Search

In the present paper a high-order Discontinuous Galerkin method is presented for the numerical simulation of the separated turbulent flow around complex geometries using unstructured grids. Bassi and Rebay extended the Discontinuous Galerkin method to solve the Navier-Stokes equations for laminar and 3D turbulent flows. Especially, an extension will be provided to calculate unsteady separated flows with a Detached Eddy Simulation, which is a hybrid method between the Unsteady Reynolds averaged Navier-Stokes approach and the Large Eddy Simulation. Some results, like flows over a flat plate and around a sphere, which could not be predicted with an Unsteady Reynolds averaged Navier-Stokes calculation, are calculated with high accuracy and compared with theory and experiments. [ABSTRACT FROM AUTHOR]

The need of high order boundary discretization in case of a high order code, e.g. a Discontinuous Gaierkin (DG) Code, has already been demonstrated in the literature. Bassi and Rebay extended the DG method to solve the Navier-Stokes equations for laminar and 3D turbulent flow. In the present paper an extension will be provided to include both three-dimensional flows and curved elements to properly represent three-dimensional bodies with curvature. Some results, like Sows around a sphere and around an Onera M6 wing, are calculated and compared with experiments. [ABSTRACT FROM AUTHOR]

In the Collaborative Research Center SFB 401 at RWTH Aachen University, the numerical aeroelastic method SOFIA for direct numerical aeroelastic simulation is being progressively developed. Numerical results obtained by applying SOFIA were compared with measured data of static and dynamic aeroelastic wind tunnel tests for an elastic swept wing in subsonic flow. [ABSTRACT FROM AUTHOR]

In this paper we present a weak fluid-structure coupling method for the aeroelastic simulation of isolated helicopter main rotors. The CFD Code FLOWer (by DLR) is coupled to the flight mechanics code HOST (by Eurocopter). HOST is used to compute the blade dynamics and the rotor trim, whereas the aerodynamic loads are determined by FLOWer. The method has been applied to two different rotors: the advanced EC145 rotor in fast forward flight and the well known Bo105 rotor in slow descent flight. [ABSTRACT FROM AUTHOR]

A numerical and experimental analysis of a Scramjet intake flow has been initiated at RWTH Aachen University as part of the Research Training Group GRK 1095: "Aerothermodynamic Design of a Scramjet Engine for a Future Space Transportation System". This report presents an overview of the ongoing work on numerical simulations using Reynolds averaged Navier-Stokes solvers. Two different geometry concepts in 2D and 3D are investigated using several turbulence models to point out the influence of the geometry on the flow behaviour. One with a double ramp/convex curve configuration, the other with a double ramp/convex corner configuration. The data obtained will be compared with results from experiments which will be started in autumn 2006. It has to be said that not all results presented here were achieved using the NEC computing cluster. For comparison several calculations were conducted on the IBM Jump system of the Jülich Research Centre and at the SUN cluster of RWTH Aachen University. At the end of this report the computational performance will be compared. [ABSTRACT FROM AUTHOR]

A series of Large Eddy Simulations (LES) of non-reacting flow in a swirl burner has been performed. The configuration consists of two unconfined co-annular jets at a Reynolds number of 81500. The flow is characterized by a Swirl number of 0.93. Two cases are studied differing with respect to the axial location of the inner pilot jet. It was observed in a companion experiment (Bender and Büchner, 2005) [1] that when the inner jet is retracted the flow oscillations are considerably amplified. The present simulations allow to understand this phenomenon: the recirculation zone and the jet interact in such a way that large scale coherent structures are generated. The resulting spectra correspond well to the experimental data. [ABSTRACT FROM AUTHOR]

The effect of arrays of single-body, delta shaped vortex generators on heat transfer and flow field has been investigated numerically using RANS (Reynolds averaged Navier-Stokes) methods. The Reynolds number based on the hydraulic diameter of the channel in which the vortex generators are positioned is fixed at 300,000. For the closure of the equation system of the flow field a full differential Reynolds stress model has been used to capture the anisotropic effects of the induced vortex structures. To gain realistic results for the heat transfer the common approach for the closure of the Reynolds-averaged energy equation using a turbulent Prandtl number has been abandoned for explicit algebraic models which deliver more realistic results for complex flows. Simultaneously to the calculations measurements have been performed on some of the geometries to validate the numerical results. [ABSTRACT FROM AUTHOR]

Within this project, we perform lattice Boltzmann simulations of spinodal decomposition and structuring effects in binary immiscible and ternary amphiphilic fluid mixtures under shear. We use a highly scalable parallel Fortran 90 code for the implementation of the lattice Boltzmann method. We demonstrate that the domain growth mechanisms in ternary amphiphilic fluid mixtures strongly depend on the amphiphile concentration. For systems under constant and oscillatory shear we analyze domain growth rates in directions parallel and perpendicular to the applied shear and find that these systems undergo structural transitions with tubular and lamellar structures appearing. [ABSTRACT FROM AUTHOR]

