See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/267485936

Ziegler-Nichols Type Tuning Rules for Fractional PID Controllers

Conference Paper · January 2005

DOI: 10.1115/DETC2005-84344

CITATIONS READS

52 902

2 authors, including:

Duarte Valério
Technical University of Lisbon
112 PUBLICATIONS   2,137 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Duarte Valério on 03 October 2016.

The user has requested enhancement of the downloaded file.


ASME 2005 International Design Engineering Technical Conferences and
Computers and Information in Engineering Conference
Long Beach, California, USA, September 24–28, 2005
ZIEGLER-NICHOLS TYPE TUNING RULES FOR FRACTIONAL PID CONTROLLERS
Duarte Valério ∗
Department of Mechanical Engineering, GCAR
Instituto Superior Técnico, Technical Univ. of Lisbon
Av. Rovisco Pais 1, 1049-001 Lisboa
Portugal
Email: dvalerio@dem.ist.utl.pt
ABSTRACT
This paper presents two sets of tuning rules for fractional
PIDs that rely solely on the same plant time-response data used
by the first Ziegler-Nichols tuning rule for (usual, integer) PIDs.
Thus no model for the plant to control is needed; only an S-
shaped step response is. These rules are quadratic and their
results compare well with those obtained with rule-tuned integer
PIDs.
INTRODUCTION
Controllers whose output is a linear combination of the
input, the derivative of the input and the integral of the in-
put, known as PID (proportional—derivative—integrative) con-
trollers, are widely used and enjoy significant popularity, because
they are simple, effective and robust.
One of the reasons why this is so is the existence of tuning
rules for finding suitable parameters for PIDs, rules that do not
require any model of the plant to control. All that is needed to ap-
ply such rules is to have a certain time response of the plant. Ex-
amples of such sets of rules are those due to Ziegler and Nichols,
those due to Cohen and Coon, and the Kappa-Tau rules [1]. It
is true that PIDs tuned with such rules often perform in a non-
optimal way. But even though further fine-tuning be possible and
sometimes necessary, rules provide a good starting point. Their
usefulness is obvious when no model of the plant is available,
∗ Address
all correspondence to this author.
1 c 2005 by ASME
Proceedings of IDETC/CIE 2005
DETC2005-84344
José Sá da Costa
Department of Mechanical Engineering, GCAR
Instituto Superior Técnico, Technical Univ. of Lisbon
Av. Rovisco Pais 1, 1049-001 Lisboa
Portugal
Email: sadacosta@dem.ist.utl.pt
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 PIλ Dµ controllers, where λ and µ are the in-
tegration and differentiation orders; if both values are 1, the re-
sult is a usual PID (henceforth called “integer” PID as opposed
to a fractional PID). They have been increasingly used over the
last years, but 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 sets of
tuning rules for fractional PIDs. Proposed rules bear similarities
to the first rule proposed by Ziegler and Nichols for integer PIDs,
making use of the same plant time response data.
The paper is organised as follows. Next section sums up the
fundamentals of fractional calculus needed to understand frac-
tional PIDs. Then 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. The last two sections
give some examples of application and draw some conclusions.
FRACTIONAL ORDER SYSTEMS
Definitions
Fractional calculus is a generalisation of ordinary calculus.
The main idea is to develop a functional operator D, associated
Copyright °
to an order ν not restricted to integer numbers, that generalises The approximation is valid in the frequency range [ωl ; ωh ]; gain
the usual notions of derivatives (for a positive ν) and integrals k is adjusted so that the approximation shall have unit gain at
(for a negative ν). The most usual definition of D is due to Rie- 1 rad/s; the number of poles and zeros N is chosen beforehand
mann and Liouville (although there are others) and generalises (low values resulting in simpler approximations but also causing
the equalities the appearance of a ripple in both gain and phase behaviours);
frequencies of poles and zeros are given by
(x − t)n−1
Z x
−n
c Dx f (x) = f (t) dt, n ∈ N (1) √
c (n − 1)! ωz,1 = ωl η (6)
Dn Dm f (x) = D n+m
f (x) , m ∈ Z−
0 ∨ n, m ∈ N0 (2) ω p,n = ωz,n α, n = 1 . . . N (7)
ωz,n+1 = ω p,n η, n = 1 . . . N − 1 (8)
which are easily proved for integer orders. The full definition of ν

D becomes α = (ωh /ωl ) N (9)

1−ν
η = (ωh /ωl ) N (10)
x (x − ξ)−ν−1
Z
f (ξ) dξ, if ν < 0


Γ (−ν)

