Composite Materials in Piping Applications PDF

Composite Materials in
Piping Applications
Design, Analysis and Optimization of
Subsea and Onshore Pipelines from FRP Materials
Dimitrios G. Pavlou, Ph.D.

Technological Institute of Halkida (TEI-Halkida)
Mechanical Engineering Department
Greece
DEStech Publications, Inc.

Composite Materials in Piping Applications

4 3 9 N o r t h Du k e S t r e e t
Lancaster, Pen nsyl van ia 1 7602 U.S.A.
Copyrig ht © 2013 by DEStech Publications, Inc.

A l l r i g h t s r es e r v e d
No part of this pub licati on may be reprodu ced, sto red in a

retrieval system, o r transmitt ed, in any fo rm or by an y means,
electro nic, mechan ical, photo cop ying, recording, or otherwi se,
without the p rior wri tten permi ssion of the publ ish er. The p ublisher
i s n o t l i a b l e f o r a n y e r r o r s r e s u l t i n g f r o m t h e u s e o f t h e C D- R O M .
Prin ted in the United States of Ameri ca

10 9 8 7 6 5 4 3 2 1
M ai n en t r y u n d e r t i t l e :
Composite Material s in Piping Ap plicat ions: Desig n, Analysis and Optimizatio n of Subsea
a n d On s h o r e P i p e l i n e s f r o m F R P M a t e r i a l s
A DE S t e c h P u b l i c a t i o n s b o o k
Bibliog rap hy: p .
In c lu d e s in d e x p . 3 9 5
Library of Congress Contro l Nu mber: 20139 35448

ISB N N o . 9 7 8 -1 -6 0 5 9 5 -0 2 9 -7
HOW TO ORDER THIS BOOK

BY PHONE: 877-500-4337 or 717-290-1660, 9AM–5PM Eastern Time
BY FAX: 717-509-6100
Order Department
BY MAIL:

439 North Duke Street
Lancaster, PA 17602, U.S.A.
BY CREDIT CARD: American Express, VISA, MasterCard, Discover
BY W W W SIT E: http://www.destechpub.com
Acknowledgments
The author wishes to express his appreciation to colleagues who helped in

developing this book. Dr. Rui Miranda Guedes of the Department of Mechanical
Engineering at the University of Porto and Dr. Hugo Faria of INEGI-Instituto de
Engenharia Mechânica e Gestão Industrial at Porto contributed the material in
Chapter 6 on creep design of piping applications using composites. I am indebted
to them for providing their expertise in this area. My former student (currently
Dipl. Mechanical Engineer) Stavros Lykakos was the tester of the specially de-
veloped software. His help was very valuable. Moreover, my students Michalis
Bazanos, Andronikos Miniatis, Giorgos Papastefanos, Panagiotis Bouyioukos,
Giorgos Macheras, Giorgos Roussos, Giorgos Vagalis, Tassos Tsitsakis, Apostolis
Kalaris provided considerable assistance with the figures, graphics and text typ-
ing. Particular thanks go to Asimina Kehagia and Giorgos Sioris for their valuable
linguistic comments.
Especially, the author wishes to express his enormous appreciation to Dr.
Joseph Eckenrode, publishing director of DEStech Publications for his kind in-
vitation to develop this book and for his encouragement during the efforts of the
last two years. Finally, of immense help in refining the book were the reviewers
who provided their helpful comments. For their time and comments I am deeply
grateful.
D.G. Pavlou
xi
Introduction
The Minoans in 27th century BC were the first civilization known to use un-
derground clay pipes for sanitation, heating systems and water supply [1]. At
Knossos, on the island of Crete, Greece, pipes having a diameter of 4.0–6.0 inches
with perfect socket joints are the oldest ever discovered. The first known applica-
tion of metal for manufacture of pipes is “hydraulis,” a 3rd century Greek pipe
organ that operated by converting the dynamic energy of water into air pressure to
drive the pipes. Later, in the first century BC, the Romans used lead to fabricate
pipes with diameters from 0.5–22.0 inches for urban plumbing.
Fiber-reinforced pipes were first developed after World War II [2]. They con-
sisted of glass fiber cloth and resin applied over a mandrel by hand. An evolved
form of the handmade pipe is the filament-wound pipe consisting of tensioned
fibers properly oriented to bear the combination of hoop and axial forces.
Manufacturing of the first commercial FRP pipes started in the mid 1950s [2].
From 1960–1980 a continuous process for manufacturing of FRP pipes was de-
veloped and its efficiency improved to the point where large quantities of pipe
were produced for the chemical and oil industry. After the peak of 1981, a slight
decrease in the productivity rate occurred until 1986. At present, FRP pipes find
many applications around the world since they combine high resistance to corro-
sive fluids with a capacity for increased mechanical loads. FRP pipes are used on
offshore platforms in Alaska and the Persian Gulf, as well as in water-flood proj-
ects in Saudi Arabia and in a saltwater/crude oil/gas line in the deserts of South
Oman. In 1993 the largest FRP pipe, with a length of 390Km and a diameter of
350 mm, was installed in Algeria [2] and is used for oilfield applications.
Since the unit price of carbon steel is about 14 times cheaper than the unit
price of Glass Fiber Reinforced Polymers (GFRP), steel pipes are still the main
type of pipes uses for fluid transmission. However, the unit price is not the only
parameter controlling the material cost. The allowable tensile force for a mate-
rial sheet subjected to tension is P = ASy, where A is the cross section carrying
the load and Sy is the material’s yield stress. The above formula can also be
written as P = (V/L)Sy, where V is the volume and L is the length of the material
xiii
xiv Introduction
sheet. Since density is ρ = Mass/V, the allowable force P can be formulated as

P = (Mass/ρL)Sy.
The material unit price is UP = Cost/Mass. Therefore, the allowable force can
now be written P = (Cost/UP) (1/ρL)Sy. Using this formula, a cost index demon-
strating the normalized cost, i.e., the cost for a unit length of the material sheet
per unit allowable force can be used for a material selection. With the aid of the
last formula providing P, the above cost index has the form CI = Cost/PL or CI
= UP ρ/Sy. From this formula it can be concluded that apart from the unit price
UP, the total material cost is strongly influenced by the material density ρ and the
yield stress Sy. The following figures illustrate the unit price ($/ton), density (Kg/
m3) and yield stress (MPa) for carbon steel, stainless steel, glass fiber-reinforced
polymers (GFRP) and carbon fiber reinforced-polymers (CFRP). Comparing for
example carbon steels and GFRPs, which are the main materials used for pipe
manufacturing, it can be shown that even though the GFRPs have a unit price that
is 11 times more expensive than that of steel, their density is about 4 times lower
and their yield stress is about 3 times higher. Taking into account the definition
of the cost index CI, it can be concluded that the lower density and the higher
strength of GFRPs reduce the total material cost significantly. Indeed, as shown in
the last figure that provides the cost index (CI) of four materials, the material cost
of GFRP is comparable to that of carbon steel.
UNIT PRICE $/ton

Introduction xv
DENSITY Kg/m3
YIELD STRESS Mpa

xvi Introduction
Cost index CI
Taking into account additional reductions in maintenance costs due to better

corrosion resistance, it is apparent that the use of FRP pipes is advantageous com-
pared that of steel pipes. However, the mechanical design of FRP pipes is much
more complicated due to the fact that FRP materials are anisotropic. Apart from
the allowable service loads, the anisotropy also influences their dynamic behavior
and stability, fatigue and creep life, spaces between supports, joining etc. The sub-
sequent chapters are intended to address the complexities presented by anisotropy
and thus to provide composite materials designers with the analytical methods
and computer programs necessary for detailed, quantitative mechanical design of
filament-wound FRP pipelines.
References
[1] http://www.historywiz.com
[2] Oswald Kenneth, Thirty years of fiberglass pipe in oilfield applications: A
historical perspective, Materials Selection and Design, MP/May, 1996.
Preface
After World War II, offshore and continental oil, gas and water transmission
infrastructure, as well as chemical, sewage and irrigation installations, benefited
from the development of fiber reinforced polymeric (FRP) pipes. However, be-
cause the unit price of composite materials was historically expensive, steel pipes
remained in use for the transmission of liquid commodities. Today, high main-
tenance costs due to corrosion of aging steel pipelines, as well as the reduction
of the unit price of composite materials, have led to reconsidering the optimum
material cost for pipeline applications. Moreover, as will be shown in the intro-
duction, the final material cost of piping is strongly influenced by material density
and strength. Since FRP materials exhibit a much lower density and much higher
strength than carbon steel, the final cost of such materials is currently comparable
to the cost of carbon steel. Moreover, the lower maintenance costs of composite
pipelines, which results from their excellent resistance to corrosion and fatigue,
leads to the conclusion that the use of composite materials for pipeline applica-
tions has now become advantageous when compared to the use of carbon steel
pipelines.
Since FRPs are anisotropic materials, the methods and theoretical tools for
the mechanical design of composite pipelines are completely different from the
design procedures for steel pipelines. The existing design standards are rather
semi-empirical and cover simple loading cases.
The aim of the present book is to provide detailed analytic and numerical tools
for the analysis and design of FRP composite pipelines under pure and combined
loading conditions (e.g., bending, external pressure and axial tension). Failure
prediction in creep and fatigue conditions, design of joints and supports, esti-
mation of flow capacity for liquid gas and multi-phase fluids are major topics
investigated in the following chapters. A strong feature of the book is the devel-
opment of Mathematica-based computer algorithms corresponding to any load-
ing conditions, in order to facilitate direct design. Moreover, nomographs for a
wide range of loading cases, pure and combined, have been derived for multi-
layered filament-wound pipes made from the common composites E-glass/epoxy
ix
x Preface
and S-glass/epoxy for quick dimensioning. Since in the first chapter principles
of the mechanics of anisotropic elasticity are briefly explained, a reader unfamil-
iar with composite materials can appreciate and understand the design principles
presented.
The book is organized as follows:
Chapter 1 provides a brief overview of the mechanical behavior of compos-

ite materials covering the anisotropic elasticity equations for laminae and
laminates, as well as the widely used Tsai-Wu failure criterion.
Chapter 2 explains the classification and properties of composite materials
and gives a brief description of the filament-winding method.
Chapter 3 develops theoretical tools for the mechanical design of pipelines
including pure bending, external pressure, axial tension, torsion and their
combinations in multilayered filament-wound pipelines by using failure and
buckling models.
Chapter 4 provides original models for the dynamic analysis of composite
pipes carrying fluids. Special emphasis is given to the estimation of critical
flow velocities that cause instability as well as the fluid hammer-induced
wave propagation.
In Chapter 5, methods are presented for designing joints as well as estimat-
ing the supporting spacing, hanger width and sizing of expansion loops.
Techniques for estimating the safe depth of underground pipelines situated
under in-service streets and railroads are also included.
Chapter 6 presents models for the life-time prediction of composite pipe-
lines under creep and fatigue conditions.
Chapter 7 explains flow models for estimating the flow capacity of compos-
ite pipelines that deliver liquid, gas or multi-phase fluids.
In Chapter 8, the parameters for optimum design of composite pipelines
with a view to reducing material costs are discussed.
Chapter 9 clarifies current quality control methods for the manufacture of
composite materials and composite pipelines.
Chapter 10 presents a collection of nomographs for direct mechanical
design of GFRP composite pipes under a wide range of pure and combined
loading conditions.
Dimitrios G. Pavlou
Technical Institute of Chalkida
January 2013
Contents
Preface ix
Acknowledgments xi
Introduction xiii
1 Mechanical Behavior of Fiber Reinforced

Composite Materials.........................................................1
1.1 Mechanical Behavior of Laminae 1
1.1.1 Generalized Hooke’s law 1
1.1.2 Effects of free thermal strains 6
1.1.3 Effects of free moisture strains 8
1.1.4 Plane stress constitutive relations 10
1.1.5 Coordinate transformation of stress and strain
components 13
1.1.6 Transformation of engineering properties 16
1.1.7 Free thermal and free moisture strains in the global
coordinate system 19
1.2 Mechanical Behavior of Laminates 24
1.2.1 Classical lamination theory 24
1.2.2 Laminate nomenclature 24
1.2.3 The Kirchhoff Assumption 27
1.2.4 Laminate strains 28
1.2.5 Laminate stresses 30
1.2.6 Laminate stiffness matrix 30
1.2.7 Classification of laminates 36
1.3 The Tsai–Wu Failure Criterion 39
References 43
iii
iv Contents
2 Classification, Properties and Production

Technology of FRP Materials..........................................45
2.1 The Composite Matrix Material 45
2.1.1 Thermosets 46
2.1.2 Thermoplastics 47
2.2 Fiber Materials 48
2.2.1 Glasses 48
2.2.2 Carbon fibers 49
2.2.3 Synthetic fibers 50
2.3 Production Technologies for FRP Composite Pipes 50
2.3.1 Filament winding 50
2.3.2 Fiber placement process 52
References 52
3 Mechanical Design of Composite Pipelines.................53

3.1 Types of Loading Cases 53
3.1.1 Installation loads 53
3.1.2 Operation loads 56
3.2 Pure Bending 56
3.2.1 Failure analysis 56
3.2.2 Buckling model 73
3.3 External Pressure 77
3.4 Combination of Bending and External Pressure 81
3.5 Axial Tension 88
3.6 Combination of Bending and Axial Tension 91
3.7 Combination of External Pressure and Axial Tension 94
3.8 Torsion 97
Contents v

References 102
4 Dynamic Stability of Composite Pipelines..................105

4.1 Free Vibration of Composite Pipes 105
4.1.1 Structural characteristics of composite pipes 105
4.1.2 Forces and bending moments acting on a composite
pipe element 111
4.2 Accelerations of the Fluid and Pipe Elements 114
4.3 Equation of Motion 116
4.3.1 Solution of equation of motion 117
4.3.2 Types of instability 122
4.4 Transfer Matrices Method (TMM) 124
4.5 Estimation of Critical Velocity for Composite Pipes
Conveying Fluid 129
4.5.1 Cantilever pipe 130
4.5.2 Fixed-fixed pipe 130
4.5.3 Pinned-pinned pipe 131
4.5.4 Fixed-pinned pipe 131
4.6 Effect of Temperature (Thermal Load) 132
4.7 Effect of Additional Mass 135
4.7.1 Transfer matrix of the segment 1-2 (pipe) 139
4.7.2 Transfer matrix of the segment 2-3 (collar-pipe) 139
4.7.3 Global transfer matrix 140
4.8 Effects of an Elastic Foundation 141
4.9 Effect of Additional Supports 142
4.9.1 Example 146
4.10 Estimation of Critical Flow Velocity in Relation to
Divergence 147
4.10.1 Elastic foundation effect 147
4.10.2 Thermal load and elastic foundation effects 147
4.10.3 Fixed-fixed pipe 148
4.10.4 Pinned-pinned pipe 149
4.10.5 Fixed-pinned pipe 149
4.11 Hydraulic Hammer 149
4.11.1 Shock pressure in a branched pipe 155
4.12 Wave Propagation Due to Hydraulic Hammer 155
vi Contents
4.12.1 Example 161

References 168
5 Connection and Supports of Composite Pipelines....169

5.1 Joining of Composite Pipelines 169
5.1.1 Approximate mechanical model for axial
loading 170
5.1.2 Approximate mechanical model for bending 178
5.2 Above-Ground Pipes 179
5.2.1 Maximum spacing between supports 180
5.2.2 Minimum hanger widths 182
5.2.3 Sizing of expansion loops 186
5.3 Underground Pipelines 192
References 196
6 Creep Design of Piping Applications Using

Composite Materials.....................................................197
6.1 Introduction 197
6.2 Creep Damage Accumulation Mechanisms in Composite
Materials 198
6.3 Short and Long-Term Static Failure of Composite
Pipes 206
6.3.1 Damage modeling 206
6.3.2 Creep rupture 212
6.3.3 An example of preliminary design for the
long-term 218
6.4 Lifetime of Composites Pipes Under Cyclic Loading 220
6.5 Applicable Standards 225
6.5.1 Identification and comparison of main standards 225
6.5.2 Long-term qualification tests of four different types of
GRP 230
6.6 Practical Design: A Case Study 232
6.7 Conclusions 237
6.8 Acknowledgements 238
References 238
7 Flow Capacity of Composite Pipelines........................243

7.1 Gas Transmission 243
7.1.1 Estimation of gas flow rate 249
Contents vii
7.2 Liquid Transmission 252

7.2.1 Flow capacity for laminar liquid flow 252
7.2.2 Flow capacity for turbulent flow 253
7.3 Multiphase Flow 255
7.3.1 Multiphase flow regimes for inclined pipelines 255
References 269
8 Optimization of Material Cost......................................271

8.1 Fiber Orientation and Loading Forces 271
8.1.1 Optimum fiber orientation for the combination of
axial tension and external pressure 272
bending and axial tension 276
bending and external pressure 277
9 Quality Control of Composite Pipe Systems...............279

9.1 Test Methods and Material Characterization 279
9.1.1 Thermal analysis DSC (Differential Scanning
Calorimetry) 279
9.1.2 Measurement of residual stresses 280
9.1.3 Creep strain and creep rupture tests 280
9.1.4 Impact testing 282
9.1.5 Fatigue testing 284
9.2 International Standards for Composite Pipes 287
9.3 Detection of Defects and Structural Health
Monitoring 295
9.3.1 Piezoelectric techniques 295
9.3.2 Optical fiber-based techniques 295
9.3.3 Ultrasonic testing 296
References 297
10 Case Studies..................................................................299
Introduction 299
10.1 Axial Tension 299
10.1.1 Results of failure model for axial tension 299
10.2 Pure Bending 310
10.2.1 Results of failure model for pure bending 310
10.2.2 Results of buckling model for pure bending 316
viii Contents
10.3 External Pressure 326

