ANNAMALAI UNIVERSITY
Dr. S. SIVAPRAKASAM, Annamalai Nagar – 608 002
Associate Professor, Tamil Nadu, India.
Department of Mechanical Engineering. Date:
CERTIFICATE
This is to certify that the thesis entitled “AN EXPERIMENTAL STUDY ON
HEAT TRANSFER CHARACTERISTICS AND MODELING OF SPIRAL
HEAT EXCHANGER” is the bonafide work of Mr. A. MOHAMED SHABIULLA
(Roll No: 1032050001), Research Scholar, Department of Mechanical Engineering,
Annamalai University, Annamalai Nagar, under my guidance for the award of the
degree of Doctor of Philosophy and that this thesis has not been previously formed the
basis for the award of any Degree, Diploma, Associateship, Fellowship or other similar
title to the candidate.
This is also to certify that the thesis represents the independent work of the
candidate.
Signature of the Research Guide
(Dr. S. SIVAPRAKASAM)
iii
ACKNOWLEDGEMENTS
I am immensely thankful to the Authorities of Annamalai University for
permitting me to do this research work.
Words cannot express the author’s profound gratitude to his guide,
Dr. S. Sivaprakasam, Associate Professor, Department of Mechanical
Engineering, Annamalai University, for providing him with the nucleus of this
research work. Dr. S. Sivaprakasam’s personal interest, enthusiasm, guidance
and encouragement have been the mainstay of this onerous task.
The author is deeply indebted to Dr. N. Krishna Mohan, Head of the
Department of Mechanical Engineering, Annamalai University, for his generous
support in making available the necessary facilities for carrying out this research
work.
The author is sincerely grateful to Dr. N. Karunakaran, Associate Professor,
Department of Mechanical Engineering, Annamalai University, but for whose
sustained guidance and spontaneous help throughout the course of his research,
this thesis would not have taken such a shape.
The author also wishes to place on record, his gratitude to
Dr. C. Karthikeyan, Professor and Dr. M. Rajasimman, Associate Professor,
Department of Chemical Engineering and Dr. E. Sivaraman, Assistant Professor,
Department of Electronics and Instrumentation Engineering, Annamalai
University, for their invaluable suggestions and encouragement.
The author records his deepest sense of gratitude to the authorities of Kongu
Engineering College Perundurai, in general, and Dr. K. Saravanan, Professor
and Head, Department of Chemical Engineering, in particular, for granting
permission to carry out the experimental work.
Last but not the least, the author acknowledges the moral and intelligent
support extended to him by his best half, Dr. S. Arulselvi and appreciates his
daughter, M. Sharmila, for having borne all the deprivation caused, patiently.
Needless to say, but for the grace of the Almighty, it would have been
impossible to envisage this research work.
(A. MOHAMED SHABIULLA)
iv
ABSTRACT
The transfer of heat to and from process fluids is one of the most basic unit
operations in the processing industries. Heat exchangers are devices that facilitate
efficient heat transfer between two media, thereby changing the temperature
distribution of the media when they are in direct or indirect contact. They can transfer
heat between a liquid and a gas (e.g., a liquid-to-air heat exchanger) or two gases
(e.g., an air-to-air heat exchanger), or they can function as a liquid-to-liquid heat
exchanger. They thus form an integral and inevitable equipment in a wide range of
industries, including power generation, food processing, refrigeration, desalination,
air conditioning, automobiles and electronics cooling.
A wide range of heat exchangers are available in the market to cater to the needs
of different industries, like shell and tube heat exchangers, double pipe heat
exchangers, plate type heat exchangers and spiral heat exchangers etc. Depending
upon the process requirements, the type of fluid, its phase, operating temperature,
density, viscosity, pressure, chemical composition and various other thermodynamic
properties, the appropriate type and size of the heat exchanger can be selected.
Heat exchangers are characterised based on their ability to achieve the maximum
possible heat transfer coefficient within their operating range of variables. This can be
achieved by forcing both the fluids at the maximum possible mass flow rates and also
by supplying the hot fluid at higher temperatures. As a rule of thumb, doubling the
mass flow rate of the fluids may reduce the initial cost by half, but will increase the
pumping power requirements by a factor of roughly eight. Hence, the operating cost
of the heat exchanger will increase drastically. Therefore, a compromise shall be
made in order to achieve the maximum possible overall heat transfer coefficient with
the minimum possible consumption of pumping power.
Compared to Shell-and-Tube heat exchangers, Spiral plate Heat Exchangers
(SHE) are characterized by their large heat transfer surface area per unit volume,
resulting in a reduced space, weight, support structure, energy requirements and cost,
as well as improved process design, plant layout and processing. Spiral plate Heat
Exchangers have the distinct advantages over other plate type heat exchangers in the
aspect that they are self cleaning equipments with low fouling tendencies, easily
accessible for inspection or mechanical cleaning. They are best suited to handle
v
slurries and viscous liquids. The use of spiral heat exchangers is not limited to liquid-
liquid services. Variations in the basic design of SHE makes them suitable for Liquid-
Vapour or Liquid-Gas services.
