Identification and Classification of Imbalance Faults in Induction Motors Using Surrogate Models

Induction motors, with their robust structures, low maintenance costs, and high reliability, have a wide range of applications in the industry. However, these motors are susceptible to electrical and mechanical faults caused by environmental and operational conditions. Fault types include issues such as bearing problems, stator winding faults, and rotor bar breakages, with mechanical imbalance faults emerging as a critical issue that adversely affects motor performance. This study aims to compare the performance of surrogate models (RBF and KRG) with deep learning models (RNN, GRU, LSTM) for diagnosing imbalance faults in induction motors. For this purpose, the experimentally collected current (Ia, Ib, Ic) and vibration (X, Y, Z) signals were analyzed in the frequency domain, and the features obtained through FFT were used in the classification processes for three classes (Healthy, DA_1, DA_2). According to the results, the RBF model exhibited the best performance with 97.78% accuracy and 97.64% precision, while the KRG model achieved a notable success with 93.89% accuracy and 93.71% precision. In contrast, the highest-performing deep learning models, RNN and LSTM, demonstrated lower performance with 87.22% accuracy and 87.23% precision. The RBF model outperformed the highest-accuracy deep learning model, RNN, by achieving a 12.11% improvement in accuracy and an 11.93% improvement in precision, proving to be a superior tool for diagnosing imbalance faults. Particularly, the RBF model achieved 100% accuracy in the DA_2 class, effectively distinguishing it from other classes due to its distinct features. These findings demonstrate that surrogate models offer an effective solution for diagnosing faults in induction motors by providing high accuracy and precision with limited data requirements and low computational cost.