
[Sponsors] 
March 20, 2000, 16:05 
Time Taken for the Whole CFD Process

#1 
Guest
Posts: n/a

Dear CFD users,
I'm interested how much time people spend in each area of CFD, for example: Pre Processing / Solving / Post Processing. Can you give the results for a typical example as set out here; Area : Powertrain Pre Processing : 70% Solving : 20% Post Processing: 10% Also what pre / solver / post software you use. Thanks for your time. regards David 

March 20, 2000, 17:47 
Re: Time Taken for the Whole CFD Process

#2 
Guest
Posts: n/a

(1). Although this is a seemingly valid question, it is extremely difficult to answer. (2). I'll try to answer your question based on my experience. (3). For flow over a single airfoil or through a cascade 2D problem, I can use PreBFC, tgrid, and fluent, and get the problem done in 4 hours, including geometry, mesh, solution and graphic output.(you can use GAMBIT also) At this level, the UNS would be used and the point of interest is the pressure field. (4). For the same problem, at the transonic speed, with shock formation, the compressible RAMPANT must be used. In this case, the time used in the geometry and mesh generation would remain the same. But the mesh refinement through the adaptive meshing, and the slow convergence of compressible code would take several rounds of iterations, which normally require a couple of weeks to reach the stable flow field with the correct shock formation. At this level, the point of interest is still the pressure and velocity fields. (5). For the same problem, at subsonic speed, using fine mesh and low Re models, using the UNS code, it still require several rounds of iterations to obtain the proper Y+ =< 1 condition on the surface of the airfoil based on the computed results. Since it is difficult to obtain accurate viscous flow solution relative to the test data, various options will used in an attempt to get better(?) results. The mesh is likely to stay unchanged, and the numerical scheme, turbulence models will be changed to provide a series of solutions. This exercise will normally take several weeks. (6). For internal flows with separation, say a simple turbine engine cavity flow, it can easily take several months to get satisfactory results (if it exists at all). (7). So, in 2D problems such as flow over an airfoil or through a cascade, the major portion of the work is still in the solution step. (8). For an inhouse 3D design code for viscous, turbulent, compressible flow over a turbine blade, it shouldn't take more than three days, to prepare the formated input file, the overnight solution and the graphic output. The geometry input and mesh control is minimum, and the only uncertainty is the convergence of the iteration process. (9). For 3D complex geometry, say a turbine inlet scroll with a couple dozen of radial nozzles included, the time to construct the complete geometry and the mesh will normally take several weeks, unless one has already done it a few times. In this case, we are still talking about the pressure and velocity fields only. The low Re model and fine mesh for a complex 3D problem like the above example will depend on the hardware available, especially, the accurate prediction of the viscous loss is still very difficult. (10). For new users of CFD , the geometry and the meshing problem may seem to be the biggest issue, so the userfriendly codes will definitely reduce the time in the process. But since each product oriented company already has a good definition of its product geometry and designrelated problem, it is likely that the geometry and the mesh generation steps have been studied many times, and thus have produced efficient inhouse solutions. It is unlikely that every engineer will have to study the geometry and meshing from ground zero. (11). For completely new design and geometry, it is really hard to provide any information, because in any analysis, the geometry must be known in the first place. In this advanced design and analysis phase of work, normally there is a group of experienced engineers or a department to handle it. And various inhouse codes will be developed for handling of geometry, meshing, etc. (12). The current general CFD codes are more suitable for consulting engineers who are facing a lot of different types of problems. So, in this case, the onefitall concept will save the investment in software. (13). At some stage of development, the geometry handling and meshing could be the major issue, but eventually, it is the accuracy and the reliability of the solution which will determine whether the CFD approach is a viable one or not. Hope this will provide some useful information. (14). By the way, a week ago, I was asked whether a complex 3D problem with two inlets, three outlets , two chambers and several interconnecting holes, can be solved in one month or not. A sixmonth estimate was suggested by my friend which is six times longer than the project allowed. So, the estimate of the time to solution depends also on the specific person involved. (15). The only reliable approach is to establish a data base first. But if it is a new design or a onetimeonly problem, then any estimate is possible.


Thread Tools  
Display Modes  


Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Moving mesh  Niklas Wikstrom (Wikstrom)  OpenFOAM Running, Solving & CFD  122  June 15, 2014 06:20 
calculation diverge after continue to run  zhajingjing  OpenFOAM  0  April 28, 2010 04:35 
Could anybody help me see this error and give help  liugx212  OpenFOAM Running, Solving & CFD  3  January 4, 2006 19:07 
ASME CFD Symposium, Atlanta, July 2001  Chris R. Kleijn  Main CFD Forum  0  September 13, 2000 04:48 
ASME CFD Symposium, Atlanta, July 2001  Chris R. Kleijn  Main CFD Forum  0  August 21, 2000 04:49 