Vous êtes sur la page 1sur 15

ISA Transactions

Volume 45, Number 4, October 2006, pages 529543

Design of PI controllers for achieving time and frequency domain


specifications simultaneously
Serdar Ethem Hamamci* Nusret Tan
Department of Electrical and Electronics Engineering, Engineering Faculty, Inonu University, 44280 Malatya, Turkey

Received 19 September 2005; accepted 20 March 2006

Abstract
This paper deals with the design of PI controllers which achieve the desired frequency and time domain specifications
simultaneously. A systematic method, which is effective and simple to apply, is proposed. The required values of the
frequency domain performance measures namely the gain and phase margins and the time domain performance
measures such as settling time and overshoot are defined prior to the design. Then, to meet these desired performance
values, a method which presents a graphical relation between the required performance values and the parameters of the
PI controller is given. Thus, a set of PI controllers which attain desired performances can be found using the graphical
relations. Illustrative examples are given to demonstrate the benefits of the method presented. 2006 ISAThe
Instrumentation, Systems, and Automation Society.
Keywords: Frequency and time domain performancesmap; PI control; Control system design; Gain margin; Phase margin; Overshoot;
Settling time

1. Introduction
PID controllers are used extensively in the process industries because of their robust performance and simplicity. Indeed, more than 90% of
the control loops are PID and most loops are in
fact PI since derivative action is not used very
often 1. Several methods see Refs. 26 and
references therein for determining the parameters
of these controllers have been developed during
past 60 years. Although most of these methods
provide acceptable performances for some transfer
functions of the systems, there is not a general
method for tuning the parameters of these controllers. For example, the Ziegler-Nichols method
which is still widely used in industries for tuning
*Corresponding

author.

E-mail

shamamci@inonu.edu.tr, ntan@inonu.edu.tr

address:

gives a high overshoot and a long settling time.


Therefore, other methods such as the refined
Ziegler-Nichols method, pole-zero cancellation
method, and ISTE performance criteria have been
proposed to improve the performance of control
systems which especially have a time delay. It is
considered in these methods directly either frequency domain specifications such as gain margin
GM and phase margin PM or time domain
specifications like maximum overshoot MO and
settling time ts. However, one cannot specify all
of these specifications before the design while using these methods.
The basic approach for the design of the controller in the modern control theory is first to obtain
the entire set of stabilizing controllers as in the
Youla parameterization, and then to search over
the set to get a controller that meets the given
performance criteria. Several methods have been
recently reported on the computation of all the sta-

0019-0578/2006/$ - see front matter 2006 ISAThe Instrumentation, Systems, and Automation Society.

530

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

bilizing P, PI, and PID controllers 710. However, beyond stabilization it is important to design
controllers that guarantee specified performance
measures such as gain and phase margins, settling
time, and overshoot. Although there are some classical formulas that exist which link frequency and
time domain performances such as PM/ 100,
these approximate formulas do not work for time
delay systems or high order control systems. It
seems that the coefficient diagram method CDM
combined with the methods for the computation of
stabilizing controllers can be a good candidate for
designing controllers which attain the time and
frequency domain specifications simultaneously.
CDM is a polynomial approach which was developed and introduced for a good transient response
of the control systems by Manabe 11. The most
important properties of this method are the adaptation of the polynomial representation for both
the plant and controller, the use of the two-degreeof-freedom 2DOF control system structure, the
nonexistence or very small of the overshoot in
the step response of the closed-loop system, the
determination of the desired settling time and the
maximum overshoot at the start and to continue
the design accordingly, and good robustness of the
control system with respect to the parameter
changes.
In this paper, a new method is proposed for the
computation of a set of PI controllers which give
the prescribed frequency and time domain performance criteria such as gain margin, phase margin,
overshoot, and settling time simultaneously. In this
method, using the stability boundary locus approach of Refs. 10,12, all the stabilizing values
of the parameters of the PI controller are first obtained. This stability region is called the global
stability region. In addition to this, all the stabilizing PI controllers within the global stability region
which provide desired frequency domain performance measures are identified. These subsets of
the global stability region are called the local stability region. Then, the settling time and overshoot
which are very important time domain performance measures are chosen for time domain
specifications. Using the CDM, some explicit formulations are obtained between the PI controller
parameters and the time domain specifications.
For this, a method is given for removing errors
due to approximation used for the time delay term.
Finally, a graphical relation on GM, PM, MO, and

Fig. 1. A standard block diagram of the CDM control


system.

ts in the same plane, k p , ki plane, is obtained.


