R. Borrell

Low-frequency unsteadiness in the vortex formation region of a circular cylinder

The presence of low-frequency fluctuations in the wake of bluff bodies have been observed in several investigations. Even though the flow past a circular cylinder at Re = 3900 (Re = UrefD/ν) has been the object of several experimental and numerical investigations, there is a large scattering in the average statistics in the near wake. In the present work, the flow dynamics of the near wake region behind a circular cylinder has been investigated by means of direct numerical simulations and statistics have been computed for more than 858 shedding cycles. The analysis of instantaneous velocity signals of several probes located in the vortex formation region, point out the existence of a low-frequency fluctuation at the non-dimensional frequency of fm = 0.0064. This large-scale almost periodic motion seems to be related with the modulation of the recirculation bubble which causes its shrinking and enlargement over the time. Two different configurations have been identified: (i) a high-energy mode with larger ...

Parallel direct Poisson solver for discretisations with one Fourier diagonalisable direction

In the context of time-accurate numerical simulation of incompressible flows, a Poisson equation needs to be solved at least once per time-step to project the velocity field onto a divergence-free space. Due to the non-local nature of its solution, this elliptic system is one of the most time consuming and difficult to parallelise parts of the code. In this paper, a parallel direct Poisson solver restricted to problems with one uniform periodic direction is presented. It is a combination of a direct Schur-complement based decomposition (DSD) and a Fourier diagonalisation. The latter decomposes the original system into a set of mutually independent 2D systems which are solved by means of the DSD algorithm. Since no restrictions are imposed in the non-periodic directions, the overall algorithm is well-suited for solving problems discretised on extruded 2D unstructured meshes. The load balancing between parallel processes and the parallelisation strategy are also presented and discussed. The scalability of the solver is successfully tested using up to 8192 CPU cores for meshes with up to 10^9 grid points. Finally, the performance of the DSD algorithm as 2D solver is analysed by direct comparison with two preconditioned conjugate gradient methods. For this purpose, the turbulent flow around a circular cylinder at Reynolds numbers 3900 and 10,000 are used as problem models.

UNSTEADY FORCES ON A CIRCULAR CYLINDER AT CRITICAL REYNOLDS NUMBERS

It is well known that the flow past a circular cylinder at critical Reynolds number combines flow separation, turbulence transition, reattachment of the flow, and further turbulent separation of the boundary layer. The transition to turbulence in the boundary layer causes the delaying of the separation point and an important reduction of the drag force on the cylinder surface known as the drag crisis. In the present work, large-eddy simulations of the flow past a cylinder at Reynolds numbers in the range 2.5 × 105-6.5 × 105 are performed. It is shown how the pressure distribution changes as the Reynolds number increases in an asymmetric manner, occurring first on one side of the cylinder and then on the other side to complete the drop in the drag up to 0.23 at Re = 6.5 × 105. These variations in the pressure profile are accompanied by the presence of a small recirculation bubble, observed as a small plateau in the pressure, and located around ϕ = 105∘ (measured from the stagnation point). This small recir...

Parallel algorithms for Sn transport sweeps on unstructured meshes

The Boltzmann Transport Equation is solved on unstructured meshes using the Discrete Ordinates Method. The flux for each ordinate is swept across the computational grid, within a source iteration loop that accounts for the coupling between the different ordinates. In this paper, a spatial domain decomposition strategy is used to divide the work among the available CPUs. The sequential nature of the sweep process makes the parallelization of the overall algorithm the most challenging aspect. Several parallel sweep algorithms, which represent different options of interleaving communications and calculations in the solution process, are analysed. The option of grouping messages by means of buffering is also considered. One of the heuristics proposed consistently stands out as the best option in all the situations analyzed, which include different geometries and different sizes of the ordinate set. With this algorithm, good scalability results have been achieved regarding both weak and strong speedup tests with up to 2560 CPUs.

Parallel load balancing strategy for Volume-of-Fluid methods on 3-D unstructured meshes

Volume-of-Fluid (VOF) is one of the methods of choice to reproduce the interface motion in the simulation of multi-fluid flows. One of its main strengths is its accuracy in capturing sharp interface geometries, although requiring for it a number of geometric calculations. Under these circumstances, achieving parallel performance on current supercomputers is a must. The main obstacle for the parallelization is that the computing costs are concentrated only in the discrete elements that lie on the interface between fluids. Consequently, if the interface is not homogeneously distributed throughout the domain, standard domain decomposition (DD) strategies lead to imbalanced workload distributions. In this paper, we present a new parallelization strategy for general unstructured VOF solvers, based on a dynamic load balancing process complementary to the underlying DD. Its parallel efficiency has been analyzed and compared to the DD one using up to 1024 CPU-cores on an Intel SandyBridge based supercomputer. The results obtained on the solution of several artificially generated test cases show a speedup of up to ~12i? with respect to the standard DD, depending on the interface size, the initial distribution and the number of parallel processes engaged. Moreover, the new parallelization strategy presented is of general purpose, therefore, it could be used to parallelize any VOF solver without requiring changes on the coupled flow solver. Finally, note that although designed for the VOF method, our approach could be easily adapted to other interface-capturing methods, such as the Level-Set, which may present similar workload imbalances. A new parallelization strategy for Volume-of-Fluid methods is presented.Based on a load balancing process complementary to the domain decomposition.Suitable for Cartesian and unstructured 3-D meshes.Speedup up to 12i? with respect to the domain decomposition approach.Easily adaptable to other interface-capturing methods.

