Beginner's approach to CFD software?

[deltapapazulu]

I am currently an "intern" at a company for a position primarily involving drafting cell site equipment. That is going quite well. But the owner of the company (an old friend of mine) has asked me to look up computational fluid dynamics softwares and begin to tinker with one, because he is interested in it for the sake of either hobby or perhaps expanding his horizons occupation-wise.

His motive is not finally the point.

I do, in fact, have a LOT of time on my hands and would like to start messing with some CFD softwares. I downloaded trial version of Autodesk CFD and haven't the slightest idea where to begin with it. There is NOT a good internet resource base/community for jumping into this cold turkey. I am not even sure if I am supposed to have Inventor installed first or whatever (I don't at the moment).

A few years ago I installed Inventor and Solidworks and learned quite a bit of those with no problem and no need for advanced engineering education.

BUT with this CFD stuff I have kind of hit a wall.

Any suggestions for a total beginner in CFD ? I currently have trial Autodesk CFD installed on my system and am looking for some absolute beginner tutorials.

 

[cjl]

Unless it's dramatically improved recently, I have never been very impressed with any of the CAD software CFD packages (SolidWorks or Autodesk). Ansys or OpenFOAM is much better. OpenFOAM has a pretty good tutorial guide and wiki too which should help get you started. Be aware that results are very sensitive to problem setup and mesh generation in CFD, and just because you get a result out doesn't mean that the result is anywhere close to right. What kind of flows are you looking to model?

 

[deltapapazulu]

This is mostly in relation to just hobby and that of aerodynamics for cycling, e.g. handlebars, wheels, body position, etc.

 

[cjl]

Yeah, I'd definitely recommend OpenFOAM for that. I'd recommend taking a look at their beginner/tutorial guides and see if that makes sense to you.

 

[deltapapazulu]

Thank you for your reply. It is a good lead.

 

[meandcfd]

what was your impression of openfoam?

 

[Anachronist]

I'm not the original poster, but my impression of OpenFOAM is that it's a powerful software package that has broad industry support, but isn't intended for anyone new to CFD... and neither is the documentation. I echo the OP's statement "There is NOT a good internet resource base/community for jumping into this cold turkey" -- and that includes the OpenFOAM Wiki and other associated documentation.

I've spent two solid weeks using OpenFOAM with SimFlow (free version) as the GUI front end for OpenFOAM, and that helps a lot, particularly SimFlow's cookbook step-by-step tutorials. But even that is lacking any explanatory clarity in what I'm doing and what all those options and settings mean.

For example: I'm trying to do a CFD analysis of a simple rocket body with an annular indentation and protrusion. This, obviously, is an axisymmetric problem. It took me a while to figure out that I needed a wedge mesh -- but there is ZERO information on:

• Is the wedge angle adjustable? (doesn't seem to be)
• If not, what is the fixed angle being used? Do I even need to know this?
• How large should the mesh volume be compared to my rocket body?
• Is the surface area in the force monitoring supposed to be the entire cross sectional area incident on the airflow, or just the area enclosed by the wedge mesh? And why is surface area and length ignored by the tutorials?
• How do I adjust the drag coefficient to represent the whole body? (The Cd result seems consistently ridiculously low with a wedge mesh, in 6 different geometries I've simulated.)
• If the simulation doesn't converge after a few thousand iterations, what do those initial condition parameters mean, so I can set them? (I had a problem that didn't converge so I tweaked some initial conditions at random and then the solution converged, but I had no idea what I was doing, and that is disturbing.)
• Why is it when I set up the problem with some turbulence, I see no turbulence in the streamlines generated in ParaView, especially since I know the tail end of the body generates a lot of turbulence all on its own?

I could go on and on. This is a partial list of questions yet to be answered. Some others I have already found answers. I get the impression that a simple axisymmetric wind tunnel problem is too trivial for those "in the know" to bother writing about.

I was once an undergraduate physics major, over 30 years ago. Since then, my career has been mostly project management and software development rather than physics, so there's much physics I've forgotten -- and anyway we never learned fluid dynamics back then, let alone CFD. So I'm approaching this as a total newbie who has the background to understand things, if only they were explained somewhere. The documentation assumes a lot of prior knowledge that a newbie doesn't have.

