Parallel Computational Fluid Dynamics Research Articles

Performance analysis of large scale parallel CFD computing based on Code_Saturne

In order to run computational fluid dynamics (CFD) codes on large scales, parallel computing has to be employed. For instance, on Petascale computing, general parallel computing without any optimization is not enough, especially for complex industrial issues that employ a large number of mesh cells to capture the details of the geometry. How to distribute these mesh cells among the multi-processors for Terascale and Petascale systems to obtain a good performance on parallel computing is really a challenge. Some mesh partitioning software packages, such as Metis, ParMetis, PT-Scotch and Zoltan, were chosen as the candidates ported into Code_Saturne to test if they can lead Code_Saturne towards Petascale and Exascale parallel CFD computing. Through the studies, it was found that mesh partitioning optimization software packages based on the graph mesh partitioning method can help the CFD code obtain good mesh distributions for high performance computing (HPC).

MODELING AND SIMULATION OF MICRO-SCALE WIND FARMS USING HIGH PERFORMANCE COMPUTING

Onshore wind farms usually consist of numerous horizontal axis wind turbines closely placed in clusters, and they are often cited on complex terrain. This paper proposes a computational framework for modeling and simulation of wind flow over micro-scale (and early meso-scale) wind farms using distributed memory, massively parallel high performance computing platforms. The present framework uses the Reynolds Averaged Navier–Stokes (RANS) to model the wind flow over the wind farms, as the flow is considered to be fully turbulent, isothermal and incompressible. The wind turbines installed in the wind farm are modeled by the virtual blade model (VBM). This technique considers the presence of a wind turbine's rotor implicitly through momentum sources placed in an actuator disc, yielding indirectly a pressure jump across the disk, which varies with its radius and azimuth. The nonlinear, aerodynamic interaction between the rotor wakes with each other and with the terrain of the wind farm is simulated by coupling the VBM with the governing flow field equations. In this manner, an efficient parallel algorithm for implementing the VBM was developed and integrated with parallel computational fluid dynamics (CFD) core simulation engine. The accuracy and performance of the proposed framework were confirmed through several test cases carried out on the IBM Blue Gene ultra-scale supercomputer.

A CFD-based wind solver for an urban fast response transport and dispersion model

In many cities, ambient air quality is deteriorating leading to concerns about the health of city inhabitants. In urban areas with narrow streets surrounded by clusters of tall buildings, called street canyons, air pollution from traffic emissions and other sources may accumulate resulting in high pollutant concentrations. For various situations, including the evacuation of populated areas in the event of an accidental or deliberate release of chemical, biological and radiological agents, it is important that models should be developed that produce urban flow fields quickly. Various computational techniques have been used to calculate these flow fields, but these techniques are often computationally intensive. Most fast response models currently in use are at a disadvantage in these cases as they are unable to correlate highly heterogeneous urban structures with the diagnostic parameterizations on which they are based. In this paper, a novel variant of the popular projection method for solving the Navier–Stokes equations has been developed and applied to produce fast and reasonably accurate parallel computational fluid dynamics (CFD) solutions for flow in complex urban areas. This model, called QUIC-CFD represents an intermediate balance between fast (on the order of minutes for a several block problem) and reasonably accurate solutions. This paper details the solution procedure and validates this model for various simple and complex urban geometries.

High performance parallel computing of flows in complex geometries

Efficient numerical tools taking advantage of the ever increasing power of high-performance computers, become key elements in the fields of energy supply and transportation, not only from a purely scientific point of view, but also at the design stage in industry. Indeed, flow phenomena that occur in or around the industrial applications such as gas turbines or aircraft are still not mastered. In fact, most Computational Fluid Dynamics (CFD) predictions produced today focus on reduced or simplified versions of the real systems and are usually solved with a steady state assumption. This article shows how recent developments of CFD codes and parallel computer architectures can help overcoming this barrier. With this new environment, new scientific and technological challenges can be addressed provided that thousands of computing cores are efficiently used in parallel. Strategies of modern flow solvers are discussed with particular emphases on mesh-partitioning, load balancing and communication. These concepts are used in two CFD codes developed by CERFACS: a multi-block structured code dedicated to aircrafts and turbo-machinery as well as an unstructured code for gas turbine flow predictions. Leading edge computations obtained with these high-end massively parallel CFD codes are illustrated and discussed in the context of aircrafts, turbo-machinery and gas turbine applications. Finally, future developments of CFD and high-end computers are proposed to provide leading edge tools and end applications with strong industrial implications at the design stage of the next generation of aircraft and gas turbines.