10.3.1 Results of failure model for external pressure 326
10.3.2 Results of buckling model for external
pressure 334
10.4 Torsion 340
10.4.1 Results of failure model for torsion 340
10.5 Butt Joints of Multilayered Filament-Wound Pipes 345
10.5.1 E-glass epoxy material 346
10.5.2 S-glass/epoxy material 356
10.6 Hanger Width 366
10.6.1 E-glass/epoxy material 366
10.6.2 S-glass-epoxy material 367
10.7 Spaces Between Supports 369
10.7.1 E-glass/epoxy material 369
10.7.2 Material: S-glass/epoxy material 371
10.8 Installation Depth for Underground Pipelines vs. the
Vertical Load F 374
10.8.1 E-glass-epoxy materials 374
10.8.2 S-glass/epoxy material 384
Index 395
About the CD-ROM 399
Chapter 1
Mechanical Behavior of Fiber Reinforced

Composite Materials
1.1 Mechanical Behavior of Laminae

1.1.1 Generalized Hooke’s law
Figure 1.1 shows a part of a fiber reinforced composite material. Since the ma-
terial properties of the fibers have different values from those of the matrix, it is
convenient to use an orthogonal coordinate system that has one axis aligned with
the fiber orientation. The axis x1 is aligned with the fiber direction, and perpen-
dicular to the fibers are the axes x2 (lying in the plane of the layer) and x3 (per-
pendicular to the plane of the layer). The orientations of the axes x2 and x3 are
called matrix directions. The coordinate system x1 , x2 , x3 is called the principal
coordinate system and the corresponding directions are called principal direc-
tions. From a macroscopic point of view, the mechanical behavior of each indi-
vidual fiber or matrix element is of no practical importance. For design purposes,
the two-material fiber-matrix system is treated as a homogeneous anisotropic ma-
terial. Obviously, this material does not have the same properties in all directions,
i.e., its stiffness and strength in the fiber’s direction are higher than in the matrix
directions. The study of the elastic response of the above material is equivalent
to determining the relations between the stresses applied to the bounding sur-
faces and the corresponding deformations of the material as a whole. Denoting by
E1 , E2 , E3 the modulus of elasticity in the directions x1 , x2 , x3 respectively, and by
νi j (i = 1, 2, 3 , j = 1, 2, 3) the Poisson’s ratios given by
εj
νi j = − ι = (1.1)
1, 2, 3 j = 1, 2, 3
ει
1
2 Mechanical Behavior of Fiber Reinforced Composite Materials
Figure 1.1 Principal coordinate system.
the relationship between the stresses and strains in the directions x1 , x2 , x3

(Fig. 1.2) is given by the following generalized Hooke’s law:
 σ1   C11 C12 C13 0 0 0   ε1 

 σ  C C22 C23 0 0 0  ε 
 2   21   2
 σ 3  C31 C32 C33 0 0 0   ε3 
 =    (1.2)
 τ 23   0 0 0 C44 0 0   γ 23 
 τ13   0 0 0 0 C55 0   γ13 
     
 τ12   0 0 0 0 0 C66   γ12 
where Ci j {i = 1 − 6 , j = 1 − 6} , are material constants.

The above equation can be written in terms of strains:
Mechanical Behavior of Laminae 3
Figure 1.2 Stresses acting on the principal directions.
 ε1   S11 S12 S13 0 0 0   σ1 

 ε  S S 22 S 23 0 0 0   σ 2 
 2   21  
 ε3   S31 S32 S33 0 0 0   σ3 
 =    (1.3)
 γ 23   0 0 0 S 44 0 0   τ 23 
 γ13   0 0 0 0 S55 0   τ13 
     
 γ12   0 0 0 0 0 S66   τ12 
where
1
S11 = (1.4)
E1
v21
S12 = − (1.5)
E2
v31
S13 = − (1.6)
E3
v12
S 21 = − (1.7)
E1
1
S 22 = (1.8)
E2
v32
S 23 = − (1.9)
E3
v13
S31 = − (1.10)
E1
v23
S32 = − (1.11)
E2
1
S33 = (1.12)
E3
1
S55 = (1.13)
G13
1
S55 = (1.14)
G13
1
S66 = (1.15)
G12
Combining equations (1.2) and (1.3), it can be concluded that Ci j can be written
in terms of the above engineering constants Si j . For shorthand notation, the equa-
tions (1.2) and (1.3) can be abbreviated as:
{σ1 } = [C ] {ε1 } (1.16)

{ε1} = [ S ]{σ1} (1.17)
From eqs. (1.16) and (1.17), it is obvious that
[C ] = [ S ]−1 (1.18)
According to the Maxwell-Betti Reciprocal Theorem, the following relationships

among material properties can be obtained [e.g., 1, 2]:
v12 v21
= (1.19)
E1 E2
v13 v31
= (1.20)
E1 E3
v23 v32
= (1.21)
E2 E3
yielding
S21 = S12 (1.22)
S31 = (1.23)
S13
S32 = (1.24)
S 23
Therefore, with the aid of eqs. (1.18) and (1.22–1.24), the members Cij of the
matrix [C ] are given by:
S 22 S33 − S 232
C11 = (1.25)
S
S13 S 23 − S12 S33
C12 = (1.26)
S
S12 S 23 − S13 S 22
C13 = (1.27)
S
C21 = C12 (1.28)
S33 S11 − S132

C22 = (1.29)
S
S12 S13 − S 23 S11

C23 = (1.30)
S
C31 = C13 (1.31)
C32 = C23 (1.32)
S11 S 22 − S122
C33 = (1.33)
S
1
C44 = (1.34)
S 44
1
C55 = (1.35)
S55
1
C66 = (1.36)
S66
where
S = S11 S 22 S33 − S11 S 232 − S 22 S132 − S33 S122 + 2S12 S 23 S13 (1.37)
The matrices [C ] and [ S ] correlating the stress and strains are called the stiffness
matrix and the compliance matrix, respectively.
1.1.2 Effects of free thermal strains

Temperature changes in a fiber reinforced composite material element can
cause significant stresses in the fiber and matrix. However, when the stresses are
integrated over the composite medium, the net result is zero. Taking into account
the assumption that the fiber–matrix system can be treated as a homogeneous

anisotropic material, the simplification that free thermal strains of a material ele-
ment with no constraints on its bounding surfaces do not cause stresses has to be
reinterpreted. To this end, in order to maintain consistency with the definition of
stress and strain, the term “mechanical strains” will be used in the stress–strain
relations. According to this concept, mechanical strains are the total strains (i.e.,
the changes in length per unit length of a material element) minus the free thermal
strains. Therefore, the stress–strain relations can be written in the form:
 ε1 − α1 ∆Τ   S11 S12 S13 0 0 0   σ1 

 ε − α ∆Τ   S S 22 S 23 0 0 0  σ 
 2 2   12  2
 ε3 − α3 ∆Τ   S13 S 23 S33 0 0 0   σ3 
  =     (1.38)
 γ 23  0 0 0 S 44 0 0   τ 23 
 γ13  0 0 0 0 S55 0   τ13 
     
 γ12   0 0 0 0 0 S66   τ12 
or
 σ1   C11 C12 C13 0 0 0   ε1 − α1 ∆Τ 

 σ  C C22 C23 0 0 0   ε − α ∆Τ 
 2   12  2 2 
 σ 3  C13 C23 C33 0 0 0   ε3 − α3 ∆Τ 
  =     (1.39)
 τ 23   0 0 0 C44 0 0   γ 23 
 τ13   0 0 0 0 C55 0   γ13 
     
 τ12   0 0 0 0 0 C66   γ12 
The mechanical strains used in the above equations are given by:
 ε1mech   ε1 − α1 ∆Τ 
 mech   ε − α ∆Τ 
 ε2   2 2 
 ε3mech   ε3 − α3 ∆Τ 
 mech  =   (1.40)
 γ 23   γ 23 
 γ13
mech   γ13 
 mech   
 γ12   γ12 
where α1 , α2 , α3 are the thermal expansion coefficients in the principal directions,

and ΔΤ is the temperature change.
The above definitions take into account the cases of: (a) stress with no thermal
effects; (b) thermal effects but no stresses; and (c) thermal effects with stresses.
For each of these three cases, the corresponding stress–strain relationships can be
written in the following form:
Case (a) Stress with no thermal effects:
 ε1   S11 S12 S13 0 0 0   σ1 

ε  S
 2  12 S 22 S 23 0 0 0  σ 
 2
 ε3  S S 23 S33 0 0 0   σ3 
  =  13    (1.41)
0 0 0 0 0 
γ
 23  S 44  τ 23 
 γ13  0 0 0 0 S55 0   τ13 
     
γ
 12   0 0 0 0 0 S66   τ12 

Case (b) Thermal effects of a material element with no constraints on its
bounding surfaces:
 ε1 − α1 ∆Τ  0
 ε − α ∆Τ  0
 2 2   
 ε3 − α3 ∆Τ  0
  =   (1.42)
 γ 23   0
 γ13   0
   
 γ12   0
Case (c) Stress caused by a temperature change of a fully restrained (against

deformation) material element:
 σ1   C11 C12 C13 0 0 0   − α1 ∆Τ 

 σ  C  − α ∆Τ 
C22 C23 0 0 0   2 
 2   12
 σ 3  C13 C23 C33 0 0 0   − α3 ∆Τ 
  =     (1.43)
 τ 23   0 0 0 C44 0 0   0 
 τ13   0 0 0 0 C55 0   0 
     
 τ12   0 0 0 0 0 C66   0 
1.1.3 Effects of free moisture strains

When a piece of composite material is exposed to a liquid, a certain amount
of that liquid is absorbed, yielding an increase in the composite’s weight and
expansion. Although the weight gain is negligible (usually less than 4%), the ex-
pansion can be important. The created free moisture strains can be considered to
be proportional to the amount of liquid absorbed. Therefore, in analogy to the

coefficient of thermal expansion, the term coefficient of moisture expansion can
be used in order to describe the free moisture expansion of a composite element.
Since the moisture expansion in the fiber direction is small (compared to the mois-
ture expansion of the polymer), the assumption of treating the fiber–matrix system
as a homogeneous anisotropic material and the concept of mechanical strains will
be again adopted. Using the notations β1 , β 2 , β3 for the moisture-expansion coef-
ficients in the principal directions and ΔΜ for the change of absorbed moisture,
the generalized Hooke’s law can be written in the form:
 ε1 − β1 ∆Μ   S11 S12 S13 0 0 0   σ1 

 ε − β ∆Μ  S S 22 S 23 0 0 0  σ 
 2 2   12  2
 ε3 − β3 ∆Μ   S13 S 23 S33 0 0 0   σ3 
  =     (1.44)
 γ 23  0 0 0 S 44 0 0   τ 23 
 γ13  0 0 0 0 S55 0   τ13 
     
 γ12   0 0 0 0 0 S66   τ12 
or
 σ1   C11 C12 C13 0 0 0   ε1 − β1 ∆Μ 

 σ  C C22 C23 0 0 0   ε − β ∆Μ 
 2   12  2 2 
 σ 3  C13 C23 C33 0 0 0   ε3 − β3 ∆Μ 
  =     (1.45)
 τ 23   0 0 0 C44 0 0   γ 23 
 τ13   0 0 0 0 C55 0   γ13 
     
 τ12   0 0 0 0 0 C66   γ12 
The mechanical strains used in the above equations are given by:
 ε1mech   ε1 − β1 ∆Μ 
 mech   ε − β ∆Μ 
 ε2   2 2 
 ε3mech   ε3 − β3 ∆Μ 
 mech  =   (1.46)
 γ 23   γ 23 
 γ13
mech   γ13 
 mech   
 γ12   γ12 
In the case of interaction between moisture strains and thermal strains, superposi-
tion of eqs. (1.38) and (1.44), as well as eqs. (1.39) and (1.45), yields:
 ε1 − α1 ∆Τ − β1 ∆Μ   S11 S12 S13 0 0 0   σ1 

 ε − α ∆Τ − β ∆Μ  S S 22 S 23 0 0 0  σ 
 2 2 2   12   2
 ε3 − α3 ∆Τ − β3 ∆Μ   S13 S 23 S33 0 0 0   σ3 
  =     (1.47)
 γ 23  0 0 0 S 44 0 0   τ 23 
 γ13  0 0 0 0 S55 0   τ13 
     
 γ12   0 0 0 0 0 S66   τ12 
and
 σ1   C11 C12 C13 0 0 0   ε1 − α1 ∆Τ − β1 ∆Μ 

 σ  C C22 C23 0 0 0   ε − α ∆Τ − β ∆Μ 
 2   12  2 2 2 
 σ 3  C13 C23 C33 0 0 0   ε3 − α3 ∆Τ − β3 ∆Μ 
  =     (1.48)
 τ 23   0 0 0 0 0 
C44  γ 23 
 τ13   0 0 0 0 C55 0   γ13 
     
 τ12   0 0 0 0 0 C66   γ12 
1.1.4 Plane stress constitutive relations

Structures made by composite materials are utilized in plates, beams, thin–
walled sheets, etc. The main characteristic of the above components is that the
value of at least one of their characteristic geometric dimensions is much smaller
than the other two dimensions. Therefore, three of the six components of stress are
much smaller than the other three. For plane structural components (e.g., plates
and thin-walled sheets), the in-plane stresses are much larger than the stresses
perpendicular to their plane. Usually, for design purposes, the stress components
perpendicular to the plane of these structures can be set to zero, which can simpli-
fy many problems. According to the above assumption, in a fiber-reinforced plate
with the principal directions shown in Figure 1.2, the stresses σ 3 , τ13 , τ 23 can be
set to zero, yielding the following simplification in the generalized Hooke’s law:
 ε1   S11 S12 S13 0 0 0   σ1 

ε  S σ 
 2  12 S 22 S 23 0 0 0   2
 ε3  S S 23 S33 0 0 0  0
  =  13    (1.49)
 γ 23  0 0 0 S 44 0 0  0
 γ13  0 0 0 0 S55 0  0
     
 γ12   0 0 0 0 0 S66   τ12 
It is obvious from the above equation that:
γ13 = 0 (1.50)
γ 23 = 0 (1.51)
Equations (1.50) and (1.51) show that planes x2 − x3 and x1 − x3 can be consid-
ered as free of shear strains.
Although the normal strain ε3 is not zero, the plane–stresses assumption
described above yields a simplified Hooke’s law involving only σ1 , σ 2 , τ12 and
ε1 , ε2 , γ12
 ε1   s11 s12 0   σ1 
    σ 
 ε2  =  s12 s22 0  2 (1.52)
γ   0 0 s66   τ12 
 12  
where
1
S11 = (1.53)
E1
v12 v
S12 = − = − 21 (1.54)
E1 E2
1
S 22 = (1.55)
E2
1
S66 = (1.56)
G12
Interpolating the plane-stresses assumptions σ 3 = 0 , τ13 = 0 , τ 23 = 0 into

eq. (1.2), the corresponding relation between stresses and strains can be written as:
 σ1   Q11 Q12 0   ε1 
    
 σ 2  = Q12 Q22 0   ε2  (1.57)
τ   0 0 Q66   γ12 
 12  
where the parameters Qij are called reduced stiffnesses, and are given by the
following equations:
C132
Q11 = C11 − (1.58)
C33
C13C23
Q12 = C12 − (1.59)
C33
C232
Q22 = C22 − (1.60)
C33
Q66 = C66 (1.61)
The above equations, with the aid of eqs. (1.25)–(1.33), (1.36), (1.37) and eqs.
(1.4)–(1.15), yield:
E1
Q11 = (1.62)
1 − v12 v21
v12 E2 v E
Q12 = = 21 1 (1.63)
1 − v12 v21 1 − v12 v21
E2
Q22 = (1.64)
1 − v12 v21
Q66 = G12 (1.65)
Taking into account the plane–stresses assumption, equations (1.47) and (1.48),
describing the generalized Hooke’s law for the cases of the effects of free thermal
and moisture strains, can now be written as:
 ε1 − α1 ∆Τ − β1 ∆Μ   s11 s12 0  σ1 
    
 ε2 − α2 ∆Τ − β 2 ∆Μ  =  s12 s22 0  σ2  (1.66)
   0 τ 
 γ12  0 s66   12 
and
 σ1   Q11 Q12 0   ε1 − α1 ∆Τ − β1 ∆ΜΜ

     
 σ 2  = Q12 Q22 0   ε2 − α2 ∆Τ − β 2 ∆Μ  (1.67)
τ   0 0 Q66   γ12 
 12   
where the mechanical strains are given by:
 ε1mech   ε1 − α1 ∆Τ − β1 ∆Μ 
 mech   
 ε2  = =  ε2 − α2 ∆Τ − β 2 ∆Μ  (1.68)
 γ mech   γ12 
 12   
1.1.5 Coordinate transformation of stress and strain components

Composite laminates are made of multiple fiber-reinforced laminae, and each
lamina has its own specific fiber orientation. Therefore, in a multilayered, fiber-re-
inforced material, multiple x1 − x2 − x3 principal coordinate systems exist. On the
other hand, the loads acting on the multilayered laminate have a common direc-
tion for all layers composing the laminate. If we are dealing with an x − y − z or-
thogonal coordinate system (Fig. 1.3) to describe the directions of the loads acting
on the laminate, then the stresses and strains in each principal system x1 − x2 − x3
should be defined with respect to the stresses and strains to the x − y − z system
(global coordinate system).
Figure 1.3 Definition of the principal coordinate system x1 − x2 − x3 with respect to the
global coordinate system x − y − z .
For the case of plane-stress, the above transformation is given by the following
matrix equation:
 σ1   cos 2 θ sin 2 θ 2 sin θ cos θ  σ x 

   2 2   
 σ 2  =  sin θ cos θ −2 sin θ cos θ  σ y  (1.69)
 τ   − sin θ cos θ sin θ cos θ cos 2 θ − sin 2 θ   
 12     τ xy 
where ϑ is the angle defining the fiber orientation with respect to the x axis
(Figure 1.3).
Using the notations m = cos θ and n = sin θ , the above equation is usually
written in the following form:
 σ1  σ x 
   
