Wyniki wyszukiwania - BazTech

1

Fault Diagnosis of a Ship’s Permanent Magnet Propulsion Motor Based on CBAM and Multi-Input CNN

Yan Guohua, Wang Bingyang, Jiang Jiawei, Kang Jingran, Yu Xinglin, Li Weibin, Cheng Xianguo

Polish Maritime Research

|

2026

|

nr 1

107--119

EN

The fault diagnosis of marine propulsion motors has been an important focus of attention in the marine industry. Various types and numbers of sensors have been used to monitor and diagnose faults in permanent magnet propulsion motors, and the comprehensive application of multi-sensor signals has played a key role in improving fault diagnosis performance, but the issue of how to efficiently exploit this multi-source information remains a difficult problem. In this paper, we propose a multi-sensor-signal feature-level fusion method based on a multi-input convolutional neural network and a CBAM attention mechanism, which fully utilises the end-to-end learning capability of deep learning and the interpretability and domain expert knowledge of traditional methods. A synchrosqueezing wavelet transform is used to extract the highresolution feature information of the current and vibration signals; the multi-input neural network extracts the high-level abstract features in the current and vibration signals; the CBAM attention mechanism is introduced to make the network more targeted, to deal with the key feature information; and a Bayesian optimisation algorithm is used to automatically determine suitable combinations of hyperparameters for training of the network. A fault test of a permanent magnet synchronous motor shows that the diagnostic accuracy of the proposed method reaches 99.08%, a value 1.29% higher than a scheme without the CBAM attention mechanism, with an increase in the detection time for each sample of only 2.33 ms. Our approach also has better anti-noise interference ability and generalisation performance.

2

Local Entropy Selection Scaling-extracting Chirplet Transform for Enhanced Time-Frequency Analysis and Precise State Estimation in Reliability-Focused Fault Diagnosis of Non-stationary Signals

Zhi Shaodan, Niu Yuli, Ma Limei, Wu Hengshan, Shen Haikuo, Wang Tianyang

Eksploatacja i Niezawodność

|

2026

|

Vol. 28, no. 1

art. no. 205977

EN

Under diverse conditions, the vibration signals of complex rotating machinery exhibit non-stationary behavior, multi-component characteristics, closely spaced frequencies, and non-proportionality, posing challenges to conventional time-frequency analysis (TFA) methods. These limitations hinder accurate instantaneous frequency (IF) estimation and time-frequency representation (TFR) construction, directly impacting machinery fault diagnosis. As such, we propose the Local Entropy Selection Scaling-Extracting Chirplet Transform (LESSECT), which optimizes entropy-based chirp rate (CR) selection to match non-proportional fundamental frequencies. By adaptively selecting multiple CRs at the same time center, LESSECT enhances TFR resolution and energy concentration, leading accurate IF identification. Experimental validation on bat echolocation, bearing fault, and planetary gearbox signals shows its superior performance in resolving nonproportional, closely spaced IFs. This significantly improves state estimation and enhances machinery diagnostics, contributing to greater system reliability.

3

Intelligent diagnosis and fault prediction of rock breaking process in milling machines based on big data analysis

Wang Jingde, Duan Jian, Sun Qingsong, Han Lu, Tang Jian, Li Jian

Diagnostyka

|

2026

|

Vol. 27, No. 1

art. no. 2026105

EN

This study develops an intelligent method for diagnosing and predicting disc cutter deflection wear faults during rock excavation processes. The proposed method integrates wavelet time-frequency graph encoding with an Inception-Bidirectional Gated Recurrent Unit (Bi-GRU) model through big data analysis techniques. The approach utilizes large-scale vibration signal data to effectively extract time-frequency characteristics of non-stationary signals through wavelet time-frequency graphs, enhancing the model's fault identification capability under complex working conditions. The one-dimensional vibration signal is mapped into a two-dimensional time-frequency graph and input into the Inception-BiGRU model. The spatial features are extracted by using the multi-scale convolution of the Inception module. Meanwhile, the dynamic evolution law is captured by combining the Bi-GRU bidirectional time series modeling to achieve the deep integration of spatial-temporal features. The Inception-BiGRU model demonstrates excellent robustness in multi-condition tests, maintaining a diagnostic accuracy consistently exceeding 98% and reaching a maximum of 99.35%. In multi-class fault identification, the accuracy for normal and severely deflected cutters reaches 99%. Meanwhile, minor misjudgments occur in cases of mild and moderate deflection, with overall identification precision exceeding 97%. A comparative evaluation against various encoding methods and traditional algorithms demonstrates that the Inception-BiGRU model outperforms baseline models in both accuracy and stability, achieving an average accuracy of 98.63% with a variance of only 0.03. The proposed model improves average accuracy by 1.84% and reduces variance by 50% compared to baseline models. The research results verify the effectiveness and engineering application potential of the proposed method in diagnosing and predicting disc cutter faults. This study provides technical support for advancing the intelligent operation and maintenance of milling machines while offering a feasible path for future multimodal sensor fusion and early fault warning.

