Underdetermined Blind Source Separation Based On Compressive Sensing Of Signals Of Bearing Fault In Reciprocating Compressor

Reciprocating compressor is increasingly large-scale,complicated,automation,continuous production of petrochemical,metallurgy and other process industry "heart equipment",is widely used in oil,refining,metallurgy and other industries,but its complex structure,the working environment,in the event of failure,not only affect the industrial production process,cause huge economic losses,and will produce a series of safety problems.Therefore,it is of great significance to study the fault diagnosis method to ensure its safety,stability and efficient operation.The vibration signal of reciprocating compressor is strong and non-stationary and nonlinear,which leads to the limitations of the traditional signal processing technology for reciprocating compressor fault diagnosis.This paper,on the basis of consulting a large number of literature,comprehensively analyzes the basic theory and method of mechanical equipment fault diagnosis technology,introduces the development of fault diagnosis technology of reciprocating compressor,introduces a novel signal processing technology —— blind source separation technology(Underdetermined Blind Source Separation,UBS),and proposes a blind source separation method based on compression sensing.The specific research contents are as follows:First,the mathematical model of UBSS and the different solution methods are analyzed.After comparative analysis,this paper,sparse component analysis(Sparse Component Analysis,SCA)is used to solve the processing principle of SCA in detail,and summarizes the general process of SCA algorithm and the commonly used methods in each process.Next,the improved fuzzy C-mean(Fuzzy C-Means,FCM)clustering was used to estimate the mixing matrix.The SCA algorithm is mainly divided into two steps,namely,the estimation of the mixture matrix and the recovery of the source signal,and the mixture matrix estimates adopt the clustering algorithm.Two main defects of FCM clustering algorithm:(1)random selection of initial cluster center;(2)clustering results are susceptible to outlier interference.This proposed with density peak clustering to improve FCM clustering algorithm.Simulation experiments show that the improved FCM clustering algorithm can improve the mixing matrix estimation accuracy better.Then,an improved compressed-sensing reconstruction algorithm is introduced to recover the source signal.In the SCA algorithm,the recovery of source signals is generally performed by the shortest path method,which has some obvious disadvantages.In this regard,the reconstruction algorithm in compressed sensing technology is introduced to recover the source signal of undetermined blind source separation,and the existing norm-based compressed sensing reconstruction algorithm is appropriately improved.The analysis of the bearing composite fault shock signal shows that the improved reconstruction algorithm has higher accuracy and better separation effect.Finally,the proposed UBSS based on compression sensing is applied to the fault diagnosis of reciprocating compressor bearings.Specifically,the composite fault signal of the reciprocating compressor collected by the sensor is separated from the undetermined blind source,and the different single faults in the composite fault signal are separated,and then each fault is judged and analyzed to provide guarantee for the extraction and diagnosis of the composite fault feature of the reciprocating compressor.The results show that the proposed method has more advantages than the traditional signal processing technology in processing composite fault signals.