Print Thesis - PDF 179 May28

STUDY OF MACHINABILITY
CHARECTERISTICS OF GFRP COMPOSITES

WITH DIFFERENT FIBRE ORIENTATIONS
DURING ENDMILLING
A THESIS
submitted by
M.P. JENARTHANAN
for the award of the degree

of
DOCTOR OF PHILOSOPHY
DEPARTMENT OF PRODUCTION ENGINEERING

NATIONAL INSTITUTE OF TECHNOLOGY
TIRUCHIRAPPALLI-620015
MAY 2013
THESIS CERTIFICATE
This is to certify that the thesis entitled STUDY OF MACHINABILITY

CHARECTERISITICS OF GFRP COMPOSITES WITH DIFFERENT FIBRE
ORIENTATIONS DURING ENDMILLING submitted by M.P.Jenarthanan (Roll
No: 414110002) to the National Institute of Technology, Tiruchirappalli for the award
of the degree of Doctor of Philosophy is a bonafide record of research work carried
out by him under my supervision. The contents of this thesis, in full or in parts, have
not been submitted to any other Institute or University for the award of any degree or
diploma.
Tiruchirappalli-620 015.
Date:
(Dr.R. Jeyapaul)
Research Supervisor
Associate Professor
Department of Production Engineering
National Institute of Technology
Tiruchirappalli-620 015.
India.
ii
ABSTRACT
With the increasing use of Glass Fibre Reinforced Polymer (GFRP)

composites outside the defence, space and aerospace industries, machining of these
materials is gradually assuming a significant role. The current knowledge of
machining GFRP composites is in transition phase for its optimum economic
utilisation in various fields of applications. Therefore, material properties and
theoretical mechanics have become the predominant research areas in this field. With
increasing applications, economical techniques of production are indeed very
important to achieve fully automated large-scale manufacturing cycles. Although in
common occurrence, composite products are manufactured with near-net shapes,
secondary processes involving machining are often necessary as to achieve the
required geometrical shapes and dimensional tolerances. Indeed, various machining
processes such as turning, drilling and milling have been used to machine composite
materials for different product requirements. Despite the existing experience and
knowledge in machining traditional materials such as metals, it has been a challenge
to maintain consistent results in terms of machining quality for composite materials.
First, due to fibre arrangements and/or orientation, poor surface finish which includes
fibre pullout, matrix delamination, sub-surface damage and matrix polymer failures is
usually observed if improper machining conditions were employed. Second, due to
their anisotropy, and non-homogeneity, FRP composites face considerable problems
in machining like fibre pull-out, delamination, burning, etc. There is a remarkable
difference between the machining of conventional metals and their alloys and that of
composite materials. Further, each composite differs in its machining behaviour since
its physical and mechanical properties depend largely on the type of fibre, the fibre
iii
content, the fibre orientation and variabilities in the matrix material. Considerable
amount of literature is readily available on the machinability of conventional
metals/alloys and also polymers to some extent; with very limited work on GFRP
composites. However, machinability evaluation of glass fibre-reinforced polymer
(GFRP) composites in end milling has not yet received its due attention in the
research community despite the extensive industrial use of this process. This work
aims to elucidate the end milling machinability of GFRP composites with respect to
surface roughness, delamination and machining forces.
Milling GFRP composite materials is rather a complex task owing to its
heterogeneity and the number of problems, such as surface delamination, which
appears during the machining process, associated with the characteristics of the
material and the cutting parameters. Due to two phases of materials with drastically
distinguished mechanical and thermal properties, they bring in complicated
interactions between the matrix and the reinforcement during machining. Thus,
measuring and characterising surface properties represent one of the most important
aspects in manufacturing processes. For this study, workpiece material of GFRP
composites with five different fibre orientations are manufactured in the shape of
plates by means of hand lay-up technique.
This research was carried out in three phases. In the first phase, endmilling
experiments on GFRP specimens were carried out to evaluate the influence of fibre
orientation, cutting speed, feed rate and depth of cut on surface roughness and
delamination factor using solid carbide tool coated with PCD by using Response
Surface Methodology (RSM). Milling experiments were designed based on four
factors, five level central composite, rotatable design matrix. Response surface
models were developed. Analysis of variance (ANOVA) was used to check the
iv
validity of the model and for finding the significant parameters. Also, an analysis of
the influences of the entire individual input machining parameters on the responses
was carried out. An Artificial Neural Network (ANN) model was developed to predict
surface roughness and delamination on the milled surface. In the development of
predictive models, cutting parameters of cutting speed, fibre orientation, depth of cut
and feed rate were considered as model variables. Good agreement was observed
between the predictive models results and the experimental measurements. The ANN
and RSM models for GFRPs milled surfaces were compared with each other for
accuracy and computational time. Cutting with minimum machining force is a prerequisite for endmill cutters, especially for those used for milling of composite
materials. Since the helix angle of the end mill cutter has a major influence on the
magnitude of the machining force, modifying its geometry produces improvements in
performance. In the second phase, milling experiments were performed on 15, 60
and 105 GFRP plates based on Taguchi L9 orthogonal array to evaluate the influence
of helix angle, spindle speed, feed rate and depth of cut on machining force.
Regression models were developed to predict the influence of each parameter on the
machinability of GFRP composites and examine which parameter has higher effect
and becomes more important using ANOVA technique. In third phase, Taguchi
method with fuzzy logic was employed to optimise the multiple performance
characteristics of the GFRP endmilling process. A Multi-Response Performance Index
(MRPI) was formulated for optimisation purpose. Experiments were conducted based
on L27 orthogonal array. The machining parameters viz., tool geometry (helix angle of
the endmill cutter), spindle speed, feed rate, depth of cut and work piece (fibre
orientation),
were
optimised
with
consideration
of
multiple
performance
characteristics viz., machining force, delamination, and surface roughness. The results
from confirmation runs indicated that the determined optimal combination of

machining parameters improved the performance of the machining process.
Keywords: GFRP composites; Surface roughness; Delamination factor; Response
surface methodology; ANN; Machining force; Fuzzy logic.
vi
ACKNOWLEDGEMENTS
In the first place, I would like to express my heartfelt gratitude to my Research
Supervisor Dr.R.Jeyapaul, Associate Professor, Department of Production
Engineering, National Institute of Technology Tiruchirappalli (NITT), for his
committed guidance, valuable suggestions, constructive criticisms and sustained help
from the very early stage of this research work that made me to look every nook and
cranny of this research work, for the successful completion.
I take the opportunity to express my heartfelt gratitude to the doctoral
committee
chairman,
Dr.T.Selvaraj,
Professor,
Department
of
Production
Engineering, NITT, for his valuable suggestions and encouragement.

I would like to show my gratitude to the doctoral committee member
Dr. P.Sathiya, Associate Professor, Department of Production Engineering, NITT,
for his constant encouragement and helpful suggestions. I am also grateful to the
doctoral
committee
member
Dr.S.P.Sivapirakasam,
Associate
Professor,
Department of Mechanical Engineering, NITT, for his support and encouragement.

I would like to place on record my sincere gratitude to the Director, NITT,
Dr. S. Sundarrajan and the Dean Academic, Dr.P.Asokan, NITT for providing a
perfect platform to execute this research work. It is my immense pleasure to convey
thanks to Dr.S. Kumanan, Head, Department of Production Engineering, NITT, for
his encouragement and kind concern.
I wish to thank all faculty members and supporting staff members of
Department of Production Engineering, NITT, for their support and cooperation.
I extend my thanks to my family members and fellow friends for their support,
especially the research scholars of Production Engineering and Mechanical
Engineering departments for their kind support and encouragement.
M.P. Jenarthanan
vii
TABLE OF CONTENTS
Title
Page No.
ABSTRACT...
iii
ACKNOWLEDGEMENT....
vii
TABLE OF CONTENTS..
viii
LIST OF TABLES.
xiv
LIST OF FIGURES...
xvi
ABBREVIATIONS....
xviii
NOTATIONS..............
xx
CHAPTER 1
INTRODUCTION
1.1
COMPOSITES: A BRIEF HISTORY...
1.1.1
Matrix phase .
1.1.1.1
Thermoset resins ...
1.1.1.2
Thermoplastic resin ...
1.1.2
Reinforcement ..
1.1.2.1
Continuous strand roving.
1.1.2.2
Woven roving
1.1.2.3
Woven fabrics
1.1.2.4
Reinforcing mat
1.1.2.5
Surfacing mat
1.1.2.6
Chopped fibres .
1.2
CLASSIFICATION OF COMPOSITES .
1.2.1
According to geometries ..
1.2.1.1
Fibrous composite
1.2.1.2
Particulate composite
1.2.2
According to matrix materials .
1.2.2.1
Polymer matrix composites (PMC) .
1.2.2.2
Ceramic matrix composites (CMC) .
1.2.2.3
Metal matrix composites (MMC)
viii
1.2.3
Classification of composites based on fibre

reinforcement .
1.3
PRINCIPLES FOR USING GFRP COMPOSITES ..
1.3.1
Amount and orientation of glass fibre used ..
1.3.2
Rein mix
10
1.3.3
Processes ..
10
1.3.4
Economy ..
10
1.4
MAJOR FABRICATION PROCESSES OF GFRP

COMPOSITES .
11
1.4.1
Hand lay-up process
11
1.4.2
Filament winding process
12
1.5
SIGNIFICANCE OF FRP MACHINING ..
12
1.5.1
Non- Traditional Machining .
12
1.5.2
Conventional machining ..
13
1.6
INTRODUCTION TO MODELLING AND

OPTIMISATION TOOL ...
13
1.6.1
Response Surface Methodology (RSM).
13
1.6.2
Taguchi Methodology .......
14
1.6.3
Analysis of Variance (ANOVA) ...
14
1.6.4
Artificial Neural Network (ANN)
15
1.6.5
Fuzzy logic
15
1.7
THE PRESENT WORK ..
16
1.7.1
Motivation for the present work .
16
1.7.2
Aim
17
1.7.3
Scope .
17
1.8
OVERVIEW OF CHAPTERS .
18
CHAPTER 2
LITERATURE REVIEW
2.1
INTRODUCTION
20
2.2
MACHINING OF GFRP COMPOSITES ..
20
2.3
NON-TRADITIONAL MACHINING (NTM) ..
22
2.4
CONVENTIONAL MACHINING .
23
ix
2.5
MECHANICS OF MACHINING
28
2.5.1
Endmilling operation
29
2.5.2
Fibre orientation
30
2.5.3
Cutting tool geometries
31
2.5.4
Factors influencing surface roughness
36
2.5.5
Factors influencing delamination factor .
38
2.5.6
Factors influencing the machining force .
40
2.6
LITERATURE ON MODELLING AND

OPTIMISATION ..
42
2.6.1
Statistical tools .
43
2.6.2
Non-traditional techniques ..
43
2.6.2.1
Response Surface Methodology ..
45
2.6.2.2
Artificial Neural Network .
46
2.6.2.3
Fuzzy logics .
47
2.7
SUMMARY .
48
CHAPTER 3
ENDMILLING OF GFRP USING SOLID CARBIDE TOOL

COATED WITH POLYCRYSTALLINE DIAMOND (PCD)
3.1
PREPARATION OF TEST SPECIMENS .
49
3.1.1
Introduction .
49
3.2
MATERIAL SELECTION PROCESS ..
49
3.2.1
Glass fibres ..
50
3.2.2
Matrices
51
3.3
FABRICATION PROCEDURE .
51
3.3.1
Hand lay-up technique
51
3.4
MECHANICAL PROPERTIES OF TEST

SPECIMENS
54
3.4.1
Tensile testing ..
54
3.5
EXPERIMENTAL SETUP .
57
3.5.1
Design of experiments .
57
3.5.2
Measurement of surface roughness (Ra) .
61
3.5.3
Measurement of delamination factor (Fd) .
62
3.6
RESPONSE SURFACE METHODOLOGY ..
63
3.6.1
Results and discussions
66
3.6.2
Influence of machining parameters on surface

roughness .
3.6.3
Influence of machining parameters on delamination

factor
3.7
3.8
CHAPTER 4
73
COMPARISON
OF
RSM
PREDICTION
77
AND
EXPERIMENTAL VALUES
81
SUMMARY ........................................
82
ARTIFICIAL NEURAL NETWORK
4.1
INTRODUCTION
83
4.2
ARCHITECTURE
84
4.2.1
Single layer net .
85
4.2.2
Multi layer net ..
86
4.3
SETTING THE WEIGHTS .
86
4.3.1
Supervised Training .
87
4.3.2
Unsupervised Training
87
4.3.3
Fixed-weight nets
87
4.4
ACTIVATION FUNCTION
88
4.5
BACK PROPAGATION NEURAL NETWORK
88
4.6
MODEL DESCRIPTION .
91
4.6.1
Data normalisation ..
91
4.7
NEURAL NETWORK DESIGN
92
4.8
NEURAL NETWORK TRAINING
92
4.9
TESTING AND PERFORMANCE OF BPNN ..
97
4.10
COMPARISON OF RSM AND ANN MODELS
98
4.11
SUMMARY ..
100
CHAPTER 5
EFFECTS OF MACHINING PARAMETERS ON

MACHINING FORCE
xi
5.1
INTRODUCTION
5.2
MACHINE TOOL SETUP AND DATA
102
ACQUISITION
103
5.3
TOOL GEOMETRY (HELIX ANGLE)
104
5.4
TAGUCHI
EXPERIMENTAL
DESIGN
AND
SELECTION OF PARAMETERS .
105
5.5
MEASUERMENT OF MACHINING FORCE
106
5.6
RESULTS AND DISCUSSIONS
108
5.6.1
Influence of helix angle on machining force
112
5.6.2
Influence of fibre orientation on machining force
112
5.6.3
Influence of feed rate on machining force
115
5.6.4
Influence of depth of cut on machining force
115
5.6.5
Influence of spindle speed on machining force ..
115
5.7
MULTIPLE REGRESSION ANALYSIS .
117
5.8
VALIDATION TESTS ..
117
5.9
SUMMARY .
118
CHAPTER 6
OPTIMISATION OF MACHINING PARAMETERS USING

TAGUCHI METHOD WITH FUZZY LOGIC
6.1
INTRODUCTION .
6.2
MULTI-RESPONSE OPTIMISATION USING
119
FUZZY LOGIC .
119
6.3
EXPERIMENTAL DESIGN
120
6.3.1
Experimental parameters
121
6.3.2
Planning the experiments
121
6.3.3
Conducting the experiments
123
6.4
MEASUREMENTS .
123
6.5
TRANSFORMATION OF EXPERIMENTAL
RESULTS INTO S/N RATIO .
124
6.6
FUZZY LOGIC UNIT .
126
6.7
ANALYSIS OF EXPERIMENTAL RESULTS
6.8
USING MRPI AND ANOVA ..
130
CONFIRMATION TEST
133
xii
6.9
DISCUSSION ..
134
6.10
SUMMARY .
135
CHAPTER 7
CONCLUSIONS
137
7.1
SCOPE FOR FUTURE WORK
139
REFERENCES
140
LIST OF PAPERS SUBMITTED ON THE BASIS OF THE THESIS
155
CURRICULUM VITAE.
157
DOCTORAL COMMITTEE.
158
xiii
LIST OF TABLES
Table
Title
No.
Page
No.
3.1
Specifications of Universal testing machine
56
3.2
Mechanical properties of GFRP specimens .
56
3.3
Specification of the CNC milling machine ..
57
3.4
Process control parameters and their limits .
58
3.5
Specifications of the tool ..
60
3.6
Layout of central composite rotatable design with results ..
65
3.7
ANOVA for surface roughness (before elimination) ...
67
3.8
ANOVA for delamination factor (before elimination) .
68
3.9
ANOVA for surface roughness (after backward elimination) .
70
3.10
ANOVA for delamination factor (after backward elimination) ..
70
3.11
Response table for surface roughness .
72
3.12
Response table for delamination factor
73
4.1
Correlation coefficient between the network prediction and the

experimental values using the entire dataset of different network
architecture and trained output parameters .
94
4.2
Comparison between RSM and ANN models
100
5.1
Process control parameters and their levels
106
5.2
Experimental machining forces for 15 fibre orientation GFRP
109
composite ..
5.3
xiv
5.4
composite ..
109
109
composite ..
5.5
ANOVA for the machining force for the three composite materials
116
5.6
Validation test results with error percentage ..
118
6.1
Machining parameters and their levels
121
6.2
Layout of L27 orthogonal array
122
6.3
Experimental results with S/N ratio
125
6.4
MRPI values
129
6.5
Response table for MRPI
131
6.6
Results of ANOVA .
132
6.7
Results of machining performance using the initial and optimal
134
machining parameters
xv
LIST OF FIGURES
Figure
Title
No.
Page
No.
1.1
Typical hand lay-up process ..
11
1.2
Filament winding process ..
12
1.3
Scheme of Research ...
19
3.1
Hand lay-up process step by step procedure ..
52
3.2
Tensile test specimen .
55
3.3
Universal testing machine (UTM) .
55
3.4
Notation of the fibre orientation with respect to cutting tool

movement ..
3.5
59
Fixation of GFRP composite material by using clamps in the

machining center
59
3.6
Solid carbide tool coated with PCD ..
60
3.7
Milled GFRP plates with PCD tool ..
61
3.8
Diagram of surface roughness measurement
62
3.9
Delamination factor measurement
63
3.10
Correlation graph for (a) surface roughness (b) delamination ..
72
3.11
Illustration of factor effects on surface roughness
73
3.12
Illustration of factor effects on delamination factor ..
73
3.13
3D Response surfaces & Contour Plots for Surface Roughness

(a) Response surface for Cutting speed Vs Feed (b) Contour Plot
for Cutting speed Vs Feed (c) Response surface for Cutting speed
Vs Fibre orientation (d) Contour Plot for Cutting speed Vs Fibre
orientation (e)Response surface for Depth of cut Vs Feed and
(f) Contour Plot for Depth of cut Vs Feed .
3.14
74
3D Response surfaces & Contour Plots for delamination factor

(a) Response surface for Cutting speed Vs Feed (b) Contour Plot
for Cutting speed Vs Feed (c) Response surface for Cutting speed
Vs Fibre orientation (d) Contour Plot for Cutting speed Vs Fibre
orientation (e)Response surface for Depth of cut Vs Feed and
(f) Contour Plot for Depth of cut Vs Feed ...
3.15
Comparison of experimental surface roughness values with RSM
xvi
78
prediction ...
3.16
81
Comparison of experimental delamination factor values with RSM

prediction .
82
4.1
Simple neural network model ..
84
4.2
Single neural network model
86
4.3
Multi neural network model .
86
4.4
Flow diagram
90
4.5
Schematic illustration of neural network model
91
4.6
Experimental surface roughness values with ANN prediction .
96
4.7
Experimental delamination factor values with ANN prediction ..
97
4.8
Comparison of RSM prediction error with ANN prediction error ...
99
5.1
Fixation of GFRP composite plate on dynamometer by using

clamps in the machining centre .
103
5.2
Experimental setup with data acquisition system
104
5.3
Solid carbide endmill with different helix angles ..
105
5.4
Effect of factors on machining force for 15 fibre orientation GFRP

plate 110
5.5
Effect of factors on machining force for 60 fibre orientation GFRP

plate ..
5.6
111
Effect of factors on machining force for 105 fibre orientation

GFRP plate
111
5.7
Influence of fibre orientation on machining force .
113
5.8
Influence of fibre orientation on machining force for 15 GFRP

plate ..
5.9

plate .
5.10
113
114

plate .
114
6.1
Membership functions for machining force .
127
6.2
Membership functions for surface roughness ..
128
6.3
Membership functions for delamination factor ....
128
6.4
Membership functions for MRPI .
129
6.5
MRPI response graph ..
132
xvii
ABBREVIATIONS
AFRP
Aramid Fibre Reinforced Polymer
AISI
American Iron and Steel Institute
ANN
Artificial Neural Network
ANOVA
Analysis of variance
ASTM
American Society for Testing and Materials
AWJ
Abrasive Water Jet
BP
Back-Propagation
BPNN
Back-Propagation Neural Network
CBN
Cubic Boron Nitride
CFRP
Carbon Fibre Reinforced Polymer
CMC
Ceramic Matrix Composites
CNC
Computer Numerical Control
CVD
Chemical Vapour Deposition
DOE
Design of Experiments
DOF
Degrees of Freedom
EDM
Electric Discharge Machining
FEM
Finite Element Method
FRP
Fibre Reinforced Polymer
GA
Genetic Algorithm
GFRP
Glass Fibre Reinforced Polymer
ISO
International Dimensional Precision
LEFM
Linear Elastic Fracture Mechanics
MMC
Metal Matrix Composites
MRA
Multiple Regression Analysis
MRO
Multi-Response Optimisation
MRPI
Multi-Response Performance Index
MSE
Mean Square Error
NTM
Non-traditional Machining
OA
Orthogonal Array
PCD
Poly-Crystalline Diamond
PMC
Polymeric Matrix Composites

xviii
RSM
Response Surface Methodology
UD-CFRP
Uni-Directional Carbon Fibre Reinforced Polymers
UD-GFRP
Uni-Directional Glass Fibre Reinforced Polymers
xix
NOTATIONS
d
Depth of cut
Fd
Delamination factor
Fm
Machining force
F- ratio
Variance ratio
Fx
Feed Force, N
Fy
Cutting Force, N
Fz
Thrust Force, N
Feed rate, mm/min
Spindle speed, rpm
Ra
Surface roughness, m
Cutting speed, m/min
Width of cut, mm
Wmax
Maximum width of damage, mm
x1
Input variable
Xi
Required coded value of parameter
Xmax
Upper level of parameter
Xmin
Lower level of parameter
Response variable
Fibre orientation, deg
Helix angle, deg
Random error
xx
xxi
CHAPTER 1
INTRODUCTION
1.1 COMPOSITES: A BRIEF HISTORY
The technologically advancing society is continuously challenging the limits
of conventional materials and is placing newer demands on material performance.
Extreme and sometimes conflicting requirements are forcing us to engineer materials
that are not possible to manufacture by conventional alloying methods. Composite
materials (or composites, for short) are engineering materials which are made from
two or more constituent materials that remain separate and distinct on a macroscopic
level while forming a single component. Composite materials form a material system
composed of a mixture or combination and are insoluble in each other. Composite
material comes under one class of engineered material developed specifically to meet
such a challenge. Mankind has been aware of composite materials several hundred
years Before Christ and since has been applying with innovations to improve the
quality of life. Contemporary composites resulting from research and innovation from
the past few decades have progressed from glass fibre for automobile bodies to
particulate composites for aerospace and to a range of other applications. The volume
and number of applications of composite materials have grown steadily, penetrating
and conquering new markets relentlessly. Modern composite materials constitute a
significant proportion of the engineered materials in the market ranging from
everyday products to sophisticated niche (hollow in a wall or statues) applications.
Glass fibre reinforced resin matrix composites were first introduced in the
early 1940s. Since then, the use of composites has grown rapidly in various industries
such as aircraft, marine, automobile, sporting goods, etc. While composites have
already proven their worth as weight-saving materials, the current challenge is to
make them cost effective. The efforts to produce economically attractive composite
components have resulted in several innovative manufacturing techniques which are
currently being used in the composites industry. Composites that form heterogeneous
structures which meet the requirements of specific design and function with desired
properties limit the scope for classification.
1
Some of the advantages of composites include high specific strength, high

specific stiffness, fatigue strength and impact resistance, thermal conductivity,
corrosion resistance, and good dimensional stability. Composite materials are usually
designed to possess certain specific properties desirable in that application. Unusual
combination of properties not easily obtainable with alloys such as higher fracture
roughness, higher oxidation and corrosion resistance, directional properties, good
resistance to heat, cold and moisture, ease of fabrication and low cost could be
brought out of course, not all together simultaneously.
A Composite Material consists of two phases:
1) Matrix phase
2) Reinforcement
1.1.1 Matrix phase
The matrix or resin is the major component of an FRP/Composite. Matrix is
usually more ductile and less hard. Resin systems are selected for their chemical,
electrical and thermal properties. The two major classes of resins are thermoset and
thermoplastics.
1.1.1.1 Thermoset resins
Thermosetting polymers are usually liquid or low melting point solids that can
easily combine with fibres or fillers prior to curing. Thermoset feature cross-linked
polymer chains become solid during a chemical reaction or cure with the application
of a catalyst and heat. The high level of cross-linking provides for reduced creep
compared to thermoplastics. The thermoset reaction is essentially irreversible.
Among the thermoset resins for FRP /Composites, the family of unsaturated
polyesters is by far the most widely used. These resins are practically suitable for
every moulding process available for thermoset. Polyesters offer ease of handling,
low cost, dimensional stability, and a balance of good mechanical, chemical, and
electrical properties. They can be formulated for high resistance to acids, weak
alkalies and organic solvents. They are not recommended for use with strong alkalis.
Other formulations are designed for low or high temperature processing, for room
temperature or high-temperature cure, or for flexible or rigid end products.
Vinylesters provide excellent resistance to water, organic solvents and alkalis, but
less resistance to acids than polyesters. Vinylesters are stronger than polyesters and
2
more resilient than epoxies. Moulding conditions for Vinylesters are similar to those
for polyesters. Epoxies are another family of thermoset resins used in
FRP/Composites. They have excellent adhesion properties and are suited for service
at higher temperatures some as high as 500F. Epoxy-matrix FRP/Composites are
processed by any of the thermoset methods. Epoxies are more expensive than
polyesters, have longer cure times, but their extended range of properties can make
them the cost/performance choice for critical applications. Epoxy/fibre structures
have generally higher fatigue properties than polyesters.
Polyurethanes are a family of resins that offer very high toughness, high
elongation, faster cure times and good coupling to a variety of reinforcements.
Polyurethanes are easily foamed in a controlled process to produce a wide range of
densities. Additives are easily incorporated into resin systems to provide
pigmentation, flame retardance, weather resistance, superior surface finish, low
shrinkage and other desirable properties. Gel coats consisting of a special resin
formulation provide an extremely smooth next-to-mould surface finish on
FRP/Composites. They are commonly applied in hand lay-up and spray-up processes
to produce a tough, resilient, weather-resistant surface. Gel coats, which may be
pigmented, are sprayed onto the mould before the reinforcement and resin are
introduced. Other thermosetting resin systems, generally formulated with chopped
strand or milled fibre reinforcement for compression or transfer moulding are:
Phenolics: Good acid resistance, good fire/smoke, and thermal properties.
Silicones: Highest heat resistance, low water absorption, excellent dielectric
properties.
Melamines: Good heat resistance, high impact strength.
Diallyl phthalates: Good electrical insulation, low water absorption.
1.1.1.2 Thermoplastic resins
Thermoplastic polymers can soften and become viscous liquids when heated
for processing and then become solid when cooled. The process is reversible allowing
a reasonable level of process waste and recycled material to be reused without
significant effect on the end product. Thermoplastic resins allow for faster moulding
cycle times because there is no chemical reaction in the curing process. Parts may be
formed as fast as heat can be transferred into and out of the moulding compound.
3
Polypropylene and polyethylene are the most common thermoplastic resins

used in FRP/Composites. They have excellent resistance to acids and alkalies and
have good resistance to organic solvents. Their relatively low melting points allow for
rapid processing at lower cost. Nylon and Acetyl are highly resistant to organic
solvents and may also be used where increased mechanical properties are required.
1.1.2 Reinforcement
Reinforcements are supplied in several basic forms to provide flexibility in
cost, strength, compatibility with the resin system, and process requirements.
Regardless of the final form, all fibre reinforcements originate as single filaments. A
large number of filaments are formed simultaneously and gathered into a strand. A
surface treatment is then applied to facilitate subsequent processing, maintain fibre
integrity, and provide compatibility with specific resin systems. After this treatment,
the strands are further processed into various forms of reinforcements for use in
moulding FRP/Composites.
1.1.2.1 Continuous strand roving
This basic form of reinforcement is supplied in the form of untwisted strands
wound into a cylindrical package for further processing. Continuous roving is
typically chopped for spray-up, sheet moulding compounds. In the continuous form,
it is used in pultrusion and filament-winding processes.
1.1.2.2 Woven roving
Woven from continuous roving, a heavy, drapable fabric is available in
various widths, thicknesses and weights. Woven roving costs less than conventional
woven fabric and is used to provide high strength in large structural components such
as tanks and boat hulls. Woven roving is used primarily in hand lay-up processing.
1.1.2.3 Woven fabrics
Made from fibre yarns, woven fabrics are of a finer texture than woven
roving. They are available in a broad range of sizes and weights. Various strength
orientations are also available.
1.1.2.4 Reinforcing mat

