Chassis FEA results

[rwolcott23]

We have always run FEA analysis on our chassis and backed up those results with physical testing. On our torsional stiffness, the FEA has predicted the chassis to be 50% stiffer than it actually turned out to be.

I know that FEA is more of a precise tool, and not necessarily accurate, and was curious what margin of error other teams were experiencing. We idealize all tubes as beam or pipe elements. If our error is out of line, what could be the most probable source?

[rwolcott23]

We have always run FEA analysis on our chassis and backed up those results with physical testing. On our torsional stiffness, the FEA has predicted the chassis to be 50% stiffer than it actually turned out to be.

I know that FEA is more of a precise tool, and not necessarily accurate, and was curious what margin of error other teams were experiencing. We idealize all tubes as beam or pipe elements. If our error is out of line, what could be the most probable source?

[Charlie]

In 2003 our frame structure was independent of the engine structure, and tested 6% less than FEA results. Our 2004 car had an almost entirely stressed engine, and had significantly more error. I don't recall the exact figure, but it was maybe 3 or more times our 2003 error. Not near 50% though.

From our experience I'd say your main source of error is from bolt-in members or engine mounts (if contributing to torsional stiffness).

But you can easily find where you are off with multiple indicators on the frame during your test.

[Kevin Hayward]

Pretty much agree with Charlie.

Our errors in FEA versus physical tests never seem to be bigger than 10%. For our frames (old spaceframes ... monocoque FEA is a different matter) the FEA would always under-predict.

Kev

[MikeWaggoner at UW]

Use dial indicators along the length of the chassis. Find out which specific areas are deflecting more than they should.

[RagingGrandpa]

FEA * .6 = physical for us in 2003 and 2004

[Denny Trimble]

From my experience, you have to really screw up to get "bad results" (>10% error) from beam element FEA of a steel spaceframe. Perhaps your loads and constraints are very different between FEA and your physical testing.

[Lash]

Does anyone know whats going on here?

We perform FEA torsional rigidity test on our frame by constraining the rear box and applying a 500 ft*lb moment on the front suspension rocker arm nodes and get a 2000 ft*lbs/degree stiffness. Then we switch the constraints so that the front bulk head is constrained and the rear box has a 500 ft*lb moment. The stiffness shoots up to 5000 + ft*lbs/degree...which is ridiculous!! The values should be the same !!

Help!

[Didier Beaudoin]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by Lash:
Does anyone know whats going on here?

We perform FEA torsional rigidity test on our frame by constraining the rear box and applying a 500 ft*lb moment on the front suspension rocker arm nodes and get a 2000 ft*lbs/degree stiffness. Then we switch the constraints so that the front bulk head is constrained and the rear box has a 500 ft*lb moment. The stiffness shoots up to 5000 + ft*lbs/degree...which is ridiculous!! The values should be the same !!

Help! <HR></BLOCKQUOTE>

I don't agree the values should be the same, except if you have a symmetrical (front/rear) frame, which I doubt. Correct me if I'm wrong, but the way the moment is distributed through the structure depends a lot on the local geometry arund the moment, which must be quite different at the front than at the rear.

Also, I suppose that, by applying a moment, you apply a single force on a single keypoint? Then I think it's normal to have a significal difference between your two methods, although the difference you got seems pretty huge indeed.

Like I said, correct me if I'm wrong, I'm pretty new to chassis testing.. http://fsae.com/groupee_common/emoticons/icon_biggrin.gif

[Ben Beacock]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by Lash:
Does anyone know whats going on here?

We perform FEA torsional rigidity test on our frame by constraining the rear box and applying a 500 ft*lb moment on the front suspension rocker arm nodes and get a 2000 ft*lbs/degree stiffness. Then we switch the constraints so that the front bulk head is constrained and the rear box has a 500 ft*lb moment. The stiffness shoots up to 5000 + ft*lbs/degree...which is ridiculous!! The values should be the same !!

