CFD with FreeCAD!

The Rocketry Forum

Help Support The Rocketry Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Joe Barnard posted a good overview of rocket CFD CFD is Better Than a Wind Tunnel I wonder what package he is using?

"CFD is better than a wind tunnel." No arguments, here!

The package looks unfamiliar to me. Not OpenFOAM/Paraview. OF has far better meshing.

Yeah, Joe's CFD is in the ball park, but the trend of low speed to high speed is essentially opposite to the "fancy" CFD runs. Trends are usually more important than the absolute numbers in these kind of studies.
 
Last edited:
I started looking for an OpenFOAM UI freeware a few months ago. FreeCAD with the CfdOF workbench seemed ideal. Coincidently, David Carter made a nice presentation at NARCON 2024 2 weeks ago on this very software.

I built up a simple test case of the Estes Fatboy design using the parametric Rocket workbench in FreeCAD plus a rear bulkhead to seal the fin can. Super easy.

View attachment 629880


From, there I built out a simulation with CfdOF. The seamless integration is fabulous!

There is a nice set of robust meshing and simulation options. I ran through about 20 mesh design iterations until I got something I liked. Meshing took about 1 minute on my Windows PC. 1-2 mm resolution on the rocket including boundary layers. The domain area is large enough such that blockage is less than 1%. Total cell count is about 2 million.

View attachment 629878View attachment 629879
View attachment 629876View attachment 629877

For the solver, I started out simple with steady, incompressible, turbulent RANS. I think this is a good model up to Mach 0.3. The solver took about 3 hours on 1 core. I couldn't get the MPI version working, but when I do, the solve time should drop to about 30 min on 8 cores. Here is the nice steady state convergence of forces. The drag results make sense.

View attachment 629882View attachment 629888

Some flow images. The isosurface of total pressure = 0 shows the base wake as well as some flow loss off the square edges of the fins.

View attachment 629883View attachment 629884
View attachment 629885View attachment 629886

This is a game changer for hobbyists! Granted, I have a background in CFD, so the setup and analysis was very intuitive. Still, the workbenches in FreeCAD make it very easy. If you can CAD it, you can simulate it in minutes. Make design changes and everything in the workflow updates automatically. With some development of the CFD best practices and templates, I see this replacing Barrowman, RockSim, and OpenRocket for accurate aero analysis of any shape.

More to come in another post where I use this CFD model for center of pressure analysis.
Ondsel engineering suite is a game changer as well. One of the devs for freecad created a business around freecad so he pays his employees to make code for freecad and then pipes it into new releases of freecad as well as ondsel.

I wish i could get paraview to work with my cfdof but it is not doing the cool things like yours is.

also forgot to mention, you can change the amount of cpus in the properties tab of the solver. Just set it to the amount you have and it will run it. I set mine to 12.

1718661917023.png
 
Last edited:
Just stumbled across this thread. I'm the guy who gave the talk at vNARCON, but I'm far from a CfD expert. My goal was to make people aware that it was there so that they could play with it. I'd say that was a success!

I do want to make this process easier in terms of workflow. CfDOF is a great add on for the generic case, but I want a workflow that will allow a beginner user to do some meaningful analysis of their rockets, from mesh generation to analysis. In particular, no one should have to learn paraview to see some basic results. But I know enough to be dangerous and not enough to know if I'm doing it right. So I'm looking for some help.

What I need is a "consultant" who can help with the CfD part. I can do the coding although you're welcome to help with that if you have the expertise. If you're interested in helping send me a note!

Great thread @Buckeye! Glad to see this discussion!
thats gonna be a challenge, even simscale is tough for beginners. I think the best bet is a very fast and straight forward tutorial with no tangents. That is what i like about the free version of simscale and their tutorials... right to the point. Its not snowballing you with data in ONE video. For example, sim scale has 3 major tutorials you go through when you first start. then you can try an advanced one such as flow over a wing.