Made from either continuous strands laid down in a swirl pattern or from
chopped strands, reinforcing mat is held together with a resinous binder or stitched
mechanically. These mats are used for medium strength FRP/Composites.
Combination mat, consisting of woven roving and chopped strand mat bonded
together, is used to save time in hand lay-up operations. Hybrid mats of glass and
carbon and aramid fibres are also available for higher strength reinforced products.
1.1.2.5 Surfacing mat
Surfacing mat or veil is a thin fibre mat made of monofilament and is not
considered a reinforcing material. Rather, its purpose is to provide a good surface
finish because of its effectiveness in blocking out the fibre pattern of the underlying
mat or fabric. Surfacing mat is also used on the inside layer of corrosion-resistant
FRP/Composite products to produce a smooth, resin-rich surface.
1.1.2.6 Chopped fibres
Chopped strands or fibres are available in lengths from 1/8 to 2 for blending
with resins and additives to prepare moulding compounds for compression or
injection moulding and other processes. Various surface treatments are applied to
ensure optimum compatibility with different resin systems.
1.2 CLASSIFICATION OF COMPOSITES
1.2.1 According to Geometries
Most composite materials developed thus so far have been fabricated to
improve mechanical properties such as strength, stiffness, toughness, and high
temperature performance. It is natural to study together the composites that have a
common strengthening mechanism. The strengthening mechanism strongly depends
on the geometry of the reinforcement. Therefore, it is quite convenient to classify
composite materials on the basis of the geometry of a representative unit of
reinforcement.
1.2.1.1 Fibrous composite
A fibre is characterised by its length being much greater compared to its
cross-sectional dimensions. The dimensions of the reinforcement determine its
5
capability of contributing its properties to the composite. Fibres are very effective in
improving the fracture resistance of the matrix since a reinforcement having a long
dimension discourages the growth of incipient cracks normal to the reinforcement
that might otherwise lead to failure, particularly with brittle matrices. Man-made
filaments or fibres of non polymeric materials exhibit much higher strength along
their length since, large flaws, which may be present in the bulk material, are
minimised because of the small cross-sectional dimensions of the fibre. In the case of
polymeric materials, orientation of the molecular structure is responsible for high
strength and stiffness.
Fibrous composites can be broadly classified as single layer and multi layer
composites on the basis of studying both the theoretical and the experimental
properties. Single layer composites may actually be made from several distinct layers
with each layer having the same orientation and properties and thus the entire
laminate may be considered a single layer composite. Most composites used in
structural applications are multilayered; that is, they consist of several layers of
fibrous composites. Each layer or lamina is a single layer composite and its
orientation is varied according to the design. Several identical or different layers are
bonded together to form multilayered composite usable for engineering applications.
When the constituent materials in each layer are the same, they are simply called
laminates. Hybrid laminates refer to multilayered composites consisting of layers
made up of different constituent materials. Reinforcing fibres in a single layer
composite may be short or long compared to its overall dimensions. Composites with
long fibres are called continuous fibre reinforced composites and those with short
fibres, discontinuous fibre reinforced composites. The continuous fibres in single
layer composites may be all aligned in one direction to form a unidirectional
composite. Such composites are fabricated by laying the fibres parallel and saturating
them with resinous material. The bidirectional reinforcement may be provided in a
single layer in mutually perpendicular directions as in a woven fabric. The
bidirectional reinforcement may be such that the strengths in two perpendicular
directions are approximately equal. The orientation of discontinuous fibres cannot be
easily controlled in a composite material. So fibres can be either randomly oriented or
preferred oriented. In most cases the fibres are assumed to be randomly oriented in
the composites. However, in the injection moulding of a fibre reinforced polymer,
6
considerable orientation can occur in the flow direction and it is a case of preferred
oriented fibres in the composites.
1.2.1.2 Particulate Composites
As the name itself indicates, the reinforcement is of particle nature (platelets
are also included in this class). It may be spherical, cubic, tetragonal, a platelet, or of
other regular or irregular shape, but it is approximately equi-axed. In general,
particles are not very effective in improving fracture resistance but they enhance the
stiffness of the composite to a limited extent. Particle fillers are widely used to
improve the properties of matrix materials such as to modify the thermal and
electrical conductivities, improve performance at elevated temperatures, reduce
friction, increase wear and abrasion resistance, improve machinability, increase
surface hardness and reduce shrinkage. Also, in case of particulate reinforced
composites the particle can be either randomly oriented or preferred oriented.
1.2.2 According to matrix materials
Modern composite materials, depending on the matrix materials used, can be
classified as Polymer Matrix Composites (PMC), Ceramic Matrix Composites (CMC)
and Metal Matrix Composites (MMC).
1.2.2.1 Polymer Matrix Composites (PMC)
Polymer matrix composites are found to be the most commonly used
advanced composites because of their low cost, high strength, and simple
manufacturing techniques. These composites consist of a polymer reinforced by
smaller-diameter fibres. PMCs can be grouped into three different categories. The
grouping is based to a large degree, on the type of fibre reinforcement utilized in the
composite matrix. A variety of polymers may be used for each type of PMC. The
three groups are Glass Fibre-Reinforced Polymer (GFRP), Carbon Fibre-Reinforced
polymer (CFRP) and Aramid Fibre-Reinforced Polymer Composites (AFRP). Glass
fibre reinforced plastics are most commonly used materials in view of their relatively
high specific strength and low cost. Other materials provide higher specific strength,
higher specific stiffness and light weight. They are, however, expensive and are used
only for those applications where performance rather than cost is the major
consideration. Aramid is used instead of carbon where strength, lightness and
7
flexibility are major considerations and are not so important on stiffness and high
temperature performance. Polyester resins are lower in cost and are not as a strong as
the epoxy. Their use in composite includes boat hulls, structural panels, appliances,
etc. Epoxy, in addition, has a lower shrinkage after cure. It is used commonly in
carbon and aramid fibre composites. Maximum usable temperature of polymeric
matrix composites is relatively low, as the matrix materials is prone to softening or
chemical decomposition at moderate temperatures.
1.2.2.2 Ceramic Matrix Composites (CMC)
Ceramic matrix composites are on the rise mainly to improve fracture
toughness, in addition to their higher specific modulus and elevated temperature and
mechanical properties that are superior to metal. Continuous fibre, discontinuous
fibre or particulates can be used as reinforcing material. The common fibre materials
used are alumina and silicon carbide. Other CMCs include carbon/carbon composite
in which high strength carbon fibres are embedded in a graphite matrix. The low
density of carbon in combination with the extraordinary strength of carbon fibres
offers potential for the development of high specific strength material.
1.2.2.3 Metal matrix composites (MMC)
Metal matrix composites are used for applications which require higher
operating temperature than that are possible with polymer matrix composites. Most of
these alloys are developed especially for use in aerospace industry; but newer
applications are found in auto industry such as in automobile engine parts, brake
drums, brake shoes etc. Continuous fibres provide the highest stiffness and strength
for metal matrix composites.
The discontinuous and particulate MMCs are of low cost that provides higher
strength, stiffness and better dimensional stability over metal alloys. They provide
increased wear resistance and contribute towards the difficulty in machining these
materials. These alloys are used for sporting equipment, automobile engine parts,
missile guidance parts, etc. even though they are costlier today.
1.2.3
Classification of composites based on fibre reinforcement

Fibres are the most important class of reinforcements, as they satisfy the
desired conditions and transfer strength to the matrix constituent influencing and
8
enhancing their properties as desired. The performance of the fibre composites is

judged by its length, shape, and orientation, composition of the fibres and the
mechanical properties of the matrix. Based on the reinforcement shape and position,
the composites can be classified as
1. Laminar Composites
2. Flake composites
3. Filled composites
4. Particulate composites
1.3 PRINCIPLES FOR USING GFRP COMPOSITES

There are four major principles that should be considered in using glass fibres
as composite reinforcement. Mechanical properties depend on the combined effect of
the amount of glass fibre reinforcement used and its arrangement in the finished
composite. Chemical, electrical and thermal performance is influenced by the resin
system used as the matrix. Material selection and design, and production
requirements determine the proper fabrication process to be used. The cost,
performance and the value achieved in the finished composite depend upon good
design and judicious selection of raw materials and process.
1.3.1 Amount and orientation of glass fibre used
The strength of finished product is directly related to the amount of glass fibre
in the finished product. Generally speaking, strength increases directly in proportion
to the amount of glass fibre used. A component containing 80% glass fibre and 20%
resin by weight is almost four times stronger than a part containing opposite amounts
of these two materials. Equally important is the arrangement of glass fibre in the
finished product. When all the strands are laid parallel to each other, maximum
strength modulus is obtained in the fibre direction. Such a parallel arrangement is
used in the design of rocket motor cases, golf and fishing rods. When half the strands
are laid at right angles to the other half, the strength is the highest in those two
directions. Although the strength is lesser in parallel arrangement, it is still
considerable. Woven fabric/Chopped strand and mat laminates find application in
boats, airplane wing tips and swimming pools. When glass fibres are arranged in a
random manner, strength is no longer concentrated in one or two directions. Safety
helmets, chairs, electrical parts, luggage and machine housings utilise this
arrangements. This random arrangement results in equal but lower strength in all
directions. This condition is called isotropic.
9
1.3.2 Resin mix

The major resins used in glass fibre reinforced plastics vary in resistance to
corrosion and heat. Formulation of the resin mix also influences corrosion and heat
resistance but has a less pronounced effect. By varying ingredients such as filler,
pigment and catalyst system, each resin mix can be made to vary in performance.
Resin also helps to prevent abrasion of the glass fibres by maintaining the position of
the fibres and keeping them separated. Polyester resins are used in approximately
85% of all glass fibre reinforced plastics because they are economical. Other resins in
use are epoxies, phenolics, silicones, melamines, acrylics and polyesters modified
with acrylics. Some thermoplastic resins such as nylon, polystyrene, polycarbonate
and fluorocarbons are reinforced with fibre glass.
1.3.3 Processes
Processes vary in their capacity to utilise different arrangements of glass,
different amounts of glass and different resins. A given combination of raw materials
required to meet performance criteria in a given application narrows the choice of
processes to those which can successfully and economically form the raw material
into a completed part. Production flexibility of a process is often the single most
important economic factor. For example, if a large number of parts are made from
one mould, the lowest total cost is achieved by using presses and moulds and
automating materials handling. Conversely, if only a few parts are required, a process
minimising investment in moulds and other equipment would be the logical choice.
Continuous forming which are being used for bench slats, FRP panelling etc., can
greatly improve the economics of large volume items.
1.3.4 Economy
Economy in terms of cost and performance results from good design based on
judicious selection of both raw materials and process. Proper materials must be
combined in a process or processes so that potential performance is realised at an
economical cost of manufacturing. Design of the part must take advantage of the
material's maximum capabilities. The Wide application of glass fibre reinforced
plastic in engineering field and the less processing cost are the reasons for selecting
this material for the present investigation.
10
1.4 MAJOR FABRICATION PROCESSES OF GFRP COMPOSITES

Many processes are available to produce the desired combination of design
performance and economics of glass fibre composites. Each process has its own
usefulness for combining different kinds and amounts of glass and resin. The basic
processes can be categorized into two classes: open mould and closed mould
processes. The open mould process includes hand lay-up, spray-up, vacuum pressure
bag, autoclave, filament winding and continuous pultrusion. The closed mould
process includes matched die moulding, injection moulding and continuous
laminating. The predominant fabricating processes used in making laminates and
structures are briefly explained hereunder:
1.4.1 Hand lay-up process
Hand lay-up process is the oldest and the simplest glass fibre reinforced
plastic forming process. In fabrication, the glass fibres with resin are placed in or on
the mould and the entrapped air is removed with squeezing rollers. Layers of glass
and resin are added to build up a desired thickness; if a high quality surface is desired,
a gel coat is applied on the mould prior to lay-up. Figure 1.1 shows the typical hand
lay-up process.
Figure 1.1 Typical Hand Lay-up processes
The lay-up normally cures at room temperature but heating may be used to
accelerate care. The exposed surface is generally rough but it can be made smoother
by wiping on cellophane or other suitable releasing films such as Mylar or Polyvinyl
alcohol. Resins used in hand lay-up are usually polyesters epoxies.
11
1.4.2 Filament winding process

Filament winding uses continuous reinforcement to achieve efficient
utilisation of glass fibre strength. Roving or single strands are red from a creel
through a bath of resin and wound on a suitably designed mandrel. Pre-impregnated
roving may also be used. Special winding machines lay down glass in a
predetermined pattern to give maximum strength in the direction required. When the
desired number or layers have been applied, the wound mandrel is cured at room
temperature or in an oven. A method of filament winding is illustrated in Figure 1.2.
Figure 1.2 Filament winding process
1. 5 SIGNIFICANCE OF FRP MACHINING

1.5.1
Non -Traditional Machining (NTM)

In recent years the fibre reinforced composite materials are used in various
fields of science and engineering because of their unique properties. As a result, there
is a strong need to understand issues associated with the manufacture of composite
components. The existing manufacturing technique of fabricating to near-net shape is
incomplete unless the component is subjected to secondary machining operations like
milling, trimming, finish grinding, drilling etc. based on the requirement.
Non-Traditional Machining (NTM) techniques were used in many of the
applications as a secondary machining operation. However, NTM techniques also
pose some limitations. As the equipment costs are very high, the process is
considered to be highly expensive for small-scale production. Moreover, they are
12
material oriented. Normally, NTM process cannot be used to work on any FRP
material. Aramid is the suitable material for laser machining, because of closeness in
properties of fibre and resin. To use Electric Discharge Machining (EDM) process,
the material should be electrically conductive. Abrasive water jet machining produces
higher noise levels. In laser machining problems like delamination, uneven kerfs and
thermal cracks are observed. These processes are incapable of producing blind
holes/cavities. As these processes are highly expensive, they are suitable for
application in industries where manufacturing cost is of secondary importance.
1.5.2 Conventional machining
Alternatively, conventional machining on FRP composite material plays a
vital role in meeting dimensional accuracy and good surface quality requirement. The
process is cheaper than NTM techniques. As a result of owing to the potential
application there is a strong need to understand the issues associated with
conventional machining of FRP composite materials.
Among the machining processes, milling is most frequently used in industries
due to the need for assembly components in mechanical structures. However, the
properties of FRP materials like anisotropy, non-homogeneity, and abrasiveness
provide problem like excessive tool wear, poor surface finish, delamination, fibre
pullout, dimensional variation, etc. These limitations motivated the present research
which revolves around the study on the problems of milling on FRP materials.
1.6 INTRODUCTION TO MODELLING AND OPTIMISATION TOOLS
1.6.1 Response Surface Methodology (RSM)
Response surface methodology is a series of mathematical and statistical
techniques used for modelling and analysing problems and has the objective of
optimising the responses. It is a sequential experimentation strategy for empirical
model building and optimisation. By conducting experiments and applying regression
analysis, a model of the response to some independent input variables can be
obtained. Based on the model of the response, a near optimal point can then be
deduced. RSM can determine and represent the effect of relationships among input
control factors that influence the responses as a two- or three-dimensional hyper
surface. Most of the work in RSM has been focused on the case where there is only
13
one response of interest. In product or process development, however, it is quite

common that several response variables are of interest. In this case, determination of
optimum conditions on the input variables would require simultaneous consideration
of all the responses; this is called a multi-response problem. Several approaches have
been proposed for Multi Response Optimisation (MRO) including the desirability
function approach and loss function approach. A number of investigations using RSM
are carried out to determine the significant factors affecting the response.
1.6.2 Taguchi Methodology
Taguchis method has been used widely in engineering analysis. It provides a
simple efficient and systematic approach to optimise design for performance quality
and cost. These techniques consist of a plan of experiments with the objective of
acquiring data in a controlled way, executing these experiments, in order to obtain
information about the behaviour of a given process. The main thrust over Taguchi
technique is the use of parameter design which is an engineering method for product
or process design that focuses on determining the parameters settings producing the
best levels of quality characteristics with minimum variation.
1.6.3 Analysis of Variance (ANOVA)
Analysis of variance is a mathematical technique which is based on the least
square approach. The purpose of ANOVA is to investigate the process parameters
which significantly affect the performance characteristics. As per this technique, if
the calculated value of the F ratio of the developed model does not exceed the
standard tabulated value of F ratio for a desired level of confidence, then the model is
considered to be adequate within the confidence limit. The variance ratio, denoted by
F in ANOVA tables, is the ratio of the mean square due to a factor and the error mean
square. In robust design, F ratio can be used for qualitative understanding of the
relative factor effects. A high value of F means that the effect of that factor is large
compared to the error variance. So, the larger the value of F, the more important is
that factor in influencing the process response.
14
1.6.4 Artificial Neural Network (ANN)

An artificial neural network is an information-processing system that has
certain performance characteristics in common with biological neural networks. ANN
has been developed as generalisations of mathematical models of human cognition or
neural biology, based on the assumptions:
1. Information processing occurs at many simple elements called neurons.
2. Signals are passed between neurons over connection links.
3. Each connection link has an associated weight, which, in a typical neural
net, multiplies the signal transmitted.
4. Each neuron applies an activation function (usually nonlinear) to its net
input (sum of weighted input signals) to determine its output signal.
In the past decades, numerous studies have been reported on the development
of neural networks based on different architectures. Basically, one can characterise
neural networks by its important features, such as the architecture, the activation
functions, and the learning algorithms. Each category of the neural networks would
have its own input, output characteristics, and therefore it can only be applied for
modeling some specific processes.
1.6.5 Fuzzy Logic
Fuzzy Logic was initiated in 1965 by Lotfi A. Zadeh, professor of computer
science at the University of California in Berkeley. Basically, Fuzzy Logic is a
multivalued logic, which allows intermediate values to be defined between
conventional evaluations like true/false, yes/no, high/low, etc. Notions like rather tall
or very fast can be formulated mathematically and processed by computers, in order
to apply a more human-like way of thinking in the programming of computers. A
fuzzy system is an alternative to traditional notions of set membership and logic that
has its origins in ancient Greek philosophy. The precision of mathematics owes its
success in large part to the efforts of Aristotle and the philosophers who preceded
him. In their efforts to devise a concise theory of logic, and later mathematics, the socalled Laws of Thought were posited. One of these, the Law of the Excluded
15
Middle, states that every proposition must either be true or false. Even when
Parmenides proposed the first version of this law (around 400 B.C.) there were strong
and immediate objections: for example, Heraclitus proposed that things could be
simultaneously true and not true. It was Plato who laid the foundation for what would
become fuzzy logic, indicating that there was a third region (beyond True and False)
where these opposites tumbled about. Fuzzy Logic has emerged as a profitable tool
for the controlling and steering of systems and complex industrial processes, as well
as for household and entertainment electronics, as well as for other expert systems
and applications like the classification of SAR data.
1.7 THE PRESENT WORK
1.7.1 Motivation for the Present Work
The performance of machine components, especially of Fibre Reinforced
Plastics (FRP) in high applications, is often critically dependent on the quality
produced by machining since the surface layer may drastically affect the strength and
chemical resistance of the material. The deterioration in the surface quality or the
damage caused by the machining operation is due to the anisotropic nature of the
fibrous composites. Milling of FRP composites to facilitate the assembly of parts is
one of the most important machining operations. Milling of composites poses
conditions and requirements that are not usually encountered in metal machining. The
mechanism of material removal in orthogonal machining of FRP is not completely
understood till date. Therefore the oblique cutting mechanism of fibrous laminated
FRP composites poses new set of problems which were never encountered earlier.
The major concern while milling the composites is the damage caused to the
workpiece. The principal mechanism of damage is largely delamination.
Delamination is generally regarded as a result of matrix dominated failure behavior,
which occurs in the interplay region. A number of research initiatives have been
undertaken for developing an up to date understanding of the milling process in the
context of FRP composite materials, but still a lot remain uncovered. The mechanism
of the material removal in an oblique cutting process such as milling, the variation or
the forces with time and their influence on accompanying damage, the optimal cutting
tool geometry for minimising the machining forces and the subsequent damage are
some of the key directions to which the research in the machining of composite
16
materials is progressively focused.

As a convention, during the application of the composites, the presence of
slots or grooves for various purposes must be considered. For example, when the slot
is produced by machining after fabrication of the composites material, a reduction in
strength is anticipated because of cutting the fibre bundle, loosening, and de-bonding
of fibre around the slot. These anticipatory aspects motivated to take up the present
investigation.
1.7.2 Aim
To study the influence of fibre orientation and tool geometry while milling of
GFRP material and thereby to suggest an appropriate geometry or modifications on
the existing geometry and suitable machining parameters so that the existing
drawbacks can be avoided.
1.7.3 Scope
To fabricate GFRP composites with different fibre orientations using hand

lay-up process.
To conduct endmilling experiments on GFRP specimens thereby to

evaluate the influence of fibre orientation, cutting speed, feed rate and
depth of cut on surface roughness and delamination factor using solid
carbide tool coated with PCD by using RSM and ANN models.
To perform the milling experiments on 15, 60 and 105 GFRP plates

based on Taguchi L9 orthogonal array to evaluate the influence of helix
angle, spindle speed, feed rate and depth of cut on machining force.
To develop regression models to predict the influence of each parameter

on the machinability of GFRP composites and examine which parameter
has higher effect and becomes more important using ANOVA technique.
To optimise the machining parameters using fuzzy logic with Taguchi

method. The machining parameters viz., tool geometry (helix angle of the
17
endmill cutter), spindle speed, feed rate, depth of cut and work piece (fibre
orientation), were
to be optimised with consideration of multiple
performance characteristics viz., machining force, delamination, and

surface roughness.
1.8 OVERVIEW OF CHAPTERS
The thesis is spread to seven chapters and the scheme of research is illustrated
in figure 1.3.
Chapter 1 gives a brief introduction to FRP composites, fabricating

techniques, and significance of FRP machining.
Chapter 2 highlights the literature review, observations, and dynamics of

milling and scope of present work.
Chapter 3 discusses the preparation and testing of GFRP specimens. It also

covers the experimental study on the milling performance on GFRP using
solid carbide end mill cutter coated with Polycrystalline diamond (PCD) with
various spindle speed, feed, fibre orientation and depth of cut. The
significance of surface roughness and delamination factor is highlighted with
the help of RSM (Response Surface Methodology).
Chapter 4 covers the modelling of surface roughness and delamination factor

using ANN (Artificial Neural Network) and comparison between RSM and
ANN models.
Chapter 5 discusses the effect of tool geometry (25, 35 and 45 helix angle
of the end mill cutter), fibre orientation, spindle speed, feed rate and depth of
cut on machining force and development of regression models.
Chapter 6 covers the optimisation of the machining parameters using fuzzy

logic with Taguchi method.
Chapter 7 covers the summary of the results of the present investigation,

conclusions and the scope for future work.
18
Fabrication of GFRP composites

(ZrB2-SiC)
Hand lay-up process
Fibre orientation
15
30
60
90
120
Impregnation
Post curing
Finished GFRP composite plates
Tensile strength
Mechanical Testing
Youngs
modulus
Shear strength
Endmilling of GFRP
composites
INPUT PARAMETERS
RESPONSES
Tool geometry
Spindle speed
Feedrate
Depth of Cut
Fibre orientation
Machining Force
Surface
roughness
Delamination
Factor
Modelling & Optimisation

Figure 1.3 Scheme of Research
19
Weight Wear
Ratio
CHAPTER 2
LITERATURE REVIEW
2.1 INTRODUCTION
The purpose of literature review is to provide background information on the
issues to be considered in this thesis and to emphasize the relevance of the present
study. This treatise embraces various aspects of polymer composites with a special
reference to machinability characteristics. The present study focuses on machining
characteristics of Glass Fibre Reinforced Plastics (GFRP) which also includes the
application of statistics and various non-traditional techniques. This chapter deals
with the progress so far made in the selected directions. Literature survey was carried
out on machining of GFRP composites. This includes Design of Experiments,
Taguchi method, Response Surface Methodology, Artificial Neural Network,
machining process of GFRP, machining forces and optimisation using Fuzzy logic.
Motivation for the study and the objectives of this study are presented at the end of
the literature review.
2.2 MACHINING OF GFRP COMPOSITES
A few researchers carried out experimental investigations to analyse the
machining characteristics of GFRP composites. The related studies and their reports
are briefly presented in the following contexts.
Polymer matrix composite materials are increasingly used in high
performance applications because of superior strength to weight ratio and
stiffness to weight ratio. Everstine and Rogers (1971) reported that machining of
polymers/ composites was employed when the quantity of items not justify the cost
for moulds, or when a product needs accurate dimensional accuracy, better surface
finish. As high performance polymers have been increasingly used for a large number
of industrial applications, the machining quality is becoming a predominant factor for
the development of new processes and materials. Nevertheless, the knowledge about
the polymer behaviour under machining is very limited, as well as the definition of
suitable models for the prediction of cutting forces. In the scientific literature,
20
machining of plastics is poorly treated. In the ancient references, an experimental

approach is preferred assuming that plastics behave as metals.
Kobayashi (1967) collected several experimental observations in his book
Machining of plastics. This text has been considered as a reference for a long time
in this field. Also the latest scientific reviews mention it to show the dependence of
the cutting forces on process parameters. Hocheng et al. (1993) studied the
machinability of some reinforced thermoset and thermoplastics for drilling operation.
They discussed the chip characteristics and the specific cutting energy to reveal the
mechanism of material removal. They observed that the level for fibre loading and
the deformation behaviour of matrix polymer determined the extent of plasticity in
chip formation and the chip length. In a further study, they also observed that in
drilling fibre reinforced-thermoplastics, the edge quality was generally fine except in
the case of concentrated heat accumulation at tool lips, which was generated by high
cutting speed and low feed rate.
Ramkumar et al. (2004) stated that GFRP composite material was extremely
abrasive when machined. Thus, the selection of the cutting tool and the cutting
parameters is very important in the machining process. Fibre-glass is simply a
composite consisting of glass fibres, either continuous or discontinuous, contained
within a polymer matrix. The machining of composite is different from the
conventional machining of metal due to the composites anisotropic and nonhomogeneous nature.
Owing to its abrasive nature, glass fibre may cause premature rounding of
cutting edges. Distinct difference in hardness between fibre and matrix combined
with the high cutting resistance of fibres results in edge chipping. Since machining
operations for cured laminates are mainly finishing operations, the quality achieved is
of some interest. The major parameters determining the cutting result, especially with
respect to the quality, are discussed and explained, as far as possible by Konieg et al.
(1985, 1990). Mechanical techniques like drilling and routing perform well, if carbide
or PCD cutters with sharp cutting edges are used. Tool geometries as well as cutting
conditions were discussed in this paper. Koplev et al. (1983) reported that when
GFRP composites were machined, it was clearly seen that the fibres are cut across
and along their lay direction, leaving deformed projection and partially disclosed
fibres on the machined surface.
21
2.3 NON-TRADITIONAL MACHINING (NTM)

Many non-traditional machining techniques are available for secondary
machining operation on FRP components after fabrication. However, these
techniques are highly expensive and have their own limitations. The following
literature survey discusses the application of NTM processes on FRP materials.
Godella and Cheng (1993) investigated a wire EDM of various ceramic composites.
They reported that it is possible to lower the electrical resistivity by combining a nonconducting ceramic with a more conducting ceramic. The addition of more
conducting phase may not be detrimental to its performance. Di Ilio et al. (1987 and
1991) reported that aramid fibres, the properties of which are much closer with the
resin would be a better choice for laser cutting than glass and graphite fibres. Some of
the defects in laser cutting include fibre pullout, presence of craters and delamination,
uneven kerfs width and taper in cut surfaces and the presence of charred material and
thermal cracks. Also, these processes find difficulty in machining thick work pieces,
drilling deep holes, blind holes and blind cavities. In addition, the process is highly
expensive.
Koricon (1987) investigated on Abrasive Water Jet (AWJ) machining to cut
resin-impregnated graphite, aramid and glass fibres. He reported that the optimum
cutting rates depend on the type of composite material, its thickness, type and size of
the abrasive particles and jet pressure. Boron, glass and aramid fibres with epoxy
resin are some of the candidate materials on which this process can be employed.
Water used in this process cools the workpiece and hence minimises the
delamination. Low material removal rates, inability to produce blind holes, abrasive
particles of high pressure water jet damaging the machine element, poor dimensional
accuracy and surface finish are some of the limitations of abrasive water jet cutting.
Hashish (1989) presented a technique for precision drilling of small diameter holes in
composites with AWJ and reported that drilling a hole of 0.043" diameter with an
accuracy of 0.001" is possible.
Garlasco (1969) conducted an experiment on ultrasonic machining of boron
fibre composite and found that the technique was more expensive. He recommends
this process as a potential one for applications where intricate shapes of high accuracy
and finish are to be produced. Doran and Maikish (1972) also performed rotary
ultrasonic drilling on the same material using a core drill at speeds up to 6000 rpm.
22
They reported that the process is suitable only when high finish and accuracy are
required and a little higher cost is acceptable. Jain et al. (1990, 1993) investigated on
electro chemical spark machining of Kevlar and glass fibre reinforced epoxy material.
They used copper as tool material and aqueous solution (NaCl) and sodium hydroxide
(NaOH) as electrolyte and showed that higher accuracy is feasible with this
technique.
2.4 CONVENTIONAL MACHINING
The NTM processes discussed so far are highly expensive and suitable for
specific composite. They are inadequate for general engineering products where
manufacturing cost is an important factor. Hence, a high degree of automation for
mass production of composite parts will be required to bring the costs down and
compete with other materials. Eric Persson et al. (1997) compared the performance of
KTH (Kungl Tekniska Hogskolata) method, PCD (Poly Crystalline Diamond) and
dagger type drill tools. Holes drilled by KTH method yielded highest strength and
fatigue life with line hole quality compared to the holes done by PCD and dagger
tools. Di llio et al. (1991) reported that large oscillations of the thrust force detected
during drilling of aramid composite can be interpreted in terms of non-uniform
distribution of thrust force along the tool cutting edge and poor inter-laminar strength
of composites. Sandvik Coromant (1987) reported that the time taken to drill critical
holes in resin impregnated carbon and glass fibre composite material at Westland
helicopters has been reduced from hours to minutes using long series Sandvik
Coromant delta drills. The H grade of carbide was selected as the best cutting
material for composite blades.
Jain and Yang (1994) reported that it is possible to reduce the thrust force
significantly by reducing the chisel edge width. The need to combine small chisel
edge with carbide and PCD materials is emphasized. Radhakrishnan and Wu (1981)
used the on line dynamic data to study the hole quality of composite materials. They
reported that change in the surface waviness can be used to monitor the drill
condition on-line, for appropriate replacement of the drill. Bhattacharya and Horrigan
(1998) performed drilling on Kevlar material with normal and modified tools under
cryogenic and ambient conditions. They reported that by introducing a negative point
angle and high clearance angle (Modified drill with-20 point angle), Kevlar Fibre
23
Reinforced Plastic (KFRP) could be successfully drilled with HSS bits. Application
of liquid nitrogen at the drilling site enhances the tool performance. Hoeheng and
Dharan (1990) reported that, as the end layers of the laminate has got less resistance
to deformation, maximum delamination occurs at the entry and exit side of the hole
mainly because of peel-up and push-out action of the tool. The model developed, is
useful in controlling feed rate and cutting speed in accordance with drill
displacement.
Singh and Khare (1975) performed drilling on PVC sheets; they presented a
relationship between error in hole diameter and drill speeds at different feed rates.
They revealed that very high feed rates result in large thrust and torque value. Sadat
et al. (1992) reported that the feed rate should be minimized at the exit part of the
hole with or without support; laminates get delaminated at a very small drilling load
if uncut portion is small. Wen-Chou Chen (1997) introduced the concept of
delamination factor in his experimental investigation on CFRP material. He reported
that delamination-free drilling is possible by proper selection of tool geometry and
drilling parameters. Furness et al. (1996) performed statistical analysis of the effects
of feed, speed and wear on hole quality in drilling hot rolled steel plates. The concept
of statistical analysis is utilised for the present investigation. Konig et al. (1989)
reported on the role of thrust force and torque on the surface quality.
Chandrasekharan et al. (1995) developed a model to predict force profile
along the cutting lips of the drill. Zhang et al. (2001) developed a model for
predicting mean values of thrust and torque in vibration assisted drilling of fibre
reinforced composite material. They reported that thrust and torque could be reduced
by vibratory drilling to a value of 20% - 30%. Face turning of filament wound GFRP
tubes using high speed steel tool was studied by Santhanakrishnan et al. (1989). They
found that during FRP machining, the HSS tool exhibits deformation at the nose and
cutting edge. Edge deformation was dominant on the primary cutting zone. Malhotra
(1990) reported that drilling fibrous composites presents many problems like
delamination, fibre/resin pull out, excessive tool wear etc. The author used HSS and
Carbide drills and reported the effect of cutting speed and feed on tool wear, thrust
force and torque. Optimisation of drilling process of graphite/ bismaleimide-titanium
alloy in terms of hole quality and cost was reported by Kim and Ramulu (2004).
24
Mathew et al. (1999) concluded that trepanning tools are found to give reduced thrust
while making holes in thin laminated composites.
Davim et al. (2003 and 2004) conducted experiments to select cutting
parameters for damage free drilling in carbon-fibre reinforced epoxy composite
materials. They established a correlation between cutting velocity and feed rate with
the delamination. Experimental investigation was also done on drilling Glass Fibre
Reinforced Plastic (GFRP) to study the effect of cutting parameters in a hand layup
GFRP material. (They also established a relation between cutting velocity and feed
rate with specific cutting pressure, thrust force, damage factor and surface roughness
in a GFRP material).
Tsao et al. (2003 and 2004) studied delamination factor by using twist drill,
candle stick drill and saw drill under various cutting conditions. They established a
relation between feed rate, spindle speed and drill diameter with delamination.
Analytically, the effect of various drill geometries on thrust force was predicted and
compared with twist drill. Abdul Budan and Vijayarangan (2002) conducted
experiments to study the effect of machining parameters and fibre concentration on
the surface finish, hole quality, and delamination. A finite element model was used to
predict the effect of fibre percentage on the thrust force causing delamination.
Murphy et al. (2002) conducted experiments to evaluate the effect of coatings on the
performance of Tungsten Carbide (WC) drills in the drilling of carbon fibre
reinforced epoxy. They used uncoated, Titanium nitride (TiN) coated and diamond
like material coated drills and concluded that the coatings were not found to reduce
either tool wear or damage to the composites.
Velayudham et al. (2005) investigated the dynamics of high volume fraction
glass fibre reinforced composite and correlated with peak drilling thrust, dimensions
of holes drilled and vibration induced during drilling. Aoyame et al. (1995)
investigated small diameter drilling in GFRP laminate and analysed hole quality and
the number of drilled holes. They concluded that the drilled hole quality is dependent
on the tool wear under the same drilling conditions.
25
Ravishankar et al. (2000) used acoustic emission technique to characterise

