10.

1515/acsc-2016-0001

Archives of Control Sciences

Volume 26(LXII), 2016
No. 1, pages 5–17

PID control of FOPDT plants with dominant dead time

based on the modulus optimum criterion

JAN CVEJN

The modulus optimum (MO) criterion can be used for analytical design of the PID con-
troller for linear systems with dominant dead time. However, although the method usually gives
fast and non-oscillating closed-loop responses, in the case of large dead time the stability mar-
gin gets reduced and even non-stable behavior can be observed. In this case a correction of the
settings is needed to preserve the stability margin. We describe and compare two methods of
design of the PID controller based on the MO criterion that for the stable first-order systems
with dead time preserve the stability margin, trying to keep maximum of the performance of the
original MO settings.
Key words: PID controller, process control, dead time, modulus optimum, magnitude op-
timum.

1. Introduction

The PID controller is a basic control instrument, which has been used in a majority
of control applications in industry [1]. Although this structure of a controller has been
widely utilized since the first half of 20th century, the problem of designing the PID
controller by means of its three tuning parameters is still alive and topical. This fact is
documented by the number of scientific papers on this problem, which appeared in the
last decades. This is true even in the case of control of linear time-invariant systems,
because the requirements on the closed-loop stability and robustness create complicated
constraints for the design objective [2].
This paper is focused on the problem of tuning the PID controller for the systems
with transfer function in Laplace transform
K
F (s) = e−τs (1)
Ts+1
where K is the system gain, T > 0 is the time constant and τ > 0 the dead-time param-
eter. τ is considered sufficiently large with respect to T , so the dead-time dynamics is
J. Cvejn is with Faculty of Electrical Engineering and Informatics, University of Pardubice, Studentska
95, 532 10, Pardubice, Czech Republic. E-mail: [email protected]
Received 22.04.2015. Revised 29.10.2015.

Unauthenticated
Download Date | 5/4/16 8:31 AM
6 J. CVEJN

important for the plant behavior. This simple plant is often being used for approximate
modeling of processes with non-oscillating high-order dynamics and allows simple ex-
perimental identification from the step response, which can be in most cases easily mea-
sured. Simple methods based on coincidence in one or more points and more complex
methods suitable for noisy data are described e.g. in [1] and [3].
For tuning the PID controller based on the model (1) many approaches exist, see e.g.
[1] for a description of the most important methods. A comprehensive survey of known
formulas is available in [4]. Early methods were derived from empirical requirements on
the step response, such as one-quarter decay ratio [5], [6], step-response overshoot [7] or
from integral criterions in time domain with approximation of the dead-time dynamics
[8]. These methods, however, usually work well only for a rather limited range of the
ratio τ/T .
The design based on the well-known modulus optimum (MO) criterion [9], [1], [10],
[14] is one of the approaches that allow working even with long τ with respect to T . This
design criterion requires that the closed loop frequency response modulus is as flat as
possible in the range of low frequencies, i.e.
L(iω) dk L(iω)
lim = 1, lim = 0, k = 1, 2, . . . , km (2)
ω→0 1 + L(iω) ω→0 dωk 1 + L(iω)

where L(iω) denotes the open-loop frequency response, i.e. the product of the responses
of the plant and the controller, and km is as high as possible. The MO-based design
is most natural for the reference tracking control tasks, where the closed-loop system
is to be able to respond quickly to changes of the reference input, or equivalently to
efficiently reject the disturbances influencing directly the plant output. In this case it
usually produces fast non-oscillating responses [10], [14].
The PID controller can be for the plant (1) designed in a simplified way if the factor
(T s + 1) in (1) is compensated by the controller zero. Then only the two remaining
control loop parameters are to be determined with respect to the MO criterion. This
approach, utilized in [13], has several important advantages: it preserves a sufficient
stability margin, gives simple tuning rules and it enables to propose a compensation of
the settings in the cases when the disturbance influences the plant input. The resulting
MO-based tuning rules
3 T τ
Kc = , TsI = T, Tsd = (3)
4 Kτ 3
where the PID controller is considered in the serial form
( )
1
R(s) = Kc 1 + (1 + Tsd s) (4)
TsI s
offer fast and well damped responses even for large τ/T , unlike most other tuning meth-
ods. For the parallel PID controller
( )
1
R(s) = Kc 1 + + Td s (5)
TI s