This graphical representation is called the frequency and time domain performances (FTDP)
map. Thus, one can easily obtain a set of PI controllers which attains the desired frequency and
time domain performances using the FTDP map.
The paper is organized as follows: The CDM is
revisited in Section 2. The proposed controller design method is explained in Section 3. Simulation
examples are given in Section 4. Section 5 includes some concluding remarks.
2. Coefficient diagram method
The standard block diagram of the CDM control
system is shown in Fig. 1, where r is the reference
input, y is the output, d is the disturbance, and u is
the control signal. N ps and D ps are the numerator and denominator polynomials of the
transfer function of the plant and have not any
common factors. As is the forward denominator
polynomial while Bs and Fs are the feedback
numerator and the reference numerator polynomials of the controller transfer function. Since the
transfer function of the CDM controller has two
numerators, the control system resembles to a
2DOF system structure. As and Bs are designed as to satisfy the desired transient behavior
and defined as
p

A s = l is i ,
i=0
q

B s = k is i

i=0

in the polynomial forms while prefilter Fs is determined as the zero order polynomial and used to
provide the steady-state gain. Better performance

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

531

Fig. 2. a The effect of the equivalent time constant on the rate of the closed-loop time response. b The effect of the stability
index, 1 on the time response shape.

can be expected when using a 2DOF structure,


because it can focus on both tracking the desired
reference signal and disturbance rejection. Unstable pole-zero cancellations and the use of more
numbers of integrators are also avoided in implementations with this structure. The output of the
CDM control system in Fig. 1 is defined as

y=

N p s F s
A s N p s
r+
d,
Ps
Ps

ai 0.

i=0

3
In the literature, it is reported that there are some
certain relations between characteristic polynomial
coefficients and important time domain measures
such as settling time and overshoot 1315.
Manabe integrated these studies with the basic
principles of his method and determined two important design parameters. These parameters are
equivalent time constant and stability indices
i which are defined as

= a1/a0,

i = a2i /ai+1ai1 ,

i = 1 n 1,

0 = n = .

Ptargets = a0

i=2

where Ps is the characteristic polynomial of the


closed-loop system. This polynomial is a Hurwitz
polynomial with real positive coefficients and described by

P s = A s D p s + B s N p s = a is i,

the target characteristic polynomial of the closedloop system. From Eq. 4, the characteristic polynomial in Eq. 3 can be expressed in terms of
and i as

It is important that these parameters are specified


before the design and then used for determining

i1

1
si + s + 1 .
j

j=1 ij
5

The equivalent time constant specifies the time response speed settling time, especially. The stability indices affect the stability and the shape of
the time response overshoot, especially. For example, consider a characteristic polynomial whose
degree is chosen as n = 2. According to this polynomial, choosing 1 = 3 as constants and changing
the equivalent time constant in the interval of 1
4, the unit step responses of the control system
are shown in Fig. 2a. It is seen in this figure that
is very effective on the settling time of the unitstep response. If the equivalent time constant is
increased, the settling time is also increased. On
the contrary, if is reduced, the system time response can be accelerated as desired. When the
choice = 1 is considered and 1 is changed in the
interval of 0.55, the change of the time response shapes are shown in Fig. 2b. This figure
indicates that the stability index 1 is much effective on the response shape and stability. If 1 is
made bigger, the stability of the control system is
increased and the overshoot is zero. But 1 is of a
the small size, the stability decreases and overshoot is nonzero in this case.

532

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

lim Cs = and lim C f s/Cs = 0.

s0

s0

If Cs includes an integrator then these conditions can be satisfied. Thus, Cs can be chosen as

Cs = k p +
Fig. 3. A 2DOF control system structure.

For a special case, it is recommended that standard Manabe values for the time response without
overshoot and with the smallest settling time are
used for the CDM design. Stability indices are
chosen as i = 2.5, 2 , 2 , . . . , 2 for i = 1 n 1n
3 in this form. For a third order system i
= 2.7, 2 and 1 = 3 in a second order system,
overshoot is zero. When the standard Manabe values are chosen, the settling time is about
2.5 3. The selection of the standard values can
be relaxed according to the various performance
requirements.

