Junction Temperature Prediction And Multiphysics Simulation Failure Analysis Of IGBT Module

Power semiconductor devices are widely used in various fields of daily life,and IGBT modules are highly favored by various industries due to their fast switching speed,high input impedance,and low conduction voltage drop.However,as the demand for kHz-level switching frequency increases and manufacturing processes develop,the power density of IGBT modules has been greatly improved,and the large amount of heat generated during operation causes fluctuations and rises in the internal junction temperature of the module,accelerating its aging and failure,which ultimately leads to equipment damage and even threatens personal safety.Therefore,obtaining the internal junction temperature of IGBT modules under predetermined operating conditions is of great research significance for ensuring reliable operation of the module and preventing internal device failures.The existing thermal parameter method based on deep learning for solving junction temperature has strong dependence on experimental data and high cost,while the equivalent thermal resistance network in the electro-thermal coupling model method is usually one-dimensional.With the emergence of Physics-informed Neural Networks,the constraint of physical information reduces the need for data samples and facilitates the solution of multidimensional complex models,providing a new research idea for the estimation of junction temperature in IGBT modules.This paper proposes a new method for estimating the junction temperature of IGBT modules,which uses the Conservative Physics-informed Neural Networks to solve the heat conduction equation to obtain the module junction temperature.Based on the package structure and heat dissipation method of IGBT modules,this algorithm decomposes the different material layers inside the module into different solution regions,and each subdomain is solved by a separate Physicsinformed Neural Networks.Finally,the module junction temperature is obtained by adding the subdomain boundary coupling term to increase the overall coherence.Using the electro-thermal coupling model method to estimate the junction temperature fluctuation range of the IGBT module as a reference object for the new method,the results show that the junction temperature values obtained by the new method are within the junction temperature fluctuation range,and the error between the mean values of the two methods is about 4℃,verifying the feasibility and accuracy of the new method.In order to further analyze the influence of junction temperature on the overall reliability of IGBT modules before and after device failure,this paper used COMSOL finite element simulation software based on multi-physics field coupling to conduct multidimensional comparative analysis on two common failure modes,namely bond wire failure and solder layer failure.By comparing the cases of different numbers of bond wire detachment,it was found that under fixed current and heat dissipation environment,the module would not fail immediately when there was bond wire detachment,and the current would flow through the remaining intact bond wires,resulting in an exponential increase in the junction temperature inside the module.Due to the increase of thermal stress and thermal strain values with the junction temperature,the remaining intact bond wires would fail faster,ultimately shortening the service life of the IGBT module.By comparing the different sizes and positions of voids in the solder layer,it was found that the larger the voids inside the solder layer,the poorer the heat dissipation performance of the module,and the higher the junction temperature.In addition,the thermal stress and deformation values that increase with temperature also further expand the voids.However,when the void was located far away from the heat source,the temperature rise caused by the void had little negative impact on the module.