Sukumar Chakravarthy

Convergence acceleration for unified-grid formulation using preconditioned implicit relaxation

Improved convergence rates for a unified grid framework are achieved by combining several convergence acceleration strategies, which include local implicit time-stepping, low-speed preconditioning, and relaxation methods. It is demonstrated that good convergence can be achieved on various grid types and topologies, all speed regimes, and for both inviscid and viscous flows.

Sub-grid turbulence modeling for unsteady flow with acoustic resonance

This paper proposes a novel combination of Reynoldsaveraged Navier-Stokes (RANS) and large-eddy simulation (LES) sub-grid models, which combines the best features of time-averaged and spatially-filtered models, yielding the superior near-wall stress predictions of (algebraic or full-transport) Reynolds-stress models with the ability to override any quasi-steady grid-converged RAN’S model solution in regions of sufficiently high grid density. The proposed hybrid formulation is well suited to the coupled simulation of all flow scales in resonating high-Reynolds number flows and contains no additional empirical constants beyond those appearing in the original RANS and LES sub-grid models.

Hypersonic Flow Predictions Using Linear and Nonlinear Turbulence Closures

Two- and three-dimensional hypersonic flow cases are computed using linear one-equation closures and a nonlinear two-equation model, where the anisotropy tensor is modeled as a cubic function of mean strain and vorticity tensors. The latter is found to excel in predicting bypass transition, whereas the one-equation R t model is very good at heat-transfer prediction. Both closures excel in predicting pressure distributions; however, the nonlinear model is found to overpredict heat-transfer. This suggests that in separated flow regions with simultaneously low mean-flow kinetic energy (and therefore low strain magnitude) and high temperature gradients, overpredicted levels of turbulence length scale can lead to rather small errors in the turbulent shear stress, while at the same time leading to a large overprediction of the turbulent heat fluxes

Validation Of CFD++ Code Capability For Supersonic Combustor Flowfields

Numerical simulations of several turbulent supersonic flows related to scramjet combustors are carried out using a new unified-grid computational methodology. Five problems are considered: a 2-D ramp unit problem; a reattaching turbulent shear layer, the 3-D University of Virginia two-hole supersonic transverse Air-Air injector; and the NASA P2 and P8 supersonic inlets. The numerical simulations are conducted using the Reynoldsaveraged Navier-Stokes equations along with oneequation and three-equation pointwise turbulence models. Both turbulence models enable accurate prediction of the flowfields and numerical results compare favorably with experimental data in all cases.

The CFD++ Computational Fluid Dynamics Software Suite

Computational Fluid Dynamics CFD is no longer the domain of just specialists. It is also being used by engineers and scientists in many disciplines who are interested in CFD as a tool to investigate other things and not as just an end in itself. The realization of this fact drives developers to produce user-friendly CFD products that automate most of the problem set up and solution process. The CFD++ software suite is a unified-grid, unified-solution, unified-computing CFD simulation capability that was designed from the outset to be effective from the users perspective. The details are explained in this paper.

A numerical study of the pulse detonation wave engine with hydrocarbon fuels

This paper explores some issues that arise in the analysis of pulse detonation wave engines with hydrocarbon fuels. One-dimensional and axisyrnmetric/two-dimensional simulations are employed along with reduced kinetic mechanisms to confirm the ability of the numerical approach to accurately compute relevant physical characteristics such as proper detonation wave speed, von-Neumann spike, aspiration, pressure time history and sequence of cycle events. It is shown that qualitatively and quantitatively reasonable results can be obtained with a careful treatment of the finite-rate-chemistry source terms. Some of the numerical difficulties that arise in dealing with unsteady detonation phenomena are discussed and improvements demonstrated. Onedimensional test cases with simplified H2-O2 and C3HgAir kinetics are used to verify correct detonation wave speed and testing boundary conditions. An axisymmetric case for the latter chemistry is studied with a generic inlet to illustrate the ability of the methodology to capture the relevant physics, namely, pressurization of thrust wall by the detonation wave and interaction of the reflected wave with rarefaction waves from the open end.

Smart Sub-Grid-Scale Models for LES and Hybrid RANS/LES

This paper considers the implementation of sub-grid scale (SGS) models for large eddy simulation (LES) or hybrid Reynolds-averaged Navier-Stokes (RANS)/LES methods. The paper describes how to reconcile Smagorinsky-type eddy viscosity SGS models with monotonically-integrated large eddy simulation (MILES) approaches by using ‘Smart’ SGS models that understand, and compensate for, the inherent diffusion in the underlying numerical transport algorithm. An improved ‘burst’ model of synthetic turbulence is also introduced for applications to inhomogeneous turbulence.

A unified-grid approach for propulsion applications

The unified-grid approach includes the integrated treatment of structured and unstructured grids as well as single and multiblock grids. Various threedimensional, two-dimensional and one-dimensional cell shapes can be utilized. It helps in building solutionadaptive meshes. The unified-grid approach includes the ability to automatically connect multiblock meshes (including patched-aligned, patched-nonaligned, and overset grids). This feature can help deal with turbomachinery problems where grids attached to the rotor stage move with respect to other stationary grid blocks. In this paper, we present these building blocks using several unit problems as illustrative examples.

Afterbody flowfield computations at transonic and supersonic Mach numbers

The flowfield around various axisymmetric afterbody configurations is computed with a finite-volume NavierStokes code, incorporating a total variation diminishing implicit upwind-biased scheme for high accuracy and using alternatively the k-e or the Baldwin-Lomax turbulence model. Computations are done for both solid plume simulators and real jet flows. Results for two geometries at several combinations of jet and freestream conditions are shown. Agreement with the experimental data is very good.

Flow Prediction around the SACCON Configuration Using CFD

A three-dimensional numerical simulation of a vortex breakdown, using the UCAV half wing configuration, has been performed. The vortex-dominated flow-field, where non-linear effects have a significant impact over the wing, causes loss of lift. These nonlinear regimes are the areas where linear stability and control methods normally fail. These tests, using CFD++, were performed primarily at the low-speed side of the static stability and control characteristics of this configuration. Two turbulence models, used for validation purposes, were the SA (Spalart-Allmaras) and kω-SST(Shear-Stress Transport) closures to capture these complex flow fields at angle-of-attack ranges of 15 to 22°. Detailed results compare the numerical to experimental data in terms of pressure coefficient, vorticity components, lift, drag and pitching moment coefficient (Cl, Cmy, Cd). The numerical results indicate that the one-equation turbulence model (Spalart-Allmaras) yields better moment predictions than those by the two-equation model (SST), but the latter predicts lift and drag better.