Learning to use CFD for a specific car project.

[Kambu]

I am currently involved in designing a body for a carbon fiber monocoqued one off supercar. A 2 seated LMP car for the streets. Although, it might sound like a very expensive build, we have managed to do it with a very small budget. - Currently there are just 2 people working on the car, myself and an engineer who has experience building experimental aircraft and various formula cars. So far we have completed the monocoque and tested the rigidity of it with quite decent results.

So far I have looked into OpenFOAM, somehow managed to install it but then went quite berserk on it when I got to the part in the Youtube tutorial where the guy started coding stuff. I am not a Linux guy, and have used UBUNTU for about an hour total.
So the first important question would be how long do you guys rate the learning curve for OpenFOAM to be for someone to run basic car aerodynamics simulations? Is it even doable on a decent home computer?
Are there any simpler more user friendly freeware or payware programs available that dont cost a fortune?

Here is a Concept CAD image of the car.

[Caito]

Welcome to the forum Kambu.

I'm sorry I can't help you with your specific problem, but more than one user here will surely be able to give you hand.

All I can say for the moment is that I like the looks of that car

[flynfrog]

Good, Free, Easy to use are sort of like a triangle you can really only pick two. When it comes to CFD you ussally only get one. I have never really seen any CFD that is easy to use. Flowworks for SW is OK but I wouldn't trust anything that comes out of it.

The real question I guess is what do you hope to get out of CFD.

[Greg Locock]

Given that even the best CFD packages in the hands of trained and experienced analysts with access to windtunnels to test their ideas on, fail to predict stuff on a regular basis, it seems a bit unlikely that a cheap easy alternative will exist.

But, what are you really trying to do? So long as it sticks to the road and has reasonable yaw stability nobody is really going to care about the odd 0.1 on Cd. Unfortunatley those two numbers, pitch coefficient and yaw coefficient, are among the trickier ones to get right.

So make sure you have lots of vertical area behind the cg, and a trimmable front wing and rear wing (and/or use rake)

Incidentally as a suspension engineer I'm a bit surprised to see you don't need any wheel travel.

[Kambu]

Learning CFD at first, just seems interesting to me, but overall the things I would like to check with it are: The air exiting from the front diffuser to the sides of the car - I want to get it to flow back over the upper part of the car. Also it would probably help to find the optimal cooling inlet size. Also there are some parts of the rear diffuser I am still not sure about.
How much should it rise without the flow separating, how much do I gain if it rises steeper or lower etc.

The model is just a quick concept since I am not that good at drawing, just something that would help me visualize all the ideas, and shouldnt be looked as anything final. As for wheel travel, in this model, I have just really eyeballed everything.
The prototype car will be designed in a way that allows big changes later on if something doesnt quite work as expected. Alot of mounting points for suspension or other parts etc..
Here is an image of the monocoque and the engine test fitted.

[flynfrog]

Since it looks like you are already on your way with a prototype and don't have a good model anyway. Some hand calcs and real world testing might be a better use of your time. Body parts can be made pretty quick out of foam bondo tape and spit. Tuft testing it pretty easy coast downs can tell you your drag. Since you only have two people on the project time is probably better spend building and testing rather than trying to learn CFD.


Cool car by the way care to share more details?

[Jersey Tom]

IMO, any engineer worth his salt knows that you can't just grab a tool off the shelf and dive into it. As an old timer machinist and instrument maker once told me, "Don't trust anything, check it yourself first."

If you want valid results, you don't just download OpenFOAM or Fluent or something and throw geometry at it. You have to know what the hell you're doing and at least some of the theory behind it. You have to really know the tool, the strengths and limitations. You have to have validation cases. No different than mechanical design for stresses or anything else. No YouTube tutorial session in an afternoon is going to teach this.

If all you want are pretty, curvy lines to make cool looking flow vis... you can do that in MS Paint.

[Kambu]

For me its alot easier to learn software, by learning the functions first, and then look at some visual results. It motivates me alot more than reading school books about theory, its boring to learn theory first, and then practice it. Trying to do something practical and then looking at what went wrong seems to be alot more motivational for me. Yes, even those pointless visual lines that dont represent reality are better motivators than theory books. Currently we are waiting for the custom clutch housing and some other bits to arrive. So I have some time to use for learning purposes.

As for the car, the chassis weights 120kg, is powered by a GM LS7 engine, and uses a Hewland Sequential transaxle transmission with a custom Outer shell to fit the rear suspension etc. The cockpit canopy is going to be sliding like on fighter aircraft.

