2*, Ivan Kuric2, Miriam Kuricová1
National Institute of Endocrinology and Diabetology, Ľubochňa, Slovakia1
University of Žilina, Faculty of Mechanical Engineering, Žilina, Slovakia2
cillik@nedu.sk
Abstract: The application of artificial intelligence (AI) in medicine has emerged as a topic of global interest. Since the introduction of the
term in the mid of 20th century, there has been considerable progress in the development of computer systems, leading into their integration
into healthcare. At present, AI is increasingly being adopted across a growing number of medical specialties, where it contributes not only to
diagnostic processes but also to the selection of appropriate treatments and the prediction of patient outcomes. AI prediction is particularly
useful in the management of chronic conditions. Furthermore, AI demonstrates significant potential to expedite routine procedures, thereby
allowing healthcare professionals to dedicate more time to cognitively demanding tasks, in which AI systems continue to present certain
limitations. Nevertheless, despite notable advancements in recent years, several challenges must still be addressed in future research. These
include the formulation of standards and guidelines for AI implementation, the assurance of cybersecurity to safeguard sensitive data, and
the continuous education and training of healthcare practitioners. In conclusion, AI holds considerable promise for enhancing the quality
and efficiency of healthcare delivery. Its role is not to replace human professionals, but rather to augment their performance and optimize the
use of their time and expertise.
Keywords: ARTIFICIAL INTELLIGENCE, MEDICINE, CLINICAL PRACISE
1 Introduction, Historical Background
The terminology associated with the science of artificial
intelligence (AI) refers to systems designed to simulate human
cognitive behaviors. More specifically, AI involves training
computers to replicate certain human abilities, such as learning,
judgment, and decision-making. The term "artificial intelligence"
was first introduced by John McCarthy in the mid-20th century, a
period that is commonly referred to as the "Founding Period" of AI
[1, 2]. During this period, and particularly throughout the 1940s and
1950s, the development of computing technology facilitated the
emergence of early AI systems. Subsequently, several universities
established dedicated AI laboratories, largely supported by
government funding. Consequently, expert systems experienced
rapid development, yielding significant benefits. Nevertheless, this
progress was accompanied by notable challenges. One such issue
was the difficulty of acquiring and encoding expert knowledge into
these systems. The diagram (Fig. 1) shows some of the milestones
of AI development.
This phase of AI development is often characterized as the
"First Golden Period.” [1] Owing to these technological
advancements, the field of medicine quickly recognized the
potential applications of AI. [2, 3] Its primary advantage was its
capacity to assist physicians by managing substantial workloads and
enhancing the accuracy of diagnostic assessments. Furthermore,
there was growing optimism that AI could eventually replicate
certain intellectual functions of medical professionals. However, by
the end of the 1970s, it became evident that the field had not
achieved the level of success that had initially been anticipated,
particularly in relation to rule-based systems and pattern
recognition. One of the primary reasons for this shortfall was the
absence of pathophysiological knowledge in the design and
implementation of AI systems. This limitation hindered the systems’
ability to effectively model and interpret complex medical data [1].
Following the introduction of the Hopfield neural network
algorithm in 1982, a significant surge in system development
ensued. This era, often referred to as the "Second Golden Period,"
was primarily marked by advances in speech translation and speech
recognition systems. Nevertheless, despite these technological
achievements, AI during this time remained largely absent from
everyday social life [1]. By the late 1980s, AI systems had begun to
find applications in various commercial tasks within the healthcare
sector. However, they had not yet been integrated into clinical
medicine [1]. In a notable publication, William B. Schwartz posited
that computer science would eventually assume roles traditionally
associated with intellectual functions. He predicted that by the year
2000, computers would play a significantly new role in the medical
field, acting in support of physicians' cognitive processes [2,5].
Fig. 1 A short way through evolution of AI [5, 6]
INNOVATIONS 2025
21
Between the 1990s and the early 2000s, the development of
AI continued to progress steadily. This advancement was driven by
substantial financial investments and significant intellectual effort.
As a result, computers became capable of performing several
medical tasks, including the automated interpretation of white blood
cell differential counts, electrocardiograms, cutaneous lesions, and
retinal images. Furthermore, numerous medical image-processing
applications were gradually implemented in clinical practice. Even
though many of these AI-driven tools have been incorporated into
routine medical workflows, they still require skilled professionals to
interpret and validate the results. [2] The " Third Golden Period" of
AI, beginning in 2006 and continuing to the present day, has been
marked by a rapid acceleration in development. This surge is largely
attributable to the growing use of Graphics Processing Units
(GPUs), which significantly enhanced computational power and
efficiency across various domains. [1] Consequently, AI and
machine learning (ML) programs have become increasingly
integrated into medicine. They are now used to identify outbreaks of
infectious diseases, and to combine clinical, genetic, laboratory, and
other data sources to improve the detection of both common and
rare medical conditions, many of which might otherwise go
unnoticed. Nevertheless, beyond diagnostics, AI also contributes
daily to a wide range of technical and administrative operations
within hospitals and healthcare systems. [2]
2 Role of AI and Machine Learning in Clinical
Practice
In the context of medical diagnostics, artificial intelligence,
particularly machine learning ML and deep learning (DL), has
shown significant promise in transforming diagnostic processes
across a wide range of medical fields. These technologies have
contributed to improved accuracy in disease detection within
several medical specialties. [6]
2.1 Dermatology
Deep learning algorithms have demonstrated a remarkably
high level of accuracy in predicting skin malignancies within the
field of dermatology. One such algorithm, trained on over 200,000
images representing 174 distinct diagnoses, not only excelled in
diagnostic classification but also in recommending primary
treatment options. Consequently, this contributed to a significant
enhancement in the performance of medical professionals,
increasing the sensitivity of malignancy prediction by more than
83%. Algorithms of this kind thus serve as powerful tools that
enhance the capabilities of dermatologists. [7] Moreover, another
algorithm surpassed the diagnostic accuracy of nearly all 58
dermatologists in a validation study using dermatoscopic images,
further underscoring the transformative potential of DL in clinical
dermatology. [8]
2.2 Radiology
The use of AI in radiology has proven to be highly efficient
across various domains. Radiology plays a crucial role in supporting
other medical specialties by contributing to accurate diagnosis and
facilitating the monitoring of chronic disease progression.
Moreover, determining prognosis based on radiographic changes is
particularly important in the long-term management of chronic
illnesses. In the fields of gynecology and oncology, modern AI-
powered image interpretation systems have demonstrated
unprecedented accuracy in reading mammograms. For instance, a
study in the United Kingdom utilized a training set comprising over
25,000 image datasets from multiple screening centers. This system
was tested, and its performance compared with the diagnostic
assessments of six radiologists, with final confirmation based on
histological findings. Notably, the AI system achieved a 10%
reduction in both false-negative and false-positive results. [9]
Similarly, another system trained on more than 170,000
mammographic images from institutions in South Korea, the United
States, and the United Kingdom also outperformed radiologists in
diagnostic accuracy. [10] Furthermore, AI has shown remarkable
effectiveness in detecting pneumonia-related changes in
radiographic images, achieving a sensitivity greater than 95% and a
specificity of approximately 60%. These findings underscore the
growing potential of AI to enhance diagnostic precision and
efficiency in radiology. [11]
2.3 Ophthalmology, Diabetic Retinopathy (DR)
Screening
Various forms of AI, particularly convolutional neural
networks (CNNs), have proven highly effective in image-based
medical specialties, with ophthalmology being one of the most
prominent. AI applications are widely utilized in DR screening
through fundus photography and optical coherence tomography
(OCT) analysis, as well as in glaucoma detection and other
ophthalmic diagnoses. Notably, AI-based software such as IDx-DR
and EyeArt, both of which are approved by the U.S. Food and Drug
Administration (FDA), are commonly used for DR screening. These
systems offer significant advantages, including high diagnostic
accuracy and efficiency, reduced workload for ophthalmologists,
and improved accessibility for patients. Furthermore, their ability to
automatically learn relevant features from large datasets enhances
their diagnostic potential. A Figure 2 shows a smartphone-based
fundus imaging method that uses AI-enabled software to detect
retinal vessels abnormalities. Nevertheless, several limitations and
future challenges remain. These include the need for larger and
more diverse datasets, the absence of standardized classification
systems, and unresolved issues related to liability and accountability
in cases of misdiagnosis. [12, 13]
Fig. 2 Fundus image captured using a smartphone with 20 diopter condensing [13]
INNOVATIONS 2025
22
2.4 Cardiology
A novel approach in the application of AI within cardiology
is centered on the detection of arrhythmias, particularly using Echo
State Networks for electrocardiogram (ECG) analysis across
multiple databases. In one study, the system achieved a sensitivity
of 92.7% for detecting ventricular ectopic beats using lead II, and
an even higher sensitivity of 95.7% when using lead V1. Among the
primary advantages of this system are its rapid computation time,
simplified feature extraction, suitability for online processing,
effective single-lead classification, and robust generalization
capability. [14] Another significant area of AI implementation in
cardiology involves the detection and prediction of atrial fibrillation
(AF). A study published in 2021 demonstrated the potential of AI to
predict new-onset AF using data from 12-lead ECGs, even in
patients without a prior history of the condition. This advancement
offers promising prospects for the early identification of high-risk
individuals and the implementation of preventive strategies. [15]
2.5 Microbiology
In recent years, several ML-based systems have been
developed not only for the identification of microorganisms but also
for antibiotic susceptibility testing. [16] These advancements have
significantly enhanced diagnostic accuracy and facilitated more
effective treatment selection, which is particularly beneficial in
combating multi-drug-resistant bacteria. AI contributes to clinical
diagnostics by enabling automated image and spectral analysis,
supporting disease management in conditions such as sepsis, and
improving infection surveillance and control strategies.
Nevertheless, several challenges persist in the implementation of AI
in this domain. Key concerns include the quality and heterogeneity
of data, the interpretability of algorithmic outputs, seamless
integration into clinical workflows, efficient data management, and
a range of ethical considerations. Addressing these issues will be
essential for the responsible and effective deployment of AI in
infectious disease diagnostics and treatment planning. [17]
2.6 Drug Development and Clinical Trials
AI has also demonstrated significant potential in the field of
drug discovery and development. It plays a key role in designing
novel drug compounds and molecular structures, leveraging ML and
DL to analyze vast datasets with the aim of predicting drug efficacy
and toxicity. Moreover, AI supports virtual screening, quantitative
structure-activity relationship analysis, and the prediction of drug-
target interactions, thereby streamlining early-phase research. It also
introduces innovative approaches to quality control within
pharmaceutical development. A major advantage of AI in this
context is its ability to accelerate the overall drug discovery process,
potentially reducing both time and cost. Furthermore, AI contributes
to the emerging field of targeted drug delivery, such as the control
of nanorobots. In nanomedicine, AI aids in the formulation and
development of therapies by enhancing drug efficacy and ensuring
consistent quality. Additionally, in pharmaceutical product
management, AI enables the creation of effective marketing
strategies, analyzing market trends, predicting sales, understanding
customer demands, and determining optimal pricing models.
However, despite its benefits, the use of AI in drug development
faces several limitations. These include the need for large and
diverse datasets, ethical and regulatory concerns, and the fact that
AI cannot fully substitute human expertise or experimental
validation. Looking ahead, close collaboration between AI
researchers and pharmaceutical scientists will be essential for
maximizing the potential of these technologies. [18] In the broader
research context, AI enhances patient selection processes, improves
clinical trial success rates, reduces failures, and aids in identifying
promising lead compounds. By analyzing patient-specific data, it
contributes to personalized treatment approaches. Furthermore, AI-
powered monitoring systems improve patient adherence and reduce
dropout rates during clinical trials. Despite ongoing challenges, AI
is poised to become an indispensable tool in the future of drug
development, clinical research, and pharmaceutical management.
[19]
2.7 AI in the Emergency Department (ED) and
Hospital Management
In the context of ED, AI offers several valuable
applications. One of its primary contributions is in triage assistance,
where it helps assess the severity of cases and categorize patients
accordingly. Another important advantage is AI’s ability to predict
patient outcomes, which supports more effective allocation of
medical resources. Furthermore, AI enhances diagnostic accuracy
through the interpretation of imaging studies such as CT scans,
MRIs, and X-rays. By rapidly identifying critical findings, AI
facilitates faster diagnosis and timely initiation of treatment. As a
result, the overall efficiency of emergency care systems can be
significantly improved, ultimately leading to better patient
outcomes. However, for these systems to be fully optimized, further
validation of training datasets and careful consideration of
implementation costs are essential for future advancements. [20]
Beyond the emergency setting, AI presents numerous opportunities
in hospital management. It can predict equipment failures in
advance, thereby minimizing downtime and reducing maintenance
costs. In addition, AI can optimize patient flow and resource
allocation across departments. Within human resource management,
AI systems support automated scheduling, effectively balancing
staff availability with patient demand. Clinical decision support
systems, powered by AI, further enhance diagnostic accuracy and
improve clinical efficiency. Nevertheless, the risk of diagnostic
errors remains a significant concern for medical practitioners, as
such mistakes can have severe, even fatal, consequences. [5] AI has
the potential to mitigate this risk by providing real-time decision
support, including diagnostic suggestions and treatment
recommendations. Collectively, these capabilities contribute to
increased efficiency in daily clinical operations and a more
responsive healthcare system. [21]
2.8 AI role in Precision Medicine and Individual
Treatment Personalization
Precision medicine, also referred to as personalized
medicine, represents a modern approach to healthcare that aims to
tailor medical treatment to individual patients based on their unique
characteristics. These characteristics encompass not only genetic
and biomarker profiles but also environmental factors and lifestyle
choices. The overarching goal is to enhance the effectiveness of
medical interventions while simultaneously increasing their safety
and efficiency. Within this context, AI plays a critical role by
predicting treatment outcomes and informing adaptive therapeutic
strategies. [5, 22]
For instance, in one study, researchers developed a machine
learning algorithm using Support Vector Machines to predict
individual responses to chemotherapy based on gene expression
profiles of tumors in oncology patients. The model was trained and
validated using data from The Cancer Genome Atlas. In a test group
of 23 female patients with ovarian cancer treated with eight
different drugs, the algorithm achieved an accuracy of 83% and a
positive predictive value of 85%. These results suggest that such AI-
based tools could be particularly valuable in guiding treatment
decisions for patients who do not respond to initial therapies. [23]
Similarly, another research team developed a predictive system to
determine whether adult patients with depression would respond to
INNOVATIONS 2025
23
one of four commonly prescribed first-line antidepressant classes:
selective serotonin reuptake inhibitors, so called SSRIs, serotonin-
norepinephrine reuptake inhibitors (SNRIs), mirtazapine, or
bupropion. The system utilized large-scale electronic health records
and various machine learning models, including deep neural
networks. The best-performing models achieved approximately
70% accuracy. In addition to predicting individual treatment
response, the system offered the added benefit of estimating patient
responses to alternative antidepressants, by identifying the most
influential factors driving the predictions. These advances highlight
the transformative potential of AI in precision medicine, offering
more individualized, effective, and data-driven healthcare strategies
for a wide range of clinical conditions. [24]
2.9 AI Medical Assistants
AI-based medical assistants represent a rapidly evolving
area of application, particularly within lifestyle and preventive
medicine. These systems are increasingly being used to support
home-based patient monitoring, improve adherence to treatment
protocols, and guide lifestyle-related decisions such as nutrition and
physical activity. Some applications have received approval from
regulatory bodies such as FDA for use in remote patient monitoring.
For instance, the Onduo platform integrates continuous glucose
monitoring data, food recognition, and physical activity tracking to
deliver personalized lifestyle recommendations for individuals with
type 2 diabetes. Other recently developed startups like Virta,
Wellpepper, and Accolade, also focus on various aspects of home-
based care. The DayTwo platform provides individualized dietary
recommendations aimed at supporting a healthy gut microbiome.
[25] In cardiology, wearable technologies such as smartwatches are
increasingly used for remote monitoring, especially in detecting AF
and other arrhythmias. The AliveCor system utilizes a deep learning
algorithm combined with single-lead ECG data and accelerometer-
based activity tracking to monitor cardiac rhythms. [26] Similarly,
ResApp provides a respiratory monitoring system capable of
evaluating chronic lung conditions, including asthma, chronic
obstructive pulmonary disease, and pneumonia [27]. Despite their
promise, these systems face significant challenges, particularly in
the integration and analysis of multidisciplinary data. Effective
performance relies on the systems’ ability to synthesize data from
various sources and contexts to provide accurate and meaningful
recommendations. Thus, improvements in data interoperability and
model training remain critical to enhancing precision and clinical
utility. Nevertheless, significant progress has been made in recent
years, indicating a promising future for these technologies. [22]
AI is also making inroads into the field of mental health
support. AI-powered tools are being developed to complement the
work of psychiatric professionals by facilitating early detection of
mental health conditions and offering personalized treatment
approaches. Several studies have demonstrated the effectiveness of
web-based cognitive behavioral therapy as a psychotherapeutic
intervention. For example, programs such as the Sadness program
and the text-based conversational agent Woebot have shown
promising results in reducing depressive symptoms, promoting
remission, and accelerating recovery. [28, 29] However, while these
tools offer valuable support, they may lack the ability to capture the
full complexity and individual variability inherent in mental health
conditions. This limitation may result in reduced personalization
and an absence of emotional empathy. Despite this, AI-powered
therapy systems have proven to be effective supplemental tools
when used alongside care provided by healthcare professionals. [5]
2.10 AI Use in the Field of Public Health
Through population health-based demographic analyses, AI
can identify clusters of risk factors that, when combined,
significantly increase the likelihood of developing certain diseases.
ML and other advanced algorithms are currently employed in the
development of predictive models aimed at enhancing patient
outcomes and reducing healthcare costs. In the context of primary
prevention, AI plays a crucial role in identifying high-risk groups,
thereby allowing for targeted interventions before disease onset. In
secondary prevention, it is particularly valuable in predicting and
preventing hospital readmissions, ultimately improving care
continuity and resource efficiency. However, a major limitation in
this field remains the quality of the underlying data. Predictive
analytics require large volumes of high-quality, standardized, and
comprehensive datasets to generate accurate predictions. Without
such data, the effectiveness of AI-driven interventions may be
compromised, potentially leading to suboptimal healthcare
decisions. Therefore, ensuring robust data infrastructure is essential
for realizing the full potential of AI in population health
management. [5, 30]
2.11 Chatbots in Medicine
The emergence of conversational agents, particularly
ChatGPT developed by OpenAI and launched in late 2022, has
generated considerable attention across various domains, including
education, research, and healthcare. In the medical field, ChatGPT
has sparked significant interest regarding its potential to improve
patient care, enhance quality of life, and streamline healthcare
delivery. Its current applications span a wide range of functions,
including support for clinical decision-making, medical education,
therapeutic communication, and administrative tasks. Notably,
studies have demonstrated its effectiveness in analyzing complex
medical data, aiding in the formulation of differential diagnoses,
and even interpreting medical images for surgical planning and
diagnostics. [31] Furthermore, ChatGPT has been explored as a
tool to assist with documentation and triage, potentially improving
workflow efficiency. [32] Its integration into clinical practice is not
without challenges. Key concerns include risks of research fraud,
lack of originality in generated content, ethical and copyright issues,
and unresolved legal considerations. Moreover, there is concern that
excessive reliance on AI-generated outputs may hinder critical
thinking and creative problem-solving among healthcare
professionals and researchers. Although ChatGPT presents
promising opportunities, particularly in addressing disparities in
access to healthcare resources, its use in real-world clinical
environments remains limited due to concerns about precision and
accountability. In medicine, even minor errors can have serious,
potentially irreversible consequences, raising important questions
regarding responsibility and liability. [31] Another critical limitation
is the system’s lack of specificity. According to one study, while a
chatbot could provide a general diagnosis and comment on different
management options, it failed to identify key indicators that
determine the appropriateness of each option. Although AI chatbots
such as ChatGPT may assist in medical decision-making, they are
currently unable to deliver fully personalized recommendations
based on a patient’s unique clinical profile. [33] While the potential
of ChatGPT in medicine is significant, further research and
development are essential to refine these systems. Enhanced
capabilities in data interpretation, clinical reasoning, and
personalized response generation are necessary for chatbots to
handle complex medical inquiries effectively and safely within
clinical settings. [31]
INNOVATIONS 2025
24
3 Further Prospects and Challenges
Looking ahead, AI holds substantial promise to revolutionize
healthcare across a wide array of disciplines. Although different
medical fields will be impacted to varying degrees, AI is expected
to permeate virtually every area of clinical practice. Currently, the
analysis of large datasets, including medical imaging and genomic
information, has already led to improvements in diagnostic
accuracy, clinical efficiency, and the development of personalized
treatment strategies, particularly in specialties such as oncology,
radiology, and pathology. In oncology, future advancements in AI
algorithms are expected to play a critical role in predicting
treatment outcomes and supporting therapeutic decision-making.
[34] Similarly, in radiology, the ongoing AI revolution is evidenced
by the development of systems capable of detecting cervical lymph
node metastases, identifying pulmonary nodules from computed
tomography-CT scans, and automating image interpretation
processes. [35, 36] DNNs have shown potential to unify related
medical specialties through integrated diagnostic capabilities. Those
specialties including radiology, nuclear medicine, and surgical
pathology. A notable example is a glioma grading system based on
magnetic resonance imaging-MRI, which has achieved a
classification accuracy of 92.86%. [37] The long-term goal is to
harness AI to recognize the unique biological and clinical
characteristics of individual patients, thereby enabling a truly
personalized approach to healthcare. However, in fields such as
pathology and surgical pathology, implementation remains
dependent on the availability of resources, including updated
workflows, trained personnel, and sufficient data storage capacity.
Nonetheless, current AI systems analyzing histological images, such
as those used in lung cancer diagnostics, are already demonstrating
improvements in diagnostic precision and in prognosis
determination, an area often more challenging than initial diagnosis
itself. [38, 39] In the future, AI is expected to take ver repetitive
tasks in pathology, thus allowing pathologists to focus more on
high-level cognitive functions, such as formulating prognoses and
guiding complex therapeutic strategies. [34]
Despite these advancements, several obstacles remain. Major
concerns include issues related to data privacy, availability, and
security. Additionally, the propensity of AI models to generate
fabricated or inaccurate information, particularly when lacking
appropriate training data or oversight, raises questions about their
reliability in clinical contexts. [40] These limitations are particularly
pronounced in complex clinical scenarios, such as the diagnosis of
rare conditions, interpretation of multifaceted diagnostic tests, and
performance of surgical procedures. Another significant concern is
the lack of emotional intelligence and empathy, which are essential
components of mental health care and patient counseling. [41]
Finally, financial constraints continue to represent a significant
barrier to the widespread adoption of AI in healthcare, especially in
underfunded medical sectors and in low- and middle-income
countries. Overcoming these challenges will require not only
technological advancement, but also thoughtful policy development,
interdisciplinary collaboration, and equitable resource distribution.
4 Conclusion
The development and integration of AI into healthcare have
recently become hot topics of discussion. The incorporation of AI
into various medical disciplines has the potential to enhance
efficiency, precision, and accuracy not only in the diagnostic
process but also in therapeutic decision-making. AI can expedite
diagnosis and improve its accuracy, thereby allowing physicians to
devote more time to cognitive tasks, such as determining optimal
treatment strategies. This is especially valuable in the context of
rare diseases, where human expertise remains essential and where
AI-based recommendations still have notable limitations.
Nevertheless, for AI to be effectively and safely implemented in
routine clinical practice, several critical measures must be
addressed. First, collaboration between healthcare professionals and
researchers is vital to establish clear standards and regulatory
guidelines for AI algorithms and their decision-making processes.
Moreover, continued investment in advanced AI research is
necessary to ensure the development of models that are both reliable
and generalizable for broader clinical use. Equally important are the
ethical and legal dimensions, particularly regarding data security.
Comprehensive cybersecurity strategies and robust data protection
protocols must be established to safeguard patient information and
prevent potential breaches. Furthermore, the perception of AI in
healthcare varies widely among both patients and healthcare
providers. Therefore, time and effort will be required to foster trust
in AI-driven tools. Educational programs aimed at both medical
professionals and the public are essential to facilitate the successful
integration of AI systems and promote collaboration between
clinicians and technologies. Ultimately, while AI holds immense
promise for improving healthcare delivery and democratizing
access to expert knowledge, it is not intended to replace human
clinicians. Rather, it should be seen as a tool to augment their
capabilities. The healthcare system is increasingly recognizing this
potential and is investing significantly in AI technologies, with the
aim of transforming medical practice in the years to come.
Acknowledgment
This work was supported by project VEGA 1/0470/23 - “Research
into methods and means of implementing artificial intelligence in
automated quality control systems for products with volatile quality
parameters.”
5 References
1. Zhang, C.; Lu, Y.: Study on artificial intelligence: The
state of the art and prospects. Journal of Industrial
Information Integration, 2021, 23: 100224.
2. Haug, Ch. J.; DRAZEN, Jeffrey M. Artificial intelligence
and machine learning in clinical medicine, 2023. New
England Journal of Medicine, 2023, 388.13: 1201-1208.
3. Yu., K.-H.; Beam, A. L.; KOHANE, Isaac S. Artificial
intelligence in healthcare. Nature biomedical engineering,
2018, 2.10: 719-731.
4. Schwartz, W. B. Medicine and the computer: the promise
and problems of change. In: Use and impact of computers
in clinical medicine. New York, NY: Springer New York,
1970. p. 321-335.
5. Alowais, S. A., et al. Revolutionizing healthcare: the role
of artificial intelligence in clinical practice. BMC medical
education, 2023, 23.1: 689.
6. Gou., F., et al. Research on artificial-intelligence-assisted
medicine: a survey on medical artificial intelligence.
Diagnostics, 2024, 14.14: 1472.
7. Han, S. S., et al. Augmented intelligence dermatology:
deep neural networks empower medical professionals in
diagnosing skin cancer and predicting treatment options
for 134 skin disorders. Journal of Investigative
Dermatology, 2020, 140.9: 1753-1761.
8. Haensele, H. A., et al. Man against machine: diagnostic
performance of a deep learning convolutional neural
network for dermoscopic melanoma recognition in
comparison to 58 dermatologists. Annals of oncology,
2018, 29.8: 1836-1842.
9. Kim, H.-E., et al. Changes in cancer detection and false-
positive recall in mammography using artificial
INNOVATIONS 2025
25
intelligence: a retrospective, multireader study. The
Lancet Digital Health, 2020, 2.3: e138-e148.
10. Mckinney, S. M., et al.: International evaluation of an AI
system for breast cancer screening. Nature, 2020,
577.7788: 89-94.
11. Becker, J.: et al. Artificial intelligence-based detection of
pneumonia in chest radiographs. Diagnostics, 2022, 12.6:
1465.
12. Grzybowski, A., et al. Artificial intelligence for diabetic
retinopathy screening: a review. Eye, 2020, 34.3: 451-
460.
13. Vaughan, N.: Review of smartphone funduscopy for
diabetic retinopathy screening, Survey of Ophthalmology,
Volume 69, Issue 2, 2024, Pages 279-286, ISSN 0039-
6257,
14. Alfaras, M.; Soriano M.; Ortím, S.. A fast machine
learning model for ECG-based heartbeat classification
and arrhythmia detection. Frontiers in Physics, 2019, 7:
103.
15. Ragnuath, S., et. al. Deep neural networks can predict
new-onset atrial fibrillation from the 12-lead ECG and
help identify those at risk of atrial fibrillation–related
stroke. Circulation, 2021, 143.13: 1287-1298.
16. Peiffer-Samadia, N., et al.: Machine learning in the
clinical microbiology laboratory: has the time come for
routine practice? Clinical Microbiology and Infection,
2020, 26.10: 1300-1309.
17. Tsitou, V-M., et. Al.: Microbiology in the era of artificial
intelligence: transforming medical and pharmaceutical
microbiology. Biotechnology & Biotechnological
Equipment, 2024, 38.1: 2349587.
18. Blanco-Gonzales, A., et al.: The role of AI in drug
discovery: challenges, opportunities, and strategies.
Pharmaceuticals, 2023, 16.6: 891.
19. Mak, K.-K., Wong, Y.-H., Pichka, M. R.: Artificial
intelligence in drug discovery and development. Drug
discovery and evaluation: safety and pharmacokinetic
assays, 2024, 1461-1498.
20. Berlyand, Y., et al.: How artificial intelligence could
transform emergency department operations. The
American journal of emergency medicine, 2018, 36.8:
1515-1517.
21. Malik, A., Solaiman, B.: AI in hospital administration and
management: Ethical and legal implications. Research
handbook on health, AI and the Law, 2024, 20-40.
22. Subramanian, M., et. Al.: Precision medicine in the era of
artificial intelligence: implications in chronic disease
management. Journal of translational medicine, 2020, 18:
1-12.
23. Huang, C., et al.: Machine learning predicts individual
cancer patient responses to therapeutic drugs with high
accuracy. Scientific reports, 2018, 8.1: 16444.
24. Sheu, Y., et al.: AI-assisted prediction of differential
response to antidepressant classes using electronic health
records. NPJ Digital Medicine, 2023, 6.1: 73.
25. Mendes-Soares, H., et al.: Assessment of a personalized
approach to predicting postprandial glycemic responses to
food among individuals without diabetes. JAMA network
open, 2019, 2.2: e188102-e188102.
26. Hall, A., et al.: Effectiveness of a single lead AliveCor
electrocardiogram application for the screening of atrial
fibrillation: a systematic review. Medicine, 2020, 99.30:
e21388.
27. Potter, P., et al.: A prospective multicentre study testing
the diagnostic accuracy of an automated cough sound
centred analytic system for the identification of common
respiratory disorders in children. Respiratory research,
2019, 20: 1-10.
28. Luxton, D. D.: Artificial intelligence in psychological
practice: Current and future applications and implications.
Professional Psychology: Research and Practice, 2014,
45.5: 332.
29. Fitzoatrick, K. K.; Darcy, A.; Vierhile, M.: Delivering
cognitive behavior therapy to young adults with
symptoms of depression and anxiety using a fully
automated conversational agent (Woebot): a randomized
controlled trial. JMIR mental health, 2017, 4.2: e7785.
30. Amarashingham, R., et al.: Implementing electronic
health care predictive analytics: considerations and
challenges. Health affairs, 2014, 33.7: 1148-1154.
31. Chakraborty, Ch., et al.: Overview of Chatbots with
special emphasis on artificial intelligence-enabled
ChatGPT in medical science. Frontiers in artificial
intelligence, 2023, 6: 1237704.
32. Tan, S.; Xin X.; Wu, D.: ChatGPT in medicine: prospects
and challenges: a review article. International journal of
surgery, 2024, 110.6: 3701-3706.
33. Seth, I., et al.: Exploring the role of a large language
model on carpal tunnel syndrome management: an
observation study of ChatGPT. The Journal of Hand
Surgery, 2023, 48.10: 1025-1033.
34. Ahmad, Z., et al.: Artificial intelligence (AI) in medicine,
current applications and future role with special emphasis
on its potential and promise in pathology: present and
future impact, obstacles including costs and acceptance
among pathologists, practical and philosophical
considerations. A comprehensive review. Diagnostic
pathology, 2021, 16: 1-16.
35. Lee, J. H.; Ha, E. J.; Kim, J. H.: Application of deep
learning to the diagnosis of cervical lymph node
metastasis from thyroid cancer with CT. European
radiology, 2019, 29: 5452-5457.
36. Liu, X.: et al.: Clinical application of artificial intelligence
recognition technology in the diagnosis of stage T1 lung
cancer. Zhongguo Fei Ai Za Zhi, 2019, 22.5.
37. Kv, A. M.; Rajendran, V.R.; K, P. J.: Glioma tumor grade
identification using artificial intelligent techniques.
Journal of medical systems, 2019, 43: 1-12.
38. Heeke, S. et al.: The age of artificial intelligence in lung
cancer pathology: between hope, gloom and perspectives.
In: Annales de pathologie. 2019. p. 130-136.
39. Yu, K.-H., et al.: Predicting non-small cell lung cancer
prognosis by fully automated microscopic pathology
image features. Nature communications, 2016, 7.1:
12474.
40. Maynez, J., et al.: On faithfulness and factuality in
abstractive summarization. arXiv preprint
arXiv:2005.00661, 2020.
41. Altamini, I., et al.: Artificial intelligence (AI) chatbots in
medicine: a supplement, not a substitute. Cureus, 2023,
15.6.
INNOVATIONS 2025
26
Photophysical properties of some phthalocyanine derivatives using ultrafast spectroscopy
Ionut Radu Tigoianu1*, Anton Airinei1, Carmen Gherasim1, Natalia Simionescu1, Tamara Potlog2, Victor Suman2, Ion Lungu2, Giulio
Cerullo3, Stefano Dal Conte3, Edoardo Carraro3
1Petru Poni Institute of Macromolecular Chemistry, Physical Chemistry of Polymers laboratory, 41A Grigore Ghica Voda Alley, 700487
Iasi, Romania
2 Moldova State University, Laboratory of Organic/Inorganic Materials for Optoelectronics, Chisinau, Republic of Moldova
3Politecnico di Milano, Physics Department, Piazza Leonardo da Vinci 32, Milano, Italy
tigoianu.radu@icmpp.ro
Abstract: Transient absorption spectroscopy uses short laser pulses which can be up to tens of femtoseconds. These pulses can effectively
“freeze” the motion of the molecules at specific times during a reaction or a dynamic process, and by modification of the time delay between
the pump and probe pulses snapshots of the molecular system can be obtained at different stages of its evolution, providing insights into the
underlying dynamics. For this purpose, the nanosecond or femtosecond transient absorption techniques can be utilized.
In this presentation ultrafast spectroscopy (transient absorption) and emission (steady state and time-resolved fluorescence) spectroscopy
were applied for to study and characterize the photophysical properties of some phthalocyanine derivatives. For this, new compounds
(phthalocyanine derivatives) have been synthesized and the quantum yields and lifetimes of excited states were estimated. Also, to
demonstrate excited-state processes from the transient absorption map, we obtained a ground state bleaching band (GBS), absorption in
excited state (ESA) and at longer wavelength, stimulated emission (SE).
Keywords: TRANSIENT ABSORPTION, PHTHALOCYANINE DERIVATIVES, EXCITED STATE
1. Introduction
Phthalocyanine derivatives have been utilized extensively in
biomedical applications due to their unique physico-chemical
properties such as high thermal and chemical stability. The
introduction of metal ions in the molecular structure of
phthalocyanine as well as the peripheral substituents determines
significant changes in the photophysical and photochemical
behavior [1, 2]. Phthalocyanines are typically second-generation
photosensitizers utilized in the photodynamic therapy (PDT) for the
cancer treatment. For photodynamic therapy different components
are necessary such as photosensitizer drugs that accumulate in
tumor cells, light irradiation with wavelength corresponding to the
absorption band of the photosensitizer, and molecular oxygen [3].
Changing the central metal atoms and the positions of the peripheral
and non-peripheral substituents, a control of their optical and
electrochemical characteristics can be realized and the therapeutic
effect can be improved. The introduction of the hydrophilic
substituents to pyrrole rings enhanced their water solubility and
stability [4].
The advantages of the metallophthalocyanines are strong
absorption in red and near infrared ranges, high singlet oxygen
generation yield, low dark toxicity, selective accumulation in
tumors, fast vanishing from healthy tissues, good solubility in
aqueous solutions and high storage stability [5, 6].
The present work was centered on the combination of
photosensitizers of tetra- and octa- carboxy phthalocyanine type
with some nanomaterials that can improve photodynamic therapy
efficiency and eliminate its side effects as well.
2. Materials and methods
All solvents were purchased from Sigma-Aldrich and were of
spectrophotometric grade, being used as received. The synthesized
materials were characterized by nuclear magnetic resonance
spectroscopy, electronic absorption and emission spectra.
The electronic absorption spectra of all the solutions were collected
on a SPECORD 210Plus spectrometer (Analytik Jena, Germany).
An Edinburgh FS5 spectrometer (Edinburgh, Ltd., England) and a
Perkin Elmer LS55 luminescence spectrometer were applied to
determine the steady-state fluorescence spectra. Time-correlation
single photon counting system (FLS 980, Edinburgh Instruments,
England) was used to conduct the time-resolved photoluminescence
analysis by using a nanosecond diode laser at 375 nm as excitation
source. Experiments were carried out at room temperature in 10 nm
quartz cells. The absolute fluorescence quantum yield (Φ) was
determined by FLS980 integrated sphere using solutions having the
absorbance value bellow 0.1 and excitation wavelengths
corresponding to the absorption band maxima.
INNOVATIONS 2025
27
The “Long” setup for ultrafast transient absorption spectra
offers an average temporal resolution of 100 fs, an amplified
Ti:Sapph laser emitting 100 fs pulses at 800nm, with a repetition
rate of 2 kHz (fig. 1). It can be turned into several layouts across
ultrabroad bandwidths and ultralong temporal windows, ranging
from few hundreds of fs to 2 ns. We can use the fundamental laser
beam at 800 nm, we can frequency double the fundamental, second
harmonic generation (SHG) at 400nm or we can exploit a frequency
tunable NOPA with the excitation beam that can continuously cover
all the wavelengths from 480 nm to 1.6 μm, with a bandwidth of 10-
20nm.
Also, for the broadband probe, obtained by White Light
Generation (WLG) onto thin highly nonlinear plates. 800 nm
pumped Sapphire plate generates light in the range of 430-780 nm
(Visible) and 840-1200 nm (Near IR). In 800 nm pumped CaF2
crystals the WL can be extended to the UV region, down to 340 nm.
Figure 1. Setup of ultrafast transient absorption spectroscopy
General chemical formulae of ZnPc derivatives are presented
below.
Figure 2. General chemical structures of ZnPc (COOH)4 (a), ZnPc
(COOH)8 (b) and ZnPc(SH)4(c)
The synthesis of ZnPc(COOH)4 was performed from a mixture
consisting of trimellitic anhydride; Zn(CH3COOH)2·2H2O;
(NH4)6Mo7O24·4H2O; anhydrous Na2SO4; urea and 1-
bromonaphthalene, heated at (200-205) ℃ for 8 h under continuous
stirring. After 8 hours, the reaction mixture is cooled and treated
with methanol. The obtained suspension is filtered. The solid
reaction product is washed on the filter with methanol, chloroform
and finally with acetone. Further after drying, it was refluxed for
one hour in 5% HCl solution. After drying, the same procedure was
carried out with 5% NaOH solution for one hour at 90℃. The basic
solution is acidified with HCl until a pH of 2 and finally a
precipitate product is reached. This product settles, filters and dries
in the open air. A quantity of 0.68 g of ZnPc(COOH)4 product is
obtained, with a yield of 70%. The following characteristic
vibrations (cm-1) were observed in the FTIR spectra: 3300 (νO-H);
3103, 3078, 3042 (νC-H); 1710 (νC=O); 1654, 1590 (νC-C); 1484,
1446, 1284, 1173 (νC-O); 1228, 1088 (νC-H); 1004, 880, 770, 705,
570, 499, 435.
3. Results and discussion
Transient absorption (TA) spectroscopy, steady state, and time-
resolved fluorescence spectroscopy were used to investigate and
characterize the tetra- and octa- carboxy zinc phthalocyanine
derivatives with applications in medicine. For this purpose, the new
compounds were been synthesized in order to have a good
emission, stability and sensitivity. We choose the investigation of
some phthalocyanine derivatives for the theoretical information and
applications resulting from this study, and to develop new
technological innovations for the health sciences, a new pathway for
the treatment of cancer.
In this paper, the synthesis, photophysical and biological
characterization of ZnPc(COOH)4, ZnPc(COOH)8 and ZnPc(SH)4
are displayed. The unsubstituted metallphthalocyanines (MPcs)
have very low solubility in the majority of organic solvents, and the
formation of substituted ZnPc is necessary to improve solubility in
the manufacturing process.
The optical properties of the compounds under study were
assessed by absorption and emission spectra (figs. 3-5),
fluorescence lifetimes (fig. 6) and fluorescence quantum yield
determinations.
INNOVATIONS 2025
28
Figure 3. Absorption spectra of ZnPc (COOH)4, ZnPc (COOH) 8
and ZnPc(SH)4
Figure 4. Absorption and emission spectra of ZnPc (COOH) 8
Figure 5. Fluorescence spectra of ZnPc (COOH)4 , ZnPc (COOH)8
and ZnPc(SH)4
Figure 6. Fluorescence (a) and phosphorescence (b) decay of ZnPc
(COOH)4
The values of the fluorescence lifetimes vary between 1.46 and
10.21 ns, whereas the phosphorescence lifetimes are in the range of
1.07 to 9.33 μs. Also, emission quantum yields under 15% were
determined [2].
In order to demonstrate the excited-state processes and the
involvement of the higher energy electronic states (Sn > 1),
transient absorption spectroscopy (TA) was used. In the transient
absorption map at ns, for ZnPc (COOH)8 in H2O/DMSO at pump
pulse 355 nm, it can be noticed ground state bleaching bands (GBS)
at 270, 290 nm, absorption in the excited state (ESA) at 260, 290,
and 320 nm, respectively and more than one excited state (Sn > 1).
At longer wavelengths, such as 460 nm, 500 nm, 530 nm,
stimulated emission (SE) occurs, and at wavelengths of 640 nm and
680 nm, other stimulated emissions appear, attributed to the triplet
manifold, confirmed by time-resolved fluorescence experiments,
such as phosphorescence. Also, at 510 nm, the triplet excited state
absorption is found (fig. 7).
Figure 7. Transient absorption of ZnPc (COOH)4 in H2O/DMSO
Figure 8. Ultrafast dynamics of ZnPc(SH)4 in DMSO
INNOVATIONS 2025
29
By using ultrafast spectroscopy, transient absorption at fs, for
ZnPc(SH)4 at pump pulse 680 nm it can be observed photoinduced
absorption (PA) at 2.2 eV and stimulated emission (SE) at 1.95 eV
(figs. 8, 9).
Figure 9. Transient absorption of ZnPc(SH)4 in DMSO
The antimicrobial activity of the synthesized ZnPc (COOH)4
and ZnPc (COOH)8 in H2O/DMSO compounds was studied in vitro
by the serial dilution method. The results of the bactericidal tests are
present in Table 1. For clarity, they have been compared with well-
known bactericidal drug, such as chlorhexidine. As can be seen
from the table, the bactericidal activity of ZnPc (COOH)4 is much
lower than that of ZnPc (COOH)8 and chlorhexidine.
Table 1. Bacterial tests
Samples
Microorganisms,
(μg/mL)
S. aureus
(t. 209)
E. faecalis
(t. ATCC 19433)
E. coli
(t. ATCC 25922)
Ps. Aerugi-nosa
(t. ATCC 27853)
Pr. Vulgaris
(t. HX 19222)
ZnPc(COOH)8
300
75
110
75
80
ZnPc(COOH)4
38
89
58
›300
›300
chlorhexidine
125
115
230
›300
›300
4. Conclusion
From the transient absorption map of the phthalocyanine
derivatives, it can be observed ground state bleaching bands (GBS),
absorption in excited state (ESA) and more than one excited state
(Sn>1). At longer wavelengths stimulated emission (SE), and other
stimulated emissions can appear ascribed to the triplet manifold,
confirmed by time-resolved experiments to be phosphorescence.
Also, the triplet excited state absorption can occur.
5. References
[1]. Brilkina A. A., Dubasova L. V., Sergeeva E. A., Pospelov A.
J., Shilyagina N. Y., Shakhova N. M., Balalaeva I.
V., Photobiological properties of phthalocyanine photosensitizers
Photosens, Holosens and Phthalosens: A comparative in
vitro analysis, J. Photochem. Photobiol. B: Biol. 191, 128-134, 2019
[2]. Potlog T., Lungu I., Airinei A., Tigoianu R. Photophysical
Properties of Substituted Zinc Phthalocyanine-Dextran Systems,
ChemPhotoChem, e202400385, 2025
[3]. Wenger O. S., A bright future for photosensitizers, Nat. Chem.
12, 323-324, 2020
[4]. Tang J. L. Y., Moonshi S. S., Ta H. T., Nanoceria: an
innovative strategy for cancer treatment, Cell. Mol. Life Sci. 80, 46,
2023
[5]. Gokçil G., Atmaca G. Y., Şen P., Şahin F., Erdoğmuş A.,
Preparation of cysteine-functionalized graphene quantum dots–Zinc
phthalocyanines supramolecular hybrid system and their sono-
photochemical studies, J. Photochem. Photobiol. A: Chem., 459,
116108, 2025
[6]. Babu N., Rahaman S. A., et. al. Photosensitizer Anchored
Nanoparticles: A Potential Material for Photodynamic Therapy,
ChemistrySelect, 7, e202200850, 2022
ACKNOWLEDGMENTS
This presentation was supported by a grant from the Romanian
Ministry of Research, Innovation, and Digitization, CNCS-
UEFISCDI, project number PN-IV-P8-8.3-ROMD-2023-0048, and
by a mobility project of the Romanian Ministry of Research,
Innovation, and Digitization, CNCS-UEFISCDI, project number
PN-IV-P2-2.2-MC-2024-0823 within PNCDI IV.
INNOVATIONS 2025
30
Simulation of solar-coal hybrid power plant based on the Rankine cycle
Paweł Józef Madejski, Isyna Izzal Muna
AGH University of Krakow, Faculty of Mechanical Engineering and Robotics, Department of Power Systems and Environmental Protection
Facilities, Al. Mickiewicza 30, 30-059 Kraków, Poland
madejski@agh.edu.pl
Abstract: The use of solar energy to generate electricity in Power Plant (PP) based on the Rankine cycle is currently a strongly developed
CSP system (Concentrating Solar Power) technology. Combining CSP systems with an existing Rankine cycle in coal-fired Power Plant (PP)
or Combined Heat and Power Plant (CHP) can increase the maximum output of a power plant and effectively provide a transition for fossil
fuel-based technologies to the use of renewable energy sources such as hybrid or renewable energy-only systems. The paper presents
simulation results of a steam Rankine cycle with solar power plant components. The hybrid power system makes it possible to reduce the use
of coal fuel and provide high efficiency and output power. The calculations were carried out using Ebsilon Professional software.
Keywords: RENEWABLE ENERGY SOURCES, HYBRID ENERGY SYSTEM, THERMODYNAMIC MODELING, ENERGY TECHNOLIGES
1. Introduction
Coal continues to play a vital role in the energy mix of many
countries, particularly in the developing countries, where power
generation infrastructure is still largely dependent on this resource
[1-3]. However, the environmental challenges posed by fossil fuel-
based electricity production - most notably the emission of
greenhouse gases - have become a pressing global issue,
contributing significantly to climate change and global warming.
These concerns underscore the urgent need to modernize existing
coal-fired power plants and accelerate the integration of cleaner,
renewable energy sources into national energy systems.
Concentrated Solar Power (CSP) systems and Thermal Energy
Storage (TES) offer a promising solution for reliable power
generation under varying solar radiation conditions. This advantage
positions CSP above other renewable technologies, such as
photovoltaics and wind power, which are often regarded as leading
alternatives to fossil fuels. Concentrated Solar Power (CSP)
technology has gained prominence for its potential to provide a
sustainable, efficient, and diverse energy mix, making it an essential
component of future energy systems [4]. CSP systems face
significant challenges, including high initial investment costs, lower
efficiency compared to fossil fuel-based systems, and the technical
complexities associated with large-scale TES implementation.
Integrating solar energy into coal-fired power plants forming a
Solar-Coal Hybrid Power Plant (SCHPP), presents a viable
approach to address these issues. This hybrid system has substantial
potential to reduce coal consumption while mitigating the
limitations associated with standalone CSP plants [5].
2. Analyzed energy systems
2.1. Steam power plant
The analyzed steam power plant is a power unit consisting
mainly of a steam boiler, high-pressure steam turbine, low-pressure
steam turbine, generator, condenser, three low-pressure feedwater
preheaters, three high-pressure feedwater preheaters, a feedwater
pump, and a deaerator (Figure 1). The plant operates to generate
electricity, utilizing thermal energy produced by the combustion of
solid fuel in a pulverized coal-fired boiler. The power output,
efficiency, and thermal energy input depend on the current
operating conditions and can vary across the power plant’s load
range. Under selected operating conditions, the main parameters of
the analyzed steam power plant are presented in Table 1.
Table 1. Main factors of steam power plant.
Parameter
Unit
Value
Power output
MW
207.8
Thermal power input
MW
501.26
Thermal efficiency
%
41.45 %
Boiler inlet water temperature
°C
221
Steam mass flow rate
t/h
717.5
Steam temperature
°C
537.8
Steam pressure
bar
104.2
Vapour pressure in condenser
bar
0.067
2.2. Concentrated Solar Power (CSP) technology
Solar power collectors have proven to be an innovative and
essential technology in the shift towards renewable energy. These
devices harness sunlight and convert it into usable electricity,
offering a sustainable alternative to traditional energy sources. Their
growing popularity stems from several benefits, including the
reduction of greenhouse gas emissions and the potential to lower
electricity costs significantly. Utilizing solar energy, these systems
contribute to global efforts to combat climate change, making them
an attractive option for both residential and commercial use. One of
the key advantages of solar power collectors is their versatility.
They come in various forms, such as photovoltaic (PV) panels, solar
power collectors (SPT), and concentrated solar power (CSP)
systems. PV panels are the most widely recognized, directly
converting sunlight into electricity using semiconductors. On the
other hand, solar thermal collectors capture and store the sun’s heat
to produce hot water or heating for homes and businesses. CSP
systems take a more advanced approach, concentrating sunlight to
generate high-temperature heat, which drives turbines for electricity
production on a larger scale. This diversity allows for various
applications, from small household systems to large-scale solar
farms. More than 6 GW of concentrated solar power (CSP) plants
were installed in 2020 [6], and the largest TES from installed CSP
can generate clean electricity for 24 hours by reducing CO2
emission. The International Energy Agency (IEA) estimates that the
CSP will contribute up to 11% of global electricity production in
2050 [6].
Fig. 1. Schematic layout of the Hybrid Solar Coal Power Plant general
concept
3. Modeling of energy systems
3.1. Solar tower receiver and heliostat field
The sun angles and incident power calculations are essential for
solar energy applications, particularly in Concentrated Solar Power
(CSP) systems. Using the standard DIN 5034, the solar angles, such
as the solar altitude and azimuth, is calculated [7]. For solar
irradiance, the Direct Normal Irradiance (DNI) is considered and its
estimation using the Clear Sky model by Hottel [8] or Clearness
Index originally developed by Liu and Jordan [9]. The Clear Sky
model estimates DNI for cloudless conditions based on atmospheric
INNOVATIONS 2025
31
parameters and solar geometry, while the Clearness Index considers
cloud cover and atmospheric clarity. These parameters determine
the usable solar irradiance and incident power on the receiver
aperture.
The total usable solar irradiance
usable
Q
is calculated using the
formula:
usable heliostat
Q DNI A
(1)
where:
heliostat
A
- is the total aperture area of the heliostat field, m2.
The incident power on the receiver aperture area is calculated as
in Eq. (2):
( , )
incident heliostat field s s wind
Q DNI A
(2)
where:
DNI - Direct Normal Irradiance, W/m2
- the average field reflectivity, -
( , )
field s s
- solar field optical efficiency depends on the sun
azimuth
s
, and the sun elevation
s
wind
- wind correction.
The amount of thermal energy transferred to the Heat Transfer
Fluid (HTF) by the receiver to raise fluid temperature is calculated
using the following formula:
eff incident loss
Q Q Q
(3)
, , ,loss loss opt loss conv loss rad
Q Q Q Q
(4)
where:
eff
Q
- power of effective heat, kW
loss
Q
- losses composed of optical
,loss opt
Q
, convective
,loss conv
Q
and radiation losses
,loss rad
Q
, kW.
3.2. Rankine steam cycle
The Rankine cycle power unit is modeled assuming pseudo-
steady-state energy balances. The solar and coal-fired components
(i.e. boiler, superheater, etc.) are modeled dynamically to account
for dynamics in DNI. The entire plant is modeled in Ebsilon
Professional software [10, 11]. For all dynamic systems, the built-in
integrator block of Ebsilon is used to solve differential equations
arising from governing equations. Other major assumptions include
neglecting conductive heat transfer, non-axial temperature
distributions, and constant condenser operation. Constant values are
assumed for saturation pressure during steam generation and
isentropic efficiencies for turbomachinery.
3.3. Hybrid energy system
To investigate improvement of steam cycle performance by the
use of additional heat source in the cycle, the power output and
efficiency of the process were defined. The efficiency of the power
plant can be calculated using the following equations:
el
F
N
Q
(5)
where:
- steam cycle fuel-driven efficiency, %
el
N
- power output, MW
F
Q
- heat rate to the cycle through boiler (energy
from fuel), MW
4. Results and discussion
4.1. Solar field simulation
In the operation of a solar field within a CSP system, solar
energy is captured, stored, and transported to enable electricity
generation. The input data of Solar Field on 15th June 2021 is
presented in Table 2. The process begins with the collection of solar
radiation, which is concentrated and transferred to a working fluid.
This fluid serves as the medium for storing and transferring thermal
energy within the system. The thermal energy is stored in
designated units to ensure continuous operation, even during
periods when solar radiation is unavailable, such as nighttime or
cloudy weather.
Table 2. Solar Field data on 15th June 2021.
Date and time
Power
absorbed by
the fluid
Effective receiver
temperature
Effective field
efficiency
12:00:00 AM
0.00 MW
274.71 ◦C
0 %
2:00:00 AM
0.00 MW
274.71 ◦C
0 %
4:00:00 AM
0.00 MW
274.71 ◦C
0 %
6:00:00 AM
46.44 MW
304.78 ◦C
35.72 %
8:00:00 AM
75.25 MW
323.41 ◦C
57.25 %
10:00:00 AM
83.91 MW
328.76 ◦C
63.72 %
12:00:00 PM
87.37 MW
330.86 ◦C
66.30 %
2:00:00 PM
83.89 MW
328.75 ◦C
63.70 %
4:00:00 PM
75.22 MW
323.39 ◦C
57.22 %
6:00:00 PM
46.44 MW
304.78 ◦C
35.72 %
8:00:00 PM
0.00 MW
274.71 ◦C
0 %
10:00:00 PM
0.00 MW
274.71 ◦C
0 %
Table 2 shows that the peak value is obtained during the middle
of the day between 12:00 PM and 01:00 PM when the sun is visible
at its peak temperature. During the nighttime, the value comes down
after the sunset. During these hours, hot storage has to be used to
supply heat to the power plant.
4.2. Steam cycle simulation
The steam Rankine cycle power plant operation simulation was
designed using Ebsilon® Professional software. The components of
the steam cycle in Ebsilon are the steam turbine, steam generator,
condenser, pump, deaerator, and preheaters (Figure 2). The three
model variants were developed for the simulations, where part of
the feedwater before each HP preheater is diverted to the heat
exchanger (HE). The hot medium in the HP preheaters is the bleed
from stages of HP turbine, and the cold medium is the feedwater
from the deaerator to the boiler. The diverted feedwater to the heat
exchanger is again mixed with the feedwater line before the boiler
inlet, keeping boiler feedwater temperature at the same level.
Fig. 2. Model of hybrid solar coal power plant in Ebsilon®
Professional.
4.3. Hybrid Solar-Coal Power Plant simulation
The simulation then calculates heat balance to model energy
flows through the solar field, thermal energy storage, and steam
cycle. Turbine efficiency, fluid properties, and flow rates are
meticulously modeled to align with design specifications and ensure
efficient system operation.
INNOVATIONS 2025
32
Three different analyses based on the variants were classified
according to the extraction of feed water. A part of the feedwater is
extracted before the HP heater and flows through the heat
exchanger to utilize solar energy. The flow rate of extracted
feedwater is controlled from 10 kg/s to 100 kg/s. The simulated
result data are presented and discussed below. The feedwater
temperature at the boiler inlet is constant and equal to 221°C. The
energy consumed by the steam boiler is also constant in all three
variants regardless of the increase in the power generation of the
electric generator. Extraction point of the variant 1 (a), variant 2 (b)
and variant 3 (c) before HP preheater is shown in Figure 3.
Fig. 3. Extraction point of the variant 1 (a), variant 2 (b) and variant 3
(c) before HP preheater in developed HSCPP model
The hybrid solar-assisted steam power plant simulation
demonstrates the successful integration of solar energy with a
conventional coal-fired power system. The increase in solar energy
input leads to a consistent improvement in thermal efficiency across
the analyzed configurations. Compared to results from variants 1
and 2, with efficiencies of 42.58% and 43.25%, this simulated result
from variant 3 achieves the highest efficiency of 43.75%.
With a solar contribution of 37.62 MW, the coal consumption
needed for generating electricity can be reduced. Although the coal
input remains constant, the improved thermal efficiency indicates
reduced fuel-specific energy requirements. The increased
integration of solar energy reduces the environmental footprint of
the power plant by lowering coal consumption and greenhouse gas
emissions. This highlights the potential for hybrid systems to
contribute to cleaner and more sustainable power generation.
Table 4. Simulation results for variant 1.
Solar heat input
Power output
Efficiency
MW
MW
%
1.60
208.47
41.59
3.21
209.02
41.70
4.81
209.57
41.81
6.42
210.13
41.92
8.02
210.68
42.03
9.63
211.23
42.14
11.23
211.79
42.25
12.84
212.34
42.36
14.44
212.89
42.47
Table 4. Simulation results for variant 2.
Solar heat input
Power output
Efficiency
MW
MW
%
2.76
208.80
41.66
5.51
209.69
41.83
8.27
210.58
42.01
11.03
211.47
42.19
13.79
212.36
42.36
16.54
213.24
42.54
19.30
214.13
42.72
22.06
215.02
42.90
24.82
215.91
43.07
Table 5. Simulation results for variant 3.
Solar heat input
Power output
Efficiency
MW
MW
%
3.76
209.06
41.71
7.52
210.19
41.93
11.29
211.33
42.16
15.05
212.47
42.39
18.81
213.61
42.62
22.57
214.75
42.84
26.34
215.89
43.07
30.10
217.03
43.30
33.86
218.17
43.52
The relationship between mass flow rate and the efficiency of
solar-assisted steam power plant, solar contribution for three
variants are illustrated in Figure 4. Overall, the efficiency
increases steadily as the mass flow rate rises. At lower mass flow
rates (10–30 kg/s), all three variants show relatively low and similar
efficiency levels, with Variant 1 performing the least efficiently.
Fig. 4. Efficiency and solar contribution as a function of mass flow rate
for all analyzed variants
Variant 1 achieves an efficiency starting at 40.5% at 10 kg/s and
peaks just below 43.0% at 100 kg/s. Variant 2 performs moderately
better, reaching around 43.25% efficiency at the highest mass flow
rate. Variant 3 outperforms both, starting at approximately 41.0% at
10 kg/s and reaching nearly 44.0% at 100 kg/s.
