ARTICLE IN PRESS

Signal Processing 86 (2006) 27712784

www.elsevier.com/locate/sigpro

Tuning of fractional PID controllers with

ZieglerNichols-type rules
Duarte Valerio,1, Jose Sa da Costa
Department of Mechanical Engineering, Instituto Superior Tecnico, Technical University of Lisbon, GCAR,
Av. Rovisco Pais 1, 1049-001 Lisboa, Portugal
Received 12 April 2005; received in revised form 29 October 2005; accepted 6 December 2005
Available online 9 March 2006

Abstract
In this paper two sets of tuning rules for fractional PIDs are presented. These rules are quadratic and require the same
plant timeresponse data used by the rst ZieglerNichols tuning rule for (usual, integer) PIDs. Hence no model for the
plant to control is neededonly an S-shaped step response is. Even if a model is known rules quickly provide a starting
point for ne tuning. Results compare well with those obtained with rule-tuned integer PIDs.
r 2006 Elsevier B.V. All rights reserved.
Keywords: Fractional controllers; PID; ZieglerNichols tuning rules

1. Introduction
The output of PID controllers (proportional
derivativeintegrative controllers) is a linear combination of the input, the derivative of the input and
the integral of the input. They are widely used and
enjoy signicant popularity, because they are
simple, effective and robust.
One of the reasons why this is so is the existence
of tuning rules for nding suitable parameters for
PIDs, rules that do not require any model of the
plant to control. All that is needed to apply such
Corresponding author. Tel.: +351 218 419 119.

E-mail addresses: dvalerio@dem.ist.utl.pt (D. Valerio),

sadacosta@dem.ist.utl.pt (J.S. da Costa).
URLs: http://mega.ist.utl.pt/
dmov, http://www.gcar.dem.
ist.utl.pt/pessoal/sc/.
1
Partially supported by Fundac- ao para a Ciencia e a
Tecnologia, grant SFRH/BPD/20636/2004, funded by POCI
2010, POS_C, FSE and MCTES.

rules is to have a certain time response of the plant.

Examples of such sets of rules are those due to
Ziegler and Nichols, Cohen and Coon, and the
KappaTau rules [1].
Actually, rule-tuned PIDs often perform in a nonoptimal way. But even though further ne-tuning be
possible and sometimes necessary, rules provide a
good starting point. Their usefulness is obvious
when no model of the plant is available, and thus no
analytic means of tuning a controller exists, but
rules may also be used when a model is known.
Fractional PIDs are generalisations of PIDs:
their output is a linear combination of the input,
a fractional derivative of the input and a fractional
integral of the input [2]. Fractional PIDs are
also known as PIl Dm controllers, where l and m
are the integration and differentiation orders; if
both values are 1, the result is a usual PID
(henceforth called integer PID as opposed to a
fractional PID).

0165-1684/$ - see front matter r 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.sigpro.2006.02.020

ARTICLE IN PRESS
D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

2772

Even though fractional PIDs have been increasingly used over the last years, methods proposed to
tune them always require a model of the plant to
control [3,4]. (An exception is [5], but the proposed
method is far from the simplicity of tuning rules for
integer PIDs.) This paper addresses this issue
proposing two sets of tuning rules for fractional
PIDs. Proposed rules bear similarities to the rst rule
proposed by Ziegler and Nichols for integer PIDs,
and make use of the same plant time response data.
The paper is organised as follows. Section 2 sums
up the fundamentals of fractional calculus needed to
understand fractional PIDs. Then, in Section 3, two
analytical methods for tuning fractional PIDs when
a plant model is available are addressed; these are
used as basis for deriving the tuning rules given in
Section 4. Section 5 gives some examples of
application and Section 6 draws some conclusions.
2. Fractional order systems
2.1. Definitions
Fractional calculus is a generalisation of ordinary
calculus. The main idea is to develop a functional
operator D, associated to an order n not restricted
to integer numbers, that generalises the usual
notions of derivatives (for a positive n) and integrals
(for a negative n).
Just as there are several alternative denitions of
(usual, integer) integrals (due to Riemann, Lebesgue, Steltjes, etc.), so there are several alternative
denitions of fractional derivatives that are not
exactly equivalent. The most usual denition is due
to Riemann and Liouville and generalises two
equalities easily proved for integer orders:
Z x
x  tn1
n
f t dt; n 2 N,
(1)
D
f
x

c x
n  1!
c
Dn Dm f x Dnm f x;

m 2 Z
0 _ n; m 2 N0 .

