Paint Flow Control Project, Fuzzy Logic Revisited, 3.15

Me with a painting by my wife Heide in the background.

Paint Flow Control Project Status, August 15, 2015:
We're still in limbo, no, not dancing Limbo, but awaiting the Robot system from Ontario. I understand that money has been paid or put in some sort of escrow, but the Robot system is locked up in a warehouse somewhere in southern California.

So I am continuing to work on the Fuzzy Logic algorithm, after getting my Python simulation code running pretty well. I had been fairly optimistic about my FL algorithm but checking it again, I see that the errors are often rather large, just less than 10%, and I would like to reduce the errors to near 1%. I don't think I can admit that my code only controls the paint flow to a 10% accuracy.

Paint Flow Control Simulation from my Python Code.

I will look at means to improve the algorithm and this may take a week or two to make significant improvements.

There are a couple of ways I was thinking of going. I could use the two-stage approach described by Li and Lau in their IEEE Control Systems paper on Fuzzy Logic ("Development of Fuzzy Algorithms for Servo Systems" by Y.F. Li and C. C. Lau, IEEE Control Systems Magazine, April 1989, pp 65-71), possibly use NeuroFuzzy ideas from the book by Constantin Von Altrock (Fuzzy Logic NeuroFuzzy Applications Explained) or maybe just roll my own method.

Although I had done some simulation of Li and Lau's Coarse Look-Up-Table (LUT) and found considerable overshoot in the response for my Paint Flow Control simulations, I will first take a look at it again and first see if I can find the cause of the overshoot and correct it.

So stay tuned.