Torsional Rigidity Measurement...warning pseudoscience involved
 

Wow, 11 replies based just on pictures before I even add the text, I'm surprised!

I am planning on a bunch of changes to this car's shell including shaving the gutters and seam welding along with some reinforcements and rust repair. I do not want to install a cage for a few reasons, not least being weight.

I thought it would be an interesting exercise to see if I can measure whether these changes along the way make the chassis measurably stiffer or more flexible. This is not an attempt to calculate with laboratory precision anything at all, just a measure of better/worse and hopefully with some measure of magnitude (ie. a little or a lot.) I am not going to calculate torsional loads or claim any level of precision or relate it to how other cars might perform.

I did a 'baseline' measurement yesterday afternoon. AS it would be the one to which I compare all my future measurements I was as careful as possible. I got repeatable results.

I'll load a diagram of the overall measurement when I get home but here are some narration of the pictures.

I started by stripping the car back to its bare shell. I then put it on the lift and strapped the rear torsion bars to the rear lift arms. I dont want the car to rock on the lift and I want the rear pivot to approximate the rear suspension loadpoints. The straps are very tight, there is no play or motion.

http://forums.pelicanparts.com/uploa...1405287571.jpg

Then I lowered the car so that the front of the tub came to a rest at its front/center on a jack which was in turn resting on a scale. I believe I want this front weight (54kgs) the same when I go and do future measures because a difference could indicate the car isnt resting on the pivot points the same as in the baseline case. I noted the front arms of the lift were 2 centemeters clear of the car so they wouldnt interfere with any twisting of the chassis.

http://forums.pelicanparts.com/uploa...1405287586.jpg

http://forums.pelicanparts.com/uploa...1405287602.jpg

Then I attached a 5mm (6 gauge?) angle iron piece across the front suspension at the crossmember points. Again, this approximates one of the suspension load points. Plus it was easily accessible.

I measured out exactly 2 meters from the centerline of the car and attached a ruler to the bottom of the angle iron.

Then I placed a laser level and recorded its unloaded measurement.

Then I put all the extra brake discs I could find on the end of the bar (41 kgs) and measures how much the bar had deflected. In this case it deflected 6.1cm.

http://forums.pelicanparts.com/uploa...1405287635.jpg

http://forums.pelicanparts.com/uploa...1405287648.jpg

http://forums.pelicanparts.com/uploa...1405287660.jpg

http://forums.pelicanparts.com/uploa...1405287683.jpg


Posted here for my recollection and your (constructive) comments!

I see what you're trying to do. However, suggest using rigid bar to anchor the chassis to the lift instead of nylon strapping.

Please continue.

Sherwood

Really interesting thread, very curious what you find out.

....but the really interesting thing is what to do to increase rigidity. There have been so many ideas, some of which make great sense (welding the drip rails), X brace in the front end, later G50 bracing of torsion tube, but anything new in that regard would be outstanding.

Dennis

That rig looks way too soft to me. Are you calculating the deflection of that beam, the rubber rear bushings, the nylon straps, and correcting your measurements? The rig looks to be less rigid than the chassis to me.

I'll be following this to see what mods will make a noticeable improvement.

I think the front crossmember mounting points are the right ones, although it would be interesting to see if there is any material difference between that location and the front A arm mounting point. It supposedly is part of the looseness of the early 911 chassis.

I couldn't tell what you were using at the rear. The logical point would be the torsion tube, but I think you would need to bolt some sort of plate to the chassis where the spring plate bushing is retained - in pictures of Celette mounting, they seem to use that location as one of the major points of construction.

Dennis

Dennis - I suppose he could bolt his beam to the front A arm mounts. But why:

My take on the load paths is that just about the loads go through those two bolts he is using, which is where the reinforcing cross member bolts into place. It holds the steering rack, the rear A arm mount, is where the torsion bars load the chassis, and is where most of the side loads and braking force are taken up.

About the only loads the front bushing of the A arm sees are the very attenuated (because of the long lever arm) side to side loads. Maybe the steering loads? Nothing vertical that I can see, except for what have also to be very small forces created by the front axle not always being directly in line with the reinforcing crossmember mounting points.

