A Multi-Parameter Statistical Inference Model for the Thermal Burnout Effect of LNA Excited by Repetitive Electromagnetic Disturbances

Abstract

Repetitive electromagnetic disturbances (REDs), characterized by power level, repetition frequency, pulsewidth, and pulse number, would couple to the low noise amplifier (LNA) via the antenna. The four parameters collectively determine the heat accumulation process and the subsequent permanent damage in LNA. In this case, the scientific vulnerability assessment for LNA is important. In this article, the multiparameter statistical inference model is put forward to assess the thermal burnout effect of LNA excited by REDs. In the physical part of the model, the internal temperature rise is analyzed using the heat conduction equation, which takes into account the effect mechanism and thus the four parameters. In the statistical part of the model, the internal temperature rise is chosen as the latent variable. The model can be robustly and efficiently determined by two sets of single-parameter effect test samples, which is validated by the simulation case. Finally, a case study is carried out on LNA. The model parameters calculated from two pair of test sets are consistent, which indicates the model is reasonable. The inference results given by the model are validated by the validation test results. In addition, the model performs better than the logistic regression model by comparing the confusion matrix, accuracy, and cross entropy.