drilling of composite laminates and concluded that it is possible to evaluate the
drilling induced damages in composites through AE signal characterisation. Konig
and Grab (1989) concluded that in the case of drilling uni-directionally reinforced
laminates the causes of thrust force and torque permit to draw conclusions on the
surface quality. They further concluded that apart from physical properties of the
materials, cutting energy and time of contact affect the course of temperature. These
depend on workpiece, tool and cutting conditions. Brinksmeier et al. (2002) drilled
holes on multi-layer materials and concluded that adapted step drills improve
diameter tolerances, surface quality and tool wear. Improvements are found as for as
tool life is concerned when coatings of Ti32 or diamond are used. Houjiang Zhange et
al. (2001) found that spalling and fuzzing are the major mechanism at exit defects.
Spalling is mainly by chisel edge. Empirical relationships were developed for
assessing characteristic dimension of the spalling damage. Dipaola et al. (1996)
carried out drilling experiments on a unidirectional carbon fibre/epoxy laminates to
understand the crack growth phenomenon as the drill emerges from the exit side of
the workpiece. The damage in the drilled hole and the instantaneous cutting forces
has been correlated with the fibre orientation. Stone et al. (1996) developed a thrust
force controller using critical force curves as a guide, to minimise the delamination
while drilling graphite epoxy-laminates.
Ramulu
et
al.
(2001)
conducted
experiments
on
drilling
Graphite/Bismaleimide (Gr/Bi) Titanium (Ti) stacks, and observed that at the

interface of Gr/Bi-Ti high temperatures induced material damage near and around the
hole region and affected the tool life; Carbide drill outperformed all other tools in
terms of tool life, minimal surface damage, and heat induced damage on both
workpiece materials.
Frederic Lachand et al. (2001) reported that a close correlation exists between
experimental and analytical values when the thrust force of the drill is modelled by
taking into account the geometrical nature of the contact between the tool and a
laminate composite material. Tsao et al. (1997) proposed an analytical approach for
the determination of the position of the onset of delamination during the drilling of
composite laminates based on Linear Elastic Fracture Mechanics (LEFM). Geon Woo
26
Kim et al. (2005) proposed a critical thrust force at propagation of delamination zone
due to drilling of an infinite FRP/ metallic strip by the definition of energy release
rate. Lin et al. (1999) concluded that drill wear is the main problem to be overcome in
high speed drilling of glass fibre reinforced materials. They also reported (1996) that
increasing cutting speed on drilling of carbon fibre reinforced composite materials
accelerates tool wear. Dharan et al. (2000) proposed an intelligent machining scheme
to drill holes below critical thrust force to avoid delamination. Konig et al. (1990)
reported that high strength combined with low weight, are some of the characteristics
of new materials. Full exploitation of the outstanding properties requires machining
processes which are based on respective material features and which preserve the
advantages of the material as well as cost efficiency.
Mariatti et al. (2004) investigated the effect of hole drilling - technique on
tensile properties of Continuous Fibre Impregnated Thermoplastic (COFIT) plain
weave composite using circular hole. They concluded that the tensile properties of the
woven system decrease with increasing hole size. Moreover, the lamination drilled by
orbital techniques showed higher tensile properties. Mansour Mohamed et al. (1993)
evaluated 3D orthogonal woven carbon/epoxy composites with holes for joint
locations. Based on testing results, they concluded that 3D woven composites are not
sensitive for machining because the structure is well bound by Z yarns. They all
concluded that there is no reduction in tensile strength due to drilled hole.
Drilling tests were carried out by Caprino and Tagliaferri (1995) on GFRP
composites in order to verify the effects of machining parameters on the cut quality
and on the mechanical behavior of the material under examination. The experimental
data showed that the width of this damage zone is correlated to the ratio between
drilling speed and feed rate. Particularly, the higher the speed/feed value the better
would be the out quality. The tensile strength of specimens with holes (notched
laminate) is not influenced by the quality of the material at the edge of the hole,
whereas the bearing strength suffers a marked reduction with increasing damage zone
widths. For small damaged zone widths, no correlation seems to exist between the
bearing strength and the damaged zone.
27
2.5 MECHANICS OF MACHINING

The machining process consists of removing a layer of material from blank to
obtain a workpiece of the required shape and dimensions and with specified quality of
surface finish. During the process of material removal there is a relative motion
between the workpiece and cutting tool. Such a relative motion is produced by a
combination of rotary and translatory movement either of the work pieces or of
cutting tool or both. The relative motion depends on the type of machining operation.
Shaw (1992) reported that the machining operation carried out on any machine tool is
based on the theory, which is the same for all processes.
The main aspect of machinability is to discuss in detail the surface finish, tool
geometry and machining force. Though there are a number of interrelated factors
which affect the machinability of the material, the most important factors are the
machining parameters, the properties of the workpiece and the geometry of the
cutting tool. Since material removal is a workpiece related activity, where very strong
economic rate or production rate constraints prevail, it is important that the selection
of cutting tool and work design parameter are based upon a rational optimisation
procedure. The machining parameters are the spindle speed, feed and depth of cut.
The tool parameter involves the tool material and tool geometry. Though the
machining technique on FRP composite material is similar to that of metals, the fibre
proportion and position in FRP composite (fibres in continuous or chopped strands,
fibres in unidirectional or bi-directional) play an important role in deciding the
machinability of the material.
It is a typical feature in the production of FRP composite component to their
net shape almost completely in the transition from soft to hard material stage. This
takes place by a phase change or by the curing of the thermoplastic or thermosetting
matrix inside or on the top of the mould. In spite of this, near net shape production
technology being general, a finish machining of cured FRP will become necessary,
because the required dimensional accuracy and the surface quality may not be
achieved at a reasonable effort in the curing cycle. Due to the inhomogeneity,
anisotropy and abrasive nature of the FRP composite material, the machining
behavior of FRP composite is quite different from that of metals. The present section
28
of the chapter highlights a comparative discussion about the mechanism of milling,

and also the influence of the various factors on the machinability of the FRP
composite material.
2.5.1 Endmilling operation
Machining of polymers/ composites is employed when the quantity of items
does not justify the cost for moulds, or when a product needs dimensional accuracy
and better surface finish. Nevertheless, the knowledge about the polymer behaviour
under machining is very limited, as well as the definition of suitable models for the
prediction of cutting forces.
Jamal Y. Sheikh-Ahmad (2009) stated that milling is one of the most
important machining processes in manufacturing parts made out of FRPs. However,
unlike the milling of metals, which is characterised by high material removal rates,
milling of FRPs is conducted at much lower scale. The reason for this is that FRP
components are largely made near net shape and any subsequent milling is limited
mainly to deburring and trimming as well as achieving contour shape accuracy.
Konig (1985) investigated the phenomenon of machining of FRP composites
using different processes like drilling, routing, milling, water jet cutting and laser
cutting. The machining of FRP was seemed different from that of metal working in
many aspects because of the inhomogeneous material behaviour, dependence on fibre
and matrix properties, fibre orientation and the type of weave.
Ramulu et al. (1993) and Erisken (1999) reported that the surface roughness
and delamination are the characteristics that could influence the dimensional
precision, the performance of mechanical pieces and production costs. For these
reasons there has been research and development with the objective of optimising
cutting conditions, to obtain desired machinability. The works of various authors,
[Koplev et al. (1983), Kaneeda (1989), Puw and Hocheng (1995), Santhanakrishnan
et al. (1988) and Ramulu et al. (1989)], while reporting on milling composite
materials, have shown that the surface quality (surface roughness), machining force
and delamination factor are strongly dependent on cutting parameters, tool geometry
and cutting forces.
29
2.5.2 Fibre orientation

GFRP have gained the greatest importance among the composites since glass
fibre reinforced polyester emerged in the early thirties. Enlarging the field of
technical applications mainly for light weight components, which are mechanically
highly loaded, finally resulted in the development of more efficient reinforcing
materials. Fibre orientation plays a vital role in the surface roughness during
machining. Peaks of roughness are generated with zero degree cut. Owing to the
combined load of bending and compression at 45 cut, fibres are pulled out by
kinking and breaking which resulted in the poorest surface quality. Smoothened
surface can be obtained through 90 cut. The main parameters concerning machining
quality, especially surface roughness and material damage, in drilling fibre reinforced
thermoset were discussed and explained in detail by Konieg and Grab (1989).
Bhatnagar et al. (1995) studied how the fibre orientation influences both the
quality of the machined surfaces and tool wear. The machinability of composite
materials is influenced by the type of fibre embedded in the composites, and more
particularly by the mechanical properties. On the other hand, the selection of cutting
parameters and the cutting tool is dependent on the type of fibre used in the
composites, which is very important in the machining process. Wang and Zhang
(2003) mentioned that fibre orientation is a key factor that determines the surface
integrity of a machined surface and 90 is a critical angle, beyond which a severe
subsurface damage will occur. If the fibre orientation angle is greater than 90, then,
these three distinct deformation zones appear: chipping, pressing and bouncing.
Wang and Zhang (2003) aims to understand the machinability of epoxy
composites reinforced by unidirectional carbon fibres when subjected to orthogonal
cutting. It was found that there was subsurface damage and it is the mechanism of a
machined component which is greatly influenced by fibre orientation. The materials
bouncing back is a characteristic phenomenon associated with the cutting of fibrereinforced composite. Singh and Bhatnagar (2006) presented some observations on
the orthogonal machining of unidirectional GFRP laminates with different fibre
orientations. Isoipescu shear test was adopted to evaluate the inplane shear strength of
varied fibre angle test specimens. A model for predicting the cutting forces and the
dependence of cutting direction on machinability requirements is also presented in the
study.
30
Venu Gopala Rao et al. (2007) investigated the effect of fibre orientation, tool
parameters and operating conditions on fibre and matrix failure and chip size and
reported that the degradation of the matrix adjacent to the fibre occurs first, followed
by failure of the fibre at its rear side. The extent of sub-surface damage due to matrix
cracking and interfacial debonding was also determined in the study. Venu Gopala
Rao et al. (2003) simulated the orthogonal machining of UD-CFRP and UD-GFRP
composites using Finite Element Method (FEM). A two phase micro mechanical
model with fibre assumed elastic and the matrix elasto-plastic was used to estimate
the cutting force during machining. The model assumed plane strain and quasi-static
condition. The failure of fibre was found to be a combination of crushing and
bending, with the bending effect becoming more significant as the fibre orientation
changed from 90 to 15. Eriksen (1999) carried out experiments to study how the
surface roughness of short fibre reinforced thermoplastic machined by turning, was
influenced by variations of the feed rate, cutting speed, tool radius, the fibre
orientation and the interactions between them and finally concluded that the surface
roughness increased for an increasing feed rate, decreasing with the tool radius and
independent of the fibre orientation.
2.5.3 Cutting tool Geometries

Many researchers and scientists developed a lot of models to study the
performance of the cutting tool and their suitability of applications. A detailed list of
their findings is given below:
Ema and Davies (1988) investigated the cutting performance of several end
mills with right and left hand helix angles. Cutting tests were performed on
aluminium alloy L65 for three milling processes in which cutting force, surface
roughness and concavity of a machined plane surface were measured. The cutting
performance of the end mills was assessed using variance analysis. The investigation
showed that end mills with left hand helix angles were generally less effective than
those with right hand helix angles, except for the case where the helix angle was
small. The effects of spindle speed, depth of cut and feed rate on the cutting force and
the surface roughness were generally significant. There was no significant difference
31
between down milling and up milling with regard to the cutting force, although the
difference between them regarding the surface roughness was large.
Paulo Davim (2002) presented an experimental machinability study of the
Metal- Matrix Composites (MMCs) A356/SiC/20p with brazed PCD tools or
Chemical Vapour Deposition (CVD) diamond coated tools. The obtained results
showed that PCD tools were important in cutting these composite types of reduced
machinability. Actually, CVD diamond coated tools show short life, as tools wear
evolution becomes very fast after coating rupture.
Hooper et al. (1999) reported on the machining of the highly abrasive
composite system of aluminium reinforced by silicon carbide particles and/or alumina
fibres using PCD tools. The results indicated that PCD offered superior performance
over carbide, both in wear resistance and the quality of the surface finish produced.
Ding et al. (2005) reports on experiments carried out to study the performance of
various types of PCBN and PCD cutting tools during machining of aluminium alloy
reinforced by silicon carbide metal matrix composite. PCD tools exhibited better
performance than PCBN tools because they possessed higher abrasion and fracture
resistance, and lesser adhesion property with the work material.
El-Gallab et al. (1998) carried out a series of dry high-speed turning tests to
select the optimum tool material, tool geometry and cutting parameters for turning of
20% SiC/Al metal-matrix composite. Polycrystalline tools with zero rake angle and
large tool nose radii were recommended for the roughing operations.
Chung-Shin Chang (2006) studied the machinability of GFRP materials using
chamfered main cutting edge carbide tools. It suggests that K type tool is better than
P type of chamfered main cutting edge tools. Pramanik et al. (2007) analysed tool
particles interaction during orthogonal cutting using Finite Element Method (FEM). It
was found that the tool-particle interaction and stress-strain distribution in the matrix
were responsible for particle debonding, surface damage and tool wear. Hung and
Zhong (1996) developed models for cumulative wear for facing and turning tools.
Experimental data shows that flank wear does not depend on the order of different
cutting speeds. Santhosh Ranganath (2007) developed a model to calibrate and
predict forces in machining with honed cutting tools or inserts. The developed model
showed promising results. Alameda et al. (2005) investigated the performance of
CVD diamond coated ceramic tools on machining hard metal. The effect of tool edge
32
geometry is also reported comprehensively. Senthilkumar et al. (2006) attempted to

analyse the important wear mechanisms like abrasive wear, adhesion wear and
diffusion wear of ceramic cutting tool materials and the performance of these cutting
tools are investigated with respect to surface finish.
Demand for machining of GFRP materials always exists and thus the study on
machining plays a great importance in composite industry. Machining parameters,
which are not carefully adjusted to the task of cutting GFRP composite materials also
lead to severe damage of machine tools. Glass fibre exhibits higher resistance to
cutting and hence GFRP demands polycrystalline diamond and carbide tools. Due to
higher cost and associated equipments, diamond tool is not often recommended.
Carbide tools give better finish of acceptable level at lower cost.
Davim and Mata (2007) studied machinability of GFRP plastics using
polycrystalline diamond and cemented carbide (K15) tools; the GFRP was
manufactured by hand layup method. A statistical technique using orthogonal array
and ANOVA was employed to know the influence of cutting parameters on specific
cutting pressure and surface roughness of machined composite product. Abrao et al.
(2007) focused the effect of cutting tool geometry and material on thrust force and
delamination produced while drilling GFRP composites.
Durao et al. (2008) studied on drilling of fibre reinforced plastic laminates for
unique machining process, characterised by the existence of two different
mechanisms: extrusion by the drill chisel edge and cutting by the rotating cutting lips.
Krishnaraj (2008) studied the effect of drill points on glass fibre reinforced plastic
composite at high spindle speed. The most effective way of achieving good quality
holes while drilling fibre reinforced plastics was found by reducing thrust as well as
torque. Kishore et al. (2009) focused on drilling of [(0/90)/0]S GFRP plastics using
Taguchi method for examining significance of drill point geometry and the operating
variables on drilling force and drilling induced damage. Singh et al. (2009)
highlighted modelling and analysis of thrust force and torque in drilling GFRP
composites by multi-facet drill using fuzzy logic and 8 facet solid carbide drills based
on L27 orthogonal array. The process parameters investigated were spindle speed,
feed rate and drill diameter.
A new machinability index was proposed by Paulo Davim and Mata (2005
and 2007) for the turning of hand laid up GFRP materials using PCD and cemented
33
carbide (K15) cutting tools. The investigation revealed that the PCD tool performed
well compared to cemented carbide (K15) tool in terms of surface roughness and
specific cutting pressure. Praveen Raj and Elaya Perumal (2010) carried out a study
of surface roughness, precision and delamination factor in use of Ti-Namite carbide
K10 end mill, Solid carbide K10 end mill and Tipped Carbide K10 end mill on GFRP
composite material. The investigation revealed that, Ti-Namite coated carbide end
mill and tipped carbide end mill produced less damage on GFRP composite material
than the solid carbide end mill i.e., the delamination factor was smaller.
Suleyman et al. (2011) reported that the tool nose radius was the dominant
factor on the surface roughness in turning of AISI steel. Sakuma et al (1984) drilled
holes using four drill materials and investigated drill wear pattern, flank wear width
and cutting forces. They concluded that K10 drill material had the highest wear
resistance among the drill material investigated. They further studied (1978, 1981,
1983, and 1985) the temperature distribution along the cutting edge, and the relation
between the physical property (thermal conductivity) of tool material on cutting speed
and the temperature at the cutting edge. They also performed face turning test on the
CFRP pipes with several kinds of tool material, and the wear patterns and wear land
growth rates were analysed. They concluded that K10 is superior in turning of CFRP
pipes. Sakuma and Seto concluded that when cutting speed was increased, the rate of
wear on the nose in every tool material started to increase remarkably at a certain
speed. Thermal conductivity of the tool material increases the critical speed. Face
turning test was performed by Sakuma and Seto on GFRP cylindrical pipes which
contained uni-directional fibres. The cutting resistance and surface roughness were
measured to analyse the machinability and the tool wear mechanism in GFRP cutting.
Experimental investigation was carried out in drilling of GFRP using, HSS
drills, by varying cutting conditions and drill geometry namely cutting speed, feed
rate, point angle and helix angle by Gindy et al (1998). They developed a simple
monogram to predict drill wear level from the thrust level or the amount of material
removed. The effect of cutting speed and feed on drilling CFRP was studied by
Davim and Reis (2003) using standard twist drill and Spur drill.
The machinability of GFRP by means of tool materials and geometries was
investigated experimentally by Sang-Ook An (1997) and reported that by proper
selection of cutting tool material and geometry, excellent machining of the work
34
piece could be achieved. The surface quality related closely to the feed rate and
cutting tools. He also concluded that the surface roughness was not related to the
depth of cut and cutting speed with respect to various tools.
Lee (2001) investigated experimentally the machinability of GFRP by means
of tools made of various materials and geometries. Experimental trails were carried
out by turning using single crystal diamond, Polycrystalline diamond and Cubic
boron nitride cutting tools. The study also recommends single crystal diamond tool
for GFRP machining.
Ferreira et al. (2001) reported practical experiments on turning carbon fibre
reinforced composite materials to study the performance of different tool materials
such as ceramic, cemented carbide, Cubic Boron Nitride (CBN) and PCD. The results
showed that only diamond tools were suitable for the finish turning. An optimisation
methodology was used to determine the best cutting conditions. The tests were
carried out with cemented carbide tool at various cutting speeds.
An attempt was made by Singh and Bhatnagar (2006) to correlate the drilling
induced damage with drilling parameters of uni-directional GFRP composite
laminates. The drill point geometry is a well known factor that influences the damage
during drilling. Among the drill point geometries tested, four facet drills were not
recommended. The results also reaffirm and agree with the earlier results that the
cutting speed to feed ratio being an important variable which influences drilling
induced damages.
Paulo Davim et al. (2004) presented a study with the objective of evaluating
the machining parameters (cutting speed and feed) and the influence of the matrix
under specific cutting force, delamination and surface roughness in two types of
matrix (Viapal VUP 9731 and ATLAC 382-05). A plan of experiments, based on the
orthogonal array, was established considering drilling with prefixed cutting
parameters in two hand lay-up FRPs materials. Finally the analysis of variance
(ANOVA) was performed to investigate the cutting characteristics of composite
material using a cemented carbide (K10) drill with appropriate geometry.
Sreejith et al. (2000) made a comprehensive study on machining carbon
phenolics ablative composites using coated carbide and PCD tools. Mathematical
models were also developed to describe the effects of cutting parameters. Paulo
Davim et.al (2003) studied the machinability on polyethere-therketone composites.
35
PCD cutting tool was used for machining and analysis of variance was preformed to
investigate the machining characteristics.
Ferreira et al. (2001) studied the machinability aspects of CFRP composites
using different tool materials like ceramics, cemented carbide, boron nitrate and PCD.
It concluded that diamond tool were suitable for finished turning and carbide tools
could be used for rough turning. Paulo Davim and Pedro Reis (2004) studied the
machinability on polyethere-therketone reinforced with 30% glass fibre using PCD
and cemented carbide (K20) tools. The investigation revealed that both PCD and K20
tools presented similar cutting characteristics and hence K20 grade cemented carbide
cutting tools were most preferred since the cost of carbide tools are much cheaper
than PCD tool.
2.5.4 Factors Influencing Surface Roughness
Ramulu et al. (1993) reported that surface roughness and delamination were
the characteristics that could influence the dimensional precision, the performance of
mechanical pieces and production costs. For these reasons there has been research
and development with the objective of optimising cutting conditions, to obtain a
determined surface roughness and delamination. Surface roughness has been received
serious attentions for many years. It has formulated an important design feature in
many situations such as parts subject to fatigue loads, precision fits, fastener holes
and aesthetic requirements. Jack fang and Wang (2002) reported that in addition to
tolerances, surface roughness imposed one of the most critical constraints for
selection of machines and cutting parameters in process planning. Sreejith et al.
(1994) mentioned that for achieving the desired surface finish, it is necessary to
understand the mechanisms of the material removal, and the kinetics of machining
processes affecting the performance of the cutting tool.
Palanikumar et al. (2003) highlighted machining of GFRP composite pipes for
cost effective implementation and the machinability became a major parameter. For
successful application of these composites, the surface finish and surface integrity
were seemed most important especially for surface sensitive parts subjected to fatigue
or creep.
Eriksen (1999) carried out experiments to study how the surface roughness of
short fibre reinforced thermoplastic machined by turning, was influenced by
36
variations of the feed rate, cutting speed, tool radius, the fibre orientation and the
interactions between them. He concluded that the surface roughness increased for an
increasing feed rate, decreasing tool radius and independent of the fibre orientation.
Santhanakrishnan et al. (1988) carried out face turning trials on GFRP, CFRP
and KFRP cylindrical tubes to study their machined surfaces for possible application
as friction surfaces. The surface roughness obtained and observed morphology of the
machined surface of FRP composites was compared and it is reported in this work.
Among the machined composite surfaces CFRP exhibited the best surface finish
value. The best surface structure of CFRP was due to the least fibre pull out with
insignificant loose fibres on the surface.
Palanikumar et al. (2006) made an attempt to assess the influence the
machining parameters on the machining of GFRP composites. Design of experiments
(full factorial design) concept has been developed to assess and optimise the chosen
factors to attain minimum surface roughness by incorporating (i) response table and
response graph (ii) normal probability plot (iii) interaction graphs (iv) ANOVA. The
results suggested that the feed rate was the factor, which had the greater influence on
surface roughness, followed by cutting speed. This was because of the nonhomogeneous microstructure of the GFRP composites.
Paulo Davim and Francisco Mata (2005) studied the machinability in turning
process of FRP using PCD tools. Controlled machining experiments were performed
with cutting parameters prefixed in the work piece. A statistical technique, using
orthogonal arrays and analysis of variance, was employed to investigate the influence
of cutting parameters on specific cutting pressure and surface roughness. The
objective was to evaluate the machinability of these materials as a function of the
manufacturing process (hand lay-up and filament winding). A new machinability
index was also proposed.
Paulo Davim and Francisco Mata (2007) carried out an optimisation study of
surface roughness on turning FRP tubes manufactured either by filament winding or
by hand lay-up, using PCD cutting tools. A plan of experiment was performed with
cutting parameters prefixed in the FRP tubes. The objective was to establish the
optimal cutting parameters to obtain certain surface roughness corresponding to
International dimensional precision (ISO) IT7 and IT8 in the FRP work pieces, using
Multiple Regression Analysis (MRA).
37
2.5.5 Factors Influencing Delamination factor

Machining parameters, which are not carefully adjusted to the task of
machining GFRPs may also lead to severe damage of machine tools. Owing to its
abrasive nature, glass fibre may cause premature rounding of cutting edges. Distinct
differences in hardness between fibre and matrix combined with the high cutting
resistance fibres result in edge chipping. Since machining operations for cured
laminates are mainly finishing operations, the quality achieved is of some interest.
Few authors attempted to study the damages on machined surfaces and their details
are presented below:
Delamination, fibre pull-out, fibre-fragmentation, burning and fuzzing are
some of the damages caused by machining on GFRP as reported by Wang X.M and
Zhang L.C (2004).
Davim et al. (2004) studied the influence of cutting parameters (cutting,
velocity and feed) while machining GFRP with two different matrixes in order to
study the influence of those parameters on delamination. Singh et al. (2004)
highlighted on drilling induced damage in FRP composite laminates for the high
degree of intricacy in composite structures which necessitated special process to
create holes in them for the purpose of assembly. Numerous methods were used, but
conventional drilling still remained the unavoidable process for making holes in
composite laminates.
Camposrubio et al. (2008) studied the influence of process parameters on
delamination in drilling of composite materials at high speed using cemented carbide
drills. Durao et al. (2008) studied the effect of drilling characteristics of hybrid carbon
glass/epoxy composites. They validated the influence of delamination in bearing
stress of drilled hybrid carbon glass/epoxy quasi-isotropic plates. Paulo Davim and
Pedro Reis (2005) evaluated the cutting parameters (cutting velocity and feed rate)
under the surface roughness, and damage in milling laminate plates of CFRP by
Taguchi method.
Davim et al. (2003, 2004) conducted experiments to select cutting parameters
for damage free drilling in carbon-fibre reinforced epoxy composite materials. They
established a correlation between cutting velocity and feed rate with the delamination.
Experimental investigation was also done on drilling glass fibre reinforced plastic
(GFRP) to study the effect of cutting parameters in a hand layup GFRP material. Tsao
38
et al. (2003, 2004) studied delamination factor by using twist drill, candle stick drill
and saw drill under various cutting conditions. They established a relation between
feed rate, spindle speed and drill diameter with delamination. Enemuoh et al. (2001)
conducted experiments to select cutting parameters for damage free drilling in carbon
fibre reinforced epoxy material. They used Taguchi's experimental analysis technique
and a multi-objective optimisation criterion. A process map based on the results is
presented as a tool for drilling process design and optimisation for the investigated
tool/material combination.
Tsao et al. (2005) further concluded that saw drills and core drills produced
less delamination than twist drills by distributing the drilling thrust toward the hole
periphery. They also concluded that delamination could be effectively reduced or
eliminated by slowing down the feed rate when approaching the exit and by using
back-up plates to support and concentrate the deflection of the composite laminate
leading to exit side delamination. Drilling tests were carried out on GFRP composites
by Tagliaferri et al. (1991), in order to verify the effect of machining parameters on
the cut quality and on the mechanical behaviour of the material under examination. A
novel method was adopted to measure the width of the damage zone. The
experimental data showed that the width of the damage zone was correlated to the
ratio between drilling speed and feed rate, in particular, the higher the ratio value the
better the cut quality.
Mohan et al. (2007) carried out delamination analysis on GFRP composite
material and the main objective of the study was to determine the factors and
combination of factors that influenced the delamination using Taguchi and Response
Surface Methodology (RSM) to achieve the optimisation machining conditions that
would result in minimum delamination. From the RSM analysis it was evident that
among all significant parameters, specimen thickness and cutting speed have
significant influence on peel up delamination and the specimen thickness and the feed
rate have more significant influence on push down delamination.
Ramkumar et al. (2004) attempted a newer technique of superimposing
oscillatory vibration to minimise the damages during drilling of GFRP composites.
Compared to conventional drilling, this method resulted with reduced cutting forces,
damages and reduced tool wear. Machining of fibre reinforced composite differs
significantly from machining conventional metals and alloys, owing to the behaviour
39
of matrix material, reinforcement and diverse properties of fibre, matrix, and

orientation of fibre and volume fraction of fibres. Li et al. (2002) proposed a
numerical model for simulating the process of low-velocity impact damage on
composite laminates using finite element method. The proposed method can simulate
various types of damages such as delamination, matrix cracking, fibre breakage etc.
Eduardo Capello (2001) analysed the differences in delamination mechanisms
when drilling with and without a support under the work piece. This investigation
leads to hypothesize two main differences in the mechanism. On the basis of this
analysis, a new device was designed that counters the hypothesized delamination
mechanism. A simple prototype of this device was built, and its effectiveness
verified. The results showed that the proposed device could drastically reduce
delamination.
2.5.6 Factors Influencing the Machining force
Paulo Davim et al. (2004) reported that machining force also plays a key role
in analyzing the machining process of FRPs. The value of machining force in the
work piece is determined using the resultant of forces acting on all the three
orthogonal directions. Generally, machining force increases with the feed rate and it
decreases with the cutting velocity.
Pramanik et al. (2006) presented a mechanics for predicting the forces of
cutting during machining metal matrix composites. The model proposed is also
compared with the theoretical model and the experimental values are also correlated.
Suleyman Yaldiz and Faruk Unsacar (2006) designed a dynamometer for predicting
the cutting force on turning. The results obtained showed that the dynamometer could
be used reliably to measure static and dynamic cutting forces.
Khashaba (2004) studied the influence of material variables on thrust force,
torque and delamination while drilling of GFRP composites with different types of
fibre structures. They carried out the experiment with cross winding/polyester,
continuous winding/polyester, woven polyester and woven/epoxy. It was found that
woven epoxy showed best results in terms of torque, thrust force.
Godfrey et al. (2006) developed a model for correlating the interactions of
drilling control parameters and their effects on axial force and torque acting on the
cutting tool during drilling by means of RSM and they reported that the drill torque
40
varies non-linearly with all the control parameters, viz., speed, feed rate, and
diameter. Machining force also plays a key role in analyzing the machining process
of FRPs. The value of machining force in the work piece is determined using the
equation:
......................................................................... (2.1)
Where Fm- Resultant machining force, Fx- Feed force, Fy - Cutting force and Fz Thrust force. Generally, machining force increases with feed rate and it decreases
with cutting velocity. Evaluation of machining parameters of hand layup GFRP
related to machining force was also carried out by Paulo Davim et al. (2004) on
milling using cemented carbide (K10) end mill.
Paulo Davim et al. (2004) evaluated the cutting parameters (cutting velocity
and feed rate) related to machining force in the workpiece, delamination factor,
surface roughness and international dimensional precision in two GFRP composite
materials (Viapal VUP 9731 and ATLAC 382-05).
Won and Dharan (2002) conducted drilling experiments on CFRP to
determine quantitatively, the effect of chisel edge and pre drilling on thrust force.
They also developed semi-empirical relationships, between drilling forces and cutting
parameters and established critical feed rate for aramid and carbon fibre polymer
composites.
Mohan et al. (2005) used Taguchis optimisation methodology to optimise the
cutting parameters of GFRP materials in drilling. The effect of process parameters
was investigated using ANOVA. The results indicated that among all the significant
parameters, speed and drill size influence more on cutting thrust than the specimen
thickness and the feed rate. They also analysed the influence of machining parameters
on cutting force during drilling of GFRP composite with the help of a commercially
available software package MINITAB14.
Arul et.al (2006) reported that the defects in drilling of composites were due
to thrust force experienced by the work piece. The parametric influence on cutting
force was also experimentally evaluated. The experimental results showed that the
defects tolerance drilling could be attained by proper selection of cutting parameters
and tool material.
41
2.6 LITERATURE ON MODELLING AND OPTIMISATION

