PID/FUZZY competition: I pay US $50!!

R

Thread Starter

Rudi

[Editor's note: We are running this post because of its unique subject matter and potential interest to our forum users. Please refrain from posting similar pleas for help.]

Gentlemen!

I see there are many clever brains out there and therefore I am raising my plea to you for a control solution to our "plant": The plant is a so-called Rolling Road (chassis dynamometer). This is a means of performance testing a road car: the driving wheels are running in a roller set, which is usually embedded in the ground. This roller set can be "braked" using a so-called Eddy current brake (retarder), the type which is also used on heavy trucks/tractor units. I am struggling to get a really good controller working for this. The requirements are:
- minimal overshoot or no overshoot
- fast response

The roller set needs to be governed to run at a constant speed. A speed signal 0-5V is avialable. A load signal (how much the retarder is "braking" is also available as
0-5V).

As controller I have available:
MCU PIC 16F877
analog inputs to set GAIN, INTEGRAL, DERIVATIVE
analog inputs to sense speed, load
analog input to sense SETPOINT

I program the PIC in C (as I am not clever enough for Assembler :( )

I would like to try different control algorithms to fulfill my control needs. The person who priovides me with a solution which will find to work to my requirements I shall award 50 (fifty) US$. This I promise by my honour. This is only for solutions which work well!! For a poor
solution I have already (my own :( ...).

If you would kindly send me some algortihms (Basic would be fine) or C for me to try, that would be great. My e-mail is [email protected]. It will take me a while to try them out, but I shall keep you posted on this interesting project. This Rolling Road is for the means of performance tuning engines and we'll have Porsche, BMW, etc. on it. It
will be great :) It is important that the control ensures that the roller speed does not drift off, when the load (braking) is varied. The roller speed must be kept constant as best as possible.

Maybe a "Fuzzy" would be better? Whatever, I will try all possibilities using the PIC 16F877 MCU (8 analog inputs, 20MHz clock). The MCU control output is an analog signal,
by the way (digital outputs load a DAC with the output control signal).

Gentlemen, many thanks for your valued help. Ladies, also invited, of course!

Many thanks !! :) :))

Rudi

P.S. example: if the setpoint was to be 50mph, I would like the rollers to run freely with no braking on at all until maybe 45mph, then the brake should be activated so that there will only be a minimal overshoot of say 10 percent (55mph) and a fast settling. Also, a "learning"
controller... Fuzzy? would be a good idea... what do you think?
 
B

Bill Clemons

Set up your controller for P-I with large gain and large integral settings. The controller should read speed and be fast and sensitive as likely the control loop (speed control) has small deadtime and time constant.

Bring in your load signal from the brake as a feed-forward trim to the controller output. Allow the trim to add/subtract output around your -/+ 5 mph band.

Keep your money, I drive a Chevy ;o)
 
F
Is this $50.00 for steering you in the right direction or do I have to give you something original? I have an engineering book somewhere on my shelves that has a real world example of a Self Organizing Fuzzy Logic Controller (I think) controlling the steering of a bus by following a buried cable. It was done by the Brits. I will have to look for it.

Frank
 
Top