Approximating the effects of a roll cage in a production car

[Crazy Bored]

I would like to model, and possibly do some FEA, on a roll cage that I've used in a Honda Civic that was transformed into a short track stock car. I am using Pro E on my university computers. I've made some sketches on paper, and simplified things to the point where I made 30 datum planes and modeled lines connecting their intersections. This allows me to easily extrude round tubes (1.5" with 0.095" wall thickness in my case) along the lines and curves I've modeled.

My real issue here is, what can I do about the unibody frame that this roll cage is attached to? Can I simply approximate it with simple tubes, or maybe rectangular extrusions? Technically of course, I can, but how much does this compromise the accuracy of a torsional stiffness analysis, for example? Is there an efficient way to do this, or am I wasting my time even trying?

How do car companies model the frames used in modern production cars? I have slightly introduced myself to the sheet metal application in Pro E. Is it worth investigating that? Or should I just abandon this idea, and instead model and analyze frames made entirely from tubing? I am mainly interested in this because it is a car I've raced for a couple years, and just made up a roll cage design from nothing and stuck it together with low quality welds.

The cage I am interested in modeling:


One of my primary questions is how useful (or useless!?) is it to connect the cage to the rear bumper?

The rear bumper is welded to the two main parts of the frame that extend all the way to the rear of the car:


I apologize for the lack of current pictures, but I ended up adding a tube between the shock towers, as well as a straight tube from the top of the main hoop to the front of the shock towers. You can see both features in some of the pictures. I'm at the point where I doubt adding any more material will help me. I'd like to have a better understanding of what I've actually done. It is a good place to experiment because it was a $300 car to begin with and competition is not very challenging. I will race this car for one more season and I want to make sure it's the best Civic that has ever raced on this track. Last year, with a completely stock 1.5L engine, my Civic was faster than some of the street stock cars at the track.

This is simply a fun side experiment, which I don't mind failing at. I'm just trying to gather information that may help me in my future, which hopefully will involve some sort of automotive design. My big question is, how useless/useful is the concept of extending the roll cage behind the rear shock mounts?

[Tim.Wright]

I'd say not so useful. Safety aside, the main job of the roll cage is to provide a stiff platform which connects the front and the rear suspension together. So basically you only need the roll cage starting at the rear suspension and ending at the front suspension. Going any further is only adding weight.

Regarding the stiffnes of the stock chassis, I've spoken to a V8 supercar engineer some years ago about this and he said they basically forget the stiffness of the chassis in their analysis. All of the forces should be going through the cage if its designed properly.

Basically the approach could be something like forget the body, attach everything to the cage, and then cut away as much of the body as the rules will allow to reduce weight.

[Greg Locock]

"How do car companies model the frames used in modern production cars? "

When i started FEA 30 years ago we built car bodies out of beam elements about 100 mm long. The joints were modelled as springs. This, basically, didn't work as estimating the stiffness of the joints such as A pillar to cantrail to header rail (corner of windscreen) is hugely complex. We tried modelling each joint and then coming up with an equivalent spring, at which point I waved goodbye to FEA and worked on more interesting stuff.

Now the guys build them build them at several resolutions, the coarsest of which is plate elements about 10mm across. These would be used for static stiffness modelling such as you are proposing, and linear dynamics. The finer resolution models are used for crash and durability work. The latter include very detailed models of each spotweld.

[gixxer_drew]

I'd say not so useful. Safety aside, the main job of the roll cage is to provide a stiff platform which connects the front and the rear suspension together. So basically you only need the roll cage starting at the rear suspension and ending at the front suspension. Going any further is only adding weight.

Regarding the stiffnes of the stock chassis, I've spoken to a V8 supercar engineer some years ago about this and he said they basically forget the stiffness of the chassis in their analysis. All of the forces should be going through the cage if its designed properly.

Basically the approach could be something like forget the body, attach everything to the cage, and then cut away as much of the body as the rules will allow to reduce weight.

I believe that would be the case with a V8, but I dont think so with all production based competition cars. In my experiences the rigidity of the original chassis and the details of how you tie it together is quite important... I would say though that its mostly concentrated in a few specific areas and the rest is throw away.

[Tim.Wright]

I think it would (should) be the case for any vehicle where a full cage is permitted to attach to the front and rear suspension mounts.