ν c
c Dx f (x) = The case ν < 0 may be dealt with inverting Eqn. (5). But if
f (x) , if ν = 0
|ν| > 1 approximations become unsatisfactory; for that reason, it

D [c Dν−n
 n
f (x)] , if ν > 0, n = min {k ∈ N}

x is usual to make
(3)
It is worth noticing that, when ν is positive but non-integer, oper-
ator D still needs integration limits c and x; in other words, D is sν = sn sδ , n + δ = ν ∧ n ∈ Z ∧ δ ∈ [0; 1] (11)
a local operator for natural values of ν (usual derivatives) only.
The Laplace transform of D follows the usual rules
and then approximate only the latter term.
 ν If a discrete transfer function approximation is desired, the
 s F (s) , if ν ≤ 0 above approximation (or any other alternative approximation)
n−1

L [0 Dνx f (x)] = sν F (s) − ∑ sk 0 Dxν−k−1 f (0) , may be discretised using any usual method (Tustin, Simpson. . . ).

(4)

 k=0 But there are also formulas that directly provide discrete approx-
if n − 1 < ν < n ∈ N imations. None shall be needed in what follows. See for in-

stance [9] for more on this subject.
and thus, if zero initial conditions are assumed, systems with a
dynamic behaviour described by differential equations involving
fractional derivatives give rise to transfer functions with frac- ANALYTICAL TUNING METHODS
tional powers of s. A fractional PID has a transfer function given by
“Fractional” calculus and “fractional” order systems are the
usual names though ν may assume irrational values in Eqn. (4) I
also. Thorough expositions of these subjects may be found in C (s) = P + + Dsµ (12)
sλ
[2, 6, 7].
In this section two methods published in the literature for analyt-
Integer order approximations ically tuning the five parameters of such controllers are given.
The most usual way of making use, both in simulations and
hardware implementations, of transfer functions involving frac-
tional powers of s is to approximate them with usual (integer or- Internal model control
der) transfer functions with a similar behaviour. Integer transfer The internal model control methodology may, in some cases,
functions would require an infinite number of poles and zeros to be used to obtain PID or fractional PID controllers. It makes use
perfectly mimic fractional transfer functions, but it is neverthe- of the control scheme of Fig. 1, where G is the plant to control,
less possible to obtain reasonable approximations with a finite G∗ is an inverse of G or at least a plant as close as possible to the
number of zeros and poles. One of the best-known approxima- inverse of G, G0 is a model of G and F is some judiciously chosen
tions is due to Oustaloup and is given by [8] filter. If G0 were exact, the error e would be equal to disturbance
d. If, additionally, G∗ were the exact inverse of G and F were
unity, control would be perfect. Since no models are perfect, e
N 1 + ωsz,n
sν = k ∏ , ν>0 (5) will not be exactly the disturbance. That is also exactly why F
n=1 1 + ω sp,n exists and is usually a low-pass filter: to reduce the influence

2 c 2005 by ASME
Copyright °
 e
d
If one of the two integral parts is neglectable, Eqn. (18) will
F G* G be a fractional PID controller. Obviously, should µ ∈ Z, both
Eqn. (16) and Eqn. (18) become usual PIDs.
G'

Tuning by minimisation
Monje et al. [4] proposed that fractional PIDs be tuned by
Figure 1. BLOCK DIAGRAM FOR INTERNAL MODEL CONTROL. requiring them to satisfy the following conditions (C being the
controller and G the plant):

e
d
1. The gain-crossover frequency ωcg is to have some specified
C G
value:

Figure 2. BLOCK DIAGRAM EQUIVALENT TO THAT OF FIG 1.

|C (ωcg ) G (ωcg )| = 0 dB (19)

of high-frequency modelling errors. It also helps ensuring that

2. The phase margin ϕm is to have some specified value:
product FG∗ is realisable. The interconnections of Fig. 1 are
equivalent to those of Fig. 2 if
−π + ϕm = arg [C (ωcg ) G (ωcg )] (20)
FG∗
C= (13)
1 − FG∗ G0 3. So as to reject high-frequency noise, the closed loop transfer
function must have a small magnitude at high frequencies;
Controller C is not, in the general case, a PID or a fractional PID, thus it is required that at some specified frequency ωh its
but it will if magnitude be less than some specified gain:

K
G= e−Ls (14) ¯ C (ωh ) G (ωh ) ¯
¯ ¯
1 + sµ T ¯ 1 +C (ωh ) G (ωh ) ¯ < H
¯ ¯ (21)

