International Journal of Advanced Statistics and IT&C for Economics and Life Sciences

Ground fault location analysis on transmission lines using MATLAB/Simulink environment

Doru Pastina — 2025-12-05

With the continued expansion of the power systems, the accuracy of fault diagnosis on power transmission lines is crucial to ensure their safe and stable operation. A single-phase-to-ground fault can occur anywhere along an overhead transmission line. The amplitude of the fault current and its variation over time depend on the distance between the transformer substation and the fault location. This paper introduces a MATLAB/Simulink programmatic modelling framework for a faster simulation of fault scenarios in a given power system. It also presents the single-phase-to-ground fault current signal features for various fault locations. Both the amplitude of the current and the voltage decreases when the distance from the busbars of the station increases, with the current declining steeper than the voltage.

Subsummation automata in conceptual graphs

Daniel-Cristian Crăciunean — 2025-12-05

One of the most important constraints faced by the development of artificial intelligence systems is the limited computing power. The logic problems involved in the construction of artificial intelligence systems are most often NP-complete in terms of computational complexity. In the case of conceptual graphs, logical inference is based on the subsummation relation between conceptual graphs. In this paper, we use homomorphisms of conceptual graphs as a model for the subsummation relation. Therefore, the problems related to the existence of relations between two conceptual graphs such as homomorphisms, isomorphisms or generalization, specialization relations become fundamental problems. Although all these decision problems are, in general, NP-complex, they can be efficiently solved in many cases, by certain strategies, without making too much compromise in terms of generality. In this paper, we propose a method to find and manipulate homomorphisms of conceptual graphs based on the concept of finite automaton. Thus, using the finite automaton to recognize predicate homomorphisms and moving a large part of the complexity of finding a homomorphism to the knowledge base construction phase represents our main contribution in this work.

Implementation of Heuristic Algorithms for Simulating Crisis Situations in the Medical System

Andrei Patrausanu — 2025-12-05

Healthcare systems face significant challenges during crises such as pandemics or mass-casualty events, where resource shortages and patient overflow require rapid, optimized decisions. This paper proposes a simulation-based approach to model and improve hospital resource allocation using a heuristic method—Genetic Algorithms (GA). The aim is to explore how intelligent algorithms can support decision-making under pressure by assigning patients to limited ICU beds and available doctors, considering constraints such as treatment duration, medical priority, and resource availability.
The core method involves evolving allocation strategies over multiple generations, using tournament selection, crossover, mutation, and fitness-based evaluation to optimize both resource usage and patient coverage. The simulation is implemented as a desktop application in C#, with a SQL Server database and an interactive GUI that allows users to run scenarios, configure parameters, and visualize outcomes.
Compared to a First-Come-First-Served (FCFS) baseline, the GA consistently achieves higher efficiency, treating more high-priority patients and reducing resource bottlenecks. The original contribution lies in the dual-resource optimization model and its integration into a flexible, user-friendly tool. Results demonstrate that heuristic-driven simulations can support better planning and training in emergency healthcare environments.

Achieving Clinical Reliability in Suicide Risk Detection: A Low-Resource Benchmark of RoBERTa vs. DistilBERT

Stefania Berghia — 2025-12-05

Detecting suicidal ideation in social media text is a critical public health objective, demanding high-accuracy, deployable models. This study addresses the challenge of achieving clinical reliability within severe hardware constraints. We conduct a comparative fine-tuning benchmark of two Transformer models, DistilBERT and RoBERTa, for binary classification of suicidal risk. The models were optimized on a balanced, 10,000-sample subset of the Reddit Suicide Detection Dataset under CPU-only, low-resource constraints. Prediction robustness is achieved through a simple selection process that chooses the output with the highest confidence score from the two models. RoBERTa achieved a peak F1-Score of 97.4% and 97.40% accuracy, substantially outperforming DistilBERT (94.1% F1-Score). Crucially for safety-critical applications, error analysis confirmed RoBERTa's ethical superiority by achieving a 40% reduction in False Negatives compared to DistilBERT. The validated framework establishes that high-performance, ethically robust risk detection is feasible under resource limitations, enabling the safe integration of the final classification system with a constrained LLaMA 3 conversational module for proactive support.

