PID Tuning

PID Parameters
The Ziegler-Nichols Rules
Ziegler-Nichols: Method 1
Ziegler-Nichols: Method 2
Computational Search

Unit 8: Part 3: PID Tuning

Engineering 5821:
Control Systems I

Faculty of Engineering & Applied Science

Memorial University of Newfoundland

March 31, 2010

ENGI 5821 Unit 8: Design via Root Locus

 PID Tuning
PID Parameters
The Ziegler-Nichols Rules
Ziegler-Nichols: Method 1
Ziegler-Nichols: Method 2
Computational Search

1 PID Tuning

1 PID Parameters

1 The Ziegler-Nichols Rules

1 Ziegler-Nichols: Method 1

1 Ziegler-Nichols: Method 2

1 Computational Search

ENGI 5821 Unit 8: Design via Root Locus

PID Tuning

Designing a PID compensator using the analytical methods

discussed so far requires a mathematical model of the system.
However, PID controllers (compensator and controller are
synonymous in control systems) are often used even when no such
model exists. The process of adjusting the parameters of the PID
controller in this situation is known as tuning.
We will consider two options for tuning a PID (or PI or PD)
controller in the absence of a system model:

Ziegler-Nichols rules (rules of thumb)

Computational search
PID Parameters

Recall the parameters for a PID compensator:

1
Gc (s) = K1 + K2 + K3 s
s
We will use the constants Kp , Ki , and Kd which stand for
Proportional, Integral, and Derivative.
1
Gc (s) = Kp + Ki + Kd s
s
If Ki and Kd are zero we have a simple P controller.
If only Kd is zero we have a PI controller.
If only Ki is zero we have a PD controller.
The Ziegler-Nichols Rules
Ziegler and Nichols came up with two methods for setting the
parameters of PID controllers. These are rules of thumb and
there is no guarantee that the resulting system behaves optimally.
Ziegler-Nichols provides only a starting point for further tuning.

Method 1: Applies if the system’s response to a unit-step is

S-shaped, indicating that the plant involves no pure
integration and the system response is not dominated by a
pair of complex-conjugate poles:

Notice that this method is applied on the plant itself, without

feedback.
Method 2: The system appears to involve some pure
integration and/or dominant complex-conjugate poles (i.e.
the response is similar to an underdamped 2nd order
response). This method is applied on the closed-loop system,
with feedback.
Ziegler-Nichols: Method 1
The system’s S-shaped response can be characterized by two
constants, the delay time L and time constant T . These
parameters can be obtained by drawing a tangent line at the
inflection point of the curve:

L is the intersection of the tangent line with the time axis. L + T

is the time at which the tangent line intersects the steady-state
value.
The following table gives the gains for Method 1.

Controller Kp Ki Kd
P T /L 0 0
PI 0.9T /L 0.27T /L2 0
PID 1.2T /L 0.6T /L2 0.6T

A PID controller tuned by this method has a pole at the origin and
double zeros at s = −1/L:

1
Gc (s) = Kp + Ki + Kd s
s
1
= 1.2T /L + 0.6T /L2 + 0.6Ts
s
 2
2
s + L1
= 0.6T
s
Ziegler-Nichols: Method 2

To apply the second method we do a test on the system that varies

Kp while keeping Kd = 0 and Ki = 0. The system being tested is
as follows:

Kp is increased from 0 until it reaches a critical value Kcr at which

the output exhibits sustained oscillations...
At Kp = Kcr the system’s output will oscillate with period Pcr .
These two values are used to determine the PID gains:

Controller Kp Ki Kd
P 0.5Kcr 0 0
PI 0.45Kcr 0.54Kcr /Pcr 0
PID 0.6Kcr 1.2Kcr /Pcr 0.075Kcr Pcr
Computational Search
If we have a model of the system or if the system can somehow be
simulated, we can simply search for the PID parameters that best
satisfy our design criteria.
e.g. Assume we have the following system:

Like the controllers produced by Ziegler-Nichols, this PID controller

has a pole at the origin and a pair of double zeros at −a. This
gives us a 2-D parameter space which is relatively easy to search.
The goal is to satisfy the following requirements:

%OS < 10%

2≤K ≤3
0.5 ≤ a ≤ 1.5

We choose a step size of 0.2 for both K and a.