
CFD 

Nye metoder og værktøjer 

Grundfos A/S, Poul Due Jensen Academy, Bjerringbro 

Onsdag den 6. oktober 2004, kl. 9.00  16.00 

Det er dagens mål at give deltagerne et samlet overblik over væsentlige aspekter ved anvendelse af CFD. Lige fra anvendte koder og logistikken omkring koderne så som pre og postprocessering, hardware platforme, og til kvalitetskontrol af beregningerne. 




09.00 
Registrering 




09.15 
Velkomst 

Dr. Peter Elvekjær, Forskningsdirektør, Grundfos Management 


09.30 
"Latest Advances in Grid Generation" 

Dr. Andreas Kolms, ICEM CFD Engineering, Hannover, Germany 

Creating an adequate discretization of the geometry is an integral part of CFD and FEA, and this apparently simple task has proved to be rather challenging. Advanced meshing includes model diagnosis, geometry cleanup, and feature detection in a common environment. ANSYS ICEM CFD traditionally provides structured and unstructured grid generation for more than hundred CFD codes. CAD surface models to be analyzed by the Finite Element Method are often characterized by general nsided, strongly curved boundaries, as well as by holes. Hybrid meshing, i.e. composition of different types of grids as well as automatic volume filling, requires an "outsideinside" approach. In addition, prescribed curves and points within the structure need to be taken into account. A recursive meshing scheme based on the "looping" algorithm satisfying the listed constraints will be presented. Finally, multiple automatic and semiautomatic pathways from CAD geometry to high quality meshes will be discussed. 
10.00 
Pause 




10.15 
"Power Flow, LatticeBoltzman Approach to Fluid Dynamics" 

Mr. G. Lang, EXA Power Flow, Europe, Germany 


11.00 
"Status and Progress in CFD Simulation Methods and Applications" 

Dr. Georg Scheuerer, Managing Dir. ANSYS Germany, Otterfing 

Computational Fluid Dynamics (CFD) has become an increasingly important technology for the improvement of existing and the design of new products and processes. The increased use of CFD software has been promoted by rapid advances in computer hardware and by improvements in the mathematical models and solution algorithms in the CFD software. These new developments permit much more reliable simulations of industrial flow problems in much shorter turnaround times. Recent developments in the field of numerical methods were: 1) Highresolution discretisation schemes, which permit accurate and bounded solutions. 2) Implicit coupled solution of the conservation equations describing single and multiphase flows (as contrasted with the traditional sequential solution of the equations); this technology provides increased robustness and faster convergence of the equation set. 3) Use of acceleration techniques such as algebraic multigrid methods, which lead to reduced calculation times and a linear increase of calculation times with the number of grid points. 4) Scalable parallelisation of the solution algorithms on shared and distributed memory computers, including PC clusters and network configurations. These modern techniques, combined with advanced statistical turbulence models, like the SST turbulence model, define the current stateoftheart in industrial CFD. They will be briefly reviewed in the presentation. Currently, most industrial CFD applications are for steadystate flow. However, in the future there will be an increasing demand for predicting the unsteadystate behavior of components efficiently. Examples of such unsteadystate flows are startup and operational transients, and prediction of instabilities or component hysteresis. Also, some processes like acoustics, cavitation, and many multiphase flows are inherently unsteadystate. The efficient prediction of unsteadystate phenomena poses new challenges for CFD methods. On the numerical methods side, techniques like grid and timestep adaptation, coupled with error estimation, will become more important. On the mathematical model side, the currently used statistical turbulence models have to be extended to predict unsteadystate phenomena accurately. This will require more than just running steadystate turbulence models in an unsteadystate mode (URANS). A suitable technology from an accuracy point of view for predicting unsteadystate, threedimensional flows are Large Eddy Simulations (LES). However, performing LES for industrial flows at high Reynolds numbers can be computationally prohibitive. There are, however, recent developments to blend URANS and LES models, thus exploiting the advantages of both. One of the most promising techniques in this area are the socalled ‘Scale Adaptive Simulations’ which will be discussed in the presentation. 
11.45 
"Discrete Vortex Simulation" 

Dr. Jens H. Walther, Research Associate, ETH, Zurich, Switzerland 


12.15 
Frokost 




13.00 
Rundvisning i Grundfos cirkulationspumpeproduktion 



Her ses Grundfos produktion af UPpumper. Cirkulationspumperne finder anvendelse til opvarmning, ventilation og aircondition i beboelseshuse, kontorbygninger osv. Produktionen har en kapacitet på 20.000 stk dagligt og har en høj grad af automatisering, der sikrer både kapaciteten og lønsomheden ved produktion her i landet. Grundfos er i dag verdens største producent af cirkulationspumper og dækker omkring 50% af verdensmarkedet for disse. Det er ikke tilladt at tage billeder under rundvisningen. 
14.00 
"CFD complete vehicle simulations at Volvo Car Corporation  today and in the future" 

Dr. Hans Envald, Volvo Cars, Sweden 

Volvo Car Corporation has a long tradition of using Computational Fluid Dynamics for product development, driven by the need to produce more different cars with a drastically reduced product development time. In order to achieve a controlled accuracy of the simulations the CFD processes for aerodynamics, thermo&cooling, climate comfort have been validated versus experimental data and documented in computational procedures. The shortterm strategy is to reduce the turnaround time of the simulations and to widen the range of applications of CFD to aeroacoustics and twophase flow applications such as dirt deposition. The next step is to introduce more accurate turbulence models (LES/DES) for critical applications. 
14.45 
Pause 




15.00 
"Virtuel Reality and CFD" 

Dr. Michael Grønager, Konsulent UNI·C, Lyngby 

One of the visions for the Center Contract: “MaskinAkustik” is to be able view and interact with a tree dimensional fluid dynamics field in real time, however, traditional CFD solvers usually have no or poor support for geometry changes during a transient calculation, and hence steering becomes cumbersome. On the other hand a 3D Lattice Boltzmann solver with support for geometry changes can be written quite easily, and besides for supporting interaction the Lattice Boltzmann algorithm is famous for its linear scalability when running on even a high number of CPUs. In the talk I will demonstrate our experimental OpenLB solver coupled to an interactive 3D interface enabling geometry changes while the actual computation is done in parallel on the Nordic production grid facility NorduGRID.’ 
15:45 
Afslutning 