INNOVATIONS 2025
33
5. Conclusions
The heat absorbed by the fluid from the solar tower receiver
varies significantly, ranging from 0 to 86.41 MW, depending on the
intensity of solar radiation during operating hours. This variability
highlights the intermittent nature of solar energy and the critical role
of solar thermal systems in effectively capturing and utilizing solar
heat. Despite this variability, the feedwater temperature at the boiler
inlet remains constant at 221 ◦C, ensuring stable operation of the
steam cycle and preventing thermal stresses within the boiler
system. This stability is achieved by integrating the solar thermal
component into the conventional Rankine cycle without disrupting
the boiler’s operational parameters.
This integration strategy demonstrates the effective utilization
of solar energy to supplement and enhance the performance of
traditional coal-fired power plants without altering the boiler’s
energy input, ensuring operational consistency. Such designs
improve thermal efficiency and allow for increased electricity
generation during peak solar hours, making the hybrid power plant
a promising solution for reducing fossil fuel dependence and CO2
emissions while maintaining reliable power output.
References
[1] Y. Zhu, R. Zhai, J. Qi, Y. Yang, M. Reyes-Belmonte, M.
Romero, Q. Yan, Annual performance of solar tower aided coal-
fired power generation system, Energy 119 (2017) 662–674.
[2] F. Liu, T. Lyu, L. Pan, F. Wang, Influencing factors of
public support for modern coal-fired power plant projects: An
empirical study from china, Energy Policy 105 (2017) 398–406.
[3] M. Hofmann, G. Tsatsaronis, Comparative exergoeconomic
assessment of coal-fired power plants–binary rankine cycle versus
conventional steam cycle, Energy 142 (2018) 168–179.
[4] S. Gourvenec, F. Sturt, E. Reid, F. Trigos, Global
assessment of historical, current and forecast ocean energy
infrastructure: Implications for marine space planning, sustainable
design and end-of-engineered- life management, Renewable and
Sustainable Energy Reviews 154 (2022) 111794.
[5] J. Wu, H. Hou, Y. Yang, E. Hu, Annual performance of a
solar aided coal-fired power generation system (sacpg) with various
solar field areas and thermal energy storage capacity, Applied
Energy 157 (2015) 123–133.
[6] C. S. Power, et al., Technology roadmap concentrating solar
power, Current 5 (2010) 1–52.
[7] J. A. Duffie, W. A. Beckman, Solar engineering of thermal
processes, Wiley New York, 1980.
[8] H. C. Hottel, A simple model for estimating the
transmittance of direct solar radiation through clear atmospheres,
Solar energy 18 (2) (1976) 129–134.
[9] B. Y. Liu, R. C. Jordan, The interrelationship and
characteristic distribution of direct, diffuse and total solar radiation,
Solar energy 4 (3) (1960) 1–19.
[10] https://www.ebsilon.com
[11] P. Madejski, P. Żymełka, Introduction to computer
calculations and simulation of energy systems operation in STEAG
Ebsilon®Professional (in Polish). The AGH University of Science
and Technology Press, Kraków, 2020.
Acknowledgements
The research project is supported by the program “Excellence
Initiative – Research University” for AGH University.
The research project results are supported by grant no. 8735
under the 2nd edition (2023) of the competition "Providing
conditions for independent scientific work for postdoctoral fellows"
(Activity D21 in the Project "Excellence Initiative - Research
University" at AGH)
INNOVATIONS 2025
34
Comparative Study of Bayesian-Optimized 1-D CNN, Bi-LSTM and MLP for Bearing Fault
Classification from Raw Vibration Signals
Paweł Knap1,*, Urszula Jachymczyk1
AGH University of Krakow, Poland
pknap@agh.edu.pl, delhi@student.agh.edu.pl
Abstract: This study evaluates the performance of newly designed deep-learning model—bidirectional long short-term memory network (Bi-
LSTM), with baseline to a conventional multilayer perceptron (MLP)—for classification faults of rolling-element bearings from raw
vibration signals. The models are benchmarked against a previously optimised one-dimensional convolutional neural network (1-D CNN),
originally obtained via Bayesian hyperparameter search. A carefully selected dataset of 3600 one-second segments was captured under
varying speed conditions and dynamically enhanced with Gaussian noise during processing. On the test set, the Bi-LSTM achieves 100 %
accuracy, the 1-D CNN 97.9 %, and the MLP 53.3 %. Training dynamics, confusion patterns, and model complexity were thoroughly
analysed, highlighting the trade-offs between accuracy, latency and deployment cost in edge-computing scenarios.
Keywords: PREDICTOVE MAINTENANCE, VIBRATION ANALYSIS, DEEP LEARNING, BEARING FAULYT CLASSIFICATION, NEURAL
NETWORK BENCHMARKING
1. Introduction
The transition towards Predictive Maintenance (PdM) in Industry
4.0 compels the development of monitoring systems capable of
forecasting failures well before catastrophic breakdowns occur.
Among the available sensing modalities, vibration signals carry
particularly rich information about incipient bearing defects [1].
Classical vibration-based diagnosis, however, hinges on hand-
crafted features and extensive domain knowledge [2, 3]. It also
common to use machine learning algorithms for the processing of
calculated condition indicators to determine the state of the machine
[4, 5, 6]. Deep learning eliminates much of this manual effort by
enabling the extraction of end-to-end characteristics, achieving
state-of-the-art results in numerous PdM tasks [7, 8, 9].
Convolutional neural networks (CNNs) dominate modern
computer vision [10] and are increasingly adopted for rotating-
machine diagnostics - either on time–frequency images [11, 12] or,
more recently, directly on raw one-dimensional signals. Their
performance nonetheless depends sensitively on a large set of
hyperparameters that are often tuned heuristically. In earlier work
[13] we introduced a compact 1-D CNN for raw-signal
classification, that was further optimized with Baysian search [14].
Here we benchmark that architecture against two strong alternatives
- a multilayer perceptron (MLP) and a bidirectional LSTM (Bi-
LSTM) - and employ Bayesian optimization to refine the CNN
hyperparameters.
Our contributions are threefold:
i)
we design and implement two deep-learning baselines - an
MLP and a Bi-LSTM - specifically adapted to raw vibration
data for bearing-fault classification;
ii)
we benchmark these models against a previously optimized
1-D CNN architecture under consistent conditions, including
input format, preprocessing, and train–test splits;
iii)
we analyze learning dynamics, confusion patterns, and
resource efficiency, providing practical insights for
deployment on embedded hardware.
2. Dataset
The dataset was collected using a real-world test rig designed
for bearing fault analysis and for comparing the effectiveness of
predictive maintenance models.
2.1. Measurement campaign
Experiments were carried out on the servo-driven rig shown in
Fig. 1. Three bearing conditions were investigated: healthy, outer-
race defect, and inner-race defect.
Fig. 1 Experimental test stand [13].
Radial vibration was recorded by a piezoelectric
accelerometer at 11.025 kHz while the motor executed a
triangular speed profile ranging from 700 to 1000 rpm with 30-s
acceleration and deceleration ramps (Fig. 2). Each condition was
measured for 20 min, yielding 60 min of raw time-domain data.
Fig. 2 Motor-speed trajectory during data acquisition [13].
2.2. Segmentation and normalization
The raw vibration signals were acquired as continuous time
series during a 60-minute recording session, with 20 minutes per
bearing condition. To convert these into a format suitable for super-
vised learning, the signals were segmented into non-overlapping
one-second windows, each containing 11 025 samples due to the
11.025kHz sampling rate. This procedure resulted in 1 200
segments for each of the three classes (healthy, outer-race defect,
and inner- race defect), with a total of 3 600 segments.
To eliminate scale variations and ensure consistent input
distributions across the dataset, each segment x was standardized
using the z-score transformation:
= (
)/,
INNOVATIONS 2025
35
where
is the mean and S is the standard deviation of the segment.
This normalization was applied independently to each example,
following the recommendation in [?]. It allowed to preserve the relative
shape of the signal while removing absolute amplitude differences.
2.3. Train-test Split
To evaluate model generalization, the dataset was split into
training and test sets using a stratified random partition, preserving
the class distribution across both subsets. A standard 80/20 split was
applied, with a fixed random seed (42) to ensure reproducibility.
This resulted in 2 880 training segments and 720 test segments. No
data leakage was allowed between sets, and all reported
performance metrics in later sections refer strictly to the unseen test
data. Validation during training was performed using a holdout
subset from within the training dataset.
3. Data Augmentation
To improve generalization, the training batches were augmented on-
the-fly using Gaussian jittering: for each segment x a noisy
counterpart
=+(,.
) is generated, where is the
per-sample standard deviation. The procedure, summarized in
Algorithm 1, effectively doubles the number of training examples
without altering validation or test splits.
Algorithm 1. Gaussian-jitter augmentation (noise level 5 %).
4. Model Architectures
4.1. Multilayer Perceptron (MLP)
The baseline MLP flattens the one-dimensional input of length
11,025 into a vector, discarding any local structure in the signal.
This vector is then processed through three fully connected (dense)
layers of 1,024, 512 and 128 neurons, respectively. Each dense
layer is followed by batch normalization, a ReLU activation, and
dropout regularization with rates 0.4, 0.4, and 0.3 in sequence. This
block-wise design encourages stable training and mitigates
overfitting. The output is fed into a final dense layer with a soft-
max activation to produce class probabilities. Due to the high
dimensionality of the input and lack of parameter sharing, the
model comprises approximately 11 million trainable parameters,
making it by far the largest among the compared architectures.
4.2. Bidirectional LSTM (Bi-LSTM)
The Bi-LSTM model begins by reshaping the raw input (11025,
1) into a 2D tensor of shape (147, 75), treating the signal as a
sequence of 147 time steps with 75 features per step. Layer
normalization is applied to stabilize the input distribution before
temporal modelling. The core of the network consists of two
stacked bidirectional LSTM layers, with 128 and 64 neurons
respectively, both using dropout at a rate of 0.3. This bidirectional
structure allows the model to incorporate both past and future
context at each time step. The resulting sequence is pooled into a
fixed-size vector via global average pooling, which is then passed
through a dense layer with 128 neurons, batch normalization, ReLU
activation, and 0.4 dropout. The final prediction is made by a soft-
max classifier. In total, the network includes approximately 1.1
million trainable parameters.
4.3. Bayesian-optimized 1-D CNN
The optimized 1-D CNN architecture processes the raw input
(11025,1) through a series of five identical convolutional blocks.
Each block applies a 1-D convolution with 88 filters of kernel size 5
(using 'same' padding), followed by batch normalization and a
ReLU activation. This design enables the network to detect local
patterns in the vibration signal across multiple layers, effectively
building hierarchical feature maps. After the final convolutional
block, the feature maps are aggregated using a global max pooling
layer, which selects the most salient activation across time for each
filter. The resulting feature vector is then passed directly to a dense
soft-max layer for classification. The model contains approximately
160,000 parameters, making it the most compact of the three, yet it
still achieves high accuracy thanks to its inductive bias and depth.
5. Training Procedure
All models were trained using the Adam optimiser, which
combines the advantages of momentum and adaptive learning rates
to achieve fast and stable convergence. For the MLP and Bi-LSTM
models, the learning rate was fixed at 103, while the CNN used
the optimiser's default settings unless modified by Bayesian
optimisation. Training was conducted in mini-batches of size 64 to
balance convergence speed and memory efficiency. To prevent
overfitting and reduce unnecessary computation, early stopping was
employed: if the validation loss did not improve for 15 consecutive
epochs, training was halted. The dataset was consistently partitioned
using the same 80/20 training/validation split across all
experiments, ensuring a fair and directly comparable evaluation of
the different architectures.
6. Result and Discussion
6.1. Training Dynamics
Figure 3 compares the training and validation learning curves
for all three model architectures. The Bi-LSTM exhibits the fastest
convergence, reaching near-zero validation loss within
approximately 30 epochs and maintaining this performance
thereafter, suggesting both effective learning and strong
generalization. The 1-D CNN converges more gradually, plateauing
around epoch 60, but achieves a comparably low validation loss,
reflecting the model's ability to capture relevant features despite its
compact architecture. In contrast, the MLP shows early signs of
instability: its training accuracy continues to improve while the
validation loss stagnates near a value of 1.2, accompanied by a
growing gap between training and validation metrics. This pattern is
characteristic of overfitting, likely caused by the MLP's large
parameter count and lack of inductive bias.
Fig. 3 Training (solid) and validation (dashed) curves for accuracy (left)
and loss (right).
6.2. Test-set performance
Table 1 presents macro-averaged classification metrics
computed on the held-out test set. The Bi-LSTM achieves perfect
performance across all evaluated criteria, with an accuracy and F1-
score of 1.000 and near-zero cross-entropy loss. This result reflects
the model’s strong capacity to capture temporal dynamics in the
vibration data. The 1-D CNN also performs remarkably well,
attaining 97.9% accuracy and high precision and F1 scores, despite
being significantly smaller in size and simpler in structure. In
contrast, the MLP reaches only 53.3% accuracy and exhibits poor
generalization, with both precision and F1 falling below 0.50. These
results confirm that models with temporal or convolutional structure
INNOVATIONS 2025
36
- capable of exploiting the signal’s local or sequential properties -
are much better suited to the classification task, performed on raw
vibration signal, than purely dense alternatives.
Table 1. Test-set metrics (macro-averaged).
Model
Accuracy
Loss
Precision
F1
1D CNN
0.979
0.035
0.987
0.979
MLP
0.533
2.090
0.480
0.462
Bi-LSTM
1.000
4.9 105
1.000
1.000
6.3. Error Analysis
Fig. 4 Confusion matrix — 1-D CNN.
Fig. 5 Confusion matrix — Bi-LSTM.
Fig. 6 Confusion matrix — MLP.
Figures 4-6 visualize class-wise errors. The 1-D CNN mis-
labels five inner-race defects as outer-race, whereas all other pairs
are perfectly separated. The Bi-LSTM exhibits zero mis-
classifications. The MLP confuses all classes, especially outer-race
versus inner-race.
6.4. Comparative Discussion
The superior performance of the Bi-LSTM underscores the
benefit of modelling long-range temporal dependencies. The
lightweight 1-D CNN nonetheless achieves near-perfect accuracy
with 300× fewer parameters and lower inference latency, making it
attractive for embedded deployment. The densely connected MLP
fails to generalize; its 11 million parameters vastly outnumber the
training samples and lack inductive bias towards local patterns.
7. Conclusion
Through Bayesian Optimization it was possible to develop a
compact 1-D CNN that attains 97.9% test accuracy on raw vibration
data with only 0.16 million parameters. Even though the deeper Bi-
LSTM reached 100%, it comes at the cost of increased latency and
memory usage with 1.1 million trainable parameters. These results
highlight that strong inductive biases and systematic
hyperparameter tuning are crucial for reliable bearing fault
diagnosis from raw sensor data. A large number of parameters and
complex architectures do not inherently ensure high effectiveness;
in fact, they often lead to poor generalization and overfitting,
particularly in relatively simple tasks such as three-class
classification. Future work will focus on further exploring
lightweight attention mechanisms and enabling on-device
deployment.
FUNDING
This research was funded by the Excellence Initiative –
Research University, Action. 12 – Integration of educational
process with scientific research – Support for student research clubs
– Rector's grants, 5rd edition of the competition (ID 8980).
8. References
1. R. B. Randall and J. Antoni, “Rolling element bearing
diagnostics—a tutorial,” Mechanical Systems and Signal
Processing, vol. 25, no. 2, pp. 485–520, [2011]. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S0888327010002
530
2. R. B. Randall, “Vibration-based condition monitoring:
industrial, automotive and aerospace applications”, John Wiley &
Sons, [2021].
3. A. Jabłoński, “Condition Monitoring Algorithms in MATLAB”.
Kraków: Springer, [2021].
4. U. Jachymczyk, P. Knap, and K. Lalik, “Improved intelligent
condition monitoring with diagnostic indicator selection,” Sensors,
vol. 25, no. 1, p. 137, [2024].
5. U. Jachymczyk and P. Knap, “Review of feature selection
methods for predictive maintenance systems,” Industry 4.0, vol. 9,
no. 3, pp. 97–100, [2024].
6. M. Gharavian, F. Almas Ganj, A. Ohadi, and H. Heidari
Bafroui, “Comparison of fda-based and pca-based features in fault
diagnosis of automobile gearboxes,” Neurocomputing, vol. 121, pp.
150–159, [2013], advances in Artificial Neural Networks and
Machine Learning.
7. Y. Ran, X. Zhou, P. Lin, Y. Wen, and R. Deng, “A survey of
predictive maintenance: Systems, purposes and approaches,” arXiv
preprint arXiv:1912.07383, pp. 1–36, [2019].
8. T. P. Carvalho, F. A. Soares, R. Vita, R. d. P. Francisco, J. P.
Basto, and S. G. Alcalá, “A systematic literature review of machine
learning methods applied to predictive maintenance,” Computers &
Industrial Engineering, vol. 137, p. 106024, [2019].
9. G. Singh and S. Ahmed Saleh Al Kazzaz, “Induction machine
drive condition monitoring and diagnostic research—a survey,”
Electric Power Systems Research, vol. 64, no. 2, pp. 145–158,
[2003]. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S0378779602001
724
INNOVATIONS 2025
37
10. Z. Li, F. Liu, W. Yang, S. Peng, and J. Zhou, “A survey of
convolutional neural networks: analysis, applications, and
prospects,” IEEE transactions on neural networks and learning
systems, vol. 33, no. 12, pp. 6999–7019, [2021].
11. C. Gianoglio, E. Ragusa, P. Gastaldo, F. Gallesi, and F.
Guastavino, “Online predictive maintenance monitoring adopting
convolutional neural networks,” Energies, vol. 14, no. 15, p. 4711,
[2021].
12. W. Silva and M. Capretz, “Assets predictive maintenance
using convolutional neural networks,” in 2019 20th IEEE/ACIS
International conference on software engineering, artificial
intelligence, networking and parallel/distributed computing
(SNPD). IEEE, [2019], pp. 59–66.
13. P. Knap, K. Lalik, and P. Bałazy, “Boosted convolutional
neural network algorithm for the classification of the bearing fault
form 1-d raw sensor data,” Sensors, vol. 23, no. 9, p. 4295, [2023].
14. P. Knap and U. Jachymczyk, “Bayesian-tuned convolutional
neural networks for precise bearing fault classification,” in 2024
25th International Carpathian Control Conference (ICCC). IEEE,
[2024], pp. 01–05.
INNOVATIONS 2025
38
Correlation-Based Sensor Pruning and Malfunction Detection in Multi-Sensor Condition
Monitoring
Urszula Jachymczyk1, Paweł Knap1,*
AGH University of Krakow, Poland
delhi@student.agh.edu.pl, pknap@agh.edu.pl
Abstract: This study presents a correlation-based approach for both detecting sensor malfunctions and identifying redundant sensors in a
multi-sensor condition monitoring system. Sensor malfunctions were detected using a threshold-based method that flagged correlation
drops, with persistence criteria applied to eliminate false positives. While no persistent malfunctions were observed during the study, the
developed algorithm remains suitable for real-time deployment. Correlation analysis also revealed that the 3axis_Y signal exhibited the
highest average correlation with others, indicating redundancy. Five machine learning models were trained and evaluated with the Leave-
One-Run-Out strategy to guarantee generalization across acquisition sessions. The findings demonstrated that correlation-driven sensor
selection and anomaly detection are effective tools for optimizing predictive maintenance systems, improving model generalization, and
simplifying sensor networks without sacrificing reliability.
Keywords: PREDICTIVE MAINTENANCE, SENSOR MALFUNCTION DETECTION, CORRELATION ANALYSIS, SENSOR
REDUNDANCY, VIBRATION ANALYSIS
1. Introduction
Modern condition monitoring systems rely on a wide array of
sensors to capture real-time data on machine and process states.
Ensuring the reliability of such multi-sensor setup is essential for
the success of Predictive Maintenance (PdM) strategies. However,
sensor malfunction - whether - resulting from drift, noise, or signal
loss (Fig. 1) - can compromise model accuracy and system safety
[1]. An increasingly adopted approach to detect such faults relies on
monitoring the correlation between sensors, either from raw time
domain data or from condition indicators extracted from those
signals. Sensors that typically show strong functional relationships
should maintain high correlation across time. A malfunctioning
sensor often deviates from these established relationships, resulting
in a measurable drop in correlation. This makes correlation a useful
unsupervised indicator for detecting anomalies in multi-sensor
environments [2].
Fig. 1 Sensor malfunction encountered during one of the measurements.
Redundant sensors are frequently used in industrial setups for
fault tolerance and system reliability. By comparing their outputs, a
faulty sensor can be detected when its readings diverge from the
correlated group. For example, in [3], voting and correlation-based
approaches were used to detect faulty sensors by identifying values
that no longer aligned with their peers. In another study [4], the
Piecewise Aggregate Approximation (PAA) technique was used to
aggregate the data in fixed intervals of time, followed by the
creation of a correlation matrix from baseline data. An anomaly was
detected when the correlation between typically aligned sensors
dropped below a defined threshold. If such events occurred in
sequence, the data was labeled abnormal.
Various methods further leverage correlation-based methods to
detect sensor faults. In [5], variables with correlation coefficients
ranging from 0.5 to 0.95 were organized into two-dimensional
clusters. Anomalies were then identified using Mahalanobis
distance to detect deviations from the established normal behavior.
In Wireless Sensor Networks (WSNs), multi-variable correlation is
used in ensemble methods to enhance detection reliability in
complex or high-volume systems [6]. An outlier was confirmed if at
least two methods agreed. A sensor reading was considered
anomalous if two or more models agreed. This approach has been
shown to maintain high detection accuracy and low false alarm
rates. In [7] correlation method was applied to differentiate between
events (true incidents) and noise (sensor errors). After detecting
anomalies using exponential moving averages, a correlation matrix
determined whether anomalies in correlated sensors were due to a
real event or just isolated sensor noise.
Simultaneously, correlation analysis can be used to identify
redundant sensors and indicators. By examining pairwise
correlations among sensor signals or performance indicators,
variables carrying overlapping information can be identified. High
correlations indicate overlapping (redundant) information.
Removing such redundancy can simplify system complexity,
improve computational efficiency, and lower deployment and
maintenance costs without significant loss of information [8]. While
correlation-based feature reduction is commonly applied to
condition indicators [9], it can be extended to entire signal channels
or even physical sensors. Authors of [8] demonstrated a correlation-
driven method for optimal sensor placement in structural
monitoring: they computed a correlation matrix of all candidate
sensor locations and used a clustering algorithm to choose a subset
of sensors with minimal redundancy. If two sensors consistently
yield highly similar indicator patterns across time windows, they
can be considered functionally redundant. In such cases, removing
or consolidating one of the sensors can reduce system complexity,
minimize computational overhead, and even lower the system’s
overall risk - fewer sensors mean fewer components that can fail.
Only a few well-placed sensors can often capture most of the
variability if chosen intelligently [10]. This type of sensor pruning
or fusion is gaining interest in industrial PdM applications, where
data volume and sensor failure management are growing
challenges.
The dual use of correlation - both as an anomaly signal and a
redundancy measure, represents a powerful and interpretable tool
for sensor network optimization.
2. Pearson’s and Spearman’s correlation
Both Pearson’s and Spearman’s correlation coefficients are used
in sensor-based data analysis for feature/sensor selection, but they
have different assumptions and strengths.
Pearson’s correlation - measures linear association between
sensor readings [11] and is widely used to identify collinear
(redundant) sensors or features. In [12] Persons correlation
coefficient was deployed to identify the most relevant
meteorological inputs, from which Machine Learning (ML)
models should learn.
INNOVATIONS 2025
39
Spearman’s correlation - nonparametric measure based on
data ranks, which captures the strength of any monotonic
(not necessarily linear) relationship [11]. Spearman’s method
is often chosen for sensor data that exhibit nonlinear trends
or outliers – for instance, it has been used to select sensors in
a machine tool monitoring setup by detecting nonlinear
dependencies (reducing 16 sensors to 7) [13]. It’s also useful
when comparing different sensor types or calibrations,
focusing on the agreement in ranked trends rather than exact
values [14].
The key difference is that Pearson assumes a straight-line
relationship and works best with continuous, normally-distributed
data, whereas Spearman makes no normality assumption and is
robust to non-linear monotonic patterns. Consequently, Pearson is
preferred when a clear linear relationship is expected, while
Spearman is preferred for capturing monotonic correlations in
skewed or ordinal sensor data and in detecting trend-like
relationships that Pearson might miss.
3. Methodology
The measurements for this study were conducted on the
experimental setup shown in Fig. 2. Vibration data was collected
from four accelerometers: one triaxial sensor and three uniaxial
sensors, resulting in a total of six signal channels. The sampling
frequency was set to 30 kHz. All sensors were rigidly mounted to
the bearing housings and machine table to ensure stable and
representative signal acquisition across different mechanical
components.
Measurements were performed under three different servomotor
speeds: 745, 750, and 755 rpm, to simulate slight process
variability. The training and validation datasets included several
independent acquisition runs to ensure robustness against data
distribution shifts between sessions. This approach was designed to
reflect real-world operating conditions and enhance model
generalization.
Fig. 2 Experimental measurement stand.
In this study, two fault classes were considered, as shown
in Table 1. Class 0 represents healthy operation, while Class 1
corresponds to shaft imbalance. The imbalance fault was simulated
by attaching an additional weight to the disks mounted on the
rotating shaft, which introduced characteristic vibration patterns.
Table 1. Fault Type to the Classes Mapping.
Fault type
0
No fault
1
Unbalanced shaft
3.1. Data analysis
In the preprocessing phase, the raw vibration signals were
divided into time windows of 1.5 seconds each. Each segment was
filtered using a 5th-order Butterworth bandpass filter with the
following cutoff frequencies:
• 0.5 - 5000 Hz for the triaxial accelerometer,
• 0.8 - 13000 Hz for the uniaxial accelerometers.
In addition, the envelope of each signal was calculated using the
Hilbert transform to enable demodulation. Filtering in the kilohertz
range (to capture the machine’s resonance) is beneficial because the
impulsive events due to bearing faults are more pronounced in that
band [15]. Prior to envelope extraction, signals were high-pass
filtered above 3 kHz to isolate machine resonances. The filtered
signals were then transformed into the frequency domain (from both
time domain and envelope) using the Fast Fourier Transform (FFT).
From each signal window, a wide range of time and frequency
domain condition indicators were calculated in order to assess the
machine’s operational state. These include the following indicators,
which were described in [16]:
• Peak-to-Peak (PP),
• Root Mean Square (RMS),
• Crest Factor (CF),
• Standard Deviation (STD),
• Kurtosis,
• Shape Factor (SF),
• Mean Frequency (MF),
• Frequency Center (FC),
• 3 maximum amplitudes and their corresponding frequencies,
• Energy Operator (EO),
• Fourth Order Figure of Merit (FM4),
• Sixth-Order Figure of Merit (M6A),
• Eight-Order Figure of Merit (M8A),
• Clearance Factor (ClF),
• Impulse Indicator (II),
• Root Mean Square Frequency (RMSF),
• Standard Deviation Frequency (STDF),
• Fourth Order Normalized Power (NP4).
In total, 276 indicators were computed per time window
across all six sensor channels, resulting in a comprehensive
condition vector for each time segment. These vectors were
subsequently used for both sensor malfunction detection and sensor
selection based on correlation analysis.
3.2. Sensor malfunction detection
The correlation between sensors was monitored across time
windows to detect deviations indicative of sensor malfunctions. For
each time window, Pearson (Fig. 4) and Spearman (Fig. 3)
correlation coefficients were calculated between all pairs of sensors
based on their computed condition indicators. The average of both
correlation types was used to increase robustness, as Pearson
captures linear relationships while Spearman captures monotonic
trends and is more resistant to outliers.
A sensor was flagged as malfunctioning in a time window if its
mean of Pearson and Spearman correlation with at least four other
sensors dropped below predefined threshold of 0.85. To avoid false
positives, a sensor was considered definitively malfunctioning only
if this behavior occurred in at least 4 windows and showed signs of
persistence (i.e., occurred in a sequence of consecutive or near-
consecutive windows, within the duration of 6 seconds in
summary). Tab. 2 shows the number of windows where each sensor
was flagged and whether a persistent failure was detected.
INNOVATIONS 2025
40
Fig. 3 Spearman’s correlation between sensors across time windows.
Fig. 4 Pearson’s correlation between sensors across time windows.
Table 2. Sensor malfunction.
Sensor
Malfunction windows
Persistent failure
1axis_3
34
False
1axis_2
5
False
3axis_X
4
False
1axis_1
0
False
3axis_Y
0
False
3axis_Z
0
False
3.3. Sensor selection
To reduce redundancy in the sensor network and improve model
generalization, correlation-based sensor selection was applied. The
average correlation of each sensor with all other sensors was
calculated using the mean of Pearson and Spearman coefficients
across all time windows. The sensor exhibiting the highest average
correlation to others was considered the most redundant and thus a
candidate for removal.
This method assumes that a highly correlated sensor contributes
less unique information to the condition monitoring system and may
be safely excluded without significant loss of predictive power. In
this case, the 3axis Y sensor had the highest mean correlation,
suggesting it could be omitted from model training to reduce
dimensionality (Tab. 3).
Table 3: Mean correlation of each sensor with others (averaged across all
windows).
Sensor
Avg mean correlation
3axis_Y
0.9494
3axis_X
0.9432
3axis_Z
0.9418
1axis_2
0.9301
1axis_3
0.9221
1axis_1
0.8785
3.4. Machine Learning models
Models were trained using BayesSearchCV with 5-fold cross-
validation to optimize hyperparameters and reduce the risk of
overfitting. For final evaluation, the Leave-One-Run-Out (LORO)
strategy was employed to ensure the models generalized well across
independent measurement runs. During hyperparameter tuning, a
subset of the training data was held out for internal validation.
The dataset consisted of measurements collected for two fault
classes at different time intervals using a dedicated test stand. Since
models tend to overfit to specific noise patterns or temporal
characteristics of a single acquisition session, special care was taken
to test their robustness on data gathered during entirely separate
sessions. Without this precaution, models can perform well on
cross-validation but generalize poorly to new conditions,
particularly in time-sensitive applications.
In this study, the following machine learning models were
trained and evaluated: K-Nearest Neighbors (KNN), Decision Tree
(DT), Random Forest (RF), Support Vector Machine (SVM), and
XGBoost. All models were initially trained and tested using the full
set of measurement signals. Subsequently, the process was repeated
after removing the most redundant signal, identified through
correlation analysis, to assess the impact of dimensionality
reduction on model performance.
3. Results and Discussion
The most correlated sensor was 3axis Y, with an average mean
correlation of 0.9494 when compared to all other sensors (Tab. 3).
This indicates that 3axis Y provides highly redundant information
and may not contribute significantly to model generalization. Based
on this redundancy analysis, the sensor was excluded from a second
round of model training and evaluation.
No persistent sensor malfunction was detected in any of the six
accelerometers used in this study. While some sensors, such as
1axis_3, showed correlation drops in several windows, none met the
criteria for persistent failure. The absence of malfunction events
indicates that all sensors operated reliably throughout the data
acquisition period (Tab. 2).
All five classifiers were trained using BayesSearchCV and
evaluated using both cross-validation and the Leave-One-Run-Out
method. The results before and after sensor reduction are
summarized in Tab. 4 and Tab. 5, respectively.
Before removing 3axis Y, all models performed well on the
validation set, with SVM achieving perfect accuracy. However,
when evaluated on the test set using unseen data from different
acquisition sessions, KNN exhibited a noticeable drop in
performance, highlighting its limited generalization capability. In
contrast, the remaining models—Decision Tree, Random Forest,
SVM, and XGBoost—achieved equal or even higher accuracy on
the test set compared to their cross-validation results during
BayesSearch optimization.
Table 4. Comparison of Accuracy of Models on Validation and Test
Datasets.
KNN
DT
RF
SVM
XGBoost
Validation
96.63%
98.08%
99.04%
100.00%
99.52%
Test
74.30%
99.53%
100.00%
100.00%
100.00%
After removing the 3axis Y sensor, the accuracy of most models
on validation dataset remained high. On Leave-one-run-out test
dataset, KNN and SVM showed slight improvements in
generalization, with test accuracy increasing to 77.57% and 99.53%,
respectively. However, Decision Tree performance dropped
significantly on the test set after sensor removal, highlighting its
sensitivity to feature reduction. Importantly, the best-performing
INNOVATIONS 2025
41
model – XGBoost – maintained high accuracy, demonstrating
robustness even with reduced input dimensionality.
Table 5. Comparison of Accuracy of Models on Validation and Test Datasets
after reduction of 3axis Y sensor.
KNN
DT
RF
SVM
XGBoost
Validation
97.12%
96.63%
99.04%
100.00%
99.52%
Test
77.57%
43.93%
97.66%
99.53%
100.00%
5. Conclusion
This study addressed both sensor malfunction detection and
sensor selection using correlation-based analysis within a machine
learning framework for fault classification. Pearson’s and
Spearman’s correlation coefficients were calculated between all
sensor pairs over time windows to monitor system behavior and
identify redundancy.
None of the sensors exhibited persistent signs of malfunction.
However, the malfunction detection algorithm will be integrated
into the measurement stand for real-time monitoring (to prevent
occurrences as shown in Fig. 1). This mechanism can serve as an
early warning system, identifying potential underperformance in the
measurement chain and reducing the risk of invalid predictions and
false alarms.
The Leave-One-Run-Out (LORO) strategy proved its relevance
for evaluating model generalization. While models often perform
very well on validation data derived from the training set, they may
perform significantly worse on data collected in separate acquisition
sessions. This confirms the importance of building comprehensive
and diverse datasets—including multiple experimental runs—to
develop noise-resistant and robust models.
After feature selection, most models maintained high accuracy
on validation dataset, with XGBoost continuing to achieve 100%
performance. Notably, the Decision Tree classifier suffered a
significant, above 50%, drop in test accuracy after sensor reduction,
highlighting its sensitivity to input dimensionality. In contrast,
models like SVM, XGBoost and Random Forest remained stable,
suggesting that carefully chosen models can operate effectively
even with a reduced sensor set.
In summary, correlation-based sensor selection can maintain
model performance while simplifying the sensor setup. Combined
with ongoing malfunction detection and robust evaluation
strategies, this approach provides a reliable pathway toward
efficient and trustworthy machine learning-based condition
monitoring systems.
FUNDING
This research was funded by the Excellence Initiative –
Research University, Action. 12 – Integration of educational
process with scientific research – Support for student research clubs
– Rector's grants, 5rd edition of the competition (ID 8980).
6. References
1. V. G. Biju, A.-M. Schmitt, and B. Engelmann, ―Assessing the
influence of sensor-induced noise on machine-learning-based
changeover detection in cnc machines‖, Sensors, vol. 24, no. 2: 330,
[2024].
2. P. Nunes, J. Santos, and E. Rocha, ―Challenges in predictive
maintenance – a review,‖ CIRP Journal of Manufacturing Science
and Technology, vol. 40, pp. 53–67, [2023].
3. A. Kajmakovic, K. Diwold, K. Römer, J. Pestana, and N.
Kajtazovic, ―Degradation detection in a redundant sensor
architecture,” Sensors, vol. 22, no. 12: 4649, [2022].
4. Z. Li, D. Xiaoou, and H. Wang, ―An Effective Constraint-
Based Anomaly Detection Approach on Multivariate Time Series‖,
vol. 10, pp. 61–69, [2020].
5. Y. Yoo, ―Data-driven fault detection process using correlation
based clustering,‖ Computers in Industry, vol. 122, p. 103279,
[2020].
6. M. Roig, M. Catalan, and B. Gastón, ―Ensembled outlier
detection using multi-variable correlation in wsn through
unsupervised learning techniques,‖ vol. 01, pp. 38–48, [2019].
7. D. ElMenshawy and W. Helmy, ―A correlation based
approach to differentiate between an event and noise in internet of
things,‖ International Journal of Advanced Computer Science and
Applications, vol. 9, no. 12 [2018].
8. W. Lu, R. Wen, J. Teng, X. Li, and C. Li, ―Data correlation
analysis for optimal sensor placement using a bond energy
algorithm,‖ Measurement, vol. 91, pp. 509–518, [2016].
9. U. Jachymczyk, P. Knap, and K. Lalik, ―Improved intelligent
condition monitoring with diagnostic indicator selection,‖ Sensors,
vol. 25, no. 1: 137, 2025.
10. A. Krause, A. Singh, and C. Guestrin, ―Near-optimal sensor
placements in gaussian processes: Theory, efficient algorithms and
empirical studies.‖ Journal of Machine Learning Research, vol. 9,
no. 2, [2008].
11. J. S. Do, A. B. Kareem, and J.-W. Hur, ―Lstm-autoencoder for
vibration anomaly detection in vertical carousel storage and
retrieval system (vcsrs),‖ Sensors, vol. 23, no. 2: 1009, [2023].
12. I. Jebli, F.-Z. Belouadha, M. I. Kabbaj, and A. Tilioua,
―Prediction of solar energy guided by pearson correlation using
machine learning,‖ Energy, vol. 224, p. 120109, [2021].
13. T.-H. Hsieh, H.-Y. Lai, Y.-H. Chou, and T.-H. Wu,
―Differences in applications of various neural network algorithms
for thermal error compensation,‖ Sensors and Materials, vol. 36, p.
3573, no. 08, [2024].
14. J. Ye, C. Xiao, R. M. Esteves, and C. Rong, ―Time series
similarity evaluation based on spearman’s correlation coefficients
and distance measures,‖ in Cloud Computing and Big Data,
Springer International Publishing, pp. 319–331, [2015].
15. R. B. Randall and J. Antoni, ―Rolling element bearing
diagnostics—a tutorial,‖ Mechanical Systems and Signal
Processing, vol. 25, no. 2, pp. 485–520, [2011].
16. V. Sharmaa and A. Parey, ―A review of gear fault diagnosis
using various condition indicators,‖ Elsevier, no. 144, p. 253–263,
[2016].
INNOVATIONS 2025
42
Control System Concept for an Omnidirectional Mobile Platform: Modeling and Design
Aspects
Patryk Bałazy,*, Julia Jeleńska1
AGH University of Krakow, Poland1
balazy@agh.edu.pl
Abstract: This paper presents a comprehensive concept for the control system of a four-wheeled omnidirectional mobile platform equipped
with mecanum wheels, intended for industrial applications under the Industry 4.0 paradigm. The platform is modeled both kinematically and
dynamically, with a nonlinear rigid-body formulation that incorporates Coriolis effects and rolling resistance. Particular attention is paid to
the challenges arising from strong coupling between translational and rotational motion. To overcome the limitations of conventional
control methods in complex and dynamic environments, a reinforcement learning strategy based on an actor–critic architecture is proposed.
The agent is trained in a virtual warehouse scenario using simulated lidar data as sensory input, allowing it to learn effective policies for
collision-free navigation. The continuous action space is mapped to wheel angular velocities through scaled hyperbolic tangent activations,
enabling direct and fine-grained control of the platform. The proposed control system is designed for modularity, robustness, and scalability,
making it a promising candidate for autonomous logistics and adaptive robotic applications in smart manufacturing environments.
Keywords: OMNIDIRECTIONAL MOBILE ROBOT, MECANUM WHEELS, ACTOR-CRITIC, DYNAMIC MODELING, LIDAR
PERCEPTION
1. Introduction
Innovations within the framework of Industry 4.0 have
fundamentally transformed the expectations for autonomous mobile
systems in modern industrial ecosystems. Increasingly, there is a
demand for intelligent, flexible, and reconfigurable robotic
platforms that can operate safely and efficiently in dynamic, semi-
structured, and human populated environments. Among various
robotic configurations, omnidirectional mobile robots (OMRs)
equipped with mecanum wheels stand out due to their holonomic
motion capabilities. These platforms possess the unique ability to
perform translational and rotational movements independently,
allowing full planar mobility without requiring reorientation. Such
characteristics are invaluable for tasks involving tight navigation,
rapid maneuvering, and synchronized motion in settings such as
smart factories, automated warehouses, and collaborative
workspaces. However, while the kinematic versatility of OMRs is
well understood and attractive from a motion planning perspective,
their effective control under real-world conditions remains a non-
trivial challenge. The dynamic behavior of mecanumwheeled
systems is governed by complex nonlinear interactions between the
platform body, wheel actuation, and roller-ground contact forces.
Additionally, practical implementations are often affected by
unmodeled dynamics, wheel slippage, time-varying disturbances,
and uncertainties in payload and surface properties. These factors
collectively lead to degraded control performance when using
simplified models or control laws that neglect the physical
intricacies of the system. As a consequence, conventional control
strategies that rely solely on idealized kinematic models prove
insufficient in achieving high-precision tracking and robust
stability. To overcome these limitations, a number of advanced
control approaches have been proposed in the literature. One well-
established method is Trajectory Linearization Control (TLC),
which partitions the control architecture into a dual-loop structure
comprising an outer-loop kinematic controller and an inner-loop
dynamic controller [1]. This architecture allows the system to
account for nonlinearities along predefined trajectories and has
shown promising results in both simulations and real-time
hardware-in-the-loop (HIL) testing. Nevertheless, the performance
of TLC-based controllers remains sensitive to deviations in model
parameters and may suffer from control saturation under
highdemand maneuvers. Nonlinear Model Predictive Control
(NMPC) represents another powerful paradigm, offering the ability
to optimize control inputs in the presence of constraints and
predictive objectives. Its application to mecanum-based platforms,
particularly in dynamically changing environments, has
demonstrated robust performance in trajectory tracking and real-
time collision avoidance [2]. The inherent capacity of NMPC to
incorporate dynamic system models and obstacle representations
makes it well-suited for collaborative Industry 4.0 tasks. Yet, its
computational cost poses challenges for embedded implementation,
requiring efficient solvers or hardware acceleration to achieve real-
time responsiveness. In parallel, efforts have been made to enhance
control robustness using strategies rooted in nonlinear systems
theory. For instance, inverse input–output linearization techniques
have been applied to decouple platform motion from actuator
dynamics, thereby enabling improved trajectory tracking even in the
presence of saturation constraints [3]. Recent developments also
include the application of mixed finite-time boundedness and H∞
performance criteria to omnidirectional mobile platforms, allowing
for disturbance rejection and guaranteed transient performance
under bounded uncertainties [4]. These contributions underline a
trend toward integrating formal robustness metrics into the control
design process. At the foundation of these methodologies lies the
necessity for accurate and physically consistent modeling of the
platform’s behavior. Studies such as [5] have developed
comprehensive multibody models that couple the dynamics of a
mecanum-wheeled platform with an onboard manipulator, thereby
revealing the impact of inertial and interaction forces on system
response. Likewise, the work in [6] provides insights into actuator-
level dynamics and highlights how saturation effects influence
overall stability and performance. These investigations reaffirm that
controller design cannot be decoupled from the underlying physics
of the system. Experimental validation continues to be a cornerstone
of mobile robot research. Several studies have confirmed that
performance on real platforms can differ significantly from
simulation predictions unless hardware-specific factors—such as
drivetrain backlash, roller misalignment, and friction modeling—are
taken into account. For example, [7] and [8] presented full design
and control implementations of omnidirectional platforms,
demonstrating how mechanical configuration and sensor integration
influence trajectory accuracy. Similarly, [9] introduced novel
mechanical designs aimed at mitigating common sources of
disturbance and enhancing motion fidelity. In light of the challenges
and findings described above, this paper introduces a conceptual
control system architecture tailored for a four-wheeled
omnidirectional mobile robot equipped with mecanum wheels. The
primary objective is to systematically identify and formalize the
modeling assumptions, control structures, and design trade-offs that
must be considered when targeting high-performance and robust
behavior in realistic industrial scenarios. Particular emphasis is
placed on modularity and scalability, making the proposed approach
amenable to integration within modern Industry 4.0 production
pipelines. This paper is a continuation of the work described in [10].
2. Critical Aspects and Limitations in Control
The control of four-wheeled omnidirectional mobile platforms
with mecanum wheels involves numerous challenges that stem not
only from the unique mechanical structure but also from inherent
nonlinearities and uncertainties in their dynamic behavior. Although
such platforms offer holonomic motion and flexibility in navigation,
their accurate and robust control remains a non-trivial task in real-
world conditions.
INNOVATIONS 2025
43
2.1 System Nonlinearities and Dynamic Coupling
A significant limitation in controlling mecanum-wheeled
platforms arises from the strong nonlinear coupling between
translational and rotational motions. Unlike standard differential
drive robots, the motion generated by each mecanum wheel
simultaneously affects all degrees of freedom. This results in a
multi-input multi-output (MIMO) system with coupled, nonlinear,
and time-varying dynamics. Moreover, the dynamics are influenced
by interaction forces at the rollerground interface, variable contact
patterns, and load-dependent friction effects, all of which contribute
to deviations from ideal kinematic behavior. As shown in [11], even
minor deviations in wheel alignment or loading can induce
considerable trajectory drift due to nonlinear interaction forces.
These nonlinearities limit the effectiveness of linear control
techniques and demand advanced nonlinear or adaptive strategies.
However, such methods must be carefully designed to balance
robustness and computational tractability.
2.2 Vibration and Ground Interaction
Due to the discontinuous nature of ground contact from the
angled rollers, mecanum wheels inherently introduce vertical
vibrations and wheel chatter. These phenomena cause perturbations
in odometry and sensor readings, which can propagate into control
loops and degrade performance. Work such as [12] demonstrates
how vibration effects can be mitigated through mechanical redesign
and controlaware modeling, yet these issues remain relevant in
standard implementations. For high-precision tasks or when
operating on uneven terrain, the control strategy must be capable of
handling or compensating for high-frequency dynamic
perturbations.
2.3 Reinforcement Learning Limitations for Control
With the growing popularity of machine learning in robotics,
reinforcement learning (RL) has emerged as a promising paradigm
for training controllers in complex, nonlinear systems. Actor–critic
methods, in particular, offer the benefit of continuous control and
online adaptation. However, applying these methods to mecanum-
based platforms introduces multiple challenges.
2.3.1 State-Space Explosion
The full-state representation of the platform includes nonlinear
combinations of velocities, orientation, wheel speeds, and potential
contact forces. This results in a high-dimensional observation space
that requires function approximators, such as deep neural networks,
to represent the value and policy functions. Consequently, this
increases sample complexity and significantly prolongs the training
time required for convergence in reinforcement learning.
2.3.2 Reward Shaping
Accurate reward definition is critical for reinforcement learning
agents to converge toward desirable behaviors. Poorly defined or
sparse rewards can lead to inefficient learning and may encourage
unsafe or unstable policies. This challenge is particularly
pronounced in systems with partial observability, delayed feedback,
or intricate dynamic interactions—characteristics inherent to
mecanum-wheeled platforms.
2.3.3 Exploration Risk
Reinforcement learning agents explore the environment to
discover effective strategies, but in physical systems such
exploration can lead to unsafe or damaging behaviors. For highly
coupled and sensitive systems like mecanum platforms, na¨ıve
exploration strategies can result in collisions, mechanical stress, or
loss of control. This poses a major obstacle for real-world
deployment without prior training in simulation and robust sim-to-
real transfer techniques.
2.3.4 Training Instability
Actor–critic architectures, while powerful for continuous
control, are inherently sensitive to training instabilities. These
include sensitivity to learning rates, entropy coefficients, gradient
noise, and offpolicy data contamination. Such issues are further
exacerbated in systems with unmodeled nonlinearities or
unobservable state variables, making consistent and safe training on
real platforms challenging. A recent application of RL to mecanum-
wheeled platforms showed promising simulation results, but the
authors noted severe performance degradation when transferring the
policy to a real robot.
3. Platform Modeling
Accurate modeling of the mobile platform is a foundational
component in the design of an effective control system. For
omnidirectional platforms equipped with mecanum wheels (Fig.1),
the model must capture the coupling between translational and
rotational motions, as well as the mapping between wheel-level
actuation and global motion. This section presents both the
kinematic and dynamic representations of a four-wheeled mecanum
platform.
Fig. 1 Schematic model of OMR.
3.1 Kinematic Model
The kinematic model describes the relation between the
platform’s body-frame velocity vb =[vx,vy,ω]T and the angular
velocities of the four wheels. It assumes no wheel slip and ideal
contact conditions. The forward kinematic equation, mapping
platform velocity to wheel angular velocities ϕ = [ϕ1,ϕ2,ϕ3,ϕ4]T, is
given by (1).
where r is the wheel radius, a and b are the half-length and
halfwidth of the platform, and L is the lever arm corresponding to
the distance from the platform center to the wheel axis in the plane.
The inverse kinematic relation, used for estimating platform
velocity from wheel speeds, is expressed as (2).
INNOVATIONS 2025
44
The above kinematic model can be integrated over time to
update the robot pose (x, y, θ ) as (3).
with the rotation matrix (4).
3.2 Nonlinear Dynamic Model
To accurately describe the physical behavior of the platform
beyond purely kinematic assumptions, a nonlinear dynamic model
is formulated. It captures the inertial and Coriolis effects, rolling
resistance, and the contribution of the individual wheel velocities to
the global motion. The model assumes that the low-level controllers
of the wheel drives enforce the desired angular velocities ϕ, which
generate net forces and torques on the body. The platform state is
defined as (5).
where (x, y) is the position of the platform in the global frame,
(vx,vy) is the velocity in the body frame, θ is the heading angle, and
ω is the angular velocity around the vertical axis.
The nonlinear dynamic equations are then given by (6).
The input-dependent forces and torque are defined as (7).
where:
• m is the mass of the platform,
• Iz is the moment of inertia around the vertical axis,
• dx, dy, and dω are damping coefficients,
This dynamic model allows capturing the nonlinear coupling
between linear and angular motion, as well as dynamic effects
caused by the rotating wheels. It is suitable for use in simulation,
controller design, and reinforcement learning environments where
inertial effects cannot be neglected.
3.3 Modeling Assumptions
The presented models rely on the following assumptions to
maintain tractability and enable efficient computation:
• Flat, rigid surface: The platform operates on a smooth,
horizontal, and rigid surface, free from irregularities, elevation
changes, or compliance.
• Rigid-body approximation: The mobile base is treated as a
rigid body. Flexibility of the chassis and dynamic effects from
wheel suspension are ignored.
• Symmetric mass distribution: The mass of the platform is
evenly distributed, and its center of mass coincides with the
geometric center.
• Direct wheel velocity control: It is assumed that low-level
motor controllers track the desired wheel angular velocities ϕ
with negligible delay and without saturations.
• Linear damping: The resistive forces acting on the platform
due to friction and drag are modeled as viscous damping terms
proportional to the velocities in the x, y, and θ directions.
• Force approximation from kinematics: The net force and
torque generated by wheel motion are inferred directly from
the inverse kinematic transformation, scaled by the wheel
radius r, rather than being derived from detailed contact force
models.
These assumptions allow the model to balance physical realism
with analytical simplicity, making it suitable for control synthesis,
real-time simulation, and learning-based applications. However,
they should be carefully revisited when deploying the system in
environments where wheel slip, uneven terrain, or actuator
limitations are significant.
4. Control System Concept
To achieve autonomous and robust control of the
omnidirectional mobile platform in complex, unstructured
environments, a learning based approach is adopted. Traditional
model-based controllers often require precise system identification
and struggle to generalize in the presence of modeling uncertainties
or dynamic obstacles. As an alternative, a reinforcement learning
(RL) framework is proposed, in which the control policy is trained
directly through interaction with the environment. The proposed
control architecture is based on an actor–critic agent that learns to
map sensory observations to wheel velocity commands. The agent
is trained in a virtual warehouse environment with a flat 2D map,
where it perceives its surroundings via a simulated lidar sensor and
learns to reach designated goals while avoiding collisions.
4.1 Architecture Overview
The control system integrates a reinforcement learning agent
with a simulated environment representing the omnidirectional
platform dynamics. As shown in Fig. 2, the actor–critic agent
generates continuous control actions, which are preprocessed and
applied to the dynamic model of the platform. The resulting motion
is captured via simulated sensors, providing state feedback for
reward calculation and observation generation.
Observations, including lidar data and setpoints, are processed
and used as input for the next policy inference. This loop continues
over an episode, enabling the agent to learn goal-directed navigation
through interaction. The architecture supports end-to-end training
and is fully compatible with episodic RL frameworks. It also allows
modular substitution of components (e.g., dynamics model or sensor
simulation) and facilitates sim-to-real transfer through consistent
state and action interfaces.
INNOVATIONS 2025
45
Fig. 2 System-level architecture of the reinforcement learning-based control
loop.
4.2 Reinforcement Learning Formulation
The RL control strategy is formulated as a Markov Decision
Process (MDP) with continuous state and action spaces. The
observation vector consists of 38 values: 36 equidistant distance
measurements from a simulated 2D lidar sensor that span the [−π,π)
range in steps of π/18, and 2 additional components that represent
the relative displacement to the target position along the x and y
axes in the global frame.
The action space is defined by four continuous control signals,
each corresponding to the angular velocity command for one of the
mecanum wheels. These actions are produced by the actor network
using a tanh activation function, and are subsequently scaled to the
physical velocity bounds of the platform’s wheel drives.
The reward function is defined as follows (8).
where:
• dlidar – vector of lidar distances at time step t,
• dg – distance to the goal,
• v, vy – linear platform velocities in the body frame,
• ω – angular velocity around the vertical axis,
• α, β, γ1, γ2, γ3 – weighting coefficients.
The reward encourages the agent to minimize the distance to the
goal, while avoiding obstacles and suppressing erratic motion. An
episode ends upon reaching the goal, colliding with an obstacle, or
exceeding the 100-second time limit (1000 steps at 0.1s resolution).
The initial positions are randomized on a fixed obstacle map, while
the target location remains constant, fostering generalizable path
planning behavior.
4.3 Actor-Critic Policy Design
The control policy is implemented using a Deep Deterministic
Policy Gradient (DDPG) agent with separate neural networks for
the actor and the critic. Both networks are initialized with random
weights and trained from scratch using gradient-based optimization.
The actor network receives the full observation vector and outputs
four continuous actions representing the normalized angular
velocity commands for each mecanum wheel. The network
architecture consists of two hidden layers with 50 neurons each and
ReLU activations, followed by a final fully connected layer with
tanh activation to constrain outputs to the [−1,1] range. These values
are then scaled to the maximum wheel angular speed, equal to
104.72 rad/s (corresponding to 1000 RPM), ensuring smooth and
bounded control. The critic network estimates the Q-value of a
state–action pair and follows a dualinput architecture. The state and
action vectors are processed through separate fully connected layers
and merged via an addition layer. The merged features pass through
a ReLU activation and a final linear output layer. This structure
enables the critic to model complex interactions between the system
state and the action space. Training stability is enhanced using
standard regularization techniques, including L2 weight penalties
and gradient thresholding. The optimizer uses separate learning
rates for the actor and critic (1e-4 and 1e-3, respectively), and
training is performed using mini-batches sampled from a replay
buffer of size 106. Exploration noise is added via decaying Gaussian
process noise on the action outputs. The DDPG agent uses a
discount factor of 0.995 and is trained in a continuous action space
using episodic learning.
5. Advantages and Challenges
The RL-based control framework introduced in this work shows
promising capabilities for complex mobile platforms. However, its
deployment in practical scenarios requires a careful assessment of
both system-specific behavior and general methodological
limitations.
5.1 Application-Specific Observations
The use of RL for the presented mecanum-wheeled platform
introduces several practical advantages. Most importantly, the
control policy is trained directly through interaction with the
environment, bypassing the need for precise system identification or
handcrafted control strategies. This approach inherently captures the
coupled nonlinear dynamics of the platform, including the
interaction between translational and rotational motion, which are
difficult to address using classical control techniques. The actor–
critic structure with continuous outputs enables smooth modulation
of wheel speeds, contributing to efficient and natural motion. The
training process incorporates the full sensing–actuation loop, with
lidar-based perception and physical velocity constraints, allowing
the agent to learn feasible and collisionfree navigation. The
modularity of the simulation architecture facilitates rapid
prototyping, hyperparameter testing, and potential extensions to
more complex tasks. Nonetheless, several challenges remain. The
performance of the trained agent is sensitive to the fidelity of the
simulation environment. Simplified modeling assumptions—such as
perfect wheel actuation and a flat surface—may cause discrepancies
when transferring the learned policy to a real-world system.
Furthermore, the training process is computationally intensive and
may require substantial tuning to achieve consistent convergence
and generalization across randomized initial conditions.
5.2 General Insights on Learning-Based Control
In a broader context, reinforcement learning offers a flexible
and scalable control paradigm for omnidirectional mobile robots,
particularly those operating in unstructured or dynamic
environments. Unlike model-based methods, RL does not require
explicit system inversion or trajectory planning, making it well-
INNOVATIONS 2025
46
suited for systems with highdimensional, coupled dynamics—such
as mecanum- or omniwheelbased platforms. The policy can be
optimized to fulfill multiple objectives simultaneously, including
obstacle avoidance, energy minimization, or goal-reaching, all
within a unified reward framework. The continuous nature of the
action space allows for smooth actuator commands and better
adaptation to perturbations or environmental variability. However,
these benefits are accompanied by fundamental limitations.
Training efficiency is often low, exploration may result in unsafe or
unstable behavior, and policy transfer between simulated and
physical domains remains a major obstacle. Reliable deployment of
RL-based controllers demands high-quality simulation
environments, careful reward shaping, and often additional
techniques such as domain randomization, safety constraints, or
fine-tuning on real hardware. Another open challenge lies in the
theoretical analysis of stability and robustness for learning-based
controllers. Unlike classical methods, which offer formal guarantees
under specific assumptions, reinforcement learning policies are
typically represented by neural networks with limited
interpretability and no inherent stability constraints. As a result,
ensuring safe behavior under disturbances, time delays, or model
mismatches remains an active area of research, particularly for
safety-critical applications involving physical interaction with the
environment.
6. Conclusion
This work presented a comprehensive control concept for a
mecanum-wheeled omnidirectional mobile platform, combining
dynamic modeling with learning-based control via deep
reinforcement learning. The proposed architecture leverages an
actor–critic structure and a simulated training environment to enable
autonomous navigation using lidar perception and continuous wheel
velocity control.
The modeling approach includes both kinematic and nonlinear
dynamic formulations, facilitating realistic motion behavior for
training and evaluation. The reinforcement learning agent
successfully learns to navigate toward a fixed goal while avoiding
collisions in a cluttered environment, without requiring explicit
trajectory planning.