3. Controller design
The proposed method which is used to design a
PI controller satisfying the required time and frequency domain specifications consists of two
steps. In the first step, a method is proposed to
compute the global and the local stability regions
using the stability boundary locus approach
10,12. In the second step, the CDM method is
used to design PI controllers for which the step
responses have a required overshoot and an acceptable settling time. As a result of combining
these two steps, the FTDP map is obtained. The
FTDP map, which is a graphical tool, shows the
relation between the stabilizing parameters of the
PI controllers and the chosen frequency and time
domain performance criteria on the same k p , ki
plane. Thus, one can choose a PI controller providing all of the desired GM, PM, MO, and ts
specification values together.
A general schema of the 2DOF control system is
shown in Fig. 3. This representation is the rational
equivalence of the control system given in Fig. 1.
Here, Cs = Bs / As is the main controller and
C f s = Fs / Bs is the set point filter 16. It can
be shown that the steady-state error to the unit step
change and the unit step disturbance become zero
robustly if

k i k ps + k i
=
s
s

in the type of the conventional PI element and


C f s is an appropriate element satisfying Eq. 6.
The proposed design approach consists of a number of distinct steps, which can be summarized as
follows.
I
Frequency response performance
I a Computation of global stability region
Consider a 2DOF control system shown in Fig.
3 where

G a s = G s e s =

N s s
e
Ds

and a PI controller of the form of Eq. 7. The


problem is to compute the global stability region
which includes all the parameters of the PI controller of Eq. 7 which stabilize the given system.
The closed-loop characteristic polynomial Ps of
the system, i.e., the numerator of 1 + CsGas,
can be written as

Ps = sDs + k ps + kiNses
= ansn + an1sn1 + + a1s + a0 ,

where all or some of the coefficients ai, i


= 0 , 1 , 2 , . . . , n are the function of k p, ki, and es
depending on the order of Ns, and Ds polynomials. In the parameter space approach, there are
three possibilities for a root of a stable polynomial
to cross over the imaginary axes to become unstable:
a

Real Root Boundary: A real root crosses


over the imaginary axis at s = 0. Thus, the
real root boundary can be obtained from
substituting s = 0 in Ps of Eq. 9 which
gives a0 = 0.
Infinite Root Boundary: A real root crosses
over the imaginary axis at s = . Thus, the
infinite root boundary can be characterized
by taking an = 0 from Eq. 9.
Complex Root Boundary: Eq. 9 becomes
unstable at s = j when its roots cross the
imaginary axis which means that the real

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

and the imaginary parts of Eq. 9 become


zero simultaneously. Thus, the complex
root boundary can be obtained as follows:
Decomposing the numerator and the denominator polynomials of Gs in Eq. 8 into their even
and odd parts, and substituting s = j , gives

G j =

N e 2 + j N o 2
.
D e 2 + j D o 2

10

k pNe cos + 2No sin


+ kiNo cos Ne sin = De .
12b
Let

Q = Ne sin 2No cos ,


R = Ne cos + No sin , 13a
X = 2D o ,

For simplicity 2 will be dropped in the following equations. Thus, the closed-loop characteristic polynomial of Eq. 9 can be written as

P j = kiNe k p2Nocos

S = Ne cos + 2No sin ,


U = No cos Ne sin,

13b
Y =

De .

+ kiNo + k pNesin Do
2

Then, Eqs. 12a and 12b can be written as


k pQ + k iR = X ,

+ jkiNo + k pNecos kiNe


2k pNosin + De = R P + jI P = 0.
11
Then, equating the real and imaginary parts of
Pj to zero, two equation are obtained as

k p 2No cos + Ne sin

k pS + k iU = Y .

+ kiNe cos + No sin = Do ,


12a

14

From these equations


XU Y R
kp =
,
QU RS

ki =
2

Y Q XS
.
QU RS

15

Substituting Eqs. 13a and 13b into Eq. 15, the


PI controller parameters are obtained as

2NoDo + NeDecos + NoDe NeDosin


,
N2e + 2N2o

16

2NoDe NeDocos NeDe + 2NoDosin


.
N2e + 2N2o

17

kp =

ki =

533

The stability boundary locus, lk p , ki , can be


constructed in the k p , ki plane using Eqs. 16
and 17. If at any particular frequency value the
denominator of Eqs. 16 and 17 Ne + 2N2o = 0,
then this value of frequency must not be used. In
this case, a discontinuous stability boundary locus
will be obtained and this will not be a problem for
the computation of stabilizing controllers. Once

the stability boundary locus has been obtained


then it is necessary to test whether stabilizing controllers exist or not since the stability boundary
locus, lk p , ki , , the real root and infinite root
boundary lines if they are exist may divide the
parameter plane k p , ki plane into stable and unstable regions. It can be found that the line ki = 0 is
the real root boundary line obtained from substi-