Unauthenticated
Download Date | 5/4/16 8:31 AM
PID CONTROL OF FOPDT PLANTS WITH DOMINANT DEAD TIME
BASED ON THE MODULUS OPTIMUM CRITERION 7

the settings corresponding to (3) can be rewritten as

( )
1 3T τ τ
Kc = 1+ , TI = T + , Td = . (6)
4K τ 3 3 + τ/T

In this paper the design of the PID controller for the plant (1) is considered in full
form, where all the three controller parameters are determined with respect to the MO
criterion and no pole/zero compensation is used. In comparison with the simplified ap-
proach mentioned above it can be assumed that the performance is enhanced in the sense
that the closed-loop magnitude response is more flat in the origin, which means that the
bandwidth is increased. On the other hand, we show that if the ratio τ/T is sufficiently
large, even non-stable closed loop can be obtained in this way. Therefore, a modification
of the settings is needed to preserve the stability margin. This problem has been treated
already in [12], where a simple correction of the MO settings was proposed.
In [11] the problem of the MO design is analyzed for the more general plant
K
F(s) = e−τs (7)
(T1 s + 1) (T2 s + 1) · · · (Tn s + 1)

where T1 > T2 > · · · > Tn > 0 and τ > 0. It was observed that the MO-optimal controller
parameters are positive and the corresponding open-loop Nyquist plot for the plants (7)
lies in the half-plane {z | Re z > −0.5} if τ is not too large. This property ensures a large
stability margin and good robustness of the MO settings, although in the cases of long
dead time a correction of the MO settings is needed. The correction of the settings pro-
posed in [11] is designed to preserve maximum of the performance, which in a certain
sense corresponds to the flatness of the closed-loop magnitude response for low frequen-
cies. However, the algorithm of this correction is iterative, although the computation is
simple and very efficient.
Based on the results in [11], in this paper we propose a modification of the MO-
based PID controller settings for the FOPDT plant (1), which guarantee stability and the
stability margin specified above for any τ > 0 and are fully explicit, i.e. can be obtained
without any iterative computations.

2. The MO tuning of PID controller

If we rewrite the transfer function (7) as

K
F(s) = e−τs (8)
An sn + · · · + A 1s + 1

the corresponding MO-optimal PID controller (5) settings for the system (7) are in [11]
obtained in the form
r0 r0 τ r1 τ
Kc = , TI = , Td = (9)
K r−1 r0

Unauthenticated
Download Date | 5/4/16 8:31 AM
8 J. CVEJN

where the parameters ri , i = −1, 0, 1 are obtained as the solution to the following system
of linear equations:
 
1 a
 ∑ j!
1− j
−1 0    
 j=0  r−1 1
 3 
 a3− j 2 a 1 a
j   1 
 ∑ j! − ∑ j! ∑ 1−j!   r0  =  2a2 − a21
2− j
 (10)
 j=0 j=0 j=0  2
 5  r1 2a4 + a2 − 2a1 a3
2
 a j 4 a 3 a
3− j 
∑ 5− j! − ∑ 4− j
j! ∑ j!
j=0 j=0 j=0

where
Aj
aj =, a0 = 1. (11)
τj
For the plant (1) a1 = T /τ and a2 = a3 = · · · = 0 in (10), so we obtain
    
η+1 −1 0 r−1 1
 η 1 ( )   1  
 2 + 6 − η + 12 η + 1   r0  =  −η2  (12)
η ( η ) η
2
4! + 5! − 6 + 4!
1 1 1 r1
2+6 0

where η = T /τ, which can be rewritten as follows:

    
η+1 −1 0 r−1 0.5
    
 3η + 1 −3 (2η + 1) 6 (η + 1)   r0  =  −3η2  . (13)
5η + 1 −5 (4η + 1) 20 (3η + 1) r1 0

The solution to (13) can be expressed in the explicit form:

1( )
r0 = 180η4 + 240η3 + 135η2 + 42η + 7 (14)
D
1( )
r1 = 60η4 + 60η3 + 27η2 + 7η + 1 (15)
D
15 ( )
r−1 = 12η3 + 12η2 + 5η + 1 (16)
D
where ( )
D = 16 15η3 + 15η2 + 6η + 1 . (17)
Note that the settings exist for any η > 0. The actual PID controller parameters are
obtained by substitution into (9).
Fig. 1 shows the open-loop Nyquist plots for the plants (1), for three values of T
and τ = 1. Fig. 2 shows the corresponding closed-loop step responses. In the simulations
there is considered the 1st-order low-pass filter with time constant T f = 0.02 [sec] by

