Psychological Factors Influencing University Students' Trust in AI-Based Learning Assistants
Abstract
Artificial intelligence (AI)-based learning assistants and chatbots are increasingly integrated into higher education. While these tools are often evaluated in terms of technical performance, their successful and ethical use also depends on psychological factors such as trust, perceived risk, technology anxiety, and students’ general attitudes toward AI. This paper adopts a psychology-oriented perspective to examine how university students form trust in AI-based learning assistants. Drawing on recent literature in mental health, human AI interaction, and trust in automation, we propose a conceptual framework that organizes psychological predictors of trust into four groups: cognitive appraisals (perceived competence, reliability, and usefulness), affective reactions (anxiety, fear, comfort), social-relational factors (anthropomorphism, perceived empathy, and autonomy support), and contextual moderators (privacy, transparency, and ethical norms). A narrative review approach is used to synthesize empirical findings on AI tools in education and mental health support, with a specific focus on college and university populations. Based on this synthesis, we derive research questions and hypotheses for future survey and experimental studies with students. The paper highlights that trust in AI is not a purely technical outcome but a psychological process shaped by individual differences and learning environments. Finally, practical implications are discussed for instructors, university administrators, and designers of educational AI systems, including guidelines to foster appropriate reliance rather than blind trust or total rejection.
Full text
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) PROCEEDINGS BOOK December 12-15, 2025 Bayburt, Türkiye ISBN: 978-625-7960-84-7
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 i 2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) Onsite – Online (Hybrid) Conference December 12-15, 2025 | Bayburt, Türkiye Editor-in-Chief Yunus Kaya Editors Erman Kadir Oztekin Hamdullah Ozturk Intan Helina Hasan Nurhafizah Hasim Selahattin Kosunalp Umit Yildirim ISBN: 978-625-7960-84-7 December 2025
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 ii Papers reviewed by at least two reviewers presented at the 2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) are included in this proceedings book. The full responsibility for all papers presented at the ICMUSTED 2025 and published in this proceedings book belongs solely with the respective authors of the papers. All rights reserved. This publication is free of charge and cannot be sold for money. No part of this publication may be reproduced, stored, retrieved or transmitted, or distributed in any other binding or cover form, without the written permission of the publisher. It can be used by citing the source. ISBN: 978-625-7960-84-7 Publication Date: 19.12.2025
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 iii Dear Participants, We would like to thank all of you for your participation and interest in the 2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025), which was held as Onsite/Online (Hybrid) in Bayburt, Türkiye on December 12-15, 2025. The aim of ICMUSTED 2025 is to provide an international forum for researchers, academics, people in industry, and students to consider the latest research results and to present and discuss their ideas, theories, technologies, systems, tools, applications, work in progress. In this regard, participants will experience all theoretical and practical problems and technological developments that arise in multidisciplinary topics. Onsite and online presentations were made by invited speakers and other participants within the scope of the ICMUSTED 2025. ICMUSTED 2025, where 195 oral presentations prepared by 475 participants from 25 different countries, took place and opened a direction to new cooperation opportunities. Therefore, we would like to thank the invited speaker and all other participants, the members of the scientific committee, the session chairs, and all those who contributed to make this conference a great success. Hope to see you at the next ICMUSTED. Best Regards, On behalf of the ICMUSTED 2025 Organizing Committe Organizing Committee Chairman Assoc. Prof. Dr. Yunus Kaya
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 iv COMMITTEES Organizing Committee Conference Chairman Assoc. Prof. Dr. Yunus Kaya, Bayburt University, Türkiye Members Prof. Dr. Jorge Montanari, National University of Hurlingham, Argentina Assoc. Prof. Dr. Emre Tekce, Bayburt University, Türkiye Assoc. Prof. Dr. Mehmet Uyar, Bayburt University, Türkiye Assoc. Prof. Dr. Mustafa Ozdemir, Bayburt University, Türkiye Assoc. Prof. Dr. Selahattin Kosunalp, Bandirma Onyedi Eylul University, Türkiye Assoc. Prof. Dr. Umit Yildirim, Bayburt University, Türkiye Asst. Prof. Dr. Erman Kadir Oztekin, Bayburt University, Türkiye Asst. Prof. Dr. Hamdullah Ozturk, Gaziantep Islam Science and Technology University, Türkiye Dr. Intan Helina Hasan, Universiti Putra Malaysia, Malaysia Dr. Nurhafizah Hasim, Universiti Teknologi Malaysia, Malaysia Secretary Asst. Prof. Dr. Meltem Kizilca Coruh, Ataturk University, Türkiye Asst. Prof. Dr. Sebahat Oztekin, Ataturk University, Türkiye Scientific Committee Prof. Dr. Ahmet Cansiz, Istanbul Technical University, Türkiye Prof. Dr. Alyani Ismail, Universiti Putra Malaysia, Malaysia Prof. Dr. Grigor Yordanov Mihaylov, University of Telecommunications and Post, Bulgaria Prof. Dr. Halim Kovaci, Ataturk University, Türkiye Prof. Dr. Jorge Montanari, National University of Hurlingham, Argentina Prof. Dr. Mehmet Ertugrul, Karadeniz Technical University, Türkiye Prof. Dr. Mohd Nizar Hamidon, Universiti Putra Malaysia, Malaysia Prof. Dr. Teodor Iliev, University of Ruse “Angel Kanchev”, Bulgaria Prof. Dr. Ugur Cem Hasar, Gaziantep University, Türkiye Prof. Dr. Zehra Can, Bayburt University, Türkiye Assoc. Prof. Dr. Adem Korkmaz, Bandirma Onyedi Eylul University, Türkiye Assoc. Prof. Dr. Bora Goktas, Bayburt University, Türkiye Assoc. Prof. Dr. Emre Tekce, Bayburt University, Türkiye Assoc. Prof. Dr. Fatih Yilmaz, Bayburt University, Türkiye Assoc. Prof. Dr. Gokhan Komur, Bayburt University, Türkiye Assoc. Prof. Dr. Gokhan Ozturk, Ataturk University, Türkiye Assoc. Prof. Dr. Kubilay Demir, AISOFT Software Corporation, Türkiye Assoc. Prof. Dr. Mehmet Uyar, Bayburt University, Türkiye Assoc. Prof. Dr. Mustafa Ozdemir, Bayburt University, Türkiye Assoc. Prof. Dr. Mustafa Tolga Yurtcan, Ataturk University, Türkiye Assoc. Prof. Dr. Ruhsen Aldemir Engin, Kafkas University, Türkiye Assoc. Prof. Dr. Samsuri Abdullah, Universiti Malaysia Terengganu, Malaysia Assoc. Prof. Dr. Selahattin Kosunalp, Bandirma Onyedi Eylul University, Türkiye Assoc. Prof. Dr. Umit Yildirim, Bayburt University, Türkiye
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 v Assoc. Prof. Dr. Yunus Kaya, Bayburt University, Türkiye Assoc. Prof. Dr. Yusuf Esmer, Bayburt University, Türkiye Asst. Prof. Dr. Duygu Tekin, Bayburt University, Türkiye Asst. Prof. Dr. Erdal Igman, Bayburt University, Türkiye Asst. Prof. Dr. Erman Kadir Oztekin, Bayburt University, Türkiye Asst. Prof. Dr. Hamdullah Ozturk, Gaziantep Islam Science and Technology University, Türkiye Asst. Prof. Dr. Ibrahim Cengiz, Ataturk University, Türkiye Asst. Prof. Dr. Latif Akcay, Erzurum Technical University, Türkiye Asst. Prof. Dr. Meltem Kizilca Coruh, Ataturk University, Türkiye Asst. Prof. Dr. Mustafa Alptekin Engin, Bayburt University, Türkiye Asst. Prof. Dr. Ramazan Simsek, Bayburt University, Türkiye Asst. Prof. Dr. Sebahat Oztekin, Ataturk University, Türkiye Asst. Prof. Dr. Zeynep Ozturk, Erzurum Technical University, Türkiye Dr. Hafize Hasar, Gaziantep Directorate of Provincial Agriculture and Forestry, Türkiye Dr. Intan Helina Hasan, Universiti Putra Malaysia, Malaysia Dr. Lawal Mohammed Bello, Bayero University Kano, Nigeria Dr. Norhazren Izatie Mohd, Universiti Teknologi Malaysia, Malaysia Dr. Nurhafizah Hasim, Universiti Teknologi Malaysia, Malaysia
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 vi TOPICS Agriculture, Forestry, and Aquaculture Architecture, Planning, and Design Education Sciences Engineering Fine Arts Health Sciences Law Philology (Language and Literature) Research & Development and Technological Developments Science and Mathematics Social, Humanities, and Administrative Sciences Sport Sciences Theology
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 vii INVITED SPEAKER Assoc. Prof. Dr. Samsuri Abdullah Universiti Malaysia Terengganu (Terengganu, Malaysia) Samsuri Abdullah is an Associate Professor at Universiti Malaysia Terengganu (UMT), Malaysia. He has been teaching environmental technology since 2018 and was promoted to Associate Professor in 2022. He obtained his Bachelor of Technology (Environment) with First Class Honours in 2014 and completed a fast-track PhD in Environmental Technology and Management in 2017, specializing in air quality. His research focuses on ambient and indoor air quality, as well as noise pollution. Samsuri has published 162 works, including journal articles, book chapters, conference papers, and a technical report. He has led or co-led 20 research projects with total funding of approximately MYR 1.4 million and has conducted data analysis workshops for applied sciences. He is also involved in consultancy projects with Tenaga Nasional Berhad Research and Enviro Excel Tech Sdn. Bhd. At UMT, he contributes to sustainability initiatives related to the Times Higher Education (THE) Impact Rankings and the UI GreenMetric World University Rankings. Currently, he serves as the Coordinator for the Undergraduate Final Year Project Dissertation, overseeing academic quality and research supervision at the faculty level.
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 viii ICMUSTED 2025 PROGRAM Friday, December 12, 2025 Conference Opening Conference Hall (Bayburt Teacher’s House, Bayburt) 09:30 – 10:00 Registration and Tea/Coffee Service 10:00 – 10:45 Opening Ceremony Asst. Prof. Dr. Erman Kadir Oztekin – On Behalf of the Organizing Committee Opening Session Air Quality and Sustainable Development: The Role of Machine Learning in Shaping Future Cities Invited Speaker – Assoc. Prof. Dr. Samsuri Abdullah Onsite Session Conference Hall (Bayburt Teacher’s House, Bayburt) Onsite Session – 1 (in Turkish) Head of Session: Asst. Prof. Dr. Erman Kadir Oztekin 10:45 – 12:00 A Modified Two-Sided Approximation Method for a Special Boundary Value Problem with Delayed Arguments Arzu Aykut*, Fulya Koc Simsek AI-Based Intelligent Document Classification in the Logistics Sector Selcuk Zereyalp*, Kemal Sogukcesme, Bahadir Fatih Yildirim History of Mobile Communication from 1G to 6G Hanifi Turgut, Tevhit Karacali, Tarik Bugra Ala* Design and Analysis of Metamaterial Absorber for Microwave X-Band Applications Yunus Kaya*, Mehmet Ertugrul Multi-Band and Inexpensive Linear and Circular Polarization Converter Using a Single-Layer Reflective Metasurface Yunus Kaya*, Ugur Cem Hasar 12:00 – 13:15 Lunch (Bayburt Teacher’s House) Online Sessions Virtual Hall Online Session – 1 (in Turkish) Head of Session: Asst. Prof. Dr. Duygu Tekin 13:15 – 14:45 Design, Prototype Production, and Testing of a 12(16) MVA 115/6.3 kV Power Transformer with Twin-Wire Transposed High-Voltage Winding for Loss Reduction Zuhal Isikli*
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xv Benamara Structural Properties of Ni-Doped ZnO Thin Films Investigated by XRD, AFM, and Profilometry Maya Hanane Rezoug*, Chewki Zegadi, Abdelkader Nouri, Nasr-Eddine Hamdadou, M’hamed Guezzoul 17:30 – 17:45 Break Online Session – 11 (in English) Head of Session: Dr. Nurhafizah Hasim 17:45 – 19:15 Immobilized Microalgae for Sustainable Soilless Agriculture in the Green Transition Burcu Simsek Uygun, Gizem Demirel* Implementation of Open Science Framework for Materials Engineering Elisabeth Viviana Lucero Baldevenites*, Jose Rogelio Fung Corro, Yorlenis Martínez Separation of Toluene – Cyclohexane Mixture Using Intensified Extraction Process by Imidazolium-Based Ionic Liquids Mohammed Djamel Eddine Allali*, Hassiba Benyounes, Nesrine Amiri Study of the Effectiveness of Steel Shavings as a Foaming Agent for the Production of Foams Glass Fayrouz Benhaoua*, Djalila Aoufi, Nacira Stiti, Mehdia Toubane, Djedjiga Bousalah Electromagnetic Frequency and Time Reaction in Electromechanical Systems Under Low Energy Mechanical Faults Azeddine Ratni*, Ali Damou, Djamel Benazzouz, Mohamed Tsebia Microstructural and Electrochemical Assessment of Co–Cr Dental Alloys in Ringer Solution Alberto Daniel Rico-Cano, Adriana Saceleanu, Anca Fratila, Julia Claudia Mirza-Rosca* Electrospinning of Polyacrylonitrile-Nanocellulose Composites Reinforced with Carbon Materials for Advanced Fiber Performance Mohd Ali Mat Nong*, Juraina Md Yusof, Suzila Sabil, Mohd Hafizuddin Ab Ghani, Che Azurahanim Che Abdullah 19:15 – 19:30 Break Online Session – 12 (in English) Head of Session: Asst. Prof. Dr. Latif Akcay 19:30 – 21:00 Hybrid Voltage Regulation Strategy Combining AVR and Microgrid Coordination for Enhanced Power System Stability Fazia Ahcene*, Hamid Bentarzi, Mohammed Tsebia, Djamila Talah
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xvi Next-Generation Deep Learning Architectures for Satellite Image Classification: Integrating Capsule Networks with CNNs and Transformers Boucif Beddad*, Samiha Mezrar, Postaire Jack-Gerard Harvesting Waste Kinetic Energy from Vehicle Suspensions for Enhanced Mileage and Power Generation Touqeer Aslam*, Shoukat Ali Mugheri, Ali Azam, Manthar Ali, Abbas Raza Development of an Algorithm for Determining the Characteristics of the Ejection System Oleksandr Panevnyk* Virtual Students in Programming Fundamentals: Comparing Large Language Models with Vocational Computer Programming Students in Classical Exams Selma Bulut, Adem Korkmaz* AI-Based Anomaly Detection in Agricultural Farms Using Drone Data and Deep Learning Berrimi Fella* Bone Fracture Classification Analysis Base Machine Learning Algorithms Ei Phyu Sin Win* Large Language Model–Assisted Hardware Design: Insights from the Ascon-128 Implementation Latif Akcay* Sunday, December 14, 2025 Online Sessions Virtual Hall Online Session – 13 (in English) Head of Session: Dr. Nurhafizah Hasim 09:00 – 10:30 Indoor Particulate Matter Levels in the Department of Environmental Engineering at Eskisehir Technical University, Türkiye Melike Cengel, Ozlem Ozden Uzmez* Seismic Earth Pressures on Retaining Walls: A Comprehensive Review of Analytical, Numerical, and Experimental Approaches Ayman Gharbi*, Fadoua Elkhannoussi, Bouraida Elyamouni, Abdellatif Khamlichi Numerical Evaluation of Seismic Performance in Earth Dams: The Case of Fontaine Gazelles Dam, Biskra, Algeria Alaoua Bouaicha*, Aissam Gaagai, Mosbah Ben Said, Ali Hachemi Influence of Carbonation on Reinforced Concrete Structures in Southern Algeria Ben Ammar Ben Khadda*
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xvii Numerical Study to the Enhancement of Pressure Distribution within the Alveolus of an Aerostaic Bearing Faiza Ghezali* A Comparative Finite Element Analysis of Static Coil Geometries for Electromagnetic Eddy Current Separation Amir Merahi*, Kaouther Boutarfa, Adel Benabboun, Meriem Boumehed Contribution to the Recovery of Paint Sludge in Wastewater Treatment Aicha Metali*, Mounir Ziati, Ikram Rebiai, Khedidja Bachi 10:30 – 10:45 Break Online Session – 14 (in English) Head of Session: Dr. Nurhafizah Hasim 10:45 – 12:15 Microservice Transformation for Infrastructure Modernization of Payment and Electronic Money Institutions Ahmet Ulker, Cansu Barisici, Eren Capraz, Ilker Ogutcu, Mehmet Bulent Muslu, Zafer Golgeli, Ceren Ulus, M. Fatih Akay* Experimental Investigation of Concrete by Using Marble Waste as Replacement of Fine Aggregate Hassaan Amjad*, Faisal Ahmed, Allah Noor Investigating the Properties Ultra High-Performance Concrete Using Silica Fume and Ground Granulated Blast Furnace Slag Hassaan Amjad*, Waseem Asghar, Dawood Jan Production of Bioethanol Using Lower-Quality Algerian Sugar Dates BoxBehnken Kaouther Zeghida*, Sarra Guilane, Leila Benmansour Artificial Intelligence Control of Active and Reactive Power for a Three-Level NPC Inverter Connected to Grid Utility Ghrissi Tahri*, Fatima Tahri, Ali Tahri Mechanical Properties of Concrete Reinforced with Steel Fiber Noshad Ali*, Waqar Ali, Muhammad Hessib Influx of Different Growth Geometries on Titanium Thin Film for Medical Applications Matteo Bertapelle, Joel Borges, Julia Claudia Mirza Rosca, Filipe Vaz* 12:15 – 12:30 Break Online Session – 15 (in English) Head of Session: Dr. Nurhafizah Hasim 12:30 – 14:00 Universal Design of Apparel Labels for People with Visual Impairment Ilkay Ozsev Yuksek*, Busra Ozdemir, Izel Kabaagac, Pelin Altay, Sukriye Yuksel Filiz, Nevin Cigdem Gursoy
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xviii Systematic Study of Design and Operating Parameters in Forced Circulation Solar Water Heaters Ahmed Remlaoui*, Driss Nehari Comparative Study of the Corrosion Resistance of Magnesium and Zinc in Simulated Body Fluid Cristina Jimenez-Marcos, Francisco Miguel Sanchez-Sosa, Julia Claudia MirzaRosca, Ionelia Voiculescu*, Victor Geanta Numerical Investigation of the Behavior of Strip Footings under Eccentric Loading in Non-Homogeneous Clay Soils Nassima Zatar*, Alaoua Bouaicha Oxidation Behaviors of Nickel-Based Superalloys at 1050 °C Saida Bouyegh*, Samira Tlili Recent Progress on Nanomaterial Application for Improving Water-Based Drilling Fluid Siti Zulaika Razali*, Norizah Abdul Rahman, Mohd Hafizuddin Ab Ghani, Siti Hajar Othman, Tan Sin Tee, Robiah Yunus, Umer Rashid Valorization of Coal Bottom Ash in Sustainable Lightweight Self-Compacting Concrete Ibtissam Boulahya*, Abedlkadir Makkani 14:00 – 14:15 Break Online Session – 16 (in English) Head of Session: Asst. Prof. Dr. Sebahat Oztekin 14:15 – 15:45 Teaching Strategies, Methods, and Techniques Used by Science Teachers Ulas Kubat* The Impact of Out-Door Learning Environments on the Attainment of Objectives in the Teaching-Learning Process Ulas Kubat*, Matthew Price Use of Artificial Intelligence by Nursing Students: What Impact on Research Integrity? Imane Bettane*, Mohammed-Yassine Takzima, Amine Mohamed, Asma Sbai, Latifa Adarmouch Valorization of Waste Rosehip Seeds – A Green and Novel Procedure Abdulkadir Keskin, Betul Akhoroz, Gulin Amasya, Zerrin Sezgin Bayindir, Zekiye Goksel, Seda Kayahan, Yasin Ozdemir, Serpil Takac, Ayse Ezgi Unlu* Exploring the Determinants of Green Innovation Adoption Among SMEs in Albania Anxhela Bakiasi*, Oljam Dervishi, Merzai Bakiasi A Comparative Analysis of Drought Indices Using Remote Sensing and
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xix Geospatial Data Processing Amar Benlakhdar*, Abdelmoutia Telli, Nourelhouda Dekhili Environmentally Friendly Synthesis, Structural Characterization, Computational Analysis, and Biological Assessment of Benzodiazepine Derivatives Samir Hmaimou*, Marouane Ait Lahcen, Wiam Lahrich, Mohamed Adardour, Mohamed Maatallah, Abdesselam Baouid 15:45 – 16:00 Break Online Session – 17 (in English) Head of Session: Asst. Prof. Dr. Sebahat Oztekin 16:00 – 17:30 Enhancing Diesel Desulfurization via Oxidation over Modified USY Zeolite Catalyst Louiza Aichaoui*, Soraya Aidene, Boudjema Hamada Volatile Composition and Antioxidant activity of Blue Safflower Oil Yasmine Ouali*, Nacera Dahmani-Hamzaoui, Zahia Ghouila Formulation and Characterization of an Ointment Based on Hot Pepper Vegetable Oil Saida Touzouirt*, Selma Kadja, Sylia Badja Nanoencapsulation of Eucalyptus Oil for Aedes Aegypti Repellents Laura Astorga*, Gisela Romero, Marina Turrado, Emiliano Nicodemo, Jorge Montanari Antimicrobial Evaluation and Phytochemical Analysis of Asphodelus Microcarpus Extract Azziza Chabane Chaouch*, Farid Benkaci-Ali, Samira Tata Coagulation-Flocculation of Humic Substances: Effectiveness of Aluminum Sulfate and the Role of Sulfate and Phosphate Salts Lynda Hecini*, Fedia Bekiri, Naima Bacha, Wahida Kherifi Citric Acid-Crosslinked CMC Bioplastics: Tuning Flexibility, Strength, and Biodegradation for Sustainable Food Packaging Nur Alya Maisarah Binti Mohamad, Nurhafizah Binti Hasim* 17:30 – 17:45 Break Online Session – 18 (in English) Head of Session: Dr. Nurhafizah Hasim 17:45 – 19:15 On the Thermodynamic Stability of Some Metal Complexes within the Framework of a Density Functional Theory Approach Investigation Boulanouar Messaoudi*, Baian Alkassas Inverse Partial Least Squares Regression for the Prediction and Optimization of Carboxyl Group Content in Graphene Oxide Soraya Aidene*, Abdelsattar Osama Elemam Abdelhalim
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xx Biofunctionalized Gold Nanocomposites from Spatholobus Litoralis for HighSensitivity Refractive Index Sensing Firdausi Nuzulah, Sumini Sumini, Titin Syufairoh, Shelgiana Hayu Krisnanda, Nurul Hidayat* Mercury Sensing in Fish via Gold-Based SPR and Chitosan Composite Functionalization Kada Abdelhafid Meradi*, Mohamed Esseddik Ouardi, Fatima Tayeboun Enhancing the Thermal Stability of Deep Eutectic Polymer Electrolytes Through Nanoparticle Incorporation Hazimah Binti Hazman, Norshahirah Binti Mohamad Saidi* Comparative Green Synthesis of Zinc Nanoparticles via Pulsed Laser Ablation in Different Liquid Mediums Nur Nadirah Binti Mohd Nor, Fairuz Diyana Binti Ismail*, Maisarah Binti Duralim Study of Electrical Conductivity in Deep Eutectic Solvent Based on Hydrogen Bond Donor and Hydrogen Bond Acceptor Combinations Farhan Shazwan Shah Shahbani*, Norshahirah Mohamad Saidi, Nurhafizah Hasim, Nur Hidayah Ahmad, Muhammad Amirul Aizat Mohd Abdah 19:15 – 19:30 Break Online Session – 19 (in English) Head of Session: Asst. Prof. Dr. Latif Akcay 19:30 – 21:00 Grid-Connected Photovoltaic System with a Three-Level NPC Inverter for Power Quality Enhancement Fatima Tahri*, Ghrissi Tahri, Ali Tahri Dynamic Energy Management of Smart Microgrids Considering Renewable Variability Mohammed Tsebia*, Hamid Bentarzi, Fazia Ahcene, Djamila Talah, Azeddine Ratni Inexpensive, Thin, and Dual-Band Metamaterial Absorber with Inner-Nested Split Ring Resonators Yunus Kaya*, Ugur Cem Hasar, Mehmet Ertugrul Discriminating Natural Modes from Defect Signatures in Rolling Bearings Through Hybrid Vibration Modelling Azeddine Ratni*, Ali Damou, Djamel Benazzouz, Mohamed Tsebia Quantitative Structure–Toxicity Relationship Prediction of Ionic Liquid Toxicity Using an GWO-Optimized Support Vector Machine Hayet Abdellatif*, Maamer Laidi, Cherif Si-Moussa, Widad Benmouloud, Imane Euldji Bibliometric Analysis of AI-Driven Energy Harvesting and Prediction in
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxi Resource-Constrained IoT and Edge Systems Sami Acik, Selahattin Kosunalp*, Mustafa Tasci Numerical Analysis of the Seismic Bearing Capacity of Offshore Shallow Skirted Foundations on Sand Using the Pseudo-Static Approach Alaoua Bouaicha* Leveraging Large Language Models for Transport-Triggered Architecture Processor Design Latif Akcay* Monday, December 15, 2025 Online Sessions Virtual Hall Online Session – 20 (in English) Head of Session: Dr. Intan Helina Hasan 09:00 – 10:30 Evaluating the Role of Plant Growth-Promoting Rhizobacteria Strains in Boosting the Nutritional Quality of Trifolium Alexandrinum Yousra Debbah*, Mohamed Bencherchali, Saida Messgo-Moumene Sustainable Utilization of Mentha Pulegium by-Products: From Cellulose Microfibers to Fermentable Sugars Fatma Bhiri*, Feriel Lessig, Amir Bouallegue, Samira Abidi, Aida Ben Hassen Trabelsi Unlocking Aromatic Potential: Tripartite Rhizosphere Interactions Steer Metabolism for Stable Essential Oil Biosynthesis in Citronella Grass Ferota Larasati*, Sudiarso Sudiarso, Nunun Barunawati Potential of Plant Waste Ash Supplemented Culture Media for in Vitro Microtuber Induction in Solanum Tuberosum L. Amina Belguendouz*, Benamar Benmahioul Effects of Activated Charcoal Supplementation on in Vitro Microtuber Production and Quality in Potato Amina Belguendouz*, Benamar Benmahioul Lignocellulosic Biomass Valorization for Biogas Production Samira Abidi*, Arwa Masrouhi, Fatma Bhiri, Aida Ben Hassen Trabelsi Nanoencapsulation of Polyphenols from Native Schinus Molle and Non-Native Fruits for Sustainable Plant Growth Promotion Matias Aguilar*, Jorge Montanari, Luciano Gabbarini 10:30 – 10:45 Break Online Session – 21 (in English) Head of Session: Dr. Siti Zulaika Razali
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxii 10:45 – 12:15 Nurses’ Perceptions of Spirituality and Spiritual Care: A Cross-Sectional Study in Kabul, Afghanistan Farzana Mortazavi, Hatice Sutcu* The Knowledge and Attitudes of Iranian Nurses About Pain Management Elahe Mohammadian, Hatice Sutcu*, Fatemeh Bahramnezhad Exploring the Interplay Between Sulpiride and Physical Activity: Health Sciences Perspectives Hassina Fisli*, Mohamed Lyamine Chelaghmia Personalized Nutrition and Food Design Ali Khalfa*, Azzeddine Senouci, Djahira Hamed, Mounir Chihab, Sofiane Bouazza, Bensalah Fatima, Farid Bennabi Antibacterial Activity of Lactic Acid Bacteria Strains on Uropathogens Djamila Amamra*, Fadela Chougrani, Mansouria Belhocine, Abdelkader ElAmine Dahou Synthesis, Structural Elucidation, and Molecular Modeling Studies of 1,2,4Triazolo-1,5-Benzodiazepine Diastereoisomers as Promising Anti-Ebola Candidates Marouane Ait Lahcen*, Nouhaila Ait Lahcen, Saad Zekri, Samir Hmaimou, Mohamed Adardour, Ismail Hdoufane, Driss Cherqaoui, Abdesselam Baouid Detection of Multi-Drug-Resistant Extended Spectrum β-Lactamase Producing Enterobacteriaceae in Patients with Urinary Tract Infection Anfal Kara*, Naouel Boussoualim, Feryal Belfihadj, Meriem Elkolli 12:15 – 12:30 Break Online Session – 22 (in English) Head of Session: Dr. Intan Helina Hasan 12:30 – 14:00 Examining the Satisfaction Levels of Patients with Nursing Care in the Internal Medicine Ward Cigdem Koc, Halise Coskun* Impact of Child-Centered Empowerment on Lifestyle Behaviors in Children with Leukemia Pouran Varvani Farahani*, Candan Ozturk, Aziz Eghbali, Atefeh Rezapoor Identification of Potent Sulfonamide Derivatives Targeting MMP2 Through Pharmacophore Ligand-Based Modeling Saad Zekri*, Nouhaila Ait Lahcen, Adnane Ait Lahcen, Wissal Liman, Ismail Hdoufane, Driss Cherqaoui Valorization of Moroccan Medicinal Biodiversity as a Source of Novel Antileishmanial Agents Mohammed-Yassine Takzima*, Mohamed Echchakery, Mohamed Hafidi, Loubna El Fels
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxiii Epidemiological Profile and Phenotypic Characterization of AminoglycosideResistant Clinical Strains in the Setif Region Iman Krache*, Anfal Kara, Naouel Boussoualim, Zineb Daoudi, Noussaiba Douadi, Fatma Gridi Anti-Bacterial and Anti-Inflammatory Activities of the Mucus of the Snail Helix Aspersa Muller Imene Yahla* Computer-Aided Drug Discovery of Novel Ebola Virus Glycoprotein Inhibitors: Integrating QSAR, Fragment-Based Design, and Molecular Dynamics Nouhaila Ait Lahcen*, Wissal Liman, Saad Zekri, Mehdi Oubahmane, Ismail Hdoufane, Mohammed Mater Alanazi, Mohamed Maatallah, Driss Cherqaoui 14:00 – 14:15 Break Online Session – 23 (in English) Head of Session: Asst. Prof. Dr. Sebahat Oztekin 14:15 – 15:45 Biopolymeric Films of HEC/PAADDA Crosslinked with Glutaraldehyde as Controlled Release Systems of Tannic Acid for Antitumoral and Wound Healing Applications Isabela Viudes Rossatto Ferrarezi*, Ingrid Cristine de Sousa Everton, Júlia Monteiro Fernandes, Luca Kiichi Suzuki Trancolin, Lucas Gomes Nascimento, Marcos Elias da Silva Almeida, Maria Carolina Rodrigues Garcia, Mariana Aparecida Vieira, Weslley Aparecido Vicente Luiz, William Capellari Fumegali Infertility Among Women in Tebessa (Northeastern Algeria): Anthropometric and Biological Risk Determinants Khalida Abla*, Nassima Toumi-Halaimia, Sawssane Ziani, Ines Benamer, Asma Kraidia Impact of Traditional Phytotherapy on Hematological Parameters and Chemotherapy Tolerance in Breast Cancer Patients: A Case-Control Study in Tebessa, Algeria Khalida Abla*, Nassima Toumi-Halaimia, Sawssane Ziani Effects of Twin Hearts Meditation Versus Mandala Coloring on Practical Examination Anxiety in Undergraduate Nursing Students Pouran Varvani Farahani*, Samineh Esmailzadeh, Precious Chisom Uzoeghelu, Hatice Sutcu Body Image Perception Among Women After Mastectomy Mohamed Khalyfa*, Mohammed-Yassine Takzima, Imane Bettane, Youssef El Allam, Aya Ait Benaim In Silico Prediction of the Biological Activity of New Oxazole Derivatives Nadia Hadhoum*, Nassim Haddouche, Yasmine Hannachi, Sabine Herbi Circular Economic Model in the Hotel Industry for the Optimization of Solid and
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxiv Recyclable Waste: Case Study of the City of Essaouira, Morocco Assala Loukili*, Mohamed Hafidi, El Hassan El Mouden, Abdallah Nassour, Nour-El Houda Chaher, Loubna El Fels 15:45 – 16:00 Break Online Session – 24 (in English) Head of Session: Dr. Mohd Hafizuddin Ab Ghani 16:00 – 17:30 Food Safety and Microbial Control Ali Khalfa*, Azzeddine Senouci, Djahira Hamed, Mounir Chihab, Sofiane Bouazza, Bensalah Fatima, Farid Bennabi Knowledge of Leishmaniasis: A Comparative Review of Morocco and Ethiopia Mohammed-Yassine Takzima*, Mohamed Echchakery, Mohamed Hafidi, Loubna El Fels Chemical Profiling and Neuroprotective Activity in Elaeagnaceae Leaves Rayene Bouaita*, Randa Djemil, Samira Bouahlit, Saber Boutellaa, Chourouk Babouche, Zineb Bouamrane Nickel Sulfate Induced Hepatotoxicity Mediated Through Reactive Oxygen Species Generation and Impairing the Antioxidant Defense in Albino Rats Mohamed Khiari*, Youcef Bougoutaia, Nadjette Bourafa, Zine Kechrid A Green Nanotechnology Approach: Rice Husk-Derived Carbon Quantum Dots (CQDs) for Sustainable Applications Mohd Hafizuddin Ab Ghani*, Siti Hajar Othman, Siti Zulaika Razali, Mohd Ali Mat Nong, Josephine Liew Ying Chyi, Nishata Royan Rajendran Royan, Chen Ruey Shan, Johari Abdu Rahim Biochar Amendment and PGPR Inoculation Improve Growth, Nutrient Uptake, and Soil Fertility in Millet Asma Dahani*, Elmostapha Outamamat, Khalid Oufdou, Loubna El Fels Mapping the Physico-Chemical and Microbiological Characteristics of Olive Mill Wastewater Across Morocco: Advancing a Waste-to-Resource Circular Strategy Oumaima Dahbane*, Mohamed Hafidi, Mohammed Rhazi, Youness Bouhia 17:30 – 17:45 Break Online Session – 25 (in English) Head of Session: Asst. Prof. Dr. Mustafa Alptekin Engin 17:45 – 19:15 Key Scoring Models Used for Performance Evaluation of Employees: A Systemic Review and Its Applicability in the Albanian Banking Sector Sllavka Kurti, Katerina Vasili* Albania’s Perspective in Fighting High-Level Corruption Compared to Bulgaria, Before the Technical Closing of EU Accession Negotiations Anila Shehi*
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxxi Mansouria Sekkal, Badir Benkrelifa Lahouaria Alginate/Chitosan-Based Nanostructures as Sustainable Nanopriming Agents for Wheat Seeds Arruje Hameed, Muhammad Mujahid, Tahir Farooq, Amjad Hameed 174 A Natural Ally Against Pine Aphids: Pauesia Silana as a Key Biocontrol Agent of Cinara Maghrebica Leila Bourouba 175 Antagonistic Yeasts as Biocontrol Agents Against Phytopathogenic Fungi Kheira Hiba Benghaffor, Hadri Zouheyr 176 Morphological and Cultural Characterization of Verticillium Dahliae Kleb., the Causal Agent of Verticillium Wilt in Olive (Olea Europaea L.) Saliha Ogab, Houria Chaalal, Tahar Maza, Fatima Zohra Zoudji 177 Role of Edible Coating on Post-Harvest Management of Fruits and Vegetables Tusneem Kausar, Ashiq Hussain 178 Comparison the Values of Reverse Harmonic Index of Two Types Hexagonal Cactus Chain Graphs Mukaddes Okten Turaci 179 Chitosan-Modified Silver-Doped Phosphate Glass: Physical, Structural, and Mechanical Improvements Siti Norfariza Farhana Binti Mohd Razak, Nurhafizah Binti Hasim, Nur Hidayah Binti Ahmad, Norshahirah Binti Mohamad Saidi, Mohd Fuad Bin Mohamad 180 Methylammonium Iodide Doped CMC Electrolyte: A Safer and Environmentally Friendly Solution for Sustainable Power Storage Na’imah Husna Nasaruddin, Nur Hidayah Ahmad 181 A Green Laser Ablation Approach to Ag/Cinnamon Cassia Nanohybrid for Sensitive Colorimetric Glucose Sensing Applications Fitriyatun Naldiyah, Wulandari Dwi Lestari, Adeka Delvis Tama, Muhammad Dimas Adytia Airlangga, Nurul Hidayat 182 Composition-Dependent Thermal, Mechanical, and Electrical Responses of CMC–CA–Glycerol Biopolymeric Films Noorul Ain Binti Kamal Ariffin, Nurhafizah Binti Hasim 183 High-Sensitivity Atrazine Sensor Using Aluminum Plasmonics and Molecular Imprinting Mohamed Esseddik Ouardi, Kada Abdelhafid Meradi, Fatima Tayeboun 184 Tailoring the Structural, Electrical, and Magnetic Properties of Ni0.35Zn0.25Cd0.4Fe1.97Ce0.03O4/ GNPs Composites Muhammad Ajaz Un Nabi 185 Cadmium Sulfide Thin Film-based Photodetector: Fabrication and Photoresponse Evaluation Erman Erdogan 186 Experimental Evidence of Thermally Activated Transport at the Mobility Edge in 4H Silicon 187
