Enhancing Predictive Maintenance Accuracy for Rotary Machine Vibration Signals with XGBoost-RFE Based Feature Selection

This study proposes a model of a rotary machine's fault diagnosis system based on vibration signal analysis under Improved eXtreme Gradient Boosting-Recursive Feature Elimination (XGBoost-RFE). A 3D dataset of vibration signals is collected from healthy and faulty induction motors. The Empirical Mode Decomposition (EMD) technique is used to perform signal conditioning. The de-noised signals are obtained to extract the multi-domain features. Finally, multiple classifiers, including K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and eXtreme Gradient Boosting (XGBoost), are performed with different kernel settings at the classification step. The results indicate that a hybrid approach that combines time and frequency domain features and classifies them using XGBoost with a Gaussian kernel achieves the highest accuracy (99%) with the lowest error rate <1.3%.