4

Bearing condition monitoring: a review of feature extraction in temporal, spectral and joint temporal-spectral domains

Thajeel Mohammed Mutar, Al-Bugharbee Hussein Razzaq

Diagnostyka

|

2026

|

Vol. 27, No. 1

art. no. 2026103

EN

Rolling element bearings are critical components in rotating machinery. Their failure can lead to catastrophic consequences. Therefore, effective condition monitoring is very necessary to avoid the occurrence of such unexpected breakdowns and ensure safety. This review focuses on the recent advances in vibration-based feature extraction techniques for bearing fault diagnosis. More than 70 peer-reviewed journal articles published since 2019 are analysed. The analysis covers feature extraction techniques in the temporal domain, spectral domain, and joint temporal–spectral domain. Then, the reviewed features are critically assessed in terms of their diagnostic sensitivity, robustness to noise, and applicability under different operating conditions. The review aims to adopt a featurecentric and decision-oriented perspective and provides guidance for selecting suitable health indicators. It can serve as a useful reference for researchers and practitioners working in rolling element bearing fault diagnosis.

5

Transformer fault diagnosis method based on multilevel acoustic information

Li Xuan, Wu Peng, Shao Jian, Liang Jiabi, Li Qun, Lu Yuncai, Wang Tonglei

Archives of Electrical Engineering

|

2026

|

Vol. 75, nr 1

23--40

EN

To more accurately obtain the feature information embedded in the acoustic pattern of transformers, a transformer fault diagnosis method is proposed based on multilevel acoustic information of 14 state types. In this method, a parallel dual-channel fault diagnosis model, CNN-BiLSTM-Transformer, is established. First, the modified Mel inversion coefficients and Mel spectrograms are extracted from the original acoustic pattern data. The modified Mel inversion coefficients and Mel spectrograms are then input into the parallel dual-channel model. In the first channel, a convolutional neural network model is used to extract the feature information of maps. In the second channel, a bidirectional long- and short-term memory network and a Transformer encoder are used to partially extract the temporal features in the MFCCs. Finally, the temporal features extracted from the two channels are fused through multimodal fusion for training. The experimental results show that the proposed diagnostic method can achieve an average accuracy of 99.5% in multiple fault diagnosis. Compared with current mainstream acoustic single-channel diagnostic models, the diagnostic rate of this model is improved by an average of 4.8%, exhibiting higher accuracy and robustness.

6

Abnormal noise detection of three-phase induction motor using linear predictive coding and neural networks

Glowacz Adam, Czechowski Miroslaw, Brumercik Frantisek, Gutten Miroslav, Glowacz Witold

Archives of Electrical Engineering

|

2026

|

Vol. 75, nr 1

103--116

EN

The paper presents an abnormal noise detection method for a three-phase induction motor. The following motor conditions were analyzed: healthy (H), motor with one broken rotor bar (1BRB), motor with two broken rotor bars (2BRB), and motor with three broken rotor bars (3BRB). The dataset was split into 48 training samples (12 per class) and 168 test samples (42 per class) for the training and evaluation of the neural networks. Linear predictive coding (LPC) was used for feature extraction. The next three original neural networks were proposed for classification: Neural Network V01, V02, and V03. The authors of the paper also used ResNet-50. The proposed approach achieved a recognition efficiency of 100%.