534

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

tuting = 0 into Eq. 9 and equating it to zero


since a real root of Ps of Eq. 9 can crossover
the imaginary axis at s = 0. Generally, the order of
Ds is greater than order of Ns for proper transfer function. Therefore, there will be no infinite
root boundary line.
It can be seen that the stability boundary locus is
dependent on the frequency which varies from 0
to . However, one can consider the frequency
below the critical frequency, c, or the ultimate
frequency since the controller operates in this frequency range. Thus, the critical frequency can be
used to obtain the stability boundary locus over a
possible smaller range of frequency such as
0 , c. Since the phase of G ps at s = jc is
equal to 180, one can write

tan1

No
Do
tan1
=
Ne
De
18

or

N oD e N eD o
= f .
N eD e + 2N oD o

19

Thus, c is the solution of Eq. 19 in the interval


0 , . By plotting tan and f vs , it can
be seen that c is the smallest value of at which
plots of tan and f intersect with each
other. If there is more than one real value of
which satisfies Eq. 19 then the frequency axis
can be divided into a finite number of intervals
and by testing each interval the stability region can
be computed.
I b Computation of local stability regions
Phase and gain margins are two important frequency domain performance measures which are
widely used in the classical control theory for the
controller design. Consider Fig. 3 with a gainphase margin tester 17, Gcs = Aej, which is
connected in the feed forward path. From Eqs.
16 and 17,

2NoDo + NeDecosh + NoDe NeDosinh


,
AN2e + 2N2o

20

2NoDe NeDocosh NeDe + 2NoDosinh


,
AN2e + 2N2o

21

kp =

ki =

tan =

where h = + . Here, the gain-phase margin


tester is not an actual part of the system which is
included in the system in order to obtain the stability regions for prespecified values of the gain
and phase margins. To obtain the stability boundary locus for a given value of gain margin A, one
needs to set = 0 in Eqs. 20 and 21. On the
other hand, setting A = 1 in Eqs. 20 and 21, one
can obtain the stability boundary locus for a given
phase margin . Thus, the local stability regions
for specified gain and phase margins can be identified within the global stability region.
II Time response performance
The CDM technique is used to obtain time domain
performances. However, in this case, it is necessary to use the first-order plus dead time FOPDT
model of the plant. The processes encountered in

the industry can be mostly described by a FOPDT


model such as

G m s =

N m s s
K s
e =
e ,
D m s
Ts + 1

22

where K is the gain, T is the time constant, and


is the dead time. The experimental identification
of these models using many techniques is well described in Ref. 4. It is necessary to use the
FOPDT model for the proposed method, since the
CDM requires high order controllers for high order models. The term es which also represents
the time delay in Eq. 22 is approximated as
es 1 s using the first-order Taylor numerator
approximation. The Taylor numerator approximation is used in order to not affect the denominator

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

535

Fig. 4. Time delay approximation for corrected and uncorrected cases: a selection of k p and ki parameters and b the step
responses for selected k p and ki values.

of Eq. 22. If the Taylor denominator approximation or Pade approximation is used, theirs denominators lead to a higher order of the approximate
transfer function of the plant and consequently to
more complex resulting controllers. The results
obtained in Section 4 show that the first-order Taylor numerator approximation is acceptable and
gives good results. Thus, for the model transfer
function of Eq. 22, plant polynomials are

N ps = Ks + K,

D ps = Ts + 1.

23

For a selection of good controllers, the degrees of


the controller polynomials in Eq. 1 get importance. The most important fact that affects the degrees is the existence of a disturbing signal and its
type. It is advised that the minimum degree polynomials are chosen depending on the type of the
disturbance. In this paper, the controller polynomials are chosen for the step disturbance signal. In
this case, the controller polynomials have the
forms

As = l1s,

B s = k 1s + k 0 .

24

Thus, the configuration in Fig. 3 is transformed to


a PI-based control system. From Eqs. 7 and 24,
it can be seen that k1 = k p, k0 = ki and l1 = 1.
II a Computation of the PI controller parameters
for the time domain performance.
In the design, a feedback controller is chosen by
the pole-placement technique and then, a set point
filter is determined so as to match the steady-state
gain of the closed-loop system. According to this,
the controller polynomials which are determined

by Eq. 24 are replaced in Eq. 3. Hence, a polynomial depending on the parameters k p and ki is
obtained. For the FOPDT model transfer function,
the characteristic polynomial of the control system
is determined as

Ps = T Kk ps2 + 1 Kki + Kk ps + Kki .


