Roll bar materials

edited January 2015 in IET Formula 24
How should we make our roll hoops?
Now that filler & streamlining are being banned, there is suddenly an unpleasant compromise to be made between aerodynamics and safety.
It seems inevitable that the rules will eventually stipulate exactly what tube must be used (diameter, wall thickness, material etc).
Independently of the current furore over this, CAUC have been setting up a series of material tests for rollbars (we should be ready to do some violently destructive testing this coming week :) but as the subject is of current interest it seemed like a good idea to get ideas from the forum before destroying all our test structures.

First up - what do the professionals say....?? here's a quote from FIA for sports/rally car roll structures.

FIA Appendix J 2009, Article 253, p14

8.3.3 Material specifications
Only tubes with a circular section are authorised.

• Minimum Material: Cold Drawn Seamless Carbon Steel
• Minimum Yield Strength: 350 N/mm2
• Minimum Dimensions (Ø mm):
•45x2.5 or 50x2.0 (main rollbar or lateral rollbars)
•38x2.5 or 40x2.0 (lateral half-rollbars and other parts of the safety cage)

The tubing must be bent by a cold working process and the centreline bend radius
must be at least three times the tube diameter. If the tubing is ovalised during bending, the ratio of minor to major diameter must be 0.9 or greater.

The above is for cars exceeding a tonne at speeds of over 100mph, and which might fly down ravines (any number of these incidents on youtube) so our requirements are considerably less. The structure generally has to withstand a car rolling over, not landing upside down from a great height.
Furthermore, I remember the old MSA blue book specifying that the roll structure inside an amateur rally car should withstand the weight of the vehicle on any node without taking a permanent 'set' - that's the weight of the vehicle, not 2 or 3 times it.

OK, back to details. FIA, MSA both say round tube. not square. Both say use cold drawn seamless carbon steel (that's DOM in USA).

The 'impact' loading on a structure can be described in 3 ways:
1) static loading. For us, let's say the car weighs 150kg. Nobody is going to argue against static load capability of 1500N
2) Impulse test; collision imparts momentum. M*v = force integrated wrt time. I'm going to assume a free fall of 300mm of the complete 150kg car, that's 364Ns
3) energy of collision: in our 300mm case that's m*v^2/2 = m*g*h = 441J

As this is clearly a low speed but high momentum impact, our test is designed to get as close as we can by dropping a large weight from a low height. We have procured an old 56lb weight (that's about 25kg) and I've made a release mechanism to drop it in a controlled way. We've made a massively strong jig into which sample rollbars can be fixed by wedges, and we'll do the tests on the concrete base of some fire escape steps round the back of the school.
If we drop the weight 1m, this would be an impulse of 111Ns and energy = 246J
If we drop the weight 2m, this would be an impulse of 157Ns and energy = 492J

I expect we should be able to do the drops accurately and consistently at 2m.
We have 4 test rollbars made from pieces of tube I was able to beg/borrow/steal, I'll give details of these next installment (one of then is electrical conduit - I don't expect the results to be pretty)

Please give me your thoughts on the validity of what we're trying to do & anything you would change. And if you want a test doing on a sample rollbar PLEASE send us one!!!

I might make a couple of solid rollbars just to see how they behave... that must be best for aero for a given level of strength...

Final thought - I actually want the rollbar to deflect to some degree, I want it to absorb the energy of the impact, rather than pass it directly to the chassis. Less than 50mm though.... :)


  • Wow, Bob, good proactive stuff! I'm still spitting feathers over this roll bar nonsense, having to destroy a perfectly beautiful piece of engineering for no apparent good reason. We only use aluminium tubing for our roll bars......are you planning on using that as well as the steel stuff? If I can get a new ones made rather than set the angle grinder to the existing ones (3 of them!) I'd be happy to let you have one to see how it fares although it will have the so called 'fairing' could lop it off with said angle grinder!!!

