See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/283107686

Design of a Fuzzy Logic process Controller for flow applications and

implementation in series tanks Pilot Plant

Conference Paper · July 2015

DOI: 10.1109/IIC.2015.7150784

CITATIONS READS

2 230

2 authors:

Tareq Aziz Hasan AL-Qutami Rosdiazli Ibrahim

Universiti Teknologi PETRONAS Universiti Teknologi PETRONAS
8 PUBLICATIONS   76 CITATIONS    154 PUBLICATIONS   1,061 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Fractional-order controllers for Process Plants View project

Vector Born, GIS, Dengue View project

All content following this page was uploaded by Tareq Aziz Hasan AL-Qutami on 17 January 2016.

The user has requested enhancement of the downloaded file.

 Design of a Fuzzy Logic Process Controller for Flow
Applications and Implementation in Series Tanks
Pilot Plant

Tareq Aziz Hasan AL-Qutami, Rosdiazli Ibrahim

Department of Electrical & Electronic Engineering
Universiti Teknologi PETRONAS
Seri Iskandar, Perak, Malaysia
[email protected], [email protected]

Abstract— Flow control is essential in many industrial PID is a linear-type controller and hence is only efficient for a
applications such as chemical reactors, heat exchangers and limited operating range when used to control non-linear
distillation columns. Most industrial processes exhibit processes[5]. But most of the industrial processes exhibit
nonlinearities and inherit dead time, which limit the performance nonlinear characteristics and inherit long dead-time such as
of conventional PID controllers. This Project aims to design and
flow, PH and temperature [5, 6]. Moreover, parameters tuning
implement Fuzzy Logic Controller (FLC) for flow control
applications. The objective is to overcome problems inherited is sensitive and has to be changed with changing process
with conventional PID control scheme such as handling conditions and requires expertise in order to be tuned properly
unpredictable disturbance, non-measurable noise as well as [7]. These reasons led to the development of advanced
further improve the transient state and steady state response nonlinear controllers such as MPC, Fuzzy logic and Neural
performance. The proposed control scheme is implemented in Networks based controllers.
Flow Control and Calibration Pilot Plant. The design is done The objectives of this project are to design advanced controller
using Matlab/Simulink software package and is connected to the for flow process using Fuzzy Logic principles then implement
Pilot Plant through USB-type DAQ cards. Simulation and the designed strategy onto a Pilot Plant followed by
implementation results showed that the developed controller has
comparison between current PID control strategy and the
less overshoot, good control performance, better disturbance
handling ability, great robustness and is more flexible and proposed advanced controller in terms of control performance
intuitive to tune. It is expected that this advanced controller and handling disturbances.
improves efficiency and productivity of industrial processes
through proper handling of any disturbance or noise and Using limited experimental data to generate a reliable empirical
increase the robustness of controller actions. model is a challenging task especially for MIMO and nonlinear
processes. Traditionally, statistical models are developed using
Index Terms—PID, Fuzzy Logic control, flow rate control, these experimental data. However it is still hard to choose the
process control, disturbance handing, series tanks flow process,
most appropriate model structure and parameters.
Fuzzy Logic Controller tuning.
Mathematical modeling for some processes like flow is not
easy and therefore model based control is not possible [8]. It
I. INTRODUCTION seems that new methodologies for efficient control and
Precise measurement of flow is essential in the process modeling are required. This is where Fuzzy Logic and Neural
industries to ensure proper rate of supply of ingredients during Networks come in [7].
mixing and blending of materials and to avoid any high rate
that may cause temperature or pressure to reach dangerous
levels, or spilling of materials. Furthermore, Flow II. FUZZY LOGIC
measurement is used in custody transfer and is essential for Fuzzy Logic is a mathematical system that can deal with
accounting and payment purposes [1, 2] . Flow measurement is uncertain and ambiguous information which is complex to
classified into volumetric flow rate (Liters/min, ft3/s, gallons/s), calculate by means of conventional mathematics [8]. Fuzzy
mass flow rate (fluid mass per unit time like kg/h) and flow Logic uses Fuzzy sets in continuous interval [0, 1] rather than
velocity (kg/h or Ib/h) [1]. two-valued logic (0, 1) or crisp set. ‎Fig. 1. shows the general
The most widely used controllers in industry are PID configuration of Fuzzy Logic system. Fuzzification interface
controllers in eq. (1) because of their simple structure and converts crisp input to a linguistic variable using membership
satisfactory performance as long as they are properly tuned [3, functions while Defuzzification interface converts the fuzzy
4]. (1) output into crisp number using membership functions.
Common Defuzzification methods are Centroid of Area, Mean continuously stirred tank reactor. Prior knowledge divides the
of Maximum and Largest of Maximum. process into Fuzzy regions that are used to propose Fuzzy
control scheme, besides secondary variable is used to identify
the operating region. As well as several other Fuzzy controllers
were developed for several processes involving disturbance,
nonlinearity and long dead-time and proved to have
improvements over the conventional controller.
Hybrid Fuzzy-PID controllers were developed for several
processes such as polymerization process which was proven to
be more reliable than conventional PID, batch time and the
energy consumption were reduced for batch distillation column
process [15]. Simple tuning mechanisms for PID controllers
were developed using Fuzzy logic for SISO and MIMO
Fig. 1. Basic configuration of Fuzzy Logic system systems and were tested on coupled double tank plant where
water level is the controlled variable and results showed better
Rule base is set of if-then statements that represent expert’s
performance and rejection of disturbance associated with tank
linguistic knowledge while Decision-making unit converts
nonlinearity than conventional PID [16]. Lastly, a Fuzzy Logic
Fuzzy inputs to Fuzzy outputs using the Rule Base and
incremental-type controller was implemented to regulate flow
determines how rules are activated and combined using
rate between two water tanks using 9 MFs 2 inputs Mamadani
operators such as AND, OR, and NOT operators. Fuzzy
FIS [17].
Inference System (FIS) combines rule base and decision-
making unit and most common FIS are Mamadani [9] and IV. FUZZY LOGIC CONTROLLER STRUCTURE
Takagi and Sugeno[10]Fuzzy Inference Systems.
Different structures of FLCs have been developed, most
General FLC block diagram is shown in ‎Fig. 2. The inputs to
common ones are:
most of FLCs are the error and error rate of change; the outputs
1. FPD Controller: Fuzzy PD controller in ‎Fig. 3. has
go through scaling functions and then fed to the final element
advantage of less overshoot due to the use of derivative term
(e.g. Control Valve).
however it is sensitive to noise and an abrupt change of the
reference causing a derivative kick [18].