Let
4. So as to reject output disturbances and closely follow refer-
1 1 + sµ T K ences, the sensitivity function must have a small magnitude
F= , G∗ = , G0 = (1 − sL) (15)
1 + sTF K 1 + sµ T at low frequencies; thus it is required that at some specified
frequency ωl its magnitude be less than some specified gain:
Notice that the delay of G was neglected in G∗ but not in G0 .
Then Eqn. (13) becomes 1
¯ ¯
¯ 1 +C (ωl ) G (ωl ) ¯ < N (22)
¯ ¯
¯ ¯
1 T
K(TF +L) K(TF +L)
C= + (16)
s s1−µ 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
that can be viewed as a fractional PID controller with the propor- roughly) constant around the gain-crossover frequency:
tional part equal to zero. And if the model of G0 in Eqn. (15) is
improved to d
¯
arg [C (ω) G (ω)]¯¯ =0 (23)
¯
dω ω=ωcg
K 1 − sL/2
G0 = (17)
1 + sµ T 1 + sL/2
Conditions are five because five are the parameters to tune. To
satisfy them all those authors proposed the use of numerical op-
then Eqn. (13) becomes
timisation algorithms, namely those implemented in Matlab’s
function fmincon (the condition in Eqn. (19) is assumed as the
1 T
L K(TF +L) K(TF +L) TL condition to minimise; conditions in Eqns. (20) to (23) are as-
C= + + + sµ (18)
2K (TF + L) s s1−µ 2K (TF + L) sumed as constraints). This is effective but allows local minima

3 c 2005 by ASME
Copyright °
 K First set of rules
A first set of rules is given in Tab. 1. This is to be read as

t
in
po
n
tio

P = −0.0048 + 0.2664L + 0.4982T

c
fle
in

+0.0232L2 − 0.0720T 2 − 0.0348T L (25)

at
nt
ge

inflection
n
ta

point
and so on. They may be used if 0.1 ≤ T ≤ 50 ∧ L ≤ 2 and were
0 L L+T designed for the following specifications:
Figure 3. S-SHAPED UNIT-STEP RESPONSE.
ωcg = 0.5 rad/s (26)
o
to be obtained. In practice most solutions found with this op- ϕm = 2/3 rad ≈ 38 (27)
timisation method are good enough, but they strongly depend ωh = 10 rad/s (28)
on initial estimates of the parameters provided. Some may be ωl = 0.01 rad/s (29)
discarded because they are unfeasible or lead to unstable loops, H = −10 dB (30)
but in many cases it is possible to find more than one acceptable
fractional PID; in others, only well-chosen initial estimates of the N = −20 dB (31)
parameters allow finding a solution.
Recall that specifications are only approximately verified.

TUNING RULES Second set of rules

The first Ziegler-Nichols rule for tuning an integer PID as-
sumes 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 pole. The method cannot be applied if the unit-step response
is shaped otherwise. The simplest plant with such a response is
A second set of rules is given in Tab. 2. These may be ap-
plied if 0.1 ≤ T ≤ 50 ∧ L ≤ 0.5. Only one set of parameters is
needed in this case because the range of values of L these rules
cope with is more reduced. They were designed for the following
specifications:

K ωcg = 0.5 rad/s (32)

G= e−Ls (24)
1 + sT ϕm = 1 rad ≈ 57o (33)
ωh = 10 rad/s (34)
The minimisation tuning method presented above was applied to ωl = 0.01 rad/s (35)
plants given by Eqn. (24) for several values of L and T , with
K = 1. The parameters of fractional PIDs thus obtained vary in H = −20 dB (36)
a regular manner. Using the least-squares method it is possible N = −20 dB (37)
to translate that regularity into formulas to find acceptable values
of the parameters from L and T . Of course, other sets of rules might have been found in the same
Two comments. Firstly, to implement the minimisation tun- way for different specifications.
ing method the last condition was verified numerically, evaluat-
ing argument in Eqn. (23) at two frequencies, equal to ωcg /1.122
and 1.122 ωcg (this corresponds to 1/20 of a decade). It is of EXAMPLES
course possible to evaluate the argument at other frequencies In what follows these rules are applied to three different
around ωcg ; actually, the larger the interval where the argument plants, and the goodness of the results asserted and compared
is constant (or nearly so) the better, and thus using more than to those obtained with integer PIDs tuned with the first Ziegler-
two points might ensure that. However, it was verified that such Nichols rule.
stronger requirements often prevent a solution from being found. Two comments. Firstly, as stated above, rules usually lead to
Secondly, least-square method-adjusted formulas cannot ex- results poorer than those they were devised to achieve. (The same
actly reproduce every change in parameters. This means that happens with Ziegler-Nichols rules: they are expected to result
fractional PIDs tuned with the rules presented below never be- in an overshoot around 25%, but it is not hard to find plants with
have as well as those tuned analytically, neither are they so ro- which the overshoot is 100% or even more.) Secondly, Ziegler-
bust. Conditions in Eqns. (19) to (23) will only approximately be Nichols rules make no attempt to reach always the same gain-
verified. crossover frequency, or the same phase margin. Actually, these

