Research On Igbt Condition Monitoring Based On Transient Analysis And Global Feature Information

The health condition of insulated gate bipolar transistors（IGBTs）is of great significance for the efficient and safe operation of traction converters.According to industry research and field failure data,IGBT is one of the most failure-prone devices in converters.Therefore,using the effective information of external characteristic parameters to achieve accurate monitoring and evaluation of IGBT operation health status is important to the safe operation of traction converters.However,the existing condition monitoring method is mostly based on local characteristic parameters,which is easily disturbed by factors such as temperature and load current,and make it difficult to apply in practice effectively.Aiming at this problem,this paper mainly studies the condition monitoring method based on transient analysis and global feature information,reduces the adverse effects of temperature,load current and other factors,and provides new ideas for monitoring the IGBT bonding wires ageing and IGBT junction temperature.Firstly,in order to study the switching transient of IGBTs,the equivalent circuit models for trench-gate field-stop IGBT chips and soldered modules are established separately.In which,the static characteristics of IGBT are represented by darlington structure composed of metal-oxide-semiconductor field-effect transistor（MOSFET）and bipolar junction transistor（BJT）,and the dynamic characteristics of the IGBT are represented by three inter-electrode capacitances composed of various parasitic capacitances.Then,a bonding wires aging monitoring method for IGBT based on the global feature information of module gate voltage(u_gem)is proposed.at first,a systematic theoretical analysis for the IGBT turn-on transient is carried out by using the equivalent circuit model,and the internal mechanism why the u_gem affected by the bonding wires ageing is studied.Secondly,a double-pulse test platform is built for theoretical verification,and the influence of temperature and other factors is further studied.Moreover,aiming the problem that the aging monitoring used u_gem local characteristic parameters is easily affected by factors such as temperature,a bond wire fracture detection method based on the u_gem global feature information is proposed,in which the supervised linear discriminant analysis（LDA）algorithm is used for global features reduction,and the support vector machine（SVM）is used to realize the bond wires fracture detection.Experiments are used to verify the effectiveness of the monitoring method studied.Finally,a junction temperature monitoring method for IGBT based on the global feature information of turn-off voltage and convolutional neural networks（CNN）algorithm is proposed.First,a systematic theoretical analysis for the IGBT turn-off transient is carried out by using the equivalent circuit model,and the internal mechanism why turn-off voltage affected by temperature is studied.Secondly,the theoretical verification is carried out based on the double-pulse experimental data,and the influence factors such as load current is further studied.Futhermore,aiming the problem that the junction temperature monitoring used turn-off voltage local characteristic parameters is easily affected by factors such as load current,a junction temperature monitoring method using graph samples of turn-off voltage and convolutional neural network is proposed.Experiments are used to verify the effectiveness of the monitoring method studied.