Browsing by Author "Dr. Jack R. Edwards, Committee Chair"

Now showing 1 - 8 of 8

Algorithmic Enhancements to the VULCAN Navier-Stokes Solver
(2003-08-15) Litton, Daniel; Dr. Jack R. Edwards, Committee Chair; Dr. D. Scott McRae, Committee Member; Dr. Ashok Gopalarathnam, Committee Member
VULCAN (Viscous Upwind aLgorithm for Complex flow ANalysis) is a cell centered, finite volume code used to solve high speed flows related to hypersonic vehicles. Two algorithms are presented for expanding the range of applications of the current Navier-Stokes solver implemented in VULCAN. The first addition is a highly implicit approach that uses subiterations to enhance block to block connectivity between adjacent subdomains. The addition of this scheme allows more efficient solution of viscous flows on highly-stretched meshes. The second algorithm addresses the shortcomings associated with density-based schemes by the addition of a time-derivative preconditioning strategy. High speed, compressible flows are typically solved with density based schemes, which show a high level of degradation in accuracy and convergence at low Mach numbers (M < 0.1). With the addition of preconditioning and associated modifications to the numerical discretization scheme, the eigenvalues will scale with the local velocity, and the above problems will be eliminated. With these additions, VULCAN now has improved convergence behavior for multi-block, highly-stretched meshes and also can accurately solve the Navier-Stokes equations for very low Mach numbers.
Hybrid LES/RANS Simulation of a 10-degree Double-Fin Crossing Shock Flow at Mach 8.28
(2007-02-28) Boles, John Arthur; Dr. Richard D. Gould, Committee Member; Dr. Jack R. Edwards, Committee Chair; Dr. D. Scott McRae, Committee Member
The simulation of a Mach 8.28 10-degree double-fin crossing shock flow using a hybrid large-eddy ⁄ Reynolds-averaged Navier-Stokes (LES⁄RANS) solver is presented in this work. The solver blends a Menter two-equation model for RANS with a Yoshizawa one-equation subgrid model for the LES calculations. The solver uses a flow-dependent transition function based on wall distance and a modeled form of the Taylor microscale. Turbulent boundary layers are initiated and sustained in the inflow region using a recycling⁄rescaling technique applied to the fluctuation fields. The hybrid LES⁄RANS model is tested using both Menter's Baseline (BSL) and Shear Stress Transport (SST) models for the RANS closure. These results are compared to pure Menter BSL and SST RANS results as well as with the experimental data of Kussoy and Horstman(1992). This study concludes that while the hybrid LES⁄RANS model outperforms RANS calculations in the inflow region where the flow is nominally two-dimensional, it significantly overpredicts the wall heat transfer rates in the region of the crossing shock interaction. Possible explanations for this behavior as well as plans for future attempts at solutions to these shortcomings are provided.
Hybrid Reynolds-Averaged / Large-Eddy Simulations of Ramped-Cavity and Compression Ramp Flow-fields
(2002-07-24) Fan, Thomas Chen-Chuan; Dr. Jack R. Edwards, Committee Chair
A procedure for simulating wall-bounded,separated flows utilizing hybrid large-eddy / Reynolds- Averaged strategies is presented in this work. Following the zonal concept, the proposed hybrid method uses a distance-dependent blending function to shift the turbulence closure from Menter's two-equation model near wall surfaces to a one-equation subgrid model away from walls. The code is parallelized using domain-decomposition / MPI message-passing methods and is optimized for operation on the 720 processor IBM SP-2 at the North Carolina Supercomputing Center. The capabilities of the hybrid method are examined on two benchmark flows: a ramped-cavity flow that is representative of the internal flow field of a high speed propulsion device, and a compression ramp flow that features the classical problem of a shock wave / boundary layer interaction. Results indicate that the hybrid method provides generally good predictions for the ramped-cavity configuration, but less satisfactory predictions for the compression ramp configuration. Nevertheless, the strength of the hybrid method in capturing the recovery of the boundary layer downstream of reattachment is found in both cases, and it is a major improvement over the simulations produced by RANS alone. The weaknesses in simulating the compression ramp flow are also discussed and possible remedies are provided for further investigation in the future.
