Automating Error Quantification and Reduction for Computational Fluid Dynamics (CFD)

Solution errors are inherent in any Computational Fluid Dynamics (CFD) simulation. Sources of error include spatial and temporal discretization, inadequate physical models, and human errors in the setup and use of the CFD code. Systematic identification, reduction, and control of these various error sources is crucial if the results of CFD simulations are to be trusted for design and performance assessment of air vehicles. While grid refinement studies may verify the spatial accuracy of a solution, these studies are generally laborious and time intensive. Continued development of a standalone Error Transport Equation (ETE) solver is proposed. This tool will provide numerical error bars, quantifiable levels of uncertainty in both local and globally integrated variables. To automate error reduction, the proposed program exploits an existing mesh adaptation package. The CRISP CFD® mesh adaptation code, used with several unstructured Navier-Stokes solvers including the Air Vehicles Unstructured Solver (AVUS), when combined with the ETE solver, provides a promising, viable path for automated error reduction and solution verification. In addition, the proposed program will address uncertainty quantification for detailed studies where the effects of model selection, solver algorithm, boundary conditions, inputs, etc. are systematically varied, through the use of Response Surface models.