Robust grid adaptation for efficient uncertainty quantification
In the recent past, adjoint methods have been successfully applied in error estimation of integral outputs (functionals) of the numerical solution of partial differential equations. The adjoint solution can also be used as a grid adaptation indicator, with the objective of optimally targeting and reducing the numerical error in the functional of interest below a prespecified threshold. In situations where we seek to quantify the effect of aleatory uncertainties on statistical moments of the output functional, it becomes necessary to evaluate the functional accurately at multiple sample points in probability space. If the numerical accuracy of these sample evaluations is not uniform, variations in the numerical error can affect the evaluation of the statistical moments. Although it is possible to independently adapt the meshes to obtain more accurate solutions at each sample point in stochastic space, such a procedure can be both cumbersome and computationally expensive. To improve the efficiency of this process, a new robust grid adaptation technique is proposed that is aimed at minimizing the numerical error over a range of variations of the uncertain parameters of interest about a nominal state. Using this approach, it is possible to generate computational grids that are insensitive to small variations of the uncertain parameters that can both locally and globally change the solution and, as a result, the error distribution. This is in contrast with classical adjoint techniques, which seek to adapt the grid with the aim of minimizing numerical errors for a specific flow condition (and geometry). It is demonstrated that flow computations on these robust grids result in low numerical errors under the expected range of variations of the uncertain input parameters. The effectiveness of this strategy is demonstrated in problems involving the Poisson equation and the Euler equations at transonic and supersonic/hypersonic speeds.