The 2016 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)

Design and Implementation of Water Level Control

Using Gain Scheduling PID Back Calculation
Integrator Anti Windup

Sony Cahya Pratama Erwin Susanto, Agung Surya Wibowo

Electronics Engineering School of Electrical Engineering
Telkom University Telkom University
Bandung Indonesia Bandung Indonesia
[email protected] erwinelektro;agungsw{@telkomuniverity.ac.id}

Abstract—Conventional PID controller is a simplest known system consist of many constraints such as nonlinearity, inertial
controller and has been used in almost all process industries for lag, time delay, time varying parameter and many more factors.
controlling process parameter at desired set point. However, the Gain scheduling is a technique that deals with nonlinear
weakness of the controller is largely depend on the parameter of processes, processes with time variations or situations where
the controlled object. As the dynamics process of a water tank is the requirements on the control change with the operating
nonlinear, it also has nonlinear behaviour between the input and conditions [1]. Since the process of nonlinear behavior
output. This paper demonstrates that Gain Scheduling PID with performing at various operating points, the objective of
Back Calculation Integrator Antiwindup enables to enhance the designing this controller is to obtain an enhanced control
response of automatic water level control system performance. A
performance at all operating points.
mathematical model of a first order tank system is considered
and simulated to determine fine tunes and schedules for the This paper organized as follows. Section 1 gives a brief
controller parameter based on trial and error experiment so as to introduction of the experiment. Section 2 explains a brief base
adapt with all operating points. The system performance theory of Gain Scheduling and Back Calculation Integrator
comparison of various setpoint given to the system is performed Anti Windup controller. Section 3 provides mathematical
to prove that the Gain Scheduling PID with Back Calculation model of the tank process in order to simulate and observe the
Integrator Antiwindup outpaces conventional PID controller. characteristic of the system. Section 4 describes the
experimental set up and its implementation of conentional PID
Keywords— conventional PID, non linear behaviour, Back
Calculation Integrator Anti Windup, Gain Scheduling.
and Gain Scheduling PID with back calculation integrator anti
windup controller. Section 5 shows the result the system
response of both controller. Conclusion is given in section 6.
I. INTRODUCTION
PID controller has been widely used for industrial control in II. SYSTEM DESCRIPTION
any application of controlling process since a long time ago.
According to the survey in 1989, 90% of process in industries A. Gain Scheduling Controller
uses the conventional PID controller [1]. The reason of why the In several conditions it is known that the dynamics of a
use of PID controller in the industry is wide spread is due to its process is varied with the operating condition. The parameter
simplicity and ease of returning online feature of PID controller changes in dynamics are what we called nonlinearities. Gain
[2-8]. Although many aspect of a control system can be Scheduling is a method that deals with nonlinear processes,
understood based on linear theory, some nonlinear effect must needed to change the control parameter with the operating
be accounted [1]. All actuator have limitations e.g a motor has conditions. Also, it is required to find measurable variable that
limited speed, a valve cannot be more than fully open, water correlated with changes in dynamic process, called scheduled
pump has maximum flow rate, etc. In some cases, such a variable. It can be the control signal or measured system
control with wide range of operating condition, it may happen response. Thus, gain scheduling can be understood as a
that control signal from controller reaches the actuator limits. feedback control system which the feedback gains are adapted
When this condition occurs, the feedback loop is broken by using feedforward compensation. When scheduling
because the actuator will stay at its limit independently of the variables are determined, the controller parameters are
process output. PID controller, which has integrating action calculated by using some design method e.g PID controller.
will continue to be integrated and the integral term may The controller then tuned for each operating condition.
become very large. It is then required a long period before it
returns to normal condition. One of the weaknesses of PID
controller is that it largely depends on the parameters of the
controlled object [5], whereas an industrial process control

978-1-5090-0744-8/16/$31.00 ©2016 IEEE 101

 The 2016 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)

cm. Water from reservoir is pumped by CBA pump with

specification of 5 bar maximum pressure, 1.6 L/min of
maximum cap flow, operating voltage of 12-15 volt, and 2.1 A
of maximum current. The level of the tank is measured by the
PING Parallax ultrasonic sensor having a measurement range
of (3-200) cm. Arduino UNO board is used as the main
controller combined with motor driver L298P to drive the
PWM of water pump.
Fig. 1. Block Diagram of Gain Scheduling controller

B. Back Calculation Integrator Anti Windup

Integrator windup may occur because the change of large
setpoint, large disturbance, or equipment error. It also
happened when the selectors are used to control one actuator.
One of several types of integrator windup is back calculation.
It works as follows : when the control signal work in actuator
limits, the integral is recalculated so that its new value gives
an output at the saturation limit. The system has extra
feedback path used to measured error signal (es) by measuring
the different between the output of the controller (v) and the Fig. 2. Experimental set up of water tank system
actuator output (u). The signal is zero if there is no saturation,
so it will not have any effect if the actuator does not saturate.

