Accepted Manuscript

A new design method for PI-PD control of unstable processes with dead
time

Cem Onat

PII: S0019-0578(18)30325-2
DOI: https://doi.org/10.1016/j.isatra.2018.08.029
Reference: ISATRA 2901

To appear in: ISA Transactions

Received date : 22 June 2018

Revised date : 15 August 2018
Accepted date : 28 August 2018

Please cite this article as: Onat C. A new design method for PI-PD control of unstable processes
with dead time. ISA Transactions (2018), https://doi.org/10.1016/j.isatra.2018.08.029

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
*Title page showing Author Details

A New Design Method for PI-PD Control of Unstable

Processes with Dead Time

Cem Onat
[email protected]
Firat University, Civil Aviation High School, Department of Airframe and Power-plant,
Elazig, Turkey
*Highlights (for review)

HIGHLIGHTS
 The paper represents a novel method for PI-PD control of the unstable
systems.
 The proposed method uses a new concept named centroid of convex
stability region.
 Ability of being extended to uncertain plants of the proposed method by
means of Kharitonov theorem is an important advantage.
*Blinded Manuscript - without Author Details
Click here to view linked References

A New Design Method for PI-PD Control of Unstable

Processes with Dead Time

Abstract:
PID controllers are still widely practiced in the industrial systems. In the literature, many
publications can be found considering PID controller design for unstable processes. However,
owing to the structural limitations of PID controllers, generally, good closed loop
performance cannot be achieved with a PID for controlling unstable processes and usually a
step response with a high overshoot and oscillation is obtained. On the other hand, PI-PD
controllers are proved to give very satisfactory closed loop performances for unstable
processes. The paper presents a simple design method to tune parameters of a PI-PD
controller for the control of the unstable processes with time delay. The proposed method is
based on plotting the stability boundary locus, which is a locus dependent on the parameters
of the controller and frequency, in the parameter plane. The method uses a new concept
named centroid of the convex stability region. Simulation examples and an experimental
application are given to illustrate the superiority of the proposed method over some existing
ones.

Keywords: PI-PD controller, convex stability region, dead time, unstable, uncertain plant.

1. Introduction

Proportional-Integral-Derivative (PID) controllers are low-order controllers which have been

simply used for decades to control higher-order processes. PID controllers still have wide
usage area due to their simple structure. Their widespread utilization in practice motivates
researchers in developing better PI/PID adjustment methods. The controller has three
parameters (kp, ki and kd) to be tuned. In general, the PI/PID controller is structurally suitable
and shows an acceptable control performance for many open loop stable processes with dead
time [1–9]. Many studies have focused on the tuning of PI/PID controller parameters [10–24].
The design studies are classified into three categories. These categories are iterative
optimization studies [10–16], tuning formulae studies for certain class systems [3–9, 17–19]
and studies based on stabilizing controller parameters region [20–24]. It is seen from the
literature that recent PID tuning studies are mostly based on stability boundary locus method.
The suggested methods have effectively provided solutions to the control of certain class
systems. However, PID controllers have structural limitations in controlling certain types of
unstable processes with dead time [25–29]. A modified form of PID controller, a PI-PD
controller structure, which corresponds to a PI controller of a plant transfer function changed
by the PD feedback can produce improved control in several situations [30]. Using an internal
PD feedback loop can convert an open loop unstable process to an open loop stable process
and for resonant or integrating processes can ensure appropriate locations of the open loop
stable process poles. The PI-PD controller has four parameters (kf, kd, kp and ki) to be tuned.
Therefore, the PI-PD controller structure has advantages over the conventional PID and PID-P
controllers [25, 31–34]. On the other hand, conventional design procedures are based on a
plant with fixed parameters although most practical system models have uncertainties [35–
39]. Therefore, the design of a satisfactory control scheme has various requirements from the
consideration of robustness to parameter uncertainties, stability and performance [40–43].

Recently, PI-PD controller design methods have been reported in the literature [28, 30, 44,
45]. Generally, the studies have focused on the unstable systems with dead time. Kaya and
Atherton have introduced a simple approach to tune parameters of a PI-PD controller for the
control of integrating and unstable systems [44]. For this purpose, they have used tuning
formulae for the ISE (Integral of Squared Error) and ISTE (Integral of Squared Time
weighted Error) minimizations. Kaya has proposed a model based PI-PD controller for stable
systems, where PD feedback is used to change the poles of the plant transfer function to more
desirable locations for control by a PI controller [28]. Tan has proposed a method which is
based on plotting the stability boundary loci and combined with the Kharitonov theorem [46]
to design robust PI-PD controllers for a control system with uncertain parameters [30]. The
method proposed by Tan benefits from the damping ratio which is an important concept for
resonance systems. Therefore, application of the method to different systems is limited.
Furthermore, the method has certain arbitrariness through the choice of the damping ratio.
Another design study on PI-PD controller has been recently done by Ozyetkin [45]. Ozyetkin
has designed integer order and fractional order PI-PD controllers by using the weighted
geometrical center (WGC) method which was firstly suggested by Onat for PI control of time
delay systems [21]. The design procedure used in [45] is not precise and includes trial and
error manipulations. To compute the WGC, it is necessary to use all the frequency dependent
points forming the stability boundary locus. So it brings an extra computational load and a
disadvantage for practical purposes. In addition, WGC might be kept out of the stabilizing
controller parameters region in case of deviating the region for different design cases. The
WGC method cannot be used in the cases. This weakness of the WGC method can be seen
from the figure numbered 21 in the reference 45.

