chassis FEA + ANSYS

[Tudor Miron]

Hi All,
I'm new to this forum. First I want to say that this is great forum and great knowledge base!
It really seems that you guys (despite competing in FSAE) are not infected by spy mania that usually surrounds racing not to mention big manufacturers.
I'm sorry if things were asked/answered before.

I'm trying to design chassis. I want to learn analyze it in ANSYS. So this question is for those of you (Rob Davies may be?) who have experience with this prog and those who have learned that FEA black art.

I have some questions (sorry if they sound stupid).

1 )What is sufficient mesh size for such problem? For example in CFD there are so called "best practices – for different problems – general mesh size, close to wall refinement etc. Not to mention requirements asked by chosen turbulence model y+ etc. So what is general "proper"� approach here?

2) How you do if you need to simulate square tubing?

3) Do you import geom. from your CAD (if so what format) or model it in ASYS?

4) Constraining the model – what is the proper way so to not overconstrain it and so overpredict stiffness?

5) I'd love to be able to plot than compare graphs - like displacement of predefined points along the chassis length so I could see where it has higher rate so I can pin point weak areas – how can I do it?

6) I have a feeling that after general design optimization (adding triangulation here and there, changing tube D. Etc.) I have to setup the model as an assembly to simulate reality a bit closer – I mean applying loads to hubs and than find what load is acting at bellcrank pivot, shock/spring attachment point and what actually happens to chassis under this scenario. I saw a screenshot posted by Frank in "SolidEdge or Solid works" thread – that's exactly what I'd love to learn how to do! PLEASE HELP!

7) How you go when there's stressed skins in the chassis? Metal or composite panel bonded like many teams do?


I have to say that I want to use FEA more as qualitative tool rather than quantitative. Thought if done right it have to be very close. Hell! it's possible to get within 2-5% with CFD if you know what you're doing and have enough PC power!

Thank you
Ted

[Will Steed]

I have been using Ansys for the last 4 years. I modeled our frame using beam elements. You can associate each beam element with a section property. The mesh density should be somewhat fine. For modeling a chassis with beam elements just set a global element size and mesh. There are section properties already built into Ansys so it should be straight forward. Square section are there too. When you become fluent in Ansys it might be useful to write an input deck that creates the geometry, meshes it, and solves. Writing an input deck like this will help keep you from losing your database.

[Denny Trimble]

Will is correct, beam elements are the way to go. If you're not familiar with them, you only need to send the centerlines of your tubes into ANSYS (or set up keypoints and draw lines in ANSYS). Then you assign tube sections and material properties.

[Tudor Miron]

Thank you Will and Denny,
Sorry for my dumbness but I still have questions. First I have to say that FEA math is totally beyond me. I'm not in school and there's no real possibility to start learning it now. Also English is not my native language (by far) so I kind of struggling when reading software manuals etc. So again I'm sorry if my questions will sound dumb.

Will wrote: "The mesh density should be somewhat fine" – can you please be more specific here?


Modeling 1"x1"x0.049 L= 1000mm rectangular tube ANSI 1020.

When reading ANSYS help about say BEAM4 element there says that I have to input the following:

"The element is defined by two or three nodes, the cross-sectional area, two area moments of inertia (IZZ and IYY), two thicknesses (TKY and TKZ), an angle of orientation (θ) about the element x-axis, the torsional moment of inertia (IXX), and the material properties. If IXX is not specified or is equal to 0.0, it is assumed equal to the polar moment of inertia (IYY + IZZ). IXX should be positive and is usually less than the polar moment of inertia. The element torsional stiffness decreases with decreasing values of IXX. An added mass per unit length may be input with the ADDMAS value."�

First question arising - are all above listed (torsional inertia etc.)inputs necessary?!?!?

1. Cross sectional area – I didn't figure yet where I can find it in ANISYS.
Will wrote:
"There are section properties already built into Ansys so it should be straight forward. Square section are there too."� - Would you please point me where exactly are they build in? What am I missing?


I can get section properties from SolidWorks :

Area = 116.16 square millimeters
Moments of inertia, of an area, at the centroid: ( millimeters ^ 4 )
Lxx = 22177.49 Lxy = 0.00 Lxz = 0.00
Lyx = 0.00 Lyy = 11088.75 Lyz = 0.00
Lzx = 0.00 Lzy = 0.00 Lzz = 11088.75

2.Thicknesses are obvious - 25.4mm

3.Angle of orientation...a bit confused here.

4. Torsional moment of inertia (IXX) - where to get it?

5. Added mass per unit length – where I get it?



OK now I model it in ANSYS

Created keypoints, line, element type MEAM4, Real constants as above but no torsional stiffness and no added mass per length unit.

Constrained one keypont in all DOF and applied 5000N force at another keypoint.

Got max displacement of keypoint 751mm.

Repeated same the same problem in Cosmos/W result is within 2.5%. (OK considering that node quantities where different)

So I assume that this indicates I did things properly ?

Do you use Beam4 element for such problems?

I repeated this problem using round tube 1"� x 0.049"�

Got 236mm max displacement. Did same in Cosmos/W and got same 236... but NO! I got 2360mm! Guess I messed something with units...

How you do with units – is it OK to use "mmgs"� and N?

Thank you
Ted

PS: Sorry for my English

[Tudor Miron]

I'm still looking for questions 4 to 7 from 1st post http://fsae.com/groupee_common/emoticons/icon_smile.gif

Thank you
Ted

[jonnycowboy]

1. Cross-sectional area just means the square area (ie 1"x1"=1" squared) minus the inside area (1-0.902^2)
2. IZZ and IYY are standard, and are the same since you have square tubing and not rectangular. You take teh moment of intertia of the whole tube minus the inside.
IZZ=IYY=1/12*b*h^3-1/12*b*h^3=1/12(1*1^3-.902*.902^3)=.0282in^4
3. thicknesses are usually your wall thickness, ie 0.049"
4. angle of orientation is probably either 0 or 90 degrees
5. IXX is probably the same as IZZ and IYY (or IZZ might be zero, depending on where are you axes)
6. this is the link you want for info on how not to overconstrain your model (http://locost7.info/files/chassis/Final_Thesis_B.doc)
7.by looking at the stress diagram in post-processing you can see using different colors the most stressed and least stressed elements, and reinforce those that need it.
8. importing a full solid model into ansys is a bit of a waste of time, the point-to-point model is pretty good, if you need to apply a force at the bellcrank, just make a member there and apply the force. Ansys will help you in triangulation, etc etc but it'll probably easier to test once the frame is fabricated.
9. for stressed skins, there is an area element in ANSYS to do this, you can either bond it using loctite adhesive or rivet it (both for better shear!)