Uncertainty quantification of vista charring ablator material database using Bayesian inference

During hypersonic trajectory through a planetary atmosphere a high heat flux environment is generated due to the friction between gas particles and the vehicle. To protect it from the excessive heat energy that is transferred to it, Thermal Protection Systems are implemented in the spacecraft’s design. Current modeling tools used for the design of heat shields, however, have been shown to be unable to fully replicate material response data recorded during flight. Collaborative efforts aimed at improving current models are also difficult to establish due to restrictions placed on the access to material response data. In response, a material model free of access restrictions dubbed VISTA was devised by a research group at University of Kentucky upon which synergistic projects aimed at studying performance of charring ablators can be readily organized. In the present thesis a sensitivity study of the VISTA material model is performed with both Pearson correlation coefficients and the method of Sobol; Sobol indices are shown to be a much more robust sensitivity metric in the context of charring ablators. Uncertain parameters of the material database are then calibrated through the use of Bayesian inference rather than basic deterministic methods often used throughout scientific works. The calibrated parameters, as well as quantified uncertainty due to model structure errors and data inaccuracy, are finally propagated through onto the output where uncertainty is seen to be reduced by a large margin. An in-house developed tool named SMUQ is used to perform analyses contained in this thesis which features a PID controller modified version of the Delayed Rejection–Adaptive Method sampling algorithm