Torsional Stiffness of 1988 911 coupe

[moneymanager]

Burgermeister, you don't have a Targa in that garage somewhere do you? WOuld be interesting to know how bad they really are (or aren't.)

[jester911]

Seeing a Targa and a Cab as compared to a coupe would be interesting as well.

[burgermeister]

Originally Posted by Bullet Bob

Burgermeister, I didn't have any issues with your setup, that was Peter Bull. The slope of the graph looks pretty linear so I would say the setup worked well.

Whoops! Apolpogies ... I am a flake.

As penance I'll share a good flaky story unrelated to the posted topic.
Took my 911 out for the first time this year - did a lot of work, hard & soft oil lines, swapped spring plates on rear suspension, pre-muffler, some rust repair in the rocker area. Car starts, all checks out & dry, Wife & kids hop in, we go out for dinner. Get there 20 minutes later, get out. Oil puddle by hard oil lines - uh oh. Explains why a few people were passing me in an irritated fashion - probably sick of the oil fog. Also explains why the tail was so easy to slide. Everyone back in, drive home. Cursing my decision to solder copper into the hard oil line as a repair. Oil pressure is increasingly flaky the whole way home, but never below 1 bar (also car always leaked, so I didn't quite run all 13 quarts out ...), driving like a grandma. Pour in 3 quarts bought at a passing gas station. Get home, all sorts of obscenities (poor kids!). I decide to look wherefrom the mess comes. Turns out I forgot to tighten one of the hard oil line fittings to the t-stat.... crap crap crap!

Moral of the story: I am an idiot. I feel better now though.

[Walt Fricke]

So enlighten me about the area where you are not an idiot: engineering.

The only setups I have heard or read about for testing torsional rigidity are ones where you lock down hard three of the four corners of the chassis, then use a long beam to apply the force and then measure. Like those shown in reply to your pictures. Since I could figure out no way to lock even one of the rears down - the one diagonally opposite where some kind of beam would extend out - I kind of wrote this off for a DIY effort. Seemed like it would take a frame bench or something equivalent.

Unlike corner balancing (where in essence jacking one side up or down just reshuffles the weight deck but doesn't change the total, the long beam method imposes a lot of extra torque and some extra weight.

I see you are using force up instead of down. And the car's weight/mass itself to do the twisting?

Does the knife edge balance point midway between the front contacts somehow equalize the forces so that the only thing moving is the left front point where you are applying and measuring force and movement? Does the right front go down an amount equal to the amount the left front goes up? Is it the car's own weight that holds it at the same height on each of the rear jack stands? Did you have to keep a careful eye on the left rear jack stand and end the experiment when you saw a micron or so of daylight between pad/stand and chassis?

If you lived near me, I'd let you borrow my fleet of four dial indicators (one is digital electronic, as that was the cheapest way to get one that reads in metric - amazing considering how inexpensive these dial indicators are that no one sells an inexpensive metric one). More to the point, I'd come over with the set of corner balance electronic scales I own as part of a group.

Anyway, love your methods. I just don't quite follow how this one works.

Walt

[burgermeister]

Walt, I NEVER implied I was not an idiot in engineering...
And quite clearly, this is a DIY effort... 80% of the answer for 20% of the work...
But I haven't read of anyone trying something more elaborate either.

Even though we're not neighbours, and I can't borrow any dial indicators, I'd like to meet you sometime - if you're ever in Michigan, give me a holler!

Gravity indeed holds the rear of the car down. The only criteria is that the car stays on the jackstands in the rear. The small amount of torque I was applying was significantly less than what is needed to lift one of the rear corners off the jackstands. In fact, the car can be jacked up completely on one front corner without either jackstand in the rear loosing contact, thank to the rearward weight distribution bias. Though that would grossly exceed the bathroom scale capacity.

In the front, all the vertical load is initially carried by the knife edge pivot, which I was careful to center in the car. When I then apply an upward load on one side of the car, that same load is no longer carried by the center pivot. The front - rear weight balance of the car didn't change, so the sum of the pivot vertical load and the applied vertical load must stay the same. So that's where the couple (moment) comes from - the added vertical load off to one side, and the requisite reduction in vertical load at the center pivot.

The other (unmeasured) side of the front beam I used would move down an equal amount as the loaded side moves up. That's not quite true, as I am reducing the load on the center of the beam, and thus it bends a bit less, and this is one of the errors in my setup. 4 dial indicators would have been the way to go. Then again, using that hysteretic (?) bathroom scale as a $8 force transducer pretty much precludes accuracy better than +- 5 to 10%.

