Peter Nachtwey
Member
Andybr, if you look at the Advanced Conrol thread you can see all I needed is one ( initial and final ) set of data from a step response. I can the change the closed loop time constant to make the response faster or slower using the initial set of data. Plus I can simulate the response quickly like Gerry did. On processes that are dangerous or take a long time, similations likes Gerry's are the way to go since they can be done off line.
The ZN methods we developed before we all had PCs with lots of processing power. The ZN methods are not statistically sound because one is using just a few subjective points to determine the gain.
Parting shot. I think it is a crime that the people who build the plants don't provide a good estimate of what the gains should be. If they truely engineered the system they should know. As a customer I wouldn't accept the system without a model. If the engineers don't have a model then they really don't know how their system works.
The ZN methods we developed before we all had PCs with lots of processing power. The ZN methods are not statistically sound because one is using just a few subjective points to determine the gain.
Parting shot. I think it is a crime that the people who build the plants don't provide a good estimate of what the gains should be. If they truely engineered the system they should know. As a customer I wouldn't accept the system without a model. If the engineers don't have a model then they really don't know how their system works.
