Research On Intelligent Fault Diagnosis And Testability Verification And Evaluation Methods Of Equipment

The ability of fault diagnosis and the level of the testability are important indicators to measure the overall performance of aviation equipment,which are the basis of the equipment acquisition,acceptance,scientific decision-making,health-management.In order to improve ability of fault diagnosis and credibility of testability validation and evaluation,the intelligent fault diagnosis technology of analog circuit with tolerance,the optimal allocation method of fault samples,and the comprehensive evaluation method of testability are studied.At present,the accuracy of fault diagnosis for airborne electronic equipment is relatively low,in order to resolve the problem,a new method of fault diagnosis based on Ada Boost is proposed.Firstly,the effective point extraction method is used to extract the fault characteristic of the analog circuit.Secondly,GA algorithm is used to optimize BP neural network to construct GABP classifier.Finally,the GABP network was boost by the AdaBoost algorithm to form a set of complementary classifiers,and then use the combination classifier to identify the fault pattern.The example shows that the method proposed has higher accuracy and lower error.Aiming at the problem that the result of failure samples allocation is unreasonable,a multiple index integrated weighted method is proposed.Firstly,the effect of the failure attribute and environment factor on the accuracy of failure samples allocation is analyzed.Secondly,the prediction model of neural network is used to improve the precision of fault rate,fuzzy pattern recognition method is used to determine the fault severity level and its damage degree,and the grey correlation analysis method is used to calculate the correlation degree between failure and environment.Finally,an integrated weighting method is proposed to obtain weights.The example shows that the method proposed improve the confidence of the failure samples allocation result.The data of testability growth test has characteristics of "small sample,multi stage,varying population",which brings the difficulty of prior information fusion and the problem of evaluating the testability level.To solve the above-mentioned problem,an optimized method based on dynamic Bayes is proposed.Firstly,the new Dirichlet distribution is used to establish a new dynamic growth model of testability level.Secondly,the D-S interval evidence reasoning theory is introduced to fuse the expert information.After that,the nonlinear optimization theory is used to fit prior information to calculate hyper-parameters of the model.Finally,the Bayes information fusion theory is used to infer the multiple joint posterior distributions of testability,and Gibbs sampling algorithm is used to obtain the high dimensional posterior integrals.The example shows that the method proposed can fuse interval expert information effectively and has higher accuracy and lower error.