4 c 2005 by ASME
Copyright °
 Table 1. PARAMETERS FOR THE FIRST SET OF TUNING RULES.
Parameters to use when 0.1 ≤ T ≤ 5 Parameters to use when 5 ≤ T ≤ 50
P I λ D µ P I λ D µ
1 −0.0048 0.3254 1.5766 0.0662 0.8736 2.1187 −0.5201 1.0645 1.1421 1.2902
L 0.2664 0.2478 −0.2098 −0.2528 0.2746 −3.5207 2.6643 −0.3268 −1.3707 −0.5371
T 0.4982 0.1429 −0.1313 0.1081 0.1489 −0.1563 0.3453 −0.0229 0.0357 −0.0381
L2 0.0232 −0.1330 0.0713 0.0702 −0.1557 1.5827 −1.0944 0.2018 0.5552 0.2208
T2 −0.0720 0.0258 0.0016 0.0328 −0.0250 0.0025 0.0002 0.0003 −0.0002 0.0007
LT −0.0348 −0.0171 0.0114 0.2202 −0.0323 0.1824 −0.1054 0.0028 0.2630 −0.0014

Table 2. PARAMETERS FOR THE SECOND SET OF TUNING RULES.

Nichols rule are

P I λ D µ 0.5158
C1 (s) = 0.4448 + + 0.2045s1.0202 (39)
s1.4277
1 −1.0574 0.6014 1.1851 0.8793 0.2778 1.3106
C2 (s) = 1.2507 + 1.1230 − 0.2589s0.1533 (40)
L 24.5420 0.4025 −0.3464 −15.0846 −2.1522 s
60.0000
T 0.3544 0.7921 −0.0492 −0.0771 0.0675 CZN (s) = 12.0000 + + 0.6000s (41)
s
L2 −46.7325 −0.4508 1.7317 28.0388 2.4387
T2 −0.0021 0.0018 0.0006 −0.0000 −0.0013 (Notice that due to the approximations involved one of the gains
is negative; this will not, however, affect results.) Corresponding
LT −0.3106 −1.2050 0.0380 1.6711 0.0021 step responses are given in Figs. 4, 5 and 6. These show what
happens for several values of K, the plant’s gain, assumed as
known with uncertainty. It should be noticed that fractional PIDs
can deal with with a clearly broader range of values of K. This is
two performance indicators vary widely as L and T vary. This likely because the specifications the integer PID tries to achieve
adds some flexibility to Ziegler-Nichols rules: they can be ap- are different: that is why responses are all faster, at the cost of
plied for wide ranges of L and T and still achieve a controller greater overshoots. More important is that the overshoot is fairly
that stabilises the plant. Rules from the previous section always constant with fractional PIDs, at least for those values closer to 1.
aim at fulfilling the same specifications, and that is why their ap- This is because fractional PIDs attempt to verify Eqn. (23), which
plication range is never so broad as that of Ziegler-Nichols rules. the integer PID does not. Data on these responses is summed
Bode diagrams presented are exact; all time-responses up in Tab. 3. (In this and the following tables, the rise time is
involving fractional derivatives and integrals were obtained reckoned according to the 10%–90% rule and the settling time is
with simulations making use of Oustaloup’s approximation de- reckoned according to the ±5% rule.)
scribed in the subsection dealing with approximations, with Those figures also give the corresponding open-loop Bode
ωl = 10−3 rad/s, ωh = 103 rad/s and N = 7. diagrams and the gains of sensitivity and closed-loop functions.
They show that the desired conditions—given by Eqns. (19) to
(23)—are reasonably—though not exactly—followed. Differ-
First-order plant with delay ences are due to the approximations incurred by the least-squares
The plant considered was fit.

K −0.1s Second-order plant

G (s) = e (38)
1+s The plant considered was

The nominal value of K is 1. Controllers obtained with the two K K

tuning rules from the previous section and with the first Ziegler- G (s) = ≈ e−0.2s (42)
4.3200s2 + 19.1801s + 1 1 + 20s

5 c 2005 by ASME
Copyright °
 0
1.5 50

gain / dB
gain / dB
−20
0

1 −40
output

−50 −2 −1 0 1 2 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10
ω / rad ⋅ s−1 ω / rad ⋅ s−1
0 0
0.5 −200

gain / dB
−20

phase / º
K
−400 −40
−60
0 −600
0 10 20 30 40 50 −2 −1 0 1 2
−80 −2 −1 0 1 2
(a) time / s (b) 10 10 10 10 10 (c) 10 10 10 10 10

Figure 4. (a) STEP RESPONSE OF EQN. (38) CONTROLLED WITH EQN. (39) WHEN K IS 1/32, 1/16, 1/8, 1/4, 1/2, 1 (THICK LINE), 2, 4 AND
8. (b) OPEN-LOOP BODE DIAGRAM WHEN K = 1. (c) SENSITIVITY FUNCTION AND CLOSED-LOOP FUNCTION GAINS WHEN K = 1.

