Search

A Galilean-invariant field equation is proposed and tested on standard turbulence model test cases. The field equation provides an additional non-dimensional outer scale which allows the turbulence models to reproduce the axial normal stress increase with Re seen in high Reynolds numbers experiments. The field equation provides a Reynolds number for every point based on the length of turbulent flow upstream of that point in the domain. This outer scale equation can be considered an odometer that gives a length scale conjectured to be related to the large stream wise structures that are seen in turbulent flow and that require run length to develop. A new RANS model using this additional scale is able to match the Reynolds number variation of the normal stresses seen at high Reynolds number. Furthermore, the good attached flow prediction capabilities of current RANS models appears to be attained. Using this scale equation, the entire Reynolds-stress state appears to be predicted correctly, over a large run length Reynolds number range such as experienced in aircraft design.

Adaptive Mesh Refinement (AMR) promises a much more computationally efficient means to obtain a discrete approximation to a continuous boundary value problem of a specified accuracy than classic isotropic grid refinement. The AMR capability of OVERFLOW (a computational fluid dynamics (CFD) code) is utilized to provide estimates of the exact analytical solutions to problems of interest to turbulence modeling. Predictions of surface pressure and skin friction, essentially the state of stress at the surface, shows little difference with grids believed to be "grid resolved." Velocity profiles, on the other hand, show marked differences in flows with shocks. The AMR method, as implemented in OVERFLOW 2.2k, appears to provide the ability to produce arbitrarily accurate solutions at a predictable cost much smaller than classic uniform mesh refinement.

Revised versions of Lag methodology Reynolds-stress and triple product models are applied to accepted test cases to assess the improvement, or lack thereof, in the prediction capability of the models. The Bachalo-Johnson bump flow is shown as an example for this abstract submission.

Predictions for Reynolds-stress and triple product turbulence models are compared for flows with significant rotational effects. Driver spinning cylinder flowfield and Zaets rotating pipe case are to be investigated at a minimum.

*† ‡ § ** †† ‡‡ Two general purpose Reynolds Averaged Navier-Stokes (RANS) flow solvers, OVERFLOW and BCFD, are used to analyze the NASA Common Research Model (CRM) in a wing-body configuration. The codes are run on structured and unstructured common grid families built specifically for the 5 th AIAA CFD Drag Prediction Workshop (DPW-5) allowing for meaningful comparison of data. There are six grid sizes in the family ranging from a 0.6 million cell “Tiny” mesh up to a 138 million cell “Super-Fine” mesh. Results from a grid convergence study are evaluated for each solver and grid type with focus on isolating individual effects of turbulence model and differencing scheme on computed forces, moments and wing pressures. A “Medium” mesh consisting of 5.1 million cells is used to run the wing-body configuration through an angle-of-attack sweep as part of a buffet onset study. The solutions are used to better understand variations in high speed wing separation prediction driven by the strengthening shock and by corner flow physics at the wing-body juncture. Numerical simulation of side-of-body separation continues to be a challenge for RANS methods where solutions are sensitive to grid density and turbulence model, amongst other variables. However, a newly developed quadratic constitutive relation (QCR) is employed with favorable results. Two additional studies are conducted to: a) investigate how well common grid solutions compare with those on a grid built using best practices for a given flow solver, and b) quantify the effects of transition and wing twist to provide additional corrections needed for comparisons of CFD results with experimental data.

This study presents a new class of turbulence model designed for wall bounded, high Reynolds number flows with separation. The model addresses deficiencies seen in the modeling of nonequilibrium turbulent flows. These flows generally have variable adverse pressure gradients which cause the turbulent quantities to react at a finite rate to changes in the mean flow quantities. This "lag" in the response of the turbulent quantities can t be modeled by most standard turbulence models, which are designed to model equilibrium turbulent boundary layers. The model presented uses a standard 2-equation model as the baseline for turbulent equilibrium calculations, but adds transport equations to account directly for non-equilibrium effects in the Reynolds Stress Tensor (RST) that are seen in large pressure gradients involving shock waves and separation. Comparisons are made to several standard turbulence modeling validation cases, including an incompressible boundary layer (both neutral and adverse pressure gradients), an incompressible mixing layer and a transonic bump flow. In addition, a hypersonic Shock Wave Turbulent Boundary Layer Interaction with separation is assessed along with a transonic capsule flow. Results show a substantial improvement over the baseline models for transonic separated flows. The results are mixed for the SWTBLI flows assessed. Separation predictions are not as good as the baseline models, but the over prediction of the peak heat flux downstream of the reattachment shock that plagues many models is reduced.

This paper describes the addition of an overset grid capability to the DPLR flow solver for hypersonic flow in thermochemical nonequilibrium. Modifications to the preexisting flow solver were simplified through the use of DiRTlib, a “solver neutral” library of overset utilities. The new capability is demonstrated on a series of examples, including the Orion Crew Module and other reentry vehicles. For the overset grids used in these examples, the hole cutting and interpolation stencils were determined using SUGGAR, a generalized grid assembly code that can naturally accommodate both the three-dimensional and true two-dimensional cell-centered discretization schemes in DPLR. First a series of building-block examples are presented which highlight aspects of the new capability and assess the technique with comparisons to baseline, block-structured discretizations. The new capability is then exercised for the specific case of a tension tie geometry protruding from the Orion heatshield at both wind tunnel and flight conditions. The addition of overset capability to the DPLR flow solver is seen to be an essential feature for analyzing increasingly complex geometries in thermochemical nonequilibrium.

