Reynolds-averaged navier-stokes computations of the nasa juncture flow model using expert-crafted and adapted grids

Two methods of creating Computational Fluid Dynamics (CFD) grids are used with the Stabilized Finite Element (SFE) library within NASA’s FUN3D CFD solver to predict the flow over the NASA Juncture Flow (JF) model. A CFD study of the JF model using one solver with grids created by expert hand-crafting and automatic grid adaptations has not been previously published. This CFD study considers Reynolds-averaged Navier-Stokes solutions produced on four grid families: 1) expert-crafted mixed-elements, 2) expert-crafted all-tetrahedral, 3) automatic adaptations to control interpolation error of a solution variable, such as Mach number, and 4) automatic adaptations that control the error in an output functional, such as drag. Using the FUN3D+SFE solver and the Negative Spalart-Allmaras One-Equation Model with Quadratic Constitutive Relation, 2000 version, to produce solutions on these four grid families enables comparisons of grid and turbulence model interplay that are not found in literature. Free-air simulations in a half-span domain were performed at an angle-of-attack of 5.0 degrees and compared with the corresponding data from the NASA JF experiment’s wind tunnel campaign. CFD results produced using the four grid families are found to be consistent with each other and in good agreement with experiments in attached regions. However, results in the separated region of the wing-body juncture show grid sensitivities and less agreement with experiments. Plots of q-criterion and the continuity costate field from an adjoint solution are presented and discussed to highlight the significance of the juncture region’s influence on the total drag force. The predictions of turbulent normal stresses near walls show the least agreement with experiments on all grids. Boundary layer resolution provided by the grids was compared using the spacing of element intersections with sampling lines at four locations where measurements were made in experiments. The drag-adapted grid family closely approximates the grid spacings and predictions of the crafted all-tetrahedral grids along the sampling lines. Both expert-crafted and automatically adapted grids enable FUN3D+SFE to predict flow phenomenon around the JF model at the frontiers of engineering accuracy while using 5× and 25× smaller grids than those in literature.