applied

sciences
Article
Discrete-Time First-Order Plus Dead-Time
Model-Reference Trade-off PID Control Design
Ryo Kurokawa 1 , Takao Sato 1, * , Ramon Vilanova 2 and Yasuo Konishi 1
1 Department of Mechanical Engineering, Graduate School of Engineering, University of Hyogo,
Hyogo 671-2280, Japan
2 Department of Telecommunications and Systems Engineering, Universitat Autònoma de Barcelona,
08193 Barcelona, Spain
* Correspondence: [email protected]; Tel.: +81-792674983




Received: 19 July 2019 ; Accepted: 5 August 2019; Published: 7 August 2019


Abstract: The present study proposes a novel proportional-integral-derivative (PID) control design
method in discrete time. In the proposed method, a PID controller is designed for first-order plus
dead-time (FOPDT) systems so that the prescribed robust stability is accomplished. Furthermore,
based on the control performance, the relationship between the servo performance and the regulator
performance is a trade-off relationship, and hence, these items are not simultaneously optimized.
Therefore, the proposed method provides an optimal design method of the PID parameters for
optimizing the reference tracking and disturbance rejection performances, respectively. Even though
such a trade-off design method is being actively researched for continuous time, few studies have
examined such a method for discrete time. In conventional discrete time methods, the robust stability
is not directly prescribed or available systems are restricted to systems for which the dead-time in
the continuous time model is an integer multiple of the sampling interval. On the other hand, in the
proposed method, even when a discrete time zero is included in the controlled plant, the optimal PID
parameters are obtained. In the present study, as well as the other plant parameters, a zero in the
FOPDT system is newly normalized, and then, a universal design method is obtained for the FOPDT
system with the zero. Finally, the effectiveness of the proposed method is demonstrated through
numerical examples.

Keywords: PID control; model-based design; discrete time system; sensitivity function;
robust stability; FOPDT system

1. Introduction
Proportional-integral-derivative (PID) [1–9] control has few tuning parameters: proportional gain,
integral time, and derivative time, and its structure is simple. Hence, PID control has been widely
used in industry, and numerous tuning methods have been proposed.
In the model-based approach, optimal tracking and robust stability are achieved. However,
the control system must be redesigned whenever the controlled system is changed. As a simple
model-based optimal design method, Ziegler and Nichols proposed the step response method (ZN
method) [10]. Using the ZN method, the PID parameters are decided based on the step response
trajectory so that the tracking performance is optimized. However, the robust stability is not taken
into account [11,12]. Therefore, robust control designs have been proposed [13,14]. Using robust
design methods, robust stability is obtained, but the tracking performance is not optimized. In [15],
H2 optimal design was proposed using internal model control (IMC) design, and hence, stability was
assured. Although robust PID control systems are also designed in the discrete time domain [16,17],
the stability margin cannot be prescribed.

Appl. Sci. 2019, 9, 3220; doi:10.3390/app9163220 www.mdpi.com/journal/applsci

Appl. Sci. 2019, 9, 3220 2 of 19

The relationship between the tracking performance and the robust stability is a trade-off
relationship [18]. Therefore, trade-off PID design methods have been proposed [19,20]. In the trade-off
design methods, the PID parameters are decided such that the tracking performance is optimized
subject to the prescribed robust stability, where servo or regulator mode is selected for performance
optimization. A trade-off optimization approach has been designed in the continuous time domain
that consists of a first-order plus dead-time (FOPDT) system [21,22], a second-order plus dead-time
(SOPDT) system [23,24], and a two degrees-of-freedom (2DOF) system [25,26].
For a discrete time system, trade-off approaches have been proposed [27,28]. Discrete-time
methods are useful for controlling discrete time systems, in which the controller is implemented
with a digital computer. However, the robust stability was not assigned in [27], and the controlled
plant was restricted to a non-zero system [28]. Therefore, in the present study, a new discrete time
domain approach is proposed, in which the constraint is relaxed and the prescribed robust stability is
accomplished. Specifically, the FOPDT model with a zero is normalized, and a universal design method
is obtained. As a result, the trade-off design strategy is available for a large class of discrete time systems.
Finally, the effectiveness of the proposed method is demonstrated through numerical examples.

2. Control System
The continuous time FOPDT transfer function is described as follows:
K
P(s) = e− Ls (1)
Ts + 1

where K is the gain, T is the time constant, and L is the dead time. In the present study, the discrete
time control model is designed with sampling interval Ts , and hence, the continuous time model is
expressed as follows:

b0 + b1 z−1 −(d+1)
Pd (z−1 ) = z (2)
1 − a 1 z −1

where a1 , b0 , and b1 are the coefficient parameters, d is the dead-time in the discrete time, and z−1
denotes the backward shift operator. The discrete time plant parameters correspond to the continuous
time plant parameters as follows:
Ts
a1 = e − T (3)
 
L0
b0 = K 1 − a1 e T (4)
 
L0
b1 = K a1 e T − a1 (5)

d = L1 (6)

where a non-negative integer L1 satisfies the following equation:

L = L1 Ts + L0 ( L0 < Ts )

Therefore, the input/output relationship in discrete time is given as follows:

y(k) = Pd (z−1 )u(k) (7)

where y(k ) is the system output (plant output) and u(k) is the control input.
Appl. Sci. 2019, 9, 3220 3 of 19

The present study proposes a new trade-off design method for the following PID control law:

u(k) z−1 )e(k) − Cy (oz−1 )y(k )

= Ce (n
Ts
Ce (z−1 ) = K p 1 + T (1− zo−1 )
i
n
Td (1−z−1 )
(8)
− 1
Cy (z ) = K p Ts
e(k) = r (k) − y(k)

where r (k ) is the reference input and K P , Ti , and Td are the proportional gain, the integral time, and
the derivative time, respectively. A block diagram of the control system is shown in Figure 1, in which
the control input is disturbed by disturbance d(k ). Equation (8) is the discrete version of the next
continuous time control law:
n  o
U (s) = K p 1 + T1s E(s) − Td sY (s)
i (9)
E(s) = R(s) − Y (s)

where L[·] means the Laplace transform, U (s) = L[u(t)], Y (s) = L[y(t)], and R(s) = L[r (t)].