General literature evaluation refers to two results. First, PI-PD controller schema is more
proper than PID controller schema for the control of unstable or integrating systems. Second,
number of studies on PI-PD controller design is comparatively shorter than PID controller
design studies. Although many existing methods have presented acceptable performance, it
needs to be improved. In this perspective, the paper represents a new basic method to design
parameters of a PI-PD controller for control of the unstable systems with dead time. The
paper originally presents a new concept named centroid of the convex stability region. The
main contributions of this paper are the presentation of a new design procedure for PI-PD
controller structure. The proposed procedure has numerically determined the PI-PD controller
parameters without any iterative optimization process and guaranteed the stability of the
closed-loop. It is precise and does not involve trial and error manipulation in any way. The
procedure uses only the coordinates of a few special points pertaining to the curve
determining the stabilizing controller parameters region. Therefore, computational load of the
proposed design procedure is very low compared to the WGC method. This is a significant
advantage for practical purposes. Because, engineers can simply design PI-PD controller by
means of the stabilizing controller parameters region. In addition, the proposed design
procedure guarantees stability in design cases where the region has been tilted and can be
applied to all processes whose stability region can be calculated as a closed region.
Furthermore, the ability of being extended to uncertain plants of the proposed method by
means of Kharitonov theorem is another important advantage. It provides a good tool for
robust PI-PD controller design. The proposed method is applied to different unstable plants
with dead time. Simulation examples show that the proposed method is superior over some
existing ones. In addition to this, the method has also been applied in real time to the magnetic
ball levitation system. Real time experimental results demonstrate the effectiveness of the
proposed method in practice.

The paper is organized as follows: The proposed method is described in Section 2. Extension
of the method to uncertain systems is presented in Section 3. Simulative and experimental
application examples are given in Section 4. Concluding remarks are given in Section 5.
 NOMENCLATURE

NOTATION DEFINITION
PID Proportional Integral Derivative
WGC Weighted Geometrical Center
IAE Integral of Absolute Error
ISE Integral of Squared Error
ISTE Integral of Squared Time Weighted Error
CPI Transfer function of the PI Controller
CPD Transfer function of the PD Controller
s Laplace Variable
Δ Characteristic Equation
ω Frequency
τ Time Delay or Dead Time
RΔ Real Part of Characteristic Equation
IΔ Imaginary Part of Characteristic Equation
kp Proportional Gain of the PI Controller
kf Proportional Gain of the PD Controller
kd Derivative Gain
ki Integral Gain
kpc Proportional Gain of the Centroid PI Controller
kfc Proportional Gain of the Centroid PD Controller
kdc Derivative Gain of the Centroid PD Controller
kic Integral Gain of the Centroid PI Controller
n Number of Cusp Points
m Number of Corner Points
j
v Cusp Points Number of Common Region
w Corner Points Number of Common Region
r Set Point Reference Signal
y Output Signal of the Closed Loop System
 2. The Proposed Method
The PI-PD control system is shown in Fig. 1. The closed loop system consists of an inner loop
with a PD controller and outer loop with a PI controller.

r + +
CPI(s) Gp(s) y
- -

CPD(s)

Fig. 1. PI-PD control system

Plant transfer function GP (s) is defined as follows.

(1)

CPD (s) and CPI (s) refer to the transfer functions of the PD and PI controllers, respectively.
They are defined by (2) and (3).

(2)

(3)

Here, kf and kd are proportional and derivative gains of the PD controller and, kp and ki are
proportional and integral gains of the PI controller respectively. The closed loop
characteristic equation of the internal feedback loop with PD controller is given in (4).

(4)

The proposed design method consists of seven steps.

Step 1: Stabilizing parameters of CPD are computed for the inner loop. In this way, the
stability region is graphically determined in kd –kf plane. For this purpose, substituting s=jω
and e-τjω=cos(τω)-jsin(τω) in (4), one can obtain

. (5)
Decomposing ΔPD into its real and imaginary parts, one can write

(6)

Here, RΔ,PD and IΔ,PD are functions of kd, kf and ω. Equating real and imaginary parts of ΔPD to
zero, two equations with two unknown parameters (kd, kf) are obtained. The equation system
is given (7).

(7)

Solving the equation system depending on the frequency (ω), and then plotting obtained kd
and kf parameters in kd –kf plane, stability region is determined. The details can be found in
[20, 21, 47, 48].
Remark 1: If more than one real value of ω which satisfies (7) exist, then frequency axis can
be divided into a finite number of divisions and by testing each division the stability region
can be computed.

Step 2: Convex stability region is described by using of the cusp and corner points of the
stability boundary loci. Coordinates of the cusp and corner points in kd –kf plane are
( ), ( ),…,( ) and ( ), ( ),…,( ) respectively. Here,
n and m are number of cusp and corner points respectively.

Step 3: Centroid of the convex stability region is calculated in kd –kf plane. The formulas of
the centroid coordinates are given in (8) and (9). The centroid controller parameters (kdc and
kfc) are determined as PD controller parameters.

