System Health Awareness in Total-Ionizing Dose Environments

Abstract

There is increasing interest in using commercial-off-the-shelf (COTS) electronics in radiation environments, such as robotic systems for remediation after the nuclear accident at Fukushima or in low-cost CubeSats. Commercial electronics have varying levels or robustness to radiation environments, and without extensive testing and redesign, the survivability of COTS systems in radiation environments is unknown and potentially insufficient. This work identifies characteristics of robotic commercial-off-the-shelf component degradation, primarily part-to part variability and the interactions between the degradation of multiple components. Insight into the health of a class of components, micro-controllers, was developed using timing characteristics. A framework using either continuous or discrete Bayesian networks was developed to model the degradation observed in sensors and other electronic components. The Bayesian network can be incorporated with deterministic models to produce a robust and scalable analysis methodology, granting awareness of how the system will behave in the radiation environment and providing insight into areas for improvement in the system hardware and software.