Figure 1. Discrete-time PID control system.

The closed-loop systems from r (k) and d(k) to u(k) and y(k), respectively, are obtained as follows:

Ce (z−1 ) Cd (z−1 ) Pd (z−1 )

u(k) = − −
r (k) − d(k) (10)
1
1 + Cd (z ) Pd (z )1 1 + Cd (z−1 ) Pd (z−1 )

Ce (z−1 ) Pd (z−1 ) P ( z −1 )
y(k) = 1+Cd (z−1 ) Pd (z−1 )
r (k) + 1+C (zd−1 ) P (z−1 ) d(k )
d d (11)
Cd (z−1 ) = Ce (z−1 ) + Cy (z−1 )
From the above equations, the relationship between the reference tracking and the disturbance
rejection is a trade-off relationship.

3. Design Objective
In the proposed method, the PID parameters of the discrete time control law are designed based on
a constrained optimization problem. The constrained optimization problem consists of the constraint
condition as the prescribed stability margin and the objective function defined by the index with respect
to the servo and regulator performances, respectively. The constraint condition and the objective
function are defined in Sections 3.1 and 3.2, respectively.
Appl. Sci. 2019, 9, 3220 4 of 19

3.1. Constraint Condition

The constraint condition is used to obtain the stability margin. In the proposed method, the
stability margin is designed using the following sensitivity function:

1
S f ( z −1 ) = (12)
1 + Cd (z−1 ) Pd (z−1 )

Using the sensitivity function, the constraint condition is defined as follows:

| Ms − Msd | = 0 (13)
− jω
Ms = max |S f (e )| (14)
ω

where Ms denotes the maximum value of the sensitivity function S f (e− jω ) and Msd is the designed
value of Ms .
The design range of Msd is recommended to be from 1.4–2.0 [4]. The relationship between Msd and
the stability margin is in inverse proportion.

3.2. Objective Function

The objective function is defined based on the sum of absolute errors (SAE) as follows:
∞
Jd = ∑ |r(k) − y(k)| (15)
k =1

In a one-degree-of-freedom (1DOF) system, the relationship between the servo and

regulation performances is a trade-off relationship. Therefore, the objective functions in
the reference tracking optimization and the disturbance rejection optimization are given by
Equations (16) and (17), respectively:

∞
Ce (z−1 ) Pd (z−1 )
 
Jds = ∑
 s 
 r ( k ) − r ( k )  (16)
k =1
1 + Cd (z−1 ) Pd (z−1 ) 
∞ 
Pd (z−1 )

Jdr = ∑ r d (k) −

− 1 − 1
d ( k )  (17)
k =1
1 + Cd (z ) Pd (z ) 

where r s (k) = 1, r (k) = 1, r d (k) = 0, and d(k) = 1.

In the discussed control system, the stability and the tracking performance are simultaneously
decided because they are in a trade-off relationship.

4. Proportional-Integral-Derivative Parameter Design

In order to obtain a universal design method for arbitrary FOPDT models, first, in Section 4.1,
the plant model and the PID control law are normalized. Second, in Section 4.2, optimal PID parameters
are obtained subject to the pre-established stability margin. Third, in Section 4.3, the parameter tuning
rule is derived based on the optimal parameters calculated in Section 4.2. Finally, the algorithm for the
proposed design method is summarized in Section 4.4.
Appl. Sci. 2019, 9, 3220 5 of 19

4.1. Normalization
The plant model and the control law are normalized to obtain a universal design method.
The normalization parameters for the plant model, τ0 and τa , are given as follows:
 
b0 a1 + b1
τ0 = −d log a1 + log (18)
a1 (b0 + b1 )
τa = − log a1 (19)

Moreover, the normalization parameters for the control law, κ p , τi , and τd , are given as follows:

b0 + b1
κp = Kp (20)
1 − a1
T log a1
τi = − i (21)
Ts
T log a1
τd = − d (22)
Ts

The derivation of the normalization parameters is shown in Appendix A.

The expression of the plant model using the normalization parameters has merits for optimization
analysis and controller design. Since the optimization problem is stated in terms of just two
parameters τ0 and τa , the analysis and elaboration of coefficient parameters are simplified. Furthermore,
the controller parameters are tuned independent of the process gain and time constant.

4.2. Optimization
The controller parameters are optimized for the servo and regulation operations, respectively,
subject to the established stability margin.
As the optimization tool, the fmincon function in MathWorks MATLAB software was used, where
the prescribed robust stability is set: Msd ∈ {1.4, 1.6, 1.8, 2.0}, and the optimal controller parameters
are calculated based on the normalization parameters for the controlled plant: τ0 ∈ {0.3, · · · , 1.7} and
τa ∈ {0.01, · · · , 0.1}. As an example, the optimized controller parameters for Msd = 1.4 in the servo are
shown in Figure 2.

2 1.4

1 1.2
κp

i
τ

0 1
0.1 0.1
1.5 1.5
0.05 1 0.05 1
τa 0.5 τa 0.5
τ0 τ0

0.5
τd

0
0.1
1.5
0.05 1
τa 0.5
τ0

Figure 2. Relationships among optimized controller parameters and the normalized parameters τ0 and
τa in servo design.
Appl. Sci. 2019, 9, 3220 6 of 19

4.3. Controller Parameter Decision

The tuning rule for the controller parameters is proposed in Section 4.3.1, and the prescribed
robust stability using the proposed parameter tuning rule is evaluated in Section 4.3.2.

