Research On Feature Extraction And Pattern Recognition Method Of Fault Signal Of Axial Piston Pump

The axial piston pump is exposed to harsh environments such as high pressure,high temperature and corrosive gas for a long time,and the failure rate of the axial piston pump is greatly improved.Therefore,it is of great significance to diagnose the early failure of the plunger pump.Aiming at the fault signal shortcoming of axial piston pump under running states such as non-linearity,instability and large interference,the wavelet threshold,local mean decomposition(LMD)and elite chaotic particle swarm optimization(ECPSO)support vector machine(SVM)algorithm are used in the preprocessing and characteristics of the fault signal to optimize the recognition accuracy rate of the fault diagnosis method.The early fault diagnosis of the axial piston pump is mainly studied from the following four aspects:(1)Research on vibration mechanism of axial piston pumpThe mechanical structure and working principle of the axial piston pump are analyzed,and the vibration mechanism and failure modes of the axial piston pump are studied to find out the failure causes.In addition,the final stress point of the piston pump vibration is found according to the internal vibration transmission path,which lays the foundation for early fault diagnosis.(2)Research on the method of extracting fault features of axial piston pumpFirstly,according to the analysis of the vibration transmission path of the plunger pump,the vibration acceleration sensors are arranged at the corresponding final stress point,and the fault vibration signal is collected at multiple points.Secondly,for the shortcomings of the hard and soft wavelet threshold functions such as the discontinuity and the existence of constant errors,the adaptive wavelet threshold function(AWTF)signal processing method is proposed.And the AWTF can adaptively adjust the wavelet threshold by calculating the correlation coefficient between the decomposed data of each layer and the original data.Finally,the LMD algorithm is used to decompose the signal after noise reduction.Furthermore,the impact of LMD end effects and false components are reduced by the waveform matching continuation and component correlation methods,and the integrity of fault feature extraction is improved.(3)Research on the method of identifying the failure mode of plunger pump based on ECPSO-SVMAiming at the problem of particle swarm optimization(PSO)such as low accuracy,long time-consuming optimization and sticking at local optima,the improved elite chaotic particle swarm optimization SVM algorithm is proposed.Firstly,the elite chaos search strategy is introduced to screen elite individuals and reduce the search dimension.Secondly,the optimization perturbation χ and cross mutation operations are added to the particle position update type to avoid sticking at local optimization in the neighborhood.Finally,the penalty factor C and the kernel function parameter θ in the SVM are optimized by the ECPSO algorithm,and a new fault classification model for plunger pumps is got.In addition,the axial piston pump is the research object,and the failure test platform is established to diagnose the fault of the axial piston pump.The experimental results show that the diagnostic accuracy of the ECPSO-SVM plunger pump diagnostic model is 98%,and it can effectively and quickly classify faults.