  • That sounds great BobC, it would be good to get some real figures. I assume you will be hitting the top of the rollbar. Which ways will you hit it and how far down will be its chassis fixing points ?
    When RR8 rolled many years ago at 30 MPH, the roll-bar (25.4mm 1.62mm steel tube, seemed) bent forward at the top by about 25mm. Unfortunately no idea which ways it got hit. I guess in a roll the friction against the ground will have a large part to play in if the roll-bar bends and fails. If this digs into soft ground then there could be a high forward, backward or sideways impact. Probably for most structures the front-rear bending would be the worst ? We might have our first pass roll-bar for RR9 which did not come up to scratch still around. if so you can have that one. Will have a look at next our GP session to see if is still around.
    Would be interesting to get some mechanical engineers input.
  • I have always felt that roll bars should be made from 3/4" x 2mm steel or 1" x 3mm ali. However careful consideration should be made as to the grade of material being used. This would actually be quite an interesting thing for one or a couple of students to look at. Investigate the weight benefits of steel v ali then look into the various British standards that govern tube grades and decide which is best for a given cost.
  • As long as the rollbar is fit for purpose, I can't see what it's made of makes any difference. BobC's thoughts are excellent, relevant and pioneering, as always. Most organizations don't have the ability or facilities to think about stuff like this. Looks as if BobC's work could become the Greenpower standard...should they choose to adopt it.
  • edited January 2015
    Hi Bob,
    I've put my Mechanical engineering hat back on after 14 years which is when I last ran any Finite Element Analysis (FEA) models. Things have certainly got a lot easier to use! When I last did this sort of work for my dissertation I had to run multiple UNIX machines overnight to generate solutions. My desktop PC is now able to spit out solutions in a few minutes!

    I'm using Autodesk Simulation Mechanical 2015 which is available free to educators and students
    (Sorry Mike - Solidedge doesn't seem to make their FEA tools available to students...). Within this package it's possible to set up a drop test where the part is dropped a height of your choosing and the resultant stresses are resolved. At present I've only run a vertical drop with no initial velocity, although it is possible to simulate a more realistic impact by specifying initial velocities along the xyz coordinate system.

    I've run two scenario's to help validate your test method:

    1) A mass of 150kg dropping vertically onto the roll hoop from a height of 300mm
    2) A mass of 25kg dropping vertically onto the roll hoop from a height of 1750mm.

    Because the simulation requires dropping the part itself rather than dropping a weight onto it, I've modeled a lump onto the base of the roll hoop to bring the part up to the appropriate mass:


    This is the 25kg mass. The roll hoop itself is 25.4mm diameter 2mm wall thickness. The radius of curvature of the hoop itself is 70mm through the centreline of the tube.

    The 150kg model is just a bigger block on the bottom.

    For setup 1, the following Von Mises stress ( is produced:


    The peak stress here is 2210 N/mm2 - This is way beyond the yield strength of 4130 alloy steel (around 435N/mm2), and approaching 10 times the yield strength of mild steel.

    For setup 2, the following Von Mises stress is produced:


    The peak stress is 2215 N/mm2. As far as your testing is concerned, your method seems valid although 2 metres may be a little high for the drop.
    Of course a computer simulation is only an approximation, but I'm already thinking that we are going to not run anything less than a quality alloy steel for our roll hoop from now on.

    I do think perhaps a number of teams should look a little more closely at their hoops, and rather than whinging on about the loss of fairings (it is pretty obvious why this is happening isnt' it?), perhaps there is a great deal more to understand about the engineering of safety into our cars.

    I'll post a few more model results up...

  • Continued...

    The effect of adding a tie bar at 200mm below the top of the hoop:


    Stress is reduced from 2215 N/mm2 to 1593 N/mm2 - heading in the right direction...

  • edited January 2015

    The effect of increasing wall thickness from 2mm to 2.64mm, all other variables kept constant (height of drop 1750mm, alloy steel etc...)


    The stress has decreased from 1593 N/mm2 to 675 N/mm2. This should now be within the elastic limit of a high quality alloy steel, meaning that there shouldn't be any permanent deformation of the structure given the impact. This would still present problems for standard mild steel tube however.

    What's the kit car roll hoop dimensions and steel grade?

  • Continued...

    A further increase in wall thickness of the tube up to 3.05mm, all other variables are constant...


    Stress falls again as expected to 551 N/mm2.