While the simulation-based framework offers clear advantages in
adaptability and system modularity, challenges such as sim-to-real
transfer, training instability, and lack of formal stability guarantees
remain open. Future work will address these limitations through
domain randomization, real-world deployment, and the integration
of safety-aware learning mechanisms.
FUNDING
This research was funded by the AGH University of Krakow –
Excellence Initiative – Research University, Action. 12 –
Integration of educational process with scientific research – Support
for student research clubs – Rector's grants, 5rd edition of the
competition (ID 12443).
References
[1] Y. Liu, J. J. Zhu, R. L. Williams, and J. Wu, ―Omni-
directional mobile robot controller based on trajectory
linearization,‖ Robotics and Autonomous Systems, vol. 56, no.
5, pp. 461–479, 2008. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S092188900
7001431
[2] M. R. Azizi, A. Rastegarpanah, and R. Stolkin, ―Motion
planning and control of an omnidirectional mobile robot in
dynamic environments,‖ Robotics, vol. 10, no. 1, 2021.
[Online]. Available: https://www.mdpi.com/2218-6581/10/1/
48
[3] X. Li and A. Zell, Motion Control of an Omnidirectional
Mobile Robot. Berlin, Heidelberg: Springer Berlin
Heidelberg, 2009, pp. 181–193. [Online]. Available:
https://doi.org/10.1007/ 978-3-540-85640-5 14
[4] F. Nesci, F. Amato, D. Dragone, C. Cosentino, and A.
Merola, ―Control of a three-wheeled omnidirectional mobile
robot via a mixed ftb/h approach,‖ International Journal of
Dynamics and Control, vol. 12, no. 9, pp. 3461–3471, 2024.
[5] S. G. Tzafestas, T. Krikochoritis, and A. Melfi, Modeling and
Control of an Omnidirectional Mobile Robot. Boston, MA:
Springer US, 2000, pp. 391–402. [Online]. Available:
https://doi.org/10.1007/978-0-387-35529-0 37
[6] X. Li and A. Zell, ―Motion control of an omnidirectional
mobile robot,‖ in Informatics in Control, Automation and
Robotics. Springer, 2008, pp. 181–193.
[7] C. Caceres, J. Ros´ ario, and D. Amaya, ―Design, simulation,
and´ control of an omnidirectional mobile robot,‖
International Review of Mechanical Engineering, vol. 12, no.
4, p. 382, 2018.
[8] S. Ma, C. Ren, and C. Ye, ―An omnidirectional mobile robot:
Concept and analysis,‖ in 2012 IEEE International
Conference on Robotics and Biomimetics (ROBIO), 2012, pp.
920–925.
[9] T. Terakawa, M. Komori, K. Matsuda, and S. Mikami, ―A
novel omnidirectional mobile robot with wheels connected by
passive sliding joints,‖ IEEE/ASME Transactions On
Mechatronics, vol. 23, no. 4, pp. 1716–1727, 2018.
[10] P. Bałazy, P. Gut, and P. Knap, ―Positioning algorithm for
agv autonomous driving platform based on artificial neural
networks,‖ Robotic Systems and Applications, vol. 1, no. 2,
pp. 41–45, 2021.
[11] Z. Zhewen, Y. Hongliu, W. Chengjia, H. Pu, and W. Jiangui,
―A comprehensive study on mecanum wheel-based mobility
and suspension solutions for intelligent nursing wheelchairs,‖
Scientific Reports, vol. 14, no. 1, p. 20644, 2024.
[12] P. S. Yadav, J. Mohanta, V. Agrawal, and M. F. Ahmed,
―Vibration-reducing modelling of a mecanum wheel based on
equivalent stiffness and strength analysis,‖ Proceedings of the
Institution of Mechanical Engineers, Part E: Journal of
Process Mechanical Engineering, p. 09544089231212594,
2023.
INNOVATIONS 2025
47
Engineering tool integration for complex system simulation and optimization
Szymon Podlasek1*, Urszula Jachymczyk1
AGH University of Krakow, Poland1
podlasek@agh.edu.pl, delhi@student.agh.edu.pl
Abstract: The integration of engineering support tools is essential for the efficient modeling, simulation, and optimization of complex
technical systems. This paper presents a dynamic model of a micro-combined heat and power (mCHP) system, developed to validate the
feasibility of integrating various computational environments. The approach leverages modular architectures, enabling seamless data
exchange between distinct software platforms, thus supporting both detailed thermodynamic analysis and real-time performance
optimization. The flexibility of this approach allows for the inclusion of diverse analytical frameworks, including neural network-based
optimization, data-driven control strategies, and alternative programming languages, without being limited to a single computational tool.
This adaptability makes the proposed architecture particularly suitable for evolving engineering applications, where rapid prototyping and
iterativ e refinement are critical. The study highlights the potential of such integrated environments to enhance the design and operational
efficiency of energy systems, providing a scalable foundation for future expansions.
Keywords: ENGINEERING TOOL, INTEGRATION ENERGY SYSTEMS, DATA EXCHANGE, SIMULATION, OPTIMIZATION
1. Introduction
In response to the increasing complexity of energy systems,
effective modeling and optimization require the cooperation of
various computational tools and simulation environments. Modern
engineering approaches increasingly rely on the integration of
multiple platforms - from programming languages such as Python
to engineering environments like MATLAB, Simulink, or ANSYS,
and data analysis tools such as Excel [1-3]. A key aspect of this
integration is the ability to exchange data flexibly between system
components while maintaining consistency in simulation and
control algorithms.
The integration of different computational environments
enables the exchange of variables and supports further data analysis,
visualization, or optimization. This architecture allows users to
benefits of various tools within a single, unified workflow, without
being limited to a specific piece of software (Fig.1).
Fig. 1 Idea diagram showing the possibilities of data exchange between
different engineering tools.
A common approach involves using one tool to simulate the
dynamic behavior of the physical system, while a separate
environment is responsible for generating control signals based on
the simulation results. This setup enables the use of advanced
algorithms, including optimization methods and control strategies,
to dynamically adjust the system‘s behavior. Increasingly, neural
algorithms such as convolutional neural networks (CNNs) or
reinforcement learning (RL) methods are applied for this purpose,
successfully supporting event detection and enabling process
optimization [4-6].
These methods require real-time access to system data and the
ability to quickly iterate and validate results. A modular integration
framework also facilitates future system expansion and adaptation
to changing operational and technological requirements [7-8]. The
use of inter-environment communication supports coupled
simulations and the implementation of hybrid optimization
strategies, which are widely discussed in recent literature on
advanced energy systems [9-11].
2. Model of the Analyzed Object
The dynamic model of the steam generator in the micro-
combined heat and power (mCHP) system enables the calculation
of thermodynamic parameters of the working fluid in the form of
water vapor and its condensate within the Rankine cycle, as well as
global system parameters [12, 13]. The system is based on the
combustion of heating oil, which provides the energy required to
generate steam in the steam generator. The working fluid is first
preheated in a tank that also functions as a deaerator, and then
further heated by hot flue gases to reach the boiling point at a
specified pressure. In the system, the generated steam is first
dehydrated and then directed to a steam engine connected to an
electric generator. Subsequently, the expanded steam is passed to a
condenser and a condensate tank in order to restore the initial
parameters of the working fluid at the steam generator inlet and to
remove gases dissolved in the condensed water.
The modeling and simulation of the mCHP system were
conducted using both built-in components from the software library
and user-defined components. The components available in the
software library were experimentally validated and/or based on
actual operational data, ensuring high reliability of the simulation
results. The system model developed in TRNSYS provides dynamic
data related to system operation. Temperature profiles, flow rates,
power levels, and control functions for all system components are
calculated through the simulation.
Fig. 2 A model of mCHP installation in TRNSYS software.
The developed model of the mCHP installation, based on a fire-tube
steam generator and discussed in the earlier sections of this work,
consists of the following components (Fig.2):
INNOVATIONS 2025
48
Steam generator, producing steam at a specified pressure using
heating oil combustion: component Type 638;
Steam engine with generator, generating electrical energy by
expanding steam in pistons: component Type 592a;
Condenser, removing heat from the expanded working fluid in
order to restore the steam parameters at the steam generator
inlet: component Type 598;
Pump, pressurizing the condensed water to the desired pressure
level: component Type 597;
Condenser circulation pump, driving the flow in the condenser
loop: component Type 3d;
Heat recovery, simulating a heat exchanger that extracts thermal
energy from the condenser's cooling stream: component Type
5b.
In addition, the mCHP installation model includes
supplementary components used to carry out the system simulation
and to generate output data. These elements include calculators for
data input and processing, as well as printers and plotters.
Specifically, two calculators were employed to input the main
parameters into the model, perform unit conversions, and calculate
efficiency. The model parameters will be presented later in terms of
kilowatts (kW), while the equation implemented in the model is
defined by the function eta_ele_cycle =
P_TURB_kW/max(Q_BOIL_kW,1) which corresponds to equation
(e1), where the turbine power output 𝑁𝑠𝑝 is related to the thermal
power of the steam generator 𝑄𝑤𝑡 :
𝜂𝑘𝑜𝑔 = 𝑁𝑠𝑝
𝑄𝑤𝑡 (e1)
Additionally, the model includes a component used to calculate
thermodynamic parameters based on the specified steam conditions
at the outlet of the steam generator. This component, referred to as
―Enthalpy,‖ is implemented using Type 58. It is worth noting that
the calculation of system efficiency (eta_ele_cycle) does not
account for the power consumption of the pumps in the steam and
condenser loops.
Fig. 3 Structure of connection and transmission of parameters between
blocks in Trnsys environment.
Figure 3 shows an example of the connection structure used to
develop the mCHP system model for the pump and steam generator.
Dynamic parameter exchange between model components is
observed during the simulation process.
3. Model Parameters and Preliminary Simulation
The system was simulated by setting the duration to 1.0 hour
with a time step of 0.01 hour. Under such conditions, the dynamic
model was assumed to operate in a static mode to determine the
operating conditions at the set conditions. The parameters used for
the developed model are shown in Table 1.
The developed model was used to evaluate the performance and
efficiency of the presented Rankine cycle that can be achieved
under nominal operating conditions. In particular, the model made it
possible to determine the power and efficiency of the system
depending on the main operating parameters, such as the mass flow
of steam and the maximum pressure of the steam system.
To enable data exchange, the model was extended with the
Type155 block. The connection scheme be presented in Section 4.
Table 1: Operating Parameters of the Modeled Cogeneration System.
Main Parameters of the Steam System
Maximum pressure: 10 bar
Maximum temperature: 180.0 °C
Steam Generator
Thermal power: 100.0 kW
Generator efficiency: 0.90
Combustion efficiency: 0.99
Condensate Pump
Nominal mass flow rate: 2000 kg/h
Power: 0.5 kW
Power conversion factor: 0.05
Steam Engine
Nominal power: 15.0 kW
Steam outlet pressure: 1.0 bar
Isentropic efficiency: 0.75
Heat Recovery
Working fluid: water
Inlet temperature: 25.0 °C
Flow rate: 1500 kg/h
Condenser
Pinch point temperature difference: 15.0 °C
Minimum pressure: 0.5 bar
Cooling degree: 3.0 °C
Main Pump
Nominal mass flow rate: 150 kg/h
Overall efficiency: 0.60
Motor efficiency: 0.90
The results of the simulation are shown in Figure 4, where the
relationship between the flow rate and the system's output and
efficiency is depicted. It can be seen from the graphs that maximum
power generation is achieved for a given value of flow (150 kg/h),
and with a decrease in this value, electrical power decreases
sharply. In this range, the power generated is proportional to the
mass flow, since the steam temperature at the steam turbine inlet is
constant at 185°C. On the other hand, with a flow value greater than
150 kg/h, the temperature at the turbine inlet drops significantly,
which affects the generated power, as well as the efficiency of the
system.
Fig. 4 Analysis of the effect of changing the maximum pressure on the power
and efficiency of the system.
The effect of maximum cycle pressure on performance is shown
in Figure 4. Trends in both power and efficiency highlight that the
system performs better at the highest possible pressure (taking into
account the temperature of the heat source). Increasing the nominal
pressure from 5.0 to 10.0 bar allows a 27.8% increase in power
output. Simplified simulations developed using TRNSYS software
have shown that the developed mCHP system based on a flame
INNOVATIONS 2025
49
steam generator, operating at a pressure of 10.0 bar and a maximum
steam temperature of 185°C, can achieve power and efficiency of
more than 10kW and 10%, respectively, with optimized steam flow
values. The developed model allows to carry out dynamic
simulations of the mCHP system under selected operating
parameters, but further work needs to be expanded and the
constraints and boundary conditions need to be made more precise.
The model does not take into account factors that can affect the
final power and efficiency of the system. It provides a basis for
further work on the design of a trigeneration system, where the heat
can also be used for cooling purposes.
4. Data Exchange Structure
In the context of control system research, a key component is
the Type155 block, which enables data exchange between the
TRNSYS environment and MATLAB. This functionality was
implemented to integrate custom control algorithms into the
simulation environment modeling the operation of a trigeneration
installation. As a result, the TRNSYS simulation can be controlled
using optimization algorithms developed in MATLAB. A graphical
representation of the data exchange mechanism and the structure of
parameter mapping is shown in Figure 5. In addition to interaction
between software environments, it is also possible to import input
data from auxiliary files, such as weather data or consumption
profiles. All constant parameters, input variables, and external data
are processed, and the results of the computational operations are
provided as output variables.
Fig. 5 Data transfer structure between environments using the Type155.
To enable training and simulation of a reinforcement learning
(RL)-based algorithm, additional scripts were developed to facilitate
data exchange between decision-making programs. In this
configuration, the program invoked by the Type155 block serves
solely as a bridge between the neural network training environment
and the simulated operation of the cogeneration unit. On the
MATLAB side, data transmission was established using a
‗localhost‘ connection, which is used to set up network
communication within the same device. In this case, it allows two
separate MATLAB programs to communicate with each other.
The communication was implemented using the tcp_client and
tcp_server tools, which enabled the transfer of data from the
TRNSYS simulation and the return of computed results. This setup
allowed the RL network to learn based on observations from the
dynamic simulation of the trigeneration system and to generate
actions in response. The performance of these actions was then
rewarded or penalized. The internal structure and logic of the RL
algorithms implemented in the MATLAB environment are
described in [14, 15].
The main component of the reward function (R1) is based on the
difference between the power demand (𝑝𝑜𝑤𝑒𝑟𝐷𝑒𝑚𝑎𝑛𝑑) and the
power generated (𝑝𝑜𝑤𝑒𝑟𝐺𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑) (e2).
𝑅1= − 𝑎𝑏𝑠(𝑝𝑜𝑤𝑒𝑟𝐷𝑒𝑚𝑎𝑛𝑑 − 𝑝𝑜𝑤𝑒𝑟𝐺𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑) (e2)
The use of the absolute value allows penalizing the agent‘s
actions in cases of both power surplus and deficit. Applying the
square root to the difference intensifies the impact of deviations,
both when the mismatch is large and when approaching zero error.
An important aspect is also the inclusion of a penalty in situations
where no variation is detected in the decision variables relative to
the moving average over the last five iterations. This penalty is
applied only if the deviation exceeds 0.1 kW. This threshold was
defined as the acceptable deviation from the network's target
performance. It helps prevent excessive oscillations near the target
value and reduces the risk of misleading the agent during training—
especially in scenarios where a zero error is achieved, but the agent
is still required to take action through its decision variables.
In the case of fixed settings, the trajectory of this parameter
follows a straight line, while in the learning scenario, each decision
variable independently affects power generation. The ranges of the
decision variables were defined within 55–95% to avoid operation
in undesired regions. The variable demand defined in the model
falls within the range of 5 to 10 kWe, while control of the decision
variables allows the system to generate power within the range of 4
to 10 kWe. The remaining simulation parameters result from the
decision variable settings and are recalculated during each program
iteration (every 3 minutes in a 1000-hour simulation). The main
objective of the algorithm is to match the generated power to the
consumer‘s demand over time.
Fig. 6 Data transfer structure between environments using the Type155.
Figure 6 presents selected curves illustrating the operational
parameters of the simulated trigeneration system over time. The
startup process of the installation is clearly visible, along with the
initial interactions of the neural network, reflected in the
fluctuations of the generated power (P_TURB).
5. Conclusion
The developed simulation model of the trigeneration energy
system enables full integration of external control algorithms with
the dynamic structure of the thermodynamic system. A key role is
played by the Type155 component, which facilitates bidirectional
real-time data exchange between the TRNSYS and MATLAB
environments. This makes it possible to incorporate a decision-
making layer based on reinforcement learning, which dynamically
adjusts control signals to match the variable demand profile. The
model allows for the observation of how control decisions affect the
INNOVATIONS 2025
50
physical parameters of the system, such as turbine power and
efficiency, creating a foundation for the analysis and development
of control strategies in a virtual environment. Importantly, the tool
fulfills all key aspects of proper optimization: it attempts to achieve
the objective function by maximizing the reward and responds by
adjusting parameters within the allowed ranges.
The proposed architecture confirms the relevance of using
multiple engineering tools for comprehensive energy system design.
The learning process, based on both simulation and experimental
data, allows for safe testing of algorithms without the risk of
destabilizing a real installation. The model can serve as a research
tool for the implementation and evaluation of adaptive strategies in
the energy sector.
Funding
This research was funded by the Excellence Initiative –
Research University, Action. 12 – Integration of educational
process with scientific research – Support for student research clubs
– Rector's grants, 5rd edition of the competition (ID 8980).
6. References
1. Nichols, Daniel. "Arduino-based data acquisition into Excel,
LabVIEW, and MATLAB." The Physics Teacher 55.4 (2017)
2. Papkov, Vyacheslav, Nikita Shadymov, and Dmitry
Pashchenko. "CFD-modeling of fluid flow in Ansys Fluent using
Python-based code for automation of repeating calculations."
International Journal of Modern Physics C 34.09 (2023)
3. Elomari, Youssef, et al. "A hybrid data-driven Co-simulation
approach for enhanced integrations of renewables and thermal
storage in building district energy systems." Journal of Building
Engineering 104 (2025)
4. Knap, Paweł, Krzysztof Lalik, and Patryk Bałazy. "Boosted
convolutional neural network algorithm for the classification of the
bearing fault form 1-d raw sensor data." Sensors 23.9 (2023)
5. Balazy, Patryk, Kamil Pieprzycki, and Paweł Knap. "Robust
Reinforcement Learning For Overhead Crane Control With
Variable Load Conditions." 2024 25th International Carpathian
Control Conference (ICCC). IEEE, 2024
6. Lalik, K., et al. "Q-Learning neural controller for steam
generator station in micro cogeneration systems. Energies 14, 5334
(2021)
7. Meiers, Josef, and Georg Frey. "Interfacing TRNSYS with
MATLAB for Building Energy System
Optimization." Energies 18.2 (2025)\
8. Adesanya, Misbaudeen Aderemi, et al. "Deep reinforcement
learning for PID parameter tuning in greenhouse HVAC system
energy Optimization: A TRNSYS-Python cosimulation
approach." Expert Systems with Applications 252 (2024)
9. Podlasek, Szymon, et al. "Application of ANN control
algorithm for optimizing performance of a hybrid ORC power
plant." Energy 306 (2024)
10. Bakeer, Abualkasim, et al. "Effect of mission profile resolution
on photovoltaic energy yield prediction in Python and
MATLAB." 2021 IEEE 15th International Conference on
Compatibility, Power Electronics and Power Engineering (CPE-
POWERENG). IEEE, 2021
11. Sudhakar, K., et al. "Modelling of a solar desiccant cooling
system using a TRNSYS-MATLAB co-simulator: A
review." Journal of Building Engineering 24 (2019)
12. Żołądek, Maciej, Rafał Figaj, and Krzysztof Sornek. "Energy
analysis of a micro-scale biomass cogeneration system." Energy
Conversion and Management 236 (2021)
13. Imran, Muhammad, et al. "Dynamic modeling and control
strategies of organic Rankine cycle systems: Methods and
challenges." Applied energy 276 (2020)
14. https://www.mathworks.com/products/reinforcement-
learning.html
15. Ciaburro, Giuseppe. MATLAB for machine learning. Packt
Publishing Ltd, 2017.
INNOVATIONS 2025
51
Prototype of a Wireless MEMS-Based Sensor Node within a Wireless Sensor Network Concept
Julia Jeleńska1,*
AGH University of Krakow, Poland
jelenska@student.agh.edu.pl
Abstract: The aim of this work is to implement a wireless communication system for MEMS-based sensors within the framework of Internet
of Things (IoT) applications, specifically in the context of Predictive Maintenance (PdM). The focus is placed on developing a functional
prototype of a wireless sensor node that enables efficient data acquisition and transmission from commercially available MEMS vibration
sensors. The solution leverages an ESP32 microcontroller for data handling and Wi-Fi communication, forming the basis of a scalable
wireless sensor network (WSN). The project emphasizes their integration into a wireless system architecture suitable for industrial
monitoring scenarios. This approach aims to demonstrate how low-cost MEMS sensors, when combined with IoT technologies, can
contribute to accessible and modular condition monitoring solutions aligned with Industry 4.0.
Keywords: MEMS, WIRELESS SENSOR NETWORK, PREDICTOVE MAINTENANCE, INDUSTRY 4.0, IOT
1. Introduction
The ongoing transformation of industrial systems towards the
principles of Industry 4.0 has increased the demand for intelligent,
interconnected, and self-monitoring infrastructure [1]. Although
many existing predictive maintenance solutions are technically
capable, they remain financially out of reach or overly complex for
small and medium-sized enterprises (SMEs). This concept paper
aims to bridge this gap by proposing a scalable, cost-effective, and
easy-to-integrate wireless sensor network (WSN) architecture. At
the heart of this system are low-cost microelectromechanical
systems (MEMS) accelerometers paired with widely available
microcontrollers [2]. MEMS sensors offer a compact and affordable
alternative to traditional solutions, making the technology suitable
for a wider range of industrial applications [3]. Effective PdM
practices not only reduce unplanned maintenance costs, but also
enhance equipment reliability, directly impacting production
continuity and quality [4]. This is particularly relevant in industries
where machinery failure can lead to costly production halts or even
safety hazards.
By integrating MEMS accelerometers with the ESP32 [5]
microcontroller, this wireless sensor node prototype achieves a
balance between cost efficiency and technical performance. The
proposed nodes are compact, energy efficient, and capable of
collecting high-resolution vibration and temperature data. Designed
with modularity in mind, the system can scale seamlessly by adding
synchronized nodes to maintain a precise time correlation between
measurements.
To enhance accessibility, the system uses Wi-Fi for data
transmission, enabling rapid deployment without the need for
extensive cabling. The data collected are transmitted to a centralized
database for processing, anomaly detection, and visualization. The
use of a time series database ensures efficient handling of high-
frequency data streams. In addition, the modular design of the
system allows it to adapt to a wide range of industrial scenarios [6].
By democratizing access to PdM technologies, this concept offers
SMEs an affordable path to implement intelligent maintenance
practices. The following sections discuss related research, system
architecture, potential use cases, and the anticipated impact on
industrial maintenance strategies.
2. System Architecture
The proposed architecture is designed to be both modular and
scalable, enabling easy integration and future expansion. Ensures
that it can evolve alongside technological advancements and
accommodate growing industrial demands.
2.1. Sensor Node
The WSN is built around a custom-designed sensor node that
integrates the IIS3DWB MEMS accelerometer with the ESP32
NodeMCU microcontroller. A prototype of the sensor node is
presented in Fig. 1.
Each sensor node has [7]:
• three-axis acceleration measurement,
• temperature measurement,
• data at sampling rates of up to 6 kHz,
• wide selectable bandwidth of up to 2.67 kHz,
• low noise density,
• full-scale acceleration ranges from ±2g to ±16g.
These characteristics make the IIS3DWB a highly suitable
choice for PdM use cases where accurate temporal resolution and
low-latency data handling are essential.
Each node operates autonomously and is powered by a 3.7V
rechargeable lithium-ion battery. To support energy autonomy and
operational safety, a dedicated power management circuit is
integrated into the PCB, enabling efficient power regulation and
secure battery charging.
Fig. 1 Prototype of MEMS-based sensor which could be use in WSN
architecture.
2.2. Wireless Communication Architecture and Data
Infrastructure
The proposed system architecture relies on Wi-Fi for data
transmission, eliminating the need for physical cabling. This
simplifies deployment, particularly in environments with limited
accessibility or complex machine layouts. In this concept, the
wireless connection is a fundamental design choice that supports
scalability, modularity, and ease of integration.
Each sensor node is built around an ESP32-based platform,
which includes an integrated Wi-Fi module. This component,
sourced from the ESP32 development ecosystem, offers a cost-
effective solution for wireless communication in IoT systems. As
the network may consist of numerous sensor nodes deployed
throughout a facility, one of the primary design considerations is
the temporal synchronization of data acquisition. To achieve this,
the Network Time Protocol (NTP) is implemented on all nodes.
Synchronization through NTP ensures that all sensor
measurements are time correlated, which is essential when
assessing vibration patterns from multiple angles or comparing
data across different machines on the same production line.
Data communication between the sensor nodes and the
central server is carried out using the HTTP protocol. Basic
INNOVATIONS 2025
52
diagram is shown in Fig. 2. This choice is motivated by the need
to handle relatively large volumes of high-frequency data
continuously generated by multiple nodes. HTTP provides a
scalable framework for transmitting these data efficiently and
securely. Each node sends structured measurement packets to a
back-end server, where the data is parsed and stored in an
InfluxDB time-series database. This database is optimized for
handling large-scale sensor data streams with high write
throughput and time-based query performance.
Fig. 2 Basic diagram for multi-node PdM architecture.
Data synchronization across multiple nodes is achieved using
the Network Time Protocol (NTP), ensuring accurate time
correlation among the distributed sensor nodes. This
synchronization is crucial for multi-point vibration analysis and
comparative diagnostics, as it guarantees that data collected from
various locations remains temporally aligned. Without proper
synchronization, phase shifts between measurements could lead to
inaccurate assessments, particularly in applications involving
vibration pattern analysis or machine condition monitoring. By
integrating NTP synchronization directly into the startup procedure,
the system minimizes human error and ensures that each
measurement cycle begins with accurately synchronized nodes.
Fig. 3 Schematic for nodes synchronization.
To initiate the measurement process, an application is provided
that allows operators to start, stop and manage data acquisition from
a PC or mobile device. When the measurement is triggered, the
application sends a synchronization signal to all sensor nodes,
which in turn requests the current time from an NTP server. The
nodes then adjust their internal clocks to match the received
timestamp, achieving precise temporal alignment before data
collection begins (Fig. 3). Once synchronized, the sensor nodes
begin data acquisition, packaging the collected data into structured
packets. These packets include a timestamp, sensor ID, and
measurement data, ensuring that each data point can be accurately
linked to a specific moment in time and a particular sensor location.
The proposed architecture envisions broader applications for the
collected data. Beyond standard monitoring, the infrastructure
supports integration with advanced systems such as virtual reality
environments, real-time diagnostics, and AI-based predictive
maintenance platforms.
3. Use Case and Applications
The proposed wireless sensor network system is designed to be
flexible and adaptable, catering to diverse industrial environments
ranging from traditional machine workshops to modern smart
factories. One of its key advantages is the ease of scalability and
customization, which allows it to be tailored to various machine
setups without significant adjustments. This makes it particularly
valuable for small and medium enterprises that often face financial
constraints when adopting proprietary condition monitoring
systems.
Using MEMS technology, sensor nodes offer a compact and
cost-effective alternative to piezoelectric-based solutions [8].
MEMS sensors are approximately ten times cheaper while
providing sufficient accuracy and sensitivity for PdM applications.
This cost advantage allows SMEs to deploy comprehensive
monitoring systems without incurring prohibitive expenses.
The WSN architecture is modular. Sensor nodes can be easily
mounted on industrial assets such as motors, compressors, CNC
machines, or other rotating equipment. Each node collects high-
resolution acceleration data across the XYZ axes and temperature
readings, which are important to identify early-stage mechanical
problems such as imbalance, misalignment, or bearing faults.
A fundamental feature of this system is its ability to scale
without requiring complex configurations. New sensor nodes can be
added seamlessly, with automatic synchronization. The plug-and-
play nature of the ESP32 firmware further enhances usability,
allowing rapid expansion or reconfiguration to accommodate
different machine layouts.
Furthermore, the wireless nature of the WSN eliminates the
need for extensive cabling, making it particularly advantageous in
facilities where machinery configurations frequently change or
where mobility is required. This allows for quick deployment in
both permanent and temporary monitoring setups.
Real-time data can be integrated with digital twins of the
factory floor, developed using platforms like Unity or Unreal
Engine. These virtual environments enable operators and engineers
to view the health status of machines through color-coded overlays
or animated vibration patterns. Such interactive visualizations
promote an intuitive understanding of the conditions of the
equipment, facilitating routine inspections.
In addition, maintenance personnel can receive proactive
notifications and detailed status reports through mobile apps or
smart devices, such as augmented reality glasses. These alerts are
based on anomaly detection algorithms running on historical data
stored in a time series database. By identifying deviations from
normal operational patterns, the system can prompt timely
maintenance actions, reducing downtime and prolonging the useful
life of the machine.
Integration of the system with real-time diagnostic platforms
and AI-based PdM algorithms also supports predictive maintenance.
Operators can remotely analyze machine performance, receive early
warnings of possible failures, and even simulate fault conditions
within a Virtual Reality training module [9]. This approach to
machine health management not only improves operational
INNOVATIONS 2025
53
efficiency, but also enhances workforce training, allowing a deeper
understanding of mechanical behaviors and maintenance strategies.
4. Discussion
The MEMS-based wireless sensor node presented within a
WSN concept demonstrates significant potential for scalable, cost-
effective predictive maintenance applications, especially in small
and medium-sized enterprises. Using MEMS accelerometers, the
system achieves a substantial reduction in cost compared to
traditional piezoelectric sensors, making high-frequency vibration
monitoring more accessible to a broader range of industrial settings.
One of the primary advantages of this architecture is its
modularity and ease of integration. The use of the ESP32
microcontroller with built-in Wi-Fi enables seamless data
transmission without the need for physical cabling, significantly
reducing installation time and costs. Additionally, NTP-based time
synchronization ensures that multinode data collection remains
coherent, which is important when analysing vibration data from
multiple measurement points.
However, some challenges remain to ensure data integrity and
consistency as the system scales up. Network congestion and
potential delays in NTP responses could affect the accuracy of
synchronization, particularly in environments with fluctuating
connectivity. To address this, it might be necessary to implement
local time correction algorithms or to use alternative
synchronization protocols in critical cases.
Moreover, reliance on Wi-Fi for data transmission could pose
limitations in industrial environments with high electromagnetic
interference or large physical obstructions. Future work could
explore the incorporation of alternative wireless technologies to
improve communication robustness and extend the operational
range of the system.
Despite these challenges, the proposed system demonstrates a
promising pathway for integrating PdM techniques into industrial
processes cost-effectively. Its flexible architecture supports real-
time monitoring, visualization, and scalable deployment, providing
a foundation for the next generation of intelligent maintenance
solutions.
5. Conclusion and Future Work
This paper posits a novel conceptual framework for a modular
and scalable wireless sensor network based on accelerometer nodes
based on MEMS, designed to facilitate predictive maintenance in a
variety of industrial environments. By addressing key impediments
to adoption, namely cost, architectural complexity, and proprietary
systems, this concept aims to enable Small and Medium Enterprises
to initiate their transition towards Industry 4.0 through a pragmatic
and extensible instrumentation paradigm.
The proposed framework exhibits significant theoretical
potential to transform conventional machinery condition monitoring
practices via the provision of comprehensive data analytics and
visualization modalities. The integrated aggregation of data streams
from heterogeneous machinery within a unified monitoring
architecture is theorized to enhance industrial process operational
efficiency and underpin proactive maintenance strategies, thereby
mitigating unscheduled downtime and associated maintenance
expenditures.
Future work will involve empirical validation through physical
instantiation and testing within operational environments to
ascertain system performance characteristics and reliability metrics.
Planned enhancements include the theoretical exploration and
potential implementation of energy optimization strategies, such as
adaptive sampling rates and duty cycling, alongside the
investigation of alternative network connectivity protocols. The
conceptual development of the Virtual Reality interface will be
extended to incorporate real-time interactive capabilities and guided
maintenance protocols, further augmenting the theoretical utility of
the framework.
Subsequent efforts will aim to conceptualize a user-centric
interface for on-site configuration and data retrieval, potentially
incorporating mobile application interfaces and augmented reality
overlays for real-time maintenance assistance. By synergistically
combining sensor-derived data with Artificial Intelligence-driven
PdM models, the framework has the potential to evolve into a
holistic platform for intelligent maintenance, supporting decision-
making processes with actionable analytical insights.
In parallel, the conceptual development of a cloud-based
management portal \cite{Wang2017} enabling remote monitoring
and configuration of distributed deployments, thereby theoretically
paving the way towards fully integrated digital twins and intelligent
factory ecosystems.
FUNDING
This research was funded by the AGH University of Krakow –
Excellence Initiative – Research University, Action. 12 –
Integration of educational process with scientific research – Support
for student research clubs – Rector's grants, 5rd edition of the
competition (ID 12443).
6. References
1. M. Achouch, M. Dimitrova, K. Ziane, S. Sattarpanah Kargan-
roudi, R. Dhouib, H. Ibrahim, M. Adda, “On Predictive Mainte-
nance in Industry 4.0: Overview, Models, and Challenges Ap-
plied Sciences”, 12, 8081 [2022], https://www.mdpi.com/2076-
3417/12/16/8081.
2. Hartzell, A. L., da Silva, M. G., Shea, H. R. “MEMS
Reliability”.
3. R. Keith Moble, “An Introduction to Predictive Maintenance”,
Elsevier, 24 Oct [2002] - 437.
4. WANG, K. “Intelligent predictive maintenance (IPdM)
system–Industry 4.0 scenario”. WIT Transactions on Engineering
Sciences, [2016], 113.1: 259-268.
5. ESP32 NodeMCU, “ESP32 NodeMCU Development Board”,
Espressif Systems
6. A. Khademi, F. Raji and M. Sadeghi, ”IoT Enabled Vibration
Monitoring Toward Smart Maintenance,” 2019 3rd International
Conference on Internet of Things and Applications (IoT), Isfahan,
Iran, [2019], pp. 1-6, doi: 10.1109/IICITA.2019.8808837.
7. STMicroelectronics, IIS3DWB: “High-performance 3-axis
digital vibration sensor”, Datasheet, DocID031727 Rev 3, March
[2022].Availableat:
https://www.mouser.de/datasheet/2/389/iis3dwb-1761424.pdf
8. Varanis, M., Silva, A., Mereles, A. et al. “MEMS
accelerometers for mechanical vibrations analysis: a
comprehensive review with applications”. J Braz. Soc. Mech. Sci.
Eng. 40, 527 [2018]. https://doi.org/10.1007/s40430-018-1445-5
9. Choi, S., Jung, K., Noh, S. D. “Virtual reality applications in
manufacturing industries: Past research, present findings, and
future directions”. Concurrent Engineering, 23(1), 40-63, [2015].
doi:10.1177/1063293X14568814
10. Wang, J., Zhang, L., Duan, L. et al. “A new paradigm of cloud-
based predictive maintenance for intelligent manufacturing”.
J. Intell. Manuf. 28, 1125–1137 [2017].
https://doi.org/10.1007/s10845-015-1066-0
INNOVATIONS 2025
54
ИЗУЧЕНИЕ ВЛИЯНИЯ ТЕРМОМЕХАНИЧЕСКОЙ ОБРАБОТКИ ЛАТУНИ,
ВКЛЮЧАЮЩЕЙ ПРЕДВАРИТЕЛЬНУЮ ТЕРМИЧЕСКУЮ ОБРАБОТКУ И
РАДИАЛЬНО-СДВИГОВУЮ ПРОКАТКУ, НА ИЗМЕНЕНИЕ ЕЕ МЕХАНИЧЕСКИХ
СВОЙСТВ
STUDY OF THE EFFECT OF THERMOMECHANICAL TREATMENT, INCLUDING PRELIMINARY HEAT
TREATMENT AND RADIAL SHEAR ROLLING, ON CHANGES OF BRASS MECHANICAL PROPERTIES
Abdrakhman Naizabekov1, Evgeniy Panin1,2, Sergey Lezhnev1, Pavel Tsyba2
1Rudny industrial University, Rudny, Kazakhstan
2 Karaganda Industrial University, Temirtau, Kazakhstan
Abstract: This work is devoted to the study of the effect of thermomechanical treatment, including preliminary heat treatment (annealing at
500°C) and deformation on a radial-shear rolling mill, on the change in the mechanical properties of brass rods of grades L63 and
LZHMTS59-1-1. The conducted studies have shown that the mechanical properties of brass rods of both brands, previously subjected to heat
treatment according to the selected mode during their deformation at the radial-shear rolling mill, change significantly. In particular, the
strength properties of these materials, averaged over the cross-section of the rods, are increasing, while the plastic properties, on the
contrary, are falling. At the same time, the decrease in plastic characteristics, namely elongation, for both grades of brass during radial-
shear rolling is within the normal range for the above materials subjected to severe plastic deformation during the implementation of various
metal forming methods.
Keywords: THERMOMECHANICAL TREATMENT, PRE-HEAT TREATMENT, ANNEALING, RADIAL SHEAR ROLLING, BRASS,
MECHANICAL PROPERTIES.
1. Введение
В настоящее время при производстве сплошных катаных
прутков круглого поперечного сечения из цветных и черных
металлов и сплавов на практике начали находить широкое
применение трѐхвалковые станы радиально-сдвиговой
прокатки. Основной особенностью процесса радиально-
сдвиговой прокатки является возможность управления схемой
напряженно-деформированного состояния металла в
достаточно широких пределах, что и обеспечивает получение
высококачественного круглого проката с необходимой
структурой и заданным уровнем механических свойств. В
настоящее время имеется множество научных работ, которые
посвящены исследованию влияния радиально-сдвиговой
прокатки на эволюцию микроструктуры и изменение
механических свойств не только различных черных и цветных
металлов, но и современных композиционных материалов. Вот
некоторые из этих научных работ [1-4]. Авторами этих работ
был доказано, что с помощью радиально-сдвиговой прокатки
можно получит высококачественные прутки разного
типоразмера по диаметру из различных материалах, которые
будут иметь градиентную ультрамелкозернистую структуру и
заданный уровень механических свойств.
Также давно уже доказано, что и термической обработкой
можно добиться дополнительного измельчения зерна в
различных материалах, если правильно подобрать режимы ее
проведения, что также положительно сказывается на свойствах
подвергаемых термической обработке металлоизделий. В том
числе в ряде научных работ было доказано, что объединение в
единый технологический процесс предварительной
термической обработки по различным режимам и разных
способов обработки металлов давлением позволяет добиться
лучших результатов в измельчении исходного зерна различных
материалов [5].
На основе анализа выше приведѐнных работ, а также и
целого ряда других работ в данном направлении исследований,
нами ранее уже были проведены исследования влияния
комбинированной технологии обработки латуни, включающей
предварительную термическую обработку, а именно отжиг при
температуре 500°С и радиально-сдвиговую прокатку при
аналогичной температуре и при температуре 700оС, на
эволюцию структуры данного материала. Результаты данных
исследований приведены в работе [6]. Но при этом нами не
было изучено влияние рассмотренной в данной работе
комбинированной технологии на изменение механических
свойств латуни. Поэтому целью данной работы является
именно исследование комбинированной технологии обработки
латуни марок Л63 и ЛЖМц59-1-1, включающей термическую
обработку и радиально-сдвиговую прокатку, на изменение
механических свойств данного материала.
2. Материалы и методика
Для достижения поставленной цели нами был проведен
физический эксперимент. В качестве исходных заготовок были
взяты латунные (латунь марок Л63 и ЛЖМц59-1-1) заготовки
диаметром 25 мм и длиной 150 мм, которые, на основе
предварительно полученных в работе [6] результатов,
подвергли предварительной термической обработке, а именно
отжигу при температуре 500°С. Предварительную
термообработку обработку латунных заготовок проводили в
камерной печи сопротивления КЭП 12/1400. Далее данные
заготовки деформировали на стане радиально-сдвиговой
прокатки РСП 10-30 (рисунок 1) при температуре 500°С.
Деформирование заготовок осуществляли до диаметра 16 мм и
12 мм с шагом абсолютного обжатия по диаметру 3,0 мм при
деформировании до диаметра 16 мм и 2,0 мм при
деформировании до диаметра 12 мм (т.е.
252219161412 мм) по стандартной схеме
деформирования.
Рис. 1. Стан радиально-сдвиговой прокатки РСП 10-30
Для определения механических характеристик полученных
прутков из латуни марок Л63 и ЛЖМц59-1-1 после проведения
INNOVATIONS 2025
55
предварительной термической обработки и многопереходного
деформирования на стане радиально-сдвиговой прокатки было
решено провести испытаний на разрыв.
Для проведения испытаний на разрыв из всех
продеформированных прутков (как до диаметра 16 мм, так и 12
мм) на высокоточном автоматизированном отрезном станке
GTQ-5000 были нарезаны образцы для испытаний в виде
полосок размерами h×b×l=0,3×3×30мм. Данные полоски
вырезались в продольном направлении прутка из
поверхностной, промежуточной областей и центральной зоны.
Также были подготовлены аналогичные образцы для
определения механических свойств латуни марок Л63 и
ЛЖМц59-1-1 после предварительно проведенной термической
обработки.
3. Результаты и обсуждение
Испытания подготовленных образцов проводили на
двухколонной цифровой машине для испытания на растяжение
усилием 1000 кН компании Qingdao Guangyue Rubber
Machinery Manufacturing Co., Ltd (Китай). При этом испытания
для каждого материала дублировались по три раза для
исключения ошибок. По полученным и статистически
обработанным данным испытаний на разрыв, были определены
среднестатистическое значение механических свойств и
построены соответствующие графики зависимости предела
прочности σв и относительное удлинение δ от вида обработки
(рисунки 2 и 3).
а)
б)
Рис. 2. Механические свойства латуни марки Л63: а - предел
прочности; б - относительное удлинение
а)
б)
Рис. 3. Механические свойства латуни марки ЛЖМц59-1-1: а - предел
прочности; б - относительное удлинение
Анализ результатов механических испытаний,
приведенный на графиках зависимости предела прочности σв и
относительное удлинение δ от вида обработки (рисунок 2 и 3)
показал, что в процессе радиально-сдвиговой прокатки
прочностные свойства предварительно подвергнутой
термической обработке по выбранному режиму латуни марок
Л63 и ЛЖМц59-1-1 растут, как в поверхностной, так и в
промежуточной и центральной областях продеформированного
прутка, а значение относительного удлинения,
характеризующего пластические свойства данных материалов,
наоборот падает. Причем чем больше обжатие, тем больше эти
изменения.
Также из графиков, приведенных на рисунках 2 и 3 видно,
что наибольшие значения предела прочности и наименьшие
значения относительного удлинения наблюдаться в
поверхностных слоях заготовки. И наоборот, наибольшие
значения относительного удлинения и наименьшие значения
предела прочности наблюдаться в центральных слоях
заготовки. Это полностью согласуется с ранее приведенными
результатами исследования эволюции микроструктуры латуни
марок Л63 и ЛЖМц59-1-1в ходе деформирования их на стане
радиально-сдвиговой прокатки, которые показали, что в
процессе РСП в данных материалах формируется градиентная
структура по сечению прутка [6]. Также из полученных
результатов исследования механических характеристик латуни
марок Л63 и ЛЖМц59-1-1 видно, что на первом этапе
деформирования, т.е. до диаметра 16 мм, значения
механических свойств (прочностных и пластических) в
промежуточной области более приближены к значениям этих
же механических свойств в центральной зоне деформируемого
прутка. При дальнейшем деформировании данных прутков
значения механических свойств в промежуточной области
становятся уже более приближены к значениям этих же
механических свойств в поверхностной области
деформируемого прутка. Что также полностью согласуется с
результатами исследования эволюции микроструктуры
различных материалов, которые приведены в том числе в
работах [1-4]. А именно с тем фактом, что при радиально-
сдвиговой прокатке с увеличением обжатия, количество
INNOVATIONS 2025
56
вытянутых зерен в промежуточной зоне уменьшается и при
этом они становятся разориентрованными. Дальнейшее
увеличение обжатия приводит к тому, что поверхностная зона
деформируемых прутков, как бы поглощает промежуточную
зону и в ней формируется равноосная мелкозернистая
структура, с сохранением небольшого количества
разориентрованных вытянутых зерен.
4. Заключение
Проведенные исследования механических свойств латуни
марок Л63 и ЛЖМц59-1-1 показали, что в процессе ее
деформирования на стане радиально-сдвиговой прокатки
усредненные по сечению прочностные свойства данных
материалов, предварительно подвергнутых термической
обработке по выбранным режимам, растут, а пластические
наоборот падают. Так для латуни марки Л63 усредненное по
сечению значение предела прочности после радиально-
сдвиговой прокатки при температуре 500°С до диаметра 16 мм
выросло на 33% по сравнению со значением данного
показателя после отжига при температуре 500°С, а после РСП
до диаметра 12 мм на 60%. Усредненное значение
относительного удлинения, характеризующего пластические
свойства, в прутке снизилось на 54% после РСП до диаметра 16
мм и на 86% после РСП до диаметра 12 мм. Для латуни марки
ЛЖМц59-1-1 усредненное по сечению значение предела
прочности после радиально-сдвиговой прокатки при
температуре 500°С до диаметра 12 мм выросло на 69% по
сравнению со значением данного показателя после отжига при
температуре 500°С, а усредненное значение относительного
удлинения соответственно снизилось на 77%. Снижение
пластической характеристики, а именно относительного
удлинения, для латуни марок Л63 и ЛЖМц59-1-1 в ходе
радиально-сдвиговой прокатки находиться в пределах нормы
для выше приведенных материалов, подвергнутых
интенсивной пластической деформации при реализации
различных способов обработки давлением.
Данное исследование финансировалось Комитетом науки
Министерства науки и высшего образования Республики
Казахстан (Грант № AP14869128).
5. Список литературы
1. Gamin Y.V., Galkin S.P., Romantsev B.A., Koshmin A.N.,
Goncharuk A.V., Kadach M.V. Influence of Radial-Shear Rolling
Conditions on the Metal Consumption Rate and Properties of D16
Aluminum Alloy Rods./ Metallurgist. 2021. Vol.65. pp. 650–659.
2. A. Arbuz, A. Kawalek, A. Panichkin, K. Ozhmegov, F.
Popov, N. Lutchenko. Using the radial shear rolling method for fast
and deep processing technology of a steel ingot cast structure./
Materials, 2023, Vol. 16, Iss. 24, 7547.
3. Gamin, Y., Akopyan, T., Galkin, S. et al. Effect of radial
shear rolling on grain refinement and mechanical properties of the
Al–Mg–Sc alloy./ Journal of Materials Research (2023) 38, 4542–
4558.
4. A.N. Petrova, D.Y. Rasposienko, V.V. Astafyev, A.O.
Yakovleva. Structure and strength of Al-Mn-Cu-Zr-Cr-Fe ALTEC
alloy after radial-shear rolling. Lett. Mater., 2023, 13(2) 177-182.
5. Lezhnev S., Volokitina I., Koinov T. Research of influence
equal channel angular pressing on the microstructure of copper./
Journal of Chemical Technology and Metallurgy, 2014, 49(6), pp.
621–630.
6. Найзабеков А.Б., Волокитина И.Е., Панин Е.А. Изучение
влияния совмещенной термомеханической обработки на
эволюцию микроструктуры и изменение микротвердости
латуни / Сборник научных трудов международной научно-
практической конференции «Перспективные
машиностроительные технологии», 2024. – С. 486 – 491.
INNOVATIONS 2025
57
НОВАЯ ТЕХНОЛОГИЯ РЕЦИКЛИНГА ПРУТКОВОГО ЛОМА ЧЕРНЫХ
МЕТАЛЛОВ, ПОЗВОЛЯЮЩАЯ ПОЛУЧАТЬ ИЗ НЕГО УПРОЧНЕННЫЙ
ВИНТОВОЙ АРМАТУРНЫЙ ПРОФИЛЬ
NEW TECHNOLOGY FOR FERROUS METALS BAR SCRAP RECYCLING TO OBTAIN A REINFORCED
SCREW REINFORCEMENT PROFILE
Sergey Lezhnev1, Evgeniy Panin2, Elena Shyraeva3, Maxim Bogachev1
1 Rudny Industrial University, 50 let Oktyabrya str. 38, Rudny, 111500, Kazakhstan
2 Karaganda Industrial University, 30 Republic Ave., Temirtau, 101400, Kazakhstan
3 Nosov Magnitogorsk State Technical University, 38 Lenin Street, Magnitogorsk, 455000, Russia
Abstract. This work is devoted to the development and research of a new technology for processing bar scrap from 40X steel to produce
high-quality finished metal products in the form of a screw reinforcement profile. The developed technology included two stages. At the first
stage, the round bar scrap is pre-formed on a radial-shear rolling mill to obtain conventional bars of the required diameter and create
initial conditions for the formation of an ultrafine grained gradient structure in the resulting screw profile. The second stage is the direct
production of a screw profile with a gradient ultrafine-grained structure on a combined installation. Metallographic studies have shown that
after the implementation of the second stage of obtaining a screw reinforcement profile, a gradient structure is observed in 40X steel, with
equiaxed grains, the size of which lies within 1.2-1.4 μm in the peripheral region and to a greater extent with elongated directional grains in
the central region of the resulting reinforcement.
Keywords: RADIAL SHEAR ROLLING, COMBINED INSTALLATION, BAR SCRAP, RECYCLING, SCREW REINFORCEMENT PROFILE,
MICROSTRUCTURE.
1. Введение
На данный момент в мире производиться почти 2
миллиарда тонн стали, которая используется для изготовления
различной металлопродукции, используемой в различных
отраслях промышленности. При этом необходимо признать,
что металлургия чаще всего является «грязной» отраслью, так
как в данном случае идет достаточно сильное загрязнение
окружающей среды. Поэтому перед металлургическими
компаниями уже не одно десятилетие стоит задача по
«озеленении» металлургии. И в настоящее время есть
достижения в данном направлении. В тоже временя, хочется
отметить, что и сама металлопродукция, отслужившая свой
срок службы, превращается в отходы, так называемый лом,
который также необходимо утилизировать. Одним из самых
простых и часто применяемых способов утилизации лома
черных металлов и сплавов, как и цветных, является его
переплавка и дальнейшее вторичное использование. Если
говорить о металлоломе черных металлов, то в некоторых
странах мира в практику уже давно вошел и другой способ его
переработки, а именно рециклинг некоторых металлоизделий,
отслуживших свой срок службы различными способами
горячей обработки давлением с получением готового товарного
продукта. На наш взгляд, данный способ переработки лома
черных металлов и сплавов можно отнести к «зеленным»
технологиям, так как он позволяет внести свой небольшой
вклад в улучшение экологической обстановки в мире.
На данный момент во многих странах мира разработан
целый ряд различных технологических процессов по
переработке некоторых металлоизделий, отслуживших свой
срок службы с помощью горячей обработки давлением в
готовый товарный продукт. Основоположником данного
направления переработки металлоизделий, отслуживших свой
срок службы горячей обработкой давлением, является
американским ученым E.E. Slick, который еще в начале
прошлого столетия разработал технологию переработки
железнодорожных рельсов горячей прокаткой в калибрах с
целью получения фланцевых профилей [1]. Но тогда
предложенная им технология не нашла широкого применения
на практике. И только в конце 20 века данное направление
заинтересовало мировое научное сообщество и производителей
металлопродукции. Вот только некоторые работы [2-5],
посвященные разработке и изучению технологий по
переработке железнодорожных рельсов горячей прокаткой с
получением различной готовой металлопродукции. Например,
одна из этих технологий, а именно технология перекатки
железнодорожных рельсов в строительную арматуру, нашла
применение на металлопрокатном заводе в городе Тула
(Россия).
Помимо перекатки железнодорожных рельсов в различную
готовую металлопродукцию находят развитие и технологии
рециклинга немерных отрезков труб с получением товарной
продукции в виде тонкой ленты. Так в работах [6-8]
рассмотрено два варианта переработки отрезков труб в ленту в
холодном состоянии: первый вариант - это сплющивание –
прокатка – разделение, а второй вариант это сплющивание –
разделение – прокатка. По обоим предложенным вариантам
была получена качественная тонкая стальная лента длиной 1,5
м.
Авторы работы [9] была предложена технология
использования техногенных отходов металлургического
производства немерной обрези медных труб, изготовленных
методом прессования, для получения композитной ленты
системы медь-сталь-медь. Данная технологи включает в себя
следующие этапы деформирования: осадка отрезка трубы
между бойками, внутрь которого помещена пластина из стали;
последующая продольная прокатка полученной заготовки.
Еще одним перспективным способом рециклинга
некоторых металлоизделий, отслуживших свой срок службы
является их перекатка на станах радиально-сдвиговой прокатки
[10]. Но чаще всего данный способ применим только для
рициклинга металлоизделий цилиндрической формы,
например, насосных штанг [11] и бывших в эксплуатации
железнодорожных осей [12], а также пруткового лома черных
металлов и сплавов [13], в том числе в виде арматуры [14]. В
основе данного метода во всех случаях лежит принцип горячей
радиально-сдвиговой прокатки бывшего в употреблении так
называемого прутка до меньшего диаметра и уже
последующего использования полученного прутка меньшего
поперечного сечения по назначению. Хочется отметить, что
радиально-сдвиговая прокатка позволяет получать
длинномерные изделия из различных материалов с
градиентной ультрамелкозернистой структурой [15] и, кроме
этого, данный способ наиболее технологичен и прост в
осуществлении по сравнению со многими другими способами
обработки металлов давлением, реализующих в процессе
деформирования интенсивные пластические деформации.
Как уже было отмечено выше радиально-сдвиговая
прокатка позволяет перерабатывать прутковый лом черных
металлов с получением товарного продукта в виде прутков
INNOVATIONS 2025
58
различного назначения. В последующем данные прутки можно
использовать, в том числе, и для получения различной
высококачественной металлопродукции, например, метизной
продукции или стяжного винта для опалубки, но это потребует
использование дополнительных технологических решений.
Развивая направление рециклинга пруткового лома с
помощью радиально-сдвиговой прокатки нами был предложен
новый технологический процесс для переработки пруткового
металлолома, совмещающий радиально-сдвиговую прокатку и
прессование через винтовую матрицу (рис.1). Данная
технология позволит получать из пруткового металлолома
круглого поперечного сечения высококачественный товарный
продукт в виде винтового арматурного профиля с градиентной
ультрамелкозернистой структурой и с заданным уровнем
механических свойств.
Рис. 1. Совмещенный способ получения арматурного профиля
Ранее в работах [16-17] нами с помощью компьютерного
моделирования в программном комплексе DEFORM уже была
доказана возможность реализации данного совмещенного
процесса и определены наиболее рациональные
геометрические и технологические параметры для его
реализации. Целью данной работы является подтверждение на
практике возможности получения винтового профиля с
градиентной ультрамелкозернистой структурой в ходе
реализации новой совмещенного способа, представленного на
рисунке 1.
2. Материалы и методика
Лабораторная совмещенная установка для получения
арматурного профиля была собрана на базе стана радиально-
сдвиговой прокатки РСП 10-30, на который была установлена
винтовая матрица (рис. 2, а) с максимальным диаметром
рабочего винтового канала 20 мм.
В качестве исходных заготовок для проведения
физического эксперимента был использован прутковый лом, а
именно шпильки (ГОСТ 9066-75, тип-А.1) размером М36х150
мм из стали 40Х, отслужившие свой срок службы. Перед
деформированием данные шпильки были подвергнуты
гомогенизирующему отжигу.
Предложенная технология рециклинга пруткового лома в
виде шпилек включает в себя два этапа. На первом этапе
предварительно подвергнутые гомогенизирующему отжигу
шпильки диаметром 36 мм нагревали до температуры
1150±10ºС и выдерживали при данной температуре. Далее эти
шпильки деформировали на стане радиально-сдвиговой
прокатки СВП 14-40 по реверсивной схеме [13] до диаметра 24
мм, то есть было осуществлено 4 прохода по схеме
3633302724 мм.
а)
б)
Рис. 2. Стан радиально-сдвиговой прокатки РСП 10-30 и собранная на
нем установка для получения арматурного винтового профиля: а –
винтовая матрица; б - рабочая клеть стана и установленная
винтовая матрица
Перед реализацией второго этапа полученные на первом
этапе прутки диаметром 24 мм были подогреты до
температуры 1000±10 ºС. Второй этап технологии рециклинга
шпилек был реализован на собранной на стане радиально-
сдвиговой прокатки РСП 10-30 совмещенной установке (рис. 2,
б). Процесс деформирования предварительно нагретых прутов
на собранной совмещенной установке заключался в
следующем: сначала пруток прокатывался в рабочей клети
стана РСП 10-30 с абсолютным обжатием по диаметру 5 мм, на
выходе из которой он входил в винтовую матрицу и за счет
заталкивающего усилия создаваемого валками, и
проталкивался через винтовой канал. В результате на
деформированных прутках был получен винтовой профиль.
3. Результаты и обсуждение
Металлографические исследования полученного винтового
арматурного профиля, как и полученных на стане радиально-
сдвиговой прокатки СВП 14-40 прутков, были осуществлены с
помощью сканирующего электронного микроскопа Quanta 200i
3D. Микроструктура исследовалась в центре и на периферии
сечения полученных прутков диаметром 24 мм и винтового
профиля. Результаты металлографических исследований
представлены на рис. 3.
а)
INNOVATIONS 2025
59
б)
в)
г)
Рис. 3. Микроструктура стали 40Х: а, б – периферия (а) и центр (б)
прутка после радиально-сдвиговой прокатки до диаметра 24 мм; в, г –
периферия (в) и центр (г) полученного винтового арматурного
профиля
Анализ эволюции микроструктуры показал, что исходная
структура шпилек, подвергнутых гомогенизирующему отжигу,
имеет типичную крупнозернистую феррито-перлитную
структуру со средним размером зерен лежащем в интервале 45-
60 мкм. Микроструктура периферийной области прутка
диаметра 24 мм, полученного радиально-сдвиговой прокаткой,
имеет преимущественно равноосный
субультрамелкозернистый характер со средним размером зерен
в пределах 1,6-1,8 мкм (рис. 3, а). В центральной зоне данных
прутков наблюдается структурная полосатость, т.е.
наблюдаются вытянутые в направлении прокатки зерна с
размерами, лежащими в интервале 6÷10,5х1,1÷1,7 мкм (рис. 3,
б). Т.е. в прутке из стали 40Х после радиально-сдвиговой
прокатки, как и предполагалось, была получена градиентная
структура.