7

A fuzzy logic-based multi-sensor diagnostic system for traction motor bearings in railway applications

Ahmadov Heybatulla, Manafov Elshan, Guliyev Huseyngulu, Huseynov Farid

Transport Problems

|

2025

|

T. 20, z. 2

73--84

EN

DOI: 10.20858/tp.2025.20.2.06 Keywords: traction motor bearings; fault diagnosis; fuzzy products; linguistic variables; expert system Heybatulla AHMADOV1, Elshan MANAFOV2*, Huseyngulu GULIYEV3, Farid HUSEYNOV4 A fuzzy logic-based multi-sensor diagnostic system for traction motor bearings in railway applications This article focuses on the diagnosis of the bearings of the traction motors of electric railway and subway trains. One of the main sources of mechanical failures in a traction motor is its bearings. The failure of traction motor bearings, the factors that cause these failures, and the diagnostic methods for detecting them are investigated. At this time, faults in traction motor bearing monitoring systems are determined only by temperature. In this work, it is proposed to use a system with temperature, vibration, and noise to determine the technical condition of bearings. Such a multi-parameter system, unlike traditional ones, will help determine specific defects at an early stage. The expert system’s model, based on fuzzy logic and diagnostic parameters, can accurately predict the likelihood of bearing faults in real-time under changing operating conditions. A fuzzy expert system represents knowledge in the form of fuzzy productions and linguistic variables. The expert system model was developed using the Mamdani fuzzy inference algorithm of the Fuzzy Logic Toolbox package in the MATLAB computing environment. The application of fuzzy logic in generating a knowledge base and inference processes enables the formalization of a process for evaluating technical conditions based on incomplete, faulty, and potentially erroneous information and for making decisions about fault identification.

8

Experimental results of inverter faults impact on the performances of SPWM and SVPWM control techniques

Horri Noureddine, Boudinar Ahmed Hamida, Aimer Ameur Fethi, Khodja Mohamed Amine

Przegląd Elektrotechniczny

|

2025

|

R. 101, nr 9

117--126

EN

Induction motors are widely used in industry due to the development of electronic power switch technology used in variable speed drives and controls. Indeed, several control strategies have been developed to control the speed of these motors such as sinusoidal PWM control and space vector PWM control techniques. Several studies have shown that space vector PWM control is more efficient than sinusoidal PWM technique based on the low THD obtained. However, this advantage can be affected by the failure of electronic switches such as IGBTs. In this paper, a time and spectral analysis are applied on experimental signals in order to investigate the impact of the IGBT open circuit (OC) fault on the performance of these control strategies.

PL

Silniki indukcyjne są szeroko stosowane w przemyśle ze względu na rozwój technologii elektronicznego przełącznika mocy stosowanej w napędach o zmiennej prędkości i układach sterowania. Rzeczywiście, opracowano kilka strategii sterowania w celu kontrolowania prędkości tych silników, takich jak sinusoidalne sterowanie PWM i techniki sterowania PWM wektorem przestrzennym. Kilka badań wykazało, że sterowanie PWM wektorem przestrzennym jest bardziej wydajne niż technika sinusoidalnego PWM w oparciu o uzyskane niskie THD (Total Harmonic Distortion, Współczynnik Zawartości Harmonicznych). Jednak ta zaleta może zostać naruszona przez awarię przełączników elektronicznych, takich jak IGBT. W niniejszym artykule, zastosowano analizę czasową i widmową sygnałów eksperymentalnych w celu zbadania wpływu awarii obwodu otwartego IGBT (OC) na wydajność tych strategii sterowania.

9

AUV Fault Diagnosis Based on Multidimensional Temporal Classification Transformer

Chen Weifeng, Xu Xiaoyan, Chen Mingzhi, Tarasiuk Tomasz

Polish Maritime Research

|

2025

|

nr 4

97--106

EN

Autonomous underwater vehicles (AUVs) play a critical role in marine resource exploration, maritime patrol, and rescue operations, requiring reliable fault diagnosis for robust operation. This paper proposes a novel AUV fault diagnosis method based on the multidimensional temporal classification transformer (MTC-Transformer) deep learning algorithm. The MTC-Transformer enhances traditional transformer architectures by optimising them for multidimensional time-series data processing and fault classification. It features adaptive automatic feature extraction, superior multi-modal data handling, and improved scalability. Key innovations include gated fusion for combining features from outlier-processed data and kernel principal component analysis-reduced time-series data, a dedicated fault feature amplification mechanism, and a multi-layer perceptron head for classification. Trained and validated on the ‘Haizhe’ small quadrotor AUV fault dataset using double-layer randomised sampling to ensure generalisation, the model achieves exceptional accuracies of 99.51% (training) and 99.44% (validation). Comparative experiments demonstrate its superiority over WDCNN, LSTM-1DCNN, and other benchmarks in handling variable-length sequences and capturing long-range dependencies, confirming its high accuracy, robustness, and practical efficacy for AUV fault diagnosis.