(8)

(9)
Step 4: The inner loop is reduced to only one transfer function by using the selected PD
controller parameters (kdc and kfc). The reduced transfer function of the inner loop is given in
(10).

(10)

Step 5: Stability region is calculated by means of reduced transfer function in kp –ki plane for
the outer loop with PI controller. Using the procedure given for the computation of the
parameters of CPD explained above, stabilizing PI controller parameters can be calculated.
Substituting s=jω into the characteristic equation of the outer loop is given in (11).

. (11)

Decomposing ΔPI into its real and imaginary parts, one can write

. (12)

Here, RΔ,PI and IΔ,PI are functions of kp, ki and ω. Equating real and imaginary parts of ΔPI to
zero, two equations with two unknown parameters (kp, ki) are obtained. The equation system
is given (13).

(13)

Solving the equation system depending on the frequency (ω), and then plotting obtained kp
and ki parameters in kp –ki plane, stability region is determined.
Step 6: Convex stability region is described by using of the cusp and corner points of the
stability boundary loci of the stabilizing PI controller parameters. Coordinates of the cusp and
corner points in kp–ki plane are ( ), ( ),…,( ) and ( ),

( ),…,( ) respectively.
Step 7: Centroid of the convex stability region is computed in kp–ki plane. The formulas of
the centroid coordinates are given in (14) and (15). The centroid controller parameters (kpc
and kic) are determined as PI controller parameters. Here, v and w are number of cusp and
corner points respectively.

(14)

(15)

In practice, parameters of the systems cannot be certainly determined or they change

according to the operation conditions. For this reason, the design methods should work out
controlling of the systems with uncertain parameters. In the next section, the method is
extended to the systems with uncertain parameters.

3. Extension of the method to uncertain systems

Analysis and design of the control systems with uncertain parameters is an important issue in
control theory. Many recent works on the systems with uncertain parameters are based on
Kharitonov’s result [46] on the stability of interval polynomials. The proposed method can be
extended to uncertain systems by using Kharitonov’s plants. Using the procedure given for
the computation of the parameters of CPD and CPI explained in the previous section, PI-PD
controller parameters for uncertain systems can be designed. For this purpose, Kharitonov
plants of the uncertain system are used. Consider the control system given in Fig. 1 with an
uncertain transfer function of the form,

, (16)

where , i=0,1,…,m and , j=0,1,…,n. Kharitonov plants of the system

are given in (17). The details can be found in [21, 30]. It is clear that the number of the
Kharitonov plants is mxn.

(17)
The proposed method can be extended to the uncertain systems by means of the Kharitonov
plants. For this target, all steps of the proposed method are individually operated for all of the
Kharitonov plants. Thus common stability region of the Kharitonov plants and their convex
stability region are computed. Finally, by computing the centroid of the common convex
stability region, robust PI-PD controller is designed. Extension of the method to uncertain
systems can be outlined in seven steps as follows.

Step 1: For the inner loop, stabilizing parameters of CPD are computed for all cases of the
Kharitonov plants. In this way, the common stability region is graphically determined in kd –kf
plane.
Step 2: Common convex stability region is described by using the cusp and corner points of
the stability boundary loci.
Remark 2: Obtained common stability region is robust region for the parameter uncertainties.
In some cases, the intersection points of the stability boundary curves determining the
common stabilizing controller parameters region for all Kharitonov plants might be faced. In
this case, the intersection points as well as the corner and the cusp points of the common
boundary curve are used to determine the common convex region.
Step 3: Centroid of the common convex stability region is calculated in kd –kf plane.
Step 4: Reduced transfer functions of the inner loop are individually calculated by using the
selected PD controller parameters for all cases of Kharitonov plant family of the uncertain
system.
Step 5: Common stability region for all cases of the reduced transfer functions is determined
for the outer loop in kp –ki plane.
Step 6: Common convex stability region is described by using the cusp and corner points of
the boundary loci of the common stability region of the PI controller parameters.
Step 7: Centroid of the common convex stability region is simply computed in kp –ki plane.
The centroid controller parameters are determined as robust PI controller parameters for the
outer loop.

Unstable systems such as humanoid robotics, magnetic levitation and transportation systems
have been encountered in practice. In the next section, simulative applications to different
systems and one real time implementation are performed to show the efficiency of the
method.
 4. Applications
This section deals with application examples. Five simulative examples and one experimental
application have been considered to illustrate the usefulness of the proposed method. A first
order unstable plant with time delay, a second order unstable plant with time delay, an
unstable plant with uncertain parameters and time delay, a long time delay unstable system, a
resonant system and a real time magnetic ball levitation system have been respectively studied
in the following examples. In the simulations, settling time, overshoot, rise time, peak time
and IAE values are used as evaluation criterions.