The Idea is to model the concept first, and then have a local yacht factory mill out the body with their 5 axis mill.

[flyboy2160]

..... I have never really seen any CFD that is easy to use. Flowworks for SW is OK but I wouldn't trust anything that comes out of it. ....

i disagree. the program solidworks now calls flow simulation* was the only program of several tested where i work that came within 30% of the subsonic wind tunnel data in a demanding separated viscous flow analysis (it was amazingly within 1%.)

it also confirmed real aero-optical distortion data from flight testing and the flow properties behind a supesonic shock within a few %.

its adaptive meshing (now copied by alogor) makes initial meshing much less critical.

you always must do a hand calc (and validations as jersey tom suggests) to verify any engineering computer answer, but i trust its results more than any other cfd program.

*this has also been sold as nika and as efd - which is used with pro, solidworks and catia.

[flynfrog]

..... I have never really seen any CFD that is easy to use. Flowworks for SW is OK but I wouldn't trust anything that comes out of it. ....

i disagree. the program solidworks now calls flow simulation* was the only program of several tested where i work that came within 30% of the subsonic wind tunnel data in a demanding separated viscous flow analysis (it was amazingly within 1%.)

it also confirmed real aero-optical distortion data from flight testing and the flow properties behind a supesonic shock within a few %.

its adaptive meshing (now copied by alogor) makes initial meshing much less critical.

you always must do a hand calc (and validations as jersey tom suggests) to verify any engineering computer answer, but i trust its results more than any other cfd program.

*this has also been sold as nika and as efd - which is used with pro, solidworks and catia.

It must have come a long way since I last used it ( back when it was part of Cosomos.) We struggled to get anything resembling real world out of it.

That being said I did find is somewhat easier to use than the starCD suite I was using before.

[Tim.Wright]

Trying to do something practical and then looking at what went wrong seems to be alot more motivational for me.

The problem with cfd is is that you get pretty looking colour plots even when the results are garbage. Same with any type of fea

Tim

[PNSD]

Learning curve on openfoam?

You're always learning, always, always and always. I've given up on it as I have two versions of Fluent and have been using them for 3 years now. I have found CFD is something where you always learn something new whenever you use. Specifically considering the whole system, meshing is something where I have learnt something new everytime I open up ICEM, or Gambit.

The first question is, what is your background? If you are the going to be using it then it's crucial you have strong knowledge on the relevant aerodynamic theory.

Personally I would not go about learning openfoam with the aim to do this project. I would learn openfoam with the aim of... learning openfoam. In a few weeks I am going to be learning a few programming languages and hopefully get more experienced with other operating systems. I am hoping this will again open the door for me to start learning, successfully openfoam.

In answer to the main question though, if you are completely new give yourself 4-5 months before you can successfully run a simulation and be sure the results are of adequate quality.

edit - Noticed that you're looking at the air exiting for front diffuser? Well rotating wheel's and small ground clearance will prove to be a meshing nightmare for the whole model. As well as computationally expensive. If your engineer did experimental aircraft that means he must have sufficient knowledge. In that case I would avoid CFD and simply use his experience. This will not a performance machine where tenths will matter, correct? Im sure he'll have the experience and know-how to determine what will work and what won't.

[slimjim8201]

Trying to do something practical and then looking at what went wrong seems to be alot more motivational for me.

The problem with cfd is is that you get pretty looking colour plots even when the results are garbage. Same with any type of fea

Tim

Nothing a little training can't help. CFD is a tool and anyone can use it, but they'll want to have some fundamental training before being truly confident in results. Probably more so than any other type of simulation software today, the adage "crap in = crap out" is so true with a CFD tool. Fortunately, some CFD tools have largely automated many of the tasks which were once required of the user. Automatic mesh sizing, mesh adaptation, intelligent solution control, automatic selection of solvers and turbulence models.

Kambu, I'd be happy to run a simulation or two for you if you have a CAD model you can share. Check out my blog for some fun stuff I've worked on in years past. http://turboeng.blogspot.com/

[stez90]

Kambu i'm in the same situation, i'm trying to learn OpenFoam.. At the moment i can only make some meaningless colorful pictures, but i think them are a good approach to enter this world and go on learning.
I suggest you to start from "motobike tutorial" in the Openfoam directory, modifying the settings and geometry to your needs. It is a basic wind tunnel setup, a good point to start from.