We've dabbled with CFD but as a relatively new team haven't yet had the opportunity to build a new car from scratch based on CFD findings; however we did use it to arrive at decisions about certain components on one of our cars, so I'd have to say that it probably is worth using.

We use Autodesk Algor which is available as a free download for students whose school has bought the full Autodesk Inventor licence, we tried Openfoam which as the name suggests is open source (i.e. free) but didn't have a great deal of success; although I believe some teams make extensive use of it.

You need to be aware that none of the software is simple to use or install, they're basically "solvers" for use on a wide range of scenarios with CFD being just one of them, so be prepared to spent many hours or even days finding the correct settings. Others may disagree but I'd also say the results should only be used for comparison rather than taken as absolute truth, by which I mean if you run one version of the car and the software says the Cd is 0.3 and then a modified model of the car returns a Cd of 0.25 then you can be fairly confident you're got an improved design; however the actual Cd is likely to be somewhat different.

One of my students did a project based on Greenpower and used the built in CFD in Solid Works. After initial problems setting it up (see previous posts on here!) it was pretty good. One big problem is the time it takes to generate the animations and the size of the processor and memory you need. An advantage is you can design in Solid Works and transfer the design straigt into the CFD part with no effort. We never made the car in the end so I cannot say if it works scaled up but my guess is that it will work reasonably well. ( the finished car ended up looking like Zebedee!)

We have a couple of quad core linux computers we can run openfoam on. I must say we have not yet designed a car using this but have used it to try to mprove both Brian and Zebedee. A bit frustrating because everything we've tried on zebedee makes it worse..... It takes typically a couple of days to get a decent answer out of openfoam on reasonable quad core PC. (that's after several months trying to make it work, during the course of which the PC is attacked with an axe several times....)

We have used the OpenFOAM open source and free software for a few years now. There is some info on our usage of it at: http://www.vwt.beamweb.co.uk/
Certainly using CFD software is worth it if you have the team members who are interested and willing to spend the time and effort on it. In the case of OpenFOAM it does take significant effort to get the software working in the manner needed and a reasonable amount of Linux based computing experience and resource (8 or more processors is best).
With all CFD software it requires knowledge to use it well. We are only basic beginners but have got useful results. As Farnorth says we find it best for comparison not getting an absolute Cd values and it does produce pretty pictures :)

Openfoam is now available as a download for Windows with a user interface that looks as though it would be less intimidating that the console used in Linux. I've no doubt it will run significantly slower in Windows than Linux, but if you have a spare high end pc doing nothing, and no interest in the football, then learning to use it may be a good way of spending the next month!

I don't think the Microsoft Windows OpenFOAM port has a GUI interface ?
I think you use the command line/text files under windows just as with Linux (although with the really poor MSWindows command line interface).
For display of the results you can use Paraview, just as in Linux which is a nice GUI application to draw you those pretty pictures :)

We set up OpenFOAM on a small Linux cluster with a web front end making it really easy to use. This allows the younger and older team members to use it from whatever PC they wanted (Linux/MSWin) at school or at home across the Internet. I have been looking for some resource/funding so that I could open this up to other Greenpower teams. No avail so far though.

Terry, It uses something called "Openfoam Designer" as the front end, it has limited functionality but you can just about set up a simple case without having to use the command line window. And as you say Paraview will draw all the nice wavey, coloured lines!

I looked at openfoam a while ago, but I don't have the IT knowledge (or more importantly time) to get it up and running. I'm sure once it's set up it'd be useful.

Our school runs autodesk products so I will enquire about Algor. Yet another package to learn though...

I use solidworks flow simulation (not the really basic built in one, but the one they supply for F1 in schools teams) and like everyone else I get relative results, not absolute. That'll do me though! The streamlines are as important as you can see issues like where the flow is separating, so you know you need to change the angle/curve/whatever to try and keep the flow attached.

My question is: what settings do people use? I'm using the following:
Half-car model
Working volume from 3m in front of the front axle to 12m behind it, 1m wide and 1.5m tall - gives a blockage ratio of approx 10%
18m/s inlet air speed (~40mph)
Ambient pressure (1atm) at the exhaust
Mesh detail step 7 of 8

The big issue I think I have with solidworks is that the mesh is cubic, and is constant size regardless of location. So when the car has detailing it seems to use approx 75x75x75mm cells, and out 1m above the car it still uses the same size cells. When we did CFD at uni the meshing software would use whatever shapes it wanted (virtually never cubes) and it would shrink them near obstructions and enlarge them in the open areas.

