DESIGNING INTELLIGENT PROCESS CONTROL SYSTEM USING FUZZY … · 2016-09-09 · DESIGNING...

DESIGNING INTELLIGENT PROCESS CONTROL SYSTEM USING FUZZY-NEURAL NETWORK Harmony Nnenna Nwobodo (Dept of Computer Engineering, Enugu State University of Science and Technology, Enugu) and Hyacinth Chibueze Inyiama (Dept of Electronic and Computer Engineering, Nnamdi Azikiwe University, Awka)

Abstract-This paper presents the design of a process control system using Fuzzy

Neural Network (FNN). A Fuzzy Neural Network (FNN) or NeuroFuzzy system is a learning

machine that finds the parameters of a Fuzzy System by exploiting approximation

techniques from neural networks. The FNN is based on a Fuzzy System which is trained

by a learning algorithm derived from Neural Network theory. A PID continuous water

process control system that lets water be at a set level as water flows in and out of the

tank was considered. In this paper, the hybrid of the Fuzzy Logic and Neural Network was

implemented to fine tune and optimize the Fuzzy output. The output of the Fuzzy Logic

alone and the output of corresponding Neural Network System were compared with the

output of the hybrid: Fuzzy Neural Network. It was observed that the Intelligent

Process Control Systems Using Fuzzy Neural Network gave a steep rise time of 4s

and stabilized after 6s. The Intelligent Process Control Systems using Fuzzy

Network rose sluggishly after an undershoot at 5s and stabilized after 9s while

the Intelligent Process Control Systems using Neural Network alone had a rise

time at 3s and stabilized after 9s. The Fuzzy Neural Network a hybrid of the two

performed better than the Fuzzy System or the Neural Network alone, considering

the rising time and stabilizing time. The Fuzzy Neural Network performed best

with a rise time of 1s and stabilizing time of 9s and showed a smoother overall

performance.

KEYWORDS: Neurofuzzy, fuzzy sets, Neuron, brain, connection weight, feed-forward, back-propagation, Membership Functions, Fuzzy Rule

International Journal of Scientific & Engineering Research, Volume 5, Issue 9, September-2014 ISSN 2229-5518

1202

IJSER © 2014 http://www.ijser.org

IJSER

(I) INTRODUCTION

Both Neural Networks and Fuzzy Systems have some things in common. They

can be used for solving a problems (e.g. pattern recognition, regression or

density estimation) if there does not exist any mathematical model of the given

problem. They solely do have certain disadvantages and advantages which

almost completely disappear by combining both concepts. Neural networks can

only come into play if the problem is expressed by a sufficient amount of

observed examples. These observations are used to train the black box. On the

one hand no prior knowledge about the problem needs to be given. On the other

hand, however, it is not straightforward to extract comprehensible rules from

the neural network's structure. On the contrary, a fuzzy system demands

linguistic rules instead of learning examples as prior knowledge. Furthermore

the input and output variables have to be described linguistically. If the

knowledge is incomplete, wrong or contradictory, then the fuzzy system must

be tuned. Since there is no formal approach for it, the tuning is performed in a

heuristic way. This is usually very time consuming and error-prone

Zhou and Quek [1] showed that Neural Networks with their remarkable ability

to derive meaning from complicated or imprecise data can be used to extract

patterns and detect trends that are too complex to be noticed by either humans

or other computer techniques. A trained neural network can be thought of as an


1203


IJSER

expert in the category of information it has been given to analyze. This expert

can then be used to provide projections given new situations of interest and

answer what if questions. Seema, Mitra and Vijay [2] designed a Neural

Network Tuned Fuzzy controller for Multiple Input Multiple output (MIMO)

system featuring a neural network based tuned fuzzy controller for controlling

the degree of freedom for MIMO systems. The coupling effect of the system

was added into the main fuzzy controller for each step to improve the

performance. A data set generated was partitioned into a set of clusters based on

subtractive clustering method. A fuzzy IF-THEN rule was then extracted from

each cluster to form a fuzzy rule base from which the fuzzy neural network is

designed. The neural network designed contains only three layers and the hybrid

learning algorithm was used to refine the parameters on fuzzy rule base.

