2026· IEEE Access· Vol 14, pp. 99482-99500· 1 citation
Computer Science
Abstract
The fault diagnostics in Brushless Direct Current (BLDC) motor drive system is critical for operational safety and system lifespan in propulsion system applications. However, signature parameters such as currents, voltages, speed, and torque have provided nonlinear behavior, which limits the usefulness of traditional model-based approaches. This research provides a deep learning based intelligent system to monitor the failures in marine propulsion system. Each signal feature is represented as a structured token, with a specific class token used to collect global contextual information. The proposed model captures both local temporal dynamics and global inter-feature interdependence multi-layer self-attention processes, allowing for the thorough modeling of complex fault patterns. The framework is tested on datasets including healthy conditions and fault conditions, such as motor faults and drive switch failures. The impact of signal-to-noise ratio in the different case structure on the various signals are investigated. With clean signal conditions (SNR = 100 dB), the proposed model achieves a motor fault classification accuracy of 89%, with a weighted precision of 0.91, recall of 0.89, and F1-score of 0.89 on the hardware testbed. Drive switch fault classification under clean simulation conditions achieves an accuracy of 97%, with a macro-averaged F1-score of 0.98 including perfect classification (F $1=1.0$ ) for four out of six switch fault types. Robustness evaluation under additive white Gaussian noise reveals motor fault accuracies of 89%, 80%, and 27% at SNR levels of 100 dB, 50 dB, and 10 dB respectively, and drive fault accuracies of 93%, 51%, and 56% at the corresponding SNR levels. Compared to CNN and LSTM based baselines, the proposed model improves diagnostic accuracy by (4–8%) with a significantly reduced false positive rate, confirming its suitability for intelligent real-time fault diagnosis in marine propulsion systems.
To maintain the safety and reliability of industrial applications, it is important to analyse faults in three-phase induction motors. Traditional methods have difficulty in identifying the complex characteristics of errors in three-phase induction motors because the signals associated with three-phase induction motors are often noisy and exhibit non-linear dynamics. In this paper, a comprehensive method to analyse errors in three-phase induction motors are proposed that utilises advanced signal preprocessing, feature extraction, and Deep Learning (DL) methods to address these issues. Initially, an Entropy-Driven Empirical Wavelet Filter (EDEWF) are applied to clean the raw motor signal, and a Min-Max scaling is used to normalise the signal to create useful features. Next, an integrated hybrid model combining a Radial Basis Function Neural Network (RBFNN) with a Multi-Head Adaptive Transformer (MHAT) are used for capturing local non-linear characteristics and global context. The features extracted from both models are combined into one feature vector and passed through a fully connected layer for classification. The Starfish Optimization Algorithm (SFOA) is implemented to adjust the hyperparameter of the proposed integrated hybrid model to improve the overall performance. The results obtained from the proposed method showed that it achieved a greatest accuracy, precision, recall, F1-Score of 0.9960 and robustness in the diagnostic of three-phase induction motors compared to traditional diagnostic methods, with respect to classifying the motor states as normal, healthy, and Broken Rotor Bar (BRB) faulted.
Killamsetti Vijeta, Perumal B Sri Suyambulinga, R. T. Kumar et al.· ITEGAM- Journal of Engineeri...· 0 citations
Industrial motor fault diagnosis is significantly affected by environmental noise and sensor degradation, which reduce the reliability of conventional deep learning models. This paper proposes a multi-sensor fault diagnosis framework based on Multi-Head Attention and LSTM networks enhanced with a Reliability-Gated fusion mechanism. The proposed framework dynamically evaluates the reliability of current, vibration, and stray-flux signals before feature fusion. Experimental results demonstrate superior diagnostic performance compared with conventional LSTM-Attention models, with an accuracy of 97.67% when the motor is operating under full load. Furthermore, under severe noise conditions (20 dB SNR), the proposed model maintains 91.5% accuracy. The reliability-gating strategy preserves robust diagnostic performance during total sensor failure, achieving an accuracy of 92.4%, while maintaining sensitivity to incipient winding faults with only 2% severity. The proposed framework provides a fault-tolerant solution for predictive maintenance in industrial environments.