 σ 2  = [T ]  σ y  (1.70)
τ   
 12   τ xy 
where the transformation matrix [T ] is given by:
 m2 n2 2mn 
[T ] =  n m −2mn 
 2 2
(1.71)
 − mn mn m 2 − n 2 
 
Using the inversed matrix [T ]

−1
 m2 n2 −2mn 
 2 
[T ]−1
= n m 2
2mn  (1.72)
 mn − mn m 2 − n 2 
 
The stresses σ x , σ y , τ xy can be expressed with respect to σ1 , σ 2 , τ12 :
σ x   σ1 
   
 σ y  = [T ]
−1
σ 2  (1.73)
  τ 
 τ xy   12 
1.1.5.1 Transformed reduced compliances

Concerning the strains involved in the plane-stress assumption, the following
transformation equation can be written as:
 ε1   εx 
   
 ε2  = [T ]  ε y  (1.74)
1 γ  1 γ 
 2 12   2 xy 
With the aid of equations (1.70), (1.74), equation (1.52) yields
 εx   S 11 S 12 S 16   σ x 
    
 ε y  =  S 12 S 22 S 26   σ y  (1.75)
γ    
 xy   S 16 S 26 S 66   τ xy 
where
S 11 = S11m 4 + ( 2 S12 + S66 ) n 2 m 2 + S22 n 4 (1.76)
(
S 12 = ( S11 + S 22 − S66 ) n 2 m 2 + S12 n 4 + m 4 ) (1.77)
S 16 = ( 2 S11 − 2 S12 − S66 ) nm3 − ( 2 S 22 − 2 S12 − S66 ) n3 m (1.78)
S 22 = S11n 4 + ( 2 S12 + S66 ) n 2 m 2 + S 22 m 4 (1.79)
S 26 = ( 2 S11 − 2 S12 − S66 ) n3 m − ( 2 S 22 − 2 S12 − S66 ) nm3 (1.80)
(
S 66 = 2 ( 2 S11 + 2 S 22 − 4 S12 − S66 ) n 2 m 2 + S66 n 4 + m 4 ) (1.81)
The above parameters S i j are called transformed reduced compliances, and the
corresponding equation (1.75) is a fundamental equation for analysis of fiber-
reinforced layers. An expanded form of equation (1.75) and the corresponding
S i j parameters are given in the Appendix to Chapter I.
1.1.5.2 Transformed reduced stiffnesses

Taking into account equations (1.70), (1.74), equation (1.57) can now be writ-
ten as:
 σ x   Q11 Q12 Q16   εx 

    
 σ y  = Q12 Q 22 Q 26   ε y  (1.82)
    
 τ xy  Q16 Q 26 Q 66   γ xy 
where
Q11 = Q11m 4 + 2 (Q12 + 2 S66 ) n 2 m 2 + Q22 n 4 (1.83)
(
Q12 = (Q11 + Q22 − 4Q66 ) n 2 m 2 + Q12 n 4 + m 4 ) (1.84)
Q16 = (Q11 − Q12 − 2Q66 ) nm3 + (Q12 − Q22 + 2Q66 ) n3 m (1.84)
Q 22 = Q11n 4 + 2 (Q12 + 2Q66 ) n 2 m 2 + Q22 m 4 (1.86)
Q 26 = (Q11 − Q12 − 2Q66 ) n3 m + (Q12 − Q22 + 2Q66 ) nm3 (1.87)
Q 66 = ( Q11 + Q22 − 2Q12 − 2Q66 ) n 2 m 2 + Q66 ( n 4 + m 4 ) (1.88)
The parameters Q i j correlating the stress with respect to the strains in the global
coordinate system are called transformed reduced stiffnesses. Like the trans-
formed reduced compliances, the transformed reduced stiffnesses vary signifi-
cantly with fiber orientation ϑ .
1.1.6 Transformation of engineering properties

Apart from the engineering properties in the principal coordinate system, en-
gineering properties can also be defined within the x − y − z global coordinate
system.
Considering again a state of plane stress, when a stress σ x is acting on an FRP
plane element (Fig. 1.3), the element stretches in the x direction, contracts in the
y direction, and since the fibers are lying in a direction with orientation ϑ the
right corner angles do not remain as right angles. Denoting the modulus of elas-
ticity in the x direction by Ex , the Hooke’s law for this direction can be written
in the form:
σ x = Ex εx (1.89)
( )
For the above situation σ x ≠ 0 , σ y = 0 , τ xy = 0 equation (1.75) yields
εx = S 11 σ x (1.90)
Therefore, the combination of equations (1.89) and (1.90) provides the definition
of the elasticity modulus in the x direction.
1
Ex = (1.91)
S 11
Taking into account equations (1.76), (1.52–1.56), the above equation can be writ-
ten as:
E1
Ex = (1.92)
 E  E
m 4 +  1 − 2v12  n 2 m 2 + 1 n 4
 G12  E2
The Poisson’s ratio in the x − y direction is defined by the ratio of the contraction
strain ε y in the y direction over the extensional strain εx in the x direction or:
εy
vxy = − (1.93)
εx
Since (for the case of σ x ≠ 0, σ y = 0, τ xy = 0 ) equation (1.75) yields:
ε y = S 12 σ x (1.94)
equation (1.93) with the aid of equations (1.90), (1.94) can be written as:
S 12
vxy = − (1.95)
S 11
Using the definitions of S 12 , S 11 and Sij, the above equation provides the follow-
ing formula:
 E E 
( )
v12 n 4 + m 4 − 1 + 1 − 1  n 2 m 2
 E2 G12 
vxy = (1.96)
 E  E
m +  1 − 2v12  n 2 m 2 + 1 n 4
4
 G12  E2
For evaluating the modulus of elasticity in the y direction, the stress situation
σ x = 0, σ y ≠ 0 , τ xy = 0 will now be considered. In that case the modulus E y is
given by the following formula:
σy
Ey = (1.97)
εy
Since
ε y = S 22 σ y (1.98)
Equation (1.97) yields
1
Ey = (1.99)
S 22
Using the definition of S 22 given by eq. (1.79) and the definitions of Sij given by
equations (1.53)-(1.56), equation (1.99) can now be written as:
E2
Ey = (1.100)
 E  E
m 4 +  1 − 2v12  n 2 m 2 + 2 n 4
 G12  E1
Due to the stress in only the y direction, the Poisson’s ratio v y x is
εx
vy x = − (1.101)
εy
For the situation σ x = 0, σ y ≠ 0 , τ xy = 0 , equation (1.75) yields
εx = S 12 σ y (1.102)
Therefore, with the aid of eqs. (1.98), (1.102), equation (1.101) provides
S 12
vy x = − (1.103)
S 22
Using the definitions of S 12 and S 22 the above equation can now be written as:
 E E 
( )
v21 n 4 + m 4 − 1 + 2 − 2  n 2 m 2
 E1 G12 
v yx = (1.104)
E  E
m +  2 − 2v21  n 2 m 2 + 2 n 4
4
 G12  E1
For the derivation of the formula providing the shear modulus Gxy , the stress situ-
ation σ x = 0 , σ y = 0 , τ xy ≠ 0 will be considered. In that case
τ xy
γ xy = (1.105)
Gxy
However, for above stress situation the equation (1.75) provides
γ xy = S 66 τ xy (1.106)
Therefore, from equations (1.105) and (1.106) the following can be derived:
1
Gxy = (1.107)
S 66
With the aid of the definition of S 66 the above equation yields:
G12
Gxy = (1.108)
 G G 
n + m + 2  2 12 (1 + 2v12 ) + 2 12 − 1 n 2 m 2
4 4
 E1 E2 
A schematic representation of the variation in fiber orientation ϑ of the elastic

properties in the global coordinate system is shown in Figure 1.4 (a)-(d). Portions
of this figure indicate that: (i) the modulus Ex becomes greatest when the fi-
bers are oriented in the x axis (i.e. ϑ = 0 ) and decreases rapidly with ϑ ; (ii) the
modulus E y has its smallest value at ϑ = 0 and increases rapidly as the fiber ori-
entation approaches the angle ± 90 ; (iii) the shear modulus Gxy takes its maxi-
mum value at ϑ = ± 45 and its minimum occurs at ϑ = 0 ; (iv) the maximum
value of vxy occurs in the area 0 < ϑ < 45 and the minimum at ϑ = 90 ; (v) the
Poisson’s ratio v yx has a minimum at ϑ = 0 while its maximum appears in the
area 45 < ϑ < 90 . The above conclusions correspond to the behavior of a typical
graphite-reinforced composite.
1.1.7 Free thermal and free moisture strains in the global coordinate system
(a) Transformation of thermal and moisture expansion coefficients.
Following the concept described in previous paragraphs, the relations of stress
and strain components in the x − y − z system, including the effects of free ther-
mal and moisture strains, can be derived. To this end, the thermal and moisture
expansion coefficients in the global coordinate system will be correlated initially
with the thermal and moisture expansion coefficients in the principal coordinate
system. Inverting equation (1.74) leads to the following formula:
Figure 1.4 Variation with fiber orientation ϑ of: (a) the elastic modulus Ex , (b) the elas-
tic modulus E y , (c) the shear modulus Gxy , (d) the Poisson’s ratios vxy and v yx .
 εx   ε1 
   
 ε y  = [T ]
−1
 ε2  (1.109)
1 γ  1 γ 
 2 xy   2 12 
Since in the case of free thermal strains,
ε1 = α1 ∆Τ (1.110)
ε2 = α2 ∆Τ (1.111)
γ12 = 0 (1.112)
equation (1.109) yields
 εx   α1 
   
 ε y  = [T ]
−1
 α2  ∆Τ (1.113)
1 γ  0
 2 xy   
Considering the following definitions of free thermal strains in the global coordi-
nate system
εx = α x ∆Τ (1.114)
ε y = α y ∆Τ (1.115)
γ xy = α xy ∆Τ (1.116)
Equation (1.113) can now be written as:
 αx   α1 
   
 α y  = [T ]
−1
 α2  (1.117)
1 α  0
 2 xy   
With the aid of equation (1.72), the above equation provides the correlation of the
thermal expansion coefficients in the global coordinate system versus the thermal
expansion coefficients in the principal coordinate system:
α x = α1 cos 2 θ + α2 sin 2 θ (1.118)
α y = α1 sin 2 θ + α2 cos 2 θ (1.119)

α xy = 2 ( α1 − α2 ) cos θ + sin θ (1.120)
A schematic representation of the variation of the above coefficients with fiber

orientation ϑ is shown in Figure 1.5.
Following the same concept, the coefficients of moisture expansion in the global
coordinate system are given by the following equations:
 βx   β1 
   
 β y  = [T ]
−1
 β2  (1.121)
1 β  0
 2 xy   
Figure 1.5 Schematic representation of the variation of thermal expansion coefficients in

the global coordinate system versus fiber orientation ϑ .
yielding
β x = β1 cos 2 θ + β 2 sin 2 θ (1.122)
β y = β1 sin 2 θ + β 2 cos 2 θ (1.123)
β xy = 2 ( β1 − β 2 ) cos θ sin θ (1.124)
It should be recalled that β 1, β 2 are the moisture expansion coefficients in the

principal coordinate system.
(b) Transformation of free thermal and moisture strains.

Using the concept of mechanical strains given by equation (1.66) and the in-
verted forms of equations (1.117), (1.121), i.e.,
 α1   αx 
   
 α2  = [T ]  α y  (1.125)
0 1 α 
   2 xy 
and
 β1   βx 
   
 β 2  = [T ]  β y  (1.126)
0 1 β 
   2 xy 
the following matrix equation can be obtained:
 ε1   αx   β x   S11 S12 0   σ1 
   
 2
ε = − [ ] y 
T α ∆Τ − [ ]  β y  =  S12
T S 22 0 
  
 σ 2  (1.127)
γ  1 α  1 β   0  τ 
0 2 S 66 
1
 12   2 xy   2 xy    12 
Taking into account equations (1.69), (1.74) the above equation yields
 εx − α x ∆Τ − β x ∆M   S 11 S 12 S 166  σ x 
     
 ε y − α y ∆Τ − β y ∆M  =  S 12 S 22 S 26  σ y  (1.128)
     
 γ x y − α x y ∆Τ − β x y ∆M   S 16 S 26 S 66   τ xy 
The inverse of the above equation provides the stress components σ x , σ y , τ xy ver-
sus the mechanical strains can be expressed as:

εmech
x = εx − α x ∆Τ − β x ∆Μ , εmech
y = ε y − α y ∆Τ − β y ∆Μ , γ mech
xy = γ x y − α x y ∆Τ − β x y ∆Μ
 σ x   Q11 Q12 Q16   εx − α x ∆Τ − β x ∆M 

     
 σ y  = Q12 Q 22 Q 26   ε y − α y ∆Τ − β y ∆M  (1.129)
     
 τ xy  Q16 Q 26 Q 66   γ x y − α x y ∆Τ − β x y ∆M 
1.2 Mechanical Behavior of Laminates

1.2.1 Classical lamination theory
As was already stated, composite laminates are composed of multiple fiber-re-
inforced laminae (Fig. 1.6). Each lamina has a very small thickness and usually a
different fiber orientation. The stacking sequence of layers, their fiber orientation
and their number influence the mechanical behavior of the multilayered medium.
Since multiple combinations of stacking arrangements of layers are possible,
the nomenclature associated with the definition of the manner in which the lami-
nate is structured will be presented first.
1.2.2 Laminate nomenclature

A multilayered composite plate consisting of N layers (Figure 1.7) will be used
to describe the laminate nomenclature. The x axis of the global coordinate system
is located in the geometric midplane of the laminate, while the axes z and y are
aligned with the directions of the thickness and the width respectively (Fig. 1.7).
The specification of the fiber angles of the various layers starts with layer 1.
For example the four–layered laminate described in Table 1.1 is denoted as a
[0 / 45 / 90 / 0] laminate.
Table 1.1
Example of the stacking sequence of a laminate.
Layer’s location Fiber orientation θ
layer 1 0
layer 2 45°
layer 3 90°
layer 4 0
Mechanical Behavior of Laminates 25
Figure 1.6 Multiple laminae composing a laminate.
Figure 1.7 Definition of the global coordinate system for a multilayered FRP laminate.
In cases where the stacking sequence for the side z ≥ 0 is a mirror image of
the stacking sequence of the side z ≤ 0 , the stacking notation can be abbrevi-
ated. For example, the symmetric six-layered laminate of Table 1.2 is denoted as
a [ 0 / 30 / 60]s laminate, where the subscript s means symmetric. In order for a
laminate to be characterized as symmetric, apart from the symmetric order of lay-
ers, the material properties, fiber orientation and thickness of the layer at a specific
location within the side z ≥ 0 should be identical to the material properties, fiber
orientation and thickness of the corresponding layer at the same location of the
side z ≤ 0 .
In cases where the symmetric laminate involves adjacent layers of opposite
orientation, the stacking notation can also be abbreviated. For example the six-
layered laminate of Table 1.3 can be defined using the notation [ ±30 / 0]s
Finally, when a group of layers is repeated within a symmetric laminate, fur-
ther shorthand notation is used. For the twelve-layered symmetric laminate de-
scribed in Table 1.4, the notation [ (±60 / 0) 2 ]s may be used.
Table 1.2
Example of stacking sequence of a symmetric six-layered laminate.
Layer 1 0
Layer 2 30°
Layer 3 60°
Layer 4 60°
Layer 5 30°
Layer 6 0
Table 1.3
Example of stacking sequence of a symmetric six-layered laminate
exhibiting adjacent layers of opposite orientation.
Layer 1 +30°
Layer 2 -30°
Layer 3 0
Layer 4 0
Layer 5 -30°
Layer 6 +30°
Table 1.4
Example of a stacking sequence of a symmetric twelve-layered
laminate displaying repeated groups of layers.