 

[cjl]

I definitely agree that it isn't very beginner friendly. CFD is very sensitive to a lot of inputs, and unfortunately, as you said, there's a lot of knowledge that isn't well documented and is just common among people who work with it frequently. Initial conditions and boundary conditions are important, and turbulence models are a massive subject even just on their own. My recommendation for OpenFoam above wasn't because it was particularly easy or beginner friendly (it's not, but it's not any worse than any of the other ones in my experience) but because it's a significantly better CFD package than any of the ones built into Solidworks or similar CAD packages. As for your specific questions (with the caveat that it's been a while since I've run anything in openFOAM, so I make no promises here):

• Is the wedge angle adjustable? (doesn't seem to be)
• If not, what is the fixed angle being used? Do I even need to know this?

I've never cared about wedge angle (I also haven't done much with axisymmetric problems), but I believe it should be 5 degrees, and can be checked with checkMesh. From my quick googling, apparently it's configurable in extrudePatchDict .

• How large should the mesh volume be compared to my rocket body?

Depends a bit on your flow geometry, but I'd think 8 or 10 radii would be more than sufficient.

• Is the surface area in the force monitoring supposed to be the entire cross sectional area incident on the airflow, or just the area enclosed by the wedge mesh? And why is surface area and length ignored by the tutorials?

Should be the whole wetted area of the object, though as I said, I haven't done much axisymmetric stuff

• How do I adjust the drag coefficient to represent the whole body? (The Cd result seems consistently ridiculously low with a wedge mesh, in 6 different geometries I've simulated.)

Have you checked what it's using for a reference area? My guess is that it's not using the cross section, but rather the complete surface area (or something like that).

• If the simulation doesn't converge after a few thousand iterations, what do those initial condition parameters mean, so I can set them? (I had a problem that didn't converge so I tweaked some initial conditions at random and then the solution converged, but I had no idea what I was doing, and that is disturbing.)

Which parameters were you messing with, and which ones are you curious about?

• Why is it when I set up the problem with some turbulence, I see no turbulence in the streamlines generated in ParaView, especially since I know the tail end of the body generates a lot of turbulence all on its own?

What turbulence model are you using, and what's your reynolds number? I would expect some turbulence in the object's wake, so it's surprising if you aren't seeing that.

 

[Anachronist]

Thanks for responding. Yes, I'll agree that OpenFOAM is a better CFD engine than the ones built into CAD packages; I surmised that much from my online research before deciding to go with it. But I also knew I would need some sort of GUI front end to make sense of it.

I've never cared about wedge angle (I also haven't done much with axisymmetric problems), but I believe it should be 5 degrees, and can be checked with checkMesh. From my quick googling, apparently it's configurable in extrudePatchDict .

OK, looks like I'll have to delve into OpenFOAM proper. I've been avoiding that, instead exposing myself only to the SimFlow front end. As far as I can tell, SimFlow doesn't tell you the wedge angle or let you adjust it, but it does show you a 3D view of the mesh you're setting up.

As far as the area used for force calculations:

Should be the whole wetted area of the object, though as I said, I haven't done much axisymmetric stuff

In that case, the drag coefficient is way too small. OpenFOAM is giving me 0.001 and I know it should be somewhere between 0.1 and 0.2. It's been consistently giving me unreasonably low numbers for drag coefficients for several different geometries I've run, some with a box mesh rather than a wedge mesh.

Which parameters were you messing with, and which ones are you curious about?

Well, all of them. The initial condition U with three components seemed obviously velocity in the x,y,z directions, so I set them as my initial flow velocity. The other three, p, k, $\omega$, and v_t default to 0, 1, 1, 1 respectively, and I have fiddled with those randomly and got a solution to converge but didn't know what I was doing.

With respect to not seeing turbulence in the flow lines:

What turbulence model are you using, and what's your reynolds number? I would expect some turbulence in the object's wake, so it's surprising if you aren't seeing that.

I'm using RANS modeling k-$\omega$ SST, because that's what this SimFlow tutorial suggested using. I looked up the different turbulence models available and agreed it was appropriate for my needs.

The Reynolds number is about 2,000,000. It's a water rocket body (picture attached), made from a 2-liter Coke bottle with a nose cone, the whole thing being about 0.4 m in length traveling at 77 m/s through air at 300°K (kinematic viscosity of air is about 1.57e-5 m²/s).

I see smooth laminar flow, although now I see there are a couple of small whorls close to the bottle opening (nozzle) end. I had to position the ParaView initial flow line in exactly the right place at the nozzle to see these. But I would expect separation at the shoulder of the bottle. I attached a picture of the nozzle end flow lines.

I fully admit I don't really know what I am doing. After struggling with this for 2 weeks I was ecstatic that I finally got any result at all... I just now question the results, or more accurately, the steps I took to get them.