Dominic Chandar

CU++: an object oriented framework for computational fluid dynamics applications using graphics processing units

The application of graphics processing units (GPU) to solve partial differential equations is gaining popularity with the advent of improved computer hardware. Various lower level interfaces exist that allow the user to access GPU specific functions. One such interface is NVIDIA’s Compute Unified Device Architecture (CUDA) library. However, porting existing codes to run on the GPU requires the user to write kernels that execute on multiple cores, in the form of Single Instruction Multiple Data (SIMD). In the present work, a higher level framework, termed CU++, has been developed that uses object oriented programming techniques available in C++ such as polymorphism, operator overloading, and template meta programming. Using this approach, CUDA kernels can be generated automatically during compile time. Briefly, CU++ allows a code developer with just C/C++ knowledge to write computer programs that will execute on the GPU without any knowledge of specific programming techniques in CUDA. This approach is tremendously beneficial for Computational Fluid Dynamics (CFD) code development because it mitigates the necessity of creating hundreds of GPU kernels for various purposes. In its current form, CU++ provides a framework for parallel array arithmetic, simplified data structures to interface with the GPU, and smart array indexing. An implementation of heterogeneous parallelism, i.e., utilizing multiple GPUs to simultaneously process a partitioned grid system with communication at the interfaces using Message Passing Interface (MPI) has been developed and tested.

A GPU-based incompressible Navier–Stokes solver on moving overset grids

In pursuit of obtaining high fidelity solutions to the fluid flow equations in a short span of time, graphics processing units (GPUs) which were originally intended for gaming applications are currently being used to accelerate computational fluid dynamics (CFD) codes. With a high peak throughput of about 1 TFLOPS on a PC, GPUs seem to be favourable for many high-resolution computations. One such computation that involves a lot of number crunching is computing time accurate flow solutions past moving bodies. The aim of the present paper is thus to discuss the development of a flow solver on unstructured and overset grids and its implementation on GPUs. In its present form, the flow solver solves the incompressible fluid flow equations on unstructured/hybrid/overset grids using a fully implicit projection method. The resulting discretised equations are solved using a matrix-free Krylov solver using several GPU kernels such as gradient, Laplacian and reduction. Some of the simple arithmetic vector calculations are implemented using the CU++: An Object Oriented Framework for Computational Fluid Dynamics Applications using Graphics Processing Units, Journal of Supercomputing, 2013, doi:10.1007/s11227-013-0985-9 approach where GPU kernels are automatically generated at compile time. Results are presented for two- and three-dimensional computations on static and moving grids.

On the integral constraint of the pressure Poisson equation for incompressible flows

We illustrate, using analytical and numerical proofs, how a conservative discretisation of the pressure Poisson equation arising out of the discretisation of the incompressible Navier–Stokes equations (on a two-dimensional unstructured non-staggered grid) satisfies the integral constraint on the pressure boundary condition without any additional treatment. When discretised in a non-conservative manner, it is seen that the integral constraint is not exactly satisfied, but only to an order , where is an appropriate velocity scale. When solved using an iterative method, such as the Bi-Conjugate Stabilised method, it is proved that the vanishing sum of residuals on all points inclusive of the boundary is a consequence of this integral constraint. This result can then be used as a tool to identify whether the discrete integral constraint has been satisfied or not, especially when the pressure is solved as a Neumann problem.

A Hybrid Multi-GPU/CPU Computational Framework for Rotorcraft Flows on Unstructured Overset Grids

The advent of General Purpose Graphics Processing Units (GPGPUs) has spawned a lot of interest in computing high resolution flows in a shorter span of time. In combination with existing parallel programming techniques such as MPI or openMP, one is able to obtain at least one order increase in speed-up for CFD applications on stationary grids. Due to a high throughput/cost ratio, GPUs are increasingly becoming popular among the CFD community. Flow fields on helicopter systems rank among one of the most challenging, and hence are computationally demanding to simulate. Further, presence of multiple bodies moving relative to each other require use of overset grid systems, which in turn, require efficient overset grid assembly methods to support the flow solution. In this context, a computational framework for flow simulation across multiple CPU cores and multiple GPUs has been developed and tested. It is shown that the overall wall-clock time for the simulation can be considerably reduced by using multiple GPU cards. The parallel GPU framework is initially tested using both explicit and implicit schemes for the flow past a sphere. Further, a two-blade hovering rotor test case adopted from literature is used to demonstrate the capability of the code towards simulating rotorcraft flows in a short span of time.