Help! <HR></BLOCKQUOTE>

My guess is that by restraining an already well triangulated rear box, you will be flexing alot more at the front. And, conversely, by restraining the front you are adding stiffness between the front members that doesn't really exist. Hope that makes sense.

[GTmule]

Thr front/rear thing may stem from the fact that you're having to "guess" a neutral axis at each end. By applying appropriately sized force loads through the suspension (in your case, 500ft*lb/one half the balljoint width), the computer will work out the neutral axeeez, that's why it's always done that way.

Obviously this requires some link elements for the lower A arms, springs and pushrods (or uppers and pullrods)

This also makes your constraining a bit easier, you'll just fix all three ball jopints that don't have a load attached.

[Didier Beaudoin]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by Ben Beacock:
<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by Lash:
Does anyone know whats going on here?

We perform FEA torsional rigidity test on our frame by constraining the rear box and applying a 500 ft*lb moment on the front suspension rocker arm nodes and get a 2000 ft*lbs/degree stiffness. Then we switch the constraints so that the front bulk head is constrained and the rear box has a 500 ft*lb moment. The stiffness shoots up to 5000 + ft*lbs/degree...which is ridiculous!! The values should be the same !!

Help! <HR></BLOCKQUOTE>

My guess is that by restraining an already well triangulated rear box, you will be flexing alot more at the front. And, conversely, by restraining the front you are adding stiffness between the front members that doesn't really exist. Hope that makes sense. <HR></BLOCKQUOTE>

That's pretty much what I was trying to say... :P

[chinesefonz]

where ideally should you contrain the frame then when testing at one end or the other? Close to the area of the rocker arm nodes ( or just where suspension ties into the frame ) ?

maybe a better question would be, HOW ideally should you constrain the frame?

[Denny Trimble]

At the hubs, with dummy links in place of the shocks:

http://students.washington.edu/dennyt/fsae/torsion_screenshot_DT-5-10-04.jpg

[Cement Legs]

Denny if your not careful your going to go and get yourself a job posting stuff like that.

[fade]

"Husky the love Bug" the next animated superhero http://fsae.com/groupee_common/emoticons/icon_biggrin.gif

[Lash]

Denny, Thanks for the picture. It was a great help for setting up actual torsional testing. Does anyone have other ideas? I wanted to also check torsional rigidity using just the frame (without suspension in). Has anyone done this?

[Mike T.]

You can get numbers for the frame itself, in addition to the frame and the suspension. Measuring deflection at some location on the torque arms gives you the stiffness of the entire car with the suspension, and measuring the deflection at many locations along the chassis will give you the stiffness of the chassis alone, without suspension. You just need to know the lateral distance from the center of the car in order to find the angle from a linear displacement. At least that was how we went about it.

[Denny Trimble]

Yup, we strapped some 3' long tubes to the bottom of the car, laterally, at a few locations along the length of the chassis, including right under the torque arm and right under the rear suspension. That way, as Mike said, you can look at the entire mechanism stiffness, or just the displacement of the frame. It gives a good picture of whether or not you have a suspension installation stiffness issue or not.

Maybe we should also do it with real springs installed this year...

[Angry Joe]

I like that picture. Testing an inboard suspension chassis without loading it through the suspension is complete garbage, if you ask me.

[Frank]

http://www.uq.edu.au/fsae/2004photostorsion.htm

50%-75% is normal

the worst offenders are the shock mounts / rocker mounts / and the bearings, bushes, and rod ends used

[Denny Trimble]

Frank, is there anything holding the front right corner of the test stand down? Looks good by the way.

[Jonathan D]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by Denny Trimble:
At the hubs, with dummy links in place of the shocks:

http://students.washington.edu/dennyt/fsae/torsion_screenshot_DT-5-10-04.jpg <HR></BLOCKQUOTE>


Stupid question, but isn't this giving you a different force at each hub? If you sum the moments about the pivot point, the forces are each hub aren't equal an opposite. Is this not an issue?

