Research On Testability Design And Fault Diagnosis For Switched Mode Power Supply

Switched mode power supply(SMPS),as an important energy supply equipment for electronic systems,of which accurate health condition monitoring and evaluation is the key to ensure the efficient operation of the system.Therefore,the research of testability design and fault diagnosis of SMPS,which can improve the state observability and diagnostic ability,has the great value to ensure the safe operation of SMPS and even the whole system.Recently,the existing researches on testability design of electronic systems lacks the combination with fault mechanisms,and mainly focuses on qualitative testability modeling,which cause problems such as lack of online testable features of faults and low diagnostic accuracy under the influence of uncertainties in testing process.With the development of integrated and intelligent electronic devices,it is more difficult for the existing testability and fault diagnosis methods to meet requirements due to devices' characteristics of high dimensional test points,strong non-linearity and interference.On the basis of existing researches,this dissertation takes SMPS as an example,and studies the testability design and corresponding fault diagnosis method with the combination of online testable feature mining based on component-level degradation mechanism,system-level testability modeling,quantitative analysis and optimization of test uncertainty and multi-interval dictionary-based fault diagnosis,attempting to establish a general testability design and fault diagnostic method.First,an online sensitive feature analysis method of power MOSFET combined with degradation mechanism is proposed.The formation mechanism of two kinds of defect charges causing the MOSFET gate oxide degradation is studied,and the influence differential of two defects on electrical characteristics is analyzed.With the accelerated test,the time-varying law of components' electrical performance during the gate oxide degradation is explored,thus specifying the dominant position of two kinds of defects.Combined with the typical operation mode analysis of MOSFET,the Miller platform voltage,which is sensitive to defect charges,is extracted to identify gate oxide degradation.Components with different degradation levels are then prepared for online condition monitoring to verify the feasibility.Second,combining graph theory with simulation technology,a hierarchical modeling method for testability of SMPS is studied.In view of the characteristics of high dimensional test points and strong nonlinearity,testability modeling is divided into two levels: test point screening and test item(feature)mining.With established multi-signal flow graph model and information entropy method,a method for screening available test points is proposed,and test points are prioritized.According to determined test point priority,feature mining based on Saber simulation model,as well as test set supplementation method,is then studied to generate the candidate test set and the corresponding fault-test dependency matrix,so as to implement the testability modeling of SMPS.Then,the system testability optimization method constructed by the Copula function is studied considering the influence of imperfect tests.The common sources of uncertainties are analyzed,and the test response model with consideration of SMPS inherent uncertainty is built to prove the non-independent influence of homologous interference on multivariate test.Based on the assumption of the non-independent multivariate test,a testability optimization model in the form of joint distribution is established,and the model construction method based on Copula theory is studied to improve the testability estimation accuracy.With the developed model,the genetic algorithm is used to solve the optimization problem,and obtain the optimal test subset with consideration of SMPS inherent uncertainty.Finally,according to the optimized test subset before,the optimized fault dictionary based on dictionary type improvement and dictionary feature transformation is proposed to improve the diagnostic performance of traditional fault dictionares.A new multi-interval dictionary is developed,and the method of feature interval partition based on density peak clustering algorithm is studied to reduce the influence of "ambiguous space".Based on the new dictionary type,linear discriminant analysis-based feature transformation method is further studied,to improve the diagnostic accuracy and its stability.The optimized fault dictionary is then used for SMPS to achieve the fault diagnosis.