Torsionator

[Captain Redbeard]

There always seem to be questions about torsion testing so I figured I would post a pic of our new setup. We took a lot of cues from Denny's famous design but put a hinge in instead of just a piece of angle metal. We got around 750 ft-lbs./deg. once we got past the non-linear stage.
http://homepage.mac.com/matthodgson/FSAE/Torsionator_2.jpg

[Matt Gignac]

Looks pretty sweet (and not too complicated)

We've thought about something similar to this setup, but the main thing we're hung up on is finding something to keep the upright diagonally opposite the loaded upright down. I have a feeling that having it under some railing, staircase, table, lift, etc would lowball the torsional stiffness.

I guess you could do a test before the torsion test that would determine the deflection due to the loading beam/support for the upright that wants to go up, and then substract that from what you get after.

Matt Gignac
McGill Racing Team

[LCheung]

What's the stiffness of the torsionator itself?

-Leon

[Denny Trimble]

Good question Leon. The deflection of the beam should be taken directly under the upright, unless extra calcs are done for the .002" that beam will deflect at that point (guessing).

If I had it to do again, I'd allow free horizontal (x and y) motion of each left-side (or right-side) hub bracket, to avoid over-constraining the frame. But the fairly flexible 1/4" plates probably aren't too bad.

I can see why you added a hinge, you'd hate to have your car fall off that pedestal. But the angle iron is so simple!

Matt, it looks like the three pedestals are bolted to the ground, preventing the lift-off you're talking about. The next question is, how stiff is the concrete slab? http://fsae.com/groupee_common/emoticons/icon_smile.gif

Also, we strapped pieces of tubing to the car, laterally oriented at various locations along the length of the car, and measured them on both sides to get an angle of twist vs. chassis location, to see which sections of the car were the best/worst performing, and to see frame stiffness vs. hub-to-hub stiffness.

[ben]

The plates actually bolted to the hubs seems quite thin, could buckling be a problem?

Ben

[Captain Redbeard]

Denny, when I was out there we were taking measurements from right under the hub, not sure why they moved it out when they went out at the time the pic was taken. The next step will be to do the measurements at different points along the frame. With this rig it is also easy to strap on a bare frame for testing although I like the idea of letting it move in the x and y direction, perhaps with slotted plates of some sort.

Ben, I actually said the same thing when the guys were building it. The concensus was that at the low angles we are testing the upright supports are mostly in compression or tension so it isn't as much of a concern.

I think our next step will be to veryify exactly what is flexing (beam, concrete bolts, etc.) and how much. In general the stifness numbers were actually better than I expected so I would like to get the measurement as acurate as possible to work on wieght reduction in the suspension for this year, and the frame for the 2007 car.

[John_Burford]

Another approach to accomplish the same test is to machine up some square wheels (1/2 inch alumimum 20"x20"). Put three square wheels on the car (2 on the rear and 1 on the front), and let the fourth wheel hang free(substitute solid rods for the sock of course). Add weights across the diagonal (to keep a wheel from lifting off the ground).

Many teams should alright have the square wheels for doing alignments, so the test can be set and run just before doing an alignment. You probably only need 100 lbs of weight to get good test data, and to not risk damaging any components.

John Burford

[Fred]

Hey Capt'n

You said :

In general the stifness numbers were actually better than I expected

Would you mind to elaborate on that please ?

I'm doing some torsion tests right now and the preleminary numbers i've got are approx 30% under what FEA predicted.

Thanks

[Z]

Captain,

I agree with Denny, you should allow the wheelhubs to move freely in X and Y, otherwise you are including the lateral stiffness of the frame and test rig in your measurements.

Deflections should be measured against something that is not being subject to changing stresses - eg. another frame (two intersecting diagonal beams) sitting unstressed under the car. Or measure deflections against the fairly constant gravitational field of the Earth http://fsae.com/groupee_common/emoticons/icon_wink.gif - use a carpenter's bubble level to measure the change from horizontal of the front and rear pairs of hubs.

Z

[LSU Dave]

This seems like a good place to ask.

How are you actually supposed to load the car in FEA? Our guy is currently holding the left rear wheel in place, restraining the front left and right rear to Z motion only, and putting a load on the front right.

It doesn't seem to be completely correct to me, but I'm sure the answer is fairly simple. Thanks in advance.

[ben]

I'd assume you need to constrain 6dof therefore x,y,z at one rear wheel, x and z at the other rear and z at one of the fronts. That should prevent any rigid body translation or rotation. Applying a displacement in z at the other front wheel should twist the chassis.

Ben

[Frank]

We got around 750 ft-lbs./deg. once we got past the non-linear stage.

thats the bit that matters

you only get 75 ft-lbs or so while driving

[Boofa]

capt, i hat e to tell u this but ur dial guage for the front right upright looks like it's mounted on your load bearing device.
or is that just some elaborate suspending device?
if it isn't ur results will be tottally whacked

Boofa
Chassis Team Leader, Swinburne Uni
Australia

[Z]

Originally posted by ben:
I'd assume you need to constrain 6dof therefore x,y,z at one rear wheel, x and z at the other rear and z at one of the fronts.
I agree that Ben's suggestion, or something similar, is the way to do it both in FEA and in real testing.


