Bose Suspension

[DaveW]

My guess would be that the Bose Suspension has (or could have) all the capabilities of the Lotus system, apart, perhaps, from the ability to carry a steady state load. That is not true for the Mumford system nor, may be, the similar Mercedes system, because in those cases both the springs and dampers are real (not simulated).

The "bandwidth" required for an active system is not easy to define succinctly. The system must control the rigid body modes, of course, but ideally should not interfere with structural modes. On average, I guess, for each structural mode that will be stabilized, there will be at least one that will be destabilized. On most road cars there is no clear frequency separation between the two.

To (try to) explain: The sprung mass of a vehicle is often modeled as a monolith. In order to generate a load, the suspension (any suspension) requires a sprung mass to act against. The monolithic sprung mass assumption is (far) from adequate for road cars. For example, the power train, radiators, fuel tank, exhaust system, etc. are often suspended from the chassis by "soft" mounts. Each will have one or more (up to six) natural modes and each of those can affect the apparent mass that the suspension has to work with.

One active car we converted (a Buick Park Lane, I recall) had an apparent front sprung mass at 10 Hz just 10 % of static (implying that a suspension load at that frequency would increase the chassis acceleration by tenfold, compared with the expected monolithic value). The front hub mode was higher than 10 Hz. So although we no difficulty making the car handle, achieving a good "ride" was difficult, requiring modifications to the base vehicle. Incidentally, that would apply equally to the passive version, & probably explained why it was an "Olley-tuned boulevard cruiser" (or the reverse, perhaps).

The issue is not confined to road cars. When developing the T99 system I inadvertently destabilized a wish bone structural mode at around 200 Hz, I recall. That was easy to fix.

[Greg Locock]

FWIW I now think that emergency manual reversion on EPAS would only involve disconnecting the power cabling
this would reduce the backdriving force required for any EPAS electrical fault state
although the starter might still be disabled if the system thinks the line voltage is too low to work the EPAS

I have tried disabing EPAS while maneuvering. At speeds above 60 kph so long as you aren't pulling any huge gs it is ok, but at 30 kph I found it very hard to maeuver around normal suburban streets at normal speeds.

The problem is that the step-up ratio is enormous and the inertia and friction of the motor/belt/ballscrew are high.

[Richard]

I have tried disabing EPAS while maneuvering. At speeds above 60 kph so long as you aren't pulling any huge gs it is ok, but at 30 kph I found it very hard to maeuver around normal suburban streets at normal speeds.

The problem is that the step-up ratio is enormous and the inertia and friction of the motor/belt/ballscrew are high.

The extreme case is trying to turn the wheel when stationary with the power off. It's impossible.

[Tim.Wright]

yes but there aren't really any safety implications to that.

Losing precise steering control at 30km/h could mean hitting a parked car or a pedestrian.

I've lost hydraulic steering at about that speed (ironically probably in a car that Greg worked on - 94 Falcon) and it wasn't too bad (although not completely safe). Seems the EPAS systems are worse in this regard.

[Greg Locock]

Yup that was one of mine, tho that was when i worked for the dark side (NVH).

Back driving an HPAS that has lost its fluid is really just seal friction, plus whatever restriction the remaining fluid offers as it pumps out.

an EPAS has a step up ratio of say 40:1, driven by a ballscrew.

[DaveW]

an EPAS has a step up ratio of say 40:1, driven by a ballscrew.

My group once developed an "after market" power assisted steering that used a direct (rotational) drive printed circuit motor with almost no inertia & minimal cogging, I recall. Probably too expensive - but what price safety?

[Tim.Wright]

...but what price safety?

USD200k said Ford (ok this was 40 years ago)

http://web1.calbaptist.edu/dskubik/pinto.htm

On topic, I was at an automotive expo today and I am seeing a large move away from hydraulics and towards electrics for linear actuation. Of all the simulators and test benches I saw today the only hydraulic devices of note were from MTS and MOOG (who are a hydraulic company). I think the bandwidth problems that Tommy mentions are slowly being overcome at the moment as brushless drives are becoming more common place. It seems to be an area where there are some definite changes taking place.

