Thomas Hauser

Numerical Simulations of a Scramjet Isolator Using RANS and LES Approaches

The internal, compressible, turbulent flow through a scramjet-isolator configuration is presented, with the primary goal being to better determine the shock train leading edge location of a typical Mach 2 nozzle-isolator configuration. Both 2D and 3D approaches are utilized, in conjunction with a variety of different turbulence models taken from both RANS and filtered models. The effects of inlet turbulence, as well as the use of grid adaption techniques are evaluated under the 2D assumption and render certain simplifying assumptions valid for the 3D cases. Experimental comparisons reveal that the RANS approach best conform to experimental observations, while the LES approach showed the most degree of disparity. Further LES simulations are warranted however, particularly since these were performed without the aid of density based solvers, which greatly facilitates the prior to the inception of a new FLUENT release, (version 6.3) which contains several improvements over previous releases with respect to compressible, turbulent flows.

Application of Genetic Algorithms and Neural Networks to Unsteady Flow Control Optimization

Evolutionary algorithms have been successfully used as a design optimization tool in several engineering optimization problems. Here, a genetic algorithm is linked with a computational fluid dynamics code in a GA-CFD system to optimize the configuration of a dual synthetic jet arrangement. The test problem is a two-dimensional NACA 0012 at a high angle of attack. The optimal configuration significantly reduces the separation region over the airfoil, yielding higher lift and lower drag. The data generated from this evolution are also used to test a possible neural network replacement for CFD computations which if successful could significantly accelerate the GA-CFD process for these types of optimizations.

Design of a Portable Cluster Supercomputer for Particle Image Velocimetry Data Processing

Inthispaper,wepresentthedesignofaportableclustersupercomputercreatedspecifically forprocessingparticleimagevelocimetryimagedata.Tomakethiscomputersystemportable for a laboratory environment, it is designed to run on minimal power, to be lightweight, and portable. The hardware configuration consists of 12 processing nodes with a total of 48 processor cores, one master node with five terabytes of disk storage, and a gigabit ethernet interconnect.The total configuration along with a rugged transportation case has an approximate gross weight of 390lb and works on one standard 120V, 20A electric circuit. With this cluster computer, a speedup of 28 relative to standard serial processing of a large particle image velocimetry data set was achieved.

Simulation of Aerodynamic Influences on Rocket-Mounted Oxygen Sensors

Over the past several decades, atomic oxygen measurements taken from sounding rocket sensor payloads in the altitude range of 80-140 kilometers have shown marked variability. Many sounding rocket payloads contain atomic oxygen sensors that are located in close proximity to the payload surface, and are thus significantly influenced by flow field disturbances. Although several additional factors including Doppler shift and sensor contamination may also play a significant role in the accurate measurement of atomic oxygen concentrations, this work focuses solely on the effects due to the flow field. The present study utilizes the three-dimensional, steady-state, direct simulation Monte Carlo technique. In addition, the lower altitudes corresponding to near-continuum flow are solved via the Navier-Stokes equations with slip wall boundary conditions. The flow is simulated at 13 different altitudes, each with three separate rocket orientations, along both the rockets upleg and downleg trajectory for a total of 75 simulations. The numerical simulations show conclusively that the relative magnitudes of undisturbed versus disturbed atomic oxygen concentrations are highly dependent upon rocket orientation, and provide a quantitative means by which existing atomic oxygen concentration data sets may be corrected for aerodynamic influences.

Large Eddy Simulation of Strongly Heated Internal Gas Flows

This work is concerned with the modeling of a strongly heated, low Mach number, gas flowing upward within a vertical tube with constant heat flux boundary conditions. Four llarge eddy subgrid models are compared to two different RANS turbulence models in their ability to predict the temperature distribution in the heated pipe. All LES models predict the mean velocity profile reasonably well. The RMS values of the Smagorinsky-Lilly model and the WALE model show the same shape but different magnitude while the standard Smagorinsky model shows the maximum at a different location. The mean temperature profiles along the wall in the section with the prescribed heat flux are underpredicted by all LES models. This work is concerned with the modeling of a strongly heated, low Mach number, gas flowing upward within a vertical tube with (nearly) constant heat flux boundary conditions, in which forced convection is dominant. The heating rate is sufficiently high such that the fluid properties vary significantly in both the radial and axial directions. Since the flow is continually adjusting to the changing properties, fully developed velocity and temperature profiles never occur. Large Eddy Simulation (LES) and Direct Numerical Simulation (DNS) provide means for determining detailed information about these heated flows that may be difficult to obtain experimentally. The motivation for this problem stems from considerations that gas may be effectively employed as a coolant for advanced power reactors. In this case, it is desirable to achieve high thermal efficiencies which require high gas exit temperatures. To achieve high temperatures the mean velocity of the gas may be low enough that the relevant Reynolds number of the flow is less than 10,000. Experimental data for this class of flows is sparse. Perkins 1 obtained mean temperature distributions for dominant forced convection flow through a circular cylinder with significant gas property variations. Using essentially the same experimental apparatus, Shehata 2 obtained mean velocity distributions under some of the same flow conditions. They employed three different constant wall heat flux boundary conditions: “low” and “intermediate” heating rates at inlet Reynolds numbers of approximately 6000, and a “high” heating rate case at an inlet Reynolds number of approximately 4200. Perkins 1 characterized these three cases as turbulent, “subturbulent,” and laminarizing, respectively. The results of Perkins 1 and Shehata 2 have since been reported in detail by Shehata and McEligot. 3,4 Consequently, subsequent references to the Shehata/McEligot data refer to the data as presented in the PhD. thesis of Shehata 3 and the report of Shehata and McEligot, 4 derived from the results of Perkins 1 and Shehata. 2 These experimental data provide an opportunity to verify turbulence and subgrid scale models under the influence of strong heating and, consequently, large variations in fluid properties.

