Wyniki wyszukiwania - BazTech

1

Large language models as diagnostic interpreters of numeric data from industrial refrigeration systems in industry 4.0.

Pędzik Robert Mirosław

Diagnostyka

|

2026

|

Vol. 27, No. 1

art. no. 2026111

EN

Industry 4.0 enables industrial appliances to generate vast operational datasets, yet their interpretation often remains inaccessible to end-users. This study evaluates Large Language Models (LLMs) as "Interactive Engineering Assistants" to democratize the analysis of raw telemetry from refrigeration units, aligning with EU Data Act (2023/2854) transparency mandates. Data were extracted from the Smart Shop Control (SSC) ekosystem -a proprietary IIoT platform architected by the author, managing over 35,000 active devices. A research gap was addressed regarding the "zero-shot" interpretation of semi-structured CSV sensor data by models optimized for natural language. Two experiments utilized 2-hour telemetry segments in anonymized and overt formats to evaluate SOTA models (GPT-5.1, Copilot, Gemini) under a 'Stateless Human-Orchestrated Sequential Prompting' paradigm. Results demonstrate that LLMs autonomously identify thermodynamic anomalies (e.g., condenser fouling) and correlate them with physical phenomena, establishing a new 'Product Truth' standard. The study introduces the LLM as a 'Mirror of Competence', where diagnostic efficacy reflects the operator's engineering logic. Furthermore, integrating the Unconscious Waste Indicator (UWI) within LLM reasoning identifies hidden energy losses. Public LLM interfaces provide a practical 'Privacy-by-Anonymity' layer, democratizing industrial diagnostics for non-expert stakeholders.

2

FDaaS: fault detection as a service in industry 4.0 environment

Bouhentala Aicha, Mahadaoui Rafik, Mouss Leila-Hayet, Chouhal Ouahiba

Diagnostyka

|

2026

|

Vol. 27, No. 1

art. no. 2026112

EN

Modern industry 4.0 technologies are critical for improving the efficiency of transactions and sharing distributed resources and big data for optimal utilization within manufacturing networks. Therefore, they play a paramount role in various industrial sectors, especially in assets maintenance. The concept of Maintenance as a Service is inspired from the principle of cloud manufacturing, that seeks to provide on-demand manufacturing through the use of manufacturing resources, thereby matching the cloud-computing paradigm’s goal of delivering everything as a service. Reliable maintenance refers to the early detection of faults to prevent assets from suffering unexpected downtime, thus reducing the high cost of maintenance. The main objective of this work is to present a fault diagnosis architecture that includes spans from fault detection to the conception of comprehensive and intelligent maintenance architecture. This concept intends to assure asset reliability and availability at the industry level, while also meeting the evolving demands of Industry 4.0 technologies.

3

Fault detection of slewing bearings in engineering cranes based on adaptive algorithms

Liu Tiejun, Chen Ke

Diagnostyka

|

2026

|

Vol. 27, No. 1

art. no. 2026110

EN

Slewing bearings play a crucial role in the operational efficiency and security of engineering cranes by supporting rotational movements under heavy loads. Over time, these components wear and degrade, making early fault detection critical to avoiding mechanical failures, costly downtime, and security risks. Conventional condition monitoring methods frequently struggle with inconsistent data patterns, sensor noise, and dynamic operating conditions. There is an urgent need for intelligent, adaptive fault detection mechanisms that can precisely predict slewing bearing failures under varying load and operational circumstances. This study aims to build a robust, adaptive fault detection algorithm - Slewing Bearing Fault Detection (SBFDetect) - capable of identifying early signs of faults in slewing bearings using real-time sensor data. The goal is to improve maintenance planning and reduce unexpected failures in engineering cranes. A dataset called the Slewing Bearings Fault (SBF) Dataset was created, which includes key parameters such as vibration intensity, temperature, noise levels, rotation speed, load pressure, lubrication levels, metal debris levels, hours of operation, and sensor drift. The proposed SBF Detect Algorithm starts with preprocessing steps like categorical encoding and normalization, then trains a Random Forest classifier on the processed dataset. The model is assessed using standard performance metrics, such as accuracy, precision, recall, F1-score, and Matthews Correlation Coefficient (MCC). An adaptive update mechanism is also included to enable incremental learning with new sensor data. The SBF Detect algorithm produced promising results on the SBF dataset, with an accuracy of 90.0%, precision of 88.9%, recall of 88.9%, F1-score of 88.9%, and MCC of 0.80. These metrics demonstrate the model's ability to correctly classify faulty and healthy bearings, even with a limited number of samples. The SBF Detect Algorithm provides a practical and scalable solution for predictive maintenance of slewing bearings in cranes. By utilizing adaptive machine learning methods, the proposed technique enhances the dependability and security of crane operations while reducing unplanned downtime.

