ABSTRACT. The application of low-cost automation is a permanent trend in modern production. It is especially promising in control and sorting operations, characterized by labor-intensiveness and low productivity. The paper presents methods and approaches for implementing a system for sorting parts of the type "Connector" by size. The suitability of the parts for automated sorting is assessed, a method for control and sorting is selected, an original passive contact system for automatic orienting is described, the main functional parameters of the system are determined. By unifying the transporting, control and sorting device, a simplified structure, increased reliability and low cost of the system are achieved. The main technical characteristics of the proposed system are indicated.

ABSTRACT. Fault-tolerant distributed systems are implemented in safety-critical applications where a system failure could cause severe damage and threaten human lives. To guarantee their flawless operation, their dependability attributes must be embedded early and continuously throughout system design, rather than treating them as an afterthought. This paper presents a conceptual framework for integrating the approach of adjustable reliability into the System Development Life Cycle (SDLC). The approach of adjustable reliability provides a way to distribute structural hardware redundancy and achieve the system reliability required by the application. The proposed framework shifts reliability from a design add-on to a core architectural decision variable in the design of dependable distributed systems. Opportunities and challenges involved are discussed, and some future research directions are outlined.

ABSTRACT. The paper discusses some important microcontroller board features in the application context of laser projection systems, which include measuring temperature and humidity by analog and digital sensors, controlling internal and external devices according to sensor values and time schedules, as well as communicating the system state to remote computers. Accordingly, the hardware architecture of the microcontroller board is designed, specific implementation details are illustrated and experimental results are discussed showcasing a custom desktop application for monitoring and control. Future development will build on the flexibility and extensibility of the board to include additional digital sensors and connectivity options

ABSTRACT. This paper discusses the creation of a microcontroller firmware that focuses on the control of lighting, temperature and external devices and enables sensor data acquisition and transmission to remote systems in industrial applications. Relevant economic and technological aspects are outlined, a suitable firmware architecture is proposed, related implementation details are presented and experimental results are summarized after gathering practical experience with three microcontroller boards developed for real-world application. The proposed firmware achieves a good balance between development cost, usability and reliability and provides enough room for future integration in new application scenarios and configuration adjustments requested by clients.

ABSTRACT. A Data fusion methods play a key role in modern multisensor systems, yet their selection remains largely heuristic due to the absence of a unified classification framework. This paper reviews existing classification approaches based on the level of data abstraction, system architecture, information interaction and mathematical foundations, taking into account their conceptual features and practical limitations. It also proposes a new classification approach based on partial information decomposition. This approach enables the objective categorisation of sensor systems and fusion methods according to the dominance of unique, redundant, and synergistic information in the acquired data.

ABSTRACT. The current study presents the design, implementation, and evaluation of a real-time face recognition system for automated access control. The system uses Python libraries to build an accurate and secure identification platform that incorporates dedicated stages for facial data processing and recognition. During data preparation, 128-dimensional facial embedding vectors are generated for authorized users through a command-line interface and protected using authenticated encryption. In real-time operation, the system captures video frames, detects faces, and verifies identities by matching them against the encrypted database. Experimental results demonstrate high recognition accuracy, real-time throughput and robust performance, highlighting the system’s suitability for GDPR-oriented deployment in small institutional environ-ments.

ABSTRACT. This article presents the results of a study on implementing a service level metrics system (SLI/SLA/SLO) in a high-load web application on the ASP.NET 8 platform. A practical implementation of a monitoring system using OpenTelemetry Metrics and a Prometheus endpoint is considered. A comparative analysis of system performance and reliability indicators before and after implementing the metrics is conducted. The mechanisms by which metrics influence operational indicators are described in detail: identifying hidden performance issues, transforming the alerting system, prioritizing engineering efforts, managing the balance between development speed and reliability through an error budget, and automating response processes. The study results showed an improvement in the average time to detect incidents by 73%, a reduction in service restoration time by 58%, and an increase in overall system availability from 98.2% to 99.7%. A methodology for determining target SLO values based on business requirements and architectural constraints is presented.

