Another Human Powered Vehicle! Part 4 - Testing Interconnected Airbags

This article was first published in AutoSpeed.

Last week in Part 3 of this series we looked at the ways in which vehicle suspension systems can use interconnections between the wheels. Most cars run lateral linkages between the suspensions at each end (they’re called anti-roll bars) and a few cars have been produced with separate front/rear interconnections (eg the Austin 1800). As described in that story, in terms of resisting pitch and roll, the front/rear interconnection approach is better than lateral interconnection, but on a 3-wheel vehicle, front/rear interconnection loses most of its advantages.

However – and especially with interconnected air springs – working out exactly what will happen with an interconnected suspension is quite difficult. So it was time to return to the bench testing...

Bench Testing

I’d set up a simple bench rig that allowed the airbags to be tested. This comprised two pieces of timber that were each drilled horizontally at one end to take a long, snug-fitting (but not tight) through-pivot. The pivot was inserted through the two timbers and was then clamped in a large vice. At various points along the two pieces of timber I drilled holes in which the top mounts of the air springs could sit. Further along the timbers I marked the spot where the wheels will go in the finished design.

By placing various amounts of heavy steel weights on the top of the arms I could realistically create different loads, including static deflection, 1g vertical acceleration in bump (ie weight doubling) and other loads. This set-up also allowed the spring to be placed at various positions along the arms, so changing the motion ratio.

In addition, by bouncing the arms on the spring, I could directly measure natural frequency and see how much damping was present. By making plumbing connections between the springs, linking could be tested.

Interconnected Airbags on the Test Bench

My first thought was to interconnect all three airbags. My logic was this:

• Front-right bump results in front-left increase in pressure, and a smaller, delayed increase in rear pressure. Pitch and roll reduced (as per Austin 1800 Hydrolastic suspension)

• Rear bump results in delayed front-left and front-right increase in pressure – pitch reduced

But the above system didn’t work.

Take the linking of just the two front airbags. As described above, a front-right bump results in front-left increase in pressure. So the front-right airbag contracts and the front-left airbag expands. But that’s where they stay! There’s no restoring force trying to return the airbag springs back to normal ride height! It was quite interesting having the system set up on the bench, pushing down on one weighted lever (indicative of roll) and watching the other airbag smoothly extend. In fact, you could say that the resistance to slow roll was just about zero.

And that led me to the next idea. What about adding an anti-roll bar to the system? That way, there’d be a restoring force trying to bring the airbag extensions back to the same level – in other words, resisting roll. I added a steel anti-roll bar (arrowed) to the bench mock-up and found that the interconnected front airbags then worked well, returning to level when released. But the roll stiffness, provided only by the anti-roll bar, was still quite soft.

Then I had another thought. What about adding a shut-off valve in the tube linking the two airbags, a valve that closed when the steering was turned (or a lateral acceleration was experienced)? That way, roll would also be resisted by the air stiffness within the two springs and the anti-roll bar.

So to summarise so far:

Interlinked front airbags, anti-roll bar, auto disconnection valve on cornering
Pitch	Not resisted
Roll	Resisted by anti-roll bar and both front airbags (one in extension and the other in compression)
Two wheel bump	Resisted by both airbags only
One wheel bump	Resisted by anti-roll bar only

But – and this why the bench testing is so important – the situation gets even more complex than this. For example, I’ve written above that a one-wheel bump is resisted by the anti-roll bar, but it is also working with a spring that is extending on the opposite side to the bump wheel. So the one-wheel resistance to bump is actually being provided by the combination of the extending opposite side airbag “diluted” through the less-than-perfect stiffness of the anti-roll bar. So what would be the actual outcome?

I added a shut-off valve to the bench system and did a lot of testing. This is what was found:

• With the interconnection valve closed, the system was very stiff in roll (great!)

• With the interconnection valve open, the system was very soft in one wheel bump (great!)

• With the interconnection valve either open or closed, the system was of medium stiffness in two wheel bumps (good!)

So things were looking very promising indeed. But the next step improved things even further - adding user-adjustable control of the airflows.

User-Adjustable Flows

The springiness of the airbags is governed by both their internal volume and the pressure within them. The latter’s easy to understand – pump them up harder and they have a higher spring rate. However, the former is also pretty damn interesting. If you link the airbag to another closed volume, there is an increased amount of air available to be compressed with airbag deflection. This has the affect of softening the spring rate (and so, incidentally, reducing the natural frequency as well).

But it gets even more complex. If, when the airbag is compressed, the air is forced to flow through a restriction into a closed volume, the spring rate will be lower (because there will be more air than just in the airbag resisting compression) and the return of the spring back to standard height will also be slower.