Model Based Control of Laminar Wake Using Fluidic Actuation

Control of fluid-structure interaction is of practical importance from the perspective of wake modification and reduction of vortex-induced vibrations (VIVs). The aim of this study is to design a control to suppress vortex shedding. We perform a two-dimensional simulation of the flow past a circular cylinder using a parallel Computational Fluid Dynamics (CFD) solver. We record the velocity and pressure fields over a shedding cycle and compute the proper orthogonal decomposition (POD) modes of the divergence-free velocity and pressure, respectively. The Navier–Stokes equations are projected onto these POD modes to reduce the dynamical system to a set of ordinary-differential equations (ODEs). This dynamical system exhibits a limit cycle with negative linear damping and positive nonlinear damping. The reduced-order model is then modified by placing a pair of suction actuators and applying a control strategy using a control function method. We use the pressure POD mode distribution on the cylinder surface to optimally locate the actuators. We design a controller based on the linearized system and make it positively damped using pole-placement technique. The control-input settles to a constant value, suggesting constant suction through the actuators. We validate the results using CFD simulations in an open-loop setting and observe suppression of the hydrodynamic forces acting on the cylinder.

Parametric Design and Numerical Simulation of the Axial-Flow Fan for Electronic Devices

The procedure of the parametric design and numerical simulation is developed to enhance greatly fan design efficiency, which can generate high quality hexahedral mesh automatically. The numerical simulation can be performed automatically and the aerodynamic performance of the fan can be characterized without using any other tools. The developed procedure can be employed to the optimization design of the fan. Based on the procedure of the parametric design and numerical simulation, the influences of various design parameters on the aerodynamic performance of the fan are investigated deeply. Such influences can provide physical insight into the flow characteristics of the fan. The multiparameter constrained optimization procedure based on the combination algorithms of design of experiments, response surface models, genetic algorithm, and mixed integer optimization, is carried out aiming at the fan to maximize its coefficient of flow rate over the operating range rather than at a single operating point with parallel computational fluid dynamics. Results show that the aerodynamic average flow rate coefficient of the optimal fan is increased by 14.5% compared to the initially designed fan. This paper demonstrates the potential of the proposed procedure to be applied widely in industry for the purpose of improving work efficiency among engineers.

Dynamic tuning of algorithmic parameters of parallel scientific codes

We present a dynamic method for tuning algorithmic parameters of parallel scientific programs. By treating tuning as a separate concern in the software development process, our method supports personalized development of optimization schemes for existing programs that are not easily supported by conventional tuning techniques. We use a compositional framework to transparently combine tuning code with the original program without direct modification of an existing code base. In this way, the inserted tuning module can dynamically search for optimal values of algorithmic parameters, accounting for runtime factors such as input problem size and parallel characteristics of a given execution platform, as well as the architectural or runtime properties of a single machine of the platform. Applying our method to a parallel CFD (computational fluid dynamics) simulation, we demonstrate how a set of performancecritical parameters can be dynamically tuned, achieving up to 26% performance improvements over average cases.

High performance parallel computing of flows in complex geometries: I. Methods

Efficient numerical tools coupled with high-performance computers, have become a key element of the design process in the fields of energy supply and transportation. However flow phenomena that occur in complex systems such as gas turbines and aircrafts are still not understood mainly because of the models that are needed. In fact, most computational fluid dynamics (CFD) predictions as found today in industry focus on a reduced or simplified version of the real system (such as a periodic sector) and are usually solved with a steady- state assumption. This paper shows how to overcome such barriers and how such a new challenge can be addressed by developing flow solvers running on high-end computing platforms, using thousands of computing cores. Parallel strategies used by modern flow solvers are discussed with particular emphases on mesh-partitioning, load balancing and communication. Two examples are used to illustrate these concepts: a multi-block structured code and an unstructured code. Parallel computing strategies used with both flow solvers are detailed and compared. This comparison indicates that mesh-partitioning and load balancing are more straightforward with unstructured grids than with multi-block structured meshes. However, the mesh-partitioning stage can be challenging for unstructured grids, mainly due to memory limitations of the newly developed massively parallel architectures. Finally, detailed investigations show that the impact of mesh-partitioning on the numerical CFD solutions, due to rounding errors and block splitting, may be of importance and should be accurately addressed before qualifying massively parallel CFD tools for a routine industrial use.

