Just because the model was made "with incredible detail" doesn't mean it is correct.

Boeing had a similar problem with the C-17, bad element formulation was one problem.

In the end, any good structural analyst will tell you that the model should be validated with physical test ... but the point of FEA is to minimize reliance on physical test, since it is very expen$ive. Ideally the structure is modeled and optimized ... 1 prototype is built, is tested, and validates the model. In practice -not surprisingly - it doesn't work this way.

I would also add that many other problems Boeing is having with the 787 have little to do with structural problems ... they basically are trying to sub out their core competency, and it isn't working out that well. Vought/Alenia (IIRC) is a perfect example, they make fuselage structures very well. They don't do a great job populating the fuselage(s) with hydraulics, electrical, pneumatic, and all the other "stuff" that needs to be installed in the fuse.

The cynicism about FEA is fine. However, here's a comparison between my models' results, and the measured results, for a modification to a ladder chassis, showing the deflection along each rail. The x axis is distance along the car, y is vertical deflection, the car was restrained (carefully) at three points and a load applied to one spring tower. I'd add there were fewer than 300 elements in that model. There again I've been dabbling in FEA for 27 years.

http://www.geocities.com/greglocock/gallery/hebel_mod.png

I probably did that model in OpenFEM, as I wrote a beam element optimiser in OpenFEM, which is a bit unusual, most optimisers use solid elements. This selected the optimum beam for each part of the chassis from a palette of supplied properties, and found the combination that gave the lightest weight with the required stiffness (10000 Nm/deg). This was a 25% weight increase over the original frame which was 3700 Nm/deg. Incidentally the original frame was designed by hand analysis and showed that somebody had taken great care in the selection of the section sizes, the flow of loads through the thing was very neat.

Also, to state the bleeding obvious, if your model can be used to predict crash to a usable extent then its estimate of torsional rigidity is a given. The correlation between the FEA models I see at work and the real world dynamic tests is gob smacking.

SO far as targets, the Stevens Esprit was too soft, at 4000 Nm/deg. I added a stiffening frame (for my own mysterious reasons) and almost doubled that, this allowed them to vastly increase the shock setting, and made it understeer, a good thing.

However, there are plenty of race winners that were down at 1500, and in a race the virtues of light weight may well outweigh(haha) the virtues of better suspension.

Incidentally quite why anyone regards the physical testing stuff as a mystery is beyond me, a weekend is enough to do it yourself on a bare chassis. There are some gotchas. So quit whining, get testing. Publish your results.

One other neat trick I saw at Lotus was they built a one sixth scale model out of transparent plastic of a full sheetmetal bodyshell. We didn't even need to test it, the designers would look at it and say, oh shit, that joint is rubbish, and go away and have another go. The material properties of the plastics used were selected so that it was easy to scale the static deflections and natural frequencies from the plastic model up to the real world. That's a proprietary GM technique, won't see anyone do that again, nowadays the FEA and visualisation tools are good enough that the designers can see where the joints are bad.

The second degree will be the same as the first. Even coil springs will always remain linear in the working range, unless they are specifically designed to "bind" some coils making their rate progressive.

Agreed. If the design is fully triangulated, all members are (effectively) in pure axial loading and the joint stiffness along the same axis will be >> the series stiffness of the attached member.

Just the vagaries of different rigs will give wildly varying value for stiffness (and always higher than true value)...it is very difficult to setup a rig that doesn't introduce constraints in the object being measured...so as a rule of thumb when comparing stiffness values the rig design comes into it. Same rig, closer values. Different rigs, beware...re validation of FEA against real world, have achieved quite good correlation (with FEA always better than measured...)

...lovely if you can achieve it, but most chassii have compromises dictated by the need to fit a driver/powerplants/suspensions in them...

Re reading this super thread made me think that maybe there isa bit of pre procesing needed before your do all the spaceframe design. It is rather stating the obvious but here goes.

If you are not worried about NVH or statutory crash testing ( as most racers are not) then the first question is is "torsional rigidity for what".The answer I think is to make the springs and dampers flex instead of the chassis so they can accurately control the car. Therefore the chassis only needs to be stiff enough to make that happen with a reasonably good safety margin over the length of time you want to use the chassis. Back in the glory days of spaceframes racecars cars were light,spring rates were low and chassis life was short so quite low torsional stiffness was adequate. Once you increase car weight and spring rates ( espeically to GE levels) the requirement soars so spaceframes have a hard time.