[Frank]

denny,

its bolted to the floor

"dynabolts"

[Lyn Labahn UW-Madison]

For teams that are not running a fully stressed engine, have you tried removing the elements that represent your engine to see how much stiffness the engine is contributing in your model? I think my model seems to be slightly over predicting its contribution based upon my seat of the pants estimation, but I have no real reason to change it until I verify it with my physical test in the spring (with and without engine in place).

[Denny Trimble]

Jonathan,
Since the "pivot point" is just a piece of angle iron on the table, I assumed it can't resist any lateral forces. In fact, it might, but they should be small (~.1 CF * 300lb max load = 30lbs). I originally designed for a roller there but thought the angle iron would be simpler and more repeatable to set up.

Since it can't resist any significant lateral forces, the lateral component of hub forces will be equal and opposite, and cancel each other out.

As for the vertical forces, the input force from the weight multiplied by the lever arm distance equals the input torque. And, the hub plates are equally spaced from the pivot, so they should equally resist that input torque.

If the car is much stiffer than the beam, things might get a little more interesting. But, in my opinion, this isn't the case.

Feel free to share your analysis of why this isn't the case.

Lyn,
It depends on the geometry of your frame. We've seen factors of 25-50% in physical testing with and without engine support members, on spaceframes with basically no triangulation in the engine bay except the engine supports.

[Jonathan D]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by Denny Trimble:
Jonathan,
Since the "pivot point" is just a piece of angle iron on the table, I assumed it can't resist any lateral forces. In fact, it might, but they should be small (~.1 CF * 300lb max load = 30lbs). I originally designed for a roller there but thought the angle iron would be simpler and more repeatable to set up.

Since it can't resist any significant lateral forces, the lateral component of hub forces will be equal and opposite, and cancel each other out.

As for the vertical forces, the input force from the weight multiplied by the lever arm distance equals the input torque. And, the hub plates are equally spaced from the pivot, so they should equally resist that input torque.

If the car is much stiffer than the beam, things might get a little more interesting. But, in my opinion, this isn't the case.

Feel free to share your analysis of why this isn't the case.

Lyn,
It depends on the geometry of your frame. We've seen factors of 25-50% in physical testing with and without engine support members, on spaceframes with basically no triangulation in the engine bay except the engine supports. <HR></BLOCKQUOTE>

Denny:

I'm not talking about the horizontal Forces. If you sum the vertical forces, and sum the moments about the pivot point, you the vertical forces at the hub are not equal, atleast by my Free Body Diagram. See below...

http://www.we-todd-did-racing.com/wetoddimage.wtdr/wNzAwNTYxNnM0MTNkZmQzMXk1NDE%3D.jpg

Please excuse the poor quality image... my scanner at home isn't cooperating. The bottom of the image cut off by the We Todd Did Racing watermark simply says FL =/ FR. Double check my calculations, since my brain is now on Winter break.

[Denny Trimble]

So you're assuming there's no vertical reaction force from the pivot point?

[Jonathan D]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by Denny Trimble:
So you're assuming there's no vertical reaction force from the pivot point? <HR></BLOCKQUOTE>

That would give me 3 unknowns and 2 equations... Only way to solve it then would be to set FL = FR as a couple moment, which invalidates what I'm trying to solve for.

Too much thinking for christmas break denny! Anyrate, I'm not trying to give you hard time... just asking questions.

[Denny Trimble]

Well, if you're doing a free body diagram for forces acting on the beam, then the reaction from the pivot has to be there. Yes, it gives you 3 unknowns with 2 equations.

I know you're not giving me a hard time, that's the value of this forum.

Happy Holidays, and for those of you on the UW team, get your ass in the shop!