So why does the latter happen? When the airbag is compressed, there’s a spike in air pressure that allows air to be pushed past the restriction. For example, on a quick compression, the pressure within the air spring might rapidly rise from 30 to 50 psi. The air that makes its way past the restriction into the closed volume will lift the pressure within the closed volume by a bit (not the full 20 psi increase though) and when the pressure in the airbag has dropped (the bump is past), the air from the reservoir will gradually find its way back into the airbag, so extending it back to normal ride height more slowly than would otherwise occur. (This quick compression and slow extension is normally call ‘rebound damping’.)

So connecting an airbag to a closed volume through a restriction allows alteration of both the airbag spring rate and the airbag rebound damping – though note that both are changed at the same time. Furthermore, these changed characteristics will be dependent on the speed of suspension movement.

Phew! It’s all getting very complicated – too complicated, I think, to work out on paper what will occur in any given system. So I went back to my bench set-up and did further testing. The set-up was as before – two linked airbags, shut-off valve in the connecting tube, steel anti-roll bar. But this time I altered the position of the valve so it wasn’t just fully open or fully shut.

With the valve opening adjustable, something very interesting happened. The stiffness of the suspension on one-wheel bumps could be varied at will, from being as stiff as provided by the airbag and anti-roll bar working by themselves, to being as soft as described above. Furthermore, the rebound damping also varied, increasing to a maximum with the valve just cracked open.

So what about having a system with two valves, mounted in series? One valve could be the user-adjustable variable control valve, and the other could be the valve that automatically shut when steering lock was applied. That way the suspension would always be stiff in roll, but the softness with which one wheel bumps (ie most bumps) were met could be user adjusted.

Forgive me for getting excited here, but at this point I was just blown away. One of the really difficult things about HPV suspension design is keeping things light. And since there’s no electrical power supply available, these two aspects mean you can’t have an on-board air compressor. And of course, in cars and trucks and trains, an on-board air compressor is usually used when airbag springs need to have their stiffness changed....

But here, with just two lightweight valves, it appeared possible to have user-adjustment of one-wheel bump stiffness and damping and also have a system that automatically provided strong roll stiffness (an especially important requirement when the rear wheel on a trike has zero roll stiffness).

So to summarise:

Interlinked front airbags, anti-roll bar, auto disconnection valve on cornering, user-adjustable series flow control valve
Pitch	Not resisted
Roll	Resisted by anti-roll bar and both front airbags (one in extension and the other in compression)
Two wheel bump	Resisted by both airbags only
One wheel bump	Resisted by anti-roll bar and opposite side spring; user adjustable softness and rebound damping

The Rear

With the front suspension working so well (on the bench only, remember – nothing had hit the road), I was reluctant to link it to the rear suspension as well. But the front design had shown the worth of connecting the airbags to a separate volume (in that case, each airbag connected to the other) via a restrictor. What could I do for the single rear airbag that would allow similar user-adjustable control of spring rate and rebound damping?

The answer appeared to be to connect it to a closed pressure vessel – but what could I use as that vessel? It had to be super light but still cope with a pressure oscillating from 30 – 35 psi with peaks as high as perhaps 60 psi. I sourced an aluminium fire extinguisher which was light and (when buffed-up) looked really good. It had an internal volume of about 1 litre.

But when on the test bench this pressure vessel was connected to a single airbag and loads appropriately applied, the system didn’t act as I’d expected. With the restriction sized to just allow flow, the airbag did not return back to normal ride height after a bump was experienced. Instead, it just stayed in its bump position.

Guessing that perhaps the volume of air in the reservoir was too great (and so the pressure rise within it wasn’t high enough) I gradually filled the fire extinguisher with water, so reducing its internal volume. When the internal volume approached 200cc, the system started to act properly. By adjusting the opening of the valve, the effective spring rate and rebound damping rate could be changed over a wide range. A suitably small container (arrowed) that could cope with the required pressure was found to replace the water-filled fire extinguisher and the system bench again tested well.

Conclusion

The use of airbags that are suitably interconnected and controlled appears to give the following advantages:

• Rugged, well proven spring design

• Sufficient travel to allow a low motion ratio (ie about 1.3:1 wheel:spring ratio) while still giving adequate wheel travel in both bump and droop (about 115mm total travel)

• Low spring mass (about 350 grams per spring)

• For a given vehicle weight, ride height and spring stiffness adjustable over a small range when installed

• Easy compensation for differing loads

• Lower natural frequency than achievable in a similar sized steel spring

• Passive or active interconnection allows dynamic roll control and user-adjustable stiffness and rebound damping

Some disadvantages remain:

• High cost

• Very effective progressive bump stops needed, especially in bump

• Monitoring and maintaining of air pressures likely to be very important

OK, that’s it for the first section of this new series. From here we’ll move onto the design and construction of the front and rear suspensions.