0
1.5 50

gain / dB
gain / dB

−20
0

1 −40
output

−50 −2 −1 0 1 2 −2 −1 0 1 2
10 10 10 −1 10 10 10 10 10 10 10
ω / rad ⋅ s ω / rad ⋅ s−1
0 0
0.5 −200

gain / dB
−20
phase / º

K
−400 −40
−60
0 −600
0 10 20 30 40 50 −2 −1 0 1 2
−80 −2 −1 0 1 2
(a) time / s (b) 10 10 10 10 10 (c) 10 10 10 10 10

Figure 5. (a) STEP RESPONSE OF EQN. (38) CONTROLLED WITH EQN. (40) WHEN K IS 1/32, 1/16, 1/8, 1/4, 1/2, 1 (THICK LINE), 2 AND 4.
(b) OPEN-LOOP BODE DIAGRAM WHEN K = 1. (c) SENSITIVITY FUNCTION AND CLOSED-LOOP FUNCTION GAINS WHEN K = 1.

0
1.5 50
gain / dB
gain / dB

−20
0

1 −40
output

−50 −2 −1 0 1 2 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10
ω / rad ⋅ s−1 ω / rad ⋅ s−1
K 0 0
0.5 −200
gain / dB

−20
phase / º

−400 −40
−60
0 −600
0 2 4 6 8 10 −2 −1 0 1 2
−80 −2 −1 0 1 2
(a) time / s (b) 10 10 10 10 10 (c) 10 10 10 10 10

Figure 6. (a) STEP RESPONSE OF EQN. (38) CONTROLLED WITH EQN. (41) WHEN K IS 1/32, 1/16, 1/8, 1/4, 1/2 AND 1 (THICK LINE). (b)
OPEN-LOOP BODE DIAGRAM WHEN K = 1. (c) SENSITIVITY FUNCTION AND CLOSED-LOOP FUNCTION GAINS WHEN K = 1.

with a nominal value of K of 1. The approximation stems from 300.0000

CZN (s) = 120.0000 + + 12.0000s (45)
the values of L and T obtained from its step-response. s
Controllers obtained with the two rules given above and with
the first Ziegler-Nichols rule are The step-responses obtained (together with open-loop Bode dia-
grams and sensitivity and closed-loop functions’ gains) are given
in Figs. 7, 8 and 9 and in Tab. 4. This time, since there is no
6.5185 delay, the plant is easier to control and a wider variation of K is
C1 (s) = 0.0880 + + 2.5881s0.6957 (43) supported by all controllers. But fractional PIDs still achieve an
s0.6751
12.4044 overshoot that is more constant, in spite of the different structure
C2 (s) = 6.9928 + 0.6000 − 4.1066s0.7805 (44) of the plant.
s

6 c 2005 by ASME
Copyright °
 Table 3. DATA ON STEP-RESPONSES OF FIG. 4, FIG. 5 AND FIG 6.
Controller of Eqn. (39) Controller of Eqn. (40) Controller of Eqn. (41)
K rise time overshoot settling time rise time overshoot settling time rise time overshoot settling time
1/32 22.1 s 26 % 94.6 s 28.1 s 5% 78.2 s 1.0 s 23 % 3.7 s
1/16 13.8 s 27 % 59.1 s 15.1 s 5% 19.2 s 0.6 s 33 % 3.9 s
1/8 8.9 s 28 % 36.5 s 8.1 s 5% 10.2 s 0.4 s 40 % 2.8 s
1/4 5.9 s 30 % 22.6 s 4.4 s 6% 12.4 s 0.2 s 45 % 1.9 s
1/2 4.0 s 30 % 19.8 s 2.4 s 8% 7.7 s 0.1 s 48 % 1.3 s
1 2.6 s 27 % 14.6 s 1.3 s 9% 4.7 s 0.1 s 74 % 0.7 s
2 1.7 s 20 % 7.4 s 0.7 s 8% 2.8 s — — —
4 0.9 s 12 % 5.5 s 0.3 s 8% 1.4 s — — —
8 0.2 s 7% 3.9 s — — — — — —

