2017 IEEE 2nd International Conference on Automatic Control and Intelligent Systems (I2CACIS 2017), 21 October 2017, Kota

Kinabalu, Sabah, Malaysia

Modeling and Comparative Study of PID Ziegler

Nichols (ZN) and Cohen–Coon (CC) tuning method
for Multi-Tube Aluminum Sulphate Water Filter
(MTAS)

Ahmad Aftas Azman1, Mohd Hezri Fazalul Rahiman*2, Nurul Nadia Mohammad3, Mohd Hezri
Marzaki4, Mohd Nasir Taib5, Mohd Fozi Ali6
1-5
Faculty of Electrical Engineering,
6
Faculty of Civill Engineering,
Universiti Teknologi MARA,
40450 Shah Alam, Selangor, Malaysia
[email protected]

Abstract—Floods are a natural phenomenon that disperse of light in a water, furthermore the amount of these
frequently occur in Malaysia. Flooding will cause water light dispersing are influence from particles (clay, sand and
to become cloudy or turbid. In this situation, the supply silt), plankton, microbes, algae and other substances [2].
of clean drinking water is the most affected component. There are various techniques used to treat drinking water,
Therefore, many researchers have designed a water the most popular is to use a water filter. Correspondingly, a
filter system to overcome these problems. One of them is convenient water filter are easy to install, portable and
used chemical as purifying agen, or known as coaguland lightweight are important. However, if the turbidity of raw
and flocculation process. In this study researcher used water is too high, water filters life span will decrease due to
Aluminium Sulphate Al (SO ) as a coagulant agent, and frequent backwashing activities.
2 4 3
To overcome this problem, the turbidity of raw water may
Multi-tube aluminum sulphate water filter (MTAS) pilot
be reduced, by using the flocculation and coagulant process
plan as flocculation process. Thus, the MTAS pilot plan
technique. Therefore, the aluminum sulfate as a coagulant
will be control by open loop controller and a discrete
agent are mixed with raw water for flocculation process [3].
plan modelling transfer function be developed in light of,
However, consumer health should also be taken into
the open loop controller output. Subsequently, PID
consideration, indeed the use of over dose aluminum sulfate
simulation controller be implemented based on Ziegler
will affect human health [4]. Thus, the proper dosage of
Nichols (ZN) and Cohen–Coon (CC) tuning method.
aluminum sulfate is important for water purification.
Conclusively, the robusness and efficiencies of PID
Therefore, in this study, researcher used Aluminium
controller is evaluated, and a comparative study has
Sulphate Al (SO ) as a coagulant agent and a water filter
done. Finally, researcher proposed the best PID tunning 2 4 3
controller that will be adopted in real time process pilot plan has been developed named as Multi-tube
control for further investigation. aluminum sulphate water filter (MTAS) plan. World Health
Organization (WHO) has a standard been outlined to
Keywords- Turbidity, Aluminium Sulphate Al (SO ) , determine the water quality parameter (WQP) for each
2 4 3
water usage [1]. Hence, turbidity is the featured parameter
MTAS, PID, Ziegler Nichols and Cohen–Coon
for drinking water, the best turbidity level is 1
I. INTRODUCTION Nephelometric Turbidity Units (NTU) or below [5]. In
Malaysia, the turbidity level is set below than 5 NTU
River is a source of human development; furthermore, accordance to standard of Ministry of Health [6]. Various
river is an important source for irrigation and domestic used of control devices have been designed for water purification
of water for human consumption. Flood may occur; if the [7]. The popular controllers for water filter are ON/OFF
drainage system is incorrect planned and water resources controllers, nevertheless some researcher are used PID,
may affected. This scenario is becoming more severe as a neural network, fuzzy and SVM controller concept to
result of fresh water demand is higher than supply [1]. control the chemical dosage for water filter system
Flooding will cause water to become cloudy or turbid. [8],[9],[10]&[11].
Water turbidity can be measure by determine the amount of