I have been trying to make tutorials for cfdof the last month or so but couldnt get paraview to play nice, mostly my fault haha.

i use prepomax to do all my fem since it will mesh stl files and its extremly accurate. But prepomax can not do cfd.


below is my workflow. hopefully it helps you

Tutorial 1: Rocket workbench to create a rocket and how to install cfdof with paraview
tutorial 2: How to setup the flow domain around the rocket body either by compound parts or cutting
tutorial 3: How to setup a basic analysis in cfdof with mesh refinement and multiple processors
tutorial 4: Discussing and using more tools and options in cfdof
tutorial 5: how to generate surface LIC and streamlines in paraview
tutorial 6: how to cut and add calculations in paraview
 
"CFD is better than a wind tunnel."
A couple notes on this:

1). How does one define "better"? If its accessibility to the average person then yeah it can be better. If it's accuracy then definitely not. Proper meshing and setting of boundary conditions in CFD is critical for solutions to provide useful data and not just imaginary numbers, and usually in industry the CFD is used for design work and wind tunnel tests are performed for validation of the CFD.

2). The video Joe posted suggested every single different condition required turning the tunnel off and reconfiguring the model. In actual wind tunnel tests the models are normally attached to a movable sting which can sweep through the various angles of attack and sideslip, and many models will have actual control of any control surfaces which would allow for changing fin/elevon/rudder/aileron/flap/slat positions without leaving the tunnel. Also generally it's sufficient to just match flow similarity parameters (Mach and Reynolds number) rather than simulating different flying altitudes.

3. Wind tunnels can be extremely expensive to run, there is no doubt about that, but the amount of data that can be collected in a relatively short amount of time is enormous. When I worked in wind tunnels we used both CFD and actual test data. Usually we could get CFD to align with actual test data to within the uncertainty of our measurements, but sometimes there were weird flow features that CFD just wasn't able to adequately capture. Additionally, data that would take months on our computing clusters to model in CFD, we could collect in a single shift with the actual wind tunnel. Literally hundreds of various sweeps could be run in a day, with each containing hundreds of individual data points like forces and moments and surface pressures and temperatures, etc.

CFD is a tool and wind tunnels are tools as well. When used together (CFD for diagnostics, wind tunnel data for model validation) you can have a really good idea of what is going on but CFD will not be able to outright replace wind tunnels for the foreseeable future.
 
Last edited:
A couple notes on this:

1). How does one define "better"? If its accessibility to the average person then yeah it can be better. If it's accuracy then definitely not. Proper meshing and setting of boundary conditions in CFD is critical for solutions to provide useful data and not just imaginary numbers, and usually in industry the CFD is used for design work and wind tunnel tests are performed for validation of the CFD.

2). The video Joe posted suggested every single different condition required turning the tunnel off and reconfiguring the model. In actual wind tunnel tests the models are normally attached to a movable sting which can sweep through the various angles of attack and sideslip, and many models will have actual control of any control surfaces which would allow for changing fin/elevon/rudder/aileron/flap/slat positions without leaving the tunnel. Also generally it's sufficient to just match flow similarity parameters (Mach and Reynolds number) rather than simulating different flying altitudes.

3. Wind tunnels can be extremely expensive to run, there is no doubt about that, but the amount of data that can be collected in a relatively short amount of time is enormous. When I worked in wind tunnels we used both CFD and actual test data. Usually we could get CFD to align with actual test data to within the uncertainty of our measurements, but sometimes there were weird flow features that CFD just wasn't able to adequately capture. Additionally, data that would take months on our computing clusters to model in CFD, we could collect in a single shift with the actual wind tunnel. Literally hundreds of various sweeps could be run in a day, with each containing hundreds of individual data points like forces and moments and surface pressures and temperatures, etc.

CFD is a tool and wind tunnels are tools as well. When used together (CFD for diagnostics, wind tunnel data for model validation) you can have a really good idea of what is going on but CFD will not be able to outright replace wind tunnels for the foreseeable future.
I can help.

At my school, the wind tunnel is ghetto as hell. Max speed is 20mph and can only hold a small 6 inch model rocket. So size of the chamber and max wind speed determines how good a wind tunnel is. I mean, i wish our school had the budget for a wind tunnel that could hold a 10 foot rocket. 2 years ago i legit strapped my wildman extreme to the roof of my car and hooked a guage to it so i could at least get to 80mph on the freeway to get SOME sort of data. But i ended up coding an excel sheet to get the data i needed from others on this forum sharing their flight data with me : )

i do agree that the data from a wind tunnel is best. Even if the only value i got from my small wind tunnel at lab was the Cd, it was enough for me to calculate so many things
 
Last edited:
I can help.

