DaveW wrote:With that information, it is possible to imagine an accelerometer attached to the upright (the wheel side of the actuator) measuring the vertical component of upright acceleration. If the signal generated by the accelerometer is integrated, to produce a signal proportional to absolute velocity, and that velocity is scaled to generate current that is used to drive the "local" EHSV, then the actuator can be made to faithfully (more or less) move to exactly reflect the velocity of the upright. In that case, the actuator will transmit no additional load to the sprung mass, regardless of upright motion (with caveats). The sprung mass will then remain stationary in space. That should satisfy xpensive's proposal (again with caveats), and the sprung mass could be made of cheese, or anything else that springs to mind.
But might not there be a need to affect the unsprung mass as a result of forces generated at the tires?
olefud wrote:But might not there be a need to affect the unsprung mass as a result of forces generated at the tires?
If you are referring to controlling tire load variation/grip then I believe DaveW's caveats (#1-2) would apply.
I read these conditions as relating to damping, but I’m not disagreeing with your view. More specifically, my concern is that the whole purpose of the suspension system is to accelerate the unsprung mass as in turning, braking and speed changes. To say that a stiff chassis isn’t required because of the caveats, it strikes me that it’s tantamount to asserting that a comfortable driver’s seat obviates the need for a steering wheel given the same caveats.
Must have been chewing gum and keystroking. "unsprung mass" should read -sprung mass-. Sorry.
Last edited by olefud on 18 May 2012, 23:12, edited 2 times in total.
So you are referring to road-planar, lateral/long forces, and thus I assume you meant the whole purpose of the suspension is to accelerate the sprung mass(?). I agree with that as the global purpose, with the understanding that there is a ride domain component (aero and grip) that will directly or indirectly affect that purpose. I think the comments in this thread are (so far) only addressing that ride component.
Admittedly slightly off-topic, but I would like to respond to Tommy Cookers first:
Tommy Cookers wrote:Preview was the point of the B-1 comment .
Without preview, a system can be called anything, it will have more limitations ? Should all such systems be called Predictive ?
I think you will find that the B1's foreplanes are (were) motivators, rather than sensors. Essentially, the foreplanes were driven by accelerometers signals so as to damp the body ending modes. The clever bit was to use the same motivators to damp both vertical & lateral modes. This reference might be useful.
I hadn't made the B-1 comment for about 20 years, there was a lot of Active Ride gloss and hype then.
I think there was/is some argument around, for anyone still interested in manned flight at 1000+ mph and 100' (the early B-1 concept at least was for this, with a part steel structure for the heat factor, any production plane was an all-aluminium 'derated' machine, as stealth replaced speed).
My impression of the AR days in F1 was that the Lotus was too clever to work properly, unlike the Williams which was more like active self-levelling ?
GSpeedR wrote: I think the comments in this thread are (so far) only addressing that ride component.
As usual, the thread is a bit ambiguous and wonders a bit. G-force’s original query was, “Would active suspension adjustments compensate chassis flex.” I think the simple answer is no, a stiff chassis is still needed. That said, there’s no reason not to learn something have a bit of fun with the subject. DaveW did so in describing an active system. Actually, it got me thinking that a laser array ahead of each tire could premap the road surface and, with Doppler input, provide a proactive servo response. No need to shoot this down; its’just an example of how DaveW’s post got me thinking.
My take-away is that active suspension may well work better with a flexible flyer chassis in than the car corners would be decoupled (dequaded?), but the car still would need a rigid chassis for reasonable performance.
Right, if the objective with active suspension is a constant ground-clearance in every corner of the car, then yes, it's obviously imperative with a rigid chassis.
"I spent most of my money on wine and women...I wasted the rest"
My reply to xpensive was attempt to demonstrate that an irreversible actuation system can satisfy his criterion, but still not be a suspension system. It was also a precursor, perhaps, to describing how the Lotus Active system came about, & why, perhaps, it should not be confused with the hydro-pneumatic Williams system.
As most posters have indicated, rightly, a suspension system must control loads, as well as position. The problem is how to respond to road inputs and manoeuvring disturbances in the presence of high, and rapidly varying, aero forces. The principle was to measure total downforce at the actuators with load cells, and to modify those by subtracting aero force, before using the (modified) forces to simulate a the responses required for a simple "no-downforce" suspension.
The algorithms required to simulate the suspension, including the corrections required to compensate for lateral & longitudinal acceleration, were relatively straightforward and were refined using a road vehicle. Measuring downforce proved rather more difficult. In the end downforce was "scheduled" using airspeed. Errors introduced when following another car were minimised by using two coupled pitot tubes.
The system failed "active", courtesy of the "helper" springs, it included various compensation algorithms (one of which provided an early warning of a puncture), an algorithm to maintain lateral balance (which, incidentally, Senna hated - although we didn't remove it completely), and a relatively "soft" basic suspension. Although the code became quite complex, the system itself was very simple (certainly compared to today's suspension systems). It was fairly heavy, I recall that the "add on" weight was stated as around 25 lb. As a said in a previous post, the result could best be described as a simulation of the '88 vehicle. It was reasonably competitive at first, but became less so during the season because, I think, it didn't work its tyres hard enough.
