Research On PHM Technology Of Avionics Analog Circuit Module

As requirement of Modern defense technology for the function integration of avionics system continues to improve,the probability of the failure occurrence is becoming bigger and bigger.And how to guarantee the fighting efficiency turns to be the most serious problem at present.The application of the Prognostics and Health Management(PHM)can improve the reliability of the system,and can realize the fault early warning,fault diagnosis and on-condition maintenance.Therefore,researching the fault diagnosis theory and the remaining useful life(RUL)prediction algorithm has a great theoretical value and practical application significance.This paper studies the failure diagnosis algorithm for the typical circuit of the avionics circuit module,and studies the RUL prediction method concerning the typical component in the avionics circuit module.The main tasks are shown as follow:(1)That the classification accuracy is not stable is caused by the randomly generated parameters of the extreme learning machine(ELM),for this problem,the differential evolution algorithm(DE)is applied to search the optimal solutions of the two random parameters.Then by comparing the diagnostic accuracy with applying ELM and DE-ELM to verify the better fault diagnostic ability of the optimized diagnostic algorithm.(2)Research the degradation mechanism of the frequently-used component which has a high damage rate in the avionics circuit under the high electrical stress,and apply the particle filter(PF)to predict RUL of the electrolytic capacitors whose degradation dataset has been published by NASA.Then,to address the problem of particle degeneration,the unscented Kalman filter algorithm(UKF)is applied to generate the appropriate proposal distribution to optimize the step of importance sampling based on the particle degradation occurred in PF,and enhance the RUL prediction accuracy.(3)For the drawbacks,particle impoverishment,occurred in the resampling step of particle filter,the paper proposes to apply two method,linear optimization resampling and particle swarm optimization algorithm(PSO),to improve the problem of particle impoverishment.In addition,for raising the prediction efficiency of the algorithm,we apply a method to adjust the particle numbers adaptively considering the importing of PSO increases the computation burden and lengthens the running time.The paper applies the optimization algorithm to improve ELM,and realizes precisely diagnosis for the typical avionics circuit with the optimized diagnosis algorithm,the accuracy can be improved more than 2% compared with the traditional ELM.Besides,proposes the improved RUL prediction algorithm based on UKF,PSO and LOR,and the method that adaptively adjusted the particle number,when we apply the algorithm to predict the RUL of the electrolytic capacitor which has been worked under the high voltage stress for 60 hours,the prediction error decreased about 10% compared with the prediction error applying particle filter,that is to say,the prediction accuracy has been increased to a great extent.And the comparison of the prediction results verify that the method can well improve the drawbacks of PF,and can raise the ability of RUL prediction.