Research On Extraction And Identification Method Of Piston Pump Fault Features

As the power source of common mechanical equipment in industrial production,piston pump is the key element that determines whether industrial production can proceed normally.Therefore,its operating condition plays a very important role in the entire industrial production.With the continuous advancement of science and technology,in order to meet the needs of modern industrial production,the piston pump is constantly developing towards high power and high speed.At the same time,in order to realize the stability of the output flow pressure,the complexity of its internal parts is correspondingly increasing,resulting in more contact movement between these.Because of this,the piston pump is prone to failure during operation,and once a failure occurs,it will have a serious impact on industrial production.Therefore,timely and effective fault diagnosis of the piston pump is a necessary guarantee for industrial production.Early fault diagnosis of piston pumps often requires manual identification by experienced workers,but this method requires prior knowledge and the accuracy rate is not high enough.In modern times,the method of collecting and analyzing the signal during the operation of the piston pump is used to determine the failure of the piston pump.However,due to the complexity of the operational process of the piston pump,the signal often exhibits nonlinearity and non-stationarity.The traditional signal analysis method is not effective in analyzing the signal of the piston pump,leading to such the shortcomings as low accuracy,slow speed and the like,which makes it unable to meet the demand of industrial production.In order to ensure high speed and high precision when diagnosing the fault of the piston pump,this paper conducts an analysis and research on the modern method of fault diagnosis of the piston pump.First,according to the characteristics of the piston pump signal,the present study attempts to find out the most suitable signal analysis method for the piston pump signal through simulation;then,taking the axial piston pump as the research object,a fault test platform designed to collect and analyze the online fault signals will be set up to construct the fault characteristics Vector;finally,the optimization performance of particle swarm optimization(PSO)is combined with traditional support vector machines(SVM)to identify several common faults of piston pumps so as to achieve high precision and high speed.The main research content and work arrangement of this paper are as follows:(1)Calculate the failure frequency of the piston pump by analyzing the working principle of the piston pump.Firstly,by analyzing the wide application of the piston pump and the flow pulsation rate,it is determined that the 9 piston hole swash plate axial piston pump model A10VSO45 should be the research object.Next,analyze its working principle as well as the contact movement between various internal parts,after which several common failure types that piston pumps are prone to are determined,and the vibration frequencies of these distinct failures of the piston pump are calculated at a given speed through formulas.Finally,analyze other vibration frequencies that are generated by the piston pump during operation are also introduced,Noticeably,the applicability of the Hilbert spectrum in signal demodulation is introduced,which lays the foundation for subsequent analysis.(2)Build a mathematical model that belongs to the characteristics of the piston pump.Considering that the characteristics of the piston pump signal in the working process are non-linear and non-stationary,common signal analysis methods prove ineffective for the analysis of this type of signal.Therefore,this study builds two nonlinear non-stationary signals respectively to simulate the characteristics of the piston pump which it exhibits when it fails.The mathematical model in question is designed and built,and then the simulation analysis is carried out.Eventually the algorithm suitable for this type of signal is found through simulation analysis and comparison.(3)Build a piston pump failure test bench and collect online failure signals: complete the production of the failure parts of the piston pump according to theanalysis of the piston pump;determine the equipment required for the experiment and establish piston pump failure test bench based on the needs of collecting signals;determine the installation position of the sensor according to the signal transmission path;and finally,collect the failure signals of the piston pump in different states.After that,the fault signal of the piston pump is used to further determine the best signal analysis method.(4)In order to solve the problem that it is difficult to extract the fault characteristics of the piston pump,a method combining the variational modal decomposition(VMD),kurtosis criterion,and fuzzy entropy is proposed.First,the collected signals in five different states are decomposed by VMD.Then,the IMF with the most information is selected using kurtosis theory.Lastly the complexity of the time series can be measured by fuzzy entropy.In this way,the fuzzy entropy is calculated and the data is normalized,serving as Composition feature vector.(5)Aiming at the problem that the recognition speed and accuracy of the traditional SVM is not high enough,a fault recognition method using PSO to optimize the penalty parameters and kernel function parameters in the SVM kernel function is proposed.First,build the PSO-SVM intelligent fault diagnosis system in MATLAB,input the training samples,and use PSO to optimize the SVM parameters;then use the optimized parameters to train the SVM,perform fault identification on the test samples,and take the average after multiple verifications.In the present study,the accuracy reaches 99.5%,and the diagnosis time reaches 5.1s.Finally,in order to verify the advantages of PSO-SVM in terms of speed and accuracy,the SVM with default parameters,SVM optimized by cross-validation(CV),and the SVM optimized by genetic algorithm(GA)are compared to the proposed method respectively to demonstrate the advantages of the method proposed in this paper in terms of speed and accuracy,In order to verify the advantages of the VMD proposed in this paper compared with the common EMD,the EMD fuzzy entropy is used as the feature vector and input into the PSO optimized SVM to compare with the method in this paper to verify the feasibility and superiority of the VMD method.