Torsional stiffness : Tube Steel Spaceframe

I am interested to know if anyone has done any work on chassis torsional stiffness.

A finite element model that I created indicates 1907Nm/deg (1406lbft/deg) for an unclad spaceframe.

I intend to progress the model with cladding. Riveted aluminium appears a standard practice.

I cannot really predict the added contribution of cladding; but 1907Nm/deg (1406lbft/deg) is a very low torsional stiffness.

RHS used in the model : 38*38*2 (1.5"*1.5"*0.080") (Red) and 25*25*2 (1"*1"*0.080") (Blue).
 

Attachments

  • Baseline_Spaceframe_Torsional_Stiffness.jpg
    Baseline_Spaceframe_Torsional_Stiffness.jpg
    134.3 KB · Views: 417
  • Baseline_Spaceframe_Torsional_Stiffness_02.jpg
    Baseline_Spaceframe_Torsional_Stiffness_02.jpg
    174.9 KB · Views: 741
Clayton i would say at a guess that was cladded, i'm waiting for the results of the first chassis i did, its being tested to ADR regs by a registered certifier, so will be accurate. will post up the results when they come to hand.

cheers Kaspa
 
Mark,

does FEA indicate where the flexion is happening? Might be that a significant chunk of it is happening in a relatively small area, which would be the focus of any efforts to stiffen the chassis.
 
Hi Mark,

Back in 2007, I asked a chassis engineer who helped design the Saleen S7 (tube frame w/ AL honeycomb cladding/stiffening panels) what sort of torsional stiffness increase would result from bonding/riveting the honeycomb panels to the tube frame.

He answered that it depended on how good the uncladded tube frame chassis is, and that one could expect an up to 20% increase with the panels.

I am interested to hear how much your bare tube frame improves with aluminum paneling attached.

Jack
 
Mark,

does FEA indicate where the flexion is happening? Might be that a significant chunk of it is happening in a relatively small area, which would be the focus of any efforts to stiffen the chassis.

yes Tom, it can indicate which individual tubes are flexing a lot in which case you could increase tube size or add extra braces.

It can also indicate where tubes are lightly loaded. Indicating that the tube could be optimised to make it smaller or re orientate it to perform better.
 
My last post was on this subject, back in 2009. Yes, I'm a lurker more than a participant these days as my main project is an SN95 Mustang Autocross / Open Track car.

Anyway, I've watched the 'spaceframe' chassis conversation for years and there has been a lot of mis-information written on it. My first kit car was a Sylva Striker Mk.II, a 'spaceframe' Lotus 7esque machine.

However, it wasn't really a spaceframe, but semi-spaceframe. It is virtually impossible to build a true spaceframe that you can fit a human in.

In a true spaceframe, all the tubes are in tension/compression. Should it be assembled with a ball joint at every tube join, the chassis would still retain it's shape and structural integrity. However, that requires 3-D triangulation of the whole frame, something never achieved in the real world (well, rarely anyway).

In the case of a semi-spaceframe, we rely on sheer panels to transfer some loads in non-triangulated structure and on the beam strength of some tubes where open areas prevent triangulation (cockpits/engine bays etc). In the case of my Sylva or FFR's Mk.IV a substantial backbone structure stiffens up the cockpit section.

All GT40 'spaceframe' chassis are actually semi-spaceframes. Analysing their stiffness without sheer panels attached is largely pointless, the chassis is a system that needs to be analysed in the structural form that will see the actual loads of road/track use.

In the case of my Striker, I bonded the panelling to the frame, in addition to riveting. This resulted in a fairly stiff frame, but likely only in the 1500-2000 ft.lb/deg region. Unpanelled, the same frame is probably only in the 800-1000 ft.lb/deg bracket (estimated).

That may seem fairly 'wet noodle' for a car that dominated 750 Motor Club kit car racing in the UK for over a decade, but absolute stiffness figures don't mean a lot.

What is far more important is the 'Chassis Torsional Stiffness Ratio' (CTSR).

CTSR = 'Chassis Torsional Stiffness' / (Front roll resistance + Rear roll resistance)

Tunable handling comes from minimising the 'sprung mass elastic weight transfer' (SMEWT). As you increase chassis stiffness, the SMEWT reduces as an exponential decay function.

Long story short, with a low CTSR (wet noodle), changing spring and/or anti-roll bar (sway bar) stiffness has little effect on the understeer/oversteer behaviour of a vehicle under lateral G loads. A stiff chassis allows front and rear roll rates to be tuned to produce the handling characteristics desired. A CTSR in excess of 4 is generally considered to be needed for acceptable tunability for competition at the clubman level of motorsport (F1 and comparable standards will be higher) for single seaters.

When using FEA, the problem is that real world CTSR figures for semi-spaceframes tend to be only 70-80% ish of FEA calculated values (the thicker the tubing wall, the closer the FEA figure to the actual, or where square tubing is joined at 90deg in at least one plane). This is because the tubing joints don't occur at a single point, but around the circumference of the welded tube. The material around the weld is not as stiff as a solid joint (and we don't want to incorporate a solid steel block at every tubing joint) so the theoretical stiffness of the tube in compression/tension is not the same as the stiffness of the joint-tube-joint system. Therefore, a CTSR value of 6 should be the goal at the design stage.