Where is the rudder of a fish?: the mechanism of swimming and control of self-propelled fish school

Numerical simulation and control of self-propelled swimming of two- and three-dimensional biomimetic fish school in a viscous flow are investigated. With a parallel computational fluid dynamics package for the two- and three-dimensional moving boundary problem, which combines the adaptive multi-grid finite volume method and the methods of immersed boundary and volume of fluid, it is found that due to the interactions of vortices in the wakes, without proper control, a fish school swim with a given flapping rule can not keep the fixed shape of a queue. In order to understand the secret of fish swimming, a new feedback control strategy of fish motion is proposed for the first time, i.e., the locomotion speed is adjusted by the flapping frequency of the caudal, and the direction of swimming is controlled by the swinging of the head of a fish. Results show that with this feedback control strategy, a fish school can keep the good order of a queue in cruising, turning or swimming around circles. This new control strategy, which separates the speed control and direction control, is important in the construction of biomimetic robot fish, with which it greatly simplifies the control devices of a biomimetic robot fish.

A van der Pol–Duffing Oscillator Model of Hydrodynamic Forces on Canonical Structures

Numerical simulations of the flow past elliptic cylinders with different eccentricities have been performed using a parallel incompressible computational fluid-dynamics (CFD) solver. The pressure is integrated over the surface to compute the lift and drag forces on the cylinders. The numerical results of different cases are then used to develop reduced-order models for the lift and drag coefficients. The lift coefficient is modeled with a generalized van der Pol–Duffing oscillator and the drag coefficient is expressed in terms of the lift coefficient. The parameters in the oscillator model are computed for each elliptic cylinder. The results of the model match the CFD results not only in the time domain but also in the spectral domain.

On the stability and extension of reduced-order Galerkin models in incompressible flows

Proper orthogonal decomposition (POD) has been used to develop a reduced-order model of the hydrodynamic forces acting on a circular cylinder. Direct numerical simulations of the incompressible Navier–Stokes equations have been performed using a parallel computational fluid dynamics (CFD) code to simulate the flow past a circular cylinder. Snapshots of the velocity and pressure fields are used to calculate the divergence-free velocity and pressure modes, respectively. We use the dominant of these velocity POD modes (a small number of eigenfunctions or modes) in a Galerkin procedure to project the Navier–Stokes equations onto a low-dimensional space, thereby reducing the distributed-parameter problem into a finite-dimensional nonlinear dynamical system in time. The solution of the reduced dynamical system is a limit cycle corresponding to vortex shedding. We investigate the stability of the limit cycle by using long-time integration and propose to use a shooting technique to home on the system limit cycle. We obtain the pressure-Poisson equation by taking the divergence of the Navier–Stokes equation and then projecting it onto the pressure POD modes. The pressure is then decomposed into lift and drag components and compared with the CFD results.

An Approach for Inclusion of a Nonlocal Transition Model in a Parallel Unstructured Computational Fluid Dynamics Code

The implementation of an integral transition model in a parallel unstructured computational fluid dynamics code is described. In particular, an algorithm for gathering the nonlocal boundary layer values (momentum thickness and shape factor) from parallel distributed computational domains is presented. Transition modeling results are presented for a flat plate and for a low pressure turbine, covering a large variation of Reynolds numbers, Mach numbers, turbulence intensities, and incidence angles. Contrary to fully turbulent simulations, transitional predictions are in very good agreement with measurements. Furthermore, the computational overhead of transitional simulations is only 7% for one multigrid cycle.

