EE617A

Industrial Automation and Control

(Units: 3-0-0-0-[9]) (2014-15, Second Semester)

• Lecture-20: Tuning of PID Controllers

 PID-Controller (Proportional-Integral-Differential)

integral factor integrator

proportional factor
1 1
Ti command
error s
variable
set-point Kp plant

Td s
process value
derivative PID controller
factor measurement

The proportional factor Kp generates an output proportional to the error, it requires a non-
zero error to produce the command variable.
Increasing the amplification Kp decreases the error, but may lead to instability
The integral time constant Ti produces a non-zero control variable even when the error is
zero, but makes the system instable (or slower).
The derivative time Td speeds up response by reacting to an error change with a control
variable proportional to the steepness of change.
 PID response summary

Plarge (Kp = 15)

less error, but unstable
PI: no remaining error,
but sluggish response
(or instable, if Kp increased)

Psmall (K=5) asymptotic error

proportional only
differential factor
increases responsiveness
load change (load decreases)
0
0 1 2 3 4 5 6 7 8 9 10
Solicited Psmall Plarge PI PID U1
Play with Matlab: http://ctms.engin.umich.edu/CTMS/index.php?example=Introduction&section=ControlPID
or control.xls
 PID-Controller: empirical settings

Rise time Overshoot Settling time Steady-State

Error
increasing
Kp Decrease Increase Small Change Decrease
Ti Decrease Increase Increase Eliminate
Td Small Change Decrease Decrease Small Change
See examples on http://en.wikipedia.org/wiki/PID_controller

Empirical formula of Nichols (1942 !)

step response (open loop)
1.4 delay time constant
1.2 1.2 T
1 Kp = Ti = 2.0 d Td = 0.5 d (Nf = 10)
0.8 KV
0.6
0.4
0.2
gain
Caution: Nichols assumes that the plant
0
0 1 2 3 4 5 6 7 8 9 10 is first order with time delay
d ~ 0.2, T= 1.5s
 Tuning of PID Controllers

The importance of P-I-D controller and the features of P, I and D actions were
elaborated in the last lecture.

It was also seen in the last lecture that the controller could be easily incorporated in a
process, whatever be the type of a process: linear or nonlinear, having dead time or
not. It is needless to say that the controller parameters influence heavily the
performance of the closed loop system.

Again, needless to mention that the choice of the values of the P, I and D parameters
is very much process dependent. Therefore, thorough knowledge about the plant
dynamics is important for selection of these parameters.

In most of the cases, it is difficult to obtain the exact mathematical model of the plant.
therefore, we have to rely on the heavy experimentation for finding out the optimum
settings of the controller for a particular process.

The process of experimentation for obtaining the optimum values of the controller
parameters with respect to a particular process is known as controller tuning.
Nowadays digital computers are replacing the conventional analog controllers.

P-I-D control actions are generated through digital computations.

Digital outputs of the controllers are converted to analog signals before they are fed
to the actuators.

In many cases, commercial software are available for Auto tuning the process. Here
the controller generates several commands those are fed to the plant. After
observing the output responses, the controller parameters are selected, similar to
the cases discussed here.
Since controller tuning is very much process dependent and any improper selection
of the controller settings may lead to instability, or deterioration of the performance
of the closed loop system.

In 1942 two practicing engineers, J.G. Ziegler and N.B. Nichols, after carrying out
extensive experiments with different types of processes proposed certain tuning
rules, these were readily accepted and till now are used as basic guidelines for
tuning of PID controllers.

Subsequently, G.H. Cohen and G.A. Coon in 1953 proposed further modifications of
the above techniques. Still then, the methods are commonly known as Ziegler-
Nichols method.

Substantial amount of research has been carried out on tuning of P-I-D controllers
since last six decades. Several other methods have also been proposed.

Most of these are model based, i.e. they assume that the mathematical model of the
system is available to the designer.