(2)

The full denition of D becomes

n
c Dx f x

8
R x x  xn1
>
>
f x dx
>
> c
>
Gn
>
>
<
f x
>
>
>
Dn c Dxnn f x
>
>
>
>
:

It is worth noticing that, when n is positive but noninteger, operator D still needs integration limits c
and x; in other words, D is a local operator for
natural values of n (usual derivatives) only.
The Laplace transform of D follows rules rather
similar to the usual ones:
L0 Dnx f x
8 n
s F s
>
<
n1
P k nk1

n
>
s 0 Dx
f 0
: s F s 

4
This means that, if zero initial conditions are
assumed, systems with a dynamic behaviour described by differential equations involving fractional
derivatives give rise to transfer functions with
fractional powers of s.
Even though n may assume both rational and
irrational values in (4), the names fractional
calculus and fractional order systems are commonly used for purely historical reasons. Some
authors replace fractional with non-integer or
generalised, however.
Thorough expositions of these subjects may be
found in [2,6,7].
2.2. Integer order approximations
The most usual way of making use, both in
simulations and hardware implementations, of
transfer functions involving fractional powers of s
is to approximate them with usual (integer order)
transfer functions with a similar behaviour. So as to
perfectly mimic a fractional transfer function, an
integer transfer function would have to include an
innite number of poles and zeroes. Nevertheless, it
is possible to obtain reasonable approximations
with a nite number of zeroes and poles.
One of the best-known approximations is due to
Manabe and Oustaloup and makes use of a
recursive distribution of poles and zeroes. The
approximating transfer function is given by [8]
N
Y
1 s=oz;n
n1

if n 0;
if n40;
n minfk 2 N : k4ng:

if n  1onon 2 N:

k0

sn  k

if no0;

if np0;

1 s=op;n

n40.

(5)

The approximation is valid in the frequency range

ol ; oh . Gain k is adjusted so that both sides of (5)
shall have unit gain at 1 rad/s. The number of poles
and zeroes N is chosen beforehand, and the good
performance of the approximation strongly depends

ARTICLE IN PRESS
D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

thereon: low values result in simpler approximations, but also cause the appearance of a ripple in
both gain and phase behaviours; such a ripple may
be practically eliminated increasing N, but the
approximation will be computationally heavier.
Frequencies of poles and zeroes in (5) are given by
p
oz;1 ol Z,
6
op;n oz;n a; n 1 . . . N,
oz;n1 op;n Z; n 1 . . . N  1,

7
8

a oh =ol n=N

Z oh =ol

1n=N

n d

s s s ;

n n d ^ n 2 Z ^ d 2 0; 1.

e
F

+
-

G*

+
+

Fig. 1. Block diagram for internal model control.

10
+

The case no0 may be dealt with inverting (5). But if

jnj41 approximations become unsatisfactory; for
that reason (among others), it is usual to split
fractional powers of s like this:
n

2773

+
+

Fig. 2. Block diagram equivalent to that of Fig. 1.

(11)

In this manner only the latter term has to be

approximated.
If a discrete transfer function approximation is
sought, the above approximation (or any other
alternative approximation) may be discretised using
any usual method (Tustin, Simpson, etc.). But there
are also formulas that directly provide discrete
approximations. None shall be needed in what
follows. See, for instance [9] for more on this
subject.

were unity, control would be perfect. Since no

models are perfect, e will not be exactly the
disturbance. That is also exactly why F exists and
is usually a low-pass lter: to reduce the inuence of
high-frequency modelling errors. It also helps
ensuring that product FG  is realisable.
The interconnections of Fig. 1 are equivalent to
those of Fig. 2 if the controller C is given by

3. Analytical tuning methods

Controller C is not, in the general case, a PID or a

fractional PID, but in some cases it will, if

The transfer function of a fractional PID is given

by

I
Dsm .
(12)
sl
In this section, two methods published in the
literature for analytically tuning the ve parameters
of such controllers are given.

Cs P

3.1. Internal model control

The internal model control methodology may, in
some cases, be used to obtain PID or fractional PID
controllers. It makes use of the control scheme of
Fig. 1. In that control loop, G is the plant to control,
G  is an inverse of G (or at least a plant as close as
possible to the inverse of G), G 0 is a model of G and
F is some judiciously chosen lter. If G0 were exact,
the error e would be equal to disturbance d. If,
additionally, G were the exact inverse of G and F

FG 
.
1  FG  G 0

K
eLs .
1 sm T

(13)

(14)

Firstly, let
1
,
1 sT F
m
1s T
,
G
K
K
G0
1  sL.
1 sm T
F

15
16
17

Note that the delay of G was neglected in G  but not

in G0 , where an approximation consisting of a
truncated McLaurin series has been used. Then (13)
becomes
C

1=KT F L T=KT F L

.
s
s1m

(18)

This can be viewed as a fractional PID controller

with the proportional part equal to zero.

