New Hybrid Invasive Weed Optimization and Machine Learning Approach for Fault Detection

Fault diagnosis of induction motor anomalies is vital for achieving industry safety. This paper proposes a new hybrid Machine Learning methodology induction-motor fault detection. Some the parameters such as stator currents and vibration signals provide great deal information about motor’s conditions. Therefore, these were selected to test proposed model. The was assessed in laboratory under healthy, mechanical, electrical faults with different loadings. In this study model developed using collected signals, an optimal features selection mechanism proposed, machine learning classifiers trained classification. procedure extract some statistical from raw signal Matching Pursuit (MP) Discrete Wavelet Transform (DWT). Then, Invasive Weed Optimization algorithm (IWO)-based subset reduce data dimension increase average accuracy fed into three classification algorithms: k-Nearest Neighbor (KNN), Support Vector (SVM), Random Forest (RF), which k-fold cross-validation distinguish between faults. A similar strategy performed by applying Genetic Algorithm (GA) compare performance method. suggested detection model’s evaluated calculating Receiver Operation Characteristic (ROC) curve, Specificity, Accuracy, Precision, Recall, F1 score. experimental results have proved superiority IWO selecting discriminant features, has achieved more than 99.7% accuracy. successfully its robustness diagnosing load