[Didier Beaudoin]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by Denny Trimble:

Happy Holidays, and for those of you on the UW
team, get your ass in the shop! <HR></BLOCKQUOTE>

I think it's a pretty general statement. Maybe we should get a christmas tree in the shop? http://fsae.com/groupee_common/emoticons/icon_biggrin.gif

[GTmule]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by Denny Trimble:
So you're assuming there's no vertical reaction force from the pivot point? <HR></BLOCKQUOTE>

What alot of people do is have the pivot point be the other hub. In other words, the vertical support and the rotation happen at the Driver's left hub, while the load is applied through the driver's right hub.

Also, in the past, I've seen a jackstand used at two corners, then the third corner (in your example, the left rear) is bolted to a ceiling beam, or something that can resist the upward load. Then ya just need a long beam attached to the 4th hub (FR), and the beam's flex and all the lateral forces are moot.

[Wham Bam]

You're still over constraining the chassis thus, all the above test methods will over-predict. Think about pin connections in many places that you have welds in your fixtures. As the chassis flexes the points will want to move around and if you hold them rigid, as show above, you'll essentially stiffen the structure.

Regardless, it's pivotal (haha, I'm a loser) that your FEA is setup the same way you test the car. If you want results that might actually reflect the car's on-track performance you need to test hub-to-hub with the actual suspension with fixed dampers in place. You may want to do some hand-calcs to make sure your pullrods won't buckle under the load, if applicable. If you use solid ones in the test, use solid ones in the analysis. (solid = solid steel/"effectively solid")

I've had mixed luck with dial gauges along the length of the car during the twist tests. The 'center of torsion' jumps around so you only get a rough idea of the angular deflections. Still, it can be useful.

And of course, how stiff is stiff enough? I won't even comment other than Cornell's paper explains it well and ultimately you have to use some judgement. At least the paper gives you some indication where you need to be. Leeds has a paper where they use ADAMS to investigate chassis torsion, but I was wondering if anyone else has tried any sensitivity analysis through dynamic simulation?

[Denny Trimble]

Wham Bam,

Thank ya ma'am (couldn't resist).

You're right that a few of my welded joints on the hub plates should be pinned, I'll put it on my list to analyze the stiffening effect of that fixture vs. a pinned fixture for a range of chassis stiffnesses.

[BryanH]

Wouldn't a more accurate and faster method of testing all cars at comp be measuring the natural frequency of the chassis?
Avoiding resonance has turned out to be a major factor in getting consistant behaviour out of a V8 Supercar and stiffness is related to but is not a measure of freq. also parts of the chassis would maybe have their own freq. and some, like long skinny susp. links may have low freq in around the same range produced by the tyres?
Merry Christmas Bryan H

[alfordda]

Halfast-

Yes, that is a possibility, but it has some difficulties as well. We have done it several times in the past, to varing degrees of accuracy. If you can get your FE model to correlate with the test data, you can then feel comfortable with your model and do just about what ever simulation you want. I don't think I would feel comfortable until somewhere between 5-10 mode shapes and frequencies matched. We have had very good results with just the frame, but once you get everything else on there, it can become a very noisy enviroment, which means its difficult to find the modes. Avoiding resonances should be a very beneficial move, but it is not always that easy. As you probably know a perfect impact will excite an infinte number of frequecies, so all of your modes, in that direction, will be excited.

Just currious, but what was the test setup for the V8 Supercar like? What kind of frequency range did you go to?

[Lash]

Hey Guys,

I've looked and searched for forms pertaining to torsional rigidity testing. We are about to design a Jig that will allow us to calculate the torsional rigidity of our frame. We want to mount it to the wheel hubs with solid links for the shocks.

I am concerned about what is the best method.

A. By applying a moment at the hubs and constricting the 2 back tires.
B. By constricting 3 tires and applying a vertical load onto on of the front tire.

Option B seems much easier but would It give me the same results as A ?

Any help would be great. Thanks guys.