ARTICLE IN PRESS
2774

D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

Secondly, let
F 1,
1 sm T
,
G
K
K
1
G0
.
m
1 s T 1 sL

19
20
21

Then (13) becomes

1 1=KL T=KL T m

1m s .
(22)
K
s
s
K
If one of the two integral parts is neglectable, (22)
will be a fractional PID controller. (The price to pay
for neglecting a term is some possible slight
deterioration in performance.)
Finally, if a Pade approximation with one pole
and one zero is used in G 0 ,

F 1,
1 sm T
,
G
K
K
1  sL=2
G0
,
m
1 s T 1 sL=2

23
24

(4) So as to reject output disturbances and closely

follow references, the sensitivity function must
have a small magnitude at low frequencies; thus
it is required that at some specied frequency ol
its magnitude be less than some specied gain:




1


(30)
1 Cjo Gjo oN.
l

(5) So as to be robust in face of gain variations of

the plant, the phase of the open-loop transfer
function must be (at least roughly) constant
around the gain-crossover frequency:


d
argCjoGjo
0.
(31)
do

25

1
1=KL
T=KL
T m

s 1m
s .
(26)
2K
s
s
2K
Again, (26) will be a fractional PID if one of the two
integral parts is neglectable.
Obviously, should m 2 Z, Eqs. (18), (22) and (26)
become usual PIDs.
C

3.2. Tuning by minimisation

Monje et al. [4] proposed that fractional PIDs be
tuned by requiring them to satisfy the following ve
conditions (C being the controller and G the plant):
(1) The gain-crossover frequency ocg is to have
some specied value:
(27)

(2) The phase margin jm is to have some specied

value:
p jm argCjocg Gjocg .

(29)

oocg

then (13) becomes

jCjocg Gjocg j 0 dB.

tude be less than some specied gain:



 Cjoh Gjoh 


1 Cjo Gjo oH.
h
h

(28)

(3) So as to reject high-frequency noise, the closedloop transfer function must have a small
magnitude at high frequencies; thus it is required
that at some specied frequency oh its magni-

Conditions are ve because ve are the parameters

to tune. To satisfy them all the authors proposed the
use of numerical optimisation algorithms, namely of
NelderMeads simplex method as implemented in
Matlabs function fmincon (the condition in (27) is
assumed as the condition to minimise; conditions in
(28)(31) are assumed as constraints). This is
effective but allows local minima to be obtained.
In practice most solutions found with this
optimisation method are good enough, but they
strongly depend on initial estimates of the parameters provided. Some may be discarded because
they are unfeasible or lead to unstable loops, but in
many cases it is possible to nd more than one
acceptable fractional PID. In other cases, only wellchosen initial estimates of the parameters allow
nding a solution.
4. Tuning rules
The rst ZieglerNichols rule for tuning an integer
PID assumes the plant to have an S-shaped unit-step
response, as that of Fig. 3, where L is an apparent
delay and T may be interpreted as a time constant,
such as the one resulting from a pole. The method
cannot be applied if the unit-step response is shaped
otherwise. The simplest plant with such a response is
Gs

K
eLs .
1 sT

(32)

ARTICLE IN PRESS
D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

2775

4.1. First set of rules

po

in
t

P  0:0048 0:2664L 0:4982T

ge

nt

output

at

in
fle
ct
io
n

A rst set of rules is given in Table 1. This is to be

read as

ta
n

0:0232L2  0:0720T 2  0:0348TL

33

and so on. They may be used if

inflection point

0:1pTp50

Lp2

and

(34)

and were designed for the following specications:

0

L+T
time

Fig. 3. S-shaped unit-step response.

The minimisation tuning method presented above in

Section 3.2 was applied to plants given by (32) for
several values of L and T, with K 1. The parameters
of fractional PIDs thus obtained vary in a regular
manner with L and T. Using the least-squares method,
it is possible to translate that regularity into polynomial formulas to nd acceptable values of the
parameters from the values of L and T. These are
given in the two following subsections.2
Two comments. Firstly, to implement the minimisation tuning method the last condition was
veried numerically, evaluating argument in (31) at
two frequencies, equal to ocg =1:122 and 1:122 ocg
1
(this corresponds to 20
of a decade for each side of
ocg ). It is of course possible to evaluate the
argument at other frequencies around ocg ; actually,
the larger the interval where the argument is
constant (or nearly so) the better, and thus using
more than two points might ensure that. However,
it was veried that such stronger requirements are
so difcult to meet that they often prevent a
solution from being found.
Secondly, least-squares method-adjusted formulas cannot exactly reproduce every change in
parameters. This means that fractional PIDs tuned
with the rules presented below never behave as well
as those tuned analytically (including those they are
based upon), neither are they so robust. In other
words, conditions in Eqs. (27)(31) will only
approximately be veried.