And it did demand a relatively stiff chassis, but the system did compensate for compliance.
DaveW wrote:With that information, it is possible to imagine an accelerometer attached to the upright (the wheel side of the actuator) measuring the vertical component of upright acceleration. If the signal generated by the accelerometer is integrated, to produce a signal proportional to absolute velocity, and that velocity is scaled to generate current that is used to drive the "local" EHSV, then the actuator can be made to faithfully (more or less) move to exactly reflect the velocity of the upright. In that case, the actuator will transmit no additional load to the sprung mass, regardless of upright motion (with caveats). The sprung mass will then remain stationary in space. That should satisfy xpensive's proposal (again with caveats), and the sprung mass could be made of cheese, or anything else that springs to mind.
Dave, although you clearly have a lot of knowledge in this area, I have to say I do not think this is a stable control scheme. Imagine your vehicle is traveling down the road and your accelerometer gets a little impulse at one corner. The actuator there would try to compensate, putting a force where the chassis meets the suspension. Fine, you've accomplished what you wanted at that wheel, but now the opposite corner has some reaction. The local actuator there goes to compensate for that, sending a reaction force back to the original corner. Now that one moves again. It's easy to imagine this going into unstable oscillation and blowing up.
The problem is that you've got a single-input-single-output model and what you need is a multiple-input-multiple-output model. What you describe in your most recent post sounds a lot more reasonable, and I imagine it used full-state feedback to ensure stability. If, after that, you wanted to throw in a controller to maintain the correct ride height that'd be pretty easy, but you need all those states, they're too coupled in the dynamics for the actuators to act individually.
Using supply (or input) measurements to drive an actuator directly is usually called "feedforward". If velocity feedforward gain is changed from zero (when the actuators don't move) to unity (when the actuators move at exactly the velocity of their accelerometers), the movement of the actuator sprung mass attachment points are successively reduced, ending up at zero. When the movement is zero, the sprung mass remains fixed in space - at least until the actuators run out stroke. The latter can be prevented by stroke limiting algorithms (at a cost of disturbing the sprung mass, slightly). The scheme does't go unstable, as such, but any motion induced in the sprung mass remains undamped. It is, however, relatively easy to add load-based feedback to take control over such disturbances.
I (royal I) used the scheme to control an active truck cab, which generated "ride" numbers, according to the customer, better than an S class Mercedes ("we couldn't market this, the drivers would break the trucks"). A similar scheme is used to control downforce actuators in 7-post rigs (using velocity transducers in the case of Servotest, and accelerometers for IST, and MTS), & I have good reason to believe the EH101 uses the scheme (now) to isolate the passenger cabin.
You are correct (in general) about multi-input/multi-output. In fact, a "modal" suspension was adopted for the Lotus suspension. In the first place this was an exercise (to do something "different") but it proved to offer advantages the were not available in a conventional simulation, and it was not changed whilst I was involved.
I would tend to think of an AR system with position as the controlled variable, following against strongly varying loads, a time-varying position demand signal from hybrid real/synthetic sources, and using real position sensors (surely the actuators are more naturally position controlled devices).
Don't we all know that a good AR could get away with a somewhat less rigid chassis, but why go that way ?
Regarding optimising roadholding by 'on track' optimising front/rear distribution of 'overturning moment' loads I'm thinking that current F1 has intentionally higher roll centres (to reduce OM effects on springs ie roll), and that AR could adjust roll centre rather than the spring(s) currently performing the traditional 'anti-roll bar' effect (this effect has won a lot of WDCs over 50+ years).
A car with a 99mph back end on a 101mph front end corners at 99mph, a car with a 100mph capability at both ends corners at 100mph, ask (Sir) Jack Brabham !
However, adjusting a roll centre also works better with a stiff chassis.
The more capable the AR system is, eg the greater is the frequency response, the more capable it is of doing something very bad, (that is why AR is greatly limited in civil aircraft certification). The stiffer chassis would reduce scope for this ?
One thing missing in the discussion are the delays, I think. I really dont know whats that for these sensor/processing/actuator systems. As our system has 4 "legs", I guess a relatively rigid chassis would help one end (or side) actuador to do its work on time. (Remember it takes time to compress "the spring" the chassis is).
The same could be thought regarding dimensional precision and again, I ignore whats top-notch in the market right now, but what would be the point of having 0,1mm pricise actuators when your chassis cedes 1mm.
"You need great passion, because everything you do with great pleasure, you do well." -Juan Manuel Fangio
"I have no idols. I admire work, dedication and competence." -Ayrton Senna
I think it all depends on from where the feed-back, or "is-value", is taken, ground-clearance sensor or something else, but either way, the draw-back with hydraulics is always speed. Ideally electro-magnets would probably the way to go, which is something that has been tampered with on valve-trains, but at a terrible energy-price.
"I spent most of my money on wine and women...I wasted the rest"
Aren't we looking at a high force, lowish speed situation, ideal for hydraulics ? They will beat anything else for all round performance'. Really the hardware is the easy bit, trust me !
The real problem is having at all times the right (position?) demand signal for each actuator ? This will take some development I think. This was the critical factor with the Lotus ?
