This paper demonstrates a methodology for taking variation into account in thermal and fatigue analyses of the die attach for an inverter of an electric traction drive vehicle. This method can be used to understand how variation and mission profile affect parameters of interest in a design. Three parameters are varied to represent manufacturing, material, and loading variation: solder joint voiding, aluminum nitride substrate thermal conductivity, and heat generation at the integrated gate bipolar transistor, or IGBT. The influence of these parameters on temperature and solder fatigue life is presented. The heat generation loading variation shows the largest influence on the results for the assumptions used in this problem setup.
INTRODUCTIONPower inverters are key components of the electric traction drives of hybrid electric, plug-in hybrid electric, fuel cell electric, and pure electric vehicles [1, 2]. Research and development of power inverters is currently focused on reducing the cost, weight, and volume of the technology while maintaining and/or increasing reliability [1, 2]. If the research goals related to cost, weight, volume, and reliability can be met, more advanced vehicles can enter the marketplace and reduce the amount of petroleum required for transportation.Current-generation inverter designs used in today's hybrid vehicles are built on proven technology that has demonstrated high reliability. In contrast, the next generation of inverter concepts combines new materials, new heat transfer methods, and new packaging concepts under increased loads and in smaller spaces. Furthermore, there is interest in expanding the range of temperatures and stresses under which these technologies can operate [3, 4]. For these new technologies and conditions, reliability is, as yet, unproven.In a production environment, reliability data can be obtained from statistical measures of empirical data or from product testing. However, design changes during production to fix reliability problems are quite expensive. In contrast, during the early phases of research, design changes are cheap to make. However, prototypes may not even be available. Therefore, at the very earliest stages of design, other methods must be used to estimate reliability (see page 3 of [5] for further discussion).