Анализ микроструктуры стали 40Х после реализации
второго этапа деформирования, т.е. после получения винтового
профиля, показал, что за счет установки на стан радиально-
сдвиговой прокатки винтовой матрицы, мы не только получаем
винтовой арматурный профиль, но и дополнительно
модифицируем ранее полученную градиентную структуру
стали 40Х. При этом в центральной и периферийной области
полученного винтового арматурного профиля размер зерна еще
дополнительно уменьшился. В периферийной области размер
зерна уменьшился до среднего размера 1,2-1,4 мкм (рис. 3, в).
В центральной области полученного винтового профиля, в
большей степени также наблюдаются вытянутые в
направлении прокатки зерна, что обусловлено
предварительной деформацией исходных шпилек по
стандартной реверсивной схеме радиально-сдвиговой
прокатки, но за счет дополнительного скручивания прутка и
возникающего со стороны винтовой матрицы противодавления,
произошло частичное изменение ранее полученной на первом
этапе структуры. В частности в центральной зоне начало
происходить частичное перестроение имеющейся полосчатой
текстуры: большая часть зерен в данной области также
остались вытянутые, но появилась их разориентация по
направлениям, также в данной области наблюдаются и
равноосные зерна (рис. 3, г).
4. Заключение
Проведенные исследования доказали на практике ранее
полученные в ходе компьютерного моделирования результаты,
а именно то что предлагаемый совмещенный технологический
процесс получения винтового профиля реализуем. Данный
процесс может быть использован для рециклинга пруткового
лома черных металлов с получением готового товарного
продукта в виде винтового профиля и соответственно его
можно отнести к «зеленным» технологиям.
Данное исследование финансировалось Комитетом науки
Министерства науки и высшего образования Республики
Казахстан (Грант № AP14869135).
5. Литература
1 Patent USA 1086789, Method of rolling flanged shapes,
Edwin E. Slick, 1914.
2 Смирнов В.К., Шилов В.А., Михайленко А.М. Технология
переработки железнодорожных рельсов на сортовой прокат //
Сталь. - 1995. - №2. - С. 46-48.
3 Patent USA 4982591, Rail recycle process, B. Darrell
McGahhey, 1991.
4 Бахтинов Ю.Б. О целесообразности перекатки
изношенных рельсов в сортовые профили // Производство
проката. - 2000. -№7. - С. 2-4.
5 Бадюк С.И., Лещенко А.И. Получение сортовых
профилей проката из изношенных железнодорожных рельсов //
Обработка материалов давлением. - 2010. - №4. - С. 162-167.
6 Патент РФ 2786705, Способ переработки немерных
отрезков труб, Логинов Ю.Н., Шимов Г.В., 2022.
7 Шимов Г.В., Логинов Ю.Н., Бушуева Н.И. Переработка
немерных отрезков труб с получением холоднокатаной ленты.
Черные металлы, 2023, Т.1101, №9. - С. 29-33.
8 Бушуева Н.И., Шимов Г.В., Логинов Ю.Н. Переработка
обрези труб с получением холоднокатаной ленты. сборник
тезисов докладов Международной научной конференции
«Современные материалы, передовые производственные
технологии и оборудование для них», Санкт-Петербург, 2023. -
С. 87-88.
9 Бушуева Н.И. Разработка технологии получения
биметаллической ленты из немерных отрезков медных труб.
Сборник статей международной научно-практической
INNOVATIONS 2025
60
конференции «Перспективные машиностроительные
технологии», Санкт-Петербург, 2024. - С. 422-425.
10 Патент РФ 2293619, Способ винтовой прокатки, Галкин
С.П., 2007.
11 Галкин С.П., Романцев Б.А. Инновационная технология
рециклинга насосных штанг с применением технологии и
министанов радиально-сдвиговой прокатки в условиях ОАО
«Очерский машиностроительныи завод» // Инженерная
практика. - 2014. - №9. – С. 58-61.
12 Гревцева В.В., Галкин С.П. Экспериментальное
опробование технологии повторного использования
железнодорожных осей с применением радиально-сдвиговой
прокатки // 72-е Дни науки студентов НИТУ "МИСиС". - 2017.
– C. 27-30.
13 Lezhnev S., Naizabekov A., Panin E., Volokitina I., Kuis D.
Recycling of stainless steel bar scrap by radial-shear rolling to
obtain a gradient ultrafine-grained structure// METALURGIJA 60
(2021) 3-4, 339-342.
14 Lezhnev S.N., Naizabekov A.B., Volokitina I.E., Panin
E.A., Kuldeyev E.I. Radial-shear rolling as a new technological
solution for recycling bar scrap of ferrous metals// Complex Use of
Mineral Resources. №1 (316), 2021, 46-52.
15 S.P. Galkin, Y.V. Gamin, A.S. Aleshchenko, B.A.
Romantsev. Modern development of elements of theory, technology
and mini-mills of radial-shear rolling. Chernye Metally, 2021, Vol.
2021, Iss. 12, P. 51-58.
16 S. Lezhnev, A. Naizabekov, E. Panin, A. Tolkushkin, D.
Kuis, A. Kasperovich, R. Yordanova. Development and computer
simulation of the new combined process for producing a rebar
profile./ Modelling and Simulation in Engineering, Volume 2023,
Article ID 7348592.
17 S. Lezhnev, E. Panin, A. Tolkushkin, D. Kuis, A.
Kasperovich. Development and computer simulation of a new
technology for forming and strengthening screw fittings./ Journal of
Chemical Technology and Metallurgy, 58, 5, 2023, 955-960.
INNOVATIONS 2025
61
Innovative technical solution for emergency repair of a pressure tunnel water
supply pipeline during its air passage over a river bed
Valeriy Naidenov
Institute of Mechanics
Bulgarian Academy of Science, Sofia, Bulgaria
valna53@mail.bg
Abstract: The pressure water supply pipeline from the „Iskar Dam“, which has existed since 1981, is the main one for the water
supply of the city of Sofia with drinking water. The water pipeline, for the most part executed as a concrete vault tunnel (mantel), passes
through complex mountainous terrain with a large difference in elevation. During the long-term active operation, the steel part of the air
passage of the water pipeline over the „Porkolitsa River” suffered a serious accident - internal pressure implosion of the steel pipe. The
latter prevented the normal water supply to the city and an urgent repair was required by replacing the air-passing steel pipe of the water
pipeline. The report discusses a proposed and successfully implemented innovative technical solution for monolithic installation of a steel
pipeline in existing tunnel excavations with concrete lining by using self-compacting concrete with shrinkage compensated admixture and
with the participation of special an internal crystallization chemical admixture, tailored to the specific requirements of the site. A prescribed
composition of self-compacting concrete has been developed, the necessary control has been carried out during the batching plant
production of the fresh concrete with monitoring of technical indicators. A technological regulation has been developed for the transport,
laying and care of the concrete. During the site laying, standard sampling and testing have been carried out to prove the compressive
strength at different ages of the concrete. Based on the above, in conclusion, the effectiveness of the developed and implemented innovative
technical solution is certified.
Keywords: PRESSURE WATER SUPPLY, SELF-COMPACTING CONCRETE, SHRINKAGE-COMPENSATING CHEMICAL
ADMIXTURE, INTERNAL CRYSTALLIZATION CHEMICAL ADMIXTURE, COMPRESSIVE STRENGTH OF HARDENED
CONCRETE
1. Introduction
The pressure water supply pipeline from the Iskar Dam, which
has been in existence since 1981, is the main one for the water
supply of the city of Sofia with drinking water. The water pipeline
passes through complex mountainous terrain with a large elevation
difference. Most of it is built using a tunnel method with a concrete
lining (mantle) with a large diameter of about 3 m. In some of its
sections, there is an air passage through ravines and mountain
rivers, and in these areas the same passes into steel pipes with a
similar, but smaller diameter, anchored to the underground concrete
sections built in the slopes of the route. Ensuring the watertightness
between the concrete arch lining and the steel pipe is carried out by
embedding it in the depth of the tunnel part of the water pipeline
and performing a concrete monolithic. In the process of many years
of active operation, the steel part of the air passage over the
Porkolitsa River suffered a serious accident - internal pressure
implosion of the steel pipe. The latter prevents the normal water
supply to the city and urgent repairs are required according to a
previously developed project.
This project does not include a specific technical solution for
the monolithic construction of the newly constructed steel pipe
route in the area of the existing concrete tunnel excavations
(mantle).
In this regard, the technical solution presented below is for the
implementation of a monolithic completion of the concrete lining in
the sections of the two opposite tunnels around the new steel pipe
parts of the overhead part of the pipeline with a total length of about
38 m.
2. Technical solution
From the inspection of the construction site, it is evident that the
length of the necessary additional monolithic construction after the
construction of the new steel pipe distribution in the area of "tunnel
1" is of the order of 3 m, respectively about 2.5 m in the area of
"tunnel 2".
In this case, the technical proposal contains reasons for the
construction of the monolithic construction in question by using a
prescribed mix design of self-compacting concrete (SCC - self
compacting concrete), through which the desired monolithic
construction can be performed in an original innovative way using
simplified technology without the use of sealing agents (given the
inaccessibility of the monolithic construction area) at an optimal
price.
2.1. Self-compacting grouting concrete
Using the method of dense volumes and successive
approximations, a prescribed mix design of innovative self-
compacting high-strength concrete with fully compensated
shrinkage*) and with the participation of an internal-
crystallizing chemical admixture**) for achieving full
waterproofness of the constructed concrete section to the tunnel
contour, has been designed, in accordance with the requirements of
BDS EN 206:2013+A2:2021 Concrete. Specification, properties,
production and conformity (item 3.1.1.10) and BDS EN
206:2013+A2:2021 Concrete, updated in 2021. Specification,
properties, production and compliance/National Application
NA:2021, with compressive strength class C 35/45 MPa.
KEPTONITE*) is an innovative product capable of
counteracting the natural shrinkage processes of cement composites,
by reducing their micro-porosity and causing a sealing effect. The
specific action of KEPTONITE compensates for volumetric
shrinkage, and in special cases can completely overcome it to a
degree of controlled self-tensioning. In this way, when designing
the composition of the concrete, a certain controlled level of
expansion can be achieved, through which the physics-mechanical
properties of the concrete can be increased, as well as its corrosion
resistance. After homogenization in the composition of the concrete,
KEPTONITE directly affects the ongoing hydration processes,
causing volumetric expansion, which counteracts the shrinkage
processes in its various phases. In this way, the intermolecular
structural parameters are increased, which in turn increases the
physis-mechanical and structural characteristics of the hardened
concrete, while also increasing its corrosion resistance and
durability [1,2].
KRYSTALIN Add1**) is an innovative product of the latest
generation - a crystallizing waterproofing admixture for cement
concrete and mortars with permanent action, designed to waterproof
and increase the durability of concrete by applying a new internal
crystallization technology based on hydrophilic development of
additional hydration processes in the concrete structure. It has the
ability to reduce the water-cement ratio and self-fill cracks up to
0,5-0,7 mm wide in the concrete cross-section (self-healing ability)
[3].
INNOVATIONS 2025
62
The designed composition of the concrete also complies with
the requirements for resistance to corrosive environmental and
production factors.
Given the location of the site (mountainous location and
remoteness from a concrete manufacturer), all specific
environmental and operational environment impact factors for
external application should be taken into account during the design,
complying with the requirements of BDS EN 206:2013+A2:2021
Concrete. Specification, properties, production conformity and BDS
EN 206:2013+A2:2021/NA:2021 Concrete. Specification,
properties, production conformity. National Annex (NA).
In this case, in accordance with the classification in BDS EN
206:2013 + A2:2021 (Table 1 - Impact classes, p. 24), the possible
aggressive factors of the environmental and operational
environment are:
- Corrosion caused by carbonation XC4 - cyclic wetting and
drying of concrete surfaces in contact with water, which do not
belong to impact class XC2;
- Impact of cyclic "freeze/thaw" without de-icing agents XF3 -
strong water saturation without de-icing agent, concrete surfaces
exposed to rain and freezing.
The limit values of the concrete composition for each of the
specified aggressive environments are determined in accordance
with the requirements of BDS EN 206:2013 + A2:2021/NA:2021
(Table NA.F.1a - Limit values for composition and properties of
concrete).
The dominant requirements for the composition of the concrete
are: - Compressive strength class C30/37;
- Maximum water-cement ratio 0,50;
- Minimum cement content (without CRM) 320 kg/m3;
- Frost resistance Cfr150 (150 cycles);
- Mineral aggregates with frost resistance F1, MS18;
- Air content 5%.
In addition to the above, there are additional specific
requirements arising from the specifics of the site. They are
determined by the location of the site, long transport conditions in
mountainous conditions, climatic conditions during execution,
mandatory laying with a truck-mounted concrete pump, if it is
impossible to use conventional dencification equipment (vibrators
of different types), as well as the need to ensure an absolutely
impermeable connection between the existing concrete lining, resp.
newly constructed steel pipe, and newly laid self-compacting
concrete sealing the contour.
The above also requires:
- alternative inclusion in the composition of the concrete, in
addition to highly range water-reducing chemical admixtures, and
such according to atmospheric conditions - retardants at high
temperatures;
- compensated free deformations of the concrete from
shrinkage, which is achieved by using a special anti-shrinkage agent
Keptonite*) with an optimal consumption rate and ability to achieve
a certain degree of self-tensioning for sealing the concrete section;
- to ensure the required degree of frost resistance Cfr150, in
accordance with the requirements of the standards, instead of
ensuring air entrainment in the concrete mix (5% with a maximum
size of the additives Dmax 25 mm), it is permissible to work with
other means to ensure such a degree of frost resistance.
In this case, given the transport distance, I decide to use an
innovative internal crystallization chemical admixture Krystaline
Add1**) instead of air entrainment in the composition of the
concrete mix, provided with air-entraining chemical admixtures,
which has a proven highly positive effect in increasing the density
of the section and its frost resistance:
- increasing the compressive strength class of the concrete to
C35/45, as well as ensuring a consistency of the concrete mix
typical of self-compacting concrete, in accordance with the
requirements of BDS EN 206:2013 + A2:2021, Annex G
(Guidelines for self-compacting concrete requirements in the fresh-
state – G1 to G2) and SF2 (slump-flow test in accordance with EN
12350-8).
In addition to the above, there are additional specific
requirements arising from the specifics of the site. They are
determined by the location of the site, long transport conditions in
mountainous conditions, climatic conditions during execution,
mandatory laying with a truck-mounted concrete pump, if it is
impossible to use conventional sealing agents (vibrators of different
types), as well as the need to ensure an absolutely impermeable
connection between the existing concrete lining, resp. newly
constructed steel pipe, and newly laid self-compacting concrete
sealing the contour.
In accordance with all of the above, and based on the materials
used by HYDROBETON OOD as the nominated manufacturer and
supplier of concrete, I specify the following concrete composition
(Table 1):
Table 1: Prescribed concrete mix design
Concrete component
Quantity ***),
kg/m3
Portland cement cement CEM II A-LL 42,5R
Cement plant "Devnya"
420
Fly ash TPP "Sliven"
40
River sand, fr. 0-4 mm, "Chelopechene" site
400
Crushed sand, fr. 0-4 mm, "Hydromineral" site,
village of Studena
444
Crushed stone, fr. 4-11,2 mm, "Hydromineral" plant,
village of Studena
407
Crushed stone, fr. 11,.2-22,4 mm, "Hydromineral" plant,
village of Studena
442
Polycarboxylate high range water-reducing chemical
admixture PC 130 Don Construction Products
4,20
Setting and hardening retarder R50 Don Construction
Products
1,05
Internal crystallization chemical additive Krystaline
Add1
1,00
Shrinkage compensator Keptonite
15,00
Mixing water
190
Water-cement ratio
0,45
Consistency (slump-flow test), mm
SF2 –
660-750
***) for dry components
2.2. Ready-mix fresh concrete production
The production of the concrete mix was organized and carried
out at the concrete batching plant HYDROBETON OOD,
Gorublyane base, adhering to the developed recipe (see Table 1),
and the special additives Krystaline Add1 and Keptonite (Photos 1
and 2) were added at the specified consumption rates during the
homogenization of the concrete mix (Photos 3 and 4).
Photo 1
Photo 2
INNOVATIONS 2025
63
Photo 3
Photo 4
2.3. Specialized sampling and testing for concrete quality
During laying, specialized standard sampling, laboratory
standard aging and testing were organized and conducted with
analysis of the results obtained and issuance of a report by an
accredited testing laboratory for the characteristics of the self-
compacting monolithic concrete.
Sampling and testing of the test specimens (cubes 15x15x15
cm) was carried out by the UNIVERSITY CONSTRUCTION
TESTING LABORATORY (USIL) of the UNIVERSITY OF
ARCHITECTURE, CIVIL ENGINEERING AND GEODESY,
Sofia, with an Accreditation Certificate in accordance with BDS EN
ISO/IEC 17025:2018 of the Bulgarian Accreditation Service - reg.
No. 239 LI of 10.01.2023. with validity until 10.01.2027 (Annex 3),
as well as the relevant Order No. A10/10.01.2023 on the scope of
accreditation of the Bulgarian Accreditation Service (Annex 4),
which includes all the necessary competencies on the topic.
Standard sampling is illustrated in Photos 5-10.
Photo 5
Photo 6
Photo 7
Photo 8
Photo 9
Photo 10
The results obtained for the compressive strength of concrete at
7 and 28 days of age are and are presented in Table 2.
Table 2 Compressive strength of concrete
Compressive strength, MPa
Value and tolerance of the
indicator
At
7 days of age
41,70
36,90
42,40
Mean: 40,20
No requirements
At
28 days of age
58,60
56,60
52,50
Mean: 55,90
55,90 for class 35/45, acc. to BDS
EN 206:2013 + A2:2021, Appl.
NA.B. with number of results 2 to
4, each single result must be ≥ 0.95
fck (42.5 MPa), resp. average result
≥ 0.95 fck +1 (46 MPa)
Check completed!
2.4. Transportation, laying, handling and care after
concreting
Before the concrete is delivered, the opening of each of the two
tunnels is closed with a tight formwork, and technological measures
have been taken to prevent the cement milk from leaking out of the
concrete mixture by installing a sealing strip along the periphery of
the contour, plus additional installation foam at individual positions.
The laying is carried out by inserting the flexible hose of the
positioned pump as far inward as possible through the technological
opening left in the formwork. The concrete mixture is discharged
and compacted by gravity, without the use of sealing agents. The
consistency of the concrete mixture allows such an action while
ensuring the quality of the concrete (Photos 11-19). For each of the
two tunnels, the concrete is laid in two stages – initially up to about
half of the volume of the opening, after which, the next day, the
remaining amount necessary to fill the entire contour is delivered
and laid. In this way, additional security is achieved regarding the
exact positioning of the steel pipe axially and at the design level.
Concrete care after placement consists solely of maintaining the
formwork for a minimum of 2 days, after which the de-moulded
concrete surface is periodically watered with water for up to 1
week.
Photo 11 Tunnel 1 –
preparing for formwork
Photo 12 Tunnel 2 –
preparing for formwork
INNOVATIONS 2025
64
Photo 13
Formwork completed
Photo 14
Formwork completed
Photo 15 Concreting
Photo 16 Concreting
Photo 17
Tunnel 1 contour completed
Photo 18
Tunnel 2 contour completed
Photo 19 Site completed
3. Conclusion
In conclusion of the above, it can be confirmed that the
developed innovative technical solution for grouting of steel
pipeline in existing tunnel works with concrete lining by using self-
compacting shrinkage-compensated concrete was implemented
qualitatively by main-contractor Promenergomontazhe AD fully
corresponding to the specific requirements of the site.
3. References
1. V. Naidenov, Innovative self-compacting fiber-reinforced
concrete with compensated shrinkage for machinery steel anchors
grouting, VI International Scientific Conference “INDUSTRY 4.0”,
Varna, June 2021, Year V, Volume 1/11, ISSN (PRINT) – 2535-
0153, ISSN (on line) -2535-0161, p. (12-15.)
2. V. Naidenov, I. Rostovski, M. Mironova, Hybrid reinforced
concrete with controlled volume deformations for hydrotechnical
facilities, VII International Scientific Congress “INNOVATION
2022”, Varna, 20-23. June, 2022, p. (41-45) ISSN 2603-3771
(Online) ISSN 2603-3763 (Print)
3. V. Naidenov, Extended research on the efficiency of internal
crystallization chemical admixtures for cement concrete -
mechanical and structural characteristics, Proc. of the IX
International Scientific Congress “INNOVA’ 2023”, ISSN 2603-
3771 (on line), ISSBN 2603-3763 (print), p. (40-45).
INNOVATIONS 2025
65
Stress intensity factor in rods under tension with twin semielliptical cracks
Pejo Konjatić*1, Ana Konjatić2, Mato Jakus2, Željko Blavicki2
University of Slavonski Brod, Croatia1
Industrial and Trade School Slavonski Brod, Croatia2
*pkonjatic@unisb.hr
Abstract: This study investigates the influence of distance and relative orientation of semi-elliptical twin surface cracks in cylindrical bars
under tensile loading on the stress intensity factors (SIF). Finite element analysis was used to analyze the interactions between two identical
cracks with different spacing (5-20 mm) and rotation angles (0°-180°). The results show that the crack orientation has a significant effect on
the stress distribution patterns and stress intensity factor values, with a 45° rotation producing the highest SIF values, while a parallel
orientation has a favorable shielding effect with significantly lower values. All configurations gradually converge to the reference value for a
single crack at larger distances (20 mm), indicating decreasing interaction effects. These results can provide insights for the assessment of
structural integrity, especially for components with multiple closely spaced defects, and provide a basis for the development of more
accurate predictive models for fracture behavior in engineering applications.
Keywords: ROD, TWIN CRACK, STRESS INTENSITY FACTOR
1. Introduction
Structural elements often contain defects that occur during
manufacturing or exploitation. Among them, cracks represent
critical stress concentration points that under appropriate loading
can lead to fracture and consequent damage or complete destruction
of the structure. Therefore, understanding the behavior of cracks,
and especially the interaction between multiple cracks, is of key
importance for assessing the integrity and reliability of structures
[1-3].
Fracture analysis based on linear-elastic fracture mechanics
(LEFM) is a widely accepted method for predicting the behavior of
materials with cracks. The basic concept of LEFM is based on the
evaluation of the linear-elastic fracture mechanics parameter the
stress intensity factor (SIF) K, which quantifies the stress at the
crack tip and serves as a criterion for predicting crack initiation and
propagation. Although much research focuses on individual cracks,
in practice, multiple cracks that interact with each other are often
encountered, changing the stress distribution and stress intensity
factors. Crack interaction can significantly affect the load-bearing
capacity and service life of the structure, so it is necessary to
consider these interactions when assessing safety [4-9].
This paper focuses on the analysis of twin cracks in cylindrical
bar under tensile loading. The aim is to investigate the influence of
the distance between cracks and the influence of angle of rotation
between twin cracks on the stress intensity factor. The results of this
research can contribute to a better understanding of the behavior of
multiple cracks and improve methods for assessing the integrity of
structures. Research related to single cracks in rods and cylinders is
numerous and well documented and stress intensity factor solutions
are available for various crack geometries and loading conditions,
often obtained by analytical, numerical, or experimental methods
[10-14].
However, the analysis of multiple cracks presents a more complex
problem due to the interaction between cracks where interaction
depends on size and shape of cracks, their positions, distance and
type of load. Some researchers have dealt with the analysis of the
interaction of twin surface cracks in the context of structural
integrity assessment. The interaction of surface cracks exposed to
different stress distributions is investigated in [10] and interaction
of twin surface cracks on a cylindrical rod under combined tensile
and torsional loading is analyzed in [15].
There is also research that has dealt with the impact of crack
interaction on crack growth due to fatigue where fatigue growth
analysis of interacting and coalescing surface defects is investigated
in [16].
Despite significant progress in understanding crack interaction,
there is still a need for further research, especially regarding the
influence of geometric parameters such as distance and angle of
rotation between cracks.
2. Finite element modelling
In this paper, the influence of twin cracks on the stress intensity
factor in a tensile loaded cylindrical bar will be investigated using
finite element analysis (FEM). Numerical simulations were
performed in Ansys Workbench software [17]. A three-dimensional
model of a cylindrical bar with two identical semi-elliptical surface
cracks was developed. The cracks are embedded in parallel planes
perpendicular to the longitudinal axis with crack depth a = 12,5
mm, and crack aspect ratio a/c = 2,5. The influence of the distance
between cracks d and angle of rotation θ between cracks on stress
intensity factor was analyzed. The distance between cracks varied in
the range from 5 to 20 mm and the of rotation θ between cracks
varied in the range from 0° to 180° in steps by 45° (Fig 1).
Fig. 1 Geometry of rod with twin cracks
To compare the results alongside the model with a single crack,
an additional geometry was modelled in which two identical cracks
lie in the same plane with a rotation angle of 180° between them
(Fig 2).
Fig. 2 Rod with single and double crack
Crack depth a, crack aspect ratio a/c and diameter of rod D = 25
mm were kept constant. A linear-elastic material with Young's
modulus of 200 GPa and Poisson's ratio of 0,3 were used for
modeling. Tensile load of 10 kN was applied to the rod. The
INNOVATIONS 2025
66
boundary conditions are defined in a way that they ensure a realistic
simulation of the load.
A high-quality finite element mesh is generated, with special
attention to the area around the crack fronts, where a high stress
concentration is expected. A fine mesh is used around the crack tip
to ensure the accuracy of the results. The size of element is
determined from convergence of results ensuring accurate results
for stress intensity factor solutions (Fig. 3).
Fig. 3 Finite element mesh
A static linear-elastic analysis is performed to determine the
stress distribution and values of stress intensity factor for different
combinations of parameters. Special attention was done to
determining the maximum SIF value (KI) along the crack tip since
dominant opening mode for crack opening is mode I (Fig. 4).
Fig. 4 Detail of finite element mesh around crack tip
3. Results and analysis
As a results of finite element analysis values of stress intensity
factors along crack fronts are obtained and maximum values of
stress intensity factor for first mode of crack opening are shown in
Table 1.
Table 1: Stress intensity factor results obtained by numerical analysis
d, mm
θ, °
KI, MPam1/2
5
0
69,200
5
45
96,213
5
90
94,310
5
135
81,218
5
180
77,752
10
0
76,251
10
45
88,802
10
90
90,560
10
135
83,574
10
180
80,388
15
0
81,312
15
45
86,173
15
90
87,926
15
135
84,760
15
180
82,756
20
0
83,853
20
45
85,482
20
90
86,421
20
135
85,188
20
180
84,218
Fig. 5 shows the distribution of the stress intensity factor along
the crack front for three configurations: a single crack, twin cracks
with an angle of 0°, and twin cracks with an angle of 45°. The
results show distinct variations in SIF depending on the crack
configuration and angular orientation.
Fig. 5 Stress intensity factor distribution along crack front for single
and twin cracks with angle of rotation between cracks of 0° and 45°
The single crack exhibits a relatively uniform SIF distribution,
peaking at 85,3 MPam1/2 near the center of the crack front. This
behavior aligns with theoretical expectations for isolated cracks
under tensile loading, where stress concentration is highest at the
midpoint due to symmetry and load distribution.
For twin cracks aligned at 0°, the SIF distribution is notably
lower compared to the single crack, with a maximum value of 69,2
MPam1/2. This reduction can be attributed to the interaction between
the two cracks, which redistributes stresses along their fronts and
reduces peak intensity. The parallel alignment minimizes stress
amplification, as both cracks experience similar loading conditions
without significant interference effects.
The twin cracks oriented at 45° exhibit a markedly higher SIF
distribution, peaking at 96,2 MPam1/2. This configuration introduces
a synergistic effect between the cracks due to their angular
misalignment, which amplifies stress concentration at specific
points along their fronts. The increased SIF values suggest that
angular orientation plays a critical role in determining crack
interaction and overall structural integrity.
The Fig. 6 illustrates the variation of the maximum values of
stress intensity factor as a function of the distance between two
identical semi-elliptical cracks in parallel planes under tensile
loading.
Fig. 6 Maximum values of stress intensity factor dependent of distance
between twin cracks
INNOVATIONS 2025
67
The graph includes results for different rotation angles (0°, 45°,
90°, 135° and 180°) of the second crack relative to the first, as well
as reference lines for a single isolated crack ("single") and two
cracks in the same plane rotated by 180° ("double").
The graph shows that the distance between cracks significantly
affects value of SIF. For all rotation angles between cracks, SIF
decreases as the distance increases, converging toward the reference
value for a single crack at larger distances at approximately 20 mm.
This trend indicates that crack interaction diminishes with
increasing crack separation, as the stress fields around each crack
become less coupled.
For angles of 45° and 90° SIF values steadily decrease,
approaching the reference value for a single crack (~85 MPam1/2).
This reduction suggests a gradual weakening of interaction between
cracks.
For angles of 0°, 135° and 180° SIF values gradually increase.
At large distances (20 mm) all configurations converge toward
approximately the same value, as interaction effects become
negligible.
The "single" line remains constant at ~85 MPam1/2, regardless
of distance, while the "double" line maintains a constant value of
~77 MPam1/2, indicating a specific effect of crack arrangement in
the same plane.
The reference "single" line represents a scenario with a single
isolated crack where no interaction occurs. The SIF value of ~85
MPam1/2 serves as an asymptotic limit toward which all
configurations converge at larger distances. The "double" line,
representing two cracks in the same plane rotated by 180°, shows a
constant SIF value of ~77 MPam1/2, reflecting a specific geometric
distribution of stress that is more favorable than other two-crack
configurations.
Fig. 7 illustrates the variation of the maximum values of stress
intensity factor as a function of the angle of rotation between two
identical semi-elliptical cracks in parallel planes under tensile
loading.
Fig. 7 Maximum values of stress intensity factor dependent of angle of
rotation between cracks
When the cracks are orientated without rotation (θ = 0°), the
SIF value is consistently lower compared to other configurations.
This indicates that a parallel orientation minimizes stress
concentration due to an even redistribution of stresses along both
crack fronts.
The configuration with a 45° rotation exhibits the highest SIF
values at smaller spacings, reaching a peak value of 96,2 MPam1/2.
This result emphasizes the significant interaction effects caused by
the angular misalignment, where the stresses are amplified due to
the geometric interference between the cracks.
The intermediate rotation angles of 90° and 135° show
moderate SIF values that gradually decrease with distance. In these
configurations, the stress distribution and interaction effects balance
each other out, resulting in less pronounced peaks compared to the
45° case.
When the cracks are rotated by 180° and lie in the same plane,
the SIF value remains relatively low across all distances and
approaches the reference value for a single crack at larger distances.
This configuration minimizes the interaction effects due to the
symmetrical alignment of the stress fields.
4. Conclusions
The results clearly demonstrate that interaction between semi-
elliptical cracks significantly influences the stress intensity factor at
smaller distances. The relative orientation defined by rotation angle
plays a crucial role in determining stress concentration levels:
The results highlight that twin cracks can significantly alter
stress intensity distributions depending on their relative orientation
with parallel cracks (0°) reducing stress concentrations through
redistribution while angled cracks (45°) amplify them due to
geometric interference effects.
Rotation angles of 45° and 90° represent the most critical
configurations due to high KI values.
Rotation angles of 0°, 135° and 180° show more favorable
stress distributions, with 0° exhibiting the most pronounced
shielding effect.
At larger distances, interaction effects become negligible, and
all configurations tend toward the reference value for a single
isolated crack.
At smaller distances, it is essential to consider not only the
presence of multiple cracks but also their relative orientation. At
larger distances, interaction effects weaken, allowing for simplified
assessments using single-crack models.
These results provide a basis for optimizing structural geometry
to minimize stress concentration and increase the material's
resistance to crack propagation.
This research provides a more detailed insight into the
interaction of twin cracks in tensile-loaded rods, which will
contribute to improving methods for assessing the integrity of
structures and more reliably predicting their service life.
In addition, future analyses that take into account variations in
crack dimensions (depth to length ratio, aspect ratio) and additional
orientation parameters (including non-planar configurations and
mixed loading conditions) would provide even more comprehensive
insights. Investigating how the asymmetry of crack size affects the
interaction mechanisms could shed light on whether smaller cracks
behave as stress reducers or concentrators in the presence of larger
defects.
Furthermore, investigating the influence of material anisotropy
on crack interaction would extend the applicability of these findings
to composites and directionally reinforced materials commonly
used in critical load-bearing applications. Such extended parametric
studies would provide a more robust basis for the development of
predictive models that accurately capture the complex interplay
between multiple defects in engineering components.
3. References
1. M.H. Aliabadi, D.P. Rooke, Numerical Fracture Mechanics
(Solid Mechanics and Its Applications (8)) (Springer, 1991)
2. M. Kuna, Finite Elements in Fracture Mechanics: Theory –
Numerics - Applications (Solid Mechanics and Its Applications
(201)) (Springer, 2013)
3. P. Konjatić, D. Kozak, N. Gubeljak, Key Eng. Mat., 488-489,
367 (2012)
4. BSi, BS 7910:2013+A1 - Guide to methods for assessing the
acceptability of flaws in metallic structures (BSi, 2013)
5. EDF Energy, R6: assessment of the integrity of structures
containing defects, revision 4, amendment 11 (EDF Energy,
2015)
INNOVATIONS 2025
68
6. K. Hasegawa, K. Miyazaki, u 10th international conference on
the mechanical behavior of materials, Busan, 2007
7. M.A. Barrinayaa, M.N. Alfiyurandaa et al., Engineering Solid
Mechanics, 10, 399 (2022)
8. H.E. Coules, Int. J. Press. Vessels. Pip., 162, 98 (2018)
9. Le Van, J. Royer, Int. J. Fract., 61, 71 (1993)
10. M.M. Hamdan, M.K. Awang, A.E. Ismail, Research Progress in
Mechanical and Manufacturing Engineering, 1, 69 (2020)
11. J. Predan, V. Močilnik, N. Gubeljak, Eng. Fract. Mech., 105,
152 (2013)
12. M. Fonte, M. Freitas, Int. J. Fatigue, 21, 457 (1999)
13. Esm, Engineering Solid Mechanics, 10, 399 (2022)
14. M. Aursand, B.H. Skallerud, Theor. Appl. Fract. Mech., 112,
102904 (2021)
15. H.E. Coules, Int. J. Press. Vessels. Pip., 157, 20 (2017)
16. X.B. Lin, R.A. Smith, Int. J. Fatigue, 19, 461 (1997)
17. ANSYS, Inc., Ansys Workbench, Release 2024 (ANSYS, 2024)
INNOVATIONS 2025
69
Nanosized BaTiO3 powder prepared via mechanochemical activation
Mihaela Aleksandrova, Bojidar Jivov, Vladimir Petkov
Bulgarian Academy of Science, Institute of Metal Science Equpment and Technology with Hidroaerodinamic Centre “ Acad, Angel
Balevski”, Sofia, Bulgaria
mihaela.krasimirova@mail.bg
Abstract: The technological possibilities for the synthesis of barium-titanate phases were investigated by applying mechanochemical
activation (from 30 min. to 2 h) of the starting charges and thermal treatment of the compositions up to 900°C (with an isothermal delay of 1
h). The applied laboratory regime for the preparation of the experimental samples is in accordance with the preliminary thermodynamic
calculations. The identification of the obtained phases was carried out by X-ray phase analysis (XRD). Based on the experimental results, the
necessary technological conditions for the synthesis of titanate monophasic product by heat treatment at lower temperature values than
those typical for standard classical synthesis have been established. The potential possibilities of application of the synthesized phases for
the deposition of thin layers on metal surfaces and the preparation of various coatings with different functional purposes are considered.
Keywords: MECHANOCHEMICHAL ACTIVATION, BARIUM TITANATE PHASES, LOW TEMPERATURE SYNTHESIS
1. Introduction
The development and application of various synthesis
methods allows obtaining materials with diverse properties,
structure and functional characteristics [1-5]. A number of author
collectives investigate the processes of phase formation in
experimental polycomponent compositions subjected to
mechanochemical treatment [6-17] under different technological
conditions. In systems subjected to mechanochemical activation, a
significant decrease in temperature [10-15] and isothermal holding
time necessary for the synthesis of the investigated compounds
(compared to classical methods) was found. It has been established
that the amorphization [14] of the reagents under the influence of
mechanochemical processing favors the formation of nanosized
crystallite products. By applying direct mechanochemical synthesis
in laboratory conditions, crystalline phases with different structure
and properties were obtained [6-16].
In mechanochemical activation of the reagents, the main
technological method is the application of intense friction and
impact [18-20]. To achieveto this effect, the batches prepared are
subjected to processing in high-energy mill facilities (planetary
mills equipped with suitable grinding bodies,vibrating millsand
other). The nature of the ongoing processes and the structure of the
obtained products are determined by a number of factors [11-19]:
specificity of the individual components, presence of additional
technological additives, initial composition of the batches, initial
humidity of the batches, mass of the batches,ratio between the mass
of the grinding grinding bodies and the mass of the charges,duration
of applied mechanochemical treatment (from minutes to
days),speed ofthe mechanochemical processing of the reagents,
specificity of the gas environment during processing, technical
characteristics of the used mill, type of grinding bodies (agat,
corundum, porcelain and others), mass, volume and number of
grinding bodies and others.
In the specialized literature experimentally established
reaction stages characteristic of the systems subjected to
mechanochemical treatment are presented and some theoretical
ideas and hypotheses about the mechanism of the processes are
presented.For reagents subjected to mechanical processing particle
sizes decrease, specific surface area increases, structural defects are
generated, active centers are formed, structural and thermodynamic
instability is initiated, and elevated reactivity. It is assumed that in
the process of intense tribomechanical impact separate localized
zones with elevated temperature values are distinguished [13-20],
which provokes the thermal decomposition of thermally unstable
components, accompanied by the release of other reaction products
that actively participate in the processes. In accordance with the
specificity of the reagents and the mode of tribochemical treatment,
conditions arise for various phase and structural transformations in
the reaction system: amorphization, change in the temperature
intervals of conversion and polymorphic transitions, lowering the
temperature of chemical reactions, complex of chemical reactions,
identification of new reaction stages uncharacteristic of identical
compositions not subjected to tribochemical processing. With some
reaction compositions, even with relatively short periods (10 min)
of tribochemical treatment, conditions are created for the formation
of new phases [14].
Due to the technological advantages of low-temperature
syntheses, carried out after mechanochemical activation, and
especially of direct mechanochemical syntheses (without heat
treatment), the perspective of applying these methods in obtaining
materials of significant applied importance acquires significant
importance [16-18]. In this aspect, barium titanate BaTiO3 is of
interest, which finds wide and permanent application in electronics
and other technical fields. As an alternative technological approach,
some authors investigate the possibilities of obtaining BaTiO3 by
applying mechanochemical methods [10-20].
2. Experimental
The aim of the present work is to study the phase composition
of experimental compositions obtained from batches prepared from
BaCO3, (Alfa Aesar, pa 99.9 %) TiO2 (anatase) (Alfa Aesar, pa 99.9
%), Bi2O3 (Alfa Aesar, pa 99.8 %) and doping components
introduced into the general composition in the form of oxides,
subjected to mechanochemical activation (up to 2 h) and heat
treatment up to 900°C (with an isothermal hold of 1 h).
In a Fristche Pulverisette 7 planetary mill with steel pots and
grinding bodies, the powdery starting components, calculated in
advance with the required stoichiometric composition, are placed at
a rotation speed of 500 rpm (revolutions per minute). The powders
are mixed for up to 2 hours in the presence of isopropanol in order
to good homogenization of the mixture and destruction of
agglomerates.
After analyzing the obtained data, the guidelines for
optimizing the technological regime for obtaining a single-phase
final product from BaTiO3 (tetragonal phase) doped with Bi by
applying mechanochemical activation and heat treatment at
temperatures significantly lower than those required for classical
synthesis at standard conditions.
For the purpose of mechanochemical synthesis, preliminary
thermodynamic calculations were carried out in order to determine
temperatures at which phases with desired properties can be
obtained. The possibilities of obtaining a single phase of BaTiO3
from different starting substances are presented in table 1.
Table 1: Thermodynamic parameters for barium titanate
synthesis
Compounds
∆Ho298
cal/mol
∆So298
cal/mol grad
Cp25
BaCO3
-295000
26,8
20,40
BaO
-135000
16,08
10,82
Ba(OH)2
-258
24,8
BaTiO3
-397,600
25,82
TiO2(анатаз)
-218,100
11,93
13,23
CO2
-94052
51,06
H2O
-57798
45,106
INNOVATIONS 2025
70
3. Results and Discussion
DTA (Differential Thermal Analysis) and
Thermogravimetry (TG) are presented in figures (Fig. 1).
Fig.1 DTA and TG of BaTiO3 raw materials.
Fig. 1 shows the DTA and TG curves of the barium titanate
parent component BaTiO3. The DTA curve shows one exothermic
peak in the range 188°C to 223°C and four endothermic peaks in
the range 598÷693°C; 830÷862oС, 899÷921°С. Endothermic
behavior is accompanied by a weight loss of approximately 14%.
From the figure, endothermic and exothermic peaks are found
during heat treatment up to a temperature of 1000°C, while the
thermo-gravimetric (TG) curve shows a weight loss of 14% upon
annealing to the same temperature. In the temperature range from
430 to 1000°C, the TG curve shows a 15.8% weight loss, which
takes place in three consecutive stages; characterized by three DTA
peaks at 633, 862, and 921°C. Maximum weight loss is observed at
915°C.
X-ray analysis was performed using a Bruker D8 Advance
automatic powder X-ray diffractometer with CuKα radiation (Ni
filter) and registration by a LynxEye solid-state detector. The X-ray
spectrum was recorded in the angular range from 10 to 80° 2θ with
a step of 0.02° 2θ and a counting time of 17.5 s/step. Qualitative
phase analysis was performed using the International Center for
Diffraction Data (ICDD) PDF-2(2009) database.
Fig.2 XRD analysis of the synthesized crystalline phases
BaTiO3, Bi12TiO20, Bi2O3
The presence of the crystalline phases was established:
BaTiO3 (dominant phase, red marking), Bi2O3 (blue marking) and
Bi12TiO20 (green marking). The recorded predominant barium
titanate (tetragonal shape) represents the synthesized target phase.
From the presented radiographs (fig. 2 a, b, c, d), it is possible to
trace the dependence and change of some of the characteristics of
the compositions with increasing homogenization time and the
applied mechanochemical treatment in the interval from 30 min to 2
hours.
The intensity of the main peak at 31.5o in the four
diffractograms increases significantly as in Fig. 2 d (2 hours of
mechanochemical activation) we have an intense peak in contrast
with fig. 3 a) activation after 30 minutes. A decrease in the
crystallite size from 118 nm to 20 nm of the target phase BaTiO3
was found. At 28o, a glass-ceramic phase Bi12TiO20 is formed.
After the analysis of the obtained experimental data, as a
main recommendation for the purpose of a more efficient and
complete course of the reaction processes, synthesis of a
monophasic end product and the absence of residual amounts of
starting reagents, it can be recommended to further increase the
period of mechanochemical treatment by 1-2 hours. Such a partial
change of the used technological mode is perceived as more
expedient and cost-effective than applying an increase in the
temperature of maximum heat treatment and the time of isothermal
loading.
The study of phase transformations in different oxide
systems subjected to mechanochemical treatment is not only of
fundamental scientific interest, but is related to the development of
innovative technological solutions potentially applicable in practice.
Obtaining different products at lower temperatures than those
required for classic high-temperature syntheses would significantly
ease the technological process and reduce costs related to the need
for appropriate equipment, inevitable depreciation of equipment and
increased consumption of energy carriers.
. 4. Conclusion
The participation of bismuth in the starting products for
synthesis contributes to a more complete preparation of barium
titanate as a final product. By applying prolonged mechanochemical
activation, the degree of homogenization of the prepared batches
increases significantly, the contact active surface of the introduced
reagents increases, the crystallite sizes decrease, energetically
excited states are formed, which determines the effective increase in
the reactivity of the system and the lowering of the temperature
values necessary for phase synthesis. The synthesized titanate phase
is potentially applicable in practice in the form of nanosized
powdered products with high specific surfaces or molded
monolithic products prepared by applying appropriate technological
methods. Another current possibility is the deposition of the
obtained phase in the role of thin films, single-layer or multi-layer
coatings on metal surfaces and the preparation of details with a
variety of functional purposes.
References
[1] Kandasamy S., Prema R., S., Methods of synthesis of nano
particles and its applications, Review Article, Journal of Chemical
and Pharmaceutical Research, 2015, 7(3), pp. 278-285.
[2] Dhand Ch., Dwivedi N., Loh X. J., Ying A.,Verma N.K.,
Beuerman R. W., Lakshminarayanan R., Ramakrishna S., Methods
and strategies for the synthesis of diverse nanoparticles and their
applications: a comprehensive overview, The Royal Society of
Chemistry - RSC Adv., 5, 2015, pp. 105003–105037.
[3]. Satyanarayana T., Reddy S., A Review on Chemical and
Physical Synthesis Methods of Nanomaterials, International Journal
for Research in Applied Science & Engineering Technology
(IJRASET), Vol. 6 Issue I, 2018, pp. 2885-2889.
[4]. Yordanov S., Bachvarova-Nedelcheva A., Iordanova R.,
Stambolova I., "Sol-gel Synthesis and Properties of Sm Modified
INNOVATIONS 2025
71
TiO2 Nanopowders",Bulgarian Chemical Communications 50,
2018, pp. 42-48.
[5]. Montoro-Leal P., García-Mes J., Mar López Guerrero M.,
Alonso E., Comparative Study of Synthesis Methods to Prepare
New Functionalized Adsorbent Materials Based on MNPs–GO
Coupling, Nanomaterials 2020, 10, 304, pp. 1-18.
[6]. Tsuzuki, T., & McCormick, PG (1999). Mechanochemical
synthesis of metal sulphide nanoparticles. NanoStructured
Materials, 12, pp 75–78.
[7]. Stojanovic DB., Mechanochemical synthesis of ceramic
powders with perovskite structure. J Mater Process Technol 2003,
143-144: 78-81.
[8]. Tsuzuki T., McCormick P., Mechanochemical synthesis of
nanoparticles, Journal of materials science, 3, 9 (2004), pp. 5143-
5146.
[9]. Tojo T, Zhang QW, Saito F. Mechanochemical synthesis of
FeSbO4-based materials from FeOOH and Sb2O5 powders. Powder
Technol 2008, 181: 281-284.
[10]. Zyryanov V., Mechanochemical synthesis of complex oxides,
Uspekhi khimii, 77, (2), 2008, 107-137–monographs.
[11]. Achimovičová M., Daneu N., Rečnik A., Ďurišin J., Baláž P.,
Fabián M., Kováč J., Šatka A., Characterization of
mechanochemically synthesized lead selenide, Chemical Papers 63
(5) 562–567 (2009).
[12]. Hao WU, Qiang LI, Application of mechanochemical
synthesis of advanced materials, Journal of Advanced Ceramics
2012, 1(2): 130-137.
[13]. Gancheva M., Iordanova R., Dimitriev Y., Nihtianova D.,
Stefanov Pl., Naydenov A., "Mechanochemical synthesis,
characterization and catalytic activity of Bi2WO6 nanoparticles in
CO, n-hexane and methane oxidation reactions", J. Alloy. Compd.,
570 (2013) 34-40.
[14]. Gancheva M., Aleksandrov L., Iordanova R., Dimitriev Y.,
"Synthesis of amorphous and crystalline LaBWO6 using
mechanochemical activation", J. Chem. Techn. Metall., 50(4)
(2015) 467-473.
[15]. Bulina N.V., Chaikina M.V., Vinokurova O.B, Prosanov I.Y.,
Lyakhov N.Z., Low-temperature mechanochemical synthesis of
zinc-substituted hydroxyapatite, Chem. Sustain. Dev. Vol. 27, 2019,
pp. 251-256.
[16]. Ivanov E., Suryanarayana C., Materials and Process Design
through Mechanochemical Routes, Journal of Materials Synthesis
and Processing, Vol. 8, Nos. 3/4, 2000, pp. 235-244.
[17]. Ohara S, Abe H, Sato K. Effect of water content in powder
mixture on mechanochemical reaction of LaMnO3 fine powder. J
Eur Ceram Soc 2008, 28: 1815-1819.
[18]. Stojanovic B.D., Simoes A.Z, Paiva-Santos C.O., Jovalekic
C., Mitic V.V., Varela, JA: Mechanochemical synthesis of barium
titanate. JE Sci. Europ. Ceram. Soc. 25 (2005) 1985-1989.
[19]. Ohara S., Kondo A., Shimodo H., Sato K., Abem H., Naito
M., Rapid mechanochemical synthesis of fine barium titanate
nanoparticles, Materials Letters, 62 (2008), pp. 2957-2959.
[20]. Mikherdjee S., Ghosh S., Chandra G.,Mitra M.,
Mechanochemical synthesis of barium titanate and Mn-doped
barium titanate,InterCeram: International Ceramic Review, 62 (1),
2013, pp. 45-49.
INNOVATIONS 2025
72
Изследване устойчивостта на междукристална корозия на образци от стомана 316L,
създадени на база на метода „Metal Injection Molding”
Research of the resistance to intergranular corrosion of 316L steel samples created on the
method “Metal Injection Molding” technology
Kalin Anastasov
Technical University Gabrowo, Bulgaria
kalinanastasov@abv.bg
Abstract: This paper presents research of austenitic steel 316L to intergranular corrosion. The study was conducted on experimental
samples produced by the MIM technology at different values of the technological process. The sintering process was carried out under the
provision of an inert environment of nitrogen process gas.
Keywords: MIM, STEEL 316L, INTERGRANULAR CORROSION, PROCESS GAS.
1. Въведение
Неръждаемата аустенитна стомана 316L е често прилаган
материал за производството на детайли и части, работещи в
агресивни среди, в химическата, автомобилната, военната,
часовникарската промишлености и медицината. Процесът на
експлоатация на тези детайли, често е свързан със значителни
термични и химични претоварвания свързани с особеностите
на работната среда, които те трябва да понесат, без да загубят
експлоатационните си качества. Тази устойчивост се залага в
процеса на проектирането на детайлите, като значителна роля
оказва и технологията за производството им. Аустенитните
стомани от този клас при нагряване до температури от 4500С –
8000С, по границите на зърната се образуват хромови карбиди
и нитриди и губят способността си за пасивиране [1, 2]. При
експлоатация в агресивна среда, започва развитие на
междукристална корозия, която се изразява в анодното
разтваряне на карбидите и образуване на пукнатини в
повърхностният слой на детайлите [3, 4].
Неръждаемите стомани спадат към труднообработваемите
материали. Обработката, чрез класическите методи със
стружкоотнемане при производството на сложни като
геометрия и точни размери детайли, понякога е силно
затруднено и свързано със значителен разход на инструменти и
повишен брак. Изработването на такъв тип детайли е
възможен, чрез MIM (Metal Injection Molding) технология –
инжекционно формоване на метали. Методът се осъществява
на четири основни етапа – смесване и гранулиране на полимера
и металният прах, инжекционно формоване, депластификация
и спичане (синтероване). Синтероването се извършва при
високи температури – 13700С – 13900С, в безокислителна среда
създадена от инертни газове – водород H2 или азот N2 или
вакуум в зависимост от типа на съоръжението. Именно при
този етап се създават условия за създаване на хромови карбиди
или нитриди по границата на зърната. Не е препоръчително
спичането да се извършва в условия на прилагане на N2 като
инертен газ, поради повишената вероятност от образуване на
нитриди, поради неговата ниска цена и безопасност при
експлоатация, в сравнение взривоопасният H2.
След преминаване на етапите, готовите детайли получават
съответните форма, размери, физико-механични и
експлоатационни качества. Върху тях съществено влияние
оказват параметрите на технологичния процес - температура,
време, работна среда на съоръжението осигуряваща
безокислителната среда.
Цел на статията е изследване и оценка на
корозоустойчивостта и някои механични параметри, на пробни
тела изработени от аустенитна неръждаема стомана, марка
316L, по Metal Injection Molding технология, синтеровани в
азотна среда, при различни температури. За постигане на
основната цел е проведено експериментално изследване на
образци изработени от аустенитна неръждаема стомана марка
316L получени чрез метода MIM съгласно БДС 7039-85,
модификация на БДС ISO 3651-2, метод А[5].
2. Материали, оборудване и методи
2.1 Материал
За изработването на пробните тела необходими за
провеждане на изследванията е използван гранулат с марка
Catamold 316L, производство на фирмата BASF.
2.2 Пробни тела (епруветки)
За нуждите на изследването и оценка устойчивостта на
междукристална корозия на аустенитна неръждаема стомана
316L са изработени 6 групи пробни тела (епруветки) за
изпитание на статичен опън, с форма и размери съгласно ISO
2740:2009 [6], за праховометалургични материали, представени
на фигура 1.
Фиг. 1 Експериментален образец за статичен опън по ISO 2740:2009,
за праховометалургични материали
2.3 Оборудване
2.3.1 Производствено оборудване
Технологичния процес за изработване на пробните тела
(епруветки), включва три етапа на производство, със
съответното оборудване:
- Инжекционно формоване – за етапа на инжекционното
формоване е използван автомат за инжекционно
формоване „BATTENFELD”, mod. SmartPower 60/350,
снабден със специализиран шприцагрегат и
инструментална екипировка „Шрицформа“, проектирана
за производство на пробните тела.
- Депластифициране и синтеровне – за етапа на
депластифициране и синтероване е използвана поточна
автоматична линия марка “CREMER”, представена на
фигура 2.
INNOVATIONS 2025
73
Фиг. 2 Автоматична поточна линия за депластифициране и
синтероване „CREMER”
2.3.2 Лабораторно оборудване
За нуждите на изследванията е използвано следното
лабораторно оборудване
- За изследване химичният състав на пробните тела е
използван спектрометър BELEG LAB 3000s.
- За изследване грапавостта на пробните тела е
използван грапавомер “DIAVITE”.
- За изследване на физикомеханичните характеристики
е използвана машина TESTOMETRIC M500-100AT.
- За провеждане изпитването на междукристална
корозия е използвана лабораторна установка, описана в [1, 2],
която е представена на фигура 3.
- За изследване на плътността по хидростатичен метод
и загубата на маса след теста за изпитание на междукристална
корозия е използвана аналитична везна, снабдена със стенд за
измерване на плътност по хидростатичен метод.
- За провеждане на макро и микроструктурните
изследвания е използван инвенторен металографски
микроскоп, mod. BS-6030.
2.4 Експериментални образци
За нуждите на изследването са произведени шест групи
експериментални образци, при различни параметри на режима
на синтероване – време и температура, като за осъществяване
на инертна среда в съоръжението е използван газ азот N2, с
чистота 99,998. Общ вид на образците е представен на фигура
4.
Фиг. 4 Експериментални образци
3. Експериментални резултати и коментари
3.1 Изходни параметри на пробните тела след
синтероване
От всяка група синтеровани образци са подбрани на
случаен принцип по два броя и са измерени, грапавост, маса и
плътност, данните от измерванията и параметрите на режима
на синтероване са систематизирани в таблица 1.
Резултатите от химичния анализ на пробните тела са
представени в таблица 2.
Фиг. 3 Лабораторна установка
Таблица 1: Параметри на пробните тела след синтероване
Група
№
Експериментален
образец №
Температура
на
синтероване,
ОС
Време на
цикъла, min
Маса след
синтероване, g
Плътност
1
1-1
1375
25
28,0908
7,46
1-2
28,0815
7,48
2
2-1
1375
30
28,1242
7,52
2-2
28,1481
7,54
3
3-1
1375
35
28,1113
7,50
3-2
28,1119
7,48
4
4-1
1390
25
28,1000
7,50
4-2
28,1119
7,51
5
5-1
1390
30
28,1440
7,51
5-2
28,1208
7,45
6
6-1
1390
35
28,0916
7,54
6-2
28,1221
7,54
INNOVATIONS 2025
74
Таблица 2: Химичен състав на пробните тела след синтеровоане
Химичен
състав%
Групи
1
2
3
4
5
6
C
0,043
0,047
0,038
0,053
0,051
0,031
Si
0,678
0,707
0,642
0,743
0,647
0,731
Mn
0,479
0,516
0,576
0,438
0,448
0,464
P
0,006
0,006
0,005
0,006
0,007
0,006
S
0,001
0,001
0,001
0,001
0,001
0,001
Cu
0,027
0,024
0,022
0,025
0,027
0,025
Al
0,002
0,002
0,001
0,003
0,001
0,002
Cr
16,30
16,15
16,24
16,04
16,33
16,13
Mo
2,234
2,237
2,206
2,239
2,214
2,227
Ni
10,31
10,26
10,26
10,25
10,10
10,24
V
0,028
0,026
0,022
0,028
0,026
0,025
Ti
0,009
0,009
0,008
0,009
0,009
0,009
Nb
0,001
0,001
0,001
0,001
0,001
0,001
Co
0,012
0,011
0,010
0,010
0,011
0,011
W
0,001
0,001
0,001
0,001
0,001
0,001
3.2 Методика на провеждане на изпитанието
Изпитването на устойчивост срещу корозия е проведено
по БДС 7039-85, метод А. Методът се основава на задържане
на пробните тела в кипящ воден разтвор на CuSO4 в
концентрирана H2SO4, поместени в стъклена колба с медни
стружки и хладник. Методът е адаптиран за провеждане на
ускорени изпитания. За провеждане на изследването са
използвани цитираните в стандарта реактиви и оборудване.
След провеждане на изпитанието, образците са оценени
на база заложените стандартни критерии за устойчивост
срещу обща и междукристална корозия, като допълнително
са подложени на изпитание на опън.
3.3 Параметри на пробните тела след изпитанието на
устойчивост на корозия
- Загуба на маса – резултатите са поместени в таблица 3.
Таблица 3: Загуба на маса след изпитанието
Група
№
Епруветка
№
Изходна
маса, g
Маса след
изпитанието,
g
ΔМ,
%
1
1-1
28,0908
27,7380
1,26%
1-2
28,0815
27,9045
0,63%
2
2-1
28,1242
27,8483
0,98%
2-2
28,1481
27,9479
0,71%
3
3-1
28,1113
27,9494
0,58%
3-2
28,1119
27,8609
0,89%
4
4-1
28,1000
27,7310
1,31%
4-2
28,1119
27,8256
1,02%
5
5-1
28,1440
27,8841
0,92%
5-2
28,1208
27,9106
0,75%
6
6-1
28,0916
27,4490
2,29%
6-2
28,1221
27,7631
1,28%
- Изпитание на статичен опън – резултатите са поместени в
таблица 4.
