A Hybrid Uncertainty Propagation Scheme for Convective Heat Transfer Problems

A computational analysis of convective heat flux around an array of cylinders in a high Reynolds number flow is presented; we assume that both the inflow and the wall heat flux conditions are specified with a margin of uncertainty and our objective is to quantify the resulting effect on some functional of interest, mainly the cylinder wall temperature. We introduce a hybrid uncertainty propagation technique that combines the accuracy and convergence properties of intrusive stochastic Galerkin with the non-intrusive nature of stochastic collocation. Additionally, it dramatically reduces the overall cost of computing the statistics of the stochastic output quantities. The success of the hybrid technique suggests future directions in loosely coupled multi-physics applications.