LTI Filtering- and Multiple Resonance-Based Signal Decomposition for Semi-Quantitative Fault Diagnosis of Rolling Bearings

Abstract

Rolling bearing fault diagnosis is of great significance to ensuring the safe operation of rotating machinery, and vibration analysis based signal processing methods have become a mainstream of rolling bearing fault diagnosis technologies. Aiming at the separation of different signal components induced by rolling bearing composite defects, a novel signal decomposition based on linear time-invariant (LTI) filtering and multiple resonance is proposed in this paper, which can decompose the fault vibration signal with composite defects into high-, middle-, low-resonance components and the low-frequency component. The high- and middle-resonance components sparsely represent the damped responses induced by severe and slight defects, respectively. The low-resonance component represents transient component induced by some random interferences, and the low-frequency component contains the components of shaft rotation rate and harmonics caused by shaft bending or imbalance. Compared with conventional dual-Q-factor resonance-based signal sparse decomposition (RSSD), this method can not only detect the feature frequency, realize semi-quantitative analysis of defects’ amounts and severities, but also provide a monitor for shaft bending and imbalance. The effectiveness and practicability of this method has been validated by the experimental signal with dual defects on outer race, which explores a new way to apply RSSD to the diagnosis of rolling bearing composite defects.