The Mark Ortiz Automotive
Presented free of charge as a service
to the Motorsports Community
Reproduction for sale subject to restrictions. Please inquire for details.
Mark Ortiz Automotive is a chassis consulting service primarily serving oval track and road racers. This newsletter is a free service intended to benefit racers and enthusiasts by offering useful insights into chassis engineering and answers to questions. Readers may mail questions to: 155 Wankel Dr., Kannapolis, NC 28083-8200; submit questions by phone at 704-933-8876; or submit questions by e-mail to: email@example.com. Readers are invited to subscribe to this newsletter by e-mail. Just e-mail me and request to be added to the list.
SIMULATION PROGRAMS, AND TIRE DATA FOR THEM
Are there any simulation programs available to the short track racer that can be used to model setups before going to the track? Track testing is not much of an option due to cost and practice time is short.
Along with that, what if anything can be used for tire data, when none is available from the maker.
Big, well funded teams write their own simulation programs, and pay degreed engineers a lot of money to do it. The programs are then closely guarded.
I have not seen any affordably priced software that even claims to simulate anything more than suspension displacements and tire loads for given lateral and longitudinal accelerations. Those I have seen presented have generated clearly anomalous outputs, and the people promoting them have displayed serious qualitative misunderstandings of vehicle dynamics. I have never had a client using any simulation program.
The basic idea is to start with a set of accelerations, perhaps derived from instrumented testing on the track we are setting up for, and estimate what the normal and ground plane forces at the four tires would be. Knowing that, if we are good, we can calculate actual normal forces at the four contact patches, and see if they look like our initial estimates.
Once we have a set of displacements, normal forces, and ground plane forces that look believable, we might then go to tire data and try to predict what slip angles and percent slip values the tires would have to have at those forces, and/or if they would be incapable of generating those forces. That would tell us if the car as modeled is oversteering, understeering, or in the wall.
Even if we suppose that everything but the tires is properly modeled and analyzed, the behavior of the tires themselves is very complex. The tire companies have good reasons for not publishing data.
First of all, to get tire data, the process is to test the tire on a TIRF machine, which rolls the tire in a controlled manner against an abrasive belt that simulates a road surface, and measures the forces. This has to be done over a range of camber angles, normal forces, slip angles, percent slip values, and so on. Ideally, we want to use a variety of inflation pressures as well, and maybe a variety of rim widths.
Thatís a lot of runs. And every run changes the tire. The tire gets heated and cooled. It gets scuffed and worn. To minimize these changes, it is common to run the tests at very low speed. But does the tire act the same at roughly room temperature being rolled on a belt at 3 or 5 miles an hour as it does at racing speed and temperature? Probably not. On an oval, the tire doesnít even have identical properties during the first and last parts of a turn, because it heats up appreciably during the turn.
Even when the process is simplified, say by testing at only one inflation pressure, the raw data exhibit a lot of scatter. It then falls to engineers to curve fit this raw data, nowadays typically using the Pacejka equation format. It is possible to do this, but when the raw data have a lot of scatter, even a good curve fit will have large deviation values. This does not mean the process is worthless or futile, but it does mean that predictions based on the model will have a sizeable uncertainty window.
And thatís just the tire. The tire doesnít do anything by itself. Not only does it need a car, it also needs a road surface. The road surface is commonly represented in tire modeling simply by a coefficient indicating the relative grippiness of the surface. Reasonable enough, but how precisely can we know that? We can back-calculate from test data, and try to figure out what reasonable grip coefficient values might be for that test, to make the rest of the modeling make sense. But do we then know what those values will be at a different temperature, or after a rain, or after a few seasons of use?
Bottom line: there is inevitably considerable uncertainty in even the best modeling, because there is considerable uncertainty in the behavior of tires and pavement. This does not mean modeling and simulation are junk, but it does mean that they cannot reduce car behavior to precise predictability. Rather, they are an enhanced means of getting to the thing some of us imagine them to be a way of avoiding: qualitative insight.
ROLL CENTER DETERMINATION WITH BEAM AXLE SUSPENSION
Should the rear roll center be determined like the front, using a resolution line and average height at the resolution line? This is with a solid axle and Panhard bar in front of housing, with Panhard bar mounted to right side of chassis further forward from axle than its left end in plan view.
The method I have described, in other newsletters and in my video, applies to independent suspensions. It is a way of allowing for the fact that the right and left wheels connect to the sprung structure through two individual uprights and sets of links or arms. Each of these separate systems has its own force line slopes and corresponding jacking coefficients. When these are unequal (which is most of the time), the difference of the jacking forces that create geometric anti-roll and anti-pitch moments will vary according to the distribution of ground plane forces.
With a beam axle, at least as regards lateral accelerations and forces, thatís not so. The right and left wheels are connected to a common rigid structure (the axle), and the suspension linkage connects this whole assembly as a unit to the sprung structure. The linkage forces then depend on the whole lateral force exerted by the sprung structure upon the axle and wheel assembly. This means that the amount of geometric roll resistance generated is sensitive not to the right/left distribution of the ground plane forces, but only to their sum.
To find the roll center for a beam axle with a Panhard bar, to be really accurate we need to find the axis of rotation about which the axle moves in roll, with respect to the sprung structure. We then find the point where this axis intersects the axle vertical plane Ė the y-z plane containing the axle centerline. As an approximation, when the Panhard bar is a full-length one rather than a shorty, and itís fairly close to level, we can take the height of the bar midpoint as the roll center height. We can at least be fairly confident that if we raise that point, we will have more geometric roll resistance, and if we lower it we will have less.