  • And some more - a comparison between a typical alloy steel (4130) and an aluminium alloy (6061). Both experiencing the same loading, dropped from 1750mm, and the same dimensions of tube - diameter 25.4mm, wall thickness 3.25mm (10 SWG)

    First steel:

    Peak Von Mises stress is 527 N/mm2.


    Peak Von Mises stress is 726 N/mm2. The yield strength of un-tempered 6061 aluminium alloy is just 55N/mm2.

    We will be removing the aluminium roll hoop currently installed in RLR2!!

  • Very impressive presentation and thanks for that. How do you determine the loading to use? Would a GP car normally or even in extremis experience the type of loadings you have used?


    No doubt you are fully aware that I'm one of the ones 'whinging on' about fairings and I'm afraid it isn't obvious to me why they have been banned. I don't think anyone has stated why unless I missed something.
  • Hi Chris - I'm as guilty as anyone of having a whinge over various issues, please don't be offended! It says in the regs the aim is to introduce a small aerodynamic inefficiency, which according to Terry's analysis is exactly what would happen. This would slow the cars down a bit, which is the aim of the regulation change.
    The loadings used in the simulation are in line with BobC's test that he is carrying out - which is to simulate flipping the (150kg) car and driver over and dropping it from 300mm onto the top of the roll hoop. The first 2 models confirm the equivalence of dropping 25kg from a height of 1750mm with dropping 150kg from a height of 300mm.
    This seems reasonable in my view - a car losing it at 30 mph and tripping over its tyres could launch itself into the air sufficiently to gain 300mm of air between roll hoop and tarmac.
    The simulations can be setup to include an initial velocity in the x and z axis components but I've not tried this yet.
    We've been running an aluminium roll hoop with a diameter of 25.4mm and 3.25mm wall thickness for the last 5 years, but we will be chopping it out and bolting in an alloy steel hoop of the same section instead following the simulations that I've run here.
  • Quick progress note -
    first - thanks Ben, that FEA work is fab, and I'm pleased to see that the 25kg for 2m drop is reasonably similar to the whole car for a foot drop - reassuring!
    We had fun doing destructive testing in car club last night. Our sample rollbars were
    1) 19mm diam 1.4mm wall ERW
    2) 25.4mm diam 1.4mm wall ERW
    3) 20mm diam 2.0mm wall electric conduit
    4) 25.4mm diam 3.2mm wall CDS
    we actually did drop about 1.7m. The rollbar samples were a U shape 230mm across (centres) and 250mm high. We filmed the procedure but I don't know if it came out - it was outside in the dark....
    1) and 2) were damaged by the impact, the shape of the bar changed and parts started to collapse. A severe indentation happened at the point of impact. On the other hand both retained considerably more strength than would be needed to support a static load of 150kg and the deformed shape only moved 1cm or so.
    3) was surprisingly good IMHO, an indentation occurred at the point of impact but the general shape remained unchanged.
    4) on this you have to look quite carefully to see where the weight hit it. The alarming thing about this impact was the noise it made - it was clear that the impact was transmitted straight to the concrete floor with no attenuation in the rollbar itself, In a car this material would not fail but whatever it was connected to would for sure.....
    Finally, if you search ont' net for rollbar specifications, the F1 FIA specs are quite easy to find. Those guys only specify static loading tests (9 tonnes vertically & 5 or 6 tonnes sideways or fore-aft). Now an F1 car is 500kg or so, & those guys frequently get quite a lot of air during a (200mph) 'incident' so I'm surprised that they only have static tests.
    I did have a piece of ally tube but I couldn't cold form it without breakage. The other issue with ally is that there are SO many alloys with such different properties.
    Hopefully we'll get one of the girls to write up the experiment & put it on our website. I'd be tempted to recommend 20x2 CDS although I must say that the quality of the ERW tubes does seem to be really good these days,
  • edited January 2015
    further note - the real results seem to be better than the simulation predictions. It seems likely to me that this is due to the fact that in real life the tube actually gets indented at the point of impact and this soaks up the impact energy as the metal yields (a mini crumple zone if you like). I think the simulations carry on assuming the metal will be elastic....
    this is the 25.4 x 1.4 one. see the impact zone (near camera) and a bit of tubular squashage on the inside further away. Note the weld (put on the neutral zone when bending) seam behaves just like the rest of the tube wall, even beyond the yield point.
    (ps I give up - how do you post pictures....?)
  • edited January 2015
    I've just spent 10 minutes trying to post the code but all you will see is image