Hence, in this work, a systematic experimental analysis is carried out, in a spiral
plate heat exchanger, to study the performance characteristics of the six fluid systems
namely,
1) Water - Water system
2) Water - Sea Water (3%) system
3) Water - Sea Water (12%) system
4) Water - Methanol system
5) Water - Butanol system
6) Water - Biodiesel system.
RSM is a collection of statistical and mathematical methods that are useful for the
modelling and analysing of engineering problems. In this technique, the main
objective is to optimise the response surface that is influenced by various process
parameters. RSM also quantifies the relationship between the controllable input
parameters and the obtained response surfaces.
Hence, in this research, an attempt is made to optimise the process parameters
using response surface methodology (RSM) with Box- Behnken design for all the six
fluid systems in a Spiral plate Heat Exchanger (SHE). Experiments are conducted by
varying the mass flow rates of the cold fluid and hot fluid (Water) and the inlet
temperature of the hot fluid by keeping the cold fluid temperature at ambient
conditions. The effects of variables on the process parameters of SHE are studied. The
process parameters viz. cold fluid flow rate ( cm ), hot fluid flow rate ( hm ) and hot
fluid inlet temperature ( inh,T ) are optimised by using RSM by solving the regression
model equation with Design Expert (Version 7.15) software and also by analysing the
three-dimensional surface plots in order to maximise the overall heat transfer
coefficient (U) and to minimise the pumping power (Wp) in SHE. The most influential
factor on each experimental design is identified from the analysis of variance
(ANOVA).
vi
The optimum conditions of cm , hm and inh,T for all the six systems in SHE,
proposed by RSM, are obtained. In these optimised conditions, the optimum values of
‘U’ and ‘WP’ are found for all the six systems. The optimised values are also verified
by experimentation, to show the effectiveness of the proposed RSM model with Box-
Behnken design. The results are presented and discussed for all the six cases.
Computational fluid dynamics (CFD) is the science of predicting fluid flow, heat
and mass transfer, chemical reactions, and related phenomena. The results of CFD
analyses are relevant in conceptual studies of new designs, detailed product
development, trouble shooting and redesign.
In this work, it is proposed to create a three dimensional meshed model of a spiral
heat exchanger to verify the experimental results. Spiral plate exchanger geometry is
created in GAMBIT 2.4.6 environment and the boundary conditions and material
properties are given as input data. The meshed model with relevant input data, is
imported into the FLUENT 13.0 software for post processing in order to predict the
temperature profiles of the respective fluid systems. Simulations are carried out for all
the six fluid systems for all the cases (15 cases per fluid system). The temperature
data for the outlet conditions are computed and the corresponding temperature profile
of the particular experimental condition is also obtained. The extracted temperature
data are compared with those of the experimental data in order to validate the fitness
of the CFD model. The results are presented and discussed for all the six cases.
Industrial processes naturally exhibit nonlinear behaviour and complex dynamics.
It is well known that virtually all processes of practical importance exhibit a certain
degree of nonlinear behaviour. The ability of fuzzy logic and artificial neural network
to represent nonlinear systems makes them powerful tools for process modelling and
much work has been reported over the last decade. The main disadvantage of
Artificial Neural Networks (ANN) is that the rules of operation are completely
unknown and it takes considerable time to train an ANN for certain functions.
Recently, Adaptive Network based Fuzzy Inference System (ANFIS) modelling and
fuzzy modelling based on fuzzy clustering techniques are strongly emerging in the
field of industrial process modelling and their application to Spiral plate Heat
Exchanger (SHE) are not much reported.
vii
The ANFIS model possesses the robustness of fuzzy systems, the learning ability
of neural networks and can adapt to various situations. An effective approach to the
identification of complex systems is to partition the available data into subsets and
approximate each subset by a simple model. Fuzzy clustering can be used as a tool to
obtain a partitioning of data where the transitions between the subsets are gradual
rather than abrupt. Although ANFIS combines the advantages of both neural and
fuzzy systems, the cluster formation of the data during the pretreatment is fixed and
the cluster shape and orientation are not considered while training the model. Instead,
Gustafson-Kessel (G-K) fuzzy clustering offers a better way of modelling the
nonlinear system by taking the above mentioned factors into consideration.
Hence, in this research work, an attempt is made to propose Artificial Neural
Network (ANN), Fuzzy Clustering technique and Adaptive Network based Fuzzy
Inference System (ANFIS) based models for the analysis of Spiral plate Heat
Exchanger (SHE). For every method, two different models are developed for the
overall heat transfer coefficient (U) and pumping power (Wp). The data required to
train the models are obtained from the experimental data based on Response Surface
Methodology (RSM). The ANN models are developed using Back Propagation
Network (BPN) algorithm, incorporating Levenberg-Marquardt (L-M) training
method. The fuzzy clustering models are developed based on Gustafson-Kessel (G-K)
algorithm. The ANFIS models are developed based on the advanced neural-fuzzy
technology. The accuracy of the ANN, G-K clustering and ANFIS models are verified
according to their ability to predict unseen data by minimising the performance
measures like root mean square error (RMSE), average percentage error (APE) and
correlation coefficient (%R2) value. The prediction of the parameters can be obtained
without using charts and complex formulae.
