Machinery Fault Diagnoses in High-Voltage Circuit Breakers Using Empirical Mode Decomposition-Support Vector Machine Model

The circuit breakers have a flaw in their trip mechanism. Early mechanical faults like coil-plunger jam are hard to report using traditional fault diagnosis based on the coil current. However, users still have difficulty understanding and trusting the diagnosis results since the standard EMD is opaque in terms of interpretability. In this work, an Empirical Mode Decomposition with Support Vector Machine (EMP-SVM) proposed for machinery fault diagnosis. This work uses the vibration signal recovered from the trip process as an intrinsic time series of the data. At first, The Improved EMD is a data-adaptive multiresolution approach that breaks down signals into physically relevant components based on vibration and coil current signals. Lastly, utilizing the characteristics that were retrieved, the SVM is used to diagnose the defect. The implemented improved EMP-SVM model demonstrates superior performance, achieving Precision of (98.37
^%
), recall of (91.03%) and F1-Score of (96.34%). When compared to previous models like Improved Dempster-Shafer Evidence Theory-Echo State Neural Network (IDSET-ESNN) and Phase Space Reconstruction with Support Vector Machine (PSR-SVM).