10

Temporal Constrained Dynamic Uncertain Causality Graph for Root Cause Analysis of Intermittent Faults

Ma Qiqi, Long Jiang, Shi Xiangnan, Liu Zun, Guo Yangming

Eksploatacja i Niezawodność

|

2025

|

Vol. 27, no. 1

art. no. 192169

EN

The diagnosis of intermittent faults is crucial in the field of maintenance support. Unfortunately, most existing studies focus on the analysis of intermittent faults in single components, ignoring the more complex intermittent failures of equipment functions caused by the coupling of multivariate anomalous states in the fault propagation process. Existing diagnostic methods based on fault propagation models, which mainly focus on one-dimensional temporal or logical relationships, fall short in representing and reasoning about intermittent faults caused by temporal and state coupling. In this paper, a Temporal Constrained Dynamic Uncertain Causality Graph (TC-DUCG) model is developed to fill this gap and effectively model intermittent faults. Our model not only considers the probability of fault propagation among variables but also integrates temporal constraints. It also presents a diagnostic reasoning process to investigate potential causes of intermittent faults. An illustrative example is proposed to demonstrate the effectiveness of the proposed method in diagnosing intermittent faults.

11

Integrated Multi-Scale Analysis and Advanced Prototype Model for Early Detection of Gearbox Failures Using Infrared Thermal Image Data Under Dynamic Conditions

Ge Jian, Zhou Di, Zhuang Xiao, Wang Xiaomin, Xiang Jiawei

Eksploatacja i Niezawodność

|

2025

|

Vol. 27, no. 2

art. no. 194151

EN

To address the issues of low data quality and poor adaptability in deep learning methods for infrared image analysis in gearbox fault diagnosis, this paper introduces an enhanced deep prototype network model (MSPNet). This model employs a multi-scale strategy to improve fault diagnosis accuracy and algorithm generalization, especially with small sample sizes. First, infrared image data of six fault types under five operating conditions are collected using a rotating test bed. Gaussian noise is added to simulate real operating conditions. Next, the fault data are processed using a multiscale module to extract multiscale fault features and reduce feature value fluctuations. Finally, the proposed model is used to process the image data and is experimentally compared with five other algorithms. The experimental results demonstrate that the proposed method outperforms the other algorithms under various operating conditions.

12

Improved Graph Convolutional Neural Networks-based Cellular Network Fault Diagnosis

Gao Zongzhen, Liu Wenlai

Eksploatacja i Niezawodność

|

2025

|

Vol. 27, no. 2

art. no. 194672

EN

To solve the problem of upstream and downlink interference in cellular networks, a graph convolutional neural networks-based novel fault diagnosis method for semi-supervised cellular networks is proposed. In the research design method, the extreme gradient enhancement technique is first used to enhance the fault diagnosis feature data of cellular networks. Then, the graph convolutional neural network is used to train and learn the fault diagnosis feature dataset of cellular networks, achieving fault diagnosis prediction of cellular networks. In the process of training the cellular network fault diagnosis model, data augmentation techniques were used to enhance the training level of the model, while Bayesian networks were used for pre diagnosis to improve the diagnostic accuracy of the modified model. The experimental results show that the cellular network fault diagnosis model constructed in the study can achieve a classification accuracy of 90% for training samples during training and testing, while other models can only achieve a maximum of about 85%. The model constructed by the research can achieve a diagnostic accuracy of over 90% in the practical application of cellular network fault diagnosis, while taking only 6 seconds. This algorithm can diagnose faults in complex cellular network environment, which has high accuracy and practicability, and can effectively improve user experience.

13

Fault diagnosis method of automobile rolling bearing based on transfer learning and improved DenseNet

Lu Xinxin, Xiao Yang

Eksploatacja i Niezawodność

|

2025

|

Vol. 27, no. 2

art. no. 194675

EN

Aiming at the problems caused by ignoring the time series characteristics, the scarcity of labeled data and the long diagnosis time in the fault diagnosis of one-dimensional vibration signals of automobile bearings, a new method combining improved DenseNet and transfer learning is proposed in this study. This method uses Recurrent Plot (RP) technology to convert one-dimensional vibration data into two-dimensional images to fully tap the potential value of time series. By optimizing the DenseNet network structure, the fault features are extracted effectively.Lightweight network design and MobileViT Attention mechanism are used to reduce the number of parameters and improve computing efficiency. With the help of transfer learning technology, the fault features in the source domain are transferred to the target domain, which solves the problem of cross-condition diagnosis and greatly reduces the diagnosis time. The experimental results show that the proposed method can improve the accuracy of fault identification and diagnosis efficiency, and achieve accurate classification.