Of course if this isn't the case (like with a half cage) then you need to include the body effects.

Depends what you why to analyse it for though. If you want to see the effects of different bar placements, then you can forget the body because the trend will be the same with and without the body there.

If you are trying to find what the overall torsional stiffness is, then you will need a lot of time, a lot of meshing skills and possibly an accurate (or at least representative) body model.

[neilbah]

surely with the design like that you have no crumple zone and a rear shunt accident load would go straight into the frame. Its surely better to have the horizontal bar between the rear struts, certainly more conventional(re reading the original post it sounds like youve done this). i suppose if you were spaceframing the car and chopping out floor plans to change the car to rear engine for example then you'd want a sturdy rear section to bear load of the engine, body and bumpers etc but even then the cage structure doesnt extend that far normally. it would be good fun to have 2 or 3 shells and crash test them

[gixxer_drew]

I think it would (should) be the case for any vehicle where a full cage is permitted to attach to the front and rear suspension mounts.

Of course if this isn't the case (like with a half cage) then you need to include the body effects.

Depends what you why to analyse it for though. If you want to see the effects of different bar placements, then you can forget the body because the trend will be the same with and without the body there.

If you are trying to find what the overall torsional stiffness is, then you will need a lot of time, a lot of meshing skills and possibly an accurate (or at least representative) body model.

I dont disagree with you about FEA, however I have found big gains by doing exactly the opposite and seeking an integrated approach between the cage and chassis. I'll leave it at that though.

[Roland Ehnström]

I do not like the idea of attaching the roll cage to the rear bumper. First of all it just adds weight for no gain in chassis stiffness (anything behind the rear axle is just dead weight). Second, it looks very bad for safety.

[Crazy Bored]

Depends what you why to analyse it for though. If you want to see the effects of different bar placements, then you can forget the body because the trend will be the same with and without the body there.

This is what I was intending, and hoping for. So, for example, for the part of the frame that runs along the floor between the rear shock towers and the main hoop, I can just use a simple tube like I'm using everywhere else?

I do not like the idea of attaching the roll cage to the rear bumper. First of all it just adds weight for no gain in chassis stiffness (anything behind the rear axle is just dead weight). Second, it looks very bad for safety

I don't have any argument against this, I should probably remove it before this season starts. The large bar on the bumper is going to stay though, because the fuel cell is just ahead of it.

When I was questioning how to model modern cars, I was referring to something at this level of detail:


I certainly don't intend on trying that, I was just curious, especially about the parts that use a plate with complex bends and spot welded to other plates or structures. See the top of the B pillar in this picture. I guess you'd have to create each part individually and somehow get all the bends and angles right.

[Greg Locock]

Here's what i'd do.

1) find/estimate/guess the trsional stiffness of the body. 7000 lb ft/deg if nothing else

2) build a stick model of the main box and tube sections in the body, and some sort of representative structure out to your roll cage mounting points and the spring towers. use the same wall thickness for all the sections

3) Apply the correct constraints and forces to the spriong towers for a torsion test and check that nothing looks crazy in the deflected shape. Increase wall thickness as appropriate or add detail to suppress craziness. Try and find pictures of other FEA models in torsion to check what craziness looks like.

4) Multiply the wall thickness of all the sections by whatever factor you need to get the number in 1)

5) Start work on the roll cage.

Incidentally in a weekend you should be able to measure the real tosrional stiffness of your bodyshell, and the displacement along the rails.



I have heard wildly different estimates ofthe contribution of the roll cage and body to torsional stiffness, I suspect a lot depends on how well the fixtures between the two are managed.

[Jersey Tom]

Let's rewind and take a step back here real quick. First question: What are you trying to achieve or what question are you trying to answer by FEM?

You touch on this with, "How useless/useful is the concept of extending the roll cage behind the rear shock mounts?" How are you defining usefulness? Useful from a driver safety standpoint? From a handling / performance standpoint?

In the former case, probably not at all useful. Ultimately I think of it as you have a driver survival cell, and then you have everything else. Series rules typically lay out the survival cell aspect of it pretty clearly.

In the latter case... I'd agree with Tim and say not at all useful either. That's just fundamentals with no need for analytical models. As Tim gets at, your inertial loads (let's say they all originate at the CG) are reacted by your suspension pick ups, and that's that. Anything past that point is just dead weight (and having an extremely rigid structure in your crash zone would be the opposite of what you want from a safety standpoint).