Flow Control Optimization Using Neural Networks and Genetic Algorithms

Evolutionary algorithms have now been used as tool to optimize complex design spaces in aerospace applications, notably in the areas of Multidisciplinary Design Optimization (MDO) [4, 2] and flow control [9]. However, in the latter area a limiting factor has been the cost of evaluating the performance of each tested flow control configuration. This process is conventionally accomplished through computational fluid dynamics (CFD) simulation of the design, but in cases that require a full, viscous Navier-Stokes solver, the computational demands still can be high even for simple flow control applications like steady suction or blowing. More challenging flow control applications typically require even more expensive simulations.

validation and Enhancement of Computational Fluid Dynamics and Heat Transfer Predictive Capabilities for Generation IV Reactor Systems

Nationwide, the demand for electricity due to population and industrial growth is on the rise. However, climate change and air quality issues raise serious questions about the wisdom of addressing these shortages through the construction of additional fossil fueled power plants. In 1997, the Presidents Committee of Advisors on Science and Technology Energy Research and Development Panel determined that restoring a viable nuclear energy option was essential and that the DOE should implement a R&D effort to address principal obstacles to achieving this option. This work has addressed the need for improved thermal/fluid analysis capabilities, through the use of computational fluid dynamics, which are necessary to support the design of generation IV gas-cooled and supercritical water reactors.

Evaluating LES Subgrid-Scale Models for High Heat Flux Flows

This work is concerned with the modeling of a strongly heated, low Mach number, gas flowing upward within a vertical tube with constant heat flux boundary conditions. Four large eddy subgrid models are compared in their ability to predict the temperature distribution in the heated pipe. All LES models predict the mean velocity profile reasonably well. The RMS values of the Smagorinsky-Lilly model and the WALE model show the same shape but different magnitude while the standard Smagorinsky model shows the maximum at a different location. The mean temperature profiles along the wall in the section with the prescribed heat flux are underpredicted by all LES models.Copyright

Parallel, Steady/Unsteady DSMC Simulations of the Aerodynamics of Sounding Rockets

Over the past several decades, atomic oxygen (AO) measurements taken from sounding rocket sensor payloads in the Mesosphere and lower Thermosphere (MALT) have shown marked variability. AO data retrieved from the second Coupling of Dynamics and Aurora experiment (CODA II) have shown that the data is highly dependent upon rocket orientation. Many sounding rocket payloads including CODA II, contain AO sensors that are located in close proximity to the payload surface and are thus significantly influenced by compressible, aerodynamic effects. These effects serve to inhibit the AO sensors’ ability to accurately determine undisturbed atmospheric conditions. The present research numerically models the influence caused by these aerodynamic effects using a newly developed parallel, steady/unsteady, three-dimensional, DSMC solver entitled foamDSMC. The solver’s parallel capabilities as well as it’s unsteady functionality demonstrates significant improvements over previous research conducted by the present authors’. The solver is used to simulate the steady flow regime at two kilometer intervals along both the up-leg and down-leg trajectories. Unsteady results are also presented and simulated near apogee. The results are used to create correction functions based on the ratio of undisturbed to disturbed flowfield concentrations. The numerical simulations verify the experimental results showing the strong influence of rocket orientation on concentration. The correction functions, when applied to uncorrected CODA II data sets, show a significant improvement in terms of minimizing the effects of compressible flow aerodynamics. A rarefied gas may be divided into several different flow regimes in accordance with its level of rarefaction as quantified by the Knudsen number (Kn). A significantly large number of flows may be classified within the transition regime (0.1 < Kn < 10), and constitute numerical simulation limits well outside that of conventional NavierStokes, continuum based solvers. Traditionally, the Boltzmann equation, based on kinetic theory, remained the only viable option for solution of these high Kn number flows. Recently however, direct particle simulation methods, not relying on the quantification of the velocity distribution function have become mainstay. Particularly relevant, is the direct simulation Monte Carlo (DSMC) method of G.A. Bird. The DSMC method is a direct simulation method based on kinetic theory, and may be regarded as a numerical solution to the Boltzmann equation in the limit of very large numbers of simulated molecules. 1,2 The method, as its name connotes may be categorized as a Monte Carlo method in that it makes extensive use of random numbers to help in the stochastic generation of continuum variables. The primary objective of the method is to approximate these variables by means of modeling the interactions of a statistically significant number of simulated molecules, each representing a much larger subset of the real, un-modeled gas. The deterministic motions and probabilistic collisions of these simulated molecules are modified over sufficiently small time steps, which is governed by the mean collision frequency. Since the tracking and molecular interactions are conducted on a particle by particle basis, the conservation of mass, momentum, and energy may be enforced to machine accuracy. 3

Determination of Laser Welding Process Parameters Through a Combined Neural Networks and Computational Fluid Dynamics Approach

In this paper weld pool shapes of type 304 stainless steel obtained through computational fluid dynamics simulations (CFD) will be used to train artificial neural networks. The neural network is then used to determine the experimental parameters of the ND:YAG laser system on 304 steel by feeding it with the parametrized, experimentally obtained weld pool parameters. The output of the neural network are the process parameters like laser spot diameter and laser power on the surface of the weld pool. These parameters are very difficult to determine accurately through experiments but are necessary inputs for CFD simulations of the weld pool. Accurate CFD simulation is required for simulation and modeling of the solidified micro structure of the material.Copyright