These rules were already presented in [10].

ocg 0:5 rad=s,

35


jm 2=3 rad  38 ,

36

oh 10 rad=s,
ol 0:01 rad=s,

37
38

H 10 dB,
N 20 dB.

39
40

Recall that specications are only approximately

veried.
4.2. Second set of rules
A second set of rules is given in Table 2. This
second set of rules may be applied if
0:1pTp50

and

Lp0:5.

(41)

Only one set of parameters is needed in this case

because the range of values of L with which these
rules cope with is more reduced. They were designed
for the following specications:
ocg 0:5 rad=s,

42

Table 1
Parameters for the rst set of tuning rules
P

Parameters to use when 0:1pTp5

1
0:0048
0:3254
1:5766
L
0:2664
0:2478
0:2098
T
0:4982
0:1429
0:1313
0:0232
0:1330
0:0713
L2
0:0720
0:0258
0:0016
T2
LT
0:0348
0:0171
0:0114

0:0662
0:2528
0:1081
0:0702
0:0328
0:2202

0:8736
0:2746
0:1489
0:1557
0:0250
0:0323

Parameters to use
1
2:1187
L
3:5207
T
0:1563
1:5827
L2
0:0025
T2
LT
0:1824

1:1421
1:3707
0:0357
0:5552
0:0002
0:2630

1:2902
0:5371
0:0381
0:2208
0:0007
0:0014

when 5pTp50
0:5201
1:0645
2:6643
0:3268
0:3453
0:0229
1:0944
0:2018
0:0002
0:0003
0:1054
0:0028

ARTICLE IN PRESS
D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

2776

Table 2
Parameters for the second set of tuning rules
l

1:0574
24:5420
0:3544
46:7325
0:0021
0:3106

0:6014
0:4025
0:7921
0:4508
0:0018
1:2050

1:1851
0:3464
0:0492
1:7317
0:0006
0:0380

0:8793
15:0846
0:0771
28:0388
0:0000
1:6711

0:2778
2:1522
0:0675
2:4387
0:0013
0:0021

jm 1 rad  57 ,

output

43

oh 10 rad=s,
ol 0:01 rad=s,

44
45

H 20 dB,

46

N 20 dB.

47

0.5

10

(a)

gain / dB

1
L
T
L2
T2
LT

1.5

Of course, sets of rules other than those two above

might have been found in the same way as these
were for different sets of specications.

20

30

40

50

time / s

50
0
-50
10-2

10-1

100
/ rad s

101

102

101

102

101

102

101

102

-1

-200
-400
-600

(b)

10-2

10-1

100

10-1

100

0
gain / dB

In this section the rules from Section 4 are applied

to three different plants. The performance of the
results is then asserted and compared to the
performance obtained with integer PIDs tuned with
the rst ZieglerNichols rule.
Two comments. Firstly, as stated above, rules
usually lead to results poorer than those they were
devised to achieve. (The same happens with
ZieglerNichols rules: they are expected to result
in an overshoot around 25%, but it is not hard to
nd plants with which the overshoot is 100% or
even more.) Secondly, ZieglerNichols rules make
no attempt to reach always the same gain-crossover
frequency, or the same phase margin. Actually,
these two performance indicators vary widely as L
and T vary. This adds some exibility to Ziegler
Nichols rules: they can be applied for wide ranges of
L and T and still achieve a controller that stabilises
the plant. Rules from the previous section always
aim at fullling the same specications, and that is
why their application range is never so broad as that
of ZieglerNichols rules.
Bode diagrams presented are exact; all timeresponses involving fractional derivatives and integrals were obtained with simulations making use
of Oustaloups approximation described in the
subsection dealing with approximations, with

-20
-40
10-2

/ rad s

-1

0
gain / dB

5. Examples

phase /

(c)

-20
-40
-60
-80
10-2

10-1

100

1
Fig. 4. (a) Step response of (51) controlled with (52) when K is 32
,
1 1 1 1
16, 8, 4, 2, 1 (thick line), 2, 4 and 8. (b) Open-loop Bode diagram
when K 1. (c) Closed-loop function gain (top) and sensitivity
function gain (bottom) when K 1.

ol 103 rad=s,

48

49
50

oh 10 rad=s,
N 7.