At my school, the wind tunnel is ghetto as hell. Max speed is 20mph and can only hold a small 6 inch model rocket. So size of the chamber and max wind speed determines how good a wind tunnel is. I mean, i wish our school had the budget for a wind tunnel that could hold a 10 foot rocket. 2 years ago i legit strapped my wildman extreme to the roof of my car and hooked a guage to it so i could at least get to 80mph on the freeway to get SOME sort of data. But i ended up coding an excel sheet to get the data i needed from others on this forum sharing their flight data with me : )

i do agree that the data from a wind tunnel is best. Even if the only value i got from my small wind tunnel at lab was the Cd, it was enough for me to calculate so many things
There are some tricks for dealing with scaling in wind tunnels where you can use a much smaller tunnel and scale model of the rocket. For rockets that remain below Mach 0.3 it's probably sufficient to treat the drag coefficient as a constant (so changing windspeed doesn't really change the drag coefficient, only the total forces/moments, and a small low speed tunnel would give reasonable data). Since turbulence is a function of Reynolds number you can add bumps to the surface of the model at a set distance from the leading edges of fins and the tip of the nosecone to force turbulent transition early on and simulate drag for a much larger model, usually some kind of grit. This can be as simple as just gluing coarse sand in a ring around the nose and in lines parallel to the fin leading edges. Form/pressure drag doesn't really care about the scale of the model, and turbulence really only affects drag while lift and pitching moment are driven predominantly by pressure.

As far as amateur efforts are concerned though given wind tunnels tend to be a bit of a luxury item for schools/amateurs, I'd trust results from RASAero far more than CFD in most cases. Charles Rogers has wind tunnel test data validating his RASAero predictions on his website: https://rasaero.com/
From what I can tell RASAero uses a lot of the same approaches as Missile DATCOM for its aerodynamic predictions (not terribly surprising as Charles did a stint with AFRL which currently maintains DATCOM), and Missile DATCOM is pretty much the industry standard for low-fidelity/rapid iteration on rockets/missiles. Lots of idealized but validated models built in, with experimentally derived fudge factors to get closer to the real answer.
 
1). How does one define "better"?

I believe the title of Joe's video was tongue in cheek.

2). The video Joe posted suggested every single different condition required turning the tunnel off and reconfiguring the model.

Yeah, he really belabored this point, and he is simply wrong. Joe probably never attended a professional wind tunnel shift and doesn't know better.

CFD is a tool and wind tunnels are tools as well. When used together (CFD for diagnostics, wind tunnel data for model validation) you can have a really good idea of what is going on but CFD will not be able to outright replace wind tunnels for the foreseeable future.

I know one large automotive OEM who is winding off wind tunnel tests. CFD is used for all design loops and development. A wind tunnel has been decommissioned, and another tunnel is to be used mainly for production vehicle certifications and competitor benchmarking. The CFD models don't have to be validated.

Wind tunnels have their problems too, but I am sure you know that.

In the context of hobby rocketry, who actually uses wind tunnels with accurate force balances and flow diagnostics? Not many. Homemade contraptions made of cardboard and a box fan may be good for a simple demo, but nothing else. I find it funny that people on the forum will suggest wind tunnel tests for complex models or student R&D projects, as if building one or getting access to one is a trivial matter.

That's where CFD comes in and hence this thread.
 
There are some tricks for dealing with scaling in wind tunnels where you can use a much smaller tunnel and scale model of the rocket. For rockets that remain below Mach 0.3 it's probably sufficient to treat the drag coefficient as a constant (so changing windspeed doesn't really change the drag coefficient, only the total forces/moments, and a small low speed tunnel would give reasonable data). Since turbulence is a function of Reynolds number you can add bumps to the surface of the model at a set distance from the leading edges of fins and the tip of the nosecone to force turbulent transition early on and simulate drag for a much larger model, usually some kind of grit. This can be as simple as just gluing coarse sand in a ring around the nose and in lines parallel to the fin leading edges. Form/pressure drag doesn't really care about the scale of the model, and turbulence really only affects drag while lift and pitching moment are driven predominantly by pressure.