It is observed that various researchers in the past employed various techniques
for modelling and optimising the machining characteristics of any machining process.
Some of the optimisation techniques employed are regression analysis, desirability
function analysis, neural networks, genetic algorithm, fuzzy logics etc. This section
presents a brief survey of the techniques employed in the available previous literature
regarding the techniques and tools used for optimising the machining parameters of
any machining process. Fiju Antony and Frenie Fiju Antony (2001) stated that the
Taguchi method is a powerful problem solving technique for improving process
performance, yield and productivity.
Ko-Ta Chiang and Fu-ping Chang (2006) presented an effective approach for
the optimisation of the wire EDM process of Al2O3 particle-reinforced material (6061
alloy) with multiple performance characteristics based on the grey relational analysis.
It is clearly shown that the above performance characteristics in the WEDM process
are greatly improved together through this approach.
Noorul Haq et al. (2007) presented a new approach for the optimisation of
drilling parameters on drilling Al/SiC metal matrix composite with the multiple
responses based on orthogonal array with grey relational analysis. Experimental
results have shown that the responses in drilling process can be improved effectively
through the new approach.
Shaji and Radhakrishnan (2003) analysed the process parameters such as
speed, feed, infeed and mode of dressing as influential factors based on Taguchis
experimental design methods for a surface grinding process. Basavarajappa et al.
(2007) applied Taguchi technique to study the dry sliding wear behaviour of metal
matrix composites. The results show that the technique applied in this study is most
effective and efficient.
Chih-Wei Chang and Chun-Pao Kuo (2007) evaluated the surface roughness
in laser assisted machining of aluminium oxide ceramics with Taguchi method. The
method facilitated the identification of optimum machining conditions.
42
2.6.1 Statistical tools

Statistical tools and numerical models are found to be most effective in
optimising the machining parameters on turning composite materials. Few of such
reports are reported below.
Paulo Davim and Rosaria Cardoso (2006) used statistical technique to study
the effects of pv factor, temperature and sliding distance on the dry sliding
tribological behaviour of 30% Carbon Fibre Reinforced Polyethere-therketone
composite (PEEK-CF30). The obtained results showed that friction coefficient was
highly influenced by the temperature factor and by sliding distance. Paulo Davim and
Conceicao Antonio (2001) attempted a numerical study for optimising the drilling
conditions on machining metal matrix composites. The developed model proved to be
effective in optimising cutting conditions in drilling.
Paulo Davim and Pedro Reis (2003) studied the optimisation of cutting
parameters of CFRP using statistical tools. They also performed ANOVA and MRA.
The results were compared and correlated between factors. Paulo Davim and Luis
Figueira (2007) investigated the machinability of cold work tool steel with ceramic
cutting tools using statistical techniques. The results proved that statistical techniques
were more appropriate for selecting cutting parameters.
Paulo Davim et al. (2004) made experimental investigations to study the
cutting characteristics of hand lay-up of GFRP materials. ANOVA was used to
optimise the machining parameters and it was found to be more effective.
2.6.2 Non- traditional techniques
Wong and Hamouda (2003) studied the feasibility of using neural network in
representing machinability data. The feed-forward neural network was used to predict
optimum machining parameters under different machining conditions. The Back
Propagation (BP) method was used to optimise the network component
representation. An object-oriented neural network handling library was developed
and implemented in the turning process. The proposed network unveils the possibility
of developing an expert system for machinability data selection based on neural
networks. Comparisons with fuzzy logic representation in the literature are also made.
Development of a surface roughness prediction model for machining mild
steel using Response Surface Methodology (RSM) was carried out by Suresh et al.
43
(2002). The experimentation was carried out with TiN-coated Tungsten carbide
(CNMG) cutting tools, for machining mild steel work pieces covering a wide range of
machining conditions. An attempt was also made to optimise the surface roughness
prediction model using Genetic Algorithms (GA) to optimise the objective function.
The GA program gives minimum and maximum values of surface roughness and
their respective optimal machining conditions.
Jeyapaul et al. (2006) presented the use of GA and ANOVA to the Taguchi
method for the optimisation of the gear hobbing process with multiple performance
characteristics. It has been demonstrated that a multiple response optimisation
problem can be effectively tackled by using GA to generate a single Weighted S/N
ratios (WSN) as a performance indicator. Statistical analysis is then carried out on the
WSN to identify the key factors which affect process performance and then determine
the optimal factor setting to optimise process performance.
Two models were developed to predict tool wear for different values of
cutting speed, feed and time. One of them was based on statistical regression, and the
other was based on a multilayer perception neural network by Ramon Quiza et al.
(2007). Parameters for the design and the training process for the neural network have
been optimised using the Taguchi method. Outcomes from the two models were
analysed and compared. The neural network model showed better capability to make
accurate predictions of tool wear under the conditions studied.
A hybrid technique based on an evolutionary search over the design space
defined by the experimental results was considered to optimise cutting condition by
Conceicao Antonio and Paulo Davim (2002). Optimal cutting conditions were
searched using a genetic algorithm based on an elitist strategy. The chromosomes
composed by random keys represented cutting conditions defined according to a
temporal scale. Franci cus and Joze Balic (2003) proposed a new optimisation
technique based on GA for the determination of the cutting parameters in machining
operations. It presented a new methodology for continual improvement of machining
conditions with GA. It performs the modification of recommended cutting conditions
obtained from a machining data and learning of obtained cutting conditions for those
learnt previously by a proposed GA. Experimental results show that the proposed GA
based procedure for solving the optimisation problem is both effective and efficient,
44
and can be integrated into an intelligent manufacturing system for solving complex
machining optimisation problems.
Sudarsana Rao (2005) presented the stress-strain response of Al2O3/SiC
ceramic composites using a back propagation neural network, which incorporates the
effect of interface shear strength in the analysis. The developed model is capable of
predicting the stress-strain relationship for new interface strength even with limited
information. Paulo Davim and Conceicao Antonio (2001) used GA for the optimal
search of cutting conditions. The chromosomes which represent cutting conditions
defined according to a temporal scale and were compared by random keys. The
evolution of the solutions is based on an elitist strategy.
2.6.2.1 Response Surface Methodology
Response surface methodology is a series of mathematical and statistical
techniques used for modelling and analysing problems and has the objective of
optimising the responses. It is a sequential experimentation strategy for empirical
model building and optimisation. By conducting experiments and applying regression
analysis, a model of the response to some independent input variables can be
obtained. Based on the model of the response, a near optimal point can then be
deduced. RSM can determine and represent the effect of relationships among input
control factors that influence the responses as a two- or three-dimensional hyper
surface. Most of the work in RSM was focused on the case where there was only one
response of interest. Khuri (1996) reported that in product or process development, it
is quite common that several response variables are of interest. In this case,
determination of optimum conditions on the input variables would require
simultaneous consideration of all the responses. This is called a multi-response
problem. As of now, several approaches have been proposed for Multi Response
Optimisation (MRO) including the desirability function approach by Derringer and
Suich (1980) and loss function approach [Pignatiello (1993), Vining (1998) and Ko et
al. (2005)]. A number of investigations using RSM were carried out by George et al.
(2004), Kansal et al. (2005) and Matoorian et al. (2008) determined the significant
factors affecting the response.
Sahin and Motorcu (2004) stated that in order to know surface quality and
dimensional properties, it is necessary to employ theoretical models for prediction
45
purpose. For prediction, the Response Surface Method (RSM) is practical,

economical and relatively easy for use. Palanikumar (2007) investigated towards
modelling and analysis for surface roughness in machining glass fibre reinforced
plastics using Response Surface Methodology (RSM). Hussain et al. (2010) carried
out investigations on machining of GFRP composites by carbide tools (K20) for
development of a surface roughness prediction model using RSM.
Asif Iqbal et al. (2010) investigated the relationships and parametric
interactions between the three controllable variables (voltage, rotational speed of
electrode and feed rate) on the material removal rate, electrode wear ratio and surface
roughness in EDM (Electrical Discharge Machining) milling of stainless steel by
RSM. Mohan et al. (2005) carried out delamination analysis on GFRP composite
material and the main objective of the study was to determine the factors and
combination of factors that influenced the delamination using Taguchi and RSM to
achieve the optimisation machining conditions that would result in minimum
delamination. From the analysis it is evident that among all significant parameters,
specimen thickness and cutting speed have significant influence on peel up
delamination and the specimen thickness and the feed rate have more significant
influence on push down delamination.
Development of a surface roughness prediction model for machining mild
steel using RSM was carried out by Suresh et al. (2002). The experimentation was
carried out with TiN-coated Tungsten carbide (CNMG) cutting tools, for machining
mild steel work pieces covering a wide range of machining conditions. An attempt
was made to optimise the surface roughness prediction model using Genetic
Algorithms (GA) to optimise the objective function. The GA program gives
minimum and maximum values of surface roughness and their respective optimal
machining conditions.
2.6.2.2 Artificial Neural Network
Omer Erkan et al. (2012) stated that in recent years researchers have been
attempting to develop ANN model, which has universal approximation capability and
more accuracy in predicting any system. This ANN is also applied for machining of
FRP composites. Tsao et al. (2008) developed an ANN model to evaluate the thrust
force and surface roughness in drilling composite material using Taguchi analysis and
46
neural network to predict the delamination of CFRP composites. Karnik et al. (2007)
developed a ANN predictive model to predict the burr height and burr thickness
using a multilayer feed forward neural network, trained using back propagation
algorithm in high speed drilling of Carbon Fibre Reinforced Plastics (CFRP). The
performance of this ANN model was compared with
the second order RSM
mathematical model and the accuracy of ANN prediction was clearly proved .
Palanikumar et al. (2006) predicted the tool wear using back propagation neural
network. This work has considerable implications in the real time monitoring of tool
wear. Stone and Krishnamurthy (1996) applied the neural network concept to
improve the hole quality and reported that drilling with desired thrust force profile
with acceptable error margin was possible with minimum delamination. Ezugwu et al.
(1995) developed a model for predicting tool wear using Artificial Neural Network
(ANN). The proposed model predicted 58.3% correctly the tool life and 87.5%
correctly the failure mode prediction.
Wong and Hamouda (2003) studied the feasibility of using neural network in
representing machinability data. The feed-forward neural network was used to predict
optimum machining parameters under different machining conditions. The Back
Propagation (BP) method was used to optimise the network component
representation. An object-oriented neural network handling library was developed
and implemented in the turning process. The proposed network unveils the possibility
of developing an expert system for machinability data selection based on neural
networks. Comparisons with fuzzy logic representation in the literature are also made.
Two models were developed to predict tool wear for different values of cutting speed,
feed and time. One of them was based on statistical regression, and the other was
based on a multilayer perception neural network by Ramon Quiza et al. (2007).
Parameters of the design and the training process, for the neural network were
optimised using the Taguchi method. Outcomes from the two models were analysed
and compared. The neural network model showed better capability to make accurate
predictions of tool wear under the conditions studied.
2.6.2.3 Fuzzy Logic
The theory of fuzzy logic, initiated by Zadeh (1965) has proven to be useful
for dealing with uncertain and vague information. Raghukandan et al. (2004)
47
explained that fuzzy theory has proved to be an effective means for dealing with
objectives that are linguistically specified. Linguistic terms, such as low, medium
and high may be defined by fuzzy sets. Ross (1992) revealed that since its
introduction, fuzzy set theory has attracted the attention of researchers in
mathematical and engineering fields.
Singh et al. (2009) highlighted modelling and analysis of thrust force and
torque in drilling GFRP composites by multi-facet drill using fuzzy logic using 8
facet solid carbide drills based on L27 orthogonal array. The process parameters
investigated were spindle speed, feed rate and drill diameter. Fuzzy rule based model
was developed to predict thrust force and torque in drilling of GFRP composites. The
results indicated that the model could be effectively used for predicting the response
variable by means of which delamination could be controlled. Latha and
Senthilkumar (2009) successfully applied fuzzy logic for the prediction of
delamination in drilling of glass fibre reinforced plastics.
2.7 SUMMARY
While a considerable amount of literature on the various aspects of
conventional machining of metals is available, only a limited amount of literature
exists on the machining of composite materials. However, from the above survey it is
observed that the investigations on the following few areas have been covered to a
lesser extent only. Experimental investigation on the effect of fibre orientation and
tool geometry on machining force and delamination factor during milling of GFRP
laminates is found inadequate. Experimental analysis on the surface roughness,
delamination factor and machining forces during endmilling of GFRP composites
with different fibre orientations has not been reported adequately. Endmilling
experiments on GFRP specimens to evaluate the influence of fibre orientation, cutting
speed, feed rate and depth of cut on surface roughness and delamination factor using
solid carbide tool coated with PCD by using RSM and ANN models has not been
attempted previously. Optimisation of the machining parameters using fuzzy logic
with Taguchi method has not been reported previously.
48
CHAPTER 3
ENDMILLING OF GFRP USING SOLID CARBIDE TOOL
COATED WITH POLYCRYSTALLINE DIAMOND (PCD)
3.1 PREPARATION OF TEST SPECIMENS
3.1.1 Introduction
Major constituents in a fibre-reinforced composite material are the reinforcing
fibres and a matrix, which act as a binder for the fibres. Other constituents are
coupling agents, coatings, and fillers. Coupling agents and coatings are applied on the
fibres to improve their wetting with the matrix as well as to promote bonding across
the fibre-matrix interface. Both, in turn, promote a better load transfer between the
fibres and the matrix. Primarily fillers are used with some polymeric matrices to
reduce cost and improve their dimensional stability. In this chapter, the fabrication of
laminates is discussed along with some of their mechanical properties.
3.2 MATERIAL SELECTION PROCESS
Material selection is one of the most important and critical steps in the
structural or mechanical design process. If the material selection is not done properly
the design may show poor performance and may require frequent maintenance,
repair, or replacement; and in the extreme, may fail, causing damage or injuries. The
material selection process requires the knowledge of the performance requirements of
the structure or component under consideration. Usually materials are selected based
on type of loading, mode of loading, service life, operating or service environment,
cost, and so on. In this work, the type of loading is axial and the mode of loading is
low-velocity impact. The material properties to be considered in the material
selection process depend on the performance requirements (mechanical, thermal, etc.)
and the possible mode or modes of failure (e.g., yielding, brittle fracture, ductile
failure, buckling, excessive deflection, fatigue, creep, corrosion, thermal failure due
to overheating, etc.).
49
3.2.1 Glass Fibres

Fibre is the principal constituent in a fibre-reinforced composite material. It
occupies the largest volume fraction in a composite laminate and shares the major
portion of the load acting on a composite structure. Proper selection of fibre type,
fibre volume fraction, fibre length, and fibre orientation is very important, since it
influences the density, tensile strength and modulus compressive strength and
modulus. Fatigue strength as well as fatigue failure characteristics of a composite
laminate glass fibre are the most common of all reinforcing fibres for polymeric
matrix composite. The main advantages of glass fibres are low cost, high tensile
strength, high chemical resistance and excellent insulating properties The
disadvantages are: relatively low tensile modulus and high density (among the
commercial fibres), sensitivity to abrasion during handling (which frequently
decreases its tensile strength), relatively low fatigue resistance, and high hardness
(which causes excessive wear on moulding dies and cutting tools). The two types of
glass fibre commonly used in the Fibre-Reinforced Plastics (FRP) industry are Eglass and S-glass. Another type known as C-glass is used in chemical applications
which have greater corrosion resistance to acids. E-glass, or electrical glass, is the
most common glass fibre and is used extensively in commercial composite products.
E-glass is a low-cost, high-density, low-modulus fibre that has good corrosion
resistance and good handling characteristics. S-glass, or silica glass originally
developed for aircraft components and missile casings, has the highest tensile
strength among all fibres in use. However, the compositional difference and the
higher manufacturing cost make it more expensive than E-glass.
Hand lay-up process was used to prepare the GFRP specimens. Fibre
percentage was maintained by controlling the weight of the fibres (number of layers).
E-glass seven mill fibre mats (360 GSM) which were compatible to epoxy resins (LY
556) were used as reinforcement. Hardener (HY 951) was mixed with epoxy for
curing at room temperature. Necessary care was taken, using the rollers, to remove
the entrapped air during the lay-up process. A gel coat was applied on the mould prior
to the lay-up process. The specimens were macroscopically tested to confirm the
absence of defects like voids or delamination. The specimen thickness and the fibre
percentages selected for this analysis are within the applicable range. The thicknesses
50
of the specimens varied in the range of 3 mm to 3.2 mm. The volume fraction of glass
fibres in all laminates were kept at 63%.
3.2.2 Matrices
The function of the matrix is to hold the fibres in their proper position, protect
the fibres from abrasion, transfer loads between fibres, and provide interlaminar shear
strength. The most prevalent thermoset resins used for composite matrices are
polyesters, vinyl esters, epoxies, bismaleimide, polyamides and phenolics. Matrices
or polymeric composites can either be thermoset or thermoplastics. Thermoset resins
usually consist of a resin (e.g., epoxy) and a compatible curing agent. When the two
are initially mixed, they form a low-viscosity liquid that cures as a result of either
internally generated exothermic or externally applied heat. Thermoset resins do not
soften on heating and cannot be remelted. Here, the epoxy resin Lapox L-12 grade
was used with Araldite LY 556 as binder. Epoxies are the most common matrix
materials for high-performance composites. They have an excellent combination of
strength, adhesion, low shrinkage and processing versatility. Hardener HY 951 grade
with diluent DY 021 was added to the mixture during fabrication of laminates. The
raw materials were procured from Hindustan Ciba Geigy, India. GFRP Laminates of
300 mm x 300 mm x 3 mm thick with 12 lay-up with desired fibre orientation (15,
30, 60, 90, 105 and 120) were prepared for the experimental analysis.
3.3 FABRICATION PROCEDURE
All laminates investigated in the present study were fabricated by dry hand
lay-up technique since it was ideally suited in manufacturing at low-volume with
minimum tooling cost. Fabrication of composite laminates involves four basic steps,
lay-up, wetting/impregnation, consolidation and solidification. All composites
manufacturing processes involve these same four steps although they are
accomplished in different ways.
3.3.1 Hand lay-up technique

The step by step procedure of hand lay-up technique is illustrated in Figure
3.1. In the first step, woven glass fibre cloth was hand-laid up one ply at a time in the
desired angles required. The desired composite thickness was built-up by placing
51
successive layers of the fibre and resin mixture. The purpose of this step was to
achieve the desired fibre architecture.
52
Figure 3.1 step by step procedure of hand lay-up technique

Performance of a composite structure relies heavily on fibre orientation and
lay-up sequence. Prior to the lay-up, the worktable was coated with Tedlar (polyvinyl
fluoride) which is a release film to facilitate easy removal of the layup after
consolidation. In the second step, fibres and resins were mixed together to form a
lamina. During or before lay up, each ply was impregnated with epoxy resin using a
hand roller. This hand roller was used to remove excess resin and air, which resulted
in compaction of the plies. During by-up, each ply was impregnated with epoxy resin
mixture. In order to make sure that the resin flowed entirely around all fibres.
Viscosity, surface tension, and capillary action are the main parameters affecting the
impregnation process.
The third step was to create intimate contact between each layer of the lamina.
This step ensured that all the entrapped air between layers was removed.
Consolidation is a very important step in obtaining a good quality part. Poorly
consolidated pans will have voids and dry spots. Consolidation of continuous fibre
composites involves two important processes: resin flow through porous media and
53
elastic fibre deformation. During the consolidation process, applied pressure is shared
by both resin and fibre structure. Initially, however, the applied pressure is carried
solely by the resin (zero fibre elastic deformation). Fibres go through elastic
deformation when the compressive pressure increases and resins flow out toward the
boundary. A breather was placed onto the matrix and then a peel ply [nylon vacuum
bag] was placed over the lay-up. Vacuum of 550 /2 was applied through
vacuum valve at one corner of the system, to enable air evacuation between the plies.
The system was allowed for ambient cure along with vacuum pressure. The bagged
part was then placed in an oven and cured under the specified time, temperature and
pressure. The final step was solidification of the laminate. Vacuum and pressure were
maintained during this process for 24 hrs. In thermoset composites, the rate of
solidification depends on the resin formulation and cure kinetics. Heat [Post curing]
was supplied during processing to expedite the cure rate of the resin. In thermoset
resins, usually the higher the cure temperature the faster is the cross-linking process.
The temperature was maintained at 80C for the first 2 hrs and then at 120C for
another 2 hrs and later allowed to cool at room temperature.
3.4 MECHANICAL PROPERTIES OF GFRP SPECIMENS
Mechanical test was performed in order to evaluate the mechanical properties
as well as the failure resistance properties of GFRP laminates. Mechanical properties
such as, tensile strength, Youngs modulus and shear strength were investigated with
a Universal Testing Machine (UTM) in accordance with ASTM standards.
3.4.1 Tensile testing
Tensile properties, such as tensile strength and tensile modulus of laminates,
were determined by static tension test in accordance with ASTM D309. The tensile
specimen was straight-sided and had a constant cross section with bevelled tabs
adhesively bonded at its ends (Figure 3.2). A complaint and straincompatible
material was used for the end tabs to reduce stress concentrations in the gripped area
and thereby promote tensile failure in the gauge section. Tensile specimens 250 mm
long and 2 mm thick with a gauge length of 50 mm and width of 25 mm were
prepared with end tabs. The tensile specimen was held in a servo- hydraulic universal
testing machine as shown in Figure 3.3 by wedge action grips and pulled at a
54
recommended crosshead speed of 5 mm/min, which corresponds to a strain rate of

0.2% per second. The specification of the UTM is shown in Table 3.1.
Figure 3.2 Tensile test specimen
Figure 3.3 Universal testing machine (UTM)
55
Table 3.1 Specification of UTM

Maximum capacity 60 Ton
Test speed range
0.01-500 mm/min
Width * depth
1000 * 500 mm
Height
1800 mm
Cross head speed
5 mm / min
Data acquisition
ASTM E 185
The strains were recorded with strain gauges. Three tests were carried out for
each case. From the tensile test results (Table 3.2), it was concluded that the tensile
strength of the 90 fibre oriented GFRP laminate was higher than the other fibre
orientations and shear strength of the 90 fibre oriented GFRP laminate was less than
the other laminates.
Table 3.2 Test results
Fibre
Tensile
Youngs
Shear
Orientation,
strength,
modulus,
strength,
(deg)
(MPa)
(GPa)
(MPa)
ASTM
ASTM
ASTM
ASTM
CODE
D3039-76
Sample No.
D3039-76 M D2344-84
15
153
8.164
7.513
30
112
12.105
8.127
60
109
12.475
10.054
90
229
14.412
9.17
105
117
12.941
11.014
120
133
13.169
11.968
56
This is due to that 90 fibre orientation GFRP specimen when subjected to

tensile loading, experience more stiffness and when it is subjected to shear loading, it
experiences low stiffness when compared to other fibre oriented GFRP specimens.
3.5 EXPERIMENTAL SETUP
GFRP Laminates of 300 mm x 300 mm x 3 mm thick with 12 lay-up with
desired fibre orientation (15, 30, 60, 90, 105 and 120) were used for the milling
operations. The experiments were conducted based on central composite rotatable
design using a CNC milling machine. The CNC milling machine is an automated
machinery, and its specifications are given in Table 3.3. The fixation of the composite
material is as shown in Figure 3.5, to make sure that vibrations and displacement are
eliminated.
Table.3.3 Specification of the CNC milling machine
Type of machine
Specification
Power
Maximum speed
Vertical machine
center
Hass-Us(Brand
New)
Hass No. 3 Machine
25 KW
12,000 rpm
X axis
Y axis
Z axis
Table area
1016 mm
520 mm
508 mm
520 x 1200
Make
3.5.1 Design of Experiments

Based on the literature [Lee, (2001) and Zhong et al. (2006)] and the previous
work done in this field by Palanikumar, K. (2004), the independently controllable
predominant machining parameters, which are having greater influences on the
surface finish and delamination of GFRP composite work piece has been identified.
They are cutting speed, work piece fibre orientation angle, depth of cut and feed rate.
A detailed analysis has been carried out to set the lower and upper limits of the
parameters. Based on the analysis, the upper and lower limits of the parameters were
identified and are given in the Table 3.4. Previous studies indicated that the surface
roughness increased with the increase of cutting speed. But a very high cutting speed
57
was found to cause a large deformation rate of glass fibre and it produced
imperfection in surface and hence the cutting speed in the current study was fixed
between 50 and 100 m/min [Palanikumar et al. (2006)]. The fibre orientation angle
plays an important role for deciding the surface roughness. The previous results
indicated that fibre orientation angle was one of the important aspects, which affected
the surface roughness. For the present study, the fibre orientation angle was
considered between 15o and 120o. The depth of cut plays only a small role in
composite machining process compared to cutting speed and feed rate, but it affects
the machining process in a significant way. The depth of cut was fixed between 0.05
and 0.25 mm. Feed rate is the main parameter, which affects the machining process.
The increase in feed rate increases the chatter and it produces incomplete machining
at a faster traverse, which leads to higher surface roughness and hence low limit of
feed rates are advisable in machining GFRP composites. In the present study, the feed
rate was selected between 0.04 and 0.12 mm/rev. The identified parameters and their
lower and upper limits are given in Table 3.4. The cutting tool is made up of solid
carbide coated with polycrystalline diamond (PCD) as shown in Figure 3.6 and its
specifications are shown in Table 3.5.
Table 3.4 Process control parameters and their limits
Process parameters
Cutting speed
Fibre orientation
angle
Depth of cut
Feed rate
Units
m/min
o
(deg)
mm
mm/rev
Notation Variable
V
A
B
d
f
C
D
Levels
-2
-1
50
60
15
30
0
75
60
1
90
90
2
100
120
0.05 0.08 0.15 0.22 0.25

0.04 0.05 0.08 0.11 0.12
The fibre orientations are defined in clockwise with reference to the cutting
direction as shown in Figure 3.4. Fixation of GFRP composite material by using
clamps in the machining centre as shown in Figure 3.5. Figure 3.7 shows the milled
GFRP plates with the PCD tool.
58
Figure 3.4 Notation of the fibre orientation with respect to cutting tool
movement
Figure 3.5 Fixation of GFRP composite material by using clamps in the

machining centre
59
Figure 3.6 Solid carbide tool coated with PCD
Table 3.5 Specifications of the tool

Rake angle
Clearance angle 30
Helix angle
30
Flute length
5 mm
Overall length
50 mm
Tool diameter
3 mm
Manufacturer
JarV tools, Chennai
60
Figure 3.7 Milled GFRP Plates with PCD tool

3.5.2 Measurement of Surface roughness (Ra)
The surface roughness (Ra) was evaluated using stylus type profilometer
Mitutoyo SJ-201. In composite machined surface, the result of the roughness depends
mainly on the stylus path with respect to fibre direction since the main direction of
fibres may change from layer to layer. For this reason, the roughness has been
measured several times and averaged as shown in Figure 3.8.
61
Figure 3.8 Diagram of surface roughness measurement

3.5.3 Measurement of Delamination factor (Fd)
The damage caused on the GFRP composite material was measured
perpendicular to the feed rate with a shop microscope Mitutoyo TM-500, as in Figure
3.9. The composite material was positioned and was fixed on the XY stage glass of the
microscope, and then the alignment of an initial measuring point with one of the
cross-hairs was made on the machined feature. Moving the XY stage glass by turning
the micrometer head with a Digital Counter to the final point with the same cross-hair
has been measured the damage (maximum width). After the measurement of the
maximum width of damage (Wmax) suffered by the material, the damage normally
assigned by delamination factor (Fd) was determined. This factor is defined as the
62
quotient between the maximum width of damage (Wmax), and the width of cut (W).
The value of delamination factor (Fd) can be obtained by the following equation:
Fd =
........................................
(3.1)
Wmax being the maximum width of damage in m and W be the width of cut in m
Figure 3.9 Delamination factor measurements