ARTICLE IN PRESS

1.5

1.5

output

output

D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

2777

K
0.5

0.5
K

10

20

40

50

50
0
-50
10-2

-1

10

10

10

10

phase /

phase /

-400

10-1

100

101

102

(b)

gain / dB

gain / dB

-20

50
0

10-1

100

101

102

100

101

102

100

101

102

101

102

-200
-400

10-2

10-1

-20
-40

-40
10-1

100

101

10-2

102

0
gain / dB

0
-20
-40
-60
-1

10

10

10-1

/ rad s-1

/ rad s-1

gain / dB

50

(c)

40

-600

-600

-80
10-2

30

-200

10-2

20

/ rad s-1

/ rad s-1

10-2

10

time / s

-50
10-2

(b)

(a)

time / s

gain / dB

gain / dB

(a)

30

10

(c)

-20
-40
-60
-80
10-2

10-1

100

10

1
Fig. 5. (a) Step response of (51) controlled with (53) when K is 32
,
1 1 1 1
,
,
,
,
1
(thick
line).
(b)
Open-loop
Bode
diagram
when
K

1.
16 8 4 2
(c) Closed-loop function gain (top) and sensitivity function gain
(bottom) when K 1.

1
Fig. 6. (a) Step response of (51) controlled with (54) when K is 32
,
1 1 1 1
,
,
,
and
1
(thick
line).
(b)
Open-loop
Bode
diagram
when
16 8 4 2
K 1. (c) Closed-loop function gain (top) and sensitivity
function gain (bottom) when K 1.

ARTICLE IN PRESS
D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

2778

5.1. First-order plant with delay

The plant considered was
Gs

K 0:1s
e
.
1s

(51)

The nominal value of K is 1. Controllers obtained

with the two tuning rules from the previous section
and with the rst ZieglerNichols rule are
0:5158
0:2045s1:0202 ,
s1:4277
1:3106
C 2 s 1:2507 1:1230  0:2589s0:1533 ,
s
60:0000
0:6000 s.
C ZN s 12:0000
s

C 1 s 0:4448

52

according to the 1090% rule and the settling time

is reckoned according to the 5% rule.)
The corresponding open-loop Bode diagrams and
the gains of sensitivity and closed-loop functions are
also given in those gures. They show that the
desired conditionsgiven by Eqs. (27)(31)are
reasonablythough not exactlyfollowed. The
approximations incurred by the least-squares t
are responsible for the conditions being only
approximately veried.
5.2. Second-order plant

53
54

(Note that due to the approximations involved one

of the gains is negative. This will not, however,
affect results.) Corresponding step responses are
given in Figs. 46. These show what happens for
several values of K, the plants gain, which is
assumed to be known with uncertainty. It should be
noticed that fractional PIDs can deal with a clearly
broader range of values of K. This is likely because
the specications the integer PID tries to achieve are
different: that is why responses are all faster, at the
cost of greater overshoots. More important is that
the overshoot is fairly constant with fractional
PIDsat least for those values closer to 1. This is
because fractional PIDs attempt to verify (31),
which the integer PID does not. Data on these
responses are summed up in Table 3. (In this and
the following tables, the rise time is reckoned

The plant considered was

Gs

4:3200s2

K
K
e0:2s

19:1801s 1 1 20s
(55)

with a nominal value of K of 1. The approximation

stems from the values of L and T obtained from its
step response.
Controllers obtained with the two rules given
above and with the rst ZieglerNichols rule are
6:5185
2:5881s0:6957 ,
56
s0:6751
12:4044
C 2 s 6:9928 0:6000 4:1066s0:7805 ,
57
s
300:0000
12:0000 s.
58
C ZN s 120:0000
s
The step responses obtained (together with openloop Bode diagrams and sensitivity and closed-loop
functions gains) are given in Figs. 79. Table 4
presents data on the step responses. This time, since
C 1 s 0:0880

Table 3
Data on step responses of Figs. 46
K

1
32
1
16
1
8
1
4
1
2

1
2
4
8

Controller of Eq. (52)

Controller of Eq. (53)

Controller of Eq. (54)

Rise time
(s)

Overshoot
(%)

Settling time
(s)

Rise time
(s)

Overshoot
(%)

Settling time
(s)

Rise time
(s)

Overshoot
(%)

Settling time
(s)

22.1
13.8
8.9
5.9
4.0
2.6
1.7
0.9
0.2

26
27
28
30
30
27
20
12
7

94.6
59.1
36.5
22.6
19.8
14.6
7.4
5.5
3.9

28.1
15.1
8.1
4.4
2.4
1.3
0.7
0.3

5
5
5
6
8
9
8
8

78.2
19.2
10.2
12.4
7.7
4.7
2.8
1.4

1.0
0.6
0.4
0.2
0.1
0.1

23
33
40
45
48
74

3.7
3.9
2.8
1.9
1.3
0.7

ARTICLE IN PRESS
D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

there is no delay, the plant is easier to control and a

wider variation of K is supported by all controllers.
But fractional PIDs still achieve an overshoot

that is more constantin spite of the plant having

a structure different from that used to derive the
rules.

