<![CDATA[This study evaluates and compares the performance of three machine learning–based classification methods, namely Support Vector Machine (SVM), Gaussian Naive Bayes (GNB), and K-Nearest Neighbor (KNN), for diagnosing stator faults in induction motors using electrical and mechanical parameters, including phase voltage, phase current, power factor, rotational speed, torque, power, and temperature. The experimental results demonstrate that all three methods achieve an accuracy of 100% on the entire test dataset, while the KNN method maintains stable performance across all tested values of K. The consistency of results obtained from algorithms with fundamentally different principles margin-based, probabilistic, and distance-based indicates that classification performance is more strongly influenced by the quality and relevance of the selected features than by algorithmic complexity. Further analysis reveals that the primary differences among the methods lie in their computational characteristics and interpretability, where GNB offers superior efficiency, SVM provides robust decision boundaries, and KNN exhibits scalability limitations. These findings confirm that accurate and reliable induction motor stator fault diagnosis can be achieved using conventional classification algorithms when supported by appropriately selected and physically representative features.]]>
Copyrights © 2026
