## Bicycle analyses

About the development of the FEM module/workbench.

Moderator: bernd

JeanPierreLumiere
Posts: 14
Joined: Fri Mar 29, 2019 4:30 am

### Bicycle analyses

Hello everyone,

I make a final work as a bicycle mechanics student.

I build a long recumbent bicycle. The shop in which I make my apprenticeship make such bicycles since the 80's.

The aim is to integrate a new gearbox to the frame. In order to do this, I already used the wonderful FREECAD to design the "Bridge" (Link between GearBox and Frame).

I would like to:
• First, run a Fem analysis on the original frame (at the bottom);
• second, I would like to compare the results with the other frame, with the Bridge;
• Finally, I would like to make several tests for several designs of this Bridge.
I have high hopes..

FEMFame.png (53.46 KiB) Viewed 195 times

Method

I've made 3d meshes on the Frame with hollow tubes (that was without the crank), i obtained negative Jakobians. is it that bad (what is it anyway)? can we see for which region of the mesh the Jakobian is negativ? would I profit from this parameter (or was it implemented at all ?)

I've read that the use of 3d elements is not good in this case, as the lengths are far bigger as the widths.
We can then use shell elements and overlook the shearing forces.

As Bernd suggested I've read the study from HarryvL about Tubular connection and got a few questions:
• I've read somewhere that the information of the width of the material is embedded in the shell element. Is it the case with this study, and where?
• what kind of mesh is applied to the model, 3D? because the model consists already of shell elements (through the downgrade of the 3d model)
• Should I use a similar method to do my work?

2 Objects

There are 2 objects : the frame and the crank.

I thought I would make the crank rigid (with a very hight young's modulus), because I want to test only the frame. The frame consists of 0.9mm dick 25CrMo4 tubes. at some place dicker.

So I guess I have to use this method to combine the 2 objects

What I tried

I tried the "Boolean fragment + Compsolid option + Compoundfilter" on the 2 objects. here's what I got :
Capture du 2020-01-04 10-31-13.png (16.44 KiB) Viewed 195 times

I tried to run a "from shape" and 3d mesh with Gmsh and it didn't work ( Unexpected error when creating mesh: <class 'RuntimeError'> )

Here's the result with the 2d mesh from gmsh (part of the tubes are hollow, other not)

Capture du 2020-01-04 10-36-46.png (12.33 KiB) Viewed 195 times
Capture du 2020-01-04 10-37-05.png (13.54 KiB) Viewed 195 times

Do you have any idea, Is the shape too complicated? Is it possible? Any books to read? I want to learn.

Unfortunately I can't share the original file.

OS: Ubuntu 18.04.3 LTS (GNOME/gnome)
Word size of OS: 64-bit
Version: 0.19.19107 (Git) AppImage
Build type: Release
Branch: master
Python version: 3.7.6
Qt version: 5.12.5
Coin version: 4.0.0a
OCC version: 7.3.0
Last edited by JeanPierreLumiere on Sat Jan 04, 2020 2:43 pm, edited 1 time in total.
toralf
Posts: 44
Joined: Fri May 03, 2019 3:54 am

### Re: Bicycle analyses

I’m not familiar with FEM in FreeCAD, but I had my fair share of FEM with LS-DYNA, ABAQUS and Optris.

For tubes and sheet metal shells are ok. For the reason you mentioned. In general, the mesh should be of uniform size. Meaning all sides of the elements are ideally equally long. For regular shapes like your tubes I would recommend to use quad elements. Maybe 6 to eight around the tube.
I would to recommend to use a mesh generator for the whole model. Better build the mesh manually or use the generator on simpler shapes and combine them manually.

Also remove small features that are ore no relevance to the results you are interested. E.g. holes and small radii.

For the gearbox and crank I assume you’ll have to use 3D elements. Be careful with the element size. The smallest element length will impact your calculation time. So start of croase to make the model work and then build a more detailed mesh when the rest is fine tuned. Then you do not wait for ages to find that some boundaries are wrong.

In your model several elements seem to intersect each other. That should not be the case.

### Who is online

Users browsing this forum: ebrahim raeyat and 1 guest