Fig. 2. Fuzzy Logic Controller block diagram

III. FUZZY LOGIC IN PROCESS CONTROL Fig. 3. FPD controller

Fuzzy Logic principles have been used in industry and have
managed to overcome problems inherited in conventional PID 2. FPI (Incremental) Controller: Fuzzy PI controller or
controllers. For example NOX reduction in a power plant and incremental controller in ‎Fig. 4. will ensure zero steady state
plasma-etching are non-linear processes that were modeled by error. The change in control signal cu is added to the current
Lou and Huang [8] using Fuzzy logic through limited control signal, A disadvantage is that it cannot include D-
experimental data. action well [4, 18].
Takagi and Sugeno who developed Fuzzy controllers for water
cleaning process and a converter in a steal-making process
which solved the decomposition and realization problems of
the controllers [11]. Another application is in boiler systems of
power plants where level control of water inside drum is
crucial, and PID control fails in the presence of process Fig. 4. Incremental controller
disturbance while Fuzzy logic controller performed better in
disturbance handling [3]. Another Fuzzy controller was also 3. FPD+I Controller: Fuzzy PD+I controller in ‎Fig. 5. is a
developed for temperature control in oil steam distillation and three input controller that combines FPD controller with I
results achieved were better than PID in terms of step point controller. The integral error is computed as
tracking [12]. (2)
A Type-2 Fuzzy controller was developed for aerobic alcoholic
fermentation nonlinear bioprocess characterized by It offers all advantages of PID; nonetheless it inherits its
parameter uncertainty proved to reduce the negative effects of drawbacks of integration windup and derivative term kick.
system parameters with a minimum computational load [13].
a multiregional FLC was proposed by Qin and Borders [14]
for control of nonlinear processes such as pH titration in
 Fig. 5. FPD+I controller

Fig. 8. Proposed Fuzzy Logic controller structure

V. METHODOLOGY
Flow of project activities is shown in ‎Fig. 6. The Fuzzy PD part will ensure faster reaction and dampen the
oscillations and its effect can be controlled through GU gain
that controls the amount of derivative action contribution to the
output. A Proportional gain that is dependent on the error has
been added and it can be used to increase the initial kick of the
controller however increasing this value may lead to harder
tuning of other parameters and bigger overshoots.
B. Fuzzy Inference System (FIS) and Membership Functions
Both Sugeno-Takagi FIS and Mamadani FIS were considered
Fig. 6. Project Flowchart initially however Sugeno-Takagi type FIS has been chosen for
this controller since it is computationally efficient and can be
A. Tools and Software Required optimized during implementation. ‎Fig. 9. and ‎Fig. 10. show
 Computer. the membership function of input variables.
 2 of USB-1208FS cards or 1 USB-1208FSPLUS card. NB
1
NM NS ZE PS PM PB