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxxii Carbide Metal-Oxide-Semiconductor Field Effect Transistor Sabrina Meguellati, Mustapha Sarra Normalized Differential Conduction Analysis of Temperature-dependent Gate Conduction in 4H-SiC MOS Capacitor Sabrina Meguellati, Mustapha Sarra 188 Tailoring Phosphate Glass Performance: Structural, Mechanical, and Optical Enhancement via Chitosan Doping Nurhafizah Hasim 189 Effect of Incorporating Graphene Oxide Nanoparticles on Ion Mobility in Deep Eutectic Polymer Electrolyte Siti Norfazleen Farhana Binti Mohd Razak, Norshahirah Binti Mohamad Saidi 190 Composability of Nanofillers Carboxymethyl Cellulose Doped Ammonium Thiocyanate for Bioplastic Packaging Hazwani Nadirah Zamri, Nur Hidayah Ahmad 191 Study of the Dynamic of Benthic Diatom Communities in Response to Water Fluctuations Along Specific Rivers in Skikda Province, North East Algeria Hadjer Kaddeche 192 Discharge Prediction Using Artificial Neural Networks: A Case Study for the Enoree River, South Carolina, USA Betul Mete, Sinan Nacar, Adem Bayram 193 Stream Water Quality Assessment in a Dammed Watershed: A Case Study from North East Türkiye Adem Bayram, Betul Mete 200 Influence of Single-Layer Tile Waste Sand on Concrete Properties Yasmine Mohamed Bouteben, Leila Kherraf 207 Optimization of Unconfined Compressive Strength in Stabilized Soils Using Taguchi L9 Design Fadila Benayoun, Moufida Moussaoui, Souhila Rehab Bekkouche 208 Enhancing Facility Layout Design: An Integrated AHP–NLP Approach Muhammad Waqas Aslam, Zeqiang Zhang 209 Study of Earthquake Ground Motion Duration Recorded in Soft Soils Issam Aouari, Aicha Rouabeh, Benahmed Baizid, Rachid Bakhti 210 Approaches to Enhancing Energy Efficiency Through Building Envelope Insulation in Algeria Sadia Laidi, Sidi Mohamed Karim El Hassar, Achour Mahrane, Rabah Sellami, Ilyas Khelifa Kerfah 211 Utilising Graph Neural Networks for Research Paper Category Prediction and Similarity Search Bengisu Sahin, Sedanur Ozer, Emrah Inan 212 Leveraging Large Language Models for Event Detection in Water Resources Literature Bengisu Sahin, Ozge Yaren Turkseven, Emrah Inan 219
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxxiii Degradation Behavior of PET Polymers: Mechanical and FTIR Insights Ivana Salopek Čubrić, Goran Čubrić 225 Structural Performance of Hybrid Composite Patch Repairs in Double Lap Configurations with Unidirectional and Woven Fiber Reinforcements Faycal Mili, Walid Halloufi, Sarah Guenifa 226 Solar-Powered Pneumatic Water Pump: A Sustainable Solution for Water Supply in Off-Grid Areas Shoukat Mugheri, Touqeer Aslam, Mazhar Ali, Muhammad Kaleem, Umair Mehmood, Ayaz Ali, Ammar Asghar 227 Optimizing the Effect of Cow Dung on Swelling Soil Marwa Feligha, Souhila Rehab Bekkouche, Fadila Benayoun, Fatima Zohra Benamara 235 Structural Properties of Ni-Doped ZnO Thin Films Investigated by XRD, AFM, and Profilometry Maya Hanane Rezoug, Chewki Zegadi, Abdelkader Nouri, Nasr-Eddine Hamdadou, M’hamed Guezzoul 236 Immobilized Microalgae for Sustainable Soilless Agriculture in the Green Transition Burcu Simsek Uygun, Gizem Demirel 237 Implementation of Open Science Framework for Materials Engineering Elisabeth Viviana Lucero Baldevenites, Jose Rogelio Fung Corro, Yorlenis Martínez 244 Separation of Toluene – Cyclohexane Mixture Using Intensified Extraction Process by Imidazolium-Based Ionic Liquids Mohammed Djamel Eddine Allali, Hassiba Benyounes, Nesrine Amiri 245 Study of the Effectiveness of Steel Shavings as a Foaming Agent for the Production of Foams Glass Fayrouz Benhaoua, Djalila Aoufi, Nacira Stiti, Mehdia Toubane, Djedjiga Bousalah 246 Electromagnetic Frequency and Time Reaction in Electromechanical Systems Under Low Energy Mechanical Faults Azeddine Ratni, Ali Damou, Djamel Benazzouz, Mohamed Tsebia 247 Microstructural and Electrochemical Assessment of Co–Cr Dental Alloys in Ringer Solution Alberto Daniel Rico-Cano, Adriana Saceleanu, Anca Fratila, Julia Claudia Mirza-Rosca 248 Electrospinning of Polyacrylonitrile-Nanocellulose Composites Reinforced with Carbon Materials for Advanced Fiber Performance Mohd Ali Mat Nong, Juraina Md Yusof, Suzila Sabil, Mohd Hafizuddin Ab Ghani, Che Azurahanim Che Abdullah 249 Hybrid Voltage Regulation Strategy Combining AVR and Microgrid Coordination for Enhanced Power System Stability Fazia Ahcene, Hamid Bentarzi, Mohammed Tsebia, Djamila Talah 250 Next-Generation Deep Learning Architectures for Satellite Image Classification: Integrating Capsule Networks with CNNs and Transformers 251
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxxiv Boucif Beddad, Samiha Mezrar, Postaire Jack-Gerard Harvesting Waste Kinetic Energy from Vehicle Suspensions for Enhanced Mileage and Power Generation Touqeer Aslam, Shoukat Ali Mugheri, Ali Azam, Manthar Ali, Abbas Raza 252 Development of an Algorithm for Determining the Characteristics of the Ejection System Oleksandr Panevnyk 261 Virtual Students in Programming Fundamentals: Comparing Large Language Models with Vocational Computer Programming Students in Classical Exams Selma Bulut, Adem Korkmaz 262 AI-Based Anomaly Detection in Agricultural Farms Using Drone Data and Deep Learning Berrimi Fella 270 Bone Fracture Classification Analysis Base Machine Learning Algorithms Ei Phyu Sin Win 271 Large Language Model–Assisted Hardware Design: Insights from the Ascon-128 Implementation Latif Akcay 272 Indoor Particulate Matter Levels in the Department of Environmental Engineering at Eskisehir Technical University, Türkiye Melike Cengel, Ozlem Ozden Uzmez 277 Seismic Earth Pressures on Retaining Walls: A Comprehensive Review of Analytical, Numerical, and Experimental Approaches Ayman Gharbi, Fadoua Elkhannoussi, Bouraida Elyamouni, Abdellatif Khamlichi 288 Numerical Evaluation of Seismic Performance in Earth Dams: The Case of Fontaine Gazelles Dam, Biskra, Algeria Alaoua Bouaicha, Aissam Gaagai, Mosbah Ben Said, Ali Hachemi 289 Influence of Carbonation on Reinforced Concrete Structures in Southern Algeria Ben Ammar Ben Khadda 290 Numerical Study to the Enhancement of Pressure Distribution within the Alveolus of an Aerostaic Bearing Faiza Ghezali 291 A Comparative Finite Element Analysis of Static Coil Geometries for Electromagnetic Eddy Current Separation Amir Merahi, Kaouther Boutarfa, Adel Benabboun, Meriem Boumehed 292 Contribution to the Recovery of Paint Sludge in Wastewater Treatment Aicha Metali, Mounir Ziati, Ikram Rebiai, Khedidja Bachi 293 Microservice Transformation for Infrastructure Modernization of Payment and Electronic Money Institutions Ahmet Ulker, Cansu Barisici, Eren Capraz, Ilker Ogutcu, Mehmet Bulent Muslu, Zafer Golgeli, 294
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxxv Ceren Ulus, M. Fatih Akay Experimental Investigation of Concrete by Using Marble Waste as Replacement of Fine Aggregate Hassaan Amjad, Faisal Ahmed, Allah Noor 302 Investigating the Properties Ultra High-Performance Concrete Using Silica Fume and Ground Granulated Blast Furnace Slag Hassaan Amjad, Waseem Asghar, Dawood Jan 303 Production of Bioethanol Using Lower-Quality Algerian Sugar Dates Box-Behnken Kaouther Zeghida, Sarra Guilane, Leila Benmansour 304 Artificial Intelligence Control of Active and Reactive Power for a Three-Level NPC Inverter Connected to Grid Utility Ghrissi Tahri, Fatima Tahri, Ali Tahri 305 Mechanical Properties of Concrete Reinforced with Steel Fiber Noshad Ali, Waqar Ali, Muhammad Hessib 306 Influx of Different Growth Geometries on Titanium Thin Film for Medical Applications Matteo Bertapelle, Joel Borges, Julia Claudia Mirza Rosca, Filipe Vaz 307 Universal Design of Apparel Labels for People with Visual Impairment Ilkay Ozsev Yuksek, Busra Ozdemir, Izel Kabaagac, Pelin Altay, Sukriye Yuksel Filiz, Nevin Cigdem Gursoy 308 Systematic Study of Design and Operating Parameters in Forced Circulation Solar Water Heaters Ahmed Remlaoui, Driss Nehari 312 Comparative Study of the Corrosion Resistance of Magnesium and Zinc in Simulated Body Fluid Cristina Jimenez-Marcos, Francisco Miguel Sanchez-Sosa, Julia Claudia Mirza-Rosca, Ionelia Voiculescu, Victor Geanta 313 Numerical Investigation of the Behavior of Strip Footings under Eccentric Loading in NonHomogeneous Clay Soils Nassima Zatar, Alaoua Bouaicha 314 Oxidation Behaviors of Nickel-Based Superalloys at 1050 °C Saida Bouyegh, Samira Tlili 315 Recent Progress on Nanomaterial Application for Improving Water-Based Drilling Fluid Siti Zulaika Razali, Norizah Abdul Rahman, Mohd Hafizuddin Ab Ghani, Siti Hajar Othman, Tan Sin Tee, Robiah Yunus, Umer Rashid 316 Valorization of Coal Bottom Ash in Sustainable Lightweight Self-Compacting Concrete Ibtissam Boulahya, Abedlkadir Makkani 317 Teaching Strategies, Methods, and Techniques Used by Science Teachers Ulas Kubat 318
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxxvi The Impact of Out-Door Learning Environments on the Attainment of Objectives in the Teaching-Learning Process Ulas Kubat, Matthew Price 319 Use of Artificial Intelligence by Nursing Students: What Impact on Research Integrity? Imane Bettane, Mohammed-Yassine Takzima, Amine Mohamed, Asma Sbai, Latifa Adarmouch 320 Valorization of Waste Rosehip Seeds – A Green and Novel Procedure Abdulkadir Keskin, Betul Akhoroz, Gulin Amasya, Zerrin Sezgin Bayindir, Zekiye Goksel, Seda Kayahan, Yasin Ozdemir, Serpil Takac, Ayse Ezgi Unlu 321 Exploring the Determinants of Green Innovation Adoption Among SMEs in Albania Anxhela Bakiasi, Oljam Dervishi, Merzai Bakiasi 322 A Comparative Analysis of Drought Indices Using Remote Sensing and Geospatial Data Processing Amar Benlakhdar, Abdelmoutia Telli, Nourelhouda Dekhili 327 Environmentally Friendly Synthesis, Structural Characterization, Computational Analysis, and Biological Assessment of Benzodiazepine Derivatives Samir Hmaimou, Marouane Ait Lahcen, Wiam Lahrich, Mohamed Adardour, Mohamed Maatallah, Abdesselam Baouid 332 Enhancing Diesel Desulfurization via Oxidation over Modified USY Zeolite Catalyst Louiza Aichaoui, Soraya Aidene, Boudjema Hamada 333 Volatile Composition and Antioxidant activity of Blue Safflower Oil Yasmine Ouali, Nacera Dahmani-Hamzaoui, Zahia Ghouila 334 Formulation and Characterization of an Ointment Based on Hot Pepper Vegetable Oil Saida Touzouirt, Selma Kadja, Sylia Badja 335 Nanoencapsulation of Eucalyptus Oil for Aedes Aegypti Repellents Laura Astorga, Gisela Romero, Marina Turrado, Emiliano Nicodemo, Jorge Montanari 336 Antimicrobial Evaluation and Phytochemical Analysis of Asphodelus Microcarpus Extract Azziza Chabane Chaouch, Farid Benkaci-Ali, Samira Tata 337 Coagulation-Flocculation of Humic Substances: Effectiveness of Aluminum Sulfate and the Role of Sulfate and Phosphate Salts Lynda Hecini, Fedia Bekiri, Naima Bacha, Wahida Kherifi 338 Citric Acid-Crosslinked CMC Bioplastics: Tuning Flexibility, Strength, and Biodegradation for Sustainable Food Packaging Nur Alya Maisarah Binti Mohamad, Nurhafizah Binti Hasim 339 On the Thermodynamic Stability of Some Metal Complexes within the Framework of a Density Functional Theory Approach Investigation Boulanouar Messaoudi, Baian Alkassas 340 Inverse Partial Least Squares Regression for the Prediction and Optimization of Carboxyl Group Content in Graphene Oxide 341
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxxvii Soraya Aidene, Abdelsattar Osama Elemam Abdelhalim Biofunctionalized Gold Nanocomposites from Spatholobus Litoralis for High-Sensitivity Refractive Index Sensing Firdausi Nuzulah, Sumini Sumini, Titin Syufairoh, Shelgiana Hayu Krisnanda, Nurul Hidayat 342 Mercury Sensing in Fish via Gold-Based SPR and Chitosan Composite Functionalization Kada Abdelhafid Meradi, Mohamed Esseddik Ouardi, Fatima Tayeboun 343 Enhancing the Thermal Stability of Deep Eutectic Polymer Electrolytes Through Nanoparticle Incorporation Hazimah Binti Hazman, Norshahirah Binti Mohamad Saidi 344 Comparative Green Synthesis of Zinc Nanoparticles via Pulsed Laser Ablation in Different Liquid Mediums Nur Nadirah Binti Mohd Nor, Fairuz Diyana Binti Ismail, Maisarah Binti Duralim 345 Study of Electrical Conductivity in Deep Eutectic Solvent Based on Hydrogen Bond Donor and Hydrogen Bond Acceptor Combinations Farhan Shazwan Shah Shahbani, Norshahirah Mohamad Saidi, Nurhafizah Hasim, Nur Hidayah Ahmad, Muhammad Amirul Aizat Mohd Abdah 346 Grid-Connected Photovoltaic System with a Three-Level NPC Inverter for Power Quality Enhancement Fatima Tahri, Ghrissi Tahri, Ali Tahri 347 Dynamic Energy Management of Smart Microgrids Considering Renewable Variability Mohammed Tsebia, Hamid Bentarzi, Fazia Ahcene, Djamila Talah, Azeddine Ratni 348 Inexpensive, Thin, and Dual-Band Metamaterial Absorber with Inner-Nested Split Ring Resonators Yunus Kaya, Ugur Cem Hasar, Mehmet Ertugrul 349 Discriminating Natural Modes from Defect Signatures in Rolling Bearings Through Hybrid Vibration Modelling Azeddine Ratni, Ali Damou, Djamel Benazzouz, Mohamed Tsebia 354 Quantitative Structure–Toxicity Relationship Prediction of Ionic Liquid Toxicity Using an GWO-Optimized Support Vector Machine Hayet Abdellatif, Maamer Laidi, Cherif Si-Moussa, Widad Benmouloud, Imane Euldji 355 Bibliometric Analysis of AI-Driven Energy Harvesting and Prediction in Resource-Constrained IoT and Edge Systems Sami Acik, Selahattin Kosunalp, Mustafa Tasci 356 Numerical Analysis of the Seismic Bearing Capacity of Offshore Shallow Skirted Foundations on Sand Using the Pseudo-Static Approach Alaoua Bouaicha 362 Leveraging Large Language Models for Transport-Triggered Architecture Processor Design Latif Akcay 363
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxxviii Evaluating the Role of Plant Growth-Promoting Rhizobacteria Strains in Boosting the Nutritional Quality of Trifolium Alexandrinum Yousra Debbah, Mohamed Bencherchali, Saida Messgo-Moumene 368 Sustainable Utilization of Mentha Pulegium by-Products: From Cellulose Microfibers to Fermentable Sugars Fatma Bhiri, Feriel Lessig, Amir Bouallegue, Samira Abidi, Aida Ben Hassen Trabelsi 369 Unlocking Aromatic Potential: Tripartite Rhizosphere Interactions Steer Metabolism for Stable Essential Oil Biosynthesis in Citronella Grass Ferota Larasati, Sudiarso Sudiarso, Nunun Barunawati 370 Potential of Plant Waste Ash Supplemented Culture Media for in Vitro Microtuber Induction in Solanum Tuberosum L. Amina Belguendouz, Benamar Benmahioul 371 Effects of Activated Charcoal Supplementation on in Vitro Microtuber Production and Quality in Potato Amina Belguendouz, Benamar Benmahioul 372 Lignocellulosic Biomass Valorization for Biogas Production Samira Abidi, Arwa Masrouhi, Fatma Bhiri, Aida Ben Hassen Trabelsi 373 Nanoencapsulation of Polyphenols from Native Schinus Molle and Non-Native Fruits for Sustainable Plant Growth Promotion Matias Aguilar, Jorge Montanari, Luciano Gabbarini 374 Nurses’ Perceptions of Spirituality and Spiritual Care: A Cross-Sectional Study in Kabul, Afghanistan Farzana Mortazavi, Hatice Sutcu 375 The Knowledge and Attitudes of Iranian Nurses About Pain Management Elahe Mohammadian, Hatice Sutcu, Fatemeh Bahramnezhad 376 Exploring the Interplay Between Sulpiride and Physical Activity: Health Sciences Perspectives Hassina Fisli, Mohamed Lyamine Chelaghmia 377 Personalized Nutrition and Food Design Ali Khalfa, Azzeddine Senouci, Djahira Hamed, Mounir Chihab, Sofiane Bouazza, Bensalah Fatima, Farid Bennabi 378 Antibacterial Activity of Lactic Acid Bacteria Strains on Uropathogens Djamila Amamra, Fadela Chougrani, Mansouria Belhocine, Abdelkader El-Amine Dahou 379 Synthesis, Structural Elucidation, and Molecular Modeling Studies of 1,2,4-Triazolo-1,5Benzodiazepine Diastereoisomers as Promising Anti-Ebola Candidates Marouane Ait Lahcen, Nouhaila Ait Lahcen, Saad Zekri, Samir Hmaimou, Mohamed Adardour, Ismail Hdoufane, Driss Cherqaoui, Abdesselam Baouid 380 Detection of Multi-Drug-Resistant Extended Spectrum β-Lactamase Producing Enterobacteriaceae in Patients with Urinary Tract Infection 381
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xxxix Anfal Kara, Naouel Boussoualim, Feryal Belfihadj, Meriem Elkolli Examining the Satisfaction Levels of Patients with Nursing Care in the Internal Medicine Ward Cigdem Koc, Halise Coskun 382 Impact of Child-Centered Empowerment on Lifestyle Behaviors in Children with Leukemia Pouran Varvani Farahani, Candan Ozturk, Aziz Eghbali, Atefeh Rezapoor 383 Identification of Potent Sulfonamide Derivatives Targeting MMP2 Through Pharmacophore Ligand-Based Modeling Saad Zekri, Nouhaila Ait Lahcen, Adnane Ait Lahcen, Wissal Liman, Ismail Hdoufane, Driss Cherqaoui 384 Valorization of Moroccan Medicinal Biodiversity as a Source of Novel Antileishmanial Agents Mohammed-Yassine Takzima, Mohamed Echchakery, Mohamed Hafidi, Loubna El Fels 385 Epidemiological Profile and Phenotypic Characterization of Aminoglycoside-Resistant Clinical Strains in the Setif Region Iman Krache, Anfal Kara, Naouel Boussoualim, Zineb Daoudi, Noussaiba Douadi, Fatma Gridi 386 Anti-Bacterial and Anti-Inflammatory Activities of the Mucus of the Snail Helix Aspersa Muller Imene Yahla 387 Computer-Aided Drug Discovery of Novel Ebola Virus Glycoprotein Inhibitors: Integrating QSAR, Fragment-Based Design, and Molecular Dynamics Nouhaila Ait Lahcen, Wissal Liman, Saad Zekri, Mehdi Oubahmane, Ismail Hdoufane, Mohammed Mater Alanazi, Mohamed Maatallah, Driss Cherqaoui 388 Biopolymeric Films of HEC/PAADDA Crosslinked with Glutaraldehyde as Controlled Release Systems of Tannic Acid for Antitumoral and Wound Healing Applications Isabela Viudes Rossatto Ferrarezi, Ingrid Cristine de Sousa Everton, Júlia Monteiro Fernandes, Luca Kiichi Suzuki Trancolin, Lucas Gomes Nascimento, Marcos Elias da Silva Almeida, Maria Carolina Rodrigues Garcia, Mariana Aparecida Vieira, Weslley Aparecido Vicente Luiz, William Capellari Fumegali 389 Infertility Among Women in Tebessa (Northeastern Algeria): Anthropometric and Biological Risk Determinants Khalida Abla, Nassima Toumi-Halaimia, Sawssane Ziani, Ines Benamer, Asma Kraidia 401 Impact of Traditional Phytotherapy on Hematological Parameters and Chemotherapy Tolerance in Breast Cancer Patients: A Case-Control Study in Tebessa, Algeria Khalida Abla, Nassima Toumi-Halaimia, Sawssane Ziani 402 Effects of Twin Hearts Meditation Versus Mandala Coloring on Practical Examination Anxiety in Undergraduate Nursing Students Pouran Varvani Farahani, Samineh Esmailzadeh, Precious Chisom Uzoeghelu, Hatice Sutcu 403 Body Image Perception Among Women After Mastectomy Mohamed Khalyfa, Mohammed-Yassine Takzima, Imane Bettane, Youssef El Allam, Aya Ait Benaim 404
2nd International Conference on Multidisciplinary Sciences and Technological Developments ICMUSTED 2025 December 12-15 xl In Silico Prediction of the Biological Activity of New Oxazole Derivatives Nadia Hadhoum, Nassim Haddouche, Yasmine Hannachi, Sabine Herbi 405 Circular Economic Model in the Hotel Industry for the Optimization of Solid and Recyclable Waste: Case Study of the City of Essaouira, Morocco Assala Loukili, Mohamed Hafidi, El Hassan El Mouden, Abdallah Nassour, Nour-El Houda Chaher, Loubna El Fels 406 Food Safety and Microbial Control Ali Khalfa, Azzeddine Senouci, Djahira Hamed, Mounir Chihab, Sofiane Bouazza, Bensalah Fatima, Farid Bennabi 407 Knowledge of Leishmaniasis: A Comparative Review of Morocco and Ethiopia Mohammed-Yassine Takzima, Mohamed Echchakery, Mohamed Hafidi, Loubna El Fels 408 Chemical Profiling and Neuroprotective Activity in Elaeagnaceae Leaves Rayene Bouaita, Randa Djemil, Samira Bouahlit, Saber Boutellaa, Chourouk Babouche, Zineb Bouamrane 409 Nickel Sulfate Induced Hepatotoxicity Mediated Through Reactive Oxygen Species Generation and Impairing the Antioxidant Defense in Albino Rats Mohamed Khiari, Youcef Bougoutaia, Nadjette Bourafa, Zine Kechrid 410 A Green Nanotechnology Approach: Rice Husk-Derived Carbon Quantum Dots (CQDs) for Sustainable Applications Mohd Hafizuddin Ab Ghani, Siti Hajar Othman, Siti Zulaika Razali, Mohd Ali Mat Nong, Josephine Liew Ying Chyi, Nishata Royan Rajendran Royan, Chen Ruey Shan, Johari Abdu Rahim 411 Biochar Amendment and PGPR Inoculation Improve Growth, Nutrient Uptake, and Soil Fertility in Millet Asma Dahani, Elmostapha Outamamat, Khalid Oufdou, Loubna El Fels 412 Mapping the Physico-Chemical and Microbiological Characteristics of Olive Mill Wastewater Across Morocco: Advancing a Waste-to-Resource Circular Strategy Oumaima Dahbane, Mohamed Hafidi, Mohammed Rhazi, Youness Bouhia 413 Key Scoring Models Used for Performance Evaluation of Employees: A Systemic Review and Its Applicability in the Albanian Banking Sector Sllavka Kurti, Katerina Vasili 414 Albania’s Perspective in Fighting High-Level Corruption Compared to Bulgaria, Before the Technical Closing of EU Accession Negotiations Anila Shehi 425 Exploratory Factor Analysis of Mobile Application Usage Among Youth Associations in Malaysia Mohd Hazlami Jusoh, Noor Aisyah Abdul Aziz, Mohd Yusri Ibrahim 429 Understanding the Role of AI Literacy in Work Engagement: The Mediating Effects of Job Crafting and Job Insecurity, and the Moderating Role of Regulatory Focus 430
6 The cellular network model depicted in Figure 1 explained is the network structure that forms the foundation of mobile communication technologies and enables user equipments (UEs) to communicate with core networks without any location constraints. Young’s work [18] addresses the working principle of this system, explaining that the service area is divided into hexagonal-shaped cells, and the mobile telephone switching office (MTSO), located at the center of the system, is the hub that manages location information and call details, simultaneously implements billing and charging services, and controls the cell sites. Figure 1. 1G architecture [19] 3.1.2. Global Standards and Comparative Analysis Firstly, the NTT standard was launched in Tokyo in 1979. It quickly became the country’s first national 1G network. On the other hand, the NMT system, an analog standard, was initiated by public telephone operators in Finland, Sweden, Norway, and Denmark, creating the world’s first multinational standardized mobile telephone system. The two most popular systems in Europe were the NMTs and TACS. Apart from NMT and TACS, other analog systems were also introduced across Europe in the 1980s. Each of these systems had capabilities such as handoff and roaming; however, the biggest disadvantage of this era was that the cellular networks did not operate internationally [1, 19]. 3.2. 2G Mobile Communication Services 3.2.1. The Advent and Motivation of 2G Technology 2G was introduced in the early 1990s, and the dominant 2G standard became the global system for mobile communications (GSM). GSM was first launched in Finland in 1991 with the aims of supporting roaming across national borders and having a higher capacity compared to 1G. While initially used only for voice transmission, it was later also utilized for data communication, such as the SMS and internet access [20, 21]. In essence, 2G is the first digital mobile communication technology developed to overcome the limitations of 1G, primarily the restriction on international roaming and the constraints imposed by its analog structure [21]. This transition from analog to digital was made possible by advancements in Metal-Oxide Semiconductor Field-Effect transistor (MOSFET) technology [1]. 3.2.2. Core 2G Architecture: GSM Within each cell, there are stations called base transceiver station (BTS) that allow users to access the network. Control centers, called base station controller (BSC), are connected to the mobile switching center (MSC), which facilitates call establishment. This center ensures the establishment and termination of all telephone calls [22]. The first GSM systems utilized the 900 MHz band and a 25 MHz frequency spectrum [23]. Frequency division multiple access (FDMA) was used to divide the available bandwidth into 124 carrier frequencies, each of which was 200 kHz. Subsequently, each frequency was divided into eight time slots using time division multiple access (TDMA), which allowed the same frequency to accommodate eight different simultaneous calls [21].
7 Figure 2. GSM architecture [1] 3.2.3. Data Rate and Content Limitations (Performance Issues) All 2G systems are voice-centric. While GSM includes SMS in addition to voice, it actually supports data transfer over voice channels; however, it typically does so at slow speeds, such as 9.6 kb/s or 14.4 kb/s. Consequently, voice remained the dominant service in the 2G world; as a result, 2G eventually failed to meet changing demands [24, 25]. 3.2.4. The Evolution of 2G: High-Speed Circuit-Switched Data (HSCSD), 2.5G (General Packet Radio Service (GPRS)), and 2.75G (Enhanced Data rates for Global Evolution (EDGE)) HSCSD is an application designed to increase 2G’s speed by combining time slots and utilize the existing structure for bursty internet data. A maximum data rate of 14.4 kbit/s could be transferred over a single existing time slot. HSCSD could combine and allocate multiple channels to the user, and thus it was calculated that speeds up to 115.2 kbit/s could be reached with the use of 8 channels; however, this remained theoretical because the MSC could only allow speeds up to a maximum of 64 kbit/s for each connection. In practical application, it was limited to a speed of 57.6 kbit/s using 4 channels [21, 22]. 3.2.5. GPRS and the Packet Switching Revolution The GPRS system served as a bridge between third-generation and second-generation technologies and was named as a packet-switched radio service. This sub-version utilized Packet Switching, meaning it added IP support to the existing GSM infrastructure [22]. With 2.5G also proving insufficient, the system moved away from the currently used Gaussian minimum shift keying (GMSK) modulation technique to a more efficient one: 8-phase shift keying (8PSK). However, the biggest innovation in this regard was the adoption of the adaptive modulation and coding (AMC) technique. With this technique, the appropriate modulation scheme is used according to the channel conditions: 8PSK, which is very sensitive to noise, is preferred under good channel conditions; but if suitable conditions are not met and the noise is high, the system automatically switches to GMSK, which is more noise-resistant but less spectrally efficient than 8PSK [4]. This evolution of 2G extending to 2.75G is essentially a story of an architecture, originally designed for voice calls and messaging, attempting to adapt to the different nature of data traffic brought by the internet. However,
8 regardless of the efforts made, the progress of mobile communication and the increasing data demands of users could not be halted. 3.3. 3G Mobile Communication Systems 3.3.1. The Global Standards Quest and 3G’s Objectives Initial plans made for 3G in the 1980s were for video conferencing applications using mobile phones. The 3G concept evolved when it was understood that the true focus was the internet. In response to this data demand, the ITU created an international standard for 3G in the 90s. This standard was named IMT-2000; a minimum speed target of 200 kbit/s was set [5, 26]. However, just like with 2G, a kind of standards war broke out for 3G. While European and Japanese GSM operators supported the WCDMA standard, which was the natural evolution of the GSM infrastructure, operators in North America began using the CDMA2000 standard, based on their own 2G evolution. China, meanwhile, developed its own standard, time division - synchronous code division multiple access (TD-SCDMA) [27, 28]. Table 1. The evolution of 3G [28] No Generation 3G 3.5G 3.75G 1 Year 2000 2003 2003 2 Frequency 1.6–2 GHz 1.6–2.5 GHz 1.6–2.5 GHz 3 Technology FDD/TDD GSM/2GPP GSM/3GPP 4 Service Voice and Data Voice and Data High Speed Internet and Multimedia FDD: Frequency division duplex and TDD: Time division duplex Due to this mixed market and commercial competition, a global standard still could not be adopted. As Mendes et al. pointed out: It was harder to agree on frequency bands for IMT-2000, and the resulting compromise included five different radio standards and three separate frequency bands. This situation caused 3G to be born in the form of various standards [24]. 3.3.2. The GSM Path and Evolution: From WCDMA to High Speed Packet Access Plus (HSPA+) As mentioned, the gateway to 3G for Europe and Asia was universal mobile telecommunications system (UMTS), which emerged from the evolution of GSM. The UMTS system was built upon the packet-switched foundations of 2.5G. In this improved architecture, the node B replaced the BTS of 2G, and the radio network controller (RNC) replaced the BSC [20]. Figure 3. UMTS architecture [29]
9 3.3.3. UMTS Radio Layer Details (Physical Layer) The main air interface of UMTS is WCDMA, which is referred to as universal terrestrial radio access (UTRA). The operation of the UTRA network (UTRAN) is asynchronous, and both FDD and TDD modes can be configured to achieve data rates up to 2 Mbps. The WCDMA carrier spacing is approximately 5 MHz, with a chip rate of 3.84 Mcps and a frame length of 10 ms. WCDMA utilizes the radio spectrum efficiently because the CDMA technique allows all base stations to use the same frequency. Furthermore, WCDMA has additional advantages such as high transfer speed, increased system capacity through statistical multiplexing, and improved communication quality [24, 30]. 3.3.4. Core UMTS and Speed Tiers The initial trials of UMTS were designed to offer speeds that varied according to the user. Data rates targeting up to 144 kbps for a user traveling in a vehicle, 384 Kbps for a pedestrian, and up to 2 Mbps for a stationary, indoor user were targeted [26]. Higher data rates (such as 2 Mbps) utilize more complex modulation and coding schemes that require flawless signal transmission. These complex schemes are more sensitive to signal distortions that occur while moving at high speeds. This is why system engineers were driven to implement this tiered speed format. 3.3.5. Factors Triggering 4G and the Advent of a New Era The new consumption perception initiated by 3G exceeded its scope and gave rise to the need for a new technology. 4G, which would address these needs, does not just mean fast internet; it also means a technology capable of providing smart, ubiquitous, seamless, and personalized service. Jiang and Han stated that the exponential growth in data traffic was fueled by the increase in mobile broadband traffic [2]. Furthermore, Ekström expressed that both the number of mobile broadband subscribers and the traffic volume per subscriber were increasing rapidly [18]; however, this created a major challenge for cellular networks, which were primarily designed for voice communication. The goal of 4G was precisely to overcome this major challenge: the necessity for cellular networks to operate both a packet-switched network for data delivery and a circuit-switched network for voice calls. With 4G, the inefficient, complex, and dual-network structure of 3G was ended, and a transition to an all-packet-switched network architecture was made. 3.3.6. Establishing Official Standards for True 4G: IMT-Advanced The ITU-R group outlined the basic technical requirements for the 4G system in a report published in 2008. However, it was understood that Release 8, which we call LTE and which was the initial equivalent of 4G, could not fully meet these requirements. For instance, the expected peak data rates of 1 Gbps for low mobility and 100 Mbps for high mobility were not entirely satisfied [2]. As a result of this situation, Release 10, known as LTE-Advanced, began to be developed by 3GPP and was approved as an IMT-Advanced technology in November 2010 [31, 32]. 3.4. 4G Market: LTE, HSPA+, and the Rival Worldwide Interoperability for Microwave Access (WiMAX) 3.4.1. The Two Main Market Rivals: LTE and WiMAX The market journey of 4G, similar to previous generation technologies, began once again with standards competition. As Khan et al. pointed out, while operators were trying to decide on the best standard for long-term investment, there were two distinct standards: WiMAX, which had a technological head start, on one side, and LTE, which offered a clean transition from 3G, on the other [25]. As also noted by Jiang and Han, WiMAX was commercialized earlier than LTE and enjoyed a period of performance superiority between 2005 and 2009; it was at the forefront of adopting MIMO and OFDM technologies [2].
