Chassis Stiffness Values

[mark_mcd]

Hi,

What is the best way to get chassis stiffness values using COSMOSWorks. I appreciate that there may not be a firm rule for this but I am keen to see what methods people have used.

Thanks in advance.

[Wes Burk]

Like any FEA program, fix your rear suspension point or frame nodes and apply offset forces in the front nodes or mounts, measure your deflection by probing the displacement plot.

We've only been measuring stiffness since last year, I have found that COSMOSWorks outputs torsional stiffness values 40% high. Found this by analyzing last years frame model in COSMOSWorks and comparing it to our measured physical value. Will know if this year's frame gives the same 40% difference from the physical value.

I would be interested in testing someones older frame models if they have physical testing values for them to see how it stacks up further.

[billywight]

Like any FEA program, fix your rear suspension point or frame nodes and apply offset forces in the front nodes or mounts, measure your deflection by probing the displacement plot.

If your FE model is restrained like your physical model, you shouldn't see more than a 10% difference. I would actually be suprised if it were more than 5% off (40% is huge!). The hard part is getting the constraints to match between the FE model and the physical model. Mismatching the constraints can have a huge impact on the results. For example, the last chassis I did a torsional analysis on had a stiffness of ~5500 ft*lbs/deg when restrained as you state here (improper way of constraining the chassis to represent reality on the track), and ~1100 ft*lbs/deg when properly constrained. That's a factor of 5 difference!

Think about how the loads are reacted on the racing track and where, then use this to come up with proper restraints in your FE model. It's nowhere near as simple as fixing the rear suspension and twisting the front...

[mark_mcd]

Sounds like I am on the right track. I have the frame fixed at the rear with opposite vertical forces (one up one down) going through the front suspension pickups. Sound right?

Just to clarify if I want to apply a 1kN force across 4 pickups - I just enter the 1kN and click the 4 required nodes. Is this distributing the force over the 4 or is it actually putting 1kN on each for example (thus more than the 1kN total).

Thanks for the replys.

[Wes Burk]

I know that for a solid study it will apply the entire designated force to each surface you click on. Try it with a separate force to each node then all as one and see if ti differs.

billywight out of curiosity where did you restrain your chassis in FEA in the end? We based ours off the Cornell paper put out several years ago, and we also found that changing the rear restraint nodes in several ways varied the deflection output very little.

In the end it is still a good tool for increasing torsional stiffness and comparing different chassis configurations, we are basing our final value on the physical testing.

[mark_mcd]

Well I had a go and the output was approx 4000lbs-ft/deg which seems a bit high...

The method I have used is based on that from the book 'Chassis Engineering: Chassis Design, Building and Tuning for High Performance Handling' by Herb Adams.

In that, the frame is fixed at the rear, the left front corner of the chassis is supported on a stand. A torque is applied to the front of the frame via a beam with a guy standing on the end to provide the twist. Resultant deflection is measured on the right front.

The method seems ok to me, but from the values I am getting I am perhaps not applying the force correctly?

The formula given is -

Torsional Stiffness = (Torque/57deg) x (Spread distance/deflection)
Spread being the distance from support.

Anyone else got this book?

Thanks

[SNasello]

We made a finite element model in ANSYS this year and performed a torsion test to validate it. Our finite element model predicted a stiffness of 1550 Nm/deg and we measure approximately 1100 Nm/deg. We just used the rocker posts to contstrain our model and apply loads. Additionally, our torsion test was a hub-to-hub test with the suspension installed and we measured the deflection in the chassis independently to come up with the chassis stiffness value.

[mark_mcd]

OK have refined this a little.

Used the formula at the botton of http://rileydynamics.com/m-eng%20web/sec3.htm

Bearing in mind I have analysed this without the suspension installed i.e. frame only, I got an output of 2950lbs-ft/deg. Still a bit on the high side perhaps from what I have been reading.

That is with the frame fixed at rear, support under front left of nose and downwards force of 1.5kN through front right suspension pickups. Vertical deflection measured at base of nose, front right.

[billywight]

billywight out of curiosity where did you restrain your chassis in FEA in the end?

You need to look at how loads are reacted into the chassis from the track surface... By simply fixing the suspension points, you are introducing stiffness between those nodes that is not really there. This works fine for testing the torsional stiffness between the suspension points (and is relatively easy to replicate physically), but is not at all accurate for the real loading situation.

What you need to do is constrain the chassis through the suspension arms (in the FE model). Model the a-arms, push/pull rods, rocker, and shock as truss elements (this can get tricky in the rocker area) such that all the loads are reacted into the chassis as they really are physically. (truss elements can't support a moment at their ends so the ends act as spherical bearings). Since you're doing this to test the chassis, either make these elements rigid, or very stiff so their deformations don't affect the results much.

Once this is set up, apply BC's as follows:

1=X (Lateral), 2=Y (Longitudinal), and 3=Z (Vertical)

Rear lower ball joint area, both sides = 3
Front lower ball joint area, both sides = 2
Anywhere near the back, centre of the chassis = 1 (to prevent rigid body motion)

Front lower ball joint area, +/- both sides = perscribed displacement in 3

The perscribed displacement needs to be calculated such that it will create 1 degree of rotation. After running the analysis, simply probe for SPC reaction forces and use this value to calculate the torsional stiffness per degree.

This is the proper method to determine torsional stiffness of the chassis as it is used on track, however, this is difficult to replicate with a physical test. Using the same model, you need to come up with BC's that are a compromise between what is happening on track and what you can replicate in a physical test to validate the model. Optimize the chassis design for stiffness/weight (to whatever goal you've set) using the constraints described above. When you've finished design and manufactured the chassis, physically test it as best you can and re-run the model with BC's that match your physical test to validate the model.

[Mikey Antonakakis]

So I am currently trying to figure out how to accurately constrain our chassis for FEA. I am trying to determine the compliance in our suspension during cornering at the moment, and I can't think of any way that really convinces me it's accurate. What I am trying to simulate is steady state cornering at whatever lateral acceleration will cause the inside tires to lift. I am basically looking to find the change in camber due to strain as this happens, as well as to see how the chassis is stressed in this situation. Using pretty simple mechanics, I can solve for a resultant force and moment on the spindle point of the upright (modeled as a rigid body). I can apply this force and torque without a problem. However, I cannot figure out where to put my constraints. Right now I am applying the force and moment that the left uprights would see during a right-hand turn at the tipping point. I applied my constraints and the lower a-arm mounts, as it is the relative displacement between upper and lower ball joints on the upright that I am looking for. Does this sound reasonable? Here is a picture if it helps (ignore the pushrod mounting, I'll go back and fix that once I feel confident I have properly constrained the chassis):

http://i7.photobucket.com/albums/y294/hhspunter/chassisconstraining1.png

If you can't tell from the picture, there is a force and moment applied at each spindle point. The constraints constrain all translation and rotation.