Multifidelity HDMR Surrogates for Thermal Protection System Heat Flux Hybrid Uncertainty Quantification

This work seeks to exploit the multiple available models of differing fidelity that can be used to predict hypersonic flow physics to construct a combined multifidelity High Dimensional Model Representation (HDMR) surrogate combining aleatory/epistemic uncertainty quantification for the study of a thermal protection system for atmospheric entry. A multifidelity method for HDMR models which can explore a wider range of subspaces is introduced and investigated before being applied in an aeroshell re-entry scenario. Freestream conditions are considered to vary aleatorily whilst material properties of the vehicle surface are described epistemically as interval values. Combined aleatory and epistemic uncertainty is described by a P-Box representation of hybrid uncertainty. A “double-loop” uncertainty propagation approach is applied where aleatory sample points are selected according to a Monte Carlo campaign and epistemic intervals defining optimisation subproblems are solved. The methodology presented here gives results that can be used to ensure robustness in the presence of hybrid uncertainty.