Research On Health Assessment Of Electro-Hydraulic Servo Pump Control System Based On Deep Neural Network

Electro-hydraulic servo pump control system adopts volumetric servo integrated powertrain,which has the technical advantages of high power/weight ratio and high efficiency and energy saving compared with traditional valve control system.Electrohydraulic servo pump control system is an important industrial equipment,the system failure may cause serious accidents,serious casualties and economic losses.Therefore,it is of great significance to realize the health evaluation of electro-hydraulic servo pump control system for the safety and reliability of manufacturing.However,the pump control system involves the infiltration and integration of mechanical,hydraulic,electrical and control disciplines,and has the characteristics of nonlinear and time-varying.The factors causing the degradation of the system performance are usually very complex and hidden,which brings many technical challenges for realizing the rapid and accurate evaluation of the system running state.In this paper,rapid and accurate health evaluation of electro-hydraulic servo pump control system is taken as the research goal.Through analyzing the factors that affect the system running state under low speed conditions,a system health evaluation method based on LSTM-GRNN-ANN(LGA)deep neural network is proposed.Specific research contents include the following aspects:Firstly,according to the basic composition and working principle of the electrohydraulic servo pump control system,the mathematical models of the servo motor,hydraulic pump,valve and hydraulic cylinder of the system are established.The feedforward PID controller based on genetic algorithm is designed to realize the closedloop control of the system pressure.On this basis,a complete simulation model of the electro-hydraulic servo pump control system is built in the Matlab/Simulink environment.It provides theoretical support for studying the health evaluation of hydraulic system.Secondly,a mathematical model was established to analyze the factors affecting the operation state of the electro-hydraulic servo pump control system.Considering the compressibility of the oil,the low-speed characteristics of the servo motor and the leakage of the system,the oil gas content,the magnetic density of the motor air gap and the leakage coefficient of the system were taken as the main performance indicators to evaluate the health state of the system and the degradation threshold was set.Based on the mathematical model,the characteristic parameters affecting the performance index were extracted,and the quantum particle swarm optimization algorithm(QPSO)was used to select the number of neural network neurons,and the health evaluation model of LGA deep neural network was constructed.Finally,the simulation experiments of oil gas content fault,motor air gap eccentric ity fault and hydraulic system leakage fault were carried out by the simulation model of electro-hydraulic servo pump control system and the experimental platform,and the sample data required by the health evaluation model were collected to train and test the model,so as to verify the accuracy and reliability of the LGA deep neural network health evaluation model method.It lays a theoretical foundation for long life and high reliability operation of electro-hydraulic servo pump control system.