25
Using Eq. 4, 1 and can be obtained as

1 = 1 + Kk p Kki2/KkiT Kk p ,
26a

= 1 + Kk p Kki/Kki .

26b

For Eqs. 26a and 26b, ki can be found for 1


and separately as
Table 1
1 and k values for some overshoot values.

MO
%

1 values

k values ts k

20
10
5
0a
0

0.8
1.4
1.9
3a
5

11.2
6
5.2
3a
3.9

Standard Manabe values.

536

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

Fig. 5. For K = 1, T = 1, and = 2 in Eq. 22: a ki-k p curves of 1 and , b the step responses for = 4 and different values
of 1 selected from Table 1.

ki = K2 + 2Kk p + 1T 1Kk p
/2K22

for 1 ,

27a

where
= K22 + 2Kk p + 1T 1Kk p2
421 + Kk p2 and

ki = 1 + Kk p/K + K

for .

27b

From the global stability region obtained in step


I a, an interval for k p can be found as k pmin
-k pmax. Using this interval and Eqs. 27a and
27b, ki-k p curves can be plotted. The values of k p
and ki parameters at the intersection of two curves
provide desired vs 1 values. At this point one
should be careful that desired -1 specifications

may not be exactly obtained since an approximation for time delay has been used. Therefore, a
correction process should be implemented. To do
this, a second order Taylor numerator approximation, es 1 s + 0.52s2 is used for the same
controller of Eq. 24. Repeating the same mathematical derivations for and 1, it can be seen
that Eq. 26b remains same but Eq. 26a is
changed to the form

1 = 1 + Kk p Kki2/KkiT Kk p + 0.5K2ki .
28
Using the higher order approximations for correction does not effect the relations for and 1. ki

Fig. 6. For K = 1, T = 1, and = 2 in Eq. 22: a ki-k p curves of 1 and , b the step responses for 1 = 3 and different values
of .

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

Fig. 7. The global stability region including all of the stabilizing PI parameters for the given plant in Eq. 32.

uncorrected cases are shown in Fig. 4a. From


this figure, it can be seen that ki, k p values which
give the desired specifications are different. The
step responses for PI controllers obtained from
corrected and uncorrected ki-k p curves are shown
in Fig. 4b which indicate that the step response
for the uncorrected case have a 3.2% overshoot
and settling time of 18.2 s, however, the step response for the corrected case has no overshoot and
a settling time of 11.9 s. These values of the settling time have been computed using a tolerance
band of 2% for the steady-state value 18. Thus,
the error due to the time delay approximation can
be removed by this way.
The reference numerator polynomial Fs which is
defined as the prefilter element is chosen to be

Fs = Ps/Nmss=0 = ki .
values for this case can be obtained as

ki = K2 + 2Kk p + 1T 1Kk p
/K222 1

for 1 ,

29a

where
= K22 + 2Kk p + 1T 1Kk p2
2
+ 2 2 11 + Kk p2 and

ki = 1 + Kk p/K + K

for .

29b

Thus, the PI controller parameters which nearly


give the desired and 1 values can be obtained.
For example, let us choose K = 1, T = 1, and = 2
in Eq. 22 and set 1 = 3 and = 4 for a step response which has no overshoot and approximately
12 s settling time. ki-k p curves for corrected and

537

30

This way, the value of the error that may occur in


the steady-state response of the closed-loop system is reduced to zero. Finally, the set point filter
in Fig. 3 is obtained by

C f s =

Fs
ki
=
.
B s k ps + k i

31

Note that the parameters of C f s depend on the PI


parameters directly. Therefore, the designer does
not need the extra calculation for the set point filter.
II b Choice of the key parameter values for the
desired time domain properties.
One of the most important properties of the pro-

Fig. 8. Local stability regions: a for some GM values and b for some PM values.