100 0
gain / dB

gain / dB
1.5 50 −20
0 −40
−60
−50
1
output

−4 −2 0 2
−80 −4 −2 0 2
10 10 10−1 10 10 10 10−1 10
ω / rad ⋅ s ω / rad ⋅ s
0 0
0.5 −50
gain / dB
phase / º

K −50
−100
−150 −100
0
0 10 20 30 40 50 −4 −2 0 2 −4 −2 0 2
(a) time / s (b) 10 10 10 10 (c) 10 10 10 10

Figure 7. (a) STEP RESPONSE OF EQN. (42) CONTROLLED WITH EQN. (43) WHEN K IS 1/32, 1/16, 1/8, 1/4, 1/2, 1 (THICK LINE), 2, 4, 8, 16
AND 32. (b) OPEN-LOOP BODE DIAGRAM WHEN K = 1. (c) SENSITIVITY FUNCTION AND CLOSED-LOOP FUNCTION GAINS WHEN K = 1.

100 0
gain / dB

gain / dB

1.5 50 −20
0 −40
−60
−50
1
output

−4 −2 0 2
−80 −4 −2 0 2
10 10 10−1 10 10 10 10−1 10
ω / rad ⋅ s ω / rad ⋅ s
0 0
0.5 −50
gain / dB
phase / º

K −50
−100
−150 −100
0
0 10 20 30 40 50 −4 −2 0 2 −4 −2 0 2
(a) time / s (b) 10 10 10 10 (c) 10 10 10 10

Figure 8. (a) STEP RESPONSE OF EQN. (42) CONTROLLED WITH EQN. (44) WHEN K IS 1/32, 1/16, 1/8, 1/4, 1/2, 1 (THICK LINE), 2, 4, 8, 16
AND 32. (b) OPEN-LOOP BODE DIAGRAM WHEN K = 1. (c) SENSITIVITY FUNCTION AND CLOSED-LOOP FUNCTION GAINS WHEN K = 1.

Fractional-order plant with delay with a nominal value of K of 1. The approximation is derived
The plant considered was from the plant’s step-response at t = 0.92 s (the step response at
t = 0.5 s cannot be used since it has an infinite derivative).
K K
G (s) = √ e−0.5s ≈ e−0.1s (46)
1+ s 1 + 1.5s Controllers obtained with the two rules given above and with

7 c 2005 by ASME
Copyright °
 100 0

gain / dB

gain / dB
1.5 50 −20
0 −40
−60
−50
1
output

−4 −2 0 2
−80 −4 −2 0 2
10 10 10 10 10 10 10 10
ω / rad ⋅ s−1 ω / rad ⋅ s−1
0 0
0.5 −50

gain / dB
phase / º
K −50
−100
−150 −100
0
0 10 20 30 40 50 −4 −2 0 2 −4 −2 0 2
(a) time / s (b) 10 10 10 10 (c) 10 10 10 10

Figure 9. (a) STEP RESPONSE OF EQN. (42) CONTROLLED WITH EQN. (45) WHEN K IS 1/32, 1/16, 1/8, 1/4, 1/2, 1 (THICK LINE), 2, 4, 8, 16
AND 32. (b) OPEN-LOOP BODE DIAGRAM WHEN K = 1. (c) SENSITIVITY FUNCTION AND CLOSED-LOOP FUNCTION GAINS WHEN K = 1.
Table 4. DATA ON STEP-RESPONSES OF FIG. 7, FIG. 8 AND FIG 9.
Controller of Eqn. (43) Controller of Eqn. (44) Controller of Eqn. (45)
K rise time overshoot settling time rise time overshoot settling time rise time overshoot settling time
1/32 31.4 s — 45.5 s 21.7 s — 148.0 s 1.5 s 75 % 41.2 s
1/16 15.6 s 8% 38.2 s 10.9 s 5% 23.3 s 1.0 s 74 % 25.9 s
1/8 9.1 s 19 % 47.1 s 6.3 s 13 % 19.2 s 0.7 s 71 % 14.0 s
1/4 5.7 s 28 % 32.1 s 3.8 s 17 % 13.3 s 0.5 s 66 % 8.2 s
1/2 3.8 s 34 % 22.1 s 2.4 s 19 % 9.0 s 0.4 s 61 % 4.6 s
1 2.6 s 36 % 15.3 s 1.5 s 19 % 5.9 s 0.3 s 53 % 2.0 s
2 1.7 s 35 % 10.5 s 0.9 s 16 % 3.8 s 0.2 s 43 % 1.3 s
4 1.2 s 31 % 7.2 s 0.5 s 13 % 2.5 s 0.2 s 31 % 0.8 s
8 0.8 s 23 % 3.3 s 0.3 s 9% 1.5 s 0.1 s 22 % 0.8 s
16 0.5 s 15 % 2.5 s 0.2 s 6% 0.7 s 0.1 s 17 % 0.8 s
32 0.2 s 8% 1.8 s 0.1 s 6% 0.3 s 0.1 s 15 % 0.8 s

