Robust Fault Classification Of Three Phase Induction Motor Using GWO-SVM With Simulation And Hardware Validation

This paper presents a comprehensive fault classification framework for three-phase Induction Motors (IMs) using a novel Grey Wolf Optimization-enhanced Support Vector Machine (GWO-SVM) approach. Addressing critical limitations of traditional methods—including narrow fault scope, computational inefficiency, and reliance on simulation-only validation—the framework targets high-impact electrical faults under no-load conditions: single-phasing fault, winding break fault, winding short fault, under-voltage fault, and over-voltage fault. A rigorous dual-validation methodology combines MATLAB/Simulink 2023a with experimental hardware implementation using an XPO MFS fault trainer kit and a three-phase IM. Current signatures are acquired via a Digital Storage Oscilloscope (DSO) at 15 kHz meeting Nyquist criterion and processed through MATLAB’s Diagnostic Feature Designer (DFD) to extract nine transient-sensitive time-domain features (e.g., RMS, crest factor, zero-crossing rate) from residual stator currents, enhancing fault signature discrimination. The GWO-SVM algorithm autonomously optimizes SVM hyper parameters (C, γ), achieving superior classification accuracy. Validation of simulation results with hardware results for five fault classification, the trained model achieves significant efficiency and the approach demonstrates robust performance. Comparative analysis with conventional SVM, K-nearest neighbor algorithm (KNN), Naive Bayes (NB), Random Forest (RF), Logistic Regression (LR) , and Decision Tree (DT) classifiers confirms the GWO-SVM’s efficacy, computational efficiency, and deployment readiness for resource-constrained edge applications.