4

Inter-turn short circuit fault detection in induction motor using XGBoost, KNN and Random Forest

Chahmi Abdelghani, Benkaihoul Said, Naas Toufik Tayeb, Yildirim Özüpak

Przegląd Elektrotechniczny

|

2025

|

R. 101, nr 2

248--253

EN

In this research endeavor, the focus was directed towards investigating a specific fault occurrence within an induction motor, namely an inter-turn short circuit (ITSC), intentionally induced within phase A of the motor. The employed dataset encompassed both correct operational states and instances afflicted with the aforementioned fault, with parameters such as current flows and torque outputs meticulously recorded and analyzed. When employing a methodology rooted in machine learning, a suite of algorithms was applied to discern and identify the presence of the fault. From among the array of algorithms utilized, the notable contenders included Random Forest (RF), k-nearest neighbors (KNN), and Extreme Gradient Boosting (XGBoost), each meticulously trained and tested on the dataset to gauge their efficacy in fault detection. The outcomes obtained in the mentioned study unequivocally demonstrate the superiority of the Random Forest algorithm in terms of accuracy assessment, boasting a remarkable accuracy rate of 99.7%. In the stark contrast, both KNN and XGBoost algorithms exhibited comparatively lower accuracy rates, standing at 96.6% and 96.5%, respectively.

PL

W tym przedsięwzięciu badawczym skupiono się na badaniu konkretnego wystąpienia usterki w silniku indukcyjnym, a mianowicie zwarcia międzyzwojowego (ITSC), celowo indukowanego w fazie A silnika. Zastosowany zbiór danych obejmował zarówno prawidłowe stany operacyjne, jak i przypadki dotknięte wyżej wymienionymi usterkami, przy czym parametry takie jak przepływy prądu i wyjściowy moment obrotowy były skrupulatnie rejestrowane i analizowane. Stosując metodologię opartą na uczeniu maszynowym, zastosowano zestaw algorytmów w celu rozpoznania i zidentyfikowania obecności usterki. Wśród szeregu wykorzystywanych algorytmów godnymi uwagi konkurentami byli Random Forest (RF), k-najbliżsi sąsiedzi (KNN) i Extreme Gradient Boosting (XGBoost), każdy skrupulatnie przeszkolony i przetestowany na zbiorze danych w celu oceny ich skuteczności w wykrywaniu usterek. Wyniki uzyskane w tym badaniu jednoznacznie wskazują na wyższość algorytmu Random Forest pod względem oceny dokładności, który może pochwalić się niezwykłym współczynnikiem dokładności wynoszącym 99,7%. Dla kontrastu, zarówno algorytmy KNN, jak i XGBoost wykazywały stosunkowo niższe wskaźniki dokładności, wynoszące odpowiednio 96,6% i 96,5%.

