Rule-of-Thumb Methods
Rule-of-thumb or heuristic methods were developed for cases in which modelling gives poor results or simply no model is available. They are used when the way the system works is not precisely known. We treat the system as a black box and control it without knowing its internal structure and behaviour. The heuristic methods are summarised in a very good Wikipedia article.
The inflection tangent method, which you got to know in the chapter Non-oscillating systems, falls into the category of rule-of-thumb methods. In this tutorial I only go into the Ziegler–Nichols method in more detail. There are others that are also quite good.
Ziegler–Nichols method
You can only use the Ziegler–Nichols method if your system is capable of oscillation (i.e. has at least two storage elements). It is important that the system cannot be damaged by overshoot. If these conditions are not met, use another method. With Ziegler–Nichols you proceed in the following steps:
1. Control the system with a P controller
2. Increase KPR until a stable sustained oscillation sets in. The minimum value of KPR required for this is called KPR,Krit (critical gain).
3. Measure the period of the sustained oscillation. It is called TKrit (critical period).
4. Use these two parameters to design the controller parameters according to the following table:
| \(K_{PR}\) | \(K_{IR}\) | \(K_{DR}\) | |
|---|---|---|---|
| P controller | \(0.5 \cdot K_{PR,\mathrm{Krit}}\) | – | – |
| PI controller | \(0.45 \cdot K_{PR,\mathrm{Krit}}\) | \(0.53 \cdot \dfrac{K_{PR,\mathrm{Krit}}}{T_{\mathrm{Krit}}}\) | – |
| PID controller | \(0.6 \cdot K_{PR,\mathrm{Krit}}\) | \(1.2 \cdot \dfrac{K_{PR,\mathrm{Krit}}}{T_{\mathrm{Krit}}}\) | \(0.075 \cdot K_{PR,\mathrm{Krit}} \cdot T_{\mathrm{Krit}}\) |
The table gives numerical values of the controller parameters for pure P controllers, PI controllers and PID controllers. A PI controller is a PID controller with KDR = 0.
If you need the parameters for the alternative representation of the controllers, use the following table:
| \(K_{PR}\) | \(T_N\) | \(T_V\) | |
|---|---|---|---|
| P controller | \(0.5 \cdot K_{PR,\mathrm{Krit}}\) | – | – |
| PI controller | \(0.45 \cdot K_{PR,\mathrm{Krit}}\) | \(0.85 \cdot T_{\mathrm{Krit}}\) | – |
| PID controller | \(0.6 \cdot K_{PR,\mathrm{Krit}}\) | \(0.5 \cdot T_{\mathrm{Krit}}\) | \(0.125 \cdot T_{\mathrm{Krit}}\) |