Large-Eddy Simulation of Particulate Resuspension and Transport Under Influences of Human-Body Motion in an Indoor Setting
(2007-05-15) Oberoi, Roshan C.; Dr. Jack R. Edwards, Committee Chair; Dr. Hassan A. Hassan, Committee Member; Dr. Pierre A. Gremaud, Committee Member
A methodology is presented for simulating particulate resuspension and transport under influences of human-body motion in an indoor setting. The simulations in this study mirror experiments performed by the U.S. Environmental Protection Agency (EPA), which funded the present study, and the Research Triangle Institute (RTI) at the EPA test facility in Cary, NC. A large-eddy simulation (LES) framework is implemented to obtain the time-dependent flow field within a room. An artificial compressibility method with low-diffusion upwinding and weighted essentially non-oscillatory (WENO) variable extrapolation is employed to obtain an incompressible Navier-Stokes solution. Unresolved fluctuations are accounted for by a Smagorinsky sub-grid scale stress model. A human body is modeled as an immersed boundary within the Cartesian grid domain. This body is comprised of several immersed components, representing separate body parts. Interpolation methods force the fluid and particle properties near the immersed surface to respond to the motion of the bodies, which is governed by prescribed rate laws. The particle phase is assumed to be dilute, and thus, does not affect the solution of the carrier fluid. An Eulerian viewpoint is taken to model the particle fields, requiring separate solutions for each size of particle simulated. Size classes are determined by taking sectional averages of a lognormal probability density function, extracted from experimental data. The motion of the particle fields, subject to hydrodynamic drag forces, is determined by solving mass and momentum conservation equations for each size class. A second-order TVD upwind scheme is used for the advection of particle fields, and a point-implicit sub-iteration method is used for time-advancement. The present simulations involve a human body walking and stamping its feet for about 20 seconds — causing particles initially contained within a carpet to resuspend — then standing still for the remainder of the simulation. In order to account for the porous structure of the carpet, Darcy-type resistance terms are applied to the solution of the carrier fluid. Micro-scale surface effects acting on the particles, such as van der Waals and electrostatic forces, are modeled by applying a size-dependent sticking force to particles contained by the carpet. This sticking force is approximated in a parametric fashion by comparing simulated particle emission factors with those obtained experimentally. Effects of an HVAC system are also modeled by applying inflow boundary conditions of measured velocity at two known vent locations. These simulations are performed on a computational domain of approximately 5.4 million grid points and are mapped to 36 Intel Xeon processors on an IBM Blade Center Linux Cluster using the MPI message passing standard. The simulations produced similar levels of particulate mass resuspension to those observed experimentally. Results indicated that a large majority of the particles resuspended originated from regions of the carpet very near where the immersed-body "feet" penetrated, while particles elsewhere in the room were mostly undisturbed. Despite the fact that most of the mass resuspended was due to large particles, much more small-particle mass remained airborne over the duration of the 7-minute simulations due to much lower settling rates. A relatively "well-mixed" state was achieved in the room after about 3 minutes of physical time. This made it possible to identify steady particle-decay trends over the last few minutes of the simulations in order predict concentrations in the room beyond this extent of time.
Numerical Simulation of Injection of Supercritical Ethylene/Methane into Nitrogen
(2005-12-13) Star, Ana Maria; Dr. Jack R. Edwards, Committee Chair; Dr. Richard D. Gould, Committee Member; Dr. Hassan A. Hassan, Committee Member
The present study investigates the physical mechanisms of supercritical fluid injection for pure ethylene and ethylene/methane mixtures, as well as onset of condensation upon fluid expansion. These mechanisms are considered a key enabling technology in the design of hydrocarbon-fueled scramjet engines. The numerical method combines a solution of the compressible Navier-Stokes equations for the supercritical fluid with two different approaches for condensate growth: one based on a homogeneous equilibrium assumption and the other on classical aerosol dynamics. The thermodynamic behavior of the supercritical fluid is described using the Peng-Robinson equation of state. Computational results are compared with shadowgraph and direct-lighting imaging data, mass flow measurements, mole fraction measurements and temperature measurements in the jet mixing zone, and pressure distributions within a three-dimensional injector geometry. Qualitative results involving jet structure, the appearance of a condensed phase, and the general effects of back pressure and injectant temperature are in good agreement with experimental results for pure ethylene injection. Quantitative results also display reasonable agreement with experimental results but do indicate the need for improving the model. Qualitative trends for ethylene/methane mixture injection are in moderate agreement with the experimental data, suggesting that the thermodynamic interaction between ethylene and methane as modeled by the chosen mixing rules are not sufficiently accurate. For conditions where both are applicable, a finite-rate (nucleation/growth) phase-transition model presents essentially the same bulk fluid response as a homogeneous equilibrium model with additional predictions of number density and average droplet size.