5

Artificial neural networks based detection of power transmission line faults

Bouraghda Skander, Chelli Zoubir, Bechaouat Mohcen

Przegląd Elektrotechniczny

|

2025

|

R. 101, nr 6

123--127

EN

As manufacturing, transportation, distribution, and usage systems improve, so does the requirement for a dependable power supply. It is crucial to continue providing customers with high–quality, safe electricity in order to satisfy these demands. The biggest danger to the uninterrupted supply of energy is electrical system failures. Electric power system faults are an inevitable issue. Therefore, to limit damage and disturbance to the electrical system, a well–coordinated protection system must be installed to quickly identify and isolate problems. The proper functioning of protective relays depends on the identification of defects in electrical networks. When a fault occurs in a transmission line, the fault current is always greater than the rated load current. Several methods and conventional numerical techniques have been used and proposed for the detection of faults. In this paper, artificial intelligence has been used, namely artificial neural networks (ANN). We develop a program, under the Matlab environment, based on the method of ANN using the sampled values of signal currents & voltages. These allow us to detect different types of shunt faults in transmission lines.

PL

Wraz z udoskonalaniem systemów produkcji, transportu, dystrybucji i użytkowania, wzrasta również zapotrzebowanie na niezawodne zasilanie. Aby sprostać tym wymaganiom, konieczne jest ciągłe dostarczanie klientom wysokiej jakości, bezpiecznej energii elektrycznej. Największym zagrożeniem dla nieprzerwanego zasilania są awarie systemów elektrycznych. Awarie systemów elektroenergetycznych są nieunikn ionym problemem. Dlatego też, aby ograniczyć uszkodzenia i zakłócenia w systemie elektrycznym, należy zainstalować dobrze skoordynowany system automatyki zabezpieczeniowej, aby szybko identyfikować i izolować problemy. Prawidłowe działanie przekaźników zabezpieczeniowych zależy od identyfikacji usterek w sieciach elektrycznych. Gdy w linii przesyłowej wystąpi awaria, prąd zwarciowy jest zawsze większy niż znami onowy prąd obciążenia. Do wykrywania usterek zastosowano i zaproponowano kilka metod i konwencjonalnych technik numerycznych. W tym ar tykule zastosowano sztuczną inteligencję, mianowicie sztuczne sieci neuronowe, mianowicie ANN. Opracowujemy program w środowisku Matlab, oparty na metodzie ANN, wykorzystujący próbkowane wartości prądów i napięć sygnału. Pozwalają nam one wykrywać różne rodzaje usterek bocznikowych w liniach przesyłowych.

6

Fault detection based on interval observers and dynamic event-triggered fault-tolerant control for morphing aircraft

Fu Xingjian, Wang Xiaohan

International Journal of Applied Mathematics and Computer Science

|

2025

|

Vol. 35, no. 3

429--442

EN

In this paper, a fault detection mechanism using interval observers and a robust fault-tolerant control strategy with dynamic event-triggered mechanism are designed for the switched control problem during the transformation process of a morphing aircraft. Firstly, an interval observer design method for the nonlinear switched system is given. It is converted by coordinate transformation into the form of solving Sylvester’s equation in the absence of actuator faults. Secondly, by using the output of the interval observer, the upper and lower bounds of the system output under the no actuator faults condition are constructed, and the design of the fault detection mechanism is achieved by monitoring whether the system output exceeds the bounds. Thirdly, in order to save communication resources, a robust fault-tolerant control strategy based on dynamic event-triggered mechanism is designed. Based on fault detection results, two different controllers are utilized for switched control, ensuring the boundedness of the closed-loop system signal, and conditions for the asymptotic stability of the closed-loop system are offered. Finally, a nonlinear model of morphing aircraft system with variable wing curvature is used to verify the validity of the designed scheme.