The performance measures like percentage error, APE, RMSE and Correlation
coefficient (% R2) are calculated and tabulated for ‘U’ and ‘Wp’ for all the fluid
systems by implementing ANN, G-K clustering and ANFIS models and compared
with those of the experimental values and RSM model outputs. From the results, it is
observed that the G-K model outperforms all the models, by producing negligible
%Error. Also, it produces a negligible value for APE and RMSE and 100% value for
R2, thereby performing superior results when compared to RSM, ANN and ANFIS
models for ‘U’ and ‘Wp’. The G-K clustering algorithm may be a viable alternative
for the modelling of other heat exchangers and industrial processes.
viii
TABLE OF CONTENTS
CHAPTER
NO
TITLE PAGE
NO
ACKNOWLEDGEMENTS iii
ABSTRACT iv
LIST OF TABLES xv
LIST OF FIGURES xviii
LIST OF SYMBOLS xxix
LIST OF ABBREVIATIONS xxxi
1. INTRODUCTION 1
1.1 COMPACT HEAT EXCHANGERS 1
1.2 SPIRAL PLATE HEAT EXCHANGERS 2
1.3 ADVANTAGES OF SPIRAL PLATE HEAT
EXCHANGERS OVER CONVENTIONAL SHELL
AND TUBE HEAT EXCHANGERS
4
1.4 APPLICATIONS OF SPIRAL PLATE HEAT
EXCHANGER
6
1.5 SELECTION OF HEAT EXCHANGERS 7
2. LITERATURE SURVEY 8
2.1 INTRODUCTION 8
2.2 HEAT EXCHANGERS 8
2.2.1 Geometries other than Spiral Configuration 8
2.2.2 Spiral Geometry 11
2.3 RESPONSE SURFACE METHODOLOGY 17
2.4 NUMERICAL ANALYSIS AND CFD MODELLING 23
2.5 INTELLIGENT MODELLING 30
2.5.1 ANN Modelling 30
2.5.2 ANFIS Modelling 33
2.5.3 Fuzzy Clustering 34
ix
2.6 MOTIVATION FOR THE PRESENT WORK 35
2.7 PRESENT WORK 36
3 OBJECTIVES 37
3.1 OBJECTIVES 37
3.2 ORGANISATION OF THE THESIS 39
4 METHODOLOGY 42
4.1 FLOW CHART FOR EXPERIMENTATION AND
OPTIMISATION
42
4.2 FLOW CHART FOR CFD MODELLING 44
4.3 FLOW CHART FOR INTELLIGENT MODELLING 44
5 OPTIMISATION OF PROCESS
PARAMETERS USING RESPONSE
SURFACE METHODOLOGY
47
5.1 DESIGN OF EXPERIMENTS (DOE) 47
5.2 RESPONSE SURFACE METHODOLOGY (RSM) 48
5.2.1 Box-Behnken Design (BBD) 49
5.2.2 Experimental Design and Analysis using BBD 50
5.2.3 Potential Advantages of Box-Behnken Design 51
5.2.4 Experimental Conditions Proposed by BBD for
SHE
52
6. EXPERIMENTAL SETUP AND RESPONSE
SURFACE METHODOLOGY (RSM) BASED
OPTIMISATION OF SPIRAL PLATE HEAT
EXCHANGER
55
6.1 EXPERIMENTATION 56
6.1.1 Experimental Set-up and Procedure 56
6.1.2 Variables to be Optimised 58
6.1.2.1 Overall heat transfer coefficient (U) 58
6.1.2.2 Pumping power (Wp) 58
6.2 RESULTS AND DISCUSSION 60
6.3 WATER-WATER SYSTEM 60
x
6.3.1 Model for Overall Heat Transfer Coefficient (U)
(W-W)
61
6.3.2 Model for Pumping Power for (Wp) (W-W) 66
6.3.3 Optimisation Plot (W-W) 70
6.3.3.1 Optimisation of Overall Heat Transfer
Coefficient (W-W)
70
6.3.3.2 Optimisation of Pumping Power (W-W) 71
6.3.4 Concluding Remarks 72
6.4 WATER-SEA WATER (3%) SYSTEM 72
6.4.1 Model for Overall Heat Transfer Coefficient
(U) (W-SW(3%))
72
6.4.2 Model for Pumping Power for Water – Sea
Water (3%) System
77
6.4.3 Optimisation Plot (W-SW (3%)) 82
6.4.3.1 Optimisation of Overall Heat Transfer
Coefficient (W-SW (3%))
82
6.4.3.2 Optimisation of Pumping Power (W-SW
(3%))
83
6.4.4 Concluding Remarks 83
6.5 WATER-SEA WATER (12%) SYSTEM 83
6.5.1 Model for Overall Heat Transfer Coefficient
(U) (W-SW(12 %))
84
6.5.2 Model for Pumping Power for Water – Sea
Water (12 %) System
89
6.5.3 Optimisation Plot (W-SW (12 %)) 94
6.5.3.1 Optimisation of Overall Heat Transfer
Coefficient (W-SW (12 %))
94
6.5.3.2 Optimisation of Pumping Power (W-SW
(12 %))
95
6.5.4 Concluding Remarks 95
6.6 WATER-METHANOL SYSTEM 95
6.6.1 Model for Overall Heat Transfer Coefficient
(U) (W-M)
96
6.6.2 Model for Pumping Power for Water –
Methanol System
102
6.6.3 Optimisation Plot (W-M) 106
xi
6.6.3.1 Optimisation of Overall Heat Transfer
