Worklog
3/3/05
So I've been a bad boy and I haven't updated this in a while. I have gotten lots of stuff done on the project. Here's a rough list:
-
Plant model - The linearized model that Dr. Anakwa gave me initially was not appropriate for
development of a fuzzy logic controller. The assumptions are that the plant is stabilized around a
certain set point. This assumption is not valid for a fuzzy logic controller that does not stabilize
the system. All that time I spent trying to develop the controller using that model was fairly
pointless. Now, I have the state-space model, which acts just like the physical system.
-
Using this model, I have been able to create a fuzzy logic controller that stabilizes and controls the
system with the specifications
- Ess average = 50mv
- %OS = 18%
- Ts ~60ms
Here's an image:
-
I've entered this project in the ASEE Student Poster Competition. It will be held at NIU on April 2nd.
- All that's left to do is tune really.
In the following weeks I'll tune, and address some issues. I might add an integrator. We'll see. Here's my schedule: Schedule
11/13/04
Well, Thursday was the third week of lab. I made my first fuzzy controller that was capable of stabilizing the system. It didn't have zero steady state error, but it was a start. The system had 8 membership functions for error, and 7 for change in error, and a total of 56 rules. It turned out that the system was entirely too complex for the xPC. The initial step time of 1ms didn't work, neither did 10 or 100. A step time of 1s just wouldn't do, so i scrapped that idea. I decided to go with a system with four MFs for error and four for change in error. I haven't gotten to the correction voltage MFs yet, but I believe that there should be as few as possible. I may have to look at several options to make the system faster. I can see if the simpler system works better for starters. Or, I can code the entire system in C or assembly. That would definitely speed the system up. It would also make it esier to incorporate genetic algorithms or neural networks into the system. I'm also going to read my book on fuzzy logic control systems to see if there are any ways it knows of to ascertain the stability of a system or make it faster. I also have to check out those books for the entire semester, or i'll start racking up fines.