7

Fault Detection in Power Transmission Lines : Comparison of Chirp-Z Algorithm and Machine Learning Based Prediction Models

Alpsalaz Feyyaz

Eksploatacja i Niezawodność

|

2025

|

Vol. 27, no. 4

art. no. 203949

EN

Fast and accurate detection of faults in power transmission lines is of great importance for the safety and continuity of power systems. This study develops a predictive model using chirp-z transform and machine learning algorithms to locate single-phase-ground faults. During the study, 39 different fault locations were modelled, current and voltage signals of these locations were analysed and frequency spectra were obtained. The fault signals were decomposed into their components using the modal transformation matrix and then spectral analysis was performed using the Chirp-Z algorithm. The resulting spectra were used as input data for the prediction algorithms. Gradient Boosting Ensemble, Support Vector Regression and Random Forests algorithms were used for fault prediction and the performance of the models was compared. The accuracy of the models was evaluated using various metrics. The results show that the Gradient Boosting Ensemble model has the lowest error rates and the highest accuracy, which is important for early fault detection, maintenance and repair processes.

8

Intelligent fault detection and classification in overcurrent protection systems based on artificial neural networks

Hadid Ibrahim N., Ibrahim Mohammed Ahmed

Diagnostyka

|

2025

|

Vol. 26, No. 4

art. no. 2025415

EN

An ANN-based intelligent overcurrent relay is proposed for simultaneous fault detection and fault-type classification in a selected subsection of the IEEE-9 bus transmission system (Bus-7-Bus-8). Two neural network modules are implemented: the first performs fault detection and directly issues the trip command based on three-phase current features, while the second classifies the fault types into AG, BG, CG, AB, BC, CA, and ABC categories. Fault current signals are generated in MATLAB/Simulink under diverse operating conditions, including variations in fault resistance, location, and inception angle. The detection network achieved a correlation coefficient of R ≈ 0.993, whereas the improved classification network achieved R ≈ 0.998, demonstrating a substantial enhancement in accuracy and generalization. Time-domain tests demonstrated consistently faster tripping performance compared with the conventional inverse-time relay namely, improvements of 0.6 ms (AG), 0.7 ms (BG), 1.25 ms (CG), 0.6 ms (AB), 1.5 ms (BC/BC-G), 1.6 ms (AC/ACG), and 0.5 ms (ABC/ABC-G). These results confirm the superior dynamic response and adaptability of the proposed intelligent relay, highlighting its suitability for modern protection applications and its potential as a foundation for next-generation smart-grid relaying systems.

9

Explainable fault detection and diagnosis based on an IDEOA as applied to an industrial process

Benbrahim Meriem, Hamaidi Brahim, Haouassi Hichem, Mahdaoui Rafik, Mouss Leila-Hayet

Diagnostyka

|

2025

|

Vol. 26, No. 2

art. no. 2025203

EN

Even with all measures approved by industrial sector specialists to avoid faults leading to major accidents, this field still suffers from some issues. Therefore, the safety and reliability of these industrial systems become necessary, leading to focus more on anticipating fault occurrence by giving fault detection and diagnosis a high priority. To solve this problem, a large set of reliable methods has been developed. Machine learning-based methods have gained significant importance as they have achieved promising results. However, the black-box nature of the generated fault detection models has restricted their investigation by users. Thus, explainable models aim to show features that influence the detection model decision. In this study, an Improved Discrete Equilibrium Optimizer Algorithm (IDEOA), which aims to solve different discrete optimization problems, was proposed to generate a rule-based fault detection model easily explainable by reading its classification rules. To this end, the Opposition-Based Learning (OBL) strategy is adopted in the IDEOA to avoid being stuck in local optima. A key contribution of this study is the novel application of the methodology to the Tennessee Eastman Process. The result of this study is a fault diagnosis model that consists of 16 rules, six of them belong to normal operating conditions and the rest reveal fault occurrence (F4). Then, an accuracy value is calculated to assess the effectiveness of our approach by contrasting it with other algorithms described in the literature. The findings indicate that the proposed approach outperforms other methods.