4.3.1. Tuning Rule

On the pair of plant parameters τ0 and τa with the established robust stability Msd , the controller
parameters are expressed in terms of the normalization parameters for the plant model as:

κ p = α0 + α1 τ0α2 (23)
τi = β 0 + β 1 τ0 + β 2 τ02 + β 3 τ03 (24)
τd = γ0 + γ1 τ0 + γ2 τ02 (25)

where αi (i = 0, 1, 2), β j ( j = 0, 1, 2, 3), and γk (k = 0, 1, 2) are defined as follows:

α0 = a00 + a01 τa (26)

α1 = a10 + a11 τa (27)
α2 = a20 + a21 τa (28)
β 0 = b00 + b01 τa (29)
β 1 = b10 + b11 τa (30)
β 2 = b20 + b21 τa (31)
β 3 = b30 + b31 τa (32)
γ0 = c00 + c01 τa (33)
γ1 = c10 + c11 τa (34)
γ2 = c20 + c21 τa (35)

where the coefficient parameters in Equation (26)–Equation (35) in the servo optimization are shown
in Table 1, and those in the regulation optimization are also shown in Table 2.

Table 1. Coefficient parameters in Equations (26)–(35) in servo design.

Msd 1.4 1.6 1.8 2.0

a00 0.2130 0.2778 0.3281 0.3098
a01 −0.4643 −0.6376 −0.8185 −0.7722
a10 0.4361 0.5803 0.6932 0.8100
a11 −0.3767 −0.4236 −0.3308 −0.4577
a20 −1.0067 −1.0169 −1.0150 −0.9861
a21 1.7509 1.7951 1.9003 1.8503
b00 1.1368 1.1451 1.2097 1.3995
b01 −1.6140 −1.1310 −0.7911 −1.9403
b10 −0.0394 0.3152 0.4516 0.1364
b11 1.4393 0.0802 −1.2593 2.0622
b20 0.1724 −0.0447 −0.1094 0.1498
b21 −0.9219 0.3521 1.6861 −1.2358
b30 −0.0326 0.0265 0.0354 −0.0201
b31 0.2070 −0.1725 −0.5677 0.2429
c00 −0.0190 0.000066 0.0047 0.0091
c01 −0.1314 −0.0898 −0.0615 −0.0129
c10 0.3193 0.2819 0.3377 0.3596
c11 0.3330 0.0381 0.0363 0.0514
c20 0.0056 −0.0100 −0.0242 −0.0090
c21 −0.0527 −0.0124 0.0078 −0.0046
Appl. Sci. 2019, 9, 3220 7 of 19

Table 2. Coefficient parameters in Equations (26)–(35) in regulator design.

Msd 1.4 1.6 1.8 2.0

a00 0.2085 0.2718 0.2999 0.3672
a01 −0.6075 −0.8871 −0.6490 −1.4148
a10 0.4445 0.5897 0.7267 0.7914
a11 −0.3597 −0.3261 −0.7568 −0.1116
a20 −1.0048 −1.0010 −0.9840 −1.0107
a21 2.4219 2.5022 2.1738 2.7688
b00 0.2175 0.1208 0.1676 0.1793
b01 1.0142 1.4350 0.5152 0.5668
b10 1.3058 1.5359 1.4478 1.3845
b11 −4.3025 −4.9006 −1.6551 −1.4977
b20 −0.7838 −0.8310 −0.6531 −0.4397
b21 3.7862 4.0734 0.9992 0.8169
b30 0.2250 0.2067 0.1519 0.0589
b31 −1.0977 −1.1117 −0.2245 −0.1967
c00 −0.0031 0.0139 0.0152 0.0314
c01 0.0802 0.1103 0.0765 0.1761
c10 0.4456 0.3783 0.3607 0.3006
c11 0.3391 0.0800 −0.0139 −0.3791
c20 −0.0467 −0.0296 −0.0374 −0.0100
c21 −0.1076 −0.0107 0.0186 0.2333

The controller parameters are optimized in the limited range. Although the range can be expanded,
the tuning rule is redesigned when the range is changed.

4.3.2. Evaluation of the Prescribed Robust Stability

Non-obtained controller parameters are interpolated, where using the tuning rule proposed in
Section 4.3.1, the controller parameters are calculated on the pair of the normalization parameters
for the controlled plant: τ0 ∈ {0.3, 0.31, · · · , 1.7} and τa ∈ {0.01, 0.011, · · · , 1.0}, where Msd is varied
in {1.4, · · · , 2.0}. The calculated controller parameters for Msd = 1.4, merged into the preliminary
calculated parameters shown in Figure 2, are shown in Figure 3. The calculated parameters using the
proposed tuning rule are sufficiently close to the optimized parameters.

Figure 3. Relationship between approximated controller parameters in servo design.

Appl. Sci. 2019, 9, 3220 8 of 19

Next, Ms values calculated using the calculated controller parameters in the servo and regulation
optimization are shown in Figures 4 and 5, respectively. The minimum and maximum calculated Ms
values are summarized in Table 3. Since the errors between Msd and Ms are within ±5%, the proposed
tuning rule has sufficient precision.

Figure 4. Relationships among the obtained Ms , τ0 , and τa in servo design.

Figure 5. Relationships among the obtained Ms , τ0 , and τa in regulation design.

Appl. Sci. 2019, 9, 3220 9 of 19

Table 3. Maximum and minimum values of obtained Ms .

