Believe it or not this is fairly rudimentary and only took about a full day to throw together, in between getting assorted other crap done. It does allow you to do a lot, though.The most important thing is that I can do engineering as opposed to tinkering (see aside*) and do high-level design work rapidly. I don't want to spend weeks bullshitting around with suspension points to see what they do. I can pick my kinematic curves, roll stiffness, lateral load transfer distribution and all those basic macro parameters... and once a lot of these are defined they really lock in where your points have to be to achieve them. There's fine tuning to be done with the realities of what rates are possible given constraints of chassis and wheel dimensions, etc... but this gets you pretty close, pretty quickly.
Also makes it much easier to anticipate things like, "If I buy a couple sets of tires to test, when I take off Set A and bolt on Set B, what if anything do I need to change?"
As an *aside... and I'm not sure if I mentioned this earlier... when it comes to design like this there are two ways you can go about it. The way we did at CU for many years was not suspension engineering. It was tinkering. You play with suspension points, move one up, another one over, and see what it does. Maybe you have an idea of where you're trying to go with the design, maybe not. This is a ground-up design approach. Can take days or weeks to get where you want to be.
The other method is to engineer a solution from the top down. You pick the performance parameters you want. Start at a high level. Once you start defining what you need to get the car to handle how you want, the small stuff falls into place. If I want a caster angle of 'X' and a mechanical trail of 'Y' there's one unique solution for how that works in the side view. It does the work for you!
8 comments:
hey Tom,
this is starting to get really interesting. but how do you go about choosing things like caster angle from your tyre data? Does it not have to be considered along with kpi and link lengths/angles?
Yes and no.
In that blurred out plot, for each corner of the car I make 3 traces. One at a constant camber/inclination angle that the car would actually see with the setup I input. I make another trace at +1 deg IA, and one at -1 deg IA. That way I can see across the board where I have to go with camber compensation to get the most grip.
If the RF shows it needs more (negative) inclination and the LF doesn't care if it goes more positive, then I can do it with more static camber.
If both the RF and LF want more (negative) inclination, then I have to do it with something like caster or the roll-camber curve.
The roll-camber curve, and the actual linkage assembly including the control arm lengths and angles.. are interchangable. One defines the other pretty uniquely given a handful of constraints. In other words they are coupled.
It is a hell of a lot easier to simulate and to adjust that roll camber curve as 3 points that define a 2nd order polynomial.. than it is to go crazy with actual suspension points. Saves time. Later on I can go back and say "Ok for this particular camber curve, what do I need from link lengths and angles to achieve it" and it should be pretty straight forward.
If that makes any sense...
OK, I understand how you can get your kinematics to match up to a given camber curve - be it in roll or bump.
But how can we decide on a finite amount of camber compensation?
I see why we want more (-ive) camber if the graph is telling us that we will achieve a higher lateral force as a result, but I'm still none the wiser as to how we quantify the amount. Sorry - I've not been doing this very long and it seems that I've been doing it the "hard" way...
Also, if you're going to do some more on this then don't worry about trying to answer all this, I'll just keep reading and ask when all has been revealed!
Thanks a bunch, ed
No worries. And yea, we can't right out the bat pick camber curve "X" on our first iteration. There is some playing to do with it, it's an iterative process.
In the example of this piece of software, by plotting baseline, +, and - inclination I get a direction but not necessarily a concrete amount of how far I have to go. In this case I pick some arbitrary amount (maybe 0.5 degree static camber at a time) and iterate in that direction until I find a performance peak where going further hurts more than helps. To know where that peak is, you have to look at the whole vehicle all together, or at least one complete axle (which I believe is called Steady State Pair Analysis in RCVD).
But still, it's a more direct and less time consuming approach of
I need more outside front camber compensation --> I change camber curve directly --> I see performance result.
Compared to...
I need more outside front camber compensation --> I play around with chassis or upright points or whatever --> Maybe that moves me in the direction I want without screwing anything else --> I see result on camber curve --> I see performance result.
Alternatively you can take the "brute force" approach and write more in-depth code that iterates on say:
5 different front toe settings...
5 different rear toe settings...
5 different front static cambers...
5 different rear static cambers
15 different front roll curves
15 different rear roll curves
5 different caster angles
5 different KPI's
5 different TLLTD's
etc etc (already 17+ million combinations) to find the "best."
Becomes difficult though to
(a) Constrain the problem appropriately
(b) Define "best"
(c) Intelligently decide which combinations to eliminate
(d) Run everything
That's great thanks. I see you go about this now - it's just a different way round doing what was done at our team last year, although I had nothing to do with it! (We went tire testing, added a a bunch of different cambers and measured the lap times wrt. camber i.e. camber sensitivity and then when we saw low sensitivity decided on a high camber compensation setup)
I quite fancy having a bash with some of the ttc data from round three and trying the "brute force" method - if it goes anywhere I'll repost and let you know!
Anyway, thanks again
Ed
Hey man
Just wanted to say that when I read this nearly 2 months ago, a light came on in my head and I realised what was bugging me ( even with no engineering background, and only 'seriously' learning about this kinda thing for about 8 or 9 months) about the way a lot of books, build diaries, magazines etc explain how to approach suspension design. It seems so logical... once you know the tyre data is out there, know that you can get your hands on it, and learn how to use it.
So yeah... huge thanks for the epiphany!
Btw...how nicely do you have to ask the manufacturers to get the data?
That depends on who you know. No matter who it is though, no harm in asking politely and seeing what you can scrounge up.
Post a Comment