 Simulink/MATLAB software package. 0.8

Degree of membership

 PcA SimExpert Flow Control and Calibration Process 0.6

Mobile Pilot Plant. 0.4

 250 ohm resistors to convert 4-20mA into 1-5V. 0.2

General setup Block diagram is shown in ‎Fig. 7.

-30 -20 -10 0 10 20 30
E

Fig. 9. Membership functions for input E

NB NM NS ZE PS PM PB
1

0.8
Degree of membership

0.6

0.4

0.2

-30 -20 -10 0 10 20 30

Fig. 7. Implementation Setup diagram CE

Fig. 10. Membership functions for input CE

VI. FUZZY LOGIC CONTROLLER DESIGN
Error and Change of Error have 7 Gaussian-type MFs and
A. Structure ranges are set to [-30 30], the Pilot Plant operating limits. The
The Fuzzy Logic Controller (FLC) for this project has Fuzzy output membership functions for the output are discrete values
PD +Fuzzy PI structure shown in ‎Fig. 8. and equation 3. The rather than fuzzy membership functions.
inputs are Error (E) and Derivative of error (CE) while the Output membership Functions are: Positive Big (PB), Positive
output of FIS is change in control action (CU). GE and GCE Medium (PM), Positive Small (PS), Zero (ZE), Negative Small
are input gains used to map input ranges to the range of FIS. (NS), Negative Medium (NM), Negative Big(NB). All outputs
The Fuzzy PI part (incremental controller) accumulates CU are distributed equally in the range of [-0.4 0.4].
and ensures zero steady state error; it has GCU gain to control C. FIS Rules
the accumulation rate and it is dependent on the process dead ‎ ABLE I. and ‎TABLE II. show the 45 FIS rules developed for
T
time and rise time. the Fuzzy Logic Controller and the different operating regions,
they are based on the behaviors of PID controller.
 TABLE I. FLC RULES AND RELATIONSHIP BETWEEN INPUTS If we change the tuning to have a more aggressive response in
AND OUTPUT the expense of overshoot in order to see how the controller
E/CE NB NM NS ZE PS PM PB will handle overshoot we get the responses shown in ‎Fig. 13.
NB NB NB NB NB NB NB NB and ‎Fig. 14. due to step point change and random steps
NM NB NB NM NM NM NS NS respectively. We can observe that FLC still has superior
NS NM NS NS NS ZE PS PS performance, faster settling time and less overshoot than PID.
ZE NM NS ZE ZE ZE PS PM SP
PS NS NS ZE PS PS PS PM 20 PID
FLC
PM PS PS PM PM PM PB PB

flow rate (l/min)

PB PB PB PB PB PB PB PB 15

10
TABLE II. OPERATING REGIONS AND APPROPRIATE ACTIONS

I: Starting up, response to the V: Error changing very fast, 5

Set Point change adjust output a bit if necessary.
II: Error is not changing, VI: reached equilibrium 0
0 100 200 300 400 500 600
change output accordingly 100 samples/s
III: Moving along, Maintain VII: Error is zero but changing Fig. 13. Response of PID and FLC to step change (aggressive tuning)
output
IV: getting worse, reverse VIII: Error is zero but changing SP
40
output somewhat. insignificantly, action maintained. PID
35 FLC
30

flow rate (l/min)

VII. SIMULATION RESULTS 25

20
Simulations were based on the FODT model in Eq. 3, a result
from empirical modeling on the Pilot Plant. 15

10

0
0 1000 2000 3000 4000 5000 6000
100 samples/s
‎ ig. 11. shows the results of a step change of 20 l/min for PID
F Fig. 14. Response of PID and FLC to Random steps (aggressive tuning)
and FLC, FLC clearly performs better in both transient and
steady state. Both controllers were tuned such that no We conclude from the simulation results that FLC has better
overshoot happens. ‎Fig. 12. shows response due to several performance and that is due to the nonlinearity inherited
random changes to ensure FLC has same performance within the design and the proper handling of error rate and its
throughout the whole range and using different step size. influence on the change in control action.
VIII. EXPERIMENTAL RESULTS
SP
20
PID This section aims to validate the controller design through
FLC real system implementation in the Pilot Plant which is two
15
flow rate (l/min)

interacting series tanks with the objective of controlling the

fluid transfer rate between the two tanks. Results are shown in
10
‎Fig. 15. and ‎Fig. 16.
5
SP
40 PID
0 FLC
0 100 200 300 400 500 600
100 samples/s
flow rate (l/min)