Unauthenticated
Download Date | 5/4/16 8:31 AM
PID CONTROL OF FOPDT PLANTS WITH DOMINANT DEAD TIME
BASED ON THE MODULUS OPTIMUM CRITERION 9

the derivative term of the controller (not considered at the controller design stage). This
means, the actual PID controller transfer function is
( )
1 Td s
R(s) = Kc 1 + + . (18)
TI s 0.02s + 1

The same modification of the controller, corresponding to a real situation, was used for
obtaining the other simulated results in this paper as well (Figs 4 and 6).
It can be seen from Fig. 1 that the optimal open-loop Nyquist plot is flat and curving
towards the origin for low frequencies. Note that the performance gets enhanced with
decreasing T in the sense that the closed-loop magnitude response is more flat in the
origin. On the other hand, for sufficiently low T the open-loop magnitude increases for
higher frequencies, which has ill effect on the stability margin, although it is true that
this phenomenon can be observed only for rather large values of τ/T . In the time domain
this problem corresponds to undesirable oscillations in the step response

Figure 1. The MO-optimal open-loop Nyquist plots for plants (1) where τ = 1 [sec] and T = 1 [sec] (solid
line), T = 0.2 [sec] (dashed line) and T = 0.05 [sec] (dotted line).

Unauthenticated
Download Date | 5/4/16 8:31 AM
10 J. CVEJN

Figure 2. The MO-optimal step responses for plants (1) where τ = 1 [sec] and T = 1 [sec] (solid line),
T = 0.2 [sec] (dashed line) and T = 0.05 [sec] (dotted line).

3. The correction of the MO settings

3.1. The simplified approach

If we use (9), the open-loop transfer function can be written as
( )
1 K
R (s) F (s) = Kc 1 + + Td s e−τs = (19)
TI s Ts+1
( ) ( ) −τs
r0 r−1 1 r1 τ K −τs 1 e
= 1+ + s e = r0 + r−1 + r1 τs .
K r0 τ s r0 Ts+1 τs Ts+1
If we put ξ = τω, the open-loop frequency response, corresponding to (19), can be writ-
ten in the form [ ( )] −iξ
r−1 e
L (ξ) = r0 + i r1 ξ − (20)
ξ iηξ + 1
where η = T /τ and ξ has the meaning of dimensionless frequency. Note that regardless
the used transformation of the frequency the Nyquist plots preserve their shape. In [11]
it is shown that MO optimality for the more general class of plants (7) implies that
1 + 2 Re L (ξ) → 0 for ξ → 0. (21)
This fact can be also seen in Fig. 1. Moreover, it was shown in [11] that the parame-
ters rk are positive. Consequently, the open-loop Nyquist plot comes out from the point
(−0.5, −∞) and it tends towards the right half-plane. It can be also seen in Fig. 1 that
for the plant (1) Re L (ξ) is always increasing for ξ 6 ξu , where ξu is the frequency when
∠L (ξ) = −π. Therefore
L (ξ) ∈ {z | Re z > −0.5} (22)

Unauthenticated
Download Date | 5/4/16 8:31 AM
PID CONTROL OF FOPDT PLANTS WITH DOMINANT DEAD TIME
BASED ON THE MODULUS OPTIMUM CRITERION 11

holds if |L (ξ)| 6 0.5 for all ξ > ξu . In this case the closed-loop system is stable by the
Nyquist criterion. However, if τ/T is sufficiently large, this condition is not fulfilled. In
such cases |L (ξ)| is increasing for ξ > ξ† , where ξ† is some frequency larger than ξu .
Since from (20) it follows that
√
r02 ξ2 + r12 ξ4 + r−1
2 − 2r r ξ2
1 −1 r1
lim |L (ξ)| = lim = (23)
ξ→∞ ξ→∞ ξ (1 + η ξ )
2 2 2 η

it is possible to conclude that the correction of the settings is needed if r1 /η > 0.5. This
observation has been used in [12], where the modified MO settings was proposed in the
simple form
r1 = min {r1∗ , 0.5η} (24)
where r1∗ denotes the MO-optimal value of r1 obtained by solving (13). The remaining
coefficients r0 , r−1 are obtained by solving the system of the first two equations in (13)
for fixed r1 :
( )( ) ( )
η+1 −1 r−1 0.5
= (25)
3η + 1 −3 (2η + 1) r0 −3η2 − 6r1 (η + 1)

i.e.
[ ] [ ]
r−1 = −1.5 (2η + 1) − 3η2 − 6r1 (η + 1) /D1 = −3 η2 + η + 0.5 + 2r1 (η + 1) /D1
[ ( ) ]
r0 = (η + 1) −3η2 − 6r1 (η + 1) − 0.5 (3η + 1) /D1 =
[ ]
1 1 2
= −3 η + η + η + + 2r1 (η + 1) /D1
3 2
(26)
2 6

where

D1 = −3 (η + 1) (2η + 1) + 3η + 1 = −6η2 − 6η − 2. (27)