978-1-5386-0846-3/17/$31.00 ©2017 IEEE 25

Authorized licensed use limited to: Universitas Brawijaya. Downloaded on January 22,2024 at 02:12:46 UTC from IEEE Xplore. Restrictions apply.
 2017 IEEE 2nd International Conference on Automatic Control and Intelligent Systems (I2CACIS 2017), 21 October 2017, Kota Kinabalu, Sabah, Malaysia

There are many previous studies of PID tuning gains to control in manual or automatic mode. Meanwhile, turbidity
reach preferred performance, and diverse PID controller sensor will measure the Nephelometric Turbidity Unit
tuning methods has proposed for process models, especially (NTU) level of water after filtering. Instead, water pump are
for the first order plus time (FOPDT), subsequently explain used to regulate the retention time for the coagulant and
the wide range of processes behavior. Two of the popular flocculent process, while proportional solenoid valve are
classical techniques are Ziegler-Nichols tuning and Cohen- used to control the Aluminium Sulphate Al (SO ) dossing
2 4 3
Coon tuning which still be implement in many industrial that will be mixed with artificial clay water.
applications [12]. This two method is easy and simple to
tune PID gains, however its performance is deficient in Meanwhile, figure 2 show the pilot plant for Multi-tube
nonlinear systems [13]. Generally, many review studies water filter (MTWF), where this plant is located at ‘Fakulti
show a good performance in linear systems; their Kejuruteraan Elektrik’ (FKE) Universiti Teknologi Mara.
performance is unpredictable and frequently deficient in While, in figure 3 shows the block diagram of Multi-tube
nonlinear systems. Hence, it is difficult to choose PID gains aluminum sulphate water filter (MTAS) pilot plant.
for nonlinear plants [14]. However, in this study, MTAS is Furthermore, this block diagram shows how the main
control using PID controller with Ziegler Nichols (ZN) and component is connected to each other’s. Conversely, water
Cohen–Coon (CC) tuning method. Then, from ZN and CC pump and proportional solenoid valve is an output of
tuning method, Matlab PID simulation will be implement, controller, while turbidity sensor is an input for controller.
and a step test, set point change and load disturbance test Aluminum staging is a Polyvinyl chloride (PVC) pipe, with
been performed. Finally, result will be evaluated and the 50mm diameter x 4’ long. These pipes are connected and
better tuning method will be used for a real time process. stacked vertically by five cylinders. While, aluminum
This paper is starting with simple explanations of flood, staging cylinder function is to trap the suspended particles at
turbidity and PID controller in section 1. Subsequently, in the bottom of the cylinder. Besides, its purpose is to
section 2 are the plant description. Section 3 is the separate the dissolved and suspended particles from the
explanation of Open loop followed by PID structure in water by coagulation and flocculation processes.
section 4. Section 5 is the Result and Discussion and lastly,
the paper is concluded in section 6.
II. PLANT DESCRIPTION
Main component for Multi-tube aluminum sulphate water
filter (MTAS) consist of computer, arduino mega board,
turbidity sensor module, proportional solenoid valve and
water pump. Conection of all these component is shown in
figure.1 bellow:

Figure 2: Multi-tube aluminum sulphate water filter (MTAS) Pilot Plant

DATA
ACQUISITION PLANT

WATER
OUTPU PUMP ALUM
Fig 1. Illustrations of Multi-tube water filter (MTWF). STAGING
PROCESS
COMPUTER VALVE
using
The computer is installed by MATLAB software that be MATLAB
ARDUINO
MEGA CARD
used to control and analyze the designed control system. For
this system arduino mega card is used as data acquisition TURBIDITY
INPUT
module acting as intermediary between the computer and SENSOR

the control panel. While the control panel used as a medium

between the arduino mega, water pump and proportional Figure 3: Block Diagram Multi-tube water filter (MTWF)
solenoid valve. In addition, the control panel allows system

26
Authorized licensed use limited to: Universitas Brawijaya. Downloaded on January 22,2024 at 02:12:46 UTC from IEEE Xplore. Restrictions apply.
 2017 IEEE 2nd International Conference on Automatic Control and Intelligent Systems (I2CACIS 2017), 21 October 2017, Kota Kinabalu, Sabah, Malaysia

Moreover, flooding will cause the supply of treated water ( )

