Fault Diagnosis of Rolling Bearings Under Variable Load Conditions Based on Multi-domain Features and Random Forests

Abstract

Vibration signals of rolling bearings collected under variable load conditions often have complex dynamic properties which pose a huge challenge for its effective fault diagnosis.To solve this problem, a novel fault diagnosis method based on multi-domain features and random forests is proposed in this paper.In features extraction, the fast ensemble empirical mode decomposition method is first used to decompose the original signals into a collection of intrinsic mode functions (IMFs).After signal decomposition, the singular values of the matrix formed by the row vectors of IMFs can be obtained by singular value decomposition.On the other hand, to obtain a comprehensive description about vibration signals, the statistical analysis method and Fourier transform are employed to extract 10 time domain features and 10 frequency domain features.As for the automatic diagnosis of bearing faults, a novel combined classifier algorithm named as random forests is used to classify the multi faults under different load conditions.Finally, the proposed method is evaluated by experiments with 10 fault types and some comparative studies are also given.The experimental results indicate its effectiveness and robustness for rolling bearing fault diagnosis under variable load conditions.