In order to know surface quality and dimensional properties, it is necessary to
employ theoretical models for prediction purpose. For prediction, the response
surface method (RSM) is practical, economical and relatively easy for use.
3.6 RESPONSE SURFACE METHODOLOGY
Response Surface Methodology (RSM) is a collection of mathematical and
statistical techniques that are useful for modelling and analysing of problems in
which an output or response influenced by several variables and the goal is to find the
correlation between the response and the variables. The steps involved in RSM
technique are as follows: (i) designing a set of experiments for adequate and reliable
measurement of the true mean response of interest, (ii) determination of mathematical
model with best fits, (iii) finding the optimum set of experimental factors that
produce maximum or minimum value of response, and (iv) representing the direct
and interactive effects of process variables on the best parameters through two
dimensional and three dimensional graphs. If all variables are assumed to be
measurable, the response surface can be expressed as follows:
y = f (x1, x2,....,xk) ............................................... ....................... (3.2)

63
The goal is to optimise the response variable y. It is assumed that the

independent variables are continuous and controllable by experiments with negligible
errors. It is required to find a suitable approximation for the true functional
relationship between independent variables and the response surface. Usually a
second-order model is utilised in response surface methodology.
= 0 +
=1
2
=1
=1
+ ....................................(3.3)
Where is a random error, is the coefficients, which should be determined

in the second-order model obtained by the least square method. An experiment is a
series of trials or tests, which produces quantifiable outcomes. Owing to slightly
wider ranges of the factors, it was decided to use a five level, central composite,
rotatable design matrix to optimise the experimental conditions. Central composite
rotatable designs of second order were found to be the most efficient tool in RSM to
establish the mathematical relation of the response surface using the smallest possible
number of experiments without losing its accuracy. In the present case, the size of the
experiment is 31 for four machining parameters.
Table 3.6 shows the 31 set of coded conditions used to form the central
composite rotatable design matrix. It comprises full replication of 24 = 16 factorial
design plus 7 centre points and 8 star points. All chosen variables at the intermediate
level (0) constitute the centre points and the combinations of each of the variables at
either lowest (-2) or the highest (+2) with the other three variables of the intermediate
levels constituting the star points. Thus, the 31 experimental runs allowed the
estimation of the linear, quadratic and two way interactive offers of the variables on
the surface roughness and delamination factor.
64
Table 3.6 Layout of central composite rotatable design with results
Trial
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Coded variables
A
-1
1
-1
1
-1
1
-1
1
-1
1
-1
1
-1
1
-1
1
-2
2
0
0
0
0
0
0
0
0
0
0
0
0
0
B
-1
-1
1
1
-1
-1
1
1
-1
-1
1
1
-1
-1
1
1
0
0
-2
2
0
0
0
0
0
0
0
0
0
0
0
C
-1
-1
-1
-1
1
1
1
1
-1
-1
-1
-1
1
1
1
1
0
0
0
0
-2
2
0
0
0
0
0
0
0
0
0
Average
Surface
roughness
(m)
Uncoded variables
D
-1
-1
-1
-1
-1
-1
-1
-1
-1
1
-1
1
-1
1
-1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
V
60
90
60
90
60
90
60
90
60
90
60
90
60
90
60
90
50
100
75
75
75
75
75
75
75
75
75
75
75
75
75
30
30
90
90
30
30
90
90
30
30
90
90
30
30
90
90
60
60
15
120
60
60
60
60
60
60
60
60
60
60
60
d
0.08
0.08
0.08
0.08
0.22
0.22
0.22
0.22
0.08
0.08
0.08
0.08
0.22
0.22
0.22
0.22
0.15
0.15
0.15
0.15
0.05
0.25
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.15
f
0.05
0.05
0.05
0.05
0.05
0.05
0.05
0.05
0.11
0.11
0.11
0.11
0.11
0.11
0.11
0.11
0.08
0.08
0.08
0.08
0.08
0.08
0.04
0.12
0.08
0.08
0.08
0.08
0.08
0.08
0.08
Delamination
Factor
Ra
Fd
1.736
1.543
1.660
1.233
1.150
0.956
1.740
1.313
2.026
1.833
2.226
1.796
1.440
1.246
2.303
1.876
2.040
1.320
1.286
1.906
1.883
1.376
1.146
2.056
1.513
1.550
1.270
1.593
1.473
1.593
1.630
1.0420
1.0020
1.0844
1.0577
1.0011
1.0035
1.0825
1.1160
1.1445
1.0681
1.1482
1.0867
1.1020
1.0433
1.1416
1.1260
1.1054
1.0327
1.0135
1.1340
1.1300
1.1083
1.0020
1.1270
1.0673
1.0643
1.0632
1.0550
1.0571
1.0598
1.0692
For the convenience of recording and processing the experimental data, the
upper and lower levels of the parameters are coded as +2 and -2. The coded value of
any intermediate levels can be calculated by using the following expression.
65
Xi =
[2 + ]

2
.......................................... (3.4)
where, Xmax is the upper level of the parameter, Xmin is the lower level of the parameter
and Xi is the required coded values of the parameter of any value of X from Xmin to
Xmax.
3.6.1. Results and discussions

Surface roughness evaluation is very important for many fundamental
problems such as friction, contact deformation, heat and electric current conduction,
tightness of contact joints and positional accuracy. For this reason, surface roughness
has been the subject of experimental and theoretical investigations for many decades.
Surface roughness imposes one of the most critical constraints for the selection of
machines and cutting parameters in process planning. The study of surface roughness
characteristics of GFRP composites depends on many factors, and is more influenced
by the cutting parameters like cutting speed, fibre orientation angle, depth of cut, feed
rate etc., for a given machine tool and work piece set-up.
The fit summary recommended that the quadratic model is statistically
significant for analysis of surface roughness and delamination. The results of the
quadratic model for surface roughness and delamination in the form of ANOVA are
given in Table 3.7 and 3.8. The value of R2 for surface roughness and delamination
are 97.24% and 98.89% respectively. This means that regression model provides an
excellent explanation of the relationship between the independent factors and the
responses. The associated p value for the model is lower than 0.05 (i.e., =0.05 or
95% confidence) which shows that the model is considered to be statistically
significant. Further, factor A, B, C and D, their square effects A2 and their
interactions AB, BC and BD only have significant effects on surface roughness.
Factor A, B and D, their square effects C2 and their interactions AB, AC, AD, BC and
BD only have significant effects on delamination.
66
Table 3.7 ANOVA for surface roughness (before elimination)

Source
Model
Sum of
squares
3.32
F value
14
Mean
Square
0.24
40.25
p-value
(Prob>F)
<0.0001
Significant
Percentage
Contribution
97.41
0.64
0.64
109.06
<0.0001
Significant
18.67
0.50
0.50
84.60
<0.0001
Significant
14.56
0.39
0.39
65.55
<0.0001
Significant
11.43
1.14
1.14
194.00
<0.0001
Significant
33.43
AB
0.055
0.055
9.32
0.0076
Significant
1.41
AC
1.062E005
9.557E005
9.557E007
12.73
insignificant
0.00
0.9923
insignificant
0.00
<0.0001
Significant
12.90
6.250E008
5.263E007
5.625E007
0075
0.9974
BD
6.250E008
5.263E007
5.625E007
0.075
0.0026
Significant
2.19
CD
0.44
0.44
75.30
0.9923
insignificant
0.00
A2
0.075
0.075
12.73
0.0111
Significant
1.41
B2
9.557E007
3.96
insignificant
0.35
5.625E007
0.075
0.1788
C2
5.625E007
0.075
0.0640
insignificant
0.67
D2
0.013
0.048
2.23
0.1548
insignificant
0.38
Error
0.094
16
0.012
5.886E003
insignificant
2.59
Total
3.41
30
AD
BC
df
1
1
Effect
R2 =0.9724, Adjusted R2 =0.9482

The result shows that the feed rate is more significant parameter for the
surface roughness and delamination, compared to the cutting seed, fibre orientation
angle and depth of cut because of higher F value. The other model terms are said to
be insignificant.
67
Table 3.8 ANOVA for delamination (before elimination)

Source
Sum of
squares
Model
0.061
df
Mean
square
F value
p-value
(Prob>F)
Effect
Percentage
contribution
14 4.358E003
101.96
< 0.0001
significant
98.84
6.285E003
6.285E003
141.03
< 0.0001
significant
10.14
0.019
0.019
447.61
< 0.0001
significant
30.64
1.552E004
1.552E004
3.63
0.0748
insignificant
0.25
0.022
0.022
507.07
< 0.0001
significant
35.48
AB
6.558E004
6.558E004
15.34
0.0012
significant
1.06
AC
1.726E003
1.726E003
40.38
< 0.0001
significant
2.78
AD
2.056E003
2.056E003
48.10
< 0.0001
significant
3.31
BC
2.395E003
2.395E003
56.04
< 0.0001
significant
3.86
BD
1.357E003
1.357E003
31.76
< 0.0001
significant
2.19
CD
1.666E004
1.666E004
3.90
0.0659
insignificant
0.27
A2
3.520E005
3.520E005
0.82
0.3776
insignificant
0.05
B2
1.492E004
1.492E004
3.49
0.0801
insignificant
0.24
C2
5.316E003
5.316E003
124.37
< 0.0001
significant
8.57
D2
2.236E008
2.236E- 5.231E008
004
0.9820
insignificant
0.00
68
Error
6.839E004
Total
0.062
16 4.274E005
0.004
insignificant
1.16
30
R2 =0.9889, Adjusted R2 =0.9792

The adequacy of the model is checked by using the ANOVA technique. As
per this technique, if the calculated value of the F ratio of the developed model does
not exceed the standard tabulated value of F ratio for a desired level of confidence
(say 99%), then the model is considered to be adequate within the confidence limit.
The variance ratio, denoted by F in ANOVA tables, is the ratio of the mean square
due to a factor and the error mean square. In robust design, F ratio can be used for
qualitative understanding of the relative factor effects. A high value of F means that
the effect of that factor is larger compared to the error variance. So, the larger the
value of F the more important is that factor in influencing the process response.
The adequacy of the model was checked by using the analysis of variance
(ANOVA) technique. As per this technique, if the calculated value of the F ratio of
the developed model does not exceed the standard tabulated value of F ratio for a
desired level of confidence (say 99%), then the model is considered to be adequate
within the confidence limit. The lack of fit was found to be less than F0.01 (10, 6) in
the present research study and, hence the developed model may be accepted.
To fit the quadratic model for surface roughness and delamination
appropriately, the insignificant terms were eliminated by backward elimination
process. The ANOVA tables for the reduced quadratic model for surface roughness
and delamination are shown in Table 3.9 and Table 3.10. The reduced model results
for surface roughness indicate that the model is significant (R2 and adjusted R2 are
96.42% and 92.84%, respectively).
69
Table 3.9 ANOVA for surface roughness (after backward elimination)

Source
Sum of
square
3.288
df
74.1147
p-value
(Prob>F)
<0.0001
significant
Percentage
contribution
96.42
Cutting
speed(A)
0.64
115.4098
<0.0001
significant
18.77
Fibre
orientation
angle(B)
Depth of
cut(C)
Feed rate
(D)
AB
0.50
90.1639
<0.0001
significant
14.66
0.39
0.39
70.3279
<0.0001
significant
11.44
1.14
1.14
205.5738
<0.0001
significant
33.43
0.055
0.055
9.9180
0.0076
significant
1.61
BC
0.44
0.44
79.3443
<0.0001
significant
12.90
BD
0.075
0.075
13.5246
0.0026
significant
2.20
A2
0.048
0.048
8.6557
0.0111
significant
1.41
Error
0.122
22
0.006
Total
3.41
30
Model
Mean
square
0.411
0.64
0.50
F value
Effect
3.58
R2 =0.9642, Adjusted R2 =0.9284

Table 3.10 ANOVA for delamination (after backward elimination)
Source
Model
Cutting
speed(A)
Fibre
orientation
angle(B)
Feed rate
(D)
AB
AC
AD
Sum of df
squares
0.0608
9
6.285E003
0.019
1
1
Mean
square
6.755E003
6.285E003
0.019
0.022
0.022
385
< 0.0001
significant
35.49
6.558E004
1.726E003
2.056E003
6.558E004
1.726E003
2.056E003
11.5
0.0012
significant
1.06
30.2
< 0.0001
significant
2.78
36.0
< 0.0001
significant
3.32
1
1
F
p-value
value (Prob>F)
118.2 < 0.0001
Effect
significant
Percentage
contribution
98.06
110
< 0.0001
significant
10.14
332.5 < 0.0001
significant
30.65
70
BC
BD
C2
Error
Total
2.395E003
1.357E003
5.316E003
1.200E003
0.062
2.395E003
1 1.357E003
1 5.316E003
21 5.714E005
30
42.0
< 0.0001
significant
3.86
23.7
< 0.0001
significant
2.19
93.0
< 0.0001
significant
8.57
1.94
R2 =0.9806, Adjusted R2 =0.9613

The reduced model results for delamination show that the model is significant
and their corresponding R2 and adjusted R2 values are 98.06% and 96.13%,
respectively. It is seen that the regression model is fairly well fitted with the observed
values. After eliminating the insignificant terms, the final response equation for
surface roughness and delamination is given as follows in uncoded units as:
Ra = 1.51743 0.16354 V + 0.14404 0.12679 d + 0.2182 f 0.058562 V* +
0.16644*d+0.068437*f+0.041174V2............................................................. (3.5)
Fd = 1.06227 0.016182 V + 0.028235 + 0.030051 f + 6.40188E-003 V* +
0.010386 V*d0.011336 V*f+ 0.012236 *d 9.21063E-003 *f + 0.013635 d2
.......................................... (3.6)
Figure 3.10 (a) and (b) shows the correlation between the predicted and
experimental values for surface roughness and delamination. The influence of
different cutting parameters on machining of GFRP composites were studied by using
response graph and response table. The influence of cutting parameters on surface
roughness and delamination are shown in Figure 3.11 and Figure 3.12 and their main
effects are shown in Table 3.11 and Table 3.12. From the Figure 3.11, it is observed
that surface roughness increases with increasing the feed rate and fibre orientation
angle, whereas the surface roughness decreases with increasing of the cutting speed
and depth of cut. From the Figure 3.12, it is observed that delamination factor value
increases with increasing the feed rate and fibre orientation angle, whereas, the
delamination factor value decreases with increasing of the cutting speed. In case of
depth of cut, the delamination factor value first decreases and then increases. Based
on the observations from Table 3.11 and Table 3.12, it is asserted that feed rate is the
main parameter which affects the surface roughness and delamination factor. Figure
71
3.13 and Figure 3.14 show the response surfaces and contour plots for the effect of
cutting speed, fibre orientation angle, and depth of cut and feed rate on surface
roughness and delamination factor.
(a)
(b)
Figure 3.10 Correlation graph for (a) surface roughness, (b) delamination
Table 3.11 Response table for surface roughness
Level
1
2
3
4
5
Delta
Rank
Cutting Speed
(V), m/min
2.04
1.785
1.559
1.474
1.32
0.720
2
Fibre orientation angle

() deg
1.286
1.491
1.572
1.768
1.906
0.620
3
72
Depth of cut
(d), mm
1.883
1.757
1.567
1.503
1.376
0.507
4
Feed rate (f),

mm/rev
1.146
1.416
1.572
1.843
2.056
0.910
1
Figure 3.11 Illustration of factor effects on surface roughness

Table 3.12 Response table for Delamination
Level
1
2
3
4
5
Delta
Rank
Cutting Speed
(V), m/min
1.1054
1.0933
1.0731
1.0629
1.0327
0.0727
3
Fibre orientation
angle () deg
1.0135
1.0508
1.0724
1.1054
1.1340
0.1205
2
Depth of cut
(d), mm
1.1300
1.0792
1.0655
1.0770
1.1083
0.0645
4
Feed rate (f),

mm/rev
1.0020
1.0486
1.0738
1.1075
1.1270
0.1250
1
Figure 3.12 Illustration of factor effects on delamination factor (Fd)

3.6.2 Influence of machining parameters on surface roughness
The effect of feed and cutting speed on surface roughness is shown in Figure
3.13a. As seen from the Figure 3.13a, the surface roughness tends to increase steadily
with an increase in feed rate and decreases considerably with an increase of cutting
speed. This is due to an increase in the feed per tooth (resulting from increasing the
feed rate) causes a significant rise in surface roughness. This may be attributed to the
heat generated as a result of higher friction. When the cutting speed varies from 50 to
61 m/min and the feed rate ranges from 0.06 to 0.12 mm/rev, the surface roughness
73
values lie within 2.000m as seen from the contour graph in Figure 3.13b. Further, it
is also seen that while the cutting speed is between 50 to 53 m/min and the feed rate
in the range of 0.11 and 0.12 mm/rev, the surface roughness appears to rise above
2.200m.
Figure 3.13 (a) Response surface graph shows the effect of feed rate and cutting
speed on surface roughness
Figure 3.13 (b) Contour plots shows the effect of feed rate and cutting speed on
surface roughness
Figure 3.13c shows the estimated response surface for the surface roughness
in relation to the individual parameters of cutting speed and fibre orientation. As seen
74
from the Figure 3.13c, the surface roughness tends to decrease with an increase of
cutting speed and increases with an increase of fibre orientation angle. From the
contour graph (Figure 3.13d), when the fibre orientation ranges from 60o - 120o and
the cutting speed varies from 50-68 m/min, the surface roughness values lie within
2.000m. Surface roughness increases with an increase in feed rate, but the effect of
fibre orientation is magnified at higher feed rates.
Figure 3.13 (c) Response surface graph shows the effect of fibre orientation angle
and cutting speed on surface roughness
Figure 3.13 (d) Contour plot shows the effect of fibre orientation angle and
cutting speed on surface roughness
75
Figure 3.13e shows the effect of depth of cut and feed rate on surface
roughness. As seen from the Figure 3.13e, the surface roughness tends to increase
steadily with an increase in feed rate and decreases slightly with an increase of depth
of cut. As read from the contour graph (Figure 3.13f), the surface roughness value lies
within 1.600m when the depth of cut varies from 0.04 to 0.25 mm and the feed rate
ranges from 0.049 to 0.082 mm/rev.
Figure 3.13 (e) Response surface graph shows the effect of feed and depth of cut
on surface roughness
Figure 3.13 (f) Contour plot shows the effect of feed and depth of cut on surface
roughness
76
3.6.3 Influence of machining parameters on delamination factor

Figure 3.14a shows the estimated response surface for the delamination factor
in relation to the individual parameters of the cutting speed and feed rate. As seen
from Figure 3.14a, the delamination tends to increase steadily by increase in feed rate
and decreases slightly with an increase of cutting speed. This is due to the increase in
the feed rate causing a sharp rise in the feed force which, in turn, causes higher
friction and produces more damage on the surface. As read from the contour graph
(Figure 3.14b), the delamination value lies within 1.1000, when the cutting speed
varies 67 to 88 m/min and feed rate ranges from 0.08 to 0.11 mm/rev.
77
Figure 3.14 (a) Response surface graph shows the effect of feed rate and cutting
speed on delamination factor
Figure 3.14 (b) Contour plot shows the effect of feed rate and cutting speed on
delamination factor
Figure 3.14c shows the effect of cutting speed and fibre orientation on
delamination factor. As seen from the Figure 3.14c, the delamination tends to
increase considerably with an increase of fibre orientation angle and decreases
slightly with an increase of cutting speed. As seen from the contour graph (Figure
3.14d), the delamination factor value lies within 1.0400, when the cutting speed
varies from 68 to 100 m/min and fibre orientation angle ranges from 15o to 48o.
78
Figure 3.14 (c) Response surface graph shows the effect of fibre orientation angle
and cutting speed on delamination factor
Figure 3.14 (d) Contour plot shows the effect of fibre orientation angle and
cutting speed on delamination factor
The effect of feed and fibre orientation angle on delamination is shown in fig
3.14e. As seen from the Figure 3.14e, the delamination tends to increase considerably
with an increase of both fibre orientation angle and feed rate. When the feed ranges
between 0.04 to 0.058 mm/rev and the fibre orientation angle is from 15o to 40o, the
delamination value lies within 1.0200 as observed from the contour graph (fig 3.14f).
79
Figure 3.14 (e) Response surface graph shows the effect of fibre orientation angle
and feed rate on delamination factor
Figure 3.14 (f) Contour plot shows the effect of fibre orientation angle and feed
rate on delamination factor
At fibre orientation 15o the fibres are subjected to slight bending and tensile
loading which cause the fibres to break in tension and little subsurface delamination.
At fibre orientation 40o the fibres are subjected to severe bending and compression
loading, which cause the fibres to break in compression shear. Significant subsurface
delamination and cracking in the fibres take place in this mode of chip formation.
80
3.7 COMPARISON OF RSM PREDICTION AND EXPERIMENTAL VALUES

In order to check the accuracy of the Response surface model, the
experimental values were compared with the RSM predicted values. Figure 3.15 and
3.16 show the comparison between the predicted (RSM) and experimental values for
surface roughness and delamination and it is observed that there is very low ( 3%)
variation between experimental results and RSM prediction.
Figure 3.15 Comparison of Experimental surface roughness values with RSM

prediction
81
Figure 3.16 Comparison of Experimental Delamination values with RSM

prediction
3.8 SUMMARY
From the experimental results presented, the following conclusions were
drawn from milling GFRP composite materials with solid carbide coated with PolyCrystalline Diamond (PCD) tool using response surface methodology. The feed rate
is the cutting parameter which has greater influence on surface roughness (33.43%)
and delamination (35.49%) for GFRP composite materials. The surface roughness
and delamination increase with the increase of feed rate and increase of fibre
orientation angle, which means that the composite damage is larger for higher feed
rate and for higher fibre orientation. The surface roughness and delamination
decrease with the increase of cutting speed. The surface roughness decreases with the
increase of depth of cut. The developed second order response surface model is used
to calculate the surface roughness and delamination of the machined surfaces at
different cutting conditions with the chosen range of 95% confidence intervals. Using
such model, one can obtain remarkable savings in time and cost. This technique is
convenient to predict the main effects and interaction effects of different influential
combinations of machining parameters.
82
CHAPTER 4
ARTIFICIAL NEURAL NETWORK MODELLING
4.1 INTRODUCTION
An artificial neural network is an information-processing system that has
certain performance characteristics in common with biological neural networks.
A neural network is characterised by
(1) Its pattern of connection between the neurons (called its
architecture)
(2) Its method of determining the weights on the connections (called
its learning or training algorithm)
(3) Its activation function.
A neural net consists of a large number of simple processing elements called
neurons, units, cells or nodes. Each neuron is connected to other neurons by means of
directed communication links, each with an associated weight. These weights
represent information being used by the net to solve a problem. Neural nets can be
applied to patterns, or finding solutions to constrained optimisation problems.
Each neuron has an internal state called its activation or activity level, which
is a function of the inputs it has received. Typically, a neuron sends its activation as a
signal to several other neurons. It is important to note that a neuron can send only one
signal at a time, although that signal is broadcast to several other neurons.
The neural network can be trained to perform a particular function by
adjusting the values of the connections (weights) between elements. Commonly,
neural networks are adjusted or trained, so that a particular input leads to a specific
target output. Such a situation is shown in the Figure 4.1. There, the network is
adjusted based on a comparison of the output and the target, until the network output
83
matches the target. Typically many such input/target pairs are used in this supervised
learning, to train a network.
A VERY SIMPLE NEURAL

NETWORK
z1
z2
Figure 4.1 A simple neural network model

In this chapter, an artificial neural network (ANN) model was developed to predict
surface roughness and delamination on the milled GFRP composites. In the
development of predictive models, cutting parameters of cutting speed, fibre
orientation, depth of cut and feed rate were considered as model variables.
4.2. ARCHITECTURE:
Often, it is convenient to visualise neurons as arranged in layers. Typically,
neurons in the same layer behave in the same manner. Key factors in determining the
behavior of a neuron are its activation function and the pattern of weighted
connections over which it sends and receives-signals. Within each layer, neurons
usually have the same activation function and the same pattern of connections to
other neurons. To be more specific, in many neural networks, the neurons within a
layer are either fully interconnected or not interconnected to a neuron in another layer
(say, the output layer), then each hidden unit is connected to every output neuron.
84
The arrangement of neurons into layers and the connection patterns within and
between layers is called the net architecture. Many nets have an input layer in which
the activation of each unit is equal to an external input signal.
Neural nets are often classified as single layer or multilayer. In determining
the number of layers, the input units are not counted as a layer, because they perform
no computation. Equivalently, the number of layers in the net can be defined to be the
numbers of layers of weighed interconnect links between the slabs of neurons. This
view is motivated by the fact that the weights in a net are extremely important for
information.
4.2.1 Single layer net
A single layer net has one layer of connection weights and it is shown in
Figure 4.2. Often, the units can be distinguished as input units which receive signals
from the outside world, and output units from which the response of the net can be
read. In the typical single layer net as shown in figure, the input units are fully
connected to output units but are not connected to other input units and the output
units are not connected to other output units.
4.2.2 Multi layer net
A multilayer net is a net with one or more layers (or levels) of nodes (then so
called hidden units) between the input units and the output units and is shown in
Figure 4.3. Typically, there is a layer of weights between two adjacent levels of units
(input, hidden, or output). Multilayer nets can solve more complicated problems than
single layer nets, but training may be more difficult. However, in some cases training
may be more successful, because it is possible to solve a problem that a single layer
net cannot be trained to perform correctly at all.
85
SINGLE LAYER NET
Figure 4.2 Single layer network model
MULTI LAYER NET
Figure 4.3 Multi layer network model

4.3 SETTING THE WEIGHTS:
In addition to the architecture, the method of setting the values of the weights
(training) is an important distinguishing characteristic of different neural nets. There
86
are two distinguished types of training for a neural network-supervised and

unsupervised. In addition, there are nets whose weights are fixed without an iterative
training process.
There is some ambiguity in the labelling of training methods as supervised or
unsupervised, and there is also a third category, self-supervised training, which seems
to be useful. However, in general, there is a useful correspondence between the type
of training that is appropriate and the type of problem to be solved. The basic
characteristics and the types of problems for which each, as well as the fixed-weight
nets is typically used are summarised.
4.3.1. Supervised Training
Perhaps in the most typical neural net setting, training is accomplished by
presenting a sequence of training vectors or patterns, each with an associated target
output vector. The weights are then adjusted to a learning algorithm. This process is
known as supervised training.
4.3.2 Unsupervised Training:
Self-organising neural nets group similar input vectors together without the
use of training data to specify what a typical member of each group looks like or to
which group each vector belongs. A sequence of input vectors is provided, but no
target vectors are specified. The net modifies the weights so that most similar input
vectors are assigned to the same output (or cluster) unit. The neural net will produce a
representative vector for each cluster formed.
4.3.3 Fixed-weight Nets:
Still, other types of neural nets can solve constrained optimisation problems.
Such nets may work well for problems that can cause difficulty for traditional
techniques, such as problems with conflicting constraints (i.e. not all constraints can
be satisfied simultaneously). Often, in such cases, a nearly optimal solution (which
the net can find) is satisfactory. When these nets are designed, the weights are set to
represent the constraints and the quantity to be maximised or minimised.
87
4.4 ACTIVATION FUNCTION

The basic operation of an artificial neuron involves summing its weighted
input signal and applying an output, or activation function. For the input units, this
function is the identity function. Typically, the same activation function is used for all
neurons in any particular layer of a neural net, although this is not required. In most
cases, nonlinear activation function is used. In order to achieve the advantages of
multilayer nets compared with the limited capabilities of single layer nets, nonlinear
functions are required.
Some of the activation functions normally used are:
(i)
Identity function
(ii)
Binary step function
(iii)
Binary sigmoid
(iv)
Bipolar sigmoid
4.5 BACK PROPAGATION NEURAL NETWORK (BPNN)

The back propagation algorithm has made it possible to design multi-layer
neural networks for numerous applications, such as adaptive control, classification of
sonar targets, stock market prediction and speech recognition. Also, BPNN has the
advantage of fast response and high learning accuracy. One of the advantages of
using the neural network approach is that a model can be constructed very easily
based on the given input and output and trained to predict process dynamics
accurately. This technique is especially valuable in processes where a complete
understanding of the physical mechanisms is very difficult, or even impossible to
acquire, as in the case of porous powder performs during upsetting. Neural network is
a logical structure with multi-processing elements, which are connected through
interconnection weights. The knowledge is presented by the interconnection weights,
which are adjusted during the learning phase. There are several algorithms available
among which the LevenbergMarquardt algorithm (trainlm) will have the fastest
convergence. In many cases, trainlm is able to obtain lower mean square errors than
any of the other algorithms tested. This BP network is a multi-layer of the network
88
architecture including the input layer, hidden layer(s) and output layer. Layers include
several processing units known as neurons. They are connected with each other by
variable weights to be determined. In the network, the input layer receives
information from external source and passes this information to the network for
processing. The hidden layer receives from the input layer, and does all information
processing. The output layer receives processed information from the network, and
sends the results to an external receptor. The algorithm for the back propagation
program is described with the help of flow diagram as shown in Figure 4.4.
89
Start
Normalise the raw data (input and output)

Select two-third data for training and one-third data for testing
Feed the data into Neural Network Tool box (MATLAB 10)
Design Network type, Training function, Perform function, Number of layers, Neurons &
Transfer function
Set Training parameters
Execute network training
Yes
Error within
tolerance?
Select next architecture
Declare success/save trained weights
No
Declare failure/network did not converge within
Feed the entire data set to that corresponding architecture
Denormalise the test & entire data
Find the correlation coefficient (R)
No
Is mean R0.99?
Yes
Select optimum architecture
Figure 4.4 Flow diagram
90
4.6 MODEL DESCRIPTION

In the development of a multi-layer neural network model, several decisions
regarding the number of neuron(s) in the input layer, number of hidden layer(s),
number of neuron(s) in the hidden layer(s), and number of neuron(s) in the output
layer and optimum architectures have to be decided. Based on the experimental
condition, the important input parameters, such as the fibre orientation angle, helix
angle, feed rate and spindle speed are given as input parameters to the present ANN
model. The output parameters are the surface roughness and delamination factor. The
input/output dataset of the model is illustrated schematically in Figure 4.5.
Figure 4.5 Schematic illustration of neural network model
4.6.1 Data normalisation

In order to provide equal importance of the influence of the smaller value as
like higher valued input variables, both the input and the output variables were
normalised within the range of 1 to 1 before the training of the network. The
normalised values (xn) for each raw input/output dataset (di) were calculated as
2( )
( )
- 1 ...................................................................... (4.1)
Where dmax and dmin are the maximum and the minimum values of raw data.
91
4.7 NEURAL NETWORK DESIGN

