| The axial piston pump is an important hydraulic component in the electro-hydraulic switch machine.Once a malfunction occurs,it will have a serious impact on rail transit,ranging from affecting the normal operation of trains to endangering the safety of train operation,leading to catastrophic accidents.Therefore,the fault diagnosis of plunger pumps is highly concerned.Data driven has performed well in the field of fault diagnosis,but due to cost and safety reasons,obtaining labeled fault data samples is difficult and the signal often contains noise,resulting in low fault diagnosis rates.In order to improve the fault diagnosis rate,many scholars have proposed methods that utilize simulation data.However,the internal structure of the plunger pump is coupled with each other and is affected by oil pressure,which makes it difficult to establish a high-precision model,and it is also difficult to achieve injection of model faults.In this paper,data driven and modeling simulation methods are used for fault diagnosis of plunger pump,and a digital analog driven fault diagnosis method is proposed.This method first builds kinematics models,hydraulic models,flexible parts,and multi physical field coupling models,and then carries out simulation to obtain high-precision model building.By changing the flow area in Amesim software,fault injection of port plate wear and cylinder peeling is realized.By integrating the simulated fault data of the model with the measured data,the effect of expanding the data is achieved.Using this dataset as the training sample for fault diagnosis,the diagnostic rate is compared with only applying experimental data to the diagnostic model.The results show that this method effectively improves the diagnostic rate of plunger pump faults.(1)Build a fault model for the plunger pump.The kinematics and force law of the plunger are theoretically deduced and calculated,and the movement,force,flow area and flow of the main parts of the plunger pump are analyzed;Use Solidworks to establish the 3D model of the axial piston pump,transfer the piston pump model to the mechanical system dynamics analysis software ADAMS,add complex constraints and forces,establish the kinematics model,conduct kinematics simulation analysis,obtain the displacement curves of the main parts,and verify the correctness of the kinematics model through theoretical derivation and simulation analysis and comparison;Use Amesim software to complete the establishment of the hydraulic model of the plunger pump.According to the actual situation of the outlet flow of the plunger pump,adjust the area function of the port plate in Amesim,so that the hydraulic model can achieve high accuracy;In order to more accurately consider the nonlinearity of materials,better grid convergence,and better capture details and changes,the unit type SOLID186 was used to achieve flexibility of the front pump cover of the plunger pump in Ansys software.In order to adjust the node mass more flexibly,reduce computational time and memory requirements,the mass unit MASS21 was selected to create external nodes;In order to simulate the operation of the plunger pump more comprehensively and accurately,a rigid flexible liquid coupling model was established in ADAMS using multi physical field coupling and unidirectional joint simulation methods.Increase the distribution area function in the hydraulic model by 5% to simulate the wear fault of the distribution plate;Use cutting tools to create a cutting plane on the surface of the plunger,and create irregular shapes on the cutting plane to simulate plunger wear faults;Reduce the distribution area function by 5% to simulate cylinder peeling failure.(2)Construction of a digital analog fusion dataset under the coupling of multiple physical fields.Inject three types of faults into the dynamic model in sequence and add drivers to obtain simulated fault vibration data;Build an axial piston pump fault diagnosis test platform for data collection,obtain test data of the piston pump under three working conditions,and verify the correctness of the dynamic model by calculating the correlation between simulation data and test data;Based on a large amount of simulation data of plunger pump faults,supplemented by a small amount of measured data of plunger pumps,a digital analog fusion dataset was constructed to solve the pain point of low fault diagnosis rate caused by insufficient plunger pump fault data.(3)Analysis of fault diagnosis results.A deep learning classifier is used to compare the training sets of measured data and fused data samples.The test results show that with the addition of normal operating conditions data,the diagnostic rates of the three types of fault conditions significantly decrease when using measured data as the training set.Adding simulation data to the training set with only measured data shows that the fault diagnosis rate is better than that with only measured data,effectively improving the fault diagnosis rate,It has been fully proven that the fault diagnosis method of piston pump based on digital analog fusion can solve the problem of low fault diagnosis rate caused by insufficient fault data... |