14

Advanced Sparse Filtering-Based Domain Adaptation for Fault Diagnosis in Variable Working Conditions

Zhou Ziyou, Chen Wenhua, Qin Jiajun

Eksploatacja i Niezawodność

|

2025

|

Vol. 27, no. 2

art. no. 194181

EN

Traditional domain adaptation (DA) methods often encounter challenges with cross-domain feature extraction and the precise assessment of domain differences. To overcome these limitations, we introduce the Enhanced Sparse Filtering-Based Domain Adaptation (ESFBDA) method. This method distinguishes itself by enhancing sparse filtering (SF) with the integration of row-column normalization and a cosine penalty, specifically designed to minimize feature loss - a critical issue in existing DA techniques. Additionally, we employ Bootstrap resampling to refine domain distribution alignment, a novel step that boosts feature similarity and effectiveness in DA. This integrated approach ensures more accurate feature extraction, which is crucial for the classifier's fault detection capability. In our study, through two distinct experiments on WT-planetary gearbox fault diagnosis and bearing fault diagnosis, the ESFBDA method demonstrated remarkable accuracy, significantly surpassing traditional approaches and showcasing its robust applicability across varied diagnostic scenarios.

15

A Multi-scale Attention Mechanism Diagnosis Method with Adaptive Online Updating Based on Deep Learning under Variable Working Conditions

Lei Xue, Lu Ningyun, Jiang Bin, Wang Cunsong, Chen Chuang

Eksploatacja i Niezawodność

|

2025

|

Vol. 27, no. 1

art. no. 192975

EN

With the advance of industrial systems, the online equipment fault diagnosis has encountered many challenges such as data drift and data imbalance under varying operating conditions, thus making stable and accurate diagnosis increasingly critical. Considering the above issues, a multi-scale attention mechanism diagnosis method with adaptive model that can be updated based on deep learning has been proposed. The method is composed of four main steps: training the multi-scale offline diagnosis model, transferring the parameters of the offline model, assessing the degree of data drifting, and adaptively updating the diagnostic model. A data balance strategy with adaptive weight balances both inter-class and intra-class data. The method updates the diagnostic model flexibly according to online data status, to reduce the impact of data drifting. The method was verified on a bearing test rig, which can reproduce the common bearing faults under variable working conditions. The experimental results have shown that the proposed method can accurately and reliably identify the bearing faults.

16

A Federated Transfer Fault Diagnosis Method for Cross-Domain and Incomplete Data

Ge Yang, Zhang Fusheng, Ren Yong

Eksploatacja i Niezawodność

|

2025

|

Vol. 27, no. 2

art. no. 194182

EN

Big data-driven intelligent fault diagnosis methods for device rely on a large amount of labeled data for centralized training. However, in practical engineering, it is difficult for a single client to collect enough labeled sample data, which is one of the reasons that limit the application of these methods. In fact, multiple clients often use similar devices and collect fault data separately, so joint multi-client collaborative fault diagnosis modeling can solve the problem of data scarcity, but this poses great challenges to data privacy protection. In this paper, we propose a federated transfer fault diagnosis method based on federated learning for cross-domain incomplete data. The proposed method only exchanges the parameters of the local training model, which achieves the privacy protection of the client’s local data. We construct a multi-client collaborative learning framework to address the problem of weak generalization ability caused by the lack of terms in single client training samples. We also propose a targeted semi-supervised fine-tuning strategy based on relative distance to reduce the probability of negative fine-tuning of out-of-distribution samples and improve the accuracy of diagnostic models. The results of cross-condition and cross-equipment experiments demonstrate that the proposed method has obvious advantages over the existing fault diagnosis methods.

17

Mechanical fault diagnosis system based on genetic algorithm optimization and multi-sensor data fusion

Hu Caodi, Zhao Yumeng

Diagnostyka

|

2025

|

Vol. 26, No. 4

art. no. 2025410

EN

Industrial machinery frequently experiences machine failures during operation. Promptly diagnosing these failures can improve industrial operational efficiency to a certain extent. Existing research has many shortcomings. This paper considers two issues: data noise and the weight of multi-sensor data fusion, and designs a Genetic Algorithm-based Multi-Sensor Data Fusion Diagnosis (GA-MSDFD) method. This method first uses manually defined indicators to filter noise and uses a custom method to eliminate it. Feature extraction is then performed on the evolved data, and a genetic algorithm is used for multi-objective feature selection. This algorithm has inherent advantages over other machinery because its design considers the impact of noisy data. Experimental results show that our model achieves a fault diagnosis accuracy of 96.8%, far exceeding several other machinery models. The model also far outperforms these models in noise robustness, noise resistance, and convergence performance. The proposed model is of great significance for the maintenance of industrial machinery operations.

