F1technical.net

Hi,

I have some data from FS event for skidpad, and I have plots for Yaw rate, lat acc and steering angle from it. I was wondering if anyone had any diea how I could use this data to determine the understeer rgadient and maybe even propose some improvements using some tyre data, as I know that the understeer gradient is a function of the cornering stiffness and the weight distribution. Any ideas?

Thanks!

If you are an engineering student->If you ever go outside you'll see tall vegetables with hard brown stems and green tops. We call these trees. People cut them down and turn them into paper. Then they write stuff in them and call them books. Doug and W Milliken wrote such a book, called RCVD. I suggest you read it. If you are in an FSAE team there will be many copies floating around.

If you are just generally interested->if you''ve got data from a constant radius test then plot SWA vs latacc. If the plot curves upwards you have understeer, is flat you have neutral steer, and if it curves down you have steady state oversteer. It is unlikely you want that. This is the understeer gradient at the steering wheel, not really much use. If you divide the gradient by the steering ratio you get the vehicle level understeer. This is where you start to see some useful stuff. The deviation of the line from neutral steer shows how much 'extra' slip angle the front axle needs relative to the vehicle. Since you know the weight distribution you can work out how much force is required at each axle for a given latacc. Then look up Bundorf analysis on wiki. I'm ignoring the yaw rate stuff, that's all in the textbooks (Pacejka's book is probably the best academic ref for it). It is very useful to work out the lateral velocity at each axle. Then you can work out the slip angle at each axle. And that gives you the axle cornering stiffness. If you are going to get into this I strongly recommend analysing your data in Octave rather than Excel.

I'm not entirely sure you have enough data to do the whole job, we measure vy directly, routinely.

As Greg said if you want to take a data driven engineering approach you need to have a pretty solid understanding of vehicle dynamics fundamentals which you can get from Milliken (read at least the tyre chapter and the steady-state stability and control chapters) or Guiggiani. Avoid the Carroll Smith stuff at least until you've got the fundamentals sorted.

Second issue is that the FSAE event skid pad data won't tell you so much because it's not a standard ISO/SAE test but a race around a circular path. So it's more or less constant speed and constant radius. It will only give you a tiny portion of your understeer diagram. Additionally, a good testing methodology would call for multiple repititions of the same test (at least 3 times left and 3 times turning right) and some sort of averaging to combine the results into a single data set.

All you can really get from the competition skid pad data is the limit OS/US condition (by comparing the actual steering angle to the ackermann steer angle) and the peak lateral acceleration your driver could manage on that surface.

We actually figure on 5 good runs in each direction, each of which meets specified statistical properties.

Thank you guys, I will do what Greg says and see what I can do with it. I assume that's where I start playing with different tyre setups and looking at tyre data to make those adjustments. I'll have a proper read through Milliken, it's just there's so much in there on loads of different stuff

Yaw velocity at constant speed on a constant radius won't help much, it is just v/r.
You can probably get away with just chapter five in milliken.

Looking at Bundorf cornering complainces, biggest afctor seems to be the weight/cornering stiffness distribution (as expected). For the weight, I guess that the change in weight distribution and cornering stiffness has to be accounted for depending on what g's you're reaching throughout a turn?

Yeah sorry, I thought you had run a variable speed at constant radius, which is a lot more useful analytically than just a steady state max latacc run.

I'm not sure how much you can usefully extract given the small number of channels of data. Also, was the run tail out oversteer lots of throttle? That complicates the analysis no end, as the 'forward' thrust helps develop latacc.

Can you list what channels of data you actually have?

I can send the whole list this weekend hopefully. Not sure what you mean by the oversteer on throttle bit?

Well there's the engineers way of getting around a skid pad, which is terminal understeer, and the drivers way, which is to hang the tail out and smoke the tires and opposite lock. Drawing a freebody diagram of the latter is instructive.

I see what you mean hah. No I don't think so... I plotted the graphs of SWA vs time and there are a lot of spikes on the steering wheel (corrections), and the fact that it was done on a wet day (comp data) might not hep much either....

Hi, sorry for the late response. Below you should see a picture of the channels I have available. I also have the data for an actual track and not just skid-pad, maybe could use that instead?

Running in the wet only tells you about wet weather performance I think.

What does gpsCourse tell you ? is that heading or track? Ideally you need both

Between the gps stuff and the other relevant channels you have some redundancy (a good thing) if you were to try and build a model of the lap (a la LapSim or equivalent), but if you specifically wanted to build an understeer budget then a lap is a fairly messy way of reverse engineering it. That's why we use swept steer or constant radius. You are sorely lacking any direct measurement of axle lateral velocities (the gps can be used to synthesise them but it is a noisy calculation).

I'm sure that is the heading of the car, and I have the track as well.

SO to summarise, with the skid pad data I have from the comp, I will not be able to get much out of it?

Noisy data, in the wet, at one latacc, won't tell you much. But you should be able to work out the slip angle at each axle, although as I said it is an error prone calculation.