As far as amateur efforts are concerned though given wind tunnels tend to be a bit of a luxury item for schools/amateurs, I'd trust results from RASAero far more than CFD in most cases. Charles Rogers has wind tunnel test data validating his RASAero predictions on his website: https://rasaero.com/
From what I can tell RASAero uses a lot of the same approaches as Missile DATCOM for its aerodynamic predictions (not terribly surprising as Charles did a stint with AFRL which currently maintains DATCOM), and Missile DATCOM is pretty much the industry standard for low-fidelity/rapid iteration on rockets/missiles. Lots of idealized but validated models built in, with experimentally derived fudge factors to get closer to the real answer.
Well, RASAero is software and so is cfd. Both use the same equations for numerical methods and techniques. Granted the governing equations for cfd is the Navier Stokes equations, but cfd still uses the Runge Kuta, barowman in some cases, RE based off of surface friction, etc etc. I mean, openrocket is pretty much CFD without the visuals. It just calculates everything numerically and with an iterative process, but so does CFD. in fact the default iterations for cfdof in freecad is 2000 iterations.

I dont think cfdof or simscale or ansys fluent is better than open rocket and visa versa. I have never needed cfd for my rockets since i get all the same info from openrocket, BUT i LOVE visuals so to see it in action is why i love cfd.

Now when it comes to 100 meter wind turbine blades, openrocket doesnt work so i have to use FEM and CFD haha

I agree, wind tunnels are a luxury, fun and awesome but a luxery for amatures.
 
As far as amateur efforts are concerned though given wind tunnels tend to be a bit of a luxury item for schools/amateurs, I'd trust results from RASAero far more than CFD in most cases. Charles Rogers has wind tunnel test data validating his RASAero predictions on his website: https://rasaero.com/
From what I can tell RASAero uses a lot of the same approaches as Missile DATCOM for its aerodynamic predictions (not terribly surprising as Charles did a stint with AFRL which currently maintains DATCOM), and Missile DATCOM is pretty much the industry standard for low-fidelity/rapid iteration on rockets/missiles. Lots of idealized but validated models built in, with experimentally derived fudge factors to get closer to the real answer.

Funny you should ask:

https://www.rocketryforum.com/threads/supersonic-cfd-with-freecad.185549/post-2567125

My CFD of the basic finner was a first attempt and fairly limited in mesh size in order to run in a few hours on a 4-core PC. No tuning the model to match the other data. The supersonic portion of CFD and RASAero are spot on identical.

CFD is great for those designs that exceed Barrowman or Missile DATCOM limitations. For instance, RASAero won't allow fins to extend on a boattail.
 
Last edited:
As far as amateur efforts are concerned though given wind tunnels tend to be a bit of a luxury item for schools/amateurs, I'd trust results from RASAero far more than CFD in most cases. Charles Rogers has wind tunnel test data validating his RASAero predictions on his website: https://rasaero.com/
From what I can tell RASAero uses a lot of the same approaches as Missile DATCOM for its aerodynamic predictions (not terribly surprising as Charles did a stint with AFRL which currently maintains DATCOM), and Missile DATCOM is pretty much the industry standard for low-fidelity/rapid iteration on rockets/missiles. Lots of idealized but validated models built in, with experimentally derived fudge factors to get closer to the real answer.

Thanks for the compliments and trust in RASAero II. We do indeed do many comparisons with NACA and NASA wind tunnel data and free-flight model data.

For Subsonic Mach Numbers RASAero II uses USAF DATCOM Methods, Missile DATCOM Methods, and Methods from Hoerner Fluid Dynamic Drag.

For Supersonic Mach Numbers RASAero II uses Missile DATCOM Methods.

For Hypersonic Mach Numbers RASAero II uses Missile DATCOM Methods and Modified Newtonian Impact Theory.

For Transonic, for Mach Numbers from Mach 0.90 to Mach 1.05, RASAero II uses Transonic Curve Fits based on NACA and NASA wind tunnel data and free-flight model data.

