Research On Fault Diagnosis Method Of Axial Piston Pump Based On PSO-BP Neural Network

As the main power source of the hydraulic system,axial piston pump is widely used in various industrial fields,and the plunger pump is one of the fault frequent components in the hydraulic system.With the development of science and technology,the structure of hydraulic system is more and more complex,and there are more and more fault types.Therefore,it is very important to diagnose faults of plunger pump timely and effectively and the traditional artificial recognition fault type has been unable to meet the needs of modern mechanical fault diagnosis.In order to diagnose the several faults of axial piston pump,the method of combining the advantages of PSO and traditional BP network model is studied and the specific content is as follows:By analyzing the working principle and basic structure of the A4 VG plunger pump,the common fault and failure mechanism of the plunger pump are pointed out in this paper.It also calculates the basic vibration frequency of the plunger pump by use of the empirical formula,and analyzes the frequency spectrum of the fault vibration signal of the plunger pump with the Hilbert envelope spectrum.Meanwhile,through the analysis of the vibration transmission path of the plunger pump,the method of collecting the vibration signal of the plunger pump through the pump case is concluded.Moreover,by setting up the fault parts to simulate the failure condition of the plunger pump,four kinds of vibration signals are collected on the plunger pump fault diagnosis experimental platform,which are ones of the normal work of the piston pump,the wear of slipper,the wear of the oil plate and the wear of the inclined disk.After analyzing the shortcomings of the current feature extraction of vibration signals,a new method based on WPD and EMD is proposed.In this approach,the WPD is used to the noise reduction of vibration signals and the EMD method with adaptive decomposition characteristics is used to the decomposition of vibration signals and the method of extracting the IMF energy moments after EMD is used as the eigenvector of the fault vibration signal of the plunger pump.At the same time,it has been proved that the method of extracting the IMF energy moment has more advantages in distinguishing the difference and feature extraction of non-stationary signals,compared to the method of extracting the energy of wavelet packet coefficient and extracting IMF energy as eigenvector.The particle swarm optimization with global optimization is introduced,pointing at the disadvantages that the BP neural network is of slow speed of iteration and a relatively large performance drawback to local minimum when applied to fault diagnosis in this paper.Moreover,the method that the PSO algorithm is combined with BP network has been put forward to classify the vibration signals of the plunger pump.Finally,the PSO-BP network diagnosis model and BP neural network diagnosis model based on the fault diagnosis of plunger pump are established respectively,it has proved that the PSO-BP network has certain advantages in classifying the fault vibration signals of the plunger pump.Finally,the grey correlation degree of the characteristic vector of the standard fault and the feature vector of the position fault is calculated by the grey relational degree theory.On the one hand,it can verify the validity of extracted eigenvalues.On the other hand,it verifies the accuracy and feasibility of PSO-BP neural network as a fault diagnosis classifier.