The other factor is chassis cross section, an FF car had a cross section of about 560mm * 400mm to resist twist, a sports car could be 1500mm*500mm so , all other things being equal ( I know they never really are) the frame demands are less in the SC. To complete the picture you probably need a fudge factor to describe the abilty to triangulate the cockpit area which is the great weak point on most spaceframes. Incidentally one of the secret weapons of the Mallock chassis stiffness is that the rules of the class it was built for required a cockpit at least 32"( 800mm) wide and you were allowed to run a tube across the part of it not occupied by the driver. Hey presto a fully triangualated cockpit/chassis top bay with a recent cross section.

Anyway the mathmatically minded could maybe produce a formula for any car , even if you did not design it, by say

1)- take the number of spring coils over 150mm of the springs
2)- take the thickness of the coil wire
3) divide 2) by 1) to get a spring stiffness factor

4) adust for effective spring ratio to get the vertical rate

or just know the vertical rate to start with if possible

5) scale by wheelbase to give total required stiffness

6) calculate chassis effective cross section

7) divide 5) by 6) to give the torsional stiffness target

8) decide your safety margin of raising the chassis stiffness well above sring/damper stiffness

THEN start the FEA etc.

I appreciate that most of that is stating the obviuos that all designers probably go through but it sometimes helps to decide what you are trying to acheive. The point is that adding too many extra tubes to get more stiffness than you need does not just add weight it adds demands on fabrication skill, makes access to car systems harder and makes any crash repairs harder, all of whch can seem rahter important once the designing is done!

Your absolutely right, I miswrote what I intended sorry - so try again;


A long spring, large diameter with lots of coils say 500lbs per inch may offer resistance of 750lbs for the second inch.

A short spring, small diameter with few coils may be 300lbs per inch but the second inch may be 2000 lbs.


I'm not budging on the chassis larger diameter theory though.

Same with chassis - large diameter sections offering greater resistance for the second degree will always be better in my mind even if it starts with a lower figure initially.

Often overlooked - the importance of damper stiffness - the dampers frequently feed higher loads into the chassis than the springs.

Much higher loads depending on the event and car. I suspect these loads are an order of magnitude higher than any suspension point on the car...I've having to package some on the front of a spaceframe at the moment.
One comment I would add is that there is the frame that is satisfactory for the loads and then there is the frame that is satisfactory for the job. The latter is rarely a thing of beauty, being compromised by all of the varying demands that occur - either in reality or in the designers imagination! An example of this is the rear subframe on my own car which is now removable by undoing 12 bolts - this allows the rear suspension/chassis to be completely removed to allow access to the engine transmission. Engine is positioned in the chassis so that both exhaust manifolds can be removed without having to struggle. Dampers and springs are easy to access, cable and hose runs are as straight as possible. Yes is is heavier than it might otherwise need to be, but this is the first iteration and I do not want to worry about chassis structural performance...besides I need to loose more weight than the chassis!
I too would like the perfect book on chassis design.

Oh, memories. A year or two ago I walked into a back room and there sat a hundred or so of those bodies in storage. Interesting process, esp. the load rig.

Yes - although I was talking about the effective wheel-rate force (normal force at contact patch) contributed by the damper compared to the spring, since this is the effective force that applies the torsional moment to the chassis.

Are chassis resonant frequencies important?

That's the $64000 question. Within living memory Porsche sold a car with a 12 Hz first mode. I know cos I did the test. Jag didn't believe me, and then did some more tests, which agreed . Didn't apologise, so they dropped a few points in the credibility meter.

Are you talking about for a race car or a roadcar?

Yes I knew you were relating to the damper/spring force, but I wanted to include suspension arm forces, which do also contribute to the torsional moment I believe...though naturally not nearly to the same degree and only during cornering. I find it interesting that many home made race cars are happy to have a couple of thin tabs hanging out in the air for suspension arm mounts and also damper mounts. Apparently for some you design the chassis and then you design the suspension and then you sort of glue the two bits together with some thin sheet steel.

As long as the suspension input points/nodes are sufficiently rigid, you can neglect suspension arm forces, when evaluating the effect of chassis torsional rigidity on wheel motion. The impotant thing is vertical wheel displacement, ie what proportion is suspension movement and what proportion is chassis deflection. I'm not an expert but I would think the suspension portion should be at least one order of magnitude (perhaps several) higher than the chassis portion.