  • I think the simulations may be closer than one might think. To permanently deform the metal requires stresses greater than yield stress, which are certainly predicted by the simulations. In the case of your number 4 test Bob, the stresses would have peaked below yield, so remained within the elastic limit. Effectively, you could dribble the car upside down on it's roll hoop like a basket ball and it would be unaffected. However, in the image above (test 2), I would suggest that any more than one impact would be undesirable - a double rollover? Probably very unlikely, but test 4 would cope providing it didn't punch through the floor whereas test 2 would not. The wording in the regulations is very loose as far as the construction of the roll hoop is concerned. Should there be a static loading requirement specified?
  • I think the tests you have both done are rather fantastic and hope they are useful to teams up and down the grid (may even have inspired me to do some FEA, which I hate doing ;) ). I wouldn't dream of using aluminium that wasn't tempered. I was thinking 6082-T6 or 2014A-T6.

    Might be able to shed some light on why there are only static roll bar tests in F1, rightly or wrongly generally in motorsport safety requirements the car is not expected to dear with "worst case scenarios" to give you another example F1 front and rear crash structures are only required to give "survivable" forces and energy dissipation at 15 m/s (about 30 mph) a speed F1 cars barely go at...

    it may also be worth looking at deformation of the roll bar under expected forces as of course some deformation would be acceptable in the most severe load cases, but i'm not sure if autodesk can show that. Might be something I look at in ANSYS over lunch tomorrow...
  • edited January 2015
    Yes, some excellent work and info. Is it possible to do some tests with sideways and front to rear loading at the top of the rollbar ? I think these would be interesting.
    BobC looks like our old roll-bar was lost in a clean up sometime.

    Are there any simple tutorials/info on using Autodesk Simulation Mechanical 2015 in this sort of way.
  • Hi Terry,
    Yes it can model static loads in any direction, or through the drop test you can give the hoop initial velocities in any direction before it begins its drop.
    There are very useful setup video guides which are easy to follow. The drop test can be found under the 'non linear MES' simulation setup.
    In a nutshell:
    1) I use Solidworks to create the models themselves (sorry Mike, probably said a rude word)
    2) Start the Simulation mechanical package and create a new file.
    3) Open the Solidworks model
    4) Mesh the model (their is an automesh function which is quick and easy, but not the most efficient).
    5) Setup the load constraints for a static test, or run the 'drop test' wizzard.
    6) Run the simulation
    7) Make a cup of tea
    8) View the results - stress, displacement etc - animate them

    Compared to running Ansys 14 years ago it's a doddle.


    P.S I can run more models after school today (yesterday I was home with the lurgy so I got loads done!)
  • quick comment regarding static load tests. There are 2 obvious advantages:
    1) it's not a destructive test (on a thing that passes)
    2) it is an easy, safe, controlled and repeatable test to do
    Neither of which is really an excuse to compromise on safety considering the famously deep pockets in formula one; A rollbar made of a stiff but brittle material would pass their test but just shatter if it were needed....
    With a big strong lever I can easily put 1/2 a tonne of force into an object & easily measure deflections to 0.1mm accuracy.
  • Here's a static load test of 10000N acting backwards at the top of the hoop. The hoop is triangulated with a strut according to Greenpower regs no more than 200mm from the top. The tube is alloy steel - 25.4mm diameter and 2.64mm wall thickness. The green arrows show the 10000N load. The base is restricted in all degrees of freedom. The little orange arrow is from a drop test and does not figure in this model's setup.

    The results:

    The peak stress is a very comfortable 160N/mm2 which is well within the yield stress of most steel, particularly alloy steel. I'll do a sideways load next...

  • edited January 2015
    Next - pure sideways loading across the roll hoop top, all dimensions are as before. The base is fixed in all degrees of freedom.

