Research On Circuit Fault Diagnosis Method Based On Behavior Modal

With the popularization of electronic devices and the enhancement of their scale and complexity,especially in the fields of aviation,spaceflight and military defense,the reliability and security of electronic devices seem particularly important.As an important means to improve the security and reliability of electronic equipment,circuit fault diagnosis can not only find out whether the circuit is faulty,but also realize failure location in the faulty circuit.Therefore,the circuit fault diagnosis is of great significance.For the different scale of circuit,there are differences in the difficulty of fault diagnosis.For small-scale circuits,the traditional fault diagnosis method can get better diagnostic results;For large-scale digital-analog hybrid circuits where there are numerous fault modes and circuit failure status is complicated,therefore the circuit diagnosis difficulty is larger and the circuit needs to adopt a new diagnosis method.The advent of the big data era provides a new train of thought for the fault diagnosis of the system-level circuit.Combining machine learning,behavior model and circuit fault diagnosis,this paper puts forward the circuit fault diagnosis method based on behavior model.The main context is:The data source of fault diagnosis modal and behavior model is obtained by circuit simulation,data preprocessing,standard normalization,dimension reduction and fault mergence and other means.Among them,the principal component analysis is used to reduce the dimensionality of the data to solve the problem of feature-redundancy in the samples;The dichotomous K-means clustering is used to cluster the similar fault modes to solve the problem of fault-overlap,thus reducing the complexity of the model.The fault diagnosis model in circuit fault diagnosis includes fault detection model and fault classification model.In view of the different number of faults in the circuit at the same time,a single-fault diagnosis method and a multi-fault diagnosis method are proposed.The single-fault diagnosis method includes a single-step fault diagnosis method based on classification model and a two-step fault diagnosis method based on anomaly detection model.In the single-step fault diagnosis method based on classification model,the fault classification model is established by BP neural network or Support Vector Machine to troubleshoot the single fault of simple cicuit;In the two-step fault diagnosis method based on anomaly detection model,the fault detector model is constructed by anomaly detection algorithm,and the fault classification model is established by classification algorithm(BP neural network),and then the models are used respectively at two levels of the single fault location in the system level circuit;In the multi-fault diagnosis method,the anomaly detection algorithm is used to construct the target class filters of fault to filtrate multi-fault source step by step in system level circuit.Proved by examples and compared with the traditional methods,those three methods achieve better effect on fault diagnosis and have inproved significantly fault coverage rate and fault detection coverage rate.In order to improve the efficiency and accuracy of system-level circuit fault diagnosis,behavior model is introduced into circuit fault diagnosis,including behavior simulation model and failure behavior model.The behavior simulation model can shorten the simulation time and improve the diagnostic efficiency;The failure behavior model can improve the accuracy of circuit fault diagnosis.