Abstract: Aiming at the problem that the separation performance of the source separation algorithm was affected by the separation matrix and the compound fault characteristics of rolling bearings could not be adaptively separated, the adaptive fruit fly optimization algorithm and the denoising source separation were combined, a method of denoising source separation for rolling bearing composite faults based on AFOA algorithm was proposed. First, adaptive fruit fly optimization algorithm was used to initially optimize the separation matrix, and then the separation matrix was taken as the individual fruit fly and negative entropy was taken as the objective function, global optimization on the maximum value of the objective function was carried out. And the optimal separation matrix for denoising source separation was determined. The tangent function was used as noise reduction function for noise redution separation to estimate the vibration signal of the inner and outer ring compound fault bearings. Finally, the envelope analysis was performed to extract the inner and outer ring fault characteristics. In addition, the proposed method was compared with fast independent component analysis based on the adaptive fruit fly optimization algorithm,through the simulation and actual test of the compound faults of the inner and outer rings of the bearing. The results show that the AFOA-DSS method can more accurately separate the compound faults of the rolling bearing features, realize fault diagnosis, verify the effectiveness and practicability of the method.

Key words:  rolling bearing; compound fault diagnosis; adaptive fruit fly optimization algorithm (AFOA); denoising source separation (DSS); negative entropy; fast independent component analysis (Fast ICA)