10 The biggest reason for LTE gaining market dominance was its ability to offer a smooth transition from dominant standards like GSM and WCDMA, meaning it was an evolution of the previous technology. On the other hand, WiMAX was referred to as a disruptive technology serving small users. Ultimately, major mobile operators such as Verizon, Vodafone, China Mobile, NTT, and Deutsche Telekom chose to seamlessly upgrade their existing 3G infrastructures to LTE rather than taking on the risk and high cost of a disruptive standard [2]. 3.4.2. LTE Architecture (Release 8) LTE is a disruptive and pioneering standard, designed without regard for backward compatibility constraints. The system design has been simplified and significantly flattened to reduce latency. One of the most concrete examples of this flat architecture is that while 3G had a central RNC managing the node B (the 3G base station), this structure was eliminated with LTE Release 8. The station, now called the evolved node B (eNodeB) is much smarter and makes decisions regarding resource management and handover internally. This allows decisions to be made faster, without intermediate processing [2]. Figure 4. LTE architecture [33] Guo et al. stated that the LTE architecture, known as evolved packet system (EPS), consists of two main parts: the evolved universal terrestrial radio access network (E-UTRAN) and the EPC [34]. They summarized the EPC architecture as follows: • Mobility Management Entity (MME): The MME controls the control plane signaling between the user and the EPC. • Serving Gateway (S-GW): The S-GW is responsible for routing and forwarding data packets between the users and the eNodeB. It is also responsible for acting as the mobility anchor... • Packet Data Network Gateway (PDN-GW): The PDN-GW serves as an interconnection point between the EPC and external packet data networks. • Home Subscriber Server (HSS): The HSS is a database that contains all relevant user and subscriber-related information. 3.4.3. The Journey to True 4G, LTE-Advanced, and Beyond It was understood that LTE, upon the release of Release 8, could not fully meet the targets set by IMT. The expected peak data rates of 1 Gbps for low mobility and 100 Mbps for high mobility were not entirely satisfied. To reach these targets, the LTE-Advanced 4G technology was officially approved [32]. 3.4.4. Technologies Enabling LTE-Advanced to be True 4G i. Carrier Aggregation (CA): LTE-Advanced allows up to five component carriers with different bandwidths to be combined (Carrier Aggregation) in order to achieve a bandwidth of 100 MHz for a single terminal. However, the cost of using this feature is that the combination of different frequencies in the same device leads to engineering problems, such as the device’s own signal interfering with itself [35].
11 ii. Advanced MIMO (eMIMO) and Multi-User MIMO (MU-MIMO): Analogically speaking, lanes were added to the data highway expanded by carrier aggregation using the eMIMO technique, allowing for more data flow simultaneously. Eluwole et al. stated that this enhancement offers more layers to a single user with single-user MIMO (SU-MIMO) and also allows different layers to be assigned simultaneously with MU-MIMO [36]. They also noted that switching between SU-MIMO and MU-MIMO is much more advantageous than using only singleuser MIMO [36]. 3.5. 5G Mobile Communication Systems 3.5.1. From the Limits of 4G to the 5G Vision (IMT-2020) We are witnessing an exponential increase in the amount of mobile traffic. Saghezchi et al. noted that mobile data traffic doubled between 2010 and 2011 [6]. Furthermore, this increase is projected to be about 20,000 times from 2010 to 2030 [40]. The surge is primarily driven by the rise in mobile device usage and mobile video consumption, alongside the changes in consumption habits mentioned for 4G [6]. One of the biggest factors is that the expected end-to-end latency for 5G is one-tenth of 4G’s [36]. As stated in both references [7] and [36], 5G’s user plane latency will be even lower than 1 ms. This low latency makes it possible to handle demanding services and applications—such as autonomous cars, augmented reality, or ultrahigh-resolution multimedia streaming—that push the limits of 4G and require 100% reliability, as also indicated by reference [37], unlike 4G. 3.5.2. Three Core Use Cases Shaping 5G i. Enhanced Mobile Broadband (eMBB): This is a higher stage of the mobile broadband technology offered by 4G. The International Telecommunication Union Radiocommunication Sector noted that it offers high speeds at Gbps levels for users and can be utilized for applications such as ultra-high-definition video streaming, 3D video, augmented reality, virtual reality, and cloud-based gaming [7]. ii. URLLC: One of 5G’s most revolutionary promises, URLLC is a technology that provides extremely low endto-end latency for applications requiring ultra-high reliability, where interruption or delay is unacceptable. It is critical for applications such as industrial automation and control, autonomous vehicles, and intelligent transportation systems [7]. iii. mMTC: This is associated with communicating with a massive number of connected devices that perform lowvolume data transmission and are not delay-sensitive. References [36] and [37] mentioned a target of 1 million devices per km2 and noted that latency tolerance is high; they also mentioned that network slicing is essential for separating mMTC traffic from other traffic types. It is critical for applications such as smart cities, smart buildings, smart agriculture, and smart meters [38]. 3.5.3. 5G’s Core Technological Pillars and Architecture i. New Radio (NR) and Spectrum Utilization: The first 5G NR specifications were approved by 3GPP in December 2017, paving the way for trial experiments and commercial deployment [39]. Reference [38] explained that 5G NR’s flexible frame structure is designed to meet the diverse requirements (such as low latency, high speed, and efficiency) of 5G services like eMBB, URLLC, and mMTC, compared to the more rigid structure of 4G LTE. Reference [37] stated that the interoperability of the 4G RAN architecture with the new radio access technology is an important design requirement. ii. 5G Core Network (5GC): The Center of Flexibility and Intelligence: In 5G, network functions are partitioned based on their services and communicate with each other via a service-based interface [40]: • Core Access and Mobility Management Function (AMF): It has functions such as access, authentication and authorization, registration management, mobility management, session management function (SMF) selection, and arrangement of network slice selection [36, 41]. • SMF: It is responsible for session management, IP allocation for users, enforcement of user subscription policies, and control of quality of service (QoS) [41, 42].
12 • User Plane Function (UPF): It is responsible for functions such as packet routing and forwarding, packet inspection of traffic, and traffic usage reporting [36]. It is stated that it is geographically located closer to endusers to reduce latency [41]. • Authentication Server Function (AUSF): It is the authentication server function; it enables the AMF to verify the UE [43]. • Policy Control Function (PCF): The PCF generates and decides the control framework for policies and decisions; that is, it provides the policy frameworks for roaming, network slicing, mobility management, session management, and QoS support [36, 40]. • Unified Data Management (UDM): It contains data related to the HSS; meaning it stores user subscription information; it also performs tasks such as authentication, credential handling, and user identity management [41]. It is the functional unit that takes over the main tasks of the 4G HSS for 5G. It holds and manages the subscriber’s data. Figure 5. 5G core network [1] 3.5.4 Deployment Architectures: Standalone (SA) and Non-Standalone (NSA) Architecture In a typical 5G NSA mode, the next generation node B (gNodeB) connects directly to the EPC via the user plane, without the need to use the next generation core (NGC), i.e., the 5G core network. Thus, the gNodeB serves as a secondary service function to boost throughput and capacity, while the eNodeB connects to the EPC to execute control plane functions. Operators can choose to leverage their existing 4G deployments by combining LTE and 5G NR to offer 5G cellular services [43]. i. SA Architecture: The 5G SA model provides both user plane and control plane functions through its connection established between the NGC and the gNodeB, and it is the ultimate long-term goal for 5G deployment across operators. It will also play a key role in fully realizing true 5G capabilities. Consequently, NSA deployment provides a fast start for operators, while SA will deliver the full 5G experience. At the same time, the SA architecture is mandatory to support all of 5G’s advanced capabilities, such as URLLC, massive connectivity (mMTC), and especially network slicing. Therefore, it is a mandatory goal that must be reached to unleash 5G’s true potential. 3.6. 6G’s Core Use Cases and Objectives 3.6.1. Performance Targets and KPI Compared to 5G, 6G networks are expected to provide higher spectral, energy, and cost efficiency, higher data rates, 10 times lower latency, 100 times higher connection density, more intelligence for full automation, subcentimeter geographical location accuracy, near 100% coverage, and time synchronization with sub-millisecond latency [44]. Furthermore, Jiang et al. specified these criteria numerically, projecting the peak data rate to reach up to 1 Tbps (ten times that of 5G), the user experienced data rate to reach 1 Gbps or more (ten times that of 5G), and latency to be reduced to 100 or even 10 µs [45]. They also projected the connection density to be improved by a factor of 10, reaching 107 devices per km2. i. Core Use Cases: In November 2023, the ITU approved the IMT-2030 objectives. References [2] and [48] addressed three new use cases for 6G with these objectives: Ubiquitous Mobile Broadband (uMBB), Ultra-reliable
13 low-latency broadband communications (ULBC), and massive ultra-reliable low-latency communications (mULC). 3.6.2. New Scenarios Specific to 6G i. Integrated Sensing and Communication (ISAC): This is a revolutionary approach made possible by the presence of sensing capabilities. This can include pilot signal training, beam alignment and tracking, interference coordination, proactive resource management, and others [46]. The networks we currently use only “talk,” meaning they carry data via signals; however, in 6G, much like a radar system, the radio waves will return to the base station after interacting with surrounding objects, and the network will be able to create a physical map of its environment through these returning signals. In other words, the network will become capable of seeing and sensing, in addition to just talking. ii. Artificial Intelligence (AI) and Communication: The network, thanks to AI, is not only able to transmit data but also to recognize and understand it, allowing it to behave appropriately for changing conditions and send specialized signals to each receiver and each channel: • Semantic Communication: Instead of a security camera transmitting the video of an “empty hallway for hours” as is, the network understands the meaning of the video and transmits only the information that “a person passed through the corridor,” thereby achieving a bandwidth saving of over 99% [47]. • AI-Driven Air Interface: When the network sends data to a device, it learns how noisy the channel is. Instead of transmitting a standard signal, it sends the most efficient signal, “designed by itself,” specifically for the current conditions. It essentially learns to speak to each device in its own language [48]. This means the network now operates not through static rules, but by learning and adapting. iii. Key Technologies That Will Enable 6G: Undoubtedly, the revolutionary capabilities and advancements promised by 6G will not be possible with a single technology, but with the presence of multiple technologies working in an integrated manner. iv. New Spectrum Technologies (THz and Visible Light Communication (VLC)): The THz band is regarded as one of the most critical technologies required to meet the exponential growth in data volume. The necessity for the THz band stems from four key advantages: its contiguous bandwidth of up to hundreds of Gigahertz (HHz), symbol duration at the picosecond level, the ability to integrate thousands of sub-millimeter length antennas, and the ease of co-existence with other regulated and standardized spectrums. Complementing this, VLC operates in the vast 400–800 THz range. This spectrum is entirely unlicensed and vacant, making it thousands of times larger than the radio spectrum and offering immense potential for reaching Tbps speeds [49]. v. Network Architecture (Space-Air-Ground Integrated Network (SAGSIN) and Zero-Touch Network and Service Management (ZSM)): We previously mentioned the SAGSIN, which integrates satellite-air-ground-sea networks. When the question of how such a large and complex network will be managed comes up, ZSM emerges [50]. The main goal of the ZSM concept is to create a network system that can configure, monitor, heal, and optimize itself without human intervention. vi. Evaluation of the 6G Vision and Future Impacts: Hexa-X introduced key value indicator (KVI) metrics, which aim to make new key metrics like sustainability, inclusiveness, and trustworthiness the main theme instead of traditional KPI metrics, and also seek to measure success based on the question, “how beneficial is it to society?” [8]. Considering these metric changes, we can draw some conclusions about the 6G vision. Up to 6G, technologyfocused goals were set, and success was measured by achieving these engineering targets; the benefit of these achievements to society and the environment was considered an indirect benefit. However, upon reaching 6G, the metrics are now directly related to the environment and society; thus, the approach can be said to be humanand value-centric. 3.7. Comparative Technical Analysis Across Generations With 2G, it was realized how critical low bands (1 GHz) are in meeting the requirements for coverage and penetration, and this trend continued throughout 3G and 4G. However, the increasing data demand from 4G onwards showed that mid-bands (1–6 GHz) are ideally positioned in terms of the capacity–coverage balance, and these bands became central to mobile broadband services.
14 With 5G, a multi-layer spectrum approach was adopted for the first time; low bands began to be used for coverage, mid-bands for capacity, and high bands (mmWave) for ultra-high data speeds. This multi-layer architecture created an infrastructural transition to the future 6G networks as well. In 6G studies, the spectrum map is expanding further. While the existing low and mid-bands are preserved with increased efficiency, the sub-THz (92–300 GHz) range is predicted to be mandatory for connections capable of reaching 1 Tbps levels. This indicates that the spectrum will no longer be merely segmented, but the unified operation of multiple frequency layers will become a core capability. Table 2. Evolution of spectrum grouping from 1G to 6G [1] Generation Spectrum Range Basic Spectrum Grouping 1G 800–900 MHz Low Band 2G (GSM/GPRS/EDGE) 800–900 MHz Low Band 3G 2 GHz Low-Mid Band 4G 1–6 GHz Düşük-Orta Bant 5G (NR/IMT-2020) Below 2 GHz, 3–6 GHz, above 24 GHz Multi-Layered Approach 6G Sub-1 GHz, 1–24 GHz, and 24–300 GHz Multilayer Approach with Sub-THz The data in the graph concretely reveals the mobile networks’ transition from human-centric communication to machine-centric critical communication. While 4G LTE networks were optimized for the 30-50 ms band, which is sufficient for human perception, the 5G and 6G vision targets sub 1 ms latency required for the Tactile nternet and collaborative robots (cobots). The 1 ms end-to-end (E2E) target specified in the table validates the necessity of a new cyber-physical system (CPS) infrastructure where physical transmission limits are pushed, and processing and transmission times are minimized. Figure 6. Evolution of latency across mobile network generations When examining the evolution of mobile communication technologies, the increase in spectral efficiency is observed to follow an exponential trend alongside technological leaps, rather than a linear one. Notably, the efficiency leap initiated by the integration of OFDM and MIMO technologies during the transition from 3G to 4G is being maximized in 5G and the targeted 6G standards through Massive MIMO and Terahertz (THz) communication techniques. As seen in the graph, the value of 100 bit/s/Hz targeted under the IMT-2030 (6G) vision represents an approximately 200-fold increase in capacity compared to early digital networks (2G). This confirms that future networks will evolve into a “hyper-connected” structure, capable of accommodating significantly more devices within the same bandwidth, not just providing higher speeds.
15 Figure 7. Spectral efficiency in mobile network generations [5, 25, 26] When examining the average monthly data consumption per subscriber in Türkiye in light of BTK data, it is evident that demand is increasing at an exponential rate. The average consumption, which was only 1.4 GB in 2015 before the 4.5G era, began an upward trend with the widespread adoption of 4.5G in 2016 and reached 16.8 GB as of 2024. Figure 8. Monthly average data consumption per subscriber in Türkiye [27–29] The 12-fold increase in data consumption observed during this approximately 9-year period is the most concrete indicator of why existing spectr cum resources must be used more efficiently. This situation proves that newgeneration technologies (5G and 6G) that will enhance spectral efficiency are not a luxury but a necessity for network sustainability. 4. CONCLUSION The evolution of mobile communication technologies from 1G to the emerging vision of 6G reflects a steady transition from analog, hardware-driven systems to software-defined, fully IP-based and increasingly intelligent architectures. While the early generations focused mainly on voice transmission and basic mobility, the shift to packet-based data with 3G matured significantly in 4G/LTE, where a simplified, flat architecture enabled high speeds and low latency as standard features. With 5G, mobile networks have moved beyond being merely faster systems and have become flexible platforms capable of supporting diverse service types through features such as network slicing and virtualization.
2nd Internation al Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 22 Multi-Band and Inexpensive Linear and Circular Polarization Converter Using a Single-Layer Reflective Metasurface Yunus Kaya1, Ugur Cem Hasar2 1Department of Electronics and Automation, Bayburt University, 69010 Bayburt, Türkiye 2Department of Electrical and Electronics Engineering, Gaziantep University, 27310 Gaziantep, Türkiye Corresponding author: Yunus Kaya (e-mail: [email protected]) Abstract This paper presents a multi-functional reflective metasurface (MS)-based polarization converter (PC) design. The design is suggested as a single-layer structure consisting of a 1.6 mm thick, inexpensive, and readily available FR-4 substrate in the middle, a copper MS design on top, and a copper ground plane on the bottom. The MS-based PC was designed in CST Microwave Studio, and simulations were performed in microwave S-, C-, X-, Ku-, and K-bands (2–4 GHz, 4–8 GHz, 8–12 GHz, 12–18 GHz, and 18–27 GHz, respectively, for a total frequency range of 2–27 GHz). Based on the calculations and analyses performed according to the simulation results, the suggested PC exhibits linear polarization (LP) conversion with a polarization conversion ratio (PCR) value of over 90% at 4.62 GHz in the microwave C-band, in the frequency ranges of 9.63–10.11 GHz and 11.04–11.58 GHz in the microwave X-band, in the frequency range of 14.52–15.36 GHz in the microwave Kuband, and in the frequency ranges of 19.26–19.74 GHz and 23.58–24.18 GHz in the microwave K-band, adn also demonstrates circular polarization (CP) conversion in the frequency range of 20.87–22.21 GHz in the microwave K-band and in the frequency ranges of 12.25–13.63 GHz and 26.06–26.08 GHz in the microwave Ku-band. The suggested MS-based PC is suitable for use in technologies such as stealth, sensing, sensor, communication with its operation in microwave C-, X-, Ku-, and K-bands, multi-function (both LP and CP conversion feature), LP conversion in five frequency bands (and also at one frequency point), and CP conversion in three frequency bands with high efficiency. Keywords: Metasurface, Polarization converter, Linear, Circular, Multi-band 1. INTRODUCTION Polarization, which plays an important role in electromagnetic (EM) waves, indicates the direction of the oscillating electric field in a plane perpendicular to the wave’s propagation [1]. Manipulating the polarization of an EM wave plays an important role in many applications, such as EM stealth, detection, antenna gain enhancement, and radar cross-section reduction [2, 3]. Traditional materials formed naturally exhibit polarization modulation due to their structure, but their use, especially in practical and miniaturized applications, is limited by high power losses, limited bandwidths, and large volumes [4]. Metasurfaces (MSs), which are two-dimensional planar versions of metamaterials designed artificially and attracting the attention of many researchers today due to their extraordinary properties, are widely used in the design of compactly designed polarization converters (PCs) [4, 5]. Researchers have developed PCs that operate at various frequency ranges, such as microwave [2], terahertz [6], visible [7], and infrared [8] frequencies. In recent years, multi-functional PCs have been suggested rather than PCs that exhibit only linear polarization (LP) or only circular polarization (CP) conversion [9]. Additionally, many researchers have developed MSs that can change the polarization of the incident EM wave in transmission or reflection type. Transmissive type PCs are manufactured with a multi-level structure compared to reflective types [10]. Therefore, their production processes are more difficult compared to reflective types. Because of this,, reflection type PCs are generally preferred because higher efficiency can be achieved thanks to the full reflection of incident EM waves [11]. Reflective PCs are manufactured on metal-supported dielectric layers that have a significant effect at the resonance frequency and also restrict the transmission of EM waves [12]. In this paper, single-layer, multi-band PC using a reflective type new MS is suggested that exhibits both LP (in five frequency bands (and also at one frequency point)) and CP (in three frequency bands) characteristics. The suggested PC was designed in CST Microwave Studio (MWS), simulations were performed in the 2–27 GHz frequency range, and the PC’s performance was evaluated based on the obtained data.
23 2. UNIT CELL DESIGN Figure 1 shows the suggested single-layer reflective MS unit cell for multi-band linear and circular polarization in three-dimensional (3D). The unit cell design consists of an inexpensive and readily available FR-4 substrate in the middle, with a thickness of ts = 1.6 mm, a relative dielectric constant εr = 4.3, a loss tangent tanδ = 0.025, a square shape, and a side length of L1 = 10 mm. The top MS pattern and bottom metallic ground plane (which completely reflects EM waves) are selected from copper with a conductivity of σ = 5.8×107 S/m and a thickness of tc = 35 µm. For the MS pattern, a square patch with a side length of L2 = 5.5 mm was inserted into a circular ring with an outer radius of r1 = 4.9 mm and an inner radius of r2 = 4.4 mm, and a square cavity with a side length of L3 = 1.5 mm was remove from the center of this patch. Subsequently, a patch with a width of g = 0.5 mm, passing through the center of the copper structure and rotated 45° clockwise with respect to the y-axis, was remove from this copper structure created at the front, and finally the suggested MS was formed. All these dimensional parameters in the design were determined by optimization. Figure 1. 3D representation of a suggested single-layer reflective MS unit cell for multi-band linear and circular PCs 3. SIMULATIONS AND PERFORMANCE ANALYSIS Simulations for the suggested multi-band PC design and numerical analysis were performed using CST MWS, an EM field simulation software. In CST MWS, boundary conditions are defined as unit cells in the xand yplanes, while a Floquet port is used for excitation in the z-direction. For a y-polarized (transverse magnetic (TM) mod) incident wave, the coand cross-polarized reflection coefficients (ryy and rxy, respectively) are defined as follows [2, 6, 9, 12]. ry rx yy xy iy iy EE r and r EE = = (1) Here, E represents the electric field, while ‘r’ and ‘i’ represent the reflected and incident electromagnetic waves, respectively. Polarization conversion ratio (PCR) is used to evaluate the LP conversion performance of a PC. For an incident wave in TM mode, the PCR value is calculated using the following formula [2, 6, 9–12]. 2 xy 22 xy yy r PCR rr =+ (2) Generally, an amplitude greater than 80% ( ) yy r 0.8< or a magnitude less than −10 dB (ryy < −10 dB) is considered effective for the co-polarized reflection coefficient. In addition, an efficient performance requirement for the crosspolarized reflection coefficient is 0 dB < rxy < −3 dB [12]. The CP conversion capacity of a PC is calculated as follows using the normalized ellipticity (e) value [9−11]. ( ) xy yy 22 xy yy 2 r r sin err φ ∆ =+ (3)
24 Here, ∆ϕ = ϕxy − ϕyy, where ϕyy and ϕxy are the phases of the coand cross-polarized reflection coefficients, respectively. If e = +1, it means xy yy rr = and ∆ϕ = +90° and the TM mode incident wave exhibits right-handed CP (RHCP) characteristics [6, 9−11]. If e = −1, then xy yy rr= and ∆ϕ = −90°, and the TM mode incident wave exhibits left-handed CP (LHCP) characteristics [6, 9−11]. The effect of both the amplitude and the phase difference of the reflected waves is calculated using the axial ratio (AR) formula given below [9−11]. ( ) ( ) 22 44 22 xy yy xy yy xy yy 22 44 22 xy yy xy yy xy yy rr rr2rrcos2 AR rr rr2rrcos2 φ φ ++ ++ ∆ = +− ++ ∆ (4) The AR value is used to calculate the efficiency of CP. In the CP condition, i.e., when ∆ϕ = ±90° and xy yy rr= , the AR value is equal to 1. Figure 2. (a) Reflection coefficients in dB, (b) PCR values, (c) e values, (d) phases and phase differences of reflection coefficients and (e) AR values of the suggested PC in the frequency range of 2–27 GHz
25 To evaluate the performance of the suggested PC, the reflection coefficients, PCR values, e values, phases and phase differences of the reflection coefficients, and AR values obtained based on simulations performed in CST MWS in microwave S-, C-, X-, Kuand K-bands (2–4 GHz, 4–8 GHz, 8–12 GHz, 12–18 GHz, and 18–27 GHz, respectively, for a total frequency range of 2–27 GHz) for the TM mode wave under normal incidence (0°) are plotted against frequency in Figure 2(a)–2(e), respectively. When Figure 2(a)–2(e) is examined, the following conclusions can be reached. Firstly, from Figure 2(a), it can be seen that ryy is less than –10 dB at 4.62 GHz (C-band), 9.87 GHz (X-band), 11.34 GHz (X-band), 14.97 GHz (Ku-band), 19.47 GHz (K-band), 24 GHz (K-band), and 26.13 GHz (K-band), and rxy is greater than –3 dB at all resonance frequencies except the 26.13 GHz resonance frequency. Secondly, from Figure 2(b), it is determined that the PCR values are over 90% in the frequency ranges of 4.62 GHz (C-band), 9.63−10.11 GHz and 11.04−11.58 GHz (X-band), 14.52−15.36 GHz (Ku-band), 19.26−19.74 GHz, and 23.58−24.18 GHz (K-band). In other words, a y-polarized incident wave at these frequencies is reflected as an x-polarized wave over 90% and LP conversion is achieved. Thirdly, in the frequency ranges of 12.25−13.63 GHz and 26.06−26.08 GHz (Ku-band), it can be seen from Figure 2(c) that e = +1, from Figure 2(d) that ∆ϕ is around +90°, and from Figure 2(e) that AR < 1 dB. Therefore, the suggested PC shows RHCP feature in the frequency ranges of 12.25–13.63 GHz and 26.06–26.08 GHz. Fourthly, in the frequency range of 20.87−22.21 GHz (K-band), it is seen from Figure 2(c) that e = −1, from Figure 2(d) that ∆ϕ is around −90°, and from Figure 2(e) that AR < 1 dB. Therefore, the suggested PC shows LHCP feature in the frequency range of 20.87−22.21 GHz. 4. CONCLUSION In summary, in this study, a single-layer, multi-band, and multi-functional (both LP and CP conversion) reflective MS-based PC operating in microwave C-, X-, Ku-, and K-bands was designed. The suggested PC, designed on an inexpensive and easily accessible 1.6 mm thick FR-4 substrate, was simulated in the CST MWS and its performance was evaluated. The suggested PC design achieves LP conversion in five frequency bands (and also at one frequency point) with high PCR values exceeding 90%, and CP conversion in three frequency bands (two exhibiting RHCP characteristics while one exhibiting LHCP characteristics) with high efficiency in terms of e and AR values. The PC suggested in this study shows promising prospects in applications such as radar systems, sensing technologies, and wireless communications. References [1] M. Beruete, M. Navarro-Cia, M. Sorolla, and I. Campillo, “Polarization selection with stacked hole array metamaterial,” J. Appl. Phys., vol. 103, no. 5, art. no. 053102, Mar. 2008. [2] X. Wang, X. Y. Tong, J. L. Wang, A. Saer, J. Wang, and X. Y. Han, “A polarization conversion metasurface for reducing radar cross section and enhancing radiation performance of circularly polarized array antennas,” Opt. Commun., vol. 556, art. no. 130269, Apr. 2024. [3] J. K. Huang, J. P. Yin, Z. M. Xu, and Y. Z. Li, “Polarization scattering regions: A useful tool for polarization characteristic description,” Remote Sens., vol. 17, no. 2, art. no. 306, Jan. 2025. [4] J. B. Sun and J. Zhou, “Metamaterials: The art in materials science,” Eng., vol. 44, pp. 145–161, Jan. 2025. [5] A. Adnan, A. Mitra, and B. Aissa, “Metamaterials and metasurfaces: A review from the perspectives of materials, mechanisms and advanced metadevices,” Nanomater., vol. 12, no. 6, art. no. 1027, Mar. 2022. [6] R. M. H. Bilal, M. A. Baqir, P. K. Choudhury, M. M. Ali, and A. A. Rahim, “On the specially designed fractal metasurface-based dual-polarization converter in the THz regime,” Results Phys., vol. 19, art. no. 103358, Dec. 2020. [7] Q. Ma, W. Gao, Q. Xiao, L. S. Ding, T. Y. Gao, Y. J. Zhou, et al., “Directly wireless communication of human minds via non-invasive brain-computer-metasurface platform,” Elıght, vol. 2, no. 1, art. no. 11, Jun. 2022. [8] D. D. Wen, F. Y. Yue, G. X. Li, G. X. Zheng, K. L. Chan, S. M. Chen, et al., “Helicity multiplexed broadband metasurface holograms,” Nat. Commun., vol. 6, art. no. 8241, Sep. 2015. [9] J. Zafar, H. Z. Khan, A. Jabbar, J. R. Kazim, M. Ur Rehman, A. M. Siddiqui, et al., “Multi-band reflective metasurface for efficient linear and circular polarization conversion,” Opt. Quantum Electron., vol. 57, art. no. 149, 2025. [10] M. I. Khan, Y. X. Chen, B. Hu, N.Ullah, S. H. R. Bukhari, and S. Iqbal, “Multiband linear and circular polarization rotating metasurface based on multiple plasmonic resonances for C, X and K band applications,” Sci. Rep., vol. 10, no. 1, art. no. 17981, Oct. 2020.
26 [11] R. Dutta, J. Ghosh, Z. B. Yang, and X. Q. Zhang, “Multi-band multi-functional metasurface-based reflective polarization converter for linear and circular polarizations,” IEEE Access, vol. 9, pp. 152738– 152748, 2021. [12] S. Hafeez, J. G. Yu, F. A. Umrani, W. Yun, and M. Ishfaq, “A multiband and multifunctional metasurface for linear and circular polarization conversion in reflection modes,” Cryst., vol. 14, no. 3, art. no. 266, Mar. 2024.