Таблица 4: Физико-механични резултати на пробните тела
Група
№
Епруветка №
Сила на разрушаване F, N
Средна сила на
разрушаване на гупата Fср.,
N
Граница на провлачване R0,2,
N/mm2
Средна граница на
провлачване на групата, R0,2,
N/mm2
Якост на опън Rm, N/mm2
Средна якост на опън на
групата, Rm ср, N/mm2
Относително удължение A10,
%
Средно относително
удължение на групата, A10, %
1
1-1
12686
12271
355
349,5
597
577,5
33
31
1-2
11856
344
558
29
2
2-1
12673
12735
361
366,5
597
599,5
30
29,5
2-2
12797
372
602
29
3
3-1
11958
11839
338
335
563
557,5
27
28
3-2
11720
332
552
29
4
4-1
11753
11730
323
327
553
552
31
30
4-2
11707
331
551
29
5
5-1
11672
11898
347
351
550
560,5
26
26
5-2
12124
355
571
26
6
6-1
11579
11459
322
320
545
539
30
29,5
6-2
11339
318
533
29
INNOVATIONS 2025
75
- Макроанализ на образците след изпитание на опън –
резултатите са представени на фигура 5.
1 2
3 4
5 6
Фиг. 5 Състояние на повърхността след статичен опън на
отделните групи образци
- Микроанализ на образците - снимки на направените
микрошлифове са представени на фигура 6.
1 2
3 4
5 6
Фиг. 6 Микрошлифове на отделните групи образци
4. Заключение
Проведено е експериментално изследване с цел
определяне и оценка на корозионната устойчивост и някои
механични параметри на пробни тела от аустенитна
неръждаема стомана, марка 316L, получени чрез Metal
Injection Molding технология, синтеровани в азотна среда,
при различни температури на процеса в инертна среда
обезпечена от азот.
- След проведените изпитания може да бъдат направени
следните изводи и заключения:
- Наблюдава се, че най-ниска загуба на маса се отчита
при групи 3 и 5, а най-висока при група 6. Разликата
между групите с най-ниска и най-висока загуба на маса
изразено в процентно съотношение при взети средни
стойности е 58,8%.
- Най-високи стойности от механичните изпитанията
имат образците от група 2, а най ниски от група 6.
- Наблюдава се разлика от около 10 – 12 % в механичните
показатели сила на разрушаване, якост на опън и
граница на провлачване между образците от група 2 и
група 6, при еднакво относително удължение между тях.
- При анализа на макроструктурата след проведеното
изпитание на статичен опън при всички групи се
наблюдава нацепване на повърхностния слой, което е
характерно при наличие на междукристална корозия.
- При микроструктурния анализ се вижда ясно изразена
междукристална корозия в повърхностния слой, като
дълбочината на нейното проникване е различно при
групите. Най-голяма дълбочина има при образци от
групи 2, 3 и 6 – около 100μm, а най-малка при групи 3 и
5 – около 30μm.
- Наблюдава се също, че в участъците с повишена
пористост на образците има по-ясно изразена корозия,
което води до извода, че пористостта влияе особено
негативно, върху корозоустойчивостта на материала.
5. Литература
1. Гуляев, А. Металловедение, М., Машиностроение, 1977.
2. Томашов, Н. Теория коррозии и корозионностойкие
конструкционные сплавы, М., Металлургия, 1986
3. Желев, А. Материалознание. Техника и технология, т. ІІ:
Технологични процеси и обработваемост, Булвест 2000,
София, 2002.
4. Райчев, Р. Корозия и защита на материалите, София,
Нови знания, 2001.
5. БДС 7039-85
6. ISO 2740:2009.
INNOVATIONS 2025
76
On application of generative artificial intelligence in higher education: insights from
bulgarian students
Teodora Varbanova1, Diana Netova1, Nikolay Netov1, Kamen Spasov1
Sofia University "St. Kliment Ohridski"1
teodorav@feb.uni-sofia.bg, dianaht@feb.uni-sofia.bg, nnetoff@feb.uni-sofia.bg, kspassov@feb.uni-sofia.bg
Abstract: The rapid integration of generative artificial intelligence (Generative AI) tools into higher education has significantly influenced
student learning behaviour and academic practices. This research paper explores the ways in which undergraduate students at Sofia
University "St. Kliment Ohridski" apply these technologies for both academic and non-academic purposes. The results reveal that students
predominantly use Generative AI for academic support, such as explaining complex concepts, conducting research, and assisting in writing
tasks. At the same time, they also engage with Generative AI for personal productivity, communication, and creative exploration. Despite
having generally positive perceptions of academic effectiveness, students show critical awareness of risks related to information accuracy,
privacy, and ethical use.
KEYWORDS: GENERATIVE AI, HIGHER EDUCATION, STUDENTS, BULGARIA, AI IN LEARNING, EDUCATIONAL TECHNOLOGY,
CHATGPT
1. Introduction
The introduction of powerful generative AI tools since 2022 has
rapidly transformed student behavior in higher education
worldwide. University students began experimenting with AI for a
range of academic and personal tasks. Surveys and studies from
2023 and 2024 indicate that a majority of college students have now
used AI in their studies, often viewing it as a useful aid despite
ongoing debates about academic integrity [1]. Researchers have
characterized the rapid spread of artificial intelligence utilization
among students as a phenomenon with both potentially beneficial
and damaging implications: it offers new learning opportunities but
also raises concerns about plagiarism, skill development, and equity
[2], Early evidence suggests students are largely positive about the
technology’s potential to assist their education [3], yet they remain
cautious about its limitations and the propriety of using AI in
graded work [4]. This paper reviews the (1) academic applications
of generative AI by students, (2) non-academic uses for personal or
creative purposes, and (3) the benefits, as well as risks and ethical
concerns, identified in recent research documents. The paper
present results from a survey entitled "Use of Generative AI Tools
for Education and Other Purposes" conducted among undergraduate
students from Sofia University "St. Kliment Ohridski", Faculty of
Economics and Business Administration.
2. The Research Backgrounds
University and college students have eagerly adopted generative
AI as a study aid and productivity tool for a variety of academic
tasks. Research consistently shows that the cases of academic-
related use are the primary way students leverage AI, often more
than for personal or entertainment purposes [5], Common academic
applications of generative AI include:
Writing assistance and content generation. The most
widespread use is as a writing support for assignments and essays.
Students use AI chatbots to brainstorm ideas, generate initial drafts,
refine their phrasing, and check grammar. In a global 2024 survey,
66% of students reported using ChatGPT or similar tools for writing
tasks, including creating first drafts (24% of respondents) and
paraphrasing text (28%) [1]. Qualitative studies have found that
students primarily use AI to improve the clarity and expression of
their own ideas rather than to write entire essays for them [6].
Consistently, only a negligible proportion (approximately 5%) of
students acknowledge submitting AI-generated content without
significant modification, suggesting that the predominant utilization
pattern is as an additional academic resource rather than as a full
replacement for original authorship [7].
Study and learning support. Study Aids and Learning
Support: Generative AI applications have evolved into essential
academic companions for a significant portion of the student
population. Tools like ChatGPT are used to explain difficult
concepts, summarize readings, and provide tutoring on demand [8].
More than one-third of students in multiple surveys report using AI
as a personal tutor – asking it to rephrase confusing course material,
provide examples, or test their knowledge with practice questions.
For example, 36% of UK undergraduates surveyed in 2023 said
they use AI to help explain concepts for assessments (acting as an
AI tutor) [7]. Similarly, 34% of daily AI-using students in the U.S.
use generative AI to summarize or paraphrase academic texts and
lecture notes for better understanding [8]. Clarification and review
capabilities are seen by students as a major advantage of AI,
allowing more self-directed learning and revision outside of formal
classes [9].
Preparing homework and answering research questions.
Many students turn to AI chatbots as an alternative search engine or
Q&A resource when they face difficulties with homework. In the
global 2024 survey, the single most common usage was searching
for information (69% of students) through AI [1]. In research-heavy
assignments, students also use AI to gather background information
or identify relevant literature. However, they have noted that
chatbots sometimes produce legitimate-like but incorrect answers or
fake citations, requiring verification [4].
Coding and technical assistance. Generative AI coding
assistants (like GitHub’s Copilot or GPT-4) are becoming
increasingly popular among students in STEM fields. STEM
students have embraced AI to help write code, debug errors, and
generate algorithms for programming assignments [10]. University
surveys indicate that while these tools are used less commonly than
writing assistance, a notable subset of students (particularly in
computer science/engineering) regularly consult AI for coding help
[4].
Presentations and other academic tasks: Beyond text and
code, students use generative AI to support various other
educational activities. This includes creating presentation content
(e.g. generating slides or talking points), data analysis assistance
(using AI to interpret data or perform calculations), and even
language translation for academic materials [8]. AI image
generators are occasionally used to produce illustrations or figures
for assignments and design projects though text-based applications
dominate [7].
Beyond formal academic tasks, students are also exploring
generative AI for a range of non-educational, personal, and creative
purposes. While cases of academic use dominate overall, surveys
show that a significant fraction of students have engaged with AI
tools for everyday tasks, creative projects, or just for fun [4].
Common non-academic usage of generative AI among students
include:
Personal productivity and organization: Many students use
AI as a personal assistant to perform daily tasks. For example, AI
tools help with scheduling, reminders, and task management.
Another common use is drafting routine communications – for
INNOVATIONS 2025
77
example, using AI to draft an email or message which the student
can then refine [8].
Career development: Generative AI has become a valuable
tool for students preparing to enter the job market. Career centers
report students using AI to write or polish resumes, cover letters,
and internship applications. In a recent study, about 31% of daily
student AI users leveraged it for resume/cover letter writing [8].
Language translation and communication support:
multilingual students and non-native English speakers utilize AI for
translation and language practice. Advanced AI translators can
convert text between languages with high accuracy, which assists
international students in both academic work and everyday
communication. For example, students might write an email or
essay in their first language and use AI to translate it into English
(or vice versa) as a starting point. The Higher Education Policy
Institutes' survey found that 38% of students expect to use AI after
graduation for translating text in their jobs [7].
Personal advice and emotional support: a very interesting and
controversial aspect of students' AI usage is for advice, coaching, or
emotional support. Some of them treat AI chatbots as a
nonjudgmental listener or life coach, asking for mental health tips,
study motivation, or relationship advice. While AI is not a substitute
for professional counseling, students seem to appreciate the privacy
and availability of AI for personal advice or even stress relief [4].
In summary, generative AI has impacted students’ daily
routines in various ways. Academic usage remains predominant, but
many students also experiment with AI for personal productivity,
creativity, social connection, and career preparation.
While writing assistants offer substantial benefits for improving
students’ written work by pointing out errors, suggesting
rephrasing, improving coherence etc., and helping them produce
higher quality papers, at the same time an important concern is that
generative AI enables forms of plagiarism or cheating by allowing
students to produce work that is not their own. Both students and
faculty are concerned about the ease of having AI write papers or
solve exam questions. Surveys confirm that a large majority of
students consider certain uses of AI to be unethical – for instance,
62% of students in Sweden said that using a chatbot during an exam
is cheating [5], and 70% of UK students were opposed to peers
using ChatGPT to write an entire essay for them [4].
It is proven that Generative AI assistants tend to produce
incorrect or even fake information with a confident tone, a
phenomenon known as hallucination. This poses a serious risk for
students who are not aware of this threat. A survey shows that 35%
of students who use AI did not know how often these tools emit
incorrect information [7].
Another aspect of concern is revealed by a recent survey of
Harvard undergraduates about AI's impact on career prospects
across all fields. Approximately 55% of students report that
generative AI has changed how they think about their future
careers, while around 45% worry AI will negatively affect their
career plans. This concern remains consistent across diverse career
paths, including technology, research, finance, public health,
politics, education, and consulting, with all fields showing similar
levels of concern (approximately 45%) [11].
3. Data and Methodology
The survey entitled "Use of Generative AI Tools for Education
and Other Purposes" was implemented online in April 2025 among
undergraduate students from bachelor programs at the Faculty of
Economics and Business Administration, Sofia University "St.
Kliment Ohridski". The survey was designed using Microsoft
Forms in the internal Microsoft 365 tenant of the faculty and
consisted of 12 closed-ended questions, 8 single choice and 4
multiple choice ones. The survey aimed to unveil how students use
Generative AI for both academic and non-academic purposes, as
well to identify students’ perceptions on the impact on their future
careers. The survey was sent to 364 respondents through the
internal messaging system of Sofia University, all part of the
English language programs. The survey was sent on April 16th and
closed on April 23rd, 2025. Out of the 364 distributed
questionnaires, 45 valid responses were received (n=45), resulting
in a response rate of approximately 12.4%. The survey has 2
separate parts: use of Generative AI tools for academic and use of
Generative AI tools for non-academic purposes. Students who
reported not using Generative AI tools for non-academic purposes
(4) did not participate in this part of the survey. No incentives were
offered and participation was entirely voluntary and anonymous.
4. Results and Discussion
All respondents are active users of Generative AI tools. Chart 1
illustrates the comparative frequency of Generative AI tool usage
across academic and non-academic contexts. Three usage categories
were analyzed: daily, several times a week, and occasional use.
Chart 1. How often do you use Generative AI tools for academic purposes? /
Beyond academics, how often do you use generative AI in your daily life?
The data reveals that daily usage is the most common pattern in
both contexts, though slightly more is reported for academic
purposes. Specifically, 46.7% of respondents reported using
Generative AI tools daily for academic activities, compared to
41.5% for non-academic tasks. This suggests that for a significant
proportion of users, AI tools have become integrated into their
routine academic workflows. The second most frequent usage
pattern is "several times a week." Again, academic use remains
dominant, with 35.6% of students using Generative AI regularly
throughout the week for their studies. In contrast, only 24.4%
reported using AI with the same frequency for non-academic
purposes. This highlights a more structured and consistent reliance
on AI for education compared to a more flexible and situational use
in their personal life. The most noticeable difference appears in the
"occasionally" category. Here, non-academic use significantly
outpaces academic use: 34.1% of respondents use Generative AI
tools occasionally for personal reasons, while only 17.8% do so for
academic work. This suggests that while AI is often a core tool for
academic success, its use in non-academic domains may still be
exploratory or opportunistic.
Table 1 represents the results, showing different academic tasks
supported by Generative AI tools and measured by a multiple
choice question.
Table 1. For which of the following academic tasks do you use
Generative AI tools?
Explaining difficult concepts
84.4%
Research Assistance
75.6%
Brainstorming ideas
71.1%
Generating ideas for projects
57.8%
Solving math or coding problems
55.6%
Writing and editing papers
48.9%
Other
6.7%
INNOVATIONS 2025
78
The most frequently reported use of Generative AI is for
explaining difficult concepts, as selected by 84.4% of participants.
This indicates that students view AI tools as effective digital tutors,
helping them to clarify complex theories, ideas, or course materials.
Research assistance was the second most cited use, with 75.6% of
students leveraging AI to support activities, such as sourcing
information, formulating research questions, or structuring literature
reviews. This reflects a shift in research behaviors toward AI-
extended knowledge discovery. Brainstorming ideas was selected
by 71.1% of respondents, emphasizing the value of AI in the early,
creative stages of academic work. Similarly, 57.8% of students used
AI tools to generate ideas for projects, showing that Generative AI
is a key support mechanism in ideation and project planning. More
than half of the students (55.6%) reported using AI to solve math or
coding problems, highlighting its growing role in STEM-related
disciplines. This demonstrates a broadening of AI application from
writing support to quantitative and technical problem-solving.
Writing and editing papers was mentioned by 48.9% of participants.
While slightly lower than the other categories, this still indicates a
substantial reliance on AI for drafting, revising, or polishing
academic texts. Lastly, 6.7% of respondents indicated "other" uses,
not listed in the main categories, but did not provide an answer.
Table 2 shows the results of the answers to the question about
the effectiveness of Generative AI tools in the academic context.
Table 2. How effective do you find generative AI tools in improving your
academic performance?
Very effective
42.2%
Somewhat effective
55.6%
Neither effective nor ineffective
2.2%
Somewhat ineffective
0.0%
Very ineffective
0.0%
The majority of survey respondents expressed a positive
perception of the impact of Generative AI tools on their academic
performance. Specifically, 55.6% of participants rated the tools as
"Somewhat effective", indicating a moderate but meaningful
contribution to their academic outcomes. An additional 42.2%
considered them "Very effective", reflecting a high level of
satisfaction. Combined, these two categories represent 97.8% of the
sample, demonstrating an overwhelming approval of Generative
AI’s role in education. Only 2.2% of respondents selected "Neither
effective nor ineffective", suggesting neutrality or uncertainty.
Importantly, none of the respondents found the tools "Somewhat
ineffective" or "Very ineffective", underscoring that negative
experiences with AI in academic contexts were either minimal or
non-existent among the surveyed group.
Table 3 presents the data from a multiple-choice question
related to the most common tasks performed by the students in non-
academic domain, with the support of Generative AI tools.
Table 3. What do you most commonly use Generative AI for in your personal
life?
Getting information or explanations
(instead of using search engines)
75.6%
Personal problem-solving
43.9%
Writing emails, messages, or posts
39.0%
Health/wellness information
36.6%
Learning new skills
34.1%
Generating ideas or content (art, music,
stories)
24.4%
Making decisions (e.g., shopping,
planning)
24.4%
News summarization
24.4%
Entertainment
22.0%
Other
2.4%
The most common non-academic use, as reported by 75.6% of
respondents, is getting information or explanations instead of using
traditional search engines. This suggests that Generative AI is
increasingly perceived as a faster or more intuitive alternative to
search engine platforms for answering everyday questions. Personal
problem-solving was the second most common use, as selected by
43.9% of participants. This includes tasks such as organizing
schedules, managing life challenges, or receiving advice etc.,
showing that students see value in AI as a practical support system.
Following closely are writing emails, messages, or social media
posts, (39%) and seeking health and wellness information (36.6%).
These uses reflect how AI tools are embedded in daily digital
communication and personal care routines. About a third of
respondents (34.1%) use AI tools to learn new skills, highlighting
their role in informal education and self-improvement. A smaller
group of respondents (24.4% each) reported using AI for generating
creative content, making decisions (e.g., shopping, planning), and
news summarization. This demonstrates AI’s multifunctionality in
multiple daily tasks. "Entertainment" was cited by 22% of
respondents, while only 2.4% selected "Other," indicating that most
use cases fall within the categories presented.
Table 4 presents data on students’ concerns related to the use of
Generative AI tools based on a multiple-choice question.
Table 4. Do you have any concerns about using generative AI tools?
Accuracy of information
78.0%
Dependency on technology
56.1%
Privacy issues
39.0%
Ethical implications
29.3%
Other
9.8%
I don't have any concerns
7.3%
The most frequently reported issue, as selected by 78.0% of
respondents, is accuracy of information. This finding suggests that
users are critically aware that AI-generated content may be factually
incorrect, misleading, or lacking credibility. The second most
reported concern, as identified by 56.1% of students, is dependency
on technology. This highlight concerns overreliance on AI for tasks
that traditionally require independent thinking or manual effort,
possibly affecting cognitive development, creativity, even academic
integrity. Privacy issues were flagged by 39.0% of respondents.
This includes fears about data collection, surveillance, or misuse of
personal information provided to AI platforms. Additionally, 29.3%
of participants expressed concern about the ethical implications of
AI use. These may include fairness, bias in AI outputs, the origin of
training data, or potential social impacts such as job displacement or
manipulation of information. A small group (9.8%) selected
"Other", suggesting concerns that fall outside standard categories
and are related to ecological impact. Only 7.3% of respondents
reported that they did not have any concerns about Generative AI
use. The overall results indicate that the majority of respondents
(over 90%) approach these tools with some level of caution or
critical awareness.
Chart 2 presents students’ perceptions of how Generative AI
tools will impact their future careers.
The majority of respondents (53.7%) believe Generative AI will
have a positive impact on their future careers. This reflects a strong
sense of optimism and confidence in the potential of AI to enhance
employability, productivity, and relevance in a rapidly evolving job
market. A significant portion, 34.1%, are not sure about the impact.
This uncertainty may come from a lack of clear understanding about
AI’s future role in different industries, or the evolving expectations
of employers. A smaller group, 12.2%, expects a negative impact.
These concerns may relate to fears of jobs automation, skill-based
lay-offs, or an over-dependence on technology in the jobs domain.
Interestingly, 0% of respondents selected "No impact", suggesting
that all participants anticipate some degree of influence from
Generative AI on their career.
INNOVATIONS 2025
79
Chart 2. How often do you use Generative AI tools for academic purposes? /
Beyond academics, how often do you use generative AI in your daily life?
Chart 3 illustrates the results of the most popular Generative AI
agents used by surveyed students and are based on a multiple choice
question.
Chart 3. Which Generative AI tools do you use?
ChatGPT is the most popular tool (91.1% of respondents),
followed by Copilot (40%), Deep Seek (33.3%), and Gemini
(26.7%). Other tools like Perplexity, Claude, and Mistral are used
far less frequently. The Bulgarian language model - BgGPT is used
by 15.6%, which is more than other globally known tools and it is
an indication of a preference for the national Generative AI
solution. An interesting trend is outlined on a secondary level of
data analysis: students were able to provide more than one answer,
so the ones who reported usage of 4 and more tools represent
19.5%, 3 tools - 26.8%, 2 tools - 29.3% and a single tool - 24.4%
(and it is only ChatGPT), which is an additional evidence of the
identified concern about information accuracy, forcing students to
perform cross-checking with different tools.
Despite high levels of usage and the variety of tools, only one-
third (34.1)% of participants had received any structured instruction
or guidance on how to use Generative AI tools effectively. The
remaining 65.9% learned to use these tools independently,
highlighting a gap in formal digital literacy education. As the
Faculty of Economics and Business Administration do not provide
formal training on Generative AI technics, students most probably
define different online courses on the market as formal training.
There is no significant difference in the answers based on gender
distribution, respondents included both male and female
participants, with a balanced representation: 60% female, 40%
male.
5. Conclusions
Generative artificial intelligence has become an integral part of
students’ academic and personal lives. They perceive Generative AI
as both beneficial and transformative in improving academic
performance and enhancing career readiness. At the same time, they
raise important concerns regarding the reliability of AI-generated
content, overreliance on technology, and ethical implications of
usage, especially in assessment settings. The overview of
worldwide reports and research papers, compared with the local
research, does not show any significant differences among students
in the way they use, evaluate, and perceive generative AI tools, both
in the academic and non-academic domain. The results clearly
indicate that higher education institutions must proactively
incorporate Generative AI literacy into their curricula. Structured
guidance, training sessions, and ethical frameworks can help
students use AI technologies in a more informed and appropriate
manner. As generative AI continues to evolve, educational systems
must adapt to ensure that these tools enhance learning without
compromising academic integrity or critical thinking development.
References:
[1]
Digital Education Council, "AI or Not AI: What Students
Want," 2024.
[2]
M. J. F. &. K. T. Abbas, "Is it harmful or helpful? Examining
the causes and consequences of generative AI usage among
university students.," International Journal of Educational
Technology in Higher Education, vol. 21, February 2024.
[3]
A.-M. M. M. A. B. a. A. A. A. Q. Gasaymeh, "University
Students’ Insights of Generative Artificial Intelligence (AI)
Writing Tools," Education Sciences, vol. 14, no. 1062,
September 2024.
[4]
H. W. R. S. E. e. a. Johnston, "Student perspectives on the use
of generative artificial intelligence technologies in higher
education," International Journal for Educational Integrity,
vol. 20, February 2024.
[5]
S. C. W. O. Malmström H., "Chatbots and other AI for
learning: A survey of use and views among university students
in Sweden," Chalmers University of Technology - Department
of Communication and Learning in Science, 2023.
[6]
R. T. B. Black, "University students describe how they adopt
AI for writing and research in a general education course,"
Scientific Reports, vol. 15, March 2025.
[7]
F. J., "Provide or punish? Students’ views on generative AI in
higher education," Higher Education Policy Institute, 2024.
[8]
C. Y. L. B. D. M. S. J. N. F. K. &. B. G. Shaw, "GenAI in
Higher Education: Fall 2023 update," Tyton Partners, 2023.
[9]
J. &. Y. S. &. D. R. &. L. N. Kim, "Exploring students’
perspectives on Generative AI-assisted academic writing,"
Education and Information Technologies, vol. 30, pp. 1265-
1300, 2024.
[10]
J. K. M. G. J. e. a. Kim, "Examining Faculty and Student
Perceptions of Generative AI in University Courses,"
Innovative Higher Education, 2025.
[11]
S. &. J. R. &. J. N. &. W. G. Hirabayashi, "Harvard
Undergraduate Survey on Generative AI," 2024.
INNOVATIONS 2025
80
Systematic Approach to Design Space Exploration of Pulley Supports Using Generative
Design
Tomislav Solar1, Ivan Grgić1,*, Mirko Karakašić1, Željko Ivandić1
University of Slavonski Brod, Mechanical Engineering Faculty in Slavonski Brod, Croatia1
igrgic@unisb.hr
Abstract: By utilizing the artificial intelligence capabilities of generative design, the main task was to generate an optimal pulley support
model that satisfies the initially applied conditions while reducing the overall mass and volume. The bracket fixation method presents a novel
approach to pulley support assemblies. Through Static Stress and Generative Design modules, various angular load applications of 1500 N
were considered, and the consequent mechanical characteristics were numerically compared and interpreted. The outcomes of the
generative design process provide a firm basis for manufacturing the optimal pulley support model through casting or additive technology.
Keywords: TOPOLOGY OPTIMIZATION, GENERATIVE DESIGN APPROACH (GDA), 3D PRINTING, PULLEY, MASS REDUCTION
1. Introduction
Pulley support is an important element in the transfer of power
and motion found in various industrial environments and
mechanical settings. The prevalence and applications of such pulley
systems have been well-documented since ancient times. The oldest
pulley that was used to change the direction of the applied force
belongs to Ancient Egypt, dated in the 12th dynasty period [1]. One
of the most memorable advancements following this theme is
Archimedes’ compound pulley system. This innovation enabled
users to manage the weight of an initially applied load with the
reduced amount of force required to do the work, catalyzing further
development of pulley systems and their expanding applications.
Pulleys can be used singly or in combination depending on the
need for work. Pulleys consisted of a wheel, axle, and rope (chain,
cable or belt). The wheel has a groove so that the rope can pass
around it. There are two types of pulleys: fixed pulleys and
moveable pulleys. Fixed pulleys are engaged to the fixed place
(beam, bar, ceiling, etc.). A fixed pulley is mainly used to change
the direction of the force. It is used for extracting something from
an unreachable depth or to carry something to an unattainable
height. In such pulleys, there is no gain from the way: The further
the rope is pulled, the higher the load will have lifted. Fixed pulleys
provide ease of doing work by changing the direction of the force.
On the other hand, moveable pulleys are not engraved to any fixed
place so they can move. One end of a rope is tied to a fixed
position, while the other end of the rope is free for applying the
force. In such pulleys, the weight is attached to the wheel, which is
located between the rope's ends. When we apply force in such
pulleys, both the weight and the wheel move in the same direction
as the force [2].
In the system dynamics analysis of the cable-pulley interaction,
one must bear in mind many different effects that have a variety of
impacts on the overall system, such as mass/inertial effects, friction,
damping, friction-induced vibrations, thermal effects, etc. In order
to examine what may cause sliding between the contacting surfaces,
one has to investigate both the static and dynamic characteristics.
Slip occurs when the inertial forces of the pulley exceed static
friction thus the surfaces slide against each other [3]. This paper
explores the behavior of a pulley support under static stress
conditions and investigates the corresponding results using Static
Stress and Generative Design modules of Autodesk Fusion 360
software.
Generative design is revolutionizing the way engineers and
designers approach the creation of optimized mechanical structures.
Concerning the multiple objectives, loads, constraints, etc.,
engineers can produce a variety of lightweight, sustainable and
efficient design alternatives in a short period of time. Generative
design helps streamline product development by producing
optimized and creative solutions more efficiently than traditional
methods. By emphasizing material efficiency, the software only
adds material where it is structurally necessary, making it well-
suited for additive manufacturing processes that create organic
shapes and forms. Considering that topology optimization modifies
an existing design by optimizing material layout within a predefined
design space, generative design offers entirely new solutions and is
often applied in complex and creative product development
processes [4].
2. Problem definition
The initial pulley support was custom made in Autodesk
Inventor Professional 2024 software. The assembly consists of six
parts (Figure 1): a nut, a washer, a bracket, a sleeve, a pulley and a
shaft.
Figure 1 Pulley support assembly in Autodesk Inventor
The bracket, with an overall height of 156 mm, is fixed at its
two upper holes, while the remaining parts are assembled on the
axis of the bottom hole.
The force is applied on the groove of the pulley and set at 1500
Newton’s, acting as a loaded imaginary rope under two cases: i)
force vector directed vertically down the axis and ii) force vector
oriented at a 45° angle relative to a reference axis.
The material set to be utilized in Autodesk Fusion 360 software
for the pulley part is gray iron, while all the other parts are set to be
made of stainless steel.
The objective is to minimize the mass of the assembly, mainly
the mass of the bracket that goes through the design process in the
Generative Design module. Minimum safety factor is limited to 4
(for the final outcome). A numerical value of 3, 4 or higher for the
safety factor is commonly found in calculations and literature for
industrial lifting [5].
It is important to be cautious when refining the mesh of the
model in Autodesk Fusion 360, as there are surface transitions
(geometric discontinuities - angular transitions) that require a finer
mesh for greater accuracy of results. Inadequate meshing can have a
big impact on the results of the equivalent stress in the final
outcome.
INNOVATIONS 2025
81
In order to obtain and compare the numerical results of the
safety factor, the equivalent stress and the displacement, previously
described pulley support assembly is imported in Autodesk Fusion
360 software and analyzed in Static Stress module before and after
generative process in Generative Design module.
3. Methodology
3.1 Static Stress
After the pulley support assembly has been modeled in
Autodesk Inventor, the Static Stress module in Autodesk Fusion
360 software is chosen for the deformation and stress analysis.
Option Simplify offers to remove features that are unnecessary
for the further simulation process. Two chamfers on the top and the
bottom of the shaft are removed using this feature, as shown in
Figure 2.
Figure 2 Assembly model before (left) and after (right) Simplify
option
The next step is to define the materials for the assembly parts.
Autodesk Fusion 360 software provides an extensive library of
materials for selection. The materials are chosen following the
description in section 2, and their properties are shown in Figure 3.
Figure 3 Materials definition
The pulley part is divided into 8 bodies because it was modeled
in Autodesk Inventor with revolve and circular pattern options.
The main material properties defined for stainless steel are:
ρ=8000 kg/m3, E=193 GPa, ν=0.30 and Re=250 MPa. The main
material properties defined for gray iron are: ρ=7150 kg/m3, E=90
GPa, ν=0,30 and Re=119 MPa.
Following the material definition, structural constraints are
placed on the top two holes of the bracket. The two cylindrical faces
are fixed meaning translational and rotational movements are
disabled in every direction as shown in Figure 4.
Figure 4 Structural constraints
Structural Load, selected as Force, is placed on four faces of the
pulley’s groove in the negative Z axis for the first case scenario.
Cloning the previous load case and orienting the force vector at a
45° angle relative to a reference axis, the second case scenario from
section 2 is applied. The force value is 1500 Newton’s. Figure 5
shows the first (a) and second (b) load case scenario.
a)
b)
Figure 5 Load case scenarios: first (a) and second (b)
The load can also be applied using the Bearing Load option
under Type.
Considering that the pulley support model is an assembly,
multiple parts are in contact with each other. Option Manage
Contacts defines contacts between two identified assembly parts.
Two contact types are recognized and applied: bonded and sliding.
Contact type bonded is applied between the parts of the pulley that
have been revolved and patterned together. They are perceived as
one body. For the contact type sliding, contacting bodies cannot
penetrate or separate from each other. However, the surfaces can
freely slide in the face-tangential direction relative to each other. It
is applied, for example, to the interacting surfaces of the sleeve and
INNOVATIONS 2025
82
the shaft. A total of 40 contact sets have been recognized. Figure 6
shows examples of both bonded and sliding contact types.
a)
b)
Figure 6 Contact types: bonded (a) and sliding (b)
Next step is to define mesh properties. The element size is
selected closer to 1% than to 10%, keeping in mind that the model
has geometric discontinuities (for example, 1% meaning the
generated mesh element size will be 1% of the overall model size).
Meshed model is shown in Figure 7.
Figure 7 Meshed model
After the study setup has successfully received all the
information required, option Solve is initialized and the Static Stress
results are provided. Tabular results of the initial simulation are
given in Table 1, while the visual comparison with the final results
is given in section 4.
Table 1 Initial Static Stress simulation results
Load case 1
Load case 2
Minimum safety factor
4.555
3.776
Maximum equivalent
stress, MPa
54.886
66.202
Maximum displacement,
mm
0.105
0.098
The minimum values of the safety factor correspond to the
values described in literature [5]. Displacement values of the two
load cases barely differ from each other.
For the necessary fulfilment of the tasks described in section 2,
Generative Design module is initialized.
3.2 Generative Design
The overall Generative Design framework, which relies heavily
on cloud-computing, is depicted in Figure 8 [6].
Figure 8 Generative Design framework [6]
Following this methodology, the same assembly model is
imported in Generative Design module and Edit Model option is
initialized. All the parts, except the bracket, are hidden (Unassigned
Geometry) so that the bracket could go through the design process
(section 2).
Three sketches are made around the holes of the bracket and
extruded for the local model thickness. Inside the Extrude option,
under Operation, New Body is chosen so that Fusion understands
each part as a separate entity (body). The three extruded cylinders
are used in Preserved Geometry and are conserved from any
modification by the software. Generative design process has no
effect on them. They are necessary for the fixation to the fixating
element (top two cylinders) or for the fixation of other assembly
parts (bottom cylinder). Figure 9 depicts bodies for Preserve
Geometry.
Figure 9 Bodies for Preserve Geometry (three cylinders)
Obstacle Geometry denotes bodies that prevent the generation
of generative model proposals. Their shape and volume do not
allow the optimal model to be generated precisely in their locations
and guides the model generation process.
INNOVATIONS 2025
83
Firstly, the rectangular sketch “of a wall” is extruded as New
Body on the top part of the bracket for 10 millimeters.
Secondly, activating the Connector Obstacle option, all the hole
volumes are filled with obstacle geometry with the bottom hole
having the obstacle geometry fill both sides. In that manner
volumes are conserved for assembly parts. Figure 10 shows the
modeling of the final, fifth obstacle geometry.
Figure 10 Obstacle geometries
The next step is to finally define all the geometries (bodies) in
Design Space. For Preserve Geometry, three cylinder bodies are
selected and highlighted in green. For Obstacle Geometry, five
bodies are selected and highlighted in red. For Starting Shape, the
bracket is selected and highlighted in yellow, meaning it will be the
part that goes through generative design process, bearing the
assigned load. Figure 11 shows the application of the previously
defined Design Space on the model.
Figure 11 Model configuration through Design Space
Option Objectives in Design Criteria allows for the definition of
mass reduction (Minimize Mass) and for the input of the safety
factor with the minimum numerical value of 4 (Section 2). Option
Manufacturing allows for the definition of the manufacturing
processes. Unrestricted, Additive and Casting are selected.
Selecting all the ejection directions in Casting allows for the
generation of every casting outcome during the design process.
Figure 12 depicts the chosen manufacturing processes.
Figure 12 Manufacturing processes for GDA
Option Design Conditions allows for the definition of loads and
constraints. The faces of the two top cylinders that are in contact
with the modeled wall (Figure 13) are treated as a fixed support.
Figure 13 Fixed constraint in GDA
For the structural load applications on the model, three load
cases are considered: one where the resultant force vector is
directed in the negative Z axis and two where the resultant force
vectors are oriented at a 45° angle on both sides relative to a
reference axis (X angle = −45° and X angle = 45°). This approach
to the load case definition offers a more symmetrically correct
outcome. The definition of the third load case is shown in Figure
14.
Figure 14 Third load case definition
The type of load being used is Remote Force, placed in the
point of the resulting force vector, as the surfaces of the pulley part
cannot be selected (are hidden and Obstacle Geometry has filled the
volume). Force magnitude stays the same, 1500 Newton’s.
INNOVATIONS 2025
84
After all the information for the generative design process has
been successfully acquired and approved, the option Pre-check is
selected to determine the behavior of the process. The next step is to
generate the outcomes based on the defined inputs by selecting the
option Generate.
4. Results and discussion
The generative design process has provided 10 converged
outcomes out of which 4 outcomes (Figure 15) have been selected
based on the numerical values for maximum von Mises stress,
safety factor, maximum displacement, mass and volume.
Figure 15 The four selected outcomes
The manufacturing possibility and methods have also been
taken into consideration while analyzing the selected outcomes. The
properties of the four individual outcomes are displayed in Figure
16.
Figure 16 The properties of the selected outcomes
Outcome 9, having the highest value of the equivalent stress
(56.96 MPa), is eliminated from the optimal selection, while the
other three outcomes have similar desired properties. Outcome 1 is
chosen as the optimal bracket model having the lowest numerical
value of the desired properties: the equivalent stress (41.25 MPa),
the safety factor (minimum 4) and the maximum displacement (0,27
mm).
After the 3D Design from Outcome model has been exported
from the Generative Design module, module Design has been
initialized to smoothen out the localized edges and areas where
stress concentration occurs. In this manner, the final results could
converge better (lower desired numerical values). The optimal
bracket model in Design module is shown in Figure 17.
Figure 17 The optimal bracket model in Design module
The final pulley support assembly is constructed in Autodesk
Inventor software and imported in Autodesk Fusion 360 software.
The final Static Stress simulation starts with the same load
conditions, constraints, materials, mesh and contacts analogue to the
description in Section 3.1. Tabular results of the final simulation are
given in Table 2.
Table 2 Final Static Stress simulation results
Load case 1
Load case 2
Minimum safety factor
4.012
4.268
Maximum equivalent
stress, MPa
62.317
58.581
Maximum displacement,
mm
0.094
0.098
Both load cases satisfy the minimum safety factor of 4, the
displacement values are almost the same while the values of the
equivalent stress are different from the initial results. The
comparative numerical results for the pulley support assembly of
the initial and the final simulation are given in Table 3.
Table 3 Comparative results of the initial and the final
simulation
Load case 1
Load case 2
Initial
Final
Initial
Final
Mass, kg
0.750
0.550
0.750
0.550
Volume, cm3
750
550.2
750
550.2
Minimum safety factor
4.555
4.012
3.776
4.268
Maximum equivalent
stress, MPa
54.886
62.317
66.202
58.581
Maximum
displacement, mm
0.105
0.094
0.098
0.098
The visual representation for Load case 1 is depicted in Figures
18, 19 and 20.
Figure 18 The initial (left) and the final (right) safety factor for
Load case 1
INNOVATIONS 2025
85
Figure 19 The initial (left) and the final (right) maximum
equivalent stress for Load case 1
Figure 20 The initial (left) and the final (right) maximum
displacement for Load case 1
From the previously presented figures and tables, it is evident
that the final equivalent stress for Load case 1 increased by
approximately 7 MPa but remained within permissible limits due to
the safety factor of 4.012, while for Load case 2, the stress
decreased by about 7 MPa. The displacement of the final assembly
for Load case 1 slightly decreased compared to the displacement of
the initial assembly, whereas the displacement of Load case 2
remained almost unchanged.
5. Conclusion
The comparative results from section 4 show the task fulfilment
for the minimum safety factor in a pulley support assembly
designed for lifting applications. The calculated equivalent stress
and displacement numerical values are both beneath the permissible
stress and displacement values.
The mass of the assembly has been reduced by approximately
200 grams or 26.67%. The initial mass of the bracket (391 grams)
has been reduced to the final mass of the bracket (192 grams) by
approximately 199 grams or 50.9%. The robustness of the assembly
has been maintained. Greater assembly mass reduction is possible
through the usage of differently modeled lightweight parts like the
shaft, the pulley and the sleeve.
Generative Design Approach in the product development cycle
has been successfully demonstrated. Further experimental
manufacturing and testing is possible. Through additional shaping
methods and optimal material implementation, optimal result values
and greater mass savings can be achieved, approaching the optimal
design of the pulley support.
6. References
[1]
A. Dieter, Building in Egypt: Pharaonic Stone Masonry,
Oxford University Press (1991)
[2]
I. Aslan Seyhan, Brief History of the Pulleys and
Explanation of Chief Instructor Ishaq Efendi’s Work on
Pulleys and Pulley Systems. In: Ceccarelli, M., López-
García, R. (eds) Explorations in the History and Heritage
of Machines and Mechanisms, History of Mechanism and
Machine Science, 40, 406-418 (2022)
[3]
J. Andreas Moseid, Mathematical Modelling of Cable and
Pulley Systems, master’s thesis, Norwegian University of
Science and Technology (2017)
[4]
G. Korwar, M. Bhelkar, A. Bhombe, International Journal
for Research in Applied Science & Engineering
Technology (IJRASET), Study of Generative Design
Using Autodesk Fusion 360, 12 (2024)
[5]
Environmental Health & Safety, Lifting safety
URL: https://www.ehsdb.com/lifting-safety.php
[6]
F. Buonamici, M. Carfagni, R. Furferi, Y. Volpe, L.
Governi, Generative Design: An Explorative Study,
Computer-Aided Design and Applications, 18 (2020)
INNOVATIONS 2025
86
Effect of organosilicate application on thermo-pressure bonding of metals and composites
with thermoplastic matrix
Anna Guzanová, Dagmar Draganovská, Nikita Veligotskyi1,*
Department of Technology, Materials and Computer Supported Production, Faculty of Mechanical Engineering,
Technical University of Košice, Slovakia
anna.guzanova@tuke.sk, dagmar.draganovska@tuke.sk, nikita.veligotskyi@tuke.sk
Abstract: The paper deals with application of organosilicate sol-gel to improve the adhesion of overlapped joints of metallic non-ferrous
alloy thin sheets and continuous fibre reinforced composites by heat and pressure. The organosilicate layer showed a significant increase in
adhesion of EN AW 6082 T6 and AZ91 alloys to glass fibre reinforced polypropylene matrix composite.
Keywords: METAL – COMPOSITE JOINING, THERMO-PRESSURE BONDING, ORGANOSILICATE
1. Introduction
The joining of metal and composite materials plays a key role
in modern lightweight automotive body design. This process allows
the mechanical properties of metals, such as strength and ductility,
to be combined with the low weight and specific properties of
composites, such as corrosion resistance or the ability to absorb
vibrations. This integration of materials allows automotive
engineers to optimise the body structure in terms of safety, energy
efficiency and durability, while also achieving reductions in overall
vehicle weight and emissions. [1 - 4]
Joints of metals and composites in the manufacture of
lightweight and durable vehicles have found their application in
various body parts, e.g. as:
- Hybrid frame construction: some vehicles use metal
frames combined with composite panels to achieve higher
strength and lower weight. These joints are often made by
bonding or mechanical joining.
- Doors and hoods: Composite materials, mainly carbon
fibre reinforced, are often combined with metal parts to
reduce the weight of the door or hood while maintaining
its strength and impact resistance.
- Floor panels: Some electric vehicles use composite floor
panels bonded to metal frames, improving stiffness while
reducing the overall weight of the vehicle.
Composite materials in terms of applicability in lightweight body
design will be either thermosoftening or thermoset matrix and in
terms of the reinforcing phase, long continuous fibres, which have a
much higher reinforcing effect compared to short fibres, will be
used in particular. [5 - 6]
High-strength fibres are the most suitable lightweight materials
for the transmission and absorption of forces. The directional
orientation of the fibres in the composite and their complete
consolidation with the thermoplastic polymer allows for high
structural strength solutions. Components can therefore be designed
with small wall thicknesses.
Thermoplastic composites are produced and processed in
solvent-free processes and allow for complete recycling cycles. In
this way, it contributes to a sustainable industry that is friendly to
climate change and uses natural raw materials efficiently.
Innovative thermosoftening prepregs made from highly flexible
composite materials, organosheets consist of continuous fibres in a
matrix of various engineering thermoplastics - polypropylene,
polyamide 6, polyamide 66, polyamide 12, polycarbonate,
thermoplastic polyurethane and polyphenylene sulphide. The
reinforcing fibres are glass, carbon, aramid or natural fibres such as
cotton, flax, jute and others. This combination of fibers and matrix
gives the flat organosheets excellent strength and stiffness
combined with extremely low weight. As a result, even complex
components can be manufactured efficiently. [7 - 9]
Depending on the material thickness of the combination of
fibers and thermoplastic polymers, composites can provide material
properties varying from high flexibility to high stiffness. Compared
to the individual materials of which they are composed,
thermoplastic composites achieve higher specific strain energy
absorption rates and are an ideal solution for applications that
require dynamic properties at reduced weight.
This article discusses the potential application of organosilicates
for thermoplastic joining of metals and composites with
thermosoftening matrix in order to improve the adhesive surface
properties of the materials. This is studied in terms of selected
parameters such as surface roughness of these joints, their load-
bearing capacity and dissipated energy at joint failure.
2. Materials and methods
The following materials were used for the research:
Metals:
• Aluminium alloy EN AW 6082 T6 (AlSi1MgMn), 1 mm thick
sheet (hereafter Al).
• Magnesium alloy AZ 91, 2 mm thick sheet (hereafter Mg).
Composites (Fig. 1):
• polypropylene matrix composite reinforced with glass fibre,
1.5 mm thick organosheet (hereafter GF)
• polypropylene matrix composite reinforced with carbon fibre,
1,5 mm thick organosheet (hereinafter CF)
The melting point of the polypropylene matrix is 165°C.
a) CF b) GF
Fig. 1 Appearance of composite samples
Methodology of joint formation
To determine the adhesive bond strength of the metal and
composite, adhesive bonds were formed by pressure and heat only,
without adhesive. The function of the adhesive in this application is
fulfilled by the thermosoftening matrix of the composite. A
schematic of this type of bond is shown in Fig. 2. The materials
(metal and composite) were overlapped and preheated to a
temperature >165°C and compressed with a force of 90 N. The
pressure of the preheated materials lasted for 2 minutes in the case
INNOVATIONS 2025
87
of the glass fibre composite and 2.5 minutes in the case of the
carbon fibre composite.
Fig. 2 Adhesive bond
To strengthen the adhesion of the joint, an experimental
organosilicate formulation was applied to the metallic materials (Al,
Mg) to improve the adhesion between organic and inorganic
materials. Using the formulation, it is assumed that the organic
functional groups in the organosilicate will form a strong chemical
bond with organic materials such as organic bonding agents in
paints, adhesives or polymeric materials on the one hand, and on the
other hand, by alkoxysilane linkages, they are able to bind firmly to
materials of an inorganic nature (metals, alloys, minerals).
The technological procedure of organosilicate coating deposition
was as follows: degreasing (alkaline degreaser, concentration 30g/l,
60°C, 10 min), rinsing, immersion in a solution of organosilicate
preparation in demineralized water (concentration 250ml/l, 20°C,
10 min), hot air drying.
SEM and EDX analysis
Surfaces with the above layer were analysed using SEM. EDX
analysis was performed to identify characteristic features on the
surfaces.
Methodology of the roughness measurement
Roughness measurements were carried out according to EN ISO
21920-2:2021 before and after the application of the organosilicate
layer with a Mitutoyo Surftest SJ-301 stylus profilometer
(Mitutoyo, Japan). The following parameters were measured: Ra
(arithmetical mean height of the assessed profile), Rz (maximum
height of profile in the sampling length), RSm (mean profile
element spacing) and the non-normalized parameter RPc (mean
number of peaks per centimeter of sampling length). Five
measurements of the above parameters were taken on each material,
from which the average value was then calculated. The
profilograms of each surface were also recorded with the Abbot-
Firestone material ratio curve of the profile.
Methodology of testing the load-bearing capacity of joints
The material combinations tested were Al-CF, Al-GF, Mg-CF, Mg-
GF, where the metal parts were in as-delivered condition, without
any treatment or degreasing as well as with a layer of organosilicate
(OS) to test the possible improvement of adhesion with the
composite.
The load bearing capacity of the formed adhesive bonded joints
was expressed in the form of maximum force at the moment of
failure of the joint Fmax [N]. Forty joints of each material
combination were formed and tested.
3. Results
EDX analysis of metal part surfaces
The surface of the Al material, its elemental EDX analysis in the
original state and after the organosilicate layer deposition are shown
in Fig. 3.
a)
b)
Fig. 3 Surface Al (a) in initial state, (b) with organosilicate layer
The surface of the Mg material, its elemental EDX analysis in
the initial condition and with the organosilicate layer are shown in
Fig.4.
On the surface with deposited organosilicate layer for both Al alloy
and Mg alloy, increased C and Si content appear in the EDX
spectrum as evidence of the presence of an organosilicate layer of
organic nature with organo-modified SiO2 nanoparticles on the
metal surface.
Surface roughness assessment
The resulting values of selected roughness parameters of metallic
materials Al and Mg, determined by the stylus profilometer, are
listed in Tab. 1 as the average of five measurements.
Tab. 1 Average roughness parameters of Al and Mg
Ra [µm]
Rz [µm]
RSm [µm]
RPc [-/cm]
Al
0.38
2.20
43.6
231.98
Al OS
0.47
2.62
51.00
199.66
Mg
0.11
0.89
99.40
101.70
Mg OS
0.20
1.47
140.60
73.64
Fig. 5 shows the profilographs of the Al and Mg surfaces in
original state and with the organosilicate layer applied.
a)
b)
Fig. 4 Mg surface (a) in initial state and (b) with a layer of organosilicate
INNOVATIONS 2025
88
a) Al
b) Al OS
c) Mg
d) Mg OS
Fig. 5 Surface profilographs
Both materials were originally relatively smooth and roughness
measurements confirmed this. The Mg surface showed less
roughness Ra and Rz than Al. Tab. 1 shows the changes in surface
roughness of each material when the organosilicate gel layer was
applied. The vertical parameters of the initial roughness of both
materials (Ra, Rz) increased slightly after the application of the
organosilicate layer. The mean width of the profile elements RSm
also increased slightly, analogously the number of peaks per
centimeter of measured length decreased. This indicates that the gel
layer of organosilicate filled in some of the valleys of the surface
profile.
Load-bearing capacity of adhesive bonded joints
Tab. 2 shows the load-bearing capacity of adhesive joints of all
material combinations, expressed in the form of the maximum force
at the moment of failure of the joint Fmax [N] as an average of 40
measurements (joints) and the average value of the dissipated
energy W [J] at the failure of the adhesive joints.
Tab. 2 Load-bearing capacity and energy dissipated by joints
Fmax
[N]
Standard
deviation
[-]
Coefficient
of variation
[%]
SW
p-value
[-]
W
[J]
Al-CF
94
30.28
32
0.2815
0.01
Al-CF OS
83
25.25
30
0.3779
0.01
Al-GF
127
47.89
38
0.1294
0.02
Al-GF OS
4524
763.85
17
0.1020
1.39
Mg-CF
5001
1601.90
32
0.3782
1.16
Mg -CF OS
1982
630.52
32
0.2126
0.38
Mg -GF
1360
641.40
47
0.0550
0.22
Mg -GF OS
2432
718.71
30
0.1542
0.44
The above results lead to the following findings:
- Al-CF and Al-CF OS joints are not satisfactory, they were
already breaking during clamping, even the application of
organosilicate did not bring the expected improvement.
- The Al-GF joints are also not satisfactory, but the application
of the organosilicate layer brought a significant improvement in
adhesion and therefore also in load-bearing capacity with an
average value of 4.5 kN (maximum up to 5.8 kN)
- The Mg-CF joints had the highest load capacity, up to 5 kN
(maximum up to 8.5 kN), however, the application of organosilicate
caused a drop in load capacity to 2 kN.
- The Mg-GF joints had a load capacity of 1.4 kN, with the
organosilicate layer the load-bearing capacity increased to 2.4 kN.
The load-bearing capacity of the adhesive joints shows
considerable variance within the tested set of 40 joints, despite the
fact that the joints were formed under identical conditions,
indicating the unstable nature of the adhesive bonding of the
materials. The higher the maximum Fmax values were in each
group of joints, the larger the variation range and standard deviation
were in that group. The heterogeneity of the data can also be
assessed by the coefficient of variation.
The coefficient of variation, also known as Pearson's coefficient
of variation, is a statistical characteristic that informs about the
relative dispersion of a set of data, that is, the degree of variability
of values as a percentage of the arithmetic mean. It is calculated by
dividing the standard deviation by the absolute value of the mean of
the set and is usually expressed as a percentage for better
understanding. If the coefficient of variation is <30%, we speak of a
good variability characteristic; if it is more than 50%, the set under
consideration is considerably heterogeneous. From this point of
view, it can be stated that, except for Al-GF OS, the variability is
significant for all other groups of joints, but does not exceed 50%
and thus the data are not heterogeneous and can be further
statistically processed.
Many statistical methods assume that the basic statistical
population has a normal distribution (many real data sets do not
have a normal distribution). If the normality assumption is not met,
statistical method cannot be used. Normality tests, such as the
Shapiro-Wilk test (SW test), are used to test whether a given
population of data can be considered normal distribution. The
Shapiro-Wilk normality test tests the validity of hypotheses:
H0: the data set has a normal distribution
H1: the data set does not have a normal distribution
The SW normality test was performed online, i.e. the data under
consideration were copied and pasted into a commonly available
online calculator. The result was a p-value, whereby the following
applies:
- if p < α, the null hypothesis is rejected at the α significance
level in favour of the alternative hypothesis,
- if p > α, the null hypothesis cannot be rejected.
In terms of p-value for the SW test, the results of the load-
bearing capacity of all types of adhesive joints show normality at
the significance level α=0.05.
Fig. 6 shows the appearance of the joint surfaces after failure.
Al-GF OS Mg-GF OS
INNOVATIONS 2025
89
Al-CF Mg-CF
Fig. 6 Appearance of adhesive joints after load-bearing capacity testing
From Fig. 6 and actual observation of the joints after failure is
clear that the failure of the joints was always adhesive. Only the
impressions of the fiber orientation in the composite remained
visible on the metal plates.
The results show that the organosilicate layer should only be
used where it has been shown to improve the adhesion of the joints,
i.e. in the case of joining both types of metal plates with the GF
composite. The GF composite, due to the thickness of the glass
fibres and their volume fill (47%,) contains a greater amount of free
PP matrix which, when the heated composite is compressed,
adheres better to the metal part. CF composite contains up to 51%
fibers, so it has less free PP matrix to form the bond.
Conclusions
Based on the results obtained, the following information was
found:
- The presence of organosilicate on the surface of both
materials investigated, Al alloy and Mg alloy, was
confirmed by EDX analysis by the increased C and Si
content with organo-modified nanoparticles.
- The roughness after the deposition of the organosilicate
layer on the surface of both materials reached an
increased value for the parameters studied. This fact gives
the assumption of an improved adhesion of the surfaces of
the materials and thus the load carrying capacity in the
application of the subsequently realized joints.
- Among the realized types of joints, the influence of the
applied organosilicate resulted in an increase of adhesion
in the case of joining metal plates with GF composite. For
joints with CF composite, the application of an
organosilicate layer is of no practical significance.
Acknowledgement:
This work was supported by The Ministry of Education,
Research, Development and Youth of the Slovak Republic under
Grant VEGA 1/0229/23: Research on the applicability of thermal
drilling technology for the creation of multi-material joints in the
automotive industry.
3. References
[1] GULLINO, Alessio – MATTEIS, Paolo – D´AIUTO, Faboi:
Review of Aluminim-To-Steel Welding Technologies for Car-
Body Applications. In: Metals 2019. Roč. 19, č. 9 (2019),
https://doi.org/10.3390/met9030315.
[2] HESSE, S.H. – LUKASZEWICZ, D.H.-J.A. – GUDDECK, F.:
A method to reduce design complexity of automotive composite
structures with respect to crashworthiness. In: Composite
Structures. Roč. 15, č. 129 (2015), s. 236-249,
https://doi.org/10.1016/j.compstruct.2015.02.086.
[3] CHEN, Dongdong – SUN, Guangyong – MENG, Maozhou –
JIN, Xihong – LI, Qing: Flexural performance and cost
efficiency of carbon/basalt/glass hybrid FRP composite
laminates. In: Thin-Walled Structures. Roč. 19, č. 142 (2019), s.
516-531, https://doi.org/10.1016/j.tws.2019.03.056
[4] SONG, Yang – CAO, Huajun – ZHENG, Wei – QU, Da – LIU,
Lei – YAN, Chunping: Cutting force modeling of machining
carbon fiber reinforced polymer (CFRP) composites: A review.
In: Composite Structures. Roč. 22, č. 199 (2022),
https://doi.org/10.1016/j.compstruct.2022.116096.
[5] GORING, Mandy – SCHREITER, Katja – SCHUBERTH,
Alexandra – WINBERG, Tina – JUNG, Heike – ANDERS,
Susann – MULLER, Philipp – NICKEL, Daniela – NESTLER,
Daisy – KROLL, Lothar – WIELAGE, Bernhard – LAMPKE,
Thomas – SPANGE, Stefan: Amino Group Bearing Organic-
Inorganic Hybrid Materials for Joining Aluminium Alloys and
Thermoplastic Fiber-Reinforced Parts. In: ADVANCED
MATERIALS INTERFACES. Roč. 17, č. 4 (2017), s. 1-12,
DOI: 10.1002/admi.201601115.
[6] CÁRCEL, B. – LÓPEZ, S. – CÁRCEL, A.C. – GERARD, P.:
Hybrid laser joining of RTM acrylic glass fiber composite with
stainless steel. In: Procedia CIRP. Roč. 20, č. 94 (2020), s. 526-
530, https://doi.org/10.1016/j.procir.2020.09.176
[7] HYNES, N. Rajesh Jesudoss – PRABHU, M. Vivek –
NAGARAJ, P.: Joining of hybrid AA6063-6SiCp-3Grp
composite and AISI 1030 steel by friction welding. In: Defence
Technology. Roč. 17, č 13 (2017), s. 338-345,
http://dx.doi.org/10.1016/j.dt.2017.05.014.
[8] ENVALIOR: Tepex – processing Guidelines [online].
