Engineering FYP proposals

[tmapv]

What I was proposing was not to write a vehicle simulation in Matlab (though this is totally possible and has been done many times before). I proposed that you write a program that takes simulation or test data and analyses it in order to quantify the vehicle's handling performance.

All of the major vehicle manufactures have software like this developed in house and some students have done this already in FSAE.

Have a look at the ISO/SAE standard for:
Steady state cornering
Transient response
Power on (actually not a huge fan of this method but useful as a starting point)
Power off (not a huge fan)

Your program would then take the test data, perform the necessary sensor processing (filtering, corrections, transformations etc) and then calclulate various metrics as defined in the ISO norms, Milliken, Guiggiani etc..

That way, you can quantify all your decisions - both to appease design judges as well as fellow team members.

Otherwise - what else are you going to say when someone asks you why have you chosen this much bumpsteer? Why have you complicated your life with push rods and rockers? Then there is usually a team member telling you that beam axles are the way to go (they are probably right too). How else can you answer these questions or make these decisions in an informed way if you have no way of quantifying the handling of the car?

Like I said before its THE biggest thing missing in FSAE at the moment. Chassis/suspension choices are almost always made with an equal mix of rules of thumb and handwaving which is the polar opposite of engineering.

Ok, so I had a quick look at the standards, and I will have to use them to first of all have a vehicle dynamics simulation set up and use them as a test, so for steady state cornering I will have to make a Matlab model of a car going round a path of a constant radius with increasing speed (from what I found so far for your example). Carry out all the other tests according to the standards, and then use matlab (or some other software?) to evaluate the data and find the optimum dynamical set up so I can justify my choice of the set up?

I also had a look at some of Bill Cobbs stuff, most of it seems tyre related, but it looks pretty complicated, but at the same time I think he has written all the code to calculate the tyre performance output (for example)?

[Tim.Wright]

Bill Cobb's stuff is tyre related because 99% of handling is the tyre.

If you want to write a handling evaluation program in Matlab, I don't recommend building a matlab model of the car at the same time as both activities require around a year of work to get something half decent running. Doing both isn't a viable option you will have to do a hack job in order to finish in time and you won't learn anything.

Another point is not to aim to be doing an optimisation on your car's design and setup while you are writing the analysis program. Same problem as before. In one year you only have enough time to write the analysis environment, or use it - but not both.

To test the application you'd need either test data from a running car or a simulation model which lets you run open loop steering tests (hint most of the motorsport based laptime simulators don't have this capability). Pretty much all simulation packages have demo vehicles which run more or less ok which you can use to simulate maneuvers to generate the data that you need. The models don't need t be too accurate to give you something to work with.

I'd recommend reading Milliken chapters on (going by memory):
The problem imposed by racing
The tyre
Steady state stability and control
Transient stability and control

This will tell you a large part or what you need to be extracting from the test/simulation data. Guiggiani's book has a load of other interesting things if I remember correct.

Your direction will depend a fair bit on what you have available in terms of running vehicles, design data or simulation tools as well.

[tmapv]

I will have a look at these chapters for the data extraction. The thing I'm still unsure about is the 'post prcoessing' of the data. Are there any books you know of that deals with this too? I will do some more research after this weekend to try and put everything together in terms of the procedures I would have to carry out for an FYP of this type.

So in terms of the data gathering, I presume you mean I could replace the whole simulation process (of the car) with simple data acquisition of a real car? We have telemetry on the FS car, I could potentially gather some data also from another racing team who run a number of single seater cars.

[Tim.Wright]

The post processing IS the data extraction.

For example - you perform a frequency sweep test. You take the raw data, filter it (if it's from a running car), transform the acceleration and slip measurements to the CG of the vehicle, perform an fft on the input output pairs eg (LatAcc/swa, Yawrate/swa, roll/swa, roll/LatAcc) and then you can extract metrics such as the yawrate natural frequency, yawrate peak ratio, phase margin for yaw and LatAcc, LatAcc-Yawrate delay...

In short, the post processing is the transformation of measurement data (or simulated data) to useful handling parameters.

[tmapv]

Right that makes more sense now, I think I was just missing that bit the whole time

Okay so what I can do is aquire the data (data acquisition of a running car or simulation but less preferred at this stage?). I can go through the books, and see what handling characteristics info can be output from the data if correctly transformed. So I'll have to use the formulae which are also going to be mentioned in the book to help me do that, and subsequently make a matlab script to do all of the transformation of data to useful info. Think I got it up to there?