Low-frequency variations in the wake of a circular cylinder at Re = 3900

Flow around cylindrical structures is of relevance for many practical applications. Knowledge of flow-related unsteady loading of such structures is crucial for hydro - and aerodynamic control and design. In order to obtain a deeper knowledge of this kind of flow, a DNS have been performed at ReD = 3900 (ReD = UrefD/ν). The instantaneous velocity signals of probes located in the separated shear-layer and in the vortex formation region exhibit the presence of low-frequency variations. The statistical analysis of these signals suggest that low-frequency variations in the vortex formation length, suction base pressure and intermittencies in the shear layer are closely related. It is shown that these variations are the responsible of the large scattering of data obtained in different experimental and numerical results, as well as the U-shape and V-shape stream-wise velocity profiles observed in the very near wake of the cylinder.

Optimising the Termofluids CFD code for petascale simulations

ABSTRACT This paper presents some recent efforts carried out on the expansion of the scalability of TermoFluids multi-physics Computational Fluid Dynamics (CFD) code, aiming to achieve petascale capacity for a single simulation. We describe different aspects that we have improved in our code in order to efficiently run it on 131,072 CPU-cores. This work has been developed using the BlueGene/Q Mira supercomputer of the Argonne Leadership Computing Facility, where we have obtained feedback at the targeted scale. In summary, this is a practical paper showing our experience at reaching the petascale paradigm for a single simulation with TermoFluids.

Portable implementation model for CFD simulations. Application to hybrid CPU/GPU supercomputers

ABSTRACT Nowadays, high performance computing (HPC) systems experience a disruptive moment with a variety of novel architectures and frameworks, without any clarity of which one is going to prevail. In this context, the portability of codes across different architectures is of major importance. This paper presents a portable implementation model based on an algebraic operational approach for direct numerical simulation (DNS) and large eddy simulation (LES) of incompressible turbulent flows using unstructured hybrid meshes. The strategy proposed consists in representing the whole time-integration algorithm using only three basic algebraic operations: sparse matrix–vector product, a linear combination of vectors and dot product. The main idea is based on decomposing the nonlinear operators into a concatenation of two SpMV operations. This provides high modularity and portability. An exhaustive analysis of the proposed implementation for hybrid CPU/GPU supercomputers has been conducted with tests using up to 128 GPUs. The main objective consists in understanding the challenges of implementing CFD codes on new architectures.

Memory Aware Poisson Solver for Peta-Scale Simulations with one FFT Diagonalizable Direction

Problems with some sort of divergence constraint are found in many disciplines: computational fluid dynamics, linear elasticity and electrostatics are examples thereof. Such a constraint leads to a Poisson equation which usually is one of the most computationally intensive parts of scientific simulation codes. In this work, we present a memory aware Poisson solver for problems with one Fourier diagonalizable direction. This diagonalization decomposes the original 3D system into a set of independent 2D subsystems. The proposed algorithm focuses on optimizing the memory allocations and transactions by taking into account redundancies on such 2D subsystems. Moreover, we also take advantage of the uniformity of the solver through the periodic direction for its vectorization. Additionally, our novel approach automatically optimizes the choice of the preconditioner used for the solution of each frequency subsystem and dynamically balances its parallel distribution. Altogether constitutes a highly efficient and robust HPC Poisson solver that has been successfully attested up to 16384 CPU-cores.Problems with some sort of divergence constraint are found in many disciplines: computational fluid dynamics, linear elasticity and electrostatics are examples thereof. Such a constraint leads to a Poisson equation which usually is one of the most computationally intensive parts of scientific simulation codes. In this work, we present a memory aware Poisson solver for problems with one Fourier diagonalizable direction. This diagonalization decomposes the original 3D system into a set of independent 2D subsystems. The proposed algorithm focuses on optimizing the memory allocations and transactions by taking into account redundancies on such 2D subsystems. Moreover, we also take advantage of the uniformity of the solver through the periodic direction for its vectorization. Additionally, our novel approach automatically optimizes the choice of the preconditioner used for the solution of each frequency subsystem and dynamically balances its parallel distribution. Altogether constitutes a highly efficient and robust HPC Poisson solver that has been successfully attested up to 16384 CPU-cores.

Numerical simulation of roughness effects on the flow past a circular cylinder

In the present work large eddy simulations of the flow past a rough cylinder are performed at a Reynolds number of Re = 4.2 × 105 and an equivalent sand-grain surface roughness height ks = 0.02D. In order to determine the effects of the surface roughness on the boundary layer transition and as a consequence on the wake topology, results are compared to those of the smooth cylinder. It is shown that surface roughness triggers the transition to turbulence in the boundary layer, thus leading to an early separation caused by the increased drag and momentum deficit. Thus, the drag coefficient increases up to CD ≈ 1.122 (if compared to the smooth cylinder it should be about CD ≈ 0.3 — 0.5). The wake topology also changes and resembles more the subcritical wake observed for the smooth cylinder at lower Reynolds numbers than the expected critical wake at this Reynolds number.