Tuning means choosing $K_c$, $\tau_I$ and $\tau_D$. Model-based rules start from the FOPDT parameters $K$, $\tau$, $\theta$ of Lesson 7.
IMC (lambda) tuning for PI control
Choose a closed-loop time constant $\tau_c$. Larger $\tau_c$ gives slower, more robust control. A common choice is $\tau_c \ge \theta$ (or $\tau_c = \tau$ for very smooth control).
$$K_c = \frac{1}{K}\,\frac{\tau}{\tau_c + \theta}, \qquad \tau_I = \min\{\tau,\ 4(\tau_c + \theta)\}$$
Ziegler–Nichols (ultimate gain) method
With P-only control, raise $K_c$ until the loop oscillates steadily. Record the ultimate gain $K_{cu}$ and period $P_u$.
| Controller | $K_c$ | $\tau_I$ | $\tau_D$ |
|---|---|---|---|
| P | $0.5K_{cu}$ | – | – |
| PI | $0.45K_{cu}$ | $P_u/1.2$ | – |
| PID | $0.6K_{cu}$ | $P_u/2$ | $P_u/8$ |
Ziegler–Nichols settings are aggressive (quarter-decay response). In practice many engineers start there and then reduce $K_c$.
Judging the result
- IAE $= \int|e|\,dt$: smaller is better.
- Overshoot and settling time for set-point changes.
- Robustness: does it stay stable if the process gain or dead time changes by 20–50 %?