The value of η, where just r1 /η = 0.5, can be obtained by substitution into (15). The only
real positive solution is ηd ≈ 0.1613, which means that the correction of the settings is
applied in the cases of τ > 6.2T , which will indeed occur rarely in practice.
An advantage of this approach is its simplicity. On the other hand, the requirement
that |L (ξ)| 6 0.5 for ξ > ξu , which ensures (22), does not always give responses favor-
able from practical point of view. It can be seen in Fig. 4 that the obtained responses
can be rather oscillatory, unlike the responses when r1 /η < 0.5, which are usually well
damped. This problem is caused by the open-loop resonance - note in Fig. 3 that even
after the correction the magnitude response is not monotonic. This also means that the
correction of the settings may be meaningful even for lower values of τ/T .

Unauthenticated
Download Date | 5/4/16 8:31 AM
12 J. CVEJN

Figure 3. The open-loop Nyquist plots corresponding to the MO-optimal settings with simplified correction,
where τ = 1 [sec] and T = 1 [sec] (solid line), T = 0.2 [sec] (dashed line) and T = 0.05 [sec] (dotted line).

Figure 4. The step responses corresponding to the MO-optimal settings with simplified correction for plants
(1), where τ = 1 [sec] and T = 1 [sec] (solid line), T = 0.2 [sec] (dashed line) and T = 0.05 [sec] (dotted
line).

3.2. The enhanced method

In [11] the correction of the MO settings for the class of systems (7) was proposed
to keep |L (ξ)| monotonic. The decreasing trend of |L (ξ)| ensures (22) for ξ > ξu and

Unauthenticated
Download Date | 5/4/16 8:31 AM
PID CONTROL OF FOPDT PLANTS WITH DOMINANT DEAD TIME
BASED ON THE MODULUS OPTIMUM CRITERION 13

is natural with respect to practical requirements of well-damped closed-loop response.

For the systems (7) the correction is unfortunately more complex and requires an itera-
tive computation. However, we show that in the special case (1) it can be obtained in a
simplified way. We can write the open-loop magnitude response as
( )
2 r02 + (r1 ξ − r−1 /ξ)2 r12 ξ2 + r02 − 2r1 r−1 2
r−1
|L(ξ)| = = + =
η2 ξ2 + 1 η2 ξ2 + 1 ξ2 (η2 ξ2 + 1)
(28)
( 2 ) λξ2 + 1 2
r−1
= r0 − 2r1 r−1 2 2 +
η ξ + 1 ξ2 (η2 ξ2 + 1)
where
r12 [ ]−1
2
λ= = (r0 /r 1 ) − 2 (r−1 /r 1 ) . (29)
r02 − 2r1 r−1
It can be seen that it must be r02 − 2r1 r−1 > 0 and consequently also λ > 0 in the MO
configuration, because otherwise the rate of decrease of |L(ξ)|2 for ξ → 0 would be larger
than for the I-controller with r0 = r1 = 0, which would mean that the settings r0 = r1 = 0
was MO-optimal. The second term in (28) is always decreasing, while the first term in
(28) is increasing if λ > η2 and decreasing if λ < η2 . Note that if r1 → 0, the magnitude
response is always decreasing, because in this case

r02 + (r−1 /ξ)2 r02 2

r−1
|L(ξ)|2 = = + . (30)
η2 ξ2 + 1 η2 ξ2 + 1 ξ2 (η2 ξ2 + 1)
It is desirable to keep λ as large as possible to ensure maximal possible bandwidth. This
suggests the following modification of the MO settings:
If λ > η2 , put λ = η2 and determine the corresponding r1 . The parameters r0 , r−1 are
obtained using (26). In the other cases the MO settings (14)-(16) are left unchanged.
The remaining problem is the determination of r1 for λ = η2 . If we define ρ = r1−1 , from
equations (26) and (27) we can obtain
r−1 r0 ( )
= γ (0.5c2 ρ + c1 ) , = γ 0.5c3 ρ + c21 (31)
r1 r1
where

c1 = 1 + η
1
c2 = + η + η2 (32)
2
1 1
c3 = + η + η2 + η3
6 2
and ( )−1
1
γ = (c1 c2 − c3 )−1 = + η + η2 . (33)
3

