CFD quality assessment in NIPCM studies with "spiral" convergence criterion for DoE refinement

Here we are trying to build on both theory and practical experience, with a view to provide better confidence in non-intrusive (NIPCM) stochastic studies for CFD. This includes not only the quality of individual CFD results but also their coherence as a set, when multiple runs are performed to analyse a domain of parametric variation. One major unknown, which we try to address, is the probability for any significant feature to remain undetected in the parameter space. Such an unexpected feature is presented on the RAE2822 case. We also present some new results regarding the sensitivity of the CFD solution to the free parameters of the solution technique, in relation with our “spiral convergence” criterion for incremental DoE (Design of Experiment) refinement.