Bearing fault diagnosis using multiple feature selection algorithms with SVM

This paper presents an efficient approach to diagnose defects in various components of bearings in rotating machines using vibration signature analysis. This Automated fault investigation scheme (AFI) method integrates the Fisher Score (FS) and Genetic algorithm (GA) feature selection methods and an efficient hyperparameter tunned model with Support Vector Machine (SVM) classifier to accurately classify defects in rolling ball bearings. This approach ensures accurate classification of bearing defects through the simple machine learning models within a reduced computation time. This work is carried out with recorded vibration signals from a laboratory experimental setup on Machine Fault Simulator (MFS), focusing on rolling ball bearings with defects in inner race, outer race and ball itself, especially focusing on the combined faults. Statistical analysis based on both time and frequency domain is employed to compute feature vectors for fault investigation in ball bearings using machine learning models. The computed results demonstrate that the proposed feature selection method with hyperparameter tuning achieved remarkable maximum accuracy with 97% in FS and 99% in GA with SVM classifier. Notably, these models accuracies improved with feature selection algorithms as compared to the normal model computation. Consequently, the testing loss using this hyperparameter tuning function remains very low. Overall, this paper compares the results of time and frequency domain analysis and introduces a promising approach for both efficient and accurate fault detection in bearings of rotating machines, potentially reducing the need for extensive manpower and sensor usage. The outcomes of this study can be used to develop efficient intelligent health monitoring schemes for industrial machines that can help in smooth and cost-effective operation.