An Enhanced Modeling Framework for Bearing Fault Simulation and Machine Learning-based Identification with Bayesian-optimized Hyperparameter Tuning

Abstract

Abstract Condition monitoring of rotating machinery offers a salient tool for predictive maintenance on rolling elements, subjected to continuous working loads, wear, fatigue, and degradation. In this study, an enhanced computational tool for bearing fault simulation and feature extraction is proposed. A subsequent identification scheme is realized, through Bayesian optimization of hyperparameters, including support vector classifier (SVC), gradient boosting (GBoost), random forest (RF), extreme gradient boosting (XBoost), light gradient boosting machine (LightGBM), and categorical boosting (CatBoost). The proposed hyperparameter optimization technique stands out from traditional methods by offering a more informed and efficient pathway to optimal performance in predictive maintenance. Utilizing Bayesian optimization for hyperparameter tuning of machine learning models, which has not been extensively explored in this field, our approach demonstrates significant advancements. Typical instances of bearing faults, like inner race, outer race, and ball faults are considered. The analysis relies on extraction of statistical and engineering features form collected response signals, including kurtosis, root mean square, peak, and ridge factor. Highly influential variables are highlighted based on feature selection and importance algorithms, allowing bearing fault classification. We demonstrate that SVC and LightGBM produce over 97% of accuracy at low computational cost. This approach constitutes a scalable and robust framework for similar applications in engineering diagnostics.