Servo Regulator
Msd
min max min max
1.4 1.3923 1.4088 1.3904 1.4216
1.6 1.5836 1.6130 1.5819 1.6183
1.8 1.7725 1.8256 1.7738 1.8266
2.0 1.9518 2.0359 1.9527 2.0356

4.4. Algorithm
The proposed design procedure is summarized as the following algorithm:

1. The plant parameters and the controller parameters are normalized.

2. For the normalized systems, the constrained optimization problem is solved such that the
reference and disturbance responses are optimized, respectively, and the optimal controller
parameters are obtained.
3. Based on the obtained optimized parameters, the tuning rule for the controller parameters
is designed.
4. Using the tuning rule, the practical PID parameters are decided.

When the tuning rule is obtained once, the PID parameters are calculated without solving the
constrained optimization problem.

5. Numerical Examples
First, in Section 5.1, the trade-off design between the servo and regulator design is
confirmed. The accomplished robust stability is then shown in Section 5.2. Finally, the proposed
method is compared with two conventional discrete time trade-off design methods [27,28] in
Sections 5.3 and 5.4, respectively.

5.1. Trade-off Tracking Performance Comparison

As a controlled plant, the following transfer function is used:

0.0231 + 0.0114z−1 −14

Pd (z−1 ) = z (36)
1 − 0.9753

The transfer function is the discrete time representation of the following continuous time system
with Ts = 0.03 s:

1.4
P(s) = e−0.4s (37)
1.2s + 1

Using the proposed method, the PID parameters are decided based on Equation (36) with Msd ∈
{1.4, 1.6, 1.8, 2.0}. The simulation results obtained using the PID parameters are shown in Figure 6, where
the reference input is 1.0 and the control input is disturbed by a unit step signal after 15 s. Figure 6 shows
that the servo design is superior to the regulator design with respect to reference tracking performance.
On the other hand, the regulator design is superior to the servo design with respect to disturbance rejection.
Appl. Sci. 2019, 9, 3220 10 of 19

1.6

1.4
Servo (Md=2.0)
s
1.2 d
Regulation (Ms =2.0)
1 Servo (Md=1.8)
Output [−]

0.8 Regulation (Md=1.8)

s
d
Servo (Ms =1.6)
0.6
d
Regulation (Ms =1.6)
0.4 d
Servo (M =1.4)
s
0.2 Regulation (Md=1.4)
s
0
0 5 10 15 20 25 30
Time [sec]

3
Servo (Mds =2.0)
2.5
Regulation (Md=2.0)
s
2 d
Servo (Ms =1.8)
1.5
Regulation (Mds =1.8)
Input [−]

1 Servo (Mds =1.6)

d
0.5 Regulation (Ms =1.6)
Servo (Mds =1.4)
0
d
Regulation (Ms =1.4)
−0.5

−1
0 5 10 15 20 25 30
Time [sec]

Figure 6. Output and input responses obtained using the proposed method in both servo and regulator
design, where the system is disturbed by a unit step disturbance after 15 s.

The control performance is evaluated using Js and Jr , where Js denotes the SAE value from the
start until 15 s, and Jr also denotes the SAE value from 15 s until the end. Here, Js and Jr , as well as
the PID parameters and Ms in the servo and regulator, respectively, are summarized in Table 4. In the
servo design, Js is smaller than Jr . In the regulator design, Jr is smaller than Js .

Table 4. Proportional-integral-derivative parameter, Ms , and index values.

Msd Kp Ti Td Ms Js Jr
1.4 1.0217 1.3331 0.1048 1.3998 0.9576 1.2253
Servo 1.6 1.3709 1.4633 0.1090 1.5964 0.7638 1.1531
Optimization 1.8 1.6359 1.5879 0.1360 1.7937 0.7064 1.0688
2.0 1.8093 1.7116 0.1537 1.9936 0.6970 1.0274
1.4 1.0159 0.6876 0.1737 1.4052 1.3048 0.8667
Regulator 1.6 1.3430 0.6641 0.1681 1.5944 1.0673 0.6466
Optimization 1.8 1.6065 0.7020 0.1597 1.7913 0.9705 0.5302
2.0 1.8217 0.7174 0.1589 1.9922 0.9458 0.4565

From both Figure 6 and Table 4, the larger Msd , the better the control performance. Therefore,
the trade-off design is achieved using the proposed method.

5.2. Robust Stability

The effectiveness of the accomplished robust stability is shown. The PID controller is designed
based on the nominal model Equation (36) in Section 5.1 in the servo and regulation, respectively.
Appl. Sci. 2019, 9, 3220 11 of 19

The scenario is that the plant is Equation (36) from the start to 30 s and is changed to Equation (38)
after 30 s as the model perturbation.

0.0530 −21
Pd0 (z−1 ) = z (38)
1 − 0.9788

The transfer function is the discrete time version of the following continuous time model with the
sampling interval Ts = 0.03 s:

2.5
P0 (s) = e−0.6s (39)
1.4s + 1

The simulation results for the servo and regulator designs are shown in Figure 7, where the
reference input is given by a unit step function, and the control input is disturbed by a unit step
function after 15 s. The simulation results show that the effect of the model perturbation is suppressed
by small Msd . However, note that the control performance is superior before the model perturbation
when the value of Msd is large.

Servo optimization regulator optimization

3 3

2.5 2.5

2 2

1.5 1.5
Output [−]

Output [−]

1 1
Md=2.0 Md=2.0
0.5 s 0.5 s
d d
Ms =1.8 Ms =1.8
0 0
Md=1.6 Md=1.6
s s
−0.5 Md=1.4 −0.5 Md=1.4
s s
−1 −1
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Time [sec] Time [sec]

Servo optimization regulator optimization

3 3
Mds =2.0 Mds =2.0
2.5 2.5
d d
Ms =1.8 Ms =1.8
2 2
Mds =1.6 Mds =1.6
1.5 1.5
Mds =1.4 Mds =1.4
Input [−]

Input [−]

1 1

0.5 0.5

0 0

−0.5 −0.5

−1 −1
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Time [sec] Time [sec]

Figure 7. Output and input responses obtained using the proposed method for each Msd with
model perturbation, where the system is disturbed after 15 s; left: servo optimization; right:
regulator optimization.