Mitchell, Lopes, Fidalgo and McCalley [3] worked on using a Neural Network

to predict the Dynamic Frequency Response of a power system to an Under-

frequency Load Shedding Scenario. It showed a method to quickly and

accurately predict the dynamic response of a power system during an under

frequency load shedding. Emergency actions in a power system due to loss of

generation typically calls for under-frequency load shedding measures, to avoid

potential collapse due to lack of time in which to correct the imbalance via

other means. This is a slow and repetitious use of dynamic simulators so a

neural network was used to obtain a fast and accurate procedure during optimal

load shedding. Inyiama H.C and Nwobodo H.N [4] worked on Designing

Intelligent Process Control System using Fuzzy Logic Principles. Inyiama H.C


1204


IJSER

and Nwobodo H.N [5] also worked on Designing Intelligent Process Control

System using Neural Network. In this particular paper, the architecture of Fuzzy

Neural Network was discussed, followed by the design of Fuzzy Neural

Network for the process control system. Finally the Fuzzy Neural Network was

simulated using Proteus software. Fuzzy Neural Network when compared with

Fuzzy Logic Principles and Neural Network showed the best performance out

of the three techniques.

(II) Architecture of Fuzzy Neural Network

A fuzzy-neural system may be viewed as a 3-layer feed forward neural network

Fig 1.1. The first layer represents the input variables, the middle (hidden) layer

symbolizes the fuzzy rules and the third layer corresponds to the output

variables. Fuzzy sets are encoded as (fuzzy) connection weights. Some

approaches also use five layers where the fuzzy sets are encoded in the units of

the second and fourth layer, respectively. However, these models can be

transformed into three-layer architecture.

FIG 1.1 THREE LAYER FEED FORWARD FNN

INPUT LAYER

OUTPUT LAYER

RULE LAYER


1205


IJSER

A fuzzy-neural system can always be interpreted as a system of fuzzy rules (i.e.

before, during and after learning). It is also possible to create the system out of

training data from scratch and it is also possible to initialize it by prior

knowledge in form of fuzzy rules. The learning procedure of a Fuzzy Neural

system takes the semantical properties of the underlying Fuzzy System into

account. This results in constraints on the possible modifications applicable to

the system parameters. A Fuzzy Neural System approximates an unknown

function that is partially defined by the training data. The fuzzy rules encoded

within the system represent vague samples, and can be viewed as prototypes of

the training data.

Two different kinds of Fuzzy Neural Networks (FNN) were discussed in this

paper: They are Cooperative FNN and Hybrid FNN [6]

i) Cooperative Fuzzy-Neural Network

Figure 1.2 Cooperative Fuzzy Neural Networks

COOPERATIVE FUZZY NEURAL NETWORK

FUZZY SYSTEMS

FUZZY SYSTEMS

ERROR COMPUTATION

FUZZY SYSTEMS

FUZZY SYSTEMS

ERROR COMPUTATION


1206


IJSER

In Cooperative Fuzzy Neural Network, both Artificial Neural Network (ANN)

and Fuzzy System work independently from each other. The ANN tries to learn

the parameters from the Fuzzy System. This can either be performed offline or

online while the Fuzzy System is applied. Figure 1.2 depicts four different kinds

of Cooperative Fuzzy Neural Networks. The upper left fuzzy neural network

learns fuzzy set from given training data. This is usually performed by fitting

membership functions with a neural network. The fuzzy sets are then

determined offline. They are then utilized to form the fuzzy system by fuzzy

rules that are given (not learned) as well.

The upper right Fuzzy-Neural System determines fuzzy rules from training data

by a neural network. Here also, the Neural Networks learn offline before the

fuzzy system is initialized. The rule learning is usually done by clustering on

self-organizing feature maps. It is also possible to apply fuzzy clustering

methods to obtain rules.

In the lower left Fuzzy-Neural model, the system learns all membership

function parameters online, i.e. while the fuzzy system is applied. Thus initially

fuzzy rules and membership functions must be defined beforehand. Moreover,

the error has to be measured in order to improve and guide the learning step.

The lower right Fuzzy-Neural model determines rule weights for all fuzzy rules

by a neural network. This can be done online and offline. A rule weight is

interpreted as the influence of a rule. They are multiplied with the rule output.

The authors argue that the semantics of rule weights are not clearly defined.

