Abstract Solder voids has an important impact on healthy operation of insulated-gate bipolar transistor (IGBT) modules, but voids damage is difficult to be estimated accurately by conventional methods due to the lack of calculation in mutual thermal effect between chips. So this paper proposes a base solder voids identification method by using case temperature, takes into account both self-thermal diffusion and coupling inside module, and adopts case temperature distribution on bottom of baseplate to assess voids damage indirectly. Firstly, finite element method (FEM) of IGBT module with base solder voids was established to study junction and case temperature response with different voids distribution, the practicability of base solder voids identification based on case temperature is verified. Then the influence mechanism of voids on case temperature was discussed theoretically, voids identification is extended to dynamic load conditions, temperature-to-thermal resistance conversion and dynamic thermal resistance reference calibration with different power loss ratio were realized by improved thermal network, recursive least square (RLS) algorithm was designed for dynamic identification of real-time identifiable thermal resistance, the difference between RLS results and thermal resistance reference was compared to confirm the severity of base solder voids, two or more groups of comparison results with different power conditions are used to complete the voids analysis. Finally, validity of proposed method is demonstrated by simulation. Compared with conventional method, it shows better ability in dynamic identification.

中文翻译：

---

使用外壳温度识别 IGBT 模块的基础焊料空隙

摘要 焊料空洞对绝缘栅双极晶体管（IGBT）模块的健康运行有重要影响，但由于缺乏对芯片间热效应的计算，传统方法难以准确估计空洞损坏。为此，本文提出了一种利用壳温识别基底焊锡空洞的方法，同时兼顾模块内部的自热扩散和耦合，采用底板底部的壳温分布来间接评估空洞损伤。首先，建立了含基体焊料空洞的IGBT模块有限元法(FEM)，研究了不同空洞分布下的结和壳温响应，验证了基于壳温识别基体焊料空洞的实用性。然后从理论上讨论了空洞对外壳温度的影响机制，将空洞识别扩展到动态负载条件下，通过改进的热网络、递归最小二乘法实现了不同功率损耗比下的温度-热阻转换和动​​态热阻参考校准。 (RLS)算法设计用于实时可识别热阻的动态识别，比较RLS结果与热阻参考值的差异，以确认基焊空洞的严重程度，使用两组或多组不同功率条件的比较结果完成空洞分析。最后，通过仿真证明了所提出方法的有效性。与传统方法相比，它显示出更好的动态识别能力。