Thermal Image-Based Fault Analysis of Induction Motors using a Novel Machine Learning Model

In the industrial industry, establishing the condition of health of an induction motor is a critical task that, when done correctly, results in a more secure operation. In this study, thermal imaging is recommended as an essential component of a successful machine learning model-based fault diagnostic methodology. A number of studies have been conducted in the laboratory, including investigations into the healthy and malfunctioning states of the induction motor. To classify motor faults, thermal pictures of the motor states, Grey Level Co-occurrence Matrices (GLCM), Invasive Weed Optimization algorithm-based feature selection (IWO), and intelligence algorithms were utilized in a hybrid method. This was done to ensure that the proposed model for problem diagnostics was robust. To extract all of the available features, the GLCM algorithm is used. The degree of relationship among pairs of pixels is calculated using second-order feature statistics. The degree of correlation determines the distance and angle between pixels. Values for contrast, dissimilarity, homogeneity, ASM, energy, and correlation were returned. IWO is utilized because it can minimize data dimensionality. To distinguish between the many forms of motor defects, two separate categorization systems are used. All of these features were tested in three load conditions: no load, 50% load, and 100% load, as well as three faulty motor conditions: broken rotor bar faults, bearing faults, and stator faults. No load, broken rotor bar defects, and stator faults were among the conditions. A set of artificially intelligent classification algorithms was chosen and trained by means of a cross-validation technique to appropriately categories each fault. The metrics accuracy, precision, sensitivity, and F1 score are used to evaluate the performance of the proposed model. The results showed that the hybrid model was able to accurately forecasting and detecting motor failures. Furthermore, the proposed model here can be applied to a wide range of infrared applications for image processing.