Fault Diagnosis for Rolling Bearings Based on the Quadrature Particle Filter

• Published in: 2020 Chinese Control And Decision Conference (CCDC) pp. 2148 - 2155
• Main Authors: Zhang, Yan, Guo, Runxia
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2020
• Subjects: Bayes methods; Degradation; Fault diagnosis; Mathematical model; Paris' law; Particle filter; Particle filters; Probability density function; Quadrature particle filter; Rolling bearing; Rolling bearings
• ISSN: 1948-9447
• DOI: 10.1109/CCDC49329.2020.9164797

Timely fault diagnosis for rolling bearings is important to prevent equipment fault and reduce subsequent maintenance costs. In order to achieve fault diagnosis for rolling bearings, a novel quadrature particle filter (QPF) is developed. The Gauss-Hermite quadrature rule is selected to obtain an approximate posterior probability density function (PDF) that designed as an importance PDF for particle filter. QPF relieves the problem that PF always suffers from the particles' degradation, and helps to enhance system state estimation. Based on QPF, a fault diagnosis scheme for rolling bearings has been developed. First, an acceleration signal of the running rolling bearing is obtained by an accelerometer, and the feature extracted from the Hilbert envelope spectrum of the acceleration signal is used as a health indicator. Then, based on the Paris' law, a physical model that tracks the performance degradation of the bearing is constructed. Finally, QPF is chosen to estimate bearing status to diagnose typical faults. Applying the vibration data collected by testing platform when the bearing has an inner ring fault, the experimental results show that the fault diagnosis scheme can effectively track the bearing performance degradation and accurately realize fault diagnosis.