ABSTRACT. During the last 50 years, average wild animal populations have decreased by 73%. The present article is part of a scientific project for tracking wild animal habitats in the Bulgarian mountains. Based on the PlantNet base model, a model based on the YOLOv8n architecture has been created through transfer learning on local data to recognize animal species specific to Bulgarian geographical latitudes. In this paper, a system for assessing population dynamics and animal movement through the analysis of time series of images is proposed. The effectiveness of the proposed methodology is illustrated using a database of images from Ukraine, as it provides a sufficient number of images needed to train and test the model, monitoring changes in the behavior of wild animals over time. This will enable the detection of migration patterns, preferred habitats, and potential threats to populations.

ABSTRACT. This paper presents the design and preliminary field validation of a portable, low-cost Internet of Things (IoT) station for georeferenced soil nutrient profiling in agricultural environments. The proposed system integrates a digital RS-485 NPK soil sensor, an ESP32 microcontroller, and a SIM7000G GSM/GPS module to enable on-site acquisition and real-time transmission of nitrogen (N), phosphorus (P), and potassium (K) measurements using the MQTT protocol. Data are serialized in JSON format and transmitted to a ThingsBoard cloud platform for remote storage and visualization. The portable architecture supports manual spatial sampling across multiple locations without reliance on fixed infrastructure, making it suitable for small- and medium-scale agricultural contexts with limited connectivity. Preliminary testing in a controlled lemon plantation demonstrated stable GSM connectivity, successful geotagging, and consistent cloud-based visualization, with an average acquisition-transmission cycle of 30–45 s per measurement. Spatial heat maps generated from collected data illustrate the system’s capability for indicative nutrient mapping. Although laboratory-grade validation is ongoing, the results confirm the technical feasibility of integrating low-cost sensing, cellular communication, and georeferenced data acquisition into a compact IoT unit. The system establishes a foundation for future calibration, large-scale field validation, and decision-support applications in precision agriculture.

ABSTRACT. Small and Medium-sized Enterprises (SMEs) in the photovoltaic sector face significant challenges in managing and leveraging data for strategic decision-making. This case study examines a comprehensive database optimization initiative within a European PV company, focusing on the digitalization of sales processes through Databricks and Power BI integration. The research documents a practical methodology for data consolidation, cleaning, semantic layer design, and installer segmentation that enabled real-time an-alytics and evidence-based decision-making. Results demonstrate improvements in data quality (67% to 94% completeness), reporting efficiency (days to real-time), and meas-urable business outcomes including resource reallocation and customer win-back cam-paigns.

ABSTRACT. Traditional authentication systems rely mainly on static login checkpoints such as passwords or one-time verification. However, the expansion of cloud services, mobile devices, and distributed digital platforms has exposed significant limitations in these approaches. Modern cyberattacks increasingly exploit credential theft, phishing, and session hijacking in order to bypass login-based security mechanisms. This study examines the use of behavioral biometrics and data-driven analytics in continuous authentication systems that verify user identity throughout an active session. Behavioral interaction signals such as keystroke dynamics, cursor movement patterns, touchscreen gestures, and device usage characteristics can form distinctive behavioral profiles for individual users. Machine-learning models can analyze these signals to detect deviations from established behavioral patterns that may indicate unauthorized access. The paper develops a conceptual framework for continuous behavioral authentication that integrates behavioral monitoring, anomaly detection, and scalable data-processing infrastructures. The analysis highlights both the cybersecurity benefits of behavioral authentication and the challenges related to large-scale behavioral data collection, including privacy protection and responsible data governance.