Numerical Simulation of the Internal Two-Phase Flow within an Aerated-Liquid Injector and its Injection into the Corresopnding High-speed Crossflows.
(2002-09-17) Tian, Ming; Dr. Jack R. Edwards, Committee Chair; Dr. Semyon V. Tsynkov, Committee Member; Dr. Hassan A. Hassan, Committee Member
Aerated-liquid atomization, which is produced by the introduction of gas directly into a liquid flow immediately upstream of the injector exit orifice to generate a two-phase flow, has been shown to produce well-atomized sprays in a quiescent environment with only a small amount of aerating gas at relatively low injection pressures. A time-derivative preconditioning method using the Low-Diffusion Flux-Splitting Scheme (LDFSS) has been extended to a 'mixture' model of two-phase flow and applied to simulate the structure of internal two-phase flow for aerated-liquid injectors, with each phase governed by its own equation of state. The Continuum Surface Force (CSF) model of Brackbill, et al. is adapted to model compressible fluid flow influenced by interfacial surface tension. A sub-iterative time integration method based on a planar Gauss-Seidel partitioning of the system matrix is used with implicit source terms as a means of solving the three-dimensional, time-dependent form of the governing equations. The calculations are parallelized using domain-decomposition and Message-Passing Interface (MPI) methods, and are optimized for operation on the 720 processor IBM SP-2 at the North Carolina Supercomputing Center (NCSC). Simulation results for 2-D aerated-liquid injector flowfields at gas-to-liquid (GLR) mass ratios of 0.08% and 2.45% are discussed. In accord with experimental visualization data, the results for GLR = 0.08% indicate a combination of slugging and core-annular two-phase flow in the injector. Results at GLR = 2.45% indicate that a core-annular flow mode dominates, again in agreement with experimental results. The effects of the choice of reference velocity and the level of surface tension on the injector flowfield solutions are also examined.
Numerical Simulation of the Internal Two-Phase Flow within an Aerated-Liquid Injector and its Injection into the Corresponding High-speed Crossflows
(2005-08-16) Tian, Ming; Dr. Hassan A. Hassan, Committee Member; Dr. D. Scott McRae, Committee Member; Dr. Zhilin Li (Dept. of Mathematics), Committee Member; Dr. Jack R. Edwards, Committee Chair
The current study investigates the flow structures within an aerated-liquid (barbotage) injector, which is designed to facilitate the rapid breakup of a hydrocarbon fuel jet prior to its entering a scramjet combustor, and the spray structures in the corresponding crossflow. Simulations of the transient, three-dimensional, two-phase flow within the "out-in" injector operating at different gas-to-liquid (GLR) mass ratios and in the corresponding crossflow domain have been performed, and the results compared with experimental pressure measurements of the injector and shadowgraph images of the crossflow. The numerical method solves a "mixture" model of two-phase flow using a preconditioning strategy. High-order spatial accuracy and good interface-capturing properties are facilitated by the use of shock-capturing schemes combined with second order TVD methods. Also, an immersed boundary method is used to investigate the probe effects, and a droplet transport model is used in the crossflow simulations to get more details about effect of droplet size. The injector simulation results highlight the effects of mesh refinement and turbulence model on the predicted solutions. The pressure drop across the injector is predicted reasonably well by the computational methodology, and the trend of increasing injector pressure with increasing GLR is captured properly. Predictions of the absolute pressure level within the injector show some discrepancies in comparison with experimental data but agree well with theoretical estimates. The results of the injector simulations with plenum included are consistent with the results of the discharge tube cases. If the centerline pressure is close to the experimental data, the gas mass flow rate at outlet will approach a value below the experimental data. If the gas mass flow rate at outlet approaches the experimental data, then the centerline pressure will be higher than the experimental data, but agrees well with theoretical analyses. The intrusion of the probe has little effect on the flowfield if the probe is contained wholly within the liquid core, but does affect the flowfield if the probe tip is in the two-phase mixing region, instead of the liquid core. The results of crossflow show that the two-phase flow injects into the crossflow, bends towards the streamwise direction, disperses into a spray plume, and initiates a horse-shoe shape structure of the jet in the cross-sectional planes. The result based on the previous injector simulation at a higher inlet gas pressure shows best penetration height prediction among all freestream Mach 0.3 cases. Including the droplet transport model gives a similar spray structure in the X-Z centerplane as that of the mixture model, but gives a different spray structure in the cross-sectional planes. The horse-shoe shaped structure fades away with increases in the droplet diameter size, and the liquid mass accumulates to the X-Z centerplane.