the first Ziegler-Nichols rule are still verified to a reasonable degree, as the frequency-response
plots show.
0.6187 Comparing these results with those obtained with IMC-
C1 (s) = 0.6021 + + 0.3105s1.0618 (47)
s1.3646 tuned fractional PIDs shows that rule-tuned fractional PIDs per-
1.6486 form nearly as well as those found with this analytical method.
C2 (s) = 1.4098 + 1.1011 − 0.2139s0.1855 (48)
s In this case, by letting TF = 21 , Eqn. (18) leads to
90.0000
CZN (s) = 18.0000 + + 0.9000s (49)
s
1 1 1 1
The step-responses obtained (together with open-loop Bode dia- CIMC = + + 1/2 + s (50)
4 s s 4
grams and sensitivity and closed-loop functions’ gains) are given
in Figs. 10, 11 and 12 and in Tab. 5. The PID performs poorly
because it tries to obtain a fast response and thus employs higher It is not clear which of the two integral terms is better to discard;
gains, but what is relevant here is that fractional PIDs still achieve simulations show it should be the fractional one. Step-responses
practically constant overshoots, since, in spite of the different obtained are shown in Fig. 13 and compare well with those of
plant structure, the conditions they were expected to verify are Fig. 10 and Fig. 11.

8 c 2005 by ASME
Copyright °
 1.4
80
0
1.2 60

gain / dB

gain / dB
40
1 20 −20
0
0.8 −40
output

−20
−2 −1 0 1 2 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10
0.6 ω / rad ⋅ s−1 ω / rad ⋅ s−1
0 0
0.4

gain / dB
−20

phase / º
K −500
0.2 −40
−60
0
0 10 20 30 40 50 −1000 −2 −1 0 1 2
−80 −2 −1 0 1 2
(a) time / s (b) 10 10 10 10 10 (c) 10 10 10 10 10

Figure 10. (a) STEP RESPONSE OF EQN. (46) CONTROLLED WITH EQN. (47) WHEN K IS 1/32, 1/16, 1/8, 1/4, 1/2, 1 (THICK LINE) AND 2.
(b) OPEN-LOOP BODE DIAGRAM WHEN K = 1. (c) SENSITIVITY FUNCTION AND CLOSED-LOOP FUNCTION GAINS WHEN K = 1.

1.4
80
0
1.2 60
gain / dB

gain / dB
40
1 20 −20
0
0.8 −20 −40
output

−2 −1 0 1 2 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10
0.6 ω / rad ⋅ s−1 ω / rad ⋅ s−1
0 0
0.4

gain / dB
−20
phase / º

K
−500 −40
0.2
−60
0
0 10 20 30 40 50 −1000 −2 −1 0 1 2
−80 −2 −1 0 1 2
(a) time / s (b) 10 10 10 10 10 (c) 10 10 10 10 10

Figure 11. (a) STEP RESPONSE OF EQN. (46) CONTROLLED WITH EQN. (47) WHEN K IS 1/32, 1/16, 1/8, 1/4, 1/2 AND 1 (THICK LINE). (b)
OPEN-LOOP BODE DIAGRAM WHEN K = 1. (c) SENSITIVITY FUNCTION AND CLOSED-LOOP FUNCTION GAINS WHEN K = 1.

1.4
80
0
1.2 60
gain / dB

gain / dB

40
1 20 −20
0
0.8 −40
output

−20
−2 −1 0 1 2 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10
0.6 ω / rad ⋅ s−1 ω / rad ⋅ s−1
0 0
0.4
gain / dB

−20
phase / º

0.2 −500 −40

−60
0
0 10 20 30 40 50 −1000 −2 −1 0 1 2
−80 −2 −1 0 1 2
(a) time / s (b) 10 10 10 10 10 (c) 10 10 10 10 10

Figure 12. (a) STEP RESPONSE OF EQN. (46) CONTROLLED WITH EQN. (49) WHEN K IS 1/32. (b) OPEN-LOOP BODE DIAGRAM WHEN K = 1.
(c) SENSITIVITY FUNCTION AND CLOSED-LOOP FUNCTION GAINS WHEN K = 1.