ABSTRACT. Managing event-driven communication in large reactive codebases remains a persistent pain point: developers routinely write repetitive wiring code that quickly becomes difficult to trace and maintain. We present AutoSignal, a programming paradigm for the Dart ecosystem that automates component interconnection through a signal-slot architecture built on name and data-type matching. To ensure scalability and prevent unintended side effects in large-scale applications, the framework implements a hierarchical namespace isolation model and guarantees deterministic reactivity via topological sorting of dependency graphs. Empirical evaluations demonstrate that the AutoSignal framework reduces manual connection code by approximately 47% and decreases unintended signal propagation by 32% compared to standard manual implementation schemes. Furthermore, the framework integrates seamlessly with the Flutter ecosystem, providing high-performance reactive primitives such as specialized hooks and builders. Overall, AutoSignal proves to be a practical, type-safe option for teams building maintainable event-driven applications in the Dart/Flutter ecosystem.

ABSTRACT. In this article, we present a constructive method for designing an information security system (ISS). The method is based on the IEEE 1471 and IEEE 42010 standards. They provide an architectural framework for describing the system through conceptual mod-eling from different perspectives. The perspectives reflect the requirements of stake-holders - regulatory, normative, technological or budgetary. As result of analysis of the problem area, conceptual models are constructed. The resulting models are combined into generalized multi-layer model. The transformation of the conceptual model into a technology-independent object-oriented (OO) design model follows. Next stage is selec-tion of an appropriate technological platform and subsequent transformation of the design model into an implementation model. An essential part of the method is the creation of agent-based simulation model. It allows simulation of the ISS in different environments, with changing the input conditions. The method ensures technological independence of the ISS, due to the fact that the resulting conceptual model reflects the requirements of the system and the methods for implementing the tasks of the ISS without using a specific technological solution. The method also ensures universal communication between the individual stakeholders and unification of the used terminology.

ABSTRACT. The primary goal of this project is to develop a robust model for forecasting electric power consumption in Bulgaria. Leveraging historical forecast data from Open-Meteo for weather-related features and Entsoe data, our objective is to create an accurate prediction tool that can assist in optimizing energy management within the country. By achieving this goal, we aim to improve energy reliability, support data-driven decision-making in energy policy, and promote sustainable energy practices in Bulgaria. This predictive model will empower us to proactively address fluctuations in energy demand, particularly during extreme weather conditions, and will contribute to the efficient allocation of electrical resources.

ABSTRACT. Supply chain traceability often fails in practice because relevant records are scattered across production, warehouse, transport, laboratory, and document systems. When a recall or audit is needed, firms must manually collect and reconcile evidence from many sources. Existing standards and blockchain platforms address parts of this problem, but prior work still reports recurring weaknesses in governance, confidentiality management, interoperability, and performance measurement. This paper presents a governance-aware, privacy-preserving, event-driven conceptual model for supply chain traceability. The model uses five event types—Create, Transform, Transfer, Verify, and Recall—linked through explicit lineage. It stores compact signed event headers on-ledger and anchors detailed off-ledger payloads and governance policy text through hashes. It also links data-sharing choices to consortium governance, defines validation invariants, embeds key performance indicators, and produces two regulator-ready outputs: a product passport and an audit pack. The contribution is a standards-informed conceptual artifact that integrates event semantics, provenance reconstruction, selective disclosure, governance, validation, performance measurement, and regulator-ready outputs in one cross-sector traceability model.

ABSTRACT. EUPHEMIA, the Pan-European day-ahead electricity market coupling algorithm, operates in a centralised manner that limits independent auditability and has been characterised as exhibiting pseudo-transparency. This paper proposes a blockchain-based architecture that enhances the transparency and verifiability of the market-coupling process without compromising participant confidentiality. The architecture introduces an off-chain computation model with selective on-chain publication, applied separately to the three principal data categories of the EUPHEMIA pipeline: order books, network constraints, and clearing outputs. Zero-knowledge proofs based on zk-STARKs verify the integrity of the order book aggregation process and targeted network constraint sub-processes, while Merkle commitments provide tamper-evident anchoring of publicly disclosed data. zk-STARKs are selected over CRS-based alternatives to eliminate the trusted setup governance overhead inherent to EUPHEMIA's multi-jurisdictional structure. Circuit complexity is estimated at approximately 2,225,000 arithmetic constraints in the worst-case NEMO (EPEX SPOT) scenario. A targeted capacity reduction proof requires 391 constraints when a Transmission System Operator (TSO) challenges the capacity constraints calculated by the Regional Coordination Centre (RCC). The computational effort of the prover and verifier follows established theoretical bounds. Welfare maximisation cannot be verified end-to-end due to the absence of a full public algorithm specification. Empirical benchmarking of proof generation times remains for future work.