So I have not understood why the 934s and 935s or thereabouts had those elaborate X shaped tubes running forward over the fuel tank, and down. Perhaps they are just to protect the tank some, or make the front end distort somewhat less in the inevitable racing crash?

Chuck Moreland must have thought about this when he designed his piece which connects to his strut tower brace stuff. Best I have heard is that they help in a crash. But that's not to say that they contribute to torsional rigidity.

Think about it - if the front mounts move some, they would move up and down torsionally. That would alter by a little bit the plane of the A arm. But not change camber, and I don't think even change toe, at least not much.

But this is a noble experiment. I don't think he needs to figure in the bending of the beam. If he has dial or other indicators under each mount, he can get the twist and easily convert that to degrees per unit of force, no?

Perhaps I can help with a vintage thread:
http://forums.pelicanparts.com/porsche-911-technical-forum/259557-structural-reinforcement.html

Sherwood

Interesting...subscribed.

The 934 and 935 had coilovers that changed somewhat the points at which loads were fed into the tub.

Study Porsche race cars long enough and you'll conclude that about the last thing on the engineers' minds was crashworthyness.

I would agree with the comments about the test rig set-up.

JR

That is a good point- if you are only measuring the chassis itself and taking a difference in angle between the front and rear then the deflection of the rubber bushings at the rear anchor points or the beam in the front doesn't matter.

Agreed, I wonder though if the front A arm see's lateral movement, which being somewhat "hinged" with the inner fender and trunk floor could cause a bit of toe in/out on bump. The X frame would stiffen it laterally more than anything.....

Which in turn means that for this noble experiment it may be not all that useful.

Dennis

Yes, thanks Serwood. That one was one of the inspirations to try and get some quantifiable measurement.

Agree. Suspension deflection, while important, is an independent variable with chassis rigidity. Should measure the basic structure first and not muddy the calcs by including pieces that hang from the chassis.

John,
I overlooked you were the original poster. I hope all is well in the PNW.

Sherwood

The angle will fail by buckling the leg pointing toward the floor in compression with very little load application; it would be better if rotated 90 degrees and mounted using spacers to clear the leg from the chassis...better yet to use a rectangular tube. However, be aware that the mounting points were not designed to be used for torsional chassis stiffness measurements and as such may be limited to what load they can take. Your long lever arm will apply quite a force to the bolts in tension and compression and they were designed more for lateral and compressive loading; not for loading in tension.

I like the digital scale; simple to measure load applied to the lever arm.

Use two digital levels and take the difference in level readings between them to calculate chassis twist.

I expect the plot will be non-linear until you generate enough twist to get the entire structure working as a unit.

It would also be good to know what stiffness you might want to achieve. This can be deduced by estimating the ultimate grip at a front tire and then use this as a load condition for which you want to control chassis twist to some acceptable level such as 1 degree. These two numbers (tire grip and torsional wind-up) will provide a target for chassis stiffness. The distance from the road to the torsional centroid axis of the chassis could be assumed to be about 24" unless a better number is known. This is my guess for a coupe but the centroid height for a Targa would be more like 12". Certainly there are those with greater technical contribution for this but I'm just thinking the process through for your awareness.

I expect the 935 cross brace was to help stiffen top of shock tower by providing another load path to a support point. Also, by cross-bracing the open front compartment it becomes more rigid as a unit and helps reduce deflections at the suspension pick-ups.

It would be VERY interesting to have a coupe to compare your torsional stiffness numbers to. At least there would be a standard to aim for.

Still digesting your other comments but note thisis a coupe...a 1978 US market sunroof coupe.

For the same reason measuring the beam and bushing deflection isnt necessary I dont think doors are necessary...Im just looking to make internal comparisons to the same car over and over as I cut gutters, add reinforcements, add lightness, etc.

The doors have clearance at the latches so a torsional test would not have them contributing stiffness until the clearances at the latches were closed which would allow for the doors to share some load. Better to forget their contribution.

Glad to hear it is a coupe instead of a flexible flyer (Targa).

I have a 71 targa, next time its on the lift ill see how it compares.

The straps don't bother me as long as there is no gap to the lift. I would however agree with not measuring the same bar that's applying the load. Perhaps you could apply the load elsewhere? Or do two parallel bars with a gap in between, load one and measure the other?

