Return of active suspension - 2017

[Tim.Wright]

The main advantage of active ride in F1 would be that you could decouple, to a degree, loads coming from aero, load transfer and road irregularities and deal with each of them seperately. You can then setup something which is very "stiff" against downforce and load transfer in order to keep the car at the optimum ride height while cornering, braking and accelerating but soft over road irregularities. This gives you a double edged performance gain in aero and mechanical grip.

This decomposition of the tyre loads is not practically possible with a passive system. Then add to that the fact that in a passive system each part can only move in the direction that an external force is pointing. I.e. having the car lift itself up as downforce increases is not possible.

Other ways it can improve performance is by running very low ride heights to lower the CG without worrying about hitting the ground in pitch and roll. Another trick could be to lift the body to a low drag attitude under acceleration to improve top speed.

I'd say seconds of laptime improvement is realistic. Though such a system would be definately be seen as a breach of the moveable aero rule. So either the rule will change or the active ride controllers will be extremely limited.

Dave, based on a few job postings I've seen around, your company seems to be gearing up for some active ride work

[DaveW]

From a team perspective, what sort of performance benefits would a fully active system bring compared to the former interconnected systems?

I've heard it's worth seconds to have a full active suspension on

Be careful who you listen to Hovepeter.

The answer depends on A) What the FIA allows, B) How mechanically-aware the team is, C) The capabilities of interconnected systems.

Suppose that the Lotus active system was permitted. Its "modal" suspension strategy controlled rigid body modes independently, and separated attitude control from vehicle dynamics control (at least in the working range of the actuators). That would imply that interconnected "fixes" were redundant, and (within limits) mechanical control did not have to be compromised by the need to control aero. You might think that could yield a performance benefit in good hands.

It goes further than that, however. Parameter based control laws could be changed without touching the vehicle. That implies that set-up changes could be made and assessed (in free practice) almost without stopping. It also implies that suspension changes could be made between qualifying and a race, even between the out lap & a race. That would probably help performance in today's "no test" rules, even when tyre heaters are banned.

p.s. Tim stated it well...

[Tommy Cookers]

An active system will driven by measurements, none of which is perfect. That reality has important consequences.
So, for example, one of the most important parameters would be load. I believe that PRL cells drift (vary with temperature) and are sensitive to more than just the vertical component of load. Hence they must be manipulated before they can be used to drive the suspension. Adding the ability to remove acceleration effects comes for free.
The same to be said about other measurements....

any production load cell will surely be ready compensated for (bulk) temperature effects on no-load reading and loaded reading (span)
but has unquantified and excessive sensitivities to load components other than the one intended and calibrated
(basically because the cell has load paths common to all components of load, ie nominally 6 d.o.f)
this can be solved by designing a cell having multiple load paths, isolating the desired component from the undesired, more 1 d.o.f

but .... clearly other measurements eg position and/or velocity are easier to make and 'better' (in themselves)
and as we are primarily looking for body attitude/height control
surely in 2014 an active system does not need to be based on load measurement ??
a 1974 control system with digital elements would likely have a lower frequency response than would a good analogue system
actuator behaviour minimising load variation is equivalent to behaviour minimising variation of position, velocity or acceleration ?
now the digital element will allow the same actuator behaviour to be realised regardless of the types of sensors used ?

[DaveW]

... any production load cell will surely be ready compensated for (bulk) temperature effects on no-load reading and loaded reading (span) but has unquantified and excessive sensitivities to load components other than the one intended and calibrated (basically because the cell has load paths common to all components of load, ie nominally 6 d.o.f) this can be solved by designing a cell that has multiple load paths, isolating the desired component from the undesired eg 1 d.o.f

That is exactly why we designed and built our own load cells, but we have already discussed this I think.

... surely in 2014 an active system does not need to be based on load measurement ?? actuator behaviour minimising load variation is equivalent to behaviour minimising variation of position, velocity or acceleration ?

I am tempted to reply "What is so special about 2014?"

It is true that a "fast ride height" system such as Dernie's probably does not require load to be measured, but then spring rates and damper settings would not be exposed to FIA scrutiny, something that they appear to require. If spring rates and damping curves are to be simulated, then accurate load measurement is a requirement.

Edit: There you go again, Tommy.

a 1974 control system with digital elements would likely have a lower frequency response than would a good analogue system actuator behaviour minimising load variation is equivalent to behaviour minimising variation of position, velocity or acceleration ? now the digital element will allow the same actuator behaviour to be realised regardless of the types of sensors used ?

The Lotus system was designed in 1981. It appeared on an F1 vehicle at the start of 1983, controlled by a fixed architecture analogue computer with multiplying DAC's acting as potentiometers. The system became digital in 1984, using a TMS320 DSP (and an assembler designed by us, because TI had yet to publish their version). Back-back tests demonstrated that the two systems had the same bandwidth, but the DSP version provided much more flexibility.