ABSTRACT. Quantum computing has emerged as a powerful tool for solving complex problems in various fields. Personalized medicine, tailoring medical treatment to patients based on their genetic and health data, is one area where predictive analytics can be useful. This work explores the application of quantum algorithms for predictive analytics, specifically in the context of predicting outcomes of cardio-vascular disease based on patient data. The research is focused on the quantum-based predictive models for the case study of cardiovascular disease. The models are based on Quantum Support Vector Machines, Quantum Neural Networks, and Variational Quantum Eigensolver algorithms. The software implementation is based on the Python programming language, including an integrated quantum algorithms. A dataset of cardiovascular disease from an online platform is used to train and evaluate the models.

ABSTRACT. This work examines the implementation and evaluation of shortest path algorithm in a Geographic Information System environment integrating QGIS with a Post-greSQL/PostGIS spatial database. Spatial edge and junction layers stored in the Post-greSQL/PostGIS define a directed weighted network in which edge costs are derived from spatial distance and modified through a gradient-based cost function. The rout-ing algorithm is implemented in Python using the PyQGIS library and a binary heap priority queue. Network data are loaded from the geodatabase and processed in memory to compute the optimal route between selected nodes. The resulting path is reconstructed as a dissolved polyline feature stored in the geodatabase and visualized within the GIS environment. The study also includes a theoretical comparison of sev-eral classical shortest-path algorithms—Dijkstra, A*, Bellman–Ford, and Floyd–Warshall—with respect to their applicability to sparse spatial graphs typical of trans-portation networks. The analysis confirms the suitability of Dijkstra’s algorithm for such networks and demonstrates that routing outcomes depend directly on the defini-tion of the edge cost function. The proposed workflow relies exclusively on open-source GIS and database technologies.

ABSTRACT. Automated recognition of identity documents has become a critical component in digital identity verification systems, including banking, e-government services, border control, and security applications. Despite their importance, the development of reliable ID card recognition systems faces several challenges. The development of robust recognition models is often constrained due to ID card recognition systems rely heavily on the limited availability of large, diverse, high-quality, and publicly accessible ID card datasets. Collecting and annotating real-world ID card images is costly, time-consuming, and often restricted due to privacy, legal, and security concerns. Additionally, real-world images suffer from variations in lighting, background, wear-and-tear, and capture conditions, which degrade recognition performance. This paper proposes a novel approach for generating a large-scale synthetic dataset for ID card recognition by merging real ID card images with a texture dataset through a structured data fusion pipeline. The proposed method introduces realistic visual variations by integrating background textures, illumination effects, geometric distortions, and noise patterns while preserving the semantic integrity of the original ID card content. Experimental evaluations using a deep learning-based recognition model demonstrate that training with a synthetic dataset generated utilizing the suggested approach significantly improves model generalization, robustness, and accuracy.

ABSTRACT. Heart disease continues to rank as the foremost contributor to premature death across the globe, and developing reliable automated screening tools has become a pressing priority in clinical informatics. This study presents a comparative analysis of three machine learning classifiers — Logistic Regression (LR), Random Forest (RF), and Gradient Boosting (GB) — for binary heart disease prediction using the UCI Cleveland Heart Disease dataset. The dataset comprises 297 patient records with 13 clinical features following removal of missing values. Models were evaluated using an 80/20 stratified train-test split with Min-Max normalisation. Logistic Regression achieved the highest AUC-ROC of 0.9554 and accuracy of 83.33%, followed by Random Forest (AUC-ROC = 0.9375, accuracy = 85.00%) and Gradient Boosting (AUC-ROC = 0.8828, accuracy = 76.67%). Feature importance analysis identified chest pain type (cp), thalassemia (thal), and maximum heart rate (thalach) as the most influential predictors. Results indicate that well-calibrated linear classifiers can match or exceed ensemble methods in structured clinical settings, a practically significant finding for risk-stratification tools where probability estimates must be trusted.