5.3. Comparison with the IMC-Based Method

Consider the following discrete time system:

0.0100
Pd (z−1 ) = z−25 (40)
1 − 0.99z−1
Appl. Sci. 2019, 9, 3220 12 of 19

The system is the discrete time representation of the following continuous time system with a
sampling interval of 0.01 s:

1 −0.25s
P(s) = e (41)
s+1

In the conventional discrete time trade-off design method [27], the IMC controller is designed as:

(1 − a1 )(1 − λc )
C I MC (z−1 ) = (42)
b0 (1 − λc z−1 )

where λc is the trade-off design parameter. The obtained IMC controller is approximated by the
following discrete time PID control law:

Td (1 − z−1 )
 
1
CPID (z−1 ) = K p 1+ + ( 1 − αTd ) (43)
Ti (1 − z−1 ) 1 − αTd z−1

where λc is designed in the range of [0.8,0.99]. However, the designed control systems are unstable
when λc ≤ 0.96, and thus, the conventional control law is designed such that λc is set to 0.97, 0.98, and
0.99, respectively, where α = 0.1. The designed PID parameters for the conventional method and the
proposed method are shown in Table 5.

Table 5. Proportional-integral-derivative parameters, Ms , and sum of absolute errors (SAE) values

using the proposed and internal model control (IMC)-based methods for Equation (40).

Kp Ti Td Ms Js Jr
IMC-based λc = 0.99 0.8248 103.10 4.6696 1.1625 1.2500 1.2493
method [27] λc = 0.98 1.3978 104.83 7.5346 1.2937 0.7533 0.7500
λc = 0.97 1.8184 106.07 9.4658 1.5659 0.5919 0.5833
Msd = 1.4 1.9120 1.1242 0.0606 1.4014 0.5936 0.5878
Proposed Msd = 1.6 2.5724 1.2107 0.0690 1.5919 0.4733 0.4705
Servo Msd = 1.8 3.0641 1.3062 0.0871 1.7869 0.4367 0.4260
Msd = 2.0 3.3823 1.4277 0.0984 1.9851 0.4313 0.4216
Msd = 1.4 1.9193 0.5000 0.1070 1.3999 0.8079 0.3352
Proposed Msd = 1.6 2.5319 0.4605 0.1079 1.5858 0.7646 0.2432
Regulator Msd = 1.8 3.0234 0.4928 0.1038 1.7778 0.6951 0.1975
Msd = 2.0 3.4407 0.5012 0.1069 1.9940 0.6642 0.1689

The simulations were conducted using the conventional and proposed methods, where the
reference input was set to 1.0, and the control input was disturbed by a unit step function after 30 s.
The conventional method is compared with the proposed servo and regulator optimization methods
in Figure 8. The obtained Ms values and the evaluated values Js and Jr are also summarized in Table 5,
where Js denotes the SAE value from the start until 30 s, and Jr also denotes the SAE value from 30 s
until the end. Table 5 shows that the conventional method provides a trade-off design by selecting
λc even though no value is assigned to Ms . Moreover, the tracking performances obtained using the
proposed method are superior to those obtained using the conventional method.
Appl. Sci. 2019, 9, 3220 13 of 19

Servo optimization Regulator optimization

1.5 1.5

1 1
Output [−]

Output [−]
Proposal (Mds =2.0) Proposal (Mds =2.0)
Proposal (Mds =1.8) Proposal (Mds =1.8)
Proposal (Mds =1.6) Proposal (Mds =1.6)
0.5 0.5
Proposal (Mds =1.4) Proposal (Mds =1.4)
Cui (λc=0.97) Cui (λc=0.97)
Cui (λc=0.98) Cui (λc=0.98)
Cui (λc=0.99) Cui (λc=0.99)

0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time [sec] Time [sec]

Servo optimization Regulator optimization

4.5 6
Proposal (Md=2.0) Proposal (Md=2.0)
4 s s
Proposal (Mds =1.8) 5 Proposal (Mds =1.8)
3.5 d d
Proposal (Ms =1.6) Proposal (Ms =1.6)
3 Proposal (Md=1.4) 4 Proposal (Md=1.4)
s s
Cui (λc=0.97) Cui (λc=0.97)
2.5
3
Input [−]

Cui (λc=0.98) Input [−] Cui (λc=0.98)

2 Cui (λc=0.99) Cui (λc=0.99)
2
1.5

1 1

0.5
0
0

−0.5 −1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time [sec] Time [sec]

Figure 8. Results of the IMC-based method and the proposed servo and regulator optimization
methods, where the system is disturbed by a unit step disturbance after 10 s; left: servo optimization;
right: regulator optimization.

5.4. Comparison with the Conventional Discrete-Time Method

The conventional discrete time design method [28] is compared with the proposed method.
Here, two scenarios are conducted, in which a non-zero plant and a zero-included plant, respectively,
are controlled.
In the first simulation, we consider the following non-zero discrete time plant:

0.05127z−1 −10
Pd1 (z−1 ) = z (44)
1 − 0.9486z−1

where Equation (44) is the discrete time representation of Equation (45) with a sampling interval of
Ts = 0.05 s.

1
P1 (s) = e−0.5s (45)
0.95s + 1

Equation (44) has no zero since the continuous time dead-time is an integer multiple of the
sampling interval.
In the second simulation, the controlled discrete time system is given as follows:

0.0201 + 0.02473z−1 −7
Pd2 (z−1 ) = z (46)
1 − 0.9552z−1
Appl. Sci. 2019, 9, 3220 14 of 19

where Equation (46) is the discrete time version of the continuous time system given by Equation (47)
with a sampling interval of Ts = 0.061 s.