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 27 Design, Prototype Production, and Testing of a 12(16) MVA 115/6.3 kV Power Transformer with Twin-Wire Transposed High-Voltage Winding for Loss Reduction Zuhal Isikli1 1R&D Department, Beta Enerji ve Teknoloji A.S., Adana, Türkiye Corresponding author: Zuhal Isikli (e-mail: [email protected]) Abstract Power transformers are essential devices that enable the safe and efficient transmission of electrical energy. Their design critically affects efficiency, durability, and thermal performance. In high-voltage windings, electromagnetic imbalances and localized heating can negatively impact transformer performance. In this context, the use of twinwire transposed windings in 12(16) MVA 115/6.3 kV power transformers represents an innovative approach to balance current distribution and reduce energy losses. Combined with disk-type winding structures, this method ensures uniform thermal distribution, enhancing thermal performance while minimizing electrical and mechanical stresses within the windings. The project aims to develop a transformer design capable of reliable and efficient operation under harsh ambient temperatures (+55 °C), while providing ease of manufacturing and assembly and maintaining cost-effectiveness. By implementing the twin-wire transposition technique, current circulation due to proximity and skin effects is reduced, and electromagnetic imbalances are minimized. This approach improves energy efficiency, maintains stable operation under high thermal stress, and provides a practical alternative to more complex designs. Overall, the design achieves a balance between performance, reliability, and economic feasibility, offering a solution suitable for demanding operating conditions. The expected outcomes include reduced winding losses, improved mechanical strength, better thermal stability, and a simplified production process compared to conventional designs, contributing to enhanced transformer performance and system-level efficiency. Keywords: Twin-wire transposition, Power transformer, Disk winding, Thermal stability
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 28 Design, Manufacturing, and Testing of a 60 Hz Frequency Distrubition Transformer Nadya Atakan1 1R&D Department, Beta Enerji ve Teknoloji A.S., Adana, Türkiye Corresponding author: Nadya Atakan (e-mail: [email protected]) Abstract This study covers the design, manufacturing, and testing processes of a 1600 kVA distribution transformer with voltage levels of 0.48/0.69 kV, developed based on design criteria specific to a 60 Hz grid frequency. The higher efficiency of transformers operating at 60 Hz serves as a key motivation for this work. The higher operating frequency allows for a reduction in core size, which in turn decreases both hysteresis and eddy current losses. Therefore, frequency-based optimization strategies have been applied in determining the core material, lamination thickness, and magnetic circuit cross-section. In general, the effects of 60 Hz frequency on core and winding design are analyzed in detail, with design criteria such as minimizing loss components, improving thermal performance, and maintaining mechanical strength integrated into the design process. The project aims not only to enhance energy efficiency but also to develop optimal magnetic circuit and winding configurations that ensure the transformer’s long-term and reliable operation. As a result, this study aims to present a 60 Hz-compatible distribution transformer design that is low-loss, thermally efficient, and mechanically balanced. Keywords: Distribution transformer, 60 Hz frequency, Energy efficiency
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 29 Forecasting Daily Mean Temperature in Izmir Using Machine Learning Algorithms Pelin Coskun1, Onur Ugurlu2, Orhan Er2 1Department of Smart Systems Engineering, Izmir Bakircay University, Izmir, Türkiye 2Department of Computer Engineering, Izmir Bakircay University, Izmir, Türkiye Corresponding author: Pelin Coskun (e-mail: pelin97[email protected]) Abstract This study focuses on forecasting daily mean temperatures for Izmir, Turkey, using machine learning (ML) algorithms over the period 2014–2023. Three regression-based models—support vector regression (SVR), random forest (RF), and extreme gradient boosting (XGBoost)—were trained and evaluated using daily temperature data obtained from the Turkish State Meteorological Service. Model performances were compared using the mean absolute error (MAE) and coefficient of determination (R2) metrics. According to the results, the RF model achieved the highest predictive accuracy with an R2 of 0.89 and an MAE of 1.37 °C, outperforming both SVR and XGBoost. These findings indicate that ensemble-based approaches, particularly RF, can effectively capture the nonlinear temperature dynamics of İzmir’s complex coastal climate and provide reliable short-term forecasts. Keywords: Temperature forecasting, Machine learning (ML), Random forest (RF), Support vector regression (SVR), Extreme gradient boosting (XGBoost) 1. INTRODUCTION The impacts of global warming and climate change have become increasingly apparent in recent decades, creating significant challenges for environmental balance, economic development, and public health. One of the most prominent indicators of these changes is the persistent increase in surface air temperatures, which affects multiple sectors including agriculture, energy, transportation, and healthcare. Accurate temperature prediction is therefore essential for effective resource management, strategic planning, and early mitigation of extreme weather events. Temperature variability is a fundamental element of the climate system with wide-ranging consequences. Monitoring how mean temperatures change over time allows both researchers and policymakers to identify shortterm variations as well as long-term climatic tendencies. In countries such as Turkey—characterized by complex topography and diverse climatic zones—temperature modeling requires not only traditional statistical tools but also advanced data-driven approaches capable of capturing nonlinear relationships. In recent years, machine learning (ML) techniques have increasingly been utilized to model and forecast temperature dynamics using extensive meteorological datasets. Several studies in Turkey have explored ML-based modeling of meteorological parameters. Yilmaz et al. estimated long-term monthly average temperatures (1981– 2020) through spatial interpolation techniques such as inverse distance weighting, kriging, and radial basis function, using observations from 81 meteorological stations [1]. Their results indicated that interpolation-based models can effectively represent spatial temperature distributions, wit. Similarly, Coskun examined potential water shortages in Bursa under different climate and population growth scenarios using ML methods [2]. The findings highlighted the value of ML models in evaluating climate-induced water scarcity. Bilgic and Elbir employed Sentinel-5P satellite data together with a random forest (RF) model to estimate hourly NO2 concentrations over Izmir [3]. By integrating meteorological and remote sensing variables, they achieved improved prediction accuracy, demonstrating the capability of ML-based models in environmental applications. In the present study, daily mean temperature data from the Turkish State Meteorological Service covering the period 2014–2023 were used to evaluate the predictive performance of different ML algorithms. Support vector regression (SVR), RF, and extreme gradient boosting (XGBoost) models were implemented to forecast daily temperature averages. Model accuracy was assessed through common statistical metrics, including mean absolute error (MAE) and the coefficient of determination (R2). The findings showed that the RF model achieved the best predictive results with an R2 of 0.894, outperforming both SVR (R2 = 0.835) and XGBoost (R2 = 0.857). This outcome suggests that ensemble-based learning algorithms, particularly RF, can effectively capture nonlinear
30 patterns in temperature data and provide reliable forecasts for regions with diverse climatic characteristics such as Turkey. 2. MATERIAL AND METHOD ML, a core field of artificial intelligence, allows computer systems to learn from data and make predictions without explicit rule-based programming. By uncovering hidden relationships and patterns in large datasets, ML algorithms can model complex systems and forecast future trends. These techniques are widely used across domains such as meteorology, healthcare, finance, and industrial automation. The objective of this study is to forecast daily mean temperatures using regression-based ML algorithms. Three models were implemented: SVR, RF, and XGBoost. These algorithms were selected for their capability to capture nonlinear relationships and handle noisy meteorological data. All models were trained and evaluated on the same dataset to ensure fair comparison, and their performances were measured using MAE and the R2. SVR is a kernel-based supervised learning algorithm that aims to identify a regression function capable of approximating the target variable within a specified tolerance margin [4]. Instead of minimizing the total error, SVR focuses on fitting as many data points as possible within this margin, thereby preventing overfitting and enhancing generalization. The algorithm relies on the concept of support vectors—data points that lie closest to the regression boundary—which play a decisive role in defining the predictive function. By employing nonlinear kernel transformations such as radial basis function, polynomial, or sigmoid kernels, SVR can project input variables into higher-dimensional feature spaces where complex and nonlinear relationships become more easily separable. This property makes SVR particularly effective for meteorological forecasting tasks, where temperature patterns often exhibit nonlinear dependencies influenced by atmospheric dynamics, topographic variations, and seasonal transitions. RF, developed by Breiman, is an ensemble-based learning technique that combines multiple decision trees to generate a more accurate and stable prediction model [5]. Each tree in the ensemble is trained using a random subset of the training data and a random selection of input features—a process known as bootstrap aggregation or bagging. This randomization ensures model diversity, minimizes variance, and reduces the likelihood of overfitting. The final prediction in regression tasks is obtained by averaging the outputs of all trees, which significantly enhances robustness against noisy observations and outliers. RF can effectively capture nonlinear patterns and interactions among predictors without the need for explicit feature scaling or transformation. Its simplicity, interpretability through feature importance analysis, and computational efficiency have made it one of the most widely used algorithms in environmental and climate-related modeling. XGBoost, short for extreme gradient boosting, is an advanced implementation of the gradient boosting framework designed for both efficiency and scalability [6]. Unlike RF, which builds trees independently in parallel, XGBoost constructs them sequentially, where each subsequent tree attempts to correct the errors of its predecessors. This additive learning process minimizes a differentiable loss function through gradient descent optimization, progressively refining prediction accuracy. XGBoost incorporates several improvements over traditional gradient boosting, such as regularization terms (L1 and L2) to prevent overfitting, tree pruning strategies for optimal complexity, and efficient handling of missing data. In meteorological applications, XGBoost has proven effective in identifying subtle temporal patterns and complex variable interactions, especially when atmospheric data exhibit heterogeneity or noise. 3. RESULTS The predictive performance of the three ML models was assessed using two key metrics: The R2 and the MAE. The comparative results are summarized in Table 1. Table 1. Performance comparison of ML algorithms MAE R2 SVR 2.16 0.84 RF 1.32 0.89 XGB 1.96 0.86 As presented in Table 1, the RF achieved the best overall accuracy, with an R2 value of 0.894 and an MAE of 1.37 °C, surpassing both SVR and XGBoost. The SVR yielded an R2 of 0.835 and an MAE of 2.16 °C, while XGBoost
31 reached an R2 of 0.857 and an MAE of 1.96 °C. The RF model was optimized with 300 estimators, a minimum samples split of 5, and a minimum leaf size of 2, which improved generalization and reduced bias. Similarly, the XGBoost model performed best with 200 estimators, max_depth = 3, and a learning rate of 0.2. The consistent superiority of the RF suggests that bagging-based ensemble methods capture nonlinear dependencies and complex temporal relationships more effectively than kernel-based (SVR) or boosting-based (XGBoost) algorithms. Overall, the combination of high R2 and low MAE values demonstrates that the RF algorithm provides the most reliable and stable temperature forecasts for the 2014–2023 period. 4. CONCLUSION The results reveal that ML algorithms can effectively model and predict daily temperature variations in Izmir, a city that exhibits strong coastal influences, land–sea interactions, and distinct microclimatic conditions. The application of data-driven regression models to a decade of meteorological observations demonstrated that nonlinear approaches outperform traditional linear techniques in representing temporal temperature dynamics. Among the tested algorithms, the RF algorithm achieved the most accurate results, which can be attributed to its ensemble-based architecture that aggregates multiple decision trees to minimize variance and reduce overfitting. This structure allows the model to recognize complex interactions between temporal and climatic variables— patterns that are often missed by single-model or kernel-based methods such as SVR. The strong predictive capability of RF, indicated by its high R2 and low MAE values, highlights its suitability for handling heterogeneous meteorological data with seasonal fluctuations and noise. The findings suggest that ensemble learning algorithms, when properly tuned, can provide robust and generalizable temperature forecasts even in regions with rapidly changing atmospheric conditions. The XGBoost algorithm also performed reasonably well, showing that gradient boosting can serve as a competitive alternative in applications requiring fast computation and high scalability. However, its tendency to overfit under certain parameter configurations emphasizes the importance of careful model calibration. Overall, this study demonstrates that ML-based methods can play a key role in enhancing the accuracy of regional temperature forecasting systems. For Izmir, such predictive capabilities are particularly valuable for urban heat management, energy demand planning, and agricultural scheduling, as well as for supporting municipal climate adaptation strategies. Future research may focus on integrating additional climatic and environmental parameters—such as humidity, wind speed, solar radiation, and atmospheric pressure—to improve model precision and extend the framework toward multivariate climate prediction. Furthermore, exploring deep learning architectures like long short-term memory (LSTM) or hybrid ensemble models could provide new insights into the temporal dependencies of temperature behavior and contribute to more resilient and adaptive forecasting systems for coastal regions. References [1] C. B. Yilmaz, H. Bodu, E. S. Yuce, V. Demir, and M. F. Sevimli, “Estimation of long-term average temperature (°C) values of Turkey using three different interpolation methods,” Geomatik, vol. 8, no. 1, pp. 9–17, 2023. [2] S. Coskun, “2022 Investigation of climatic conditions of Bursa in terms of climate, climate and population scenarios using machine learning,” M.Sc. thesis, Bursa Uludag University, Turkey. [3] E. Bilgic and T. Elbir, “Estimation of hourly NO2 concentrations in Izmir atmosphere using sentinel-5P satellite data and machine learning method,” Karaelmas Journal of Science and Engineering, vol. 15, no. 1, pp. 175–194, 2025. [4] M. Awad and R. Khanna, “Support vector regression,” in Efficient Learning Machines: Theories, Concepts, and Applications for Engineers and System Designers, Berkeley, CA: Apress. 2015, pp. 67–80. [5] L. Breiman, “Random forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001. [6] T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” in 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, Aug. 2016, pp. 785–794.
38 [7] E. Bjørndal, M. Bjørndal, E. K. Bøe, J. Dalton, and M. Guajardo, “Smart home charging of electric vehicles using a digital platform,” Smart Energy, vol. 12, art. no. 100118, 2023. [8] S. S. A. Salam, V. Raj, M. I. Petra, A. K. Azad, S. Mathew, and S. M. Sulthan, “Charge scheduling optimization of electric vehicles: A comprehensive review of essentiality, perspectives, techniques, and security,” IEEE Access, vol. 12, pp. 121010–121034, 2024. [9] A. Boukhchana, A. Flah, A. Alkuhayli, R. Ullah, and C. Z. El-Bayeh, “Optimal planning strategy for charging and discharging an electric vehicle connected to the grid through wireless recharger,” Frontiers in Energy Research, vol. 12, art. no. 1453711, 2024. [10] Y. An, Y. Gao, N. Wu, J. Zhu, H. Li, and J. Yang, “Optimal scheduling of electric vehicle charging operations considering real-time traffic condition and travel distance,” Expert Systems with Applications, vol. 213, art. no. 118941, 2023.
2nd International Conference on Multidis ciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 39 Design, Prototype Production, and Testing of a Low-Impedance High Short-Circuit Current 18 MVA 33/33 kV Power Transformer Bugra Akpinar1 1Power Transformer Design Department, BETA Enerji, Adana, Türkiye Corresponding author: Bugra Akpinar (e-mail: [email protected]) Abstract This study focuses on the comprehensive design of an 18 MVA, 33/33 kV power transformer characterized by low impedance and high short-circuit current capability. The main objective of this work has been to develop a transformer concept that is capable of operating reliably under severe short-circuit conditions while maintaining superior electrical, thermal, and mechanical performance characteristics. In this context, it has been aimed to perform detailed electromagnetic and mechanical analyses and to systematically optimize the design parameters to achieve the desired performance targets. During the design process, it has been targeted to achieve the intended low impedance value without compromising efficiency or thermal performance by optimizing the winding geometry, conductor configuration, and magnetic core dimensions. It has been intended that the magnetic core be designed to minimize stray flux and core losses, while the leakage reactance be carefully controlled through appropriate winding placement and interleaving techniques. To enhance mechanical robustness under high shortcircuit stresses, it has been aimed to strengthen both the axial and radial mechanical withstand capabilities of the windings, and to dimension and arrange the supporting structures, clamping systems, and spacers to resist the electromagnetic forces expected during short-circuit events. From a thermal perspective, it has been planned to direct the oil circulation paths and cooling ducts in a way that ensures uniform heat distribution and minimizes hot-spot temperatures. The insulation system has been specially considered; it has been aimed to define the dielectric clearances between windings and structural components in accordance with IEC standards to ensure high dielectric reliability. In conclusion, this study has been intended to develop a design methodology that establishes an engineering balance between low impedance requirements and high short-circuit withstand capability. The proposed approach has been targeted to provide a technical basis for future transformer designs, demonstrating how optimized electromagnetic, thermal, and mechanical coordination can contribute to the development of compact, reliable, and high-performance power transformers suitable for modern power systems. Keywords: Low-impedance transformer, High short-circuit current, Transformer design methodology, Electromagnetic and thermal optimization, Mechanical strength of windings
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 40 Fire Safety in Electric Vehicles: An Assessment of Battery Risks, Intervention Techniques, and Policy Recommendations Mehmet Akif Er1, Omer Kaya1 1Transportation Department, Engineering and Architecture Faculty, Erzurum Technical University, 25050 Erzurum, Türkiye Corresponding author: Omer Kaya (e-mail: [email protected]) Abstract The rapid global increase in electric vehicle (EV) use introduces new risks to fire safety. The highest fire risk in EVs is observed in lithium-ion batteries, which have a high energy density. This study comprehensively examines the causes of EV fires, risk differences based on battery chemistry, safety criteria in the design of charging units, and fire response processes. Lithium iron phosphate (LFP) batteries are determined to be safer than other lithiumion chemistries, while nickel cobalt aluminum oxid (NCA) and nickel manganese cobalt oxide (NMC) types, despite their high performance advantages, have a higher risk of thermal runaway. Furthermore, design elements such as the positioning of charging units, ventilation systems, and sprinkler infrastructure in indoor parking lots are found to be decisive in fire safety. It was determined that water-based cooling is the most common approach in responding to EV fires, but long-term monitoring is required due to the ongoing chemical reactions. In this context, the study provides a holistic assessment of the prevention, response, and post-emergency measures for EV fires. Keywords: EV fires, Battery safety, Charging infrastructure, Fire response strategies 1. INTRODUCTION The use of electric vehicles (EVs) is becoming increasingly widespread both globally and in Turkey. This increase has brought the risk of fire in EVs to the forefront. Available data indicates that the likelihood of fire in EVs is lower than in internal combustion engine vehicles. For example, a study conducted by the Australian-based EV Fire Safe between 2010 and 2020 found that fires occurred in 0.1% of internal combustion vehicles, compared to only 0.0012% of EVs. This rate suggests that the risk of fire in EVs is approximately 83 times lower [1]. However, while the likelihood of fire is low, when battery-related fires occur in EVs or hybrid vehicles, the severity and difficulty of controlling these fires can reach alarming levels. The area with the highest risk of fire in EVs is the battery area. Batteries used in these vehicles can be broadly categorized into four main categories: lithium-ion batteries, nickel-metal hydride batteries, lead-acid batteries, and solid-state or flow batteries. Among these types, lithium-ion batteries are the most commonly used battery type in EVs. The primary reasons for this are their high energy density, rapid charging capability, and long cycle life [2]. In EV fires, the thermal energy released from the battery, the release of toxic gases, and the risk of high voltage significantly increase the fire's danger. Therefore, the methods used to respond to EV fires differ significantly from those used in internal combustion vehicle fires. Chemical gases emitted from the battery during a fire pose a serious health risk, and the use of protective equipment is crucial during response. Another important issue is that EV fires can originate from charging units. A recent fire in a building in Istanbul while a vehicle was charging demonstrates the serious risks posed by the design and infrastructure deficiencies of charging units [3]. To mitigate this risk in indoor parking garages, charging units must be correctly positioned, designed to facilitate easy response in the event of a fire, and appropriate safety distances must be maintained. Furthermore, it is crucial for indoor charging stations to install ventilation and security systems that comply with international standards such as National Fire Protection Association (NFPA) 69 and NFPA 70. EV-related fires that may occur in indoor parking garages can threaten not only the vehicle itself but also other surrounding vehicles. Therefore, adequate distances must be maintained between vehicles, and parking areas must be designed in accordance with fire safety criteria. Additionally, design principles that facilitate rapid and safe response by firefighters should also be considered. However, the Building Fire Regulation published in the Official Gazette dated 19.12.2007 does not address such fire risks comprehensively, as EAs were not yet widespread at that time, and therefore needs to be updated in this respect.
41 There are significant differences between firefighting responses to EV and internal combustion vehicle fires. When fires occur in EV batteries, the response time is generally longer. There is also a risk of re-ignition after the fire is extinguished. To address this risk, vehicles are typically monitored with thermal cameras and kept under observation for 6–8 hours after the incident. Disabling the high-voltage system during a fire, if possible, is critical for both the safety of firefighters and controlling the fire. Therefore, firefighters responding to EV fires require specialized training. This study presents research findings on preventive measures to be taken before EV fires break out, response methods to be implemented during the fire, and post-fire safety measures. Figure 1. Electric vehicle fire 2. FINDINGS 2.1. Types of Batteries Used in Electric Vehicles User error is one of the most significant factors contributing to fires in EVs. Lithium-ion batteries are the most commonly used battery type in EVs, and fires occurring in these batteries are generally caused by overcharging or overdischarging. In such batteries, a chemical reaction called thermal runaway occurs during a fire, posing a serious health risk due to the release of toxic gases. Another significant factor contributing to EV fires is improper charging devices and stations. The design and location of charging stations should ensure rapid and safe response in the event of a fire. Vehicles can become more dangerous, especially during charging, due to high voltages. Furthermore, modifications to parts without authorized service approval are another user error that increases the risk of fire. While EV manufacturers use a variety of battery models, studies have shown that most companies prefer lithium-ion batteries. Therefore, this study comprehensively covers the most commonly used lithium-ion battery types. 2.1.1. Lithium Iron Phosphate (LFP) Battery LFP batteries were developed by the Massachusetts Institute of Technology (MIT) in 2003. These batteries are a derivative of conventional lithium-ion technology in terms of chemical composition and utilize lithium ferrophosphate (LFP; LiFePO4) instead of lithium cobaltate [4]. Because their energy density is lower than other lithium-ion types, they are not widely preferred in high-performance EV models. However, their long cycle life and ability to be fully charged offer significant advantages [5]. An uncontrolled increase in the internal temperature of a battery increases the risk of a chain reaction. The thermal runaway temperature in LFP-type batteries is approximately 270 °C [5], making them safer than other lithium-ion battery chemistries. Consequently, the risk of fire is significantly lower with LFP batteries. In some application examples, Tesla prefers Contemporary Amperex Technology Limited (CATL)-supplied LFP batteries in some of its models [6]. Similarly, China-based BYD uses LFP chemistry in its Blade Battery system, which it introduced in 2020. This system has successfully passed rigorous safety tests such as the “nail penetration test” and has demonstrated high resistance to fire [7]. 2.1.2. Lithium Nickel Cobalt Aluminum Oxide (NCA) / Nickel Manganese Cobalt Oxide (NMC) Batteries NCA (LiNiCoAlO2) batteries are notable for their use of high-energy-density cathode materials. They are similar to NMC batteries in terms of specific power and long cycle life. However, they have disadvantages in terms of
42 safety and cost. The charge capacity of NCA batteries is higher than other battery types, which increases the risk of overheating and thermal instability, creating safety concerns. A study in China examining the thermal effects of cathode materials found the order of thermal stability as lithium cobalt oxide (LCO) > NCA > nickel cobalt manganese (NCM811) >> LFP. Panasonic, Tesla’s supplier, produces NCA-type batteries specifically for use in Tesla's high-performance models. NMC batteries are one of the highest-performance types of lithium-ion batteries. The cathode structure contains a combination of nickel, manganese, and cobalt elements [7]. The ratios of these elements determine the battery's energy density, lifespan, and safety features. Therefore, various NMC versions with varying ratios are being developed. A study conducted in the United States examined the aerosol amounts emitted during thermal runaway in lithium-ion batteries and found that NMC batteries emitted more aerosols than other battery types [8]. Research indicates that NMC batteries are widely preferred due to their high performance advantages, but they also present higher risks of fire and thermal runaway. 2.2. Charging Units Used in Electric Vehicles Unlike internal combustion engine vehicles, EVs run on electricity instead of fuel. These vehicles operate using the energy stored in their batteries after a certain period of charging. This energy is provided by charging units. Today, charging infrastructure is quite diverse, offering various alternatives, including public charging stations, individual charging points under buildings, and integrated solutions in private parking areas. EV battery charging methods can be broadly categorized into two main categories: wired charging systems and wireless (inductive) charging systems. 2.2.1. Wired Charging Method The wired charging methods used in EVs are basically divided into two main types: Alternating current (AC) and direct current (DC) charging systems. In AC charging, the vehicle converts AC from an external power source (e.g., a wall outlet or wall-mounted station) to DC through its on-board converter (an on-board charger) and charges the battery. This is because EV batteries can only store energy with DC. In contrast, in DC charging systems, the energy conversion occurs directly in the charging unit, and the DC current is delivered directly to the vehicle [9]. These systems, thanks to their high power capacity, offer fast charging and provide quick charging for longer ranges. AC charging systems, typically used in homes, public spaces, and offices, operate with an external cable connection. However, charging times are longer than with DC systems. Figure 2. Wired charging method 2.2.2. Wireless Charging Method With technological advancements, wireless charging systems are now used in many areas of daily life. For example, just as we can charge mobile devices in vehicles without using cables, the idea of wirelessly charging EVs is gaining increasing interest. In this context, studies are underway on various concepts, such as vehicles drawing energy from the road surface while in motion or charging without a cable connection while parked. In Norway, in particular, a large-scale investment plan is underway to allow vehicles to dynamically charge their batteries while driving. For this system to be implemented, the road infrastructure must be designed to accommodate static power transfer. Wireless charging stations (docks) used for smartphones or watches transfer energy to the battery via magnetic induction. Similarly, in parked EVs, energy can be transferred via receivers placed under the vehicle or on certain surfaces using systems based on this principle [10]. While wireless charging methods offer advantages such as saving time and reducing the risk of cable-related fires, they still require development in terms of efficiency, cost, and safety.
43 Figure 3. Wireless charging method 2.2.3. Charging Modes in Electric Vehicles Mode 1: In this mode, charging is performed using current drawn directly from a household AC outlet. It typically has a single-phase current limit of 16 A and a voltage limit of 250 V, and a three-phase voltage limit of 480 V [11]. An effective grounding system is required for this method. However, due to overheating and electrical safety risks, the use of Mode 1 is prohibited in Europe. Currently, this mode is only used on a limited basis for vehicles with small battery capacities—for example, electric bicycles or e-scooters. Due to these safety concerns, a more protective Mode 2 charging system has been developed. Figure 4. Mode1 type charging method Mode 2: In this mode, charging is performed using AC current from household outlets, as in Mode 1. However, Mode 2 differs in that it uses a cable with a special protection circuit during charging. This cable cuts off the circuit in cases of leakage current and overheating, reducing the risk of fire and improving user safety. Mode 2 is considered the most widely used charging method today due to its simple installation requirements and high level of safety. Figure 5. Mode2 type charging method
44 Mode 3: This mode is typically implemented using wall-mounted charging boxes (wallboxes) at public charging stations or in home parking lots. Charging is done directly via a cable integrated into the charging unit, rather than via an external cable. Mode 3 allows for more controlled energy flow, reducing charging time and improving efficiency. Furthermore, thanks to integrated safety systems, the risk of fire and electrical hazards are significantly reduced. These features make Mode 3 considered one of the most reliable charging methods, both in terms of safety and efficiency. Figure 6. Mode3 type charging method Mode 4: This charging mode transfers energy directly using DC instead of AC, enabling rapid charging. Thanks to its high current and voltage ratings, vehicles can typically be fully charged in as little as 20–40 minutes, depending on battery capacity. Mode 4 systems incorporate advanced safety protocols, temperature monitoring, and cooling systems to mitigate the risks associated with high power transfer. These measures protect against potential fires, overheating, or electrical faults, thus increasing both user safety and battery life. Figure 7. Mode4 charging method 2.2.4. Fire Safety Perspective in Parking Lot Design of Charging Units Today, EVs are mostly charged in underground parking garages under residential buildings. However, this necessitates careful design due to the risk of fire. The location and design of charging units are crucial, as heating, electrical faults, or battery-related chemical reactions that may occur during charging in parking garages can increase the risk of fire. When designing charging units, priority should be given to facilitating fire response and protecting surrounding vehicles and people against chemical explosions that may originate from batteries. Therefore, locating charging units near parking garage entrances and exits facilitates rapid access for firefighters [12]. Furthermore, the installation of fire alarm systems and the installation of camera systems near charging units are recommended for both early warning and post-fire surveillance. If the parking garage is not a separate section
45 of the building, structural separations with a fire resistance of at least one hour are required. Furthermore, ventilation systems are also critical. Chemical gases released during EV fires in indoor parking garages can accumulate, creating an explosion risk. Therefore, ventilation systems should be designed and reinforced to prevent such situations. Finally, sprinkler systems play a vital role in first aid in the event of a fire. These systems limit the growth of a fire before firefighters arrive on the scene and, in conjunction with alarm systems, provide early warning, providing a critical advantage in firefighting [13]. 2.2.5. Protective Equipment to be Used When Responding to Electric Vehicle Fires Extinguishing EV fires poses different risks than internal combustion vehicle fires. High-voltage battery systems, in particular, can pose hazards such as electric shock, arcing, and the release of chemical gases. Therefore, it is vital to use appropriate protective equipment during response. Insulated Gloves: Due to the high voltage risk inherent in EV fires, the use of insulated gloves is mandatory. These gloves typically consist of four layers: an outer fabric, a moisture barrier, a heat barrier, and an inner lining. Insulated gloves must be manufactured to withstand at least 1000 V of electrical current. When touching highvoltage components such as batteries or power systems, only gloves approved to European Norm (EN) 60903 and American Society for Testing and Material (ASTM) D120 standards are essential for safe response. Figure 8. Insulated glove Protective Helmet: Helmets used during firefighting must be resistant to heat, impact, and electrical hazards. Protective helmets must protect against factors such as sparks from batteries and high temperatures. The EN 50365 standard defines the characteristics of helmets that provide safety at voltages not exceeding 1000 V AC and 1500 V DC. Therefore, helmets must be checked for compliance with the EN 50365 standard. Figure 9. Protective helmet Heat Protective Clothing: The high heat, flames, and electric arcs generated during a fire can cause serious injuries. Therefore, it is important that protective clothing used is both heat-resistant and does not restrict mobility. The
46 Turkish Standard (TS) EN International Organization for Standardization (ISO) 11612 standard covers the evaluation of protective performance against heat and flame. It is recommended that response equipment comply with this standard. In addition, the use of additional protective equipment such as face shields, gas masks, and insulated, steel-toed shoes provides additional safety against fire gases, hot particles, and chemicals. Figure 10. Protective clothing 2.2.6. Electric Vehicle Fire Response Techniques EV fires are more dangerous and take longer to extinguish than internal combustion vehicle fires. In such fires, it is crucial for bystanders to wait for the arrival of a specialist team before responding. There are serious risks, such as exposure to chemical gases and the fire spreading due to improper intervention. First aid can be provided for EV fires with a fire blanket. However, this method is only effective for small-scale fires, and its use in large-scale battery fires is reported to be of limited benefit. Crews arriving at the scene must first accurately identify the type of vehicle, as it may not always be readily apparent upon initial inspection. Before responding, crews must wear appropriate protective equipment to protect against the risk of chemical gas release and battery-related explosions. After ensuring environmental safety, if the vehicle is still running, the crew should shut it down using the power button or ignition key. If possible, it is recommended to access the battery, which is the auxiliary power source, disconnect the connections, and deactivate the battery system cables [14]. The possibility of electrical current in the vehicle should be considered during all these procedures, and the intervention must be carried out with caution. The most commonly used extinguishing method in EV fires is water extinguishing. However, the amount of water used and the cooling time are much longer than in internal combustion vehicle fires. The responding team must also be cautious about the risk of explosion that may arise from reactions between water and battery chemicals. Following a fire, it is recommended that EVs be monitored for at least 72 hours due to the possibility of recurrence. During this period, the vehicle may be placed in a cooling pool or monitored with thermal cameras to monitor for potential thermal runaways. In one experiment, an EV fire was initially extinguished with a fire blanket. This method slowed the spread of the fire initially but was not effective in stopping thermal runaway within the battery. Compressed air foam was then used, and the fire was brought under control in approximately 90 seconds. However, it was observed that the chemical reactions within the battery remained active for a long time, requiring a long cooling period [15].
47 Figure 11. Fire blanket During an EV fire, or after a small fire has subsided, a device called an emergency plug must be used to control the vehicle’s safety (Figure 12). The emergency plug prevents unforeseen vehicle movement in an emergency and instantly and safely disables the battery system. The device allows the system to perceive the vehicle as charging, thus preventing any movement of the vehicle. Figure 12. Intervention with emergency plug [16] 3. POLICY RECOMMENDATIONS AND IMPLEMENTATION STRATEGIES The proliferation of electric vehicles necessitates a reassessment of existing fire safety regulations and infrastructure standards. Policy recommendations developed in this context should encompass not only firefighting but also preventative engineering, regulatory updates, infrastructure planning, and increased institutional capacity. Policy and practice recommendations for strengthening EV fire safety are presented below: i. Legislation and Standard Updates • Updating the Building Fire Regulation: The 2007 regulation was prepared at a time when EV use was not increasing. The new regulations should include charging infrastructure, ventilation, distance, sprinkler, and material standards specific to EVs. • Battery Safety Regulation: A dedicated “Battery Safety Regulation” should be established for EV batteries, covering topics such as thermal runaway tests, aerosol emission limits, and recycling safety. • Compliance with European Standards: International standards such as NFPA 69/70, EN 50365, and TS EN ISO 11612 should be integrated with national legislation. ii. Charging Infrastructure and Parking Lot Design • Mandatory Safety Distance: A minimum safety distance of 1.5 meters between vehicles should be mandatory in indoor parking lots.
54 Table 2. HSCA Flatness Index experiment results Material Information HSCA Mass of the Test Sample Section - g M0 = 5962 Total Mass of Samples Not Subjected to Treatment - g 35 Screening with Test Sieves Screening with Bar Screens Particle Size Range (mm) Particle Size Mass (g) Sieve Gap Width (mm) Mass Passing Through the Sieve (g) FIi = (mi/Ri)×100 25 / 31.5 0 16 0 - 20 / 25 0 12.5 0 - 16 / 20 72 10 0 - 12.5 / 16 1598 8 60 - 10 / 12.5 1186 6.3 41 3 8 / 10. 219 5 13 6 6.3 / 8 1496 4 85 6 5 / 6.3 615 3.15 16 3 4 / 5. 741 2.5 25 3 M1 = SRi = 5927 M2 = Smi = 240 FI = (M2/M1)×100 = 4.05 Flatness Index Category FI15 Table 3. LSCA flatness index test results Material Information LSCA Mass of the Test Sample Section - g M0 = 5982 Total Mass of Samples Not Subjected to Treatment - g 46 Screening with Test Sieves Screening with Bar Screens Particle Size Range (mm) Particle Size Mass (g) Sieve Gap Width (mm) Mass Passing Through the Sieve (g) FIi = (mi/Ri)×100 25 / 31.5 0 16 0 - 20 / 25 0 12.5 0 - 16 / 20 70 10 0 - 12.5 / 16 1600 8 57 - 10 / 12.5 1241 6.3 32 3 8 / 10. 256 5 15 6 6.3 / 8 1691 4 87 5 5 / 6.3 570 3.15 16 3 4 / 5. 508 2.5 14 3 M1 = SRi = 5936 M2 = Smi = 221 FI = (M2/M1)×100 = 3.72 Flatness Index Category FI15 The density and water absorption test results for LSCA and HSCA are given in Table 4 and Table 5. The test results showed that the density values of recycled aggregates were lower than those of natural aggregates, while their water absorption values were higher. This situation stems from the porous hardened cement paste in the RCA
55 structure. Table 4. LSCA density and water absorption test results LSCA Water Absorption Test Results LSCA Density Test Results Dimension 0/4 4/8. 8/16. Size 0/4. 4/8 8/16. Pycnometer + Water + Material Mass (g) M2 6916 8133 8198 Measuring Cup Mass (g) m1 3188 3188 3188 Pycnometer + Water Mass (g) M3 6311 6316 6329 Measuring Cup Volume (l) v 3 3 3 Material DKY Mass (g) M1 1013 3096 3078 Sample + Measuring Cup Mass (g) m2 7646 7166 7192 Material Oven-Dried Mass (g) M4 934 2968 2978 Loose Bulk Density (kg/dm3) ρb 1.486 1.326 1.335 Apparent Particle Density (kg/dm3) ρa 2.84 2.58 2.69 Percentage of Empty Space (%) n 35 43 46 Oven-Dried Grain Density (kg/dm3) ρrd 2.29 2.32 2.46 DKY Particle Density (kg/dm3) ρssd 2.48 2.42 2.55 Water Absorption (%) WA24 8.5 4.3 3.4 Table 5. HSCA density and water absorption test results HSCA Water Absorption Test Results HSCA Density Test Results Dimension 0/4 4/8. 8/16. Size 0/4. 4/8 8/16. Pycnometer + Water + Material Mass (g) M2 7288 8137 8161 Measuring Cup Mass (g) m1 3188 3188 3188 Pycnometer + Water Mass (g) M3 6682 6314 6289 Measuring Cup Volume (l) v 3 3 3 Material DKY Mass (g) M1 1008 3067 3038 Sample + Measuring Cup Mass (g) m2 7898 7362 7366 Material Oven-Dried Mass (g) M4 944 2950 2973 Loose Bulk Density (kg/dm3) ρb 1.57 1.391 1.393 Apparent Particle Density (kg/dm3) ρa 2.79 2.62 2.7 Porosity (%) n 33 41 45 Oven-Dried Grain Density (kg/dm3) ρrd 2.35 2.37 2.55 DKY Particle Density (kg/dm3) ρssd 2.51 2.47 2.61 Water Absorption (%) WA24 6.8 4 2.2 3.2. Fresh and Hardened Concrete Test Results The fresh concrete properties of C30 target concrete produced with both aggregate types were found to be similar. The slump value of concrete produced with HSCA was measured as 83 mm, while that of concrete produced with LSCA was measured as 85 mm. The slump test setup is shown in Figure 5. These S2 class consistency values showed that, despite LSCA’s higher water absorption rate, it did not pose a problem in terms of workability when
56 adjusted for water according to the saturated dry surface (SDS) condition. Different results were obtained from non-destructive tests performed on hardened concrete specimens. The non-destructive test results are shown in Table 6. Figure 5. Slump test setup Table 6. Schmidt hammer test and ultrasonic transit time test results Sample No. Schmidt Hammer Readings Ultrasonic Readings 1 2 3 4 5 6 7 8 9 10 Average (Rc) Strength Estimate (MPa)* Average Strength Estimate (MPa) Duration (μs) Speed - V (km/h) Strength Estimate (MPa)* Average Strength Estimate (MPa) Y1 26 24 25 26 27 26 28 24 29 24 26 30.81 26.56 33 4.2 32.84 33.03 Y2 25 25 24 28 22 29 25 25 24 26 25 27.47 34 4.4 39.69 Y3 22 27 26 19 29 22 18 23 26 25 24 24.37 35.3 4.2 32.84 Y4 21 22 26 18 26 24 24 26 20 22 23 21.51 34 4 26.92 Y5 20 21 29 22 20 20 30 28 24 27 24 24.37 34.6 4.3 36.14 Y6 28 22 28 22 30 30 26 27 20 22 26 30.81 33.5 4.1 29.77 D1 22 21 18 21 27 29 23 24 26 27 23 21.51 23.06 33.6 4 26.92 27.10 D2 20 25 24 26 25 20 24 26 19 21 23 21.51 34.2 4.2 32.84 D3 27 22 23 26 27 25 24 24 24 23 25 27.47 34 3.8 21.84 D4 19 26 22 27 20 21 22 20 18 22 22 18.89 33.9 3.9 24.28 D5 23 24 27 26 22 25 25 27 24 21 25 27.47 33.8 4.1 29.77 D6 19 24 24 24 21 25 27 18 21 19 23 21.51 33.6 4 26.92 *: Compressive strength estimates were calculated using Rc and V values. In the estimated strength measurements performed with the Schmidt hammer, HSCA concrete (26.56 MPa) demonstrated higher performance than LSCA concrete (23.06 MPa). Similarly, in the estimated strength analysis performed using the ultrasonic transit velocity, HSCA concrete (33.03 MPa) yielded significantly higher results than LSCA concrete (27.10 MPa). The transmission velocity of HSCA concrete, averaging above 4.2 km/s, indicated a denser and more homogeneous internal structure compared to the velocity of LSCA concrete, which averaged 4.0 km/s. The results of the 28-day compressive strength test are given in Table 7 and the test setup is shown in Figure 6. The compressive strength results revealed a different outcome compared to the non-destructive
57 test results. Both concrete series achieved the targeted C30 class. The average compressive strength of concrete produced using HSCA was found to be 34.59 MPa, while that of concrete produced using LSCA was 35.63 MPa. LSCA concrete was found to have approximately 3% higher strength than HSCA concrete. This indicates that nondestructive tests (particularly Schmidt) require calibration for concrete containing RCA and may be misleading. The superior performance of LSCA concrete is most likely due to the mechanical quality of the aggregate and, in particular, the better optimised fine material ratio. The improved particle distribution in LSCA fills the fine material voids lacking in HSCA, creating a denser microstructure and a stronger cement paste-aggregate interface. This effect compensates for the potential negative impact of the aggregate’s original low strength. Table 7. Compressive strength test results Compressive Strength Values (28 days) Sample No. Unit Volume Mass (kg/dm3) Fracture Load (kN) Strength (MPa) Average Strength (MPa) Y1 2.27 828.2 36.81 34.59 Y2 2.26 762.8 33.9 Y3 2.25 773 34.36 Y4 2.27 755.8 33.59 Y5 2.23 744.8 33.1 Y6 2.26 804.5 35.76 D1 2.27 804.2 35.74 35.63 D2 2.25 781.4 34.73 D3 2.28 783 34.8 D4 2.23 798.6 35.49 D5 2.24 812.8 36.12 D6 2.26 830.8 36.92 Y: Concrete produced from HSCA and D: Concrete produced from LSCA Figure 6. Compressive strength graphs of HSCA and LSCA 4. CONCLUSION In this study, the properties of recycled aggregates (LSCA and HSCA) obtained from waste concrete sourced from construction laboratories and separated according to low strength class and high strength class were compared, along with the performance of C30 target concrete produced with these aggregates.