Computational-Fluid-Dynamics-Based Kriging Optimization Tool for Aeronautical Combustion Chambers

The current state of the art in computational fluid dynamics provides reasonable reacting-flow predictions and is already used in industry to evaluate new concepts of gas turbine engines. In parallel, optimization techniques have reached maturity and several industrial activities benefit from enhanced search algorithms. However, coupling a physical computational fluid dynamics model with an optimization algorithm to yield a decision-making tool needs to be undertaken with care to take advantage of the current computing power while satisfying the gas turbine industrial constraints. Among the many delicate issues for such tools to contribute efficiently to the gas turbine industry, combustion is probably the most challenging, and optimization algorithms are not easily applicable to such problems. In our study, a fully encapsulated algorithm addresses the issue by making use of a new multiobjective optimization strategy based on an iteratively enhanced metamodel (kriging) coupled to a design-of-experiments method and a fully parallel three-dimensional computational fluid dynamics solver to model turbulent reacting flows. With this approach, the computer cost needed for thousands of computational fluid dynamics computations is greatly reduced while ensuring an automatic error reduction of the approximated response function. Preliminary assessments of the search algorithm against simple analytical test functions prove the strategy to be efficient and robust. Application to a three-dimensional industrial aeronautical combustion chamber demonstrates the approach to be feasible with currently available computing power. One result of the optimization is that possible design changes can improve performance and durability of the studied engine. With the advent of massively parallel architectures, the intersection between these two advanced techniques seems a logical path to yield fully automated decision-making tools for the design of gas turbine engines.

Simulation of entrained flow coal gasification

Integrated gasification combined cycle (IGCC) with carbon capture and storage (CCS) is a viable greenhouse gas control technology in using coal for power and/or hydrogen generation. This work describes the development of a pilot-scale pressurized entrained flow coal gasification facility and the parallel computational fluid dynamics (CFD) based simulation capability for the purpose of advancing commercial coal gasification technology. The gasification simulation approach, its present predictive performance and the potential areas for further improvement are presented.

Parallel CFD simulations of an original and redesigned hydraulic turbine draft tube

The design of hydraulic turbine draft tubes has traditionally been based on simplified analytic methods, model and full-scale experiments. In the recent years the use of numerical methods such as computational fluid dynamics (CFD) has increased in the design process, due to the rapid escalation in computer performance. Today with parallel computer architectures, new aspects of flow can be considered. However, several problems have still to be solved before CFD can routinely be applied in product development. This paper aims to investigate the parallel performance of commercial CFD software on homogeneous computer networks, as common solutions encountered in the industry today. In addition, the efficiency improvements obtained in earlier experiments by modifying the shape of the draft tube will be considered, to deduce if the improvements can be captured with aid of CFD. Results from both the steady and the unsteady CFD simulations show that almost full scalability is obtained with the commercial CFD software CFX-5.7.1. Furthermore, no noticeable improvement in the pressure recovery factor or the flow field is noticed in the CFD simulations as compared to experiments. The discrepancy may be to the applied inlet boundary conditions and/or the turbulence model.

Large eddy simulations of a turbulent thermal plume

Large eddy simulations of a three-dimensional turbulent thermal plume in an open environment have been carried out using a self-developed parallel computational fluid dynamics code SMAFS (smoke movement and flame spread) to study the thermal plume’s dynamics including its puffing, self-preserving and air entrainment. In the simulation, the sub-grid stress was modeled using both the standard Smagorinsky and the buoyancy modified Smagorinsky models, which were compared. The sub-grid scale (SGS) scalar flux in the filtered enthalpy transport equation was modeled based on a simple gradient transport hypothesis with constant SGS Prandtl number. The effect of the Smagorinsky model constant and the SGS Prandtl number were examined. The computation results were compared with experimental measurements, thermal plume theory and empirical correlations, showing good agreement. It is found that both the buoyancy modification and the SGS turbulent Prandtl number have little influence on simulation. However, the SGS model constant Cs has a significant effect on the prediction of plume spreading, although it does not affect much the prediction of puffing.

