![]() |
$ 53,900 (US) for CFD?
1) I read through a paper on Tesla Model S CFD analysis out of China.
2) They mentioned: * a 120-core server. * kappa-epsilon turbulence model * and SIEMENS STAR-CCM+ software -------------------------------------------------------------------------------------- Online, DELL Computer is offering a 120-core, 512-Gb unit for $ 4,000 ( US ) ............................................... a 120-core, 1-Tb unit for $ 6,000 ( US ) -------------------------------------------------------------------------------------- The Kappa-Epsilon turbulence model is available, free, online. -------------------------------------------------------------------------------------- It looks like, you can't get a direct price from SIEMENS, without going through a salesperson. ( It was the same for Dassault's Exa POWERFLOW ). I did find an invoice online, from the US Nuclear Regulatory Commission, which purchased the STAR-CCM+ for $ 49,900 (US). -------------------------------------------------------------------------------------- Another fly in the ointment is, that one must have the data cloud for any specific vehicle, to input into the program. The Chinese investigators had a Tesla, which was laser-scanned to create the data cloud. ------------------------------------------------------------------------------------- A single iteration required 3-hours, 25-minutes run time. -------------------------------------------------------------------------------------- Any modification of interest would require its own additional data cloud. -------------------------------------------------------------------------------------- The computer sounds cheap. One hour at Haus Racing's moving floor tunnel would pay for the small DELL computer. -------------------------------------------------------------------------------------- With something like 36,000 EcoModder members, we might swing enough donations for the software. A cheap cup of coffee per member. |
I've got a spare Dell R740 and Tesla P100 GPU laying around. Don't mind the power consumption currently as we need the heat anyhow.
Apparently I'm a Siemens employee now. Employee pricing? |
Interesting... There is an online CFD forum.
https://www.cfd-online.com/Wiki/K-epsilon_models One can acquire a point cloud with a smart phone. A render-farm of ARM Cortex-A72 4-core compute modules would be $30 x (120/4) = $900 plus supporting hardware. Quote:
Quote:
https://en.wikipedia.org/wiki/Simcenter_STAR-CCM%2B sez there exists a Linux version. I can find mention of the COMOS site licensing mechanism, but I haven't dug deeper, it's too sunny outside. EcoModder.com Statistics Members: 31,260, Active Members: 328 |
Aerohead. I was told you only ever advise people to use your template. You are blowing my mind. It's almost like someone is trying to spread misinformation on this site. You should really advise people that ten surface pressure readings will suffice to establish whether a mod has lowered drag. That's where the real money is.
Disclaimer: This post is meant to be humorous, however, it will probably fail at it's stated goal. If you have no sense of humour please disregard. |
readings
Quote:
|
https://www.cfd-online.com/Wiki/Best...automotive_CFD
Quote:
Quote:
|
Watched a fascinating video on the development of CFD, its limitations and potential. Warning. There are depictions of a lot of transonic studies of NACA wing template 0012. NoT aNoTheR TemPlaTE!!!1!
Its from 2014 so not up to date but interesting, nonetheless. http://https://www.youtube.com/watch?v=uaH91P665PI |
His conclusion is the same point I made in another thread, Moore's Law isn't enough to bridge the mesoscale gap.
Have you looked into OpenVDB? Quote:
Blender has OpenVDB and a physics engine. I struggle with just the modeling, but all the animation rigging, rendering and video compositing needed are in there. Quote:
|
No one has mentioned OpenFOAM?
I know nothing about CFD (except that I see a huge amount of amateur stuff on the web that is obviously quite wrong eg in calculated Cd figures) but I know of an aerodynamicist using a professional(?) version of OpenFOAM. Maybe I'm just dumb, but no one has quite explained how you develop the shape of your car to model in CFD - 3D scanning? OpenFoam is here. |
Sorry. I didn't go that deep in the prior post OpenFOAM appears here in the CFD Wiki:
https://www.cfd-online.com/Wiki/What...nFOAM_.28OF.29 The pipeline from image capture to usable model involves re-topologizing. An example. Intro to Retopology in Blender - Course Teaser |
Quote:
Another way is to take a measurement to your car from a plane, say a garage ceiling, take a measurement every 5-10cm along the centreline, do the same thing 5cm to the left, and then again until the whole left side is complete. You then have multiple 2D slices of your car that you interpolate between to get the correct shape. Do it every few cm if the detail is more complex. CFD doesn't need huge supercomputers contrary to what aerohead believes, I know people who run open-foam on middle range laptops. Open-foam is free. If anyone is using price or processing power as an excuse for not doing CFD then they are ignorant to the many ways of doing it. |
Usually you can buy a 3d model of generically similar to your car, or find it for free on the various repositories.
https://duckduckgo.com/?q=3d+capture...deos&ia=videos 3D capture is free on a smart phone. This generates a triangulated point cloud. The re-topo reduces the number of faces by orders of magnitude and reorders to quads. This reduces the task calculating the physics. |
Interesting article on CFD here.
|
Quote:
Quote:
|
All times are GMT -4. The time now is 05:53 PM. |
Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2025, vBulletin Solutions Inc.
Content Relevant URLs by vBSEO 3.5.2
All content copyright EcoModder.com