Where the single seaters (non-aerodynamic downforce) and Lotus 7 style roadsters have an advantage is that their vertical Centre of Gravity (V.CofG) is only just slightly above their roll centre heights. Therefore, adequate roll control (tyres/springs/shocks/anti-roll bars) and lateral grip is achieved with relatively soft suspension settings. Therefore, it doesn't take a particularly stiff chassis to achieve a CTSR of 4 or more. Hence, you have a Sylva Striker or Caterham 7 that will lap multiple seconds per lap faster (on a track dominated by corners, not straights) than a Cobra with double the chassis stiffness.

Additionally, the vehicle's intended use comes into play. A road car will be set up with a more compliant ride than a track car. Roll stiffness will be less than a track car and in proportion to the car's weight. Therefore, the CTSR of a road setup could be over 4, but the same car set up to run on the track, with stiffer springs/shocks/tyres/anti-roll bars, could end up with a CTSR of less than half that. The chassis stiffness doesn't change, but the roll stiffness does.

The nice handling road car becomes an evil handling pig on the circuit! In some cases of marginal chassis stiffness, the same car will produce faster lap times (due to more predictable handling) with a softer setup because the chassis isn't stiff enough to capitalise on the greater roll stiffness in the suspension.

For a lightweight low vertical CofG car (Lotus 7 style roadster), as a very rough rule-of-thumb, chassis stiffness in ft.lbs/deg should equal the gross vehicle weight in lbs. For a clubman level competition car, it should be double the GVW. My Striker weighed in at a bit over 1100lbs so a stiffness of just under 1500-2000 ft.lbs/deg was adequate for a road/track car.

A 'best practice' design process should follow these steps:

Estimate the GVW
Decide on roll rates/stiffness (road or track, roll centre vs V.CofG, etc)
Determine minimum stiffness to give a CTSR >6 (FEA)
Design chassis to achieve this with minimum weight
Build it and test it
Modify as required

So, back to Mark's original post, if you can't adequately model the panelling, add in triangulation in the areas that will be panelled (simulating a sheer panel for load purposes). See where you are then.

For a replica GT40, 5000 ft.lbs/deg (6779 Nm/deg) would probably be adequate ('69 Lola T70 IIIB hit this figure). Even 3500 ft.lbs/deg would probably suffice for a purely road going build as the Lola T70 Spyder did well enough at 3200 ft.lbs/deg (but then, who really builds a GT40 with that goal?).

For a competition car, I'd want to aim for the 10,000 figure, but not at the expense of too much additional weight.

As a point of reference, Factory Five's Mk.IV chassis delivers somewhere in the region of 3300 ft.lbs/deg (many estimates out there but there have been some good estimations using FEA). Much of the flex in that chassis occurs in the engine bay area. If you don't want doors, their spec racer achieves double that. David Borden's Mk.IV was extensively modified (using Solidworks to drive the modifications, google for several write-ups on the net) to achieve a similar stiffness in a car that has working doors. The FFR spec racer has proven to be a pretty capable track car and I'm sure David's would hold it's own with most modern sports cars in the same weight/hp bracket.

I hope I haven't thrown too many people off with this long post but I've read every 'torsional rigidity' threat on GT40s.com, all the FFR forums, most of the late model Mustang forums and most of corner-carvers.com and, with the exception of the latter, there is much more opinion available than real engineering.
 
Having re-read my post, I should add a couple of points for clarification.

1. The 2x lbs to ft.lbs/deg figure is a rough estimate only and, may not give a CTSR of 4+ for all setups. My 'aim' point of 10000 ft.lbs/deg would mean a GVW:stiffness ratio of 1:4, double the rule-of-thumb. Many building clubman competition cars are using the 4x figure as a rough aiming point. Time will tell whether a SylvLocCaterField with 5000 ft.lbs/deg shows significant competitive advantage over the 1500-2500 ft.lbs/deg that the unmodified variants offer.

I think that they'll find their variation in driving ability and suspension tuning ability will have far greater effect than their effort to create a stiffer chassis, beyond a certain point. A look at the BMW M3 and M5 stable over the last 20 years (often significantly less stiff than their 'competition') should be convincing proof that 20,000+ ft.lbs/deg figures are largely meaningless unless you're having to run astronomically stiff springs to deal with the down-force generated by high end 'formula' single seaters.

2. If using triangulation to 'simulate' panelling (assuming the use of Grape GBW32 software or similar that doesn't model sheer panels), you would be wise to use something like Mathcad to estimate the buckling limits of the panel in sheer in a couple of planes, to appropriately size/configure the tubing that will 'simulate' the panelled structure. Far from perfect, but if we're building a tube-frame GT40, we probably don't want to pay the licencing fees for Solidworks or similar (add the software licence fees and the cost of the spaceframe chassis together and you might be able to buy a mono tub off someone).

3. If you've got access to Solidworks or similar, ignore #2 and carry on.
 
Last edited:
Back
Top