Numerical Investigation of Flow Interference Effects in Fixed and Vibrating Tandem Square Columns Using Hybrid RANS-LES Models

Investigation of flow past tandem and side-by-side circular and square columns is of interest in offshore engineering. Flow past fixed and vibrating circular columns has received a lot of focus in the literature. However, the studies focused on square columns, especially at high Reynolds numbers are very limited due to the computational cost of large eddy simulation (LES). Unsteady Reynolds-averaged Navier-Stokes (URANS) methods are limited by their accuracy, especially for tandem columns in the wake interference regime (spacing to diameter ratio: L=D ∼ 3:0). Hybrid URANS-LES models (URANS near the solid-wall and LES away from the wall) can overcome the drawbacks of the traditional URANS methods and can provide a reasonable prediction of the flow physics at a fraction of the cost of LES without significantly sacrificing numerical accuracy.Arbitrary Lagrangian-Eulerian (ALE) methods fails when vibrating tandem bodies are in close proximity to each other or vibrate at high reduced velocities. Remeshing the domain can be expensive, especially at high Reynolds numbers (Re). Alternate strategies are necessary to efficiently simulation this problems. This study proposes the use of a non-linear URANS-LES model, coupled with an overset mesh method (for vibrating columns), for studying flow past tandem square columns. Simulations are performed at sub-critical Re to match the experimental Re. The initial results are encouraging for further investigation of fixed and vibrating tandem square column flow interference at high Reynolds numbers.Copyright

Flow Past Tandem Circular Cylinders at High Reynolds Numbers using Overset Grids in OpenFOAM

This study investigates the application of overset grids for simulating high Reynolds number flows in OpenFOAM. The problem of tandem cylinders is specifically chosen as an appropriate test model due to (a) a direct relevance to the marine industry and (b) the availability of experimental results conducted at the Basic Aerodynamics Research Tunnel at LaRC. Usage of overset grids in OpenFOAM has previously been demonstrated only using SUGGAR++, and this library is not easily available to the general public. This paper thus explores the application an in-house overset grid library (Overset Parallel Engine for Aerodynamics Applications) OPErA in OpenFOAM. For verification purposes of the overset coupling, results are presented for the inline oscillation of two cylinders in low Reynolds numbers Re 40. For the high Reynolds number case of Re 1.66×10, an assesment of the forces and the flow between the cylinders are presented for static tandem cylinders with a distance to diameter spacing of (L/D = 1.4). This specific spacing corresponds to a sub-critical spacing (short separation) where the drag on the rear cylinder is negative. For single grid computations, force predictions for (a) a non-linear URANS model and (b) a hybrid model are presented. For overset grid computations, only non-linear URANS results are presented to verify the implementation of the overset coupling. Comparisons between single grid and overset grid cases for the forces and the flow features are very encouranging and demonstrate greater potential in using overset grids for moving body problems.

Evaluation of the effect of Realistic and Synthetic Inflow on the Power and Loading Pattern of Wind Turbine

This study is focused on investigating the differences between realistic and synthetic inflow models for the large eddy simulation (LES) of the flow field and prediction of aerodynamics in an onshore and offshore wind farm . Weather Research and Forecasting (WRF) is run in LES mode for generating the realistic inflow boundary conditions while synthetic inflow is generated using Mann’s model. WindPact and NREL 5MW turbines were used as the model turbines for onshore and offshore wind farms, respectively. The geostropic wind driving the flow field in WRF-LES was adjusted to ensure the same wind speed at hub height for onshore and offshore cases, respectively to perform a comparative study. The coupling between the mesoscale and microscale codes are performed using a mesoscale microscale coupling interface (MMCI) developed as a part of our earlier work. This framework allows the efficient interpolation of data between the different codes in an automated and parallel fashion. Three different microscale computational fluid dynamics codes were