ABSTRACT. This paper studies a deep probabilistic modeling framework that learns bidirectional sequence representations from large unlabeled text and then reuses them across multiple supervised tasks. The approach trains a multi-layer Transformer encoder with masked token prediction and sentence-level discrimination objectives, yielding context-sensitive embeddings that integrate information from both left and right contexts. These pretrained representations are then fine-tuned with minimal task-specific modifications to optimize standard classification and span-prediction losses on a variety of benchmarks. Experiments show that scaling the model and training corpus leads to substantial gains over prior transfer-learning methods in natural language processing, illustrating the effectiveness of bidirectional attention-based encoders as general-purpose statistical machine learning models.

ABSTRACT. The increasing sophistication of generative artificial intelligence (AI) has led to an exponential rise in highly realistic synthetic facial media. As detection models evolve, so do adversarial techniques designed to bypass them, creating an arms race between deepfake creators and defenders. This paper investigates adversarial robustness and explainability in the detection of AI-generated faces, proposing the Robust and Explainable Detection (RED) framework that combines adversarial training with explainable AI (XAI) tools such as Gradient-weighted Class Activation Mapping (Grad-CAM). We evaluate how convolutional neural networks (CNNs), Vision Transformers (ViTs), and hybrid multi-stream detectors perform under adversarial perturbations. RED integrates multi-stream feature extraction, adversarial defense layers, and Grad-CAM-based interpretability. The methodology is formalized with FGSM and PGD adversarial training, with loss formulation and configuration options; the implementation is released for reproducibility. Experiments use the Kaggle faces dataset (real/fake structure) with stratified 70/15/15 train/validation/test splits. Metrics include accuracy, F1, AUC-ROC, Adversarial Robustness Index (ARI), and Explainability Fidelity (EF). The framework contributes toward trustworthy, transparent, and robust AI systems for detecting synthetic facial media and supports forensic and legal applications where both accuracy and interpretability are required.

ABSTRACT. This study proposes a comprehensive design framework for computer mice tailored to Vietnamese users, based on anthropometric data collected from 610 university students using a cross-sectional survey. Three important parameters—hand length, finger length, and palm width—were used to specify five optimal size groups with an overall satisfaction rate of 81.7%.The average hand length of 186 mm places Vietnamese users within the 20th–30th percentile of the U.S. population (ANSUR), suggesting that devices designed as medium-sized in Western markets may be oversized in Vietnam. Parametric modeling combined with Fitts’ Law and ISO 9241 produced geometrically precise designs that reduce discomfort at inclination angles of 25°–30%, while also enabling the development of a user-friendly tool for selecting the most appropriate mouse size for consumers.

ABSTRACT. Effective thermal management is essential for maintaining reliability and performance of modern electronic devices as heat generation increases with system miniaturisation. This study presents a numerical investigation of fluid flow and heat transfer in wavy microchannel heat sinks with circular pin inserts of varying heights ranging from 0 to 0.4 mm. A 3D conjugate heat transfer model was developed in ANSYS Fluent to evaluate thermal-hydraulic performance with Reynolds numbers ranging from 300 to 800. The microchannels were examined with wave amplitudes of 0, 150, and 250 μm. Water was used as the coolant and a constant heat flux of 50 W/cm2 was applied at the base of the heat sink. Key performance indicators including the Nusselt number, friction factor, and performance evaluation criterion were used to assess heat transfer enhancement and flow resistance. Results show that channel waviness significantly improves heat transfer, with the 250 μm amplitude channel outperforming the straight and 150 μm channels. Increasing pin height further enhances the Nusselt number, reaching about 15% higher than the smooth channel, although it also increases the pressure drop. The best overall thermal-hydraulic performance was obtained for a wavy microchannel with 250 μm amplitude and a pin height of 0.1 mm.