Germany, 2024. [cit. 2024-01-25]. Dostupné na internete:
https://www.envalior.com/content/dam/envalior/en/documents/t
epex-processing-guidelines-en.pdf
[9] ENVALIOR: Tepex – Advanced Thermoplastic Composites
[online]. Germany, 2024. [cit. 2024-03-12]. Dostupné na
internete:
https://www.envalior.com/content/dam/envalior/en/documents/t
epex-brochure-en.pdf
INNOVATIONS 2025
90
Detection of fibre continuity in joints of metallic and composite thin-walled materials
formed by thermal drilling technology by computed tomography
Anna Guzanová1,*, Nikita Veligotskyi1, Teodor Tóth2
Technical University of Košice, Faculty of Mechanical Engineering, Slovakia
Department of Technology, Materials and Computer Supported Production1
Department of Biomedical Engineering and Measurement2
anna.guzanova@tuke.sk
Abstract: This paper presents the results of scanning fiber continuity in the joints of thin sheets made of aluminum and magnesium alloys
with bi-directionally glass and carbon fiber-reinforced composites formed by thermal drilling technology using computed tomography. The
scans show the progression of the bushing forming and the closure of the fibres in the hem joint. Little delamination, but no fiber fracture are
visible in metal-composite joints.
Keywords: METAL-COMPOSITE JOINING, THERMAL DRILLING, FIBER FRACTURE, COMPUTED TOMOGRAPHY
1. Introduction
The joining of metals and composites in the form of thin-walled
materials is a current subject of research [1-4]. The need to join
such dissimilar materials arises from the needs of the automotive
and aerospace industries, where the combination of ultra-
lightweight materials and the production of tailored parts is
required. The largest group of composite materials suitable for
automotive applications are polymer composites with a
thermosoftening or thermosetting matrix. Thermoset composites
provide higher mechanical properties and can be bonded to other
materials by means of press-fit fasteners, since once cured, a change
in their shape and possible further bonding without breaking the
fibres is no longer possible. Thermosoftening matrix composites
can be purchased in the form of organosheets formed by
consolidating several layers of prepregs to achieve the desired panel
thickness. The forming of such organosheets, heated above the
melting temperature of the polymer matrix, can be carried out using
a pair of flexible membranes, a rubber die or in conventional metal
split moulds. Composite organosheets can also be processed by
combined technologies, short or long fibre plastics with the same
type of matrix can be injection moulded to form reinforcing or
stabilising ribs, force transfer elements, functional elements or
additional contours on the edges of the component, Fig. 1.
Fig. 1 Rib structure and edge overmolding
By choosing appropriate matrices and process parameters, a
homogeneous connection between two different cross-sections is
achieved. In particular, the temperature of the two materials must be
high enough to form a reliable fusion bond. For a good fusion of the
materials, the highest possible injection speed is preferred (because
at high injection speeds there are high shear stresses in the injected
mass and thus too rapid cooling is prevented), especially in areas
away from the inlet, and also a high clamping force helps to fuse the
materials. The joining of polymers and polymer composites with
compatible matrices can also be achieved by welding. Welding is
mainly used for joining thermosoftening composites, bonding and
mechanical joining can be used for joining different material
combinations, even plastics with metals.
In addition to the aforementioned technologies, the issue of
joining composites with other materials, especially metals, is also a
challenge. Composite components are often part of complex
assemblies that, in extreme cases, combine a variety of materials
including steel, lightweight metals such as aluminum or magnesium
alloys, polymers reinforced with short or long glass fibers, or
carbon fiber composites.
According to the physical principle, joining processes can be
divided into:
- homogeneous joints (welded, bonded)
- non-homogeneous joints (mechanical joints)
o non-positive joints (screw, pressure, riveted joints)
o positive joints (latches, locks, handles)
According to their demountability they are divided into:
- demountable joints (using screws, pins, wedges)
- non-demountable joints (bonded, welded, riveted).
Typical mechanical joining techniques, such as bolting and
riveting, cause the fibres in the composite to break when making
metal-to-composite joints, which significantly reduces the load-
bearing capacity of the joint. This may underutilize the high
potential of lightweight composite components in highly loaded
multi-material systems. In addition, often multi-step manual
operations, such as pre-drilling of the parts to be joined and
subsequent insertion of additional fasteners, represent time-
consuming and costly preparatory work in the joining process [5-9].
Flowdrill technology is used in joining thin-walled materials, it
uses frictional heat between the tool and the material, which softens
the material and deforms it plastically by the movement of the tool
[4]. The plasticized material is displaced below the level of the
material being drilled, forming a bushing into which sufficient
threads can be created to make a mechanical connection to another
material with a screw. The bushing is very advantageous when
joining thin workpieces because it locally increases the thickness of
the material.
The aim of the work is to experimentally verify the possibility
of using the side effect of thermal drilling technology (flowdrill) for
joining two overlapped thin-walled materials (metal-composite)
without using a screw, just by forming a metal bushing penetrating
through the fiber composite without breaking the integrity of the
reinforcing fibers. The integrity of the fibres is analysed by
computed tomography (CT).
Materials and methods
The joints were made of the following materials:
Aluminium alloy EN AW-6082 T6 (AlSi1MgMn) – Al alloy
supplied in a precipitation-hardened state with relatively high
mechanical properties. The alloy was supplied in the form of a 1
mm thick rolled sheet. Hereinafter Al.
AZ 91 magnesium alloy with the main alloying elements Al and
Zn. The alloy was supplied in the form of a 2 mm thick rolled
sheet. Hereinfter Mg.
Composite with polypropylene (PP) matrix reinforced with
continuous glass fibres in two perpendicular directions. The
thickness of the consolidated composite organosheet is 1.55 mm,
which is the result of the compaction of three layers of prepregs.
INNOVATIONS 2025
91
Hereinfter GF.
Composite with polypropylene matrix (PP) reinforced with
continuous carbon fibres in two perpendicular directions. The
thickness of the consolidated composite orgaanosheet is 1.55 mm,
the result of the compaction of seven layers of prepregs.
Hereinfter CF.
The basic characteristics of the materials are shown in Tab. 1-3.
Tab. 1 Chemical composition of alloys in wt.%
alloy
Si
Mg
Mn
Fe
Zn
Cu
Al
Al
1.0
0.7
0.44
0.4
0.08
0.06
balance
Mg
0.09
balance
0.14
0.004
0.93
0.02
8.9
Tab. 2 Mechanical and physical properties of alloys
alloy
Re [MPa]
Rm [MPa]
A50 [%]
density [g·cm-3]
Al
295
340
15
2.71
Mg
280
200
24
1.81
Tab. 3 Properties of composite organosheets
Composite
PP CF50 T200
OS
PP GF45 T600
OS
Matrix
polypropylene
Reinforcement
Carbon fibre
E Glass fibre
Weaving
Twill 2/2
Twill 2/2
Surface mass. mats
200 g·m-2
600 g·m-2
Density of weaving
3K
1200 tex
Fibre content in the
composite
51%
47%
Weight of finished product
301 g·m-2
887 g·m-2
Prepreg thickness
0.22 mm
0.5 mm
Specific weight of composite
1.46 g·cm-3
1.68 g·cm-3
Melting temperature
165C
165C
CTE (23-80°C)
3.2×10-6K-1
11×10-6K-1
The philosophy of joint formation is that the tool, as it feeds
into the material, comes into contact with the metal sheet, begins to
penetrate it by means of friction-assisted rotation, whereby an
opening is created and the material from the opening is transformed
into a bushing. As it is formed, the bushing penetrates through the
heated composite, deflecting its fibres off the axis of the opening
without breaking it. This creates a mechanical interlocking of the
two materials. However, when conducting the experiments, we
divided the process into two steps - first, a hole was formed in the
preheated composite, and in the second step, a metal sheet was
placed on the composite and the drilling was carried out once more.
a) direct drilling
b) hemming flange of the joint
Fig. 2 Joining of metals and composites by thermal drilling
This sequential joining method was designed to facilitate the
shaping of the metal bushing in the prepared hole in the composite
plate. The joint was then further secured against opening from the
opposite side with a hem flange, Fig. 2. The above materials were
then used to form the joints with the metal plate in the top position
when drilled and the composite in the bottom position. The material
combinations of the joints formed were Al-CF, Al-GF, Mg-CF, Mg-
GF.
The resulting metal bushing has an outer surface that is conical
and an inner surface that is cylindrical in shape, Fig. 3 top.
However, the outer conical shape of the bushing can cause the
composite to slide away from the metal sheet when the joint is
loaded, causing the joint to open and its load carrying capacity to
depend on the thickness of the hem flange instead of the thickness
of the bushing. Therefore, a geometric modification of the bushing
was proposed by widening the bushing from the flange side with a
larger diameter tool (Flowdrill Long 9.3 mm), Fig. 3 bottom. This
results in an inverted bushing and the elimination of the tangential
component of the force that causes the joint to open.
Fig. 3 Design of the bushing geometry modification to increase the load
carrying capacity of the connections (F - loading force, Fn - normal, Ft -
tangential component of the force F)
After the joints were made, the entire area was scanned using
CT. The fiber distribution around the opening in the composite as
well as around the bushing in joints was visualized by Carl Zeiss
Metrotom 1500 Gen1. The scanning parameters were different
when scanning the composites alone and when scanning the metal-
composite joints. The scanning parameters for the composites were:
voltage 100 kV, current 230 µA, integration time 1000 ms,
resolution 27.04 µm, no filter applied. Scanning parameters for
metal-composite joints: voltage 120 kV, current 250 µA, integration
time 1000 ms, resolution 40,29 µm, Cu filter 0,5 mm.
Results
First, the prepared openings in the composite were scanned,
which is actually the first step in sequential joining, Figs. 4 and 5.
INNOVATIONS 2025
92
Fig. 41 Fibre distribution around the opening in the GF composite
Fig. 5 Fibre distribution around the opening in the CF composite
From CT scans of the openings made with the Flowdrill 5.3 mm
tool in the GF and CF composite, fiber displacement is visible.
Heating the composite to temperatures above 165°C caused the
polypropylene matrix to melt and allowed the fibres to be displaced
from their position to a greater distance from the opening, up to 22-
25 mm in diameter. Fiber breakage is not detected. The opening in
the preheated composite, created by the 5.3 mm diameter tool,
shrank to 4.6 mm after the tool returned to its initial position due to
the tensile stress in the fibres. However, the opening is still large
enough to facilitate the forming of the metal bushing after
subsequent overlapping with metal sheet. It is also clear from the
scans that the glass fibers are thicker, as are the bundles that make
up the glass fabric. The carbon fibers are thinner, the fiber bundles
are thus also thinner, and the weaving density of the fiber mat is
greater. The higher weaving density may resist the penetration of
the forming metal bushing. However, this need not be negative. It
can help in the forming of the Al bushing in particular, which would
not have a suitable shape without the resistance of the fibers. We
consider a slightly conical bushing with a sharp end to be a suitable
bushing shape for joining purposes.
CT scans of the joints of Al and Mg plates with GF and CF
composites are shown in Figs. 6-9.
Fig. 6 CT scan of Al-GF joint
The scan (Fig. 6) shows the displacement of the glass fibers during
the formation of the bushing. Significant fiber breakage is not
visible. The hem prevents delamination of the composite. The
bushing has the shape of an inverted cone as a result of modifying
the joint opening from the opposite side with a 9.3 mm tool.
Fig. 7 CT scan of Al-CF joint
The Al hem (Fig. 7) is well formed, trapping and fixing the fibres
around the joint opening, thus preventing significant composite
delamination. There is a visible fiber rich zone around the bushing,
indicating fiber displacement during thermal drilling and fiber
accumulation at the periphery of the bushing.
INNOVATIONS 2025
93
Fig. 8 CT scan of Mg-GF joint
At the Mg-CF joint (Fig. 8), the importance of modifying the
geometry of the bushing is more apparent. The Mg bushing is
cylindrical outside and conical inside, which eliminates the
composite from sliding off the bushing when the joint is subjected
to load and could lead to desirable bushing failure by shear. The
hem is relatively small and so there was local pulling of the fibers
out of the composite.
Fig. 9 CT scan of Mg-CF joint
The Mg-CF joint (Fig. 9) again exhibits the desired inverted
cone geometry. Widening the hem with the 9.3 mm tool leads to a
reduction of the hem, as a result of which a slight pulling out of the
fibers from the composite around the joint opening is again evident.
Conclusion
Computed tomography is an essential non-destructive technique
that makes it possible to monitor the behaviour of a composite
around a joint with a metal sheet formed by thermal drilling
technology. CT confirmed the displacement of the fibers out of
position during penetration of the flowdrill tool without breaking
the fibers while maintaining their reinforcing effect. Due to the
undisturbed continuity of the fibres and also due to the geometrical
modification of the shape of the resulting bushing, the mechanical
properties of the two materials involved in the joint can be
effectively used for increasing the load carrying capacity of the
joint.
Acknowledgement:
This work was supported by The Ministry of Education,
Research, Development and Youth of the Slovak Republic under
Grant VEGA 1/0229/23: Research on the applicability of thermal
drilling technology for the creation of multi-material joints in the
automotive industry, and VEGA 1/0191/24: Development,
optimization, and application of coordinate measurement strategies
of geometric parameters and structure of additive products.
References
[1] FEKETE, J.R. – HALL, J.N.: 1 – Design of auto body:
Materials perspective. In: Automotive Steels. Roč. 17, s. 1-18,
https://doi.org/10.1016/B978-0-08-100638-2.00001-8.
[2] LIEDL, G. – BIELAK, R. – IVANOVA, J. – ENZINGER, N.
– FIGNER, G. – BRUCKNER, J. – PASIC, H. – PUDAR, M.
– HAMPEL, S.: Joining of Aluminum and Steel in Car Body
Manufacturing. In: Physics Procedia. Roč. 11, č. 12 (2011), s.
150-156, https://doi.org/10.1016/j.phpro.2011.03.019.
[3] MORI, K. – ABE, Y.: A review on mechanical joining of
aluminium and high strength steel sheets by plastic
deformation. In: International Journal of Lightweight
Materials and Manufacture. Roč. 18, č. 1 (2018), s. 1-11,
https://doi.org/10.1016/j.ijlmm.2018.02.002.
[4] GUZANOVÁ, A. – JANOŠKO, E. – DRAGANOVSKÁ, D. –
VRABEĽ, M. – TOMÁŠ, M. – HORŇAK, P. – VOJTKO, M.
– VELIGOTSKYI, N.: Investigation of Applicability
Flowdrill Technology for Joining Thin-Walled Metal Sheets.
In: Metals. Roč. 22, č. 4 (2022), s. 1-23,
https://doi.org/10.3390/met12040540.
[5] TROSCHITZ, J. – KUPFER, R. – GUDE, M.: Process –
integrated embedding of metal insert in continuous fibre
reinforced thermoplastics. In: Procedia CIPR. Roč. 19, č. 85
(2019), s. 84-89, https://doi.org/10.1016/j.procir.2019.09.039.
[6] SEIDLITZ, H. – ULKE-WINTER, L. – KROLL, L.: New
Joining Technology for Optimized Metal/Composite
Assemblies. In: Journal of Engineering. Roč. 14, s. 1-11,
http://dx.doi.org/10.1155/2014/958501.
[7] KOHLER, D. – POPP, J. – KUPFER, R. – TROSCHITZ, J. –
DRUMMER, D. – GUDE, M.: In-Situ Computed
Tomography – Analysis of a Single-Lap Shear Test with
Composite-Metal Pin Joints. In: Journal of Physics:
Conference Series. Roč. 23, č. 2526 (2023), s. 1-7,
doi:10.1088/1742-6596/2526/1/012067.
[8] TROSCHITZ, J. - FÜẞEL, R. – KUPFER, R. – GUDE, M.:
Damage Analysis of Thermoplastic Composites with
Embedded Metal Inserts Using In Situ Computed
Tomography. In: Journal of Composites Science. Roč. 22, č. 6
(2022), s. 1-9, https://doi.org/10.3390/jcs6100287.
[9] GROGER, B. – KOHLER, D. – VORDERBRUGGEN, J. –
TROSCHITZ, J. – KUPFER, R. – MESCHUT, G. – GUDE,
M.: Computed tomography inverstigation of the material
structure in clinch joints in aluminium fibre-reinforced
thermoplastic sheets. In: Production Engineering. Roč. 22, č.
16 (2022), s. 203-212, https://doi.org/10.1007/s11740-021-
01091-x.
INNOVATIONS 2025
94
Statistical analysis of modification of joints of metals and composites by thermal drilling
Anna Guzanová1,*, Nikita Veligotskyi1, Gabriela Ižaríková2
Technical University of Košice, Faculty of Mechanical Engineering, Slovakia
Department of Technology, Materials and Computer Supported Production1
Department of Applied Mathematics and Informatics2
anna.guzanova@tuke.sk
Abstract: The paper presents an innovative method for joining Al alloy sheets and polymer fibre-reinforced composites by thermal drilling. A
proposal for modifying the joint geometry by reverse drilling with a larger diameter tool is presented. The proposed modification
demonstrated a statistically significant increase in the failure energy of the resulting joints.
Keywords: METAL-COMPOSITE JOINING, THERMAL DRILLING, BUSHING GEOMETRY, REVERSE DRILLING, STATISTICAL
ANALYSIS, p-VALUE
1. Introduction
Thermal drilling represents a technique for joining dissimilar
materials through the formation of bushings, even without fastener.
In recent years, significant progress has been made in addressing
the challenges associated with joining materials possessing
fundamentally different physical and mechanical properties, such as
thin-walled metallic and fiber reinforced composite substrates [1–
7]. These hybrid material configurations are increasingly utilized in
the design of ultralight structures, particularly within the automotive
and aerospace sectors. Typical applications of metal–composite
joints include structural elements such as A-pillars, engine covers,
load-bearing plates, and front stretchers. Joints between thin-walled
metal sheets and composite plates with a thermoplastic matrix are
characterized by the preservation of fiber continuity, thereby
maintaining their reinforcing function. To ensure this, the composite
material must be heated above the melting temperature of the
thermoplastic matrix during the joining process. Common
thermoplastics used for the matrix include polypropylene,
polyamide 6, polyamide 66, polyamide 12, polycarbonate,
thermoplastic polyurethane, and polyphenylene sulfide. During the
heating and formation of the metal bushing, the fibers within the
composite are deflected, allowing the bushing to form without
damage to the fibers.
The joint between the metal sheet and the composite, created
through thermal drilling, may appear as shown in Fig. 1.
Fig. 1 Schematic representation of a possible joint failure
As seen in Fig. 1, the joint is mechanical; however, due to the
heating of the polymer matrix and the mutual compression of the
materials during the joining process, adhesive bonding also occurs
between the materials. The end of the bushing is used for hemming
flange the joint, which prevents it from opening under stress. The
outer surface of the bushing is conical, while the inner surface is
cylindrical. Fig. 1 also highlights the potential modes of failure of
the joint under tensile loading. In an ideal case, the joint should fail
by shear of the bushing, but practical experience has shown that due
to the conical shape of the bushing, failure occurs more often by the
composite slipping off the conical bushing, straightening of the
hem, or shearing of the hem, leading to the composite being pulled
out of the bushing. As a result, the materials separate without fully
utilizing the material properties of the metal (bushing), and the joint
fails.
The aim of this paper is to present the proposed change in the
geometry of the bushing and its effect on the effective use of the
material properties of the metal parts of the joint.
2. Materials and methods
The metal sheet used for the joint formation is made of EN AW
6082 T6 alloy with a thickness of 1 mm. This is a precipitation-
hardened aluminum alloy, commonly used in the manufacturing of
lightweight structures in the automotive industry or construction.
For the composite, polypropylene with a melting temperature of
165°C, reinforced with glass fibers (GF) and carbon fibers (CF),
was used. The fibers are continuous and laid in two mutually
perpendicular directions in the form of a mat. The thickness of the
composite panels was 1.5 mm, with the thickness of the glass fiber
panels achieved by consolidating three layers of prepreg, while the
carbon fiber panels consisted of seven layers.
The joint was created by thermal drilling using a Flowdrill Long
tool ø5.3 mm, with the following process parameters: RPM 4800
min⁻¹, feed rate 60 mm·min⁻¹. During drilling, the materials (Al-
CF, Al-GF) were overlapped on a length of 30 mm, and the
composite was heated to 175°C, as shown in Fig. 2. Forty joints
were produced for each material combination to allow for a detailed
statistical analysis of the results.
Fig. 2 Principle of joint formation by thermal drilling
Load-displacement dependence was recorded when testing the
load-carrying capacity of the connections.
Results
Selected load-displacement curves for Al-CF and Al-GF joints are
shown in Fig. 3.
a) Al-CF
0
1
2
3
4
5
6
7
0 0,5 1 1,5 2 2,5 3
load [kN]
displacement [mm]
INNOVATIONS 2025
95
b) Al-GF
Fig. 3 Load-displacement curves when testing load-bearing capacity of
joints
In Al-CF joint, there was no adhesive bonding, only mechanical
joining through the bushing. In the case of Al-GF joints, the curve
first shows an increase in force up to Fmax, when the adhesive part
of the joint was broken and the force dropped sharply. At this point,
the load was transferred to the joint bushing which deformed and
the composite was gradually pulled out of the joint, Fig. 4.
a) bushing deformation, straightening of the hem
b) hem shear
Fig. 4 Mechanism of bushing failure in joints
To prevent the composite from sliding out of the bushing, a
geometric modification of the bushing was proposed by forming an
inverted cone with a tool of larger diameter (9.3 mm) than the
original bore (5.3 mm), which would eliminate the tangential
component of the force that leads to the composite sliding out of the
bushing, Fig. 5. The process of modifying the geometry of the
bushing will be referred to as reverse drilling (RD) in this context.
Fig. 5 Design of the bushing geometry modification to increase the load
carrying capacity of the connections (F - loading force, Fn - normal, Ft -
tangential component of the force F)
In reverse drilling the shape of the bushing changes to the opposite -
the outer shape is cylindrical, the inner conical, Fig. 5 and 6.
a) Al-GF, thermal drilling and hemming flange
b) Al-GF, thermal drilling, hemming flange, RD
Fig. 6 Metallographic section of Al-GF joint
The joints, modified by reverse drilling (RD) were also stressed in
tension, Fig. 7.
a) Al-CF RD
0
1
2
3
4
5
6
7
0 0,5 1 1,5 2 2,5 3
load [kN]
displacement [mm]
0
1
2
3
4
5
6
7
0 0,5 1 1,5 2 2,5 3
load [kN]
displacement [mm]
INNOVATIONS 2025
96
b) Al-GF RD
Fig. 7 Load-displacement curves when testing load-bearing capacity of
joints with improved bushing geometry by reverse drilling
At first glance, the load-displacement curves appear very
similar, which is why it was necessary to analyze the resulting
dependencies in more detail using statistical methods. The load-
displacement relationships also serve as the basis for calculating
another indicator, beyond just the maximum force Fmax, namely
the energy W dissipated at the failure of the joint. It is desirable for
the joint to absorb as much energy W (in joules) as possible during
failure, thereby contributing to the overall impact absorption during
a vehicle collision in the vehicle’s deformation zones and joints.The
dissipated energy is calculated using the following formula:
W = Fds
s max
0Fsis
n
i=1 =sF1s1+ F2s2++
Fnsn
where Fi is the instantaneous value of the load in N and si is the
instantaneous value of the displacement in m.
Additionally, it can be seen that the load-displacement curve can be
divided into two parts – the adhesive and mechanical part of joint,
Fig. 8, and the dissipated energy for both components of the joint
can be calculated separately, as shown in Table 1.
Fig. 8 Typical load-displacement curve for a hybrid, adhesion-mechanical
joint
Tab. 1 Energy dissipated in the particular areas of the hybrid joints
Joint
Adhesive part
Mechanical part
Total energy
Change in
energy
[J]
[%]
[J]
[%]
[J]
[%]
[%]
Al-CF
0.00
0.00
1.07
100.00
1.07
100.00
+37.33
Al-CF RD
0.00
0.00
1.47
100.00
1.47
100.00
Al-GF
1.81
79.23
0.48
20.77
2.29
100.00
+27.96
Al-GF RD
2.13
71.66
0.80
28.34
2.93
100.00
As seen in Table 34, modifying the bushing geometry resulted
in an increase in the total energy consumed during failure for all
joint types, as well as an increase in the partial energies in the
adhesive and mechanical parts of the joint. The failure mode of the
joint also changed, from bushing deformation to bushing shear (Fig.
9), which is desirable for the efficient utilization of the material
properties of the aluminum alloy.
a) Al-GF RD, partial shear of the bushing (approx. 1/3 of the
circumference), the bushing remains part of the Al plate
b) Al-CF RD, partial shear of the bushing (approx. 2/3 of the
circumference), the bushing remains wedged in the CF
c) Al-CF RD, total shear of the bushing, the bushing remains wedged
in the CF
Fig. 9 Mechanism of bushing failure in Al-composite hybrid joints with
bushing geometry modification by reverse drilling
The percentage representation of the failure modes of the joints
before and after the geometry modification through reverse drilling,
based on testing 40 joints in each group, is shown in Fig. 10.
Thermal drilling
Thermal drilling + RD
Fig. 10 Change in the joint failure mode (blue - deformation of the bushing,
pulling out of the composite, green - shearing of the bushing)
By modifying the bushing geometry, the proportion of joints
failed by bushing shear significantly increased, which was the
objective of the proposed bushing modification.
0
1
2
3
4
5
6
7
0 0,5 1 1,5 2 2,5 3
load [kN]
displacement [mm]
INNOVATIONS 2025
97
To compare the data sets (maximum force Fmax, energy W
consumed during joint failure), statistical tests were used. The
normality condition was verified using the Shapiro-Wilk test (SW)
to determine whether it is appropriate to use a parametric or non-
parametric test. The Shapiro-Wilk normality test evaluates the
validity of the following hypotheses:
H0: the data set has a normal distribution
H1: the data set does not have a normal distribution
Subsequently, Fisher's F-test was used to test the hypothesis
regarding the equality of variances of two independent data sets.
The hypotheses tested were:
H0: the two independent data sets have equal variance
H1: the two independent data sets do not have equal variance
Based on the results of the F-test, the Student's t-test for two
independent samples was used for the comparison, either assuming
equal variances or unequal variances. This test evaluates the
hypothesis about the difference in the means of two data groups and
is used to determine whether the observed difference in means from
the samples is due to chance or is statistically significant.
Tested hypotheses:
H0: the difference in the means of two independent data sets is not
statistically significant
H1: the difference in the means of two independent data sets is
statistically significant
There are various ways to test the hypotheses H0 against H1 at the
significance level α (in our case α=0.05), and although these
methods may differ, the conclusion remains the same. The
acceptance or rejection of the null hypothesis is based on the p-
value as follows:
if p<α, the null hypothesis H0 is rejected at the significance
level α in favor of the alternative hypothesis H1,
if p>α, the null hypothesis H0 cannot be rejected.
The statistical tests mentioned above were used to compare the
different categories of joints created by thermal drilling and reverse
drilling with modified bushing geometry, both in terms of the
achieved maximum force Fmax during joint load testing and in
terms of the energy W dissipated during joint failure, as shown in
Tab. 2 and 3.
Tab. 2 Statistical comparison of Al-composite joints in terms of Fmax
Joints
Al-CF
Al-CF RD
Al-GF
Al-GF RD
SW test
p-value
0.2815
0.1619
0.2530
0.0547
conclusion
p>α
normal
distribution,
parametric
tests
p>α
normal
distribution,
parametric
tests
p>α
normal
distribution,
parametric
tests
p>α
normal
distribution,
parametric
tests
F-test
p-value
0.000621184
0.041717792
conclusion
p<α
data sets do not have equal
variance,
unequal variances T-test
p<α
data sets do not have equal
variance,
unequal variances T-test
t-test
p-value
0.062563566
0.074745448
conclusion
p>α
The difference in means is not
statistically significant
p>α
The difference in means is
not statistically significant
Tab. 3 Statistical comparison of Al-composite joints in terms of W
Joints
Al-CF
Al-CF RD
Al-GF
Al-GF RD
SW test
p-value
0.8878
0.548
0.9638
0.9547
conclusion
p>α
normal
p>α
normal
p>α
normal
p>α
normal
distribution,
parametric
tests
distribution,
parametric
tests
distribution,
parametric
tests
distribution,
parametric
tests
F-test
p-value
0.0592
0.5301
conclusion
p>α
data sets have equal variance,
equal variance T-Test
p>α
data sets have equal variance,
equal variance T-Test
t-test
p-value
0.0000
0.000373
conclusion
p<α
The difference in means is
statistically significant
p<α
The difference in means is
statistically significant
The statistical tests confirmed that, although the maximum force
at joint failure Fmax does not change statistically significantly with
the modification of the bushing geometry, the increase in the total
energy W dissipated during the failure of joints with modified
bushing geometry—across all tested material combinations—is not
random, but statistically significant at the 0.05 significance level.
Conclusion
Experimental and statistical tools have confirmed that by
modifying the tool geometry through reverse drilling with a larger
diameter tool, it is possible to achieve a more efficient utilization of
the mechanical properties of metallic parts when joining metals and
composites using the thermal drilling technology.
Acknowledgement: This work was supported by The Ministry of
Education, Research, Development and Youth of the Slovak
Republic under Grant VEGA 1/0229/23: Research on the
applicability of thermal drilling technology for the creation of
multi-material joints in the automotive industry.
References
[1]. XIN, Z. – LIU, L. – CHEN, T. – WU, L. – CHEN, K. – KONG,
L. – WANG, M.: Laser surface treatment to enhance the
adhesive bonding between steel and CFRP: Effect of laser spot
overlapping and pulse fluence. In: Optics & Laser Technology
22, 59 (2023), 1-12, 10.1016/j.optlastec.2022.109002.
[2]. LAMBIASE, F. – KO, D.: Two-steps clinching of aluminum
and Carbon Fiber Reinforced Polymer sheets. In: Composite
Structures 17, 164 (2017), 180-188,
https://doi.org/10.1016/j.compstruct.2016.12.072.
[3]. GUZANOVÁ, A. – JANOŠKO, E. – VELIGOTSKYI, N.:
Optimization of joining parameters of thin-walled materials by
flowdrill technology. In: MACHINES. TECHNOLOGIES.
MATERIALS 22, 5 (2022), 176-178, ISSN 1313-0226.
[4]. GUZANOVÁ, A. – JANOŠKO, E.: Application of flow-drill
technology for joining metal materials. In: INNOVATIONS 21,
3 (2021), 108-111, ISSN 2603-3763.
[5]. SEIDLITZ, H. – ULKE-WINTER, L. – KROLL, L.: New
Joining Technology for Optimized Metal/Composite
Assemblies. In: Journal of Engineering 14, 1-11,
http://dx.doi.org/10.1155/2014/958501.
[6]. TROSCHITZ, J. - FÜẞEL, R. – KUPFER, R. – GUDE, M.:
Damage Analysis of Thermoplastic Composites with Embedded
Metal Inserts Using In Situ Computed Tomography. In: Journal
of Composites Science 22, 6 (2022), 1-9,
https://doi.org/10.3390/jcs6100287.
[7]. GROGER, B. et al.: Computed tomography inverstigation of the
material structure in clinch joints in aluminium fibre-reinforced
thermoplastic sheets. In: Production Engineering 22, 16 (2022),
203-212, https://doi.org/10.1007/s11740-021-01091-x.
INNOVATIONS 2025
98
Effect of thermal drilling strategy on the geometrical characteristics of metal – composite
joints
Anna Guzanová*, Nikita Veligotskyi
Technical University of Košice, Faculty of Mechanical Engineering,
Department of Technology, Materials and Computer Supported Production, Slovakia
anna.guzanova@tuke.sk
Abstract: This paper deals with the change of joint geometry of non-ferrous metal sheets and composite plates reinforced with bi-directional
glass and carbon fiber by thermal drilling due to the effect of different drilling strategy. Joints formed by direct and sequential drilling were
tested. Sequential drilling represents a convenient way to minimize delamination of layered composites during joining.
Keywords: METAL-COMPOSITE JOINING, THERMAL DRILLING, DRILLING STRATEGY, DIRECT DRILLING, SEQUENTIAL
DRILLING, BUSHING GEOMETRY
1. Introduction
Thermal drilling is one way of joining materials through
bushing forming. Nowadays, the challenges of joining materials
with completely different properties such as metallic and composite
thin-walled materials are being overcome [1-6]. Such material
combinations occur in the construction of ultralight structures in the
automotive or aerospace industries. Joints of metals and composites
are found, for example, in the construction of A-pillars, engine
covers, load plates or front stretchers.
Some specific applications of composites in car bodies are shown in
Fig. 1.
a) b)
Fig. 1 Examples of the application of composite materials: a) Porsche A-
pillar, b) MQB platform engine cover
Thermal drilling provides a way to create a bond between a metal
sheet and a continuous fiber-reinforced composite with two
significant advantages: no fastener and therefore no added weight,
and no disruption to the continuity of the fibers. Making joints
without disturbing the continuity of the fibres in the composite is
possible only if the matrix of the composite is thermosoftening
polymer. It heats up when drilled, softens and the fibres are
deflected out of position without breaking when the tool penetrates
and the bushing is formed, Fig. 2.
Fig. 2 Comparison of a hole in a fibre composite formed by a) conventional
drilling, b) and c) thermal drilling of a fibre composite with a
thermosoftening matrix, adapted from [4,6]
The mechanical joining of the metal sheet and the composite plate
formed by thermal drilling can then look like this, Fig. 3:
Fig. 3 Schematic diagram of metal sheet and fibre composite joint made by
thermal drilling
The aim of this paper is to present the differences in shaping the
bushing when drilling the metal plate alone, and also when drilling
the metal-composite pair by direct and sequential thermal drilling.
The objective is to create a bushing with sufficient thickness and
length to penetrate through the full thickness of the composite plate
to form a mechanical joint and to determine which of the strategies
investigated is the most appropriate.
2. Materials and Methods
Materials used in the automotive industry were chosen for the metal
– composite joints formation using thermal drilling. Composite
materials are progressively replacing conventional metals in
automotive construction. They compete with them in terms of
comparable mechanical properties and low specific weight - i.e. an
advantageous strength-to-weight ratio, also called specific strength.
Characteristics and identification of selected materials:
- Aluminium alloy EN AW-6082 T6 (AlSi1MgMn) – Al alloy
supplied in a precipitation-hardened state with relatively high
mechanical properties. The alloy was supplied in the form of a 1
mm thick rolled sheet. Hereinfter Al.
- AZ 91 magnesium alloy with the main alloying elements Al and
Zn. The alloy was supplied in the form of a 2 mm thick rolled sheet.
Hereinfter Mg.
- Composite with polypropylene (PP) matrix reinforced with
continuous glass fibres in two perpendicular directions. The
thickness of the consolidated composite organosheet is 1.55 mm,
which is the result of the compaction of three layers of prepregs.
Hereinfter GF.
- Composite with polypropylene matrix (PP) reinforced with
continuous carbon fibres in two perpendicular directions. The
thickness of the consolidated composite orgaanosheet is 1.55 mm,
the result of the compaction of seven layers of prepregs. Hereinfter
CF.
The basic characteristics of the materials are shown in Tab. 1-3.
Tab. 1 Chemical composition of alloys in wt.%
alloy
Si
Mg
Mn
Fe
Zn
Cu
Al
Al
1.0
0.7
0.44
0.4
0.08
0.06
balance
Mg
0.09
balance
0.14
0.004
0.93
0.02
8.9
INNOVATIONS 2025
99
Tab. 2 Mechanical and physical properties of alloys
alloy
Re [MPa]
Rm [MPa]
A50 [%]
density [g·cm-3]
Al
295
340
15
2.71
Mg
280
200
24
1.81
Tab. 3 Properties of composite organosheets
Composite
PP CF50 T200 OS
PP GF45 T600 OS
Matrix
polypropylene
Reinforcement
Carbon fibre
Glass fibre
Type of fibre
Carbon HT
E glass
Weaving
Twill 2/2
Twill 2/2
Surface mass.
mats
200 g·m-2
600 g·m-2
Density of
weaving
3K
1200 tex
Fibre content by
weight
50%/50%
50%/50%
Fibre content in
the composite
51%
47%
Surface weight of
finished product
301 g·m-2
887 g·m-2
Prepreg
thickness
0.22 mm
0.5 mm
Specific weight of
composite
1.46 g·cm-3
1.68 g·cm-3
Melting
temperature
165C
165C
Operating
temperature
(short-term)
140C
161C
Operating
temperature
(long-term)
100C
100C
CTE (23-80°C)
3.2×10-6K-1
11×10-6K-1
The geometry of the test joints as well as the shape and dimensions
of the test specimens, determined in accordance with STN EN ISO
12996:2014 using a 5.3 mm diameter tool (Flowdrill Long 5.3 mm),
are shown in Fig. 4. The opening (joint) is located in the centre of
the overlapped area.
Fig. 4 Shape and dimensions of test specimens and joint for the tool
5.3 mm.
The parameters tested in thermal drilling were: rotational speed
2400 min-1 and 4800 min-1, feed rate 60 mm·min-1 and
240 mm·min-1. Thermal drilling was carried out on a bench drill,
the materials were heated and clamped in the fixture. The fixture for
setting and securing the relative position of the materials during
thermal drilling is shown in Fig. 5.
a)
b) c)
Fig. 5 Fixture for joining production: a) CAD model, b) open fixture,
inserted materials, c) closed fixture, ready for thermal drilling
The dimensions of the resulting bushing were determined when the
metal plates only were drilled in the fixture, as well as when metal
plate is placed on the composite sheet in direct and sequential
drilling. Direct drilling means the drilling of overlapped metal-
composite materials heated and 165°C at once, Fig. 6. Sequential
drilling means that in the first step a opening is made with a tool in
the preheated composite, which is overlapped with a metal plate in
the second step and drilled again, assuming easier forming the metal
bushing in the pre-prepared opening in the composite and less
delamination, Fig. 7.
Fig. 6 Direct drilling
Fig. 7 Sequence drilling
After forming the bushings in the metal sheets separately, as well as
after forming the joints with the composites, the bushing
characteristics - thickness and length - were monitored, Fig. 8. The
overall suitability of the shape of the resulting bushing was also
assessed - it should have a sharp end, conical shape to be able to
penetrate the composite without much delamination of the layers.
INNOVATIONS 2025
100
Fig. 8 Basic bushing characteristics monitored
Results
The metallographic study of the bushings was carried out on Al, Mg
metal sheets at all process parameters tested, and also for Al and
Mg joints with CF and GF composites at all process parameters
(two levels of rotational tool speed, two levels of feed rate) and
direct and sequential drilling strategies. Example metallographic
sections for Al and Mg and their joints with CF are shown in Tab. 4
and 5.
Tab. 4 Al-CF joints
thermal drilling of Al sheet
Al-CF joint, direct drilling
Al-CF, sequential drilling
Table 4 shows that thermal drilling of Al sheet only leads to
inappropriate bushing ends, often resulting in bushing
fragmentation. When drilling with the composite, either by direct or
sequential drilling, the geometry of the bushing was improved. It
can be concluded that Al shaping that is constrained by other
material or a pre-drilled hole improves the bushing shaping process.
Tab. 5 Joints Mg-composite CF
thermal drilling of Mg sheet
Mg-CF joint, direct drilling
Mg-CF joints, sequential drilling
In Mg alloy, on the other hand, a bushing of suitable shape is
formed when the Mg sheet is drilled separately, whereas when
drilling with the composite by direct drilling, the flow of Mg
material into the bushing area is significantly impeded. On the
contrary, sequential drilling improves the shaping of the bushing by
pre-drilling an opening in the composite, which allowed a sharper
and longer bushing to be formed than in direct drilling. The overall
results of the bushing characteristics for all material combinations
and drilling strategies are shown in Tab. 6 and Fig. 9 and 10.
Fig. 9 Bushings characteristics for Al-CF
Fig. 10 Bushings characteristics for Mg-CF
INNOVATIONS 2025
101
Tab. 6 Geometric characteristics of the bushing in metal drilling and direct and sequential joining
Mat.
Process parameters
Thermal drilling of
metallic sheets only
Direct drilling
Sequential drilling
CF
GF
CF
GF
RPM
[min-1]
Feed rate
[mm·min-1]
thickness
[mm]
length
[mm]
thickness
[mm]
length
[mm]
thickness
[mm]
length
[mm]
thickness
[mm]
length
[mm]
thickness
[mm]
length
[mm]
Al
2400
240
0.57
1.99
0.80
1.98
0.86
2.05
0.68
1.77
0.69
1.85
60
0.55
3.02
0.56
1.73
0.59
0.80
0.59
1.96
0.59
2.90
4800
240
0.63
2.29
0.58
1.98
0.67
2.10
0.64
2.25
0.66
2.20
60
0.49
1.65
0.62
2.96
0.61
2.60
0.56
2.07
0.56
2.46
Mg
2400
240
0.92
2.18
1.09
2.18
1.10
2.29
1.28
2.09
1.33
2.43
60
0.89
2.61
1.07
3.06
1.17
2.60
1.04
2.92
1.04
3.07
4800
240
0.98
3.03
1.09
2.65
1.23
2.38
1.28
2.84
1.19
3.02
60
0.98
3.15
1.18
2.16
1.07
2.74
0.99
2.60
1.01
3.44
Notes:
Inappropriate bushing shape
Rather inappropriate bushing shape
Rather suitable bushing shape
Suitable bushing shape
From Tab. 6 and Fig. 9 and 10, the following observations can
be made:
the thickness of the bushings is not significantly affected by the
process parameters of thermal drilling
a greater thickness of bushings is achieved for Mg alloy (0.89-
1.33 mm), which is due to a larger material volume available in
drilling area with respect to the 2 mm plate thickness
for Al alloy, the thickness of the bushings varies from 0.49 to
0.86 mm, which corresponds to a smaller sheet thickness
compared to Mg
the length of the bushings was greater than the thickness of the
composite for both Al and Mg, except in one case (Al 2400/60 -
GF), i.e. satisfactory
bushing length is more sensitive to the variation of thermal
drilling process parameters, with the feed rate having a more
pronounced effect on the bushing length than the RPM. The
effect of feed rate on bushing length is clear - slower feed rate
(especially when combined with higher speed) allows longer
bushing lengths to be shaped, both for Al and Mg.
The length of the bushings is around 2 mm for Al-CF, between
2 and 3 mm for Al-GF, the length of the bushings is between 2
and 3 mm for Mg-CF, between 2.5 and 3.5 mm for Mg-GF.
This can be justified by the less dense weave of the GF fabric,
the thicker fibres and the lower number of glass fibre layers in
GF compared to CF. This implies that the glass fibres have less
resistance to penetration of the bushing.
If we consider not only the dimensions of the bushing but also
their shape visually, and mark the cases of suitable shape in green in
Table 6, we can determine the appropriate process parameters for
the formation of the joints easier. Suitable process parameters for
joining Al and Mg sheets with CF and GF composites are
following: rotational tool speed 4800 min-1 and slower feed rate
60 mm·min-1. A suitable joining strategy is sequential drilling.
If a joint is formed with the above mentioned suitable process
parameters with sufficient thickness and length of the bushing, the
joint can be secured against opening by hemming flange, Fig. 11.
The fibres are closed under the formed hem flange, which in
addition to closing the joint also prevents delamination of the
composite.
Conclusion
Thermal drilling technology can be used for creation efficient
joints of non-ferrous metal alloys and fibre composites with
thermosoftening matrix at suitable process parameters. Good
joining is also supported by a suitable two-step drilling strategy, but
this means an extra operation and an increase in joining time when
applied under industrial practice conditions. Therefore, the load
carrying capacity of joints formed by these drilling strategies still
needs to be mapped.
Fig. 11 Al-GF joint with hem flange
Acknowledgement: This work was supported by The Ministry of
Education, Research, Development and Youth of the Slovak
Republic under Grant VEGA 1/0229/23: Research on the
applicability of thermal drilling technology for the creation of
multi-material joints in the automotive industry.
References
[1]. FEKETE, J.R. – HALL, J.N.: 1 – Design of auto body:
Materials perspective. In: Automotive Steels. Vol. 17, p. 1-18,
https://doi.org/10.1016/B978-0-08-100638-2.00001-8.
[2]. LIEDL, G. et al.: Joining of Aluminum and Steel in Car Body
Manufacturing. In: Physics Procedia. Vol. 11, No. 12 (2011), p.
150-156, https://doi.org/10.1016/j.phpro.2011.03.019.
[3]. MORI, K. – ABE, Y.: A review on mechanical joining of
aluminium and high strength steel sheets by plastic deformation.
In: International Journal of Lightweight Materials and
Manufacture. Vol. 18, No. 1 (2018), p. 1-11,
https://doi.org/10.1016/j.ijlmm.2018.02.002.
[4]. TROSCHITZ, J. – KUPFER, R. – GUDE, M.: Process –
integrated embedding of metal insert in continuous fibre
reinforced thermoplastics. In: Procedia CIPR. Vol. 19, No. 85
(2019), p. 84-89, https://doi.org/10.1016/j.procir.2019.09.039.
[5]. GUZANOVÁ, A. et al.: Investigation of Applicability Flowdrill
Technology for Joining Thin-Walled Metal Sheets. In: Metals.
Vol. 22, No. 4 (2022), p. 1-23,
https://doi.org/10.3390/met12040540.
[6]. SEIDLITZ, H. et al: New Joining Technology for Optimized
Metal/Composite Assemblies. In: Journal of Engineering. Vol.
14, p. 1-11, http://dx.doi.org/10.1155/2014/958501.
INNOVATIONS 2025
102
Optimization of femtosecond laser parameters on surface morphology of lithium disilicate
glass ceramic
Andreja Carek1, Ljerka Slokar Benić2*, Hrvoje Skenderović3, Lorna Martić1
1University of Zagreb School of Dental Medicine, Zagreb, Croatia
2 University of Zagreb Faculty of Metallurgy, Sisak, Croatia
3Institute of Physics. Zagreb, Croatia
slokar@simet.unizg.hr
Abstract: Lithium disilicate glass-ceramic is one of the most commonly used aesthetic materials in fixed prosthetics that requires surface
pre-treatment. Recent studies have begun to propose a femtosecond laser for processing to improve surface morphology and microstructure,
but without defined parameters. Therefore, the aim of this study is to determine the optimal parameters for the surface morphology of lithium
disilicate glass-ceramics. The untreated sample of lithium disilicate glass-ceramic was observed with an optical profilometer. It was then
treated with a femtosecond laser, forming squares on the surface of the sample, which were observed with an optical profilometer, and the
surface morphology was analysed with a scanning electron microscope. The results show that laser treatment with higher energy densities
leads to an increased roughness of the surface.
Keywords: LITHIUM DISILICATE GLASS CERAMIC, FEMTOSECOND LASER, SURFACE MORPHOLOGY, MICROSTRUCTURE
1. Introduction
In recent years, there has been a growing interest in improving
the strength and aesthetics of materials used in dental prosthetics in
order to offer patients a better appearance and a longer life for their
dental prosthesis. One of the most commonly used aesthetic
materials in dental prosthetics is lithium disilicate glass-ceramic [1].
Lithium disilicate glass-ceramic is the material of choice for the
fabrication of aesthetic prosthetic dental restorations due to its
excellent optical and mechanical properties, its chemical stability
and its improved translucency. Lithium disilicate glass-ceramic
offers excellent aesthetics and a natural appearance and has the best
resistance to external factors such as the consumption of tea, coffee,
cigarettes or other substances that promote pigmentation [2], [3]. It
contains a high volume fraction of about 70 % of needle-like
lithium disilicate crystals (Li2Si2O5) in the matrix [4]. The
crystallisation mechanism of lithium disilicate glass-ceramics is
bulk crystallisation, which allows heterogeneity, nucleation and
growth of crystals throughout the glass [5]. The poor properties of
lithium disilicate glass-ceramics, despite their high pressure
resistance, are fractures in thin parts of the restorations, which can
be a particular challenge in patients with parafunctional habits such
as bruxism [6].
To achieve long-term clinical success with lithium disilicate
glass-ceramics, the surface morphology must be pretreated. Lithium
disilicate glass-ceramics are conventionally processed using
chemical methods [7], [8]. Recent studies suggest femtosecond
laser (FL) treatment as an alternative pre-treatment method for
surface treatment in adhesive cementation techniques. Femtosecond
lasers belong to the category of ultrafast lasers or ultrashort pulse
lasers. The generation of such short light pulses is almost always
achieved using passive mode locking techniques. This results in
pulses with high repetition rates (MHz or GHz). The speed
combined with the limited average output power leads to relatively
low pulse energies (nJ). One of the main advantages of using FLs is
the high peak power, which makes them ideal for a variety of
applications. Due to the wave effect of optical fibres, laser systems
have very good thermal and vibrational stability and can produce
nearly diffraction-limited beam profiles. Femtosecond lasers can be
tuned to a wide range of wavelengths [9]. FL treatment is often
referred to as surface abrasion. FL treatment can effectively modify
the surface. However, the optimum FL parameters for the treatment
of lithium disilicate glass-ceramics are not yet known [1].
The aim is to analyse the influence of FL treatment on the
microstructure and surface morphology of lithium disilicate glass-
ceramic and to determine the optimum parameters for femtosecond
laser treatment of the surface of lithium disilicate glass-ceramic.
2. Materials and methods
The material used in this study was lithium disilicate glass-
ceramic (IPS e.max PRESS, Ivoclar Vivadent, Schaan,
Lichtenstein) with the dimensions 32 mm x 10 mm x 1 mm.
Femtosecond laser irradiation was performed using a laser
system that originally generates laser radiation with pulses of 190 fs
duration, a maximum energy per pulse of 1.5 mJ and an average
total power of 15 W. A repetition rate of 10 kHz was used in this
study, and the sample scanning speed (vscan) was 1 mm/s. The beam
has an initial diameter of 6 mm and the spot size on the sample was
adjusted to 2r0 = 40 µm using the optical system. The maximum
energy per unit area (peak fluence) is equal to the highest energy
emitted by the Gaussian laser beam, which was calculated
according to the following formula [1]:
𝐹𝑝𝑒𝑎𝑘 =2𝐸
𝜋𝑟0
2 (1)
where E is the energy of the laser pulse and r0 is the radius of the
beam at intensity 1/e2. The energy of a laser pulse is calculated
using the following formula :
𝐸=𝑃1
𝑓
𝑟𝑒𝑝 (2)
where P is average laser radiation power and is measured in watts
and frep is pulse repetition frequency.
The equivalent number of pulses is defined as the number of pulses
at a point, and the formula for calculation is as follows [1]:
𝑁𝑒𝑞 = 2𝑟0
𝑓
𝑟𝑒𝑝
𝑣𝑠𝑐𝑎𝑛 (3)
where vscan is laser beam speed. The corresponding number of
pulses based on these calculations is 40. After the calculation, the
following maximum energies per area (Fpeak) were used for lasering
three squares: Fpeak (A) = 12.6 J/cm2, Fpeak (B) = 9.5 J/cm2 and Fpeak
(C) = 18.09 J/cm2. 3 test squares (A, B, C) measuring 1 x 1 mm
were produced on a glass-ceramic plate.
An optical profilometer Filmetrics ProFilm 3D, KLA, San
Diego, California, with the application ProFilmOnline® was used to
analyse the surface topography, i.e. to determine the roughness as
the arithmetic mean height (Sa) of the surface. The untreated and
laser-treated surfaces of the lithium disilicate glass-ceramic sample
were viewed using a Nikon CF IC epi Plan 50x objective with a
spatial magnification of 4x zoom and a pixel area of 1. The sample
was imaged using the green light interference method, envelope
peak analysis type, and the scan length of the wafer was 22 μm in
the ProFilmOnline® application.
INNOVATIONS 2025
103
A scanning electron microscope (SEM, Tescan Vega 3 LMU,
Tescan, Brno, Czech Republic) at 3 kV was used to analyse the
surface.
3. Results and discussion
The SEM images of the untreated and laser-treated surfaces
with a scan resolution of 50 (left) and 5 (right) µm are shown in
Figure 1.
Untreated surface
Square A
Square B
Square C
Fig. 1. SEM images of tested surfaces.
SEM images show the rough and uneven surface morphology of
the lithium disilicate glass-ceramic. Laser-induced periodic surface
INNOVATIONS 2025
104
structures (LIPSS) and/or microcracks in laser-treated surfaces of
lithium disilicate glass-ceramics are visible on the lasered surfaces,
as is increased surface roughness. The investigations carried out
have shown that the number of repetitions, i.e. the pulses and the
energy emitted per unit of lasered surface, significantly affects the
surface morphology of lithium disilicate glass-ceramics.
The microstructure of the unprocessed part of the glass-ceramic
and the lasered squares A, B and C was examined with a scanning
electron microscope at an energy of 3 kV. Square A (12.6 J/cm2) is
morphologically and microstructurally similar to the untreated
surface with visible LIPSS. These lines are parallel to the
polarisation of the laser beam and depend on the total energy per
unit area and the number of repetitions, which was seven in this
study. Irregularities and a rough surface morphology can be seen on
the untreated surface of the glass-ceramic. On square B (9.5 J/cm2),
larger irregularities and unevenness of the sample are visible. On
square C (18.09 J/cm2), large pits with LIPSS and microcracks
were also found.
The increase in Fpeak led to the appearance of LIPSS, with the
microcracks increasing the most. The SEM images show that too
little Fpeak makes the surface of the highest points rougher and too
much Fpeak creates LIPSS with microcracks that increase the
likelihood of the material itself breaking. The LIPSS on square A
are just proof of the creation of periodic structures to better shape
the surface morphology and microstructure.
After calculations and analyses with an optical profilometer, it
was determined that the surface morphology changed after seven
repetitions. The surface roughness (Sa) values before and after
treatment of lithium disilicate glass-ceramics with the femtosecond
laser are shown in Table 1.
Table 1: Surface roughness
Untreated
surface
Square A (12.6
J/cm2)
Square B (9.5
J/cm2)
Square C
(18.09 J/cm2)
Sa [µm]
2.01
1.91
3.04
3.03
The investigations carried out have shown that the number of
repetitions, i.e. the pulses and the energy emitted per unit of lasered
surface, significantly affects the surface morphology of lithium
disilicate glass-ceramics.
Figure 2 shows three-dimensional optical profilometry images
and line profiles of untreated and laser treated surfaces of lithium
disilicate glass-ceramic named as squares A, B and C.
Untreated surface
Square A
INNOVATIONS 2025
105
Square B
Square C
Fig. 2. 3-D optical profilometry images and line profiles of tested surfaces.
In this study, the surface morphology and microstructure of the
untreated and FL-treated surfaces were analysed. The advantage of
lasers with ultrashort pulses compared to longer pulses or
continuous lasers is the extremely short time of interaction between
the laser radiation and the material, which significantly reduces the
thermal effects on the surrounding area. When considering FL
surface treatment, the mechanical properties should also be taken
into account. Further studies are required to determine the effects of
FL treatment on mechanical properties, crystallography and bond
strength. Yavuz and colleagues [10] showed that FL treatment of
lithium disilicate glass-ceramics increased the bond strength with
adhesive resin cement. Although FL treatment resulted in a higher
shear bond strength compared to conventional methods, it was still
lower than that achieved by etching with hydrofluoric acid or
tribochemical silica sandblasting.
4. Conclusions
Peak fluence frep = 10kHz, wavelength = 1030 nm and pulse
duration = 190 fs were used for this study. Energy of laser pulse of
9.5, 12.6 and 18.09 J/cm2 respectively and equivalent number of
pulses Neq = 40 were calculated for the three lasered squares on
untreated lithium disilicate glass-ceramic (IPS e.max PRESS,
Ivoclar Vivadent, Schaan, Lichtenstein). The surface roughness was
analysed using an optical profilometer. For the untreated surface of
the lithium disilicate glass-ceramic, Sa was 2.011 μm, while for the
surface treated with the femtosecond laser, Sa was 1.91 μm, 3.04 μm
and 3.03 μm, respectively. SEM analysis showed increase in surface
roughness as well as LIPPS and microcracks on laser treated
surfaces.
5. References
1. M. Inokoshi, K. Yoshihara, M. Kakehata, H. Yashiro, N.
Nagaoka, W. Tonprasong, K. Xu, S. Minakuchi, Materials. 15, 3614
(2022)
2. S. Schelkopf, C. dini, T. Beline, A. G. wee, V. A.R. Barao, C.
Sukotjo, J. C-C. Yuan, Materials. 15, 6901 (2022)
3. K. Bauer, A. Carek, L. S. Benić, Badel, Materials, 17, 3160
(2024)
4. M. S. Dahiya, V. K. Tomer, S. Duhan, Bioactive glass/glass
ceramics for dental applications (Elsevier Inc., 2019)
5. P. Goharian, A. Nemati, M. Shabanian, A. Afshar, J. Non.
Cryst. Solids, 356, 4–5, (2010)
6. M. Schmitter, W. Bömicke, R. Behnisch, J.L. Bermejo, M.
Waldecker, P. Rammelsberg, B. Ohlmann, J. Clin. Med. 12, 1
(2023)
7. R. R. Braga, R. Y. Ballester, M. Daronch, Dent. Mater. 16, 4,
(2000)
8. Y. Maruo, G. Nishigawa, M. Irie, K. Yoshihara, T.
Matsumoto, S. Minagi, J. Appl. Biomater. Funct. Mater. 15, 1
(2017)
9. T. Ide, T. O’Brien, Refract. Surg. (UCLA, 2009)
10. T. Yavuz, Ö. Y. Özyılmaz, E. Dilber, E. S. Tobi, H. Ş. Kiliç, J.
Prosthodont. 26, 5 (2017)
INNOVATIONS 2025
106
Electrical and hydrodynamic conditions in electrolyte-plasma treatment of internal surfaces
Aleksandr Korolyov1, Vyacheslav Tomilo2, Vladimir Niss2
Scientific and Technological Park BNTU «Polytechnic», Belarus1
Belarusian National Technical University, Belarus2
E-mail: korolyov@park.bntu.by
Abstract: A method has been developed for performing electrolyte-plasma treatment (EPT) of internal surfaces. The method is based on the
use of an electrode device that moves along the inner surface of the tubular workpiece and provides the necessary electrical and
hydrodynamic conditions. The design of the electrode device combines a counter-electrode and a nozzle for supplying polarized electrolyte
to the treated surface. The study presents the results of investigations into the electrical and hydrodynamic conditions during EPT of internal
surfaces, as well as the influence of hydrodynamic characteristics on current parameters, metal removal rate, and the quality of the resulting
surface.
Keywords: ELECTROLYTE-PLASMA TREATMENT, VAPOR-GAS SHELL, INTERNAL SURFACE, CURRENT DENSITY, FLOW
VELOCITY.
1. Introduction
As an alternative to electrochemical methods for improving
the surface quality of metal products, electrolyte-plasma treatment
(EPT) is widely used in industry. Its main advantage lies in the use
of aqueous salt-based electrolytes [1]. One of the most promising
applications of the method is in the manufacturing of medical
devices. Based on previous research, a number of new high-
efficiency processes have been developed and implemented using
EPT, ensuring improved surface quality of products made from a
wide range of materials, including corrosion-resistant steels,
titanium and titanium alloys, cobalt-chromium alloys, and Nitinol
[2–4].
