A Bayesian framework for in-flight calibration and discrepancy reduction of spacecraft operational simulation models

Abstract

Modeling and Simulation (M&S) tools have become indispensable for the comprehensive design, operations, and maintenance of products in the space industry. An example is the European Space Agency (ESA), which relies heavily on M&S throughout the entire lifecycle of a spacecraft. However, their use in operational settings poses significant challenges, mainly attributable to (i) the harsh, uncontrollable, and often unforeseen environmental conditions; (ii) the dramatic changes in operating conditions throughout a spacecraft’s lifespan, often beyond the intended designed-for lifetime; and (iii) the presence of epistemic and aleatoric uncertainty. This results in unavoidable discrepancies between the numerical simulations and real measurements, limiting their use for delicate operational tasks. To address those challenges, we present a Bayesian framework for simultaneous calibration of M&S tools, reduction of the model discrepancy, and quantification of the process and model uncertainties. The approach leverages the Kennedy and O’Hagan (KOH) calibration, tailored for a multi-objective problem. Its effectiveness is shown by its application to flying Earth observation spacecraft data and the operational simulation models.