The generalisation capability of the neural network is essentially dependent on
(i) the selection of the appropriate input/output parameters of the system, (ii) the
distribution of the dataset, and (iii) the format of the presentation of the dataset to the
network. In this study, the total number of experimental results is 31 dataset among
which 22 (two-third) dataset were considered for training, and 9 for testing. Before
training the network, the input/output dataset were normalised within the range of 1
using the Eq. (4.1). The standard multi-layer feed forward back propagation
hierarchical neural networks were designed in the MATLAB 10 Neural Network
Toolbox. The networks consist of three layers, namely, the input, hidden layer, and
output layer. Now, the designed network has four input neurons and two output
neurons. Initially, four neurons in input layer, 2 hidden neurons in one hidden layer
and 2 neurons in output layer (422) architecture was taken and trained with 1000
epochs (iterations) to predict the optimum results. The MSE achieved from different
iterations is shown in Table 4.1. It is evident that the value of MSE gradually
decreases during the progress of training.
4.8 NEURAL NETWORK TRAINING

In this study, the total number of data obtained from experimental
investigation is 31 set for each response among which 22 (two-third) data set were
considered for training, and 9 for testing. The standard multilayer feed forward back
propagation hierarchical neural networks were designed with MATLAB 10 Neural
Network Toolbox. The networks consist of three layers: the input, the hidden layer,
and the output layer. Now, the designed network has four input neurons and two
output neurons. In the network, each neuron receives total input from all the neurons
in the proceeding layer as:
=
................................................................ (4.2)
Where netj is the total or net input, Xi n is the output of the node j in the nth
layer, and Wijn represents the weights from node i in the (n-1) th layer to node j in the
nth layer. A neuron in the network produces its input by processing the net input
through an activation (transfer) function which is, usually nonlinear. There are several
92
types of activation functions used for BP. However, the tan-sigmoid transfer function
is mostly used which is assigned in hidden layer (s) for processing the inputs as:
2
1+
1 1, 1 ...................................... (4.3)
The purelin, a transfer function, calculates a hidden layers output from its net
input which is assigned for output layer as:
=
1 1, 1 ....................................... (4.4)
The weights are dynamically updated using the BP algorithm. The network
has been trained with Levenberg Marquardt algorithm. This training algorithm has
been selected due to its high accuracy in similar function approximation. Default
training parameters available in MATLAB 10 was set for training the dataset. In
order to judge the performance of the network, the average error (MSE) has been
calculated as:
=
=1
=1
............................................................ (4.5)
In the case of BPNN training, the number of iterations (epochs) to be executed is an

important parameter, which is used in the model, as presented in Table 4.1. To
determine the best configuration several structures have to be considered with
different numbers of hidden neurons.
93
Table 4.1 Correlation coefficient between the network predictions and the
experimental values using the test and the entire dataset of different network
architecture and trained output parameters
Network
Data set Surface
architecture
roughness
Delami Network performance
Mean
Mean
-nation during training
correlation
prediction
Factor
Ra
Fd
error (%)
MSE
State of
convergence
4-2-2
4-4-2
4-6-2
4-8-2
4-10-2
4-12-2
4-14-2
4-18-2
4-22-2
4-26-2
4-30-2
4-40-2
4-4-4-2
Test
0.9647
0.9863
Entire
0.9669
0.9872
Test
0.9864
0.9712
Entire
0.9919
0.9718
Test
0.9870
0.9719
Entire
0.9929
0.9798
Test
0.9901
0.9743
Entire
0.9934
0.9873
Test
0.9912
0.9763
Entire
0.9914
0.9896
Test
0.9919
0.9799
Entire
0.9923
0.9907
Test
0.9929
0.9902
Entire
0.9934
0.9912
Test
0.9927
0.9908
Entire
0.9935
0.9932
Test
0.9932
0.9944
Entire
0.9981
0.9944
Test
0.9957
0.9908
Entire
0.9958
0.9909
Test
0.9967
0.9889
Entire
0.9972
0.9868
Test
0.9974
0.9885
Entire
0.9989
0.9888
Test
0.9918
0.9809
Entire
0.9938
0.9838
4.910E-02
3.860E-02
1.440E-02
8.970E-03
3.550E-03
3.280E-03
3.270E-03
8.752E-04
9.435E-04
5.220E-04
3.929E-04
1.172E-04
4.280E-03
94
Not
0.9754
4.258
converged
0.9766
4.171
Not
0.9786
4.008
converged
0.9810
3.581
Not
0.9791
3.766
converged
0.9863
3.359
Not
0.9805
3.128
converged
0.9901
2.522
Not
0.9816
3.071
converged
0.9902
2.478
Not
0.9864
2.744
converged
0.9907
2.321
Not
0.9913
1.782
converged
0.9918
1.681
Not
0.9917
1.694
converged
0.9919
1.511
Not
0.9924
1.502
converged
0.9953
1.012
Not
0.9931
1.444
converged
0.9932
1.439
Not
0.9926
1.452
converged
0.9928
1.329
Not
0.9928
1.328
converged
0.9936
1.295
Not
0.9867
2.652
converged
0.9889
3.127
4-6-6-2
4-10-10-2
4-12-12-2
4-14-14-2
4-16-16-2
4-20-20-2
4-22-22-2
4-24-24-2
4-26-26-2
4-28-28-2
4-30-30-2
4-32-32-2
4-36-36-2
4-40-40-2
Test
0.9929
0.9887
Entire
0.9937
0.9889
Test
0.9918
0.9827
Entire
0.9988
0.9928
Test
0.9952
0.9782
Entire
0.9974
0.9929
Test
0.9902
0.9752
Entire
0.9942
0.9922
Test
0.9952
0.9808
Entire
0.9975
0.9940
Test
0.9932
0.8862
Entire
0.9915
0.9899
Test
0.9932
0.9562
Entire
0.9962
0.9915
Test
0.9919
0.9854
Entire
0.9955
0.9921
Test
0.9921
0.9021
Entire
0.9979
0.9882
Test
0.9942
0.9713
Entire
0.9969
0.9921
Test
0.9982
0.9843
Entire
0.9991
0.9934
Test
0.9959
0.9785
Entire
0.9991
0.9926
Test
0.9967
0.9662
Entire
0.9989
0.9916
Test
0.9975
0.9791
Entire
0.9991
0.9934
2.617E-03
6.771E-04
3.825E-04
9.734E-05
7.697E-05
3.364E-05
2.100E-05
4.320E-06
5.762E-07
3.750E-07
1.510E-30
3.440E-30
2.712E-30
1.930E-30
Not
0.9902
2.477
converged
0.9909
2.258
Not
0.9874
3.178
converged
0.9952
1.016
Not
0.9869
3.378
converged
0.9948
1.127
Not
0.9827
3.277
converged
0.9924
1.523
Not
0.9874
2.601
converged
0.9951
1.018
Not
0.9550
5.002
converged
0.9908
2.316
Not
0.9750
4.268
converged
0.9936
1.395
Not
0.9863
3.358
converged
0.9938
1.207
Not
0.9532
5.128
converged
0.9928
1.471
Not
0.9827
3.311
converged
0.9939
1.102
Converged
0.9884
1.978
0.9968
1.008
0.9881
2.007
0.9958
1.067
0.9795
4.007
0.9954
1.124
0.9889
2.002
0.9962
1.019
Converged
Converged
Converged
The machinability of GFRP composites have been trained with different

architecture by varying number of neurons in the hidden layer(s). After training, it has
been denormalised and compared with the experimental data as shown in Figure 4.6
and 4.7. The denormalised values (xi) for each raw output dataset was calculated as
+1 ( )
2
95
+ (4.6)
Where dmax and dmin are the maximum and minimum values of raw data.
Figure 4.6 Comparison of Experimental surface roughness values with ANN

prediction
96
Figure 4.7 Comparison of Experimental Delamination values with ANN

prediction
For testing the prediction ability of the model, prediction error in each output
node has been calculated as follows:
% =
( )

100.............. (4.7)
In order to determine the optimal architecture, a total of 28 different networks

with different number of layers and neurons in the hidden layer have been designed
and tested for GFRP composites.
4.9 TESTING AND PERFORMANCE OF BPNN
The performance capability of each network has been examined based on the
correlation coefficient, error distribution, and convergence of entire dataset within
specified error range between the network predictions and the experimental values
using the test and entire dataset. For deciding the optimum structure of neural
network, the rate of error convergence was checked by changing the number of
97
hidden neurons and number of hidden layers. The error distribution was not uniform
when there was an increase of neuron in the single hidden layer (beyond 40). Hence,
it was decided to select two hidden layers and varied the number of neurons in each
hidden layer to get an optimum one. It was observed from the Table 4.1 that the
network with 30 neurons in each hidden layer has produced the best performance for
each of the output parameters (430302). It is also observed that the architectures
432322, 436362, and 440406, do not show more difference in the mean
correlation coefficient than the previous architecture. In the process it had not been
selected as an optimum architecture, because the mean prediction error as well as the
error distribution, maximum value of error, minimum value of error were observed as
high, which can be evident from Table 4.1.
4.10 COMPARISON OF RSM AND ANN MODELS
The main advantage of RSM is its ability to exhibit the factor contributions
from the coefficients in the regression model. This ability is powerful in identifying
the insignificant main factors and the interaction factors or insignificant quadratic
terms in the model and thereby can reduce the complexity of the problem. On the
other hand, this technique requires good definition of ranges for each factor to ensure
that the response(s) under consideration changes in a regular manner within this
range.
98
Figure 4.8 Comparison of RSM prediction Error with ANN prediction Error
It is noted that ANN models perform better than other techniques, especially
RSM, when highly non-linear behaviour is the case, as shown in Figure 4.8. Further,
this technique can build an efficient model using a small number of experiments.
However, the accuracy of the technique would be better when a larger number of
experiments were used to develop a model. On the other hand, the ANN model itself
provides little information about the design factors and their contribution to the
response if further analysis has not been done.
Generation of ANN model requires a large number of iterative calculations,
whereas, it is only a single step calculation for a response surface model. Depending
upon the nonlinearity of the problem and the number of parameters, an ANN model
may require a high computational cost to create. Although computationally much
more costly than a response model, ANN model leads to comparatively accurate
predictions of responses as shown in Table 4.2. The mean errors for ANN and RS
models are about 0.258847% and 0.769831% respectively. The error against
observation order of both the models is compared in Figure 4.8.
99
Table 4.2 Comparison between RSM and ANN models

Model summary and
prediction errors
RSM
ANN
Mean Square Error (MSE)
6.75E-03
1.510E-30
R2
0.97240
0.99260
Mean error/%
0.769831
0.258847
Computational time
Short
Long
Experimental domain
Regular
Irregular or regular
Model developing
With interactions
No interactions
Understanding
Easy
Moderate
Application
Frequently
Frequently
4.11 SUMMARY
A model using Artificial Neural Networks (ANN) was developed to study the
Machinability of GFRP composite during endmilling. The feed forward neural
network was trained with training samples, using the standard back propagation
algorithm. In the development of predictive models, cutting parameters of cutting
speed, fibre orientation, depth of cut and feed rate were considered as input model
variables and surface roughness and delamination factor were considered as output
100
responses. By using input/output data of the experiments, the model was trained and
tested. Data normalisation was applied to improve the accuracy and efficiency of the
behavioural characteristics of neural networks. The network consists of one input
layer with four neurons representing the four input parameters and one output layer
with two output neurons indicating the two output parameters. The weights were
randomly generated for the first iteration. The network was trained using Levenberg
Marquardt algorithm. The network was trained initially with two neurons in the
hidden layer i.e., 4-2-2 topology was considered for training. In the subsequent
phases, the number of hidden neurons in the hidden layer were increased gradually
from 2 to 40 and then tested with two hidden layers with the same number of hidden
neurons in the second hidden layer. Feed forward back propagation neural network
(BPNN) model with one hidden layer having 30 neurons was found to be optimum
network model (4-30-30-2) which has the mean correlation coefficient of 0.9926. The
developed ANN model was compared with the RSM models for the prediction of
surface roughness and delamination factor of milled GFRP composites. A sound
performance was achieved with the neural network model, and it shows good
correlation between the predicted values of the optimum neural network model and
the experimental data for prediction of surface roughness and delamination using the
entire datasets within acceptable error limits. The predictive ANN model was found
to be capable of better predictions of responses within the range that they had been
trained. The results of the ANN model indicate it is much more robust and accurate in
estimating the values of surface roughness and delamination when compared with the
response surface model.
101
CHAPTER 5
EFFCTS OF MACHINING PARAMETERS ON

MACHINING FORCE
5.1 INTRODUCTION
Machining of composite materials is usually performed to achieve required
geometrical shapes and dimensional tolerances in the research community. However,
machinability evaluation of Glass Fibre-Reinforced Polymer (GFRP) composites in
end milling has not yet received its due attention despite the extensive industrial use
of this process. This work aims to elucidate the end milling machinability of GFRP
composites with respect to machining forces. Experiments were conducted under
different experimental parameters and their levels according to the Taguchi design of
experiment method. Taguchi analysis combined with statistical analysis of variance
(ANOVA) was performed to quantify the effects of tool geometry (helix angle), fibre
orientation, spindle speed, feed rate and depth of cut on the machining force. Multiple
Regression Analysis (MRA) was also employed to establish parametric relationships
between the experimental parameters and the machinability output.
In milling, cutting forces are exerted in three planes to deform and shear away
material in the form of a chip. Tangential cutting forces overcome the resistance to
rotation and account for 70% of the total force while feed forces account for 20%.
Radial forces tend to push away the tool and account for 10% of the cutting forces.
Consequently, this chapter investigates the effects of machining parameters on
machining force of GFRP composites using Taguchi Design of Experiment (DOE)
method. This method was employed to systematically formulate the experimental
layout, analyse the significant influence of each experimental parameter using
statistical Analysis of Variance (ANOVA) and finally predict the optimal parametric
combination to yield the best machining conditions. In addition to Taguchi analysis,
Multiple Regression Analysis (MRA) has also been employed to establish parametric
relationships between the machining parameters and the machining forces. The
102
adequacy of the developed models has been confirmed through the coefficient of
determination (R2) and confirmation tests under randomly selected conditions.
5.2 MACHINE TOOL SETUP AND DATA ACQUISITION
The end milling experiments were carried out on a CNC milling machine and
the machining force components were measured with a Kistler 9265B milling
dynamometer, and the data acquisition was carried out by appropriate software called
Dynawarekistler data acquisition system, as shown in Figure 5.2. The composite plate
was held firmly on top of the dynamometer using clamps as shown in Figure 5.1, to
make sure that vibrations and displacement were eliminated. The experiments were
designed based on Taguchi L9 (34) orthogonal array for the 15, 60 and 105 fibre
oriented GFRP composite plates.
Figure 5.1Fixation of GFRP composite plate on dynamometer by using

clamps in the machining centre
103
Figure 5.2 Experimental setup with data acquisition system
5.3 TOOL GEOMETRY (HELIX ANGLE)

The angle formed by a line tangent to the helix and a plane through the axis of
the cutter is called helix angle. It is also defined as the cutting edge angle, which a
helical cutting edge makes with a plane containing the axis of a cylindrical cutter. All
end mills have a helix angle unless they are straight fluted tools. Helix can be either
right handed or left handed. A right handed helix means the cutter will remove the
chip in an upward motion or away from the work piece. During cutting, this may have
a tendency to lift the part. This is pretty typical for most tools. The left hand helix
tools will force the chip toward the material or solid locator. The advantage of a left
hand helix is a much more rigid setup during cutting but one must have clearance for
the chip. Left-hand helix cutters are used in lot to circular interpolate a bore because
it is a more stable process. Again, it doesnt have a tendency to lift the part.
The end mill cutters which had the geometry of 25, 35and 45 helix angles
with 9 and 16 as primary relief and secondary clearance angles respectively, as
shown in Figure 5.3, were used to perform the machining.
104
Figure 5.3 Solid carbide end mill with different helix angles
5.4
TAGUCHI
EXPERIMENTAL
DESIGN
AND
SELECTION
OF
PARAMETERS
Extensive and expensive experimentations (e.g. time/labour/ materials, etc.)
would typically be required to evaluate the machinability of a material. Hence,
experimental approach of machinability assessments can be well achieved through
statistically designed tests or series, commonly known as design of experiment
(DOE). DOE methodology involves full factorial as well as partial or fractional
approaches. Full factorial experiments may provide all possible effects and
interactions, but the scale of experimentations can be prohibitive for scientific
investigations. Realistically, fractional factorial approaches such as the Taguchi
methodology which involves significantly fewer tests but with highly acceptable or
reliable results would be more attractive. In this study, Taguchi DOE method was
used to design the experimental matrix. Taguchi method systematically plans the
experiments according to a specially designed orthogonal array (OA) which can
significantly reduce the number of experiments. In Taguchi's OA, each combination
105
of factors has a balance, in which within a column of the array, each factor has equal
number of levels or appears at equal number of times. The unique characteristics of
GFRP composites affect their machinability differently to those of the traditional
homogenous materials. Physical properties of fibre reinforcements and the matrix
material, fibre orientation, types, matrix material and volume fraction greatly
influence the machinability of GFRP composites apart from processing parameters
which include cutting speed, feed rate and depth of cut, tool materials and geometries.
Such a large number of influencing factors inevitably add to the complexity of
experimental investigations. Hence, in this part of work, only machining or
processing parameters were considered for the parametric analysis of their significant
influence. The five important parameters, namely fibre orientation, , tool geometry,
, feed rate, f, spindle speed, N and depth of cut, d, that affect the resultant machining
force, Fm were chosen for this study. Three different levels, low (1), medium (2) and
high (3) of each parameter. Table 5.1, encompassing a typical range of machining
parameters employed in the industry, have been selected for the experimentation.
Justifications for selection of those parameters and the three levels setting were
twofold: (1) to incorporate all possible processing parameters which were not covered
in the literature, and (2) to investigate any nonlinear effects that they have on the
machinability output. The range of machining conditions was selected depending on
the importance of industrial applications, within the limit of the machine tool as well
as over the range of conditions employed in the reported literatures [Paulo Davim, J
and F.Mata (2005), Paulo Davim, J and F.Mata (2007), Wang, X.M and L.C. Zhang
(2003) and Palanikumar, K et al. (2006)].
Table 5.1 Process control parameters and their levels
Process parameters
Helix angle
Spindle speed
Feed rate
Depth of cut
Units
(deg)
rpm
mm/min
mm
o
Notation
N
f
d
1
25
2000
500
1
Levels
2
35
4000
750
1.5
3
45
6000
1000
2
Spindle speed has significant influence on the extent of tool wear as well as
surface roughness. Initially, the spindle speed was set at 2,000 rpm, but this resulted
106
in premature failure or chipping of the cutting tool edges. This could be, due to nature
of discontinuous cutting actions in milling as the tool encounter inhomogeneous
layers of GFRP composites. In addition, spindle speed of 2,000 rpm is deemed to be
low as far as machining productivity is concerned. On the other hand, higher spindle
speed (above 5,000 rpm) leads to rapid tool wear, which lasted the tool for only a
couple of endmilling passes to reach the predefined tool life criteria. Consequently,
3,0005,000 rpm was set as test range for the spindle speed herein. This range truly
represents the typical range of industrial applications. On the contrary, as indicated in
the previous studies [Paulo Davim, J and F. Mata (2005) and Paulo Davim, J and F.
Mata (2007)], the employed feed rate was reported to be within the range of 200800
mm/min when machining GFRP composites. Selection of feed rate range during
machining is critical because it determines the surface roughness of the machined
components. A value lower than those reported in the literature would diminish
machining productivity while a higher value accelerates heat generation, machining
forces and tool wear, hence, deteriorating the surface quality. As a result, the 500
1,000 mm/min range of feed rate was found to be appropriate for the current study. It
is important to highlight that, although the depth of cut plays a small role during
metal machining, its range was selected to be 12 mm, in accordance to a rough and
finish machining of the given material. In the traditional full factorial
experimentation, 27 trials would be needed to complete the entire experimental work
of three factors at three levels. However, based on the selected parameters and their
levels, current parametric study could be well performed using the L9 Taguchi OA in
which nine experimental runs would be required to complete the array. The Taguchi
experimental layout was arranged according to Table 5.2 5.4 with each trial
performed in a random order in a bid to minimise any chances of systematic error
during measurement of the machining force.
5.5 MEASUREMENT OF MACHINING FORCE

The output response considered in this study is machining force (Fm). The
measurement and calculation of response, based on the input parameters, are
described below.
107
Force measurement in manufacturing, especially in machining, is very

important. This is because:
Force measurement can be used for monitoring the tool conditions,

and avoiding breakage during the machining process.
Force measurement helps us to understand machining process, because

cutting force is one of the most sensitive indicators of machining
performance. Both the static and the dynamic components of the
cutting force contain information concerning the state of chip
formation and the cutting tool.
Force measurement enables engineer to optimise manufacturing

process and design proper machining tool.
The force measurement was carried out using a Kistler dynamometer. The
data acquisition was carried out by an appropriate software called Dynawarekistler.
The value of machining force in the work piece is determined using the Eq. 5.1.
(5.1)
Where Fm - Machining Force, Fx - Feed force, Fy - Cutting force and Fz - Thrust force.
5.6 RESULTS AND DISCUSSIONS
It is generally known that the cutting mechanism in GFRP composites is due
to the combination of plastic deformation, shearing, and bending rupture. The
occurrence of the above mechanisms depends on the flexibility, orientation, and
toughness of the fibres. These constitute a surface texture on the work piece. The
chips formed from machining of GFRP composites are discontinuous types in powder
form. During the machining of GFRP composites, the tool continuously encounters
alternating matrix and fibre materials, whose response to machining can vary greatly.
Normally, in machining of GFRP, the tool encounters a low temperature soft epoxy
matrix and brittle glass fibres. Due to the above facts, achieving a low machining
force with good surface finish is a tedious job.
108
The machinability in this work was evaluated by the machining force induced.
The results obtained through experiments are presented in Table 5.2 5.4.
Table 5.2 Experimental machining forces for 15 fibre orientation GFRP
composite
Experiment
number
1
2
3
4
5
6
7
8
9
Helix
angle
(deg)
25
25
25
35
35
35
45
45
45
Spindle
Feed
Speed
f
N
(rpm) (mm/rev)
3000
500
4000
750
5000
1000
3000
750
4000
1000
5000
500
3000
1000
4000
500
5000
750
Depth of
cut
d
(mm)
1
1.5
2
2
1
1.5
1.5
2
1
Machining Force
Fm (N)
15 fibre
orientation
22.4
33.7
36.6
27.5
29.3
25.2
22.1
17.5
21.2

composite
Experiment
number
1
2
3
4
5
6
7
8
9
Helix
angle
(deg)
25
25
25
35
35
35
45
45
45
Spindle
Speed
N
(rpm)
3000
4000
5000
3000
4000
5000
3000
4000
5000
Feed
f
(mm/rev)
500
750
1000
750
1000
500
1000
500
750
Depth of
cut
d
(mm)
1
1.5
2
2
1
1.5
1.5
2
1
Machining Force
Fm (N)
60 fibre
orientation
30.6
38.2
40.1
29.2
30.1
25.7
25.2
21.8
24.9
composite
Experiment
number
1
2
3
4
5
Helix
angle
(deg)
25
25
25
35
35
Spindle
Feed
Speed
f
N
(rpm) (mm/rev)
3000
500
4000
750
5000
1000
3000
750
4000
1000
109
Depth of
cut
d
(mm)
1
1.5
2
2
1
Machining Force
Fm (N)
105 fibre
orientation
32.4
35.1
44.1
31.5
33.3
6
7
8
9
35
45
45
45
5000
3000
4000
5000
500
1000
500
750
1.5
1.5
2
1
29.0
29.3
27.0
28.9
The evaluation of machining force in function of helix angle, spindle speed,

feed rate and depth of cut for different fibre orientation, can be seen in Figure 5.4 - 5.
6. It is observed that Fm increases with the feed rate and decreases with increase in
helix angle, for different spindle speeds, and the effect of spindle speed and depth of
cut are insignificant. In order to obtain a reduced machining force, the GFRP
composite with the fibre orientation 15 is essential.
Main Effects Plot (data means) for Machinig force 15 degree

Helix angle
Mean of Machinig force 15 degree
26
Spindle speed
24
22
20
25
35
45
3000
Feed rate
26
4000
5000
Depth of cut
24
22
20
500
750
1000
1.0
1.5
2.0
Figure 5.4 Effect of factors on machining force for 15 fibre orientation angle
110
Main Effects Plot (data means) for Machining Force- 60 degree

Helix angle
Spindle speed
Mean of Machining Force
35.0
32.5
30.0
27.5
25.0
25
35
Feed rate
45
3000
4000
Depth of cut
5000
500
750
1000
1.0
1.5
2.0
35.0
32.5
30.0
27.5
25.0
Main Effects Plot (data means) for Machining Force-105 Degree

Helix angle
Spindle speed
Mean of Machining Force
32
30
28
25
35
Feed rate
45
3000
4000
Depth of cut
5000
500
750
1000
1.0
1.5
2.0
32
30
28
111
5.6.1 Influence of helix angle on machining force

When machining of brittle material such as GFRP composites, fracture of
glass and epoxy matrix reduce tool/chip contact on the tool flank and rake faces
which tend to alleviate friction between tool and the workpiece material. This ends
up with lower machining forces as compared to homogenous and ductile material
such as metals.
The machining force (Fm) decreases with the increase of helix angle as shown
in Figure 5.4 - 5.6. When the helix angle increases from 25 to 45, the amount of
space between the work material & cutting tool rake surface increases, leading to
easy removal of chip from the parent material leading to lower machining force. Fibre
orientation & tool geometry are the main significant parameters that influence the
cutting force in FRP machining.
5.6.2 Influence of fibre orientation on machining force
The combination of shearing and bending rupture with little plastic
deformation is the main cutting mechanism of FRP composites resulting in the
absence of other forms of wear to exist. During end milling, chips were mostly
discontinuous type and in particular, for GFRP composites, the chips were mostly in
the form of dusts as a result of bending, buckling and brittle fractures of glass fibres
and epoxy matrix.
Machining force (Fm) increases with the increase of fibre orientation angle as
shown in Figure 5.7. For the fibre orientation above 90, the tool first bends the fibres
before cutting them and therefore expands a maximum force than other orientations.
Both cutting and thrust force have a minimum value for 15 fibre orientation and
maximum for 105 fibre orientation as illustrated in Figure 5.8- 5.10.
112
15 GFRP Plate
60 GFRP Plate
105 GFRP Plate
40
Machining Force
35
30
25
20
15
10
5
0
1
Experimental Run
Figure 5.7 Influence of fibre orientation on machining force
Figure 5.8 Influence of fibre orientation on machining force for 15

GFRP plate
113

GFRP plate

GFRP plate
At fibre orientation 15, the fibres are subjected to slight bending and tensile
loading which cause the fibres to break out of tension. Further there is little
subsurface delamination experiencing low machining force. But, at fibre orientation
105, the fibres are subjected to severe bending and compression loading, which
cause the fibres to break in compression shear and exert high machining force.
114
5.6.3 Influence of feed rate on machining force

Machining force increases with the increase of feed rate. This may be
attributed to the fact that feed rate determines the area of undeformed chip thickness.
The low feed rate produces a smaller chip per tooth and ends up with lower cutting
forces and less severe impact on the cutting edge. The increase in the feed force leads
to significant increases in cutting temperatures and undesirable thermal stresses in the
tool and the workpiece. This, in turn, increases the machining force.
5.6.4 Influence of depth of cut on machining force
Depth of cut is an equally dominant parameter that influences Fm during end
milling of GFRP composites apart from the feed rate. This may be attributed to the
fact that feed rate determines the area of undeformed chip thickness. The effect of
depth of cut is mainly due to the increase in the cross-sectional area as well as the
number of fibre layers to be cut. Hence, any changes in these parameters significantly
influence the amount of force experienced by the cutting tool during machining.
5.6.5 Influence of spindle speed on machining force
Machining force decreases with higher spindle speed. This could be justified
by the fact that increase in spindle speed generates friction, which elevates the cutting
zone temperatures. As a result of low thermal conductivities of glass fibres and epoxy
matrix, this would tend to soften the polymer matrix and hence, requiring less force to
shear the material. Nevertheless, the minor effect of spindle speed on the Fm, in terms
of decreasing machining force with higher spindle speed, contradicts the common
perception. Similar results have been reported by Wang et al. (2003) during
orthogonal cutting of graphite epoxy composites. Likewise, Lee (2001) reported a
reduction in machining forces with the increase of cutting speed during turning
operation of GFRP composites for the entire cutting tool materials tested. This could
be attributed to the increase in strain on the composite material with higher cutting
speed which accelerates brittle fracture or behaviour of the epoxy matrix and glass
fibres. ANOVA was carried out for the level of significance of 5% (the level of
confidence is 95%).
115
Table 5.5 shows the result of ANOVA with Fm during the machining of GFRP
composite plates.
Table 5.5 ANOVA for the machining force (Fm) for the three composite
materials
Source of
variance
Degree of
freedom
Sum of
squares
Variance
Test F
F=5%
Percentage
contribution
(P)
3.63
69.25
()
15 fibre orientation angle

85.803
42.902 10725.5
N (rpm)
f (mm/min)
2
2
0.123
6.811
0.061
3.406
15.25
851.5
3.63
3.63
2.72
19.72
d(mm)
0.092
0.046
11.5
3.63
1.25
Error
16
0.064
0.004
Total
24
92.893
7.06
100