= (1)
affected, while the flooding has become murky water. ( )
Therefore, the source of the flood water should be treated as
clean water. This could be resolved with coagulation Moreover, reaction curves will provide information
process in water filter treatment. Therefore, inorganic regarding proportional gain (Kp), time constant (τ) and dead
coagulants (e.g., aluminum and ferric salts) are used to time (θ) for plan transfer function modeling.
remove turbidity in water. Therefore, in this study
researcher used Aluminium Sulphate Al (SO ) as
2 4 3
coagulant agent. Likewise, turbidity is remove effectively
by using aluminum sulphate [15]. Instead, figure 4 show the
operating flow of Multi-tube aluminum sulphate water filter
(MTAS) to purify the water.

Figure 5: MTAS Simulink Open loop Control

Subsequently, in this study researcher setting a set point

for open loop controller is 256 (bit), with equivalent to 5V
that gives full power for water pump and proportional
solenoid valve. Meanwhile, turbidity sensor will collect the
data and send the data in 1023 bits and converted to NTU
values.
IV. PID STRUCTURE
Controller is the brain of a process control; there are
various types of control technique starting from P, PI, PD,
PID control to the advance control technique. About 95%
of proses control system are using PID controller, this is
because PID are easy to control and low maintenance,
beside this controller are robust. The PID terms come from
P (Proportional), I (Integral), D (Derivation) which
represents transient and steady-state response requirements.
This study will focused on PID controller, due to its
features and characteristic, further more due to the fact that
it have conceptual that easy to understand and simple
structure [17]. PID Conventional control algorithms are
most commonly used and well-established of controllers in
a water treatment process [18].
Figure 4: Multi-tube aluminum sulphate water filter (MTAS) Schematic
Diagram PID controller continuously calculates an error value
e(t) as the difference between a desired set point and
measured process variable. This PID controller will apply
the applied a correction based on proportional, integral,
III. OPEN LOOP CONTROL
and derivative terms [19]. The controller are attempts to
Open loop control has no effect or influence on minimize the error over a time by adjustment of a control
the control action of the input signal. Therefore, open-loop variable u(t), such as the position of a control valve, speed
system is defined by the output signal or condition is neither of motor pump, or flow of liquid, to a new value determine
measured nor “fed back” for comparison with the system set by equation 2.
point or input signal. Therefore, the open loop control for ( )
Multi-tube aluminum sulphate water filter (MTAS) is shown ( )= ( )+ ( ) + (2)
in figure 5. Furthermore, from the open loop controller will
Where Kp is proportional gain, Ti is integral time constant
produce the process reaction curves as shown in figure 7.
and Td the derivative time constant. Simulink PID block
The reaction curves process is probably most popular
controller in this study is shown in figure 6.
method to identify the dynamic model. Consequently, its
simple to perform and provide adequate models for many
applications. Indeed, first order plus dead time (FOPDT) is
restricted to this model, where X(s) is the input and Y(s) are
the output. Therefore, equation 1 is the expression of
modelling transfer function plant [16].

27
Authorized licensed use limited to: Universitas Brawijaya. Downloaded on January 22,2024 at 02:12:46 UTC from IEEE Xplore. Restrictions apply.
 2017 IEEE 2nd International Conference on Automatic Control and Intelligent Systems (I2CACIS 2017), 21 October 2017, Kota Kinabalu, Sabah, Malaysia

