# Thread: Traction control setting based on data acquisition

1. ## Traction control setting based on data acquisition

Hi everybody, I am Federico Scacco, I join the Race UP Team of the University of Padua, Italy.

I would like to show you a graph which I consider really interesting, although I never saw it on vehicle dynamics' book.

Wheel_slipVSacceleration.jpg

The "y" axis represents the longitudinal acceleration of the car (negative values correspond to higher longitudinal acceleration) while the "x" axis represents the wheel slip channel.
The graph is obtained for different TPS positions, in this configuration I chose the TPS between 90 and 100% which corresponds to maximum performance required by the driver.

Our car is equipped with Hoosier tires. Do you think this data representation could be useful to obtain information for the traction control setting?

Thank you a lot, don't hesitate to report suggestions or critical, which will surely be useful for everybody.

Federico Scacco - Race UP Team - University of Padua

2. Federico,

This is a pretty standard plot to look at. You might want to redo the test for pure acceleration. You can get a pretty good idea of longitudinal tyre performance from a few hard launches. The information you get can be very useful for traction control tuning.

Kev

3. Thank you a lot Kevin,
I found your answer very useful, since I never depict data in this kind of graph. Can somebody suggest me literature in which this kind of information are explained?

Could somebody post here their own data in order to compare them and make some conclusions useful for this topic? As I mentioned above, we use Hoosier tires, 13".

Wheel_slipVSacceleration_longitudinal.JPG

This graph shows the same data of the last post only for the acceleration.

Federico

4. ## Slip Mu Sing

Seriously, a Google search on just Images alone shows hundreds of similar thumbnails related to this subject. Reduce the fuzzy scatter in your plot by taking into account dynamic toe, camber and suspension travel effects. The tire data all by itself is the spine of your results. Explanations of this effect probably won't be found on Facebook ...

5. Ok Bill, thank you too. Although this trouble seems to be well known by a lot of students (or ex students) here, I am sure that a lot, like me, never obtained this kind of graph.
In formula SAE competitions the first lesson is to respect everybody who want freely explain their own troubles with the aim to share their knowledge with others.

Coming back to the matter, you suggest to modify the graph keeping into consideration the kinematic effects. If I am not wrong, we should "filter" the longitudinal acceleration by removing the effect of suspension travel, toe and camber. Did I correctly understand?

Federico

6. ## Yes !

I would much prefer you to obtain some tire data (as in FSAE- TTC for your specific tire) and use it in a model (as in Simulink for example) of a closed loop braking or traction control simulation. A simple quarter car model would have enough complexity to study theoretical situations. For example a Pacejka model with just 4 terms that produces a realistic looking mu-slip curve would be very informative. Your results, even without load transfer or elastci and kinematic suspension compliance(s) should overlay a spine on the plume of measured data you have shown us.

THEN: you step up the complexity in a graduated manner suited to the exploration of 'other' factors that scatter your 'perfect' simulation results. Ultimately, you must consider suspension characteristics, dynamics weight transfer, powertrain dynamics and launch effects, throttle response, hydraulic plumbing functions for brake activation, tire heating, axle hop and tramp responses, unsprung mass effects and even driver application methodologies. This would be an application of Engineering Principles (Da Vinci style) instead of spending time at the track "spinning your wheels", so to speak ! You will need some help from collegues to obtain subsystem and component performance parameters, etc, but the learnings should be obtained before you twiddle some settings in the TCS controller. This notion supports Team Building by your contribution(s), drives a Systems Engineering process, and makes an AWESOME presentaion to other professionals, educators and design judges.

7. I'll play devil's advocate here. The typical traction control setup in these cars is reasonably limited (although less than what a lot of students believe). A good design of experiment approach and a few parameters easily changed through rotary pots on a dash can make for incredibly quick development times.

Pretty much all of this engineering is in a constant look of testing and dynamic model development. I would argue that the first round should be the testing before the first dynamic model, at least then you have a few results to compare your dynamic models to. There is something in the seeing of these tests that is very valuable for dynamic understanding. I find with students one of the most valuable measuring devices at a test day is a high speed camera. Going over both the data and the slowed video footage helps show a lot of the reality of kinematics, compliance, reaction loads, transient motion, and so on. Doing this along with a basic understanding of tyre behaviour, and the nature of the parameters that can be changed (rather than what you might want to change) will take teams a long way. This approach saw the ECU team change their car from a 4.0-4.1s accel car to a 3.8s car in about one day (following a couple of days welding up a few extra rear suspension points).

