Stator Imbalance Defects Diagnosis of Induction Machine Using Thermography and Machine Learning Algorithms

Abstract

Identifying stator imbalance problems in induction machines (IMs) is crucial to preserving operating efficiency, reliability, and safety. This identification contributes to the IM’s overall health, prolongs its lifespan, and helps organizations meet regulatory and performance standards. Moreover, it enables the execution of maintenance tasks when required, improving resource use and decreasing shutdown periods. This research presents a Non-Destructive Technique (NDT) that utilizes a thermography-based machine learning algorithm for diagnosing Stator Imbalance Defects (SID) within IMs experimentally. This technique relies on assessing temperature distribution on the outer surface of the IM stator. Then, the evaluation of various statistical variables that provide valuable insights into the normality of temperature distribution within the designated area referred to as the Region of Interest (ROI) is presented. Initially, thermal images of the stator are captured under various operating conditions, including a healthy state, SID presence, and phase missing. Then, all thermal images are converted into thermographic data matrices for statistical analysis using the Machine Learning Algorithm Gaussian Naïve Bayes (GNB). This analysis aimed at classifying SID in IMs arising from deviations in the normalcy of temperature distribution within the ROI. To demonstrate the efficacy of the suggested GNB approach, a comparison with the Subspace Discriminant Classifier and Cubic Support Vector Machine is provided. The GNB algorithm exhibits exceptional accuracy of 95.2% in classification tasks while maintaining a minimal processing time, irrespective of the operational duration of the IM.

Keywords

• Induction machine
• stator imbalance defects
• thermography
• statistical analysis
• Gaussian naïve Bayes