Research On Bearing Fault Feature Extraction Based On Sparse Theory

With the continuous advancement of China's industrial modernization,the machinery and equipment in the vital industries of the national economy such as transportation,petrochemical,metallurgy and energy are moving in the direction of precision,automation and intelligence.The reliability of mechanical equipment is becoming one of the most important factors affecting industrial development and modernization,it is an effective way to avoid economic and personnel losses caused by accidental failure of mechanical equipment by implement the modern mechanical condition monitoring and fault diagnosis technology to monitor and diagnose mechanical equipment,in the meantime,The bearing in the equipment is the component with the highest failure rate in many machinery.If it is possible to detect the running condition of the bearing in the mechanical equipment in real time,and to extract fault characteristics and determine the health status of the bearing fault that has occurred,it is beneficial to improve the reliability level of the rotating system and the whole mechanical equipment,and reduce the unplanned downtime,which increases equipment utilization in turn.Firstly,this paper summarizes and analyzes the research status of bearing fault feature extraction technology at home and abroad,and illustrates the shortcomings of the current fault extraction and improvement of bearing fault characteristics in the background of high noise interference.In the aspect of extracting objects,the classification and failure mechanism analysis of the fault types of bearings were carried out.The mathematical modeling method was used to simulate the time domain signals of the pitting corrosion occurred on the inner/outer rings and the rolling elements of the bearing.Moreover,the simulated signal is used for further research.Secondly,the bearing fault feature extraction method based on sparse theory is proposed for the high noise interference of bearing fault signals in many mechanical equipment.The sparse representation theory is discussed in detail,and then it is presented in general.The problem of poor dictionary applicability and imperfect algorithm optimization is proposed.The feature extraction algorithm based on Laplace wavelet basis based over-complete dictionary and MP(Match Pursuit)algorithm with residual threshold optimization is proposed to verify its simulation signal,Artificial processing failure signal and Early weak fault signal in bearing life-life degradation test.Thirdly,the algorithm ACFSS which combines the attenuated cosine dictionary and the feature sign search based on the sparse BP(Basis Pursuit)method are proposed for the problem of relatively low computational efficiency and insufficient threshold in MP algorithm.The effectiveness of ACFSS is verified on the low SNR simulation fault signal and the weak lifetime fault signal object.Then the proposed algorithm ACFSS further validates its application to the composite fault feature extraction under noise interference.