1.5

1.5

output

output

2779

0.5

0.5
K

10

20

(a)

30

40

50

time / s

50
0

-4

10

-2

10

10
/ rad s

10

50

-50

50

-100

10-4

10-2

100

102

102

-1

-100

(b)

10-4

10-2

100

102

10-2

100

102

0
gain / dB

-40
-60
10-2

100

-20
-40
-60
-80
10-4

102

/ rad s

/ rad s-1

-1

0
gain / dB

0
-50
-100
10-4

100

- 150

-20

-80
10-4

10-2

/ rad s

gain / dB

40

50

10-4

-150

(c)

30
time / s

-1

phase /

phase /

20

-50

-50

gain / dB

10

100
gain / dB

gain / dB

100

(b)

(a)

10-2

100

102

1
Fig. 7. (a) Step response of (55) controlled with (56) when K is 32
,
1 1 1 1
16, 8, 4, 2, 1 (thick line), 2, 4, 8, 16 and 32. (b) Open-loop Bode
diagram when K 1. (c) Closed-loop function gain (top) and
sensitivity function gain (bottom) when K 1.

(c)

-50
-100
10-4

10-2

100

102

1
Fig. 8. (a) Step response of (55) controlled with (57) when K is 32
,
1 1 1 1
16, 8, 4, 2, 1 (thick line), 2, 4, 8, 16 and 32. (b) Open-loop Bode
diagram when K 1. (c) Closed-loop function gain (top) and
sensitivity function gain (bottom) when K 1.

ARTICLE IN PRESS
D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

2780

5.3. Fractional-order plant with delay

1.5

The plant considered was

output

Gs
1

0.5
K

10

20
30
time / s

(a)

40

50

K
K
p e0:5s 
e0:1s
1 1:5s
1 s

(59)

with a nominal value of K of 1. The approximation

is derived from the plants step response at
t 0:92 s. (It might seem more reasonable to base
the approximation on the step response at t 0:5 s,
but this cannot be done, since the response has an
innite derivative at that time instant.)
Controllers obtained with the two rules
given above and with the rst ZieglerNichols
rule are

gain / dB

100
50
0
-50
10-4

0:6187
0:3105s1:0618 ,
s1:3646
1:6486
C 2 s 1:4098 1:1011  0:2139s0:1855 ,
s
90:0000
0:9000 s.
C ZN s 18:0000
s
C 1 s 0:6021

10-2

100

102

/ rad s-1

60
61
62

phase /

0
-50
-100
-150

(b)

10-4

10-2

100

102

10-2

100

102

gain / dB

0
-20
-40
-60
-80
10-4

/ rad s

-1

gain / dB

0
50
100

(c)

10-4

10-2

100

102

1
Fig. 9. (a) Step response of (55) controlled with (58) when K is 32
,
1 1 1 1
,
,
,
,
1
(thick
line),
2,
4,
8,
16
and
32.
(b)
Open-loop
Bode
16 8 4 2
diagram when K 1. (c) Closed-loop function gain (top) and
sensitivity function gain (bottom) when K 1.

The step responses obtained (together with openloop Bode diagrams and sensitivity and closed-loop
functions gains) are given in Figs. 1012. Table 5
presents data on the step responses. The PID
performs poorly because it tries to obtain a fast
response and thus employs higher gains (and hence
1
the loop becomes unstable if K is larger than 32
), but
that is not what is most relevant. The most relevant
result here is that fractional PIDs still achieve
practically constant overshoots, since, in spite of the
different plant structure, the conditions they were
expected to verify are still veried to a reasonable
degree, as the frequencyresponse plots show.
In this case it is possible to nd IMC-tuned
fractional PIDs to compare results. Using the
parameters of plant (59) (K 1, T 1, L 0:5
and m 0:5; these are not the parameters of
the approximation used with the tuning rules),
Eqs. (22) and (26) yield the two following transfer
functions:
2
2
C IMC 1 1=2 s1=2 ,
s s
1 2
2
1
C IMC 1=2 s1=2 .
2 s s
2

63
64

In none of the two cases is clear which of the two

integral terms is better to discard. By trying both
possibilities, it is found out that it is better to keep

ARTICLE IN PRESS
D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

2781

Table 4
Data on step responses of Figs. 79
K

1
32
1
16
1
8
1
4
1
2

1
2
4
8
16
32

Controller of Eq. (52)

Controller of Eq. (53)