Layer 1 +60° Repeated
Layer 2 -60° group
of layers
Layer 3 0
Layer 4 +60°
Layer 5 -60°
Layer 6 0
Layer 7 -60°
Layer 8 +60°
Layer 9 0
Layer 10 -60°
Layer 11 +60°
Layer 12 0
In cases in the above notation where the subscript s is missing, the notation
[(±60 / 0) ]
2 represents the unsymmetrical laminate [ +60 / −60 / 0 / +60 / −60 / 0]
1.2.3 The Kirchhoff Assumption

The simplest laminated theory is an extension of the Kirchhoff isotropic plate
theory to laminated composite plates. According to this theory, the straight lines
normal to the xy plane (Fig. 1.7) before deformation remain straight and normal
to the mid-surface z = 0 after deformation. Therefore, on the basis of this as-
sumption, both transverse shear and transverse normal effects are neglected.
According to Figure 1.8 the movement of a point A in the x axis due to defor-
mation in the x-z plane is given by:
∂w0
u = u0 − z (1.130)
∂x
while its vertical movement is:
w = w0 (1.131)
Figure 1.8 Laminated plate deformation as viewed in the x-z plane.
In the above equations u0 represents the horizontal movement of the reference

point O located in geometric the mid-plane, w0 represents its vertical movement,
and ∂w0 / ∂x represents the slope in the x-z plane of the x-axis at the point O after
the deformation.
Following the same concept, the movement of the point A in the direction of
the y-axis (Fig. 1.7) can be given by:
∂w0
u = u0 − z (1.132)
∂y
where u 0 is the deflection of reference point O in the direction of the y-axis, and
∂wo / ∂x is the slope in the x-y plane of the y–axis at the point O after deformation.
1.2.4 Laminate strains

With the aid of eqs. (1.130)–(1.132), following the well-known [3] strain-
displacement relations:
∂u
εx = (1.133)
∂x
∂υ
εy = (1.134)
∂y
∂w (1.135)
εz =
∂z
∂υ ∂u
γ xy = + (1.136)
∂x ∂y
∂w ∂u
γ xz = + (1.137)
∂x ∂z
∂w ∂υ
γ yz = + (1.138)
∂y ∂z
yields:
εx = ε0x + zk x0 (1.139)
ε y = ε0y + zk y0 (1.140)
εz = 0 (1.141)
γ xy = γ 0xy + zk xy0 (1.142)
γ xz = 0 (1.143)
γ yz = 0 (1.144)
In the above equations (1.139)–(1.144) the following definitions were used:
∂u
ε0x = (1.145)
∂x
∂υ
ε0y = (1.146)
∂y
∂υ0 ∂u
γ 0xy = + (1.147)
∂x ∂y
∂ 2 w0
k x0 = − (1.148)
∂x 2
∂ 2 w0
k y0 = − (1.149)
∂y 2
∂ 2 w0
k xy0 = −2 (1.150)
∂ x∂ y
0
The strains ε0x , ε y , γ xy are called reference surface extensional strain in the x
0
direction, reference surface extensional strain in the y direction and reference sur-
face in-plane shear strain respectively. The quantities k x0 and k y are the curva-
0
tures of the reference surface in the x and y directions respectively, while the
quantity k xy0 is the reference twisting curvature.
1.2.5 Laminate stresses

If the plane stresses assumption is adopted for each lamina, the stresses of each
point of a lamina can be determined with the aid of equations (1.139), (1.140),
(1.142) by the already known stresses-strain relation in the global coordinate
system:
σx   Q11 Q12 Q16   ε0 + zk 0 
     0x x
0 
 σ y  = Q12 Q22 Q26   ε y + zk y  (1.151)
τ     γ 0 + zk 0 
 xy  Q16 Q26 Q66   xy xy 
Since the strains and the reduced stiffnesses Q ij are functions of the location
z of each lamin, and since the material properties controlling the value of Q ij
are generally different for adjacent layers, the distribution of stresses through the
thickness of the laminate is expected to be incremental.
1.2.6 Laminate stiffness matrix

Input data for the design of composite structures are often the forces and mo-
ments acting on the boundary surfaces of laminates. The goal of the designer is
to calculate the stress field on each layer in order to apply a failure criterion. To
achieve this goal, formulae correlating the stress resultants, i.e., the loads and
moments required to produce the specified mid-plane deformations—with the
mid-plane strains and curvatures should be derived. Using the notations found
in Fig. 1.9 (a) (b), the force and moments resultants are given by the following
definitions:
Figure 1.9 Nomenclature for (a) force resultants, and (b) moment resultants.
H
2
NX ≡ ∫H
σ x dz (1.152)
−
2
H
2
Ny ≡ ∫ H
σ y dz (1.153)
−
2
H
2
N xy ≡ ∫H
τ xy dz (1.154)
−
2
H
2
Μx ≡ ∫ H
σ x zdz (1.155)
−
2
H
2
Μy ≡ ∫ H
σ y zdz (1.156)
−
2
H
2
Μ xy ≡
H
∫ τ xy zdz (1.157)
−
2
The unit of the force resultants is force per unit length, while the unit of the mo-
ment resultants is moment per unit length. For example, the resultants N x , Μ χ
are
∧
Nx
Nx = (1.158)
AB
∧
Mx
Mx = (1.159)
AB
∧ ∧
where N x , M x are the force and moment acting on the segment AB in Figure 1.9,
and AB is the length of the segment AB.
With the aid of eq. (1.151), equations (1.152)–(1.154) can be written as:
 H2   H2   H2   H2 
  0   0   0  
N x =  ∫ Q11dz  εx +  ∫ Q11 zdz  k x +  ∫ Q12 dz  ε y +  ∫ 12  k y0 +
Q zdz
− H  − H  − H  − H 
 2   2   2   2 
 H2   H2 
  0   0
 ∫ Q16 dz  γ xy +  ∫ Q16 zdz  k xy (1.160)
− H  − H 
 2   2 
 H2   H2  H   H2 
  0   0 2  0  
N y =  ∫ Q12 dz  εx +  ∫ Q12 zdz  k x +  ∫ Q22 dz  ε y +  ∫ Q22 zdz  k y0 +
− H  − H  − H  − H 
 2   2   2   2 
 H2   H2 
  0   0
 ∫ Q26 dz  γ xy +  ∫ Q26 zdz  k xy (1.161)
− H  − H 
 2   2 
 H2   H2  H   H2 
  0   0 2  0  
N xy =  ∫ Q16 dz  εx +  ∫ Q16 zdz  k x +  ∫ Q26 dz  ε y +  ∫ Q26 zdz  k y0 +
− H  − H  − H  − H 
 2   2   2   2 
 H2   H2 
  0   0
 ∫ Q66 dz  γ xy +  ∫ Q66 zdz  k xy (1.162)
− H  − H 
 2   2 
H
+
2
For a laminate with N layers each integral ∫QH
ij dz has the form:
−
2
H
2
∫ Q dz
H
ij = Q ij1 ( z1 − z0 ) + Q ij 2 ( z2 − z1 ) + Q ij3 ( z3 − z2 ) + …. + Q ijk ( zk − zk −1 ) +
−
2
….+ Q ijN ( z N − z N −1 ) (1.163)
Using the notation
H
+
2
Aij = ∫ Q dz
H
ij (1.164)
−
2
equation (163.1) can be written in the following simpler form:

N
Aij = ∑ Q ijk ( zk − zk −1 ) (1.165)
k =1
Following the same concept, the integrals of the form:

+H 2
Bij = ∫
−H 2
Q ij zdz (1.166)
yields
1 N
Bij = ∑ Q ( zk2 − zk2−1 )
2 k =1 ijk
(1.167)
Therefore, with the aid of equations (1.165), (1.167), the force resultants given by
equations (1.160)-(1.162) can be written in the following matrix form:
 Nx   A11 A12 A16   ε0x   B11 B12 B16   k x0 

    0  0
 N y  =  A12 A22 A26  
 ε y  +  B12 B22 B26   k y  (1.168)
N   A16  γ0  k 0 
 xy  A26 A66   xy   B16 B26 B66   xy 
Combining equations (1.155) – (1.157) with equation (1.151), the following for-
mulas for the moment resultants can be obtained:
 H2   H2   H2   H2 
  0   0   0  
M x =  ∫ Q11 zdz  εx +  ∫ Q11 z dz  k x +  ∫ Q12 zdz  ε y +  ∫ Q12 z dz  k y0 +
2 2
− H  − H  − H  − H 
 2   2   2   2 
 H2   H2 
  0   0
 ∫ Q16 zdz  γ xy +  ∫ Q16 z dz  k xy
2
(1.169)
− H  − H 
 2   2 
 H2   H2   H2   H2 
  0   0   0  
M y =  ∫ Q12 zdz  εx +  ∫ Q12 z dz  k x +  ∫ Q22 zdz  ε y +  ∫ Q22 z dz  k y0 +
2 2
− H  − H  − H  − H 
 2   2   2   2 
 H2   H2 
  0   0
 ∫ Q26 zdz  γ xy +  ∫ Q26 z dz  k xy
2
(1.170)
− H  − H 
 2   2 
 H2   H2   H2   H2 
  0      
M xy =  ∫ Q16 zdz  εx +  ∫ Q16 z dz  k x +  ∫ Q26 zdz  ε y +  ∫ Q26 z dz  k y0 +
2 0 0 2
− H  − H  − H  − H 
 2   2   2   2 
 H2   H2 
  0   0
 ∫ Q66 zdz  γ xy +  ∫ Q66 z dz  k xy
2
(1.171)
− H  − H 
 2   2 
Using the notation
H 2
Dij = ∫
−H 2
Q ij z 2 dz (1.172)
equations (1.169)-(1.171) can now be written as:
 Mx   B11 B12 B16   ε0x   D11 D12 D16   k x0 

   0   0

 M y  =  B12 B22 B26   ε y  +  D12 D22 D26   k y  (1.173)
M   γ0   k 0 
 xy   B16 B26 B66   xy   D16 D26 D66   xy 
where
1 N
Dij = ∑ Qijk ( zk3 − zk3−1 )
3 k =1
(1.174)
Superposition of equation (1.168) and (1.173) yields
 N x   Α11 Α12 Α16 B11 B12 B16   ε0x 

N    0
 y   Α12 Α 22 Α 26 B12 B22 B26   εy 
 N xy   Α16 Α 26 Α66 B16 B26 B66   γ 0xy 
  =   0 (1.175)
 M x   B11 B12 B16 D11 D12 D16   kx 
 M y   B12  k y0 
B22 B26 D12 D22 D26   0
   
 M xy   B16 B26 B66 D16 D26 D66   k xy 
In the above equation the 6 × 6 matrix consisting of the components Aij, Bij, Dij is
called the laminate stiffness matrix or ABD matrix. Inversion of the above equa-
tion provides the relation of the strains and curvatures with the force and moment
resultants :
 ε0x   a11 a12 a16 b11 b12 b16   Nx 

 0  N 
 ε y   a12 a22 a26 b21 b22 b26   y
 γ 0xy   a16 a26 α66 b61 b62 b66   N xy 
 0 =    (1.176)
 k x   b11 b12 b61 d11 d12 d16   Mx 
 k y0   b b22 b62 d12 d 22 d 26  My 
 0   12   
 k xy   b16 b26 b66 d16 d 26 d 66   M xy 
The 6 × 6 matrix consisting of the components αij is given by:
 a11 a12 a16 b11 b12 b16   Α11 Α12 Α16 B11 B12 B16 
−1
a b26   Α12
 12 a22 a26 b21 b22
 Α 22 Α 26 B12 B22 B26 
 a16 a26 α66 b61 b62 b66   Α16 Α 26 Α66 B16 B26 B66 
=  = 
 b11 b12 b61 d11 d12 d16   B11 B12 B16 D11 D12 D16 
 b12 b22 b62 d12 d 22 d 26   B12 B22 B26 D12 D22 D26 
   
 b16 b26 b66 d16 d 26 d 66   B16 B26 B66 D16 D26 D66 
(1.177)
and is called the laminate compliance matrix or abd matrix. Knowing the force
and moment resultants acting on the bounding surface of a laminate, the principal
stresses σ1, σ2, τ12 of each layer can now be calculated by following the algorithm
shown in Figure 1.10.
1.2.7 Classification of laminates

Depending on the specific stacking sequence of the layers composing a lami-
nate, the following types can be used in engineering practice:
(a) Symmetric laminates are laminates where: (i) the stacking sequence of the
upper half side (with respect to the geometric mid-plane) is a mirror image
of the stacking sequence of the lower half side, and (ii) the material proper-
ties, fiber orientation and thickness of a pair of symmetric (with respect to
the geometric mid-plane) layers are identical.
(b) Symmetric Balanced laminates are laminates consisting of pairs of symmet-
ric (with respect to the geometric mid-plane) layers with identical material
properties and thickness but opposite fiber orientation.
(c) Cross-Ply laminates are ones where every layer has its fibers oriented at
either 0° or 90°.
(d) Symmetric Cross-Ply laminates meet the definitions of (a) and (c) above.
Taking into account the above definitions, the ABD matrix can be simplified.
These simplifications yield the following formulations for equation (1.175):
Figure 1.10 Algorithm for calculation of the principal stresses on each layer of a laminate.
For symmetric laminates:
 N x   Α11 Α12 Α16 0 0 0   ε0x 

N    0
 y   Α12 Α 22 Α 26 0 0 0   εy 
 N xy   Α16 Α 26 Α66 0 0 0   γ 0xy 
  =   0 (1.178)
 Mx   0 0 0 D11 D12 D16   kx 
My   0  k y0 
0 0 D12 D22 D26   0
   
 M xy   0 0 0 D16 D26 D66   k xy 
For symmetric balanced laminates:
 N x   Α11 Α12 0 0 0 0   ε0x 

N    0
 y   Α12 Α 22 0 0 0 0   εy 
 N xy   0 0 Α66 0 0 0   γ 0xy 
  =   0 (1.179)
 Mx   0 0 0 D11 D12 D16   kx 
My   0  k y0 
0 0 D12 D22 D26   0
   
 M xy   0 0 0 D16 D26 D66   k xy 
For cross-ply laminates:
 N x   Α11 Α12 0 B11 0 0   ε0x 

N    0
 y   Α12 Α 22 Α 26 0 B22 0   εy 
 N xy   0 0 Α66 0 0 0   γ 0xy 
  =   0 (1.180)
 M x   B11 0 0 D11 D12 0   kx 
My   0  k y0 
B22 0 D12 D22 0   0
   
 M xy   0 0 0 0 0 D66   k xy 
For symmetric cross-ply laminates:
 Nx   Α11 Α12 0 0 0 0   ε0x 

N  Α   0
 y  12 Α 22 0 0 0 0   εy 
 N xy   0 0 Α66 0 0 0   γ 0xy 
  =   0 (1.181)
 Mx   0 0 0 D11 D12 0   kx 
My   0  k y0 
0 0 D12 D22 0   0
   
 M xy   0 0 0 0 0 D66   k xy 
The Tsai–Wu Failure Criterion 39
For the special case where the laminate is a single isotropic layer of thickness H,
the equation (1.175) yields
 EH vEH 
1 − v 2 0 0 0 0 
1 − v2
 
 vEH EH
0 0 0 0 
 Nx  1 − v 2 1 − v2   ε0x 
N     0
EH
 y  0 0 0 0 0   εy 
 N xy   2(1 + v)   γ 0xy 
  =  EH 3 vEH 3
  0 
 Mx   0 0 0 0   kx 
My   12(1 − v ) 12(1 − v 2 )
2   k y0 
     0
M
 xy   vEH 3 EH 3   k xy 
 0 0 0
12(1 − v 2 ) 12(1 − v 2 )
0 
 
 EH 3 
 0 0 0 0 0 
 24(1 + v) 
(1.182)
In the above equation E, v are the modulus of elasticity and the Poisson’s ratio
respectively of the isotropic layer.
1.3 The Tsai–Wu Failure Criterion

From a micromechanical point of view, the failure of fiber-reinforced mate-
rials is a complex and multi-parametric process. However, explaining the mi-
cromechanical approach to composites failure is beyond the scope of this book.
Macroscopically, models taking into account the applied stress components in
the principal coordinate system and the corresponding tensile and compressive
strengths, have been developed in order to estimate the limits of stresses that
cause failure.
The above models are called “failure criteria.”
Since “failure” is considered to be the loss of integrity of the material itself
[ e.g., 1 ] the stress situation yielding buckling is not covered by the failure criteria.
Generally speaking, the main target of a designer aiming to estimate the safety
factor of a composite structure is the determination (e.g., according to the algo-
rithm of Figure 1.10) of the principal stresses. In analogy with the failure criteria
for isotropic materials, a number of criteria for composites have been developed
in order to answer questions such as which principal stress component controls
failure.
Does σ1 or σ 2 or τ12 or a combination of these play the predominant role in

failure?
Among existing failure criteria for composites, the Tsai-Wu [4] criterion seems
to be the most popular for design purposes and will be adopted for the case studies
in this book. Based on a concept similar to that followed for the derivation of the
von Mises criterion for isotropic materials, the Tsai-Wu criterion for composites,
assuming the plane–stresses assumption, is given by the following formula:
F1 σ1 + F2 σ 2 + F11 σ12 + F22 σ 22 + F66 τ12

2
− F11 F22 σ1 σ 2 ≤ 1 (1.183)
where
 1 1 
F1 =  Τ + C  (1.184)
 σ1 σ1 
 1 1 
F2 =  Τ + C  (1.185)
 σ2 σ2 
2
 1 
F66 =  F  (1.186)
 τ12 
1
F11 = − (1.187)
σ σ1C
Τ
1
1
F22 = − (1.188)
σ Τ2 σ C2
In equations (1.184)–(1.188) the following is the case:
σ1C : compressive failure stress in the x1 direction

σ1Τ : tensile failure stresses in the x1 direction
σ C2 : compressive failure stresses in the x2 direction
σ Τ2 : tensile failure stresses in the x2 direction
τ12F : shear failure stresses in the x1-x2 plane
The Tsai–Wu Failure Criterion 41
In Figure 1.11 the bounding surface including the allowable values of σ1 , σ 2 , τ12
in σ1 - σ 2 - τ12 Cartesian system is schematically represented, while Figure 1.12
shows the corresponding area of σ1 , σ 2 values for the case where τ12 = 0. In both
figures it can be shown that the corresponding 3-D and 2-D ellipsoids are very
long (in the direction of the σ1 axis), and very slender (in the direction of the σ2
and/or τ12 axis), indicating dependency on the high strength of the fibers.
Figure 1.11 Schematic 3-D graphical representation of Tsai-Wu failure criterion.
Figure 1.12 Schematic 2-D graphical representation of Tsai–Wu criterion for the case of
τ12 = 0.
Appendix I
General form of compliance matrix  S i j 
 εx   S 11 S 12 S 13 S 14 S 15 S 16  σ x 
ε    σ 
 y   S 21 S 22 S 23 S 24 S 25 S 26   y
 εz   S 31 S 32 S 33 S 34 S 35 S 36 
  σ z 

 =  
 2 γ yz   S 41 S 42 S 43 S 44 S 45 S 46   σ yz 
2 γ   σ 
 xz   S 51 S 52 S 53 S 54 S 55 S 56   xz 