In fact, if the mathematical model of the system is available, many of them perform
better than conventional Ziegler-Nichols method. But the strength of the ZN method
is that it does not require a mathematical model, but controller parameters can
simply be chosen by experimentation.
We would be discussing the three experimental techniques those come under the
commonly known Ziegler-Nichols method.

Now let us look back to whatever discussed already in last lectures. The closed loop
system can be described as shown below:

OR

The error signal is fed to the controller and the controller provides output u(t).

Since the capacity of the controller to deliver output power is limited or even
otherwise, an actuator is needed in between the controller and the process, which
will actuate the control signal.

It may be a valve positioner to open or close a valve; or a damper positioner to

control the airflow through a damper.

The controller considered here is a P-I-D controller whose input and output
relationship is given by the equation:
Our objective is to find out the optimum settings of the P,I,D parameters, namely
Kp, τd and τi through experimentation, which will provide satisfactory closed loop
performance, of the particular process in terms of, say, stability, overshoot,
setting time etc.

There are three methods of tuning PID Controllers.

1) Reaction Curve Technique

2) Closed Loop Technique (Continuous Cycling method)
3) Closed Loop Technique (Damped oscillation method)
 1-Reaction Curve Technique

This is an open loop technique of tuning.

Here the process is assumed to be a stable first order system with time delay.

The closed loop system is broken as shown in figure below and a step input is
applies at m’ , output is measured at b.

The input and the output response are plotted by suitable means as shown.
M,L and K are measured. Let us define the following terms corresponding to Fig. above.
 2-Closed Loop Technique (Continuous Cycling method)

The major objection to the tuning methodology using reaction curve technique is that
process has to be run in open loop that may not always be permissible.

For tuning the controller when the process is in under closed loop operation, there are
two methodologies.

The first one, continuous cycling method is explained as below:

Here, the loop is closed with the

controller output connected to the
actuator input.

amplitude continuously. The output waveform is

plotted as shown in adjacent Figure.

The time period of continuous oscillation is noted.

The recommended optimum settings are:
 3) Closed Loop Technique (Damped oscillation method)

In many cases, plants are not allowed to undergo through sustained oscillations, as
is the case for tuning using continuous cycling method.

Damped oscillation method is preferred for these cases.

Here, the closed loop system is operated initially with low gain proportional control
mode with τd =0 and τ =∞ .
i

The gain Kp is increased slowly till a decay ratio (p2/p1) of 1/4th is obtained in the
step response in the output,(this gain is called Kd) as shown in Figure.

Under this condition, the period of damped oscillation, Td is also noted. Let Kd be
the proportional gain setting for obtaining 1/4th decay ratio.
 General Comments on Controller Tuning

The different methodologies of controller tuning, known as Ziegler-Nichols method

have been illustrated.

It is to be remembered that the recommended settings are empirical in nature, and

obtained from extensive experimentation with number of different processes; there is
no theoretical basis behind these selections.

As a result, a better combination of the P, I, D values may always be found, that will
give less oscillation and better settling time.

But with no a-priori knowledge of the system, it is always advisable to perform the
experimentation and select the controller settings, obtained from Ziegler-Nichols
method.

But there is always scope for improving the performance of the controller by fine-
tuning. So, Ziegler Nichols method provides initial settings that will give satisfactory,
result, but it is always advisable to fine-tune the controller further for the particular
process and better performance is expected to be achieved.
 Integration windup and Bumpless transfer

Two major issues of concerns with the closed loop operation of P-I-D controllers
are the Integration Windup and the requirement of providing Bumpless Transfer.

These two issues are briefly elaborated below.

Integration Windup

A significant problem with integral action is that when the error signal is large for a
significant period of time. This can occur every time when there is large change in set
point. If there is a sudden large change in set point, the error will be large and the
integrator output in a P-I-D control will build up with time. As a result, the controller
output may exceed the saturation limit of the actuator. This windup, unless prevented
may cause continuous oscillation of the process whic is not desirable.
Bumpless Transfer

When a controller is switched from manual mode to auto-mode, it is desired that the
input of the process should not change suddenly.