Even if you're only trying to get relative measurements it'd be nice to see a more representative percentage difference. If bar flex is 90% of the deflection and you cut chassis deflection in half, your new measurement will still be 95% of the original. Repeatable perhaps, but better signal to noise would be nice...

Did anyone read the Pano article on the development of the Targa about a year ago?
Too much to cover here on the forum, a must read for its how P did what your trying to do with about as much resources. Very interesting and informative.

JohnJL - do I have this right: You put weights 6' out toward the end of your long bar, and measured how much closer to the floor the bar was with and without the weights?

If you did that, I would suggest that what would be a better measuring point would be from the bolts holding the bar to the floor. That is chassis twist, and you don't have to worry about how much your long bar bends. And you can measure how much down one bolt moves, and up (or whatever it does ) the other bolt moves.

I know you aren't worrying too much about quantifying and precision, but couldn't you use two dial indicators on some kind of stand to do this measuring? They are not expensive, maybe you have a couple, and surely you have friends who do and might bring some over, and maybe those magnetic mounts with rods and thumbscrew clamps which often come with them? This should be easier to read than a tape measure or the like.

As to the nylon straps, they are going to stretch. I'd be inclined to use chain somehow, and a chain tensioner like truckers use, if it could be used in the space you have. Maybe wrap it around the torsion tube?

But there is rubber bushing stretch. Perhaps you can compensate for all the stretch back there with measurement? Treat the unstressed tub as a plane. If one rear side moves up, deduct that from how much that side moves up in front. Ditto with the side moving down?

I think this is a really great thing.

I am in the GTA, if you need material let me know. I have a bit of everything...

I cant quite visualize what you mean. Both my dial indicators are gravity-based. I guess you mean mount them upside-down and push up the plungers to measure the change in distance with/without the load? I would have to build a solid rig to hold them too, something that can be moved away and stored, then brought out again in a few months as things progressed.

Use a digital level on repeatable mounts at each end of chassis, better two levels that are not moved during your measurement process. Zero them and record readings during loading intervals. Remove load and see if they return to zero. Total chassis twist is difference between level readings, no need to deal with motion at your support points. Total spring rate is difference between level readings (twist) divided by load times the distance from center-line of chassis to point of load application on the beam.

By the way, the digital scales will indicate less than the actual load applied since load is applied between the supports of the rear torsion tube and the single point at the front where the scale is located. Actual weight applied can be calculated from the digital scale reading and the geometry (location of supports and load application beam) but easier to measure individual weights.

If you could attach loading beam to the tops of the shock towers then I think there would be less risk for damaging your chassis (pulling out a threaded mount). Also, the loading at these points of application will be lower in magnitude than at the threaded mounts for the steering rack for a given torque about the chassis. The upper shock mounts are designed for vertical loads so they are a better structural interface.

I've got a longacre digital camber/caster level. I'll give that a shot too and see what angle changes I see when placed across the croaamember lcoation where the beam is bolted.

I've never thought of dial indicators (or their digital successors) as being gravity based. They have springs pushing parts around, and are independent of gravity. Yes, I envisioned some kind of moveable support - a big tool chest, just about anything. Plus you don't need the chassis high off the floor to do your measurement.

Paul's idea of dispensing with distance measurement and using angles directly makes more sense. Digital levels are fairly inexpensive, can be zeroed at any actual angle with a tangent to the earth, and would, as pointed out, go directly to what you want - degrees of twist per units of weight.

I think the crossmember mounts are strong enough for this. They are welded into the lower pan sheet metal, and again in a higher level you can't see because the fuel tank is in the way. For lateral loads, the crossmember has a recess around the bolt holes, and the chassis mounts have a collar sticking down. The two mate, and thus take up side loads. But you aren't using much weight. I suppose you could build up a structure over the shock mounts to clear the inner fender metal - but you have one side pulling up, for which the shock towers are not really designed.

So the only thing I think is wrong in your protocol is to be measuring out at the end of the flexible beam.

I agree with Walt - you're really measuring the flex in your piece of angle iron and the twist in the chassis.

Actually - you may not have reached the point where the chassis is twisting. There is really no way to tell based on the way you have things set up.

