Stochastic reconstruction of thermal protection material properties from arc-jet experiments

The focus of this work is to revisit the capability of extracting material properties from an arc-jet experimental dataset while acknowledging and modeling uncertainty using cutting edge methods and algorithms. To achieve this, efficient simulation of the experiments is crucial. Specifically, we focus on simulating the temporal evolution of in-depth temperatures for comparison with thermocouple data. In this work, we introduce a comprehensive Bayesian analysis by developing anidealsurrogatemodelbasedontheKarhunen-Loèveexpansioncapable of approximating the functional output, such as temperature versus time. This approach offers dual benefits: firstly, it relies solely on a dataset of model solutions to address uncertainty quantification (UQ) problems, including forward analysis, sensitivity analysis, and inverse analysis. Secondly, it allows us to maintain the necessary computational complexity of the model solver that best reflects the experimental data. Consequently, this methodology constitutes a pivotal advancement for facilitating more intricate Bayesian inference analyses in the future. This work presents the theoretical framework, initial findings to assess the feasibility of the proposed surrogate model, and a complete Bayesian inference analysis to refine material properties based on arc-jet experiments, leveraging state-of-the-art computational tools.