THREE-DIMENSIONAL NUMERICAL SIMULATION OF TURBULENT FLOW AND HEAT TRANSFER IN A CONTINUOUS GALVANIZING BATH

This article presents the application of a three-dimensional finite-element solution algorithm for the prediction of velocity and temperature fields in an industrial continuous galvanizing bath. The effect of line speed, strip width, strip temperature, and inductor mixing are evaluated. Simulations were carried out using a parallel computational fluid dynamics (CFD) software developed at the Industrial Materials Institute, Natural Research Council of Canada. The incompressible Navier-Stokes equations are solved for turbulent flows using the k - k model. Both forced-convection and temperature-dependent density conditions are considered in order to assess the buoyancy effect. When considering the buoyancy, the flow induced by variations in density is especially apparent near the inductors and the melting makeup ingot, while little effect is observed in the sheet and rollers region. Thermal effects are also amplified when the inductor is at high capacity, during the ingot melting. Simulations allow visualization of regions of varying velocity fields and clearly illustrate the mixed and stagnant zones for different operating conditions.

Coupling Of CFD And CSD Methodologies For Modeling BlastAnd Structural Response

This paper describes recent algorithm developments and applications of a program that couples parallel Computational Fluid Dynamics (CFD) and Computational Structural Dynamics (CSD) methodologies. FEFL098 is the CFD code used while DYNA3D handles the CSD portion. FEFL098 solves the time-dependent compressible Euler and Reynolds-Averaged Navier-Stokes equations on an unstructured mesh of tetrahedral elements. DYNA3D solves explicitly the large deformation, large strain formulation equations on an unstructured grid composed of bricks and hexahedral elements. While the initial coupled algorithm used the so-called glued-mesh approach, where the CFD and CSD faces match identically, failure of this approach to model severe structural deformation, as well as crack propagation in steel and concrete, led us to the development of the so-called embedded-mesh approach. Here, the CSD objects float through the CFD domain. While each approach has its own advantages, limitations, and deficiencies, the embedded approach was proven to be superior for the class of problems modeled here. Critical application of both approaches are described, including weapon detonation and fragmentation, airblast interaction with a reinforced concrete wall, and fragment/airblast interaction with a steel wall. The final application models the interaction of an external airblast with a generic steel ship.

Development And Applications Of An Embedded CSD ApproachFor Coupled CFD/CSD Modeling Of Blastlstructure Interactions

This paper describes recent developments and select applications of a program that couples parallel Computational Fluid Dynamics (CFD) and Computational Structural Dynamics (CSD) methodologies. FEFL098 is the CFD code used while DYNA3D handles the CSD portion. FEFL098 solves the time-dependent, compressible Euler and Reynolds-Averaged Navier-Stokes equations on an unstructured mesh of tetrahedral elements. DYNA3D solves explicitly the large deformation, large strain formulation equations on an unstructured grid composed of bricks and hexahedral elements. The initial algorithm constructed to model the coupled processes used the so-called glued-mesh approach, where the CFD and CSD faces match identically. Failure of this approach to model severe structural deformations in steel plates, as well as crack growth and propagation in steel and concrete, led us to the development of the so-called embedded-mesh approach, where the CSD mesh float through the CFD domain. While each approach has its own advantages, limitations and deficiencies, the embedded approach was proven to be superior for the class of problems modeled here. Critical applications of both approaches are described, including weapon detonation and fragmentation, airblast interaction with a reinforced concrete wall, and fragmentlairblast interactions with steel wall structures including a generic steel ship hull and a steel tower. Transactions on the Built Environment vol 71, © 2004 WIT Press, www.witpress.com, ISSN 1743-3509

A DIRECT SCHUR-FOURIER DECOMPOSITION FOR THE SOLUTION OF THE THREE-DIMENSIONAL POISSON EQUATION OF INCOMPRESSIBLE FLOW PROBLEMS USING LOOSELY COUPLED PARALLEL COMPUTERS

Parallel computers based on PC-class hardware (Beowulf clusters) provide a matchless computing power per cost unit. However, their network performance tends to be too low for standard parallel computational fluid dynamics (CFD) algorithms. A relevant example is the solution of the Poisson equations. The subject of this article is a direct Schur-Fourier decomposition (DSFD) algorithm that, for certain three-dimensional flows, produces an accurate solution of each Poisson equation with just one message, providing speed-ups of at least 24 in a low-cost PC cluster with a conventional network and 36 processors. Direct Numerical Simulation (DNS) of turbulent natural convection flow is used as a benchmark problem.