Coefficient (W-M)
106
6.6.3.2 Optimisation of Pumping Power (W-M) 107
6.6.4 Concluding Remarks 107
6.7 WATER-BUTANOL SYSTEM 107
6.7.1 Model for Overall Heat Transfer Coefficient
(U) (W-B)
108
6.7.2 Model for Pumping Power for Water – Butanol
System
113
6.7.3 Optimisation Plot (W-B) 117
6.7.3.1 Optimisation of Overall Heat Transfer
Coefficient (W-B)
117
6.7.3.2 Optimisation of Pumping Power (W-B) 118
6.7.4 Concluding Remarks 118
6.8 WATER-BIODIESEL SYSTEM 119
6.8.1 Model for Overall Heat Transfer Coefficient
(U) (W-Bio)
119
6.8.2 Model for Pumping Power for Water –
Biodiesel System
125
6.8.3 Optimisation Plot (W-Bio) 129
6.8.3.1 Optimisation of Overall Heat Transfer
Coefficient (W-Bio)
129
6.8.3.2 Optimisation of Pumping Power
(W-Bio)
130
6.8.4 Concluding Remarks 130
6.9 CONSOLIDATION 130
7 NUMERICAL MODELLING OF A SPIRAL
HEAT EXCHANGER USING CFD
TECHNIQUE
133
7.1. NUMERICAL MODELLING 133
7.1.1. Governing Equations 134
7.1.2. Discretisation of the Governing Equations 134
7.2. CFD MODELLING 135
7.2.1 Meshing 136
xii
7.2.2 Assumptions 137
7.2.3 Boundary Conditions 138
7.2.3.1 Velocity Inlet Boundary Conditions 138
7.2.3.2 Pressure Outlet Boundary Conditions 138
7.2.3.3 Thermal Boundary Conditions 138
7.2.4 Physical Properties 139
7.2.5 Simulation 140
7.3. NUMERICAL RESULTS AND DISCUSSION 141
7.3.1 Water - Water System 141
7.3.2 Water - Sea Water (3%) System 143
7.3.3 Water - Sea Water (12%) System 146
7.3.4 Water - Methanol System 148
7.3.5 Water - Butanol System 150
7.3.6 Water - Biodiesel System 153
7.4. SUMMARY OF RESULTS 155
8 INTELLIGENT MODELLING OF SPIRAL
PLATE HEAT EXCHANGER USING
ARTIFICIAL NEURAL NETWORK, ANFIS
AND FUZZY CLUSTERING ALGORITHMS
157
8.1 MOTIVATION FOR AN ARTIFICIAL NEURAL
NETWORK APPROACH TO NON-LINEAR
SYSTEMS
158
8.1.1 Architecture of Artificial Neural Networks 158
8.1.2 Back Propagation Algorithm 160
8.1.2.1 Levenberg -Marquardt (LM) Algorithm
for Training Neural Model
161
8.1.3 Advantages of Artificial Neural Network 163
8.1.4 Disadvantages of Artificial Neural Network 164
8.1.5 Implementation of Artificial Neural Network 164
8.2 FUZZY LOGIC SYSTEMS (FLS) 165
8.2.1 Mamdani Fuzzy Inference System 167
8.2.2 Takagi - Sugeno Fuzzy Inference System 168
xiii
8.3 FUZZY CLUSTERING 171
8.3.1 Cluster Analysis 171
8.3.2 Fuzzy Partition 171
8.3.3 Gustafson-Kessel (G-K) Clustering Algorithm 173
8.3.4 Implementation of Gustafson-Kessel (G-K)
Clustering Algorithm
174
8.4 ANFIS Model 175
8.4.1 Implementation of ANFIS Model 179
8.5 MODELLING USING ARTIFICIAL NEURAL
NETWORK, ANFIS AND FUZZYCLUSTERING
ALGORITHMS
183
8.5.1 Forward Model using ANN, ANFIS and Fuzzy
Clustering Techniques
183
8.6 PERFORMANCE MEASURES 183
8.7 RESULTS AND DISCUSSION 185
8.7.1 Water –Water System 185
8.7.1.1 ANFIS rules for Overall Heat Transfer
Coefficient (U)
190
8.7.1.2 ANFIS rules for Pumping Power (Wp) 190
8.7.1.3 G-K rules for Overall Heat Transfer
Coefficient (U)
194
8.7.1.4 G-K rules for Pumping Power (Wp) 194
8.7.2 Water-Sea Water (3%) System 201
8.7.3 Water-Sea Water (12%) System 209
8.7.4 Water-Methanol System 217
8.7.5 Water-Butanol System 225
8.7.6 Water-Bio Diesel System 233
8.8 CONCLUDING REMARKS 241
9 CONCLUSION 243
9.1 GENERAL 243
9.2 SUGGESTIONS FOR FUTURE WORK 245
xiv
ANNEXURES 246
I TEMPERATURE CONTOUR PLOTS FOR WATER-
WATER SYSTEM 246
II TEMPERATURE CONTOUR PLOTS FOR WATER-
SEA WATER (3%) SYSTEM 249
III TEMPERATURE CONTOUR PLOTS FOR WATER-
SEA WATER (12%) SYSTEM 252
IV TEMPERATURE CONTOUR PLOTS FOR WATER-
METHANOL SYSTEM 255
V TEMPERATURE CONTOUR PLOTS FOR WATER-
BUTANOL SYSTEM 258
VI TEMPERATURE CONTOUR PLOTS FOR WATER-
BIODIESEL SYSTEM 261
REFERENCES 264
LIST OF PUBLICATIONS 276
xv
LIST OF TABLES
TABLE
NO.
