Hi guys and gals

I have a severely non linear system that i am trying to develop a controller for. 

The non linear curve can be broken down into 3 linear-ized sections. 

With a single fine tune in the s7-1200 i can get close but when the controller set point travels from one linear region to another we experience large overshoots and large errors.

Research tells me that gain scheduling may be an idea. a set table of gives each used for a different range of the process variable. 

Has anyone implemented this before?

Recommendations on how to implement this?

Thanks

Look-ahead PID control, is the term your looking for.

Identify when to make the transition and what that curve should look like. Put the PID in manual and make sure its set for bump-less transfer. Manually play your curve out of a cyclic OB.

This is very over simplified. The first sentence is the hardest. Experience required.

Thanks for the reply

Is there anyone that can elaborate a bit more on this

I am curious how to go about implementing this.

OK some more detail: Not knowing your application, Lets say you have a curve that looks like Figure 1.You have tuned the PID and found the values to control the 3 sections of this non-linear curve will work for this control scheme. However, when you switch the PID values there is un-rest.

Bump less transfer insures that the Integral hasn't "wound up" and the start CV of the PID will pickup where you left off. Update your manual CV value in the same cyclic OB as your are calling the PID. The control will be more predictable.

Now - your transitions are the issue. So When you trigger Transition AB or BC. Read the current CV and Put the PID in manual. Modify the CV to the start position you have worked out will transition without causing disruption in product. Update the PID parameters to the next set of values: B or C respectively. And put the PID back in Auto.