Optimization of Laplace Wavelet Dictionary for Sparse Decomposition of Early Weak Signal of Rolling Bearing Based on Artificial Bee Colony Algorithm

Abstract

The transient components in the early weak fault vibration signals of rolling bearings are easily obscured by intense background noise and cannot be detected quickly. Based on the sparse representation principle, an artificial bee colony (ABC) optimization sparsity method of the Laplace wavelet dictionary is proposed to realize the transient characteristic components in the signal extraction. Sparse decomposition with Orthogonal Matching Pursuit (OMP) algorithm is a signal adaptive decomposition algorithm, and it is one of the effective methods for weak feature extraction under strong noise background. Aiming to select and construct an over-complete dictionary for the sparse representation of rolling bearing fault vibration signals, based on the analysis of fault signals' characteristics, an improved Laplace wavelet atomic library was constructed. For solving extensive calculation and low efficiency of the orthogonal matching pursuit algorithm, this paper combines the ABC algorithm's fast operation characteristics to select the improved Laplace wavelet atom that best matches the fault through the inner product operation, thereby improving the calculation efficiency. Experiments show that the method has a proper matching with the early weak fault signals of rolling bearings and can adequately characterize fault information and judge the fault type more accurately.