2 γ xy   S 61 S 62 S 63 S 64 S 65 S 66   σ xy 

Compliance coefficients:
S 11 = S11 cos 4 θ − 2 S16 cos3 θ sin θ + ( 2 S12 + S66 ) cos 2 θ sin 2 θ − 2 S 26 cos θ sin 3 θ
+ S 22 sin 4 θ
S 12 = S12 cos 4 θ + ( S16 − S 26 ) cos3 θ sin θ + ( S11 + S 22 − S66 ) cos 2 θ sin 2 θ
+ ( S 26 − S16 ) cos θ sin 3 θ + S12 sin 4 θ
S 13 = S13 cos 2 θ − S36 cos θ sin θ + S 23 sin 2 θ
S 16 = S16 cos 4 θ + ( 2 S11 − 2 S12 − S66 ) cos3 θ sin θ + 3 ( S 26 − S16 ) cos 2 θ sin 2 θ
+ ( S66 − 2 S12 − 2 S 22 ) cos θ sin 3 θ − S 26 sin 4 θ
S 22 = S 22 cos 4 θ + 2 S 26 cos3 θ sin θ + ( 2 S12 + S66 ) cos 2 θ sin 2 θ
+ 2 S16 cos θ sin θ3 + S11 sin 4 θ
S 23 = S 23 cos 2 θ + S36 cos θ sin θ + S13 sin 2 θ
S 26 = S 26 cos 4 θ + ( 2 S12 − 2 S 22 + S66 ) cos3 θ sin θ + 3 ( S16 − S 26 ) cos 2 θ sin 2 θ
+ ( 2 S11 − 2 S12 + S66 ) cos θ sin 3 θ − S16 sin 4 θ
S 33 = S33
(
S 36 = 2 ( S13 − S 23 ) cos θ sin θ + S36 cos 2 θ − sin 2 θ )
( ) (
S 66 = S66 cos 2 θ − sin 2 θ + 4 ( S16 − S 26 ) cos 2 θ − sin 2 θ cos θ sin θ )
2
+ 4 ( S11 + S 22 − S12 ) cos 2 θ sin 2 θ

S 44 = S 44 cos 2 θ + 2 S 45 cos θ sin θ + S55 sin 2 θ
( )
S 45 = S45 cos 2 θ − sin 2 θ + ( S55 − S 44 ) cos θ sin θ
2 2
S 55 = S55 cos θ + S 44 sin θ − 2 S 45 cos θ sin θ
References 43
S 14 = S14 cos3 θ + ( S15 − S 46 ) cos 2 θ sin θ + ( S 24 − S56 ) cos θ sin 2 θ + S 25 sin 3 θ

S 15 = S15 cos3 θ − ( S14 − S56 ) cos 2 θ sin θ + ( S25 − S46 ) cos θ sin 2 θ − S 24 sinn 3 θ
S 24 = S 24 cos3 θ + ( S 25 + S 46 ) cos 2 θ sin θ + ( S14 + S56 ) cos θ sin 2 θ + S15 sin 3 θ
S 25 = S 25 cos3 θ + ( − S 24 + S56 ) cos 2 θ sin θ + ( S15 − S 46 ) cos θ sin 2 θ − S 214 sin 3 θ
S 34 = S34 cos θ + S35 sin θ
S 35 = S35 cos θ − S34 sin θ
S 46 = ( 2 S14 − 2 S 24 + S56 ) cos 2 θ sin θ + ( 2 S15 − 2 S 25 + S 46 ) cos θ sin 2 θ
+ S 46 cos3 θ + S56 sin 3 θ
S 56 = ( 2 S15 − 2 S 25 + S 46 ) cos 2 θ sin θ + ( 2 S 24 − 2 S14 − S56 ) cos θ sin 2 θ
+ S56 cos3 θ + S 46 sin 3 θ
Si j = S ji
References
[1] Hyer M., Stress analysis of fiber reinforced composite materials, DEStech
Publications, 2009.
[2] Reddy J.N., Mechanics of laminated composite plates and shells, CRC
Press, 2004.
[3] Timoshenko S.P., and Goodier J.N., Theory of elasticity, McGraw-Hill,
1970.
[4] Tsai S.W., and Wu E.M., “A general theory of strength for anisotropic ma-
terials,” Journal of Composite Materials 5 (1971), pp. 58–80.
Chapter 2
Classification, Properties and Production

Technology of FRP Materials
Composite FRP materials consist of a polymer matrix material reinforced by

fibers. Since the strength of the fibers is much higher than the strength of the ma-
trix, FRP materials have directionally dependent properties.
2.1 The Composite Matrix Material

For the polymer matrices, two main classes of resins are used: thermosets and
thermoplastics. Thermosetting resins, such us epoxies, differ from thermoplas-
tics concerning their behavior during heating. At elevated temperatures, thermo-
sets transform their microstructure into an irreversible molecular chain (curing)
yielding an unchanged shape. Heating of thermosets after curing does not cause
melting. Actually, thermosets continue to retain their shape until their thermal de-
composition at very high temperatures. On the other hand, thermoplastics, such as
polyethylene, become malleable at high temperatures and solidify when cooled.
When reheated above a lower forming temperature, thermoplastics can be re-
shaped. This behavior of thermoplastics is important, since it allows them to be
repaired. For fabrication of fibers, three types of materials are in common use:
carbon (or graphite), glass and synthetics (e.g., Kevlar). The above materials are
characterized by high strength, thermal stability and low density.
45
46 Production Technology of FRP Materials
2.1.1 Thermosets
Due to their valuable properties, such as retention of mechanical behavior in
hot and moist conditions, good chemical resistance, good dimensional stability,
low processing temperatures, excellent fiber impregnation, low melt viscosity,
low cost etc., thermosets have become the most common type of resin used for
matrix fabrication for composites. Thermosets include the following types of
resins: (a) polyester resins, (b) vinyl ester resins, (c) bisphenol fumerate resins,
(d) liquid epoxy resins, (e) solid epoxy resins, (f) polyurethane resins, (g) furane
resins, (h) phenolic resins and (i) chlorenic resins. The main characteristics [1]
of the above types of resins used for thermosetting matrices are summarized in
Table 2.1, while in Table 2.2 typical values of the physical and mechanical proper-
ties of resins used in the filament winding process are presented.
Table 2.1
Main characteristics of thermosetting resins.
Types of resins used for Main characteristics
thermosets
Polyester resins (a) Orthophthalic polyester resins: Suitable for applications
requiring limited resistance to temperature or chemicals (e.g.,
water industry). It is the cheapest thermosetting resin.
(b) Isophthalic polyester and terephthalic resins: They offer
better all-around performance and molding properties com-
pared to orthophthalic resins.
Vinyl ester resins Suitable for mild to severe chemical applications. Offer very
high mechanical properties allied to an excellent corrosion
resistance.
Bisphenol fumerate resins Provide excellent corrosion resistance to strong acids and the
highest resistance to alkalis at elevated temperatures.
Liquid epoxy resins Suitable for manufacture of adhesives, laminates and coat-
ings. Widely used for manufacture of high-performance
pipe systems and pressure vessels for water treatment and
handling.
Solid epoxy resins Suitable for paints and powder coatings.
Polyurethane resins Offer high impact resistance, excellent bonding properties,
very high toughness and abrasion resistance. It is suitable for
coatings.
Furane resins Offer excellent resistance to alkaline solutions and acids
containing chlorinated solvents. However, furanes provide
poor mechanical properties.
Phenolic resins Provide excellent resistance to heat. Almost non-flammable,
phenolics also generate very limited smoke or toxic fumes.
Chlorenic resins Deliver reliable behavior in high operating temperatures and
highly oxidizing environments.
The Composite Matrix Material 47
Table 2.2
Typical properties of thermosetting resins used in filament winding process [1].
Thermosetting Specific gravity Tensile strength Tensile Flexural
resin (MPa) elongation (%) modulus (GPa)
Orthophthalic 1.10 70 2.5 3.8
polyester
Terephthalic 1.10 70 2.5 3.8
polyester
Isophthalic 1.10 70 2.5 3.9
polyester
Vinyl ester 1.11 80 5.0 3.2
Furane 1.10 36 1.0 4.1
Phenolic 1.25 40 2.0 4.0
Chlorenic 1.18 56 2.0 3.8
2.1.2 Thermoplastics
Thermoplastic resins can be amorphous or semi-crystalline. The semi-
crystalline type contains both amorphous and crystalline phases. Since thermo-
plastics are not cross-linked, they can be processed quickly, thereby reducing
their manufacturing cost. From a mechanical point of view, thermoplastics can
display and undergo large deformations before final fracture, thus offering higher
toughness compared to thermosets. Moreover, under graduated loading condi-
tions their deformation is time-dependent, due to creep. Since today new ther-
moplastics are obtained by modification or mixing of existing polymers, the full
range of these materials is extremely wide. The characteristics of a few main types
of thermoplastics, their abbreviations, and selected trade names are summarized
in Table 2.3, while Table 2.4 provides the mechanical and physical properties of
widely used thermoplastic polymers [2].
Table 2.3
Main characteristics of thermoplastics used for composite materials.
Type of thermoplastics Abbreviation Main characteristics
Poly-Benzimidazoles PBIs Offers good stability after aging. They are well
suited for use at temperatures up to 250°C but at
higher temperatures oxidate degradation occurs
accompanied with strength reduction.
Poly-Phenylene Sulfide PPS Offers excellent chemical and thermal stability.
Poly-Ether PEEK Excellent thermoplastics for engineering applica-
Ether Ketone tions due to good mechanical behavior. However,
their use for high-performance composites is limited
due to inadequate chemical resistance and adhesion.
Table 2.4
Physical and mechanical properties of widely used thermoplastic polymers.
Thermoplastic Density Tensile Tensile Tensile
Polymers (Kg/m3) strength (MPa) Elongation (%) Modulus (GPa)
Poly-Phenylene 1340 70–75 3 3.3
Sulfide(PPS)
Poly-Ether-Ether- 1320 92–100 150 —
Ketone(PEEK)
Poly-Sulfone(PS) 1240 70–75 50–100 2.5
Poly-Propylene(PP) 900 25–38 300 1.0–1.4
Nylon 6,6 (NYLON) 1140 60–75 40–80 1.4–2.8
Poly-Carbonate (PC) 1060–1200 45–70 50–10 2.2–2.4
Poly-Ether Imide 1270 105 60 3
(Ultem)
Poly-Amide Imide 1400 95–185 12–18 5
(Torlon)
2.2 Fiber Materials

2.2.1 Glasses
Glass fibers are low-cost fibers suitable for piping applications (conveying cor-
rosive fluids), vessels, crafts, playground equipment etc. Their microstructure is
based on silica SiO2 which in the form of a polymer (SiO2)n does not melt but
softens progressively up to 2000oC. Although silica is a valuable material for en-
gineering applications, high temperatures are required to form the glass fibers.
Improvements in glass processing technology have led to the following types of
glasses suitable for composites manufacturing:
(i) A-Glass: It is a soda-lime glass, which was the first used and is still retained
for minor applications.
(ii) E-Glass: E-glass is a borosilicate glass, which exhibits very good corro-
sion resistance and is suitable for operating in water and mild chemical
environments.
(iii) C-Glass: C-glass is an improvement of E-Glass, providing better durability
when exposed to acids and alkalis.
(iv) S-Glass: S-glass exhibits increased strength and stiffness and is suitable for
high-performance applications.
Physical and mechanical properties of the main types of glass used for fiber
fabrication are presented in Table 2.5 [2].
Fiber Materials 49
Table 2.5
Properties of the main glass fibers.
Property E-Glass C-Glass S-Glass
Tensile strength (MPa) 3450 3160 4590
Tensile modulus (GPa) 72.4 68.9 85.5
Elongation (%) 1.8–3.2 4.8 5.7
3
Density (Kg/m ) 2541 2492 2492
Diameter (μm) 8–13 — 10
2.2.2 Carbon fibers

The fabrication of carbon fibers starts from a precursor fiber, mainly made
from polyacrylonitril (PAN) or from pitch. Early precursor fibers were made from
rayon. Phenolics, polyvinylalcohols or polyimides may also be used but only for
limited cases. Although the final mechanical properties of carbon fibers are not
significantly affected by the type of precursor, their mechanical properties are
strongly influenced by processing techniques. Generally, by carbonizing organ-
ic precursor fibers, and then graphitizing them at very high temperatures, high
modulus carbon fibers can be produced. As provided by manufacturers, ranges
of the values of physical and mechanical properties of carbon fibers are given in
Table 2.6 [2].
Table 2.6
Values of physical and mechanical properties of carbon fibers.
Property PAN Pitch Type-P
Intermediate High modulus Ultra high

modulus modulus
Tensile strength 2410–2930 2070–2900 1720 1720
(MPa)
Tensile modulus 228–276 331–400 517 345

(GPa)
Elongation (%) 1.0 0.5 0.3–0.4 0.4–0.9

Density (Kg/m3) 1780–1820 1670–1900 1860 2020
Diameter (μm) 8–9 7–10 7–10 10–11
2.2.3 Synthetic fibers

Synthetic (or polymeric) fibers have found increased use in the fabrication of
FRP composite materials due to their good chemical resistance and low density.
However, their maximum operating temperature is relatively low and varies be-
tween 100oC and 300oC. The most widely used synthetic fiber is Kevlar, devel-
oped in 1968 by the DuPont Co. It belongs to the family of aromatic polyamides
and its chemical name is poly (paraphenylene terephthalamide). Another impor-
tant polymer for fiber production is Spectra®, which has been developed by Allied
Signal. Spectra® is based on polyethylene containing oriented polymer chains.
Synthetic materials such as poly(benzobisoxazole) (PBO), aromatic copolyesters
and polyimides are also used for fiber fabrication with limited engineering use.
Physical and mechanical properties of widely used synthetic materials for fibers
are given in Table 2.7 [2].
2.3 Production Technologies for FRP Composite

Pipes
FRP composite pipes are fabricated by using the filament winding and fiber
placement methods. The matrix can be thermosetting or thermoplastic resins,
while the fibers can be glass, carbon or polymeric material. Both methods are
fully automated and yield lightweight and high-strength products.
2.3.1 Filament winding

2.3.1.1 Winding patterns
During filament winding, continuous reinforcements in the form of rovings or
monofilaments are wound over a rotating mandrel [3] (Fig.2.1).
Table 2.7
Values of physical and mechanical properties of polymeric fibers.
Property Kevlar-29 Kevlar-49 Spectra 900
(Polyethylene)
Tensile strength 2760 2800–3792 2580
(MPa)
Tensile modulus 62 131 117
(GPa)
Elongation (%) 3–4 2.2–2.8 4–5
3
Density (Kg/ m ) 1440 1479 970
Diameter (μm) 12 12 38
Production Technologies for FRP Composite Pipes 51
Figure 2.1 The filament winding process.
This winding pattern is called helical and it is suitable for pipe production. The
main parameter controlling the mechanical behavior of the produced pipe is the
angle of the fibers with respect to the longitudinal direction of the pipe (winding
angle). It is obvious that winding angle θ = 90ο (hoop winding pattern) improves
the resistance in long pipes subjected to internal pressure, while winding angles
θ = 0 – 15o yield high-resistant pipes for axial tension or bending loading cases.
The prescribed winding angle can be achieved by controlling the rotational speed
of the mandrel and the longitudinal speed of the head that dispenses the tows
(payout head). Increasing the longitudinal speed of the payout head decreases the
winding angle. A hoop winding can be achieved by advancing the payout head
slowly along the mandrel axis so that the fiber tows are wound transversely to
the longitudinal axis. Generally, since the fibers tend to spread into bands due to
tension, a helical winding pattern does not put the tows in order on the mandrel’s
surface. Therefore, several circuits are used before the full surface of the mandrel
is covered.
Apart from the helical pattern, a polar one can be used for axisymmetric com-
posite shells. This pattern results when the mandrel does not rotate but the payout
head rotates about the longitudinal axis. This pattern is mainly used for axisym-
metric pressure vessels.
2.3.1.2 mandrel design

A critical parameter for successful filament wound products is the design of the
mandrel. The following design considerations should be taken into account [3]:
• The mandrel should be sufficiently stiff to withstand the compression im-

posed by the winding force.
• The resin should not stick to the surface of the mandrel. Release agents need
to be applied.
• The mandrel must be extractable from the part after curing.
The separation of the final product from the mandrel is achieved by using ex-
tractable, collapsible, breakable or dissolvable mandrels.
2.3.2 Fiber placement process

The main concept of the fiber placement process is similar to the filament
winding method. However, in the fiber placement process, pre-impregnated tapes
instead of distinct fibers are placed on the mandrel’s surface (Fig.2.2). At the nip
point, a heat source is directed toward the fiber to melt the tape. A roller is used
to apply pressure at the nip point to spread the fiber and to apply compaction [3].
References
[1] Derich Scott, Advanced materials for water handling: Composites and
Thermoplastics, Elsevier, 2000.
[2] Hyer M., Stress analysis of fiber reinforced composite materials, DEStech
Publications, 2009.
[3] Hoa S.V., Principles of the manufacturing of composite materials, DEStech
Publications, 2009.
Figure 2.2 Fiber placement process.

Chapter 3
Mechanical Design of Composite Pipelines
3.1 Types of Loading Cases

The loading conditions of a composite pipeline have fundamental importance
for its dimensioning. Underestimation of the initial sizing of a pipeline can cause
failure, with catastrophic environmental, economic and geopolitical consequenc-
es. On the other hand, as the length of a pipeline (especially for oil and gas trans-
mission) is extremely large, even minor overestimation in sizing (e.g., in the wall
thickness) can lead to critical consequences at the design competition stage of a
project. Therefore the type of loads to be considered is a key factor for a success-
ful design. The loads affecting the dimensions of a composite pipeline are classi-
fied into two main categories: 1. installation loads, and 2. operation loads.
3.1.1 Installation loads

Depending on the selected installation method, installation loads are very of-
ten more critical for pipeline dimensioning than operation loads. In the case of
offshore pipelines, the installation methods can be: (a) S-Lay (Fig.3.1), (b) J-Lay
(Fig.3.2), and (c) Towing (Fig.3.3 abc). In the case of continental pipelines, the
pipes are mostly embedded in soil (Fig.3.4), due to their sensitivity to ultraviolet
radiation. The installation of onshore pipelines thus requires movable cranes and
compaction machines.
53
54 Mechanical Design of Composite Pipelines
Figure 3.1 S-Lay installation process.
Figure 3.2 J-lay installation process.