TITLE PAGE
NO.
5.1. Box-Behnken Design for Three Variables 50
5.2. Experimental conditions proposed by BBD for SHE in
coded form
53
5.3 Experimental conditions proposed by BBD for SHE in
uncoded form
54
6.1 Specifications of the Spiral Plate Heat Exchanger 59
6.2 Experimental conditions 59
6.3 BBD matrix for the experimental design and predicted
responses for the overall heat transfer coefficient and
pumping power for Water-Water system.
61
6.4 ANOVA for overall heat transfer coefficient for
Water-Water system
62
6.5 ANOVA for pumping power for Water-Water system 66
6.6 BBD matrix for the experimental design and predicted
responses for the overall heat transfer coefficient and
pumping power for Water-Sea Water (3%) system
73
6.7 ANOVA for overall heat transfer coefficient for Water-Sea
Water (3%) system
74
6.8 ANOVA for pumping power for Water-Sea Water (3%)
system
78
6.9 BBD matrix for the experimental design and predicted
responses for the overall heat transfer coefficient and
pumping power for Water – Sea Water (12%) system
85
6.10 ANOVA for overall heat transfer coefficient for Water-Sea
Water (12%) system
86
6.11 ANOVA for pumping power for Water-Sea Water (12%)
system
90
6.12 BBD matrix for the experimental design and predicted
responses for the overall heat transfer coefficient and
pumping power for Water -Methanol system
97
xvi
TABLE
NO.
TITLE PAGE
NO.
6.13 ANOVA for overall heat transfer coefficient for Water-
Methanol system
98
6.14 ANOVA for pumping power for Water-Methanol system 102
6.15 BBD matrix for the experimental design and predicted
responses for the overall heat transfer coefficient and
pumping power for Water-Butanol system
108
6.16 ANOVA for overall heat transfer coefficient for Water-
Butanol system
109
6.17 ANOVA for pumping power for Water-Butanol system 113
6.18 BBD matrix for the experimental design and predicted
responses for the overall heat transfer coefficient and
pumping power for Water-Biodiesel system
120
6.19 ANOVA for overall heat transfer coefficient for Water -
Biodiesel system
121
6.20 ANOVA for pumping power for Water-Biodiesel system 125
6.21 Optimisation summary of the fluid systems 132
7.1 Specifications of the Spiral Plate Heat Exchanger for CFD
input
136
7.2 Input process parameters for CFD simulation
corresponding to the 15 cases
140
7.3 Summary of CFD simulation results 156
8.1 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for U (Water-Water)
199
8.2 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for Wp (Water-Water)
200
8.3 Comparison of performance measures of different
modeling techniques for Water-Water system
201
8.4 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for U (Water-Sea Water (3%))
207
8.5 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for Wp (Water-Sea Water (3%))
208
xvii
TABLE
NO.
TITLE PAGE
NO.
8.6 Comparison of performance measures of different
modeling techniques for Water–Sea Water(3%) system
209
8.7 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for U (Water-Sea Water (12%))
215
8.8 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for Wp (Water-Sea Water (12%))
216
8.9 Comparison of performance measures of different
modeling techniques for Water-Sea Water (12%) system
217
8.10 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for U (Water-Methanol)
223
8.11 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for Wp (Water-Methanol)
224
8.12 Comparison of performance measures of different
modeling techniques for Water-Methanol system
225
8.13 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for U (Water-Butanol)
231
8.14 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for Wp (Water-Butanol)
232
8.15 Comparison of performance measures of different
modeling techniques for Water-Butanol system
233
8.16 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for U (Water-Biodiesel)
239
8.17 Comparison of % Error of RSM, ANN, ANFIS and G-K
models for Wp (Water-Biodiesel)
240
8.18 Comparison of performance measures of different
modeling techniques for Water-Biodiesel system
241
xviii
LIST OF FIGURES
FIGURE
NO.
TITLE PAGE
NO.
1.1 Elementary construction of a spiral plate heat exchanger 3
1.2 Flow paths in the spiral plate heat exchanger 4
4.1 Flow chart for experimentation and optimisation 43
4.2 Flow chart for CFD Modelling 45
4.3 Flow chart for Intelligent Modelling 46
5.1 Position of variable points of Box-Behnken Design in cubic
region
49
6.1 Schematic diagram of the SHE experimental set-up. 57
6.2 Photographic view of the experimental set up 57
6.3 Actual values versus predicted responses of overall heat
transfer coefficient for Water-Water system
64
6.4 3D plot showing the effect of cold fluid and hot fluid flow rate
on the overall heat transfer coefficient for Water-Water system
64
6.5 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on the overall heat transfer coefficient for
Water-Water system
65
6.6 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on the overall heat transfer coefficient for
Water-Water system
65
6.7 Actual values versus predicted responses of pumping power
for Water-Water system
68
6.8 3D plot showing the effect of cold fluid and hot fluid flow
rates on pumping power for Water-Water system
69
6.9 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on pumping power for Water-Water system
69
6.10 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on pumping power for Water-Water
system
70
xix
FIGURE
NO.