COMMENTS AND CONCLUSIONS quadratic (approximations of lower order being unsatisfactory).

In this paper two analytical methods for tuning the parame-
ters of fractional PIDs were reviewed. The optimisation method
was then used for developing tuning rules similar to those of the
first set of Ziegler-Nichols rules, making use of two parameters
(L and T ) of the unit-step response of the plant (which should be
S-shaped; otherwise rules 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
And the broader the application range of the rules is to be, the
more complicated they become (the first rule needs two tables of
parameters while the second, good for a narrower interval of val-
ues 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 fine-tuning (this is relevant
if the optimisation tuning method is used, since its results depend

9 c 2005 by ASME
Copyright °
 Table 5. DATA ON STEP-RESPONSES OF FIG. 10, FIG. 11 AND FIG 12.
Controller of Eqn. (47) Controller of Eqn. (48) Controller of Eqn. (49)
K rise time overshoot settling time rise time overshoot settling time rise time overshoot settling time
1/32 25.1 s 26 % 105.5 s 26.5 s 7% 86.6 s 0.6 s 31 % 3.1 s
1/16 15.8 s 27 % 65.8 s 14.7 s 7% 47.7 s — — —
1/8 10.3 s 28 % 41.2 s 8.2 s 8% 27.1 s — — —
1/4 6.9 s 29 % 26.1 s 4.6 s 9% 16.0 s — — —
1/2 4.4 s 27 % 17.2 s 2.4 s 9% 9.6 s — — —
1 2.7 s 23 % 11.9 s 1.1 s 8% 5.6 s — — —
2 1.5 s 17 % 8.6 s — — — — — —

1.4 FCT, Ministério da Ciência e Ensino Superior, Portugal, grant number

1.2 SFRH/BD/2875/2000, and ESF through the III Quadro Comunitário de
Apoio.
1

0.8
output

0.6
0.4
0.2
K CRC Press, Boca Raton, pp. 817–826.
0 [2] Podlubny, I., 1999. Fractional differential equations: an
0 10 20 30
time / s
Figure 13. STEP RESPONSE OF EQN. (46) CONTROLLED WITH IMC
WHEN K IS 1/32, 1/16, 1/8, 1/4, 1/2, 1 (THICK LINE) AND 2.
significantly 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, fine-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. These rules compare well with integer PIDs tuned
according to the first Ziegler-Nichols rule, even though the com-
parison is made difficult because Ziegler-Nichols rules achieve
different specifications for different values of T and L while rules
developed for fractional PIDs attempt to keep always a uniform
result. (It is of course likely that carefully tuned integer PIDs
perform better than rule-tuned fractional PIDs.)
It is surely possible to improve these tuning rules. Rules
similar to the second Ziegler-Nichols rule (making use of a
closed-loop response of the plant) are certainly possible. Rules
specific for non-minimum phase plants may also be of interest.
ACKNOWLEDGMENT
Duarte Valério was partially supported by programme POCTI,
REFERENCES
[1] Hägglund, T., and Åström, K., 1996. “Automatic tuning of
pid controllers”. In The control handbook, W. S. Levine, ed.
40 50
introduction to fractional derivatives, fractional differential
equations, to methods of their solution and some of their ap-
plications. Academic Press, San Diego.
[3] Caponetto, R., Fortuna, L., and Porto, D., 2002. “Parameter
tuning of a non integer order PID controller”. In Fifteenth
International Symposium on Mathematical Theory of Net-
works and Systems.
[4] Monje, C. A., Vinagre, B. M., Chen, Y. Q., Feliu, V.,
Lanusse, P., and Sabatier, J., 2004. “Proposals for fractional
PIλ Dµ tuning”. In Fractional Differentiation and its Applica-
tions.
[5] Chen, Y. Q., Moore, K. L., Vinagre, B. M., and Podlubny,
I., 2004. “Robust PID controller autotuning with a phase
shaper”. In Fractional Differentiation and its Applications.
[6] Miller, K. S., and Ross, B., 1993. An introduction to the frac-
tional calculus and fractional differential equations. John
Wiley and Sons, New York.
[7] Samko, S. G., Kilbas, A. A., and Marichev, O. I., 1993. Frac-
tional integrals and derivatives. Gordon and Breach, Yver-
don.
[8] Oustaloup, A., 1991. La commande CRONE : commande
robuste d’ordre non entier. Hermès, Paris.
[9] Vinagre, B. M., Podlubny, I., Hernández, A., and Feliu, V.,
2000. “Some approximations of fractional order operators
used in control theory and applications”. Fractional calculus
& applied analysis, 3, pp. 231–248.

10 c 2005 by ASME
Copyright °

View publication stats