They could be replaced by modified membership functions. However, this could


1207


IJSER

destroy the interpretation of fuzzy sets. Moreover, identical linguistic values

might be represented differently in dissimilar rules.

ii) Hybrid Fuzzy-Neural Network

Hybrid Fuzzy Neural Network is homogeneous and usually resembles Neural

Networks. Here, the Fuzzy System is interpreted as special kind of neural

network. The advantage of such hybrid FNN is its architecture, since both

Fuzzy System and Neural Network do not have to communicate any more with

each other. They are one fully fused entity. These systems can learn online and

offline. Figure 1.3 shows such a Hybrid FNN, here the rule base of a Fuzzy

System is interpreted as a Neural Network. Fuzzy sets can be regarded as

weights whereas the input and output variables and the rules are modeled as

neurons. Neurons can be included or deleted in the learning step. Finally, the

Figure 1.3 Hybrid Fuzzy Neural Networks

HYBRID FUZZY NEURAL

NETWORK

DETERMINE ERROR

SYSTEM

IF

THEN

System state

Control output


1208


IJSER

neurons of the network represent the fuzzy knowledge base. Obviously, the

major drawbacks of both underlying systems are thus overcome.

In order to build a fuzzy controller, membership functions which express the

linguistic terms of the inference rules have to be defined. In fuzzy set theory, no

formal approach to define these functions. Any shape (e.g., triangular,

Gaussian) can be considered as membership function with an arbitrary set of

parameters. Thus the optimization of these functions in terms of generalizing

the data is very important for fuzzy systems. Neural networks can be used to

solve this problem by fixing a distinct shape of the membership functions, say

triangular. The neural network must optimize their parameters by gradient

descent. Thus, besides the information about the shape of the membership

functions, training data must be available as well.

(III) The Fuzzy Neural Network Design

Fuzzy Neural Networks (FNN) combines fuzzy logic with Artificial Neural

Networks. In this work, the input and output variables processed by the fuzzy

logic was fed into the neural networks for learning. (This step is called

fuzzification). The neural network then takes the fuzzified input and output

scales to derive a model which is converted back to the original input and output

scales (defuzzification). Then the “defuzzified” output becomes input to the

process under control. The FNN shown in Fig 1.4 has four feed –forward layers.

The first layer represents the input variables for liquid level, temperature and

pressure respectively, the second represents their membership sets, the third

layer represents the fuzzy rules and the fourth layer represents the defuzified


1209


IJSER

output for the three variables under consideration. The rule layer produces

seventy-five rules altogether, twenty-five rules for each of the variables and

gives three outputs at a time, one for each of the three variables and that

satisfies the output requirement at a particular instance.

Fig 1.5 shows the block diagram of the Fuzzy-Neural control system running on

PC that is interfaced to the process under control via an INTEL 8286

Figure 1.4 The Fuzzy Neural controller of tank water level

MEMBERSHIP

FUNCTION

LAYER

Vo Vop Hc

INPUT LAYER

OUTPUT LAYER

FUZZY RULE LAYER

HL T P


1210


IJSER

Bidirectional Buffer. Just like in the neural controller, when the signal line B is

active, the buffer sends information to the process under control via the output

port latch. Similarly, when the signal line A is active, the feedback signals from

the process are input via the input port latch to the bidirectional buffer and from

thence to the PC. The process under Fuzzy-neural control has a valve that

controls the inflow of the liquid into the tank and another one that controls the

outflow from the tank. The heater increases the temperature of the liquid while

the stirrer is used to ensure that the liquid is of uniform temperature. The control

system tries to keep the liquid level in the tank constant within the set-point for

level. This is achieved by fuzzy-neural control by adjusting the inflow rate and

outflow rate dynamically as appropriate. Four sensors were also used: 1) to

sense the liquid level in the tank and 2) to sense the outflow rate from the tank,

3) to sense the temperature and 4) to sense the pressure of the system. The four

sensor outputs are selected one at a time via 4X1 analog multiplexer. The

selected signal is first amplified and then converted to digital pattern via an

analog to digital converter. The digitalized sensor output are latched by the

input port latch and forwarded to the fuzzy-neural control system via the

bidirectional buffer.