iii. Load Disturbance

An open-loop controller doesn’t actually play any role in

determining how the process reacts to a disturbance, so the
controller’s tuning is irrelevant when feedback is disabled
[23][24]. Hence, load disturbance test is set at process time
Figure 6: MTAS Simulink PID Control of 2500 sec as shown in figure 10; this test was performed
by monitoring a disturbance to a process and plan recovery
A. Tuning Rule time.
The controller tuning methods are classified into two main
V. RESULT AND DISCUSSION
categories:
i. Open loop methods Consequently, from the open loop controller result by using
ii. Closed loop methods the graphical calculation from reaction curves in figure 7;
parameter of ∆y, 0.63∆, t63%, 0.28∆, t28% has be
The open loop methods are operates in manual state, means determine below;
plant operates in uncontrolled condition. Therefore, closed
loop tuning techniques refer to methods that tune the ∆y = 96.2 – 80.27 = 15.93
controller during automatic state in which the plant is 0.63∆ = 90.3059 t63% = 445
operating in closed loop [19]. Furthermore, most popular 0.28∆ = 84.7304 t28% = 136
techniques for tuning rule are Ziegler Nichols (ZN) and
Cohen–Coon (CC) tuning method. A ZN tuning rule was The time constant (τ) , Dead time (θ); and proportional gain
introduced in the mid-20th century and is the beginning of Kp be calculate as shown in equation 3,4 and 5 below:
systematic tuning rules for PID control. Another method of
tuning PID controller tuning rule is CC that has been τ = 1.5(t63 – t28) (3)
introduce since 1950 [20]. The PID tuning rule is mention = 1771.5
in table 1 for ZN and CC.
Table 1: PID Tuning rule θ = t63% - τ (4)
= -341.5
Derivative
Proportional Integral Time △
PID Tuning Time = (5)
Gain Kp Constant, Ti
Constant Td △
1.2 = 0.088
ZN 2θ 0.5θ
MTWF-ZN 193.592 37 9.25 From the parameter of Kp, θ and τ, the MTAS pilot plan
. .
CC [ + 0.185] 2.5θ x
.
0.37θ x
. for PID ZN and PID CC tuning formula was calculate, then
MTWF-CC 219.398 9.096 6.8021 parameter of Kp, Ti and Td was determine as shown in table
1. By using Matlab software coding for ARX modeling, the
B. Simulation test for PID discrete transfer function for the process is shown equation
Therefore, in this study researcher had apply three type of 6.
testing technique, namely step test, set point change and
load disturbance test. Moreover this test was performed to ( ) . ^
= (6)
evaluate the robustness of controller performance. ( ) . ^

i. Step test

Step test is performed by supplying unit step function as

input in to a system[21], where a certain set point value will
be given in this study, researcher set at ‘100’ as an input 0.63∆
value for MTAS plan as shown in figure 8.
∆y

ii. Set Point Change

0.28∆
Set point change is where multiple stages of set input value
will be will be given to the process. For this study,
researcher has set three input point values that is ‘80, 90 and
100’ as a input for MTAS as shown in figure 9. A set point
Figure 7: MTAS Simulink Open loop Control
change input will pass through both the controller and the
process, even without any feedback. As a result, the
mathematical inertia of the controller combines with the Furthermore, from the modeling process for MTAS pilot
physical inertia of the process to make the process’s plan, the discrete PID simulation is proposed as shown in
response to a set point that has been manipulated [22]. figure 6. Then, researcher had compared the performance

28
Authorized licensed use limited to: Universitas Brawijaya. Downloaded on January 22,2024 at 02:12:46 UTC from IEEE Xplore. Restrictions apply.
 2017 IEEE 2nd International Conference on Automatic Control and Intelligent Systems (I2CACIS 2017), 21 October 2017, Kota Kinabalu, Sabah, Malaysia

of PID ZN and PID CC. Next, the simulated PID controller Table 3: PID set point change
have applied at three type of testing technique, that is step Rise
test, set point change and load disturbance test to evaluate Controller/ Settling Time %
Time
the robustness and performance of the controller. Tuning (sec) Overshoot
(sec)
As been mentioning earlier the set point value for step test PID ZN 80 180.5595 180.5595* 36.6167*
are 100 and the time of simulation is set at 5000 second, the
simulation output result for step test as shown in figure 8. If PID ZN 90 54.2068 200.1796 6.0825*
been compared the two types of PID controller that is PID PID ZN 100 79.8178 284.1217 5.7361*
ZN and PID CC, found that PID ZN is much better
performance of rise time and overshoot then PID CC. PID CC 80 160.2132* 851.5043 44.1022
However PID CC is better in settling time Then PID ZN, the
result for PID simulation step test as shown in table 2. PID CC 90 39.7467* 154.4156* 8.3022
PID CC 100 54.0589* 187.6569* 7.1335
‘*’ show better performance.
PID-ZN – 4*
PID-CC - 5*

For load disturbance test, at simulation time of 2500 sec a

load disturbance of 20 NTU has been setup as a load for the
process. The simulation output result for load disturbance
test as shown in figure 10. By comparing PID ZN with PID
CC for this test, the recovery time of PID ZN is much faster
than PID CC, therefore PID ZN has better performance than
PID CC as shown in table 2.
Figure 8: PID simulation Step test output

Table 2: PID step test

Rise
Controller/ Settling Time %
Time
Tuning (sec) Overshoot
(sec)

ZN 240.018 1027.6 24.2865*

CC 240.018 1232.9* 24.4272

‘*’ show better performance.