Controller of Eq. (54)

Rise time
(s)

Overshoot
(%)

Settling time
(s)

Rise time
(s)

Overshoot
(%)

Settling time
(s)

Rise time
(s)

Overshoot
(%)

Settling time
(s)

31.4
15.6
9.1
5.7
3.8
2.6
1.7
1.2
0.8
0.5
0.2

8
19
28
34
36
35
31
23
15
8

45.5
38.2
47.1
32.1
22.1
15.3
10.5
7.2
3.3
2.5
1.8

21.7
10.9
6.3
3.8
2.4
1.5
0.9
0.5
0.3
0.2
0.1

5
13
17
19
19
16
13
9
6
6

148.0
23.3
19.2
13.3
9.0
5.9
3.8
2.5
1.5
0.7
0.3

1.5
1.0
0.7
0.5
0.4
0.3
0.2
0.2
0.1
0.1
0.1

75
74
71
66
61
53
43
31
22
17
15

41.2
25.9
14.0
8.2
4.6
2.0
1.3
0.8
0.8
0.8
0.8

the terms with the rst derivative:

2
C IMC2 1 s1=2 ,
s
1 2 1 1=2
C IMC1 s .
2 s 2

65
66

Step responses obtained are shown in Fig. 13 and

compare well with those of Figs. 10 and 11. It is seen
that rule-tuned fractional PIDs perform nearly as
well as those found with IMC.
6. Comments and conclusions
In this paper, two analytical methods (among
others published in the literature) for tuning the
parameters of fractional PIDs were reviewed. The
optimisation method of [4] was then used for
developing two sets of tuning rules similar to those
of the rst set of ZieglerNichols rules. These new
tuning rules make use of two parameters (L and T)
of the unit-step response of the plant, which should
be S-shaped: otherwise they cannot be applied.
The most obvious difference is that the new rules
are clearly more complicated than those of Ziegler
Nichols: they have to be quadratic (approximations
of lower order were tried but proved unsatisfactory). And the broader the application range of the
rules is to be, the more complicated they become:
the rst rule needs two tables of parameters, while
the second, good for a narrower interval of values of
L only, needs only one.

The usefulness of these rules is that of all sets of

rules: they may be applied even if no model of the
plant is available, provided a suitable time response
is; they may be used as a departing point for netuning (this is, for instance relevant if the optimisation tuning method is used, since its results depend
signicantly from the initial estimate provided);
they are easier and faster to apply than analytic
methods. Their drawbacks are also those of all sets
of rules: their performance is often inferior to the
one sought, ne-tuning being often needed; they
perform worse than controllers tuned analytically;
they cannot be applied to all types of plants, but
only to those with a particular sort of time response.
Fractional PIDs tuned with these new rules
compare well with integer PIDs tuned according
to the rst ZieglerNichols rule, even though the
comparison is made difcult because Ziegler
Nichols rules achieve different specications for
different values of T and L while the new rules
attempt to always keep a uniform result. The
advantage fractional PIDs provide is a roughly
constant overshoot when the gain of the plant
undergoes variations. (It is of course likely that
carefully tuned integer PIDs perform better than
rule-tuned fractional PIDsjust as carefully tuned
fractional PIDs are likely to perform better than
rule-tuned integer PIDs.)
It is surely possible to improve these tuning rules.
Rules similar to the second ZieglerNichols rule
(making use of a closed-loop response of the plant)
are certainly possible, and are currently being

ARTICLE IN PRESS
D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784
1.4

1.4

1.2

1.2

0.8

0.8

output

output

2782

0.6

0.6
0.4

0.4

K
0.2

0.2
0

10

20

(a)

30

40

50

80
gain / dB

gain / dB

20
0
-1

10

10

10

10

40

50

40
20
0
10-2

10-1

100

101

102

101

102

101

102

101

102

/ rad s-1
0
phase /

phase /

30
time / s

-20

-500

-1000
10-2

20

60

40

/ rad s-1

(b)

10

80

60

-20
10-2

(a)

time / s

10-1

100

101

102

(b)

-500

-1000
10-2

10-1

100

10-1

100

0
gain / dB

gain / dB

0
-20
-40
10-2

10-1

100

101

-20
-40

102

10-2

/ rad s-1

/ rad s

(c)

-20
gain / dB

gain / dB

-40
-60
-80
10-2

-1

-1

10

10

10

10

Fig. 10. (a) Step response of (59) controlled with (60) when K is
1 1 1 1 1
32, 16, 8, 4, 2, 1 (thick line) and 2. (b) Open-loop Bode diagram
when K 1. (c) Closed-loop function gain (top) and sensitivity
function gain (bottom) when K 1.