A. Gain Scheduling PID Controller with Back Calculation

C. Mathematical Model Integrator Antiwindup Design
The water level process is a single tank system, which
consist of a tank of uniform volume to which attached a flow
reistance R (valve). Assume :
q0 , the volumetric rate of flow through the resistance is
conneted to the head of tank.
q0 (t)= h / R (1)
The accumulation of mass in the tank is the difference between
the input flow rate and output flow rate, given by :
q(t) – q0(t) = A dh / dt (2)
Substitute (4) and (5) :
Fig. 3. Block Diagram of Gain Scheduling PID with Back Calculation
q(t) – h / R = A dh / dt (3) Integrator Antiwindup
Apply Laplace transformation,
Figure 4 shows the block diagram of the controller, consist of
Q(s) = 1 / R H(s) + As H(s) (4) three main part i.e conventional PID controller, added up by
Back Calculation Integrator Antiwindup to prevent instability,
The transfer function of system is
and Gain Scheduling to enhance the performance at any
H(s) / Q(s) = R / (RAs +1) (5) desired operating condition.
From the experiment it is observed :
A = 450 cm2, h = 30 cm, q =8.625 cm3/s, q0 = 31.275 cm3/s
Using the values of the parameter above, the transfer function
of the system derived as follows :
G(s) = H(s) / R(s) = 1 / (450 s + 1.0425) (6)

III. CONTROLLER DESIGN AND IMPLEMENTATION Fig. 4. Block Diagram of conventional PID

The experimental set up of the tank system is made up of the

crossbar-shape tank with the dimension of 15 cm x 30 cm x 30

978-1-5090-0744-8/16/$31.00 ©2016 IEEE 102

 The 2016 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)

Figure 5 is block diagram of conventional PID system. To increasing its value scaling by a factor of one until it oscillated.
implement the controller using Arduino, we need to change Then the integral parameter Ki is increased until steady
the system into discrete of PID form. Below is the equation of oscillations obtained by scaling it value by the factor of 10.
the discrete PID form : Next brings up the derivative parameter by scaling it by a
PID(z) = P(z) + I(z) + D(z) (7) factor of one as well as the proportional.
where,
P(z) = Kp . E(z) (8)
I(z) = Ki / (1-z-1) . E(z) (9)
D(z) = Kd . (1-z-1) . E(z) (10)
Let (1), (2), and (3) transform to the discrete form,
P(n) = Kp . e(n) (11)
I(n) = Ki e(n) + I(n-1) (12)
D(n) = Kd (e() – e (n-1)) (13)
Then equations above are represented to codes below,
P = Kp . error
I = Ki . Ʃerror Fig. 6. Simulation of the system
D= Kd . delta error
where, It may resulting some noises in the real implementation of
the system, but it can be used as the refference to understand
Ʃerror = present error + previous error
the characteristic of the system so that we can determining the
delta error = present error – previous error.
best tuned parameter. The tuned parameter then will be used as
scheduled variable as follows :
TABLE I. PID SCHEDULED PARAMETER
PID Parameter
Condition
Kp Ki Kd
13-16 cm 1 0.5 1

17-20 cm 1 0.4 2

21-23 cm 2 0.5 3

IV. RESULT AND ANALYSIS

Fig. 5. Block Diagram of Back Calculation Integrator Antiwindup
The designed controller is now examined for its ability to
Figure 5 shows a block diagram of PID controller with control the level of water tank system compared to the
back calculation integrator antiwindup. When the actuator conventional PID. The PID controller is tuned, where Kp, Kd,
saturates, the signal es is different from zero and the integrator and Ki as shown at table 1.
output is driven towards a value so that the integrator input
becomes zero. The integrator input is :
α . es + Ki . e (1)
es = -(Ki . e) . α (2)
where es = u – v.
Substitute (1) and (2) to find v
v = ulim + (Ki .e) . α (3)
ulim is the saturation valye of controled variable. Since es and
ulim have same sign , the value v is always greater than ulim so (a) System Response with Back Calculation Anti Windup
that it will prevent the integrator from winding up.

