Statistical Uncertainty of DNS in Geometries without Homogeneous Directions

In this paper, we present uncertainties of statistical quantities direct numerical simulations (DNS) with small errors. The are analysed for channel flow and a separation case in confined backward facing step (BFS) geometry. infinite has two homogeneous directions is usually exploited to speed-up the convergence results. As show, such procedure reduces results by up an order magnitude. This effect strongest near wall regions. over BFS, there no thus very long integration times required. individual converge square root time so, improve uncertainty factor two, simulation be prolonged four. We provide estimator that can used evaluate priori DNS relative from obtained Reynolds Averaged Navier Stokes simulation. DNS, predict averaging it required achieve certain For accurate evaluation averages their uncertainties, not use every DNS. observe uninfluenced reducing number samples point where period between consecutive measured Courant–Friedrichss–Levy (CFL) condition units below one. Nevertheless, crossing limit, estimates start exhibit significant growth.