Types of Loading Cases 55
(a)
(b)
(c)
Figure 3.3 (a) Mid-depth tow, (b) Off-bottom tow, (c) Surface tow.
Figure 3.4 Installation of a pipeline within soil.
During offshore and onshore installation procedures, the pipeline is subjected

to pseudo-static loads of the following loading types: i) bending, ii) axial tension,
iii) external pressure, iv) combination of bending and axial-tension, v) combina-
tion of external pressure and axial-tension, vi) combination of bending, external
pressure and axial tension, vii) torsion (e.g., during the wave movement of the
installation ship), viii) combination of torsion, bending, axial tension and external
pressure. The loading cases above cause plane stress in the wall of composite
pipes. Therefore the stress state for every lamina has to be analyzed by a failure
criterion. Moreover, external pressure, bending, torsion and their combinations
can cause local buckling into the wall of the composite pipe. The loading cases
cited above are summarized in Table 3.1.
3.1.2 Operation loads

During service, composite pipelines can be subjected to static, pseudo-static,
or dynamic loading conditions at ordinary or elevated temperatures. During static
conditions, the loads continuously retain constant values, while during pseudo-
static conditions changes occur in them, assuming constant static conditions over
time. During static or pseudo-static loading, stresses can be produced by, for ex-
ample: a) internal pressure; b) deformation of the pipe due to its own weight
and the methods whereby it is supported; c) thermal stresses due to temperature
gradients; d) creep effects due to uniform (or variable) elevated temperatures;
e) moisture strain effects; and f) soil-pipe interactions, which affect underground
pipes. Dynamic loads can be caused by: a) vibrations resulting from hydrodynam-
ic forces due to internal axial flow or external cross-flow (fluid-pipe interactions);
b) impacts due to fluid hammer; or c) local impact due to foreign objects. During
variable loading conditions, the material mechanisms that can result in failure are
different from those taking place during static or pseudo-static loads. Therefore,
the failure criteria developed for static loads are inadequate for design purposes.
For variable loading cases, the designer has to use fatigue damage accumulation
rules in order to ensure the sustainability of the pipeline for the projected life.
Moreover, for long-term loading histories, especially at elevated temperatures,
the designer has to take into account deformation or possible rupture due to creep.
Current damage mechanics theory provides valuable theoretical tools [e.g., 1–3]
for an effective design that prevents fatigue or/and creep damage. Table 3.1 sum-
marizes the main composite pipeline operation loads, as well as required check-
points for sustainable design.
3.2 Pure Bending

3.2.1 Failure analysis
Bending is the most common loading case in the installation of a composite
pipeline. During the procedure of, for example, S-Lay or J-Lay installation the
pipeline must be curved, which causes significant normal stresses in each lamina.
Moreover, while the pipeline is in operation, possible free-spans on the sea floor
of an offshore pipeline or soil settlement with underground piping can also cause
Pure Bending 57
Table 3.1
Classification of loads acting on composite pipelines.
Loading types Checks

Failure Buckling Fatigue Creep
criterion
Bending + +
Axial tension +
INSTALLATION LOADS
External pressure + +
Combination of bending and axial tension + +
Combination of external pressure and axial + +
tension
Combination of bending, external pressure and + +
axial tension
Torsion + +
Combination of torsion, bending axial tension + +
and external pressure
Constant internal fluid pressure + +
Fluctuation internal fluid pressure + +
Hydrodynamic forces due to internal axial flow + +
Hydrodynamic forces due to external cross flow + +

OPERATION LOADS
Impact pressure due to fluid hammer +

Thermal stresses due to temperature gradients + + +
Uniform elevated temperature effects +

Moisture strain effects +
External pressure due to pipe-soil interaction + +
(forces due to soil weight and road or rail cross-
ings of underground pipelines)
Bending due to differential settlement of soil (of + +
underground pipelines)
Local impact by foreign objects +
bending. In most bending cases, the determination of the allowable bending mo-
ment or the minimum radius of curvature of the deformed pipeline is usually
the main target of bending analysis. Since the plastic deformation of composite
materials is almost absent in their stress-strain behavior (compared to steel), the
stress and deflection analysis of composite pipelines will be based only on elastic-
ity equations.
Design data available to a designer to perform stress and failure analysis of a

pipeline under bending conditions include:
1. the engineering properties of the pipe’s material in the principal coordinate

system, i.e.:
(a) The modulus of elasticity E1, E2
(b) The Poisson’s ratio ν12
(c) The shear modulus G12
(d) The longitudinal tensile strength s 1T
(e) The longitudinal compressive strength s 1C
(f) The transverse tensile strength s 2T
(g) The transverse compressive strength s 2C
(h) The in-plane shear strength τ 12F
2. the fiber orientation θ of each layer and the corresponding stacking se-
quence of the laminate;
3. the thickness h of each layer;
4. the internal diameter D of the pipe.
After the selection of the material, the data 1 (a)-(h) can easily be found in
existing data bases (e.g. Table 3.2) containing material properties. The input data
(2)-(4) are design parameters to be chosen by the designer taking into account,
e.g., manufacturing cost parameters, the fluid supply/demand scenario, the toler-
ance of the pipeline to further loading situations, etc.
Table 3.2
Material properties of widely used composites [4].
E-Glass/Epoxy S-Glass/Epoxy AS/3501 T300/5208
Carbon/Epoxy Carbon/Epoxy
E1 (GPa) 39 43 138 181
E2 (GPa) 8.6 8.9 8.96 10.3

ν12 0.28 0.27 0.30 0.28
G12 (GPa) 3.8 4.5 7.10 7.17

1080 1280 1447 1500
s 1T (MPa)
620 690 1447 1500
s 1C (MPa)
39 49 51.7 40
s 2T (MPa)
128 158 206 246
s 2C (MPa)
89 69 93 68
τ 12F (MPa)
Pure Bending 59
3.2.1.1 mathematical model

To derive a model for dimensioning a multi-layered filament wound pipe un-
der bending, the Lekhnitskii formalism for stress and displacements of a single-
layered pipe will be used, initially.
3.2.1.1.1 Single-layered pipe

According to [5], the stress distribution on a single-layered pipe is given by the
following equations:
 C 
σ r = C1r n −1 − C2 r − n −1 + 3 + Agr  sin ϕ (3.1)
 r 
 C 
σ ϕ = C1 ( n + 1) r n −1 + C2 ( n − 1) r − n −1 + 3 + 3 Agr  sin ϕ (3.2)
 r 
1
σ z = Ar sin ϕ −
S33
( S13 σ r + S23 σ ϕ ) (3.3)
or
1
σz =  − ( S13 + S 23 + nS 23 ) r n −1C1 + ( S13 + S 23 − nS 23 ) r − n −1C2 −
S33 (3.4)
( S13 + S23 ) r C3 + ( S33 − S13 g − 3S 23 g ) rA sin ϕ
−1
τ rϕ = −  r n −1C1 − r − n −1C2 + r −1C3 + grA cos ϕ (3.5)
where [6]:
−1
 Sij  = Qij  a ij   Pij  (3.6)
1 0 0 0 0 0 
0 cos 2 θ sin 2 θ 0 cos θ sin θ 0 

0 sin 2 θ cos 2 θ 0 cos θ sin θ 0 
Qij  =  
0 0 0 cos θ 0 − sin θ 
0 2 cos θ sin θ −2 cos θ sin θ 0 cos 2 θ − sin 2 θ 0 
 
0 0 0 sin θ 0 cos θ 
(3.7)
1 0 0 0 0 0 
0 cos 2
ϑ sin 2 ϑ 0 −2 cos ϑ sin ϑ 0 

0 2
sin ϑ cos 2 ϑ 0 2 cos ϑ sin ϑ 0 
[ Pij ] =   (3.8)
0 0 0 cos ϑ 0 − sin ϑ 
0 cos ϑ sin ϑ − cos ϑ sin ϑ 0 cos ϑ − sin 2 ϑ
2
0 
 
0 0 0 sin ϑ 0 cos ϑ 
 1 ν 32 ν 21 
 E − − 0 0 0 
E2 E1
 2 
 ν 32 1 ν 
− − 21 0 0 0 
 E2 E2 E1 
 ν 21 ν 21 1 
− − 0 0 0 
E E1 E1
[aij ] =  1 

(3.9)
2(1 −ν 32 )
 0 0 0 0 0 
 E2 
 1 
 0 0 0 0 0 
 G12 
 
 0 1 
0 0 0 0
 G12 
S 23 − S13
g= (3.10)
β11 + 2 β12 + β 66 − 3β 22
β11 + 2 β12 + β 66
n = 1+ (3.11)
β 22
Si 3 S j 3
β ij = Sij − (3.12)
S33
According to [e.g., 7], the strains and stresses in directions z , ϕ , r (Figures 3.5
and 3.6) are correlated with the following relationships:
− − −
ε z = S 11 σ z + S 12 σ ϕ + S 13 σ r (3.13)
− − −
ε ϕ = S 21 σ z + S 22 σ ϕ + S 23 σ r (3.14)
− − −
ε r = S 31 σ z + S 32 σ ϕ + S 33 σ r (3.15)
Pure Bending 61
1−
γ rϕ = S 44 τ rϕ (3.16)
2
−
where S ij are given in the Appendix to Chapter 1 [7].
Figure 3.5 Geometry of a single-layered pipe.
Figure 3.6 Coordinate systems for directions of stresses acting on a lamina.

Taking into account equations (3.1)-(3.4), the above equations yield:
−  1
ε z = S 11   − ( S13 + S 23 + nS 23 ) r n −1C1 + ( S13 + S 23 − nS 23 ) r − n −1C2 −
 S 33
( S13 + S23 ) r −1C3 + ( S33 − S13 g − 3S23 g ) rA sin ϕ} +

(3.17)
{ }
−
S 12 ( n + 1) r C1 + ( n − 1) r
n −1
C2 + r C3 + 3 grA sin ϕ +
− n −1 −1
{r }
−
S 13 n −1
C1 − r − n −1C2 + r −1C3 + grA sin ϕ
or
− 1
ε z = sin ϕ  S 11  − ( S13 + S 23 + nS 23 ) r n −1C1 + ( S13 + S 23 − nS 23 ) r − n −1C2 −
 S33
( S13 + S23 ) r −1C3 + ( S33 − S13 g − 3S23 g ) rA +
−
(3.18)
S 12 ( n + 1) r C1 + ( n − 1) r
n −1 − n −1
C2 + r C3 + 3 grA +
−1
−
S 13  r n −1C1 − r n −1C2 + r −1C3 + grA }
− 1
ε ϕ = sin ϕ  S 21  − ( S13 + S 23 + nS 23 ) r n −1C1 + ( S13 + S 23 − nS 23 ) r − n −1C2 −
 S 33
( S13 + S23 ) r −1C3 + ( S33 − S13 g − 3S23 g ) rA +

−
(3.19)
S 22 ( n + 1) r n −1C1 + ( n − 1) C2 + r −1C3 + 3 grA +
− n −1
 
−
S 13  r n −1C1 − r − n −1C2 + r −1C3 + grA }
− 1
ε r = sin ϕ  S 31  − ( S13 + S 23 + nS 23 ) r n −1C1 + ( S13 + S 23 − nS 23 ) r − n −1C2 −
 S 33
( S13 + S23 ) r −1C3 + ( S33 − S13 g − 3S23 g ) rA +

−
(3.20)
S 32 ( n + 1) r n −1C1 + ( n − 1) r − n −1C2 + r −1C3 + 3 grA +
−
S 33  r n −1C1 − r − n −1C2 + r −1C3 + grA }
Pure Bending 63
1−
γ rϕ = − S 44 cos ϕ  r n −1C1 − r − n −1C2 + r −1C3 + grA (3.21)
2
The above strains ε r , ε ϕ , ε z , γ rϕ can be correlated with the displacements ur and

uϕ by the following relationships [e.g., 5]:
∂ur
εr = (3.22)
∂r
ur 1 ∂uϕ
εϕ = + (3.23)
r r ∂ϕ
∂u z
εz = (3.24)
∂z
1 ∂ur ∂uϕ uϕ
γ rϕ = + − (3.25)
r ∂ϕ ∂r r
Substitution of eq. (3.20) with eq. (3.22) yields:
ur = ∫ ε r dr + F (ϕ ) (3.26)
or
1   − −
− − −

ur = r − n  2C2  − S13 S 31 + ( −1 + n ) S 23 S 31 + S33  S 32 − n S 32 + S 33   −
2nS33    
 − −
− − −

2C1r 2 n  S13 S 31 + (1 + n ) S 23 S 31 − S33  S 32 + n S 32 + S 33   +
  
  −
 −  − −

nr n  Ar 2  S33 S 31 + g  − S 31 ( S13 + 3S 23 ) + S33  3 S 32 + S 33    +
    
 − − −
 
+2C3  − S 31 ( S13 + S 23 ) + S33  S 32 + S 33   ln ( r )   sin ϕ +
   
+ F (ϕ )
(3.27)
According to [5] the displacements should be single-valued functions.
Therefore:
C3 = 0 (3.28)
According to eq. (3.28), equation (3.27) can be simplified as:
r − n sin ϕ   − −
− − −

ur =  2C2  − S13 S 31 + ( n − 1) S 23 S 31 + S33  S 32 − n S 32 + S 33   −
2nS33    
 − −
 − − −

2C1r 2 n  S13 S 31 + ( n + 1) S 23 S 31 − S33  S 32 + n S 32 + S 33   +
  
 −
 −
 −
  
−
Anr n + 2  S33 S 31 + g  − S 31 ( S13 + 3S 23 ) + S33  3 S 32 + S 33     +
     
+ F (ϕ )
(3.29)
where F (ϕ ) is an unknown function of ϕ . Substitution of eqs. (3.19) and (3.29)

into eq. (3.23) yields:
uϕ = G (r ) + ∫ F ( ϕ )d ϕ +
r − n cos ϕ    − −
 −
 2C1r 2 n  n 2  S 23 S 21 − S33 S 22  − S 31 ( S13 + S 23 ) +
2nS33    
− −
− − −

n  S 21 ( S13 + S 23 ) − S 23 S 31 − S33  S 22 + S 23 + S 32   +
  
  −  − − −     − −

Anr n+ 2  S33 S 33 2 S+21S33+g +( S13 + 3S 23 )  2 S 21 − S 31  +
S 31 −S 32
       
   − −6 S 22 − 2 S 23 + 3 S 32 + S 33    
− − − − − −
2C2  n 2  S 23 S 21 − S33 S 22 S−33 S 31 ( 13 + S 23 ) +   
      
 − −
− −

− (3.30)
 − S 21 ( Sof
where G(r) is an unknownn function r. S 23 ) + S 23 S 31 + S33  S 22 + S 23 − S 32   +
13 +
  
Taking into account the condition of symmetry
− −

S33  S 32 + S 33   +
 uϕ (ϕ ) = −uϕ (π − ϕ ) (3.31)
eq. (3.30) yields:
G ( r ) + ∫ F (ϕ ) d ϕ = 0 (3.32)
Let
∫ F (ϕ ) d ϕ = Ε (ϕ ) (3.33)
Pure Bending 65
then eq. (3.32) can be written:
G ( r ) + E (ϕ ) = 0 (3.34)
yielding
G (r ) = 0 (3.35)
and
E (ϕ ) = 0 (3.36)
or
F (ϕ ) = 0 (3.37)
With the aid of eqs. (3.28), (3.35), (3.37) and using the following notations
  r 
−n
 −
λ1 = −2  S13 S 31 + ( n + 1) S 23 S 31 − S33  S 32 + n S 32 + S 33   r 2 n 
− − − −
 (3.38)
     2nS33 
 r 
−n

λ2 = 2  − S13 S 31 + ( n − 1) S23 S 31 + S33  S 32 − n S 32 + S 33   
− − − − −
 (3.39)
    2 nS 33 
  r 
−n
  −
λ3 = nr n + 2  S33 S 31 + g  − S 31 ( S13 + 3S 23 ) + S33  3 S 32 + S 33    
− − −
 (3.40)
       2nS33 

λ4 = 2r 2 n  n 2  S 23 S 21 − S33 S 22  − S 31 ( S13 + S 23 ) +
− − −
  
 − −
− − −

n  S 21 ( S13 + S 23 ) − S 23 S 31 − S33  S 22 + S 23 + S 32   + (3.41)
  
  r 
− n
− −
S33  S 32 + S 33    
    2nS33 

λ5 = 2  n 2  S23 S 21 − S33 S 22  − S 31 ( S13 + S 23 ) +
− − −
  
 − −
− − −

n  − S 21 ( S13 + S 23 ) + S 23 S 31 + S33  S 22 + S 23 − S 32   + (3.42)
  
  r 
− n
− −
S33  S 32 + S 33    
    2nS33 
 
λ6 = nr n + 2  S33  S 31 − 2 S 21  + g ( S13 + 3S23 )  2 S 21 − S 31  +
− − − −
     

   r 
−n
 − − − −
S33  −6 S 22 − 2 S 23 + 3 S 32 + S 33     
    2 nS 33 
(3.43)
the expressions (3.29), (3.30) can be written:
ur = sin ϕ ( λ1C1 + λ2 C2 + λ3 A ) (3.44)
uϕ = cos ϕ ( λ4 C1 + λ5C2 + λ6 A ) (3.45)
With the aid of eq. (3.28) and using the following notations
µ1 = r n −1 (3.46)
µ2 = −r − n −1 (3.47)
µ3 = gr (3.48)
µ4 = ( n + 1) r n −1 (3.49)
µ5 = ( n − 1) r − n −1 (3.50)
µ6 = − ( S13 + S 23 + nS 23 ) r n −1 (3.51)
µ7 = ( S13 + S 23 − nS 23 ) r − n −1 (3.52)
µ8 = ( S33 − S13 g − 3S 23 g ) r (3.53)
the expressions (3.1)-(3.5) can be written:

Pure Bending 67
σ r = sin ϕ ( µ1C1 + µ2 C2 + µ3 A ) (3.54)
σ ϕ = sin ϕ ( µ4 C1 + µ5C2 + 3µ3 A ) (3.55)
sin ϕ
σz = ( µ6C1 + µ7 C2 + µ8 A) (3.56)
S33
τ rϕ = − cos ϕ ( µ1C1 + µ2 C2 + µ3 A ) (3.57)
3.2.1.1.2 Multi-layered pipe

The equations (3.1)-(3.5) and (3.44), (3.45) express the stresses and deforma-
tions for a single-layered pipe. For the dimensioning of a multi-layered pipe under
bending (Fig. 3.7), we have to determine the stresses σr , σφ, σz, τrφ for each layer k.
Therefore, the next step is the determination of the unknown constants C1, C2, A
for all layers from k = 1 to k = N. To achieve this target, the following conditions
have to be satisfied:
3.2.1.1.2.1 Equilibrium equations on interfaces
τ r1ϕ ( r2 ) = τ r2ϕ ( r2 ) 

τ ( r3 ) = τ ( r3 )
2
rϕ
3
rϕ 

 
 

 
 ( N − 1) conditions (3.58)
τ rϕ ( rk +1 ) = τ rϕ ( rk +1 ) 
k k +1
 

 

 
τ rϕ ( rN ) = τ rϕ ( rN ) 
Ν −1 Ν
Figure 3.7 Geometry of a cross-section of a multi-layered pipe.
3.2.1.1.2.2 Compatibility equations
u1r ( r2 ) = ur2 ( r2 )


u ( r3 ) = u ( r3 )
2
r 
3
3

..................... 
...................... 