Sensorless Vector Control of a Surface Permanent Magnet Synchronous Motor

Gabriela Craciunas — 2025-12-05

This paper analyzes the operation of the surface permanent magnet synchronous motor (SPMSM) in a sensorless configuration, with the main objective of speed estimation. It is presented a vector control strategy independent of position sensors, implemented through a voltage inverter with prescribed currents. Advanced algorithms, including the Gopinath observer, are used to accurately determine the rotor speed and position. This plays an essential role in increasing the stability and accuracy of the control system, by reducing estimation errors and ensuring a rapid response to load variations. Performance validation was performed through simulations in the MATLAB/Simulink environment, with the results demonstrating both the accuracy and reliability of sensorless control under various operating conditions. The conclusions highlight the efficiency of the proposed method and its relevance for the development of modern electric drive technologies.

Deepfake Voice Detection for Underrepresented Languages A Romanian Case Study

Marco Olescu — 2025-12-05

The rapid advancement of artificial intelligence and its subsequent application in deepfake media present a significant global security concern. Despite this widespread issue, effective detection solutions are notably absent for less commonly used languages. This paper proposes a potential solution for identifying generated audio in Romanian. The solution centres on an SVM-based algorithm, previously demonstrated to perform effectively in English language tests, adapted with a dataset specifically tailored to the Romanian language. The resulting language-specific model exhibits better performance in differentiating between authentic and synthetic Romanian audio, thereby offering an improvement over general-purpose, Anglocentric systems. This constitutes an effective strategy for developing localised solutions for a language with limited resources.

SDR receiver for power measurement with internal noise compensation

Iulian Boulean — 2025-12-05

This paper presents a solution for implementing a power measurement receiver made with PlutoSDR. The paper proposes the dynamic elimination of the contribution of internal noise to the measured signal level by compensating with the level of internal noise produced by the receiver in the adjacent RF band. Our approach is based on the fact that the internal noise level in the working band is approximately equal to that in the frequency band in the immediate vicinity. As far as we know, our work is the first to use the idea of measuring the noise in an adjacent band to estimate the internal noise of the measuring instrument. Another aspect necessary to ensure a complete measurement solution that does not require pre-measurement calibration concerns the compensation of the nonlinearities of the reception path. The compensation was automatically performed by modifying the amplification of the reception path depending on the working frequency. For high levels of the measured CW signal (-50 dBm ... -10 dBm) the solution allows obtaining measurement errors of less than 0.55 dB. From the perspective of measurement uncertainty, with a value of U=±1 dB (k=2), the instrument with the method used fits very well into the requirements for pre-compliance EMC measurements. For levels lower than -60 dBm, the error starts to increase rapidly, so that at values of -70 dBm the signal can no longer be measured.

Forecasting and reliability

Ioan Lucian diodiu — 2025-12-05

No precise law governs the design of electrical power systems, and due to the errors inherent in the results obtained, confidence in forecasting methods is low. In reality, a large number of possible solutions may arise, and one aims to determine the limits within which they will be found — more precisely, within what boundaries a certain proportion of them lies.

GRADIS: A Comprehensive System for Academic Contribution Tracking and Evaluation at Lucian Blaga University of Sibiu

Cantemir Mihu — 2025-12-05

Academic institutions face increasing pressure to systematically track and evaluate faculty contributions across teaching, research, and administrative activities. This paper presents GRADIS (Grila de evaluare a Activitatii Didactice si Stiintifice - Grid for Evaluation of Teaching and Scientific Activity), a comprehensive web-based system developed for Lucian Blaga University of Sibiu (ULBS) to address these challenges. GRADIS provides a unified platform for managing academic contributions across multiple dimensions including scientific publications, teaching activities, student guidance, and administrative responsibilities. The system integrates with external academic databases such as Crossref, Web of Science, and Scopus to automate data collection and validation, while supporting flexible reporting and evaluation workflows tailored to Romanian academic standards. Built using Spring Boot and Vue.js, GRADIS demonstrates how modern web technologies can be effectively applied to academic administration, achieving significant improvements in data quality, process efficiency, and compliance with institutional evaluation frameworks.

IoT System for Milk Dispenser Monitoring and Management

Claudia Banciu — 2025-12-05

An Internet of Things (IoT)-like application is an advanced system for automation and data analytics that leverages communication networks, sensors, and artificial intelligence to provide a comprehensive solution for a product or service. The development of IoT technologies has facilitated advancements in automated systems across various industries like healthcare, agriculture, smart cities, smart homes, and transportation. This paper presents an IoT-based system designed for the real-time monitoring and management of a milk dispenser. The system integrates an ESP32 microcontroller, sensors for temperature and milk level measurement, and cloud storage for data analysis and visualization. By using machine learning algorithms, the system predicts milk consumption trends and detects potential dispenser blockages, enhancing operational efficiency. Experimental evaluation showed that the Random Forest and Gradient Boosting regression models achieved high predictive accuracy, with the best configurations reaching a coefficient of determination (R2) of 0.95, a Mean Absolute Error (MAE) as low as 1.05 liters, and a Root Mean Squared Error (RMSE) of 1.32 liters, confirming the strong reliability of the proposed approach.

