Massively parallel implementation of an explicit CFD algorithm on unstructured grids, II

SummaryThe high‐performance conjugate gradients (HPCG) and high‐performance geometric multi‐grid (HPGMG) benchmarks are alternatives to the traditional LINPACK benchmark (HPL) in measuring the performance of modern HPC platforms. We performed HPCG and HPGMG benchmark tests on a Cray XE6/XK7 hybrid supercomputer, Blue Waters at National Center for Supercomputing Applications (NCSA). The benchmarks were tested on CPU‐based and GPU‐enabled nodes separately, and then we analyzed characteristic parameters that affect their performance. Based on our analyses, we performed HPCG and HPGMG runs in multiple program, multiple data (MPMD) mode in Cray Linux Environment in order to measure their hybrid performance on both CPU‐based and GPU‐enabled nodes. We observed and analyzed several performance issues during those tests. Based on lessons learned from this study, we provide recommendations about how to optimize science applications on modern hybrid HPC platforms.

In this paper, a three-dimensional computational study was carried out to study the turbulent reactive flow during the combustion of an inter-turbine burner (ITB) with varied circumferential cavity volume and fuel injection temperature. The present ITB configuration, which was based on the high g-load Ultra-Compact Combustor (UCC) concept, used curved radial vanes (CRV). The emphasis of this study was to determine the ITB performance changes due to the changes in circumferential cavity volume at cold and heated fuel injection conditions. The turbulent flow in the ITB was modeled with a RANSbased realizable k-e turbulence model. The JET-A liquid fuel spray combustion was modeled using an eddy-dissipation model with one-step global chemical reaction mechanism. Numerical computations were implemented with a second-order finite volume method on an unstructured grid. Detailed results of temperature distributions, turbulence intensity, vorticity, pressure drop, and species transport and tracking are presented and discussed.

Steady turbulent viscous incompressible fluid flow in a plane channel is calculated for the case of uniform blowing and suction through opposite walls. There are no experimental data for flows of this type. The flows were calculated by two methods: a direct numerical simulation method and using a three-parameter turbulence model. Direct numerical simulation was carried out using the same (apart from the boundary conditions) algorithm for numerical solution of the Navier-Stokes equations as that used earlier for calculating flows in pipes and channels with impermeable walls. In the second group of calculations the version of the model published in 1978 was used. The results obtained by the two methods are in good agreement. The difference is within the spread of the experimental data used for determining the parameters of the model. The agreement obtained makes it possible to assert that the turbulence direct numerical simulation algorithm developed can be used for the analysis of flows with quite different boundary conditions, including cases where there are no corresponding experimental data.

CFDA parallel Method for Adaptive Unstructured Grids with Optimum Static Grid Repartitioning

Dynamic load balancing is a key problem for the efficient use of parallel systems when solving applications with unpredictable load estimates. However, depending on the underlying programming paradigm Single Program Multiple Data (SPMD) or Multiple Program Multiple Data (MPMD) the balancing requirements vary. In SPMD scenarios, a perfect load balance is desired, whereas in MPMD scenarios it might be better to quickly obtain a large reduction in load imbalance in a short period of time. We propose extending the local domain of a given processor in the load-balancing algorithms to find a better scope for each paradigm. For that purpose, we present a generalised version of the Diffusion Algorithm Searching Unbalanced Domains (called ds-DASUD), which extends the local domain of each processor beyond its immediate neighbour. ds-DASUD belongs to the iterative distributed load-balancing (IDLB) class and, in its original formulation, operates in a diffusion scheme where a processor balances its load with all its immediate neighbours (ds=1). We evaluate this algorithm for the two programming paradigms varying the domain size. The evaluation was carried out using two simulators (load-balancing and network simulators) for a large set of load distributions that exhibit different degrees of initial workload unbalancing. These distributions were applied to torus and hypercube topologies, and the number of processors ranged from 8 to 128. From these experiments, we conclude that the 1-DASUD fits well for SPMD scenarios, whereas for MPMD 3-DASUD and ((d/2)+1)-DASUD for hypercube and torus topologies, respectively, obtain the best trade-off between the imbalance reduction (up to 85%) and the cost incurred in reaching it.

In this paper we focus on the application of a higher-order finite volume method for the resolution of Computational Aeroacoustics problems. In particular, we present the application of a finite volume method based in Moving Least Squares approximations in the context of a hybrid approach for low Mach number flows. In this case, the acoustic and aerodynamic fields can be computed separately. We focus on two kinds of computations: turbulent flow and aeroacoustics in complex geometries. Both fields require very accurate methods to capture the fine features of the flow, small scales in the case of turbulent flows and very low-amplitude acoustic waves in the case of aeroacoustics. On the other hand, the use of unstructured grids is interesting for real engineering applications, but unfortunately, the accuracy and efficiency of the numerical methods developed for unstructured grids is far to reach the performance of those methods developed for structured grids. In this context, we propose the FV-MLS method as a tool for accurate CAA computations on unstructured grids.

