Standardized Torsional Stiffness Testing Method

[Matt Gignac]

Just curious what kind of tests other teams perform in FEA to get torsional stiffness numbers for their chassis. Do you model your suspension as a mechanism and get the torsional stiffness of the frame with loads being directed at bellcrank pivot and the shock (or rigid link replacing it), or do you apply a force at a wheel while constraining the other 3 and use the torque=(force applied)*(distance from application to center line of car) and divide by resulting angle, thus considering the suspension as a solid part of the frame?

I'm asking this because in the past our tests have yielded numbers that are much much lower than some other teams boast. Basically we would like to use the same tests as other schools to compare.

Matt Gignac
McGill Racing Team

[Matt Gignac]

Just curious what kind of tests other teams perform in FEA to get torsional stiffness numbers for their chassis. Do you model your suspension as a mechanism and get the torsional stiffness of the frame with loads being directed at bellcrank pivot and the shock (or rigid link replacing it), or do you apply a force at a wheel while constraining the other 3 and use the torque=(force applied)*(distance from application to center line of car) and divide by resulting angle, thus considering the suspension as a solid part of the frame?

I'm asking this because in the past our tests have yielded numbers that are much much lower than some other teams boast. Basically we would like to use the same tests as other schools to compare.

Matt Gignac
McGill Racing Team

[Brock Wilt]

Hey Matt,

Cornell posted a paper called "Design, Analysis and Testing of a Formula SAE Car Chassis." The paper explains several different tests, one where joint elements were used to model the whole suspenion, A-rods, pushrod, etc in ANSYS. The load is then applied to the tires and the force is transmitted through the push rod and not a suspension point that everyone has traditional used. I am modeling and analyzing the frame for our team but in order to compare to last year's design I will load the frame in the same manner and eventually model the suspension and its affect on the stiffness.

On a side note, I posted a message regarding the calculation of torsional stiffness. A force couple typically converts into a torque by the magnitude of the force on one side times the moment arm which is the distance between the two forces. I have seen some schools and you mentioned you use the distance from the force to the centerline. Is there a reason out of curiousity? This may be a reason why your values are lower, using the full moment arm would increase your stiffness by 2. What values are you gettting?

Brock Wilt
University of Cincinnati

[Matt Gignac]

We didn;t do it at all like this, we considered the force of the pushrod acting on the bellcrank, as we thought this would give us a better idea of how our chassis would work in real life (mind you i guess this isnt exactly torsional stiffness if you use the textbook definition). As I remember, last year's frame gave us something like 750 lb-ft per degree. However, we didn't have much of a problem tuning our suspension, so such a low figure surprises me somewhat. Is this report by Cornell in the SAE paper database?

[Ben Beacock]

I used a modified suspension for analysis last year, where the pushrod was fixed to the rotation point of the bellcrank. It seems to be a good approximation/simplification, especially when the exact bellcrank layout hadn't been figured out yet. I was using the the centreline distance in calculations, but I switched to the total moment arm for this year because it seems to make more sense. Last year's frame came in at a theoretical 1600Nm/deg weighing a hefty 50kg and the change of measurement gives me a preliminary 1800Nm/deg for a 28kg frame. I don't have any actual testing of the old frame (yet) but i'd have to guess its upwards of 3000Nm/deg just by how heavy and stiff it is.

Ben Beacock
Co-Manager
2004 Gryphon Racing - University of Guelph

[Angry Joe]

The only way to get accurate torsional stiffness is to simulate the force of the shock on the chassis. Basically I would say model the force up the pushrod and through the bellcrank, and through the shock if it were replaced by a solid link.

How do you do this in FEA? Hell if I know, I had to teach myself how to use that stupid software. We did do this in real life, however...

http://www.we-todd-did-racing.com/wetoddimage.wtdr/i=wMTgyMDE2NnM0MTNkZmQzMXk1NDE%3D

Notice the shocks are locked in place



Lehigh Formula SAE Alumni
Team Captain 2002-2003

www.lehigh.edu/~insae/formula (http://www.lehigh.edu/~insae/formula)

[Brock Wilt]

In response to Angry Joe, ANSYS has both joint elements and elements which take both spring and damper values to simulate the shocks and springs. As for the joint element, it allows rotation but no translation. I am not sure if the time spent to model these phenomena would produce much different results than just modeling the pushrod and bellcrank. It is possible for those who want to spend the time.

Brock

[V2 - Italy]

We modeled the chassis and the suspensions (with joints) in ANSYS, and we applied the forces to the wheels.
The dampers were locked.

The only simulation "error" is around the engine-chassis mounts where the strain energy is too high.

It is also necessary to take into account the virtual welding stiffness that is higher than in the reality.

If you can, validate the model with a real test.

Daniele

Firenze Race Team V2

http://www.firenzerace.too.it

DUCATI POWER at the UniversitÃ* di Firenze

[Angry Joe]

Ah... I was using I-DEAS. There may have been a way to do it, but I never found it. I spent too much time on FEA as it was.

I didn't do a very detailed model either (uniform tube thicknesses, for example). I wasn't concerned with accurate deflection since I knew it was beyond my abilities. I used it simply to compare the stiffness of different designs.



Lehigh Formula SAE Alumni
Team Captain 2002-2003

www.lehigh.edu/~insae/formula (http://www.lehigh.edu/~insae/formula)

[Erik C]

My .02...From my experience, at this level (undergrad, FSAE) the biggest gains can be made by using FEA as a comparative tool rather then a quantitative one. The reasons for this are that, the amount of time needed to accurately model each welded joint and tube in 3d (rather then wire frame) could be used elsewhere in the design. The most important aspect of this entire competition (if you want to win) is real world testing. The more testing you have with your chassis the more you learn and the faster you will be. Once you get to the point that your modeling is the aspect of your design which is holding you up, then I could see spending the extra time on it. FYI, if you are going to make a 3D model of your tubes, be careful to keep your warp angles small. Modeling a tube with 4-node quads will be easy, but as long as you are going this far, you might as well use 8-node (mid-side node) quads, which will yield much more accurate results in both strength and stiffness.

[Fox]

Since the start of our project here (3 years ago) it has never really been cleared what torsional chassis rigidity exactly means. Offcourse it is not really hard to calculate and all, but the most interesting thing is what is does and what you do with the results. For example what kind of figure/magnitude does a chassis designer tries achieve and why. What I can think of is that it all comes down to suspension ajustabilitty. If you change something to your suspension it should have affect on suspension instead of chassis. Therefore can you say that chassis torsional stiffness should be 10 (wild guess) higher than suspension roll stiffness?
Is this the bigger and main thought behind chassis torsional stiffness and besides that do you think 10 is an acceptable figure???? Or can you think of a more exact CALCULATION.

[Ben Beacock]

I think it would be the difference in front and rear roll stiffness that would be a cause for torsional stress on the frame.
I've seen the 10x multiplier thrown around in this case.
The other thing to consider would be the worst-case bump(4G?) on one wheel and the torsional stress it imposes as well. I'm using this condition as an input to measure the Y deflection at the wheel and determine the torsional stiffness.

Ben Beacock
Co-Manager
2004 Gryphon Racing - University of Guelph

[Erik C]

see http://fsae.com/eve/forums?a=tpc&s=763607348&f=125607348&m=6606051151&r=4946053861#4946053861