Then there's the question of performance gain by having a rigid structure between front and rear suspension. May sound like blasphemy but I'd kinda say there's probably not much raw performance to be had there - that is assuming that your front and rear suspension (roll) stiffness isn't anywhere near that of the frame. All it is, is springs in series determining what proportion of inertial roll moment is reacted by the front and rear suspension. If the frame is made of slinkies welded together at the ends, then being the softest rate spring in the system your load transfer distribution (mechanical balance) is a function purely of that and your springs and bars don't do much of anything. If there's a decent amount of rigidity (I would have to think this is already the case if the center section is built stoutly from a survival standpoint) there you should fundamentally be able to achieve whatever mechanical balance you want... and making the thing stiffer and stiffer I think would very quickly reach and pass a point of diminishing returns.

In any event, for simplicity's sake I like Tim's idea of just ignoring the OEM body and looking at the effects of tube placement. Even that should be fairly intuitive from a standpoint of building trusses where you can... and assuming the tubes are straight the analysis becomes relatively straightforward with simple geometries. You could get crafty maybe and try to use parts of the existing body as shear panels, but even that is probably overkill for the series.

Really I see it as a question of.. you only have so much time / money to spend - where can you get the biggest bang for the buck? I would be a bit surprised if the frame were the limiting performance factor.. but one never knows!

[Crazy Bored]

Let's rewind and take a step back here real quick. First question: What are you trying to achieve or what question are you trying to answer by FEM?

Well, now the question is comparing alternative designs after I remove the tubing connected to the rear bumper. I won't even bother modeling that, it obviously has to go. For example, I would like to compare some basic configurations of the tubing connecting the top of the main hoop to the rear shock towers:



Another question would be if it is worth adding tubing from the shock towers to the bottom of the main hoop.

You touch on this with, "How useless/useful is the concept of extending the roll cage behind the rear shock mounts?" How are you defining usefulness? Useful from a driver safety standpoint? From a handling / performance standpoint?

How much does it add to the stiffness? I realize the the question of how much does stiffness influence handing/performance is a whole other thing. I am not looking for any absolute answers, only relative ones.

... and making the thing stiffer and stiffer I think would very quickly reach and pass a point of diminishing returns.

I understand. This car was purchased in a state where it was unsafe for driving on the roads. Being in Canada, the salty roads have rusted away quite a bit of material. I had to replace the floor on my side because I could fit both of my feet through the hole there.

How stiff does this look to you? I actually wondered if it was better to get a stiffer car with a good frame, or one that was rusted out so it was lighter.

This section used to basically be a rectangular beam:

Really I see it as a question of.. you only have so much time / money to spend - where can you get the biggest bang for the buck? I would be a bit surprised if the frame were the limiting performance factor.. but one never knows!

I am sure it is not the limiting factor, I have found way more performance from moving suspension mount points to increase camber gain in the rear (limited to -1 degree of static camber in the rules). Going from ~58% front weight to 55.3% also seemed to help things. Stiffer/new bushings made it more predictable to drive. I am doing this particular thing mainly for fun, and hopefully to learn basic concepts and techniques that I can carry on to more important projects in the future.

[zonk]

This could help You

[zonk]

http://www.rallyandracing.pl/gallery/Ra ... da/22.html

[Jersey Tom]

Well, now the question is comparing alternative designs after I remove the tubing connected to the rear bumper. I won't even bother modeling that, it obviously has to go. For example, I would like to compare some basic configurations of the tubing connecting the top of the main hoop to the rear shock towers:

I see. You may be able to infer some of the performance of those configurations just from visual inspection. Ultimately it comes down to (a) what the applied load is and (b) how well that load can be resolved through tension / compression members (we'll assume that the joints are pinned rather than actually welded, which I think gives more insight anyway).

If it's a torque about the SAE x-axis we're talking about (i.e. roll torque), I don't believe two straight back bars can do anything for you. That's the easy one to eliminate right off the bat. Could make similar fundamental assessments of how well the rest of those configurations can resolve a number of different load cases, again mostly by visual inspection.

Biggest thing is to not over-complicate it. It's just like static class...