Particle Flow, Agglomeration, Mixing, Chemical and Physical Absorption in Circulating Fluidized Bed Absorbers
(2004-01-06) Mao, Deming; Dr. H. Henry Lamb, Committee Member; Dr. Hassan A. Hassan, Committee Member; Dr. Andrey V. Kuznetsov, Committee Co-Chair; Dr. Jack R. Edwards, Committee Chair
Coal-utilization for energy production poses considerable environmental concerns as it results in emission of sulfur dioxide (SO2), nitrogen oxide (NOx), fine particulate matter (PM), and trace heavy metals such as mercury vapor (Hg) during coal-combustion. Circulating Fluidized Bed Adsorbers (CFBAs) are regarded as a potentially effective technology to capture some of the above pollutants. In particular, one could use limestone to remove sulfur dioxide by chemical adsorption, and activated carbon to remove elemental mercury by physical adsorption. Also, sorbent particles could be used to capture fine PM or promote formation of clusters of larger PM. In order to analyze CFBA systems in detail, a new approach has been developed for solving the Navier-Stokes equations for a gas-mixture/solids-mixture system. Sub-models are also developed to be combined with the gas/solids hydrodynamics model to simulate capture of multiple pollutants. Specific tasks accomplished include the following. 1. A model for fine particle agglomeration in CFBAs has been developed. It can model the influence of different factors on agglomeration, such as the geometry of a CFBAs, the superficial gas velocity, initial particle size distribution (PSD), and type of agglomeration mechanism. It is found that the Brownian agglomeration mechanism can be neglected compared to agglomeration by mean shear and turbulence. Sorbent particles are shown to capture fine particles effectively for certain conditions. A simplified version of this model has been developed for coupling with the hydrodynamics model. 2. A mixing model based on a core-annulus model of a CFBA has been developed to simulate the particle residence time distribution (RTD). Thus, macrochemical reaction can be simulated by combining microchemical reaction dynamics with the particle RTD. This has been applied to simulate SO_2 removal by chemical adsorption onto dry lime. 3. A 'gas mixture' and ``solids mixture' model has been developed to simulate fine particle agglomeration onto sorbent particles, sulfur dioxide removal through chemical adsorption with lime, and mercury vapor removal through physical adsorption with activated carbon. The 'gas mixture' is composed of fine PM, sulfur dioxide, mercury vapor, oxygen and inert gas; while 'solids mixture' is composed of solids-1 and solids-2. Solids-1 is composed of lime (CaO) and CaSO_4, and solids-2 is activated carbon. These equations are similar to continuity equations appearing in the gas-solids hydrodynamics system and are integrated fully coupled with that system. 4. A new approach for solving the Navier-Stokes equations governing gas/solids two-phase flow with chemical reaction has been developed. The approach combines a time-derivative preconditioning strategy for a gas/solids two-phase flow model with extensions of low-diffusion flux-splitting upwinding techniques. The combined framework is used to simulate jet-induced bubble formation within a minimally fluidized bed, flow within a circulating fluidized bed without chemical reaction, a downward fludized bed, and gas mixture/solids mixture flow within a prototype CFBA device. For bubble bed simulation, the model gives good results for bubble formation, growth and burst. For three-dimensional CFBAs, it can give good results compared with experimental data and can capture details of solids clustering phenomena. For downward fluidized bed, the model gives plausible results regarding the influence of superficial gas velocity on particle flow and mixing. Finally, simulations of a ``bench-scale' CFBA reactor with combined SO_2, Hg and fine PM capture give reasonable results for gas species and solids species, but further validation is needed.