Getting back to the original point, the maximum wheel force (and therefore chassis deflection) may well occur at zero suspension travel - since this is where wheel vertical velocity will be near its highest value and damper force will be maximum. So it may be necessary to calculate the wheel displacement due to chassis twist under these conditions and compare that to maximum wheel travel of the suspension.

Greig, my primary interest is in race/rally cars but I am also curious about road ones. From a NVH point of view I would imagine vibration and resonance is very important but I can't help thinking it is an issue with motorsport cars in terms of grip and maybe handling. In a previous life, long long ago I have had some involvement in structural dynamics but I have forgotten most of it! What is important about 12hz?

Nothing as such, it's just that most monocoques even in the bad old days were up at 18 hz, these days 22-25 hz is more typical, and BMW claim 40, but that may be for a bare BIW. Roughly speaking the frequency is proportional to the square root of the stiffness.

Actually 12 Hz is getting far too close to wheelhop.

Interesting. Is the first mode torsional?

I wouldn't get too hung up about what the mode shape is called. If someone has optimised the structure with the aim of making pure first modes, you'll typically get bending first and then torsion a couple of Hz higher, in a FR car. before FEA it was fairly common for both of the first two modes to be combined torsion/bending modes of just one end. Mercedes in particular didn't stand for any of that nonsense and would develop the mode shapes experimentally, at least in some of their sedans.

Cheapy, sorry to do this to you but you have provided a textbook perfect example.

Back in post #3 of this thread I said, ¨Nothing really provides much info on the most important matter in mono or tube frame design and that is bracket design and how you feed the loads into the structure. I don't know anything better than spending time (lots of) looking at and studying real life examples to recognise both good and bad examples.¨

Others have alluded to this in their posts.

Well Cheapy in his thread, ¨What do you think, Guys¨, has provided a photo in post #144 of about as bad a bracket design as can ever happen.

I will post a prize of a case of beer, (to be picked up of course), to the obserever that can make a list of the most things wrong! Actually, if you pick it up I will help you drink it.

While I have fingers on keys I also will comment a bit more on torsional resistance being the most important design paprameter.

People once built some real Flexible Flyers, (remember the sleds?) that seemed to work OK from the standpoint of not breaking and seemingly holding things together. People just didn´t appreciate the importance of frame torsional rigidity to handling capabilities.

One of the first and most dramatic examples was the great advantage that Chunky found when he designed his first monocoque single seaters. They were overnight quanum leap successes despite using the same mechanicals as the chassis he was succeeding.

In my own experience with developing stiffer chassis I found that only about a 5% change in front or rear suspension settings made a very measurable difference to handling where the previous less stiff chassis needed anytning up to a 15-20% change tosee the same result.

The reason is that in the less torsionally stiff chassis the front and rear suspensions were acting quite independently. Once you start to torsionally stiffen the chassis they start to work in concert and tuneability and lap times just tumble.

There is quite a nice curve of stiffness vs improvement that is quite linear and then starts to curve flatter until becoming quite horizontally asymptopic.

This is the effect that Lotus found with the monos. Between the last of the type 24s and the first of the type 25s they picked improved torsional stiffness by a factor of 3 1/2.

The improvements keep coming as you get sttiffer up to a point at which added weight starts to negate ever smaller incremental improvements in the working of the suspensions.

Gregs comments in another post identifies favourable torsional stiffnesses of about the same magnitude as I stated in my second post.

The good thing about designing for a torsional stiffness is that you really don´t even have to worry about suspension loads or where they come from in or their direction in your calcs.

That is if you have good bracket design!!!

Regards

let's try...

upper wishbone, rod ends are vertical instead of horizontal - more unwanted axial loading.. Bracket is not properly supported for loads going along the vehicle axis... Upper wishbone is bent at teh ends so that the rodends are at 90 deg to the chasis (so that they can be used for adjusting the camber probably) which means that all wishbone forces do not end up as pure radial but you have an axial component all the time..
Bracket is welded to the middle of the vertical tube, which is a no no as it will bend..
Lower damper mount too far away from the outer lower ball joint, ball joint threads are being bent all the time..

It is useful to go to their website and look at their pics of beautiful, fully machined, billet uprights and wishbones. Some of the engineering is apalling.

Facts are facts - not my car anyway and you can view a car from the same company in post 55 in the same thread and make up your own minds (I especially like the shocker loads into the heim joints on the support bar ).

