LightGBM-Based Fault Diagnosis of Rotating Machinery Under Changing Working Conditions Using Modified Recursive Feature Elimination

This article presents an intelligent and accurate framework for fault diagnosis of induction motors using light gradient boosting machine (LightGBM). The proposed offers promising generalization ability when the testing data contains new unseen operating conditions unavailable during training process. After acquisition vibration signals feature extraction in multiple domains, we perform iterative selection (FS) approach by utilizing a modified version recursive elimination (RFE) features’ importance scores obtained LightGBM. To prevent overfitting subsequent bias, outer resampling loop encompasses whole process our RFE-LightGBM algorithm. Moreover, instead conventional methods based on K-fold cross-validation (CV) or leave-one-out CV (LOOCV), use scheme called leave-one-loading-out (LOLO-CV). Leveraging LOLO-CV, FS method identifies optimal subset, making robust under changing conditions. Then, final classification is performed with subset LightGBM model adjusted hyperparameters employing Bayesian optimization. Experimental results from two real case studies show that achieves accuracies between 98.55% 100% various scenarios. For example, worst-case scenario bearing dataset Case Western Reserve University where no-load (0hp) absent only used testing, accuracy classifier before after applying RFE-LightGBM-FS 88.04% to 97.23%, respectively. Using hyperparameter optimization further improves 98.55%.