An Experimental Comparative Evaluation of Machine Learning Techniques for Motor Fault Diagnosis Under Various Operating Conditions

The diagnosis of electric machines, such as induction motors, is one of the key tasks that needs to be performed to guarantee their right operation as electromechanical energy converters in most industrial facilities. The ability to reliably identify a mechanical fault occurrence before it becomes catastrophic can reduce risks related to the productive chain. Recently, different intelligent approaches have been proposed to develop feature-based methods for automatic rotor fault diagnosis of induction motors. This paper provides an experimental comparative evaluation of different machine learning techniques for rotor fault identifications. The classifiers are tested with data obtained under different operating conditions of the ones used to train them, as it is usual in industry. The input information is obtained from current signals of an induction motor with two states of rotor bar degradation under two preestablished load levels.