Prognostic Health Monitoring Method for Thermal Fatigue Failure of Power Modules Based on Finite Element Method-Based Lagrangian Neural Networks

Abstract

Prognostic health monitoring technologies for power electronic systems assess their performance degradation, load histories, and degrees of fatigue in order to increase maintenance effectiveness, reliability design methods, and equipment availability under conditions of actual use. To improve reliability and reduce downtime, prediction of reliability in terms of thermal fatigue life under field conditions is important, as is the use of load and health monitoring data from the field in cases of performance degradation during use, maintenance, and field failure. The fatigue life of solder joints is also affected by whether the load history waveform is symmetric or asymmetric. In this paper, we propose a novel health monitoring method for thermal fatigue failure corresponding to time-dependent inelastic strain response, such as in asymmetric cycles, by use of a surrogate model obtained by a finite element method-based thermal stress simulation. We applied this method to an insulated-gate bipolar transistor power module capable of monitoring module temperature, electrical performance, and number of revolutions of the cooling fan. With the proposed method, inelastic strain cycles and thermal fatigue life distribution of solder joints could be estimated from their temperature monitoring history. The method was judged to be useful for assessing thermal load histories and estimating thermal fatigue life in prognostic health monitoring.