Originally posted by Frank:
you only get 75 ft-lbs or so while driving
Well... yes, while driving gently on a smoothish road.

But if cornering hard so that you lift one wheel then;
Torque = ~half-weight-of-car+driver x half-track

So maybe Realistic Torque = 1/2 x 600lbs x 2ft = ~600 ft.lbs? That's close to 1 degree of flex for Captain's car!

(Torsional loading from bumps will be greater.)

Z

[Captain Redbeard]

Most of my comments about the curves/measurements are based of the paper written by Cornell about chassis design and testing. When I mentioned the non-linear region I was talking about the first tenth of a degree or so when the jig is moving a bit on the hubs to be forced on the wheel nuts and the suspension linkages are getting rid of any slop in the mounting/bearings. Obviously it would be great if we didn't have any slop in the linkages but I think most bearings have a little play in them and our bolt holes probably aren't exactly perfect.

When I say that 750 was more than I expected it is simply because we have never tested the whole chassis and I didn't expect to be so close to our target the first time. Unfortunately we never had a chance to test just the frame of that car so I can't really do calculations to determine the suspension stifness vs. chassis stifness. We will be testing this years frame which is similar to last years so I could potential use those numbers to get a ball part figure.

Z, not sure how I could get any measurements off a bubble level, seems pretty inaccurate to me. We do have a digital level but again it is only accurate to around a tenth of a degree.

Boofa, Sorry for the confusion, the dial you are talking about is not even being used.

I will be taking many more measurements today so I can let you guys know how it goes if you would like. I'll be getting deflection along the frame and deflection of the rear hubs/rear support beam. Should be fun.

-Capt

[Denny Trimble]

Capt - You have the fixture to test the frame!

Cut some 4' lengths of 1" square tubing. Attach them to the chassis in front of the front suspension, and behind the rear suspension. The difference in twist of these two bars is your frame twist. Compare that to the twist of your torque arm, and you have frame stiffness vs. hub-to-hub stiffness, with an accurately applied torque load (through the pushrods & rockers).

Add bars in the middle, and you can make a nice graph. You'll need two dial indicators on each bar, so it helps to have a bunch of shop monkeys with nothing better to do.

[Nihal]

Wouldn't there be torsion on the chassis from the different rates of front and rear roll angles and rates?

[Captain Redbeard]

Denny, thanks for the idea, I'm more interested in testing the new frame but perhaps I can convince them to try this out. Oh yeah, we call them shop ninjas up here :-)

-Capt

[Z]

Capt,

In your above pic you are testing both your frame, and frame+suspension, stiffnesses, but you are only measuring the latter. So, as Denny said, just divide applied torque by amount of frame twist (over whatever length of frame you want to measure it) and you have that frame stiffness value.

A standard hardware store bought carpenter's bubble level, about 3 or 4 ft long, is claimed accurate to 0.5mm/m from horizontal. In practice you can read relative displacements more accurately than this. So you can quite easily read angles in steps of 0.0005 radians or 0.03 degrees. Well, apprentice carpenters can, but you guys are training to be engineers, so, errr..., maybe not??? http://fsae.com/groupee_common/emoticons/icon_smile.gif

So if your track is 1.25m (~49") you should be able to tell if one hub is 0.63mm (25 thou) higher than the other. If you load the chassis in ten steps to 600ft.lbs (0, 10,.. to 100lb on 6ft lever arm), and with your nominal chassis stiffness of 750ft.lbs/deg, then each step will give a twist angle of ~0.08 degrees, or 0.0014 radians, or 1.7mm (~1/16") at each wheel hub, which is easily measured.

I would keep the bubble level horizontal, with one end of the level on the higher wheel hub, then measure to the lower wheel hub with a dial gauge clamped to the other end of the level. You can use the one bubble level for all the measurements - ie. front and rear axles, and any chassis stations inbetween - because "horizontal" doesn't change much between front and rear of chassis.

Note that a damaged or cheap level mightn't read "horizontal" accurately (because the "bubble tube" has moved in the frame), but it can still give accurate results if you keep it facing the same way (say, keep one end of level always at the left side of the chassis). Check for damaged level by placing on rigid table, then turning end for end.

Z

PS. Every team should have at least one good bubble level anyway, for doing car setup. And "digital levels"!!??? Do you have to have everything spelled out for you! http://fsae.com/groupee_common/emoticons/icon_smile.gif

PPS. "750ft.lbs/deg" sounds pretty low to me. Any comments from other teams on this figure?

[Denny Trimble]

Until 2004 our cars were in that range of stiffness (hub-to-hub). Lately we've approximately doubled that number.

[Schumi_Jr]

Originally posted by Z:
PPS. "750ft.lbs/deg" sounds pretty low to me. Any comments from other teams on this figure?

I'd shoot for 2-3x that value with the properly constrained test fixture. We've seen ~20% decrease in stiffness (FEA, hub-to hub) when we went from constrained rotations/displacments at each wheel to the constraints discussed above.