So how could I use Bill Cobbs tyre analysis if say I was to aquire all of the data from the real car anyway?

[Greg Locock]

Because you won't get real time tire data from a car, unless you spend a couple of million installing wheel force transducers.

Here's how I correlate a model to real life
0)hardpoint geometry
1) weights, cg location, moments of inertia - sprung and unsprung
2)springs sstbar jounce bumper rebound bumper bushes vs kinematics and compliance test
3)shocks (just type them in)
4)tires from Flattrac machine
So now we've got a vehicle that behaves statically like the real thing, and has some sort of tire and shock

Now do a constant radius or swept steer test. They are subtly different, either will do.

So, if you have enough instrumentation on the real car then you can compare cornering stiffnesses with your model.

Not surprisingly running on asphalt is different to running on sandpaper, so you'll get some odd results.

Then run frequency response, and/or step steer.

If you can get good agreement with that first time through then you have done well.

Notice that all the above could be done with a modified bicycle model, as shocks don't affect those tests much (slight effect on the roll metrics in FR, but not a biggy).

If you really want to get into MBD but suspect that ADAMS is going to eat your time budget, the stable and somewhat easier to use CarSim has its plus points.

In practice I have some more linear range tests, but that's because we obsess about steering feel.

Then we go above 0.6g into non linear stuff and dynamic events, but that's for me to know and you to find out,

And then we get into ride. Ha.

You have to get the correlation done in the above order, anything else is madness. You won't need to worry about ride.

[tmapv]

Okay so what if I can't get the real car data. I can still use the simulation, and the bicycle model which still can be done in matlab? I just won't have the correlation to confirm validity. And then separately (after) do the post processing of that info in matlab too?

[notsofast]

it's almost as much about demonstrating your project management ability as it is about content

This!

It's been a few decades since I got my first job. The ability to plan, execute, and finish a project ended up being perhaps the most valuable skill I learned. It has served me well in a career that has little to do with my original field of study.

[krisfx]

it's almost as much about demonstrating your project management ability as it is about content

This!

It's been a few decades since I got my first job. The ability to plan, execute, and finish a project ended up being perhaps the most valuable skill I learned. It has served me well in a career that has little to do with my original field of study.

It's what I was told all through my FYP, though I wish I'd done something CFD/computational as opposed to basic design principles and methodologies, what I learned has helped and I still get to dabble in CFD these days anyway

[Greg Locock]

"Okay so what if I can't get the real car data. I can still use the simulation, and the bicycle model which still can be done in matlab? I just won't have the correlation to confirm validity. And then separately (after) do the post processing of that info in matlab too?"

Sure you can do that. I've even see a Master's thesis that did that. It isn't engineering in my opinion, it is playing about with computer programs, so make sure your professor understands what you intend to do, and see if you can wangle some real world data into it.

[tmapv]

"Okay so what if I can't get the real car data. I can still use the simulation, and the bicycle model which still can be done in matlab? I just won't have the correlation to confirm validity. And then separately (after) do the post processing of that info in matlab too?"

Sure you can do that. I've even see a Master's thesis that did that. It isn't engineering in my opinion, it is playing about with computer programs, so make sure your professor understands what you intend to do, and see if you can wangle some real world data into it.

Great. I'm going through some of the books to get a better understanding of this before I start (officially) in a couple months. Considering I model the vehicle as a bicycle model (which I think both you and Tim were proposing), I am limited to not being able to calculate body roll or the weight transfer (lateral). Would this prevent me from calculating some major vehicle handling characteristics?

[Greg Locock]

There is a modified bicycle model that usefully includes some of the effects of 4 wheels. Specifically you can work out what the weight transfer is and then modify the tire characteristics to be the average for each axle. By then you've worked the roll out as well no doubt. I've never done it. The basic bicycle model is actually pretty good for handling up to about 0.5g

[tmapv]

I looked for something that does but couldn't find much. I found this on mathworks:

http://de.mathworks.com/help/ident/exam ... ystem.html

Was it something like this you had in mind. They still refer to it as 'bicycle model'

[Greg Locock]

That sort of thing, tho I can't tell if it rolls, at first sight, not.

[guiggiani]

The so-called bicycle model (better, single track model) can easily take into account lateral load transfers, roll steer, roll camber and many other setup aspects. Moreover, it can deal with high lateral accelerations. The only real limitation is the open differential. Unfortunately, the single track model is often presented without a clear explanation of the underlying assumptions. In my book "The Science of Vehicle Dynamics" I spent a lot of pages to state what the bicycle model is all about.