Research On Life Evaluation Method Of IGBT

With the continuous development of semiconductor technology,the research,of power modules are increasing mature and the performance achieve great improvement,which have been used widely in many fields.Working under the non-stationary conditions,improper use and long time operations make IGBT with a variety of faults,which accelerates aging failure degree of device.It will greatly reduce the life of device and affect the normal operation.Therefore,it is urgent to make evaluation of IGBT life.The main work of this paper is to analyze the failure problem of IGBT under the accelerated aging experiment.Firstly,the failure mechanisms of IGBT are described,and the life evaluation is analyzed.After studying the accelerated aging data of IGBT,this paper selects the collector-emitter off voltage spike as the characteristic parameter of IGBT lifetime evaluation,and then uses Saber simulation model to analyze the change trend of the parameter in the off process.According to the accelerated aging data of IGBT provided by NASA PCoE research center,we will deal the data with the first exponential smoothing and second exponential smoothing.Then the model of BP neural network,MEA-BP and NAR dynamic neural network are established separately to assess the life of IGBT.According to the prediction result,the evaluation index of the mean square error MSE and coefficient of determination R2 will be used to determine the prediction accuracy.The results show that the MEA-BP life evaluation model with data of second exponential smoothing gets the best prediction result,so as to provide the correlation analysis of the life evaluation of IGBT.The experiment results are carried out based on the accelerated aging data of IGBT provided by NASA PCoE research center.The results show that the method proposed in this paper can be applied to the life evaluation of IGBT,and it has certain engineering application value.