TITLE PAGE
NO.
6.11 Optimisation plot for Water-Water system 71
6.12 Actual values versus predicted responses of overall heat
transfer coefficient for Water- Sea Water (3%) system
75
6.13 3D plot showing the effect of cold fluid and hot fluid flow
rates on the overall heat transfer coefficient for Water- Sea
Water (3%) system
76
6.14 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on the overall heat transfer coefficient for
Water- Sea Water ( 3%) system
76
6.15 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on the overall heat transfer coefficient for
Water-Sea Water (3%) system
77
6.16 Actual values versus predicted responses of pumping power
for Water- Sea Water (3%) system
79
6.17 3D plot showing the effect of cold fluid and hot fluid flow
rates on pumping power for Water-Sea Water (3%) system
80
6.18 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on pumping power for Water-Sea Water
(3%) system
81
6.19 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on pumping power for Water-Sea Water
(3%) system
81
6.20 Optimisation plot for Water- Sea Water (3%) system 82
6.21 Predicted responses versus actual values of overall heat
transfer coefficient for Water-Sea Water (12%) system
87
6.22 3D plot showing the effect of cold fluid and hot fluid flow
rates on the overall heat transfer coefficient for Water-Sea
Water (12%) system
88
6.23 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on the overall heat transfer coefficient for
Water- Sea Water (12%) system
88
6.24 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on the overall heat transfer coefficient for
Water-Sea Water (12%) system
89
xx
FIGURE
NO.
TITLE PAGE
NO.
6.25 Predicted responses versus actual values of pumping power
for Water-Sea Water (12%) system
91
6.26 3D plot showing the effect of cold fluid and hot fluid flow
rates on pumping power for Water-Sea Water (12%) system
92
6.27 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on pumping power for Water-Sea Water
(12%) system
93
6.28 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on pumping power for Water-Sea Water
(12%) system
93
6.29 Optimisation plot for Water- Sea Water (12%) system 94
6.30 Actual values versus predicted responses of overall heat
transfer coefficient for Water-Methanol system
99
6.31 3D plot showing the effect of cold fluid and hot fluid flow
rates on the overall heat transfer coefficient for Water-
Methanol system
100
6.32 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on the overall heat transfer coefficient for
Water- Methanol system
101
6.33 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on the overall heat transfer coefficient for
Water-Methanol system
101
6.34 Actual values versus predicted responses of pumping power
for Water-Methanol system
103
6.35 3D plot showing the effect of cold fluid and hot fluid flow
rates on pumping power for Water-Methanol system
104
6.36 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on pumping power for Water-Methanol
system
105
6.37 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on pumping power for Water-Methanol
system
105
6.38 Optimisation plot for Water- Methanol system 106
xxi
FIGURE
NO.
TITLE PAGE
NO.
6.39 Actual values versus predicted responses of overall heat
transfer coefficient for Water-Butanol system
110
6.40 3D plot showing the effect of cold fluid and hot fluid flow
rates on the overall heat transfer coefficient for Water –
Butanol system
111
6.41 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on the overall heat transfer coefficient for
Water-Butanol system
112
6.42 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on the overall heat transfer coefficient for
Water-Butanol system
112
6.43 Actual values versus predicted responses of pumping power
for Water-Butanol system
114
6.44 3D plot showing the effect of cold fluid and hot fluid flow
rates on pumping power for Water-Butanol system
115
6.45 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on pumping power for Water-Butanol
system
116
6.46 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on pumping power for Water-Butanol
system
116
6.47 Optimisation plot for Water-Butanol system 118
6.48 Actual values versus predicted responses of overall heat
transfer coefficient for Water-Biodiesel system
122
6.49 3D plot showing the effect of cold fluid and hot fluid flow
rates on the overall heat transfer coefficient for Water-
Biodiesel system
123
6.50 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on the overall heat transfer coefficient for
Water -Biodiesel system
124
6.51 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on the overall heat transfer coefficient for
Water-Biodiesel system
124
xxii
FIGURE
NO.
TITLE PAGE
NO.
