Christopher P. Stone

Large-Eddy Simulation of Turbulent Flow over an Axisymmetric Hill

Large-Eddy Simulation (LES) of high Reynolds number turbulent flow over an axisymmetric hill of height H = 2δ, where δ is the incoming turbulent boundary layer thickness, is presented in this paper. Surface mean pressure, flow visualization and mean velocity profiles are presented and compared with experiments. In general, good agreement is obtained for the surface mean pressure and axial flow profiles. Analysis of the results show that multiple separation and reattachments occur on the lee-side of the hill. These results are also consistent with experimental observations, although there are some differences. The flow field is further analyzed to understand the dynamics of the flow.

GPGPU parallel algorithms for structured-grid CFD codes

A new high-performance general-purpose graphics processing unit (GPGPU) computational uid dynamics (CFD) library is introduced for use with structured-grid CFD algorithms. A novel set of parallel tridiagonal matrix solvers, implemented in CUDA, is included for use with structured-grid CFD algorithms. The solver library supports both scalar and block-tridiagonal matrices suitable for approximate factorization (AF) schemes. The computational routines are designed for both GPU-based CFD codes or as a GPU accelerator for CPU-based algorithms. Additionally, the library includes, among others, a collection of nite-volume calculation routines for computing local and global stable time

Modeling Rotor Wakes with a Hybrid OVERFLOW-Vortex Method on a GPU Cluster

The vortex core shed from rotorcraft blades maintains coherency—and thus dynamic relevance—many blade turns after its creation. This presents a challenge to traditional Eulerian computational methods, as fine grids are required to suppress numerical diffusion which would weaken the vortex cores after a small number of revolutions. Vortex methods have been used in the past to overcome these problems, as they require computational elements only in vorticity-containing regions, but suffer from greater computational cost per element. In the present work, we will solve these problems with a hybrid EulerianLagrangian method for modeling rotor wakes. An Eulerian OVERFLOW overset grid method computes the near-body flow, while a Lagrangian particle vortex method tracks the wake. The vortex method uses an anisotropic LES model to handle subgrid-scale dissipation explicitly. The computational cost of vortex methods is alleviated by using a parallel adaptive treecode on a cluster of machines each with multi-core CPUs and multiple costefficient graphics processing units (GPUs). Simulations of a low-Re sphere, finite wing, and 4-bladed rotor model are presented and are validated by comparisons with computational and experimental data. Rotorcraft operate in a highly complex vortex-dominated aerodynamic environment, characterized by unsteady non-homogenous turbulent flow interacting with the craft structure. The fuselage bluff body is often associated with unsteady separated flow. Further, the rotating blades generate highly energetic vortices, which invariably lead to the familiar phenomenon of blade-vortex interactions (BVI). BVI induces unsteady, non-periodic impulsive airloads along the length of the blades; thereby, increasing the vibration of the blades and the airframe. This has a strong impact on the stability of flight dynamics as well as the fatigue life of the vehicle. A comprehensive design and analysis tool that can predict the coupled fluid, structural, and vehicle dynamics of rotorcraft with high fidelity would greatly enhance the capability of the designer or analyst to understand the physics of the problem with better clarity, and it will ultimately lead to optimal aircraft designs. Eulerian computational fluid dynamics methods are very efficient in accurately resolvingthe flow in the immediate vicinity of the helicopter boundary, which is primarily anisotropic and essentially unidirectional in nature. Furthermore, mature technologies exist for Eulerian simulation of compressible flow, which, for rotor blades, is most significant within this same near body region. However, as is well known within the CFD community, the method is notoriously diffusive and tends to dampen high-intensity vortical structures within

Parallel Simulations of Swirling Turbulent Flames

The feasibility of using massively paralleled computations as an engineering design tool is evaluated. A parallel Large-Eddy Simulation (LES) algorithm which simulates turbulent reacting flows using a space and time-accurate method, is used to model the complex flow found inside a realistic gas-turbine combustor. The parallelization philosophy and its implementation as a platform-independent solver is discussed. A performance analysis is carried out to determine the communication and storage requirements, and the associated overhead. As a case study, the LES methodology is used for a parametric investigation of swirl effects on the turbulent reacting flow in the gas-turbine.

Rotor wake modeling with a coupled Eulerian and Vortex Particle Method

A new coupled Eulerian/Lagrangian CFD method is presented for rotorcraft wake ∞ow modeling. Speciflcally, the Vortex Particle Method is coupled with an overset, flnite-difierence URANS algorithm to solve the wallbounded and wake ∞ow. The coupled algorithm is presented in detail along with the necessary parallel computing algorithms. The coupled algorithm is then used to model the ∞ow over a NACA0015 airfoil wing at 12 ‐ angle-of-attack. Results from the coupled algorithm are compared to a baseline CFD solution and, where available, experimental data. Surface pressure proflles, sectional loads and tip vortex visualization and velocity proflles are compared to assess the efiectiveness of the coupling algorithm. Comparisons show that the coupled approach can capture the tip vortex far better than the baseline URANS solution. However, the particle solution can be signiflcantly impacted by the excessive dissipation through the Eulerian domain.