Jaafar Falih Hasan, A. R. Hussein· European Conference on Artif...· 0 citations
The stable detection of faults in smart power grids is essential when focusing on the stable operation and the reduction of the downtime. In this paper, the author suggests a hybrid deep learning-based model that combines convolutional neural networks (CNN) and long short-term memory (LSTM) with explainable artificial intelligence (XAI) to detect and classify faults accurately and interpretably. The model aims at capturing the spatial and time-varying attributes of multivariate electrical signatures such as voltage, current, and frequency changes. A combination of real time sensor measurements and simulated fault conditions are used which includes several fault classes including LG, LL, LLG, and three phase faults. Experimental evaluation demonstrates that the proposed model achieves a classification accuracy of 97.84%, with an F1-score of 97.08%, outperforming conventional CNN and LSTM models by more than 3%. The framework is also able to work in a noisy environment with a stable performance of less than 2% performance decrease, and inference latency of 18 ms, which can be deployed in real-time. Besides, the combination of SHAP and attention processes improves the interpretability through the detection of the crucial features that lead to fault prediction. The findings suggest that the suggested solution is a powerful, scalable, and clear solution when it comes to intelligent fault management in a contemporary smart grid.
Udit Mamodiya, Divyanshu Sinha, I. Kishor et al.· Scientific Reports· 0 citations
Deep learning frameworks, such as Simple Recurrent Neural Network, Long Short-Term Memory, Long Short-Term Memory, Bidirectional LSTM, Bidirectional LSTM, and Gated Recurrent Unit, are employed for the detection and categorization of IM.
R. Sooraj, S. Ramu, R. Sitharthan et al.· Scientific Reports· 0 citations
Hydropower systems operate under complex hydraulic, mechanical, and electrical conditions, and their operational reliability is closely related to equipment safety and maintenance efficiency. With the increasing availability of monitoring data, intelligent algorithms have been gradually introduced into hydropower fault prediction and condition assessment. This review summarizes recent developments in machine learning, deep learning, physics-informed neural networks, and digital twin-assisted approaches for hydropower fault analysis. Representative studies and typical applications are discussed, with attention given to their diagnostic performance, data dependence, and applicability under different operating conditions. Existing studies indicate that conventional machine learning methods still perform effectively in limited-sample scenarios, while deep learning models are more suitable for extracting complex features from multi-source monitoring signals and time-series data. Hybrid approaches combining physical mechanisms with data-driven analysis have shown potential for improving model robustness and reliability. In addition, digital twin frameworks provide a possible way to integrate real-time monitoring, fault diagnosis, and operational assessment within a unified platform. Despite recent progress, several challenges remain, including limited fault data, model interpretability, and differences in operating conditions among hydropower stations. Future studies are expected to place greater emphasis on multi-source data fusion, improved model adaptability, and the integration of physical knowledge with intelligent algorithms, supporting more reliable fault analysis and condition monitoring in hydropower systems.
Xu Guo, Endi Fu, Han Hu et al.· Frontiers in Water· 0 citations
Wind turbine gearbox fault diagnosis is essential for ensuring the safe and stable operation of wind energy systems. However, in industrial scenarios, limited fault data and highly complex, non-stationary operating conditions lead to significant distribution shifts, which reduce the generalization ability of traditional methods. To address this issue, we propose a task-oriented fault diagnosis framework based on large language models. Time-domain and frequency-domain features are extracted from vibration signals to construct structured representations. Inspired by few-shot learning, a support–query task construction strategy is introduced, reformulating classification as conditional task reasoning. Low-Rank Adaptation is adopted for parameter-efficient fine-tuning of the pre-trained model. In addition, a global coverage and local balance task sampling strategy is designed to enhance task diversity and mitigate sample imbalance. Experiments on a wind turbine gearbox dataset show consistent performance across multiple backbone models, including LLaMA2-7B, LLaMA3-8B, Qwen-7B, and Baichuan-7B. LLaMA3-8B achieves the best performance, with 98.00% accuracy and an F1-score of 0.9798. These results demonstrate strong robustness and cross-model generalization under few-shot and complex operating conditions.