Wind tunnel model of an animated cyclist
 

Hi cfd-online users!

I have a (long-term) project aiming to design a cyclist in a wind tunnel in order to estimate drag forces for different air speeds and wind angles.

I have access to a 3D scanner (Faro Freestyle 3D).

My first step would be to perform the analysis on a static cyclist and the next one would be to perform the analysis on a pedaling/moving cyclist.

With a very small experience in 3D softwares (my background is a PhD in Biomechanics) but a will to learn and no fear to take my time in order to get the most accurate modelization, my first question is simple: which free software would you recommend to perform my project? Is Blender good enough or are there better solutions?

Thank you for your advices!

OpenFOAM will take you through the CFD-part of your project. I'm not familar with the 3D scanner but if you can export the cyclist as a .stl file, then you can import that using snappyHexMesh. snappyHexMesh is and OpenFOAM tool for meshing your geometry. Even though I'm not the biggest fan of snappyHexMesh, it will eventually get you there (there is even an example of a motorbike).

Thank you for your answer! I just downloaded OpenFOAM. I am trying to run the "Motorbike Tutorial" now...

The scanned cyclist is a .stl file for information.

Would you be able to share the .stl file? I'm interested in seeing how many details it contains :)

Sure, check your private messages.

You are not in a easy problem with the moving mesh part since you will need finite motion. For the meshing part, check Hypermesh, it is a gem; however, it is not free and its learning curve is steep. For the flow, OpenFOAM should suit your needs. Take into account that you will need a software capable of generating boundary layer elements.

Can you share the stl file.



I created mesh without the cyclist for the purpose of demo of immersed boundary method with Wildkatze solver.



I think the set up with cyclist would be more interesting.

This is what I have for now.

Sorry for the slow answer, I haven't worked on this project for a while but I am back at it now!

Here you will find a .stl file of a cyclist obtained by 3D scan:

http://file.k-upload.com/k-upload_fr.png

There you will find what I would like to be able to do summarized in a short video:

https://www.youtube.com/watch?time_c...&v=dG-fdZDcUrk

Your problem is very complex to solve and is requiring a huge computational power to be solved properly. The main problem I see is to guarantee the grid is able to solve correctly the BL over the body while simultaneously prescribe the correct time-dependent BCs. during the cycle. This latter issue has lack of accuracy using the immersed boundary method.

Recently appeared some simulations of peloton of cyclist but on a static configuration. They used a URANS formulation.
I think you will find many difficulties if you are not an expert in CFD.

Thank you for your answer FMDenaro.

What do you call BL and BCs?

How do you think they managed to do it in the last video I posted?

I am aware of the studies by Bert Blocken that you talked about. As you said, they are static but this is going to be my starting point too.

I was thinking about using the following process:

1) static drag measurement of a scanned cyclist with OpenFOAM using snappyHexMesh to create a mesh and Paraview for post-processing.
1bis) validation in wind tunnel + velodrome testing.
2) creation of a dynamic cyclist 3D model (with Blender?). I will have to see if I am able to input some kinematic data from optoelectronic measurements by Vicon and/or if I can "morph" several scans to create a full cycle accurate enough.
3) dynamic drag measurement with OpenFOAM. I found that I may have to look at "timeVaryingUniformFieldValue" in the process, but I will have to create a good dynamic mesh first. I guess this is the part where I may need more computational power and will see if I can rent/buy it.
3bis) validation with wind tunnel measurements at several wind speeds and yaw angles.

This project should be fun!

BTW after you mentioned we did do it just for fun of it.

(Tire rotation is by immersed bnd)


https://youtu.be/uM86UoGRV60

BL -Boundary layer
BC - Boundary condition


I suspect you are not aware of the complexity of the problem... resolving the BL during the full cycle of rotation by alone requires a lot of computational resource. You need to ensure your grid is very fine both over the total bodies (and bicycle) parts and the separated flow to get a correct drag. Even using LES you get a wall resolution compared to DNS. And you need to prescribe correctly the BCs. that change in time.


However, start first solving the static problem to test you can get accurate drag ...

When you want to know how the simulation from the video was done, just ask the people who did it :-) They can also tell you, where they see some room for improvement.


My guess is, that the coupling between blender/maya/3d max with the cfd world would be an interesting topic.


https://www.youtube.com/watch?v=ONtoefgkKxg


The static bycicle is a bit boring, since it is already done since years. You can use the mororcycle tutorial from OpenFOAM as a starting point.

arjun:

Thank you for the video. The flow around the bike and the rotating wheels seem nice! I am surprised by the flow around the head/torso as it looks quite static though...

FMDenaro:

Thank you for your feedback. I am aware of the complexity of the problem. I have no limit in time for this project and as I said I am not afraid to ask for help if/when I will need more computational power. Other persons made this kind of analysis which proves it is doable. The devil is in the details and I think that you would agree that a model linking two rotating wheels with two triangles (=a bicycle!) would not be that difficult to create. However I also know that modeling a cyclist with 100% accuracy including soft tissue movements, internal fluid movements, hair and so on would be far too complex! My goal is to create a model as accurate as possible and able to detect the smallest possible meaningful changes. Even a wind tunnel has limits in this aspect, especially if you put a human and his kinematic "random" variability in it. This has been evaluated in cycling for field testing too: http://www.trainingandracingwithapow...ists.html?m=1a
I will take the challenge step by step and will start with static for sure. By the way this was already done by others too, with a certain degree of accuracy: https://hal.archives-ouvertes.fr/hal-01385084/document

JBeilke:

Thank you, I will ask them!

I think that you are not fully understanding me. The problem is not in the softwares and the technology to create the geometrical model in dynamic movement. The problem is in the goal you are pursuing, the computation of the visocus drag. That requires you accurately compute the stresses over all solid parts. In a static configuration, this is challenging for creating a good computational grid with the correct grid refinement (you need to set several nodes everywhere close to the solid parts) but if you have a large computational power you can solve it using a standard LES. In the dynamic problem you have a lot of more difficulties in computing the correct drag. This is because you need to consider the rotation of the wheels along with the rotation of the legs, all this external forcing influencing the flow. While setting this movement as a known law, the BC.s changes in time and you need to ensure that the grid still resolves correctly the stresses at any angle of the moving walls. But we can discuss also about the problem of setting the correct inflow conditions for LES or the suitable SGS model to adopt as well as the data reduction from the fields at several times.

In conclusion, the issue is in the accurate CFD simulation that can provide a physically meaningful solution. Of course, if your aim is just to get a coloured movies to show, this discussion is useless.

Thank you very much for your answer and for pointing out potential difficulties in my project. This is useful and will help me to know where to put my focus in order to get a meaningful solution.

By the way I am more interested in getting accurate CdA than fancy coloured movies.

I should have written that previously: the end goal of this approach is to optimize my cycling setup without having to go in a wind tunnel nor designing new hardware parts. For example modifying the helmet shape and being able to predict the change in CdA for a given air speed.

Optimizing the design for the helmet shape does not require at all to use the full geometry of the moving cyclette and is a much more affordable problem. You can simulate only the upper body part of the body in static condition. Shape optimization of this detail is also a quite different topic.

This only works in theory. There are aero helmets which provide a very low drag as long as you can keep your head in an optimal position. But when you get tired you will change the position on the bike and your head starts hanging more and more. Now you will see a much higher drag with your aero helmet than with some very cheap equipment.


And this is just one example of things which can go wrong during an optimisation.

Yes but let me say that if you have a wrong position the body is responsible of much more drag than any helmet.
Aero helmets are used only in few specific competitions not in classical peloton competitions.