Eye candy and engineering are not the same thing but one certainly outsells the other sad to say.

I thought the quote in the current RCE (Dakar car on the cover) by the Radical team engineer was very interesting. He claimed that they think they could make a spaceframe LMP2 chassis just as stiff as their current carbon tub but only 5kg heavier (and obviously a crapload cheaper).

Anyone else read that?

Cheapy

Glad to hear that it is not yours. (Just pictured on your thread.

Still one of the worst bracketing examples I have ever seen.

Regards

Zac I recall they could make it as stiff...I wonder if they could make it pass the impact tests etc?
FWIW Multimatic made a tub for Lola here: http://www.multimatic.com/engineering/projects/lola.shtml
You will note the 116Hz first freq...would construction materials affect freq?

Allow me to make a personal point of view in terms of frame construction. A framework is the car's skeleton, whose main role is to link all components into a complete car. The problem is to figure out where all components should sit and how they should be designed for its task. The car has an outer shell with a certain shape which limits the frame outer dimensions. When this work is clear so you can design a frame
which remains beneath the skin and keeps all parts in the right place, and that the parts can be assembled and disassembled without problems. This makes the tubes to be located, very much, in given positions. Therefore there is no room to freely design a framework, or according to a simplified strength model. In my view, the hard part of the frame structure is not the art of building a framework, but the art to understand the overall design of the car for it to cope with its task.

Of course there are things to think about in terms of "tube frame construction", for example, only using straight pipes and not curved. To see the different part of the frame as separate sections. In the case of torsional stiffness we can perform a full-scale test, measure the flexing areas and insert diagonal reinforcements.

Goran

I thought it might be of interest to show a picture of my car that shows a torsional reinforcement.The car has an original torsional stiffness of 6000Nm/dgr. It is a sheetmetal monocoque. I have mounted in an X reinforcement consisting of two (simple extruded) aluminum tubes that are 25mm in diameter, which appears in the picture. They are stuck with M6 screws. No welds, no Crm steel, but still increased the torsional strenght to 15000Nm/dgr. The Tq numbers are not exact but close enough to describe the difference.
I handed over to everyone to draw a conclusion from the above:

Goran

6000Nm/dgr to 15000Nm/dgr from 2 x long 25 mm tubes (that aren't even connected in the middle?)?

Love your car Goran, looked at your website many times, the work you have done is awesome and a great amount of respect given to you when I tell you my conclusion is grandiose claim.

I thank the warmest! I must add that it was very important that the tubes were cut to precise contact with the chassis, otherwise they bowed themselves. Only one tube could handle approximately 70% increase in torsional strenght. One can not achieve the same improvements for all cars since the Pantera has a wide open, and thus weak section at the rear. What I want to draw attention to is that if reinforcement and workmanship are performed correctly, it is more important than many other factors. In this case I loaded the chassis while study flexing. In such a test, I can study in detail the importance of the rainforcement placement and workmanship. As you mention, it's good to connect the tubes in half, but if one manages to get the forces to cut perfect centre in the tube, so reduces the importance of linking them. But when running with the car so I can imagine that there still is a great idea to connect the pipes.

I hope to be able to give the questioner some understanding from the "real life" about the dimension of importance of different factors. This is missing many times in books that are otherwise very well written.
Goran

My conclusion is the same as Cheapy's except to say, if the numbers are real, Panteras must have an extreme problem at that end of the car. To put it another way, the other 80% of the car's wheelbase must have been very stiff to start with.



Connecting the tubes will probably make little difference to the torsional stiffness but might increase the strength by reducing buckling tendency.

"My conclusion is the same as Cheapy's except to say, if the numbers are real, Panteras must have an extreme problem at that end of the car. To put it another way, the other 80% of the car's wheelbase must have been very stiff to start with. "

Yes, the car is pretty weak in original condition. I have got exact numbers together with that of a few other cars from the same era. The original Pantera GT-5 have fibre glass made rocker panel skirts, which I replaced with those made from sheet metal making the rocker panel box area very large. The original wheelrate is around 1,5 Hz so even with a pretty
weak body you will not notice very much of this during street driving. In fact even with my 3 Hz and racing shocks one must drive pretty hard to tell the difference. However, during the shock absorber testings I have been doing, driving over bumps while cornering, it is quite easy to observe much better shock setting response, at least at my level spring rate. This also shows in lap times.