1
P2 (s) = e−0.4s (47)
1.33s + 1

Equation (46) has a zero since the dead-time in the continuous time model is not an integer
multiple of the sampling interval. Since the discrete time system has a zero, the conventional method
is not directly used. Therefore, Equation (46) is hereby approximated by the next discrete time system,
and the conventional method is used:

0 0.04483z−1 −7
Pd2 ( z −1 ) = z (48)
1 − 0.9552z−1

In the simulations, the reference input is set to 1.0. Furthermore, the control input is disturbed by
a unit step function signal after 10 s. The obtained PID parameters are shown in Tables 6–9. Using
the obtained parameters, the discrete time models Equations (44) and (46) are controlled, respectively,
and the output results are plotted in Figures 9 and 10. Furthermore, the obtained Ms value and index
values Js and Jr are also shown in Tables 6–9, where Js denotes the SAE value while the control is not
disturbed, and Jr denotes the SAE value while the control input is disturbed.

Table 6. Proportional-integral-derivative parameters, Ms , and SAE values in the servo design using
Equation (44).

Msd Kp Ti Td Ms Js Jr
1.4 0.9373 1.0470 0.1445 1.4002 1.1737 1.1171
Proposed 1.6 1.2529 1.1866 0.1346 1.6009 0.9489 0.9465
method 1.8 1.5008 1.2976 0.1649 1.8023 0.8833 0.8638
2.0 1.6595 1.3763 0.1867 2.0043 0.8703 0.8282
1.4 0.9732 1.0888 0.1367 1.4224 1.1427 1.1186
Conventional 1.6 1.2996 1.2398 0.1376 1.6311 0.9539 0.9531
method 1.8 1.5332 1.3585 0.1682 1.8363 0.8955 0.8846
2.0 1.6849 1.4374 0.1894 2.0442 0.8802 0.8513

Table 7. Proportional-integral-derivative parameters, Ms , and SAE values in the regulator design using
Equation (44).

Msd Kp Ti Td Ms Js Jr
1.4 0.9239 0.6663 0.2190 1.4009 1.3680 0.8922
Proposed 1.6 1.2254 0.6837 0.2020 1.6017 1.2768 0.6949
method 1.8 1.4660 0.7268 0.1886 1.8026 1.2042 0.5834
2.0 1.6614 0.7538 0.1796 2.0054 1.1653 0.5092
1.4 0.9608 0.6652 0.2151 1.4227 1.3553 0.8618
Conventional 1.6 1.2648 0.6865 0.2002 1.6322 1.2685 0.6745
method 1.8 1.4886 0.7154 0.1818 1.8314 1.2143 0.5733
2.0 1.7045 0.7580 0.1789 2.0559 1.1583 0.4945
Appl. Sci. 2019, 9, 3220 15 of 19

Table 8. Proportional-integral-derivative parameters, Ms , and SAE values in the servo design using
Equations (46) and (48), respectively.

Msd Kp Ti Td Ms Js Jr
1.4 1.4664 1.4390 0.1009 1.4026 1.0088 0.9796
Proposed 1.6 1.9725 1.5788 0.1066 1.6010 0.8020 0.7978
method 1.8 2.3577 1.7146 0.1354 1.8014 0.7412 0.7236
2.0 2.6043 1.8463 0.1550 2.0010 0.7320 0.7037
1.4 1.3144 1.5050 0.1075 1.3571 1.1449 1.1407
Conventional 1.6 1.7619 1.6841 0.1194 1.5024 0.9553 0.9495
method [28] 1.8 2.0745 1.8297 0.1473 1.6576 0.8813 0.8737
2.0 2.2804 1.9491 0.1662 1.8076 0.8537 0.8441

Table 9. Proportional-integral-derivative parameters, Ms , and SAE values in the regulator design using
Equation (46) and (48), respectively.

Msd Kp Ti Td Ms Js Jr
1.4 1.4332 0.7274 0.1790 1.4026 1.2958 0.6457
Proposed 1.6 1.8980 0.7008 0.1744 1.6030 1.2111 0.4786
method 1.8 2.2724 0.7352 0.1645 1.8028 1.1244 0.3915
2.0 2.5759 0.7527 0.1659 2.0076 1.0778 0.3375
1.4 1.2922 0.7535 0.1982 1.3571 1.3330 0.7262
Conventional 1.6 1.6990 0.7295 0.1833 1.5124 1.2267 0.5433
method [28] 1.8 2.0191 0.7328 0.1731 1.6620 1.1631 0.4495
2.0 2.2923 0.7810 0.1659 1.8111 1.0886 0.3925

Servo optimization Regulator optimization

1.6 1.5

1.4

1.2

1
1
Proposal (Mds =2.0) Proposal (Mds =2.0)
Output [−]

Output [−]

0.8 Conventional (Mds =2.0) Conventional (Mds =2.0)

Proposal (Mds =1.8) Proposal (Mds =1.8)

0.6 d d
Conventional (Ms =1.8) Conventional (Ms =1.8)
0.5
Proposal (Mds =1.6) Proposal (Mds =1.6)
0.4
Conventional (Mds =1.6) Conventional (Mds =1.6)

Proposal (Mds =1.4) Proposal (Mds =1.4)

0.2
Conventional (Mds =1.4) Conventional (Mds =1.4)

0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time [sec] Time [sec]

Servo optimization Regulator optimization

2.5 3

Proposal (Mds =2.0)