All of this works because of "linear static superposition', quite frequently used in structural analysis. Basically, if you have a complex load set, each load can be applied and measured individually, and the sum of all the individual effects would equal the effect of the complex load. This is the basis of commonly used fatigue analysis methods. Also, a load added to an already loaded structure will have the same effect as the load added to an unloaded structure. For instance, the torsional stiffness of a car would be the same measured in outer space as it is under the influence of gravity. All of this only works so long as the structure is in its linear range, and geometric effects are negligible.

Hope this makes some sort of sense ...

As far as the oil debacle, still had 3.5 quarts left, and no metal stuck to either drainplug magnet, so I think I got lucky. And I got an oil change with significantly less used oil!

[Walt Fricke]

You mean the scale that causes you to get hysterical when you step on it and realize how much weight you have gained?

Or the one that doesn't quite show changes in weight applied accurately because of a little lag in its mechanisims (from friction, perhaps)? My dictionary shows hysteretic as an adjective, though I don't think I had ever seen it as that part of speech.

I'll be at Mosport for the PCA Club Race (alas, officiating, not racing). And at Charlotte at the Parade.

I can probably duplicate your setup, but with scales of 1% accuracy (resolution is more than that, for what that might be worth). Shoot - I could also put one of the scale pads under the center pivot.

I normally rest the 911 in the rear on the spring plate extensions. The SC has Polybronze bushings, not apt to compress much. For the track car I just use the bracketry holding the heim joint bolt in double shear, as it is located conveniently.

I suppose I can measure jack side height increase, center pivot height increase, and other side height increase, which would take care of the bending of the cross beam and other effects?

Another decent place to lift/support from is where the fronts of the A arms attach. Plenty strong enough to take the weight of the car. Might learn something from measuring twist at different places.

Lots simpler than that long torque arm. And if I get around to doing this, you or someone on this discussion can tell me what the results mean. Someone will probably beat me to it.

Now, if someone would only share the results of their successful CG height measurements!

Walt

[burgermeister]

Yes, I meant a combobulation of hysteresis and hysterical

The front of the A-arms? After doing a pan replacement, I don't see how the structure is strong enough to support the car. There's nothing there but a flat piece of 0.8mm steel, loaded in "oil canning" mode! But, empirical evidence trumps gut feel and theory, so if it works it works. The engine mounting system doesn't look like it can support the rear of the car either, but it sure does.

Torsional stiffness to the front A-arm mounts wouldn't affect the ability to transfer roll moment from rear to front, though.

What does it mean? Not so sure myself. I'd guess if the roll stiffness of the front or rear suspension gets close to the torsional stiffness of the car, tuneability starts to go by the wayside. By my figuring, a 100#/in wheel rate and a 60" track would give a roll stiffness of 520 ft*lb / deg. Now that's without any caffeine, and I may have botched the numbers ...
Edit: I did botch the numbers ... and noone called me on it. Shame shame!
260 ft*lb / deg seems more reasonable (no sway bars assumed)

And I read in the materials for a 3 day vehicle dynamics introduction class I took recently that the CG height for a 2000 911 (not sure if 4wd or 2wd) is 480mm. FWIW.

[RWebb]

- where did they come up with that value?

There were some guys who did an article in Pano on a 996 - they used 19 inches (about the same as 48 cm) - when I wrote one author tho, they said they just guessed at the Cg. Tho it seems a reasonable guess.... There is a procedure for determining the Cg using weighing at various angles - also published in an old Pano.

[Bullet Bob]

The difficult part about measuring the CG height is that the suspension has to be fixed so that it can't travel while tipping the car (solid shocks, etc...).

FWIW, I have read that the torsional stiffness of the chassis should be at least 5X of the suspension's roll stiffness. What is a typical wheel rate of a 911?

[burgermeister]

I've been thinking about how to correct for the wooden beam deflection ... without placing another Harbor Freight order and without setting the car up again ... and recently I figured, hey, why not just measure it?

I put the wooden beam used to balance the front of the car across 2 jackstands, my knee in the middle, and noted deflection, as well as my weight on said bathroom scale. I zeroed the dial indicator with my weight directly over the jackstands to get any slop out of the equation. The "procedure" was surprisingly repeatable, over 3 different setups I got the same number +- 10%.

That wooden beam was surprisingly springy, at 17000 lb / in.

I used this number to correct the original spreadsheet deflections / angles.

New, corrected torsional stiffness: 5200 ft*lb / deg.




Not quite so noodely after all ...

[RWebb]

nice - did you fit a regression line to get the slope as = 5,200 ??

[burgermeister]

Nothing that fancy - I don't even know how to set it up in XL. Just an eyeball intercept ...

[jurhip]

I kinda glanced over this topic, but had a few comments.
Chassis stiffness is mainly desired, as I believe someone noted, so that the suspension geometry can be designed (on paper) to fit the desired handling intent of the car. Then, it is tested in the real world to tune out the things that could not be assumed correctly.