This report summarizes selected results of investigations of the flow and species transport in packed beds. First of all, the difficulty of segmenting image data with respect to the correct choice of the threshold value and thus the resulting porosity is discussed. Then, the accuracy of lattice Boltzmann flow simulations is compared with CFX-5 simulations. The 3-D flow data is furthermore used to show how the pressure drop is made up by shear forces and dissipation owing to elongation and deformation. For the species transport, a random walk particle tracking algorithm is used to complement the lattice Boltzmann method thus allowing a wide range of Peclet numbers. In the last part, preliminary performance results of a new 1-D list based lattice Boltzmann implementation ("sparse lattice") are summarized which soon will replace the currently used full array based code. It is shown that outstanding performance on vector as well as cache based parallel computers can be achieved with this 1-D list based "sparse lattice" code, too. Despite sophisticated optimizations for cache based microprocessors, the sustained application performance of a single NEC SX-8 CPU is about 10-20 times higher than that of any commodity CPU. For parallel calculations, this gap even grows further. [ABSTRACT FROM AUTHOR]

Self-excited (thermo-acoustic) oscillations often occur in combustion systems due to the combustion instabilities. The high pressure oscillations can lead to higher emissions and structural damage of the chamber. For the disposal of the undesirable oscillations one must clearly know the mechanism of the feedback of periodic perturbations in the combustion system. In the last years intensive experimental investigations were performed at the University of Karlsruhe to develop an analytical model for the Helmholtz resonator-type combustion system. In order to understand better the flow effects in the chamber and to localize the dissipation large-eddy simulations (LES) were carried out. In this paper the results of the LES are presented, which show good agreement with the experiments. The comparison of the LES study with the experiments sheds light on the significant role of the wall roughness in the exhaust gas pipe. [ABSTRACT FROM AUTHOR]

The paper presents results of several Large Eddy Simulations (LES) of the flow in an open channel where the channel bed is roughened with one or two layers of spheres. The roughness height k, which corresponds to the sphere diameter d is 0.23 of the channel depth. The Reynolds number Reτ, based on the average friction velocity uτ and the channel depth h (distance from the roughness tops of the spheres to the water surface) is approximately 2820. The flow configurations were selected to correspond to recently performed laboratory experiments. Mean streamwise velocities from the LES are compared with the measured data and the distributions of the calculated turbulence intensities are evaluated by comparing them with empirical relationships for flow over rough walls suggested by Nezu [1]. The occurrence of low- and high-speed streaks is examined and their spanwise spacing is quantified. Moreover, sweeps and ejections are shown to occur as well as the amalgamation process i.e. ejection of fluid into the outer layer associated with vortex growth. It is shown that these structures occur irrespective of roughness conditions, however further studies and data analysis are needed to evaluate and quantify the effect of porosity on coherent structures. [ABSTRACT FROM AUTHOR]

Large-eddy simulations (LES) of a continuous tundish flow are carried out to investigate the turbulent flow structure and vortex dynamics. The numerical computations are performed by solving the viscous conservation equations for compressible fluids. An implicit dual time stepping scheme combined with low Mach number preconditioning and a multigrid accelerating technique is developed for LES computations. The method is validated by comparing data of turbulent pipe flow at Reτ = 1280 and cylinder flow at Re = 3900 at different Mach numbers with experimental findings from the literature. The impact of jet spreading, jet impingement on the wall, and wall jets on the flow field and steel quality is investigated. The characteristics of the flow field in a one-strand tundish such as the time-dependent turbulent flow structure and vortex dynamics is analyzed and compared with experimental results. [ABSTRACT FROM AUTHOR]

The adaptive local deconvolution method (ALDM) provides a systematic framework for the implicit large-eddy simulation (ILES) of turbulent flows. Exploiting numerical truncation errors, the subgrid scale model of ALDM is implicitly contained within the discretization. An explicit computation of model terms therefore becomes unnecessary. Subject of the present paper is the efficient implementation and the application to large-scale computations of this method. We propose a modification of the numerical algorithm that allows for reducing the amount of computational operations without affecting the quality of the LES results. Computational results for isotropic turbulence and plane channel flow show that the proposed simplified adaptive local deconvolution (SALD) method performs similarly to the original ALDM and at least as well as established explicit models. [ABSTRACT FROM AUTHOR]

In the present work, computations of flow and heat transfer in an idealized cylindrical Czochralski configuration are conducted using Large Eddy Simulation (LES) with the flow solver FASTEST-3D developed at LSTM Erlangen. The results match well with DNS data from the literature. However, detailed data for analysis of turbulent quantities are not available. Therefore, DNS computations are conducted using the code LɛSOCC, employing explicit time marching. Preliminary simulations show the high efficiency of the solver on the NEC SX-8. Furthermore, from a study of the velocity profiles at the wall, the resolution requirements had to be corrected such that the computational grid will now consist of approximately 8 × 106 control volumes. The present run of the DNS took more than 540 hours of walltime on 8 processors. With the results, the LES computations will be thoroughly validated so that appropriate models and parameters can be chosen for efficient and accurate simulations of practically relevant cases. [ABSTRACT FROM AUTHOR]

