Research On Fault Diagnosis And Prediction Technology Of Power Electronic Circuit

With the rapid development of modern technology, modern electronic system is becominglarger and more complex, the probability of fault occurrence and function failure increases gradually.Power electronic circuit, as an important part of electronic system, is widely used in a variety ofelectronic products and systems. Its failure may cause the failure of the product and the wholesystem, leading to severe damage and great loss. In order to make sure that the electronic equipmentis always in good condition, we must be able to predict the faults that may occur in a short period oftime more accurately, thus to prevent and treat on the fault. All these make Prognostic and HealthManagement (PHM) technology receives more and more attention.Based on the characteristics of power electronic circuits, this paper studies the fault predictand health management technology of power electronic circuit. With the analysis of the output signal,this thesis works on health monitoring, fault diagnosis and prediction methods, including thefollowing:1. The degradation models of power electronic circuit components are acquired, so as thedegradation of the whole circuit; After wavelet analysis, the fault characteristic parameters of powerelectronic circuits can be extracted; And then based on the Mahalanobis distance and HMMtheory, the difference or similarity between the circuit under test and the standard circuit is calculated.Then the health state of the circuit can be acquired. Verified by BUCKBOOST simulation, the methodcan monitor the degradation of the circuit over time, and accurately judge the circuit’s health state.2. After the modeling of the parametric faults and structural faults of the components, thediagnosis of single soft and hard faults in power electronic circuit can be realized. And the BUCKsimulation results show that this method achieves a high correct rate of diagnosis.3. The predict Algorithms of AR model, GM theory and BP neural network were analyzed andameliorated. Based on the above algorithms, the future health state of power electronic circuit can bepredicted, as well as its remaining useful life. After all, based on the analysis ofBUCKBOOST circuit, the prediction validity is verified and comparison of the three methods ismade.