Experimental Efforts for Predictive Computational Fluid Dynamics Validation

Ideally, Validation and Verification methods to be used to deliver Predictive Computational Fluid Dynamics (P-CFD) methods for design use, must utilize a consistent set of detailed input to perform an output response comparison to associated data that can accurately characterize the uncertainty in the calculated results. In this paper, some experimental efforts are described and discussed in relation to their use in Predictive CFD Validation. The P-CFD methodology will be used in a design-by-simulation methodology to design large, complex hydraulic system geometries comprising multiple scales of physical phenomena. Currently, comparisons are made between experiments and CFD at the same integratedeffects scale and it is difficult to identify the causes for why computational predictions do well or the reasons for discrepancies should they arise. Instead of large system or subsystem level tests, the focus for P-CFD Validation-level experimental studies is to concentrate on elucidating the fundamental fluid mechanics of unit physics phenomena. These unit and benchmark experiments can be performed in such a way that the uncertainty in the results can be systematically quantified and provide far greater experimental data density than integrated-effects scale testing. It is this systematic quantification of uncertainty in both the computational and experimental results that forms the basis for Predictive CFD. Examples of validation-level experimental efforts including a surface-mounted cube, bluff bodies in cross flow, and an impinging radial diffuser flow will be discussed. These experimental efforts have enabled us to begin to identify the needed approaches for acquiring comprehensive documentation of the input for underpinning P-CFD principles. These lessons including information about the as-built geometry, the flow conditions, as well as boundary conditions able to be transferred to the analysis will be described. One key finding is that validation-level experiments and the associated analysis must be cooperative between the testing groups and the analysis groups to be successful. Regular communications between the analysts and the experimentalists is crucial in all phases of these studies. In addition, the importance of uncertainty quantification in both the experiments and analysis relating to Predictive CFD Validation will be discussed. Finally, the vision for the design use of Predictive CFD, the key being analysis results with quantified uncertainty, is described.