Example 1: Consider
(18)

which is an unstable first order plus time delay transfer function. The model is also studied by
Tan [30], Kaya [49], and Visioli [50]. Firstly, for the inner loop, stability region of the system
is graphically obtained in kd –kf plane. For this purpose, the procedure in Step 1 is operated.
The computed stability region and its convex region are given in Fig. 2. The stability region
has one cusp point and two corner points. The convex region is described by these points.
Accordingly, convex stability region is a triangle. Reading the coordinates of these points,
centroid of the convex region is calculated by using (8) and (9). Parameters of the centroid PD
controller are kf= kfc=0.3412 and kd= kdc=0.439.
 1

0,9 Cusp point

of the stability region

0,8

Srability
boundary
0,7 locus

CONVEX
0,6
kf

STABILITY
REGION

0,5

Centroid
0.439 of the convex
stability region
0,4
Corner points
of the stability
region
0,3

0,2
-0,5 0 0.3412 0,5 1
kd

Fig. 2. Centroid of the convex stability region of inner loop for Example 1.

By embedding the centroid PD controller parameters (kdc=0.3412 and kfc=0.439) into the inner
loop, it is reduced to only one transfer function by using (10). Using the transfer function,
stabilizing PI controller parameters region is calculated for the outer loop. The stability region
in kp-ki plane and its convex region are given in Fig. 3. The stability region has one cusp point
and two corner points. These points have described the convex region. Accordingly, convex
stability region is a triangle. Reading the coordinates of these points, centroid of the convex
region is calculated by using (14) and (15). Parameters of the centroid PI controller are kp=
kpc=0.107 and ki= kic=0.0393. Thus, parameters of the PI-PD controller are determined. They
are kd=0.3412, kf=0.439, kp=0.107 and ki=0.0393.
 0,12
Cusp point
of the stability region

0,1

Srability
boundary
locus
0,08
CONVEX
STABILITY
REGION
ki

0,06

Centroid
0.0393 of the convex
stability region

Corner points
of the stability
0,02 region

0
0 0.107 0,5
kp

Fig. 3. Centroid of the convex stability region of outer loop for Example 1.

Fig. 4 compares the unit step response of the proposed PI-PD controller with the design
proposed by Tan [30], Kaya [49]and Visioli [50]for set-point tracking. Parameters of the
controllers are given in Table 1.

Table 1. Controller parameters used for Example 1.

Controllers Proposed PI-PD Tan PI-PD Kaya PID Visioli PID
kd=0.3412, kf=0.439, kd=0.20, kf=0.413 kp=0.6217, ki=0.0722, kp=0.652, ki=0.0789,
Parameters
kp=0.107 ki=0.0393 kp=0.07, ki=0.03 kd=0.5352 kd=0.6309

Settling time, overshoot, rise time, peak time and IAE values are used as evaluation criteria.
Values of these criteria of the proposed method and the other designs are given in Table 2.
According to the overall evaluation criteria, the proposed controller exhibits the best
performance. The proposed controller provides the lowest settling time, overshot and IAE.
While PID controllers response quickly, other performances are unsatisfactory. Furthermore,
It is seen from controller performances that PI-PD control structure has a superiority over the
PID structure.

3
Proposed PI-PD
Tan PI-PD
Kaya PID
Visioli PID
2.5

2
Output (y)

1.5

0.5

0
0 5 10 15 20 25 30 35 40
Time (s)

Fig. 4. Step responses for Example 1

Table 2. Values of the performance criteria for Example 1

Methods Settling Time (s) Overshoot (%) Rise Time (s) Peak Time (s) IAE
Proposed PI-PD 16.04 2.13 3.77 8.63 4.87
Tan PI-PD 20.26 6.24 4.64 10.62 5.89
Kaya PID 17.70 174.16 0.45 5.01 8.57
Visioli PID 24.15 191.65 0.32 4.00 9.18

Example 2: Consider

(19)

which is an unstable second order plus time delay transfer function. The model is also studied
by Kaya and Atherton [44]. For this example, the computed stability region and its convex
region are given in Fig. 5. Parameters of the centroid PD controller are kf= kfc=5.69 and kd=
kdc=5.46.
 16

Cusp point
of the stability region
14

12

Srability
boundary
10 locus CONVEX
STABILITY
REGION
8
kf

Centroid
5.69 of the convex
stability region

4
Corner points
of the stability
region
2

0
-2 0 2 4 5.46 8 10 12
kd

Fig. 5. Centroid of the convex stability region of inner loop for Example 2.

By embedding the centroid PD controller parameters (kdc=5.46 and kfc=5.69) into the inner
loop and reducing the inner loop to one transfer function, 5., 6. and 7. steps of the method are
operated respectively. The stability region in kp-ki plane and its convex region are given in
Fig. 6. Parameters of the centroid PI controller of the outer loop are kp= kpc=3 and ki= kic=5.3.
Thus, parameters of the PI-PD controller are determined. They are kd=5.46, kf=5.69, kp=3 and
ki=5.3.
 Cusp point
16
of the stability region

14

Srability
12 boundary
locus
CONVEX
10
STABILITY
REGION
ki

Centroid
5.3 of the convex
stability region
4 Corner points
of the stability
region
2

0
-2 0 2 3 4 6 8 10
kp

Fig. 6. Centroid of the convex stability region of outer loop for Example 2.

Fig. 7 compares the step response of the PI-PD controller with the design proposed by Kaya
and Atherton [44], which has kd=0.535, kf=3, kp=1.398 and ki=0.793, and the PID controller
designed by means of the WGC method [24], which has kp=1.398, ki=0.793 and kd=0.535 for
set-point tracking. The closed loop responses of these controllers for unit step function are
illustrated in Fig. 7. The performance criteria of the proposed controller and the other
controllers are given in Table 3. It is clearly seen that the proposed PI-PD controller is better
than the others. WGC based PID controller has surprisingly performed better than PI-PD
controller designed by Kaya and Atherton in terms of IAE and settling time performances.