538

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

Fig. 9. a ki-k p curves for different 1 values in Table 1. b The step responses of these 1 values.

posed method is that the settling time and overshoot information of the control system are determined at the beginning before starting to design. It
is generally aimed that the time response of the
control system must have no overshoot and desired settling time. Table 1 shows the relation between ts and for the desired settling time for the
various overshoot properties. These values of the
general relation are obtained by normalizing the
characteristic polynomial of Eq. 5 for n = 2 and
= 1. For a transfer function with K = 1, T = 1, and
= 2 in Eq. 22, the step responses for = 4 and
different values of 1 selected from Table 1 are
shown in Fig. 5 where it can be seen that the maximum overshoot values given in Table 1 are exactly
met. Similarly, the step responses for 1 = 3 and

different values of are shown in Fig. 6 which


clearly shows that there is no overshoot and the
settling time values determined in Table 1 are also
met.
III Building of the FTDP map
The FTDP map is built by plotting the frequency
domain stability regions and the time domain performance curves in the same k p, ki plane. The
frequency domain stability regions include the
global and local stability regions which can be obtained using Eqs. 16, 17, 20, and 21, and the
time domain performance curves which correspond to the various overshoot and settling time
values can be plotted using Eqs. 29a and 29b.
Thus, the designer can choose any points, k p, ki
values, from the FTDP map which satisfy the de-

Fig. 10. a Four different PI controllers in the GM 2 region and on the 1 = 3 curve. b The unit step responses without
overshoot for the selected PI controllers.

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

539

Table 2
The time domain and frequency domain specifications correspond to selected four points.

kp
ki
MO %
ts s
GM
PM

Step 1. Computation of the frequency response


performance regions in the k p, ki-plane for
Eq. 8:
1 construction of the global stability region
using Eqs. 16 and 17, the stability
boundary locus lk p , ki , and the real root
boundary line ki = 0;
2 obtaining the local stability regions correspond to desired GM and PM values using
Eqs. 20 and 21.
Step 2. Computation of the time performance
curves in the global stability region using
FOPTD model of actual plant given in Eq. 8:
1 obtaining ki = fk p , 1 curves which have
different values of overshoot using Eq.
29a;

3.0941
1.1573
0
6.2
2.3
38.8

2.5002
0.8216
0
8.8
2.95
47.7

2.0601
0.6142
0
10.8
3.6
54.6

1.7201
0.4768
0
12.7
4.4
60

2 finding ki = gk p , curves which have different values of settling time using Eq.
29b.

Fig. 11. A local stability region with 2.3 GM 5 and


30 PM 60 filled in gray color and 1 = 3 curve expressed no overshoot property.

sired frequency and time domain properties. After


building the FTDP map and choosing the controller parameters, the control system shown in Fig. 3
is simulated using the actual process and the designed PI controller. However, for some transfer
functions there are not any k p, ki values which
give the desired performances because of the hard
properties of the plant. For such situations, it is
necessary to relax the desired frequency and time
domain specifications.
In view of the earlier developments, the FTDPmap design procedure linking the frequency and
time domain performances is summarized as follows for the ease of reference:

Step 3. Plotting the frequency domain performance regions together with the time domain
performance curves, thus, obtaining the FTDP
map and finding the desired PI controllers from
this map.

4. Simulation examples
4.1. Example 1
A first order process with a time delay is chosen
as

G a s =

1
es ,
5s + 1

32

which has a pole near the imaginary axis. The aim


is to investigate the PI controller parameters which
have a satisfying frequency and time domain
specifications such as gain margin, phase margin,
overshoot, and settling time. For this, it will be
made use of the FTDP map explained in Section 3.
The global stability region which is shown in
Fig. 7 is computed using the procedure given in
Section 3I a. The global stability region includes
all PI controllers which stabilize the given system.
However, one needs to choose suitable k p and ki
values within this region to obtain the required
frequency and time domain performances. The local stability regions for the specified gain and
phase margins can be identified within the global
stability region using the procedure given in Section 3I b as shown in Figs. 8a and 8b. If the
local regions for the GM and PM values are com-

540

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

Fig. 12. a Global and desired local stability regions. b Time domain curves related to the overshoot and settling time.

bined, the intersection local regions can be obtained satisfying both the GM and PM together.
Now, it is important that how the PI parameters
which provide the desired frequency domain performances in the local region will exhibit time domain performances. Using Eq. 29a to get the
overshoot information of the control system for
different 1 values in Table 1, Fig. 9a is obtained. According to this 1 values, the step responses are shown in Fig. 9b for k p = 1 and corresponding ki values which are computed from the
1 curves. From the step responses, one can see
that the values of the resultant overshoots are
equal to the results given in Table 1. For example,
for 1 = 1.4, the expected value of the overshoot
from Table 1 is about 10% and this value is exactly obtained as shown in Fig. 9b. Similarly, for
1 = 3 a response without an overshoot is expected
and it can be seen that which is also met.
It is vital to point out that all the values of k p
and ki obtained over any 1 curve do not necessarily give a response having its overshoot property given in Table 1. For satisfactory performance, the gain margin should be greater than 2,
and the phase margin should be between 30 and
60 18. However, the simulation results showed
that especially the gain margin is very effective on
the overshoot property. The part of any 1 curves
within the stability region for GM 2 gives k p
and ki values for which the step responses have the
overshoot values given in Table 1. Note that GM
= 2 is a boundary value and causes a time response
with a little overshoot. For example, the PI controllers which correspond to points 1, 2, 3, and 4

