Prognostic Reliability Analysis of Power Electronics Modules

This paper describes a physics-of-failure (PoF) based prognostic method for power electronics modules (PEMs). Differing from the traditional reliability prediction methods, this approach allows the reliability performance of PEMs to be assessed in real time. Four techniques have been used to develop this method, they are: (1) Compact electro-thermal model (2) Rainflow counting algorithm (3) Compact thermo-mechanical model and (4) Lifetime consumption model. As a demonstration, this method has been applied to a typical IGBT half bridge module and solder joint fatigue was assumed as the major failure mechanism. In this application, a random electric current load profile was generated in laboratory environment and used to derive the thermal loading condition for the module. Due to the randomness of the load profile, rainflow counting method was used to reduce the continuous load profile into discrete sets of thermal cycles. The damage induced in each temperature cycle was calculated via a compact thermo-mechanical model, and used in the lifetime model to calculate the PEMs lifetimes under simple cyclic loading conditions. Based on these predicted lifetimes and the linear damage accumulation rule, the total consumed life of the PEMs over the whole period of usage was predicted.