Table 3. Values of the performance criteria for Example 2

Methods Settling Time (s) Overshoot (%) Rise Time (s) Peak Time (s) IAE
Proposed PI-PD 4.40 10.14 0.83 2.08 1.25
Kaya and Atherton PI-PD 18.23 17.52 1.33 3.11 2.79
WGC PID 7.45 115.11 0.21 1.00 1.64
 2.5
Proposed PI-PD
Kaya and Atherton PI-PD
WGC PID

1.5
Output (y)

0.5

0
0 2 4 6 8 10 12 14 16 18 20
Time (s)

Fig. 7. Step responses for Example 2

Example 3: Consider
(20)

where, a [0.5 1.5] and b [0.5 1.5]. It is an unstable first order plus time delay transfer
function with uncertain parameters. There are four Kharitonov transfer functions which are
G11(s), G12 (s), G21 (s) and G22 (s). The plants are given in (21), (22), (23) and (24)
respectively.

(21)

(22)

(23)

(24)
Firstly, for the inner loop, stability regions of the Kharitonov plants are graphically obtained
in kd –kf plane. For this purpose, the procedure in step 1 is severally operated for each of the
plants. The common stability region for all plants is given in Fig. 8. Accordingly, the common
convex stability region has one cusp point and two corner points. The convex region is shown
in Fig. 9. Parameters of the centroid PD controller are computed from here as kf= kfc=3.8027
and kd= kdc=0.2158.

18

16

14

12

10
kf

Common Stability Region

6 for Kharitonov Plants

0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
kd

Fig. 8. Common stability region of inner loop for the Kharitonov plants.
 5.5
Cusp point
of the stability region

Srability
boundary
locus
4.5 COMMON CONVEX
STABILITY REGION
FOR KHARITONOV
PLANTS
kf

Centroid
3.8027
of the convex
stability region

3.5 Corner points

of the stability
region

3
-0.3 -0.2 -0.1 0 0.1 0.2158 0.3 0.4 0.5 0.6 0.7
kd

Fig. 9. Centroid of the common convex stability region of inner loop for the Kharitonov
plants.

After this stage, common stability regions are obtained for the outer loop. The regions are
given in Fig. 10. The common region consists of one junction point and two corner points.
Convex stability region and its centroid are given in Fig. 11. Accordingly, parameters of the
centroid PI controller for the outer loop are kp= kpc=0.6027 and ki= kic=2.6357. Thus,
parameters of the robust PI-PD controller are determined. The robust controller parameters
are kd=0.2158, kf=3.8027, kp=0.6027 and ki=2.6357.
 30

25

20
ki

15

10

Common Stability Region

for Kharitonov Plants

0
-4 -2 0 2 4 6 8 10 12 14
kp

Fig. 10. Common stability region of outer loop for the Kharitonov plants.
 8
Junction point
of the stability regions
7

COMMON CONVEX
5 Srability STABILITY REGION
boundary FOR KHARITONOV
locus PLANTS
ki

3
Centroid
2.6357
of the convex
stability region
2

Corner points
of the stability
1 region

0
-1 -0,5 0 0.6027 1 1,5 2
kp

Fig. 11. Centroid of the common convex stability region of outer loop for the Kharitonov
plants.

The unit step responses of four Kharitonov transfer functions for the robust PI-PD controller
designed according to the centroid of the common convex stability region are shown in Fig.
12. The robust controller has capableness in spite of the uncertainty of the system parameters.
It stabilizes all Kharitonov plants and exhibits acceptable robust performance.
 1.6
for G11
for G12

1.4 for G21

for G22

1.2

1
Output, y

0.8

0.6

0.4

0.2

0
0 5 10 15 20 25 30 35 40
Time (s)

Fig. 12. Step responses for Kharitonov plants

Example 4: In this example, a chemical reactor with unstable dynamical behavior is

considered. System transfer function is given in (25). The closed loop system has a long time
delay of 20 s.

(25)

The model is also studied by Shariati et al [51]. PI and H∞ based PI controller have been
designed in [51]. In addition, a PID controller is tuned using the WGC design procedure
proposed in [24]. For this example, the computed stability region and its convex region are
given in Fig. 13. Parameters of the centroid PD controller are kf= kfc=1.0646 and kd= kdc=6.66.
 3

Cusp point
of the stability region

2.5

Srability
2 boundary
locus CONVEX
STABILITY
REGION
f

1.5
k

Centroid
1.0646 of the convex
stability region

Corner points
of the stability
0.5 region

0
-30 -20 -10 0 6.66 20 30 40
kd

Fig. 13. Centroid of the convex stability region of inner loop for Example 4.