†. A computational fluid dynamic analysis using the unsteady Reynolds-averaged NavierStokes equations is performed on the Apollo space capsule for subsonic to supersonic flow. The purpose of this analysis is to determine the methods and requirements of grid generation and flow solver execution that will produce high quality aerodynamic coefficients for similar space capsule geometries. A systematic approach is used to determine the grid resolution required to obtain grid-resolved results within reasonable force and moment tolerances. Different turbulence models are evaluated, and the results are compared with wind-tunnel data.

This study presents results obtained with the OVERFLOW2 code for three geometries that produce massively separated flowfields: a circular cylinder in crossflow and the Apollo and Crew Exploration Vehicle (CEV) capsules at various angles of attack. For the circular cylinder, a standard test case for massively separated flowfields, both three-dimensional Detached Eddy Simulations (DES) and Unsteady Reynolds Averaged Navier Stokes (URANS) simulations were obtained with emphasis on comparing URANS solutions using the Lag model to the DES solutions. In addition, two-dimensional URANS simulations are compared to three-dimensional DES and URANS. The Apollo and CEV configurations were simulated with only URANS simulations but with several one- and two- equation turbulence models in addition to the Lag model. For the capsule simulations, the Lag model provided more accurate predictions of experimental lift, drag and pressure distributions for subsonic and supersonic flowfields.

The blast wave resulting from an explosion was simulated to provide guidance for models estimating risks for human spacecraft flight. Simulations included effects of headwind on blast propagation, Blasts were modelled as an initial value problem with a uniform high energy sphere expanding into an ambient field. Both still air and cases with headwind were calculated.

A probabilistic risk assessment (PRA) approach has been developed and applied to the risk analysis of capsule abort during ascent. The PRA is used to assist in the identification of modeling and simulation applications that can significantly impact the understanding of crew risk during this potentially dangerous maneuver. The PRA approach is also being used to identify the appropriate level of fidelity for the modeling of those critical failure modes. The Apollo launch escape system (LES) was chosen as a test problem for application of this approach. Failure modes that have been modeled and/or simulated to date include explosive overpressure-based failure, explosive fragment-based failure, land landing failures (range limits exceeded either near launch or Mode III trajectories ending on the African continent), capsule-booster re-contact during separation, and failure due to plume-induced instability. These failure modes have been investigated using analysis tools in a variety of technical disciplines at various levels of fidelity. The current paper focuses on the roles and impacts of the higher-fidelity methods on this process and, by association, the roles and impacts of the high performance computing resources of the Columbia supercomputer system at NASA Ames Research Center.

The Lag model has shown great promise in prediction of low speed and transonic separations. The predictions of the model, along with other models (Spalart-Allmaras and Menter SST) are assessed for various high speed flowfields. In addition to skin friction and separation predictions, the prediction of heat transfer are compared among these models, and some fundamental building block flowfields, are investigated.

An implementation of both finite rate and equilibrium chemistry have been completed for the OVERFLOW code, a chimera capable, complex geometry flow code widely used to predict transonic flow fields. The implementation builds on the computational efficiency and geometric generality of the solver.

Combustion CFD calculations using fully-coupled and loosely-coupled approaches are analyzed using zero-dimensional simulations and von Neumann stability analysis. Several loosely-coupled variants are considered, including dierent choices of the primary dependent variable and dierent ways of partitioning the equations. Results indicate that only the fully-coupled scheme is always stable, while the various loosely-coupled approaches behave dierently depending upon the Mach number and the relative strengths of the combustion source terms. In general, the loosely-coupled method using internal energy as the solution variable is conditionally stable, while that using temperature as the variable is stable for subsonic flow but unstable for supersonic flow.

A validation of the matrix dissipation option in OVERFLOW was completed, and a single set of constants was found which yield good inviscid shock capturing, viscous heat and skin friction predictions, and work over a wide Mach and Reynolds number range. This scheme works well with multigrid, and produces fast convergence to steady state for 2 and 3 dimensional problems, with accuracy comparable to Roe upwinding, and without 'carbuncle' problems for the blunt body flowfields.

Recent experiences with pressure-sensitive paint (PSP) have shown that very thin paint layers on wind-tunnel models tested at high Reynolds numbers can significantly alter the pressure distributions, and thus the forces and moments, on the models. This was observed during two tests of transport-like wings: a "clean" supercritical wing at transonic cruise conditions in the High Reynolds Number Channel 2, and a high-lift wing, complete with slats and flaps, at landing conditions in the 12-Ft Pressure Wind Tunnel. The effect of paint on the cruise wing was to displace the shock wave slightly upstream from its no-paint position. Smoothing the paint, and even removing it entirely from the leading edge, decreased this displacement slightly. Applying paint to only the slats of the high-lift wing caused the wing to stall prematurely at the highest Reynolds number. This effect could be eliminated by smoothing the paint. Adding paint to other parts of the model had little effect. Paint intrusiveness was much smaller and more ambiguous at lower Reynolds numbers. The roughness of the paint in both tests did not exceed generally accepted "admissible roughness" criteria for turbulent boundary layers nor did it exceed accepted criteria for forcing premature transition of a laminar boundary layer.

The computational simulation of wind tunnel tests are central in efforts to improve the accuracy and range of applicability of CFD codes. A validation experiment in solid wall transonic wind tunnel is simulated with the inviscid and viscous tunnel wall computation models. Viscous wall modeling appears to provide the ability to reproduce the experimental conditions fairly well without wall interference corrections but specification of the actual wall geometry is crucial.