Estimating Uncertainty in Computations of Two-Dimensional Separated Flows

Abstract

The present paper investigates sources of uncertainties in two-dimensional flow computations and presents methods for estimating them. A sample problem is used for illustration. The following categories are explored in detail: (i) Uncertainty due to truncation error in numerical schemes; (ii) Uncertainty due to discretization error; (iii) Uncertainty due to outflow boundary conditions; (iv) Uncertainty due to incomplete iterative convergence; (v) Uncertainty due to computational grid aspect ratio. The error estimates are based on requirements for internal consistencies in computed results. Therefore, they provide better judgement of the numerical solution integrity than comparisons to experimental data or “benchmark” solutions whose reliability may sometimes be questionable. Ideally, both approaches should be employed. A new method is presented for estimating the optimum grid-cell aspect ratio for computational accuracy and efficiency.