By embedding the centroid PD controller parameters (kdc=6.66 and kfc=1.0646) into the inner
loop and reducing the inner loop to one transfer function, 5., 6. and 7. steps of the method are
operated respectively. The stability region in kp–ki plane and its convex region are given in
Fig. 14. Parameters of the centroid PI controller of the outer loop are kp= kpc=0.4289 and ki=
kic=0.0158. Thus, parameters of the PI-PD controller are determined. They are kd=1.0646,
kf=6.66, kp=0.4289 and ki=0.0158.
 0.05
Cusp point
of the stability region
0.045

0.04

0.035 Srability
boundary
locus
0.03 CONVEX
STABILITY
REGION
i

0.025
k

0.02

Centroid
0.0158 of the convex
stability region

0.01 Corner points

of the stability
region
0.005

0
-1 -0.5 0 0.4289 1 1.5
kp

Fig. 14. Centroid of the convex stability region of outer loop for Example 4.

Fig. 15 compares the unit step responses of the PID controller tuned by WGC method and PI
controllers with the design proposed by Shariati [51]. Parameters of the controllers used in the
example 4 are given in Table 4.

Table 4. Controller parameters used for Example 4.

Controllers Proposed PI-PD WGC PID Shariati et al H∞ PI Shariati et al PI
kd=1.0646, kf=6.66 kp=1.262, ki=0.01319, kp=0.9435, kp=0.6554,
Parameters
kp=0.4289, ki=0.0158 kd=7.00 ki=0.006 ki=0.0064

The performance values of the proposed controller and others are given in Table 5. The
simulation results clearly show that the proposed PI-PD controller has performed better than
the other controllers. It is interesting that the H∞ PI controller exhibits a better performance
than the PI controller. Because the designs based on H∞ theory have generally offered a trade-
off between robustness and performance.
 Table 5. Values of the performance criteria for Example 4
Methods Settling Time (s) Overshoot (%) Rise Time (s) Peak Time (s) IAE
Proposed PI-PD 166.34 0.55 37.57 83.47 48.55
WGC PID 162.60 76.43 13.68 67.10 76.87
Shariati et al. H∞ PI 310.14 90.11 21.33 93.50 118.22
Shariati et al. PI 868.62 112.51 28.23 128.10 246.04

2.5
Proposed PI-PD
WGC PID
Shariati et al. H PI
Shariati et al. PI

1.5
Output (y)

0.5

0
0 100 200 300 400 500 600 700 800 900 1000
Time (s)

Fig. 15. Step responses for Example 4

Example 5: The proposed method can be applied to all processes whose stability region can
be calculated as a closed region and provides acceptable controller performance. A stable
resonant system given in (26) has been studied in this example in order to illustrate the
capability of the method.

(26)
This system is also studied in [30]. If the proposed design procedure is run for (26), PI-PD
controller parameters are obtained as kf=14.47, kd=7.51, kp=10.1033 and ki=51.47. In addition,
the controller parameters for the design of PID based WGC and design of Tan are given
kp=26.29, ki=76.2, kd=10 and kf=3.9, kd=3, kp=1.40, ki=5.0, respectively. Fig. 16 compares
closed loop responses of the controllers for the unit step input. Numerical values of the
performances of the controllers are given in Table 6. The proposed PI-PD controller is the
best controller in terms of settling time and IAE performances. However, overshot
performance with 12 percent can be a problem. Furthermore, despite the oscillatory response,
the WGC based PID controller has surprisingly performed better than PI-PD controller
designed by Tan in terms of settling time and IAE performances.

2.5
Proposed PI-PD
Tan PI-PD
WGC PID

1.5
Output (y)

0.5

0
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Fig. 16. Step responses for Example 5

Table 6. The values of the performance criteria for Example 5

Methods Settling Time (s) Overshoot (%) Rise Time (s) Peak Time (s) IAE
Proposed PI-PD 1.60 12.09 0.29 0.75 0.4681
Tan PI-PD 1.81 1.43 1.10 2.21 0.9944
WGC PID 2.99 125.30 0.07 0.35 0.8112
Example 6: In this example, an experimental application to a magnetic ball levitation system
is done by using the proposed method. Transfer function of the experimental system is given
in (27). This system is also studied by Swain et al [52]. Details of the experimental system
are seen in [52,53].
(27)

If the proposed design procedure is operated for (27), PI-PD controller parameters are
obtained as kf=-0.6958, kd=-0.6958, kp=-0.0536 and ki=-2.3370. PID controller parameters
tuned by Swain et al for the same experimental system are kp=-2.3973, ki=-2.7749 and kd=-
0.4451. Fig. 17 compares closed loop responses of the controllers for a step input with
magnitude of 0.0055 m. It is clearly seen that the proposed PI-PD controller is better than the
PID controller designed by Swain et al. There is no overshot in the response of the proposed
PI-PD controller. Settling time performance of the controller is approximately measured as
1.1 s.