58 • As a result of the sieve analysis, LSCA showed a higher proportion of fine material compared to HSCA. For this reason, the amount of fine aggregate for HSCA was below the Fuller curve. • The abrasion loss at the end of the abrasion test conducted on LSCA and HSCA aggregates was found to be 41% for HSCA and 32% for LSCA. The ASTM C-33 standard states that the LA percentage should not exceed 50% in concrete production. Accordingly, it was concluded that both types of concrete waste can be used in concrete production. • The MB test was conducted to determine the quality of the fine material, and the MB values of the HSCA and LSCA materials were found to be 0.8g/kg. • As a result of the test, the density values of both recycled aggregates were found to be lower than the density values of general natural aggregates, while their water absorption values were found to be higher. • According to the Schmidt Hammer test results, the average estimated compressive strength was found to be 26.56 MPa for concrete produced from HSCA and 23.06 MPa for concrete produced from LSCA. • According to the results of the ultrasonic transit time test, the average estimated compressive strength was found to be 33.03 MPa for concrete produced from HSCA and 27.10 MPa for concrete produced from LSCA. • According to the 28-day compressive strength test results of the cube specimens, the average compressive strength was found to be 34.59 MPa for concrete produced from HSCA and 35.63 MPa for concrete produced from LSCA. When the results of the study were evaluated collectively, it was observed that the Schmidt Hammer Test results for both groups of specimens were lower than the target concrete strength, while the ultrasonic test results were close to the target strength. The compressive strength results, on the other hand, showed that both groups of specimens achieved the target design strength. The concrete strength produced from LSCA was found to be approximately 1% higher than that from HSCA. This is thought to be due to the fact that the fine material ratio in LSCA is slightly higher than in HSCA. In conclusion, it was found that aggregates obtained from concrete waste with both strengths can be used as aggregates in concrete production. Acknowledgments This paper utilizes data from the project entitled “Investigation of the Use of Concrete Wastes/Debris Produced in Different Strength Classes as Recycled Aggregate and the Differences in Mechanical Properties Among Them,” which was supported by the TUBİTAK-2209/A program (2023/2 Term). References [1] A. Koken, M. A. Koroglu, and F. Yonar, “The usability of waste concrete as concrete aggregate,” Selcuk University Technical Sciences Vocational School Technical-Online Journal, vol. 7, no. 1, pp. 86–97, 2008. [2] I. Demir, “The use of construction waste in concrete production and its effect on concrete properties,” Afyon Kocatepe University Journal of Science and Engineering, vol. 9, no. 2, pp. 105–114, 2009. [3] G. Durmus, O. Can, and O. Simsek, “Determination of the engineering properties of different classes of concrete produced from recycled aggregates," in 5th International Advanced Technologies Symposium (IATS’09), Karabük, Turkey, May 13–15, 2009. [4] T. C. Hansen and H. Narud, “Strength of recycled concrete made from crushed concrete coarse aggregate,” in Concrete International: Design and Construction, 1983, pp. 79–83. [5] A. Nealen and M. Ruhl, “Consistency aspects in the production of concrete using aggregates from recycled demolition material,” in Darmstadt Concrete, 1997. [6] H. Polat, U. E. Yurtcan, and M. N. Kolak, “Investigation of the usability of waste kerbstones as concrete aggregate,” Journal of Nature and Science, vol. 3, no. 2, pp. 37–41, 2014. [7] M. M. Tufekci, “Investigation of the reusability of recycled aggregates in concrete production,” M.Sc. thesis, Institute of Science, Yıldız Technical University, 2011. [8] C. Demirel and O. Simsek, “Investigation of the effect of using waste concrete as recycled aggregate in concrete production on compressive strength," in International Academic Research Congress, Antalya, Turkey, Oct. 18–21, 2017, pp. 1026–1031. [9] Beton - Karma suyu - Numune alma, deneyler ve beton endustrisindeki islemlerden geri kazanilan su dahil, suyun, beton karma suyu olarak uygunlugunun tayini kurallari, TS EN 1008, Türk Standartları Enstitüsü, 2003. [10] Tests for geometric properties of aggregates – Part 1: Determination of gradation – Sieve analysis, TS EN 933-1, Turkish Standards Institute, Ankara, 2012. [11] Tests for mechanical and physical properties of aggregates – Part 2: Methods for determining resistance to fragmentation, TS EN 1097-2, Turkish Standards Institute, Ankara, 2020.
59 [12] Tests for geometric properties of aggregates – Part 9: Determination of fine particles – Methylene blue test, TS EN 933-9, Turkish Standards Institute, Ankara, 2022. [13] Tests for geometric properties of aggregates – Part 3: Flatness index, TS EN 933-3, Turkish Standards Institute, Ankara, 2012. [14] Tests for physical and mechanical properties of aggregates – Part 6: Determination of particle density and water absorption, TS EN 1097-6, Turkish Standards Institute, Ankara, 2022. [15] Concrete aggregates, TS 706 EN 12620+A1, Turkish Standards Institute, Ankara, 2009. [16] Concrete – Properties, performance, production and conformity, TS EN 206+A2, Turkish Standards Institute, Ankara, 2021. [17] Concrete – Fresh concrete tests – Part 2: Slump test, TS EN 12350-2, Turkish Standards Institute, Ankara, 2019. [18] Determination of compressive strength of concrete in structures and precast concrete components in situ, TS EN 13791, Turkish Standards Institute, Ankara, 2019. [19] Concrete tests in structures – Part 4: Determination of ultrasonic pulse wave velocity, TS EN 12504-4, Turkish Standards Institute, Ankara, 2021. [20] Concrete – Tests on hardened concrete – Part 3: Determination of compressive strength of test specimens, TS EN 12390-3, Turkish Standards Institute, Ankara, 2019. [21] Standard specification for concrete aggregates, ASTM C33-07, ASTM International, West Conshohocken, PA, 2007.
2nd International Conference on Multidisciplinary Scien ces and Technological Developments (ICMUSTED 2025) December 12-15, 2025 60 Numerical Modelling of Moisture Loss and Oil Uptake in Noodle Strips During Deep-Fat Frying Using Finite Difference and Finite Element Methods Serpil Pekdogan Goztok1, Cihat Guner2, Hakan Basdogan2, Omer Said Toker3 1Department of Food Processing Programme, Siirt University, Siirt, Türkiye 2Eris R&D Center, Tekirdag, Türkiye 3Department of Food Engineering, Yildiz Technical University, Istanbul, Türkiye Corresponding author: Serpil Pekdogan Goztok (e-mail: [email protected]) Abstract This study aimed to experimentally characterize and numerically model moisture loss and oil uptake in noodle strips during deep-fat frying. Noodles of fixed geometry (≈ 10×8×2.5 cm3) were fried in refined palm oil at 135 °C for up to 240 s, and their moisture and fat contents were determined gravimetrically and by Soxhlet extraction, respectively. The initial moisture and fat contents were ≈ 19.0% and 7.8% (wb). Effective moisture diffusivity (Dw) was determined from Fick’s second law for three-dimensional finite slabs, moisture diffusion coefficient Dw = 8.80×10–7 m2/s (R2 = 0.97). The overall mass transfer coefficient was found to be kc = 0.0043–0.0054 s–1. Moisture and oil transport were modeled using coupled diffusion equations solved by two numerical models: a finite difference (FD) method implemented in MATLAB and a finite element (FE) method implemented in COMSOL Multiphysics. Oil diffusivity was linked to moisture content through an empirical relationship to capture simultaneous water loss and oil gain. The mass transfer phenomena occurring during deep-fat frying were accurately captured by both the FD and FE models. Both models reasonably reproduced the experimental trends, with rapid moisture depletion and concurrent oil penetration in the early stages, followed by a deceleration associated with crust formation. For moisture loss, the FD model showed better agreement over the entire process (average error −5.50%), whereas the FE model more accurately predicted the final moisture content (final error −2.13%). For oil uptake, the FE model provided the closest fit to experimental data (final error 0.51%; average error −7.17%). Three-dimensional simulations revealed the formation of a low-moisture depletion front and preferential oil accumulation in the crust region. The proposed modelling framework offers a useful tool for optimizing frying conditions, controlling oil uptake, and improving product quality in industrial noodle processing. Keywords: Deep-fat frying, Noodle, Moisture diffusivity, Oil uptake, Numerical modelling 1. INTRODUCTION Deep-fat frying is a key thermal processing step widely used in the industrial production of many foods, including potato products, meat products, bakery items, and snacks. During frying, simultaneous heat and mass transfer occur between the product and the hot oil phase, leading to a series of complex physical and chemical processes such as evaporation of internal moisture, development of a porous structure, crust formation, and oil penetration into the product matrix [1–3]. While frying contributes desirable quality attributes such as a crispy crust, characteristic aroma, colour, and pleasant texture, it must be carefully controlled because of its association with high fat content, oxidative degradation products, and the formation of potentially harmful compounds such as acrylamide [3, 4]. Instant noodle products have become one of the most rapidly growing carbohydrate-based foods worldwide due to their ease of preparation, low cost, and long shelf life. In typical industrial processing, noodle dough is mixed, sheeted and cut, then steam-precooked, followed by deep-fat frying in palm oil or its fractions to obtain lowmoisture, porous noodle cakes. In commercial applications, palm oil and palm olein/stearin blends are commonly preferred because of their favourable frying stability and oxidative resistance [5–7]. However, the fat content of fried instant noodles often reaches levels of around 20–25%, making the control of oil uptake through formulation and process optimisation an important technological objective [5, 8]. In recent years, fried noodle products have been increasingly examined from nutrition and public health perspectives due to their high fat and sodium contents, acrylamide formation, and oxidative stability issues during
61 storage [3, 4, 9]. Various strategies have been proposed to reduce oil uptake, including the use of alternative frying oils and oleogel systems, pre-drying or pre-cooking treatments, and incorporation of dietary fibre or hydrophilic/hydrophobic additives into the dough formulation [8, 9]. Although these approaches provide valuable practical information for reducing fat content and improving quality parameters, in many studies the underlying water–oil mass transfer mechanism is not described in terms of fundamental transport parameters such as effective diffusion coefficients and mass transfer coefficients, and remains mainly empirical [2, 9]. In frying and drying studies, moisture loss in foods is most commonly described using diffusion models based on Fick’s second law, and the effective moisture diffusivity (Deff) is used as a key parameter characterising internal mass transfer resistance. Reported Deff values for various deep-fat fried foods generally cluster within the range of 10–10–10–8 m2/s, although this range may broaden depending on product structure, temperature, geometry, pretreatments, and the properties of the frying medium [2, 10–13]. Similarly, external mass transfer at the product– oil interface is expressed in terms of convective mass transfer coefficients and empirical rate constants, which play a critical role in process design and optimisation [1, 14]. However, in many studies heat and mass transfer are treated separately, and the actual three-dimensional geometry of the product is simplified, which may lead to less realistic estimates of mass transfer coefficients [14]. Mathematical models developed for deep-fat frying have been successfully applied to describe both heat and mass transfer in various systems such as potato slices, root vegetables, composite products, and bakery items. In these studies, Fick-based differential equations were solved using different mathematical models to predict product centre temperature, moisture content, and oil uptake, and the model results were validated against experimental data [10–12, 15]. Furthermore, studies focusing on the relationship between water loss and oil uptake—especially in potatoes and chicken—have shown that rapid early-stage moisture loss is accompanied by simultaneous oil penetration, whereas at later stages the rate of oil uptake decreases due to crust formation and pore filling [16, 17]. These findings highlight the need to model frying processes not only through empirical final-point measurements, but also by capturing the distributional and time-resolved evolution of diffusion phenomena. A review of the literature on noodle products shows that most studies focus on the effects of formulation changes— such as frying oil type, oleogel use, emulsifiers, and polysaccharide/thickener additions—on oil uptake, moisture content, colour, and texture [5–8]. While these works provide important practical insights into lowering oil absorption and improving quality, they typically report bulk moisture and fat contents and rarely employ threedimensional numerical models that describe the spatial and temporal distribution of water and oil in the product during frying via effective diffusion coefficients and mass transfer coefficients. In particular, comprehensive studies applying detailed 3D diffusion-based models to steamed, pre-cooked, thin and porous noodle strips, and systematically comparing FD and FE approaches on the same geometry, are extremely limited. In this context, the aim of the present study is to investigate, both experimentally and numerically, moisture loss and oil uptake in wheat-based, steam-precooked noodle strips during deep-fat frying in palm oil, and to determine the key mass transfer parameters governing the process. To this end, (i) effective moisture diffusivity and mass transfer coefficients for noodle samples under frying conditions were determined, (ii) a three-dimensional mathematical model was developed to describe water loss and oil uptake based solely on diffusion-driven mass transport, (iii) the model was numerically solved using a finite difference (FD) code developed in MATLAB and a finite element (FE) model implemented in COMSOL Multiphysics, and (iv) the predictions were compared with experimentally measured moisture and oil contents throughout frying to evaluate the performance of the two numerical approaches. In this way, the study seeks to quantitatively describe the characteristic “rapid water loss– limited oil uptake” behaviour of deep-fat fried noodle products within a diffusion-based framework, and to provide a practical tool for the design and optimisation of industrial frying processes. 2. MATERIAL AND METHOD 2.1. Materials and Sample Preparation Wheat-based, steam-precooked noodle strips produced on an industrial instant noodle line were used in this study. After production under standard factory conditions, samples underwent a standardized pre-frying preparation: steam-precooking was completed on the line, the noodles were then briefly equilibrated at ambient conditions to stabilize surface moisture, and no batter/coating was applied prior to frying. The strip dimensions were kept constant throughout the experiments; noodles were cut to width ≈ 10 cm, length ≈ 8 cm, and thickness ≈ 2.5 cm, and this geometry was transferred unchanged to the numerical models. To define the initial state for numerical modeling, measurements performed immediately before frying (t = 0 s) indicated an initial moisture content of ≈ 19.0% (wb) and an initial fat content of ≈ 7.8% (wb). All experiments were conducted with replicate measurements
62 at each temperature–time condition, and the resulting data were subsequently used for calibration and validation of the FD and FE models presented in the following sections. 2.2. Frying Procedure Frying operations were carried out in a thermostat-controlled benchtop fryer using refined palm oil. Palm oil and its fractions (notably palm olein/stearin) are widely adopted in instant-noodle frying and offer thermo-oxidative stability advantages [6, 17–19]. To maintain bath temperature stability, the fryer was pre-equilibrated to the set point, and a high oil-to-product ratio was preserved to minimize temperature sag during processing [20–22]. Experiments were conducted at 135 °C and frying time was set within 0–240 s; during the early stage (0–60 s), sampling frequency was increased to capture crust formation and accelerated mass transfer. Throughout frying, both the oil-bath temperature and the product centre temperature were monitored simultaneously; core temperature was measured with a fine-tip thermocouple inserted into the geometric center of the noodle strip. In deep-fat frying studies, center-probe thermocouple placement is a common practice for core temperature measurement [23–25]. Upon reaching the target time, samples were removed from the bath, briefly drained to remove surface-carried free oil, cooled briefly at ambient conditions, and then subjected to the planned analyses (moisture, fat, etc.). The oilbath temperature was continuously tracked; when necessary, filtration/renewal was performed, as oil quality indices (e.g., total polar matter (TPM), free fatty acid (FFA), peroxide value (PV)) are critical to product safety and quality in deep-fat frying [21, 24]. Overall, the process entails simultaneous heat and mass transfer accompanied by crust formation phenomena [3, 26]. 2.3. Analytical Measurements Moisture content was determined by the loss-on-drying procedure in a forced-air oven at 105 °C to constant mass, following the principles of Association of Official Analytical Collaboration (AOAC) AOAC 925.10 and the American Association of Cereal Chemists (AACC) 44-15.02 air-oven method; results are reported on a wet basis (% wb) [27–29]. Total fat was quantified by Soxhlet extraction using petroleum ether as solvent, in accordance with the AOAC ether-extract family of methods (e.g., AOAC 920.39 and AOAC 945.16); the equivalence of petroleum ether and hexane for lipid extraction in food matrices is documented in the literature [30–32]. All assays were performed with n ≥ 3 replicates, and data are expressed as mean ± standard deviation (SD); routine good laboratory practices were applied for sample preparation, instrument calibration, and weighing to ensure analytical precision and repeatability. 2.4. Mathematical Modelling and Numerical Solution 2.4.1. Determination of Effective Moisture Diffusivity The effective moisture diffusivity (Dw) of noodle strips during frying was determined using Fick’s second law of diffusion (Equation (1)): 2 w2 MM D tx ∂∂ = ∂∂ (1) where M is the moisture content (kg water/kg wet product), x is the diffusion distance (m), t is time (s), and Dw is the effective moisture diffusivity (m2/s). The initial and boundary conditions were defined as Equation (2): i e t0 LxLMM dM t0 x0 0 dx t0 xL MM = −<< = >= = >= = (2) When ambient humidity is constant in the frying process, moisture content is reduced to the fractional moisture ratio (MR), the distance from the center in food is expressed by using dimensionless distance variable (n = x/L), and the time is non-dimensional time. Mass Fourier number (Fom= Dw t/L2) was converted and resolved analytically to achieve Equation (3) for frying of noodle [33].
63 ( ) ( ) 22 w te 12 n1 ie 2 2n 1 πD t MM 1 MR A exp MM A 2n 1 ∞ = −− − = = −− ∑ (3) where MR is the fractional MR, Mi is the initial moisture content (kg water/kg dry matter), Me is the equilibrium moisture content (kg water/kg drymatter), Mt is the moisture content of product at time t (kg water/kg dry matter). L was the half thickness (m) since the frying occurred from both surfaces. For long osmotic dehydration times, only first term of the sum in Equation (4) was used [33]. 2 w 1 2 πD t MR A exp A = − (4) where A1 is ( ) 3 2 8 π , A2 is ( ) 22 2 123 1L L L++ for 3-dimensional finite slab. 2 w 2 πD slope A = − (5) where Lx, Ly, and Lz are half-thicknesses along the x, y, and z directions, respectively. This approach accounts for the 3D geometry of the noodle strips during frying. Oil uptake during frying was coupled with water loss by relating the effective oil diffusivity (Do) to the water diffusivity as follows: m i ow inital X D D1 X = − (6) where Xi is the moisture content at time i, Xinitial is the initial moisture content, and m is an empirical constant determined by fitting the model to experimental oil uptake data. This relationship allowed simultaneous prediction of water loss and oil gain during frying. 2.4.2. Mathematical Model Moisture loss and oil uptake during frying were modeled based solely on diffusion-driven mass transport within the noodle matrix. Porosity changes and shrinkage were neglected, and the noodle samples were assumed to be geometrically fixed, homogeneous, and isotropic. Accordingly, two separate partial differential equations based on Fick’s second law were employed for moisture (water) and oil: 2 ww w2 CC D tx ∂∂ = ∂∂ (7) 2 oo w2 CC D tx ∂∂ = ∂∂ (8) Where, Cw and Co are represent the moisture (water) and oil concentrations of noodle (mol/m3), respectively, while Dw and Do is the effective diffusion coefficients (m2/s). At the beginning of the frying process, the initial moisture and oil distributions were assumed to be uniform: w w,0 C (x, y,z,0) C= (9) o o,0 C (x, y,z,0) C= (10) During frying, moisture evaporation at the surface and oil penetration into the noodle were described by convective mass transfer boundary conditions. The corresponding surface mass fluxes were expressed as follows:
70 appropriate method for predicting the temperature distribution in a potato model,” Potato Research, vol. 65, no. 4, pp. 933–957, 2022. [34] I. D. Demiray, H. Ergezer, E. Demiray, and O. Sufer, “Deep-fat frying of chicken nuggets: Impacts on mass transfer and some quality indices,” Food Science & Nutrition, vol. 13, no. 6, art. no. e70451, 2025. doi: 10.1002/fsn3.70451 [35] A. O. Oladejo, H. Ma, W. Qu, C. Zhou, B. Wu, X. Yang, et al., “Effects of ultrasound pretreatments on the kinetics of moisture loss and oil uptake during deep fat frying of sweet potato (Ipomea batatas),” Innovative Food Science and Emerging Technologies, vol. 43, pp. 7–17, 2017. doi: 10.1016/j.ifset.2017.07.019 [36] F. Pedreschi, P. Hernández, C. Figueroa, and P. Moyano, “Modeling water loss during frying of potato slices,” International Journal of Food Properties, vol. 8, no. 2, pp. 289–299, 2005. doi: 10.1081/JFP200059480 [37] A. M. Ziaiifar, F. Courtois, and G. Trystram, “Porosity development and its effect on oil uptake during frying process,” Journal of Food Process Engineering, vol. 33, no. 2, pp. 191–212, 2010. doi: 10.1111/j.174 5-4530.2008.00267.x
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 71 Investigation of the Usability of Bayburt Yellow Tuff in Structural Lightweight Concrete Production Omer Bayrak1, Emin Erdem2 1Civil Engineering, Bayburt University, Bayburt, Türkiye 2Department of Chemistry, Pamukkale University, Denizli, Türkiye Corresponding author: Omer Bayrak (e-mail: [email protected]) Abstract Buildings being less affected by earthquakes depends on reducing the dead load of the building as much as possible and the aggregate used in reinforced concrete elements having the desired properties and being light. Lightweight natural aggregates are generally composed of perlite, pumice, and tuffite, which are formed as a result of volcanism and are exposed to high temperatures. The yellow tuff extracted in the Bayburt region provides significant economic income to the region through decorative stone carving, fountains, and the production of load-bearing wall blocks for masonry structures. In this study, the usability of aggregates obtained from Bayburt yellow tuff (BYT) waste generated in the enterprises in the production of structural lightweight concrete was investigated. For this purpose, BYT wastes obtained from the enterprises were ground in the laboratory to a maximum particle diameter (Dmax) of 11.2 mm and separated into 9 different sieve classes (0-0.063, 0.063-0.125, 0.125-0.25, 0.250.5, 0.5-1, 1-2, 2-4, 4-8, and 8-11.2 mm). Aggregate tests were carried out. A granulometry curve was determined in accordance with Dmax = 8 mm (Turkish Standards European norm (TS EN) 802). In addition to BYT aggregate, perlite and pumice aggregates commercially used in lightweight concrete production were also used for comparison. Lightweight concrete (LC) mixes were prepared with a cement dosage of 400 kg/m3 and a water/cement ratio of 0.55. LCs produced with BYT aggregate were compared with lightweight concretes produced with perlite and pumice aggregates at the same cement dosage. It was determined that BYT aggregate cannot be used in LC production due to its lack of pozzolanic properties (2.8 MPa), its unit weight value being higher than the lightweight aggregate standard (2.0 g/cm3) (2.06 g/cm3), and its low inherent strength (23.8 MPa). However, it was determined that BYT aggregate mixtures prepared using at least 400 doses of cement, a water/cement ratio of 0.55, and plasticizer chemical additives had an average 28.2 MPa 28-day compressive strength (higher than LC produced with pumice (14.4 MPa) and lower than LC produced with perlite (41.8 MPa)), but could be successfully used as a normal concrete aggregate. Keywords: Lightweight concrete, Bayburt waste yellow tuff
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 72 Use of Bayburt White Tuff as Cement Substitute Material Omer Bayrak1, Emin Erdem2 1Civil Engineering, Bayburt University, Bayburt, Türkiye 2Department of Chemistry, Pamukkale University, Denizli, Türkiye Corresponding author: Omer Bayrak (e-mail: [email protected]) Abstract Today, approximately 8% of air pollution stems from carbon dioxide emitted into the environment during cement productio. White-colored volcanic tuffs from the Bayburt region, used for interior and exterior cladding of buildings and also for decorative purposes, have an important potential in the production of blended cement due to their pozzolanic properties. In this study, waste Bayburt white tuff (BWT) generated in the enterprises was used both as lightweight aggregate in the production of non-structural lightweight concrete (N-SLC) blocks and as cement replacement material after fine grinding. BWT wastes were ground to a maximum grain diameter (Dmax) of 11.2 mm and separated into 9 different sieve classes (0-0.063, 0.063-0.125, 0.125-0.25, 0.25-0.5, 0.5-1, 1-2, 24, 4-8, 8-11.2 mm) and aggregate tests were carried out. The granulometry curve (TS EN 802) was determined in accordance with Dmax 8 mm. 0-0.063 micron BBT was substituted into the cement at rates of 0%, 10%, 20%, 30% and 40%. N-SLC mixtures prepared with a binder dosage of 200 kg/m3 and a water/cement ratio of 0.50 were compacted in 10x10x10cm cube molds by first shaking and then applying a pressure of 12 MPa. Compressive strength tests were performed after the N-SLC blocks were stored in a humid environment for 7 and 28 days. It was determined that the dry densty values of the N-SLC block elements averaged 1.4 g/cm3, and their 28-day compressive strength values ranged from 5.6 to 7.6 MPa, exceeding the 5 MPa required by the relevant standard. It was determined that the addition of BWT powder, while slightly decreasing in strength, achieved the desired strength. As a result, it was determined that BWT aggregate can be successfully used as an aggregate in N-SLC production, and that BWT powder can be substituted for cement up to 30%. Keywords: Lightweight concrete, Bayburt waste white tuff
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 73 Investigation of Parameters Affecting Pedestrian Crossing Speeds: Empirical Findings from Türkiye Mehmet Burhan Sume1, Erdem Dogan1 1Civil Engineering, Kirikkale University, Kirikkale, Türkiye Corresponding Author: Mehmet Burhan Sume (e-mail: mehmetsume9[email protected]m) Abstract Insufficient pedestrian crossing times in intersection signal systems cause pedestrians to make dangerous crossings and reduce road safety. Therefore, accurately determining pedestrian crossing speeds is crucial. This study examined pedestrian crossing speeds and analyzed the factors affecting these speeds. The study examined the crossings of 500 pedestrians using bird’s-eye video recordings from eight different intersections in the provinces of Izmir, Kirikkale, and Samsun in Türkiye. In addition to pedestrian speeds, numerous variables, including the intersection’s signalization status, geometric characteristics, pedestrian-vehicle intersections, average vehicle speeds, pedestrian group crossings, and pedestrian crossing usage, were analyzed using statistical methods to determine their impact levels. The results obtained in this study can contribute to the current work of researchers and professionals working in this field. The purpose of this study is to help improve pedestrian crossing safety by examining the speeds at which pedestrians cross intersections and the factors affecting these speeds. Keywords: Pedestrian speed, Intersection analysis, Pedestrian behavior, Statistical analysis, Traffic safety 1. INTRODUCTION Throughout history, transportation has been a fundamental need in human life. Roads have been symbols of civilization’s development, and population and vehicle numbers have steadily increased. Today, pedestrian and vehicle density in many city centers is significantly higher than in the past. This excess has necessitated the development of traffic systems. Intersections play a crucial role in these traffic systems. As critical intersections where multiple roads intersect, they are frequently used by numerous pedestrians and vehicles, often resulting in simultaneous contact. Pedestrians are the most vulnerable road users during these encounters. Inadequate time allowed for pedestrians in intersection designs leads to dangerous crossings, which lead to various accidents at intersections. Many pedestrians are injured or killed each year due to these accidents. According to data from the United States Department of Transportation’s National Center for Statistics and Analysis, a total of 7,314 pedestrians were killed and an estimated 68,244 injured in traffic crashes in the United States in 2023. This means one pedestrian died every 72 minutes in the United States in 2023. These pedestrian deaths in the U.S. in 2023 accounted for 18% of all traffic deaths that year [1]. As another example, Poland has the second-worst pedestrian fatality rate in the European Union. Between 2007 and 2012, 9,101 pedestrians died and 71,328 were injured in traffic accidents in Poland. Approximately 30% of pedestrian injury accidents occurred at unsignalized zebra crossings [2]. In Türkiye, according to data from the Ministry of Transport’s General Directorate of Highways (KGM), a total of 6,351 people lost their lives in traffic accidents in 2024, 21% of which (1,333 people) were pedestrians. In 2024, 1,884 accidents occurred at pedestrian crossings in Türkiye, resulting in 3,318 injuries and 149 deaths [3]. Based on these data, accurate analysis of pedestrian behavior at intersections is important for the safety of road users and the efficiency of traffic systems. The primary objective of this study is to contribute to the existing literature by analyzing pedestrian crossing speeds at intersections and the factors affecting these speeds. In this context, the crossings of 500 pedestrians using eight different intersections in the provinces of Kirikkale, Samsun, and Izmir in Türkiye were examined. The data obtained from the examination of intersections of different characteristics were subjected to statistical analysis. The analyses were conducted in the Python environment and summarized in tables. The obtained results are supported by visuals and compared with various sources in the literature. The final section of the study proposes various recommendations, and the study concludes. 1.1. Literature Review
74 Because pedestrian behavior is variable, average pedestrian speeds can vary across countries and even regions. Some studies indicate that cultural differences significantly influence pedestrian behavior. Interventions to improve road safety should take these cultural factors into account [4]. Another study that clearly demonstrates the variation in pedestrian speeds across countries is conducted by Goh et al. [5]. Goh et al. summarize the average pedestrian speeds across countries in Table 1. Table 1. Average pedestrian walking speeds in different countries Country Average Walking Speed (m/s) Asia Riyadh, Saudi Arabia 1.08 Madras, India 1.20 Hong Kong 1.20 Thailand 1.22 Singapore 1.23 Colombo, Sri Lanka 1.25 Israel 1.31 Malaysia 1.39 Jordan 1.34 USA Columbia 1.32 New York 1.35 Pittsburg 1.47 Others England 1.31 Calgary, Canada 1.40 According to the data summarized in the table, average pedestrian speeds range from 1.08 to 1.40 m/s. The speed data determined in a follow-up study are as follows: In Malaysia, the average and 15th percentile pedestrian speeds at signalized pedestrian crossings are 1.31 m/s and 1.09 m/s, respectively. At non-signalized pedestrian crossings, the average and 15th percentile pedestrian speeds are 1.39 m/s and 1.15 m/s, respectively. However, the design speed applied at the intersections is 1.22 m/s, and neither intersection can ensure safe pedestrian crossing time [5]. In another study, pedestrian speed data was collected at an unsignalized intersection and then a signaling system was installed at the intersection. After the installation, the average pedestrian speed decreased by 23% compared to the pre-installation speed, from 1.37 m/s to 1.05 m/s. The recommended design speed decreased from 0.73 m/s to 0.52 m/s. Both values are lower than the pedestrian design speed applied at the intersection (1.20 m/s) [6]. Some studies have shown that the crossing speed is higher at unsignalized intersections [7, 8]. In a study conducted at a four-leg signalized intersection in Mumbai, India, the average pedestrian crossing speed was found to be 1.34 m/s, which is different from the design speed (1.20 m/s). Another study determined the 15th percentile speed of pedestrians crossing at crosswalks to be 1.24 m/s. It has been stated that this value may be suitable for the signalized design speed of pedestrian crossings [9]. In a study conducted in Türkiye, the average crossing speed was found to be 1.31 m/s and the 15th percentile crossing speed was found to be 1.07 m/s. In addition, the study stated that the recommended average walking speed in Türkiye is 1.4 m/s given by the Turkish Standards Institute [10]. In Jordan, a design speed of 1.11 m/s was recommended [11]. There is a consensus in studies that the 15th percentile speed should be the design speed. Studies generally indicate that the recommended pedestrian speed is lower than the current design speeds. This indicates that pedestrians do not have enough time to cross. However, there are also studies where the design speed is lower than the recommended speed range. In a study conducted in India, 16 signalized intersections were examined and the 15th percentile crossing speed was between 1.11 and 1.31 m/s. These values exceed the design crossing speed of 0.95 m/s [12]. When these studies in the literature were examined, it was determined that the 15th percentile pedestrian speeds determined were different from the current design speeds. The times allocated for pedestrians to cross at intersection signals do not match the speeds of pedestrians at the local level. This can cause pedestrians to have difficulty crossing the street or make dangerous crossings. Therefore, improvements in intersection designs are necessary.