ABSTRACT. The Betz limit defines the maximum theoretical conversion of the kinetic energy of an air flow into mechanical work for an idealized free-standing wind turbine. In practice, how-ever, turbines are sometimes installed in confuser–diffuser or Venturi-type channels, where the flow is geometrically constrained. In this study, a theoretical model of a wind turbine integrated into a confuser–diffuser channel is analyzed. Based on energy and momentum balance, a relationship between system efficiency and a dimensionless geo-metric parameter describing the ratio of characteristic channel cross-sections is derived. The analysis shows that the maximum kinetic energy conversion is achieved at a cross-section ratio of n = 1.299, indicating that flow profiling can influence turbine energy performance.

ABSTRACT. This study presents a numerical investigation of airflow through a G-shaped channel, fo-cusing on the influence of geometric parameters on pressure distribution and velocity field development. A three-dimensional CFD model was developed to analyze three channel configurations under identical boundary conditions, with an inlet velocity of 10 m/s and atmospheric pressure at the outlet. The results reveal significant pressure redistribution in the curved section and localized acceleration of the airflow near the inner wall, where ve-locity increases of up to 49% were observed. The findings demonstrate the strong effect of curvilinear geometry on flow energy transformation and support potential applications in adaptive wind turbine systems.

ABSTRACT. This article presents an experimental study aimed at determining the optimal print temperatures for the nozzle and the build plate when printing ABS, by evaluating the geometric accuracy of the parts. The recommended print temperatures were analyzed, and various operating temperature regimes were selected for the nozzle and the build plate. The experimental part is based on single-factor and two-factor experiments. An analysis of the manufacturing process was also performed. Despite the experimental determination of the optimal temperatures, the manufactured parts do not meet the specified tolerances. Further research is needed to determine the shrinkage coefficient and to make corrections to the initial model.

ABSTRACT. During operation, several parts on thermal reactor burners sustain heavy damage and need to be replaced. Different solutions for parts manufacturing are investigated: thermal spraying, Selective Laser Melting (SLM) and hardfacing by Directed Energy Deposition plasma arc (DED-arc). Based on the comparison to original material and the duration of usage, application of those three methods for replacement is studied in order to determine the most suitable one with Additive Manufacturing (AM) being proposed for targeting the problem. Thermal sprayed items have zirconium-oxide based outer layer, SLM gives monolithic item while with the help of DED-arc is produced a composite structure with sound metallurgical bond between the base and the added material. The microstructures with the interface zones are observed. Samples are machined and grinded and their friction characteristics taken with the help of acoustic emission (AE) and electrical contact resistance (ECR) sensors during scratching. As a result is proposed overlaying by DED-arc of the base stainless steel due to the better metallurgical stability of the added mixture in hot environment above 800оС and its hardness characteristics.

ABSTRACT. The modification of bitumen with recycled rubber materials has gained significant at-tention due to its potential to enhance pavement performance and support sustainable waste management. In this study, the rheological properties of bitumen modified with crumb rubber (CR) and devulcanized crumb rubber (DCR) were investigated. Rubber modifiers were added at concentrations of 5–25% by weight, and the rheological be-havior was evaluated using a Dynamic Shear Rheometer at 1.59 Hz over a tempera-ture range of 46–96 °C. Key parameters, including storage modulus (G′), loss modulus (G″), and complex viscosity (η*), were analyzed. The results indicate that increasing rubber content significantly enhances stiffness and viscosity, improving resistance to deformation at elevated temperatures. Moreover, DCR-modified binders exhibit higher rheological performance compared to CR systems, indicating better compatibility with the bitumen matrix. Overall, devulcanized rubber demonstrates superior efficiency as a modifier and shows strong potential for improving the durability and high-temperature performance of asphalt binders.