0
Proposed PI-PD
Swain et al PID

-0.002

-0.004
Ball position (m)

-0.006

-0.008

-0.01

-0.012

-0.014
0 5 10 15 20 25 30
Time (s)

Fig. 17. Real time step responses for Example 6

The applications have generally showed that the proposed method is effective in a broad
range of unstable processes. Thus, the proposed PI-PD controller design procedure has a big
potential to use in practice.
5. Conclusion
In this paper, a graphical method has been presented in order to tune the PI-PD controller
parameters for the unstable systems. The method is based on the stability boundary locus
which can be easily obtained by equating the real and the imaginary parts of the characteristic
equation to zero and uses a new concept named convex stability region. Centroids of the
described convex regions in controller parameters plains are used for the controller design.
The method uses only the coordinates of a few special points pertaining to the curve
determining the stabilizing controller parameters region. Therefore, the computational load of
the proposed design procedure is very low. For the engineers in the industry, simplicity of the
method is an important advantage. Extension of this method to unstable systems with
uncertain parameter has been given using the Kharitonov theorem. The extended version of
the proposed controller to parameter uncertain systems can work out varying time delay
systems such as systems with quantization and under the network based environment.
Numerical examples and an experimental application have been provided to show the benefit
of the results. The results clearly show the value of the method proposed. In the near future,
nonlinear model based gain scheduling PI-PD controller and fractional order PI-PD controller
can be studied to enlarge the proposed method. Furthermore, the effects of the gain-phase
margin tester can be investigated.

ACKNOWLEDGEMENT
Reviewers made critical and constitutive comments about this paper. Therefore, I did major
changes and additions in the revised version. I would like to thank the reviewers for their
comments.

References

[1] Shamsuzzoha M. Closed-loop PI/PID controller tuning for stable and integrating process
with time delay. Ind Eng Chem Res 2013;52:12973–92.
[2] Jeng JC, Lin SW. (2012) Robust proportional-integral-derivative controller design for
stable/integrating processes with inverse response and time delay. Ind Eng Chem Res 2012;
51:2652–65.
[3] Rahimian MA, Tavazoei MS. Application of stability region centroids in robust PI
stabilization of a class of second-order systems. Transactions of the Institute of Measurement
and Control 2012, 34(4):487-98.
[4] RP Sree, Srinivas MN, Chidambaram M. A simple method of tuning PID controllers for
stable and unstable FOPTD systems. Computers and Chemical Engineering 2004; 28: 2201–
18.
[5] Shamsuzzoha M, Skogestad S. The setpoint overshoot method: a simple and fast closed-
loop approach for PID tuning. J. of Process Control 2010; 20(10): 1220-34.
[6] Wang YG, Shao HH. Optimal tuning for PI controller. Automatica 2000, 36, 147-52.
[7] Majhi S. On-line PI control of stable processes. Journal of Process Control 2005; 15: 859-
67.
[8] Mudi RK, Dey C. Performance improvement of PI controllers through dynamic set-point
weighting. ISA Transactions 2011; 50: 220-30.
[9] Mudi RK, Dey C, Lee TT. An improved auto tuning scheme for PI controllers. ISA
Transactions 2008; 47: 45-52.
[10] Zhuang M, Atherton DP. Automatic tuning of optimum PID controllers. IEEE
Proceedings-Part D:Control Theory and Applications 1993; 140(3): 216-24.
[11] Åström KJ, Hägglund T. Automatic Tuning of Simple Regulators with specification on
phase and amplitude margins. Automatica 1984, 20(5), 645-51.
[12] Yeroglu C, Onat C, Tan N. A new tuning method for PIλDμ controller, Proc. International
Conference onElectrical and Electronics Engineering ELECO 2009; 2: pp. 312-6.
[13] Pai NS, Chang SC, Huang CT. Tuning PI/PID controllers for integrating processes with
deadtime and inverse response by simple calculations. Journal of Process Control 2010; 20(6):
726–33.
[14] Ho WK, Lim KW, Xu W. Optimal Gain and Phase Margin Tuning for PID Controllers.
Automatica 1998; 34(8): 1009–14.
[15] Ho WK, Hang CC, Cao LS. Tuning of PID Controllers Based on Gain and Phase Margin
Specifications. Automatica 1995; 31(3): 497–502.
[16] Mahmooddabadi MJ, Taherkhorsandi M, Talebipour M, Castillo-Villar KK, Adaptive
robust PID control subject to supervisory decoupled sliding mode control based upon genetic
algorithm optimization, Transactions of the Institute of Measurement and Control 2015;
37(4): 505-14.
[17] Chidambaram M, Sree RP. A simple method of tuning PID controllers for
integrator/dead-time processes. Computers and Chemical Engineering 2003; 27(2): 211–5.
[18] Luyben WL. Process design and control identification and tuning of integrating processes
with deadtime and inverse response. Ind. Eng. Chem. Res 2003; 42: 3030–5.
[19] Shamsuzzoha M. Closed-loop PI/PID controller tuning for stable and integrating process
with time delay. Ind Eng Chem Res 2013;52:12973–92.
[20] Srivastava S, Pandit VS. A PI/PID controller for time delay systems with desired closed
loop time response and guaranteed gain and phase margins, Journal of Process Control 2016;
37: 70-7.
[21] Onat C. A new concept on PI design for time delay systems: Weighted Geometrical
Center, Int J Innov Comp Inf Cont 2013; 9(4): 1539-56.
[22] Kaya I, Atiç S. PI controller design based on generalized stability boundary locus.
International Conference on System Theory, Control and Computing (ICSTCC) 2016; 20: 24–
8.
[23] Atic S, Cokmez E, Peker F, Kaya I. PID controller design for controlling integrating
processes with dead time using generalized stability boundary locus. IFAC PapersOnLine
2018; 51(4): 924-9.
[24] Ozyetkin MM, Onat C, Tan N. PID tuning method for integrating processes having time
delay and inverse response. IFAC PapersOnLine 2018; 51(4): 274-9.
[25] Atherton DP, Majhi S. Limitation of PID controller. Proc. of Amer. Contr. Conf. 1997;
3843-47.
[26] Majhi S. Relay feedback identification and controller design. Ph.D. Thesis 1999;
University of Sussex, Brighton UK.
[27] Kwak HJ, Sung SW, Lee I. On-line process identification and auto tuning for integrating
processes. Ind. Eng. Chem. Res. 1997; 36: 5329-38.
[28] Kaya I. PI-PD controller design for control of unstable and integrating processes, ISA
Trans 2003; 42: 111-21.
[29] Kaya I. PI-PD controllers for controlling stable processes with inverse response and dead
time, Electr Eng 2016; 98: 55-65.
[30] N. Tan, Computation of stabilizing PI-PD controllers, International Journal of Control
Automation and Systems 7(2) (2009) 175-84.
[31] Park JH, Sung SW, Lee I. An enhanced PID control strategy for unstable processes.
Automatica 1998; 34: 751-6.
[32] Atherton DP, Majhi S. Tuning of optimum PIPD controllers, Proc. of Int. Conf.
Control’98 1998; Caimbra Portugal: 549-54.
[33] D.P. Atherton, A.F. Boz,, Using standard forms for controller design, Proc. of
Control'98, (1998) Swansea UK 1066-71.
[34] I. Kaya, Relay Feedback Identification and Model Based Controller Design, PhD thesis,
University of Sussex (1999) Brighton, U.K.
[35] Onat C, Kucukdemiral IB, Sivrioglu S, Yuksek I. LPV model based gain-scheduling
controller for a full vehicle active suspension system. Journal of Vibration and Control 2007;
13(11): 1629–66.
[36] Onat C, Kucukdemiral IB, Sivrioglu S, Yuksek I, Cansever G. LPV gain-scheduling
controller design for a nonlinear quarter-vehicle active suspension system. Transactions of the
Institute of Measurement and Control 2009; 31(1): 71–95.
[37] Onat C, Sahin M, Yaman Y, Prasad E, Nemana S. Design of an LPV based fractional
controller for the vibration suppression of a smart beam. CANSMART CINDE IZFP 2011; 2-
4 November 2011, Montreal-Quebec-Canada.

