A VMD–mRMR–Random Forest Framework for Intelligent Detection and Severity Assessment of Broken Rotor Bar Faults in Induction Motors

Induction motors are critical and expensive components of industrial sectors; hence, providing a reliable condition monitoring scheme is a serious issue. However, mechanical maloperation due to broken rotor bar (BRB) faults may degrade the motor performance significantly. If these severe faults are left undetected, secondary damages will be inevitable. This paper presents a new diagnosis method that uses time-frequency analysis to distinguish between healthy and BRB faults as well as fault severity. To achieve this, initially, vibration signals are decomposed into intrinsic mode functions using Variational Mode Decomposition (VMD) and some features are extracted from decomposed signals. Then, the minimum Redundancy Maximum Relevance (mRMR) method is employed to select the most effective features to enhance the accuracy and generalizability of detection scheme. Finally, these features are fed into a Random Forest (RF) classifier to detect the faulty condition and determine its severity. To implement and evaluate the proposed method, real-world vibration data collected from a ۳۸۰ V, ۴-pole induction motor is utilized. The obtained results indicate that the accuracy of the proposed method is quite superior to conventional methods in detecting the number of broken bars, despite its simplicity. Therefore, the proposed intelligent method can be effectively utilized in industrial applications.

کلیدواژه ها:

Induction Motors ، Broken Bar ، Pattern Recognition ، Signal Analysis ، Feature Selection ، Vibration Signals