10

Machine learning-based fault detection in transmission lines: A comparative study with random search optimization

Özüpak Yıldırım

Bulletin of the Polish Academy of Sciences. Technical Sciences

|

2025

|

Vol. 73, nr 2

art. no. e153229

EN

Regular and fast monitoring of transmission line faults is of immense importance for the uninterrupted transmission of electrical energy. Rapid detection and classification of faults accelerate the repair process of the system, reducing downtime and increasing the efficiency and reliability of the power system. In this context, machine learning stands out as an effective solution for transmission line fault detection. In this study, fault detection is performed using machine learning techniques such as decision trees, logistic regression, and support vector machines. Random search hyperparameter optimization was applied to improve the performance of the models. The models were trained and tested with data from fault-free and faulted cases. While the support vector machines model showed the lowest performance with 74.19% test accuracy, the logistic regression model achieved 97.01% test accuracy. The decision tree model showed the best performance with low error rates. Error measures such as root mean square error (RMSE) and mean absolute error (MAE) were also used to evaluate the predictive power of the models. This research demonstrates how machine learning-based methods can be effectively used in the detection of transmission line faults and presents the performance of different algorithms in a comparative manner.

11

Description and comparison of fault detection algorithms based on a selected building automation device

Włostowska Sandra, Kawa Bartłomiej, Borkowski Piotr

Bulletin of the Polish Academy of Sciences. Technical Sciences

|

2025

|

Vol. 73, nr 3

art. no. e153832

EN

The article compares selected classification algorithms and those dedicated to anomaly detection. The models used temperature measurements in four rooms simulated in the MATLAB Simscape environment as test signals. The empirical part of the work consists of two parts. In the first one, an example data from the simulated building heating model object, models were built using unsupervised and supervised machine learning algorithms. Then, data from the facility was collected again with changed parameters (failures occurred at times other than the test ones, and the temperature patterns differed from those recorded and used to train the models). The algorithm effects and test signals (temperature changes) were saved in the database. The results were presented graphically in the Grafana program. The second part of the work presents a solution in which the analysis of the operating status of the heating system takes place in real time. Using an OPC server, data was exchanged between the MATLAB environment and the database installed on a virtual machine in the Ubuntu system. The conclusions present the results and collect the authors’ suggestions regarding the practical applications of the discussed classification models.

12

Metoda unikania fałszywych alarmów w „skojarzonych” systemach automatycznego monitorowania procesów produkcyjnych w zakładach wytwórczych przemysłu chemicznego

Mirosław Tomasz, Grabowski Paweł, Mirosław Marcin

Przemysł Chemiczny

|

2024

|

T. 103, nr 11

1312--1315

PL

Omówiono najważniejsze problemy związane z wywoływaniem fałszywych alarmów, które są związane z usterkami i awariami systemów monitorowania urządzeń stosowanych w procesach technologicznych. Zaproponowano koncepcję pozwalającą znacząco zmniejszyć liczbę takich zdarzeń, która opiera się na idei obserwacji „pozaprocesowych” parametrów pracy ciągów technologicznych.

EN

The most important problems related to false alarms raised by supervision systems operating on production lines were analyzed. These alarms were very often caused by failures or malfunctions of diagnostic devices such as transducers, conditioners and amplifiers, involved in monitoring processes. A concept of a method for reducing the problem of false alarms, based on observations of non-process parameters, was proposed. It can be applied to production processes in the chemical industry.

13

System wizyjny do analizy położenia sieci trakcyjnej względem odbieraka prądu

Konieczka Adam, Stachowiak Dorota, Garczarek Adam, Dominiak Wojciech

Przegląd Elektrotechniczny

|

2024

|

R. 100, nr 7

169--173

PL

W artykule przedstawiono system wizyjny do analizy położenia sieci trakcyjnej względem odbieraka prądu. Jest on przeznaczony do montażu na dachu pojazdu kolejowego. Wyposażono go w kamerę i minikomputer Raspberry Pi 3B+, który analizuje zarejestrowany obraz oraz wykorzystuje moduł GPS do rejestracji miejsc, w których wykryto nieprawidłowe ustawienie sieci trakcyjnej. Wyniki przeprowadzonych badań, także z wykorzystaniem pojazdu szynowego, wskazują na możliwość szerokiego zastosowania proponowanego systemu.