For set point change test, researcher has set three input point
values that are 80, 90 and 100 as a input for MTWF. This Figure 10: PID simulation load disturbance output
value will interval for 1500 sec at each set point value, the
simulation output result for set point change test as shown in Table 4: PID load disturbance test
figure 9. Form PID simulation result of rise time, settling Controller/
time and overshoot, found that PID CC (5*) has better Recover Time (sec)
Tuning
performance than PID ZN(4*) as shown in table 3.
PID-ZN 260*
PID-CC 282

VI. CONCLUSION
Conclusively, it has been proven from PID MTAS pilot
plan simulation result for step test, set point change and load
disturbance test, the PID ZN tuning method has better
performance compared to PID CC. This is due to PID ZN
had better performance on step test and load disturbance
test, whereas PID CC only performed better at set point
change. Therefore PID ZN tuning method will implement
Figure 9: PID simulation set point change output on real time mode for further study.

29
Authorized licensed use limited to: Universitas Brawijaya. Downloaded on January 22,2024 at 02:12:46 UTC from IEEE Xplore. Restrictions apply.
 2017 IEEE 2nd International Conference on Automatic Control and Intelligent Systems (I2CACIS 2017), 21 October 2017, Kota Kinabalu, Sabah, Malaysia

ACKNOWLEDGMENT [13] R. Dittmar, S. Gill, H. Singh, and M. Darby,

The authors would like to give and endless gratitude to “Robust optimization-based multi-loop PID
Malaysia Ministry of Higher Education for providing controller tuning: A new tool and its industrial
scholarship of ‘Hadiah Latihan Persekutuan’ (HLP) PhD application,” Control Eng. Pract., vol. 20, no. 4, pp.
programme, Faculty of Civil Engineering UiTM, Shah 355–370, 2012.
Alam for allowing the use of Environmental laboratory and [14] X. Li and W. Yu, “A systematic tunning method of
Faculty of Electrical Engineering UiTM, Shah Alam for the PID controller for robot manipulators,” in IEEE
support given in completing this study. International Conference on Control and
Automation, ICCA, 2011, pp. 274–279.
[15] M. G. Antov, M. B. Šćiban, and J. M. Prodanović,
REFERENCES “Evaluation of the efficiency of natural coagulant
[1] I. A. A. Ahmad Aftas Azman, Mohd Hezri Fazalul obtained by ultrafiltration of common bean seed
Rahiman, Norbaya Sidek, “Water Quality extract in water turbidity removal,” Ecol. Eng., vol.
Parameter : A Review On Dissolve Oxygen ( DO ) 49, pp. 48–52, 2012.
Control Method,” Int. J. Tech. Res. Appl. e-ISSN [16] T. Marlin, “Process Control,” pp. 1–17, 2000.
2320-8163, vol. 28, no. 28, pp. 98–102, 2015. [17] M. Korkmaz and H. Aydodu, mer Doan, “Design
[2] O. US EPA, “Turbidity,” and performance comparison of variable parameter
http://water.epa.gov/type/rsl/monitoring/vms55.cfm. nonlinear PID controller and genetic algorithm
[3] M. F. A. Ahmad Aftas Azman, Mohd Hezri Fazalul based PID controller,” INISTA 2012 - Int. Symp.
Rahiman, Mohd Nasir Taib, Ilyani Akmar Abu Innov. Intell. Syst. Appl., 2012.
Bakar, Norbaya Hj Sidek, “Characterization of [18] N. A. Wahab, R. Katebi, and J. Balderud,
Aluminum Sulphate for Control Application,” 2016 “Multivariable PID control design for activated
