Controller

[Mergenthal]

Dear experts,

we have some trubble with controllers.
If our measured variable is too far away from our setpoint we run into the same problems as they were discribed in this thread: http://www.cape-open-forum.org/viewtopic.php?f=5&t=360&p=791&hilit=controller#p791
Thus we decided to take the enthalpy of steam or liquid water as measured variable and calculate the belonging setpoint with a matlab-Unit.
But in a more complex Flowsheet, the Jacobian matrix runs into singularity as soon as the startingvalues are too far away from a good solution and our solver stops.
Is there a solution to vari the setpoint with the help of a matlab component so that this problem solvable?
We tried to let the setpoint drift if the measured variable is farer away from the setpoint than x%, but this doesn't help.
Any ideas?

Best regards
Pieter

[jasper]

This message belong in the COCO forum I think.

The controller simply adds an equation to the system of equations to solve. If you indeed did something similar as the referenced posting, the equation may be stated as: solve the temperature of a stream so that the heat balance is closed, or H(T) = Hspec. The problem here is that H(T) is a discontinuous function of T in the boiling point of a pure compound.

I am not clear on how you attempted to modify the equations using a Matlab unit operation, but what comes to mind is to reverse the problem: run a PH flash at Hspec to find T. Is this an option at all?