Bayesian Inference for Susceptibility of Electronics to Transient Electromagnetic Disturbances With Failure Mechanism Consideration

It is a necessity to evaluate the susceptibility of electronics to transient electromagnetic disturbances, such as electromagnetic pulses, lightning, electro-static discharges, and switching pulses. One can conduct experiments on the system of interest and analyze the result through statistical inference for the relationship between electromagnetic excitations and effect phenomena. However, since conventional statistical models are driven by experimental data, it might result in inappropriate explanations of the effect mechanisms due to limited samples, leading to over- or under-estimations. To address this problem, we propose a hybrid model that combines the statistical estimation with the known failure mechanism obtained from either the theoretical analysis or the empirical evidence. The physical cause can be drawn from the mechanism analysis, and then a statistical classifier is introduced to build the relationship between the causes and the effects. Based on Bayesian assessment, all the parameters involved are estimated with the Markov chain Monte Carlo method. Numerical simulation results have verified that the model is capable of estimating the susceptibility of electronics with given failure mechanism assumptions and limited test samples. In the end, the breakdown characteristic of the gas discharge tube is investigated to demonstrate the applicability of the proposed model.