B. Tuning of PID controller

Controller tuning is a compromise between the requirement
desire for fast and stable control. In this section, the parameter
of PID controllers are derived from trial and error method. PID
parameters Kp, Ki, Kd, are changes so that it will be found the
best system response at all operation condition. The trial and
(b) System Response without Back Calculation Integrator Antiwindup
error method step through the system from proportional to
integral to derivative. Each coefficient of the PID controller is Fig. 7. System Response of PID Controller
set to zero, the proportional component is considered by

978-1-5090-0744-8/16/$31.00 ©2016 IEEE 103

 The 2016 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)

Figure 7 (a) represents the response of the PID controller with Figure 9 represents the graph and prove that Gain Scheduling
back calculation integrator antiwindup which has rise time of PID controller with back calculation integrator antiwindup
34.6 second. It is different with the system response without could track the setpoint at different operating points. Based on
back calculation integrator antiwindup which has rise time of the experiment, given from 9.18 cm to 16.18 cm, the response
156 second shown in figure 7 (b) . It is evident from the are obtained detailed below
response shown in figure 7 PID controller with back TABLE III. SYSTEM RESPONSE
calculation integrator antiwindup could track the setpoint with
minimal rise time and settling time than the conventional PID. Initial point Setpoint Rise Time Settling Time
9.18 cm 10.18 cm 23.2 second 27.2 second
10.18 cm 11.18 cm 21 second 29 second
11.18 cm 12.18 cm 24.8 second 28.2 second
12.18 cm 13.18 cm 28 second 38 second
13.18 cm 14.18 cm 29.4 second 33.2 second
14.18 cm 15.18 cm 39.8 second 44.2 second
15.18 cm 16.18 cm 40.8 second 46 second

V. CONCLUSION
Gain scheduled PID controller with Back Calculation
Integrator Antiwindup can adapt with water tank which has
Fig. 8. System Response of Conventional PID Controller and Gain nonlinear characteristic at various operating points. Gain
Scheduling PID at several setpoint Scheduling PID is performed better than the conventional PID
controller i.e it has faster rise time and settling time.
Figure 8 shows the system response of Gain Scheduling PID Controller proves that it is robust enough for changes in the
versus conventional PID at several setpoint. At (14-16) cm, process variables and keep it on the setpoint given by user.
(18-20) cm, and (21-22) cm the trajectory plotted at 400, 1200,
and 1600 second respectively. Table II below represents the
details of the system response. REFERENCES
TABLE II. COMPARISON OF TRANSIENT RESPONSE
[1] Astrom K.J.and T.Hagglund, ”PID controllers, Theory, design and
Transient Response tuning”, 2nd edition”, Instrument Society of America, 1995.
Setpoint Rise Time (s) Settling Time (s) Overshoot (cm) [2] Astrom K.J.and T. Hagglund,” The Future of PID Control”,
Control Engineering Practice,Vol 9, 2001, pp 1163-1175.
Conv GS PID Conv GS PID Conv GS PID
[3] Astrom K.J.and T.Hagglund, C.C.Hang, W.K.Ho, “Automatic Control
13 cm 129.6 125 136 135.6 0 0 and Adaptation for PID Controllers - A Survey,” Control
14 cm 162.4 155.2 167 164.6 0.6 0.01 Engineering Practice ,Vol 1(4), 1993, pp 699-714.
15 cm 184.8 182 200.2 194.8 0 0 [4] Pradeepkannan, D and Sathiyamoorthy, S., “Implementation of Gain
16 cm 253.4 242.8 264 246.8 0 0.05 Scheduled PID Controller for a Nonlinear Coupled Spherical Tank
Process.” International Journal of Mechanical & Mechatronics
19 cm 410.4 396.4 413.2 400.2 0.15 0.13
Engineering, Vol. 14, 2014, pp 93-98
20 cm 462. 4 449.8 471.8 450.6 0.06 0.06
[5] Liu, F.C, Liang, L.H, and Gao, J.J, “Fuzzy PID Control of Space
21 cm 572.4 522.4 583 525 0.03 0.01 Manipulator for Both Ground Alignment and Space Applications”,
22 cm 696.8 623.2 712.8 628.2 0 0 International Journal of Automation and Computing, 11(4), August
2014, pp 353-360.
[6] Al-Mshat H. A. and Sabri L. A., "Implementation of Fuzzy and PID
It is evident from the system response of Gain Scheduling PID Controller to Water Level System using LabView," International Journal
is better than the conventional PID at rise time, settling time, of Computer Applications, vol. 116(11), 2015, pp 6-10.
and overshoot. Based on the experiment at 8 different setpoints [7] Gondaliya, P and Patel, M. C, “Modelling and system identification of
given to the system, Gain Scheduling PID has better liquid level system.” International Journal of Science, Engineering and
performance. Technology Research (IJSETR), Vol 4(5), 2015, pp 1716-1719..
[8] A Usha, A. Patel, H. and Narayana, K. V. L., “Water Level Control
System using Self-Adaptive Fuzzy-PID Control.”International Journal of
Engineering Research & Technology, vol.3 (6), 2014, pp 1992-1997.

Fig. 9. System Response of Gain Scheduling PID at different setpoint

978-1-5090-0744-8/16/$31.00 ©2016 IEEE 104