I saw 2 or 3 simulators using a Stewart platform and combination Stewart-platform + XY-table setups and they are all electric. The algorithms used to provide the feelings of sustained G are also interesting but its another topic

[Greg Locock]

..and yet every safety improvement you see on cars and aircraft and streets and industrial facilities and so on and so forth is driven by a cost benefit analysis. And I'm sure the cost no object crowd always fit the best possible replacement tires to their annually updated cars as soon as they show any sign of wear.

Dave, yes there's also a Russian system that uses a direct drive motor on the steering column. Plus of course the old BMW system that at least locked the motor out of the steering if it failed (via an epicyclic).

[DaveW]

... Stewart platform ...

Interesting & seductive device. Performance will be heavily dependent on the quality of the strut bearings, and the transformation matrix is not linear-in-parameter. The latter means, I think, that a solution is iterative, which has bandwidth implications.

[Tim.Wright]

Interesting & seductive device. Performance will be heavily dependent on the quality of the strut bearings, and the transformation matrix is not linear-in-parameter. The latter means, I think, that a solution is iterative, which has bandwidth implications.

The relationship between the actuators and the table motion is for sure non linear, but I think its pretty determinate, i.e. well behaved. I don't see why you would need iterations to arrive at a solution. If you know the 3 angles and three displacements then its easy to know the required actuator displacement.

[DaveW]

... If you know the 3 angles and three displacements then its easy to know the required actuator displacement.

I thought that too. Experience proved otherwise. A search should be fruitful....

[Greg Locock]

Lots of matlab scripts out there for stewart platforms, and similar robots. Our K&C rig is a stewart platform like this one http://www.ideasinmotioncontrol.com/200 ... ystem.html
Bandwidth problems due to indeterminate solutions could be solved by lookup tables, RAM is cheap!

The advantage for a robot is that you can get a fully placeable and controllable mounting point into position using grounded actuators, so their mass is not especially problematical. The disadvantage is that the ratio of working space to total volume is limited.

[DaveW]

... Our K&C rig is a stewart platform like this one ...

Horses for courses, I guess, assuming you do need six degrees of freedom. I was merely trying to point out that position (and trajectory) management is not straightforward, particularly when dynamics is a consideration.

The advantage for a robot is that you can get a fully placeable and controllable mounting point into position using grounded actuators, so their mass is not especially problematical. The disadvantage is that the ratio of working space to total volume is limited.

Also, the layout of the actuators often leaves something to be desired when they are used to support & control a large sprung mass (e.g. a multi-ton flight simulator motion platform).

[Greg Locock]

Incidentally the reason I got involved with Stewart platforms is that somebody instrumented a flex coupling with 6 stringpots in a stewart array and had lots of test data and no idea how to turn it into a usable description of the motion of the two ends of the coupling. In the end I just fed the measured time histories into a model with linear actuators and then we could see it animated and it was easy to measure the relative motions. The maths is non symmetrical, it is (relatively) easy to work out where the platform is and what angles it is at if you know the lengths of each actuator, it is not easy at all (but not impossible) to work out what the actuator lengths have to be to achieve a given platform position and rotation in 3d.

[olefud]

the electric or electromechanical actuator must have springing in parallel for it to control the dynamic response
so it will need to be rather capable in both force capacity and bandwidth
and can beyond its nominal bandwidth be allowed some managed 'free' backdriving ie on severe/high frequency bumps

in any parallel system there will be force differences ie some conflict (in this case between the actuator and the springs)
this happens in conventional suspension (that's what dampers do), and in the F1 Active
that's what design is about
there's no easy way out, strut compliance must be manageable by the millisecond, this demands both power and bandwidth
a low-power electromechanical system will give mostly uncontrolled Newtonian damping

The voice coil has inherent, rate-progressive damping in that movement of the coil in the magnet gap generates current in the coil. Of course, as with hydraulic dampers, the energy would have to be sunk somewhere, perhaps as rectified current to a battery.