in Fig. 10a give unit step responses without an


overshoot for 1 = 3 as shown in Fig. 10b.
Eq. 29b can be used to obtain the desired settling time. The four intersection points in Fig.
10a express four different settling time values
for 1 = 3. The intersection points of 1 and
curves give important information: 1 represents
the overshoot, and the settling time obtained from
ts = k. Thus, the values of k p and ki correspond to
the intersection points that give the values of the
overshoot and the settling time. The step responses
for these points are shown in Fig. 10b. It can be
seen that the settling time increases when moving
from point 1 to point 4 as shown in Fig. 10b. It
has been observed that the values of the settling
time and overshoot given in Table 1 are met.
The local region with properties of 2.3 GM
5 and 30 PM 60 is shown in Fig. 11.

Fig. 13. FTDP map for example 2.

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

541

Fig. 14. Set point and disturbance responses for a points 1, b 2 and c 3.

Four points are selected from this figure within the


shaded region. The frequency and time domain
properties for each point are given in Table 2.
Thus, it can be seen that all PI controllers for
which the gain margin is between 2.3 and 4.4, the
phase margin is between 38.3 and 60, the settling time is between 6.2 and 12.7 s, and with no
overshoot is between the points of 14 on 1 = 3
curve within the local region.
From the simulation studies using the FTDPmap method, it has been observed that there are
many k p , ki values which satisfy the proper time
and frequency domain performances simulta-

G a s =

neously. However, for some transfer functions


there is not any k p , ki values which give desired
performances. For example, there is not any
k p , ki value for the example studied earlier which
simultaneously provide the gain margin to be 5
and a response without the overshoot, because this
property is out of the desired region. For such situations, it is necessary to relax the desired frequency and time domain specifications.
4.2. Example 2
Consider a higher order process with a large
time delay as

1
e4s .
s + 10.5s + 10.25s + 10.125s + 1

The aim is to design PI controllers which make the


gain margin of the control system greater than 2.3
and the phase margin between 30 and 60. Using
Eqs. 16, 17, 20, and 21 the global and the

33

local stability regions are obtained as shown in


Fig. 12a. For time domain performances it is
necessary to use the the FOPDT model which is

542

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

Table 3
The time domain and frequency domain specifications correspond to the selected three points.

kp
ki
MO %
ts s
GM
PM

0.2463
0.1317
5
25.2
2.6
57

0.1295
0.1194
10
29.3
2.8
58.5

0.1034
0.1055
5
31
3.1
60

given in Ref. 19 as Gms = e4.462s / 1.521s + 1.


1 and curves which are directly related to the
time domain performances as explained earlier can
be seen from Fig. 12b. Combining Figs. 12a
and 12b, the FTDP map shown in Fig. 13 is obtained. From this figure, it is clear that it is not
possible to obtain a step response without an overshoot since the 1 = 3 curve does not pass through
the shaded region. Looking at Fig. 13, it is seen
that the shaded region is approximately bounded
by 1 = 1.9 and 1 = 0.8 which give a 5% and 20%
overshoot as stated in Table 1, respectively. Therefore, all PI controllers in the shaded region give
the desired time and frequency domain performances. Three points are selected within this region which are indicated by 1, 2, and 3 as shown
in Fig. 13. Figs. 14a14c show the time responses of the selected three points according to
the set point changes applied at t = 0 s and the disturbance applied at t = 70 s. The time and frequency domain specifications of these points are
shown in Table 3.
5. Conclusions
A new method which is simple to use has been
presented to design PI controllers giving the desired frequency and time domain performances.
One of the most important features of the proposed method is that the desired frequency and
time domain specifications can be specified before
the design. It has been shown that:
i
ii

The global and the local stability regions


can be computed using the stability
boundary locus approach.
The values of the parameters of PI controller which achieve desired values of
settling time and overshoot can be estimated using the CDM.

iii

iv

The FTDP map gives a visual chance to


see which parts of the global stability region is important for the required design
specifications.
A set of PI controllers which provide predefined performance specifications can be
computed from FTDP-map.

