Machine Learning Ball-Bearing Fault Detection Methods Using Envelope Analysis and Power Spectral Density

Kakesh, Mohammad Hatami; Rahideh, Akbar; Agah, Gholam Reza; Hedayati Kia, Shahin

doi:10.30503/eecs.2025.488460.1064

Machine Learning Ball-Bearing Fault Detection Methods Using Envelope Analysis and Power Spectral Density

Document Type : Original Article

Authors

Mohammad Hatami Kakesh ¹

Akbar Rahideh ²

Gholam Reza Agah ¹

Shahin Hedayati Kia ³

¹ Department of Electrical Engineering, Shiraz University of Technology

² Shiraz University of Technology

³ Université de Picardie JulesVerne, 80039 Amiens, France

10.30503/eecs.2025.488460.1064

Abstract

Ball-bearings are one of the most important components in rotating machinery. Due to practical importance of rotating machineries in industry, fault detection has become inevitable. Various techniques have been implemented for ball-bearing fault detection using vibration signals. In this research, vibration signal analysis methods are presented to extract suitable features for training some of the machine learning techniques in order to diagnose ball-bearing defects in different speeds. The purpose of this study is to obtain a highly accurate algorithm and compare its performance with that of other machine learning algorithms. To achieve this goal, Hilbert transform has been applied for envelope analysis to attenuate the frequencies that are not related to ball-bearing fault and perform power spectral density and descriptive statistics to extract features. Also comparison and evaluation of random forest, support vector machine, artificial neural network and k-nearest neighbour have been carried out for this study. For dataset with 1465 samples in various speed, random forest has achieved the accuracy above %97.

Keywords

Fault Detection

Ball-bearing

Machine learning

Signal Processing

Induction Motor

Subjects