Design Of IGBT Module Condition Monitoring And Life Prediction System

In recent years,high-power power electronic technology has been widely used in smart grid,new energy power generation and other emerging fields.Power electronic technology is fully promoting technological innovation and social development.Each component of power electronic system will affect the reliability and stability of the system,and the key components play a decisive role in the stability and reliability of the system.Insulated gate bipolar transistors(IGBT)are often used as the core devices of power electronic systems because of their superior performance.However,with the continuous improvement of the power capacity of the system,the stability requirements of IGBT power modules are becoming more and more strict.In order to improve the reliability of IGBT power module,people have optimized the IGBT structure and upgraded the module radiator performance,but in the harsh and unstable operating environment,the system failure is still inevitable.Therefore,it is very important to monitor the state of IGBT power module and predict the remaining service life of IGBT for the safety and stability of the whole system.Therefore,it is necessary to deeply analyze the failure mechanism of IGBT to solve the problem of condition monitoring and life prediction of IGBT power module.The main contents of this paper are as follows:(1)the structure and failure mechanism of IGBT are analyzed.The fatigue failure mechanism of IGBT silicon material is analyzed in detail at the material physical level.It is concluded that as the IGBT device is constantly subjected to the impact of thermal stress,the silicon material of the chip is fatigued,the carrier mobility changes,resulting in the change of conductivity,and the saturation voltage drop of the IGBT chip increases with the increase of the fatigue degree of the silicon material of the chip,and the collector emitter saturation voltage drop increases(2)The principle and method of IGBT accelerated aging experiment are analyzed.Taking the IGBT accelerated aging platform developed by NASA as an example,the experimental platform and IGBT accelerated aging process are analyzed.The collector emitter saturation voltage drop,collector emitter current,gate emitter voltage and package temperature smoothed by filtering are used as the input of BP neural network prediction model,At the same time,BP neural network optimized by mind evolutionary algorithm is used to monitor the status of IGBT module(3)Life prediction of IGBT.This paper introduces the model of life prediction and the method of life prediction,and puts forward the method of using particle filter to predict the life of IGBT.Through the analysis of the accelerated aging data,it is found that the collector emitter saturation voltage drop increases gradually with the increase of the fatigue degree of IGBT chip silicon material.Therefore,the collector emitter saturation voltage drop is selected as the characteristic parameter for IGBT life prediction.Based on the accelerated aging data of collector emitter saturation voltage drop,the fitting degree of various models to the data is compared.Finally,the IGBT life model is selected as the exponential form,and the state equation and observation equation of IGBT aging are established.Finally,the process of life prediction based on particle filter method is summarized,and the residual life prediction model based on IGBT Collector Emitter Saturation Voltage drop is established.With the increasing of training samples of collector emitter saturation voltage drop,the error of IGBT residual life prediction model is decreasing.According to the above research content,this paper proposes the particle filter method to predict the IGBT life,and provides a feasible method for IGBT real-time online residual life prediction.