Forgive me, but I would be grateful if you could check your facts before rushing into print. That would save us all lot of time....

I really don't understand your last statement, I'm afraid. What does ".. same actuator behaviour to be realised regardless of the types of sensors used.. " actually mean?

[Hovepeter]

From a team perspective, what sort of performance benefits would a fully active system bring compared to the former interconnected systems?

I've heard it's worth seconds to have a full active suspension on

Be careful who you listen to Hovepeter.

The answer depends on A) What the FIA allows, B) How mechanically-aware the team is, C) The capabilities of interconnected systems.

Suppose that the Lotus active system was permitted. Its "modal" suspension strategy controlled rigid body modes independently, and separated attitude control from vehicle dynamics control (at least in the working range of the actuators). That would imply that interconnected "fixes" were redundant, and (within limits) mechanical control did not have to be compromised by the need to control aero. You might think that could yield a performance benefit in good hands.

It goes further than that, however. Parameter based control laws could be changed without touching the vehicle. That implies that set-up changes could be made and assessed (in free practice) almost without stopping. It also implies that suspension changes could be made between qualifying and a race, even between the out lap & a race. That would probably help performance in today's "no test" rules, even when tyre heaters are banned.

p.s. Tim stated it well...

thank you for these replys! i think it was based on that you had no restrictions on your suspension whatsoever like the 93 williams car. its because I've have been watching these documentary, where they tell about the williams racer: https://www.youtube.com/watch?v=hkY_EeHwlcQ from 1.40 min into the video

[DaveW]

.. thank you for these replys! i think it was based on that you had no restrictions on your suspension whatsoever like the 93 williams car. its because I've have been watching these documentary, where they tell about the williams racer: https://www.youtube.com/watch?v=hkY_EeHwlcQ from 1.40 min into the video

Apologies, I thought you were being sarcastic. The mantra is still relevant, though.

[DaveW]

but .... clearly other measurements eg position and/or velocity are easier to make and 'better' (in themselves) and as we are primarily looking for body attitude/height control
actuator behaviour minimising load variation is equivalent to behaviour minimising variation of position, velocity or acceleration ?...now the digital element will allow the same actuator behaviour to be realised regardless of the types of sensors used ?

It has something to do with independent and dependent variables. Nominally, an hydraulic actuator is irreversible (that is, it moves only if it is commanded to do so).

Let's see if I can explain without diagrams. A suspension unit comprising a linear spring (rate ) and a parallel linear damper (damping coefficient ) has a transfer function:


Where is the load supported by the strut, and is the position of the strut. is the operator δ/δt.

An hydraulic actuator (a cylinder controlled by an EHSV) is an integrator. Thus the velocity of the actuator will be proportional to the current used the power the EHSV. Let Gain be the constant of proportionality. It follows (hopefully) that a control law to use the two measurements and to make the actuator behave like a suspension unit can be derived by referring to the transfer function:



An increase in Load will create a drive current that will cause an increase in strut position, until . The velocity of the actuator will depend upon the inverse of .

Thus the actuator will emulate a linear suspension using just two measurements and two parameters.

The emulation becomes interesting when the load has a deterministic component, say an aero force. This can then be subtracted from the measured load before the control law is used to drive the actuator. Then the actuator will not respond (at all) to changes in aero force provided, of course, that the estimate is accurate.....now the fun starts.....

[godlameroso]

Ah yes the wonderful world of calibration. It seems that F1 is headed to putting a greater and greater importance on software. It is the software that will create the best way to exploit the various systems. Between brake by wire, throttle by wire, electronic differentials, and active suspension, not to mention energy harvesting and deployment strategies. Getting everything to work seamlessly, give confidence, and good feel for a driver becomes just as much software related, as it is to building mechanical components.

I look forward to the day when there are 6 engineers with touch pads plugged into the car changing parameters along with the mechanics tweaking suspension settings building the car etc.

[Tommy Cookers]

Yes
when I IMO'd that the digital side had advanced usefully (notionally over these last 40 years) it caused some displeasure
usefully maybe eg increasing the hardware options for simplification and cost reduction/'road relevance'
and maybe the scope for intelligence eg the system 'teaching itself'

btw about 25 years ago the X-31 project found that digital (flight control) system designs were inferior to analogue ones

[Hovepeter]

its okay dave W! it can be hard to know when people are writing, what there meanings really are

haven't anyone go the formula one teams asked for clarity on what ''kind'' of active suspension they have to run in 2017?
we are really many people on this forum that would look forward to know what to expect!

[DaveW]

btw about 25 years ago the X-31 project found that digital (flight control) system designs were inferior to analogue ones

That is an interesting and controversial statement. Do you have a reference explaining why they were inferior?