(c)

-20
-40
-60
-80
10-2

10-1

100

Fig. 11. (a) Step response of (59) controlled with (61) when K is
and 1 (thick line). (b) Open-loop Bode diagram when
K 1. (c) Closed-loop function gain (top) and sensitivity
function gain (bottom) when K 1.
1 1 1 1 1
32, 16, 8, 4, 2

ARTICLE IN PRESS
1.4
1.2

0.2
0

10

(a)

20

30

40

50

time / s

31

0.6 s

0.4

Rise time
(s)

0.6

Overshoot
(%)

0.8

Controller of Eq. (54)

output

2783

3.1 s

Settling time
(s)

D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

40
20
0
-20
10-2

10-1

100

101

102

100

101

102

100

101

102

101

102

86.6
47.7
27.1
16.0
9.6
5.6

60

Settling time
(s)

gain / dB

80

/ rad s

-1

-20

Overshoot
(%)

7
7
8
9
9
8

26.5
14.7
8.2
4.6
2.4
1.1

gain / dB

Rise time
(s)

10-1

Rise time
(s)

-1000
10-2

Controller of Eq. (53)

(b)

-500

Controller of Eq. (52)

phase /

10-1

/ rad s-1

-20

Fig. 12. (a) Step response of (59) controlled with (62) when K is
(b) Open-loop Bode diagram when K 1. (c) Closed-loop
function gain (top) and sensitivity function gain (bottom) when
K 1.

1
32.

developed. Rules specic for non-minimum phase

plants (with which ZieglerNichols rules do not
properly deal) may also be of interest.

1
2

100

25.1
15.8
10.3
6.9
4.4
2.7
1.5

10-1

1
32
1
16
1
8
1
4
1
2

-80
10-2

(c)

Table 5
Data on step responses of Fig. 1012.

-60

26
27
28
29
27
23
17

-40

Overshoot
(%)

gain / dB

105.5
65.8
41.2
26.1
17.2
11.9
8.6

10-2

Settling time
(s)

-40

ARTICLE IN PRESS
D. Valerio, J.S. da Costa / Signal Processing 86 (2006) 27712784

2784
1.4

References

1.2

[1] T. Hagglund, K. Astrom, Automatic tuning of PID

controllers, in: W.S. Levine (Ed.), The Control Handbook,
CRC Press, Boca Raton, FL, 1996, pp. 817826.
[2] I. Podlubny, Fractional Differential Equations: An Introduction to Fractional Derivatives, Fractional Differential
Equations to Methods of their Solution and Some of their
Applications, Academic Press, San Diego, 1999.
[3] R. Caponetto, L. Fortuna, D. Porto, Parameter tuning of a
non integer order PID controller, in: Fifteenth International
Symposium on Mathematical Theory of Networks and
Systems, Notre Dame, 2002.
[4] C.A. Monje, B.M. Vinagre, Y.Q. Chen, V. Feliu, P. Lanusse, J.
Sabatier, Proposals for fractional PIl Dm tuning, in: Fractional
Differentiation and its Applications, Bordeaux, 2004.
[5] Y.Q. Chen, K.L. Moore, B.M. Vinagre, I. Podlubny, Robust
PID controller autotuning with a phase shaper, in:
Fractional Differentiation and its Applications, Bordeaux,
2004.
[6] K.S. Miller, B. Ross, An Introduction to the Fractional
Calculus and Fractional Differential Equations, Wiley, New
York, 1993.
[7] S.G. Samko, A.A. Kilbas, O.I. Marichev, Fractional Integrals
and Derivatives, Gordon and Breach, Yverdon, 1993.
[8] A. Oustaloup, La commande CRONE: commande robuste
dordre non entier, Herme`s, Paris, 1991.
[9] B.M. Vinagre, I. Podlubny, A. Hernandez, V. Feliu, Some
approximations of fractional order operators used in control
theory and applications, Fractional Calculus & Appl. Anal.
3 (2000) 231248.
[10] D. Valerio, J. Sa da Costa, ZieglerNichols type tuning rules
for PID controllers, in: Fifth ASME International Conference on Multibody Systems, Nonlinear Dynamics and
Control, Long Beach, 2005.

output

1
0.8
0.6
K

0.4
0.2
0

10

20

(a)

30

40

50

30

40

50

time / s
1.4
1.2

output

1
0.8
0.6
K

0.4
0.2
0

(b)

10

20
time / s

Fig. 13. (a) Step response of (59) controlled with (65) when K is
1 1 1 1
1
32, 16, 8, 4 and 2. (b) Step response of (59) controlled with (66)
1 1 1 1 1
when K is 32
, 16, 8, 4, 2 and 1 (thick line).