[38] Onat C, Sahin M, Yaman Y. Gain scheduling H∞ control of a smart beam with parameter
varying, VIII ECCOMAS Thematic Conference on Smart Structures and Materials SMART
2017; 5-8 June 2017, Madrid, Spain; 453-63.
[39] Bhattacharyya SP, Chapellat H, Keel LH, Robust Control: The Parametric Approach,
Prentice Hall, 1995.
[40] Ghosh BK. Some new results on the simultaneous stabilization of a family of single
input single output systems. Syst. Contr. Lett 1985; 6: 39-45.
[41] Hollot CV, Yang F. Robust stabilization of interval plants using lead or lag
compensators. Syst. Contr. Lett 1990; 14: 9-12.
[42] Barmish BR, Holot CV, Kraus FJ, Tempo R. Extreme points results for robust
stabilization of interval plants with first order compensators. IEEE Trans. on Automat. Contr.
1993; 38: 1734-35.
[43] Ho MT, Datta A, Bhattacharyya SP. Design of P, PI and PID controllers for interval
plants, Proc. of Amer. Contr. Conf.; Philadelphia USA, June 1998.
[44] Kaya I, Atherton DP. Simple analytical rules for PI-PD controllers to tune integrating
and unstable processes. International Control Conference 2006; Glasgow, Scotland, United
Kingdom, 30th August to 1st September.
[45] Ozyetkin MM. A simple tuning method of fractional order PIλ-PDμ controllers for time
delay systems. ISA Transactions 2018; 74: 77-87
[46] Kharitonov VL. Asymptotic stability of an equilibrium position of a family of systems of
linear differential equations. Differential Equations 1979; 14: 1483-85.
[47] Onat C, Hamamci SE, Obuz S. A practical PI tuning approach for time delay systems,
IFAC 2012, Boston, USA.
[48] Onat C. WGC based robust and gain scheduling PI controller design for condensing
boilers. Advances in Mechanical Engineering 2014; 6: 1-13.
[49] Kaya I. Simple and optimal PI/PID tuning formulae for unstable time delay
processes,10th International Conference on Electrical and electronics Engineering ELECO, 30
Nov-2 Dec 2017, Bursa, Turkey. 847-51
[50] Visioli A. Optimal tuning of PID controllers for integral and unstable processes. Control
Theory and Applications IEE Proc. 2001; 148(2):180-84.
[51] Shariati A, Taghirad HD, Fatehi A. A neutral system approach to H∞ PD/PI controller
design of processes with uncertain input delay, Journal of Process Control 2014; 24: 144-57.
[52] Swain SK, Sain D, Mishra SK, Ghosh S. Real-time implementation of fractional order
PID controller for a Magnetic Levitation Plant. Int. J. Of Electronics and Communications
(AEÜ) 2017; 78: 141-56.
[53] Sain D, Swain SK, Mishra SK. Real-time implementation of robust set-point weighted
PID controller for Magnetic Levitation System. Int. J. On Electrical Engineering and
Informatics 2017; 9(2): 272-83.