[Tommy Cookers]

in conference proceedings around that time (NASA iirc)
maybe it said only that the digital designs and the designs based on modern control theory were a disappointment
I have copy around somewhere (it also contains a debunking the then-current mythification of the Gurney 'flap')

another btw
some wave machines (generating complex water waves for model testing) are nominally force-controlled (others position-controlled)
(force control facilitates running the machine in wave absorbtion mode)
these machines have a tension spring in parallel with the (electromechanical) actuator, or can be centre-sprung (2 springs)
in absorbtion mode they seem to be essentially an 'active suspension' system, nullifying incoming bumps on the water surface

in absorbtion mode a position-controlled machine needs predictive position demand signals numerically derived from measurements
but would still be essentially an 'active suspension' system

[DaveW]

You made the unequivocal statement:

btw about 25 years ago the X-31 project found that digital (flight control) system designs were inferior to analogue ones

which became, after a prompt:

maybe it said only that the digital designs and the designs based on modern control theory were a disappointment. I have copy around somewhere....

The first statement has little in common with the second. Which, if any, should a young (or not so young) engineer believe?

You clearly have much knowledge and experience, Tommy, but I'm afraid I can only sigh, & repeat my earlier rebuke.....

[DaveW]

To continue....

I knew nothing about the X31 project, but I was intrigued that digital flight control systems (FCS) were compared with analogue alternatives.

It is true that digital FCS systems are (potentially) limited by quantisation and transport delay issues, but these are well documented, I think. It is also true that conservative design caused FCS designs to lag current technology by several years. My work suffered from that fact in one occasion, and I recall that one military system suffered weapon accuracy issues that were, apparently, caused by asynchronous data streams, but an analogue design was not a realistic alternative in either case. This is why I asked for a reference.

Your alternative quotation depends on what is meant by "modern control theory". A reference I found stated:

”In contrast, modern control theory is carried out strictly in the complex-s or the frequency domain, and can deal with multi-input and multi-output (MIMO) systems. This overcomes the limitations of classical control theory in more sophisticated design problems, such as fighter aircraft control. In modern design, a system is represented as a set of first order differential equations defined using state variables. Nonlinear, multivariable, adaptive and robust control theories come under this division. Being fairly new, modern control theory has many areas yet to be explored. Scholars like Rudolf E. Kalman and Aleksandr Lyapunov are well-known among the people who have shaped modern control theory.”

You might like to search for "70967.pdf X15", which is a re-visit of an X15 crash. This concluded:

"The MH-96 adaptive flight control system is an elegant design that accomplished its goal of enforcing performance across all flight conditions. Furthermore, the MH-96 showed that a satisfactory adaptive control system could be designed without having accurate a priori information about the aircraft aerodynamics, and, consequently, aircraft configuration changes could be easily accounted for. However, the MH-96 lacked an analytically based proof of stability, which was highlighted by the fatal crash in 1967. After four decades, the theoretical ground work for applying adaptive control has now made it possible to design adaptive controllers that offer high performance as well as stability guarantees in the presence of uncertainties."

I might add, cynically, until the next time.....

[gixxer_drew]

The inerter gains grip because it reduces oscillation of tire load. What this means is your minimum tire load will be higher so its harder to lose grip.

DaveW please help here. Your multimagic is needed for an explanation.

Mmm. Not sure about that.

However, for what its worth, an inerter fitted (hypothetically) to the front axle of a race car (a GP2 in my example), will increase the minimum load at the front axle up to a point, hence increasing "grip". Here is my example.

The plot show the minimum, mean and maximum values of vertical load under the front wheels whilst the vehicle is being subjected to a swept frequency vertical input. That shown in red is with no inerter, with the minimum value occurring at the heave mode. Those shown in green, blue and black are for increasing values of inerter mass.

Heave mode control is successively improved with each increase in inerter mass, but at higher values of mass another mode intrudes, and becomes dominant. The new mode is actually the unsprung mass (hub) mode. The interesting fact about the inerter is that the overall power dissipated by the front tyres increases with inerter mass.

Hence there is a balance to be made between the improvement in heave mode control, loss in control of the hub mode, and the heat generated by the power dissipated by the tyres (front tyre temperature is often an issue).

Apologies digging up an old post, I did not have a chance to log in for some time. Dave, in this example you said inerter fitted to the front axle I assume you mean to the front third, yes? So this is a heave mode only test of just the front suspension, I'm not sure what you can talk about freely here, but what was the rear doing in this test?

It seems fairly obvious that the "spinning top" will carry inertia and result in a secondary unload after it has "damped" the initial load but that should be in equal force value to the "damping" of the initial load. In the graph you posted the secondary unload event, even with the low inerter setting is larger magnitude than the forces absorbed in initial compression. This is curious to me, why? Before I read your post I assumed that the car will now be on the "gravity spring" if you will and that is totally undamped but you reference to the hub mode, has me curious if I had that backwards?

Finally I think you sort of implied here that CPL toward the edges of the spectrum is more severe than even a larger magnitude one toward mid spectrum? I wonder how would you go about characterizing the tire for something like that?