Unauthenticated
Download Date | 5/4/16 8:31 AM
14 J. CVEJN

Substituting into (29) yields

( )2
λ−1 (ρ) = γ2 0.5c3 ρ + c21 − γ (c2 ρ + 2c1 ) (34)

which for given λ−1 = d is a quadratic equation for obtaining ρ, which can be rearranged
as follows: ( ) ( )
0.25γ2 c23 ρ2 + γ2 c21 c3 − γc2 ρ + γ2 c41 − 2γc1 − d = 0. (35)
The solution to (35) is obtained in the form
√( )2 ( )
γc2 − γ2 c21 c3 ± γc2 − γ2 c21 c3 − γ2 c23 γ2 c41 − 2γc1 − d
ρ = =
0.5γ2 c23
√ ( )
c2 − γc21 c3 ± c22 + 2γ c1 c23 − c21 c2 c3 + c23 d
= (36)
0.5γc23
√
c2 − γc21 c3 ± c22 − 2c1 c3 + c23 d
= .
0.5γc23
Further, we can write
( )2 ( )
−1 r0 r−1 ( )
λ (ρ) = −2 = ρ r02 ρ − 2r−1 . (37)
r1 r1

Since λ−1 → ∞ for ρ → ∞, λ−1 (ρ) in (37) must be increasing in all the interval [ρ2 , ∞),
where ρ2 = 2r−1 /r02 > 0, and it is the only interval where λ−1 (ρ) > 0, because ρ > 0.
This means, for any given d > λ−1M > 0, where λM denotes the value of λ corresponding
to the MO-optimal settings, the equation (35) must have just one real positive solution.
Moreover, due to increasing trend of λ−1 (ρ), to larger d there must correspond a larger
value of ρ. Consequently, we can consider only ‘+’ in place of ‘±’ in (36).
The correction is to be applied only if η < ηmin , where ηmin is such that

(r0 /r1 )2 − 2 (r−1 /r1 ) = η−2

min . (38)

Substituting (14)-(17) into (38) gives ηmin as the real solution to the equation
( )2
180η5 + 240η4 + 135η3 + 42η2 + 7η 12η5 + 12η4 + 5η3 + η2
− 30 = 1.
60η4 + 60η3 + 27η2 + 7η + 1 60η4 + 60η3 + 27η2 + 7η + 1
(39)
The equation (39) can be easily rearranged into an algebraic equation, which has only
one positive real solution ηmin ≈ 0.2915. Therefore, if η < ηmin , the value of r1 is ob-
tained from
0.5γc23
r1 = √ (40)
2
c2 − γc1 c3 + c2 − 2c1 c3 + (c3 /η)
2 2

Unauthenticated
Download Date | 5/4/16 8:31 AM
PID CONTROL OF FOPDT PLANTS WITH DOMINANT DEAD TIME
BASED ON THE MODULUS OPTIMUM CRITERION 15

where ck and γ are given by (32) and (33). The remaining parameters r0 , r−1 then can be
computed from (26), or using the expressions (31).
Fig. 5 shows the Nyquist plots corresponding to the same plants like in Fig. 1, but
using this type of correction of the MO settings. Fig. 6 shows the corresponding step
responses, which are more damped for low η than the responses in Fig. 4.

Figure 5. The open-loop Nyquist plots corresponding to the MO-optimal settings with enhanced correction,
where τ = 1 [sec] and T = 1 [sec] (solid line), T = 0.2 [sec] (dashed line) and T = 0.05 [sec] (dotted line).

It is clear from Fig. 6 that the final settings are only slightly dependent on T and are
similar to the settings for T → 0 (represented approximately by the case T = 0.005 [sec]
in Fig.6). The fastest response with overshoot of about 10% was obtained for T ≈ 0.3τ,
for lower values of T /τ the response gets a bit slower due to the correction.
It can be easily seen from (40) or directly from (28) that for T = 0 the proposed
correction produces the settings such that r1 = 0, which means that the corresponding
controller is of type PI. Therefore, the closed-loop behavior for the proposed settings
will be similar to the behavior in the case of the serial-type PID controller (4) where
Tsd = T is chosen to compensate the process lag and the remaining parameters Kc and
TsI are designed with respect to the MO criterion. The corresponding settings then are
given by the simple tuning formulas (6) proposed in [13]. Therefore, it seems that for
practical purposes the settings (14)-(17) with the enhanced correction can be replaced
by the simple design rules (6). This also shows a rather surprising fact that that although
these rules were developed only as sub-optimal with respect to the MO criterion, they
produce settings very close to the optimal settings if the stability margin requirements
are included into the design objective.

