Enhancing Fault Detection in Induction Motors using LSTM-Attention Neural Networks

Induction motors carry notable importance in modern machinery and industrial equipment. Hence, the imperative to establish an early fault detection system for discerning operational states and potential faults in these motors is evident. This paper introduces a pioneering methodology for fault detection in induction motors by leveraging the power of Long Short-Term Memory (LSTM) networks and attention mechanisms within a neural network framework. By combining the temporal sequence analysis capabilities of LSTMs with the selective focus feature of attention mechanisms, the proposed architecture enhances the accuracy and timeliness of fault detection, paving the way for improved machinery health monitoring and predictive maintenance. The proposed approach is rigorously evaluated alongside established methods, including simple Artificial Neural Networks (ANN) and LSTM networks to showcase the strengths of the LSTM-Attention approach. This comprehensive analysis reveals the additional advantages of attention mechanisms and positions the LSTM-Attention architecture as a robust solution for detecting faults in induction motors.