I didn't do a stint with AFRL, although I've worked with many AFRL personnel over the years. I was out at the Air Force Flight Test Center, and then later NASA (the NASA Armstrong Flight Research Center).


Charles E. (Chuck) Rogers
 
Last edited:
Honestly @Chuck Rogers I have no intentions of competing with your great software. I've used it myself many times. If CfD couldn't handle cases that RASAero can't, I wouldn't bother.

I decided early in development that I would work with existing tools doing things they can't rather than compete by reinventing the wheel. Even with dedicated CfD support many modelers just aren't going to want to run long simulations. But for geeks like me blessed with a 24 core ThreadRipper it's an attractive option, even as an intellectual curiosity! :p

The Rocket Workbench has an incomplete RASAero importer, paused only because it was dependent on some other pieces. I should be back at that soon. I also want to create an exporter so things can be round tripped.
 
Funny you should ask:

https://www.rocketryforum.com/threads/supersonic-cfd-with-freecad.185549/post-2567125

My CFD of the basic finner was a first attempt and fairly limited in mesh size in order to run in a few hours on a 4-core PC. No tuning the model to match the other data. The supersonic portion of CFD and RASAero are spot on identical.

CFD is great for those designs that violate Barrowman or Missile DATCOM. For instance, RASAero won't allow fins to extend on a boattail.

I'm working on a comparison for the RASAero web site where RASAero II is compared to results from FLUENT CFD for CD, CNalpha, and CP for a Mach 6 finned projectile.

A good friend of mine is a CFD expert, and we have many stimulating engineering discussions on the relative merits of CFD versus semi-empirical methods (Missile DATCOM, RASAero II). He's a CFD expert. He knows I am a semi-emirical guy. He runs software with $25,000 per year licenses. He did agree that you would probably run a semi-empirical method (Missile DATCOM, RASAero II) for the majority of the Mach Number and Angle of Attack points, and then run high-end ($25,000 per year license) CFD for selected points to check the results, and/or to adjust the results (shift the curves at selected points).


Charles E. (Chuck) Rogers
 
Honestly @Chuck Rogers I have no intentions of competing with your great software. I've used it myself many times. If CfD couldn't handle cases that RASAero can't, I wouldn't bother.

I decided early in development that I would work with existing tools doing things they can't rather than compete by reinventing the wheel. Even with dedicated CfD support many modelers just aren't going to want to run long simulations. But for geeks like me blessed with a 24 core ThreadRipper it's an attractive option, even as an intellectual curiosity! :p

The Rocket Workbench has an incomplete RASAero importer, paused only because it was dependent on some other pieces. I should be back at that soon. I also want to create an exporter so things can be round tripped.

Most certainly you and Buckeye keep up the great work! CFD is quite interesting, and will certainly have uses in high power and amateur rocketry, and is worth pursuing.


Charles E. (Chuck) Rogers
 
Funny you should ask:

https://www.rocketryforum.com/threads/supersonic-cfd-with-freecad.185549/post-2567125

My CFD of the basic finner was a first attempt and fairly limited in mesh size in order to run in a few hours on a 4-core PC. No tuning the model to match the other data. The supersonic portion of CFD and RASAero are spot on identical.

CFD is great for those designs that violate Barrowman or Missile DATCOM. For instance, RASAero won't allow fins to extend on a boattail.
Well, i sure hope its accurate because openFOAM is constantly validated and benchmarked by nasa on projects. NASA really likes openFOAM
 
I'm working on a comparison for the RASAero web site where RASAero II is compared to results from FLUENT CFD for CD, CNalpha, and CP for a Mach 6 finned projectile.

Nice. Let us know how you calculate CP from the CFD data, as it is a very abstract concept!

May I suggest comparison studies from subsonic to, say, Mach 3, which would be the realm of 99.9% of hobbyists?
 
Nice. Let us know how you calculate CP from the CFD data, as it is a very abstract concept!