1211


IJSER

Outflow rate sensor

FUZZY-NEURAL

CONROL SYSTEM

RUNNING ON PC

Intel 8286 Bidirectional Buffer

Analog MU

X 1 out of 4

Analog Digital Converter

INPU

T

PORT

OU

TPUT PO

RT LATCH

Outflow

Rate

Sensor

Level sensor interface

Valve

Control

ON/OFF

VALVE

PROCESS UNDER FUZZY-NEURAL CONTROL

Stirrer

Level sensor

Temperature

Sensor

Pressure sensor interface

Heater

Control

Heater

Amplification stage

Set point for level

V

mV

A

B

Figure 1.5 Block diagram of The Intelligent Process Control System Using Fuzzy-Neural Network


1212


IJSER

IV THE INDUSTRIAL SCENARIO

It is always necessary to perform a real-time simulation of any control system

before the circuit is constructed, to ensure the workability of that system. In this

work Proteus software package was used for the simulation.

The real time simulation of the control system used as a typical industrial

scenario is activated by clicking the play button. When this is done the liquid

from the upper tank flows to the lower tank. The valve at the water inlet of the

upper tank is being controlled based on the rate of the outflow from the lower

tank. This is done to make sure that the set-point of the process under control is

maintained. The heater heats the liquid while the stirrer ensures that a uniform

liquid temperature is maintained. Fig 1.6 shows the industrial scenario of the

process control in Proteus environment. The three computer system interfaces

shown display the variables under control (liquid level/outflow rate, pressure

and temperature). It should be noted that when Proteus software is used to

implement a real-time simulation, not all the system components are placed on

the layout. Some are placed in the sub sheet for example the controller and some

components are not visible but are there by default e.g. reset circuit,

bidirectional buffer, power, the interfaces, analog to digital converter, the


1213


IJSER

multiplexer etc; while some are taken care of by the control program. In Proteus

design, this is called “referencing”.

rxd vcc

gnd

vcc

gnd

flow0

liq0

a0a1a2a3a4a5a6a7b0b1b2b3b4b5b6b7rxdga8ga9v1 v0

flow0liq0

ga7ga6ga5ga4ga3ga2ga1ga0c7c6c5c4c3c2c1c0

a6 a6

a5 a5

a4 a4

a3 a3

a2 a2

a1 a1

a7 a7

a0 a0

b7 b6 b5 b4 b3

b1

b0

b2

ga9

ga0

ga8ga7ga6ga5ga4ga3ga2ga1

ga0ga1ga2ga3ga4

ga5ga6

ga7ga8ga9

P2.0P2.1pretempP2.a

buzrxd1txd1

txd3

rxd3

txd3rxd3

P3.atxd2

tssrxd2

txd2

rxd2

txd1

rxd1

SUB?

CCT001

RXD

RTS

TXD

CTS

P2.0P2.1P2.2P2.3P2.4P2.5P2.6P2.7ga0ga1ga2ga3ga4ga5ga6ga7buzliq0

f low0v0

P0.0P0.1P0.2P0.3P0.4P0.5P0.6P0.7P1.0P1.1P1.2P1.3P1.4P1.5P1.6P1.7txdga8ga9v1

24.0

3

1

VOUT 2

43.0

3

1

VOUT 2

12345678

2019181716151413

910

1211

U2

LED-BARGRAPH-GRN

12345678

2019181716151413

910

1211

flow

0gnd

P2.0P2.1

?

?

pretempP2.a

buzrxd1txd1

buz

P2.0

P2.1

Level Sensor Over FlowRate Sensor

CollectorTank

ON/OFF ValveInterface

ValveControl

Process UnderNeuro Fuzzy

Control

Neuro fuzzy Controller Of Tank Water Level

Computer System Interface

Sound(Alarm) Sub System

SyStem Sub Circuit

RXD

RTS

TXD

CTS

36.0

3

1

VOUT 2

67.0

3

1

VOUT 2

pre

temp

P3.a txd3rxd3txd2

rxd2 tss

RXD

RTS

TXD

CTS

RXD

RTS

TXD

CTS

?

?

P2.a

P3.a

tp

tp

P3.a

Temp Sensor

Pressure Sensor

Figure 1.6 Real time simulation for a typical Fuzzy Neural controller of tank water level


1214


IJSER

V. RESULT

Table 1 shows the output of the Fuzzy Logic, table 2 shows the output of the

Neural Network while Table three shows the output of Fuzzy Neural Network

Figs 1.7, Fig 1.8 and Fig 1.9 show the graph of output of Fuzzy Logic for water

level, graph of output of Neural Network and that of Fuzzy Neural Network

respectively.

