Compound Fault Diagnosis of Rolling Bearing Based on Transformation Scale Improved BPD and MCKD

Rolling bearing is one of the major parts of a rotary machine, and its status also influences the operation of the rotary machine. This paper presents a new method to detect and separate compound faults of rolling bearing by integrating an improved basis pursuit denoising (BPD) algorithm with maximum correlated kurtosis deconvolution (MCKD). Split variable augmented Lagrangian shrinkage algorithm (SALSA) is adopted in this paper to perform BPD. Furthermore, to set appropriate values of Lagrange multipliers in BPD, transformation scale, which is determined by the energy in each sub-band of the initial denoised signal, is introduced. The vibration signal is then denoised to reveal repetitive impulses through new Lagrange multipliers. MCKD is used for further separation of compound faults in rolling bearing. Vibration signal analysis simulating compound faults of inner race fault and outer race fault verifies effectiveness of the presented method.