18

Fault diagnosis and fault tolerant control of asynchronous motor speed sensor combined with different speed estimators

Li Qiubo

Diagnostyka

|

2025

|

Vol. 26, No. 2

art. no. 2025202

EN

Accurate speed and flux estimation are important conditions for achieving high-performance and low-cost control. Therefore, this study first constructs a mathematical model and space vector pulse width modulation control method for asynchronous motors. Then, a simple speed estimator, an extended Kalman filter speed estimator, and a full-order magnetic flux observer speed estimator are established in the speed estimation module. Finally, based on the Euler voting algorithm, a fault diagnosis and fault-tolerant control method for speed sensors is designed. The results showed that under low-speed conditions, the average mean square errors of the speed estimators of the simple, extended Kalman filter, and full-order magnetic flux observer were 0.7969, 0.9134, and 2.2526, respectively. The first two speed estimators had better performance, while under medium to high-speed conditions, the latter two speed estimators had a lower average mean square error and better performance. When various faults occurred, the research method could quickly determine the best performing speed estimator for feedback and effectively display the speed fluctuations caused by the faults. Finally, it smoothly switched to the speed sensorless mode and controlled the speed error within -5r/min-5r/min.

19

Fault diagnosis analysis of gas turbine engine based on fuzzy algorithm

Peng Yana, Hai Wenjiong, Xue Peng

Diagnostyka

|

2025

|

Vol. 26, No. 1

art. no. 2025109

EN

To solve the problems of low efficiency and difficult feature extraction in traditional fault diagnosis methods, this study proposes an optimized Fuzzy C-Means clustering algorithm for diagnosing and analyzing gas turbine engine faults. This algorithm mainly introduces subtraction clustering, penalty factors, and data weights on the basis of the original fuzzy C-means clustering algorithm, thereby improving the generalization ability of the algorithm model and the credibility of the results. The optimized fuzzy C-means clustering algorithm had the highest level of accuracy value, with a value of 95.67%, which was 11.79% higher than the average accuracy of other algorithms. Meanwhile the optimized Fuzzy C-Means clustering algorithm improved the accuracy values of KNN, BP, SVM, and fuzzy C-means clustering algorithms by 19.65%, 12.26%, 3.55%, and 11.70%. The training set accuracy of the optimized fuzzy C-means clustering algorithm under four engine states was at the highest level, with an average improvement of 15.5%, 25%, 24%, and 16% in accuracy. The optimized fuzzy C-means clustering algorithm achieved an accuracy of 90.39% in the test set, with an average improvement of 16.13% in accuracy. The membership classification results indicated that the optimized fuzzy C-means clustering algorithm had a membership degree of 1.

20

Dynamically enhanced weighted network acoustic and photoelectric GIS fault diagnosis based on attention mechanizm

Fu Hui, Lu Yongling, Wang Zhen, Li Yujie, Gong Duohu, Sun Weihao

Diagnostyka

|

2025

|

Vol. 26, No. 3

art. no. 2025302

EN

In power systems, the normal functioning of gas-insulated switchgear (GIS) is essential for the security of the electrical grid. However, when a single signal is used for discharge detection and diagnosis, it will be interfered. Through joint analysis of different signals, fault diagnosis can be more accurately performed. Therefore, to address this problem, this paper proposes a dynamically enhanced weighted network model (AMB-DEWNM) based on the attention mechanism. The model first extracts fault features from the PRPD spectra of UHF, optical and ultrasonic signals through a multi-scale convolutional neural grid. Furthermore, a two-tier focus module is introduced to enhance fault characteristics that are insensitive to changes in operating conditions. Finally, a new dynamic enhanced weighted voting strategy (DEWVS) is designed. This strategy constructs a diagnostic performance index matrix by considering the diagnostic accuracy and misclassification rate of the base model to dynamically adjust the voting weight of each base model. distribution to obtain more reliable collaborative diagnostic results. Test outcomes demonstrate that the error detection precision of the AMB-DEWNM system is notably enhanced. Compared with other advanced network models, the diagnosis accuracy is as high as 95.28%. It has high stability and robustness, and provides fault detection and maintenance for GIS. strong support.