The given simulation examples clearly show


that the results presented are useful for the design
of PI controllers.
It is known that the mathematical techniques for
obtaining the exact correlation between the step
transient response and frequency response are
available but they are very laborious and of little
practical value 18. However, the method presented in this paper has a potential to show a
graphical relation between the frequency and time
domain performances of high order or time delay
systems. Therefore, the results obtained can be applied for many real applications. The extension of
the method to the PID controller will be very important in the control system design.
References
1 strm, K. J. and Hgglund, T., The future of PID
control. Control Eng. Pract. 9, 11631175 2001.
2 Ziegler, J. G. and Nichols, N. B., Optimum settings for
automatic controllers. Trans. ASME 64, 759768
1942.
3 strmK. J., Hgglund, T., Hang, C. C., and Ho, W.
K., Automatic tuning and adaptation for PID
controllersA survey. Control Eng. Pract. 1, 699
714 1993.
4 strm, K. J. and Hgglund, T., PID Controllers:
Theory, Design, and Tuning. Instrument Society of
America, Research Triangle Park, NC, 1995.
5 Zhuang, M. and Atherton, D. P., Automatic tuning of
optimum PID controllers. IEE Proc.-D: Control
Theory Appl. 140, 216224 1993.
6 Ho, W. K., Hang, C. C., and Cao, L. S., Tuning of PID
controllers based on gain and phase margins specifications. Automatica 31, 497502 1995.
7 Datta, A., Ho, M. T., and Bhattacharyya, S. P., Structure and Synthesis of PID Controllers. Springer, New
York, 2000.
8 Ackermann, J. and Kaesbauer, D., Design of robust
PID controllers. Proceedings of the 2001 European
Control Conference, 2001, pp. 522527.
9 Sylemez, M. T., Munro, N., and Baki, H., Fast calculation of stabilizing PID controllers. Automatica 39,
121126 2003.
10 Tan, N., Kaya, I., and Atherton, D. P., Computation of
stabilizing PI and PID controllers. Proceedings of the
CCA2003 IEEE Conf. on the Contr. Appl., 2003.
11 Manabe, S., Coefficient Diagram Method. Proceed-

S. E. Hamamci and N. Tan / ISA Transactions 45, (2006) 529543

12
13
14

15

16
17

18
19

ings of the 14th IFAC Symposium on Automatic Control in Aerospace, Seoul, 1998.
Tan, N., Computation of stabilizing PI and PID controllers for processes with time delay. ISA Trans. 44,
213223 2005.
Naslin, P., Essentials of Optimal Control. Boston
Technical, Cambridge, 1969.
Lipatov, A. and Sokolov, N., Some sufficient conditions for stability and instability of continuous linear
stationary systems. Autom. Remote Control Engl.
Transl. 39, 12851291 1979.
Kim, Y. C., Keel, L. H., and Bhattacharyya, S. P.,
Transient response control via characteristic ratio assignment. IEEE Trans. Autom. Control 4812, 2238
2244 2003.
Chen, C. T., Analog and Digital Control System Design: Transfer Function, State-Space and Algebraic
Methods. Saunders College, New York, 1992.
Argoun, M. B. and Bayoumi, M. M., Robust gain and
phase margins for interval uncertain systems. Proceedings of the 1993 Canadian Conference on Electrical
and Computer Engineering 1993, pp. 7378.
Ogata, K., Modern Control Engineering. PrenticeHall, Englewood Cliffs, NJ, 1970.
Mann, G. K. I., Hu, B.-G., and Gosine, R. G., Timedomain based design and analysis of new PID tuning
rules. IEE Proc.: Control Theory Appl. 1483, 251
261 2001.

543

Serdar Ethem Hamamci received B.Sc. and M.Sc. degrees


from Erciyes University, Kayseri,
Turkey, in 1992 and Firat University, Elazig, Turkey, in 1997, respectively. He obtained a Ph.D.
degree from Firat University, in
2002. He is currently working as
a research assistant in the department of electrical and electronics
engineering at Inonu University,
Malatya, Turkey. His primary area
of research is polynomial control
methods especially the coefficient diagram method and their applications.

Nusret Tan was born in Malatya,


Turkey, in 1971. He received a
B.Sc. degree in electrical and
electronics engineering from Hacettepe University, Ankara, Turkey,
in 1994. He received a Ph.D. degree in control engineering from
University of Sussex, Brighton,
UK, in 2000. He is currently
working as an associate professor
in the department of electrical and
electronics engineering at Inonu
University, Malatya, Turkey. His
primary research interest lies in
the area of systems and control.