Bearing Fault Diagnosis in Induction Motor Using Hybrid CNN Model

Abstract

Induction motors (IMs) are the prime movers for the industries. The availability of an efficient electrical drive has aided in the widespread application of IMs in different sectors, including mining, cement, textile, and many more. Bearings are the critical components of the motors. The bearing failure may cause severe accidents and production losses. The timely detection of the bearing fault is essential for the minimum downtime. Researchers have used conventional machine learning techniques for the bearing fault detection in motors. However, these approaches require input features, and selecting efficient features poses a big challenge. Deep learning (DL) algorithms have recently captured the interest of researchers all over the world. DL algorithms like convolutional neural networks (CNNs) can automatically execute feature extraction and selection. This paper proposes a hybrid CNN-based model in combination with support vector machine for bearing fault detection in IMs. Various bearing faults, such as inner race fault, outer race fault, and ball defect, have been considered in the proposed work. The proposed method has efficiently detected various bearing faults. The proposed approach has achieved a mean accuracy of more than 99%. Python was used for all of the analysis and programming. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.