Infrared Thermography-Based Induction Motor Bearing Fault Detection Using Hybrid Method

In Induction Motor (IM), bearing is one of the most important components to support the rotational movement of shafts while minimizing friction. It necessitates frequent analysis in order to ensure dependable performance. Considerably, Infrared Thermography (IRT) is a non-contact, non-destructive technique used to find bearing flaws in rotating machinery. Whereas, the IRT image's string noise and irrelevant information restrict its overall performance. So, in this research, an emergent two-dimensional Discrete Wavelet Transform (2D-DWT) based IRT approach along with an effective feature selection technique namely Independent Component Analysis (ICA) has been presented to solve the aforementioned problem and diagnose the various bearing failures in IM. The best feature set was obtained by ranking the chosen features according to their relevance using the Mahalanobis Distance (MD) approach. In order to classify faults and assess performance, these particular features have been sent to a Complex Decision Tree (CDT). The implemented model outperforms the compared models such as Artificial Neural Network (ANN), Empirical Wavelet Transforms (EWT), and hybrid motor-current data-driven approach by obtaining higher classification accuracy of 99.85%.