May I suggest comparison studies from subsonic to, say, Mach 3, which would be the realm of 99.9% of hobbyists?
The NACA ran a series of tests (wind tunnel and flight tests) on a vehicle called RM-10. I tried analyzing it with the Solidworks CFD tool, FloWorks (was it really 20 years ago?) and reported on it here. I didn't really know what tweaks to run to make a really accurate model (not sure FloWorks could handle it anyway). But the references are there for validation, if anyone is interested. I exported an stl model from my SWX model, you should be able to download it from my link. A pdf drawing of the model is on the page. Be interesting to see a modern analysis!

Edit: I didn't put the stl out there, just a pdf.
 
The NACA ran a series of tests (wind tunnel and flight tests) on a vehicle called RM-10. I tried analyzing it with the Solidworks CFD tool, FloWorks (was it really 20 years ago?) and reported on it here. I didn't really know what tweaks to run to make a really accurate model (not sure FloWorks could handle it anyway). But the references are there for validation, if anyone is interested. I exported an stl model from my SWX model, you should be able to download it from my link. A pdf drawing of the model is on the page. Be interesting to see a modern analysis!

Edit: I didn't put the stl out there, just a pdf.

Not bad for 20-year-old CFD! I think CfdOF with the HISA compressible model would sim this just fine.
 
For reference, here are the rates for using the University of Washington's wind tunnel. I'll use CfD thanks.

https://www.aa.washington.edu/AERL/KWT/rateguide
To be fair an 8'x12' tunnel is massive. A lot of universities have 1'x1' tunnels, some of them even supersonic or hypersonic blowdown tunnels, that 8'x12' tunnel has 96x the cross sectional area and roughly the same increase in power requirements.
 
On an amateur level CFD is nice to have, though in most cases the results you'd get from RASAero would be good enough and would run in less than a second compared to hours for CFD. The point of my original post was to clarify it isn't a drop in replacement for experimental data from something like a wind tunnel or even better a flight test. As the saying goes every model is wrong, some of them are useful.
 
I've occasionally wondered if I could build a pretty decent wind tunnel with a bunch of leaf blowers and a series of laminarising baffles.

Maybe some day when I retire.
 
I've occasionally wondered if I could build a pretty decent wind tunnel with a bunch of leaf blowers and a series of laminarising baffles.

Maybe some day when I retire.
You may get better results with an industrial fan. Higher CFM/$, especially if you can source a used one. Also a lot more rugged. You may find building useful instrumentation to be a bigger expense.

Sounds like an interesting project though. Oooh! Squirrel! (Enough projects for now :p)
 
To be fair an 8'x12' tunnel is massive. A lot of universities have 1'x1' tunnels, some of them even supersonic or hypersonic blowdown tunnels, that 8'x12' tunnel has 96x the cross sectional area and roughly the same increase in power requirements.
Yes. A smaller tunnel will reduce power costs and others, but it won't reduce manpower costs (setup, teardown, analysis, etc...)
 
I've occasionally wondered if I could build a pretty decent wind tunnel with a bunch of leaf blowers and a series of laminarising baffles.

Maybe some day when I retire.
A big industrial box fan may be sufficient for low speed (non-compressible) stuff. For blowdown supersonic tunnels people usually have a big tank (UTK's was a big repurposed propane tank) with a burst disc and a converging diverging nozzle leading into the test section.
 
<<snip>> ... a pretty decent wind tunnel with a bunch of leaf blowers and a series of laminarising baffles.

You may get better results with an industrial fan. Higher CFM/$, <<snip>> ...

Sounds like an interesting project though. Oooh! Squirrel! (Enough projects for now :p)
And not nearly as loud :)

-- kjh
 
Nice. Let us know how you calculate CP from the CFD data, as it is a very abstract concept!

May I suggest comparison studies from subsonic to, say, Mach 3, which would be the realm of 99.9% of hobbyists?

Correcting myself, the CFD analysis was for selected Mach numbers from Mach 0.6 up to Mach 4. Other analyses (Missile DATCOM and RASAero II) were run up to Mach 6.

The analysis results include CD, CNalpha, and CP.

A lot of the CFD analysis points are grouped from Mach 0.6 up to Mach 2, a very usable Mach number range for high power and amateur rocketeers.

Again, the analysis comparisons will be put up on the RASAero web site in the future.


Charles E. (Chuck) Rogers
 
Back
Top