Gray Box Photocurrent Models: A Data-Driven Approach for Compact Radiation Modeling

We demonstrate generalizability of compact gray box photocurrent models (GBPMs), developed in [4], to variable bias conditions. To that end we use experimental measurements of a Z5236 device collected at the Little Mountain Test Facility (LMTF), Hill AFB, Utah. The dataset contains the response of the device, held at two different bias conditions, to approximately rectangular radiation pulses with different pulse widths. Change in the bias prompts both qualitative and quantitative changes in the measured photocurrent, thereby providing a challenging test for the GBPM. We train the model using a subset of the available data and perform reproductive and predictive tests, showing that the GBPM is capable of accurately reproducing the variability of the device response under different bias conditions. These results suggest that the model form underpinning the GBPM is sufficiently expressive to generalize beyond variable dose-rates, as reported in [4], to also include variable bias conditions.