| Insulated gate bipolar transistor required drive power is smal,with a small conductive resistance and small drive current,etc.,has been widely used in electric vehicles,railway locomotives and a new generation of aircraft,etc.,although the IGBT module in high power applications Wide,but its limited number of load cycles makes its reliability a problem,and the reliability of the IGBT directly affects the reliability and performance of these vehicle systems.In recent years,a series of research work on IGBT reliability,fault mode and aging analysis has been extensively carried out.Despite the study of IGBT reliability and failure modes and so on,the remaining useful life estimation of IGBT is still the key to the work of IGBT health management system,and it is becoming more and more important for IGBT's remaining life cycle algorithm.This paper first discusses the common failure of the IGBT module in the practical applicatio n of the IGBT module,and summarizes the root cause of the failure of the IGBT and is related to the external operating environment.the result of.And the establishment of the IGBT circuit model,simulation of the aging process of IGBT devices on their circuit to determine the appropriate stress.Then,based on the experimental principle of IGBT aging,the experimental hardware architecture is designed,and the design of IGBT drive circuit module with adjustable duty cycle is completed.The collector voltage is used as the health factor,and the aging time is recorded and the data are processed.The characteristics of the data and the randomness of the aging process of the IGBT are carried out.The stochastic process model of the aging stage is fitted by five stochastic process models.The optimal parameters of the model parameters are used to characterize the distribution model.These five distribution models are used to simulate the IGBT aging process stage.And then the residual life prediction process is put forward.In this paper,we use five stochastic process models for the IGBT module data in two different states,and combine the random number simulation method to generate the aging unit of each aging stage,and then combine the discrete data characteristics to deduce the expression of life prediction formula.The residual life of IGBTs is predicted and the mean value of life expectancy of the two devices is analyzed and compared with the mean square error.The results show that the prediction results of different models are different in the prediction of the remaining life of different models.The results show that the data are suitable for fitting the five models,and the prediction results of different models have their own characteristics.The results show that the model is different from each other.The different models also have different precision for different IGBTs.The prediction error is also within the reasonable range that is acceptable,which verifies the rationality of the forecasting method.And to verify the randomness of the aging process and prove the feasibility of modeling the aging data.The two groups of control groups,five stochastic models to minimize the aging stage,set a reasonable number of random number cycle,so as to achieve predict the accuracy,and feasibility of the algorit hm.Subsequent studies can be based on the application of sensor technology for differe nt warning parameters of the remaining life prediction,while the development of accelerated aging experimental design to achieve full coverage of the test stress,but also reduce the equipment requirements and operational difficulties,and from the point of view of material and mathematics The randomness of aging and whether all stochastic models can be adapted to the distribution modeling.The feasibility of the prediction algorithm and the accuracy of the algorithm are further optimized,and the particle filter algorithm in the lithium ion battery can be further used to solve the problem of instantaneous failure prediction The method can be considered from the circuit point of view. |
PDF Full Text Request