..................... 
k  ( N − 1) conditions (3.59)
ur ( rk +1 ) = ur ( rk +1 ) 
k +1
..................... 

..................... 

...................... 
urN −1 ( rN ) = urN ( rN ) 
Pure Bending 69
uϕ1 ( r2 ) = uϕ2 ( r2 )


u ( r3 ) = u ( r3 )
2
ϕ
3
ϕ 

..................... 
...................... 

..................... 
k  ( N − 1) conditions (3.60)
uϕ ( rk +1 ) = uϕk +1 ( rk +1 ) 
..................... 

..................... 

...................... 
uϕN −1 ( rN ) = uϕN ( rN ) 
3.2.1.1.2.3 Boundary conditions on exterior cylindrical surfaces

The conditions on the exterior surfaces of the pipe located on r = r1 and r = rN+1
are:
τ r1ϕ ( r1 ) = 0 
 2 conditions (3.61)
τ rNϕ ( rN +1 ) = 0 
3.2.1.1.2.4 Boundary conditions on the cross-sections at the ends of the pipe

Due to the action of stress σ zk (r , ϕ ) on an elementary area dA = r dϕ dr lo-
cated within an arbitrary layer k (Fig. 3.7), the corresponding elementary load is
dN z = σ zk ( r , ϕ ) .rdϕ .dr . This axial force yields an elementary bending moment
dM k = ydN , where y = r sin ϕ . Therefore, the bending moment due to the cross-
section of a layer k is
π rk +1
Μκ = 2∫ ∫ σ ( r,ϕ ) r
k 2
z sin ϕ dr dϕ .
0 rk
With the aid of the last equation the equilibrium between the bending moment M
acting at the ends of pipe is equal to the summation of the bending moments due
to N layers, i.e.
Ν
Μ = ∑ Μκ
κ =1
or
Ν π rk +1
Μ = 2∑ ∫ ∫ σ ( r,ϕ ) r
k
z
2
sin ϕ dr dϕ (3.62)
κ =1 0 rk
With the aid of eq. (3.56) and taking into account eqs (3.51)-(3.53), the above
equation can be written as:
Ν
(
Μ = π∑ I 6 C1k + I 7 C2k + I 8 Ak ) (3.63)
k =1
where
rk +1
I6 = ∫
rk
µ6 r 2 dr (3.64)
rk +1
I7 = ∫
rk
µ7 r 2 dr (3.65)
rk +1
I8 = ∫
rk
µ8 r 2 dr (3.66)
Using the notations (3.51)-(3.53) the above integrals yield:
S13 + S 23 + nS 23 n + 2 n + 2
I6 = − (r − r )
( n + 2 ) S33 k +1 k
(3.67)
S13 + S 23 − nS 23 − n + 2 − n + 2
I7 = (r − r )
( −n + 2 ) S33 k +1 k
(3.68)
S33 − S13 g − 3S 23 g 4
I8 =
4 S33
( rk +1 − rk4 ) (3.69)
For N-layered pipe the number of unknowns is 3N . On the other hand, the num-
ber of available conditions shown in eqs (3.58)–(3.62) is 3 ( N − 1) + 2 + 1 = 3 N .
Taking into account eq. (3.57) the conditions (3.58) for a multilayered pipe
with stacking sequence ( ±θ ) NP can be written:
Pure Bending 71
µ1 ( r2 ,1) C11 + µ2 ( r2 ,1) C21 + µ3 ( r2 ,1) A1 = µ1 ( r2 , 2 ) C12 + µ2 ( r2 , 2 ) C22 + µ3 ( r2 , 2 ) A2 


µ1 ( r3 , 2 ) C12 + µ2 ( r3 , 2 ) C22 + µ3 ( r3 , 2 ) A2 = µ1 ( r3 , 3) C13 + µ2 ( r3 , 3) C23 + µ3 ( r3 , 3) A3 

............................................................................................................................... 
............................................................................................................................... 

............................................................................................................................... 
k +1 
µ1 ( rk +1 , k ) C1 + µ2 ( rk +1 , k ) C2 + µ3 ( rk +1 , k ) A = µ1 ( rk +1 , k + 1) C1 + µ2 ( rk +1 , k + 1) C2 + µ3 ( rk +1 , k + 1) A
k k k k +1 k +1

................................................................................................................................ 

................................................................................................................................. 
................................................................................................................................ 

µ ( r , N − 1) C1 + µ2 ( rN , N − 1) C2 + µ3 ( rN , N − 1) A = µ1 ( rN , N ) C1 + µ2 ( rN , N ) C2 + µ3 ( rN , N ) A 
N −1 N −1 N −1 N N N
1 N
(3.70)
With the aid of eq. (3.44) the conditions (3.59) for a multi-layered pipe with stack-
ing sequence ( ±θ ) NP yield:
λ1 ( r2 ,1) C11 + λ2 ( r2 ,1) C21 + λ3 ( r2 ,1) A1 = λ1 ( r2 , 2 ) C12 + λ2 ( r2 , 2 ) C22 + λ3 ( r2 , 2 ) A2 


λ1 ( r3 , 2 ) C + λ2 ( r3 , 2 ) C + λ3 ( r3 , 2 ) A = λ1 ( r3 , 3) C + λ2 ( r3 , 3) C + λ3 ( r3 , 3) A
1
2 2
2
2 3
1
3
2
3


............................................................................................................................. 
............................................................................................................................. 

............................................................................................................................. 
λ1 ( rk +1 , k ) C1k + λ2 ( rk +1 , k ) C2k + λ3 ( rk +1 , k ) Ak = λ1 ( rk +1 , k + 1) C1k +1 + λ2 ( rk +1 , k + 1) C2k +1 + λ3 ( rk +1 , k + 1) Ak +1 

............................................................................................................................ 

............................................................................................................................. 
............................................................................................................................. 

λ1 ( rN , N − 1) C1N −1 + λ2 ( rN , N − 1) C2N −1 + λ3 ( rN , N − 1) AN −1 = λ1 ( rN , N ) C1N + λ2 ( rN , N ) C2N + λ3 ( rN , N ) AN 
(3.71)
Similarly, with the aid of eq. (3.45), the conditions (3.60) can be written:
λ4 ( r2 ,1) C11 + λ5 ( r2 ,1) C21 + λ6 ( r2 ,1) A1 = λ4 ( r2 , 2 ) C12 + λ5 ( r2 , 2 ) C22 + λ6 ( r2 , 2 ) A2 


λ4 ( r3 , 2 ) C + λ5 ( r3 , 2 ) C + λ6 ( r3 ,1) A = λ4 ( r3 , 3) C + λ5 ( r3 , 3) C + λ6 ( r3 , 3) A
1
2 2
2
2 3
1
3
2
3


............................................................................................................................ 
............................................................................................................................ 

............................................................................................................................ 
k +1 
λ4 ( rk +1 , k ) C1 + λ5 ( rk +1 , k ) C2 + λ6 ( rk +1 , k ) A = λ4 ( rk +1 , k + 1) C1 + λ5 ( rk +1 , k + 1) C2 + λ6 ( rk +1 , k + 1) A
k k k k +1 k +1

............................................................................................................................ 

............................................................................................................................ 
............................................................................................................................ 

λ4 ( rN , N −11) C1 + λ5 ( rN , N − 1) C2 + λ6 ( rN , N − 1) A = λ4 ( rN , N ) C1 + λ5 ( rN , N ) C2 + λ6 ( rN , N ) A 
N −1 N −1 N −1 N N N
(3.72)
The above equations (3.70)-(3.72) regarding the interfaces of the layers have to be
completed with the boundary conditions on exterior surfaces of the multi-layered
pipe.
Taking into account the eq. (3.57), the conditions (3.61) yield:
µ1 ( r1 ,1) C11 + µ2 ( r1 ,1) C21 + µ3 ( r1 ,1) A1 = 0 

 (3.73)
µ1 ( rN , N − 1) C1N + µ2 ( rN , N − 1) C2N + µ3 ( rN , N − 1) AN = 0 
The equations (3.70)-(3.73), as well as the eq. (3.63), can be written in the follow-
ing matrix form:
[ M 1 ] [ M 2 ] [ M 3 ]
  {C } 
 [ Λ1 ] [ Λ 2 ] [ Λ 3 ]   1 
 [ Λ 4 ] [ Λ 5 ] [ Λ 6 ]  {C2 } = {L} (3.74)
   
 [ 1 ] [ 2 ] [ 3 ]  { A} 
Β Β Β
 [J ] [J ] [J ] 
 6 7 8 
where:
{C1} = {C11 , C12 , C13 , , C1k , , C1N }
T
(3.75)
{C2 } = {C21 , C22 , C23 , , C2k , , C2N }

T
(3.76)
{ A} = { A1 , A2 , A3 , , Ak , , AN }
T
(3.77)

 µi (2,1) − µi (2, 2) 0 0 0 0 0 0 0 0 
 0 µi (3, 2) − µi (3, 3) 0 0 0 0 0 0 0 
 
 0 0 µi (4, 3) − µi (4, 4) 0 0 0 0 0 0 
 
[ M i ]( N −1) xN = 0 0 0 0  0 0 0 0 0 
 0 0 0 0 0 µi (k , k − 1) − µi (k , k ) 0 0 0 
 
 0 0 0 0 0 0 0  0 0 
 0 0 0 0 0 0 0 0 µ ( N , N − 1) − µ ( N , N ) 
 i i 
(3.78)
λi (2,1) −λi (2, 2) 0 0 0 0 0 0 0 0 

 0 λi (3, 2) −λi (3, 3) 0 0 0 0 0 0 0 
 
 0 0 λi (4, 3) −λi (4, 4) 0 0 0 0 0 0 
 
[ Λi ]( N −1) xN = 0 0 0 0  0 0 0 0 0 
 0 0 0 0 0 λi (k , k − 1) −λi (k , k ) 0 0 0 
 
 0 0 0 0 0 0 0  0 0 
 0 0 0 0 0 0 0 0 λi ( N , N − 1) −λi ( N , N ) 

(3.79)
Pure Bending 73
 µi (1,1) 0 0 0 0 0 0 0 0 0
[ Bi ]2 xN =
µ ( N , N − 1) 0 0 0 0 0 0 0 0 0 
(3.80)
 i
{ J i }1xN = {I i1 I i2 I i3  I ik −1 I ik I ik +1  I iN −1 I iN } (3.81)
T
M
{ L}3 Nx1 = 0 0 0 0 0 0 0 0 0  (3.82)
 π
C1k is the coefficient C1 for the k-th layer

C2k is the coefficient C2 for the k-th layer
Ak is the coefficient A for the k-th layer
µi ( j , k ) is the coefficient µi for the k-th layer for r = rj
λi ( j , k ) is the coefficient λi for the k-th layer for r = rj
I ik is the coefficient I i for the k-th layer
The solution of the matrix equation (3.74) yields the unknown constants C1k , C2k ,
Ak ( k = 1, 2, , NP ) for all layers. Therefore, using the values C1 , C2 , ANP
NP NP
of the critical exterior layer ( k = NP ) , its stress state can now be determined
by the eqs. (3.54)-(3.57). In that case, the determination of the principal stresses
σ 1 , σ 2 ,τ 12 of the exterior layer is possible by using eq. (1.70). Finally, the allow-
able bending moment can now be obtained by using a failure criterion (e.g. [8]).
3.2.2 Buckling model

When a bending moment M y is acting on a composite pipe (Fig. 3.8a), the
narrow strip AB is subjected to maximum compressive longitudinal strain ε ξ0 (0) . In
order to calculate the critical value of M y causing local buckling to the strip AB,
we have initially to estimate the critical value of εξο (0), namely εξcο, at the corre-
sponding buckling state. To this scope, we’ll assume that the above strip is a part
of the same pipe that has reached the critical longitudinal compressive strain εξcο
due to axial compression (Fig. 3.8b).
Figure 3.8 (a) Buckling of narrow strip AB due to bending, (b) Buckling of narrow strip
AB due to axial load.
∧
We will approximate the critical value of M y for the case of Fig. 3.8a, taking
o
into account the critical strain ε ξ c derived from the model shown in Fig. 3.8b.
Because the maximum compressive stress causing buckling to the model of
Fig. 3.8b does act on the whole perimeter of the pipe’s cross-section, instead of
acting only on point A (as in the case of Fig. 3.8a), the practical assumption above
is expected to be conservative, and therefore, safe for designing purposes. This
assumption has been checked in [9] for the case of isotropic thin-walled tubes. In
that work it is seen that for pure bending the exact solution for the critical com-
pressive stress gives a value about 30% higher than that obtained by the above
assumption [10]. In regard to Figure 3.8b, the critical longitudinal strain at the
buckling state is:
ε ξοc = α11 Ν c (3.83)
where Ν c is the critical axial load per unit circumference given by
N c = λcr N o (3.84)
Pure Bending 75
In this equation No = 1. The parameter λcr is the minimum eigenvalue obtained by

the solution of the following equation:
 O L   M 0 M n  O L 
T
 Φ ⋅ [ J ] Φ 2 ⋅ [ J ] 
det   −λ  1  = 0
  L O   M n   
(3.85)

M 0   L O Φ 2 ⋅ [ J ] Φ1 ⋅ [ J ]  
The above equation, as well as the matrices [L], [O], [J] and the parameters Φ1,
Φ2, are given in ref. [11], i.e.:
 −α 0 β 0 0 0 
 0 
[ ] 
O = −β α 0 0 0  (3.86)
 0 2/ D 0 a + β 2 c22
2
β + a 2 c12
2
−2a β (1 + c1c2 ) 
 β c2 0 −ac1 0 0 0 
[ ]  0
L = ac1 − β c2 0 0 0 
 (3.87)
 0 0 0 −2a β c2 −2a β c1 2(c1a 2 + c2 β 2 ) 
 A11 A12 0 B11 B12 0 

A A22 0 B12 B22 0 
 12
 0 0 A66 0 0 B66 
[M o ] =  B B12 0 D11 D12 0 
 (3.88)
 11
 B12 B22 0 D12 D22 0 
 
 0 0 B66 0 0 D66 
Α Β 
[Μ n ] =   − [M o ] (3.89)
Β D
0 0 0 
[ J ] = 0 0 0
 (3.90)
0 0 1 
Φ1 = a 2 + β 2 c22 (3.91)
Φ 2 = −2a β c2 (3.92)
In the above matrices the four constants a, b, c1, c2 characterize the wave pat-
tern of the buckling. For different combinations of the above constants, equation
(3.85) will result in different values of λcr. The lowest value of λcr is the appropri-
ate value to be used in equation (3.84).
ο
The local strain ε ξ at any point located on the mean diameter of a cross-
∧
section of a pipe subjected to pure bending by moment M y is [11]:
1
ε ξο = z (3.93)
ρy
where
∧
1 My
= (3.94)
ρ y EI yy
π
 1 2 1 D
EI yy = 2 ∫  z + cos 2 θ  dθ (3.95)
α
0  11
d11 2
D
z= cos θ (3.96)
2
ο
The local strain ε ξ of the strip AB shown in Fig. 3.8a occurs when θ = 0.
Therefore:
ο
My D
ε =
ξ (3.97)
EI yy 2
Combining eqs. (3.83), (3.84), (3.97), the critical bending moment that causes
local buckling into the strip AB can be approximated by the following equation:
∧ 2 EI yy a11λcr
M yc = (3.98)
D
External Pressure 77
3.3 External Pressure

The long pipe under consideration (Fig. 3.9) with diameter D, thickness h,
stacking sequence [±θ NP ] , is subjected to external pressure pz. According to
Table 3.1, the first check to ensure its tolerance for this kind of loading should
be based on failure analysis. The maximum allowable external pressure satisfy-
ing the selected failure criterion can be determined by the procedure shown in
Figure 3.10.
Figure 3.9 Geometry of the problem.
Figure 3.10 Procedure for estimation of allowable external pressure pa.

