In the following, AMC is presented with some simple cases

A) A first order system. The process parameters were identified from an open loop step test
B) AMC performance in the presence of model errors
C) An extremely nonlinear process
D) A more challenging situation: Inverse response and noise
E) Advantages of AMC in case of rate-of-change clamps on the output.

These examples also address and contradict some of the typical defensive arguments against Model Based Control:
- A model is always extremely difficult to develop
- My process is non-linear – a linear model cannot cope
- MBC is always instable if the model is not perfect.

A) The principal behavior of AMC - shown with a first order system

A simple first order process with deadtime. The controllability ratio CR is around 0.5 – indicating an easy process. In the second test the deadtime is increased from 10 to 60 seconds, the CR is now 3.33, the situation not so easy any more. But the response curve is just shifted by the deadtime. No re-tuning is needed, just adjustment of the model dead time.

B) AMC performance in the presence of model errors

Even with 50% error there is no sign of any instability. The same holds true for combinations of errors. Note that the behavior of the controller output also gives a good indication of the model error – which allows further improvement of the model.

C) An extremely non-linear process

We are controlling here a process that is extremely non-linear and the operating point is actually very close to a maximum – still, AMC is very stable.

D) A process with inverse response.

The open loop test shows clearly inverse response. First the process parameters were identified and on that basis the PID settings calculated. The PID dilemma: Either the response is extremely slow or the loop oscillates. In the last step, AMC is tuned for smooth yet relatively tight response. The result speaks for itself.

E) Rate –of change clamps – an often overlooked problem

The problem with rate-of-change (speed) clamps: The performance of the PID depends very much on the initial ‘proportional kick’. If this ‘kick’ is blocked by the clamps, then the PID loses performance. More P-action cannot help since blocked by the clamps, but with more I-action we will overshoot. MBC does it clearly better - as the picture shows.

---

ACT - D.I. Hans H. Eder KEG
Wienerstr. 10, 3443 Elsbach, Austria
Brussels office: Madeliefjeslaan 13, 3080 Tervuren
Phone ++32-(0)2-767-0895, e-mail: office@act-control.com