Proposal (Mds =2.0)
Conventional (Mds =2.0) 2.5
2 Conventional (Mds =2.0)
Proposal (Mds =1.8)
Proposal (Mds =1.8)
Conventional (Mds =1.8) 2
Conventional (Mds =1.8)
1.5
Proposal (Mds =1.6)
Proposal (Mds =1.6)
Conventional (Mds =1.6) 1.5
Conventional (Mds =1.6)
Input [−]

Input [−]

1 Proposal (Mds =1.4)

Proposal (Mds =1.4)
Conventional (Mds =1.4) 1
Conventional (Mds =1.4)

0.5
0.5

0
0

−0.5 −0.5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time [sec] Time [sec]

Figure 9. Output and input trajectories for the proposed and conventional designs for Equation (44),
where the system is disturbed by a unit step disturbance after 10 s; left: servo optimization; right:
regulator optimization.
Appl. Sci. 2019, 9, 3220 16 of 19

Servo optimization Regulator optimization

1.4 1.4

1.2 1.2

1 1

Proposal (Mds =2.0) Proposal (Mds =2.0)

Output [−]

Output [−]
0.8 0.8
Conventional (Mds =2.0) Conventional (Mds =2.0)
d d
Proposal (Ms =1.8) Proposal (Ms =1.8)
0.6 0.6
Conventional (Md=1.8) Conventional (Md=1.8)
s s

Proposal (Md=1.6) Proposal (Md=1.6)

0.4 s 0.4 s

Conventional (Mds =1.6) Conventional (Mds =1.6)

0.2 Proposal (Md=1.4) 0.2 Proposal (Md=1.4)

s s

Conventional (Md=1.4) Conventional (Md=1.4)

s s

0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time [sec] Time [sec]

Servo optimization Regulator optimization

3.5 4.5
Proposal (Mds =2.0) Proposal (Md=2.0)
s
d 4
3 Conventional (Ms =2.0) Conventional (Mds =2.0)
d d
Proposal (M =1.8) 3.5 Proposal (Ms =1.8)
s
2.5
Conventional (Mds =1.8) Conventional (Mds =1.8)
3
2 Proposal (Mds =1.6) Proposal (Mds =1.6)
2.5
Conventional (Mds =1.6) Conventional (Mds =1.6)
Input [−]

Input [−]
1.5 Proposal (Mds =1.4) 2
Proposal (Mds =1.4)
Conventional (Mds =1.4) 1.5 Conventional (Mds =1.4)
1
1
0.5
0.5
0
0

−0.5 −0.5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time [sec] Time [sec]

Figure 10. Output and input trajectories for the proposed and conventional designs for Equation (46),
where the system is disturbed by a unit step disturbance after 10 s; left: servo optimization; right:
regulator optimization.

In the case of the non-zero system Equation (44), Js , Jr , and Ms obtained using the conventional
method were close to those obtained using the proposed method. On the other hand, when the
controlled plant was a zero-included system, Equation (46) was out of range of the conventional
method. Therefore, the tracking performances using the conventional method were inferior to those
using the proposed method, and hence, the SAE values using the conventional method were larger than
those using the proposed method. Furthermore, the prescribed robust stabilities Msd were achieved
using the proposed method, while on the other hand, the Ms values obtained using the conventional
method were insufficient.
The simulation results showed that both the conventional and proposed methods were useful
for the non-zero plants. However, when the zero-included plant was controlled, the design
objective was achieved using the proposed method even though the conventional method was not
available. Therefore, the proposed method is a more general method of the conventional discrete time
design method.

6. Conclusions
The present study proposed a new trade-off PID control design method for discrete time FOPDT
systems including a zero. In the proposed method, the regulator- or servo-optimal PID controller was
designed in discrete time. In the proposed method, since robust stability was a design parameter,
it was adjustable depending on the model perturbation.
In the conventional discrete time design method [28], the designable class was restricted such
that the dead-time in the continuous time system must be an integer multiple of the sampling interval,
and hence, no zero appeared. On the other hand, in the proposed method, the constraint condition
was relaxed.
Appl. Sci. 2019, 9, 3220 17 of 19

Author Contributions: Conceptualization, R.K. and T.S.; methodology, R.K., T.S., and R.V.; software, R.K.;
validation, T.S. and R.V.; formal analysis, R.K.; investigation, T.S. and R.V.; resources, R.K.; data curation, R.K. and
T.S.; writing, original draft preparation, T.S.; writing, review and editing, R.V.; visualization, R.K., T.S., and R.V.;
supervision, T.S. and R.V.; project administration, T.S., R.V., and Y.K.; funding acquisition, T.S., R.V., and Y.K.
Acknowledgments: The present study was supported by JSPS KAKENHI Grant Number 16K06425. This research
is supported by the Catalan Government under Project 2017 SGR 1202 and also by the Spanish Government under
Project DPI2016-77271-R co-funded with European Union ERDF funds.
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A. Derivation of Normalization Parameters

In the same way as the conventional continuous time design, the normalization parameters
are obtained.
Here, τ0 is derived as follows:

L
τ0 =
T
Ts L
= d+ 0 (A1)
T T
Ts
In this equation, T is derived using Equation (3):

Ts
= − log a1 (A2)
T
L0
and T is also derived by eliminating K from Equations (4) and (5):
 
L0 b0 + a1
= log (A3)
T (b0 + b1 ) a1

Equations (A2) and (A3) are substituted into Equation (A1), and Equation (18) is then obtained.
From Equation (A2), τa is derived as:

Ts
τa =
T
= − log a1 (A4)

Since K is the steady-state gain, κ p is obtained as:

κ p = KK p
b0 + b1
= Kp (A5)
1 − a1

From Equation (A2), τi and τd are calculated as follows:

Ti
τi =
T
Ti log a1
=− (A6)
Ts
Td
τd =
T
T log a1
=− d (A7)
Ts
Appl. Sci. 2019, 9, 3220 18 of 19

