On the propagation of statistical model parameter uncertainty in CFD calculations

Abstract

This work considers a new class of finite-volume approximations for scalar and system nonlinear conservation laws with multiple sources of stochastic model parameter uncertainty. The deterministic propagation of model parameter uncertainty is achieved through the introduction of additional stochastic coordinates. Particular attention is given to the construction of specialized piecewise polynomial approximation spaces well suited to the high-order accurate approximation of solution discontinuities in both physical and stochastic dimensions without exhibiting Gibbs-like oscillations characteristic of polynomial approximation. The proposed discretization easily retrofits existing finite-volume CFD codes in use today. Numerical results are presented for inviscid Burgers equation with uncertain initial data as well as the compressible Reynolds-averaged Navier–Stokes equations with uncertain boundary data and turbulence model parameters.