If you follow Walt's suggestion and measure movement at the suspension mounts, you'll only be measuring the chassis movement.

Great experiment! Can't wait to see more results!

K&C rig is needed
 

You need one of these
SPMM4000 - Moving Body Suspension Parameter Measurement Machines: Kinematics & Compliance Test Machines

Oh yeah. Let's order up a SPMM4000. :eek: Can you quote us a price.... out the door?

Agree. If X ft.lbs of torque is applied at the end of the bar (angle iron), then that is an element that must be considered in the calcs, especially if it flexes.

Here's a link to improving the chasis rigidity of an Ultima GTR, a very quick kit car. Of interest are some chassis rigidity numbers of various vehicles at top left.
Untitled Document

Sherwood

I am anxiously awaiting the gusset and bracing part 2 of the thread!
 

How the engineering study is translated, is when it gets really interesting. Maybe we can get some Las Vegas types odds and pay-outs going on strategically where we all would guess they are going to go! :D

Everything flexes unless at absolute zero when molecular movement stops.

Use two digital levels and any flexure in bars or movements at attachment points is irrelevant.

Dial indicators will give better resolution.

Dial indicators give better resolution but you need four and then the trig to resolve to angular rotations. Absolute accuracy does not seem to be the end requirement but a method to determine progress in chassis stiffness modification. Also, indicators must be mounted to a ground reference that is not affected by the torque applied. In this case that ground reference would be the garage floor which is somewhat below the points of interest so reasonably substantial mounts would need to be constructed just to locate and support the indicators.

Then again: The stiffness of a 993 is around 24,000 ft-lb/degree and a cabriolet/Targa is about 1/3 of that. A 911SC is probably somewhat below that figure but would still be pretty stiff. So, to load your chassis to generate meaningful deflections you would probably need to use something more than a few hundred pounds of weights which implies that a battery of indicators might be required for a home garage stiffness test.

So if the expected ballpark of stiffness of a 78sc was say 15000 ftlbs per degree, someone willing to estimate the weight i would have to put at the end of a 2 meter bar attached to the crossmember points? My digital level is pretty good, has 10th of degree resolution.

So if the expected ballpark of stiffness of a 78sc was say 15000 ftlbs per degree, someone willing to estimate the weight i would have to put at the end of a 2 meter bar attached to the crossmember points? My digital level is pretty good, has 10th of degree resolution.

So 15000 ftlbs/deg x 0.1 deg is 1500 ft pounds. If your lever is 10 ft long you need 150 lbs to get that 0.1 deg deflection.

And you want well over .1 degree of twist to get any sort of accuracy (otherwise you will have a result that is 7500 ftlbs/deg plus or minus 7500 ftlbs/deg since the level will round up or down). Which is why I favor the dial indicators.

Thanks guys.

Wouldn't it be 75 lbs though, since the deflection is across the plane? Half the twist is "up" on the far side of the beam and half is on the same side as the beam?

....Banker doing physics.....

No, the torque will cause one attachment point to rotate "up" and the other "down"; their total rotation will equal the torsional stiffness of the chassis (in degrees per foot-pound of applied torque) times the torque applied.

One data point is not a curve so like Flieger said you will want a few data points to plot a stiffness response curve. Your data will be scattered and most likely non-linear, especially for the first few points. If you have only the one level then you will need to have very good mounts for it. The mounts must have very good locators so when the level is shuffled between them they will provide repeatable readings. Thermal distortions will affect your readings so stay out of the sun.

The 150 pounds applied at 10 feet from the centerline of the car will generate approximately 825 pounds of vertically applied load to each of your steering rack mounting bolts; one in tension and one in compression.

It would be easier to load the chassis if you used concrete anchors to provide a point in your garage floor to attach a turnbuckle to your lever arm. Your digital scale will resolve the applied load and you don't need to shuffle weights around or have an excessively long lever arm.

By the way, since you are using the digital scale to provide loading data you might want to correct its output to show actual weight/load applied. Easiest way is to add 200 pounds to the steering rack mounting points and observe the scale's output. This will be a constant factor for whatever load is applied, the digital scale will read perhaps 85% (guess) of actual weight/load applied.