Experimental Evaluation of Multi-Sensor AI-Based Fault Diagnosis for Induction Motors in Industrial Applications

Abstract

Induction motors are critical components in industrial systems, and unexpected failures can lead to significant production losses and maintenance costs. This paper presents an experimentally validated multi-sensor fault diagnosis approach for induction motors using vibration and stator current signals under practical operating conditions. A laboratory test bench was developed to simulate common industrial faults, including bearing defects, rotor abnormalities, and stator winding faults, across multiple load levels. Conventional machine learning techniques, namely artificial neural networks and support vector machines, were employed as baseline classifiers, while an adaptive hybrid convolutional neural network–long short-term memory (CNN–LSTM) model was used to improve fault classification robustness. The proposed approach achieved a maximum classification accuracy of 98.4 %, with stable performance across varying load conditions and repeated experimental trials. The results demonstrate that integrating vibration and current measurements enhances diagnostic reliability compared to single-sensor methods. The study highlights the practical applicability of adaptive AI-based diagnostic systems for industrial predictive maintenance, offering improved fault detection capability while maintaining feasibility for real-world deployment.