Modeling And Fault Diagnosis Of Aero-engine Fuel Regulator

As an important part of the aero-engine control system,the performance of the fuel regulator determines the safety and reliability of the aircraft.Due to the complex structure of the fuel regulator and its high technical content,once it fails,it is not easy to locate,and it may even lead to serious accidents.Therefore,it is of great significance to carry out modeling and fault diagnosis research of the fuel regulator.This paper takes the A-type fuel regulator used in the XX-type aircraft engine as the research object.First of all,the structure and working principle of the fuel regulator are analyzed.Based on the AMESim software simulation platform,the component level model of the fuel regulator is established.Then,the function module model is established on the basis of the component model.Finally,using the established component model and function module model,the complete model of the fuel regulator is established,as well as simulated and analyzed.The precondition for fault diagnosis of the fuel regulator is to have sufficient fault sample data.Considering the fact that the fault data is difficult to acquire from engineering practice,the fault data used in this paper is obtained through simulation.Firstly,the mechanism leading to the hydraulic system fault is detailed analyzed.Based on the established fuel regulator model,faults are injected into the fuel regulator model by adding fault parameters.The fault curves and fault characteristics are obtained.Then,combining particle swarm optimization(PSO)with neural network,a faut diagnosis method of the fuel regulator called PSO-BP neurak network is proposed,where the PSO algorithm is utilized to optimize the connection weights and biases of the BP neural network.Finally,two cases,single fault mode and double faults mode,are used to verify the effectiveness of the proposed method.The results show that the PSO-BP neural network can make accurate diagnosis of the faults in both cases,and it has a more rapid convergence speed and higher precision than classical BP neural network,which providing a new way in the fault diagnosis field of aero-engine fuel regulator.