Rout (cm3/s) 23,5 24,3 25,1 30.4 50,5 90.3 100.8 112.2 111.0 112.3 112.5 118.3 120.6 119.9

T(s) 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Rout (cm3/s) 24 24 25 30 51 90 101 112 111 112 113 118 121 120

T(s) 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Rout(cm3/s) 23 24 24 24 110 110 111 111 112 112 119 119 120 120

T(s) 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Table 1 data output Fuzzy Logic

Table 2 data output Neural Network

Table 2 data output Fuzzy Neural Network


1215


IJSER

0

20

40

60

80

100

120

140

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Rout

in c

m3

T(s)

The fuzzy output for outflow rate (Rout (cm3/s)

Rout (cm3/s)

0

20

40

60

80

100

120

140

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Rout

cm

3

T(s)

The neural output for outflow rate Rout (cm3/s)

Rout (cm3/s)

Fig 1.7 graph of Fuzzy output for PID tank water level [4]

Fig 1.8 graph of Neural Network output for PID tank water level [5]


1216


IJSER

VI. CONCLUSION

Figs 1.7 through 1.9 show the graphs of the output of the three Networks. It was

observed that the Intelligent Process Control Systems Using Fuzzy Neural

Network gave a steep rise time of 4s and stabilized after 6s. The Intelligent

Process Control Systems using Fuzzy Network rose sluggishly after an

undershoot at 5s and stabilized after 9s while the Intelligent Process Control

Systems using Neural Network kicked off its rise time at 3s and stabilized after

9s. The Fuzzy Neural Network performed better than either the Fuzzy System

alone or the Neural Network alone, considering the rising time and stabilizing

0

20

40

60

80

100

120

140

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Rout

(cm

3 )

T(s)

the fuzzy neural output for outflow rate (Rout(cm3/s)

Rout(cm3/s)

Fig 1.9 graph of Fuzzy Neural Network output for PID tank water level


1217


IJSER

time. The Fuzzy Neural Network performed best with a rise time of 1s and

showed a smoother overall performance. The three network graphs are plotted

together in Fig 1.10 to further highlight these observations.

0

20

40

60

80

100

120

140

1 2 3 4 5 6 7 8 9 10 11 12 13 14

outf

low

rate

in c

m3

T(s)

comparison of outflow rate for fuzzy,neural and fuzzy -neural network

Rout(cm3/s) for FNN

Rout (cm3/s) for fuzzy

Rout (cm3/s) for neural

Fig 1.10 The graph of combination Fuzzy, Neural and Fuzzy-Neural Network


1218


IJSER

(VII) REFERENCES

1) Zhou, R. W. & Quek, C. (1996). "POPFNN: A Pseudo Outer-product Based

Fuzzy Neural Network". Neural Networks, 9(9), 1569-1581.

2) Seema Chopra, Mitra R, Vijay Kumar, Neural Network Tuned fuzzy

controller for Multiple Input Multiple output (MIMO),World Academy of

Science, Engineering and Technology 2007.(p.485-491)

3) Mitchell Matthew .A, Lopes Pecas J.A, Fidalgo J.N and McCalley James D,

Neural Network to predict the Dynamic Frequency Response of a power

system to an Under-frequency Load Shedding Scenario, I.E.E.E Journal 2000

(p. 346-351).

4) Inyiama H.C and Nwobodo H.N, Designing Intelligent Process Control

System Using Fuzzy logic Principles, International Journal Of Scientific

Research And Education, August 2014

5) Inyiama H.C and Nwobodo H.N, Designing Intelligent Process Control

System Using Neural Network, International Journal Of Scientific Research

And Education, August 2014


1219


IJSER

6) Lin, C.-T., & Lee, C. S. G. (1996). Neural Fuzzy Systems: A Neuro-Fuzzy

Synergism to Intelligent Systems. Upper Saddle River, NJ: Prentice Hall.


1220


IJSER

DESIGNING INTELLIGENT PROCESS CONTROL SYSTEM USING FUZZY … · 2016-09-09 · DESIGNING...

Documents

Transcript of DESIGNING INTELLIGENT PROCESS CONTROL SYSTEM USING FUZZY … · 2016-09-09 · DESIGNING...