Research On Service Performance Evaluation Of IGBT Considering Multiple Characteristic Parameters

With the development of modern technology and industrial technology,the engineering systems in the fields of electric power and energy such as rail transit,aerospace are becoming more and more complex,resulting in the continuous improvement of the degree of intelligence of the system.This brings with it an increase in the cost of system design,production,and especially support and maintenance.Circuit topology with semiconductor device as the core unit is an indispensable and important component in every complex engineering system.Among them,the Insulated Gate Bipolar Transistor（IGBT）is widely used in various industrial systems due to its advantages of low on-state voltage drop and low driving power.IGBTs usually work under frequent voltage,current stress or other severe environmental conditions.For a long time,fatigue damage will accumulate inside the device and eventually lead to failure,which may further lead to system failure and unnecessary economic losses.Therefore,research on the reliability of IGBT is meaningful.This paper takes IGBT as the research object,and completes the construction and improvement of accelerated aging test,and collects multiple aging characteristic parameters.Then,the dimensionality reduction and fusion of aging feature parameters is carried out to further effectively evaluate the health status of the device.Finally,the improved particle filter and curve fitting methods are used to predict the aging trend of the IGBT,and it gives an early warning before the aging failure of the device occurs.This prevents more serious security problems from happening.In this paper,a research on IGBT service evaluation based on multi-characteristic parameters is carried out.The main research contents include:Firstly,the basic structure and working characteristics of IGBT are expounded,and the working characteristics of IGBT devices are further analyzed and studied.The failure mode and failure mechanism of IGBT are isolated,and the changing trend of different aging parameters during the aging process of the device is discussed.It is pointed out that a single aging characteristic parameter cannot cover all the performance degradation information of the device.Therefore,considering multiple aging characteristic parameters can more accurately predict the performance degradation of IGBT.The three aging characteristic parameters of collector-emitter saturation voltage V_CE（ON）,turn-off time t_offand turn-off voltage peak Vpeak are selected to study the degradation state of IGBT.Secondly,the factor analysis method is used to realize the removal and dimensionality reduction of redundant information of multi-feature parameters,and generate a one-dimensional health factor.The method includes data standardization,moving average filtering,etc.Then,aiming at the difficulty of measuring the collector-emitter saturation voltage drop,a saturation voltage drop measurement circuit is designed,which can measure the saturated conduction voltage drop of the device.At the same time,the experimental platform was further improved,and the power cycling capability of the device was improved.On this basis,three aging characteristic parameters were collected for feature extraction and dimension reduction fusion;the construction of health factors was completed,and verification and correlation analysis were carried out.To a certain extent,this method reflects the IGBT reliability characterization of dimensional reduction fusion of various aging parameters.Finally,it is pointed out that the aging characteristic parameters of IGBT are a set of time series,and the degradation problem of IGBT can be regarded as a time series prediction problem for characteristic parameters.The particle algorithm of unscented Kalman filter is used to predict the characteristic parameters of the saturated voltage drop with noise,and the curve fitting method is used to predict the off-time toff,the off-voltage peak Vpeak,the health factor HI and so on.Aiming at the degradation phenomenon of traditional particle filter,the filtering algorithm is improved,and the number of training set cycles is increased to verify the performance of the algorithm.It is found that the improved particle filter can generally achieve the degradation prediction of IGBT.According to the fitting evaluation parameters,the other three parameters are predicted by the linear interpolation method and the polynomial fitting method.In summary,this paper completes the reliability characterization method based on IGBT multi-characteristic parameters,and collects experimental data by building and improving the accelerated aging test platform.This paper also extracts the collected data by factor analysis,and finally uses the improved particle filter algorithm to carry out the analysis.For prediction and verification,it has certain guiding significance for practical engineering applications.