6.52 Actual values versus predicted responses of pumping power
for Water-Biodiesel system
126
6.53 3D plot showing the effect of cold fluid and hot fluid flow
rates on pumping power for Water –Biodiesel system
127
6.54 3D plot showing the effect of hot fluid inlet temperature and
hot fluid flow rate on pumping power for Water –Biodiesel
system
128
6.55 3D plot showing the effect of hot fluid inlet temperature and
cold fluid flow rate on pumping power for Water –Biodiesel
system
128
6.56 Optimisation plot for Water-Biodiesel system 129
7.1 Meshed cross sectional view of the Spiral plate Heat
Exchanger grid
137
7.2 Meshed 3D view of the Spiral plate Heat Exchanger 137
7.3 Contours of static temperature (K) for Water-Water system
corresponding to the case 5
141
7.4 Comparison of Experimental and CFD simulated Cold Fluid
Outlet Temperatures of Water-Water system
142
7.5 Comparison of experimental and CFD simulated hot fluid
outlet temperatures of Water-Water system
143
7.6 Contours of static temperature (K) for Water- Sea water (3%)
system corresponding to the case 5
144
7.7 Comparison of experimental and CFD simulated cold fluid
outlet temperatures of Water- Sea Water (3%) system
145
7.8 Comparison of experimental and CFD simulated cold fluid
outlet temperatures of Water- Sea Water (3%) system
145
7.9 Contours of static temperature (K) for Water- Sea water (12%)
system corresponding to the case 5
146
7.10 Comparison of experimental and CFD simulated cold fluid
outlet temperatures of Water- Sea Water (12%) system
147
7.11 Comparison of experimental and CFD simulated hot fluid
outlet temperatures of Water- Sea Water (12%) system
147
xxiii
FIGURE
NO.
TITLE PAGE
NO.
7.12 Contours of static temperature (K) for Water- Methanol system
corresponding to the case 5
148
7.13 Comparison of experimental and CFD simulated cold fluid
outlet temperatures of Water- Methanol system
149
7.14 Comparison of experimental and CFD simulated hot fluid
outlet temperatures of Water- Methanol system
150
7.15 Contours of static temperature (K) for Water- Butanol system
corresponding to the case 5
151
7.16 Comparison of experimental and CFD simulated cold fluid
outlet temperatures of Water- Butanol system
152
7.17 Comparison of experimental and CFD simulated hot fluid
outlet temperatures of Water- Butanol system
152
7.18 Contours of static temperature (K) for Water- Biodiesel
system corresponding to the case 5
153
7.19 Comparison of experimental and CFD simulated cold fluid
outlet temperatures of Water- Biodiesel system
154
7.20 Comparison of experimental and CFD simulated hot fluid
outlet temperatures of Water- Biodiesel system
155
8.1. Generalized multilayer neural network configuration 159
8.2. Block diagram of a Fuzzy Inference System 166
8.3. A graphical representation of weighted average defuzzifier 168
8.4. Local linear model and the membership function diagram 170
8.5. Linear model of the nonlinear function and the membership
function diagram
173
8.6. (a) Two - input first-order T-S fuzzy model with two rules and
(b) its equivalent ANFIS architecture
178
8.7. Conceptual neural-fuzzy (ANFIS) model for Spiral plate Heat
Exchanger(SHE)
179
8.8. Computing flowchart of ANFIS model 181
8.9. Structure of resulting T-S Fuzzy model of SHE 182
xxiv
FIGURE
NO.
TITLE PAGE
NO.
8.10. Structure for resulting ANFIS model of SHE 182
8.11. Block diagram of a forward model scheme based on neural
network, ANFIS and fuzzy clustering for U
184
8.12. Block diagram of a forward model scheme based on neural
network, ANFIS and fuzzy clustering for Wp
184
8.13. Variation of mean square error for forward neural model
during training of the process for U (Water-Water)
186
8.14. Variation of mean square error for forward neural model
during training of the process for Wp (Water-Water)
187
8.15. Membership diagram of cm for ANFIS model (Water-Water) 187
8.16. Membership diagram of hm for ANFIS model (Water-Water) 188
8.17. Membership diagram of inh,T for ANFIS model (Water-
Water)
188
8.18. Membership diagram of U(k-1) for ANFIS model (Water-
Water)
189
8.19. Membership diagram of Wp (k-1) for ANFIS model (Water-
Water)
189
8.20. Results showing clusters formation (4 nos.*)for U by applying
G-K algorithm for Water-Water system
191
8.21. Results showing clusters formation (4 nos.*)for Wp by
applying G-K algorithm for Water-Water system
191
8.22. Membership diagram of cm for G-K model (Water-Water) 192
8.23. Membership diagram of hm for G-K model (Water-Water) 192
8.24. Membership diagram of inh,T for G-K model (Water-Water) 193
8.25. Membership diagram of U(k-1) for G-K model (Water-Water) 193
8.26. Membership diagram of Wp(k-1) for G-K model (Water-
Water)
194
8.27. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs U (Water-Water)
195
xxv
FIGURE
NO.
TITLE PAGE
NO.
8.28. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs U (Water-Water)
196
8.29. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs U (Water-Water)
196
8.30. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs Wp (Water-Water)
197
8.31. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs Wp (Water-Water)
198
8.32. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs Wp (Water-Water).
198
8.33. Variation of mean square error for forward neural model
during training of the process for U (Water-Sea Water (3%))
202
8.34. Variation of mean square error for forward neural model
during training of the process for Wp (Water-Sea Water (3%))
202
8.35. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs U (Water-Sea Water (3%))
203
8.36. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs U (Water-Sea Water (3%))
204
8.37. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs U (Water-Sea Water
(3%))
204
8.38. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs Wp (Water-Sea Water (3%))
205
8.39. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs Wp (Water-Sea Water
(3%))
206
8.40. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs Wp (Water-sea Water
(3%))
206
8.41. Variation of mean square error for forward neural model
during training of the process for U (Water-Sea Water (12%))
210
xxvi
FIGURE
NO.
