Proceedings of the
The 33rd European Safety and Reliability Conference (ESREL 2023)
3 – 8 September 2023, Southampton, UK
Analysis of Shrinkage Estimators and Bayesian Decision Rules for Bioburden Density Estimation in Planetary Protection Probabilistic Risk Assessment
1Idaho National Laboratory, USA.
2California Institute of Technology, Jet Propulsion Laboratory, USA.
ABSTRACT
The discipline of forward planetary protection aims to minimize microbial contamination on spacecraft in order to prevent the inadvertent contamination of other planetary bodies. Understanding the number of microorganisms, or bioburden, launched with the spacecraft is fundamental to achieving this outcome and is calculated using estimates of the bioburden density (bioburden per unit area or volume) across the spacecraft.
While extremely simple, the deterministic estimators based on NASA-specified and implied bioburden densities may, under certain conditions, have quadratic risk lower than data-driven estimators, with no data estimators being uniformly better (i.e., the estimators are admissible). By comparing risks of deterministic and data-driven estimators, different sampling schedules and volumes can be analyzed to optimize the performance of these estimators. This paper contrasts two approaches used for bioburden calculations-frequentist and Bayesian-and evaluates their performance using data collected from NASA's InSight mission. Specifically, we calculate quadratic risks of different types of shrinkage estimators and compare the risks with the Bayesian approach. An analysis for different regions of the parameter space found estimators with the lowest risks for bioburden values most frequently occurring in practice.
Keywords: Planetary protection, Bayesian inference, InSight mission, Loss function, Risk, Gamma-Poisson model.