If for some reason, a chassis "bent" correctly, you would not need a stiff chassis as you could use this as a suspension component. On cars, this is all but impossible as it flexes along an infinite number of axis. Motorcycle racers, on the other hand, have learned that torsionally rigid frames did not handle to their liking, apparently less complianced when leaned over where the suspension does not act along the same axis as encountered road imperfections and throttle/brake inputs.

An extremely stiff chassis actually allows more compliance suspension as you are able to get the geometry you want without trying to overcome flex.

Long story short, this test is informative, very interesting, and probably good to test cages and chassis tie-ins. However, we must remember these cars can 10-40 years old with archaic suspension to begin with. I think some stiffening helps, but you will never overcome the limitations of the pre 964 suspension weaknesses, nor the effects of age on the.

Keep doing the tests though. I am interested in how certain things help (like a strut tie bar).

[RWebb]

While trolling thru these threads recently, I noticed that Tyson posted his driving impressions of 911s with stiffened chassis -- and they were not at all subtle effects. You might say they were significant, or major effects.

2nd, there is a huge increase in torsional rigidity for modern cars as vs. those from teh 1960s designs (like the 911). I am certain that the reason is the use of Finite Element modelling - which requires a lot of computer power -- more than could be had in the 1960s. In fact, that level of computer power was not generally available in university settings until the late 1970s/early 1980s - I know b/c I killed an Amdahl with a fairly simple FEM of a hibernating marmot about that time. Luckily the computer center people were unable to catch me...

3rd, I found some additional information on torsional rigidity of the 996 - enough to back calculate the values for the 964 and 993.

The 1998 Porsche 996 coupe has a torsional rigidity of 20,120 N-m deg^-1 (Ludvigson, Carl, Excellence was Expected, page 1381). This is 45% greater than 993 (Becker, Clauspeter, Porsche 911 -- The Evolution, page 121).

Thus, the 1994 993 coupe was 13,876 N-M ^deg-1. In turn, the 993 was "20% stiffer in torsion than old design w/o adding wt. [allowing] reduction in the number of body welds [and hence] production costs." Exl. April 1994, p. 73

That means the 964 had a torsional stiffness of 11,563 when it was introduced in 1989.

I have not been able to back calculate to the original 911 and the value that Mark Donahue reportedly reported...

I added these new numbers to a spreadsheet called Stiffness.xls which is based on a variety of data that Sherwood originally compiled.

Sherwood - I'm ready to hand it back over to you now if you want it...

If anyone else wants a copy, email me and mention Stiffness.xls

[911pcars]

"Sherwood - I'm ready to hand it back over to you now if you want it..."

Randy,
You're doing fine. I'm just along for the ride. Thanks for everyone's input.

Sherwood

[hcoles]

rwebb, excellent research
I wonder if current Porsches are made with more spot welds per foot compared to other brands. This may also be a factor. I'm guessing the Porsche chassis are now all robot spot welded compared to hand spot welding. I read somewhere that "specials" are made by putting in a larger number of spot welds.

[RWebb]

In Becker's book - page 121, he notes that the 996 uses 3 types of steel: regular steel, and special high-strength steel in certain areas - about 30%, and about 1.3% of the body is "extremely strong boron steel."

Also, the 996 has special laser welded assemblies called "tailored blanks" made of different thicknesses of steel that are then incorporated at certain places into the unit body. He was a little unclear on exactly what they did - suggesting two functions: absorb energy in a collision and handle "extreme load variations."

There are also some diagrams floating around showing different steels used in different spots - they are color coded by steel type.

[hcoles]

FYI - in case we have forgotten..
high strength steel stiffness = "lower" strength steel stiffness
for same geometry.
likely they found "hot" spots where the panel was cracking and put in the higher strength stuff to control that situation - I wish BMW would have done this on E36, E46 cars where they have unibody cracking in various locations.

[RWebb]

updated last post

several articles on tailored blanks:
http://findarticles.com/p/articles/mi_m0KJI/is_7_117/ai_n14839879?tag=rbxcra.2.a.1

[911pcars]

Originally Posted by hcoles

FYI - in case we have forgotten..
high strength steel stiffness = "lower" strength steel stiffness
for same geometry.
likely they found "hot" spots where the panel was cracking and put in the higher strength stuff to control that situation - I wish BMW would have done this on E36, E46 cars where they have unibody cracking in various locations.

"high strength steel stiffness = "lower" strength steel stiffness for same geometry."

Not sure what you're trying to describe with the above statement.

Manufacturers use a combination or "regular" and "high strength" steel more to develop a prioritized crush pattern during a collision while protecting the structure around seated occupants.

Sherwood