What does Algor or openfoam do? Do you have to mesh using a separate program? Do you get finer control over the mesh? Does the mesh significantly change your results?

Sorry to recycle an old post, but thought it would be useful to keep everything together!

Openfoam comes witha meshing utility called "snappyHexMesh" which is very controllable and very stable, but no driving instructions. As a matter of course it puts finer cells near the model surface and you can set up a finer cell zone in the model's wake. I can only get the first 2 stages to work, but the resulting model looks pretty good ;^)

I can get pretty pictures: https://skydrive.live.com/redir.aspx?cid=65466c161fa1d896&resid=65466C161FA1D896!158&parid=65466C161FA1D896!147&authkey=!AO4Wo0hUr1qKhAg

I hope people who haven't used CFD can see what I meant about the half-car model. I know enough to interpret the results but it's taking 30 minutes a run so I should really set a load of variants up and leave them to run overnight/while I'm at work.

I've downloaded a windows version of openfoam, it's complex beyond belief! It says to download stl files into a case directiory, then point snappy at the directory and sit back and wait. Do you use moving boundaries? What size working volume? Too many variables!

Yes front, back, sides and top of the modelling "box" have airflow at (say)30mph in the x direction. Size of box & where you put the car in it entirely down to you - I always left plenty of room behind the car in particular....
Use of a "symmetry plane" & modelling half a car makes simulation 2x as quick but in reality you get a lot of transient side to side air movement (particularly in the turbulent air behind the car) so you probably can'r believe the results (as much.....).
BTW if you read the "bloodhound" aero stuff, they say that thrust SSC was modelled as half a car, later they found that the asymmetric back wheels added a TONNE of sideways force at the back making it almost impossible to control.

OpenFoam is complex its is a generic CFD set of tools. It has even been used for financial modelling ...
There are many possible settings, each can yield quite different results ....
From our basic knowledge/experience, the quality of meshing of the design is very important. You will note that
snappyHexMesh has many quality checks.
We use a VWT size of 8x4x4m with the car in the middle and just off the floor, and as BobC, with sides/top/floor moving at 13 m/s. Quite big to reduce effects on the sides/top. Takes a bit longer to run though.
Peter, If you send/whisper me you email address I can send you the basic config files we used to use before we implemented our easy to use web interface as a starting point. Treat all CFD results with a big pinch of salt though ...

Terry, that'd be incredibly useful. Unfortunately whispering isn't available any more and the message function is broken. Old fashioned then: peterfagg ta ohtmial dot cmo. Couple of spelling mistakes to throw the scambots off - it's quite a major email supplier if you're no good at anagrams! Admin, any chance of bringing back whispering?

I've tasked the team with sweet talking the computer techs to link all of the brand new engineering computers into a distributed network so hopefully we'll have a version of what you've got. I believe they're dual core so that should be 64 cores all working together.

I'm surprised you use such a short working volume and such a big cross section. It seems the opposite of the real wind tunnels I've worked with - most used a blockage ratio of 6% ideally, 10% in extremes. (Blockage ratio being the obstruction cross section compared to tunnel cross section) I realise real tunnels have a cost consideration - bigger costs more! As long as the cross section is a sensible size and kept constant the results will be useful. I use a long tunnel with the car near the entrance as the wake patterns take a while to develop - some trailing vortex streamlines haven't even completed a half rotation 10m behind the car with the smoother body shapes I've tested.

How come you put the car just off the floor? It's done in real wind tunnels to negate the effects of the thickness of the boundary layer that has built up by the test piece, typically as the floor doesn't move. It shouldn't be necessary in a simulation with a moving floor?

Will find the info and send it. Note we use Linux not Msw.
No idea if our VWT volume is good (note it is X.Y.Z (8m.4m.4m) with X being the wind direction), any feedback would be welcome. We have just set this up based on a bit of Internet research and playing.
Just off the floor is 1cm at the wheels just in case there are issues with the simulation between the moving floor and the wheels. We are close to the floor as we are interested in ground effects.

PeterF, we'll try and get the whispering feature back... thanks for pointing out its absence.

Admin