For Federico I will add that it is easy to get carried away with depth to tyre modelling. Bill's suggestion of a 4 term Pacejka model is a pretty good start. But in the process please remember that what you are playing with is predominantly the friction of a chunk of rubber with some sliding velocity to the road. You should develop a basic understanding of the mechanisms behind the friction between rubber and other surfaces, as well as a rudimentary understanding of the material properties of rubber, and the method of its manufacture. A slip ratio at a known speed can be back calculated to a longitudinal sliding speed. A slip angle can also be converted to a lateral sliding speed. If you haven't already, have a look into how the friction from rubber varies as a function of sliding speed.

If you understand the manufacture and chemical composition of rubber, along with the friction mechanisms you will understand the importance of temperature (including temperature cycles). This is where a little understanding of heat transfer can really help. Your rubber cools due to a delta T with atmosphere, and has a knowable delta v with the surrounding atmosphere. The amount of work going into the tyre is easily calculated with limited data. You have everything you need to have a good go at approximating a decent thermal model. If you aren't convinced about the HUGE importance of tyre temp for these cars strap on a couple of tyre temp sensors and monitor performance as a function of temperature. Wont take you long. You might just find that any tyre modelling approach that isn't accounting for temperature may be of limited use. You may even find that the perfect traction control settings at the start of a run, may not be ideal at the end.

Again for Federico please note that testing these cars should be conducted as an experiment. Ideally all testing should be well planned with clear objectives, and all data (including non DAQ data) recorded, organised and analysed. The heart of any disagreement with Bill on this matter is not that dynamic models shouldn't be done (they absolutely should), rather it is that I believe that some decent experimentation should precede it.

I think that this sort of vehicle testing can be broken up into two main classes. Sometimes when you take a car out you are trying to investigate the envelope of the vehicle and develop fundamental understanding. In these circumstances the goal is not necessarily improved performance, but rather an investigation of the sensitivity of various parameters, and the interactions between them. The second type is when you are solely trying to improve the vehicle performance from a particular starting point. The former is akin to broad investigation of the solution space looking with insights into global optimisation, the latter is only good for local optimisation. Also note that you should be adopting similar approaches to your modelling. Only looking at local gradient searches of the solution space leads to the dark side, or at least sub-optimal vehicles.

Kev

8. I apologize for the delay of the answer.

@Bill
I totally agree with you as far as concern introducing variables step by step. I think the best approach toward this problem (but this is true also for others in Engineering) is starting from the basis, as you said.
My initial ideal was not to make a model of the traction control because I don't find it pretty adherent to the reality, since it is everything but easy to keep into consideration all the parameters involved in the phenomena.

I am not sure the "theoretical way" is the best approach because it would be impossible to simulate all the parameters that the ECU (Motec, M400) uses for the traction control strategy. Moreover, the PID which control the amount of engine power reduction in relation to the AIM SLIP, which is the slip desired - the one which permit to reach the best performance- is not easy to implement into a model.

@Kevin
Thank you for the very complete answer.
The advice of the camera is very useful, I will surely keep into mind this in the following months of tests.

Our suspension division has attended some Claude Rouelle's seminars and we know something about our tyres. I am also aware that we commit an error considering the slip ratio as the relative variation of the rear and front wheel speed, but I consider this approximation negligible. Do you agree with me?
For sure, the temperature plays an important role in this matter. While the determination of temperature is possible during test session, this is not true during the dynamic events, in which the temperature sensors are removed. This means that it is not possible to define an AIM SLIP as a function of the temperature. However it is possible to select different traction aggressiveness from a steering wheel's potentiometer, for example one setting which best suits for cold tyres and the other for hot tyres.

I would like to ask you another information. Is it more important to focus our attention in defining the correct AIM SLIP (for example with a tyre model) or to reaching a deep knowledge of the PID strategy with a lot of tests (experimental approach)?

Thank you a lot for the amount of information you summarized in these few posts.

Federico Scacco