    The height of the roll hoop from base is 712mm. Results are as follows...
    Again I'm plotting Von Mises stress. The values are clearly higher and focus at the base of the support leg. The peak stress at this point is 699 N/mm2 which will be very close or in excess of yield for a typical alloy steel. A standard mild steel would likely collapse as would most Aluminium tubes of this section. The taller the hoop from it's supporting base the greater this effect would be.
    Remember this is a tonne of load acting sideways on the hoop. Is this realistic?
    There are many design changes that can alleviate this stress. A triangulated brace between the legs of the hoop; supporting struts outside the hoop etc...

    If Greenpower wish to tighten up the roll hoop specifications, then a specification is what is needed. What loads is the roll hoop expected to support? Should teams submit an FEA of their design? Given the free availability of the software (and it really is very easy to use), then this would also be a superb learning opportunity for all involved.
    The alternative would be to specify minimum tube requirements beyond 'steel or aluminium'. Diameter, wall thickness, minimum bend radius (3x diameter as mentioned above), seamless tube, British standard xxxx etc... Either way a spec is really needed otherwise people end up shooting in the dark which leads to confusion and unease. A good example of Greenpower being more specific is the brake test for 2015 - everyone knows where they stand.

  • I'm just about to have new hoops fabricated by a kind sponsor. Our current hoops are Aluminium 25.5mm OD x 3.4mm wall. Should we now be considering alloy steel instead and what would you guys be recommending? From the above tests I would assume a similar O.D. of 25.5mm (1") but what wall thickness? I'm not that familiar with the different grades of tubing so what is alloy steel? I'm guessing its' a mix! Thanks for any help.

  • edited February 2015
    Hi Chris,

    Without any further guidance from Greenpower I can't second guess what they're thinking, but we're making our new hoop from 4130 Chromoly seamless steel tube. This is what is used in the steel framed bike industry. This 'alloy steel' is certainly a lot stronger than a standard 'carbon steel'. The hoop radius will be a factor in your decision - the tighter the radius the higher the stress, so the FIA guidance Bob gave above is sensible - 3 x tube diameter as a minimum. As for wall thickness, a steel bike frame may use 1.24mm wall thickness. Bob's photo above is of greater thickness but a poorer quality tube. I think a wall thickness of 2mm would be more than adequate providing the steel is of a high quality. Those wanting to go thinner to offset the weight should do their own modelling and FEA!

  • edited February 2015
    If this is to be a useful exercise, we need to know what is a reasonable static loading?

    I think that given the all up weight of the car a load of 5000N vertically (1/2 tonne), and 2500N across and 2500N back would be a fair and significant step forwards in terms of defining the strength of the roll hoop, rather than 'it must be steel or aluminium' which is quite vague.

    What does this look like on FEA?

    The following are models of a 25.4mm diameter 1.64mm wall thickness alloy steel tube. The yield stress of 4130 Chromoly steel is 435 N/mm2. The new regs do state that the roll hoop "must be capable of taking loading in all directions." The diagonal strut beneath the hoop in this image would in my view satisfy this new wording in the regs. the top of the cross tie is 200mm from the very top of the hoop. The radius of hoop curvature is 76.2mm (3 x diameter).

    5000N Vertical load:


    2500N Cross load:


    2500N Backwards load:


    If there are any others that would like to put in their thoughts as to what is a reasonable static load then please do. I feel I should say that my thoughts and modelling work here are not a guarantee of safety, but I personally would feel happy sending out my students with the above design to the spec I have outlined above.


