Research On Intelligent Fault Diagnosis Algorithm Of Electro-hydraulic Servo Valve Based On Test Signal Fusion

Electro-hydraulic servo valve,as a precision control element in the hydraulic system,has the advantages of high control accuracy,fast response,small size,light weight,and can adapt to pulse modulation and analog modulation.It has been widely used in aerospace,metallurgical machinery and other fields.Because the faults of electro-hydraulic servo valves are diversified,non-linear and closed,and often difficult to detect in time,the health of the electro-hydraulic servo valve has a vital impact on the normal operation of the system.This paper takes the force feedback two-stage electro-hydraulic servo valve as the research object,and launches the research on the intelligent fault diagnosis algorithm of the electro-hydraulic servo valve.The main research contents are as follows:First of all,this article has conducted the research on the common faults of the electro-hydraulic servo valve.The mathematical model of the force feedback two-stage electro-hydraulic servo valve is established,and the typical failure modes and failure characteristics of the torque motor,the nozzle baffle,and the power-stage spool valve in the electro-hydraulic servo valve are analyzed.Then research on fault data preprocessing method of electro-hydraulic servo valve.In order to avoid the phenomenon of feature loss and feature confusion during the preprocessing of fault data of electro-hydraulic servo valve with traditional methods,this paper proposes a data normalization method based on basic rated parameters,which realizes the normalization of fault data of electro-hydraulic servo valve.Secondly,research on fault diagnosis algorithm based on message propagation mechanism.A Message Passing Neural Network(MPNN)algorithm based on the state collection of multi-source information systems is proposed for information fusion.Through the study of Graph Convolutional Network(GCN)and Transformer Graph Convolutional Network(GCN),the reason for the poor convergence effect of the stacked GCN classification model is analyzed,and a residual GCN classification model is proposed on the basis of the stacked GCN classification model.Compared with the stack graph integral model,this model has a higher convergence speed and accuracy rate,which shows that this model has a certain feasibility in the intelligent fault diagnosis of electro-hydraulic servo valves.Finally,research on fault diagnosis algorithm based on Feature Distillation(FD).In order to improve the operating efficiency of the algorithm model in the terminal,a fault diagnosis algorithm based on feature distillation is proposed.The residual GCN classification model is used as the teacher network,and the 1D CNN network is used as the student network,and the lightweight student network is trained through migration learning.Through experimentation and comparison,a separately trained student network is easy to fall into the local optimal solution,causing it to fail to converge to the global optimal solution,while the student network using characteristic distillation has reached a level close to that of the teacher’s network in identifying electro-hydraulic servo valve faults,and its operating speed is much faster than that of the teacher’s network.It is proved that characteristic distillation has a certain effect on improving the operating efficiency of the electro-hydraulic servo valve fault diagnosis algorithm in the terminal.In this paper,the MOOG G761-3004 electro-hydraulic servo valve has been tested to reproduce its fault conditions such as wear,seizure and coil open circuit,and verify the effectiveness of the fault diagnosis algorithm based on the message propagation network;and through the feature distillation,the model The operating efficiency of the model is improved without the accuracy rate being reduced,and the lightweight transfer of the model is realized.