35
Fig. 11. Response of PID and FLC to step change

40 SP 30
PID
35
FLC
30
25
flow rate (l/min)

25

20 1000 1100 1200 1300 1400 1500 1600 1700

100 samples/s
15
Fig. 15. Response of PID and FLC to step change
10

0
0 1000 2000 3000 4000 5000
100 samples/s
Fig. 12. Response of PID and FLC due to random steps
 40
SP X. CONTROLLERS TUNING PROCEDURE
PID
FLC Tuning stage usually requires experience and some intuition;
flow rate (l/min)

35
however there are some guidelines that can be followed in
30 order to accelerate tuning:
25
1. All gains are initialized to zero.
2. Identify the range of the system and modify GE to scale to
20 the range of FIS universe [-30 30], the maximum E must not
1000 1500 2000 2500 3000 3500 4000 4500 5000 5500
100 samples/s
go beyond limit of universe, i.e. |Emax*GE|=|Universemax|.
Fig. 16. Response of PID and FLC due to random steps 3. Increment GCU until you have zero steady state error and
the response starts to oscillate, good initial value is Ki/1.2
FLC has generally better tracking performance, faster rise from conventional PID gains.
time and settling time, less overshoot compared to PID. 4. Increment GCE until the oscillation is removed or the
‎TABLE III. Summarize the results from both simulation and response starts to undershoot, good initial value is Kp/2 from
implementation. conventional PID gains.
5. Alternate between incrementing GCU and GCE until there
TABLE III. RESPONSE CHARACTERISTICS FOR CONTROLLERS is overshoot followed by undershoot.
6. Alternate between incrementing GU and decrementing GCE
Mode Simulation Implementation
until you get the best rise time, minimum overshoot and no
Parameter/Controller PID FLC PID FLC undershoot.
Settling Time (s) 3.9 1.9 3.8 1.4 7. To improve the performance further, increment GEU
Overshoot (%) 0.7 0.5 0.7 0.6 slightly to get better rise time, and then repeat steps 4-6 to
Rise Time (s) 0 0 12.4 2.2 remove any overshoots but this time consider either
incrementing or decrementing each gain depending on the
response, this is done through observation of the effect of each
IX. DISTURBANCE HANDLING gain and intuition.
In this segment controller performance is examined when XI. CONCLUSION
subjected to sudden disturbance using the manual valves in the
pilot plant. Changing Hand Valves 101 and 121 to 20% Fuzzy Logic-based controllers exhibit far better control
opening will reduce the flow rate (MV), then waiting until the performance and stability than conventional PID controllers,
controller adjusts the Control Valve to make PV equal to MV, they are flexible and can handle any sudden changes or
then changing both Valves back to 100% as fast as possible disturbances on the system. It is expected that the efficiency of
will create a disturbance in MV that controller has to handle. the process and ability to handle any disturbances will be
‎Fig. 17. shows the results. increased dramatically. The designed system can then be
transferred into real industrial processes through the use of
FLC disturbance response modern PLCs which support Fuzzy Logic; such PLCs are
MV
37 PV
available nowadays from different manufacturers such as
36 Siemens and Omron. Furthermore, Fuzzy Logic Controller
PV(l/min), MV(%)

35 can perform well despite presence of process nonlinearities,

34
operation variability and measurements noise.
33
32
31
XII. FUTURE WORK
30 It is recommended to expand current work and develop a level
1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 controller that is cascaded with the flow controller and test how
100 Samples/sec
PID disturbance response it performs in the cascade system. Furthermore, an
PV investigation is required to develop a comprehensive and clear
MV
36 procedure to tune FLC; this will guarantee optimum
PV(l/min), MV(%)