References
1. O’Dwyer, A. Handbook of PI and PID Controller Tuning Rules; Imperial College Press: London, UK, 2003.
2. Silva, G.J.; Datta, A.; Bhattacharyya, S. PID Controllers for Time-Delay Systems; Birkhäuser: Boston, MA,
USA, 2005.
3. Johnson, M.A.; Moradi, M.H. (Eds.) PID Control: New Identification and Design Methods; Springer: London,
UK, 2005.
4. Åström, K.J.; Hägglund, T. Advanced PID Control; Instrumentation, Systems, and Automation Society:
Pittsburgh, PA, USA, 2006.
5. Visioli, A. Practical PID Control; Springer: London, UK, 2006.
6. Yu, C.C. Autotuning of PID Controllers, 2nd ed.; Springer: London, UK, 2006.
7. Vilanova, R.; Visioli, A. (Eds.) PID Control in the Third Millennium; Springer: London, UK, 2012.
8. Boiko, I. Non-Parametric Tuning of PID Controllers; Springer: London, UK, 2013.
9. Alfaro, V.; Vilanova, R. Model-Reference Robust Tuning of PID Controllers; Advances in Indstrial Control;
Springer: Cham, Switzerland, 2016.
10. Ziegler, J.; Nichols, N. Optimum settings for automatic controllers. Trans. ASME 1942, 64, 759–768.
[CrossRef]
11. Shinskey, F.G. How good are our controllers in absolute performance and robustness? Measur. Control 1990,
23, 114–121. [CrossRef]
12. Shinskey, F.G. Process-Control Systems. Application, Design, and Tuning, 4th ed.; McGraw-Hill: New York, NY,
USA, 1996.
13. Chen, C.; Huang, H.; Hsieh, C. Tuning of PI/PID controllers based on specification on maximum closed-loop
amplitude ratio. J. Chem. Eng. Jpn. 1999, 32, 783–788. [CrossRef]
14. Matušů, R.; Prokop, R. Computation of robustly stabilizing PID controllers for interval systems. SpringerPlus
2016, 5, 702. [CrossRef] [PubMed]
15. Jahanmiri, A.; Fallahi, H. New methods for process identification and design of feedback controller.
Chem. Eng. Res. Des. 1997, 75, 519–522. [CrossRef]
16. Ackermann, J.; Kaesbauer, D.; Bajcinca, N. Discrete-time robust PID and three-term control. IFAC Proc. Vol.
2002, 35, 127–132. [CrossRef]
17. Hiroaki, M.; Takayoshi, U.; Yoshiyuki, T.; Toshio, T. Guaranteed cost PI control for uncertain discrete
time systems with additive gain. In Proceedings of the 2009 European Control Conference (ECC),
Budapest, Hungary, 23–26 August 2009; pp. 2319–2324.
18. Garpinger, O.; Hägglund, T.; Åström, K.J. Criteria and trade-offs in PID design. IFAC Proc. Vol. 2012,
45, 47–52. [CrossRef]
19. Alfaro, V.; Vilanova, R.; Méndez, V.; Lafuente, J. Performance/robustness tradeoff analysis of PI/PID servo
and regulatory control systems. In Proceedings of the 2010 IEEE International Conference on Industrial
Technology, Vina del Mar, Chile, 14–17 March 2010; pp. 111–116.
20. Alcántara, S.; Pedret, C.; Vilanova, R. On the model matching approach to PID design: Analytical perspective
for robust servo/regulator tradeoff tuning. J. Process Control 2010, 20, 596–608. [CrossRef]
21. Arrieta, O.; Vilanova, R. Simple servo/regulation Proportional-Integral-Derivative (PID) tuning rules for
arbitrary Ms -based robustness achievement. Ind. Eng. Chem. Res. 2012, 51, 2666–2674. [CrossRef]
22. Kurokawa, R.; Inoue, N.; Sato, T.; Arrieta, O.; Vilanova, R.; Konishi, Y. Simple optimal PID tuning method
based on assgined robust stability—Trade-off design based on servo/regulation performance. Int. J. Innov.
Comput. Inf. Control 2017, 13, 1953–1963.
23. Alfaro, V.M.; Méndez, V.; Vilanova, R. Robust-performance tuning of 2DoF PI/PID controllers for first- and
second-order-plus-dead-time models. In Proceedings of the 2011 9th IEEE International Conference on
Control and Automation (ICCA), Santiago, Chile, 19–21 December 2011; pp. 237–242.
24. Sato, T.; Hayashi, I.; Horibe, Y.; Vilanova, R.; Konishi, Y. Optimal robust PID control for first- and
second-order plus dead-time processes. Appl. Sci. 2019, 9, 1934. [CrossRef]
25. Alfaro, V.; Vilanova, R. Robust tuning and performance analysis of 2DoF PI controllers for integrating
controlled processes. Ind. Eng. Chem. Res. 2012, 51, 13182–13194. [CrossRef]
26. Alfaro, V.; Vilanova, R. Model-reference robust tuning of 2DoF PI controllers for first- and second-order plus
dead-time controlled processes. J. Process Control 2012, 22, 359–374. [CrossRef]
Appl. Sci. 2019, 9, 3220 19 of 19

27. Jiyao, C.; Yueling, C.; Tao, L. Discrete-time domain IMC-based PID control design for industrial processes
with time delay. In Proceedings of the 2016 35th Chinese Control Conference (CCC), Chengdu, China,
27–29 July 2016; pp. 5946–5951.
28. Tajika, H.; Sato, T.; Vilanova, R.; Konishi, Y. Optimal PID control in discrete time using sensitivity
function. In Proceedings of the 2015 23rd Mediterranean Conference on Control and Automation (MED),
Torremolinos, Spain, 16–19 June 2015; pp. 268–273.

c 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).