Integrating Multi-View features into diagrammatic languages

Daniel-Cristian Crăciunean — 2025-12-05

In this paper, we define diagrammatic languages by metamodels represented by categorical sketches. Then, a model, corresponding to a categorical sketch, is a Set-value functor M:𝓖→Set, where 𝓖 is the sketch graph, and which respects the constraints imposed by the categorical sketch. A view of a model M, is another model 𝓥 in the diagrammatic language that maps to M by a natural injective transformation. This is a theoretically satisfactory definition but does not provide us with solutions to actually find these views. In this paper, we will present a practical way to construct a satisfactory set of views for a given model. Moreover, this mechanism can be implemented generically, at the modeling language level, and then used for each model separately. The categorical mechanisms that we use to solve this problem are representable functors and the Yoneda lemma. This is also the main contribution of this paper to solving problems related to multi-view modeling.

Optimization of Romanian Identity Documents Processing

Delia-Victoria Milașcon — 2025-12-05

This paper presents the development of an application based on computer vision and deep learning aimed at the automatic recognition of data from Romanian national identity documents. Its primary goal is to optimize the processing of such documents within public and private institutions. Developing such a system involves several challenges. These include the variability in the visual appearance of documents caused by wear and tear, scanning angles, camera quality, and ambient lighting conditions; the difficulty of accurately recognizing Romanian diacritics; and the correct identification of textual fields based on their positioning within the document. The system must be sufficiently robust to adapt to these variations while maintaining a high level of accuracy.

The Open-loop BLDC motor simulation using the Motor Control blockset from Matlab library

Alina Viorel — 2025-12-05

Special electrical machines play an increasingly important role in modern applications. They are designed to address situations that were previously handled by conventional machines, offering improved efficiency and sustainability. The growing need to save resources and protect the environment has encouraged the development of new strategies in motor design and control. This paper presents a BLDC behaviour using datasheets obtained from a PMSM motor. Both BLDC (Brushless DC) and PMSM (Permanent Magnet Synchronous Motor) are permanent magnet machines and their mathematical models are closely related, differing mainly in how the back-EMF is shaped (trapezoidal for BLDC, sinusoidal for PMSM). Key parameters like stator resistance, inductance, and permanent magnet flux linkage are common to both machines. This means PMSM data can be repurposed for BLDC simulations without needing a completely separate dataset. MATLAB/Simulink, provide PMSM blocks that can be adapted for BLDC by changing the EMF profile. This paper presents an open-loop control scenario for simulating the BLDC motor performance using the Motor Control Blockset from the MATLAB/Simulink library starting from a PMSM open loop control strategy.

Deep Learning Segmentation for stem cells images

Catalina Neghina — 2025-12-05

In this study, we investigate the application of deep learning techniques for automatic segmentation of the Fluo-N2DH-GOWT1 dataset, which consists of time-lapse grayscale images of rat neural stem cells. We employ the DeepLabV3+ semantic segmentation framework with a ResNet-18 backbone, chosen for its balance between accuracy and computational efficiency on relatively small biomedical datasets. To improve generalization and robustness, we apply data augmentation strategies including rotation, scaling, shear, and reflection. The performance of the proposed model is evaluated using standard metrics such as F1-score, Intersection of Union, Precision and Recall. Experimental results demonstrate that the ResNet-18–based DeepLabV3+ achieves reliable segmentation of stem cells, effectively distinguishing cells from the background.

The Emergent Strategies in Multi-Agent Systems

Cantemir Mihu — 2025-12-05

This article aims to present an overview of the emerging protocols and strategies used for achieving scalability of performance- and efficiency-centric multi-agent systems as this field is rapidly developing and focus is shifting towards architectures employing multiple, tool-enabled models that need to both split tasks, working concurrently, and follow step-based problem-solving strategies, while maintaining effective communication throughout the system.