75 In studies examining other factors affecting pedestrian speed, the most striking parameter is group crossings. Many studies have found that pedestrians crossing in groups are slower than individual pedestrians [4, 7, 8, 11, 13–19]. It is thought that distraction during group crossings may be a factor in this situation. Another study found that as pedestrians’ crossing distance increases, their crossing speed also increases [20]. This is thought to be due to pedestrians’ desire to minimize the time they spend in the intersection. Pedestrians crossing wider crosswalks are found to be faster than those crossing narrower crosswalks. This is likely due to pedestrians’ greater ease of movement in wider crosswalks [8, 11, 17]. Drivers’ behavior towards pedestrians is generally negative. Some studies have found that drivers tend not to yield to pedestrians [21, 22]. This situation causes safety problems and significant improvements are needed for pedestrian safety [16]. In this study, the investigation of pedestrian speed was conducted using video analysis. A separate study has stated that pedestrian-vehicle interactions and pedestrian behavior can be analyzed through video analysis [23]. In our study, video analysis was conducted using the open-source software Kinovea®. This software determined the speeds of pedestrians and vehicles, and calculated the geometric properties of pedestrian crossings. Some studies have been conducted to evaluate the reliability of Kinovea, and the program is reported to be accessible and easy to use based on the parameters examined, and to be a reliable tool for measuring speed-based training data [24, 25]. Numerous studies exist in the literature examining pedestrian speed. These studies examine various factors, but few studies examine multiple factors simultaneously. This study examines pedestrian speeds and a total of 20 parameters affecting these speeds. 2. MATERIALS AND METHODS 2.1. Data Collection The data in this study were obtained from a total of eight signalized and unsignalized intersections with varying numbers of branches in the provinces of Izmir, Samsun, and Kirikkale. The general characteristics of the intersections examined are given in Table 2. Table 2. General characteristics of the examined intersections Province Intersection Number of Legs Intersection Control Type Number of Inspected Intersections Kirikkale Five-Legged Non-Signalized 1 Samsun FourLegged Non-Signalized 1 Izmir FourLegged Signalized 5 Izmir ThreeLegged Non-Signalized 1 Izmir is located in western Türkiye, Samsun in northern Türkiye, and Kirikkale in central Türkiye. The cities studied have distinct characteristics. Of the eight intersections studied, five are signalized and three are unsignalized. In terms of the number of intersection legs, six of the eight are four-leg intersections, while threeand five-leg intersections have one each. The vast majority of the intersections studied are four-leg signalized. The intersections were recorded between 8:30 am and 4:30 pm, in good weather conditions, with fifteen-minute drone recordings. The recorded footage was used to examine various conditions, including pedestrian-vehicle intersections, group crossings, carrying goods or pets, visibility of pedestrian crossings, types of intersection controls, and the dates and times of video recordings for the intersections studied. Intersection features such as the width and length of pedestrian crossings, the geometric characteristics of intersections, and the presence of pedestrian crossing lines and sidewalks were also analyzed. Pedestrian speeds were calculated by measuring the distance and time pedestrians crossed. All data analyzed in the videos was grouped and listed in Table 3. The analyzed data consists of numerical or categorical variables. Numerical variables include the date and time of the video, the length and width of the crosswalk, the number of vehicles impacting the pedestrian and their average speed, group size, the distance and time the pedestrian crossed, and the pedestrian’s speed. Categorical variables are cities (Kirikkale, Izmir, and Samsun), the time of the video (morning, noon, evening), and the geometric characteristics of the intersection (3-leg unsignaled, 4-leg signaled, 4-leg unsignaled, and 5-leg unsignaled). All other groups have two variables, yes or no. The Google Earth program was used to determine the physical characteristics of the intersections. This program used the distance measurement feature to measure the geometric characteristics of the intersections based on their actual locations. (Figure 1)
76 Table 3. Analyzed data Crosswalk and Intersection Features Pedestrian Behavior Pedestrian-Vehicle Interactions City Geometric Features of the Intersection Carrying Goods or Dogs Pedestrian-Vehicle Contact at the Crossing Video Date Signalization Pedestrian Crossing Usage Number of Cars Impacting the Pedestrian Video Time Presence of Crosswalk Lines Group Size Number of Motorcycles Impacting the Pedestrian Length and Width of The Pedestrian Crossing (m) Existence of Obstacles at the Crossing Pedestrian’s Crossing Distance (m) Number of Heavy Vehicles Impacting the Pedestrian Existence of Usable Sidewalks Visibility of the Road from the Pedestrian’s Eyes While Crossing Pedestrian Speed (m/s) Average Speed of Vehicles Impacting the Pedestrian (km/h) Figure 1. Google Earth example pedestrian crossing distance measurement The obtained values were transferred to Kinovea, a video analysis program, and the necessary calibrations were performed, and measurements and analyses were continued in this program. The measurement-based data were measured using Kinovea’s video measurement tools. Since the length and width of the pedestrian crossing (m), the average speed of vehicles impacting the pedestrian (km/h), and the pedestrian’s crossing distance (m) are measurement-based data, they were measured using Kinovea’s measurement features (Figure 2). The average speed of vehicles impacting the pedestrian (km/h) was calculated using Kinovea’s moving object tracking feature (Figure 3). All other data is not based on metric measurements. It was visually extracted from the videos and converted into data. A coordinate system (red) was placed on the video screen to better identify the location of pedestrians and crossings. The origin of the coordinate system was chosen as the center of the intersection. Pedestrian speed was calculated using the following formula. n i i1 n i i1 d vt = = = (1) Here, v is pedestrian speed (m/s), di is distance traveled (m), and ti is elapsed time (s). The entire distance traveled by a total of (n) pedestrians, from the moment they first stepped onto the roadway to cross the roadway until the moment they exited the roadway, was marked with dots at short intervals. The orange dots in Figure 2 represent
77 the paths followed by pedestrians crossing the roadway (di). These dots were then connected to measure the distances each pedestrian crossed. The distances between two points are shown within the orange signs. The times at which pedestrians entered and exited the intersection were recorded, and the total crossing times were calculated ti. The program’s path-following feature calculated the approximate speeds of vehicles at the moment they contacted pedestrians (Figure 3). Figure 2. Kinovea program sample pedestrian crossing measurement Figure 3. Kinovea program path tracking feature Vehicles that impacted pedestrian speed by contacting them were tracked using the program’s path tracking feature before reaching the point of contact, and tracking was discontinued shortly after the contact was completed. The program automatically displays the distance and elapsed time the vehicle traveled during the tracking period. Using the data obtained, the vehicle’s speed was calculated as in Equation (1). Since the speed data was in m/s, it was converted to km/h. 2.2. Data Analysis The obtained data were subjected to a two-stage analysis process in the Python environment. In the first stage, descriptive statistical analysis was conducted to reveal the general distribution of the numerical data. This analysis
78 yielded general statistical values for the numerical variables, such as mean, minimum, maximum, standard deviation, and 15-85 percentiles. In the second stage, the effects of various factors on pedestrian speed were evaluated using various tests. Statistical analyses are basically divided into two groups: parametric and non-parametric tests. If the data are not normally distributed and consist of small samples, non-parametric tests are applied [26]. To determine which tests to apply to the existing data set, first, tests called normality tests are applied. The Shapiro-Wilk Normality Test was applied to determine whether the data set in this study was normally distributed. Among normality tests, the Shapiro-Wilk test is said to be the most powerful test for all distributions and sample sizes [27]. The Shapiro-Wilk test revealed that the data were not normally distributed, and therefore, non-parametric tests were applied. These tests are the Mann–Whitney and Kruskal–Wallis tests. The Mann-Whitney U test was applied to groups with two variables, and the Kruskal-Wallis H test was applied to groups with three or more variables to examine whether the variables had significant effects on pedestrian speed. For example, the Mann-Whitney U test was used to examine whether there was a significant difference between the speed of pedestrians at signalized intersections and those at unsignalized intersections. However, the Kruskal-Wallis H test was used to test whether there was a significant difference between the median pedestrian speeds for the provinces of Izmir, Kirikkale, and Samsun. When the Kruskal-Wallis test result was significant, pairwise (post-hoc) comparisons were made between the groups with significant differences to determine which two groups had the most significant difference. In these pairwise comparisons, the Bonferroni correction method was applied to correct the risk of obtaining erroneous results [28]. When the P-values obtained from these tests are <0.05, a significant difference is indicated between the groups. By examining all variables in this manner, those that significantly affected pedestrian speed were identified. 3. RESULTS 3.1. Descriptive Analysis Descriptive analysis was applied to determine the general distribution of numerical groups. The results of the descriptive analysis are presented in Table 4. Table 4. Descriptive analysis results (numerical groups) Variable Mean Std Min Max Mode P15 P50 (Median) P85 Group Size 1.16 0.48 1.00 4.00 1.00 1.00 1.00 1.00 Number of Cars Impacting Pedestrians 0.62 1.29 0.00 12.00 0.00 0.00 0.00 1.00 Number of Heavy Vehicles Impacting Pedestrians 0.04 0.21 0.00 2.00 0.00 0.00 0.00 0.00 Number of Motorcycles Impacting Pedestrians 0.06 0.28 0.00 2.00 0.00 0.00 0.00 0.00 Average Speed of Impacted Vehicles (km/h) 8.81 14.14 0.00 60.00 0.00 0.00 0.00 25.00 Length of the Pedestrian Crossing (m) 9.09 2.77 0.00 19.23 7.00 7.00 9.18 11.27 Width of the Pedestrian Crossing (m) 4.16 1.21 0.00 6.00 3.00 3.00 4.25 5.50 Pedestrian’s Crossing Distance (m) 12.04 7.72 4.30 67.19 7.00 7.33 10.18 15.00 Pedestrian Speed (m/s) 1.35 0.36 0.36 3.93 1.23 1.05 1.32 1.58 Std: Standard deviation, Min: Minimum, Max: Maximum, P15: 15th percentile, P50 (Median): 50th percentile, and P85: 85th percentile According to the descriptive analysis results, at least 85% of pedestrians cross the street alone. Group crossings are rare. More than 50% of pedestrians do not encounter any vehicles while crossing. In cases where vehicles are encountered, the vehicle speed affecting the pedestrian is sometimes greater than 25 km/h. The average length and width of the crossings used by pedestrians are 9.18 and 4.25 meters, respectively. It is believed that the length and width of the crossings do not pose any problems in terms of usability. However, there are significant differences in the lengths of pedestrian crossings (It is between 7.33 and 15 meters according to the 15-85 percentile). Pedestrian preferences played a role in this. Some pedestrians followed the rules and used the crosswalks, while others took alternative routes, making dangerous and lengthy crossings. The median pedestrian speed was 1.32 m/s, and the 15th percentile speed was 1.05 m/s, which are consistent with literature data.
79 3.2. Normality Test Determining whether the data distribution is normal serves as a guide for the remainder of the analysis. The Shapiro-Wilk Normality Test was used in this study. Based on the test results, W=0.853 and p=0.00000 was determined. Because the W-value was not equal to 1 and the p-value was less than 0.05, it was determined that the pedestrian speed data were not normally distributed. Therefore, the analysis continued with non-parametric tests. 3.3. Hypothesis Tests and Results 3.3.1 Hypothesis Testing and Results Based on Categorical Variables The Kruskal-Wallis test was applied to categorical groups with three or more variables. The test results are given in Table 5. Table 5. Kruskal-Wallis test results Variable H p Number of Observations Significance Status City (Kirikkale, Izmir, Samsun) 2.281 0.31972 500 p > 0.05 Meaningless Video Time (Morning, Afternoon, Evening) 3.761 0.15250 500 p > 0.05 Meaningless Video Date (March 2024, June 2024, March 2025) 2.281 0.31972 500 p > 0.05 Meaningless Intersection Geometry (3-, 4-, and 5-Leg Intersections) 10.406 0.01541 500 p < 0.05 Significant Here, the H statistic represents the distance between the ranked means of categorical variables. However, the key indicator of significance is the p-value. A value less than 0.05 indicates that the relevant variable significantly affects pedestrian speed. According to the results in Table 5, no significant difference was found between cities (Kirikkale, Izmir, Samsun) affecting pedestrian speeds. Similarly, no significant difference was observed between pedestrian speeds during morning, afternoon, or evening crossings. Based on these test results, it was determined that the geometric characteristics of the intersection may have an impact on pedestrian speed. There are four different intersection types. A Post-Hoc pairwise comparison test with Bonferroni correction was applied to measure the relationship between each. The test results are presented in Table 6. The number of observations (n) for each intersection type is given in the definitions section immediately below the table. Table 6. Post-Hoc test results by intersection geometry (2: Four-legged non-signalized intersection (n = 68), 3: Three-legged non-signalized intersection (n = 15), 4: Four-legged signalized intersection (n = 334), and 5: Fivelegged non-signalized intersection (n = 83)) Variable Median 1 Median 2 U p_raw p_adj (Corrected) Significance Status (2 vs 3) 1.330 1.180 743.50 0.00581 0.03487 p < 0.05 Significant (2 vs 4) 1.330 1.335 11367.50 0.98995 1.00000 p > 0.05 Meaningless (2 vs 5) 1.330 1.290 3178.50 0.18297 1.00000 p > 0.05 Meaningless (3 vs 4) 1.180 1.335 1425.50 0.00476 0.02854 p < 0.05 Significant (3 vs 5) 1.180 1.290 426.00 0.05309 0.31851 p > 0.05 Meaningless (4 vs 5) 1.335 1.290 15483.50 0.09878 0.59270 p > 0.05 Meaningless Here, Median 1 and Median 2 are the median values of the two different intersection types being compared. The nature of the test involves comparing the medians between two variables. The U value is one of the outputs of this test. As this value decreases, the probability of the difference between the groups being significant increases. The P-raw value is the raw p-value resulting from the test. Because there are more than two groups, the Bonferroni correction is applied to control the false positive rate. The final decision is made based on this p-value. According to the test results, pedestrian speeds at four-leg signalized and four-leg unsignalized intersections are higher than at three-leg unsignalized intersections. The density distribution graphs are shown in Figure 4 and Figure 5.
86 in 19th EURO Working Group on Transportation Meeting, EWGT2016, Istanbul, Turkey, 2016, pp. 3–12. doi: 10.1016/j.trpro.2017.03.002 [11] M. S. Tarawneh, “Evaluation of pedestrian speed in Jordan with investigation of some contributing factors,” J Safety Res, vol. 32, pp. 229–236, 2001. [12] A. Bansal, T. Goyal, and U. Sharma, “Modelling the pedestrian speed at signalised intersection crosswalks for heterogeneous traffic conditions,” Traffic & Transportation, vol. 31, pp. 681–692, 2019. [13] S. Marisamynathan and P. Vedagiri, “Modeling pedestrian delay at signalized intersection crosswalks under mixed traffic condition,” in 2nd Conference of Transportation Research Group of India (2nd CTRG), Mumbai, Dec. 2013, pp. 708–717. doi: 10.1016/j.sbspro.2013.11.165 [14] S. Marisamynathan and P. Vedagiri, “A new approach to estimate pedestrian delay at signalized intersections,” Transport, vol. 33, no. 1, pp. 249–259, 2018. doi: 10.3846/16484142.2016.1158208 [15] M. Giannoulaki and Z. Christoforou, “Pedestrian walking speed analysis: A systematic review,” Sustainability (Switzerland), vol. 16, no. 11, pp. 1–19, Jun. 2024. doi: 10.3390/su16114813 [16] D. Muley, M. Kharbeche, L. Downey, W. Saleh, and M. Al-Salem, “Road users’ behavior at marked crosswalks on channelized right-turn lanes at intersections in the state of Qatar,” Sustainability (Switzerland), vol. 11, no. 20, pp. 1–19, Oct. 2019. doi: 10.3390/su11205699 [17] N. M. Asmael and M. Q. Waheed, “Analysis of pedestrian characteristics crossing along roads,” Transport Problems, vol. 18, no. 4, pp. 189–200, 2023. doi: 10.20858/tp.2023.18.4.15 [18] Y. Poojari, E. Prashanth, D. Divya, and M. Kalyani, “Modelling pedestrian crossing behaviour at a midblock section,” Slovak Journal of Civil Engineering, vol. 29, no. 3, pp. 15–21, Sep. 2021. doi: 10.2478/sjce-2021-0017 [19] S. Ahmed, S. Hossain, M. E. Shaik, and A. Shakik, “Evaluation of speed characteristics and gap acceptance behavior of pedestrians of Asian countries: A review,” Transp Res Interdiscip Perspect, vol. 27, art. no. 101199, Jun. 2024. doi: 10.1016/j.trip.2024.101199 [20] D. Muley, W. Alhajyaseen, M. Kharbeche, and M. Al-Salem, “Pedestrians’ speed analysis at signalized crosswalks,” in 9th International Conference on Ambient Systems, Networks and Technologies (ANT 2018), Doha, 2018, pp. 567–574. doi: 10.1016/j.procs.2018.04.102 [21] H. Amado, S. Ferreira, J. P. Tavares, P. Ribeiro, and E. Freitas, “Pedestrian-vehicle interaction at unsignalized crosswalks: a systematic review,” Sustainability (Switzerland), vol. 12, no. 7, pp. 1–23, Apr. 2020. doi: 10.3390/su12072805 [22] C. Feliciani, L. Crociani, A. Gorrini, G. Vizzari, S. Bandini, and K. Nishinari, “A simulation model for unsignalized pedestrian crosswalks based on evidence from field observation,” J. Advanced Computational Intelligence and Intelligent Informatics, vol. 18, no. 1, pp. 1–20, 2014. [23] P. Olszewski, I. Buttler, W. Czajewski, P. Dąbkowski, C. Kraśkiewicz, P. Szagała, et al., “Pedestrian safety assessment with video analysis,” in 6th Transport Research Arena, Warsaw, Poland, 2016, pp. 2044–2053. doi: 10.1016/j.trpro.2016.05.172 [24] P. Fernández-González, A. Cuesta-Gómez, J. C. Miangolarra-Page, and F. Molina-Rueda, “Reliability and validity of Kinovea to analyze spatiotemporal gait parameters,” Revista Internacional de Medicina y Ciencias de la Actividad Fisica y del Deporte, vol. 22, no. 87, pp. 565–578, Sep. 2022. doi: 10.15366/rimcafd2022.87.009 [25] J. M. Jiménez-Olmedo, A. Penichet-Tomás, L. Villalón-Gasch, and B. Pueo, “Validity and reliability of smartphone high-speed camera and Kinovea for velocity-based training measurement,” Journal of Human Sport and Exercise, vol. 16, no. 4, pp. 1–11, 2021. doi: 10.14198/jhse.2021.164.11 [26] G. Norman, “Likert scales, levels of measurement and the ‘laws’ of statistics,” Advances in Health Sciences Education, vol. 15, no. 5, pp. 625–632, Dec. 2010. doi: 10.1007/s10459-010-9222-y [27] N. Mohd Razali and Y. Bee Wah, “Power comparisons of Shapiro-Wilk, Kolmogorov-Smirnov, Lilliefors and Anderson-Darling tests,” Journal of Statistical Modeling and Analytics, vol. 2, no. 1, pp. 21–33, 2011. [28] R. A. Armstrong, “When to use the Bonferroni correction,” Ophthalmic Physiol Opt., vol. 34, no. 5, pp. 502–508, 2014. doi: 10.1111/opo.12131 [29] S. Ascione, V. Piscopo, and A. Scamardella, “Comparison of linear and rank-correlation for the sea state estimation,” in 2024 IEEE International Workshop on Metrology for the Sea, MetroSea 2024 - Proceedings, 2024, pp. 250–255. doi: 10.1109/MetroSea62823.2024.10765705
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 87 Performance Properties of Lightweight Aggregate Hybrid Foam Concretes Selcuk Memis1 1Department of Civil Engineering, Kastamonu University, Kastamonu, Türkiye Corresponding author: Selcuk Memis (e-mail: [email protected]) Abstract Hybrid geopolymers are binder systems in which an alkali-activated binder phase (N-A-S-(H)/C-(N)-A-S-H gels formed by activating Al-Si rich sources such as fly ash, slag, metakaolin with NaOH/Na2SiO3 etc.) and portland cement (ordinary portland cement (OPC)) hydration products (especially C-S-H/C-A-S-H) develop together in the same matrix. The aim is to combine the early age strength/setting advantages of cement with the low CO2, thermal stability, and some durability advantages of geopolymer in a single material. In this study, the fresh and hardened properties of geopolymers, cement–geopolymer hybrids, and cementitious mortars with six different binder compositions produced using pumice aggregate were investigated. The fly ash (FA) and cement (OPC) ratios were systematically varied; sodium hydroxide (NaOH) and sodium metasilicate (Na2SiO3) were used for activation. The flow, unit weight, water absorption, thermal conductivity (λ), and 3-7-28 day flexural/compressive strengths were measured. The 28-day compressive strength reached 6.14 MPa in the pure geopolymer series, while the hybrid systems achieved a range of 2.90–5.53 MPa. The lowest water absorption (32%) and relatively low λ value (≈ 295 W/mK) were observed in hybrit group at 2. The results show that density and λ increase with increasing cement ratio in the hybrid systems, while water absorption decreases. The findings suggest that geopolymer and hybrid binders are suitable alternatives for the design of lightweight and fire-resistant structural elements. Keywords: Sustainability, Pumice, Geopolymer concrete, Hybrid binders, Thermo-mechanical performance
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 88 The Role of Biodiesel Addition in Reducing Vapor Pressure in Gasoline– Methanol–Ethanol Ternary Blends Abdulvahap Cakmak1 1Department of Mechanical Engineering, Faculty of Engineering and Natural Sciences, Samsun University, Ballica Campus, Ondokuzmayis, 55420 Samsun, Türkiye Corresponding author: Abdulvahap Cakmak (e-mail: [email protected]) Abstract This study examines how adding biodiesel affects the vapour pressure and other physicochemical properties of gasoline–methanol–ethanol ternary fuel blends. Vapour pressure is a key factor influencing volatility, evaporation, engine performance, and emissions, especially in blends with low-boiling-point alcohols such as methanol and ethanol. However, blending methanol and ethanol into hydrocarbon-based gasoline significantly raises the vapour pressure of the mixture due to azeotropic interactions, even though methanol and ethanol individually have lower vapour pressures. To overcome this, biodiesel, a high-boiling, low-volatility renewable fuel, was incorporated into various ternary blend formulations. Biodiesel was added to alcohol-gasoline blends containing total alcohol (methanol and ethanol) at 10%, 20%, and 30%v/v, with biodiesel concentrations of 1%, 3%, and 5%v/v, respectively. The vapour pressure, density, distillation, and sulphur content measurements were taken using standard testing methods. The highest dry vapour pressure equivalent (DVPE) value of 72 kPa was recorded at a 10% v/v alcohol content, 6.04% higher than base gasoline. However, with the addition of 1% v/v biodiesel, the DVPE decreased slightly to 71.5 kPa. Biodiesel was also found to enhance the density and distillation properties of the ternary blends. Crucially, the measured fuel properties largely complied with the European standard for gasoline (EN 228), except for some samples’ DVPE and distillation values at E70 and E100. These findings suggest that biodiesel can serve as a volatility stabiliser in alcohol-containing fuel formulations and as a renewable fuel component. The study highlights the limitations of current work and offers recommendations for further research into engine performance, combustion characteristics, and the long-term storage stability of the blends. Keywords: Alcohol-gasoline blends, Biodiesel, Fuel volatility, Physicochemical properties, Vapor pressure 1. INTRODUCTION Driven by rapid industrial growth, increasing energy demands, and heightened concern over climate change, research into alternative fuels with lower environmental impact and improved carbon performance has gained significant attention among scientists [1]. In this context, incorporating various alcohols into gasoline blends has attracted growing interest in reducing emissions, decreasing fossil fuel consumption, and enhancing fuel quality. Among these alcohols, methanol and ethanol are the most widely used in engine applications due to their advantageous properties, including renewability, broad availability, compatibility with existing engine technologies, oxygenated structure, and high octane rating. Methanol and ethanol have long been commercially blended into gasoline in small fractions, varying by country, to extend fuel supplies and improve gasoline quality, particularly its octane rating. However, growing concerns over the limited availability of fossil fuel resources and their associated environmental and health impacts have highlighted the need to gradually reduce fossil fuel consumption and increase the share of renewable fuels in the energy market. There are ongoing projections and policy initiatives, supported by government regulations and incentives, to gradually increase the share of ethanol and methanol in the transportation sector. For example, the United States renewable fuel standard (RFS) mandates blending biofuels, including ethanol, into gasoline, targeting 36 billion gallons of renewable fuel use by 2022 and beyond [2]. Similarly, the European Union’s renewable energy directive (RED II) requires member states to achieve at least a 14% share of renewable energy in transportation by 2030, with advanced biofuels playing a significant role [3]. China has also implemented E10 (10% ethanol) mandates in several provinces, with plans for nationwide expansion [4]. These initiatives reflect a global trend toward increasing alcohol fuel usage to reduce dependence on fossil fuels, improve energy security, and lower greenhouse gas emissions.
89 Methanol and ethanol are oxygenated fuels with a low carbon-to-hydrogen ratio, contributing to cleaner combustion. As a result, ethanol–methanol–gasoline blends tend to reduce emissions of carbon monoxide (CO), hydrocarbons (HCs), particulate matter (PM), and air toxics from gasoline engine vehicles. Additionally, due to their higher octane number, methanol and ethanol provide a convenient and cost-effective way to upgrade lowoctane gasoline. However, when alcohols, particularly short-chain alcohols like methanol and ethanol, are blended with gasoline, the resulting mixture exhibits reduced distillation temperatures. It deviates from ideal mixing behavior due to forming a near-azeotropic mixture (non-ideal blending). This effect is especially pronounced at low alcohol concentrations (3-10%, v/v) [5], where the final blend may exhibit a vapor pressure higher than that of pure gasoline or the alcohol alone. Therefore, although methanol and ethanol have relatively low reid vapor pressures (RVP) of about 32 kPa and 13 kPa, respectively, their blended RVP in gasoline (60 kPa) can range from 65 kPa up to as high as 200 kPa, depending on the alcohol content in the blend. When blending to meet an RVP requirement, refineries need to remove some butane from the gasoline to compensate for the RVP increase caused by the alcohol. However, at higher alcohol blending levels, both methanol and ethanol cause only a slight increase in RVP, and the RVP response curve becomes relatively flat. Vapour pressure refers to the pressure exerted by the vapour of a liquid in equilibrium with its liquid phase in a closed container. It serves as an indicator of a fuel’s volatility and is a critical property of automotive gasoline. Higher vapour pressure indicates greater volatility, and vice versa. Excessively high and low vapour pressures can affect engine operation and cold-start performance. High vapour pressure may lead to fuel line blockages, engine stalling, unstable engine operation, and increased evaporative HC emissions. Conversely, a vapour pressure that is too low can cause difficulties during cold starts. Therefore, the EN 228 gasoline specification defines allowable vapour pressures based on the type of gasoline and its ethanol content. According to this standard, the vapour pressure of winter-grade gasoline should range from 60 to 90 kPa, while summer-grade gasoline should fall between 45 and 60 kPa. However, the upper limit for summer vapour pressure increases depending on the ethanol content. As the use of methanol and ethanol in gasoline blends is expected to rise, it is crucial to lower the vapour pressures of these blends. Several researchers [6–10] have studied this challenge by investigating various strategies, such as incorporating higher alcohols or other oxygenated additives and optimizing blending ratios, to mitigate the increase in vapour pressure while maintaining or improving fuel properties and engine performance. From this context, this study examines the vapour pressures of dual-blends of methanol and ethanol with gasoline by adding biodiesel as biofuel components. Biodiesel has already been widely used as a biofuel for diesel engines worldwide. Compared to ethanol and methanol, biodiesel is a highly available fuel with better lubricity, a higher boiling point, and a lower vapor pressure. These unique properties make biodiesel especially interesting for research, particularly when mixed with dual alcohol–gasoline blends. Combining these biofuels allows for taking advantage of each, resulting in a single fuel with improved overall qualities. To the best of the authors’ knowledge, no previous study has investigated the effect of biodiesel on vapour pressure reduction in methanol–ethanol–gasoline blends, thereby underscoring the novelty of the present research. 2. MATERIAL AND METHOD To conduct this research, the fuel samples were first prepared, followed by fuel analyses performed following European standard/International Organisation for Standardisation (EN/ISO). The test fuel samples were prepared using commercial 95-octane gasoline obtained from a fuel station, methanol and ethanol (95% purity) supplied by TEKKIM, and biodiesel meeting the EN 14214 specification provided by DP Tarımsal Enerji. To prepare the methanol–ethanol–gasoline–biodiesel blends, high-quality pipettes, beakers, and volumetric burettes with maximum precision were used. First, three dual alcohol–gasoline blends were prepared by adding methanol and ethanol to gasoline at volumetric ratios of 5%, 10%, and 15% v/v each. The resulting mixtures contained total alcohol contents of 10%, 20%, and 30% v/v, designated as A10, A20, and A30, respectively. Subsequently, biodiesel was blended with these dual alcohol–gasoline fuels at volumetric ratios of 1%, 3%, and 5% v/v, producing the quaternary blends labeled A10+B1, A20+B3, and A30+B5, respectively. Furthermore, base summer gasoline was incorporated as a reference for comparative analysis of fuel properties. In this study, the base gasoline may contain up to 3% v/v methanol and 5% v/v ethanol. Ideally, a base gasoline free of ethanol and methanol would have been used; however, such a fuel is currently unavailable on the market. Figure 1 shows the composition of fuel samples. DVPE, density, sulfur content, and distillation temperature were measured using the standard test methods and the equipment listed in Table 1.
90 Table 1. The equipment used for fuel specification Property Equipment Method Accuracy Density Mettler Toledo DE40 EN ISO 3675 ±0.0001 g/cm3 RVP Herzog HVP 972 EN 13016 ±0.2 kPa Distillation Herzog OptiDist TM EN ISO 3405 ±0.1 °C Sulfur Analytik Jena Multi EA 3100 EN ISO 20846 ±1 ppm Figure 1. Composition of fuel samples 3. RESULTS The RVP indicates the front-end volatility of gasoline and is closely related to its butane content. The RVP of gasoline blends can be adjusted by adding or removing normal butane during refining [11]. It is always reported in absolute pressure units such as psi, kPa, or bar. A higher RVP value corresponds to greater fuel volatility. Since the differences between RVP and DVPE are minimal, DVPE is generally considered equivalent to RVP [12]. Figure 2 presents the measured DVPE values of the fuel samples. In this figure, the numerical values are displayed inside each bar, while the relative change compared to the base gasoline is indicated above the bar. Since the ethanol content of summer gasoline is 5% v/v, the allowable increase in vapor pressure is 8 kPa, setting an upper DVPE limit of 68 kPa. The measured DVPE was 67.9 kPa. When the alcohol fraction was increased to 10% v/v, the DVPE rose to 72 kPa, about 6% higher than the base gasoline. This behavior suggests the presence of an azeotropic mixture at such low blending ratios. Further increases in alcohol concentration led to only a slight reduction in DVPE; for example, increasing the alcohol fraction from 20% to 30% v/v decreased DVPE by just 0.5 kPa, although values remained higher than the base gasoline. Adding 1%, 3%, and 5% v/v biodiesel to the A10, A20, and A30 blends decreased the DVPE by only 0.8, 0.7, and 1.7 kPa, respectively. This shows that the azeotropic effect of the hydroxyl group stays strong at low alcohol levels. As a result, biodiesel’s vapor pressurereducing ability is limited in these conditions, implying that a higher biodiesel blending ratio might be necessary to achieve a more substantial reduction. However, the DVPE values of all fuel samples, except for the base gasoline, fall outside the EN 228 specification, which defines an RVP range of 45–68 kPa for summer gasoline. Fuel density is a crucial factor that affects engine performance, emissions, durability, and the operation of fuel system parts. High-density fuels generally provide more energy per volume, improving fuel economy and power output. However, high or low density can hamper spray atomization and air-fuel mixing, reducing combustion efficiency. Density influences flow rates, injector and pump calibration, and fuel metering accuracy in fuel systems. Deviations from the expected density can lead to improper injection, lowering efficiency, and raising emissions. Fuels must meet standard density ranges to ensure optimal engine performance and reliability. According to the EN 228 specification, gasoline density should be between 720 and 775 kg/m³ at 15 °C. Figure 3
91 presents the density measurement results at 15 °C for the prepared fuel samples. Since methanol, ethanol, and biodiesel each have higher densities than gasoline, the ternary (dual alcohol–gasoline) and quaternary (dual alcohol–biodiesel–gasoline) fuel blends exhibited higher density values than neat gasoline. As the proportion of alcohols and biodiesel in the blends increased, the density values rose accordingly, as expected. The highest density, measured at 767.3 kg/m3, was observed for the A30+B5 fuel sample, representing a 3.2% increase relative to gasoline. Despite these increases, the density values of all fuel samples remained within the acceptable range defined by the EN 228 specification limits. Figure 2. DVPE measurement results Figure 3. Density measurement results The lower heating value (LHV) is a critical parameter in fuel selection for internal combustion engines, as it directly influences energy efficiency, combustion characteristics, and fuel handling [13]. Although no specific fuel regulation currently mandates a minimum LHV, it remains a key performance indicator. The LHVs of the fuel samples were estimated using the weighted average method, based on the mass fraction of each component and 67.9 72 71.5 71.4 71.2 70.8 69.7 G A10 A20 A30 A10+B1 A20+B3 A30+B5 10 30 50 70 90 0 20 40 60 80 EN 228 upper limit: 68 kPa +2.7% +4.3%+4.7% +5.2% +5.3% DVPE (kPa) Fuel +6.0% 743.6 749.4 754.4 759.9 751.2 759.2 767.3 G A10 A20 A30 A10+B1 A20+B3 A30+B5 50 150 250 350 450 550 650 750 850 0 100 200 300 400 500 600 700 800 +3.2% +2.1% +1.0% +2.2% +1.5% y ( g ) Fuel EN 228 upper limit: 775 kg/m 3 +0.8%
92 their respective LHVs. Figure 4 illustrates the LHV results for the various fuel blends. Since methanol (50 wt% oxygen), ethanol (34.7 wt% oxygen), and biodiesel (11 wt% oxygen) are oxygenated fuels, they inherently possess lower calorific values compared to gasoline. Consequently, both ternary (dual alcohol–gasoline) and quaternary (dual alcohol–biodiesel–gasoline) blends exhibit reduced heating values, which may lead to lower engine power output unless compensated by higher volumetric fuel delivery or improved combustion efficiency. In engine applications, the fuel injection system meters and delivers fuel based on volume rather than mass. Therefore, the heating value of a fuel should be evaluated on a volumetric basis. The energy densities of the tested fuels G, A10, A20, A30, A10+B1, A20+B3, and A30+B5 were calculated as 32.3, 30.9, 29.6, 28.2, 31.0, 29.7, and 28.5 MJ/L, respectively. Compared to gasoline, the volumetric calorific values of these blends were lower by 4.1%, 8.4%, 12.6%, 4.0%, 8.0%, and 11.7%, respectively. Although methanol, ethanol, and biodiesel have significantly lower heating values than gasoline by approximately 54%, 38%, and 14%, respectively, the overall decrease in the fuel blends’ volumetric energy content was less severe due to the higher density of these oxygenated components, which helps offset their lower energy content per unit volume. Figure 4. Lower heating values of fuel samples The sulfur content of the fuel samples was measured as 1.6, 1.4, 1.2, 1.1, 1.3, 1.3, and 1.4 mg/kg for G, A10, A20, A30, A10+B1, A20+B3, and A30+B5, respectively. These values are well below the maximum limit of 10 mg/kg specified in the EN 228 standard, indicating full compliance. The observed reduction in sulfur content with increasing alcohol content is attributed to methanol and ethanol being inherently sulfur-free. Consequently, binary alcohol-gasoline blends and ternary blends with biodiesel exhibit improved sulfur profiles, contributing to cleaner combustion and reduced SOx emissions. The volatility characteristics of gasoline were further investigated through distillation profiling, which offers more comprehensive information over a wider temperature range than DVPE measurements. Figure 5 shows the distillation curve of fuel samples. Gasoline shows a wide range of distillation curves because it contains various hydrocarbons with different boiling points. Due to the presence of low-boiling-point components such as butanes, pentanes, and light olefins, the temperature at which 5% of the fuel is distilled was 35.7 °C, about 7 °C lower than that of the fuel blends. Between the 10% and 30% distilled fractions, the distillation curves of all fuels were nearly the same, but they diverged in the 40% to 80% distillation range. This divergence results from the low boiling points of methanol and ethanol, which, as pure compounds, boil at constant temperatures. Most of the alcohol content in the blends evaporates within this range. As the remaining alcohol content decreases, the distillation curves converge again beyond the 80% threshold. The decrease in distillation temperature within the 40% to 80% distilled volume range will negatively impact fuel economy and engine performance. Despite its high boiling point of 340–360 °C, the addition of biodiesel did not noticeably alter the distillation temperatures of the ternary blends. This suggests that alcohol is the dominant factor in shaping the distillation curves, thereby reducing the influence of biodiesel at low blending ratios.
93 Figure 5. Distillation curve of fuel samples Table 2 presents the specific distillation data for all fuel samples. In gasoline distillation analysis, specific temperatures are reference points to characterize volatility behavior. These include E70, E100, and E150, representing the percentage of fuel evaporated at 70, 100, and 150 °C, respectively. In addition, IBP and FBP denote the initial and final boiling points, corresponding to the temperatures at which the first and last drops distill. Although the EN 228 regulation does not specify a limit for IBP, it requires the FBP to be equal to or lower than 210 °C. According to Table 2, all fuel samples meet the E150 and FBP limits. However, most ternary and quaternary blends fall outside the E70 and E100 ranges specified by EN 228. This is due to the high volatility of these fuels. While such high volatility can improve cold start, warm-up, and driveability, it may also lead to vapor lock and unstable engine operation. In addition, high volatility can increase fuel consumption and evaporative hydrocarbon emissions [9]. Table 2. Detailed distillation data Unit G A10 A20 A30 A10+B1 A20+B3 A30+B5 EN 228 Limit IBP °C 35.7 32.7 34.4 36.7 36.3 37.7 35.1 - E70 %v/v 43.4 62.1 64.5 64.3 62.6 65.8 64.8 22-50 E100 %v/v 62.7 68.6 76.3 82.8 68.4 76 82.7 46-71 E150 %v/v 88.2 90.1 90.6 91.8 89.9 90.5 91.9 ≥ 75 FBP 190.9 183.9 181 182 190.6 189.4 187.1 ≤ 210 Residue %v/v 1 1 1 1 1 1 1 ≤ 2 4. CONCLUSION The results indicate that all gasoline–methanol–ethanol ternary blends had higher vapour pressure than the standard gasoline. Although adding biodiesel to these blends lowered the vapour pressure, the measured values still surpassed the EN 228 limit. The increase in RVP observed in alcohol–gasoline blends could not be reduced below the EN limit by adding 5% vol. biodiesel, indicating that a higher biodiesel fraction is required. Therefore, its effectiveness in lowering the vapor pressure of lower alcohol–gasoline blends is limited, owing to the strong azeotropic behavior of these mixtures. The other measured fuel characteristics showed slight improvements with biodiesel added to the ternary blends. Although biodiesel has a very low octane number, which could reduce the knocking resistance of the final blend, this adverse effect can be offset by the high octane numbers of methanol and ethanol. Future studies could focus on utilizing higher biodiesel fractions or reformulating methanol–ethanol– gasoline blends with novel additives to achieve better vapor pressure control. If any additive significantly improves, subsequent investigations should include fuel stability assessments, engine performance tests, and fuel– material compatibility evaluations.
