I have noticed that in my oil production plant the system engineers use trial and error method for loop tuning instead of ziegler-nichols or lambda tuning method. I checked some of the loops and they seem to be tuned properly. So my question is is it practical to enforce lambda tuning method for all future loop tuning or should we not attempt to fix something that isn't broken? What issues do you see arising from either scenario and what would you prefer?
ziegler-nichols or lambda tuning methods work in theory, but in actual plants useful only as a starting point.
Regardless of the method used, it is all driven by your process, and matters the theory does not take into account.