EPT of external surfaces of such products presents no
significant technological challenges and is typically performed
using a conventional setup with free immersion in the electrolyte.
However, the treatment of internal surfaces using this approach is
difficult due to electrostatic shielding effects [5]. In such cases, the
use of auxiliary electrode devices is required to create the necessary
electrical and hydrodynamic conditions on the internal surface.
A review of literature on EPT of internal surfaces of tubular
products has shown that existing methods suffer from several
significant drawbacks, which prevent the establishment of a stable
EPT process with consistently high surface quality [6–8]. To
address these issues, a technical solution has been developed based
on the use of an electrode device that moves along the treated
internal surface of a tubular product and ensures the required
electrical and hydrodynamic conditions [9].
1. Analysis of anodic processes and
hydrodynamic conditions
The design of the electrode device combines a counter-
electrode (cathode) and a nozzle for supplying polarized electrolyte
to the treated surface. Fig. 1 shows the design of the developed
electrode device. The electrode device is connected to the pressure
pipe (8) made of corrosion-resistant steel using a nut (7). The
negative pole of the power supply is connected to the pipe (8). The
electrolyte is supplied under pressure into the electrode device, then
flows through the cavity and openings in the nut (7) to the cathode
(3) and is flowed into the treatment zone through the adjustable gap
between tubes (4) and (5). The cathode (3) is fixed to the nut (7)
using a clamping screw (6). The gap (δ) between tubes (4) and (5),
which forms an annular slit nozzle, is adjusted by placing shims
between the cathode (3) and the nut (7).
The current-carrying parts of the electrode device are
protected from short-circuiting with the treated surface by a housing
made of electrically insulating material (fluoroplastic), which
consists of tubes (4) and (5), plugs (1, 2), and a cap (9). To connect
tubes (4) and (5) with plugs (1) and (2), respectively, pins (10)
positioned at 120° intervals are used. Additionally, the pins (10)
ensure a uniform gap between the electrode device and the treated
surface during movement inside the pipe. The used electrolyte
drains from the treatment zone through the lower annular gap by
gravity, while the generated vapor-gas shell is removed through the
upper gap.
a
b
Fig. 1 Electrode device for internal surface treatment of tubular
products: a – model of the electrode device; b – design of the
electrode device with a schematic representation of electrolyte flow
The shape and dimensions of the cathode determine the
electrical conditions of the EPT process on the treated internal
surface. They influence the distribution of electric field intensity
lines in the treatment zone and, accordingly, the current density in
the electrolyte-anode circuit, the stability of the vapor-gas shell, and
the quality of the resulting surface. One of the key conditions for
forming a stable vapor-gas shell during EPT is that the cathode
surface area Sc must exceed the anode surface area Sа. The
monograph [10] presents studies that established the influence of
voltage, temperature, and electrolyte concentration on the critical
values of the Sc/Sа ratio at which the stability of the EPT process is
disrupted, causing a reversal to cathodic mode. It was found that,
under conditions where the voltage is applied to a pre-immersed
anode, the limiting values of the Sc/Sа ratio range from 2.00 to 2.85
depending on the electrolyte temperature. Considering that internal
surface treatment using an electrode device is also carried out under
complex hydrodynamic and electrical conditions, it is necessary to
provide a significantly higher Sc/Sа ratio to avoid process inversion
and cathode damage. In addition to ensuring a larger cathode
surface area compared to the anode when using an electrode system,
it is also crucial for a stable EPT process that the electric field lines
are equidistant from the treated surface to the cathode surface. To
meet these conditions, the electrode system design should
incorporate a cathode shaped as a globoid or a hyperboloid, which
approximates a globoid in form (fig. 2).
INNOVATIONS 2025
107
Fig. 2 Electric field intensity lines when using a globoid-shaped
cathode
In the electrode device for processing the inner surfaces of
tubular products, the cathode surface area Sc remains constant.
Accordingly, to maintain the stability of the EPT process, it is
necessary to control the surface area Sа being treated. The area Sа,
which forms an annular treatment zone of height ℎ, depends on the
gap size forming the slit nozzle, and the hydraulic characteristics
– namely, the flow rate Q and the average electrolyte flow velocity
el. The flow of the electrolyte from the annular slit nozzle of the
electrode device onto the inner surface corresponds to the scheme of
a flat jet interacting with a barrier perpendicular to the flow.
Optimal conditions for forming a vapor-gas shell are created when
the jet between the electrode device and the treated surface remains
compact – i.e., the flow continuity is maintained. Furthermore, the
flow velocity must not exceed the limit at which the flow disperses
upon impact with the barrier. Flow continuity ensures stable
electrolyte conductivity, and the direction and intensity of the
electric field lines remain unchanged.
The maximum electric field intensity occurs in a narrow
section of the treated surface directly opposite the slit nozzle. In this
area, the intensity is sufficient to form a continuous vapor-gas shell
and ensure stable EPT process. As the distance from this zone
increases, the field intensity decreases and eventually reaches a
critical value at which continuous vapor-gas shell formation is no
longer possible. Due to the constant cathode size, the height of the
area where stable EPT can occur does not depend on the width of
the slit nozzle and is defined as hcr. In zones with lower field
intensity, conditions similar to the switching mode arise, in which
the continuity of the vapor-gas shell is periodically interrupted, with
the process transitioning to the switching mode, accompanied by a
chaotic increase in current, intense vapor and gas evolution, and the
emergence of hydrodynamic flows.
Fig. 3 illustrates process diagrams characterizing treatment
zone behavior under various hydrodynamic conditions. When the
height of the treated surface area is less than or equal to the critical
value (ℎ≤ hcr) and no flow dispersion occurs, a continuous vapor-
gas shell forms at the anode under the jet, and a stable EPT process
is maintained (fig. 3a and 3b). Maximum efficiency and surface
quality are achieved when ℎ = ℎcr, where the largest possible area
is treated simultaneously (fig. 3b).
As a result of increasing the width of the slit nozzle to values
at which the height of the section of the treated surface becomes
greater than the critical value (ℎ > hcr), a switching mode arises in
areas with low electric field strength, which complicates the process
of EPT of the inner surface (fig. 3c). Steam and spray bursts cause
unstable flow within the confined internal volume, leading to
current fluctuations. Under these conditions, vapor-gas shell cannot
be effectively vented through the upper annular gap. A similar
switching mode may also occur if the flow disperses upon impact
with the surface (fig. 3d), or if the flow rate increases to the point
where the electrolyte fills the entire space between the electrode
device and the surface, preventing free drainage through the lower
annular gap.
Fig. 3 Processes in the treatment zone under various hydrodynamic
conditions: а – h
hcr; b – h = hcr; c – h
hcr; d – spreading of the
continuous jet
To ensure a stable EPT, both the average velocity and the flow
rate of the electrolyte through the annular slit nozzle must be taken
into account, as they are interrelated according to relation (1). On
the one hand, the electrolyte flow during processing must be
sufficient to provide heat exchange and to remove the products of
anodic dissolution from the treatment zone. At the same time, the
flow must have a velocity that ensures its continuity in the gap
between the electrode device and the surface being treated, and
spreading of the jet upon impact with the surface should be avoided.
On the other hand, the flow rate and velocity are limited by the
condition that the spent electrolyte must be able to drain freely from
the treatment zone through the lower annular gap under the force of
gravity, while the resulting vapor-gas mixture is removed through
the upper annular gap,
el Q
, (1)
where – the live cross-sectional area of the flow, which depends
in part on the slit nozzle width δ.
The current and metal removal rate during the processing of
the inner surface are also determined by the hydrodynamic
characteristics and the width of the slit nozzle. As shown above,
increasing the slit width up to a certain value – while maintaining
other conditions necessary for a stable EPT process – allows for an
increase in the treatment zone height up to hcr. In this case, the
current density is significantly influenced by the jet pressure force
on the surface 𝑃, which is determined by formula (2). The greater
the jet pressure force, the thinner the vapor-gas shell becomes,
which in turn reduces its electrical resistance and increases the
current density.
el el
PQ
, (2)
where el – electrolyte density.
Thus, the combined effect of hydraulic conditions and
geometric parameters of the electrode device on anodic processes in
the treatment zone, current density, metal removal rate, and
INNOVATIONS 2025
108
consequently, the quality of the resulting surface necessitates
comprehensive experimental studies to establish the
interrelationships of these factors and to optimize the processing
parameters.
3. Materials and methods
The samples for the study were sections of AISI 321 steel
tubes with an inner diameter of 26.5 mm. The outer diameter of the
electrode device was 25 mm (resulting in a gap of 0.75 mm between
the electrode device and the surface being processed). The width of
the slit nozzle of the electrode device (δ) varied from 0.5 to 2.0 mm
in 0.5 mm increments. A 6% ammonium sulfate solution was used
as the electrolyte. Since the electrolyte flow processing is localized
and the treated surface has a simple shape, to improve productivity
and reduce thermostatting costs, the electrolyte temperature during
the study was set at 75 °C. The electrolyte flow rate (Q) was
controlled by adjusting the pump motor power using pulse-width
modulation (PWM). Prior to the experiments, for each δ value, the
relationship between the flow rate and the PWM duty cycle was
experimentally determined. The average flow velocity of the
electrolyte (el) from the electrode device was calculated using
equation (1). EPT was carried out at a voltage of 300 V. The current
was measured using UNIT-203 current clamps. The mass of the
samples before and after treatment was measured with an Ohaus
Pioneer 214-PA analytical balance (0.0001 g resolution). Surface
roughness was measured using a MarSurf PS1 profilometer.
4. Results and discussion
Fig. 4 shows the dependencies characterizing the effect of
electrolyte flow rate on current intensity at various slit nozzle
widths (δ). The study results revealed that for any value of δ, an
increase in flow rate leads to a directly proportional increase in
current intensity. Moreover, the dependencies for different slit
widths converge into a single line and differ only in the magnitude
of the achievable current. The maximum current was observed at δ
= 1.0 mm. It should be noted that variations in electrolyte flow rate
– and, accordingly, in current intensity – significantly affected the
nature of the treatment process. For each slit width value, at a
certain flow rate Q, instability in the electrolyte-plasma mode was
observed, resulting in current surges characteristic of the switching
mode, accompanied by electrolyte splashing.
Fig. 4 The influence of electrolyte flow rate on current intensity at
various slit nozzle widths
Fig. 5 shows the dependencies characterizing the effect of the
average electrolyte velocity from the electrode device on current
intensity at various slit nozzle widths. Unlike the dependencies
shown in fig. 4, these graphs allow for a clearer assessment of the
parameter ranges that ensure the feasibility of internal surface
treatment. In the graphs, solid lines indicate regions corresponding
to the stable existence of the vapor-gas shell with a stable
electrolyte-plasma mode, while dashed lines mark regions where a
switching mode occurs. From the fig. 4, it is evident that stable EPT
is not possible at slit widths of δ = 1.5 mm and δ = 2.0 mm.
Relatively wide operational ranges of average electrolyte velocity
(el) are achieved with slit nozzle widths ranging from δ = 0.25 to
1.0 mm. The broadest range (el = 0.38–0.78 m/s) corresponds to
the smallest slit width (δ = 0.25 mm). It is apparent that the working
range of average electrolyte velocity can be expanded by increasing
the electrolyte temperature. However, this approach would reduce
processing productivity, which is undesirable given the small area
of the surface being treated at any one time.
Fig. 5 The influence of electrolyte flow velocity from the electrode
device on current intensity at various slit nozzle widths: 1 –
=
0,5 mm; 2 –
= 1,0 mm; 3 –
= 1,5 mm; 4 –
= 2,0 mm
Since a stable electrolyte-plasma mode could not be achieved
at slit nozzle widths of 1.5 mm and 2.0 mm, further research was
conducted at δ = 0.25, 0.5, and 1.0 mm. To determine the current
density during processing at various electrolyte flow velocities from
the electrode device (and correspondingly at different current
values), the processing area had to be established. The processing
area was defined as the product of the circumference of the
sample’s inner surface and the height of the processing zone h. To
determine the height h, processing was performed with a stationary
electrode device. Depending on υₑₗ, the average values of h were:
1.9–2.4 mm for δ = 0.25 mm; 2.5–3.2 mm for δ = 0.5 mm; and
0.30–0.43 mm for δ = 1.0 mm.
Figure 6 shows the current density as a function of the
average electrolyte velocity for various slit nozzle widths. The
current density was calculated as the ratio of the current during
sample processing to the area of the processing zone. As υₑₗ
increases within the ranges that ensure a stable EPT, current density
also increases: from 1.14 to 3.67 A/cm² for δ = 0.25 mm and from
1.76 to 3.04 A/cm² for δ = 1.0 mm, which is related to the
increasing pressure force of the jet on the vapor-gas shell according
to equation (2). These dependencies exhibit a parabolic character.
Fig. 6 The influence of the average electrolyte velocity on the
change in current density during the processing of internal surfaces
0
5
10
15
20
0 1 2 3 4 5
I, А
Q, dm3/min
- 0,25 mm
- 0,5 mm
- 1,0 mm
- 1,5 mm
- 2,0 mm
0
5
10
15
20
0,0 0,2 0,4 0,6 0,8 1,0 1,2
I, А
el, m/s
2
3
4
5
1
0,0
0,5
1,0
1,5
2,0
2,5
3,0
3,5
4,0
0,0 0,2 0,4 0,6 0,8 1,0
i, А/сm2
el, m/s
= 1 mm
= 0,5 mm
= 0,25 mm
INNOVATIONS 2025
109
The metal removal rate ΔG during the processing of internal
surfaces of tubular components was evaluated based on the change
in mass per unit surface area per unit time (mass removal rate). To
determine the removal, the sample mass was measured before and
after processing for a duration of 3 min. The dependencies
characterizing the effect of current density on the change in removal
rate for slit nozzle widths of δ = 0.25 mm and δ = 1.0 mm are
shown in fig. 7. The relationships are linear in nature. For
comparison, a similar dependence of removal rate on current
density obtained during EPT using the traditional method
(immersion in electrolyte) for the external surface of a flat sample is
also presented. This processing was carried out under the same
voltage and electrolyte concentration as for the internal surface
treatment. To control the current density, the electrolyte temperature
was varied from 60 to 80 °C in 5 °C increments.
Fig. 7 The influence of current density on the change in removal
rate during surface processing: 1 – δ = 0.5 mm; 2 – δ = 1.0 mm; 3
– processing of the external surface of a flat specimen
Due to the low temperature of the electrolyte, treatment the
external surface by immersion in the electrolyte resulted in a
relatively high current density – ranging from 0.605 to 0.305 A/cm².
However, as the dependencies show, even these values are
significantly lower than those achieved during internal surface
processing using the electrode device. For example, at an electrolyte
temperature of 75 °C, the processing performance for internal
surfaces at δ = 0.25 mm is 1.2 to 3.0 times higher than that of
traditional external surface processing by immersion in the
electrolyte. The dependence of the removal rate on current density
differs from similar dependencies for internal surfaces by its slope.
The dependencies for internal surface processing are more gradual,
which is due to the lower current efficiency during processing with
electrolyte jets compared to traditional EPT of external surfaces by
immersion.
When processing internal surfaces, the metal removal rate
and, consequently, the final surface roughness will be determined
by the movement speed of the electrode device inside the
workpiece. In developing the technological process, this parameter
must be selected individually depending on the condition of the
initial surface and the surface quality requirements of the finished
product. Since the flow rate and flow velocity of the electrolyte
significantly affect the current and the performance of the internal
surface EPT process, it is evident that the resulting surface quality
will also depend on these hydrodynamic characteristics.
Fig. 8 presents the dependencies of changes in the surface
roughness parameter Ra of the internal surface on the average
electrolyte flow velocity from the electrode device, for a slit nozzle
width of δ = 0.25 mm, obtained from processing at electrolyte
temperatures of 75, 80, 85, and 90 °C. The initial surface roughness
was Ra = 0.586 µm. During processing, the electrode device moved
inside the workpiece at a speed of 1 mm/min. The total length of the
processed section was 12 mm. From the presented dependencies, it
is evident that the higher the average electrolyte velocity, the
greater the change in the Ra parameter achieved during processing.
The intensity of surface roughness change varies depending on the
temperature. As the electrolyte temperature increases, the surface
quality declines due to reduced metal removal, which is generally
characteristic of EPT processes.
Fig. 8 The influence of average electrolyte velocity on the change in
Ra surface roughness parameter of the internal surface at various
electrolyte temperatures (δ = 0.25 mm)
5. Conclusion
In the electrolyte-plasma treatment (EPT) of internal surfaces
of tubular products, the electrolyte flow from the annular slit nozzle
of the electrode device onto the treated internal surface corresponds
to the interaction of a flat jet with an obstacle positioned
perpendicular to the flow. The optimal conditions for forming a
vapor-gas shell are created when the jet in the section between the
electrode device and the treated surface is compact, meaning the
flow continuity within it is not disrupted. The height of the section
where stable EPT process flow is possible is independent of the slit
nozzle width. Beyond this section, the electric field decreases to
values where conditions similar to the switching mode arise. This
leads to periodic disruptions in the continuity of the vapor-gas shell
and a transition to the switching mode, accompanied by chaotic
current growth, intensive vapor and gas release, and the formation
of hydrodynamic flows.
For a tube made of corrosion-resistant steel AISI 321 with an
inner diameter of 26.5 mm, it was found that the optimal conditions
for forming the vapor-gas shell are achieved when the slit nozzle
width of the electrode device is from 0.25 to 1.0 mm, and the jet
between the electrode device and the treated surface remains
compact, ensuring flow continuity. The widest range of electrolyte
velocity (el = 0.38–0.78 m/s) that ensures a stable EPT process
without transition to the switching mode with current surges and
electrolyte splashing corresponds to δ = 0.25 mm. With an increase
in el, the current density rises from 1.14 to 3.67 A/cm², which is
due to the increased pressure of the electrolyte jet on the vapor-gas
shell, leading to a reduction in its thickness and, accordingly, the
electrical resistance of the shell. At an electrolyte temperature of 75
°C, the processing productivity for internal surfaces with δ = 0.25
mm is 1.2–3.0 times higher than the productivity of traditional
external surface treatment by immersion in electrolyte.
6. References
1. Aliakseyeu, Y., Korolyov, A., & Bezyazychnaya, A. (2006,
October 19–20). Electrolyte-plasma treatment of metal materials
surfaces. Paper presented at the 14th International Scientific
Conference "CO-MAT-TECH-2006", Trnava, Slovakia, p. 6.
2. Korolyov, A., Bubulis, A., Vėžys, J., Aliakseyeu, Y., Minchenya,
V., Niss, V., & Markin, D. (2021). Electrolytic plasma polishing of
0
1
2
3
4
5
6
7
0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0
G,
mg/(сm2·min)
i, А/сm2
External surface
= 0,25 mm
= 1,0 mm
0,00
0,05
0,10
0,15
0,20
0,25
0,30
0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4
Rа,m
el, m/s
90 С
85 С
80 С
75 С
INNOVATIONS 2025
110
NiTi alloy. Mathematical Models in Engineering, 7(4), 70–80.
https://doi.org/10.21595/mme.2021.22351
3. Aliakseyeu, Y., Bubulis, A., Korolyov, A., Niss, V., &
Kandrotaitė-Janutienė, R. (2021). Plasma electrolyte polishing of
titanium and niobium alloys in low-concentrated salt solution based
electrolyte. Mechanika, 27(1), 88–93.
http://dx.doi.org/10.5755/j02.mech.25044
4. Aliakseyeu, Y. G., Korolyov, A. Yu., & Niss, V. S. (2019).
Electrolytic-plasma polishing of cobalt-chromium alloys for
medical products. Proceedings of the National Academy of
Sciences of Belarus. Physical-Technical Series, 64(3), 296–303.
https://doi.org/10.29235/1561-8358-2019-64-3-296-303
5. Korolyov, A. Yu., Tomilo, V. A., & Niss, V. S. (2023).
Investigation of the features of electrolyte-plasma treatment of inner
pipe surfaces. Mechanical Equipment of Metallurgical Plants, (2)21,
3–16.
6. Cornelsen, M., Deutsch, C., & Seitz, H. (2018). Electrolytic
plasma polishing of pipe inner surfaces. Metals, 8(1), 12.
https://doi.org/10.3390/met8010012
7. Bagaev, S. I. (2015). Device for electrolytic-plasma treatment of
hollow metallic products (Patent BY 10686, published 30.06.2015).
8. Navickaitė, K., Nestler, K., Kain, M., Tosello, G., Calaon, M.,
Pedersen, D. B., Penzel, M., Böttger-Hiller, F., & Zeidler, H.
(2022). Effective polishing of inner surfaces of additive
manufactured inserts for polymer extrusion using Plasma
Electrolytic Polishing. In 18th Rapid.Tech 3D Conference, Erfurt,
Thuringia, Germany.
9. Alekseev, Y. G., Korolyov, A. Yu., Niss, V. S., & Parshuta, A. E.
(2016). Electrolytic-plasma treatment of internal surface in tubular
products. Nauka i Tekhnika = Science & Technique, 15(1), 61–68.
https://doi.org/10.21122/2227-1031-2016-15-1-61-68
10. Sinkevich, Y. V., et al. (2014). Electro-impulse polishing of
alloys based on iron, chromium, and nickel. Minsk: BNTU.
INNOVATIONS 2025
111
Investigation of the local permeability of filter materials by methods with the opposite
direction of local air flows
Aliaksandr Ph. Ilyushchanka 1, 2, Iryna M. Charniak 2, Aliaksei R. Kusin 2, Anastasia A. Astapenko 2, Ruslan A. Kusin 2
State research and production powder metallurgy association1 – Minsk, Republic of Belarus
State Scientific Institution “O.V. Roman Powder Metallurgy Institute” 2 – Minsk, Republic of Belarus
Email address: alexil@mail.belpak.by, irinacharniak@tut.by, 2312444@mail.ru, anasnasiy2012@gmail.com, 19081877@mail.ru.
Abstract: The results of investigations of local permeability of filter materials (PFM) made of powders of tin-phosphor bronze, titanium and
nickel of various fractions are presented by two methods: by one method the local flow was directed to the sample from the measuring head,
and by another method the local flow entered the measuring head after passing through the sample. It is established that both methods
display the uniformity of the distribution of local permeability over the surface of the filter material (filtration surface) and can be used to
assess the permeability of the filter material. The advantages and disadvantages of these methods are described.
KEYWORDS: POWDER FILTER MATERIALS, UNIFORMITY OF LOCAL PERMEABILITY DISTRIBUTION, MEASUREMENT OF
OPPOSITELY DIRECTED LOCAL FLOWS.
1. Introduction
The peculiarity of powder metallurgy is the possibility of
creating powder filter materials (PFM), the operability and
application area of which are determined by the presence of an
interconnected pore system. This pore structure provides them with
such properties as permeability to gases and liquids, filtering
ability, ability to capillary transport of liquid, its retention in the
pores, ability to interfacial interaction, etc. [1]. A significant
characteristic of PFMs, especially when their application is not
associated with the capture of impurities during operation, but is
aimed at ensuring a uniform supply of a certain amount of gas or
liquid from the working surface of the product, is the uniformity of
the local permeability distribution over the filtration area [2]. The
scientific and technical literature describes quite a large number of
methods for estimating the uniformity of local permeability
distribution over the filtration area, based on cutting out control
samples from separate PFM sections [3, 4] or scanning the surface
of samples using thermoanemometers or special measuring heads
[5-10]. However, these methods either require the destruction of the
sample (product), or are complex in hardware design, require
additional calibration and adjustment work, and cannot be operated
in draughts and in the open air, or have significant methodological
limitations [2].
In research works [2, 11] a device for local permeability
determination, simple in construction and hardware design, devoid
of the noted disadvantages, is proposed; the scheme of this device is
shown in Figure 1, a). This device differs from the device described
in the work [9] (Figure 1, b) by the direction of air flow relative to
the measured local area: in the proposed device, the air flow is
directed through the measuring head to the sample, and in the
device (fig. 1, a) – through the sample to the measuring head. The
proposed device allows to significantly simplify the design and
expand technological capabilities when using, allowing you to
study large-sized tubular, flat products and samples, due to the
absence of the need to seal their ends. It should also be noted the
reduction of gas consumption during measuring process and their
possibility of their carrying out not only in the laboratory, but also
in production conditions.
а)
b)
Fig. 1. Schemes of devices for determination of local permeability
according to [2, 11] (a) and [9] (b): 1-pressure vessel, 2-flow meter,
3-flexible hoses, 4-pressure gauge, 5-measuring head, 6-device for
fixing samples or products.
The aim of this work is to compare the results of studies of the
uniform distribution of permeability over the filtration area of samples
with different pore structures, carried out using the devices
schematically presented in Figure 1.
2. Materials and methods of the research
The studies were performed using the measuring head described in
[2], (Figure 2): the measuring head was made using 3-D printing on a
Z-BOLT S300 printer (country of manufacture – Russia); a PET – G
UMKA polymer (country of manufacture – Belarus) was used to make
the head body, the tip – a polymer of the TRU FLEX brand (the
country of origin is Russia) [2].
Fig. 2. Measuring head.
INNOVATIONS 2025
112
The scheme of the measurable sections on the samples also
corresponded to the scheme given in the work in [2] (Figure 3).
Fig. 3. Scheme of the location of the studied samples sites.
To conduct the research, there were used one-layer samples in
the form of a disk with a diameter of 30 mm and a thickness of 3
mm, made of air-sprayed tin-phosphorous bronze and titanium
powders of different fractions obtained by milling a titanium
sponge, a mixture of electrolytic nickel powder with different
content of a pore-forming agent (carbamide), and two-layer
samples of the same dimensions made of finely dispersed
aluminum oxide powder (a thin filter layer 0.3–0.4 mm thick) and a
mixture of electrolytic nickel powder with a pore -forming agent
(substrate). It should be noted that the consolidation of aluminum
oxide particles among themselves and with nickel powder particles
during samples fabrication was provided by using a binder based on
sodium liquid glass.
Permeability coefficients were determined in accordance with
the recommendations of the research work [12], and mathematical
processing of the results of local permeability coefficients
measurements (determination of the coefficient of variation) was
performed according to the accepted method [13].
3. Experimental results and their discussion
The final results of calculations of the measurement results
performed in the sample sections according to the scheme shown in
Figure 3 are presented in the Table 1. Since the intermediate values
of the permeability coefficients in local sections are unsystematic
and uninformative, they are not shown in the table.
Table 1. Results of studies of the uniformity of the permeability
distribution over the sample area
One-layer samples made of tin-phosphor bronze powder
CuSn10P1
Defined parameters
Powder particle sizes, mm
minus
0.4+0.315
minus
0.2+0.1
minus
0.1+0.063
Local
permeability
coefficient
×10-13, m2
By the method
of [2]
1446.7
228.0
99.0
By the method
of [9]
892.8
147.7
66.6
Mean square
deviation
By the method
of [2]
75.40
27.67
11.51
By the method
of [9]
62.49
14.77
7.2
Coefficient
of variation
By the method
of [2]
0.05
0.12
0.12
By the method
of [9]
0,07
0,10
0,11
Permeability coefficient of
the sample, ×10-13, m2
1147.78
140.00
62.45
One-layer titanium powder samples
Defined parameters
Powder particle sizes, mm
minus
0.63+0.306
minus
0.306+0.16
minus
0.16
Local
permeability
coefficient
×10-13, m2
By the method
of [2]
203.4
74.8
53.5
By the method
of [9]
255.7
81.4
53.6
Mean square
deviation
By the method
of [2]
14.63
7.62
5.54
By the method
of [9]
14.36
7.32
5.89
Coefficient
of variation
By the method
of [2]
0.07
0.10
0.10
By the method
of [9]
0.06
0.09
0.11
Permeability coefficient of
the sample, ×10-13, m2
151.57
49.98
28.79
One-layer samples made of a mixture of electrolytic nickel
powder with a pore-forming agent (carbamide)
Defined parameters
The ratio of powder particles and pore
forming agent
1:0.3
1: 0.6
1:0.9
Local
permeability
coefficient
×10-13, m2
By the method
of [2]
108
15.1
16.5
By the method
of [9]
2.4
3.2
4.7
Mean square
deviation
By the method
of [2]
1.62
1.82
2.14
By the method
of [9]
0.36
0.38
0.65
Coefficient
of variation
By the method
of [2]
0.15
0.12
0.13
By the method
of [9]
0.15
0.12
0.14
Permeability coefficient of
the sample, ×10-13, m2
1.29
1.95
2.81
Two-layer samples of a finely dispersed aluminum oxide
powder (a thin filter layer 0,3–0,4 mm thick) and a mixture of
electrolytic nickel powder with a pore-forming agent
(substrate). The measuring head contacts the filter layer.
Defined parameters
Ratio of powder particles and pore
forming agent
1:0.3
1: 0.6
1:0.9
Local
permeability
coefficient
×10-13, m2
By the method
of [2]
10.6
14.9
16.8
By the method
of [9]
3.6
3.9
6.9
Mean square
deviation
By the method
of [2]
0.74
0.89
1.51
By the method
of [9]
0.25
0.27
0.62
Coefficient
of variation
By the method
of [2]
0.07
0.06
0.09
By the method
of [9]
0.07
0.07
0.09
Permeability coefficient of
the sample, ×10-13, m2
1.47
2.25
2.84
Two-layer samples of fine aluminum oxide powder (a thin filter
layer 0.3–0.4 mm thick) and a mixture electrolytic nickel
powder with a pore-forming agent (substrate). The measuring
head contacts the substrate.
Defined parameters
Ratio of powder particles and pore
forming agent
1:0.3
1: 0.6
1:0.9
Local
permeability
By the method
of [2]
9.2
12.2
15.1
INNOVATIONS 2025
113
coefficient
×10-13, m2
By the method
of [9]
4.5
6.8
11.5
Mean square
deviation
By the method
of [2]
0.55
0.61
1.35
By the method
of [9]
0.22
0.34
0.92
Coefficient
of variation
By the method
of [2]
0.06
0.05
0.09
By the method
of [9]
0.05
0.05
0.08
Permeability coefficient of
the sample, ×10-13, m2
2.18
3.01
4.59
In the table, the average value was defined as the arithmetic
mean of a set of values of the local permeability coefficient; the
permeability coefficient was determined by measuring the
permeability of the whole sample (by passing the air flow through
the area of the whole sample).
As a result of the research, it was found that both methods can
be used with a sufficient degree of reliability to assess the
uniformity of permeability distribution over the filtration area: the
variation coefficients values are almost the same. As it could be
assumed, the lowest uniformity of the permeability distribution
over the filtration area is found in samples made of the mixture of
nickel powder with pore-forming agent (both single-layer and
double-layer), which is explained by the difficulties in achieving
uniform distribution of the pore-forming agent in the charge during
traditional mixing. Nevertheless, all samples have good uniformity
of permeability distribution over the filtration area, since the
coefficients of variation determined by known methods [13], are in
the range of 0.05. -.0.15, which allows, according to [13], to refer
the set of their values to homogeneous aggregates.
Similar to the data of the research work [2], the research results
indicate that in all cases the average value of the local permeability
coefficient significantly differs from the measured permeability
coefficients of samples, and, to assess the permeability of products,
it is necessary to introduce correction factors. Obviously, the latter
will be determined by such technological parameters as the size and
shape of powder particles, the presence of a pore-forming agent,
etc. However, this does not significantly affect the values of the
variation coefficient, since measurement errors due to the reasons
of the differences noted above (dispersion of a part of the local flow
due to deviation from the direction perpendicular to the filter
surface towards lower hydraulic resistances) have a systemic
character and equally affect all measurements.
The study of permeability of two-layer samples confirmed the
data given in the research [14] that when liquids flow in
inhomogeneous porous media, the flow asymmetry is observed (the
filtration rate in the direction of the porosity gradient is higher than
in the opposite direction), and the difference in the values of the
permeability coefficient for the studied samples is in the range from
1.34 to 1.62 times, if to compare the coefficient values of the
samples themselves, and from 1.11 to 1.74 times, if to compare the
average values of the local permeability coefficients.
The disadvantages of both methods include the negative impact
on the measurement accuracy of the subjective factor at manual
pressing of the measuring head, while the advantages include the
simplicity of measurements and convenience in operation. In
contrast to [9], the method of local flow direction through the head
to the sample [2], as noted above, is simpler in hardware design and
allows to study large-sized tubular flat products and samples, due to
the absence of the need to seal their ends. Examples of studies of
tubular and flat products are shown in Figure 4.
a)
b)
Fig. 4. The process of measuring the local permeability of tubular (a)
and flat (b) porous powder products.
4. Conclusion
There are presented the results of investigations of local
permeability of one-layer PFMs made of a mixture of elec tin-
phosphorous bronze powder and titanium powders of different
fractions trolytic nickel powder with different content of a pore-
forming agent (carbamide) and of tin-phosphorous bronze and
titanium powders of different fractions, and two-layer samples made
by joint pressing of finely dispersed aluminium oxide powder (a thin
filter layer 0.3–0.4 mm thick) and a mixture of electrolytic nickel
powder with a pore-forming agent (substrate). The investigations were
carried out by measuring local air flows using two methods: according
to one method, the local flow was directed to the sample from the
measuring head, in the other way the local flow entered the measuring
head after passing through the sample. It is found that both methods
reflect well the uniformity of local permeability distribution over the
surface of the filter material (filtration surface). At the same time, the
average value of the local permeability coefficient significantly differs
from the real values of permeability coefficients of the samples, and to
estimate the permeability of products, it is necessary to introduce
correction factors. It is shown that the flow asymmetry is observed
during the flow of liquids in two-layer materials.
5. Reference / Literature
1. P. Vityaz, V. Kaptsevich, V. Sheleg. Porous Powder Materials and
Products from Them (Minsk, Vysheyshaya Shkola) (1987). Russia.
2. A. Ilyushchanka, I. Charniak, A. Kusin, A. Astapenka, R. Kusin.
Research of the local powder materials permeability. Intern. Scien. J.
Nonequilibrium phase transformations. Sofia. Bulgaria. Year X. Is. 1.
(2024). 14-17.
3. A. Ph. Tretyakov. Practical application of research results of
technological processes of forming solid-phase wire joints in a
controlled gas environment in the creation of sheet porous mesh
materials with specified properties. Engineering J.: Science and
Innovations. 7. (2022). 1-17.
4. S.A. Oglezneva S.A. Material science and technologies of modern
and perspective materials (Perm, PNRPU, 2012). Russia.
INNOVATIONS 2025
114
5. A.s. 1702254 USSR. Device for determining the local gas
permeability of porous materials, B.V. Evstegneev, V.M. Ivanov,
T.Yu. Rodivilina, A.G. Kunitsyn, B.M. Volpe. 1991.
6. V.M. Kaptsevich, V.K. Sheleg, and L.P. Pilinevich. Method for
monitoring the local permeability of porous powder materials using
a thermoanemometer. Powder metallurgy. 7. (1987). 60 63.
(Kiev).
7. A.s. 1746259 USSR. Device for determining the local
permeability of porous products, V.E. Matsera, P.A. Kornienk, and
O.I. Derecha. 1992.
8. A.s. 1580227 USSR. Device for determining the local
permeability of a material, A.I. Golyakov, V.V. Smirensky,
V.E. Killich and V.V. Makartsev. 1990.
9. A.s. 735972 USSR. Device for measuring the local permeability
of porous materials, P.A. Vityaz, V.K. Sheleg, S.V. Popko and
V.M. Kaptsevich, 1980.
10. A.s. 735972 USSR. Measuring head for measuring local
permeability. V.A. Shabarov, E.I. Tyutyunshchikova. No. 44. BI.
1990.
11. BY 13543. Device for determining the local permeability of
porous products, A.Ph. Ilyushchanka, R.A. Kusin, I.N. Charniak,
A.R. Kusin and A.A. Astapenko. 2024.
12. Sintered permeable materials. Determination of the liquid
permeability. GOST 25283-93. Publishing House of Standards.
Minsk. (1993). Russia.
13. Svechnikova V.V. Statistics: textbook. Orsk: OGTI Publishing
House. 2012.
14. Siraev R.R. Liquid filtration in a heterogeneous porous
medium. Fundamental research. 11. (2013). 451-455.
INNOVATIONS 2025
115
Recent Innovations in Metal Processing with the Use of Lasers 2025
Nikolaos Papageorgiou
Ship propulsion plants Department of the Faculty of Engineering Nikola Vaptsarov Naval Academy
Abstract: Laser technology has transformed metal processing, offering unmatched precision, efficiency, and adaptability across
industries such as automotive, aerospace, and electronics. By focusing high-energy laser beams to cut, weld, or engrave metals, these
systems deliver superior results compared to traditional methods. In 2025, advancements in laser sources, artificial intelligence (AI)
integration, and applications in electric vehicle (EV) manufacturing are driving significant progress. This presentation explores these
innovations, focusing on high-power fiber lasers, AI-enhanced process control, and the growing role of laser welding in EV battery
production. It also examines market trends and challenges, drawing on the latest 2025 research to highlight the future potential of
laser-based metal processing.
1. Introduction to Laser-Based Metal Processing
Laser technology is a cornerstone of modern metal processing,
enabling precise joining, cutting, and surface modification with
minimal material waste. Laser welding, in particular, uses a
concentrated beam to fuse metals, producing strong welds with
reduced thermal distortion. This method is highly valued in
industries requiring intricate, high-quality work, such as
automotive manufacturing and medical device production. Its
non-contact nature and ability to automate processes make it
ideal for high-volume production.
In 2025, the field is evolving rapidly due to advancements in
laser systems and smart technologies. The rise of electric
vehicles, projected to reach 15 million units in global sales by
2025 [1], has spurred demand for laser welding in battery
manufacturing. Additionally, the integration of AI and the
Internet of Things (IoT) is enhancing process efficiency and
quality control. This presentation delves into these
developments, offering insights into how laser technology is
shaping the future of metal processing.
2. Advancements in Laser Sources
Fiber Lasers: The Industry Standard
Fiber lasers have become the preferred choice for metal
processing in 2025 due to their high efficiency, compact
design, and ability to deliver precise, high-energy beams.
These lasers use optical fibers doped with elements like
ytterbium to produce a focused beam, enabling deep welds and
fast processing speeds. Their low maintenance and long
lifespan (up to 100,000 hours) make them cost-effective for
industrial applications.
Recent innovations include high-power fiber laser systems
capable of welding thick metals. For example, IPG Photonics’
25-kW fiber laser, introduced in 2025, achieves weld depths of
25 mm in a single pass at 2.5 meters per minute, ideal for
heavy industries like shipbuilding [2]. Additionally, portable
handheld fiber laser welders, such as DenaliWeld’s 1500W
model, weigh only 48.5 lbs and offer dual welding and seam-
cleaning functions, making them accessible for small-scale
operations [3].
Pic 1. Fiber Laser assembly 2025
Emerging Laser Technologies
Beyond fiber lasers, other systems are gaining traction. Blue
laser systems, operating at 450 nm, are increasingly used for
welding highly reflective metals like copper, which absorb
shorter wavelengths more effectively. In 2025, Coherent
introduced a 1-kW blue laser system that achieves defect-free
copper welds at speeds of 200 mm/s, critical for electronics
and EV battery applications [4]. These advancements suggest a
trend toward specialized laser sources tailored to specific
materials and applications.
3. AI and Machine Learning in Laser
Processing
The integration of AI and machine learning is revolutionizing
laser welding by enabling real-time process optimization and
quality control. AI-powered systems analyze data from sensors
and high-speed cameras to adjust parameters like power and
beam focus during welding, ensuring consistent results. For
instance, Precitec’s Laser Welding Monitor with AI predicts
weld strength by analyzing emissions, achieving 98% accuracy
in detecting defects in automotive components [5].
Pic 2. Machine Learning cycle
Machine learning also supports predictive maintenance,
identifying potential equipment issues before they cause
downtime. A 2025 study demonstrated that AI-driven systems
could reduce maintenance costs by 20% in high-volume
welding operations [9]. Additionally, research into self-
learning laser systems, combining deep neural networks and
INNOVATIONS 2025
116
reinforcement learning, shows promise for autonomous
welding processes that improve without human intervention
[6].
4. Laser Welding in Electric Vehicle
Manufacturing
The electric vehicle (EV) industry is a key driver of laser
welding innovation, particularly in battery production. EV
batteries require thousands of precise welds to connect tabs,
busbars, and cells, often involving dissimilar metals like
aluminum and copper. Laser welding’s ability to minimize
heat-affected zones and join these materials effectively makes
it ideal for this application.
Pic 3. Frame Laser Welding
In 2025, companies like Laserax offer specialized welding
machines with up to 6 kW of power, enabling high-speed
production of battery packs for cylindrical, prismatic, and
pouch cells [7]. These systems can weld dissimilar metals with
minimal intermetallic compound formation, achieving
electrical resistance as low as 0.1 mΩ. The growing EV
market, expected to reach 23 million units by 2030,
underscores the critical role of laser welding in meeting
production demands [1].
5. Other Metal Processing Applications
Beyond welding, lasers are advancing other metal processing
techniques. Laser cutting, using high-power fiber lasers,
achieves clean, precise cuts in metals up to 50 mm thick, with
systems like those from IPG Photonics offering cutting speeds
of 10 m/min for steel [2]. Laser cleaning, another emerging
application, uses low-power lasers (50W–200W) to remove
contaminants from metal surfaces, improving weld quality and
reducing preparation time [7].
These applications are particularly valuable in aerospace,
where precision and material integrity are paramount, and in
electronics, where lasers enable micro-welding of delicate
components. The versatility of laser systems, capable of
welding, cutting, and cleaning, enhances their value in
integrated manufacturing processes.
6. Market Growth and Future Outlook
The global laser welding market is experiencing robust growth,
driven by demand from the EV sector and advancements in
automation. Industry projections estimate a compound annual
growth rate (CAGR) of 4.4% from 2025 to 2035, with the
market expected to reach $2.79 billion by 2029 [8]. The
integration of AI and IoT technologies is likely to further
accelerate this growth by enabling smarter, more efficient
production lines.
Pic 4. Huge Market Growth expected by 2034
However, challenges remain, including high initial costs,
which can exceed $8,000 for industrial handheld welders [10].
Training programs and user-friendly systems are being
developed to address these barriers, making laser technology
more accessible to small and medium enterprises.
7. Challenges and Solutions
Despite its advantages, laser welding faces hurdles such as
high equipment costs and the need for skilled operators. To
mitigate these, manufacturers are introducing cost-effective
portable welders and automated systems requiring minimal
human intervention. For example, DenaliWeld’s modular plug-
and-play design simplifies integration with robotic arms,
reducing setup time [3]. Additionally, AI-driven training tools
are helping workers adapt to advanced systems, ensuring
broader adoption across industries.
8. Conclusion
Laser technology is reshaping metal processing, offering
precision, efficiency, and versatility that traditional methods
cannot match. Innovations in fiber and blue laser systems,
coupled with AI integration, are expanding the capabilities of
laser welding, particularly in high-demand sectors like electric
vehicle manufacturing. As the market grows and challenges are
addressed, laser-based metal processing is poised to play a
central role in the future of industrial production, driving
advancements in quality, sustainability, and automation.
INNOVATIONS 2025
117
Table: Comparison of Laser Processing Techniques
Technique
Key Features
Applications
Advantages
Challenges
Fiber Laser
Welding
High efficiency,
100,000-hour
lifespan
Automotive, EV
batteries,
aerospace
Precise, low
maintenance,
versatile
High initial
cost
Blue Laser
Welding
450 nm
wavelength, high
absorption by
copper
Electronics,
battery
manufacturing
Defect-free welds,
fast speeds
Limited to
specific
materials
Laser
Cutting
High-power
beams, up to 50
mm thickness
Aerospace,
automotive,
manufacturing
Clean cuts, high
speed
Requires
precise
calibration
Laser
Cleaning
Low-power
(50W–200W),
removes
contaminants
Surface
preparation, weld
enhancement
Non-abrasive,
environmentally
friendly
Limited to
surface
applications
References
[1]. SNE Research. (2025). Global EV Sales Projections
[2]. IPG Photonics. (2025). High-Power Fiber Laser
Systems
[3]. DenaliWeld. (2024). Portable Laser Welder
[4]. Coherent. (2025). Blue Laser Systems for Copper
Welding
[5]. Precitec. (2025). Laser Welding Monitor with AI
[6]. ScienceDirect. (2024). Intelligent Laser Welding
Architecture
[7]. Laserax. (2025). EV Battery Welding Solutions
[8]. Fortune Business Insights. (2024). Laser Welding
Market Size
[9]. The AI Journal. (2025). Future of Laser Welding in
Automation
[10]. STYLECNC. (2025). Portable Handheld Laser
Welding Machine
INNOVATIONS 2025
118
Антиброкариани на четириъгълник
Станислав Стефанов1, Хаим Хаимов2
1Висше транспортно училище „Тодор Каблешков“, 2Варна, ул. Братя Шкорпил 16
stanislav.toshkov@abv.bg, , haim.haimov@abv.bg
Резюме. В статията ще разгледаме интересните свойства на една двойка забележителни точки в равнината на произволен
изпъкнал четириъгълник, наречени Антиброкариани. Те спадат към така наречените Брокарови точки на четириъгълника,
разгледани в (Tabov et al. 2025) и характеризиращи се с това, че образуват с двойка срещуположни страни или с двойката диагонали
на четириъгълника подобни триъгълници. Антиброкарианите образуват подобни триъгълници с двойка срещуположни страни на
четириъгълника (при определена наредба на върховете на триъгълниците). Те са тясно свързани с пресечните точки на
симетралите на двойките срещуположни страни на четириъгълника, които за краткост ще наричаме негови симетрални точки.
Оказва се, че правата, определена от двете Антиброкариани е успоредна на правата, определена от двете симетрални точки.
Освен със симетралните точки, Антиброкарианите са свързани и с друга двойка забележителни точки в четириъгълника, наречени
Брокариани. Дефиницията на Антиброкарианите е в известен смисъл противоположна на тази на Брокарианите, което оправдава
даденото им име. В четириъгълник с равни диагонали, Антиброкарианите съвпадат с Брокарианите. Интересни свойства има
четириъгълникът с върхове проекциите на едната Антиброкариана върху двойка срещуположни страни и проекциите на другата
върху оставащата двойка срещуположни страни в изпъкналия четириъгълник. Както ще видим още, всяка от Антиброкарианите
образува с двата диагонала на четириъгълника равнолицеви триъгълници.
КЛЮЧОВИ ДУМИ: ИЗПЪКНАЛ ЧЕТИРИЪГЪЛНИК, ЗАБЕЛЕЖИТЕЛНИ ТОЧКИ, ВРЪЗКИ МЕЖДУ ТЯХ, СВОЙСТВА,
ИЗОБРАЖЕНИЯ, ЗАБЕЛЕЖИТЕЛНИ ПРАВИ, ПЕДАЛЕН ЧЕТИРИЪГЪЛНИК.
Antibrocardians of a quadrangle
Stanislav Stefanov1, Haim Haimov2
1Todor Kableshkov University of Transport, 2Varna, 16 Bratya Shkorpil Str.,
stanislav.toshkov@abv.bg, haim.haimov@abv.bg
Abstract: In this article, we will examine the interesting properties of a pair of notable points in the plane of an arbitrary convex quadrilateral,
called Antibrocardians. They belong to the so-called Brocardian points of the quadrilateral, considered in (Tabov et al. 2025) and
characterized by the fact that they form similar triangles with a pair of opposite sides or with the pair of diagonals of the quadrilateral.
Antibrocardians form similar triangles with a pair of opposite sides of the quadrilateral (in a certain arrangement of the triangle vertices).
They are closely related to the intersections of the bisectors of the pairs of opposite sides of the quadrilateral, which for brevity we will call its
bisector points. It turns out that the line defined by the two Antibrocardians is parallel to the line defined by the two bisector points. In addition
to the bisector points, the Antibrocardians are also connected to another pair of notable points in the quadrangle, called Brocardians. The
definition of the Antibrocardians is in some ways the opposite of that of the Brocardians, which justifies their given name. In a quadrilateral
with equal diagonals, the Antibrocardians coincide with the Brocardians. The quadrilateral with vertices the projections of the one
Antibrocardian onto a pair of opposite sides and the projections of the other onto the remaining pair of opposite sides in the convex
quadrilateral has interesting properties. As we will see too, each of the Antibrocardians forms with the two diagonals of the quadrilateral,
triangles of equal areas.
KEYWORDS: CONVEX QUADRILATERAL, NOTABLE POINTS, RELATIONS BETWEEN THEM, PROPERTIES, IMAGES, NOTABLE
LINES, PEDAL QUADRILATERAL.
I. Въведение. Както споменахме в резюмето,
Антиброкарианите имат връзка с така наречените Брокариани на
четириъгълника. Подобно на Брокарианите, всяка
Антиброкариана образува със съответна двойка срещуположни
страни подобни триъгълници, само че при противоположна
наредба на върховете на триъгълниците (в сравнение с
наредбата при Брокарианите). Антиброкарианите лежат заедно
с Брокарианите на определени прави през пресечните точки на
продълженията на двойките срещуположни страни. В
четириъгълник с равни диагонали те съвпадат с Брокарианите.
Както споменахме, Антиброкарианите са тясно свързани и с
пресечните точки на симетралите на двойките срещуположни
страни, които нарекохме симетрални точки на четириъгълника.
Всяка от двете Антиброкариани има за свой образ при едно
важно изображение в равнината на четириъгълника, наречено
инверсна симетрия на Микел, една от симетралните точки на
четириъгълника. Заедно с другата симетрална точка, тя лежи на
права, минаваща през известната точка на Микел на
четириъгълника. Интересни свойства има четириъгълникът с
върхове ортогоналните проекции на едната Антиброкариана
върху съответна двойка срещуположни страни в изходния
четириъгълник и проекциите на другата – върху другата двойка
срещуположни страни в този четириъгълник. Правата,
определена от средите на диагоналите на полученият по
описания начин четириъгълник е перпендикулярна на правата,
определена от средите на диагоналите на изходния
четириъгълник и съвпада с правата, определена от двете
Антиброкариани.
II. Дефиниция на Антиброкарианите.
Преди да приведем точната дефиниция на
Антиброкарианите, да припомним дефиницията на
Брокарианите:
„Нека ABCD е изпъкнал четириъгълник. Точката
1
K
от вътрешността на ъгъла между правите AB и CD, в който лежи
четириъгълникът, такава, че са изпълнени равенствата
1 1
K AB K CD
и 1 1
K BA K DC
, се нарича
Брокариана, съответна на страните AB и DC.“ (фиг.
1
a
)
Фиг. 1а
INNOVATIONS 2025
119
Фиг. 1б
Условието 1 1
K AB K CD
и 1 1
K BA K DC
в
тази дефиниция очевидно е еквивалентно с условието
триъгълниците
1
AK B
и 1
CK D
да са подобни, като на върха
А при подобието съответства върхът С, а на върха B – върхът D.
Като разменим местата на върховете С и D при подобието, т.е.
поставим условието на върха А да съответства върхът D, а на
върха B – върхът С, то получаваме дефиницията на съответна
Антиброкариана на четириъгълника:
Определение 1. Точката
1
K
от вътрешността на ъгъла
между правите AB и DC, в който лежи четириъгълникът ABCD,
такава, че 1 1
AK B DK C
(при което на върха А
съответства върхът D, а на върха B – върхът С), се нарича
Антиброкариана на ABCD, съответна на страните AB и DC. (фиг.
1
б
)
Забележка 1. За Брокарианата
1
K
, съответна на
страните AB и DC, са подобни не само триъгълниците
1
AK B
и
1
CK D
, но и триъгълниците 1
AK C
и 1
BK D
(фиг.
1
a
)
(Haimov & Stefanov 2023).
III. Предварителни означения.
Преди да докажем съществуването на
Антиброкарианите, да въведем следните означения в
разглеждания четириъгълник ABCD, за който (за да са коректни
твърденията в изложението по-нататък) ще предполадаме, че
няма двойка успоредни или равни срещуположни страни:
Означаваме пресечната точка на продълженията на
страните AD и BC с U, а тази на продълженията на страните AB
и DC – с V (фиг. 2) (за определеност ще считаме, че върхът C
лежи между точките U и B и между точките D и V). Дължините
на страните AB, BC, CD и DA означаваме съответно с a, b, c и d,
дължините на диагоналите AC и BD – съответно с m и n, а
мерките на ъглите при върховете A, B, C и D – съответно с
,
,
и
. Средите на страните AB, BC, CD и DA ще бележим
съответно с
1 2 3
, ,
E E E
и
4
E
, а средите на диагоналите AC и
BD – съответно с E и F. Въвеждаме и означенията
,
AUB
AVD
.
Фиг. 2
IV. Брокариани, точки на Микел и Лемоан и
симетрални точки
1
P
и
2
P
на четириъгълника.
Дефиницията на Брокарианите
1
K
и
2
K
приведохме
по-горе. Сега ще приведем дефиницията на точката на Микел:
Определение 2. Точката на Микел на четириъгълника
ABCD наричаме общата точка M на описаните окръжности
около триъглниците AUB, CUD, ADV и BCV (фиг. 3).
Фиг. 3
Фиг. 4
Точката на Микел притежава следните свойства:
Свойство 1. Триъгълниците ABM и DCM са подобни и
триъгълниците ADM и BCM също са подобни (фиг. 3) (Haimov
& Stefanov 2023).
Определение 3. Окръжността (
1
c
), определена от
средите
1
E
и
3
E
съответно на страните AB и DC и пресечната
точка V на техните продължения се нарича Брокарова
окръжност, съответна на страните AB и DC.
Свойство 2. Точката на Микел M и Брокарианата
1
K
,
съответна на страните AB и CD лежат на Брокаровата окръжност
(
1
c
), съответна на тези страни (фиг. 3) (Haimov & Stefanov
2023).
Свойство 3. Ъглите AMC и BMD имат обща
ъглополовяща g, която се нарича ос на Микел на
четириъгълника ABCD (фиг. 4) (Haimov & Stefanov 2023).
Свойство 4. Изпълнени са равенствата
2
AM CM BM DM UM VM r
. Величината
2
r
се нарича константа на Микел на четириъгълника ABCD
(фиг. 4) (Haimov & Stefanov 2023).
Определение 4. Точката L от вътрешността на
четириъгълника ABCD, за разстоянията
1 2 3
, ,
h h h
и
4
h
от
която съответно до страните AB, BC, CD и DA са изпълнени
равенствата
3
1
h
h
AB CD
и
2 4
h h
BC AD
, се нарича точка на
Лемоан на четириъгълника (фиг. 5).
INNOVATIONS 2025
120
Фиг. 5
Фиг. 6
Определение 5. Правата l през пресечната точка V на
продълженията на страните AB и DC, разстоянията от точките
на която до правите AB и DC са пропорционални на дължините
на страните AB и DC и която лежи във вътрешността на ъгъл
AVD, се нарича симедиана на четириъгълника ABCD през
точката V (фиг. 5).
Свойство 5. Точката L на Лемоан и Брокарианата
1
K
,
съответна на страните AB и DC лежат на симедианата l през
точката V (фиг. 5) (Haimov & Stefanov 2023).
Определение 6. Пресечната точка
1
P
на симетралите на
страните AB и DC ще наричаме симетрална точка, съответна на
AB и CD. Пресечната точка
2
P
на симетралите на страните AD
и BC ще наричаме симетрална точка, съответна на страните AD
и BC (фиг. 6).
V. Инверсна симетрия на Микел.
Преди да дефинираме изображението инверсна
симетрия на Микел, ще припомним дефиницията на
изображението инверсна симетрия в равнината.
Определение 7. Нека M е точка в равнината и g е права
през M, а
2
r
е положително число. Композицията от
симетрията спрямо правата g и инверсията
I
с полюс M и
степен
2
r
се нарича инверсна симетрия в равнината с полюс M,
ос – g и степен
2
r
(фиг. 7). Последната се бележи с
2
; ;
g MI g
r
.
Фиг. 7
Фиг. 8
Определение 8. Инверсната симетрия
2
; ;
g MI g
r
в равнината на четириъгълник ABCD с
полюс точката M на Микел, ос – оста g на Микел и степен –
константата
2
r
на Микел, се нарича инверсна симетрия на
Микел спрямо ABCD (фиг. 4).
Свойство 6. Ако точката Y е образът на точката X при
инверсната симетрия на Микел, то е изпълнено равенството:
2
XM
r
YM
(Haimov & Stefanov 2023).
Свойство 7. При инверсната симетрия
I g
на Микел
права през полюса M се изобразява в права през M (Haimov &
Stefanov 2023).
Свойство 8. Ако точката Y е образът на точката X при
инверсната симетрия на Микел спрямо четириъгълника ABCD,
то разстоянията от Y до върховете му са свързани с разстоянията
от X до тези върхове чрез равенствата (фиг. 8) (Haimov &
Stefanov 2023):
, , , .
AY CX d BY DX a CY AX b DY BX a
BY DX b CY AX c DY BX d AY CX c
Свойство 9. Ако точката Y е образ на точката X при
инверсната симетрия на Микел спрямо четириъгълника ABCD,
то са изпълнени равенствата (фиг. 8):
, .
AYB DXC AXD BYC
(Haimov & Stefanov 2023).
VI. Съществуване на Антиброкарианите.
Вече можем да докажем съществуването и
единствеността на Антиброкарианите.
Фиг. 9
Теорема 1. (за съществуването на
Антиброкарианите). Нека ABCD е изпъкнал четириъгълник, в
който
AB CD
и AD и BC не са успоредни.
Антиброкарианата
1
K
, съответна на страните AB и CD,
съществува и е еднозначно определена, и образът й при
инверсната симетрия на Микел е симетралната точка
2
P
,
съответна на страните AD и BC (фиг. 9).
Доказателство: Лесно се съобразява, че при
направените ограничения за четириъгълника ABCD,
симетралната точка
2
P
, съответна на страните AD и BC,
съществува и е различна от полюса M на инверсната симетрия
на Микел. Означаваме образа на точката
2
P
, при инверсната
симетрия на Микел с
2
P
(фиг. 9). Нека a и c са дължините
съответно на страните AB и CD. От свойство 8 имаме, че е
изпълнено равенството: 2 2
2
2
AP CP
a
BP c
DP
. Понеже точката
2
P
лежи на симетралата на страната BC и следователно
2 2
CP BP
, оттук получаваме: 2
2
AP
a
c
DP
. Аналогично се
INNOVATIONS 2025
121
доказва, че 2
2
BP
a
c
CP
. Можем да заключим, че
2 2
AP B DP C
(при посочената наредба на върховете на
триъгълниците). Получихме, че точката
2
P
отговаря на
условието от определение 1, т.е. че Антиброкарианата
1
K
,
съответна на страните AB и CD, съществува. По обратния път се
доказва, че ако една точка
1
K
отговаря на условието от
определение 1, то тя е образ на симетралната точка
2
P
при
инверсната симетрия на Микел. Следователно
Антиброкарианата
1
K
е определена еднозначно и образът й при
инверсната симетрия на Микел е симетралната точка
2
P
.