performance and stability of the controller. Another

34
recommendation is to take this project to the next phase, which
32 is implementation in industrial PLC or DCS system and test
whether it will still have the same performance and explore
30
ways of optimization to implement the controller with
1000 1500 2000 2500 3000 3500 4000 4500 5000 5500
100 Samples/sec minimum resources. Last recommendation is to expand the
Fig. 17. PID and FLC Controllers disturbance response designed controller into other processes such as temperature
and pressure and test whether this controller can be used to
FLC shows superior handling of disturbance both in terms of control these processes with minimum changes to the structure
recovery time (5.8s) and undershoot (1.3%) while PID takes or FIS rules.
14.8s to recover and has 2.5% undershoot percentage.
 REFERENCES [10] M. Sugeno, "Fuzzy measures and fuzzy integrals: a survey,"
[1] T. L. M. Bartelt, Instrumentation and process control vol. 1: Fuzzy Automata and Decision Processes, , pp. 89-102,
Thomson Delmar learning, 2007. 1997.
[11] T. Takagi and M. Sugeno, "Fuzzy identification of systems
[2] A. K. Pandey, "Low Cost Sensor for Low Flow-Rate
and its applications to modeling and control," Systems,
Measurement," in Sensors for Industry Conference, 2005,
2005, pp. 6-8. Man and Cybernetics, IEEE Transactions on, vol. SMC-15,
[3] H. F. VanLandingham and N. D. Tripathi, "Knowledge- pp. 116-132, 1985.
[12] N. N. Mohammad, N. Kasuan, M. H. F. Rahiman, and M. N.
based adaptive fuzzy control of drum level in a boiler
system," in Southcon/96. Conference Record, 1996, pp. Taib, "Steam temperature control using fuzzy logic for
steam distillation essential oil extraction process," in
454-459.
[4] K. J. Astrom. Control System Design [Online]. Available: Control and System Graduate Research Colloquium
http://www.cds.caltech.edu/~murray/courses/cds101/fa0 (ICSGRC), 2011 IEEE, 2011, pp. 53-58.
[13] C. Bartolomeo and G. Mose, "Type-2 fuzzy control of a
2/caltech/astrom.html
bioreactor," in Intelligent Computing and Intelligent
[5] G. Zaidner, S. Korotkin, E. Shteimberg, A. Ellenbogen, M.
Systems, 2009. ICIS 2009. IEEE International Conference
Arad, and Y. Cohen, "Non linear PID and its application in
process control," in Electrical and Electronics Engineers in on, 2009, pp. 700-704.
[14] S. J. Qin and G. Borders, "A multiregion fuzzy logic
Israel (IEEEI), 2010 IEEE 26th Convention of, 2010, pp.
000574-000577. controller for nonlinear process control," Fuzzy Systems,
IEEE Transactions on, vol. 2, pp. 74-81, 1994.
[6] M. Yukitomo, Y. Baba, T. Shigemasa, M. Ogawa, K.
Akamatsu, and S. Amano, "A model driven PID control [15] A. M. F. Fileti, A. J. B. Antunes, F. V. Silva, V. Silveira Jr, and
system and its application to chemical processes," in SICE J. A. F. R. Pereira, "Experimental investigations on fuzzy
2002. Proceedings of the 41st SICE Annual Conference, logic for process control," Control Engineering Practice,
vol. 15, pp. 1149-1160, 9// 2007.
2002, pp. 2656-2660 vol.4.
[16] M. B. B. Pinto, J. G. R. Mota, and O. M. Almeida, "PID self-
[7] Z. Dingxue and X. Zhen, "Fuzzy PID control in the process
control," in Instrumentation & Measurement, Sensor adjustable fuzzy logic MIMO case: Method and
application," in Industry Applications (INDUSCON), 2010
Network and Automation (IMSNA), 2012 International
9th IEEE/IAS International Conference on, 2010, pp. 1-6.
Symposium on, 2012, pp. 587-591.
[17] E. Pathmanathan and R. Ibrahim, "Development and
[8] H. H. Lou and Y. L. Huang, "Fuzzy-logic-based process
implementation of Fuzzy Logic Controller for Flow Control
modeling using limited experimental data," Engineering
Applications of Artificial Intelligence, vol. 13, pp. 121-135, Application," in Intelligent and Advanced Systems (ICIAS),
2010 International Conference on, 2010, pp. 1-6.
4/1/ 2000.
[18] J. Jantzen, "Tuning Of Fuzzy PID Controllers," Technical
[9] E. H. Mamdani and S. Assilian, "An Experiment in Linguistic
University of Denmark, Lyngby, DENMARK 98-H 871 30
Synthesis with a Fuzzy Logic Controller," International
Journal of Human-Computer Studies, vol. 51, pp. 135-147, Sep 1998.
8// 1999.

View publication stats