Computational Fluid Dynamics of Flatback Airfoils for Wind Turbine Applications

This paper presents results from a computational study of the aerodynamic performance of various atbac k airfoils designed for wind turbines. Multiple turbulence modelings methods are used for the aerodynamic modeling: Unsteady Reynolds-Averaged Navier-Stokes (URANS) equations, Detached-Eddy Simulations (DES) and a Hybrid RANS/Large-Eddy Simulations (HRLES) method based on the k{! RANS model and k{equation sub-grid LES model. All simulations make use of overset structured grids. Turbulence modeling and grid resolution studies show that both DES and HRLES methods capture the expected qualitative turbulent behavior such as cross o w in the separated wake regions of the o w. Quantitative results include the predicted lift and drag over a range of angles-of-attack using URANS. It is shown that the atbac k airfoil design results in a substantially increased lift compared to traditional thin trailing edge airfoils. Flow-eld results from the two advanced modeling methods, DES and HRLES, are qualitatively compared and analyzed. It is observed that the DES method does not predict the transition and separation along the airfoil while the HRLES method does. This limitation is explained and the eects on the aerodynamic forces is also given.

Techniques for solving stiff chemical kinetics on GPUs

numbers of ODEs with comparable solution accuracy. The GPU implementation of the DVODE solver achieved a maximum speed-up of 7.7x over the baseline CPU run-time. The performance impact of mapping one thread to each ODE was compared to mapping an entire CUDA thread-block per ODE (i.e., multiple threads per ODE). The one-threadper-ODE approach achieved greater overall speed-up compared to the one-block-per-ODE approach but only when the number of ODEs was large: 1,000 ODEs were needed just to break even with the scalar CPU version and over 50,000 ODEs to reach maximum parallel efficiency. The performance difference is most pronounced with the RKF45 algorithm. The peak performance with the one-thread-per-ODE method was nearly 2x faster than the one-block-per-ODE approach. The one-block-per-problem implementation of RKF45 and DVODE both achieved lower peak speed-ups but outperformed the scalar CPU performance with as few as 100 ODEs. The new GPU-enabled ODE solvers demonstrate a method to significantly reduce the computational cost of detailed finite-rate combustion simulations with turn-around cost savings exceeding an order of magnitude.

LES of Partially-Premixed Unsteady Combustion

Large-Eddy Simulation (LES) methodology has been used to model the combustion dynamics in a realistic swirl-stabilized dump combustor. A novel partially-premixed flamelet combustion model is employed to capture the interaction of the unsteady flame-front with local velocity and equivalence ratio fluctuations. Two sets of simulations have been conducted which employ this combustion model. First, the model is used to simulate temporal variations in the mean inlet fuel flow-rate mimicking pressure and fuel feed-line interaction, a major source of instabilities. The second application models the effects of unmixedness as a result of spatial various in the inlet equivalence ratio. A baseline, uniform equivalence ratio simulation is also reported and is used for comparison. It is shown that the present model is capable of capturing the combustion dynamics under both temporal and spatial fluctuations in the equivalence ratio.

Performance analysis, design considerations, and applications of extreme-scale in situ infrastructures

A key trend facing extreme-scale computational science is the widening gap between computational and I/O rates, and the challenge that follows is how to best gain insight from simulation data when it is increasingly impractical to save it to persistent storage for subsequent visual exploration and analysis. One approach to this challenge is centered around the idea of in situ processing, where visualization and analysis processing is performed while data is still resident in memory. This paper examines several key design and performance issues related to the idea of in situ processing at extreme scale on modern platforms: scalability, overhead, performance measurement and analysis, comparison and contrast with a traditional post hoc approach, and interfacing with simulation codes. We illustrate these principles in practice with studies, conducted on large-scale HPC platforms, that include a miniapplication and multiple science application codes, one of which demonstrates in situ methods in use at greater than 1M-way concurrency.

Large-eddy simulations on distributed shared memory clusters

The practicality of Large-eddy simulation (LES) of turbulent combustion, as is found in gas turbine engines, on clusters of commodity PC-based symmetric multi-processor (SMP) systems in 2-, 4-, and 8-way configurations has been investigated. Bandwidth demands from both memory and networking in the benchmark LES algorithm are shown to the primary performance inhibitors. Contention in the various SMP architectures tested is shown to compound these two hardware limitations. To investigate the ability of the parallel clustered systems, low-level hardware studies are conducted in conjunction with bench-marking of the LES application. The hardware tests focus on memory and communication contention under loads found in the LES algorithm. For comparison, the benchmarks are also applied to two industry leading high-performance super-computing architectures. It is found that contention in the 4- and 8-way SMP architecture studied here limits their applicability while the 2-way systems shows competitive performance and speed-up compared to its industry counter-parts. It is concluded that design-level combustion LES on clusters of commodity hardware, when equipped with sufficient memory and communication bandwidth, are a viable substitute for more expensive super-computing platforms.