  • Chris, it feels uncomfortable advising somebody else on what to do, particularly when the roll bar is just one factor in the crashing system. I did the impact tests because it's a dynamic situation & I was genuinely interested to know what happened. Static testing is easy (even doing it for real) but sort of sidesteps the issue. The rollbar only has to take a static load of 1G - that's (say) 1500N. In a crash much higher loads are seen but they are very short duration. A roll bar that is strong enough to take an impact without yielding is pointless unless you have a similarly strong chassis beneath it (are the bits of wood making a "trug" chassis that strong?).
    It is easy to look at Ben's simulation results & conclude that no part of the structure should exceed the yield stress. This too is unpractical. A rollbar will impact on tarmac. There will be a point of impact and this point will exceed yield stress and be deformed. The structure must and will deform to some degree and this local deformation absorbs the energy of the impact so that lower loads are transmitted to the rest of the structure.
    I doubt that the autodesk inventor FEA models the impact dynamics past the yield point - this sounds like very expensive research type simulation, not giveaway amateur FEA.
    So I think it's OK and inevitable for small areas of the structure to exceed yield stress, and this is backed up by the excellent performance of even the most weedy test rollbar (19mm diameter and 1.4mm wall). Sure it got bent and squashed a bit, but nothing moved more than a centimetre or so & it would have done the job on a car (based on the equivalence under FEA of my drop test & a car falling a foot). However I think the Ally rollbar FEA showing almost all of the rollbar structure exceeding yield stress & peaking at 14x that - is really not up to scratch.
    I was thinking that 20mm x 2mmwall steel tube would be suitable. This can be bent to radius of about 80mm using an electrician's conduit bender. This bend radius is not so tight as to weaken the tube unduly by ovalising it, and not so large radius that the span of the bend means that the arch is weakened. I'd probably get CDS tube for the job even though the tests showed that modern ERW tube (at least the stuff I used) is of excellent and reliable quality.
    I must re-iterate that the most alarming test we performed was of the strongest rollbar!!
    As it might take some time to get the girls to write this up on the CAUC website, I'll get some more pictures over the coming day or 2.
    Ben - think about point loading? (when I FEA'd Zebedee's rollbar way back in 2007, I put 1/2 a ton onto it downwards and forwards onto a small square block in the midde of the arch - here's an ancient picture - the red bits are over yield stress...

  • Bob, I would expect that those high stresses are due to how FEA works with point loads. I won't go into details but it's fair to assume 2 things when working with FEA for roll bars. 1. Point loading creates unrealistically high stresses. 2. Modelling of round sections is poor. Usually when modelling something like a tubular chassis it's done using 2D FEA to counter these problems.
  • I agree Bob that these FEA models are not perfect - the setup on autodesk allows picking of surfaces to apply the force. Yes I could model a small surface on top of the hoop to better replicate point loads, but these tend to produce stress concentration factors that are unusual.
    The complexities of a real crash may produce a moving point load as the surface of the hoop rolls over the tarmac, or if the hoop digs into a soft grass verge then the load will be more spread out - who knows? What about the car sliding on the hoop - another loading condition that is difficult to model. At least the static loading condition is reasonably repeatable, but an agreement is needed with respect to point loading or surface loading.

    I think the main lesson to take from these exercises so far are:

    1) Aluminium shouldn't really be used.
    2) A seamless high strength steel is probably best - not just any old rubbish.

    What would be really useful is if a large engineering business with interest in Greenpower could dedicate some resources to providing some clear answers to the question of what are suitable roll hoop dimensions and materials for our cars.
  • Bob, the rollbar on the trug chassis is secured at the base by bolting through the rear bulkhead and also clamped to the top part of the bulkhead. Wouldn't like to say how good that is but seems substantial enough! However, as we now have to make new ones I've decided to fix the base of the hoop to the aluminium sub frame (shortening the verticals) which itself is bolted down at four points to the timber chassis. This will also allow us to remove the whole sub frame plus roll bar in one piece which we have to do just to change a tyre! (no stub axles).

    Very informative thread and I for one much appreciate every ones's input.

  • edited February 2015
    I'll just have another go at making a point (because I have been doing it really badly :).
    The FEA analysis using von mises stress is intended to be used to determine whether a part undergoing regular cycles or continuous stress will fail in use. In this case "failure" means that the part has taken on a permanent deformation as a result of its duty, and will thus eventually fatigue & snap.
    That's not how a rollbar is used. I think the analysis above is excellent and informative, but if you designed a rollbar to survive multiple accidents undamaged, it would be disproportionately strong compared to the rest of the structure.
    To my mind this is one reason why they have the rounded shape - it is a bad shape from a structural view (i.e. not a triangle) - it is designed to deform, yield and thus absorb the energy of the impact. After the crash the damaged rollbar is replaced, and the rest of the chassis is undamaged and can be re-used.
    Sorry to go on, I was just a bit worried that some might make stupendously strong rollbars, which would be pristine among the shattered remnants of the rest of their car after a roll :)
Sign In or Register to comment.