Abstract: This paper presents mean fatigue lifetime prediction of a wire-bond structure model in power electronic module using a failure physics approach that integrates high fidelity modelling and reduced order modelling. Loading current with variable amplitudes is applied to a finite element model of simplified wirebond structures. The resulting accumulated fatigue damage due to random loads is predicted by using reduced order modelling based on failure physics, a cycle counting algorithm, and various nonlinear fatigue damage models widely used in the literature. The reduced order modelling approach based on failure physics uses prediction data for the electro-thermo-mechanical behaviour of the wire-bond design of a powermodule obtained through non-linear transient finite element simulations, in particular for the fatigue life-time of the aluminium wire attached to the silicon chip of thewire in the module. The reduced order models that capture the black box function of the accumulated plastic strain are used in predicting the mean fatigue life time of the wire bond structure under random loads. One of the widely used cycle counting algorithms, rainflow counting algorithm, is used to count cycles of the temperature profile at the specific point of the wire bond structure in a power electronicmodule. The cycle data fromthe rainflow algorithmmean life time of the wire bond structure are predicted with various cumulative fatigue models. Non-linear cumulative fatigue models such as damage curve approach (DCA), double linear damage rule (DLDR), and double damage curve approach (DDCA), and linear cumulative fatigue damage model such as Palmgren–Miner rule are used to predict the mean fatigue life of the wire bond structure, and the resultsare compared.

Key words: Reduced order models , Power module , Kriging , Radial basis , Damage rule