Cost-Effective Prognostics of IGBT Bond Wires With Consideration of Temperature Swing

Abstract

This article presents a cost-effective prognostic method for the bond wires in the insulated-gate bipolar transistor (IGBT). Consider that the crack propagation in the wire bond leads to the bond wire liftoff, the corresponding state equation is established from the fracture mechanics theory, with the consideration of the uneven distribution of the temperature swings. Hence, the proposed model can work under different loading conditions. With the fact that the on -state voltage $(v_{{\rm{ce,on}}})$ of the IGBT shifts with the crack propagation, the history $v_{{\rm{ce,on}}}$ is used to predict the remaining useful lifetime (RUL), through which numerous power cycling tests are avoided, and low economical cost for doing prognosis is fulfilled. In this article, the functional relationship between the increase of $v_{{\rm{ce,on}}}$ and the crack length of each bond wire is obtained through finite-element simulation, while the effects of the temperature variation and metallization degradation to the $v_{{\rm{ce,on}}}$ are compensated. Thus, the output equation can be obtained. Then, the unknown parameters of the aforementioned equations and the current crack length can be estimated by the particle-based marginalized resample-move algorithm. Finally, the RUL can be predicted effectively by evolving the particles obtained in the algorithm. The proposed method has been validated by the power cycling test.