ABSTRACT. Chromium–nickel steels represent an important class of engineering materials due to their excellent corrosion resistance, good mechanical properties and structural stability [1] (pp. 15–20). The present study investigates the influence of chemical composition on the microstructure and hardness of three chromium–nickel steels containing constant chromium content (24 wt.%) and varying nickel concentration. Nitrogen was introduced only in the steel with the lowest nickel concentration through nitrided ferrochromium during melting in an induction furnace. Three steels were investigated: 24CrNi8N containing 0.3 wt.% nitrogen, 24CrNi14 without nitrogen addition, and 24CrNi20 without nitrogen. The measured hardness values were 360 HV, 230 HV and 190 HV respectively. Optical metallographic analysis revealed significant differences in the microstructure of the investigated steels. The nitrogen-containing steel exhibited dendritic morphology with partially transformed regions, whereas the steels with higher nickel content showed predominantly austenitic and fully austenitic structures. The results demonstrate that both nitrogen alloying and nickel concentration strongly influence phase stability, microstructure and hardness of chromium–nickel steels.

ABSTRACT. High-chrоmium whitе cаst irоns rеprеsеnt аn impоrtаnt grоup оf wеаr-rеsistаnt еnginееring mаtеriаls widеly usеd in mining, minеrаl prоcеssing аnd cеmеnt industriеs duе tо thеir еxcеllеnt аbrаsiоn rеsistаncе аnd high hаrdnеss [1] (pp. 210–215). Thе prеsеnt study invеstigаtеs thе influеncе оf chеmicаl cоmpоsitiоn аnd mаgnеsium mоdificаtiоn оn thе micrоstructurе аnd hаrdnеss оf twо high-chrоmium cаst irоns. Twо аllоys wеrе еxаminеd: а 28 wt.% Cr cаst irоn withоut mаgnеsium аdditiоn аnd а mоdifiеd аllоy cоntаining 14 wt.% Cr аnd 0.88 wt.% Mg. Оp-ticаl mеtаllоgrаphic аnаlysis rеvеаlеd significаnt diffеrеncеs in cаrbidе mоrphоlоgy bеtwееn thе invеstigаtеd аllоys. Thе аllоy withоut mаgnеsium еxhibits cоаrsе primаry M7C3 chrоmium cаrbidеs еmbеddеd in thе mеtаllic mаtrix, whеrеаs thе Mg-mоdifiеd аllоy shоws а significаntly rеfinеd еutеctic structurе with finе cаrbidе distributiоn. Hаrdnеss mеаsurеmеnts rеvеаlеd vаluеs оf аpprоximаtеly 508 HV fоr thе nоn-mоdifiеd аllоy аnd 612 HV fоr thе Mg-mоdifiеd аllоy. Thе оbtаinеd rеsults dеmоnstrаtе thе strоng rеlаtiоnship bеtwееn chеmicаl cоmpоsitiоn, mi-crоstructurе аnd hаrdnеss оf high-chrоmium cаst irоns.

ABSTRACT. The widespread application of plain S235JR steel across various industries often involves its exposure to degradative conditions, such as the organic acid concentrations typically found in various foodstuffs. This study investigates the corrosion behavior of S235JR steel immersed in several common beverages. The corrosion of steel S235JR was evaluated by gravimetric and electrochemical study. Statistical modeling via multiple linear regression and ANOVA was utilized to verify the experimental dataset. Based on the results, the highest early-stage corrosion occurs in coffee and tomato juice, though apple cider vinegar stands out for its steadily intensifying corrosive effect over the immersion period. Conversely, the presence of citric acid in lemon and orange juices appears to act as an inhibitor, leading to the lowest observed corrosion rates. The strong correlation between electrochemical parameters (Icorr and Eocp) and physical mass loss confirms that the chemical nature of each beverage significantly governs the corrosion kinetics of S235JR.