TITLE PAGE
NO.
8.42. Variation of mean square error for forward neural model
during training of the process for Wp (Water-Sea Water (12%))
210
8.43. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs U (Water-Sea Water (12%))
211
8.44. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs U (Water-Sea Water
(12%))
212
8.45. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs U (Water-Sea Water
(12%))
212
8.46. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs Wp (Water-Sea Water
(12%))
213
8.47. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs Wp (Water-Sea Water
(12%))
214
8.48. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs Wp (Water-Sea Water
(12%))
214
8.49. Variation of mean square error for forward neural model
during training of the process for U (Water-Methanol)
218
8.50. Variation of mean square error for forward neural model
during training of the process for Wp (Water-Methanol)
218
8.51. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs U (Water-Methanol)
219
8.52. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs U (Water-Methanol)
220
8.53. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs U (Water-Methanol)
220
8.54. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs Wp (Water-Methanol)
221
xxvii
FIGURE
NO.
TITLE PAGE
NO.
8.55. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs Wp (Water-Methanol)
222
8.56. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs Wp (Water-Methanol)
222
8.57. Variation of mean square error for forward neural model
during training of the process for U (Water-Butanol)
226
8.58. Variation of mean square error for forward neural model
during training of the process for Wp (Water-Butanol)
226
8.59. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs U (Water-Butanol)
227
8.60. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs U (Water-Butanol)
228
8.61. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs U (Water-Butanol)
228
8.62. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs Wp (Water-Butanol)
229
8.63. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs Wp (Water-Butanol)
230
8.64. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs Wp (Water-Butanol)
230
8.65. Variation of mean square error for forward neural model
during training of the process for U (Water-Biodiesel)
234
8.66. Variation of mean square error for forward neural model
during training of the process for Wp (Water-Biodiesel)
234
8.67. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs U (Water-Biodiesel)
235
8.68. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs U (Water-Biodiesel)
236
8.69. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs U (Water-Biodiesel)
236
xxviii
FIGURE
NO.
TITLE PAGE
NO.
8.70. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for cm Vs Wp (Water-Biodiesel)
237
8.71. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for hm Vs Wp (Water-Biodiesel)
238
8.72. Comparison of experimental values with RSM, ANN, ANFIS
and G-K model outputs for inh,T Vs Wp (Water-Biodiesel)
238
xxix
LIST OF SYMBOLS
cm Cold fluid flow rate (kg/s)
hm Hot fluid flow rate (kg/s)
inc,T Cold fluid inlet temperature (oC)
inh,T Hot fluid inlet temperature (oC)
jiw Weight on the connection from ith
input unit to jth
hidden unit of
Neural Network (ANN)
pix ith
input for the pth
input vector of ANN
jθ Bias unit for jth
hidden neuron of ANN
l Number of input vectors of ANN
jf Bipolar tan-sigmoid activation function for the hidden layer of ANN
mjw Weight on the connection from jth
hidden unit to mth
output unit of
ANN
mθ Bias unit for mth
output neuron of ANN
mf Bipolar pure linear function for the hidden layer of ANN
pmY Desired output of ANN
pmO Actual output from the neural model of ANN
pme Error vector of ANN
μ Learning rate
(x)φi Validity function for the ith
operating regime
ia Parameter of the local linear model
ib Parameter of the local linear model
iw Weight of the rule.
jv Cluster centre of the j-th cluster
ε Termination tolerance
(j)iV Cluster means
ikAD Euclidian distance
xxx
(j)ikμ Partician matrix
iF Cluster covariance matrix
U Predicted overall heat transfer coefficient (W/m2K)
pW
Predicted pumping power (W)
jO
jth
output value
jt
jth
target value
c
Center of Gaussian membership function
Width of Gaussian membership function
µij
Membership of data point j in cluster i
Ai Antecedent fuzzy set of the ith
rule
C Number of clusters
m Fuzziness parameter
R2 Correlation coefficient
U Overall heat transfer coefficient (W/m2K)
u Defuzzified output value
Uk Fuzzy set
W Weight of the network.
Wp Pumping power (W)
Z Data set
xxxi
LIST OF ABBREVIATIONS
3D Three Dimensional
ANFIS
Adaptive Network based Fuzzy Inference System
ANN Artificial Neural Networks
ANOVA Analysis of Variance
APE Average Percentage Error
BBD Box-Behnken Design
BPN Back Propagation Network
C.V Co-efficient of Variance
CFD Computational Fluid Dynamics
DPT Differential Pressure Transmitter
FCM Fuzzy C-Means
FIS Fuzzy Inference System
G-K Gustafson-Kessel
L-M Levenberg –Marquardt
MSE Mean Squared Error
OHTC Overall Heat Transfer Coefficient
RMSE Root Mean Square Error
RSM Response Surface Methodology
RTD Resistance Temperature Detector
SCG Scaled Conjugate Gradient
SHE Spiral plate Heat Exchanger
T-S Takagi-Sugeno
W- B Water- Butanol System
W- Bio Water- Biodiesel System
W-M Water- Methanol System
W-SW (12%) Water- Sea Water (12%) System
W-SW (3%) Water- Sea Water (3%) System
W-W Water-Water System