Aging Test And Rul Prediction Of IGBT Module For Motor Driver

Insulated gate bipolar transistor（IGBT）is an important electronic device in motor driver.However,the IGBT module is made of multilayer materials,and the CTE of each layer is very different.When the temperature of IGBT changes,periodic thermal stress will be produced in different layers due to different strain,which will lead to the accumulation of stress fatigue damage in welding joint,and eventually lead to failure.And its chain reaction may damage the system and other related components.In most cases,the whole equipment will be shut down,resulting in project delay and economic loss of maintenance.Therefore,the use of an IGBT（IGBT）to diagnose the remaining life of the IGBT is urgently needed.At present,most of the IGBT prediction methods use the Collector Emitter conduction voltage(V_CE,on)as an aging parameter to evaluate the aging degree and predict the remaining life of IGBT.However,V_CE,onare easily affected by temperature,and the change rule of V_CE,onwith aging time of IGBT modules of different types or manufacturers is not the same.Therefore,the residual life of IGBT can not be predicted accurately by V_CE,on.In this paper,the aging mechanism of IGBT is thoroughly studied by designing accelerated aging test.From two aspects of electrical characteristics transient V_CEand thermodynamic characteristics transient thermal resistance,combined with neural network,life prediction models are established to predict the remaining life of IGBT.（1）Build IGBT module aging experimental platform and collect aging data.Firstly,after the investigation of different aging experimental strategies and temperature measurement schemes,a feasible experimental strategy is selected according to the laboratory conditions and actual working conditions.Then,the selection of instruments,materials,and equipment is done according to the test scheme,and the accelerated aging test bench are built.Finally,the aging test is conducted,and IGBT modules are working from intact aging to failure,and collect the data needed for subsequent research.（2）From the perspective of transient V_CE,this paper proposes a method to predict the RUL of IGBT based on Deep AR.Firstly,the transient signal when IGBT is turned off is intercepted from the aging data;secondly,the features that can represent the health status of IGBT are extracted;and then the RUL of IGBT is predicted by using Deep AR after appropriate filtering.Through experimental analysis and verification,it is found that Deep AR can well predict the life of IGBT,and its results are more accurate than other similar prediction methods.At the same time,in order to further understand its aging mechanism,this paper simulates the voltage at the moment of IGBT turn off.（3）From the point of view of transient thermal resistance,this paper proposes a RUL prediction method based on attention mechanism.Firstly,data cleaning,the characteristics of transient thermal resistance curve are extracted.Then an attention mechanism neural network is trained to predict the life of IGBT.At the same time,this paper establishes the IGBT three-dimensional simulation model,it is found that the temperature fluctuation of the bonding wire is the largest,and the bonding wire falling off fault is most likely to occur.（4）An intelligent monitoring platform including edge end,user end and server side is developed to realize real-time status monitoring and fault detection of heavy load AGV motor,transmission shaft and driving components IGBT.