()
113.989
N (rpm)
0.7341
f (mm/min)
d(mm)
56.995
1838.55
3.63
61.96
0.0867
2.7968
3.63
1.64
9.880
4.940
159.35
3.63
18.91
12.812
0.406
13.097
3.63
2.91
Error
16
0.498
0.0311
14.58
Total
24
100
()
137.913
128.28
64.14
1961.47
3.63
59.44
N (rpm)
0.8502
0.0912
2.7889
3.63
1.98
f (mm/min)
10.57
5.28
161.468
3.63
23.22
d(mm)
1.162
0.581
17.768
3.63
1.63
Error
16
0.523
0.0327
Total
24
141.385
13.73
100
116
From the analysis it was observed that the helix angle of the milling cutter and
feed rate factor have statistical and physical significance on the machining force
obtained, especially the helix angle of the cutter. The effect of the factors spindle
speed and depth of cut are practically insignificant. It is also clear that the helix angle
factor has a higher influence on the results for the fibre orientation 15(69.25%) than
60 and 105 fibre orientations (61.96% and 59.44%).
5.7 MULTIPLE REGRESSION ANALYSIS (MRA)
Correlations between machining parameters (helix angle, spindle speed, feed
rate and depth of cut) and the machinability output for three different fibre oriented
GFRP plates have been modelled using MRA. The parametric dependency of
machinability output (Fm) is considered in terms of a power series equation given by:
15 fibre orientation:
Fm = 29.1 0.532 + 0.000917 N + 95.4 f + 2.9 d
(R2=93.9%)
(5.2)
Fm = 40.6 0.615 + 0.000425 N + 75.0 f + 1.8 d
(R2=94.2%)
(5.3)
(R2=93.6%)
(5.4)
Fm = 34.7 0.440 + 0.000733 N + 76.3 f + 2.67 d
Where, be the helix angle in degree, N is the spindle speed in rpm, f is the
feed rate in mm/min, and d is the depth of cut in mm.
The coefficients of determination, R2, are higher for Fm models. Thus,
Equations 5.2 -5.4 can be effectively used to obtain reliable estimates of Fm during
end milling GFRP composites within the range of experimental parameters.
Nevertheless, judging from the coefficient values of each parameter of the derived
semi-empirical models, it is further confirmed that helix angle has the most
significant effect on each of the machinability outputs studied here followed by the
feed rate.
5.8 VALIDATION TESTS
A comparison of experimental data with the calculated results from MRA
equations within the reported range of experimental conditions shows average
117
variations within 4% for Fm. A very close agreement between the predicted values
and the experimental results can be observed indicating the accuracy of the semi
empirical models developed. Hence, Eqs. (5.2) - (5.4) are demonstrated as a feasible
and effective way for the evaluation of machining force of the GFRP composites with
different fibre orientation.
Validation experiments were additionally performed under randomly selected
machining conditions (some of which were outside the range of L9 array) to evaluate
further the accuracies of the developed semi-empirical models. Experimental
conditions for confirmation tests and the results are summarised in Table 5.6,
showing the absolute percentage errors of within 59% for Fm. Hence, it is
demonstrated that the developed parametric models show good predicting capability
of the machinability output within the range of parameters studied here.
Table 5.6 Validation test results with error percentage
Test
1
2
3
4
5
6
7
8
9
15 fibre orientation
Exp.
Cal. Error
value
value (%)
22.4
33.7
36.6
27.5
29.3
25.2
22.1
17.5
21.2
22.9
33.1
35.9
26.7
28.5
24.6
21.5
18.2
20.6
-2.2
-1.2
1.9
2.9
2.7
2.4
2.7
-3.4
2.8
Exp.
Cal. Error
value
value (%)
30.6
38.2
40.1
29.2
30.1
25.7
25.2
21.8
24.9
29.8
37.5
39.6
29.8
29.7
26.3
25.9
22.5
24.1
2.6
1.8
1.2
-2.1
1.3
-2.3
-2.8
-3.2
3.2

Exp.
Cal. Error
value
value (%)
32.4
35.1
44.1
31.5
33.3
29.0
29.3
27.0
28.9
31.9
35.9
43.1
30.7
32.3
28.2
28.1
27.9
28.2
1.5
-1.4
2.3
2.5
2.3
1.4
4.1
-1.6
2.8
5.9 SUMMARY
From the experiment it is inferred that the machining force increases with the
increase of fibre orientation, feed rate and depth of cut. It is also observed that the
machining force decreases with the increase of helix angle and spindle speed. Lower
machining force was found for 15 fibre oriented GFRP composite plates. It is
observed that contribution of helix angle (P=69.25%) has more influence on the
machining force (Fm) followed by the feed rate (P=19.72%).
118
CHAPTER 6
OPTIMISATION OF MACHINING PARAMETERS

USING TAGUCHI METHOD WITH FUZZY LOGIC
6.1 INTRODUCTION
Optimisation of process parameters is an important criterion in the machining

process to achieve high quality. Normally, the Taguchi method is used to optimise the
performance characteristics of process parameters to achieve high quality. However,
most reports on Taguchi applications to date have been centered with the optimisation
of a single performance characteristic. Handling the more demanding multiple
performance characteristics is still an interesting research problem. Optimisation of
multiple response characteristics is more complex compared to optimisation of single
performance characteristics. The theory of fuzzy logic, initiated by Zadeh, L (1965),
has proven to be useful for dealing with uncertain and vague information. This theory
has proved to be an effective means for dealing with objectives that are linguistically
specified. Linguistic terms, such as low, medium and high may be defined by
fuzzy sets. Since its introduction, fuzzy set theory has attracted the attention of
researchers in mathematical and engineering fields. This chapter discusses the
application of the Taguchi method with fuzzy logic to optimise the machining
parameters for machining of GFRP composites by converting complicated multiple
performance characteristics into a single Multi-Response Performance Index (MRPI).
The machining tests were performed on a CNC milling machine using solid
carbide (K10) end mill cutting tool with 3 different helix angles. Experiments were
planned using Taguchis orthogonal array with the cutting conditions prefixed. The
machining parameters, namely, helix angle of the end mill cutter, cutting speed, feed
rate, depth of cut, and work piece fibre orientation (specially applied to the GFRP
composites) were optimised with considerations of multiple response characteristics,
including machining force, surface roughness, and delamination.
6.2 MULTI-RESPONSE OPTIMISATION USING FUZZY LOGIC
In the present work, a multiple performance characteristics optimisation

method is introduced for a composite machining process. Optimisation of the
119
multiple performance characteristics is concerned with optimising a vector of

objectives. For the composite machining process, machining force, surface roughness
and delamination have lower-the-better performance characteristics. Hence,
optimisation of such multiple performance characteristics is much more complicated
than optimisation of a single performance characteristic. In this work, the Taguchi
method with fuzzy logic is used to investigate the multiple performance
characteristics in the composite machining process. The optimisation procedure
adopted is as follows:
(i) Conduct the experiments using Taguchis orthogonal array.
(ii) Transform the experimental results into a signal- to - noise (S/N) ratio.
The S/N ratio can be used to measure the deviation of the performance
characteristics from the desired values.
(iii) Development of fuzzy rules. The loss function corresponding to each
performance characteristics is fuzzified, and then a single MRPI is obtained
through fuzzy reasoning on the fuzzy rules. The MRPI is used to optimise the
machining process.
(iv) Analyse the experimental results using the MRPI and statistical analysis
of variance.
(v) Select the optimal levels of process parameters.
(vi) Verify the optimal parameters through experiment.
6.3 EXPERIMENTAL DESIGN
The experiments for this work were planned using Taguchis Design of
Experiments (DOE). Taguchis approach to parameter design provides the design
engineer with a systematic and efficient method for determining near optimum design
parameters for performance and cost. This method can dramatically reduce the
number of experiments required to gather necessary data.
120
6.3.1 Experimental parameters

The essential step of Taguchi method was to identify the important parameters
which affect the process. From the literature and the previous work done in this field
the independently controllable predominant machining parameters which have greater
influences on the machining of GFRP composites, were identified. They are: (1) helix
angle of the cutter (tool geometry) (A), (2) cutting speed (B), (3) feed rate (C), (4)
depth of cut (D), and (5) work piece (fibre orientation) (E), out of which fibre
orientation angle has been specially applied to fibre reinforced composites. The
output responses used to measure the machinability are machining force, surface
roughness and delamination. The experiments were conducted according to Taguchis
three-level design. The parameters selected, the designated symbols, and their ranges
are given in Table 6.1.
Table 6.1 Machining parameters and their levels
Levels
Control parameters
Units
Symbol
Level
1
Level
2
Level
3
(Cutting tool geometry) Helix

angle
(deg)
25
35
45
Spindle Speed
(rpm)
3000
4000
5000
Feed rate
(mm/min)
500
750
1000
Depth of cut
(mm)
1.0
1.5
2.0
(Work piece) Fibre Orientation
(deg)
15
60
105
6.3.2 Planning the experiments

The experiments were planned according to Taguchis orthogonal array. For
conducting the experiments, a L27 orthogonal array was chosen. In the L27 orthogonal
array, there are 13 columns that can be used to assign test factors and their
interactions. To check the Degrees of Freedom (DOF) in the experimental design, for
the three-level test, the five main factors take 10 DOFs (5 2) and the remaining
DOFs are taken by interactions. The three-level L27 orthogonal array is shown in
121
Table 6.2, where the numbers 1, 2, and 3 stand for the levels of the factors. This array
specifies 27 experimental runs and has 13 columns. To avoid aliasing and overlap of
the interactions with the main factors, the factors were assigned to the L27 columns, as
in Table 6.2. The columns chosen for the main factors were 1, 2, 5, 12, and 13.
Table 6.2 Layout of L27 orthogonal array
Column numbers
Trail No.
10
11
12
13
10
11
12
13
14
15
16
17
18
122
19
20
21
22
23
24
25
26
27
6.3.3 Conducting the experiments

The performance of glass fibre reinforced polymer composite in endmilling
was studied by conducting various machinability tests. Work materials having
different fibre orientations were used for the experiments. The cutting tool insert was
a solid carbide (K10) end mill with 3 different helix angles. In the present study, the
machining performance was studied through the responses viz., machining force,
surface roughness and delamination.
6.4. MEASUREMENTS
The force measurement was carried out using a Kistler dynamometer. The
data acquisition was carried out by appropriate software called Dynawarekistler. The
value of machining force in the work piece is determined using the Eq. 6.1
(6.1)
The surface roughness (Ra) was evaluated using stylus type profilometer
Mitutoyo SJ-201. For each test five measurements were made over milling surfaces.
By considering the number of measurements to be carried out, a programmable
technique was used, by previously selecting a roughness profile, the cut-off (0.8 mm)
and the roughness evaluator parameter (Ra) specified by ISO.
123
The Delamination (Fd) was measured using the formula mentioned in Eq. 6.2.
.............................................. (6.2)
Where, W is the nominal slot width, Wmax is the maximum damaged slot
width. Maximum damaged widths of the scanned slots were precisely measured in a
CAD environment and the damage factor for each slot was calculated.
6.5 TRANSFORMATION OF EXPERIMENTAL RESULTS INTO S/N RATIO
In the Taguchi method, the S/N ratio is used. The objective of using the S/N
ratio as a performance measurement is to develop products and processes insensitive
to noise factors. The S/N ratio indicates the degree of predictable performance of a
product or process in the presence of noise factors. Process parameter settings with
the highest S/N ratio always yield optimal quality with minimum variance.
Taguchi categorised the performance characteristics of a system into three
different kinds based on the type of performance: the nominal the best, the smaller the
better, and the larger the better. In this study, smaller the better principle was
considered to minimise the machining force, delamination and surface roughness.
The corresponding loss function can be expressed as follows:
For smaller the better
1
2
=1
............................................................ (6.3)
In Equation 6.3, n represents the number of repeated experiments, Lij is the

loss function of the ith performance characteristic in the jth experiment, and yijk is the
experimental value of the ith performance characteristic in the jth experiment at the kth
test. The loss function is further transformed into an S/N ratio. The S/N ratio for
the ith performance characteristic in the jth experiment can be calculated from the
following expression:
= 10 log ..................................................................... (6.4)
High signal-to-noise ratios are always preferred in a Taguchi experiment.
Table 6.3 shows the experimental results for machining force, surface roughness and
delamination and their S/N ratios based on the experimental parameter combinations.
To consider the three different performance characteristics in the Taguchi method, the
124
S/N ratios corresponding to the machining force, surface roughness and delamination
were processed by the fuzzy logic unit. To convert the different performance into a
single performance unit, fuzzy logic has been used and is discussed in the following
section.
Table 6.3 Experimental results with S/N ratio
Trail
Fm
S/N ratio
Fd
S/N ratio
No.
(N)
25.4
-28.0543
1.51
-3.5795
1.0321
-0.3728
31.5
-29.9662
1.69
-4.5577
1.0413
-0.3913
34.2
-30.6805
1.80
-5.1055
1.0497
-0.4213
34.9
-30.8565
1.72
-4.7106
1.0468
-0.3973
33.7
-30.5526
1.59
-4.0279
1.0342
-0.3854
38.4
-31.6866
1.73
-4.7609
1.0479
-0.4064
39.8
-31.9977
1.62
-4.1903
1.0432
-0.3882
42.3
-32.5268
1.85
-5.3424
1.0491
-0.4163
36.4
-31.2220
1.58
-3.9731
1.0423
-0.3946
10
30.9
-29.7992
2.07
-6.3194
1.0623
-0.5249
11
27.5
-28.7867
1.67
-4.4543
1.0463
-0.3931
12
29.9
-29.5134
1.84
-5.2964
1.0518
-0.4387
13
28.3
-29.0357
1.81
-5.1536
1.0509
-0.4312
14
33.1
-30.3966
2.21
-6.8879
1.0678
-0.5249
15
29.6
-29.4258
1.81
-5.1536
1.0564
-0.4766
16
24.3
-27.7121
1.42
-3.3876
1.0422
-0.3938
17
28.1
-28.9741
1.80
-5.1055
1.0625
-0.5266
18
29.9
-29.5134
1.99
-5.9771
1.0699
-0.5869
19
25.8
-28.2324
1.91
-5.6207
1.0672
-0.5649
S/N ratio
Ra
(m)
125
20
28.9
-29.2180
2.21
-6.8879
1.0901
-0.7493
21
26.1
-28.3328
2.05
-6.2351
1.0825
-0.6886
22
25.3
-28.0624
1.69
-4.5577
1.0689
-0.5787
23
26.1
-28.3328
1.96
-5.8451
1.0827
-0.6902
24
29.1
-29.2779
2.12
-6.5287
1.0902
-0.7509
25
27.9
-28.9121
2.01
-6.0639
1.0898
-0.7469
26
25.3
-28.0624
1.69
-4.5577
1.0701
-0.5885
27
26.7
-28.5302
1.86
-5.3903
1.0861
-0.7174
6.6 FUZZY LOGIC UNIT

Fuzzy logic has rapidly become one of the most successful technologies for
developing sophisticated systems. With its aid, complex requirements may be
implemented in amazingly simple, easily maintained, and inexpensive ways. Fuzzy
logic technology provides decision support and expert systems with powerful
reasoning capabilities bound by a minimum of rules. A fuzzy logic unit consists of a
fuzzifier, membership functions, a fuzzy rule base, an inference engine, and a
defuzzifier. The fuzzifier uses membership functions to fuzzify the S/N ratios. The
inference engine is used for fuzzy reasoning on fuzzy rules to generate a fuzzy value.
Finally, the defuzzifier converts the fuzzy value into a multi-response performance
index (MRPI).
The concept of fuzzy reasoning for three-input-one-output fuzzy logic unit is
described as follows. The fuzzy rule base consists of a group of IF- THEN statements
with three inputs, x1, x2, and x3 and one output y, i.e.,
Rule 1: if x1 is A1 and x2 is B1 and x3 is C1 then y is D1 else
Rule 2: if x1 is A2 and x2 is B2 and x3 is C2 then y is D2 else
...............................................................................................
...............................................................................................
Rule n: if x1 is An and x2 is Bn and x3 is Cn then y is Dn.
Ai, Bi, Ci, and Di are fuzzy subsets defined by the corresponding membership
functions. In the present work, three fuzzy subsets are assigned to the three inputs, as
126
shown in Figure 6.1-6.3. Five fuzzy subsets are assigned to the output, as shown in
Figure 6.4. Twenty-seven fuzzy rules were developed based on the fact that a higher
S/N ratio gives better performance. By taking the max-min compositional operation,
the fuzzy reasoning of these rules yields a fuzzy output. Suppose that x1, x2, and x3 are
the three input values of the fuzzy logic unit then the membership function of the
output of fuzzy reasoning can be expressed as in Equation 6.5.
0
=
1 1 1 2 1 3 1 .
1 2 3 ]
..................................................................... (6.5)
Where is the minimum operation and is the maximum operation. Finally,
a defuzzification method called centre of gravity is used to transform the fuzzy output
into a non-fuzzy value y0,
0 =
0
0
..................................................................................... (6.6)
Figure 6.1 Membership functions for machining force
127
Figure 6.2 Membership functions for surface roughness
Figure 6.3 Membership functions for delamination factor
128
Figure 6.4 Membership functions for MRPI

The non-fuzzy value y0 gives MRPI. Invariably, a larger MRPI is preferred,
which gives a better performance characteristic. Table 6.4 shows the results of the
MRPI for different experiments.
Table 6.4 MRPI Values
Trail No.
MRPI
0.43
0.43
0.25
0.56
0.40
0.41
0.34
0.33
0.28
10
0.26
11
0.64
129
6.7
12
0.25
13
0.54
14
0.39
15
0.68
16
0.69
17
0.37
18
0.43
19
0.32
20
0.09
21
0.26
22
0.51
23
0.38
24
0.26
25
0.25
26
0.43
27
0.25
ANALYSIS OF EXPERIMENTAL RESULTS USING MRPI AND
ANOVA
The experimental scheme used in this work is based on Taguchis orthogonal
array, by which it is possible to separate the effect of each machining parameter on
the MRPI at different levels. For example, the mean of the MRPI for work piece
(fibre orientation) at levels 1 and 2 can be calculated by averaging MRPI values for
the experiments 1 to 9 and 10 to 18, respectively. The mean of the MRPI for each
level of other machining parameters can be calculated in a similar manner. The mean
MRPI at each level for different machining parameters are presented in Table 6.5,
which is referred to as a response table. In addition, the total mean of the MRPI was
also calculated and is given in Table 6.5.
130
The influence of each machining parameter can be more clearly presented by

means of the MRPI response graph. The MRPI graph shows the change in the
response when a given factor goes from level 1 to level 3. The response graph for the
machining parameters of the composite machining process is presented in Figure 6.5.
Table 6.5 Response table for MRPI
MRPI
Control parameters
Units
Symbol
Level
1
Level
2
Level
3
(Cutting tool geometry) Helix

angle
(deg)
0.3801
0.4822
0.3056
Spindle Speed
(rpm)
0.3244
0.4589
0.3744
Feed rate
(mm/min)
0.4633
0.3844
0.3411
Depth of cut
(mm)
0.3189
0.3844
0.3956
(Work piece) Fibre Orientation
(deg)
0.4801
0.3656
0.3278
Based on the response graph and response table, the optimal machining
parameters for the GFRP machining process can be achieved. Basically, larger the
MRPI, better is the multiple performance characteristics. It was found from
experimental results that the settings for experiment number 16 had the highest
MRPI, as seen in Table 6.4. Therefore among 27 experiments, experiment 16
machining parameter settings are optimal for attaining multiple performances
simultaneously. However, the relative importance among the machining parameters
for the multiple performance characteristics still needs to be analysed so that the
optimal combinations of the machining parameter levels can be determined more
accurately. The relative importance among the factors can be analysed through an
Analysis of Variance (ANOVA).
131
0.50
0.45
0.40
0.35
0.30
25
35
Helix angle
45
50
75
Cutting speed
0.05
0.15
Depth of cut
0.25
15
60
Fibre orientation
100
0.04
0.08
Feed rate
0.12
0.50
0.45
0.40
0.35
0.30
105
Figure 6.5 MRPI response graph

Table 6.6 Results of ANOVA
Machining Parameters
Degree of
freedom
Sum of
square
Mean
square
Fvalue
(Cutting tool geometry)

Helix angle
0.1753
0.0877
10.95
Spindle Speed
0.0702
0.0351
4.71
Feed rate
0.0912
0.0456
6.84
Depth of cut
0.0397
0.0199
2.85
(Work piece) Fibre

Orientation
0.1579
0.0790
10.03
Error
16
0.1078
0.0067
Total
26
0.6421
ANOVA is used to analyse which machining parameters significantly affect

the performance characteristics. This is accomplished by separating the total
variability of the MRPI, which is measured by the sum of the squared deviations from
the total mean of the MRPI, into contributions by each machining parameter and the
132
error. In addition, the Fishers F test can also be used to determine which machining
parameters have significant effect on the performance characteristic. Usually, a
change of a machining parameter has a significant effect on a performance
characteristic when F is large. Based on the results of analysis of variance (Table
6.6), it was determined that work piece (fibre orientation) and helix angle of the
endmill cutter were the most significant machining parameters affecting multiple
performance characteristics. Referring to the average response table and average
response graph, the variable settings for optimal machining parameters are the helix
angle at level 2, cutting speed at level 2, feed rate at level 1, depth of cut at level 3,
and work piece (fibre orientation) at level 1.
6.8 CONFIRMATION TEST

The final step of the optimisation process was to predict and verify the
improvement in the performance characteristic for machining of GFRP composites by
an endmilling process with respect to the chosen initial parameter setting. The
estimated MRPI, using the optimal level of the machining parameters, can be
calculated from following equation.
= +
(6.7)
=1
Where Mm is the total mean of the MRPI, Mo is the mean MRPI at optimal
level, and n is the number of main design parameters that affect the multiple
performance characteristics.
Table 6.7 shows the comparisons of predicted and actual machining
performance for multiple performance characteristics using their optimal machining
parameters.
133
Table 6.7 Results of machining performance using the initial and

optimal machining parameters
Initial
Optimal
machining Prediction
machining
parameters
parameters
Experimental
Setting Level
A1B1C1D1E2
A2B2C1D3E1
Machining Force, N
17.19
--
15.92
Surface roughness, m
0.6572
--
0.6107
Delamination Factor
1.0063
--
1.0045
MRPI
0.43
0.7547
0.7419
Improvement in MRPI = 0.3119

Based on the confirmation experiment results, the final optimal setting for
parameters arrived at are tool geometry (helix angle) at level 2, speed at level 2, feed
rate at level 1, depth of cut at level 3, and work piece (fibre orientation) at level 1.
6.9 DISCUSSION
It is known that the cutting mechanism in GFRP composites is due to the
combination of plastic deformation, shearing, and bending rupture. The occurrence of
the above mechanisms depends on the flexibility, orientation, and toughness of the
fibres. These constitute a surface texture on the work piece. The chips formed from
machining of GFRP composites are discontinuous types in powder form. During the
machining of GFRP composites, the tool continuously encounters alternating matrix
and fibre materials, whose response to machining can vary greatly. Normally, in
machining of GFRP, the tool encounters a low temperature soft epoxy matrix and
134
brittle glass fibres. Due to the above facts, achieving a low machining force with
good surface finish and low delamination becomes tedious job. If different cutting
parameters are increased proportionately, the surface roughness increases, but the
machining forces decreases and delamination increases. In this work, the Taguchi
method with fuzzy logic has been applied to achieve a better machining force with a
good surface finish and a delamination factor.
The Taguchi method with fuzzy logic used in this work was found to be
effective in the optimisation of machining parameters for the machining of glass fibre
reinforced polymer composites. The confirmation experiment results at optimal levels
(A2, B2, C1, D3, and E1) show that the machining force decreased from 17.19 N to
15.92 N. The Delamination factor reduced from 1.0063 to 1.0045, and the surface
finish improved from 0.6572 m to 0.6107 m. This approach can be easily
employed to optimise machining parameters with a wider range. This approach is
more flexible and superior to other techniques. Normally, in a single performance
characteristics study, spindle speed and feed rate are the factors that affect the
machining process. However, in a multiple performance characteristics study, work
piece (fibre orientation) and tool geometry (helix angle) are the parameters that show
more influence than other machining parameters. The results indicated that the
optimal performance could be achieved at a 60 fibre orientation angle and at a
middle speed of 75 m/min. The results also indicated that the multiple performance
optimisation can be achieved at a low feed rate and a low fibre orientation, while the
depth of cut should be fixed at high value. From the results, it can be asserted that the
optimisation methodology used in this study is useful in improving multiple
performance characteristics in the composite machining process.
6.10 SUMMARY
The Taguchi method with fuzzy logic was employed to optimise the multiple
performance characteristics of a machining process. The work material used was
GFRP composites with 3 different fibre orientations and the tool material used was
solid carbide end mill with different helix angles. The experimental results for
optimal settings showed that there was a considerable improvement in the
performance
characteristics
viz.,
machining force,
135
surface
roughness
and
delamination factor. Work piece (fibre orientation) and tool geometry (helix angle)
were the parameters that showed higher influence on endmilling of GFRP
composites. This technique is more convenient and economical to predict the optimal
machining parameters. The Taguchi method with fuzzy logic technique using MRPI
converts the multiple performance characteristics into single performance
characteristics and, therefore, simplifies the optimisation procedure.
136
CHAPTER 7
CONCLUSIONS
GFRP composite plates with five different fibre orientations were prepared by
means of hand lay-up process. Tensile strength, Youngs modulus and shear strength
test were conducted as per the ASTM standards. Initial machinability studies (surface
roughness and delamination study) were conducted on the GFRP samples using solid
carbide endmill coated with PCD by means of RSM and from the experimental
results, the following conclusions are drawn:
An artificial neural network and a response surface model were developed to

predict surface roughness and delamination on the milled surface. In the
development of predictive models, cutting parameters consisting of cutting
speed, fibre orientation, depth of cut and feed rate were considered as model
variables. Good agreement was observed between the predicted optimum and
the experimental measurements.
Based upon the correlation coefficient and convergence, 28 different back

propagation neural network architectures were trained/analysed using the
experimental data until an optimum architecture was identified. Out of the
different multi-layer back propagation neural network architectures trained,
the back propagation neural network architectures with two hidden layers
having 30 neurons in each (430302) trained with Levenberg
Marquardt algorithm was found to be the optimum network model. A sound
performance was achieved with the neural network model, and it showed good
correlation between the predicted values of the optimum neural network
model and the experimental data for prediction of surface roughness and
delamination using the entire datasets within acceptable error limits.
The feed rate was the cutting parameter which has greater influence on
surface roughness (33.43%) and delamination (35.49%) for GFRP
composite materials followed by the fibre orientation.
137
The Surface roughness and delamination increases with the increase of

feed rate and increase of fibre orientation angle, which means that the
composite damage is larger for higher feed rate and for higher fibre
orientation.
Milling experiments were conducted on 15, 60 and 105 GFRP plates based on
Taguchi L9 orthogonal array to evaluate the influence of helix angle, spindle speed,
feed rate and depth of cut on machining force.
Machining force increases with the increase of fibre orientation, feed rate and
depth of cut. Machining force decreases with the increase of helix angle and
spindle speed. GFRP composites with fibre orientation 15 gives smaller
value of Fm than the 60 and 105 fibre oriented GFRP composites.
Helix angle of the end mill cutter is the cutting parameter that has the highest
physical as well statistical influence on Fm followed by the feed rate.
The Taguchi method with fuzzy logic was employed to optimise the multiple
performance characteristics of the GFRP endmilling process. Milling experiments
were conducted on GFRP plates to evaluate the influence of helix angle, spindle
speed, feed rate, depth of cut and fibre orientation on machining force, surface
roughness and delamination factor based on L27 orthogonal array. The experimental
results for optimal settings showed that there was a considerable improvement in the
performance
characteristics
viz.,
machining force,
surface
roughness
and
delamination factor.
Work piece (fibre orientation) and Tool geometry (helix angle) were the
parameters that showed higher influence on endmilling of GFRP composites.
The Taguchi method with fuzzy logic technique using MRPI is more
convenient and economical to predict the optimal machining parameters. This
technique converts the multiple performance characteristics into single
138
performance characteristics and, therefore, simplifies the optimisation

procedure.
7.1 SCOPE FOR FUTURE WORK
Composite laminates can also be fabricated by using different

techniques, other than the hand layup technique.
In future work, experimental investigations on compression strength of

the GFRP specimens can be performed.
Milling of GFRP using different tool geometries with different

machining conditions (fibre orientations speed and feed etc.,) can be
initiated.
Delamination characterisation of GFRP laminates can be examined

using Finite element method with ANSYS or ABAQUS software
packages.
Milling of GFRP using CBN (Cubic Boron Nitride) tools can be

investigated with high spindle speed and feed rate.
The above mentioned study can be extended to relate the strength and
defects of the GFRP composites.
139
REFERENCES
Abdul D. Budan and Vijayarangan S (2002) Quality assessment and delamination

force evaluation in drilling glass fibre reinforced plastic laminates - a finite element
analysis and LEFM approach. Proc. of Int. Mech., 216, 173-182.
Abrate, S. Machining of composites, Composites engineering hand book, Ed. by
Mallick P.K, Marcel Deckker Inc., New York, 1997.
Alauddin, M., I.A.Choundhury E.L.Baradie and M.S.J.Hashmi (1995) Plastics
and their machining: A review. Journal of Materials Processing Technology, 54, 40
46.
Almeda, E.A., F.J.Oliveira, M.Sousa, A.J.S. Femandes., J.Sacramento and
R.F.Sitva (2005) Machining hard metal with CVD diamond direct coated ceramic
tools; effect of tool edge geometry. Diamond & Related Materials, 14, 651-656.
An, S.O., E.S.Lee and S.L.Noh (1997) A study on the cutting characteristics of glass
fibre reinforced plastics with respect to tool materials and geometries. Journal of
Materials Processing Technology, 68, 60-67.
Aoyame Inoue, Hirogaki Nobe, Kithara and C. Katayama (1995) Investigation on
damage in small diameter drilling in GFPR. Conference on Composites, Recent
advances in Japan and the United States, 18, 677-682.
Basavarajappa, S., G.Chandramohan and J.Paulo Davim (2007) Application of
Taguchi techniques to study dry sliding wear behaviour of metal matrix composites.
Materials and Design, 28, 1393-98.
Bhattacharyya, D. and P.W.Horrigan (1998) A Study of hole drilling on Kevlar
composites. Composite Science and Technology, 58, 267-283.
Brinksmeier, E. and R.Janssen (2002) Drilling of multi-layer composite materials
consisting of carbon fibre reinforced plastics (CFRP) Titanium and aluminum alloys.
Annals of the CIRP, 51, 87-90.
140
Caprino, G. and V.Tagliaferri (1995) Damage development in drilling GFRPs'. Int.