Almost, every year I've been on this team the previous year's chassis designer tells us that if we want to make the chassis stiffer it will end up being heavier. Every year we increase our stiffness target, the specific stiffness sky-rockets and the mass goes down. Basically this is accomplished through better triangulation with very thin walled tubes.

If you have a low stiffness target you will end up with a frame that isn't adequately triangulated and probably isn't any lighter. Though it should be much easier to manufacture...

[Captain Redbeard]

Z, The digital level came with a camber measurement/adjustment kit that someone bought a couple years ago. I guess its probably not a level as much as an angle indicator. I understand the bubble level thing but I'm pretty sure I can accomplish the same thing with a bunch of dial guages. I took some extra measurements for verification today. The X direction deflection in the rear is around .04" (with 1100 FtLbs applied) so it seems that the car isn't constrained too badly in that direction. There was .008" of deflection in the Z direction at the rear left hub so I factored that out of our equations.

As far as the final hub-to-hub number, given wheel rates ~ 100 lbs/in I figure we are around the right area. 1000 would follow the common "10 times the wheel rate" argument but from reading the paper from Cornell and from the numerous discussions in these forums it seems like 7-800 has been deemed acceptable by most drivers.

I'm surprised that teams are putting alot of effort into getting 2-3x that number. At some point you must be able to cut weight to bring it back down. Are your wheel rates 2-300 lbs/in? Although....I've been looking back over our dynamics calculations from this summer and we had a desired total roll gradient around 350 ft.lbs/degee so perhaps we should aim higher. Any insights would be appreciated, I've only been at this vehicle design stuff for 5 months now...

Denny, I'm interested to know how you doubled your stiffness in one year (obviously you don't have to share if you don't want to.)

Thanks for the input guys!

-Capt

[Denny Trimble]

Our old spaceframes were not very well triangulated, especially in the engine bay. We weren't doing beam element FEA back then, we were doing shell meshes, which took 20 hours to run, and so we didn't do many iterations. We were also looking at artificial constraints on the frame, instead of modeling the control arms and pushrods and rockers and dummy shocks. Basically, we were doing everything I recommend against these days.

More recently, we use beam element FEA (30 second solve time), and at the end of the 2004 competition, I did about 30 iterations on different triangulation schemes. We ended up more than doubling our hub-to-hub stiffness through proper triangulation (I'm sure you can find pictures of our '05 frame online if you search the forum). There are many more tubes, which made it harder to manufacture and engine maintenance is less than optimal. But we gained only 2-3lbs over our previous frames which had fewer, but heavier, tubes.

Since you're required to use certain tube sizes on certain members (roll hoops, braces, side impact, front bulkhead), and they usually make up more than 50% of the frame mass, adding a few 1" x .035 (or even 5/8" x .035) tubes won't add much weight, but they can dramatically improve stiffness. Also, minimizing the lengths of the mandated heavy tubes is an option.

If your FEA program has an "animate" function, that can help show you where the weak links in your chassis are, and where extra triangulation is needed.

[Captain Redbeard]

Cool, sounds pretty much like what I was doing this summer. I think a major difference may be that due to past precedence most of the non-mandated tubes on our frame are .049 instead of .035 wall thickness so when I added more tubes the weight would skyrocket. Sounds like a good suggestion for next years team. Its sweet that you were able to improve your hub-to-hub so much with only changing the frame. Thanks for sharing, you rock as always.

-Redbeard

[CMURacing - Prometheus]

where do you guys source your 0.035 tubing? is it DOM? mild steel? we're having trouble finding it from local suppliers...

[Denny Trimble]

EMJ (Earl M Jorgensen) Steel. It's 4130-N

[Fred]

We mainly order from these two :

Aircraft Spruce (California)(http://www.aircraftspruce.com/)

Leavens Aviation (Canada) (http://www.leavensaviation.com/)

[Wizard]

Mike,

One of the best sources for 4130-N tube on the east coast is Dillsburg Aeroplane Works.

Contact: Charles Vogelsong
Tel:717-432-4589

We pick-up but thats only since its close. They can UPS 8' sections.

[James Morrison]

Denny,

what program do you use for your beam elemement FEA? We crash every computer we try with our frame using cosmos xpress and nastran.

[Denny Trimble]

Algor or Ansys.

[B Hise]

Contact: Charles Vogelsong
Tel:717-432-4589

Just make sure not to irritarte Mr. Vogelsong. Another school in our state did bad business with him... it took us 2 years to get back to working with him on a civil level.

Bryan

[John_Burford]

I believe Altair still offers their software FREE to all Formula SAE teams. If you are interested send a request to "fsae@altair.com".

HyperWorks includes applications for FEA and Multi-Body Dynamics. Personally, I perfer their program MotionView for Vehicle Dynamic over Adams/CAR. HyperMesh is widely used in the Auto industry for Pre-processing. Optistruct is their solver and is identical to nastran for Linear static analysis.

If you have the intersection points of the frame tubes, you should be able to put together an analysis in an hour.

John Burford