EN

The article presents a vision system for analyzing the position of the catenary in relation to the current collector. This system is designed to be mounted on the roof of a railway vehicle. The system is equipped with a camera and a Raspberry Pi 3B+ minicomputer that analyzes the recorded image and uses a GPS module to record the locations where incorrect alignment of the overhead contact line was detected. The results of the tests carried out, also using a railway vehicle, indicate the possibility of widespread use of the proposed system.

14

Diagnosis of a stand-alone photovoltaic installation by the analytical redundancy relationship method (ARR)

Bait Fateh, Samia Latreche, Mabrouk Khemliche, Lokmane Boulemzaoud

Przegląd Elektrotechniczny

|

2024

|

R. 100, nr 4

78--82

EN

Solar energy is one of the most important renewable sources to replace fossil fuels for electric power generation. like other energy systems, it is susceptible to several faults and anomalies during operation, which can reduce its performance and productivity. This paper provides a diagnostic method based on the Analytical Redundancy Relation (ARR) method for detecting faults in off-grid photovoltaic systems. This method is based on the calculation of residues between the modules and the PV system, which enables more accurate faults detection and anomalies. Additionally, a comprehensive structure to diagnose the effect of rainfall on PV installations has been designed. The proposed approach is evaluated using a MATLAB/Simulink simulation model, and the results demonstrate the effectiveness of the ARR method in detecting faults in PV systems. The developed structure is also able to diagnose the impact of rainfall on the performance of PV installations. The results obtained can help improve the performance and reliability of PV systems, which can contribute to the wider adoption of solar energy as a clean and sustainable energy source.

PL

Energia słoneczna jest jednym z najważniejszych źródeł odnawialnych zastępujących paliwa kopalne w procesie wytwarzania energii elektrycznej. podobnie jak inne systemy energetyczne, jest on podatny na szereg usterek i anomalii podczas pracy, co może zmniejszyć jego wydajność i produktywność. W artykule przedstawiono metodę diagnostyczną opartą na metodzie analitycznej relacji redundancji (ARR) służącą do wykrywania uszkodzeń w systemach fotowoltaicznych poza siecią. Metoda ta opiera się na obliczeniu pozostałości pomiędzy modułami a systemem PV, co umożliwia dokładniejsze wykrywanie usterek i anomalii. Dodatkowo zaprojektowano kompleksową strukturę do diagnozowania wpływu opadów atmosferycznych na instalacje PV. Zaproponowane podejście zostało ocenione przy użyciu modelu symulacyjnego MATLAB/Simulink, a wyniki wykazały skuteczność metody ARR w wykrywaniu usterek w systemach fotowoltaicznych. Opracowana konstrukcja umożliwia także diagnozowanie wpływu opadów atmosferycznych na pracę instalacji PV. Uzyskane wyniki mogą pomóc w poprawie wydajności i niezawodności systemów fotowoltaicznych, co może przyczynić się do szerszego przyjęcia energii słonecznej jako czystego i zrównoważonego źródła energii.

15

An Analysis of power system faul Classifier using neural network

Kumar Venkatesh, Sujitha S., Aarthi C., Nirmala M., Samanvita N., Garapati Durga Prasad, Kumar Vinoth, Manjunatha B.