Building a power meter for cycling

Macarie BREAZU — 2025-12-05

The main goal of this research was to design and implement a simple power meter for cycling. The designed power meter had to be affordable, so we chose to use the crank arm solution. In order to achieve the objectives, we had to measure the tangential force applied on the crank arm using tensometric stamps. Also, we had to measure the cadence based on zero-crossing of the applied force, and to transmit data to external applications using Bluetooth Low Energy (BLE). The hardware implementation is based on BF350-3AA tensometric stamps, HX711 amplifier and Seeed XIAO nRF52840 Sense microcontroller, while the software implementation is based on the HX711_MP.h and bluefruit.h libraries. Experiments prove that the proposed solution (cost approx. 29€) meets the requirements. After calibration, our system achieved a deviation of ~10% compared to Tacx virtual power data, confirming its feasibility for amateur training applications.

A Deep Dive into Wavelet-Based Neural Architectures for Inpainting Forgery Detection

Adrian Barglazan — 2025-12-05

This paper provides a comprehensive analysis of several neural network architectures designed for the detection of generative image inpainting. The central challenge addressed is the increasing sophistication of inpainting technologies, driven by generative AI, which renders traditional forensic methods based on simple statistical or visual artifacts obsolete. This paper investigates a feature-centric hypothesis: that visually seamless inpainting manipulations introduce consistent, subtle, and detectable artifacts in specific high-frequency and statistical domains. The primary domain of interest is the feature space defined by the Dual-Tree Complex Wavelet Transform (DTCWT).

The core of this report details the methodology and evaluation of six novel deep learning architectures, each custom-designed to operate on these wavelet-domain features. These architectures explore diverse design philosophies, including dense feature aggregation (Custom UNet++), explicit global context integration (UNet++ with Global Average), recurrent spatial modeling (ConvLSTM), unsupervised generative anomaly detection (Variational Autoencoder), and channel-specific processing (StackDeepAll).

A rigorous experimental evaluation was conducted on a custom inpainting detection dataset. The quantitative results demonstrate the clear superiority of the StackDeepAll architecture. This model, which processes each of the 12 DTCWT channels through independent, parallel UNet-like encoders before fusion, achieved a state-of-the-art Intersection over Union (IoU) of 0.8126 in validation. This result significantly surpassed all other proposed models, including a stable Variational Autoencoder (0.3711 IoU) and overfitting-prone ConvLSTM models (0.4514 IoU). The findings establish that for high-dimensional, engineered features like complex wavelets, a channel-specific processing paradigm that avoids premature feature-blending is a powerful and highly effective design strategy. This feature-centric approach is validated as a potent alternative to contemporary end-to-end spatial-domain models.

Office Ergonomics and Ambiental Environment Monitoring System

Aida-Mihaela Manta — 2025-12-05

More and more people suffer from back health issues and other afflictions due to their position at the desk, their working ambiental environment and growing sedentarism. The proposed IoT system aims to help people and companies with such concerns and create a healthier environment to work in. The proposed system is placed on a person’s desk. The system uses a camera and a computer vision model to analyse a person’s position at the desk. It also uses a set of sensors to assess the ambiental environment’s quality in accordance with ESH (Environmental, Safety, and Health) recommendations. The information analysed can be accessed by the end-user through a visual report that emphasizes the deviations from the recommended standards. The possibility to faster detect non-conformities in working environments and bad habits in ergonomics can improve the quality of the working space and delay or prevent health issues in the long term.

The Wireless Controller for Modbus RS485

Mihai-Liviu Stroila — 2025-12-05

The purpose of this paper is to propose a modern and efficient solution for controlling and analysing RS-485 communication networks, overcoming the limitations of traditional methods. In high-voltage systems, isolating RS-485 nodes is essential to protect both circuits and human operators from high voltages and transient phenomena. Conventional solutions, such as high-speed optocouplers and transformers, are effective but require numerous external components, resulting in bulky devices with high production costs. The proposed solution introduces a rechargeable, battery-powered control device operated wirelessly via a smartphone or laptop. This ensures complete isolation for the operator while simplifying the system’s design.

Compared to traditional galvanic isolation methods, this solution significantly reduces production costs and device size, with the total size being comparable to a computer mouse. It offers a cost-effective, compact, and future-ready alternative for industrial applications requiring efficient communication and safety in high-voltage environments. Additionally, the integration of IoT technology enhances functionality by enabling remote monitoring, real-time control, and data analysis. This IoT-driven approach supports predictive maintenance and scalability, making it ideal for modern industrial and automation systems.

Keywords: RS-485, communication networks, IoT, industrial applications.