But since there is always a possibility that the decision of the manual mode of
control and the auto mode of control be different, there may be a sudden change in
the output of the controller, giving rise to a sudden jerk in the process operation.

Special precautions are taken for bumpless transfer from manual to auto-mode.
 Extract from a controller’s manual: it’s empirical !

Optimization according to Ziegler-Nichols

Assuming that the process is stable at the operating temperature:
1. Set the Parameters ‘ti’ und ‘td’ OFF.
2. The actual value differs now from the solicited value by proportional factor.
3. As soon as temperature stabilizes, reduce the value of the proportional band
‘Pb’, until the temperature starts to swing => swinging period „T“.
4. Slowly increase proportional band until temperature just stops swinging
=> value of the proportional band ‘B’.
5. Set the values of Pb, Ti and Td
according to table

But what do you do if this method does not work ?

How do you know that this plant can be controlled by a PID controller (many cannot)
How do you prevent overshoot ? (this method does not)
Trial-and-error can’t replaces a serious analysis -> see “for further reading”.
 Several controllers act together: Electricity Generator

DP

Active power Reactive power

frequency (Pf)
DQ voltage (QV)
D controller controller

f
Steam DV
Controllable
excitation
Main Valve
source
steam control
valve mechanism
Voltage
sensors
Turbine Generator

Frequency sensor

Mechanical power 3-phase Electrical Power DP + j DQ

 Generator Regulator structure

excitation
current U = k × Ie × ω
voltage PID exitation Ie

load

speed
ω
frequency PID turbine generator voltage

measure

measure
 Nested control of a continuous plant - example

Example: position control of a rotating shaft

PD torque regulation
Position PID (protection) encoder
sol Speed
cmd Torque PID
sol tacho
is cmd sol M
cmd
is
is
amplifier

Current

Velocity

Position

Nesting regulators allow to maintain the output variable at a determined value

while not exceeding the current or speed limitations
 Nested loops and time response

position control

speed control
torque control robot arm trajectory

A control system consists often of nested loops,

with the fastest loop at the inner level
 Feedforward

basic idea: bring the output on an approximate track from the start, let the regulator only correct
the small deviations. The feedforward controller knows the plant, it can also consider known disturbances

disturbances
feed-forward
controller
x
set point command
feed-back
controller plant
m

y
measurement
process value
 Advanced Control

controller disturbances
plant
model
x
economical
control
objectives, m plant
algorithms
Cost functions
command

y
measurement
process value

This is a high-level control in which the set-points are computed based on economical objectives
 Exercises

A Control System…
a) is dependent not only on current environment but on past environment as well
b) describes the direction PV moves and how far it travels in response to a change in
CO (steady state)
c) set of devices to manage, command, direct or regulate the behavior of other
device(s) or system(s)

What is the set point?

a) Variable you want to control
b) Desired value of control variable
c) Signal that is continuously updated

What has only one tuning parameter so it’s easy to find “best” tuning, but permits offset?
a) P only
b) PI
c) PD

What is proportional to both the magnitude of the error and the duration of the error?
a) P only
b) PI
c) PD
 Assessment

How does a two-point regulator works ?

How is the a wear-out of the contacts prevented ?

How does a PID regulator works ?

What is the influence of the different parameters of a PID ?

Is a PID controller required for a position control system (motor moves a vehicle)?

Explain the relation between nesting control loops and their real-time response

What is feed-forward control ?

 To probe further

"Computer Systems for Automation and Control", Gustaf Olsson, Gianguido Piani,
Lund Institute of Technology

“Modern Control Systems”, R. Dorf, Addison Wesley

“Feedback Systems”, Karl Johan Aström, Richard M. Murray

http://www.cds.caltech.edu/~murray/books/AM08/pdf/am08-complete_28Sep12.pdf