94 Acknowledgments The author wishes to sincerely thank Fatih Bilgin, Director of the Fuels Analysis Laboratory at Guzel Enerji A.S., for his technical support and assistance with this study. References [1] E. Vanzela, W. C. Nadaleti, R. A. Bariccatti, P. A. Cremonez, E. de Rossi, P. B. Filho, et al., “Physicochemical properties of ethanol with the addition of biodiesel for use in Otto cycle internal combustion engines: Results and revision,” Renewable and Sustainable Energy Reviews, vol. 17, no. 74, pp. 1181–1188, 2017. doi: 10.1016/j.rser.2017.03.053 [2] U.S. Environmental Protection Agency (EPA), “Renewable fuel standard program (RFS2) regulatory impact analysis. Office of transportation and air quality,” EPA-420-R-10-006 [Online], Feb. 2010. [3] EN. (2025, Aug. 06). European Parliament & Council. Directive - 2018/2001 - EN - EUR-Lex [Online]. Available: https://eur-lex.europa.eu/eli/dir/2018/2001/oj [4] H. Hao, Z. Liu, F. Zhao, J. Ren, S. Chang, K. Rong, et al., “Biofuel for vehicle use in China: Current status, future potential and policy implications,” Renewable and Sustainable Energy Reviews, vol. 82, pp. 645– 653, 2018. doi: 10.1016/J.RSER.2017.09.045 [5] V. F. Andersen, J. E. Anderson, T. J. Wallington, S. A. Mueller, and O. J. Nielsen, “Vapor pressures of alcohol-gasoline blends,” Energy and Fuels, vol. 24, pp. 3647–3654, 2010. doi: 10.1021/ef100254w [6] L. M. Rodríguez-Antón, F. Gutiérrez-Martín, and C. Martinez-Arevalo, “Experimental determination of some physical properties of gasoline, ethanol and ETBE ternary blends,” Fuel, vol. 156, pp. 81–86, 2015. doi: 10.1016/j.fuel.2015.04.040 [7] L. M. Rodríguez-Antón, M. Hernández-Campos, and F. Sanz-Pérez, “Experimental determination of some physical properties of gasoline, ethanol and ETBE blends,” Fuel, vol. 112, pp. 178–184. 2013. doi: 10.1016/j.fuel.2013.04.087 [8] E. Vanzela, W. C. Nadaleti, R. A. Bariccatti, P. A. Cremonez, E. de Rossi, P. B Filho, et al., “Physicochemical properties of ethanol with the addition of biodiesel for use in Otto cycle internal combustion engines: Results and revision,” Renewable and Sustainable Energy Reviews, Vol. 74, pp. 1181– 1188, 2017. doi: 10.1016/j.rser.2017.03.053 [9] N. E. Bulbul and A. Cakmak, “The effectiveness of iso-alcohols in reducing vapor pressure and enhancing fuel properties of ethanol-gasoline mixtures,” International Journal of Automotive Engineering and Technologies, vol. 14, pp. 1–10, 2025. doi: 10.18245/ijaet.1591917 [10] L. M. Rodríguez-Antón, F. Gutiérrez-Martín, and M. Hernández-Campos, “Physical properties of gasolineETBE-isobutanol (in comparison with ethanol) ternary blends and their impact on regulatory compliance,” Energy, vol. 185, pp. 68–76, 2019. doi: 10.1016/J.ENERGY.2019.07.050 [11] T. M. M. Abdellatief, M. A. Ershov, V. M. Kapustin, M. Ali Abdelkareem, M. Kamil, and A. G. Olabi, “Recent trends for introducing promising fuel components to enhance the anti-knock quality of gasoline: A systematic review,” Fuel, vol. 291, art. no. 120112, 2021. doi: 10.1016/J.FUEL.2020.120112 [12] F. Leach, R. Stone, and D Richardson, “The influence of fuel properties on particulate number emissions from a direct injection spark ignition engine,” SAE Technical Paper, 2013. doi: 10.4271/2013-01-1558 [13] N. I. Masuk, K. Mostakim, and S. D. Kanka, “Performance and emission characteristic analysis of a gasoline engine utilizing different types of alternative fuels: A comprehensive review,” Energy and Fuels, vol. 35, pp. 4644–4469, 2021. doi: 10.1021/acs.energyfuels.0c04112
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 95 Shunt Reactor (without Tap Changer) Design, Manufacturing, and Testing Ali Tas1 1R&D Department, Beta Enerji ve Teknoloji A.S., Adana, Türkiye Corresponding author: Ali Tas (e-mail: [email protected]) Abstract This study covers the design, manufacturing, and testing processes of a dual-input FR3 vegetable oil transformer, developed to enhance efficiency, safety, and sustainability of power conversion equipment in solar power plants (SPPs). Unlike conventional single-input transformers, this design allows feeding from two different high-voltage sources, providing operational flexibility and reducing the number of required equipment, thereby lowering operating costs. The transformer is designed according to the International Electrotechnical Commission (IEC) 60076 standards, incorporating low-loss silicon steel cores, high voltage (HV) and low voltage (LV) coil structures with aluminium foil, electrostatic shielding for harmonic suppression, and FR3 vegetable oil offering high dielectric strength, improved thermal performance, and environmental sustainability. Critical manufacturing processes include core stacking, foil and flat conductor winding, insulation design, corrugated tank production, and vacuum oil filling. The prototype underwent routine testing, including direct current (DC) resistance, shortcircuit losses and impedance, no-load losses, voltage ratio, and applied/induced voltage tests to validate the design. The dual-input configuration allows use across two different lines or sites, while FR3 oil reduces carbon footprint, enhances fire safety, and extends equipment lifetime, contributing to sustainable energy infrastructure. Keywords: Dual-input transformer, FR3 vegetable oil, Solar power plant (SPP)
102 Figure 5. Slump test setup The ultrasonic transit time and Schmidt hammer test results are given in Table 4, and the test images are shown in Figure 6. According to the results in Table 4, the estimated compressive strength measured with the Schmidt hammer was 28.15 MPa on average. The estimated strength calculated using the ultrasonic transit velocity is 36.77 MPa. The ultrasonic transit velocity being above 4.0 km/s indicates that the internal structure of the concrete is “very good” and homogeneous. Figure 6. Schmidt hammer test and ultrasonic wave velocity test application According to Table 5, the average unit volume mass of concrete specimens produced with RCA was found to be 2.27 kg/dm3. This value is 5-6% lower than the density of normal concrete produced with NAs. This reduction in weight is directly attributable to the low-density structure of RCA. The compressive strength test was performed on 150x150x150mm cube specimens after 28 days of curing, in accordance with TS EN 12390–3. The compressive strength test results are given in Table 5 and the test graph is shown in Figure 7. The average compressive strength of the sample was determined to be 34.17 MPa. This value is close to the average strength value required for the targeted C30 class and is very close to meeting the minimum requirement (37 MPa) for C30/37 class concrete according to the TS EN 206 standard. Although the LA abrasion value of the aggregate is 3% above the limit, the fact that the compressive strength exceeds the target class C30 strongly supports the use of paving stone waste as aggregate in structural concrete production. The differences between the Schmidt (28.15 MPa) and ultrases (36.77 MPa) estimates and the actual compressive strength (34.17 MPa) indicate that non-destructive methods may require calibration for RCA concrete, but that the Ultrases method provides a closer estimate.
103 Table 4. Schmidt hammer test and ultrasonic wave velocity test results Sample No. Schmidt Hammer Readings Ultrasonic Readings 1 2 3 4 5 6 7 8 9 10 Average (Rc) Strength Estimate (MPa)* Average Strength Estimate (MPa) Duration (μs) Speed - V (km/h) Strength Estimate (MPa)* Average Strength Estimate (MPa) P1 26 28 23 26 22 20 23 26 20 20 24 24.37 28.15 34.6 4.3 36.14 36.77 P2 24 28 28 24 22 20 22 24 24 20 24 24.37 34.1 4.4 39.69 P3 24 21 29 28 22 26 27 26 22 30 25 27.47 34.6 4.3 36.14 P4 29 24 23 28 32 23 23 27 28 24 26 30.81 33.9 4.4 39.69 P5 27 28 28 23 29 22 24 25 21 22 25 27.47 34.5 4.3 36.14 P6 28 24 29 27 28 24 28 24 28 26 27 34.42 35.7 4.2 32.84 *: Compressive strength estimates were calculated using Rc and V values. Table 5. Compressive strength test results Compressive Strength Values (28 days) Sample No. Unit Volume Mass (kg/dm3) Fracture Load (kN) Strength (MPa) Average Strength (MPa) P1 2.26 780.9 34.71 34.17 P2 2.3 756.7 33.63 P3 2.23 760.9 33.82 P4 2.27 791.4 35.17 P5 2.29 748.1 33.25 P6 2.25 774.6 34.43 Figure 7. Pressure resistance graph 4. RESULTS In this study, the usability of recycled aggregates obtained from waste concrete paving stones, which were discarded as a result of QC tests, in concrete production was investigated. Based on the findings, the following conclusions were reached: • It was determined that RCA produced from waste paving stones has a lower density and significantly higher water absorption rates ranging from 2.6% to 8.2% compared to NAs. This situation necessitates the precise adjustment of the mixing water in concrete design according to the DKY condition. • The Los Angeles abrasion loss of RCA was measured as 53%. This value is slightly (3%) above the 50% limit specified by the standards. However, the FI (FI15) and MB (0.5 g/kg) results of the aggregate are excellent.
104 • In the production of the targeted C30 class concrete using RCA, a workability value of 80 mm (S2 consistency) was achieved. After 28 days of curing, the average compressive strength of the samples was measured as 34.17 MPa. • Although the LA abrasion value was slightly above the limit, the compressive strength of the final concrete successfully met the C30 class target. This proves that aggregates obtained from waste paving stones can be an effective alternative to NAs in standard structural concrete applications (up to C30/37 class) that do not require high abrasion resistance. This study demonstrates that paving stone waste, which poses an environmental problem, can be converted into a valuable raw material in concrete production, thereby contributing to both waste management and the conservation of natural resources. Acknowledgments This paper utilises data from the TUBITAK-2209/A 2023/2 term project entitled “Investigation of the Reusability of Concrete Paving Stone Waste as Aggregate”. References [1] A. Koken, M. A. Koroglu, and F. Yonar, “The usability of waste concrete as concrete aggregate,” Selcuk University Technical Sciences Vocational School Technical-Online Journal, vol. 7, no. 1, pp. 86–97, 2008. [2] G. Durmus, O. Can, and O. Simsek, “Determination of the engineering properties of different classes of concrete produced from recycled aggregates," in 5th International Advanced Technologies Symposium (IATS’09), Karabük, Turkey, May 13–15, 2009. [3] T. C. Hansen and H. Narud, “Strength of recycled concrete made from crushed concrete coarse aggregate,” in Concrete International: Design and Construction, 1983, pp. 79–83. [4] B.T. Yuksel, “The use of blast furnace slag in the production of paving stones and kerbs,” in Earthquake Symposium, Kocaeli, 2005, pp. 870–880. [5] A. Nealen and M. Ruhl, “Consistency aspects in the production of concrete using aggregates from recycled demolition material,” in Darmstadt Concrete, 1997. [6] H. Polat, U. E. Yurtcan, and M. N. Kolak, “Investigation of the usability of waste kerbstones as concrete aggregate,” Journal of Nature and Science, vol. 3, no. 2, pp. 37–41, 2014. [7] M. M. Tufekci, “Investigation of the reusability of recycled aggregates in concrete production,” M.Sc. thesis, Institute of Science, Yıldız Technical University, 2011. [8] Tests for geometric properties of aggregates – Part 1: Determination of gradation – Sieve analysis, TS EN 933-1, Turkish Standards Institute, Ankara, 2012. [9] Tests for mechanical and physical properties of aggregates – Part 2: Methods for determining resistance to fragmentation, TS EN 1097-2, Turkish Standards Institute, Ankara, 2020. [10] Tests for geometric properties of aggregates – Part 9: Determination of fine particles – Methylene blue test, TS EN 933-9, Turkish Standards Institute, Ankara, 2022. [11] Tests for geometric properties of aggregates – Part 3: Flatness index, TS EN 933-3, Turkish Standards Institute, Ankara, 2012. [12] Tests for physical and mechanical properties of aggregates – Part 6: Determination of particle density and water absorption, TS EN 1097-6, Turkish Standards Institute, Ankara, 2022. [13] Concrete – Properties, performance, production and conformity, TS EN 206+A2, Turkish Standards Institute, Ankara, 2021. [14] Concrete – Fresh concrete tests – Part 2: Slump test, TS EN 12350-2, Turkish Standards Institute, Ankara, 2019. [15] Determination of compressive strength of concrete in structures and precast concrete components in situ, TS EN 13791, Turkish Standards Institute, Ankara, 2019. [16] Concrete tests in structures – Part 4: Determination of ultrasonic pulse wave velocity, TS EN 12504-4, Turkish Standards Institute, Ankara, 2021. [17] Concrete – Tests on hardened concrete – Part 3: Determination of compressive strength of test specimens, TS EN 12390-3, Turkish Standards Institute, Ankara, 2019. [18] Standard specification for concrete aggregates, ASTM C33-07, ASTM International, West Conshohocken, PA, 2007.
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 105 Spintronics: The Future of Electronics – A Comprehensive Overview of Dilute Magnetic Semiconductors Muhammet Arucu1 1Department of Computer Technologies, Bandirma Onyedi Eylul University, Gonen, 10900 Balikesir, Türkiye Corresponding author: Muhammet Arucu (e-mail: [email protected]) Abstract The unrelenting downscaling of conventional charge-based electronics is rapidly approaching fundamental thermodynamic and physical limits, necessitating a paradigm shift toward alternative state variables. Spin transport electronics (Spintronics)—exploiting the electron's intrinsic spin degree of freedom—emerges as the most promising candidate to overcome these bottlenecks, offering non-volatility, higher processing speeds, and reduced power consumption. This paper presents a comprehensive review of dilute magnetic semiconductors (DMSs), the pivotal material class anticipated to bridge the gap between modern semiconductor logic and magnetic storage. We critically examine the physical mechanisms underlying carrier-mediated ferromagnetism in III-V and II-VI compounds, framed within the theoretical context of the zener model and Ruderman–Kittel–Kasuya–Yosida (RKKY) interactions. Furthermore, the thermodynamic challenges hindering the realization of room-temperature ferromagnetism, including solubility limits and phase separation, are analysed. Finally, we evaluate the transformative potential of DMSs in realizing next-generation devices such as spin field-effect transistors (SpinFETs) and fault-tolerant quantum computing architectures. Keywords: Spintronics, Dilute Magnetic Semiconductors, Ferromagnetism, Spin Polarization, Carrier-Mediated Exchange. 1. INTRODUCTION For more than five decades, improvements in semiconductor technology have relied on aggressive geometric scaling of silicon-based metal-oxide semiconductor field-effect transistors (MOSFETs). As channel lengths shrink below 5 nm and oxide thicknesses reach the atomic limit, power dissipation, quantum tunneling, and short-channel effects have become prohibitive [1]. These limitations have triggered a global search for beyondcomplementary metal-oxide semiconductor (CMOS) technologies that circumvent the fundamental barriers of charge-based electronics [2]. Spin transport electronics (Spintronics) offers a transformative approach by exploiting the spin degree of freedom of electrons, enabling non-volatile, high-speed, and energy-efficient devices [3, 4]. Because electron spin orientation—up or down—can encode binary information without continuous power, spintronic systems promise orders of magnitude reductions in standby power consumption. More importantly, spin-based architectures enable the integration of memory and logic within the same device platform, overcoming the von Neumann bottleneck that currently dominates power and area consumption in modern processors [5]. Spintronics offers a fundamental solution to these challenges. While classical electronics utilizes only the charge of the electron (-e), spintronics exploits the electron's intrinsic angular momentum, or “spin” (↑ up or ↓ down). Spin provides a non-volatile state that can retain data without continuous energy consumption. The figure 1 illustrates the quantum mechanical principle of superposition in electron spin states, showing a spin-up electron (red) along a vertical magnetic field, a spin-down electron (blue) in the opposite direction, and their linear superposition resulting in a coherent state with spin oriented perpendicular to the field. Dilute magnetic semiconductors (DMSs) stand at the intersection of semiconductor physics and magnetism. By doping conventional non-magnetic semiconductors (e.g., GaAs, InAs, ZnO, GaN) with a small fraction of transition-metal ions, one may obtain materials that simultaneously exhibit semiconducting carrier transport and magnetic ordering. This unique combination makes them ideal candidates for spin injection, spin manipulation, and electrically controlled magnetism [6]. This paper provides a fully revised and coherent overview of the physical principles, material systems, fabrication techniques, challenges, and applications associated with DMSs.
106 Figure 1. The diagram shows electron spin up, spin down, and their superposition state 2. PHYSICAL FOUNDATIONS OF DMSs DMSs are fabricated by incorporating a small fraction (typically 1–10%) of paramagnetic transition-metal ions— most commonly Mn, but also Fe, Co, Cr, or V—into the lattice of a conventional non-magnetic semiconductor host (e.g., GaAs, InAs, ZnO, GaN, or Si) [7, 8]. The emergence of collective magnetism in these otherwise dilute systems arises from the indirect exchange interaction between the localized dor f-electron spins of the magnetic dopants and the delocalized charge carriers (holes or electrons) in the host bands, an interaction that can mediate long-range ferromagnetic, antiferromagnetic, or spin-glass order depending on carrier type, density, and dopant separation (see Table 1 for a comparison of key DMS material systems) [9, 10]. Table 1: Comparison of prominent DMS material systems Material System Host Dopant Curie Temp (TC) Current Status and Challenges III-V Group GaAs Mn ~180-200 K Well-understood model system. Limited by subroom temperature TC. III-V Group InAs Mn ~60-90 K Low T C , but exhibits high electron mobility. II-VI Group ZnSe Mn < 50 K Excellent optical properties; difficult to achieve p-type doping. Oxides ZnO Co, Mn > 300 K (Debated) Reports of high T C , but magnetism origin is controversial (extrinsic phases vs. intrinsic). Nitrides GaN Mn > 300 K (Theoretical) High potential for room temperature applications; challenges in phase separation. 2.1. Crystal Structure and Magnetic Doping In archetypal systems such as (Ga,Mn)As, manganese atoms substitute gallium atoms in the zinc-blende lattice, generating localized magnetic moments (S = 5/2) and introducing holes into the valence band. These holes play a critical role in mediating long-range magnetic coupling among dopants. However, solubility limits and defect formation complicate this process, often requiring non-equilibrium growth conditions such as low-temperature molecular beam epitaxy. 2.2. The Origin of Ferromagnetism / Carrier-Mediated Ferromagnetism Ferromagnetism in DMS is generally described by Carrier-Mediated Ferromagnetism. Since the magnetic ions are dilute, they are too far apart to interact directly. Consequently, the magnetic interaction is mediated by the free
107 charge carriers (typically holes) in the system. As holes delocalize and move from one Mn ion to another, they transmit spin information, aligning the Mn spins in a parallel configuration (ferromagnetic order). Theoretically, the observed spin-related phenomena in magnetic materials, particularly in DMSs and certain metallic systems, are often described within the framework of the Zener carrier-mediated ferromagnetism model [11], originally proposed by Clarence Zener, which posits that the alignment of localized magnetic moments arises from the indirect exchange interaction between localized dor f-shell electrons (typically from transition metal or rare-earth impurities) and the delocalized conduction electrons belonging to the sand p-bands; this indirect coupling is more rigorously formulated through the Ruderman–Kittel–Kasuya–Yosida (RKKY) interaction [12], a second-order perturbative mechanism in which itinerant carriers become spin-polarized by one localized moment and subsequently transmit this polarization to another distant localized moment via Friedel-like oscillations of the spin density in the conduction electron sea, thereby mediating long-range oscillatory magnetic ordering that can be either ferromagnetic or antiferromagnetic depending on the inter-moment distance and the Fermi wavevector of the host material [13]. 3. PROMINENT MATERIAL SYSTEMS AND FABRICATION TECHNIQUES Different material families exhibit varying degrees of magnetic ordering, dopant solubility, and carrier control. Among these, III–V semiconductors have been the most extensively studied, while oxideand nitride-based systems offer the potential for room-temperature ferromagnetism but remain subject to debate regarding the intrinsic origin of their magnetic properties. 3.1. III-V Group: GaMnAs and InMnAs The III–V compound family, particularly (Ga,Mn)As, remains the most thoroughly investigated and theoretically best-understood class of DMSs, serving as the benchmark system for carrier-mediated ferromagnetism in semiconductors. The equilibrium solubility limit of Mn in GaAs is below 0.01%, rendering conventional hightemperature growth inapplicable. Ferromagnetic (Ga,Mn)As is therefore synthesized exclusively by lowtemperature molecular beam epitaxy (LT-MBE) at substrate temperatures of 180–300 °C, far below the regime where surface segregation or precipitation of secondary phases (e.g., MnAs) occurs [14]. This non-equilibrium approach enables metastable Mn concentrations up to ~10–12%, well into the metallic regime required for holemediated exchange [15]. The DMS crystal structure shown in Figure 2 is a “Zinc-Blende” (Zinc-Sulfide) lattice structure. The blue spheres represent Gallium (Ga) atoms (the main structure), the red spheres represent Arsenic (As) atoms, and the green spheres represent Manganese (Mn) atoms randomly placed in place of the blue spheres (substitutional doping). Figure 2: Mn atoms substitute Ga atoms, providing both local magnetic moments and holes to the system Despite intensive optimization of growth conditions, post-growth annealing, and co-doping strategies, the highest experimentally verified Curie temperature in (Ga,Mn)As thin films and heterostructures remains approximately 200 K. The persistent gap to room-temperature ferromagnetism continues to restrict practical device applications of this material system [16]. 3.2. Oxides and Nitrides: ZnOand GaN-Based DMSs Following seminal mean-field Zener-model predictions by Dietl and co-workers in 2000, transition-metal-doped wide-bandgap semiconductors such as ZnO (Eg ≈ 3.37 eV) and GaN (Eg ≈ 3.4 eV) were proposed as prime
108 candidates for achieving robust ferromagnetism well above room temperature, owing to their large hole effective masses and strong p–d hybridization [11]. These materials are typically prepared by pulsed laser deposition (PLD), metal-organic chemical vapor deposition (MOCVD), or ion implantation followed by annealing. However, reproducible room-temperature ferromagnetism has proven elusive; reported high-TC values are frequently attributable to nanoscale secondary phases, clusters, or defect-induced magnetism rather than intrinsic carrier-mediated ordering, rendering the physical origin in most oxide and nitride DMSs highly controversial. While isolated reports claim TC exceeding 400 K, no consensus exists on intrinsic, defect-free room-temperature ferromagnetism in these systems, and their integration into functional spintronic devices remains limited [8, 17]. Table 2: Comparison of conventional and spintronic technologies Feature Conventional Electronics (Charge-Based) Spintronics (Spin-Based) Fundamental Particle Electron (Charge: -e) Electron (Charge: -e + Spin: ↑↓) State Variable Charge quantity Spin orientation (Up/Down) Data Volatility Volatile (Data loss upon power cut) Non-Volatile (Data retention without power) Energy Consumption High (Due to leakage currents) Low (Quantum state switching) Switching Speed Limited by capacitance (RC delay) Fast (governed by spin relaxation time) Material Basis Semiconductors (Si, Ge) Magnetic Semiconductors, Ferromagnets 4. KEY CHALLENGES IN DMS RESEARCH The commercialization of DMS technology remains stymied by the persistently low Curie temperature (TC), which marks the transition from ferromagnetic to paramagnetic order and must exceed 350–400 K (77–127 °C) for reliable operation in ambient environments encountered in consumer electronics, automotive systems, and data centers [3, 4]. Achieving room-temperature or above ferromagnetism in DMSs is essential to prevent thermal randomization of spin alignments, which would otherwise degrade device performance through loss of nonvolatility, reduced magnetoresistance ratios, and impaired spin injection efficiency [18]. Figure 3. The plot shows that the Curie temperature increases with Mn concentration up to an optimal doping range, but decreases beyond the solubility limit due to secondary phase formation Figure 3 illustrates the characteristic dependence of the Curie temperature on Mn concentration in DMSs such as (Ga,Mn)As. As the Mn content increases, carrier-mediated ferromagnetism strengthens and TC rises sharply, reaching a maximum within the optimal doping window. Beyond this region, however, thermodynamic solubility limits are exceeded, leading to the formation of precipitates and compensating defects that suppress long-range magnetic order. The trend underscores the fundamental challenge of achieving room-temperature ferromagnetism through conventional doping strategies. The figure illustrates the experimentally observed dependence of the Curie
109 temperature on Mn concentration in (Ga,Mn)As, showing an initial increase up to an optimal doping range (4– 8%) followed by a decline beyond the thermodynamic solubility limit where secondary phase formation suppresses ferromagnetism. 4.1. Thermodynamic Limits and Solubility High concentrations of magnetic dopants are desirable for strong ferromagnetism, yet they violate thermodynamic solubility limits. Excess dopants precipitate as secondary magnetic phases (e.g., MnAs) that degrade device performance. Non-equilibrium growth partially mitigates this issue but introduces defect complexities that must be carefully managed. However, thermodynamic constraints impose a solubility ceiling: in prototypical systems like (Ga,Mn)As, Mn concentrations beyond ~8–10% trigger phase instability during growth, manifesting as the nucleation of metallic MnAs precipitates or other secondary phases (e.g., Mn clusters or arsenide) that disrupt lattice coherence, introduce spin-glass-like disorder, and quench carrier-mediated exchange. These limits arise from the mismatch in valence, ionic radius, and formation enthalpy between the dopant (e.g., Mn2+) and host cation (e.g., Ga3+), exacerbated at elevated growth temperatures; non-equilibrium techniques such as LT-MBE partially circumvent this, but fundamental Gibbs free energy barriers preclude arbitrarily high doping without structural degradation, as corroborated by density functional theory (DFT) calculations and in situ reflection high-energy electron diffraction (RHEED) monitoring [19, 20]. 4.2. Defect Chemistry and Self-Compensation A major intrinsic limitation in III–V DMSs, especially in (Ga,Mn)As, arises from the tendency of a fraction of the incorporated Mn atoms to occupy interstitial rather than substitutional sites (Mn_I instead of Mn_Ga). Interstitial Mn acts as a double donor, releasing electrons that compensate (neutralize) the holes generated by substitutional Mn acceptors. This self-compensation strongly reduces the effective hole concentration essential for carriermediated ferromagnetism and consequently lowers the Curie temperature significantly. In addition, close Mn_Ga– Mn_I pairs exhibit antiferromagnetic coupling, which further decreases the average magnetic moment contributed by each Mn atom. To mitigate these detrimental effects, controlled low-temperature post-growth annealing is commonly employed; this process allows mobile interstitial Mn to diffuse out toward the surface or to form less harmful complexes, partially restoring the hole density and enhancing TC. Careful optimization of the annealing conditions is crucial, as excessive thermal treatment can introduce other compensating defects such as arsenic antisite, highlighting the delicate defect-engineering balance required in these metastable materials [21, 22]. Increasing the magnetic doping concentration generally raises TC; however, beyond a certain threshold, the crystal lattice degrades, leading to the formation of secondary phases (precipitates). Particularly in GaMnAs structures, a fraction of Mn atoms occupies interstitial sites rather than substitutional sites. These interstitial Mn atoms act as double donors, compensating the holes necessary for ferromagnetism and reducing the magnetic moment. Postgrowth annealing processes are standardly used to mitigate this issue by diffusing interstitial Mn to the surface. 5. APPLICATIONS AND FUTURE DIRECTIONS DMSs serve as a foundational materials platform for a wide range of emerging spintronic devices in which electrical control of magnetism, non-volatility, and low switching energies are essential. Their ability to combine magnetic ordering with conventional semiconductor properties opens a pathway toward hybrid charge–spin architectures that surpass the limitations of purely charge-based electronics. Although practical implementations remain at varying levels of technological maturity, DMS systems provide a coherent route to integrating spin functionality into semiconductor environments, enabling innovative device concepts in logic, memory, and quantum information processing. 5.1. Spin Field-Effect Transistors (Spin-FETs) The Datta–Das Spin-FET remains one of the most widely discussed paradigms in spin-based logic, relying on the controlled precession of spin-polarized carriers during their propagation through a semiconductor channel. Efficient operation of the device requires three key capabilities: the injection of highly spin-polarized carriers, the coherent manipulation of their spin states during transport, and the selective detection of the final spin orientation. DMS materials play a critical role in enabling these functionalities because they can provide ferromagnetic contacts that are lattice-matched to III–V semiconductor heterostructures and capable of generating robust spin polarization. Their carrier-mediated magnetic properties allow spin injection and spin filtering to occur through mechanisms that are inherently compatible with semiconductor band structures.
110 Within the channel region, the Rashba spin–orbit interaction generates a momentum-dependent effective magnetic field that rotates the injected spins by an angle controlled through the gate voltage. The presence of DMS layers, either as spin injectors or as gate-modulated magnetic components, enhances the tunability of this precession because their magnetic characteristics can be adjusted through doping, strain, or electric fields. At the drain contact, the device converts the accumulated spin phase into an electrical signal by exploiting the magnetization-dependent conductance. Through this mechanism, the Spin-FET circumvents conventional charge-switching limitations and offers a route to ultra–low-power logic devices. Despite the challenges in achieving room-temperature operation with long spin coherence, DMS-based structures continue to provide an attractive materials platform for nextgeneration spin logic technologies. 5.2. Magnetic Random Access Memory (MRAM) Magnetic random access memory has emerged as the leading non-volatile memory technology for replacing embedded flash and static random access memory (SRAM), owing to its high endurance, fast switching, and excellent data retention characteristics. Conventional MRAM structures rely on metallic ferromagnets and magnetic tunnel junctions, but DMS materials introduce new opportunities for electrically controlled magnetism within a fully semiconductor-based memory element. Because the magnetic ordering in DMS is mediated by carrier concentration, the magnetization can be modulated not only by conventional spin-transfer or spin–orbit torques but also by electric fields that alter the carrier density. This capability offers a fundamentally different approach to memory switching, enabling voltage-controlled magnetic anisotropy and magnetization reversal with considerably lower energy consumption than current-driven mechanisms. Moreover, the crystallographic and electronic compatibility of DMS with III–V semiconductor platforms allows MRAM structures to be integrated directly into existing semiconductor processes. This monolithic integration provides a promising pathway toward logic-in-memory architectures that mitigate the latency and power bottlenecks associated with modern processor design. DMS-based magnetic tunnel junctions (MTJs) may also support enhanced functional capabilities such as magneto-optical readout, multilevel storage, and tunable spin polarization. Although significant material challenges remain, particularly regarding Curie temperature and defect engineering, DMS materials constitute a promising avenue for next-generation MRAM technologies that combine high performance, low power, and seamless semiconductor compatibility. 5.3. Quantum Information Processing Electron spins in semiconductors represent an appealing platform for quantum information processing due to their long coherence times, compatibility with lithographic patterning, and ability to interface with photonic and superconducting systems. DMS materials provide additional advantages by offering tunable magnetic exchange interactions arising from the controlled incorporation of transition-metal dopants. This tunability enables precise engineering of spin splitting, g-factors, and qubit–qubit coupling strengths, which are essential for realizing scalable quantum gate operations. In many DMS systems, strong magneto-optical effects allow spins to be initialized, manipulated, and read out using optical techniques, giving rise to hybrid architectures that combine spin-based qubits with photonic quantum communication channels. Furthermore, DMS layers can be incorporated into quantum dots, nanowires, and heterostructures, where quantum confinement enhances spin coherence and enables electrically tunable quantum states. Their ability to host both localized magnetic moments and delocalized carriers provides a unique environment for implementing exchangebased two-qubit gates or dipolar coupling networks. The capacity to integrate these functionalities within semiconductor manufacturing frameworks offers a realistic pathway toward constructing large-scale quantum circuits. Although challenges related to magnetic noise, dopant homogeneity, and thermal stability must still be resolved, DMS systems hold considerable promise for hybrid quantum technologies that bridge the gap between condensed matter physics, spintronics, and quantum information science. 6. CONCLUSION DMS constitute a critical technological bridge, aiming to integrate the non-volatile memory capabilities of magnetism with the logic processing power of semiconductors within a single material system. While prototypical systems like (Ga,Mn)As have successfully validated the fundamental principles of spintronics, the realization of stable, homogeneous ferromagnetism at and above room temperature remains a formidable materials science challenge. Current limitations, primarily driven by thermodynamic solubility and self-compensation mechanisms, necessitate novel approaches in non-equilibrium growth and defect engineering. Future research must increasingly focus on low-dimensional structures, such as nanowires and quantum dots, where spin coherence can be more
111 effectively preserved. Ultimately, surmounting these physical barriers through advanced nanofabrication and deeper theoretical modelling will be the key to unlocking the full potential of the spintronic revolution. References [1] A. Chaudhry and M. J. Kumar, “Controlling short-channel effects in deep-submicron SOI MOSFETs for improved reliability: A review,” IEEE Trans. Device Mater. Rel., vol. 4, no. 1, pp. 99–109, Mar. 2004. [2] V. K. Khanna, “Short-channel effects in MOSFETs,” in Integrated Nanoelectronics: Nanoscale CMOS, Post-CMOS and Allied Nanotechnologies, New Delhi, India, 2016, pp. 73–93. [3] S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J. M. Daughton, S. Von Molnár, M. L. Roukes, et al., “Spintronics: A spin-based electronics vision for the future,” Science, vol. 294, no. 5546, pp. 1488–1495, Nov. 2001. [4] D. D. Awschalom and M. E. Flatté, “Challenges for semiconductor spintronics,” Nature Phys., vol. 3, no. 3, pp. 153–159, Mar. 2007. [5] Z. Guo, J. Yin, Y. Nai, D. Zhu, K. Shi, and G. Wang, “Spintronics for energy-efficient computing: An overview and outlook,” Proc. IEEE, vol. 109, no. 8, pp. 1398–1417, Aug. 2021. [6] Y. Ohno, D. K. Young, B. Beschoten, F. Matsukura, H. Ohno, and D. D. Awschalom, “Electrical spin injection in a ferromagnetic semiconductor heterostructure,” Nature, vol. 402, no. 6763, pp. 790–792, Dec. 1999. [7] J. K. Furdyna, “Diluted magnetic semiconductors,” J. Appl. Phys., vol. 64, no. 4, pp. R29–R64, Aug. 1988. [8] T. Dietl, “A ten-year perspective on dilute magnetic semiconductors and oxides,” Nature Mater., vol. 9, no. 12, pp. 965–974, Dec. 2010. [9] T. Dietl and H. Ohno, “Dilute ferromagnetic semiconductors: Physics and spintronic structures,” Rev. Mod. Phys., vol. 86, no. 1, pp. 187–251, Mar. 2014. [10] P. Parida, J. Patra, V. R. Singh, and V. K. Verma, “Dilute magnetic semiconductors and its applications— An overview,” in Proc. Int. Conf. Adv. Nanomater. Appl., Singapore, 2024, pp. 181–220. [11] T. Dietl, H. Ohno, F. Matsukura, J. Cibert, and D. Ferrand, “Zener model description of ferromagnetism in zinc-blende magnetic semiconductors,” Science, vol. 287, no. 5455, pp. 1019–1022, Feb. 2000. [12] R. Skomski, J. Zhou, J. Zhang, and D. J. Sellmyer, “Indirect exchange in dilute magnetic semiconductors,” J. Appl. Phys., vol. 99, no. 8, art. no. 08D504, Apr. 2006. [13] B. Lee, T. Jungwirth, and A. H. MacDonald, “Theory of ferromagnetism in diluted magnetic semiconductor quantum wells,” Phys. Rev. B, vol. 61, no. 23, pp. 15606–15609, Jun. 2000. [14] A. Shen, F. Matsukura, S.P. Guo, Y. Sugawara, H. Ohno, M. Tani, et al., “Low-temperature molecular beam epitaxial growth of GaAs and (Ga,Mn)As,” J. Cryst. Growth, vol. 201, pp. 679–683, May 1999. [15] K. Takamura, F. Matsukura, Y. Ohno, and H. Ohno, “Growth and properties of (Ga,Mn)As films with high Mn concentration,” J. Appl. Phys., vol. 89, no. 11, pp. 7024–7026, Jun. 2001. [16] T. Jungwirth, J. Sinova, J. Mašek, J. Kučera, and A. H. MacDonald, “Theory of ferromagnetic (III,Mn)V semiconductors,” Rev. Mod. Phys., vol. 78, no. 3, pp. 809–864, Aug. 2006. [17] K. Ueda, H. Tabata, and T. Kawai, “Magnetic properties of cobalt-doped TiO2 rutile thin films grown by PLD,” Appl. Phys. Lett., vol. 79, no. 7, pp. 988–990, Aug. 2001. [18] I. Žutić, J. Fabian, and S. Das Sarma, “Spintronics: Fundamentals and applications,” Rev. Mod. Phys., vol. 76, no. 2, pp. 323–410, Apr. 2004. [19] S. Sanvito, “First-principles materials design for spintronics,” Nature Mater., vol. 6, no. 11, pp. 803–811, Nov. 2007. [20] K. S. Burch, D. B. Shrekenhamer, E. J. Singley, J. Stephens, B. L. Sheu, R. K. Kawakami, et al., “Magnetooptical signatures of MnAs cluster formation in (Ga,Mn)As,” Phys. Rev. Lett., vol. 97, no. 8, art. no. 087208, Aug. 2006. [21] D. Chiba, A. Werpachowska, M. Endo, Y. Nishitani, F. Matsukura, T. Dietl, et al., “Anomalous hall effect in field-effect structures of (Ga,Mn)As,” Phys. Rev. Lett., vol. 104, no. 10, art. no. 106601, Mar. 2010. [22] K. W. Edmonds, P. Bogusławski, K. Y. Wang, R. P. Campion, S. N. Novikov, N. R. S. Farley, et al., “Mn interstitial diffusion in (Ga,Mn)As,” Phys. Rev. Lett., vol. 92, no. 3, art. no. 037201, Jan. 2004.