"Connecting the tubes will probably make little difference to the torsional stiffness but might increase the strength by reducing buckling tendency."

Agreed.
Goran

Ok so lets establish that the Pantra has a big problem with left and right longitudal half connection even though both halves are extremely stiff, in my mind the forces required to buckle a 25mm aluminium tube are fairly low so the obvious next question is how did you measure it?

I have seen a number of FEA studies and the effects in stiffness each stage gives and I got to say that additions of your tubes are quite unique in what extra stiffness you have achieved in comparison - in the order of 10 fold.

Connecting the tubes near the middle will give extrat support from buckling under compression to at least one plane.

The rocker panels gives stability to the drivers section of the car. The front axle section is very short and stable.
So all of the worst flex take place in between the rear firewall and coilover mountings. Theese are right under where the
tubes are standing behind the engine.

I have no images handy from my Pantera twisting, but I have from the test of my Corvette chassis.
As you see the frame is resting on a stand, but this is only in rest, when lift is applied it is strapped down by
a wire (floating) bolted to the floor. The loading points by the wooden bar is the coilover mountings.
The same at the front but the bar is solid to the ground. The trick is not to get no bindings that is creating false
readings. Then there is scaled blub level each end of the chassis so you can read the difference in between
them.



In every strategical possition I had telescopic tubes with a spring inside with a scale on the side to read deflection.





My car finally got 15500 fp/dgr. In fact at that reading the front window get cracked.
Some reference numbers...
Lamborghini Countach 1900 fp/degree. Ferrari 360 spider 6250 fp/degree. Viper gts has 9000 fp/degree.
Panoz racing car tub carbonfibre monocoque has a stiffness of 45000 fp /degree

Goran

Hi Goran, how did you decide where to put the braces? Did you just measure how far likely locations sprung apart at a give load? By that I mean, once you had a list of candidates how did you decide which ones to pursue?

The reason I ask is that there is no simple way to decide (it isn't the one that moves the most that is necessarily most important).

Incidentally I would not be in any great hurry bolt the X tubes together. There is a good chance that they are actually working as tension struts, all you'll do is weaken them. The reason is that a strut under compression actually softens as it thinks about buckling.

Thanks for taking the time to explain your method, what did the Corvette chassis achieve?

Hi Greg!
In order to work as a twist locking device, the area of bracing have to be one that is flexing. The sheet metal monocoque is more difficult to handle than the tube frame of the race car, in the respect that the sheet metal monocoque has no exact "corners of force" like the tube frame. It is also weaker at the spot where the tubes should join. It is also a question of practical location, one must be able to get in and out the car, for example. However, the location of the tubes seen on the image was quite obvious to me, so I simply fitted a tube and looked what it did. No more than 1/2 hour of work. The engine area did also show, by compare, in the order of 4 times the distance of movement than that of other places. I had some flex in the door openings, but diagonaling here was not very helpful. The window area seemed more effective,(with the racecar in mind) but not a very great place to reinforce.

Using steel tubes with welded in treaded ends and heimjoints would create more centered force lines within the tubes, probably lessen the risk of buckling. In fact these aluminium tubes was an experiment and not really thought of as permanent stuff. And as everything I do on that car, it is only performed to see the outcome, an then drive to see the importance of those outcomes. In this case driving the car with a weak body equipped with standard springrates vs racing rates, then doing the same with a stiffer chassis.
So far I have seen the greatest influence for shock absorber settings.
I also find it interesting that these simple aluminium tubes gave this great result, since there is a lot discussions about tube diameters and selection of material. Well, tying to load even an aluminium tube takes some power, coud we stand on a match if possitioned perfect?
Goran

"Thanks for taking the time to explain your method, what did the Corvette chassis achieve?"

The Corvette had 19000Nm/dgr for the bare frame with the engine block mounted. Then there is a large flat floor mounted diagonaling the whole underside.
A steel firevall front and rear of the driver section. the car was newer up for final twisting, we might look at this during the winter.
Goran

Hi Goran,

That makes sense. Yes, bracing the windscreen is a good idea.

When the Test Lab used to do this they'd plot the deflection along the rockers/rails of the car every 200mm, and then have a think about any kinks in that line.

You can see in the example I posted that there are obvious parts of the ladder frame that are weaker than others. If only we didn't need an engine and a rear suspension!