Unauthenticated
Download Date | 5/4/16 8:31 AM
16 J. CVEJN

Figure 6. The step responses corresponding to the MO-optimal settings with enhanced correction for plants
(1), where τ = 1 [sec] and T = 1 [sec] (solid line), T = 0.2 [sec] (dashed line) and T = 0.05 [sec] (dotted
line).

4. Conclusions

The modulus optimum criterion enables to design the PID controller for the FOPDT
plant (1) analytically for any value of dead time τ > 0. For larger τ/T , however, the
stability margin gets reduced and even non-stable closed-loop behavior can be obtained.
This problem can be resolved be a suitable modification of the settings, which keeps the
open-loop Nyquist plot in the half-plane {z | Re z > −0.5}. After this correction a suffi-
cient stability margin and good robustness of the settings are guaranteed. We presented
two methods of the correction, which try to preserve maximum of the performance of
the original MO settings. For practical purposes especially the correction based on the
requirement of monotonic open-loop magnitude seems to be advantageous. It was shown
that for the FOPDT plant (1) this correction can be obtained in fully explicit form, which
does not use any iterative computations, unlike the more general case (7), discussed in
[11]. It was observed that the resulting settings with the correction give responses similar
to the simple sub-optimal tuning rules (6) proposed in [13], which could be preferable
for practical purposes.

References

[1] K.J. Å STRÖM and T. H ÄGGLUND: PID Controllers: Theory, Design, and Tuning.
2nd edition. Instrument Society of America. 1995.

Unauthenticated
Download Date | 5/4/16 8:31 AM
PID CONTROL OF FOPDT PLANTS WITH DOMINANT DEAD TIME
BASED ON THE MODULUS OPTIMUM CRITERION 17

[2] K.J. Å STRÖM , H. PANAGOPOULOS and T. H ÄGGLUND: Design of PI controllers

based on non-convex optimization. Automatica, 34 (1998), 585-601.

[3] T.K. K IONG , W. Q UING -G UO , H.C H . C HIECH and T.J. H ÄGGLUND: Advances
in PID Control. Springer Verlag. 1999.

[4] A. O’DWYER: Handbook of PI and PID Tuning Rules. Imperial College Press.
2003.

[5] J.G. Z IEGLER and N.B. N ICHOLS: Optimum settings for automatic controllers.
Trans. of the ASME, 64 (1942), 759-768.

[6] G.H. C OHEN and G.A. C OON: Theoretical consideration of retarded control.
Trans. of the ASME, 75 (1953), 827-834.

[7] K.L. C HIEN , J.A. H RONES and J.B. R ESWICK: On the automatic control of gen-
eralized passive systems. Trans. of the ASME, 74, (1952), 175-185.

[8] A.M. L OPEZ , P.W. M URRILL and C.L. S MITH: Controller tuning relationships
based on integral performance criteria. Instrumentation Technology, 14(11), 1967.

[9] R.C. O LDENBOURG and H. S ARTORIUS: A uniform approach to the optimum

adjustments of control loops. Trans. of the ASME, 76 (1954), 1265-1279.

[10] D. V RANCI Ć , Y. P ENG and S. S TRMCNIK: A new PID controller tuning method
based on multiple integrations. Control Engineering Practice, 7 (1999), 623-633.

[11] J. C VEJN: The design of PID controller for non-oscillating time-delayed plants
with guaranteed stability margin based on the modulus optimum criterion. J. of
Process Control, 23 (2013), 570-584.

[12] J. C VEJN: PID control of FOPDT plants with dominant dead time - The frequency
domain approach. Proc. of the 2013 Int. Conf. on Process Control, Strbske Pleso,
Slovakia, (2013), 53-56.

[13] J. C VEJN: Sub-optimal PID controller settings for FOPDT systems with long dead
time. J. of Process Control, 19 (2009), 1486-1495.

[14] D. V RANCI Ć , S. S TRMCNIK and D. J URICI Ć: A magnitude optimum multiple

integration tuning method for filtered PID controller. Automatica, 37 (2001), 1473-
1479.

Unauthenticated
Download Date | 5/4/16 8:31 AM