The simulation of the fluid flow has been simulated on the levels l2, l3, l4. The results on level l2 and l4 are plausible and suggest convergent behavior, whereas the results on l3 does not fit in this sequence. The quantitative comparison along a line within the numerics show good agreement for all three levels. This can be explained with the fact, that the velocities in the middle of the mixer coincide in the cross section after the mixing element too. The ultimate validation with experiments show quite good agreement with respect to quality and quantity. Finally, as uniform refinement increases the complexity of the computation tremendously, effort in adaptive methods for turbulent flow has been invested already and should be applied to the static mixer soon. [ABSTRACT FROM AUTHOR]

Numerical simulations of laminar separation bubbles are carried out to investigate the so-called bubble bursting, i.e. the changeover from a short to a long bubble by means of very small variation of one governing parameter. A laminar separation bubble is formed if a laminar boundary layer separates in a region of adverse pressure gradient on a flat plate and undergoes transition, leading to a reattached turbulent boundary layer. Bubble bursting denotes a phenomenon, in which a local, in average closed region of reverse flow (the short separation bubble) suddenly becomes considerably longer as a result of only small changes in the conditions of the surrounding flow. Here, this condition is the disturbance input upstream of separation. Both, long laminar separation bubbles and bubble bursting, are not yet well understood on a fundamental level, but it is commonly accepted that the transition process plays an important role. Simulations in which transition is or is not explicitly triggered are carried out. Depending on this triggering, either a short laminar separation bubble develops or the bursting process is initiated and the flow develops towards a long-bubble state. If the flow is tripped to turbulence prior to the adverse pressure gradient, the boundary layer remains attached. Performance data on a NEC SX-8 super computer are given for two different resolutions. [ABSTRACT FROM AUTHOR]

A benchmark simulation of the jet in crossflow (JICF) configuration is presented in detail. A Direct Numerical Simulation (DNS) was carried out with a low Reynolds number equal to 275 and a jet-to-crossflow velocity ratio equal to 2.4. The benchmark is carefully selected to provide reference data concerning the following phenomena: the flowfield, the mixing process of passive scalars and three chemical reactions. The data presented concern both instantaneous and time-averaged values as well as the corresponding fluctuations. To facilitate the quantitative comparison with the data from the present work various one-dimensional plots are presented. To allow easy repetition of the present numerical benchmark, both the jet and the crossflow are supplied at laminar flow conditions. As a result of this a transition zone occurs which in turn constitutes a severe test for any simulation methodology. [ABSTRACT FROM AUTHOR]

Starting from the first experimental and analytical studies on primary breakup phenomena, many interesting results have been published in the past. It is known that in addition to the typical dimensionless groups (Reynolds and Weber number), inflow conditions can drastically influence primary breakup phenomena. Now that high computational resources are available, direct numerical simulation (DNS) has become a powerful tool in order to study primary breakup phenomena. Nevertheless only a few DNS studies concerning breakup phenomena and the influence of inflow conditions are available. This might be due to the fact that besides high demands of computational resources, sophisticated numerical models are also required in order to prescribe realistic inflow conditions and capture all length scales in the flow. This paper mainly focuses on the influence of different inflow conditions, such as the integral length scale or the fluctuation level inside the turbulent nozzle flow. For this, the breakup phenomena of water sheets at moderate Reynolds numbers injected into an quiescent air environment are considered. Since this study is performed as an numerical experiment by varying the character of the inflow velocity data, it was found that not only the mean axial velocity profile but also the integral length scale and the fluctuation level can have an influence on breakup phenomena. [ABSTRACT FROM AUTHOR]

The transition process around a short leading-edge separation bubble subjected to a sweep angle of 30° is studied in detail by means of direct numerical simulation, spatial linear stability theory and solutions of the parabolised stability equations. The combined analysis of the averaged flow field, instantaneous flow visualisations and postprocessing data as amplification curves leads to the distinction of four succeeding stages qualitatively comparable to the unswept case. It is shown that the saturation of background disturbances is the key event, after which a rapid breakdown of transitional structures occurs. The mechanism of the final breakdown of this swept scenario of fundamental resonance is best described as an "oblique K-type transition". Great care is taken to isolate and describe of typical structures within each stage as a foundation for the analysis of complex transition scenarios. [ABSTRACT FROM AUTHOR]

Numerical simulation of complex flows has always demanded the biggest com-puters both in storage capacity and in performance that were available on the market. This situation is still going on. The following paragraph repre-sents a selection of papers that were submitted as yearly demanded progress reports to the HLRS. Although most of the reports revealed a very high sci-entific standard those papers were preferably selected for publication that clearly demostrated the unalterable usage of high performance computers (HFC) for the solution of the problem. [ABSTRACT FROM AUTHOR]