Effect of parameters correlation on uncertainty and sensitivity in dynamic thermal analysis of thermal protection blanket in service

In this paper, an uncertainty propagation methodology considering correlations among the input variables for heat transfer analysis through fibrous composite insulation material subjected to severe aerodynamic conditions is presented, based on Monte Carlo (MC) simulation with Cholesky decomposition. The effects of correlation coefficient on the thermal evolution uncertainty, failure probability, and Sobol sensitivity indexes are discussed. The results show that the thermal response outputs become more scattered after taking into account the dependence of variables. The failure probability increase with the rise of the correlation coefficient, and nearly 15% increase is observed when the temperature limit is 470 K under the current uncertainty level. Sobol sensitivity analysis demonstrates that the correlated group variables of aerodynamic thermal and pressure load have the most significant first order and total effect on the thermal response simulations. However, the mean fiber diameter and the correlated group variables of density, specific heat and thermal conductivity of virgin material have weak contributions to the first order effect, but have relatively strong contribution to the interaction effect with other variables. The correlation effect of input parameters causes different influences on the sensitivity indexes of different variables. In all cases, the weakest correlation gives the highest total sensitivity index. The provided analyses demonstrate that incorporation of correlation information of stochastic variables is essential in random variable characterization, thermal evolution uncertainty level quantification, parameter importance identification and risk assessment, and it is especially imperative for thermal protection components with high reliability requirement to prevent the occurrence of catastrophic event.