IEEE 7th Control Syst. Grad. Res. Colloq. (ICSGRC sludge process with nitrification and
2016), no. August, pp. 6–11, 2016. denitrification,” Biochem. Eng. J., vol. 45, no. 3, pp.
[4] H. Gala-Gorchev, “Aluminium in Drinking-water: 239–248, 2009.
Background document for development of WHO [19] M. Shahrokhi and A. Zomorrodi, “Comparison of
Guidelines for Drinking-water Quality,” World PID Controller Tuning Methods,” Proc. 8th Natl.
Heal. Organ., vol. WHO/SDE/WS, p. 2, 1998. Iran. Chem. Eng. Congr., pp. 1–12, 2002.
[5] W. H. Organization, “Guidelines for Drinking-water [20] M. H. Marzaki, M. H. A. Jalil, H. Shariff, and R.
Quality,” no. FOURTH EDITION, 2011. Adnan, “Comparative Study of Model Predictive
[6] “Drinking Water Quality Surveillance Programme - Controller ( MPC ) and PID Controller on
Ministry of Health.” [Online]. Available: Regulation Temperature for SSISD Plant,” pp. 136–
http://kmam.moh.gov.my/public-user/drinking- 140, 2014.
water-quality-standard.html. [Accessed: 04-Oct- [21] N. N. Mohammad, H. Marzaki, A. A. Azman, R.
2015]. Adnan, and S. N. Tajuddin, “Evaluation of Energy
[7] M. Lu and F. Peng, “Correlation Analysis of Consumption in Small Scale Distillation Pot based
Flocculation Detection and Control Parameters on Integral Control Signal for Real Time
Based on Multi-target Tracking,” 2010 Second WRI Implementation,” Int. J. Electr. Electron. Syst. Res.,
Glob. Congr. Intell. Syst., pp. 389–393, 2010. 2017.
[8] M. Onat and M. Dogruel, “Effluent turbidity control [22] “Disturbance-Rejection vs. Setpoint-Tracking
in direct filtration,” Proc. 2003 IEEE Conf. Control Controllers | Control Engineering.” [Online].
Appl. 2003 CCA 2003, vol. 2, pp. 1284–1289, 2003. Available: http://www.controleng.com/single-
[9] Y. Wang, B. Han, and Y. Zhang, “A Kind of article/disturbance-rejection-vs-setpoint-tracking-
Coagulant Dosing Control Model based on ISFLA- controllers/fe8b0cf50a2f0000c3b647c7e36326af.ht
SVM,” 978-1-4799-7016-2/15/2015 IEEE, pp. ml. [Accessed: 29-May-2017].
6437–6440, 2015. [23] Dale E Seborg, D. a Mellichamp, T. F. Edgar, and
[10] M. Onat and M. Dogruel, “Fuzzy Plus Integral F. J. Doyle, “Process Dynamics and Control,”
Control of the Effluent Turbidity in Direct Process Dyn. Control, p. 2010, 2010.
Filtration,” IEEE Trans. Control Syst. Technol., vol. [24] X. Yaun, C. Wang, J. Wang, L. Ma, and L. Zhang,
12, no. 1, pp. 65–74, 2004. “Simulation research of hydraulic turbine generator
[11] Z. Song, X. Song, Y. Zhao, and C. Liu, “Research under fuzzy PID control with load disturbance and
on predictive control of coagulant dosage based on 100% load rejection,” in Proceedings of 2015
neural network,” 2011 Seventh Int. Conf. Nat. International Conference on Fluid Power and
Comput., pp. 217–221, 2011. Mechatronics, FPM 2015, 2015, pp. 1028–1033.
[12] C. R. Madhuranthakam, A. Elkamel, and H.
Budman, “Optimal tuning of PID controllers for
FOPTD, SOPTD and SOPTD with lead processes,”
Chem. Eng. Process. Process Intensif., vol. 47, no.
2, pp. 251–264, 2008.

30
Authorized licensed use limited to: Universitas Brawijaya. Downloaded on January 22,2024 at 02:12:46 UTC from IEEE Xplore. Restrictions apply.