Validation Strategies for High-Fidelity Computational Hydro-Moprhodynamic Models of Waterways

Abstract:

Turbulent flow and transport processes in riverine environments pose major challenges to computational fluid dynamics models. Such flows are dominated by arbitrary nature and man-made geometric complexity, are bounded by dynamically evolving boundaries whose shape and deformation are not known a priori, and are dominated by slowly evolving, albeit energetic, aperiodic coherent structures that often constitute the primary mechanism of turbulence production. With exponentially growing computing power and the development of powerful computational methodologies for carrying our large-eddy simulation (LES) of flow and transport processes in real-life waterways, it is now possible to simulate such flows with unprecedented realism and spatial and temporal resolution. Such sophisticated numerical simulations, however, require experimental measurements of equal sophistication and resolution that can provide reliable data to enable the validation of the numerical models. In this talk I will present a three-pronged high-fidelity modeling validation approach developed at the St. Anthony Falls laboratory based on the integration of laboratory-scale experiments (Indoor Streamlab), field-scale experiments under controlled laboratory conditions (Outdoor StreamLab), and field measurements. I will demonstrate this validation approach in the context of SAFL’s high-fidelity hydro-morphodynamic code, the SAFL Virtual StreamLab (VSL3D), and emphasize the need for tight integration of experiments and simulations. I will show how such coupling is critical prerequisite not only for validating numerical models but also for uncovering new physical insights that could not have been found by measurements alone. I will also discuss and illustrate the challenges for meaningful validation of RANS turbulence models and present metrics for establishing the grid independence of computed solutions for both hydrodynamic and morphodynamic aspects of the simulations.

This work was supported in part by NSF grant IIP-1318201. Simulations were carried out at the Minnesota Supercomputing Institute.