Bayesian Reliability Analysis of the Enhanced Multimission Radioisotope Thermoelectric Generator

Abstract

Radioisotope thermoelectric generators (RTGs) have been used as power sources for several space missions, including the multimission RTG (MMRTG) in NASA’s recent Mars Curiosity and Perseverance rovers. The enhanced MMRTG (eMMRTG) design is an MMRTG retrofitted with advanced skutterudite thermoelectric couples to improve efficiency and power output over time. The eMMRTG’s predicted reliability over the mission duration is influenced by the uncertainty in the physics parameters governing its performance. We use a Bayesian approach to account for two types of uncertainty: epistemic and aleatory. The reliability analysis has two steps: 1) uncertainty quantification from experimental data for key physics parameters, and 2) uncertainty propagation through a computational RTG life performance prediction. In particular, we use hierarchical models to separate specimen-to-specimen uncertainty and also to model the separate manufacturing batches of thermoelectric couples loaded into the eMMRTG. We use a secondary probability method to predict reliability with epistemic uncertainty intervals for the eMMRTG system at 10,000 h for which recent data are available. We show how reliability and uncertainty intervals depend on hypothetical power requirements at 10,000 h. The approach is generally applicable when experimental data and computer simulation are used to predict the reliability of a new technology.