[syoung]

Either way, this test will give you an approximate stiffness value for the whole car, not the frame. Even with solid links in the shocks, I would expect the deflection of the suspension arms to be very significant compared to the twist of the whole chassis. (I'd hope so anyway!)

This may be what you want to test, of course, but if it's a figure for the frame itself you're after, loads and constraints surely have to be applied to the suspension mounting points (even if through a very stiff moment arm).

I don't claim to have answers here - we'd like to conduct a similar test ourselves but are struggling to conceive a suitable method.

[GTmule]

I'm of the opinion that the stiffness of the frame, per se, is of no importance, it's the stiffness of the whole deal that realyl matters. If your A-Arms or your bellcrank mounts are flexible as hell, well you really need to know that. In general, though, the only part of the suspension that will really see a large deflection (with that type of load) is that part of the Lower A-Arm (assuming pushrods) which is in bending. The pushrod shouldn't deflect much (basically pure compression), nor should the dummy shock (same deal). It just seems to me that the stiffness of the frame itself is pretty much moot, because it's hard to get into a corner without A-arms.

[Chris Clarke]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR> I'm of the opinion that the stiffness of the frame, per se, is of no importance, it's the stiffness of the whole deal that realyl matters. If your A-Arms or your bellcrank mounts are flexible as hell, well you really need to know that. In general, though, the only part of the suspension that will really see a large deflection (with that type of load) is that part of the Lower A-Arm (assuming pushrods) which is in bending. The pushrod shouldn't deflect much (basically pure compression), nor should the dummy shock (same deal). It just seems to me that the stiffness of the frame itself is pretty much moot, because it's hard to get into a corner without A-arms. <HR></BLOCKQUOTE>

I agree with the fact that the overall stiffness of the car is what matters, but how are you going to know where the weakest part and most deflection of the package is if you dont measure the stiffness of the chassis?

A good analogy is "You cant study the human body by only looking at what you eat, and what comes out the other end".

In order to find out where all the deflection is occuring, I would think at the minimum you would have to test your chassis with suspension, and then test the chassis alone. Making sure you constrain and apply all the loads properly. You should then be able to see the difference in the two and have a better idea of where improvements are required.

[GTmule]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR> but how are you going to know where the weakest part and most deflection of the package is if you dont measure the stiffness of the chassis? <HR></BLOCKQUOTE>

Confuscious say; Dial indicators point upward, measure deflection at any point along frame/suspension.

if dz/dx is big in an aread, then that area is a problem.

[Chris Clarke]

Your right, neglect that last post. Too many late nights and metal filings have killed my brain over the past month.

Chris

[syoung]

I'd still say that to do it properly and get an idea of where your weaknesses are - and for future repeatability/reference - the best option would be to test with and without suspension arms.

[Lash]

Our advisor has told us to test both frame and frame with suspension so that you can determine the slack. But you're saying with multiple dial indicators you can determine the largest change in the z axis? Makes sense...but wouldn't the suspension deflect a lot more than the chassis?

Also if anyone saw my previous post about what option is more efficient A or B, please comment on it. I'm still hesitant on which one i'm going to use.

[Frank]

http://www.uq.edu.au/fsae/2004photostorsion.htm

would the suspension flex more than the chassis?

IMO its probably about the same

the biggest loss is in the rocker and shock mounts and bearings/bushes holding the rockers and shocks

the wheel bearing compliment attributes more with increased scrub radius

you'll note that we used dial indicators measuring about the uprights..

and another way is to measure angle directly (can you see the little laser pointer?). There's a piece of angle just resting on top, and a laser pointer sitting in the piece of angle

in stock cars they mount dial indicators in incriments along the chassis rail.. we tried a similar approach, but gave up.. too difficult to get "flats" attached to the chassis

another cool thing to do is attach strain guages , and measure individual beams... in an attempt to compare fea results with physical testing

[Lash]

Sweet pics frank. I see you rotated from the back of the chassis, did you try rotating the frame from the front? I'm interested in the results of that. I like the laser level idea.