Rolling bear fault recognition based on improved sparse decomposition

Abstract

Sparse decomposition is an adaptive sparse expression of the signals, which makes it easier for us to get the useful information contained in the signal. BP(basic pursuit) and MP(match pursuit) algorithm as two kinds of sparse decomposition are widely used in signal processing, pattern recognition etc, as while as in equipment fault identification. This paper use sparse decomposition to extract the fault signal from the bearing vibration signal containing with background noise. An atom dictionary which express bearing fault characteristics correct is introduced in this article. This atom dictionary incorporates the rotational speed of the bearing, the dimensions of the bearing and the bearing fault status etc combining with basic pursuit and matching pursuit algorithm to analysis the bearing fault simulation signal. The simulation and experiment result suggest that basic pursuit and matching pursuit algorithm have good effect on bearing fault recognition.