Przegląd Elektrotechniczny

|

2024

|

R. 100, nr 7

96--100

EN

Fault is obviously a significant phenomenon for energy transmission in the distribution system because of the potentially harmful consequences that finally lead to economic crises. In order to verify their sustainability error experience, MATLAB and Simulink analyse the 3-phase power system in this article. An intelligent expert like a neural network may easily identify the defect that may have happened in the transmission line and categorize transmission issues on the power supply using artificial neural network (ANN). ANN is used to categories problems and generate a change status indication for the protection relay. This work proposes design strategies for fault recognition, classification, and isolation supported by state-of-the-art artificial intelligence and signal processing. Three-phase current and voltage from one end are taken as inputs in the proposed scheme. The Various simulations and signal analysis are performed in MATLAB environment.

PL

Zwarcie jest oczywiście zjawiskiem istotnym dla przesyłu energii w systemie dystrybucyjnym ze względu na potencjalnie szkodliwe skutki, które ostatecznie prowadzą do kryzysów gospodarczych. Aby zweryfikować swoje doświadczenia związane z błędami w zakresie zrównoważonego rozwoju, MATLAB i Simulink analizują w tym artykule 3-fazowy system zasilania. Inteligentny ekspert, taki jak sieć neuronowa, może z łatwością zidentyfikować usterkę, która mogła wystąpić w linii przesyłowej i sklasyfikować problemy z transmisją w zasilaczu za pomocą sztucznej sieci neuronowej (ANN). SSN służy do kategoryzacji problemów i generowania wskazania stanu zmian dla przekaźnika zabezpieczeniowego. W pracy zaproponowano strategie projektowania rozpoznawania, klasyfikacji i izolacji uszkodzeń wspierane przez najnowocześniejszą sztuczną inteligencję i przetwarzanie sygnałów. W proponowanym schemacie jako dane wejściowe przyjmuje się prąd trójfazowy i napięcie z jednego końca. Różne symulacje i analiza sygnałów wykonywane są w środowisku MATLAB

16

An AutoML-based comparative evaluation of machine learning methods for detection of eccentricity faults in induction motors by using vibration signals

Irgat Eyüp, Ünsal Abdurrahman

Metrology and Measurement Systems

|

2024

|

Vol. 31, nr 4

831--848

EN

Induction motors (IMs) are the most widely used electrical machines in industrial applications. However, they are subject to various mechanical and electrical faults. Eccentricity faults are among the common mechanical faults of IMs. This study compares the performance of four commonly used machine learning (ML) methods, including k-nearest neighbours (k-NN), decision tree (DT), support vector machine (SVM), and random forest (RF) along with the statistical features in detecting eccentricity faults of IMs with an automated machine learning (AutoML) model. The aim of using AutoML in this study is to fully automate the process of detection of eccentricity faults of IMs by selecting the classifier with the highest accuracy rate and shortest computation time along with the most effective feature(s). The eccentricity fault analysed in this study was experimentally implemented in the laboratory. Three-axis vibration signals were collected for healthy and eccentricity-faulty IMs. In the proposed study the three-axis vibration signals are pre-processed to determine the statistical features that are used as input to the ML methods. The proposed study offers the best ML method among the four studied algorithms and the need for expert knowledge of ML and eccentricity fault detection. The proposed AutoML model offers the DT method along with the z-axis rms feature for the highest accuracy rate and the shortest computation time in detecting the eccentricity fault.

17

Improving the exploitation efficiency of cooling equipment by monitoring operational parameter

Pędzik Robert Mirosław

Diagnostyka

|

2024

|

Vol. 25, No. 2

art. no. 2024205

EN

Cooling equipment is widely used in industry, commerce and households. Due to their widespread use, they are responsible for the consumption of a significant amount of electricity. They are subject to degradation and various types of damage. Most often, their energy efficiency decreases and electricity consumption increases. Practice shows that even a specialist service is unable to diagnose damage at an early stage of its development. The paper presents a comparison of continuous monitoring of the temperature of the cooling chamber as a utility standard, with constant monitoring of the temperature of the cooling chamber and electricity consumption of a professional refrigeration cabinet with a built-in condensing unit. The comparative analysis was intended to confirm the thesis about unconscious waste resulting from assessing the correct operation of the device based on limited information. The experiment showed an increase in daily electricity consumption on average by over 30% during the period of unconscious exploitation of the device in a state of failure and an increase in daily electricity consumption on average above 300% during the period of conscious exploitation of the device in a state of failure, but still at an acceptable level of temperature of cooling chamber.

