Advanced Diagnosis of Air Gap Eccentricity in Three-Phase Induction Motor Using DWT Decomposition and AI Techniques

Early fault detection for the induction machine became a necessity to prevent the escalation of failures to severe levels, thereby avoiding unscheduled downtimes.Among the various failure modes in electrical machines, rotor-related faults, such as air gap eccentricity, require particular attention and to detect this type of defects model-based methods are extensively used in this field.However, because of the intricacies of the diagnosed model and the time-consuming investigations it renders the diagnosis process more laborious and less efficient.This article focuses on applying a non-model based approach that relies in general on feature extraction using discrete wavelet transform decomposition analysis of stator current signal for various stages of air gap eccentricity and under multiple operating conditions and as a first step of the conducted work, through performing an in-depth energy distribution analysis through all of the decomposed signal levels to extract the best sub-signal level that holds the most relevant information about the machine's condition alongside to RMS values of the signal.The second part of the research focuses on employing the extracted features as input data used for training a multi-layer perceptron algorithm such as support vector machine and decision trees.Our endeavor is to choose the most accurate algorithm for the multiclass classification.