214 covering major engineering disciplines: Computer, civil, mechanical, electrical, chemical, food, and textile engineering. For each article, the classifier predicted the top two most relevant categories, which were stored as predicted_engineering_field_1 and predicted_engineering_field_2. To optimize performance on large datasets, the classification step was parallelized using Python’s ThreadPoolExecutor. Figure 2. Algorithm 2: Keyword-based category labeling This labeled dataset was saved in CSV format and served as foundational input for the subsequent graph construction and GNN classification stages. 3.2. Node2Vec Node2Vec [17] is a semi-supervised algorithm designed for learning low-dimensional vector representations (node embeddings) of nodes in a graph. The algorithm optimizes a graph-based objective function using stochastic gradient descent (SGD), aiming to preserve the neighborhood structure of nodes in the embedding space. These node embeddings serve as compact numerical representations that capture both local and global structural features of the graph. To generate training samples for embedding learning, Node2Vec employs a second-order biased random walk strategy. This allows for flexible and scalable exploration of node neighborhoods. 3.2.1. Random Walk In a random walk, a node (e.g., v) transitions to another node (e.g., x) by considering the edge between them ( ) ,.vx E∈ The decision is made using the following rule: ( ) ( ) ( ) ( ) 11 , , if 0 return 1, if 1 neighbor , if 2 distant node pq tx tx tx α tx d d d pq = = = = This α value is multiplied by the unnormalized transition probability weight πvx to determine the likelihood of transitioning to node x in the next step [17]. Return Parameter (p) Controls the Likelihood of Revisiting the Previous Node in the Walk: • A low value (e.g., p = 0.25) increases the probability of returning to the previous node, encouraging local exploration. • A high value (e.g., p = 2) decreases the chance of revisiting, promoting outward exploration. In-Out Parameter (q) Influences the Breadth of the Walk: • q > 1: DFS-like behavior — more likely to explore further nodes. • q < 1: BFS-like behavior — more likely to remain near the current node. • q = 1: Unbiased behavior — balanced between local and global traversal.
215 In our implementation, we set p = q = 1, which corresponds to an unbiased walk. This means the walker has equal probability of returning, staying near, or moving further from the current position. As a result, the random walks balance between BFS and DFS behaviors, enabling the model to capture diverse aspects of node similarity and structure in the learned embeddings. 3.2.2. Node2Vec Parameters The following hyperparameters were used during the embedding generation: • Embedding Dimension: 64 • Walk Length: 30 • Number of Walks Per Node: 200 • Window Size: 10 • Minimum Count: 1 • Batch Size: 4 • p = 1, q = 1 (Unbiased walk) 3.2.3. Embedding Visualization with t-SNE To qualitatively assess the separation of node embeddings, we applied t-SNE to reduce the 64-dimensional embeddings to 2D. Each node was then plotted and colored based on its label. The visualizations allow us to observe cluster structures and category separation in the embedding space. Figure 3. t-SNE visualization of Node2Vec embeddings with 7 labels As shown in Figure 3, the learned Node2Vec embeddings exhibit meaningful cluster structures in settings. In the 7-label visualization demonstrates subtle and overlapping class boundaries, especially for semantically similar categories. These results suggest that the Node2Vec embeddings capture latent structural patterns that align with category information, with clearer separability when fewer classes are involved. 3.2.4. Similarity Search To demonstrate the semantic consistency of the learned embeddings, we performed a similarity search based on Node2Vec vectors. A random node (representing a paper) was selected from the graph, and the top-5 most similar nodes were retrieved using cosine similarity via the most_similar() function. Each result includes the paper’s identifier, category, and title, along with a similarity score. This experiment illustrates how the Node2Vec embeddings capture structural and semantic similarities between academic articles, enabling applications such as recommendation and clustering.
216 As an example, the model was queried with “Population-based local search algorithms for cross-domain search”. The five most similar articles based on Node2Vec embeddings are listed in Table 1. Table 1. The query article (article 10) is shown in the first row. Top 5 most similar articles are listed below based on Node2Vec embeddings Article ID Tittle Similarity 10 Population-based local search algorithms for cross-domain search Query 588 Detection and analysis of driver fatigue stages with electroencephalographic signals 0.95 473 Extruder line selection with fuzzy CRITIC and fuzzy MAIRCA for a cable company 0.95 33 System of automatic scientific article summarization in Turkish 0.95 426 Fitzhugh-Nagumo neuron model and hardware verification 0.95 377 Seismic isolation parameters optimization via crow search 0.95 All the retrieved articles belong to the same category (computer engineering), indicating that the learned embeddings effectively capture domain-specific similarity. 3.3. Node Classification Node classification (NC) aims to infer class labels of unlabeled nodes in a graph using a subset of labeled nodes. In this study, we formulate NC as a supervised learning task on a graph of academic articles, where nodes represent papers and labels denote engineering subfields (e.g., computer, chemical engineering). Node2Vec embeddings are employed to encode structural node features. The graph is constructed and processed using the PyTorch Geometric framework. Labeled nodes are split into training, validation, and test sets via stratified sampling. Three GNN models—GCN, GAT, and APPNP—are trained under identical settings with early stopping. Class imbalance is addressed through weighted cross-entropy loss. Model performance is compared using consistent graph and feature inputs and evaluated via test set metrics and learning curves. 3.3.1. GCN GCN is a spectral-based graph neural network architecture that performs convolution operations directly on graphstructured data~\cite{b1}. It aggregates information from neighboring nodes by applying a shared linear transformation followed by a non-linear activation function. In our implementation, a 3-layer GCN model is used, where each layer updates node representations by combining features from local neighborhoods. Batch normalization and dropout are applied between layers to improve generalization and prevent overfitting. The model is trained using a weighted cross-entropy loss to handle class imbalance. 3.3.2. GAT For the node classification task, we utilized the GAT [12], which introduces attention-based aggregation to assign varying importance to neighboring nodes. This allows the model to focus on more informative local structures during feature propagation. In our implementation on the academic article graph, node features obtained from metadata were processed using a two-layer GAT architecture. The first layer employed 8 parallel attention heads with ReLU activation and dropout, while the second layer used a single head to generate final class logits. Training was performed using cross-entropy loss on labeled nodes, and the entire model was built with PyTorch Geometric. By enabling adaptive weighting of neighbors, GAT provided more expressive node representations compared to uniform aggregation methods like GCN. 3.3.3. APPNP For the node classification task, we employed the APPNP [15], which addresses the over-smoothing issue in deep GCNs by decoupling feature transformation and information propagation. In our implementation on the academic article graph, node features derived from article metadata were first processed through a two-layer multi-layer perceptron (MLP), followed by batch normalization, ReLU activation, and dropout regularization. The transformed features were then propagated using APPNP’s personalized PageRank-based iterative scheme,
217 allowing effective information aggregation from semantically related articles. The model was implemented using PyTorch Geometric and achieved better performance compared to baseline models such as MLP and GCN. 4. USER SCENARIO To facilitate user interaction with the system and make the functionality of the classification and similarity components more accessible, an interactive web-based user interface was developed using the Gradio library in Python. The interface is organized into three main tabs: ● Find Similar Papers: Users can input a paper’s title or abstract to retrieve the top five most similar articles in the dataset. ● Predict Category: This section allows users to obtain the predicted engineering subfield of a given article based on its textual content. Additionally, a bar chart displays the confidence scores for all relevant categories. ● Model Performance: It displays a summary of performance metrics for each GNN model used in the node classification task, along with the model identified as the best performing. This modular interface allows for a streamlined and intuitive exploration of both the dataset and the models’ predictions, supporting both quantitative evaluation and qualitative user analysis. 5. RESULTS To assess model performance, we employed standard classification metrics: Accuracy, precision, recall, and F1score. We evaluated three GNN models—GCN, GAT, and APPNP—on the node classification task. Table 2 summarizes their overall performance. APPNP achieved perfect scores across all metrics, indicating its effectiveness in capturing class boundaries. GCN followed closely with strong and balanced results. GAT, while showing relatively high precision, suffered from lower recall, suggesting more conservative predictions. Overall, APPNP emerged as the best-performing model. Table 2. Node classification performance of different GNN models Model Accuracy Precision Recall F1-Score Accuracy GCN 0.973 0.850 0.863 0.855 0.977 GAT 0.806 0.804 0.701 0.731 0.923 APPNP 1.000 1.000 1.000 1.000 1.000 6. CONCLUSION This study introduces a GNN-based approach for classifying Turkish academic papers by constructing a custom graph where nodes represent articles and features are derived using Node2Vec. Unlike benchmark datasets, our domain-specific corpus enables topic-focused classification within the engineering field. We compared three GNN models: GCN, GAT and APPNP, under consistent settings, and observed promising classification performance. Additionally, we developed an interactive Gradio interface for real-time category prediction and paper similarity search, making the system accessible for academic use. Future work includes expanding the dataset to other fields, exploring contrastive learning, and integrating additional graph-based tasks such as link prediction. References [1] T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” arXiv, 2017. doi: 10.48550/arXiv.1609.02907 [2] H. Liu, H. Kou, C. Yan, and L. Qi, “Keywords-driven and popularity-aware paper recommendation based on undirected paper citation graph,” Complexity, vol. 2020, art. no. 2085638, 2020. doi: 10.1155/2020/2085 638. [3] C. Wu, F. Wu, M. An, Y. Huang, and X. Xie, “Neural news recommendation with topic-aware news representation,” in Proc. 57th Annu. Meeting Assoc. Comput. Linguistics (ACL), Florence, Italy, Jul. 2019, pp. 1154–1159. [4] G. Guo, B. Chen, X. Zhang, Z. Liu, Z. Dong, and X. He, “Leveraging title-abstract attentive semantics for paper recommendation,” in Proc. AAAI, New York, USA, 2020, pp. 67–74. [5] K. Haruna, M. A. Ismail, A. Qazi, H. A. A. Kakudi, M. Hassan, S. Abdullahi Muaz, et al. “Research paper recommender system based on public contextual metadata,” Scientometrics, vol. 125, no. 1, pp. 101–114,
218 Oct. 2020. doi: 10.1007/s11192-020-03642-y [6] Y. Chen, C. Gao, X. Ren, Y. Peng, X. Xia, and M. R. Lyu, “API usage recommendation via multi-view heterogeneous graph representation learning,” IEEE Trans. Softw. Eng., vol. 49, no. 5, pp. 3289–3304, 2023. doi:10.1109/TSE.2023.3252259 [7] C. Gao, Y. Zheng, N. Li, Y. Li, Y. Qin, J. Piao, et al., “A survey of graph neural networks for recommender systems: Challenges, methods, and directions,” ACM Trans. Recomm. Syst., vol. 1, no. 1, art. no. 3, pp. 1– 51, Mar. 2023. doi: 10.1145/3568022 [8] X. Wang, X. He, M. Wang, F. Feng, and T.-S. Chua, “A survey on knowledge graph-based recommender systems,” ACM Trans. Intell. Syst. Technol., vol. 10, no. 2, art. no. 1, pp. 1–37, Jan. 2019. doi: 10.1145/3235 029. [9] W. Huang, T. Zhang, Y. Rong, and J. Huang, “Graph learning: A survey,” IEEE Trans. Artif. Intell., vol. 2, no. 2, pp. 109–127, Apr. 2021. doi: 10.1109/TAI.2021.3076021 [10] L. Wu, Y. Chen, K. Shen, X. Guo, H. Gao, S. Li, et al., “Graph neural networks for natural language processing: A survey,” arXiv, 2021. doi: 10.48550/arXiv.2106.06090 [11] W. L. Hamilton, R. Ying, and J. Leskovec, “Inductive representation learning on large graphs,” in Proc. 31st Int. Conf. Neural Inf. Process. Syst. (NeurIPS), Long Beach, CA, USA, 2017, pp. 1025–1035. [12] P. Veličković, G. Cucurull, A. Casanova, A. Romero, P. Liò, and Y. Bengio, “Graph attention networks,” arXiv, 2017. doi: 10.48550/arXiv.1710.10903 [13] Q. Li, Z. Han, and X. Wu, “Deeper insights into graph convolutional networks for semi-supervised learning,” in Proc. AAAI Conf. Artif. Intell., vol. 32, no. 1, pp. 3538–3545, 2018. [14] K. Xu, W. Hu, J. Leskovec, and S. Jegelka, “How powerful are graph neural networks?,” arXiv, 2018. doi: 10.48550/arXiv.1810.00826 [15] J. Klicpera, A. Bojchevski, and S. Gunnemann, “Predict then propagate: Graph neural networks meet personalized PageRank,” in Proc. Int. Conf. Learn. Represent. (ICLR), New Orleans, LA, USA, 2019. [16] P. Sen, G. Namata, M. Bilgic, and L. Getoor, “Collective classification in network data,” AI Magazine, vol. 29, no. 3, pp. 93–106, 2008. [17] A. Grover and J. Leskovec, “node2vec: Scalable feature learning for networks,” in Proc. ACM SIGKDD Int. Conf. Knowl. Discov. Data Min. (KDD), San Francisco, CA, USA, 2016, pp. 855–864. doi: 10.1145/293 9672.2939754 [18] W. Hu, M. Fey, M. Zitnik, Y. Dong, H. Ren, B. Liu, M. Catasta, and J. Leskovec, “Open graph benchmark: Datasets for machine learning on graphs,” in Proc. NeurIPS, vol. 33, pp. 22118–22133, 2020. [19] C. B. Clement, M. Bierbaum, K. P. O’Keeffe, and A. A. Alemi, “On the use of arXiv as a dataset,” arXiv, 2019. doi: 10.48550/arXiv.1905.00075 [20] M. Łopuszyński and Ł. Bolikowski, “Tagging scientific publications using Wikipedia and natural language processing tools,” in Theory and Practice of Digital Libraries – TPDL 2013 Selected Workshops (CCIS), Ł. Bolikowski, V. Casarosa, P. Goodale, N. Houssos, P. Manghi, and J. Schirrwagen, Eds. Cham: Springer, 2014, vol. 416, pp. 25–36. doi: 10.1007/978-3-319-08425-1_3 [21] A. M. Dai, C. Olah, and Q. V. Le, “Document embedding with paragraph vectors,” arXiv, 2015. doi: 10. 48550/arXiv.1507.07998 [22] A. A. Alemi and P. Ginsparg, “Text segmentation based on semantic word embeddings,” arXiv, 2015. doi: 10.48550/arXiv.1503.05543 [23] L. van der Maaten and G. Hinton, “Visualizing data using t-SNE,” Journal of Machine Learning Research, vol. 9, no. 86, pp. 2579–2605, 2008. [24] A. Wijesinghe, Q. Wang, and S. Gould, “A regularized Wasserstein framework for graph kernels,” in 2021 IEEE International Conference on Data Mining (ICDM), pp. 739-748, 2021.
2nd International Conference on Multidisciplinary Sciences and Technological Developments (ICMUSTED 2025) December 12-15, 2025 219 Leveraging Large Language Models for Event Detection in Water Resources Literature Bengisu Sahin1, Ozge Yaren Turkseven2, Emrah Inan1 1Department of Computer Engineering, Izmir Institute of Technology, Izmir, Türkiye 2Department of International Water Resources, Izmir Institute of Technology, Izmir, Türkiye Corresponding author: Bengisu Sahin (e-mail: [email protected]) Abstract Event extraction is a critical task in information extraction that focuses on identifying event-related information within texts. Although event extraction has been researched for decades, it remains a highly challenging task. To carry out event extraction, a system must comprehend the text's semantics and ambiguity, while also organising the extracted information into structured formats. To address these challenges, we utilise open-source large language models for event extraction within a specific domain, ensuring greater flexibility. In this study, we present a new dataset designed for extracting events related to water resources. Experimental results reveal that the “gemma3:4b” model outperforms the others, which is why we choose to continue using it for few-shot settings. As we integrate samples from 0 to 5, the F1 score improves to 44.22. This encourages us to incorporate more samples into the prompt design, and with 20 samples, the model achieves the highest F1 score of 68.77. Keywords: Event extraction, Generative models, Groundwater, Water resources, Large language model 1. INTRODUCTION Event extraction is a critical task in information extraction within natural language processing (NLP). An event is an activity that takes place at a specific time and location, or it can be viewed as a change in state. A typical event extraction task involves two key subtasks: event detection, which focuses on identifying events within the text, and event classification, which categorises them into appropriate classes [1]. For event detection, it involves identifying the participants in the event through argument extraction and defining their attributes via argument role labelling. In essence, event extraction structures unstructured text by answering what, who, when, where, why, and how questions of an event, particularly in news texts. In the present study, event extraction is primarily approached as a classification problem, with the goal of identifying and categorising each event argument [2]. Previous research has demonstrated that traditional classification-based approaches to event extraction are data-hungry and suffer from data scarcity problems [3]. Despite significant progress, classification-based methods demand a large amount of training data to achieve optimal performance [4]. Furthermore, these methods typically struggle to handle new event types that have not been encountered during the training phase [5]. Recently, generative language models have become widely adopted across several NLP subfields, including event extraction. These generative approaches often differ from traditional methods of identifying and categorising events and their arguments, bringing both new opportunities and challenges, particularly in relation to training and evaluation [6]. To tackle challenges related to domain adaptation and the need for extensive training datasets, we leverage opensource large language models for event extraction within a specific domain, offering greater flexibility. In this study, we introduce a new dataset tailored for extracting events related to water resources. Experimental results reveal that the “gemma3:4b” model outperforms the others, which is why we choose to continue using it for fewshot settings. As we integrate samples from 0 to 5, the F1 score improves to 44.22. This encourages us to incorporate more samples into the prompt design, and with 20 samples, the model achieves the highest F1 score of 68.77. The paper is structured as follows: Section 2 reviews the relevant literature and examines current methodologies. Section 3 outlines the approach for extracting events within the domain of water resources. Section 4 presents the experimental results of various open-domain models applied to the generated dataset. Finally, Section 5 concludes the study and suggests potential directions for future research.
220 2. RELATED WORK In the early stages of event extraction, most methods rely on feature engineering for statistical classifiers. The features are typically derived from constituent parsers, dependency parsers, and contextual information. These models make use of statistical techniques such as nearest neighbour, maximum-entropy classifiers, and conditional random fields [7]. Sha et al. [8] proposed enhancing the bidirectional recurrent neural network (RNN) with dependency bridges, which transmit syntactic information when modelling words in a sentence. They demonstrate that using both hierarchical tree structures and sequential structures in the RNN simultaneously leads to improved performance compared to the traditional sequential structure. Liu et al. [5] investigate an approach to event extraction by explicitly treating it as a machine reading comprehension (MRC) problem. They show that this approach can improve performance by effectively utilising both the model and data within the MRC framework. Furthermore, they propose an unsupervised question generation method that connects MRC with event extraction. Their method generates questions that are both relevant to the topic and dependent on the context, thereby better guiding an MRC model in question-answering tasks. Ren et al. [9] focus exclusively on document-level event argument extraction using retrieval-augmented generation (RAG). Specifically, they first retrieve the top-k potentially relevant documents from the training corpus. The relevance of a document is assessed using a T5-encoder-based Siamese network, which compares the input text and event schema. The retrieved documents are then provided as additional input to the model, alongside the input document and schema information. Wang et al. [10] present InstructUIE as a unified information extraction framework for multiple IE tasks, aligned with UIE. Specifically, all IE tasks are redefined as natural language generation tasks, guided by expert-designed instructions that outline the required output format. InstructUIE facilitates joint training across multiple IE tasks using a collection of 32 datasets, creating a unified and semantically consistent label set. This approach enables cross-task knowledge sharing and benefits from an expanded pool of training data. Hsu et al. [11] formulate joint modelling for event extraction as a text generation task using pretrained generative language models. Leveraging these models, they employ an attention-based autoregressive decoder to generate event mentions, entity mentions, along with their labels and relationships. The task dependencies are encoded through the attention mechanism of the transformer-based decoder, enabling the model to learn the relationships between tasks and task instances in a flexible manner. In recent years, there have been two primary methods for utilizing large language models in event extraction. The first approach relies on prompting-based methods, where the model is treated as a black box and given task instructions in zeroor few-shot settings. These methods have the advantage of requiring no additional training, although their effectiveness can vary depending on the task and domain. The second approach of research emphasizes instruction-tuning of large language models, which directly adapt models to information extraction tasks through supervised fine-tuning on task-specific instructions. [12] investigated the integration of annotation guidelines with textual descriptions of event types and roles into instruction-tuned LLMs. They suggested five ways to automatically generate these annotation guidelines and evaluated how well they worked on the ACE05 and RichERE datasets [13, 14]. The results demonstrated that the use of structured guidelines leads to improved performance in low-resource settings and facilitates generalization across different annotation schemas. Domain-specific studies are crucial because general-purpose event extraction methods often face to capture the nuances of specialized contexts. These challenges in event extraction have been addressed by examining how models perform when encountering specialized language patterns and discourse. BRAD, a dataset based on nineteenth-century African American newspapers, was introduced in [15] and revealed significant performance gaps when models encountered domain-specific language and historical writing styles. Their experiments with BERT-based models highlighted the urgent need for domain adaptation, especially to handle evolving language and specialized terminology. In this study, we contribute to domain specific event extraction research by introducing a novel dataset focused on water resource related events. Instead of relying on additional fine-tuning as in instruction-tuning approaches, we adopt a few-shot prompting setting, which allows the use of open-source large language models without taskspecific retraining. By evaluating models on a specialized dataset, we demonstrate that domain-specific data can substantially enhance performance in few-shot settings, helping to bridge the gap between general-purpose event extraction methods and the requirements of domain-oriented applications.
221 3. PROPOSED METHOD 3.1. Dataset Preparation We utilise a sample study [16] to generate an event detection dataset tailored to water resources literature as a domain-specific task. Turkseven et al. [16] focus on how climate change presents a major threat to island ecosystems, highlighting the importance of sustainable groundwater management. They suggest that nature-based solutions, such as rainwater harvesting, provide innovative strategies to alleviate water scarcity and ensure the long-term sustainability of water resources. Table 1. Excerpts of the generated dataset for water resources Type Sample Sentence Climate change is one of the most crucial threats affecting ecosystems and water resources. Tokens “Climate”, “change”, “is”, “one”, “of”, “the”, “most”, “crucial”, “threats”, “affecting”, “ecosystems”, “and”, “water”, “resources” BIO “B-EVENT”, “I-EVENT”, “O”, “O”, “O”, “O”, “O”, “O”, “B-EVENT”, “I-EVENT”, “O”, “O”, “B-EVENT”, “I-EVENT” Sentence Understanding the unique status of the islands and the complex relationship between water resources and climate change is of great importance for the creation and development of sustainable policies in these areas. Tokens “Understanding”, “the”, “unique”, “status”, “of”, “the”, “islands”, “and”, “the”, “complex”, “relationship”, “between”, “water”, “resources”, “and”, “climate”, “change”, “is”, “of”, “great”, “importance”, “for”, “the”, “creation”, “and”, “development”, “of”, “sustainable”, “policies”, “in”, “these”, “areas”, “.” BIO “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “B-EVENT”, “I-EVENT”, “O”, “B-EVENT”, “I-EVENT”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O”, “O” Table 1 presents examples from the generated dataset, including the sentence, token list, and BIO format. The BIO format denotes the token position, where “B-” marks the beginning of an event, “I-” represents the continuation of an event token, and “O” indicates tokens that are not part of an event [17]. In this context, we adapt the generated dataset for use with the transformer-based named entity recognition (T-NER) library [18], which supports the exploration of cross-domain generalisation. 3.2. Event Detection Method To detect events, we utilise the Ollama library [19], a tool designed to streamline the deployment and operation of Large Language Models (LLMs) for a range of language tasks. Ollama provides a solution that improves the utility of LLMs, making them more adaptable to employ in different contexts, especially on local machines, as opensource models. From the Ollama library, we utilise Gemma 3 [20], a multimodal extension to the Gemma series of lightweight open models, ranging from 1 to 27 billion parameters. This updated version introduces vision capabilities, expanded language support, and longer context handling—up to 128K tokens. Additionally, the model's architecture has been adjusted to minimise the excessive growth of KV-cache memory when dealing with long contexts. Another model is Llama 3 [21], a series of language models developed to support multilingual tasks, including coding, reasoning, and tool usage. The largest model in the series is a dense Transformer with 405 billion parameters and a context window of up to 128K tokens. Llama 3 has been publicly released, featuring both pretrained and post-trained versions of the 405B parameter language model. We also make use of Phi-4 [22], a 14billion parameter language model designed with a training methodology focused on data quality. In contrast to most language models that mainly depend on organic data sources such as web content or code for pre-training, Phi-4 deliberately incorporates synthetic data throughout its training. 4. EVALUATION In this work, we employ the seqeval [23] as the experimental framework and evaluation metrics. This is a Pythonbased framework used for sequence labelling assessment. To evaluate the performance of the generative models,
222 we utilise precision, recall, and F1 score metrics. Figure 1 illustrates the prompt design templates for both the dataset preparation and event detection tasks (below). The first prompt uses definitions and guidelines to identify events in the literature. For the event extraction and detection task, we adapt the prompt design from the PromptNER study [24] with the following: Figure 1. Templates for prompt design used in dataset preparation and event detection Table 2 presents a comparison of the models from Ollama, where we utilise the “llama3.2” model with 1 billion parameters [25], the “phi3” model with 3.8 billion parameters [26], and the “gemma3” model with 4 billion parameters [27]. Given that these models have relatively small parameter sizes, they tend to yield poor results. Additionally, since these are zero-shot models, no sample is integrated into the chat history of the prompt design. Among these, the “gemma3:4b” model outperforms the others, which is why we choose to continue using it for few-shot settings. Table 2. Comparison of zero-shot models on the water resources dataset Model P R F1 llama3.2:1b 21.26 24.34 22.70 phi3:3.8b 19.27 24.34 21.51 gemma3:4b 37.37 24.34 29.48 Table 3 presents the performance of the “gemma3:4b” model with varying sample sizes. The zero-shot setting, with no samples, results in the lowest F1 score of 29.48. As we integrate samples from 0 to 5, the F1 score improves to 44.22. This encourages us to incorporate more samples into the prompt design, and with 20 samples, the model achieves the highest F1 score of 68.77. Table 3. Comparison of few-shot performance of the gemma3:4b model on the water resources dataset Model Samples P R F1 gemma3:4b 0 37.37 24.34 29.48 5 45.77 42.76 44.22 10 65.93 58.55 62.02 20 73.68 64.47 68.77
223 5. CONCLUSION AND FUTURE WORK In this study, we leverage open-source large language models for event extraction within a specific domain, offering enhanced flexibility. To assess the performance of these models, we introduce a new dataset tailored for extracting events related to water resources. In the zero-shot setting, where no samples are provided, the “gemma3:4b” model achieves the lowest F1 score of 29.48. As we incorporate samples ranging from 0 to 5, the F1 score improves to 44.22. This motivates us to add more samples to the prompt design, and with 20 samples, the model reaches the highest F1 score of 68.77. One limitation of this study is the use of large language models with relatively low parameter sizes, due to resource constraints. Another limitation is the relatively small size of the prepared dataset. As a future direction, we plan to parse additional water resources papers to expand the dataset. Additionally, we aim to adapt the large language model approach to automate the extraction of supply chain information [28], as achieving comprehensive supply chain visibility is essential for effective risk management. References [1] V. D. Lai, “Event extraction: A survey,” arXiv Preprint, 2022. doi: 10.48550/arXiv.2210.03419 [2] T. H. Nguyen, K. Cho, and R. Grishman, “Joint event extraction via recurrent neural networks,” in Proc. 2016 Conf. North Amer. Chapter Assoc. Comput. Linguistics: Hum. Lang. Technol., San Diego, CA, USA, 2016, pp. 300–309. [3] J. Liu, Y. Chen, K. Liu, W. Bi, and X. Liu, “Event extraction as machine reading comprehension,” in Proc. 2020 Conf. Empirical Methods in Natural Language Processing (EMNLP), Nov. 2020, pp. 1641–1651. [4] J. Liu, Y. Chen, K. Liu, and J. Zhao, “Event detection via gated multilingual attention mechanism,” in AAAI Conf. Artif. Intell., 2018. [5] J. Liu, Y. Chen, K. Liu, W. Bi, and X. Liu, “Event extraction as machine reading comprehension,” in Proc. 2020 Conf. Empir. Methods Nat. Lang. Process. (EMNLP), Nov. 2020, pp. 1641–1651. [6] É. Simon, H. Olsen, H. You, S. Touileb, L. Øvrelid, and E. Velldal, “Generative approaches to event extraction: Survey and outlook,” in Proc. Workshop Future Event Detection (FuturED), Nov. 2024, pp. 73–86. [7] A. Majumder and A. Ekbal, “Event extraction from biomedical text using CRF and genetic algorithm,” in Proc. 2015 3rd Int. Conf. Comput., Commun., Control Inform. Technol. (C3IT), 2015, pp. 1–7. [8] L. Sha, F. Qian, B. Chang, and Z. Sui, “Jointly extracting event triggers and arguments by dependencybridge RNN and tensor-based argument interaction,” in Proc. 32nd AAAI Conf. Artif. Intell., 2018. [9] Y. Ren, Y. Cao, P. Guo, F. Fang, W. Ma, and Z. Lin, “Retrieve-and-sample: Document-level event argument extraction via hybrid retrieval augmentation,” in Proc. 61st Annu. Meeting Assoc. Comput. Linguistics (Vol. 1: Long Papers), Toronto, Canada, 2023, pp. 293–306. [10] X. Wang, W. Zhou, C. Zu, H. Xia, T. Chen, Y. Zhang, et al., “InstructUIE: Multi-task instruction tuning for unified information extraction,” arXiv Preprint, 2023. doi: 10.48550/arXiv.2304.08085 [11] I.-H. Hsu, K.-H. Huang, E. Boschee, S. Miller, P. Natarajan, K.-W. Chang, and N. Peng, “DEGREE: A data-efficient generation-based event extraction model,” in Proc. 2022 Conf. North Amer. Chapter Assoc. Comput. Linguistics: Hum. Lang. Technol., Seattle, WA, USA, 2022, pp. 1890–1908. [12] S. Srivastava, S. Pati, and Z. Yao, “Instruction-tuning LLMs for event extraction with annotation guidelines,” arXiv Preprint, 2025. doi: 10.48550/arXiv.2502.16377 [13] G. R. Doddington, A. Mitchell, M. Przybocki, L. Ramshaw, S. Strassel, and R. Weischedel, “The automatic content extraction (ACE) program – tasks, data, and evaluation,” in Proc. 4th Int. Conf. Lang. Resour. Eval. (LREC), Lisbon, Portugal, May 2004. [14] Z. Song, A. Bies, S. Strassel, T. Riese, J. Mott, J. Ellis, J. Wright, S. Kulick, N. Ryant, and X. Ma, “From light to rich ERE: annotation of entities, relations, and events,” in Proc. 3rd Workshop on EVENTS: Definition, Detection, Coreference, and Representation, Denver, CO, USA, 2015, pp. 89–98. doi: 10.3115/v1/W15-0812 [15] V. Lai, M. Van Nguyen, H. Kaufman, and T. H. Nguyen, “Event extraction from historical texts: a new dataset for black rebellions,” in Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, Stroudsburg, PA, USA: Association for Computational Linguistics, 2021, pp. 2390–2400. doi: 10.18653/v1/2021.findings-acl.211 [16] O. Y. Turkseven, O. Gunduz, and A. Baba, “Innovative Solutions on Water Management for Islands Under Climate Change,” in Int. Conf. Water Problems Mediterranean Countries, Cham, Switzerland, Dec. 2023, pp. 33–40. [17] J. C. S. Alvarado, K. Verspoor, and T. Baldwin, “Domain adaptation of named entity recognition to support credit risk assessment,” Proc. Australasian Lang. Technol. Assoc. Workshop, Dec. 2015, pp. 84–90.
230 This energy output is sufficient to power small household loads such as fans, lighting, and pumps. Consequently, such systems provide a cost-effective, renewable energy solution, particularly valuable for rural communities in developing regions experiencing persistent energy shortages. Hydraulic power exhibits an increase with rising head height. At a head of 10 ft, the hydraulic power is 24 W, which increases to 57 W at 24 ft. This trend demonstrates that greater power is necessary to elevate water to higher levels, consistent with the underlying principles of pumping system mechanics. The input power at both 60% and 40% loads also increases with the head, as shown in Figs. 4(a) and 4(b). For example, at 10 ft, the input power at 60% load is 40 W, and at 40% load, it's 60 W. As the head increases, the difference between the input power at 60% and 40% load grows, with the 60% load requiring higher input power due to increased operational demand. At the highest head (24 ft), the input power at 60% load is 96 W, while at 40% load, it is 144 W. Run-time is determined by the total energy available (3.0 kWh or 3.5 kWh) and the input power required. As the head increases, the system consumes more power, thus reducing the runtime, as exhibited in Figs. 4(c) ve 4(d). For example, at 10 ft, the run-time at 3.0 kWh ranges from 75 to 50 hours, while at 24 ft, the run-time ranges from 31 to 21 hours. Similarly, for 3.5 kWh, the run-time is slightly longer but follows the same decreasing pattern with increasing head. At 10 ft, it’s between 88 and 59 hours, while at 24 ft, it’s between 37 and 24 hours. The observations clearly illustrates how system performance changes with increasing head. As head increases, more hydraulic and input power is required, which reduces the available run-time per given energy capacity. This data is essential for evaluating the efficiency and operational feasibility of the pump under varying conditions. Figure 4. Effect of head on power requirement and run-time performance 2.4. Mathematical Modelling Power required for compressor and pump is calculated by: 60 w WN P× = ⋅ (1) Low pressure cylinder work done by two stage reciprocating air compressor with inter cooler is calculated by: 1 2 1 11 1 1 1 n n p n W pv np − = × × −⋅ − (2) High pressure cylinder work done is calculated by:
[Document text truncated for crawler view.]