18

Fault detection and diagnosis of squirrel cage asynchronous motor using a new COMSOL multiphysics software 6.0.

Djaidir Benrabeh, Nail Bachir, Rouibah Abdelkader

Diagnostyka

|

2024

|

Vol. 25, No. 1

art. no. 2024102

EN

In this paper, the diagnosis of faults in squirrel cage asynchronous motor and experimental analysis process are presented. Currently there are several simulation tools, that lets users analyze and interpret the behavior of their devices. Based on this, there is a lot of researches that is working on developing models, to detect and classify 3-phase asynchronous motor faults, significantly in the early stages. This work proposed design and experimental analysis established in Comsol Multiphysics 6.0 , which implements finite element analysis software (FEM) for detecting and diagnosing broken bar rotors of this types motors and its practical application. In this case, the post processor of the COMSOL-Multiphysics makes it possible to visualize in 2D the various magnetic and mechanical quantities. Through the curves of the magnetic flux density and analysis distribution of the field with magnetic induction lines, we can draw some conclusions, where we proposed an strategy, for detecting and diagnosing faults consistent with the structure of the software.

19

Detection and identification of fault in transmission lines based on ANN

Abdullah Ahmed G., Ibrahim Mohammad A., Ibrahim Waseem Kh.

Diagnostyka

|

2024

|

Vol. 25, No. 2

art. no. 2024206

EN

Along with power transmission lines' efficiency, another crucial factor in electrical power transmission networks is reliability, which guarantees power transmission stability. One of the crucial and essential tasks for maintaining the continuity and stability of power transmission in transmission networks Capacity without any significant failures is identifying errors and malfunctions in power transmission lines as soon as possible. The goal of this article is to develop and apply ANN technology to overcome the obstacles faced by the electrical power transmission network. In order for the ANN to learn useful patterns and features from raw current measurements, pre-processing and feature extraction techniques are used during the training process. Real-time applications can benefit from the ANN's architecture, which is optimized for high accuracy, quick response times, and scalability. To validate the performance of the ANN-based fault detection system, extensive simulations are conducted using data from different transmission line scenarios, including various fault types that short-circuit. The results demonstrate the capability of the ANN model to accurately detect and classify faults, as well as disconnect the power grid after detect any fault. The results showed the accuracy and high speed of the proposed method using a neural network compared to traditional methods.

20

Detection and classification of air gap eccentricity fault in induction machine using artificial intelligence techniques

Bentrad Moutaz Bellah, Ghoggal Adel, Bahi Tahar

Diagnostyka

|

2024

|

Vol. 25, No. 4

art. no. 2024412

EN

This paper employs artificial intelligence to diagnose induction machine health by detecting air gap eccentricity under varied conditions. It addresses Model-Based Method and conventional MCSA techniques limitations, requiring extensive model knowledge. The proposed technique relies on stator current signals, simplifying data acquisition. Using Root Mean Square and raw data using the three phases of stator current from a multi winding model of a squirrel cage induction machine. The study emphasizes on employing classification and regression tasks for supervised learning as a non-model-based approach by applying several models and classifiers to choose the best one for the monitoring task. This approach allows online diagnosis, detecting defects early, even under weak load conditions by conduction a multiclassification technique for each class of the dataset. The paper's strength lies in its holistic analysis of signal fluctuations, categorizing faults based on nature and location. Overall, the proposed algorithm for the classification which is Decision Trees achieved an overall accuracy surpassing 80% against other classifiers, and for the regression task Random Forest outperformed by delivering the least values of loss error with 0.014 using mean square error evaluation metric and achieving a 98.6% accuracy.