j. Machine Tool and Mfr., 35, 817-829
Chandrasekharan, V., S.G.Kapoor and R.E.Devor (1995) A mechanistic approach
to predicting the cutting forces in drill with application to fibre reinforced composite
materials. J. Engg. Industry, 17, 559-570.
Chen, W.C. (1997) Some experimental investigation in the drilling of carbon fibre
reinforced plastic (CFRP) composite laminates. International Journal of Machine
Tools and Manufacture, 37, 1097-1108.
Cheng, P.J. and S.C Lin (2010) Using neural networks to predict bending angle of
sheet metal formed by laser. Int J Mach Tools Manuf, 40, 11851197.
Chih-Wei Chang, Chun-Pao Kuo (2007) Evaluation of surface roughness in laserassisted machining of aluminum oxide ceramics with Taguchi method. International
Journal of Machine Tools & Manufacture, 47, 141-147.
Chung-Shin Chang (2006) Turning of glass-fibre reinforced plastics materials with
chamfered main cutting edge carbide tools. Journal of Materials Processing
Technology, 172, 117-129.
Conceicao Antonio, C.A and J. Paulo Davim (2002) Optimal cutting conditions in
turning of particulate metal matrix composites based on experiment and a genetic
search model. Composites Part-A, 33, 213-219.
Davim, J.P and Pedro Reis (2003) Study of delamination in drilling carbon fibre
reinforced plastics (CFPR) using design experiments. Composite Structures, 59, 481487.
Davim, J.P., Pedro Reis and Conceicao Antonia (2004) Experimental study of
drilling glass fibre reinforced (GFRP) manufactured by hand lay-up. Composite
Science and Technology, 64, 289-297.
Demuth, H and M. Beale Neural Network Toolbox Users Guide, Version 4
(Release 12), The Mathworks, Inc., 2000
141
Derringer, G., and R.Suich (1980) Simultaneous optimization of several response

variables. J Quality Technology, 12, 214219.
Dharan, C.K.H. and M.S.Won (2000) Machining parameters for an intelligent
machining system for composite laminates. Int. J. Mac. Tools Mfr, 40, 415-426.
Dillio, A., V.Tagliaferri and F.Venially (1991) Cutting mechanism in drilling
Aramid composite Int. JI. Of Machine Tool and Mfr, 31, 155-165.
Dillio, A., V.Tagliaferri and Crivelli Visconti (1987) Laser cutting of aramid FRP
material', Proceedings of LAMP, 87, 291-296.
Dipaola, G., S.G.Kapoor and R.E.Devor (1996) An Experimental investigation Jr
the crack growth phenomenon for drilling of fibre reinforced composite material. J1.
Of Engineering for Industry, 118, 104-11.
Doran, H. and C.R.Maikish Machining boron composites Composite materials in
engineering design Ed. by B.R. Noi1on, American Society of Metals, 1972.
Douglas C. Montgomery Design and analysis of experiments, 5th edition, John
Wiley and Sons, 2004.
Durao, L.M.N., A.G.Magalhaes, A.T.Marques A.M.Baptista and M.Figueired
(2008) Drilling of fibre reinforced plastic laminates. Materials Science Forum, 58,
706-710.
Edoardo Capello and Vicenza Tagliaferri (2001) Drilling damage of GFRP and
residual mechanical behaviour-Part II: Static and cyclic bearing loads. J. Composites
Technology and Research, 14, 131-137.
Erik Persson, Ingvar Eriksson and Leif Zackrisson (1997) Effects of hole
machining defects on strength and fatigue life of composite laminates. Composites
Part A, 28, 141-151.
142
Eriksen, E. (1999) Influence from production parameters on the surface roughness of

a machined short fibre reinforced thermoplastic. Int. J. Machine Tools and Mfr. 39,
pp. 1611-18.
Everstine, G.C. and T.G.Rogers (1971) A theory of machining of fibre reinforced
materials. Int.J. Composite Materials, 15, 94-106.
Ferreira, J. R., N.L.Coppini and G.W.A. Miranda (1999) Machining optimisation
in carbon fibre reinforced composite materials. Journal of Material Processing
Fiju Antony and Frenie Fiju Antony (2001) Teaching the Taguchi method to
industrial engineers. Work Study, 50, 141-149.
Franci Cus and Joze Balic (2003) Optimization of cutting process by GA approach.
Robotics and Computer Integrated Manufacturing, 19, 113-121.
Frederic Lachand, Robert Piquet, Francis collombet and Laurent Surcin (2001).
Drilling of composite structures. Composite Structures, 52, 511-516.
Furness, R.J., Wu C.L. and A.G.Ulsoy (1996) Statistical analysis of the effects of
feed, speed and wear on hole quality in drilling. Trans. of ASME, J. Mfg. Science and
Engineering, 118, 367-375.
Garlasco, F.M (1969) Machining of Boron Fibre Composites. ASME TP EM, 69-74.
Gindy, N.N (1998), Selection of drilling conditions for reinforced plastics, Int 1.
Production Research, 11, 1317-27.
Geon Woo Kim and Kang Yong Lee (2005) Critical thrust force at propagation of
delamination zone due to drilling of FRP metallic strips. Composite Structures, 69,
137-141.
George, P.M., B.K.Raghunath, L.M.Manocha and A.M.Warrier (2004)
Modelling of machinability parameters of carbon-carbon composite-a response
surface approach. J Mater Process Technology, 12, 920924.
143
Godella, A.M. and Y.M. Cheng (1993) Recent developments in electric discharge
machining. Proc, of symposium on machining of advanced composites. Edited by
Ramulu M and Kornanduri R, ASTM, 187-205.
Godfrey, C. Onwubolu and Shivendra Kumar (2006) Response surface
methodology-based approach to CNC drilling operations. Journal of Materials
Processing Technology, 171, 4147.
Hashish, M. (1989) Machining of advanced composites with abrasive water jets.
Manufacturing Review, 2, 142-150.
Hocheng, H. and C.K.H.Dharan (1990) Delamination during drilling in composite
laminates. Trans. ASME, 112, 236-239.
Hocheng, H. and H.Y. Puw (1993) Machinability of fibre reinforced thermo Plastics
in drilling. Trans. ASME, 114, 146-149.
Hocheng, H., H.Y.Pwu and K.C Yao (1993) On drilling characteristics of fibre
reinforced thermoset and thermoplastics. Materials and Manufacturing Processes, 8,
653682.
Houjiang Zhang, Wurji chew, Dingchang Chen and Liangchi Zhang (2001)
Assessment of the exit defects ill carbon fibre-reinforced plastic plates caused by
drilling Key Engineering Materials, 196, 43- 52.
Hull, D. and T.W.Clyne. An introduction to composite materials. 2nd ed. Cambridge
University Press, 1996.
Hung, N.P and C.H.Zhang (1996) Cumulative tool wear in machining metal matrix
composites Part I: Modeling Journal of Materials Processing Technology, 58, 109113.
Hsu, C.M and D.illiao (2004) Simultaneous optimization of the broadband tap
coupler optical performance based on neural networks and exponential desirability
functions. International Journal of Advanced Manufacturing Technology, 23, 896902.
144
Hussain, S.A., A.Pandurangadu and K.Palanikumar (2010) Surface roughness

analysis in machining of GFRP composites by carbide tools (K20).
European
Journal of Scientific Research, 41, 84-98.

Jack Feng, C.X. and X.Wang (2002) Development of empirical models for surface
roughness prediction in finish turning. Int J Advanced Manufacturing Technology, 20,
348356.
Jain, S. and N.C.H Yang (1994) Delamination free drilling of composite laminates.
Transaction of ASME, J. Engg. Industry. 116, 475-461.
Jain, V.K., Sreenivas Rao, S.K. Choudry and K.P.Rajurkar (1993) Experimen.al
Investigation into traveling wire electrochemical spark machining of composites. J.
Engg. Industry, 113, 75-84.
Jain, V.K., Tandon and Kumar (1990) Experimental Investigation into
electrochemical spark machining of composites. Engg. Industry, 112, 194-197.
Jamal Y. Sheikh-Ahmad Machining of Polymer composites. Springer Science and
Business Media LLC, 2009.
Jeyapaul, R., P.Shahabudeen and K.Krishnaiah (2006) Simultaneous optimization
of multi-response problems in the Taguchi method using genetic algorithm,
International Journal of Advanced Manufacturing Technology, 30, pp.870-878.
Jinn Tsonga Tsai, Tung Kuna Liu and Jim Hong Chou (2004) Hybrid TaguchiGenetic Algorithm for Global Numerical Optimization, IEEE transactions on
Evolutionary Computation, 12, 365-377.
Kansal, H.K., S.Singh, and P.Kumar (2005) Parametric optimization of powder
mixed electrical discharge machining by response surface methodology. J Mater
Process Technology, 169, 427436.
Karnik, S.R., V.N.Gaitonde, and J.P. Davim (2007) A comparative study of the
ANN and RSM modeling approaches predicting Burr size in drilling. The Int J of Adv
Manuf Technol, 38, 868-883.
145
Khashaba, U.A (2004) Delamination in drilling GFR thermoset composite

structures. Composite Structures, 63, 3l3-327.
Khuri, A.I. (1996) Multi response surface methodology in Ghosh, A., Rao, C.R.
(ed.), Handbook of statistics: design and analysis of experiment, 13, 377406.
Kim,
D.
and
M.Ramulu
(2004)
Drilling
process
optimization
for
graphite/bismaleimide-titanium alloy steels. Composite Structures, 63, 101-114.

Ko, Y., K.Kim and C. Jun (2005) A new loss function-based method for
multiresponse optimization. J Quality Technology, 37, 5059.
Kobayashi, A. Machining of plastics. McGraw-Hill, New York, 1967.
Komanduri, R. (1993) Machining of fibre-reinforced composites, Mechanical
Engineering, 115, 58-66.
Konig, W. and P.Grab (l989) Quality definition and assessment in drilling fibre
reinforced thermo sets, Annals of CIRP, 38, 119-123.
Konig, W. WulfCh., P.Grab and H.Willercheid (1985) Machining of fibre
reinforced plastics, Annals of the CIRP, 38, 119-124.
Konig, W., L.Cronjager, G. Spur and H.K.Tonshoff (1990) Machining of new
materials, Annals of the CIRP, 39, 673-680.
Koplev, A., A.Lystrup and T. Vorm (1983) The cutting process, chips and cutting
forces in machining CFRP. Composites, 14, 371 376.
Koricon, J. (1987) Water jet and abrasive water jet cutting, ASM Engineering
Materials Hand Book, 1, 673-675.
Krishnaraj, V. (2008) Effects of drill points on glass fibre reinforced plastic
composite while drilling at high spindle speed. Proceedings of the World Congress
on Engineering Vol II WCE 2008, London, U.K.
146
Latha, B and V. S. Senthilkumar (2009) Fuzzy rule based modeling of drilling

parameters for delamination in drilling GFRP composites. Journal of Reinforced
Plastics
and
Composites,
(Online
Version
available)
DOI:
10.1177/0731684407087570.
Latha, B., and V. S. Senthilkumar (2010) Modelling and Analysis of Surface
Roughness Parameters in Drilling GFRP Composites Using Fuzzy Logic. Materials
and Manufacturing Processes, 25, 817-827.
Lee, E.S. (2001) Precision machining of glass fibre reinforced plastics with respect to
tool characteristics. Int J Advanced Manufacturing Technology, 17, 791798.
Lin, S.C. and L.K. Chen (1996) Drilling carbon fibre-reinforced composite material
at high speed, Wear, 194, 156-162.
Malhotra, S.K. (1990) Some studies on drilling fibrous composites, J. Mat.
Processing Technology, 24, 291-300.
Mansour Mohamed, Cheryl Carlson Pugh and Robert Sadler (1993) Effect of
machining on the performance of carbon/epoxy 3D woven composites, International
conference on advanced composites, Ed. by Chandra T. and Dhingra A.K. The
Minerals Metals and Materials Society, 723-728.
Mariatti, M., M.Nasir, H.Ismail and J.Backlund (2004) Effect of hole drilling
techniques on tensile properties of continuous fibre impregnated thermoplastic
(COFIT) plain weave composites, J. Reinforced Plastics and Composites, 23, 117386.
Mata, F., P.Reis P and J.P.Davim (2006) Physical cutting model of polyamide
composites (PA66 GF30). Materials Science Forum, 14, 643647.
Mathew, J., N. Ramakrishnan and N.K. Naik (1999) Trepanning on unidirectional
composites: delamination studies, Composites Part A, 30, 951-959.
147
Matoorian, P., S.Sulaiman and M.M.H.M.Ahmad (2008) An experimental study

for optimization of electrical discharge turning (EDT) process. J Mater Process
Technology, 204, 350356.
Mohan, N.S., S.M.Kulkarni and A.Ramachandra (2007) Delamination analysis in
drilling process of glass fibre reinforced plastics (GFRP) composite materials.
Journal of Material Processing Technology, 186, 265- 271.
Naveen Sait, A., S. Aravindan and A. Noorul Haq (2009) Optimisation of
machining parameters of glass-fibre-reinforced plastic (GFRP) pipes by desirability
function analysis using Taguchi technique. International Journal of Advanced
Manufacturing Technology, 43, 581589.
Noorul Haq, A., P. Marirnuthu and R. Jeyapaul (2007) Multi response
optimization of machining parameters of drilling A1/SiC metal matrix composite
using grey relational analysis in the Taguchi method, International Journal Advanced
Manufacturing Technology, 00110.1007/s00170-007-0981-4.
Omer Erkan, Birhan Isk, Adem Cicek and Fuat Kara (2012) Prediction of
Damage Factor in end Milling of Glass Fibre Reinforced Plastic Composites Using
Artificial Neural Network. Appl Compos Mater., 23, 512-518.
Palanikumar, K (2007) Modelling and analysis for surface roughness in machining
glass fibre reinforced plastic using RSM methodology. Material and Design, 28,
2611-18.
Palanikumar, K., L. Karunamoorthy, and R. Karthikeyan (2006) Assessment of
factors influencing surface roughness on the machining of glass fibre-reinforced
polymer composites. Material Design, 27, 862871.
Palanikumar, K., L.Karunamoorthy, R.Karthikeyan and B.Latha (2006)
Optimization of Machining Parameters in Turning GFRP Composites Using a Carbide
(K10) Tool Based on the Taguchi Method with Fuzzy Logics. Metals & Materials
International, 12, 483-491.
148
Palanikumar, K., L. Karunamoorthy, S.B.Ramesh and S.Jeaudeen (2006)

Application of ANN for prediction of tool wear in machining of GFRP
composites. Proceedings of International Conference on Recent Advances in
Material Processing Technology, 95-104.
Paulo Davirn, J. and C.A.Conceicao Antonio (2001) Optimal drilling of particulate
metal matrix composites based on experimental and numerical procedures,
International Journal of Machine Tools & Manufacture, 41, 21-31.
Paulo Davim, J. and Luis Figueira (2007) Machinability evaluation in hard turning
of cold work tool steel (02) with ceramic tools using statistical techniques, Materials
and Design, 28, 1186-91.
Paulo Davim, J. and A. Monteria Baptist (2000) Relationship between cutting
force and PCD cutting tool wear in machining silicon carbide reinforced aluminum
Journal of Materials Processing Technology, 103, 417-423.
Paulo Davim, J. and Pedro Reis (2005) Damage and dimensional precision on
milling carbon fibre-reinforced plastics using design experiments, Journal of
Materials Processing Technology, 160,160-167.
Paulo Davim, J. and C.A. Conceicao Antonio (2001) Optimization of cutting
conditions in machining of aluminum matrix composites using a numerical and
experimental model, Journal of Materials Processing Technology, 112, 78-82.
Paulo Davim, J., Pedro Reis and C.A Conceicao Antonio (2004) Experimental
study of drilling glass fibre reinforced plastics (GFRP) manufactured by hand lay-up,
Composites science and Technology, 64, 289-297.
Paulo Davim, J. and Rosaria Cardoso (2006) Tribological behaviour of the
composite PEEK-F30 at dry sliding against steel using statistical techniques,
Materials and Design, 27, 338-342.
Pignatiello, J.J. (1993), Strategies for robust multiresponse quality engineering. IIE
Trans, 25, 515.
149
Pramanik, A., L.C. Zhang and J.A.Arsecularatne (2006) Prediction of cutting

forces in machining of metal matrix composites, International Journal of Machine
Tools & Manufacture, 46, 1795-80.
Praveen Raj, P. and A. Elaya Perumal (2010) Taguchi Analysis of surface
roughness and delamination associated with various cemented carbide K10 end mills
in milling of GFRP. Journal of Engineering Science and Technology Review, 3, 5864
Puw, H.Y. and H. Hocheng (1995). Anisotropic chip formation models of cutting of
FRP. In: ASME Symposium on Material Removal and Surface Modification Issues in
Machining Processes, New York.
Radhakrishnan, T. and S.M.Wu (1981) Online hole quality evaluation for drilling
composite material using dynamic data, J. Engg. Industry, 103, 119-25.
Ramkumar, J., S.Aravindan, S.K.Malhotra and R. Krishnamurthy (2004) An
enhancement of the machining performance of GFRP by oscillatory assisted drilling,
Int. J. Advanced Manufacturing Tech, 23, 240-244.
Ramkumar, J., S.K.Malhotra and R.Krishnamurthy (2004) Effect of workpiece
vibration on drilling of GFRP) laminates, J. Materials Processing Technology, 152,
329-332.
Rarnulu, M. and Arola (1993) Water and abrasive water jet cutting of unidirectional
graphite/Epoxy composites, Composites, 24, 299-303.
Ramulu, M., C.W.Wern and J.L.Garbini (1993) Effect of the direction on surface
roughness measurements of machined graphite/epoxy composite. Composites
Manufacturing, 4, 3951.
Ramulu, M., D.Arola, and K.Colligan (1994) Preliminary investigation of effects
on the surface integrity of fibre reinforced plastics. In. Engineering systems design
and analysis, 2, 93101.
150
Ramulu, M., T.Branson and D. Kim (2001) A study on the drilling of composite
and titanium stacks, Composite Structures, 54, 67-77.
Ravishankar, S.R. and C.R.L Murthy (2000) Application of acoustic emission in
drilling of composite laminates, NDT &E International, 33, 429-235.
Roy P.K., and S.K.Basu (1977) Evaluation of processing factors on turning of
thermoplastics. Polymer Engineering and Science, 17, 751757.
Sadat, A.B., W.S.Chari and B.P.Wang (1992) Delamination of graphite/ epoxy
laminate during drilling operation. J. Energy Resources Technology, 114,139-141.
Sakuma SETO (1981) Flank wear characteristics of cutting tools in different types
of tests (Comparison between results of conventional turning, step turning and facing
tools), Bulletin of the JSME, 24, 1871-78.
Sakuma SETO (1983) Tool wear in cutting glass fibre reinforced plastics (The
relation between fibre orientation and tool wear), Bulletin of the JSME, 26, 1420-27.
Sakuma Yakoo Seta (1984) Study on drilling of reinforced plastics (GFRP &
CFRP)-Relation between-tool material and wear behavior, Bulletin of JSME, 27,
1237-44.
Sandvik Coromant, U.K. Sandvik tools cut composite machining time at Westland
helicopters, Aircraft Engineering, 1987.
Sahin Yusuf and A.Riza Motorcu (2004) Surface roughness prediction model in
machining of carbon steel by PVD coated cutting tools. In. J Applied Science, 1, 12
17.
Sahji, S. and V.Radhakrishnan (2003) Analysis of process in surface grinding with
graphite as lubricant based on the Taguchi, Journal of Materials Processing
151
Santhanakrishnan, G., R. Krishnamurthy and S.K.Malhota (1988). Machinability

characteristics of fibre reinforced plastics composites. Journal of Mechanical
Working Technology, 17, 195204.
Santhanakrishnan, Krishnamurthy and Malhotra (1989) High speed steel tool
wear studies in machining of glass fibre reinforced plastics. Wear, 132, 327-336.
Santhosh Ranganath, B.Alberta, W. Campbell Daniel and Gronkiewicz (2007) A
model to calibrate and predict forces in machining with honed cutting tools or inserts.
International Journal of Machine Tools & Manufacture, 47, 820-840.
Senthil kumar, A., A.Raja Durai and T.Sornakurnar (2006) Wear behavior of
alumina based ceramic cutting tools on machining steels. Tribology International, 39,
191-197.
Singh, Y.K. and M.K.Khar (1975) A note on drilling PVC sheets. Int. J. Machine
Tools Design and Research, 15, 267-269.
Singh, Y.K., D.Nayak, R.Saxena and N.Bhatnagar (2004) Drilling induced
damage in FRP composite laminates. Journal of Engineering for Industry, 85, 37-40.
Singh, R.V.S., B.Latha and V.S Senthilkumar (2009) Modeling and analysis of
thrust force and torque in drilling GFRP composites by multi-facet drill using fuzzy
logic. International Journal of Recent Trends in Engineering, 5, 66-70.
Sivasankaran, S., R. Narayanasamy, T. Ramesh and M. Prabhakar (2009)
Analysis of workability behavior of AlSiC P/M composites using back propagation
neural network model and statistical technique. Computational Materials Science,
47, 4659.
Sreejith, P.S., R. Krishnamoorthy, S.K.Malhotra and K.Narayanasamy (2000)
Evaluation of PCD tool performance during machining of carbon/ phenolics ablative
composites. J Mater Process Technology, 104, 5358.
152
Stone, R. and R.Krishnamoorthy (1996) A neural network thrust force controller to

minimize delamination during drilling of graphite-epoxy laminates. Int.J. Machine
Tools and Manu. 36, 985-1003.
Sudarsana Rao, H., G.Vaishaf Ghorpade and A.Mukherje (2005) A genetic
algorithm based back propagation network for simulation of stress-strain response of
ceramic-matrix-composites. Composites and Structures, 5, 2-10.
Suleyman Yaldiz and Faruk Unsacar (2006) A dynamometer design for measure
merit the cutting forces on turning. Measurement, 39, 80-89.
Suresh, PVS., P.Venkateswara Rao and S.G. Ceshmukh (2002) A genetic
algorithmic approach for optimization of surface -foughnes3 prediction model.
International Journal of Machine Tools & Manufacture, 42, 67S-680.
Tsao, C.C (2008) Influence of drill geometry in drilling carbon fibre reinforced
plastics. Key engineering materials, Advances in Machining & manufacturing
technology, 9, 236-240.
Tsao, C.C. and W.E.Chen (1997) Prediction of the location of delamination in the
drilling of composite laminates. JI. Of Materials Processing Technology, 70,185-189.
Tsao, C.C. and H.Hocheng (2004) Taguchi analysis of delamination associated with
various drill bits ill drilling of composite material. Int. J. Machine tools & Mfr.,44,
1085-90.
Velayudham, A., R.Krishnamoorthy and T. Soundara pandian (2005) Evaluation
of drilling characteristics of high volume fraction fibre glass reinforced polymeric
composites. Int. J. Mach Tools and Mfr., 5, 339-406.
Venu Gopala Rao, G., P.Mahajan and N.Bhatnagar (2007) Micro-mechanical
modeling of machining of FRP composites Cutting force analysis. Composites
Science and Technology, 67, 579593.
153
Vining, G. (1998) A compromise approaches to multiresponse optimization. J

Quality Technology, 30, 309313.
Wen - Chou Chen (l997) Some experimental investigations in the drilling of Carbon
fibre reinforced composite laminations, Int. J. Machines Tools and Mfr., 37, 1097 108.
Wong, S.V. and S.Hamoudam (2003) Machinability data representation with
artificial neural network, Journal of Materials processing technology, 138, 538-544.
Xiao, K.Q., and L.C.Zhang (2002) The role of viscous deformation in the
machining of polymers. International Journal of Mechanical Sciences, 44, 2317
2336.
Zadeh, L (1965) Fuzzy sets, Inform and control, 8, 338.
Zhang, L.B., L.J.Wang, X.Y.Liu, Zhao H-W, Wang X. and H.Y.Lou (2001)
Mechanical model for predicting thrust and torque in vibration drilling fibre
reinforced composite materials. Int. J. of Machine Tools and Mfr, 41, 641-657.
Zhong, Z.W., L.P.Khoo and S.T.Han (2006) Prediction of surface roughness of
turned surfaces using neural networks. Int J Advanced Manufacturing Technology,
28, 688693.
154
LIST OF PUBLICATIONS BASED ON THESIS

International Journals
1. Jenarthanan, M.P, R.Jeyapaul and N.Naresh, (2012), Modelling and
analysis of factors influencing surface roughness and delamination of milling
of GFRP laminates using RSM, Multidiscipline Modeling in Materials and
Structures, Vol.8. Issue: 4 pp.489- 504.
2. Jenarthanan, M.P and R.Jeyapaul, (2012), Evaluation of milling
characteristics of Resin hybrid GFRP laminates, Pigment and Resin
Technology, Vol.42. Issue: 5. (Accepted)
3. Jenarthanan, M.P and R.Jeyapaul, (2012) Analysis of factors influencing
Delamination in milling process of glass fibre reinforced plastic (GFRP)
composite materials, Multidiscipline Modeling in Materials and Structures
(Accepted)
4. Jenarthanan, M.P and R.Jeyapaul, Machinability Study of carbon fibre
reinforced polymer (CFRP) composites Using Design of Experiment
Technique, Pigment and Resin Technology (Accepted)
5. Jenarthanan, M.P and R.Jeyapaul, Optimisation of machining parameters
on milling of GFRP composites by Desirability Function Analysis using
Taguchi Method, International Journal of engineering, science technology
review (Under Review)
6. Jenarthanan, M.P
and R.Jeyapaul, Analysis and comparison
of
machinability behaviour of GFRP composites with different fibre orientations

using RSM and ANN, International Journal of Advanced Manufacturing
Technology (Under Review).
7. Jenarthanan, M.P and R.Jeyapaul, Modelling of Machining Force in end
Milling of GFRP Composites Using MRA and ANN, International Journal
of Advanced Manufacturing Technology (Under Review).
155
8. Jenarthanan, M.P and R.Jeyapaul, Analysis and Optimisation of Machinability

behavior of GFRP Composites Using Fuzzy Logic, International Journal of
Machining and Machinability of Materials (Under Review).
.
International Conferences
1. Jenarthanan, M.P and R.Jeyapaul (2012), Evaluation of Machinability on
Milling of GFRP Laminates Using Solid Carbide End mill Tools with
Different Helix Angles, International Conference AMPCO 2012 at IIT
Roorkee, November 2-4, 2012.
2. Jenarthanan, M.P and R.Jeyapaul (2012), Taguchi Analysis of surface
roughness and delamination associated with solid carbide end mills with
different helix angles in milling of GFRP composites, International
Conference on Emerging Trends in Manufacturing Technology, September 56, 2012 , Toc H Institute of Technology,Ernakulam,Kerala.
3. Jenarthanan, M.P, Krishna Kumar Dadsena and R. Jeyapaul (2013)
Evaluation of Machinability Behaviour in Endmilling of GFRP Composites
Based on the Taguchi Method and ANN, International Conference CPIE
2013 at NIT Jalandhar, March 29-31, 2013.
4. Krishna Kumar Dadsena, M. P. Jenarthanan and R. Jeyapaul (2013)
Analysis and Optimisation of Machinability behavior of GFRP Composites
Using Fuzzy Logic, International Conference CPIE 2013 at NIT Jalandhar,
March 29-31, 2013.
National Conferences
1. Naresh nelli, M.P.Jenarthanan and R.Jeyapaul (2012), Delamination
analysis in milling process of glass fibre reinforced plastic (GFRP) composite
materials, National conference on Emerging Trends in Mechanical
Engineering, April 2-3, K.L.N College of Engineering, Madurai.
156
CURRICULUM VITAE
Mr. M.P. Jenarthanan was born on 29th September 1980 at Paramakudi,
Ramanathapuram district, Tamilnadu, India. He graduated in Mechanical engineering
from Madurai Kamaraj University, in the year of 2003 and obtained his Masters
Degree in Engineering Design from Anna University Chennai, in the year 2005. He
joined as a lecture at K.L.N College of Engineering, Madurai during the year 2005.
He has published about 9 papers in International/National Journals and Conferences.
His area of interest includes composite machining, taguchi methodology, response
surface methodology, non-traditional optimisation techniques and numerical
simulation.
157
DOCTORAL COMMITTEE
Dr. T. Selvaraj (Chairman)
Professor, Department of Production Engineering,
N.I.T Tiruchirappalli.
Dr. R. Jeyapaul (Research Supervisor)

Associate Professor, Department of Production Engineering,
Dr. Somashekhar.S.Hiremath (External Member)

Associate Professor, Department of Mechanical Engineering,
I.I.T. Madras, Chennai.
Dr. P. Sathiya (Internal Member)

Associate Professor, Department of Production Engineering,
Dr. S. P. Sivapirakasam (Allied department Member)

Associate Professor, Department of Mechanical Engineering,
158

Print Thesis - PDF 179 May28

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Print Thesis - PDF 179 May28

Uploaded by

Copyright:

Available Formats

STUDY OF MACHINABILITY

CHARECTERISTICS OF GFRP COMPOSITES

for the award of the degree

DEPARTMENT OF PRODUCTION ENGINEERING

This is to certify that the thesis entitled STUDY OF MACHINABILITY

With the increasing use of Glass Fibre Reinforced Polymer (GFRP)

from confirmation runs indicated that the determined optimal combination of

Engineering, NITT, for his valuable suggestions and encouragement.

Department of Mechanical Engineering, NITT, for his support and encouragement.

COMPOSITES: A BRIEF HISTORY...

Thermoset resins ...

Thermoplastic resin ...

Continuous strand roving.

According to matrix materials .

Polymer matrix composites (PMC) .

Ceramic matrix composites (CMC) .

Metal matrix composites (MMC)

Classification of composites based on fibre

PRINCIPLES FOR USING GFRP COMPOSITES ..

Amount and orientation of glass fibre used ..

MAJOR FABRICATION PROCESSES OF GFRP

Hand lay-up process

Filament winding process

SIGNIFICANCE OF FRP MACHINING ..

Non- Traditional Machining .

INTRODUCTION TO MODELLING AND

Response Surface Methodology (RSM).

Taguchi Methodology .......

Analysis of Variance (ANOVA) ...

Artificial Neural Network (ANN)

THE PRESENT WORK ..

Motivation for the present work .

MACHINING OF GFRP COMPOSITES ..

NON-TRADITIONAL MACHINING (NTM) ..

Cutting tool geometries

Factors influencing surface roughness

Factors influencing delamination factor .

Factors influencing the machining force .

LITERATURE ON MODELLING AND

Response Surface Methodology ..

Artificial Neural Network .

ENDMILLING OF GFRP USING SOLID CARBIDE TOOL

PREPARATION OF TEST SPECIMENS .

MATERIAL SELECTION PROCESS ..

Hand lay-up technique

MECHANICAL PROPERTIES OF TEST

Measurement of surface roughness (Ra) .

Measurement of delamination factor (Fd) .

RESPONSE SURFACE METHODOLOGY ..

Results and discussions

Influence of machining parameters on surface

Influence of machining parameters on delamination

ARTIFICIAL NEURAL NETWORK

Single layer net .

Multi layer net ..

SETTING THE WEIGHTS .

BACK PROPAGATION NEURAL NETWORK

NEURAL NETWORK DESIGN

NEURAL NETWORK TRAINING

TESTING AND PERFORMANCE OF BPNN ..

COMPARISON OF RSM AND ANN MODELS

EFFECTS OF MACHINING PARAMETERS ON

MACHINE TOOL SETUP AND DATA