Researches On Sparse Representation And Reconstruction Method Of Bearing Fault Vibration Signal

With the rapid development of modern science and industrial automation level unceasing enhancement,as an important composition in industrial production,the structure and function of rotating machinery is increasingly complicated.Because rotating machinery plays an important role in industrial production,once the rotating equipment fails,the damage often is immeasurable.Therefore,it is great significant to monitor rotating machinery operation status and taking corresponding measures with the use of modern signal.However,because the structure of the rotating machinery is becoming more and more complicated and the amount of signal increase sharply,signal on-line monitoring and fault signal feature extraction become very difficult.Compressive sensing(CS)theory makes it possible to solve the problem.Resear-ches on sparse representation,sparsity estimation,signal reconstruction for rolling bearing which is the important component in rotating machinery have been carried out.In terms of vibration signal sparse representation,this paper studied the methods of vibration signal sparse representation based on analytic dictionary and learnt dictionary.First,this paper discusses the analytic dictionary and the learnt dictionary in detail,then compares the performance of the two dictionaries via the sparse representation experiment of rolling bearing vibration signal,finally gives a brief introduction of parametric dictionary.At the same time,this paper analyzes the running speed in sparse representation of the analytic dictionary and the learnt dictionary,because improving running speed of the algorithm is also important.In terms of sparsity estimation of vibration signal,this paper studied the sparsity estimation method based on CS and singular value decomposition(SVD).Because the previous sparsity estimation method needs to calculated sparsity through signal reconstruction process,so this paper improves the estimation method which is based on SVD.The improved algorithm can estimate the signal sparsity quickly and accurately,and it does not need to calculate sparsity through signal reconstruction process.In the end,sparsity estimation experiments prove the superiority of the improved algorithm.In terms of signal reconstruction of vibration signal,this paper studies the signal reconstruction algorithm based on compressive sensing.Reconstruction precision is difficult to guaranteed when redundant dictionary is applied to compressive sensing algorithm,so this paper propose signal reconstruction algorithm named compressive sensing algorithm based on signal space which improve the precision of signal reconstruction successfully.The improved Orthogonal Matching Pursuit algorithm and the improved Compressive Sampling Matching Pursuit are used respectively to do the compressive reconstruction experiment,then the reconstruction signal is used to extract the fault characteristic frequency which is very important to fault diagnosis.On the basis of compressive sensing theory,this paper carry on the thorough researches on sparse representation,sparsity estimation,signal reconstruction of vibration signal which provide effective support for the research and development of compressive sensing theory.