Забележка 2. Аналогично се доказва, че образът при
инверсната симетрия на Микел на Антиброкарианата
2
K
,
съответна на страните AD и BC, съвпада със симетралната точка
1
P
, съответна на страните AB и CD.
VII. Свойства на Антиброкарианите.
Антиброкарианите се характеризират с интересни и
разнообразни свойства:
Свойство 1К. В четириъгълник с равни диагонали AC и
BD, Антиброкарианите съвпадат с Брокарианите (Haimov &
Stefanov 2023).
Доказателство:
Фиг. 10
Нека ABCD е четириъгълник с равни диагонали AC и BD (фиг.
10). Ако
1
K
е Брокарианата, съответна на страните AB и DC, то
1 1
AK C BK D
и 1 1
AK B CK D
(съгласно
определението на Брокарианите и забележка 1). Тогава
1
1
AK
AC
BK BD
и понеже
AC BD
(по условие), то
1 1
AK BK
. Аналогично се доказва, че
1 1
CK DK
.
Получихме, че триъгълниците
1
AK B
и 1
CK D
са
равнобедрени. Тогава те са подобни не само при наредба на
върховете, съответна за Брокарианата
1
K
, но и при
противоположна наредба на върховете, откъдето следва, че
1
K
е и Антиброкариана на четириъгълника ABCD.
Свойство 2К. Нека
1
K
е Антиброкарианата на
четириъгълника ABCD, съответна на страните AB и CD, а
2
P
е
симетралната точка, съответна на страните AD и BC. Изпълнени
са равенствата (фиг. 9).
(1) 1 1 2
1 1 2
1
,
2
1
.
2
ABK DCK AP D
BAK CDK BP C
Доказателство: Точките
1
K
и
2
P
са образи една на
друга при инверсната симетрия на Микел спрямо
четириъгълника ABCD (по теорема 1). Тогава, ако
AVD
, от свойство 9 имаме: 2 1
AP D BK C
.
Оттук и от определение 1 получаваме последователно:
1 1 1 1
1 1 2
1
2
1 1 1
.
2 2 2
ABK DCK ABK DCK
BK C BVC BK C AP D
Доказахме първото от равенства (1). Аналогично се доказва и
второто равенство.
Свойство 3К. Ако
1
K
е Антиброкарианата на
четириъгълника ABCD, съответна на страните AB и CD, то
триъгълниците 1
AK C
и 1
BK D
са равнолицеви.
Фиг. 11
Доказателство: Имаме 1 1
AK B DK C
(по
определение 1) (фиг. 11), откъдето следва, че
1 1
1 1
AK BK
DK CK
и
1 1
AK B DK C
. Тогава
1 1 1 1
AK CK BK DK
и
1 1 1
1 1 1
,
AK C AK B BK C
DK C BK C DK B
т.е
1 1
AK C DK B
. С помощта на тези равенства
получаваме последователно:
1
1
1 1 1
1 1 1
1sin
2
1
sin .
2
AK C
BK D
S AK CK AK C
BK DK DK B S
С това свойството е доказано.
Свойство 4К. Антиброкарианата
1
K
, съответна на
страните AB и CD на четириъгълника ABCD, лежи на
симедианата през точката V (фиг. 12).
Фиг. 12
INNOVATIONS 2025
122
Доказателство: Симедианата на четириъгълника ABCD
през точката V е геометричното място на точките в ъгъл AVD,
разстоянията от които до раменете VA и VD на ъгъла са
пропорционални на дължините на страните AB и CD (по
определение 4). Означаваме ортогоналните проекции на точката
1
K
върху правите AB и DC съответно с M и P. Понеже
1 1
AK B DK C
(по определение 1), а 1
K M
и 1
K P
са
съответни височини в тези триъгълници, то 1
1
K M
AB
K P CD
.
Това означава, че точката
1
K
принадлежи на въпросното
геометрично място от точки, т.е. че Антиброкарианата
1
K
лежи
на симедианата през V.
Сега ще докажем едно свойство на Антиброкарианите
на четириъгълника, от което в частност се получава прост метод
за построяването им:
Свойство 5К. Нека ABCD е изпъкнал четириъгълник и
1
K
е Антиброкарианата му, съответна на страните AB и CD, а
0
M
е точката му на Микел. Ако продълженията на страните AB
и DC се пресичат в точката V, то е изпълнено равенството
1 0
90
K M V
(фиг. 13).
Доказателство: Означаваме ортогоналните проекции
на Антиброкарианата
1
K
върху правите AB и DC съответно с М
и Р. Триъгълниците
1
ABK
и
1
DCK
са подобни (по
определение 1) и точките М и Р са петите на съответни техни
височини. Лесно се съобразява, че тогава
AM DP
BM CP
.
Триъгълниците
0
ABM
и
0
DCM
също са подобни (съгласно
свойство 1) и от получената пропорция следва, че точките М и Р
са съответни при подобието. Тогава и триъгълниците
0
MBM
и
0
PCM
са подобни, откъдето получаваме:
0 0
BMM CPM
, т.е.
0 0
VMM VPM
.
Фиг. 13
Можем да заключим, че точките М и Р лежат на дъга от
окръжност с краища точките
0
M
и V. Изказано по друг начин,
точката
0
M
лежи на окръжността
c
, определена от точките
М, Р и V. Но понеже 1 1
90
K MV K PV
,
окръжността
c
, определена от точките М, Р и V, има за
диаметър отсечката
1
VK
. Тъй като
0
M c
(по
доказаното), то 1 0
90
K M V
.
С помощта на свойства 4K и 5K се получава следният
метод за построяването на Антиброкарианите:
Нека ABCD е изпъкнал четириъгълник и продълженията
на страните му AB и DC се пресичат в точката V (фиг. 13).
Построението на Антиброкарианата
1
K
извършваме в следния
ред:
1) Построяваме точката на Микел
0
M
като втора обща
точка на описаните околo
AVD
и
BVC
окръжности;
2) От точката
0
M
издигаме перпендикуляр l към правата
0
M V
;
3) Построяваме симедианата g на ABCD през точката
V
;
4) Определяме пресечната точка
1
K
на перпендикуляра l
и симедианата g.
Тя е търсената Антиброкариана, съответна на страните AB и DC.
Доказателството следва непосредствено от свойства 4K и 5K.
Сега ще се спрем на една интересна връзка между
Антиброкарианите, симетралните точки на четириъгълника и
точката на Микел:
Свойство 6K. Нека
1
K
и
1
P
са Антиброкарианата и
симетралната точка на четириъгълника, съответни на страните
му AB и CD, а
2
K
и
2
P
– Антиброкарианата и симетралната
точка, съответни на страните AD и BC. Ако M е точката на
Микел на ABCD, то точките
1
P
,
1
K
и M лежат на една права,
точките
2
P
,
2
K
и M също лежат на една права и
1 2 1 2
PP K K
. (фиг. 14)
Фиг. 14
Доказателство: Означаваме средите на страните AB и CD
съответно с
1
E
и
3
E
, а пресечната точка на продълженията им
– с V (фиг. 14). Понеже 1 1 1 3
90
PE V PE V
, то
четириъгълникът
1 1 3
PEVE
е вписан в окръжността
c
с
диаметър
1
VP
. Тъй като
1
E
,
3
E
, V
c
, то
c
е
Брокаровата окръжност, съответна на страните AB и CD (по
определение 3) и следователно точката на Микел M лежи върху
c
(от свойство 2). Тогава 1
90
PMV
. Но и
1
90
K MV
(от свойство 5K). Следователно точките
1
P
,
1
K
и M лежат на една права, перпендикулярна на MV. Понеже
съгласно теорема 1 и забележката след нея, точките
2
K
и
2
P
са образи съответно на точките
1
P
и
1
K
при инверсната
симетрия на Микел
2
; ;
g MI g
r
, имаща за полюс –
INNOVATIONS 2025
123
точката М на Микел, то точките
2
K
,
2
P
и М също лежат на
една права (от свойство 7). От
1 2
( )
I g P K
и
2 1
( )
I g P K
(по теорема 1) имаме още:
2
1 2
PM K M r
и
2
2 1
P M K M r
(от свойство 6).
Следователно 1 2 2 1
PM K M P M K M
, т.е.
1 2
1 2
PM P M
K M K M
, откъдето и получаваме, че
1 2 1 2
PP K K
.
Преди да изложим следващите свойства на
Антиброкарианите, ще дефинираме понятието Антиброкаров
педален четириъгълник:
Нека
1
K
е Антиброкарианата на четириъгълника ABCD,
съответна на страните му AB и CD, и М и Р са съответно
ортогоналните й проекции върху правите AB и CD (фиг. 15).
Триъгълниците
1
ABK
и
1
DCK
са подобни (по определение
1), и 1
K M
и 1
K P
са съответни височини в тях, следователно
1 1
1
K P K M
k
CD AB
, където
1
k
е някакво положително число.
Аналогично ако Q и N са ортогоналните проекции на
Антиброкарианата
2
K
върху страните AD и BC, то са
изпълнени равенствата 2 2
2
K Q K N
k
AD BC
, където
2
k
е
друго положително число.
Фиг. 15
Фиг. 16
Определение 1К. Числата
1
k
и
2
k
от горните равенства
ще наричаме Брокарови коефициенти, съответни на
Антиброкарианите
1
K
и
2
K
.
Определение 2К. Нека
1
K
е Антиброкарианата на
четириъгълника ABCD, съответна на страните AB и CD, а
2
K
е
Антиброкарианата му, съответна на страните AD и BC (фиг. 15).
Ако М и Р са ортогоналните проекции на точката
1
K
, съответно
върху правите AB и CD, а N и Q – ортогоналните проекции на
точката
2
K
, съответно върху правите BC и AD, то
четириъгълникът МNРQ ще наричаме Антиброкаров педален
четириъгълник.
Ще се спрем на две интересни свойства на
Антиброкаровия педален четириъгълник. Предварително ще
докажем следната:
Лема 1. Нека
1
K
е Антиброкарианата на
четириъгълника ABCD, съответна на страните AB и CD, и М и Р
са ортогоналните й проекции върху правите AB и CD. Нека още
1
k
е съответният на
1
K
Брокаров коефициент. Ако средите на
страните BC и AD са съответно
2
E
и
4
E
, а средата на
диагонала АC е Е, то триъгълникът
4 2
E EE
е образ на
триъгълника 1
PK M
при композицията от въртяща хомотетия
с център
1
K
, ъгъл
90
и коефициент
1
1
2
k
, и транслация на
вектор 1
K E
(фиг. 16).
Доказателство: Понеже 1
K P CD
(по условие) и
4
E E CD
(средна отсечка), то 1 4
K P E E
(фиг. 16).
Освен това 1
1
K P
k
CD
, където
1
k
е съответният на
1
K
Брокаров коефициент и 4
2CD
E E
, откъдето 1
1
4
2
K P
k
E E
,
т.е. 4
1 1
1
2
E E
K P k
. Аналогично се доказва, че
1 2
K M EE
и
2
1 1
1
2
EE
K M k
. Следователно при въртящата хомотетия с
център
1
K
, ъгъл
90
и коефициент
1
1
2
k
, триъгълникът
1
K PM
се изобразява в триъгълник с две страни съответно
успоредни и равни на страните на триъгълник
4 2
EE E
. От своя
страна, при транслация на вектор 1
K E
, последният триъгълник
се изобразява в
4 2
EE E
. Следователно
4 2
EE E
е образ на
1
K PM
при композицията от двете преобразувания.
Следствие: Нека
1
K
е Антиброкарианата на
четириъгълника ABCD, съответна на страните AB и CD, и М и Р
са ортогоналните й проекции съответно върху правите AB и CD.
Нека още
1
k
е съответният на
1
K
Брокаров коефициент. Ако
4
E
и
2
E
са средите съответно на страните AD и BC, то
4 2
MP E E
и
1 4 2
2
MP k E E
. (фиг. 16)
Доказателство: Триъгълникът
4 2
E EE
е образ на
триъгълника 1
PK M
при композицията от въртяща хомотетия
с център
1
K
, ъгъл
90
и коефициент
1
1
2
k
, и транслация на
вектор 1
K E
(по лема 1). Страните РМ и
4 2
E E
на тези
триъгълници са съответни при тази композиция от
INNOVATIONS 2025
124
преобразувания, следователно
4 2
PM E E
и
4 2
1
1
2
E E PM
k
, т.е.
1 4 2
2
k E
P EM .
Забележка 3. Нека
2
K
е Антиброкарианата на
четириъгълника ABCD, съответна на страните AD и BC, а Q и N
са ортогоналните й проекции съответно върху правите AD и BC
(фиг. 17). Нека още
2
k
е Брокаровият коефициент, съответен на
2
K
. Ако
1
E
и
3
E
са средите съответно на страните AB и CD,
то аналогично се доказва, че са изпълнени релациите
1 3
QN E E
и
2 1 3
2
k E
Q
E
N .
Фиг. 17
Вече можем да докажем следното:
Свойство 7 K. Нека
1
K
и
2
K
са Антиброкарианите на
четириъгълника ABCD, М и Р са ортогоналните проекции на
1
K
върху правите AB и CD, а Q и N – ортогоналните проекции
на
2
K
върху правите AD и BC. Нека още
1
k
и
2
k
са
съответните на
1
K
и
2
K
Брокарови коефициенти. Ако S е
лицето на ABCD, то лицето
1
S
на Антиброкаровия педален
четириъгълник MNPQ се определя от равенството
1 1 2
2
S k k S
. (фиг. 17)
Доказателство: Означаваме средите на страните AB, BC,
CD и DA съответно с
1
,
E
2
E
,
3
E
и
4
E
, ъгъла между правите
РМ и QN с
, а ъгъла между правите
4 2
E E
и
1 3
E E
с
.
Имаме:
1 4 2
2
k E
Р EМ и
2 1 3
2
k E
Q
E
N (по
следствието от лема 1). От друга страна, понеже
4 2
PM E E
и
1 3
QN E E
(по същото следствие и забележка 3), то
(ъгли с перпендикулярни рамене). Като вземем
предвид, че 1 2 3 4
1
2
E E E E
S S
(според известното свойство на
успоредника
1 2 3 4
E E E E
– теорема на Вариньон), с помощта на
горните две равенства, за лицето на четириъгълника МNPQ
получаваме последователно:
1 2 3 4
1 1 4 2 2 1 3
1 2 4 2 1 3 1 2 1 2
1 1
sin 2 2 sin
2 2
1
4 sin 4 2 .
2E E E E
S PM QN k E E k E E
k k E E E E k k S k k S
С това свойството е доказано.
Преди да разгледаме второто свойство на Антиброкаровия
педален четириъгълник, ще докажем следните две прости леми:
Лема 2. Нека ABCD е изпъкнал четириъгълник, и Е и F са
средите съответно на диагоналите му AC и BD. Нека още
E
е средата на диагонала МР на Антиброкаровия педален
четириъгълник МNPQ, а
F
– средата на диагонала му QN. Ако
1
K
и
2
K
са Антиброкарианите на четириъгълника ABCD, то са
изпълнени релациите 1
K E EF
и 2
K F EF
. (фиг.
17
а
).
Фиг.
17
а
Доказателство: Означаваме центъра на тежестта на
четириъгълника ABCD с G и с
1
k
– Брокаровия коефициент,
съответен на Антиброкарианата
1
K
(фиг.
17
а
). Лесно се
съобразява, че G е обща среда на отсечката ЕF и
4 2
E E
. При
композицията от въртяща хомотетия с център
1
K
, коефициент
1
1
2
k
и ъгъл
90
, и транслация на вектор 1
K E
,
триъгълникът
1
MK P
се изобразява в
2 4
E EE
(по лема 1).
Отсечките 1
K E
и ЕG са съответни медиани в двата
триъгълника при тази композиция от преобразувания, затова
1
K E EG
, т.е. 1
K E EF
. Доказахме първата от
горните релации. Аналогично се доказва и втората.
Лема 3. Нека ABCD е изпъкнал четириъгълник и Е и F са
средите съответно на диагоналите му AC и BD. Ако
1
P
и
2
P
са
симетралните точки на ABCD, то е изпълнена релацията
1 2
PP EF
. (фиг.
17
б
)
Фиг.
17
б
Доказателство: За да докажем, че 1 2
PP EF
, както
следва от Питагоровата теорема, е достатъчно да докажем
равенството:
INNOVATIONS 2025
125
(2)
2 2 2 2
1 1 2 2
PE PF P E P F
.
От
1
ACP
, в който 1
PE
е медиана, имаме:
(3)
2 2 2 2
1 1 1
4 2 2
PE AP CP AC
.
Аналогично от 1
BPD
имаме:
(4)
2 2 2 2
1 1 1
4 2 2
PF DP BP BD
.
Точката
1
P
е симетралната точка на ABCD, съответна на
страните му AB и CD, затова
1 1
AP BP
и
1 1
CP DP
. Като
извадим почленно равенство (4) от равенство (3) и използваме
последните две равенства, получаваме:
2 2
1 1
2 2 2 2 2 2
1 1 1 1
2 2
4
2 2 2 2
,
PE PF
AP CP AC DP BP BD
BD AC
т.е.
2 2 2 2
1 1
4
PE PF BD AC
.
Аналогично получаваме равенството:
2 2 2 2
2 2
4
P E P F BD AC
.
Оттук и следва равенство (2). С това лемата е доказана.
Следствие: Нека ABCD е изпъкнал четириъгълник и Е и F
са средите съответно на диагоналите му AC и BD. Ако
1
K
и
2
K
са Антиброкарианите на четириъгълника, то е изпълнена
релацията 1 2
K K EF
. (фиг.
17
б
)
Доказателство: Имаме
1 2 1 2
K K PP
(съгласно свойство
6K) (фиг. 14). Понеже 1 2
PP EF
(по доказаната лема 3) (фиг.
17
б
) заключаваме, че 1 2
K K EF
.
Преди да докажем второто свойство на Антиброкаровия
педален четириъгълник, ще дадем следното:
Определение 3K. Правата през средите на диагоналите на
произволен четириъгълник, който не е успоредник, ще наричаме
Гаусова права на четириъгълника.
Свойство 8K. Нека ABCD е изпъкнал четириъгълник, и
1
K
и
2
K
са Антиброкарианите му. Гаусовата права на
Антиброкаровия педален четириъгълник МNPQ съвпада с
правата
1 2
K K
и е перпендикулярна на Гаусовата права на
ABCD. (фиг.
17
в
).
Фиг.
17
в
Доказателство: Означаваме средите на диагоналите AC
и BD на четириъгълника ABCD съответно с E и F, а средите на
диагоналите PM и QN на Антиброкаровия педален
четириъгълник MNPQ – съответно с
E
и
F
. (фиг.
17
в
).
Гаусовата права на четириъгълника MNPQ е определена от
точките
E
и
F
. За да докажем, че тя съвпада с правата
1 2
K K
, е достатъчно да докажем, че
1 2
, E F
K K
. Имаме
1
K E EF
(съгласно лема 2) и 1 2
K K EF
(според
следствието от лема 3). Следователно правите 1
K E
и
1 2
K K
съвпадат. Оттук и следва, че
1 2
K
E K
. Аналогично се
убеждаваме, че и
1 2
K
F K
, с което свойството е доказано.
Заключение: В статията дефинирахме и разгледахме свойствата
на една двойка забележителни точки в изпъкнал четириъгълник,
които нарекохме Антиброкариани. Както видяхме, тя има тясна
връзка с друга двойка забележителни точки, наречени
Брокариани. В следваща статия ще се спрем на още една връзка
между Антиброкарианите и Брокарианите. Ще посочим още
един начин за построението на Антиброкарианите. Ще изведем
формули за разстоянията от тях до Брокарианите, до точката на
Микел и до върховете на четириъгълника, както и до пресечните
точки на продълженията на двойките срещуположни страни, с
помощта на които в по-следващи статии ще изведем различни
тъждества и неравенства в него. Ще покажем също, че в някои
частни видове четириъгълници Антиброкарианите съвпадат с
други забележителни точки в тях.
Литература:
[1] Haimov, H. & St. Stefanov. Geometry of the quadrilateral,
„Partner BG“ EOOD, Sofia, 2023, ISBN: 978-619-7066-32-6.
[Хаимов, Х. & Ст. Стефанов. Геометрия на четириъгълника,
„Партнер БГ“ ЕООД, София, 2023, ISBN: 978-619-7066-32-6.]
[2] Tabov, J., St. Stefanov, D. Milusheva-Boykina, A. Velchev & H.
Haimov. Set of Remarkable Points, Lines and Circles in the Plane of
the Quadrilateral, 51-th International Conference on Applications of
Mathematics in Engineering and Economics (AMEE’2025), 7-13
June, Sozopol, Bulgaria, ISSN: 0094243X. (Accepted for publication
in the AIP Conference Proceedings)
INNOVATIONS 2025
126
Други свойства на антиброкарианите
Станислав Стефанов1, Хаим Хаимов2
1Висше транспортно училище „Тодор Каблешков“, 2Варна, ул. Братя Шкорпил 16
stanislav.toshkov@abv.bg, , haim.haimov@abv.bg
Резюме. В (Stefanov & Haimov 2025) дефинирахме и разгледахме някои свойства на една двойка забележителни точки в произволен
изпъкнал четириъгълник, наречени Антиброкариани. Към тези свойства тук ще добавим нови. Ще приведем още един метод за
построението на Антиброкарианите. Ще изведем полезни формули за разстоянията от тях до други точки в четириъгълника, с
помощта на които в следващи статии ще докажем различни тъждества и неравенства в него. Ще разгледаме още една интересна
връзка на Антиброкарианите със забележителните точки Брокариани в четириъгълника. Ще покажем също, че в определени видове
четириъгълници Антиброкарианите съвпадат с определени забележителни точки в тях.
КЛЮЧОВИ ДУМИ: ИЗПЪКНАЛ ЧЕТИРИЪГЪЛНИК, ЗАБЕЛЕЖИТЕЛНИ ТОЧКИ, ВРЪЗКИ МЕЖДУ ТЯХ, ИЗОБРАЖЕНИЯ,
ФОРМУЛИ ЗА РАЗСТОЯНИЯ.
Other properties of antibrocardians
Stanislav Stefanov1, Haim Haimov2
1Todor Kableshkov University of Transport, 2Varna, 16 Bratya Shkorpil Str.,
stanislav.toshkov@abv.bg, haim.haimov@abv.bg
Abstract: In (Stefanov & Haimov 2025) we defined and considered some properties of a pair of notable points in an arbitrary convex
quadrilateral, called Antibrocardians. To these properties we will add new ones here. We will show one method more for constructing
Antibrocardians. We will derive useful formulas for the distances from them to other points in the quadrilateral, with the help of which we will
prove various identities and inequalities in it in subsequent articles. We will consider one interesting connection more of the Antibrocardians
with the notable points in the quadrilateral, named Brocaridans. We will also show that in certain types of quadrilaterals, Antibrocarians
coincide with certain notable points in them.
KEYWORDS: CONVEX QUADRILATERAL, NOTABLE POINTS, RELATIONS BETWEEN THEM, IMAGES, DISTANCES FORMULAS.
I. Въведение. Една важна двойка забележителни точки в
изпъкналия четириъгълник са неговите Антиброкариани. Както
видяхме в (Stefanov & Haimov 2025), те имат интересни връзки
с други забележителни точки в равнината на четириъгълника:
Брокарианите и пресечните точки на симетралите на двойките
му срещуположни страни. Тук ще докажем още една връзка на
Антиброкарианите с Брокарианите, базираща се на понятието
изогонални прави. Ще изведем формули за разстоянията от тези
точки до върховете на четириъгълника, до точката му на Микел,
до Брокарианите и до пресечните точки на продълженията на
двойките срещуположни страни. Ще приведем и прост метод за
построението на Антиброкарианите.
II. Предварителни означения.
За разглеждания в изложението по-нататък четириъгълник
ABCD ще предполагаме, че няма двойка успоредни или равни
срещуположни страни. Пресечната точка на продълженията на
страните му AD и BC ще означаваме с U, а тази на
продълженията на страните му AB и DC – с V (фиг. 1). При това,
за определеност ще считаме, че върхът C на четириъгълника
лежи между точките U и B и между точките D и V. Дължините
на страните AB, BC, CD и DA ще означаваме съответно с a, b, c
и d, дължините на диагоналите AC и BD – съответно с m и n, а
мерките на ъглите при върховете A, B, C и D – съответно с
,
,
и
. Средите на страните AB, BC, CD и DA ще бележим
съответно с
1 2 3
, ,
E E E
и
4
E
, а средите на диагоналите AC и
BD – съответно с E и F. Пресечната точка на диагоналите AC и
BD ще означаваме с T. Въвеждаме още означенията:
AUB
, AVD
,
2 4 1 1 3 2 3
, , =
E E l E E l EF l
.
Нека първо приведем дефиницията и едно свойство на
Антиброкарианите, което даказахме в (Stefanov & Haimov 2025)
и ще използваме по-нататък:
Определение 1. Точката
1
K
от вътрешността на ъгъла
между правите AB и DC, в който лежи четириъгълника ABCD,
такава, че 1 1
AK B DK C
(при посочената наредба на
върховете на триъгълниците), се нарича Антиброкариана на
ABCD, съответна на страните му AB и DC (фиг. 2).
Аналогично се дефинира и Антиброкарианата
2
K
,
съответна на страните AD и BC.
Фиг. 1
Фиг. 2
Нека
1
K
е Антиброкарианата на четириъгълника ABCD,
съответна на страните AB и DC (фиг. 2).
Свойство 1. Триъгълниците 1
AK C
и 1
BK D
са
равнолицеви.
Ще приведем сега дефиницията и някои свойства на
точката на Микел на четириъгълник:
Определение 2. Точка на Микел на четириъгълника ABCD
се нарича общата точка M на описаните окръжности около
триъгълниците AUB, CUD, ADV и BCV (фиг. 3) (Nenkov 2019).
INNOVATIONS 2025
127
Фиг. 3
Фиг. 4
Фиг. 5
Точката на Микел се характеризира със следните свойства:
Свойство 2. Триъгълниците ABM и DCM са подобни и
триъгълниците ADM и BCM също са подобни (фиг. 3) (Haimov
& Stefanov 2023).
Свойство 3. Разтоянията от точката на Микел M до
върховете на четириъгълника и до точките U и V се определят
по формулите (фиг. 3):
3 3 3 3
3 3
, , , ,
2 2 2 2
sin sin
, .
2 sin 2 sin
ad ab bc cd
MA MB MC MD
l l l l
ac bd
MU MV
l l
(Haimov & Stefanov 2023) (Tabov et al. 2024)
Сега ще дефинираме и приведем някои от свойствата на
едно изображение в равнината на четириъгълника, които ще
използваме по-нататък:
Определение 3. Композицията от симетрията спрямо
ъглополовящата g на ъгъл DMB (фиг. 3) и инверсията с полюс
точката M на Микел и степен 2
r BM DM
се нарича
инверсна симетрия на Микел спрямо четириъгълника ABCD с
полюс M, ос g и степен
2
r
и се бележи с
2
; ;
g MI g
r
или само с
I g
.
Свойство
4
a
. Ако точките X и Y са образи една на друга
при инверсната симетрия на Микел
2
; ;
g MI g
r
, то е
изпълнено равенството:
2
MX MY r
.
(Haimov & Stefanov 2023)
Свойство
4
б
. Разстоянието между точките
X
и
Y
–
образи на точките X и Y при инверсната симетрия
2
; ;
g MI g
r
на Микел се изразява по формулата:
2
XY r
X Y
MX MY
.
(Haimov & Stefanov 2023).
Сега ще приведем дефинициите и някои свойства на
Брокарианите и на две познати забележителни точки в изпъкнал
четириъгълник:
Определение 4. Нека ABCD е изпъкнал четириъгълник и T
е пресечната точка на диагоналите му (фиг. 4). Втората обща
точка
1
K
на описаните окръжности около триъгълниците ABT
и CDT ще наричаме Брокариана на ABCD, съответна на страните
AB и CD.
Аналогично се дефинира и Брокарианата
2
K
, съответна
на страните AD и BC.
Свойство 5. Нека
1
K
е Брокарианата на четириъгълника
ABCD, съответна на страните AB и CD. Триъгълниците
1
ABK
и
1
CDK
са подобни и триъгълниците 1
AK C
и 1
BK D
също
са подобни (фиг. 4) (Haimov & Stefanov 2023).
Свойство 6. Разстоянията от Брокарианите
1
K
и
2
K
до
точката на Микел M се определят по формулите:
2 1
1 2
1 3 2 3
, .
2 2
acl bdl
K M K M
l l l l
(Nenkov et al. 2019)
Определение 5. Пресечната точка
1
P
на симетралите на
страните AB и DC ще наричаме симетрална точка на
четириъгълника, съответна на тези страни, а пресечната точка
2
P
на симетралите на страните AD и BC – симетрална точка,
съответна на страните AD и BC (фиг. 5).
Свойство 7. Разстоянията от симетралните точки
1
P
и
2
P
до върховете на четириъгълника, до Брокарианите
1
K
и
2
K
,
до точката му на Микел M и до точките U и V се определят по
формулите:
2 2
2 2
2 2
2 2
1
2 cos ,
2sin
1
2 cos ,
2sin
P A P D a c ac
P B P C a c ac
(Nenkov et al. 2023);
2 2 2 2 2 2
1 2 1 1
3 3 1
2 2
2 1
2 2 1 1
2
, , ,
4 sin 4 sin 4 sin
, , .
4 sin sin sin
b d a c m n
PM P M PK
l l l
m n l l
P K PV PU
l
(Tabov et al. 2024)
Следното свойство определя образите на върховете А и В,
Антиброкарианите
1
K
и
2
K
, пресечната точка U на
продълженията на страните AD и BC, и Брокарианата
1
K
при
инверсната симетрия на Микел:
Свойство 8. Изпълнени са релациите:
INNOVATIONS 2025
128
1
2 1 2
2
1
, , ,
, , .
I A C B Dg I g I g K
I g K I
P
P U V
g I g K K
(Stefanov & Haimov 2025) (Haimov & Stefanov 2023)
Сега ще преминем към същинската част на статията.
III. Други свойства на Антиброкарианите.
Както ще видим най-напред, в определени видове
четириъгълници Антиброкарианите
1
K
и
2
K
съвпадат с
определени точки в тях:
Свойство 9. Във вписан четириъгълник ABCD, двете
Антиброкариани съвпадат с пресечната точка T на диагоналите.
Доказателство: Понеже
BAT CDT
и
ABT DCT
(от свойство на вписаните ъгли), то
ABT DCT
(при посочената наредба на върховете на
триъгълниците) (фиг. 6). Получаваме, че точката T отговаря на
условието от определение 1. Можем да заключим, че
Антиброкарианата на ABCD съответна на страните AB и CD
съвпада с T. Аналогично, същото се доказва и за
Антиброкарианата
2
K
на ABCD, съответна на страните му AD
и BC.
Фиг. 6
Свойство 10. Нека ABCD е изпъкнал четириъгълник, в
който
BAD CDA
. Антиброкарианата на ABCD,
съответна на страните AB и CD, лежи върху страната AD и я дели
в отношение AB : CD, считано от A. (фиг. 7)
Доказателство: Означаваме с
1
K
тази точка върху
страната AD, за която е изпълнено условието
1 1
AK DK
AB DC
(фиг. 7). По условие, в четириъгълника е изпълнено равенството
1
BAK
=
1
CDK
. Следователно
1
ABK
1
DCK
(при
посочената наредба на върховете на триъгълниците).
Получаваме, че за точката
1
K
е изпълнено условието от
определение 1. Можем да заключим, че Антиброкарианата на
ABCD, съответна на страните AB и CD, съвпада с построената
точка
1
K
. Следователно Антиброкарианата, съответна на
страните AB и DC, лежи на страната AD и я дели в отношение
AB : CD, считано от A.
Фиг. 7
Свойство 11. Нека ABCD е изпъкнал четириъгълник, в
който продълженията на страните BA и CD се пресичат в точка
V и е изпълнено равенството
VD DC VA AB
(D лежи
между V и C). Антиброкарианата на ABCD, съответна на
страните AB и DC, съвпада със симетричната точка на V спрямо
средата на страната AD.
Фиг. 8
Доказателство: Означаваме симетричната точка на V
спрямо средата на страната AD с
1
K
(фиг. 8). Четириъгълникът
1
VAK D
е успоредник, откъдето имаме:
1 1
BAK CDK
(ъгли с еднопосочни рамене). Освен това
1
VD
AK
и
1
VA
DK
, поради което от равенството
VD DC VA AB
, изпълнено по условие, следва
равенството: 1 1
B
A K
DC D A
K
. Оттук получаваме
пропорцията:
1 1
AK DK
AB DC
. Можем да заключим, че
1 1
ABK DCK
(при посочената наредба на върховете на
триъгълниците). Получихме, че точката
1
K
отговаря на
условието от определение 1. Следователно
1
K
е
Антиброкарианата на ABCD, съответна на страните AB и DC.
Свойство 12. Нека ABCD е изпъкнал четириъгълник, в
който продълженията на страните BA и CD се пресичат в точка
V (D лежи между V и C) и е изпълнено равенството
cos cos
AB BCD CD ABC
. Антиброкарианата
на ABCD, съответна на страните AB и DC, съвпада с ортоцентъра
на
VBC
. (фиг. 9)
Фиг. 9
Доказателство: Означаваме ортоцентъра на
VBC
с
1
K
(фиг. 6). Имаме: 1 1 90
ABK DCK CVB
.
Освен това, лесно се проверява, че за ортоцентъра
1
K
на
VBC
е изпълнено равенството 1
1
cos
cos
BK
VBC
CK BCV
.
Понеже от условието cos cos
AB BCV CD VBC
INNOVATIONS 2025
129
следва равенството cos
cos
AB VBC
CD BCV
, получаваме:
1
1
BK
AB
CK CD
. Можем да заключим, че
1 1
ABK DCK
(при посочената наредба на върховете на триъгълниците).
Получаваме, че за точката
1
K
е изпълнено условието от
определение 1. Следователно
1
K
е Антиброкарианата на ABCD,
съответна на страните AB и DC.
Сега ще разгледаме една връзка между Антиброкарианите
и Брокарианите на четириъгълника, базираща се на понятието
изогонални прави. Предварително ще припомним дефиницията
на това понятие.
Определение 6. Две прави се наричат изогонални спрямо
даден ъгъл XOY, ако образуват равни ъгли с ъглополовящата му,
а следователно и с раменете на ъгъла.
Както ще видим, всяка от двете Антиброкариани лежи със
съответна Брокариана на изогонални прави спрямо ъгъла,
определен от диагоналите на четириъгълника. Ще използваме
следната:
Лема 1. Нека S и T са две точки от вътрешността на даден
ъгъл XOУ, M и N са ортогоналните проекции на точката S върху
раменете му
ОХ
и
ОУ
, а P и Q – тези на точката T върху
същите рамене. Точките S и T лежат на изогонални прави спрямо
ъгъл XOУ тогава и само тогава, когато е изпълнено равенството
SM TQ
SN TP
(фиг. 10).
Фиг. 10
Доказателство: а) Нека първо точките S и T лежат на
изогонални прави спрямо ъгъл XOУ. Тогава
SON TOP
, откъдето следва, че
SON TOP
.
Оттук получаваме:
SN OS
TP OT
. Аналогично се доказва
равенството
SM OS
TQ OT
. Като сравним десните страни на
получените две равенства, стигаме до равенството
,
SM SN
TQ TP
т.е. до доказваното равенство
SM TQ
SN TP
.
б) Нека сега обратно, за точките S и T е изпълнено
равенството
SM TQ
SN TP
. Ще докажем, че точките S и T лежат
на изогонални прави спрямо ъгъл XOУ. Нека
1
T
e произволна
точка от правата l, изогонална на OS спрямо ъгъл XOУ. Ако
1
P
и
1
Q
са ортогоналните проекции на
1
T
върху раменете на ъгъла,
то по доказаното в а) следва, че е изпълнено равенството
1 1
1 1
T Q
SM
SN T P
. Сравняваме левите страни на последните две
равенства и получаваме
1 1
1 1
T Q
TQ
TP T P
. Оттук следва, че точките
T и
1
T
лежат на една и съща права през O, т.е. че правата OT
съвпада с правата
1
OT
. Но правата
1
OT
бе изогонална на OS
спрямо ъгъл XOУ. Следователно и правата OT е изогонална на
OS спрямо ъгъл XOУ, т.е. точките T и S лежат на изогонални
прави спрямо ъгъл XOУ.
Вече можем да докажем следното:
Свойство 13. Нека
1
K
и
1
K
са Брокарианата и
Антиброкарианата, съответни на страните AB и CD на
четириъгълника ABCD и T e пресечната точка на диагоналите
му. Точките
1
K
и
1
K
лежат на изогонални прави спрямо ъгъл
ATD. (Предполагаме, че точките
1
K
и
1
K
лежат едновременно
във вътрешността на ъгъл ATD или BTC – фиг. 11.)
Фиг. 11
Доказателство: Означаваме ортогоналните проекции на
Брокарианата
1
K
, съответна на страните AB и CD, върху
правите AT и DT съответно с
1
M
и
1
N
, а ортогоналните
проекции на Антиброкарианата
1
K
, съответна на AB и CD,
върху същите прави съответно с
2
M
и
2
N
(фиг. 11). Ще
покажем, че за точките
1
K
и
1
K
е изпълнено равенството
1 1 1 2
1 1 1 2
K M K N
K N K M
. От това по лема 1 и ще следва, че
1
K
и
1
K
лежат на изогонални прави спрямо ъгъл ATD, с което свойството
ще бъде доказано. Понеже 1 1
AK C BK D
(от свойство 5),
то 1 1
1 1
K M
AC
K N BD
(съответни височини в подобни
триъгълници). От друга страна, понеже 1 1
AK C BK D
S S (от
свойство 1), то
1 2 1 2
1 1
2 2
AC K M BD K N
, т.е.
1 2
1 2
K N
AC
K M BD
. Сравняваме десните части на получените две
INNOVATIONS 2025
130
равенства и стигаме до желаното равенство
1 1 1 2
1 1 1 2
.
K M K N
K N K M
Така доказахме, че точките
1
K
и
1
K
лежат на изогонални
прави спрямо ъгъл ATD.
Сега ще изведем формули за разстоянията от
Антиброкарианите
1
K
и
2
K
до върховете на четириъгълника.
Предварително ще докажем следната:
Лема 2. Нека ABC е равнобедрен триъгълник с бедра AC
и BC, с дължина a и D е произволна точка от строгото
продължение на основата му BA. Изпълнено е равенството (фиг.
20):
(1) 2 2
.
CD a AD BD
Фиг. 12
Доказателство: Означаваме BDC ADC
.
От
DAC
с помощта на косинусовата теорема определяме:
2 2 2
cos .
2
АD CD a
AD CD
От
DBC
аналогично имаме:
2 2 2
cos .
2
BD CD a
BD CD
Оттук получаваме равенството:
2 2 2 2 2 2
,
2 2
AD CD a BD CD a
AD CD BD CD
което е равносилно на равенството:
2
2
.
CD BD AD
a BD AD AD BD BD AD
Предвид на това, че
BD AD
, последното равенство води до
равенство (1).
Свойство 14. Нека ABCD е изпъкнал четириъгълник с
дължини на страните AB, BC, CD и DA съответно a, b, c и d и
мерки на ъглите при върхове A, B, C и D – съответно
, ,
и
. Ако
1
K
е Антиброкарианата на ABCD, съответна на
страните AB и CD, то са изпълнени равенствата (фиг. 13):
(2)
2 2
12 2
2 2
12 2
2 2
12 2
2 2
12 2
2 cos ,
2 cos ,
2 cos ,
2 cos .
ad
K A a c ac
a c
ab
K B a c ac
a c
bc
K C a c ac
a c
cd
K D a c ac
a c
Фиг. 13
Доказателство: За Антиброкарианата
1
K
имаме:
1
AK B
~1
DK C
(по определение 1) (фиг. 13). Затова
съществува число k такова, че 1 1
K A K D
k
AB CD
, т.е. че са
изпълнени равенствата:
(3) 1
K A ak
, 1
K D ck
.
Ще определим величината k и така ще намерим разстоянията
1
K A
и 1
K D
. Понеже 1 1
K AB K DC
, имаме:
1 1
1 1
,
K AD K DA
BAD K AB CDA K DC
BAD CDA
т.е.
(4) 1 1
K AD K DA
.
1). Ще разгледаме първо случая, когато
. От свойство 10
тогава имаме, че Антиброкарианата
1
K
лежи на страната AD и
понеже 1
1
K A
a
K D c
, лесно определяме: 1
ad
K A
a c
и
1
cd
K D
a c
. Понеже в този случай
1
cos
,
непосредствено се проверява, че последните равенства могат да
бъдат записани във вида:
2 2
12 2 2 cos
ad
K A a c ac
a c
и
2 2
12 2
2 cos .
cd
K D a c ac
a c
Получихме, че при
разстоянията 1
K A
и 1
K D
се
определят по формули (2).
2). По-нататък можем без ограничение да считаме, че
.
Тогава от (4) следва, че 1 1
K AD K DA
. Означаваме с
1
A
точката от лъча
AD
такава, че
1 1 1
K A K A
. Като
използваме последното неравенство, получаваме:
1 1 1 1 1 1
K A A K AA K AD K DA
, т.е.:
1 1 1
K A A K DA
. Оттук следва, че
1
A
е вътрешна точка
за отсечката AD (понеже, ако тя би била от продължението й,
INNOVATIONS 2025
131
бихме имали:
1 1 1
K DA K A A
). По косинусовата
теорема от 1 1
A K D
имаме:
(5)
2
1
2 2
1 1 1 1 1 1 1 1
2 cos .
A D
K A K D K A K D A K D
Като използваме (4) и равенството
1 1 1 1
AA K K AA
(следващо от
1 1 1
K A K A
), по теоремата за външен ъгъл от
триъгълника 1 1
A K D
получаваме последователно:
1 1 1 1 1 1
1 1 1 1
1 1
,
A K D AA K A DK
K AA A DK
K AD K DA
т.е.:
1 1
A K D
.
Понеже, освен това 1 1 1
K A K A ak
(от равенства (3)) и
1
K D ck
(от същите равенства), оттук след заместване в (5)
определяме:
(6)
2 2 2 2 2
1
2 cos .
A D a k c k ack
От друга страна, по лема 2, приложена към равнобедрения
1 1
AA K
и точката D от продължението на основата му
1
AA
имаме:
2 2
1 1 1
.
K D K A AD A D
Предвид равенства (3) и (6) и равенството
AD d
, оттук след
заместване получаваме:
2 2
2 2 2 2 2 2 2
2 cos .
c k
a k d a k c k ack
От последното равенство определяме:
2 2
2 2 2 cos
d
k a c ac
a c
.
Като заместим с този резултат в равенства (3), непосредствено
получаваме първото и последното от равенства (2). Така
доказахме верността на формулите за разстоянията 1
K A
и
1
K D
. По същия начин се доказват и тези за разстоянията
1
K B
и 1
K C
.
Забележка: Аналогично се извеждат и следните формули за
разстоянията от Антиброкарианата
2
K
до върховете на
четириъгълника:
2 2
22 2
2 2
22 2
2 2
22 2
2 2
22 2
2 cos ,
2 cos ,
2 cos ,
2 cos .
ad
K A b d bd
b d
ab
K B b d bd
b d
bc
K C b d bd
b d
cd
K D b d bd
b d
Преди да изведем формули за разстоянията от
Антиброкарианите
1
K
и
2
K
до точката на Микел M на
четириъгълника, ще докажем следната:
Лема 3. Константата
2
r
на Микел (т.е. степента на
инверсната симетрия
2
; ;
g MI g
r
на Микел) се
определя по формулата: 2
2
3
4
abcd
r
l
.
Доказателство: Понеже Брокарианите
1
K
и
2
K
са
образи една на друга при инверсната симетрия на Микел
2
; ;
g MI g
r
(от свойство 8), то от свойство
4
a
имаме:
2
1 2
MK MK r
.
Но
2
1
1 3
2
acl
MK
l l
и 2
2 3
2
bd
MK
l l
(от свойство 6) и като
заместим в това равенство получаваме:
22 1
2
1 3 2 3 3
.
2 2 4
acl bdl
abcd
r
l l l l l
.
Свойство 15. Разстоянията от Антиброкарианите
1
K
и
2
K
до точката на Микел M се определят по формулите:
(7) 1 2
2 2 2 2
3 3
sin sin
, .
abcd abcd
K M K M
a c l b d l
Доказателство: От свойство 8 имаме
1 2
I g K P
,
т.е. точките
1
K
и
2
P
са образи една на друга при инверсната
симетрия на Микел. Съгласно свойство
4
a
тогава е изпълнено
равенството
2
1 2
MK MP r
. Понеже
2 2
2
3
4 sin
a c
MP l
(от свойство 7) и 2
2
3
4
abcd
r
l
(по лема 3), оттук определяме:
2
3
122 2 2 2
2 3
3
4 sin
sin
.
4
l
r abcd abcd
MK MP l
a c a c l
Получихме първата от доказваните формули (7). Аналогично се
доказва и втората формула.
Сега ще изведем формули за разстоянията от
Антиброкарианите
1
K
и
2
K
до съответните Брокариани
1
K
и
2
K
:
Свойство 16. Разстоянията от Антиброкарианите
1
K
и
2
K
съответно до Брокарианите
1
K
и
2
K
на изпъкнал
четириъгълник ABCD се определят по формулите:
(8)
2 2 2 2
1 1 2 2
2 2 2 2
1 2
, .
2 2
ac m n bd m n
K K K K
l a c l b d
Доказателство: Имаме
2 1
I g P K
и
2 1
I g K K
(от свойство 8), т.е. образите на точките
2
P
и
2
K
при инверсната симетрия на Микел са съответно
1
K
и
INNOVATIONS 2025
132
1
K
. Разстоянието
1 1
K K
между тези образи се изразява по
формулата от свойство
4
б
:
(9)
2
2 2
1 1
2 2
r P K
K K
MP MK
.
От друга страна
2 2 2 2
2 2 2
2 3
,
4 sin 4 sin
m n a c
P K MP
l l
(от свойство 7),
1
2
2 3
2
bdl
MK
l l
(от свойство 6) и 2
2
3
4
abcd
r
l
(по лема 3). След заместване с тези изрази в (9) получаваме:
2
2 2
1 1
2 2
2 2 2 2
3 2 3
2
2 2 2 2
3 2 1 1
4 sin 2
. .
4 4 sin 2
r P K
K K MP MK
m n ac m n
l l l
abcd
l l bdl
a c l a c
С това получихме първата от формули (8). Аналогично се
доказва и втората формула.
Свойство 17. Разстоянията от Антиброкарианите
1
K
и
2
K
съответно до пресечните точки V и U на двойките
срещуположни страни се определят по формулите:
(10)
1
12 2
2
22 2
2 sin
,
sin
2 sin
.
sin
bdl
K V a c
acl
K U b d
Доказателство: Имаме
2 1
I g P K
и
I g U V
(от свойство 8), т.е. образите на точките
2
P
и
U при инверсната симетрия на Микел са съответно
1
K
и V.
Разстоянието 1
K V
между тези образи се изразява по
формулата от свойство
4
б
:
(11)
2
2
1
2
PU r
K V
MP MU
.
От друга страна
2 2
1
2 2
3
,
sin 4 sin
a c
l
PU MP l
(от
свойство 7),
3
sin
2 sin
ac
MU l
(от свойство 3) и 2
2
3
4
abcd
r
l
(по лема 3). След заместване с тези изрази в (11) получаваме:
2
2
1
2
3 3
1
22 2
3
1
2 2
4 sin 2 sin
4 sin sin
2 sin .
sin
r P U
K V M P M U
l l
labcd
l ac
a c
bdl
a c
Получихме първата от формули (10). Аналогично се доказва и
втората формула.
Накрая ще изложим още един метод за построяването на
Антиброкарианите. Ще използваме следните съображения:
Фиг. 14
Нека
1
K
е Антиброкарианата на четириъгълника ABCD,
съответна на страните му AB и CD (фиг. 14). Имаме
1 1
ABK DCK
(по определение 1), откъдето
1
1
K B
AB
K C CD
. Можем да заключим, че точката
1
K
лежи на
Аполониевата окръжност (
1
c
) за отсечката BC, за точките X от
която е изпълнено равенството
XB AB
XC CD
. Аналогично
получаваме, че точката
1
K
лежи и на Аполониевата окръжност
(
2
c
) за отсечката AD, за точките Y от която е изпълнено
равенството
YA AB
YD CD
. Точката на Микел M също лежи на
Аполониевите окръжности (
1
c
) и (
2
c
), понеже от
ABM DCM
(от свойство 1) следва, че
MB MA AB
MC MD CD
. Затова точката M е обща точка на
окръжностите (
1
c
) и (
2
c
). Тогава точката
1
K
може да се
построи като втората обща точка на окръжностите (
1
c
) и (
2
c
).
Заключение. В статията, явяваща се продължение на
(Stefanov & Haimov 2025) допълнихме разгледаните в
последната свойства на Антиброкарианите с нови. Най-
съществено беше допълнението към свойствата на Брокаровия
педален четириъгълник. С оглед на приложенията на
разгледаните свойства на Антиброкарианите особено важно
значение имат изведените формули за разстояния от тях до
други точки в четириъгълника. В следващи статии те ще бъдат
използвани за получаване на ред тъждества и неравенства в
четириъгълника.
Литература:
[1] Stefanov, St. & H. Haimov. Antibrocardians of a quadrangle, XI
International Scientific Congress "Innovations", June 23-26, 2025,
Varna, ISSN 2603-3763 (Accepted for publication) [Стефанов, Ст.
& Х. Хаимов. Антиброкариани на четириъгълник, XI
Международен научен конгрес „Иновации“, 23-26 юни, 2025,
Варна, ISSN 2603-3763 (Приета за печат)]
[2] Nenkov, V. Improving mathematical competencies with dynamic
geometry, "Arкhimed 2" EOOD, Sofia, 2019, ISBN: 978-954-779-
291-3. [Ненков, В. Повишаване на математически компетенции
с динамична геометрия, ”Архимед 2“ ЕООД, София, 2019, ISBN:
978-954-779-291-3.]
INNOVATIONS 2025
133
[3] Haimov, H. & St. Stefanov. Geometry of the quadrilateral,
„Partner BG“ EOOD, Sofia, 2023, ISBN: 978-619-7066-32-6.
[Хаимов, Х. & Ст. Стефанов. Геометрия на четириъгълника,
„Партнер БГ“ ЕООД, София, 2023, ISBN: 978-619-7066-32-6.]
[4] Tabov, J., St. Stefanov, Kr. Kanchev & H. Haimov. Distances
between remarkable points and inequalities in a convex quadrilateral,
Mathematics and Informatics, vol. 67, № 3, 2024, pp. 253-274, ISSN
1310-2230. [Табов, Й., Ст. Стефанов, К. Кънчев & Х. Хаимов.
Разстояния между забележителни точки и неравенства в
изпъкнал четириъгълник, Математика и информатика, том 67,
№ 3, 2024, стр. 253-274, ISSN 1310-2230.]
[5] Nenkov, V., St. Stefanov & H. Haimov. Formulas for the
distances from the Brocardians and the Miquel point in a
quadrilateral to its vertices and the midpoints of its sides and
diagonals, Mathematics and Informatics, vol. 62, № 3, 2019, pp. 305-
324, ISSN 1310-2230. [Ненков, В., Ст. Стефанов & Х. Хаимов.
Формули за разстоянията от Брокарианите и точката на Микел в
четириъгълника до върховете и до средите на страните и
диагоналите му, Математика и информатика, том 62, № 3, 2019,
стр. 305-324, ISSN 1310-2230.]
[6] Nenkov, V., S. Grozdev, A. Velchev, R. Alashka, St. Stefanov
and H. Haimov. New Formulas for Distances Between New and
Traditional Remarkable Points in a Quadrilateral, 48-th International
Conference on Applications of Mathematics in Engineering and
Economics (AMEE’2022), 7-13 June, Sozopol, Bulgaria. AIP
Conference Proceedings, Vol. 2939, Issue 1, 2023, Art. № 060003,
ISSN: 0094243X, ISBN: 978-073544763-9, DOI:
10.1063/5.0188027.
INNOVATIONS 2025
134
Synthesis of carbon adsorbents for adsorption of chlorhexidine gluconate
from aqueous solution
I.Stoycheva1*, B. Petrova1, B. Tsyntsarski1, A. Kosateva1, M. Argirova1, G. Tirolski1, N. Petrov2, P. Dolashka2, M. Kalapsazova3
1 Institute of Organic Chemistry with Centre of Phytochemistry, Bulgarian Academy of Sciences,Acad. G. Bonchev Str., BL. 9, 1113 Sofia,
Bulgaria
2 Center of Competence “Clean technologies for sustainable environment - water, waste, energy
for circular economy”, 15 Tsar Osvoboditel Blvd., 1000 Sofia, Bulgaria
3 Institute of General and Inorganic Chemistry, Bulgarian Academy of Sciences, Sofia, 1113, Bulgaria
Abstract: Water purification by adsorption of various pollutants using carbon adsorbents with different characteristics has proven to be an
effective method that is often used in purification technology. In the present work, a new method for the production of synthetic carbon
adsorbents with a large surface area and a wide range of pore sizes is used. Carbon adsorbents based on waste materials were synthesized.
The treatment of such starting materials with inorganic acids leads to suitable physicochemical properties for application in the production
of high-tech carbon materials. A new method has been developed that ensures the production of carbon adsorbents with a regulated
structure, high mechanical strength and, most importantly, with a predominant amount of macropores. The carbon materials were
characterized by SEM, X-ray structural analysis, N2 physisorption, Raman spectroscopy, etc. The obtained carbon materials were
successfully tested for the adsorption of chlorhexidine gluconate from an aqueous solution.
Acknowledgements:
The authors acknowledge financial support for this work by Bulgarian National Science Fund, grant number KP-06-М77/2 from 29.11.2023,
KP-06-N-69/7 from 15.12.2022, Project BG05M2OP001-1.002-0019:„Clean technologies for sustainable environment - water, waste,
energy for circular economy“, financed by the Operational program “Science and Education for Smart Growth” co-financed by the European
union. The authors are also grateful to the funding by Project BG-RRP-2.017-0006-C01 from Recovery Plan for Europe
(NextGenerationEU).
INNOVATIONS 2025
135
Study of the porous properties of carbon adsorbents obtained from waste materials
B. Petrova1*, I. Stoycheva1, B. Tsyntsarski1, A. Kosateva1, N. Petrov2, P. Dolashka2
1 Institute of Organic Chemistry with Centre of Phytochemistry, Bulgarian Academy of Sciences,Acad. G. Bonchev Str., BL. 9, 1113 Sofia,
Bulgaria
2 Center of Competence “Clean technologies for sustainable environment - water, waste, energy
for circular economy”, 15 Tsar Osvoboditel Blvd., 1000 Sofia, Bulgaria
Abstract: Porosity is very important characteristic of carbon materials. The adsorption properties of carbon materials are connected with
their internal pore structure. That's why it is necessary to perform complete characterization of the texture of porous materials, including
pore volume, pore size distribution, the specific surface area, etc. To evaluate the porous texture of porous carbon materials obtained in the
laboratory, it was investigated by the amount of adsorbed nitrogen at 77 K. The analysis is performed on an Autosorb iQ-MP Quantachrome
instrument.
The results obtained for pore volume, pore size distribution, and specific surface area of porous carbons and carbon materials samples
obtained from different raw materials and treatment conditions are presented. Based on the obtained results, a preliminary assessment of the
possibilities for their application was made.
Acknowledgements:
The authors acknowledge financial support for this work by Bulgarian National Science Fund, grant number KP-06-М77/2 from 29.11.2023,
KP-06-N-69/7 from 15.12.2022, Project КP-06-N59-12 from 22.11.2021, Project BG05M2OP001-1.002-0019:„Clean technologies for
sustainable environment - water, waste, energy for circular economy“, financed by the Operational program “Science and Education for
Smart Growth” co-financed by the European union. The authors are also grateful to the funding by Project BG-RRP-2.017-0006-C01 from
Recovery Plan for Europe (NextGenerationEU).
INNOVATIONS 2025
136
Synthesis and characterization of novel bio-char adsorbents
B. Tsyntsarski1*, I.Stoycheva1, B. Petrova1, A. Kosateva1, N. Petrov2, P. Dolashka2, Teodor Sandu 3 and Andrei Sarbu 3
1 Institute of Organic Chemistry with Centre of Phytochemistry, Bulgarian Academy of Sciences,Acad. G. Bonchev Str., BL. 9, 1113 Sofia,
Bulgaria
2 Center of Competence “Clean technologies for sustainable environment - water, waste, energy
for circular economy”, 15 Tsar Osvoboditel Blvd., 1000 Sofia, Bulgaria
3 National Research- Developent Institute for Chemistry and Petrochemistry INCDCP- ICECHIM, Spl. Independentei 202, sector 6,
Bucharest, Romania;
Abstract: Bio-char carbon adsorbents, derived from biomass by-products, are produced by using new original method. The resulting carbon
materials are distinguished by high BET surface area and large amount of micro-, meso- and macro-pores. The treatment of biomass waste
precursors with mineral acids results in synthesis of carbon product with suitable physicochemical properties for wide application areas.
The newly developed method ensures synthesis of carbon adsorbents with order structure, high mechanical strength and, most importantly,
with a predominant amount of macropores. The carbon materials were characterized by XRD, BET, Raman spectroscopy, etc. The obtained
carbon materials were successfully tested for removal of different pollutants from water solution.
Acknowledgements:
The authors acknowledge financial support for this work by Bulgarian National Science Fund, grant number KP-06-М77/2 from 29.11.2023,
Project BG05M2OP001-1.002-0019:„Clean technologies for sustainable environment - water, waste, energy for circular economy“, financed
by the Operational program “Science and Education for Smart Growth” co-financed by the European union. The authors are also grateful to
the funding by Project BG-RRP-2.017-0006-C01 from Recovery Plan for Europe (NextGenerationEU). The authors also appreciate the
support from Romanian-Bulgarian Academia Joint Project INNOMAG.
INNOVATIONS 2025
137