A simplified finite volume lattice Boltzmann method for simulations of fluid flows from laminar to turbulent regime, Part II: Extension towards turbulent flow simulation

Flow passing a heated square cylinder is investigated using a hybrid LES-RANS approach on unstructured grids at a moderate Reynolds number of 22, 050. The effects of inflow turbulence on the flow field as well as surface convective heat transfer are studied by adopting a grid-based random-number method (GRM). Validation of the GRM method is carried out by generating and simulating a decaying homogeneous isotropic turbulent flow. Analysis of the turbulent quantities and comparisons with the Synthetic Coherent Eddy Method (SCEM) suggest that the cheaper GRM is able to generate good quality inflow turbulence, despite a longer transition region is required. Studies of the heated square cylinder in crossflow show that the inflow turbulence results in an early breakup of the shear layer, which leads to further effects on the vortex shedding and surface heat transfer. The surface convective heat transfer is increased with inflow turbulence, especially on the front, top and bottom surfaces, while limited influence is found on the rear surface. It is shown by the spectra that the effects of the inflow turbulence mainly focus on the turbulent shear layers, as well as heat transfer of the corresponding surfaces.

Numerical investigations in the sub-and supersonic flowfield around space configurations are presented. The subsonic turbulent flow around the space plane configuration ELAC as well as the supersonic laminar flow around have been considered. The third configuration is a two-stage, i. e. EOS mounted on ELAC. system. The numerical results are obtained by solving the Navier-Stokes equations using a finite volume method based on node-centered blockstructured grids in curvilinear coordinates. The Baldwin-Lomax turbulence model was used for fully turbulent flow without transition. The results presented are compared with experimental findings and show satisfactory agreementKeywordsSpace ConfigurationSupersonic FlowLift CoefficientExpansion RampMaximum Courant NumberThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

This paper describes the current state in the development of a computational steering environment (CSE) for indoor air-flow simulation. The system is especially desiged to provide an interactive evaluation of thermal comfort. It consists of an integrated, VR-based visualization client, a parallel computational fluid dynamics (CFD) kernel and a 3D grid generator component. The CSE is intended to support engineers with an easy-to-use tool for designing HVAC systems (heating, ventilating and air conditioning), for example, in the early stages of a project. By utilizing high-performance computing, the tool enables the investigation of indoor air-flow scenarios approaching closely to "real-time". The user interface for visualizing simulation results and the "steering" of the simulator have been combined to form a single application front-end. This allows an intuitive access for steering the simulation, changing the set-up of geometry and boundary conditions and subsequently exploring the results. The simulation kernel is based on a hybrid lattice Boltzmann method with extensions for large-eddy simulations of turbulent convective flows. The next steps in the project are to improve the VR client, enhance the treatment of boundary conditions in the computational kernel and extend the simulator in conjunction with partners in order to integrate a local thermal comfort model.

Isothermal and heated subsonic jet noise using large eddy simulations on unstructured grids

In this paper, the numerical dissipation properties of the Spectral Difference (SD) method are studied in the context of vortex dominated flows and wall-bounded turbulence, using uniform and distorted grids. First, the validity of using the SD numerical dissipation as the only source of subgrid dissipation (the so-called Implicit-LES approach) is assessed on regular grids using various polynomial degrees (namely, p = 3, p = 4, p = 5) for the Taylor-Green vortex flow configuration at R e = 5 000. It is shown that the levels of numerical dissipation greatly depend on the order of accuracy chosen and, in turn, lead to an incorrect estimation of the viscous dissipation levels. The influence of grid distortion on the numerical dissipation is then assessed in the context of finite Reynolds number freely-decaying and wall-bounded turbulence. Tests involving different amplitudes of distortion show that highly skewed grids lead to the presence of small-scale, noisy structures, emphasizing the need of explicit subgrid modeling or regularization procedures when considering coarse, high-order SD computations on unstructured grids. Under-resolved, high-order computations of the turbulent channel flow at R e τ = 1000 using highly-skewed grids are considered as well and present a qualitatively similar agreement to results obtained on a regular grid.

The objective is to evaluate the applicability of the penalty method for practically high-Re flows and to investigate the influence of wall model on the quality of the predicted results. This penalty method is introduced at first time on staggered Cartesian grid and then using unstructured mesh. For validation, due to faster convergence, 2D finite volume method in a cartesian grid is used to numerically investigate the incompressible unsteady laminar (Re = 100) and turbulent (Re = 3900) flow around a circular cylinder and the turbulent flow in a channel at \(Re_{u_{\tau}}=395\). 3D simulation is performed on an unstructured grid to simulate the turbulent flow behind a circular cylinder at Re = 3900 without the use of a wall model.

The physical aspects governing accurate numerical simulation of turbulent flows having large regions of crossflow separation are re-examined. Time-accurate, three-dimensional fine-grid Navier-Stokes solutions were obtained for turbulent subsonic flows over a slender ogive-cylinder body of revolution at large angles of attack. These flowfields are complex and contain regions of crossflow separation and an organized leeward-side vortex structure. An algebraic eddy-viscosity turbulence model has been modified to correctly account for the effects of the vortices on the underlying viscous layers

Two-dimensional Finite Volume Method (FVM) based on Reynolds-averaged Navier-Stokes (RANS) equations is applied to solve the turbulent viscous flow around sphere and pod. Unstructured grid with boundary layer treatment is constructed around sphere whereas structured grid is generated around pod. Spalart-Allmaras (S-A) and Shear Stress Transport (SST) k-? turbulence models are used for sphere but SST k-? turbulence model is used only for pod to solve turbulent viscous flows at Reynold’s number of 5×106 and 3×106 respectively. The numerical results in terms of the skin friction coefficient, pressure coefficient and drag coefficient are shown either graphically or in the tabular form. Velocity vectors as well as contour of pressure and velocity distribution are also displayed. Finally, the comparative study between flows around sphere and pod is done.DOI: http://dx.doi.org/10.3329/jname.v8i1.7388