Figure 3.11 Equilibrium of a half pipe.
In order to determine the external forces acting on the pipe’s wall, the equilib-
rium equation of the half pipe shown in Figure 3.11 should be used:
2N y L = pDL (3.99)
or
1
Ny = pD (3.100)
2
Because of the absence of external loads in directions x and xy, Ny is the only load
acting on the laminate.
Therefore:
Nx = 0 (3.101)
N xy = 0 (3.102)
Mx = 0 (3.103)
My = 0 (3.104)
M xy = 0 (3.105)
Taking into account the above equations as well as the inverse ABD matrix, it can
be written:
ε x0 = a12 N y (3.106)
ε y0 = a22 N y (3.107)
γ xy0 = a26 N y (3.108)
κ x0 = b21 N y (3.109)
κ y0 = b22 N y (3.110)
κ xy0 = b26 N y (3.111)
Therefore:
ε x = ε x0 + zκ x0 (3.112)
ε y = ε y0 + zκ y0 (3.113)
γ xy = γ xy0 + zκ xy0 (3.114)
With the aid of equations (3.106)-(3.114), the stress-strain relation provides the
stresses σx, σy, τxy for every ply:
σ x  ε x 
   
 
σ y  = Q ij (θ )  ε y  (3.115)
   
τ xy  γ xy 
Using the matrix [T (θ )] , the principal stresses σ1, σ2, τ12 can be obtained by the
following well-known equation:
σ 1  σ x 
   
σ 2  = [ Τ(θ ) ] σ y  (3.116)
τ   
 12  τ xy 
Using eq. (3.116), the Tsai-Wu failure criterion for every ply yields the allowable
value pa of external pressure. Since the Tsai-Wu is a second-order algebraic equa-
tion, we are going to obtain two values of pa for every ply. From the derived dif-
ferent values of pa, the minimum one should be adopted in order to cover all cases.

A long pipe with mean diameter D is made from a multi-layered laminate with
stacking sequence [±θ ] . According to ref. [12], the critical buckling pressure can
be estimated by the relation:
 2
Aani Dani − Bani 
pcr = 3   (3.117)
 Aani ( D / 2 ) + 2 Bani ( D / 2 ) + Dani ( D / 2) 
3 2
In the above equation, the parameters Aani, Bani, Dani can be obtained by the follow-
ing matrix equation:
 Aani Bani   A22 B22  −1

− [ L1 ]  L2  [ L1 ]
T
B =
Dani   B22 
(3.118)
 ani D22 
where
 A12 B12 
A B26 
[ L1 ] =  26 (3.119)
B12 D12 
 
 B26 D26 

Combination of Bending and External Pressure 81
 A11 A16 B11 B16 

A A66 B16 B66 
[ L2 ] =  B16 B16 D11 D16 
(3.120)
11
 
B
 16 B66 D16 D66 
3.4 Combination of Bending and External

Pressure
Superposition of the stresses σx, σy, τxy due to external pressure is given by:
 h 
α12 + b21 
σ p
 2
x
 p  
 h 1
σ =  Q (
 y   ij   22θ )  a + b22  pD (3.121)
 p  2 2
τ xy   h 
a26 + 2 b26 
 
with the corresponding bending-induced stresses
 1 
σ xΜ   S ( µ6 C1 + µ7 C2 + µ8 Α 
 Μ   33 
σ y  =  µ 4 C1 + µ5C2 + 3µ3 Α  (3.122)
 Μ  0 
τ xy   
 
the following formulas for the principal stresses can be obtained where
σ 1 = (σ xΜ + σ xp ) cos 2 θ + (σ yΜ + σ yp ) sin 2 θ (3.123)
σ 2 = (σ xΜ + σ xp ) sin 2 θ + (σ yΜ + σ yp ) cos 2 θ (3.124)
τ 12 = (σ xΜ + σ xp )(− cos θ sin θ ) + (σ yΜ + σ yp ) cos θ sin θ + (τ xyΜ + τ xyp )(cos 2 θ − sin 2 θ )
(3.125)
It should be noticed that equation (3.122) provides the stresses of the exterior
lamina at the location φ = π/2 (location of the maximum bending stresses). In eq.
(3.121), h denotes the thickness of the lamina, D is the exterior diameter of the
pipe, and p is the external pressure.
Since the principal stresses σ1, σ2, τ12 are known, the Tsai-Wu failure criterion
can be used for derivation of the diagrams pα = pα(M) providing the allowable
combinations of the external pressure pα and bending moment M.

A laminated pipe of diameter D, made by a composite wall with stacking se-
quence [ ±Φ ] is subjected to combined pure bending and external compressive
pressure. The objective of this section is to determine the combination of the
critical values of the bending moment Mcr and the external pressure pcr that cause
buckling. We recall from section 3.3 that the critical bending moment Mcr can be
correlated (for design purposes) with the critical axial compression Nx cr through
the formula:
2 ⋅ EI ⋅ a11
M cr = N x cr (3.126)
D
where
π
 D 2 cos 2 θ 1 D
EI = 2 ∫  + cos 2 θ  dθ (3.127)
0
4a11 d11 2
The above formulas (3.126) and (3.127) yield:
 D 2 a11 
M cr = π  + N x cr (3.128)
 4 d11 
Therefore, the first step for the development of a buckling model for the combina-
tion of pure bending and external pressure is the solution for the case where the
compressive axial load Nx and the lateral pressure q are combined.
3.4.2.1 composite pipe under compressive axial load

Let u, v, w denote displacements along the axes x, y, z respectively. Taking into
account
(a) the geometric relation between the displacements u, v, w,

(b) the equilibrium of resultant forces and compressive axial force Nx,
(c) the assumption that all resultant forces, except NX, are very small and we
can neglect the products of these forces with the derivatives of the displace-
ments u, v, w,
(d) the assumption that the moments are very small and we can neglect the
products of moments and derivatives of the displacements u, v, w,
the following relations based on [10] can be written:
∂N x ∂N xy 
R + =0 
∂x ∂θ 
∂N y ∂N xy ∂ 2 v ∂M xy ∂M y 
+R + RN x 2 + − =0  (3.129)
∂θ ∂x ∂x ∂x R∂θ 
∂2w ∂2 M x
2 2
∂ M xy ∂ M y 
RN x 2 + N y + R 2
−2 + = 0
∂x ∂x ∂x∂θ R∂θ 2

3.4.2.2 composite pipe under uniform lateral external pressure

Taking into account: (a) the geometric relations between the displacements
u, v, w; (b) the equilibrium of resultants and pressure q acting on a wall element;
(c) the assumption that all resultant forces, except Ny, are small and we can ne-
glect the terms containing the products of these resultants with the derivatives of
the displacements u, v, w; and (d) the assumption that the bending and twisting
moments are small and we can neglect the products of these moments with the
derivatives of the displacements u, v, w, the following relations based on [10] can
be written:
Figure 3.13 Coordinate system of a pipe under compressive axial load.

∂N x ∂N xy  ∂ 2 v ∂w  
R + + qR  − =0 
∂x ∂θ  ∂x∂θ ∂x  
∂N y ∂N xy ∂M y ∂M xy 
+R − + =0  (3.130)
∂θ ∂x R∂θ ∂x 
∂ 2 M xy 2
∂ 2
M ∂ 2
M 
∂ Mx y xy  2
∂ w  
− +R + − − q  w +  = 0
∂x∂θ ∂x 2 R∂θ 2 ∂x∂θ  ∂θ 2  

3.4.2.3 composite pipe under combined lateral external pressure and axial
compressive load
When both lateral external pressure q and axial compression Nx are simulta-
neously acting on a long composite pipe, the combination of eqs. (3.129) and
(3.130) yields:
∂N x ∂N xy  ∂ 2 v ∂w  
R + = − qR  − 
∂x ∂θ  ∂x∂θ ∂x  
∂N y ∂N xy 1 ∂M y ∂M xy  ∂2v  
+R − + = −Nx R 2   (3.131)
∂θ ∂x R ∂θ ∂x  ∂x  
2 2

∂2 M x 1 ∂ M y ∂ M xy  ∂2 w   ∂2 w 
R + −2 = +q w + 2  − Nx R 2 
∂x 2 R ∂θ 2 ∂ x∂ θ  ∂θ   ∂x  
3.4.2.4 correlation of resultant forces with displacements

For a general cross-ply laminated circular cylindrical shell, i.e., a shell with
stiffnesses A16 = A26 = A45 = B16 = B26 = D16 = D26 = 0, the resultant forces are given
[7] by:
N x = A11ε x0 + A12ε y0 + B11k x0 + B12 k y0 


N y = A12ε x0 + A22ε y0 + B12 k x0 + B22 k y0 

N xy = A66γ xy0 + B66 k xy0 
 (3.132)
M x = B11ε x0 + B12ε y0 + D11k x0 + D12 k y0 
M y = B21ε x0 + B22ε y0 + D12 k x0 + D22 k y0 
M xy = B66γ xy0 + D66 k xy0 

where [7]:
∂u 
 ε x0 =
∂x

∂v 
ε y0 =
R∂θ 

∂u ∂v 
γ xy0 = +
R∂θ ∂x 
 (3.133)
∂2 w 
k x0 = 2
∂x 
2 
1 ∂ w 
k y0 = 2
R ∂θ 2 

2 ∂2 w 
k xy0 =
R ∂x∂θ 
With the aid of eq. (3.133), eq. (3.132) can be written:
∂u A12 ∂v ∂2 w B ∂2 w 
N x = A11 + + B11 2 + 122 
∂x R ∂θ ∂x R ∂θ 2 
∂u A22 ∂v ∂ w B ∂ w 
2 2
N y = A12 + + B12 2 + 222 
∂x R ∂θ ∂x R ∂θ 2 
A ∂u ∂v 2 B66 ∂ 2 w 
N xy = 66 + A66 + 
R ∂θ ∂x R ∂x∂θ 
(3.134)
∂u B12 ∂v ∂ 2 w D12 ∂ 2 w 
M x = B11 + + D11 2 + 2
∂x R ∂θ ∂x R ∂θ 2 
∂u B22 ∂v ∂2 w D ∂2 w 
M y = B21 + + D12 2 + 222
∂x R ∂θ ∂x R ∂θ 2 
B ∂u ∂v 2 D66 ∂ 2 w 
M xy = 66 + B66 + 
R ∂θ ∂x R ∂x∂θ
Taking into account the above equation (3.134), equation (3.131) yields:
 ∂ 2u A ∂ 2 v ∂ 3 w B ∂ 3 w  A66 ∂ 2 u ∂v 2B ∂3 w 
R  A11 2 + 12 + B11 3 + 122 2 
+ + A66 + 66 =
 ∂x R ∂x∂θ ∂x R ∂x∂θ  R ∂θ 2
∂θ∂x R ∂x∂θ 2 
 ∂ 2 v ∂w  
= −qR  − =0 
 ∂x∂θ ∂x  
(3.135)
 ∂ 2 u A22 ∂ 2 v ∂ 3 w B22 ∂ 3 w   A ∂ 2u ∂ 2v 2B ∂3 w 
 A12 + 2
+ B11 2
+ 2 3 
+ R  66 + A66 2 + 66 2  −
 ∂θ∂x R ∂θ ∂θ∂x R ∂θ   R ∂x∂θ ∂x R ∂x ∂θ 
1 ∂u B ∂ v 2
∂ w D22 ∂ w   B66 ∂ u
3 3 2
∂ v 2 D66 ∂ 3 w 
2
−  B12 + 22 2 + D12 + +
  + B66 + =
R  ∂θ∂x R ∂θ ∂θ∂x 2 R ∂θ3   R ∂x∂θ ∂x 2 R ∂x 2 ∂θ 
 ∂2v 
= −Nx R 2  (3.136)
 ∂x 
B66 ∂3u ∂3 v 4D ∂4 w  ∂3u B ∂v ∂4 w D ∂4 w  

−2 2
− 2 B66 2 − 66 2 2 + R  B11 3 + 12 2 + D11 4 + 122  +
R ∂x 2 ∂θ 2  
R ∂ x∂ θ ∂x ∂ θ R ∂x ∂θ  ∂x R ∂x ∂θ ∂x
1 ∂3u B ∂3 v ∂4 w D ∂4 w   ∂2 w   ∂2 w  
+  B21 2 + 22 3 + D12 2 2 + 222 = +q w + 2  − N x R 2  
4 
R ∂θ ∂ x R ∂θ ∂θ ∂x R ∂θ   ∂θ   ∂x  
(3.137)
Assuming that the origin of the coordinates is located at the one end of the pipe
(see Fig. 3.13), the following general solutions for equations (3.135)-(3.137) will
be used:
(mπ x) 
u = A ⋅ sin(nθ ) ⋅ cos
L 

(mπ x) 
v = B ⋅ cos( nθ ) ⋅ sin (3.138)
L 

(mπ x) 
w = C ⋅ sin(nθ ) ⋅ sin
L 
where A, B, C, are unknown constants. The above equation assumes that during
buckling, the generators of the shell subdivide into m half-waves and the circumfer-
ence into 2n half-waves. Combining equations (3.138) and (3.135)–(3.137) yields:
 F11 F12 F13   A 

 
 F21 F22 F23   B  = 0 (3.139)
F F32 F33  C 
 31
where:
A11m 2 π 2 R 3
F11 = + A66 n 2 R + (3.140)
L2
A12 mn πR 2 A66 mn πR 2 mn πR 3
F12 = − + − q (3.141)
L L L
B12 mn 2 πR 2 B66 mn 2 πR B11m3 π3 R 3

F13 = + + (3.142)
L L L3
B12 mn πR 2 B66 mn πR 2 A12 mn πR 3 A66 mn πR 3
F21 = − − − (3.143)
L L L L
B66 m 2 π 2 R 3 A66 m 2 π 2 R 4 m 2 π 2 R 4
F22 = B22 n 2 R − A22 n 2 R 2 − − − N x (3.144)
L2 L2 L2
B12 mn πR 2 B66 mn πR 2 A12 mn πR 3 A66 mn πR 3
F23 = − − − (3.145
L L L L
B12 mn 2 πR 2 2 B66 mn 2 πR 2 B11m3 π3 R 4
F31 = − + (3.146)
L L L3
B12 m 2 n π 2 R 3 2 B66 m 2 n π 2 R 3
F32 = B22 n3 R + − (3.147)
L2 L2
2 D12 m 2 n 2 π 2 R 2 4 D66 m 2 n 2 π 2 R 2
F33 = D22 n 4 + − + qR 3 −
L2 L2
(3.148)
2 3 m2 π2 R 4 D11m 4 π 4 R 4
n qR − Nx +
L2 L4
The equation for calculating the critical value of the pairs of external pressure qcr
and axial compression N xcr has the following form:
 F11 F12 F13 

det  F21 F22 F23  = 0 (3.149)
 F31 F32 F33 
In Figure 3.14 the curves N xcr vs qcr for several pairs (m,n) are schematically dis-
played. However, according to eq. (3.128), N x cr represents the quantity
( )
−1
 π D / 4 + a11 / d11  M cr
2
For the range of values M and q occurring in engineering applications, these

curves are almost straight lines. The safe values N xcr , qcr are included within the
polygon OABCD.
Taking the point of intersection of the polygon with the horizontal axis (point
D), we obtain the critical value of q, providing that only lateral external pressure
alone is acting. On the other hand, the intersection of the polygon with the vertical
axis (point A) indicates the critical value of bending moment M when q = 0. Since
the procedure for deriving the lines M cr vs qcr for several values of (m,n) is
Figure 3.14 Curves N xcr vs qcr for several pairs (m, n).
complicated, from a practical point of view it is adequate to determine the points

D and A and to draw the line AD. The points within the area of triangle OAD can
be used for engineering purposes and will deliver safe results.
3.5 Axial Tension

A multilayered pipe with a mean diameter D made by a wall composed by NP
layers with a stacking sequence [ ±θ ] (Fig.3.15) is considered.
∧
We shall assume that the load N is distributed uniformly around the circum-
ference of the cross-section. Therefore, the load per unit length Nx of laminate is:
∧
N
Nx = (3.150)
πD
This is the only external load acting on the laminate constituting the wall of the
pipe and thus
Ny = 0 (3.151)
N xy = 0 (3.152)
Mx = 0 (3.153)
Axial Tension 89
My = 0 (3.154)
M xy = 0 (3.155)
∧
In order to estimate the allowable force N , a failure criterion, e.g., the Tsai-Wu
criterion, should be applied. To this scope, the principal stresses σ1, σ2, τ12 have to
be determined by following the general procedure shown in Fig. 3.16.
Taking into account this procedure, as well as equations (3.150)–(3.155) and
the formula for the inverse ABD matrix, one can formulate the following:
ε x0 = a11 N x (3.156)
ε y0 = a12 N x (3.157)
γ xyo = a16 N x (3.158)
k x0 = b11 N x (3.159)

İ [
1[1\1[\
$%' H F H \J [\
.LUFKKRII
İ \
0[0\0[\ 0DWUL[

N N N K\SRWKHVLV
[ \ [\
Ȗ[\

$OORZDEOHD[LDO 6WUHVV
)DLOXUH
FULWHULRQ ııĲ >7@ ı[ı\Ĳ[\ VWUDLQ
IRUFH 1 UHODWLRQ
∧
figure 3.16 Concept for estimation of allowable axial force N .
k y0 = b12 N x (3.160)
k xy0 = b16 N x (3.161)

Therefore:
ε x = ε x0 + zk x0 (3.162)
ε y = ε yo + zk y0 (3.163)
γ xy = γ xyo + zk xy0 (3.164)
Thus, the stresses σx, σy, τxy for each lamina with fibers orientation θ can be ob-
tained by the well-known stress-strain relation:
σ x = Q11 (θ ) ⋅ ε x + Q12 (θ ) ⋅ ε y + Q16 (θ ) ⋅ γ xy (3.165)
σ y = Q 21 (θ ) ⋅ ε x + Q 22 (θ ) ⋅ ε y + Q 26 (θ ) ⋅ γ xy (3.166)
τ xy = Q16 (θ ) ⋅ ε x + Q 26 (θ ) ⋅ ε y + Q 66 (θ ) ⋅ γ xy (3.167)
With the aid of matrix [T(θ)], the principal stresses σ1, σ2, τ12 for every lamina can
be determined by the following matrix equation:
σ 1  σ x 
   
σ 2  = [ Τ(θ ) ] σ y  (3.168)
τ   
 12  τ xy 
Combination of Bending and Axial Tension 91
By applying the Tsai-Wu failure criterion for the values σ1, σ2, τ12 of every lamina,
F1σ 1 + F2σ 2 + F11σ 12 + F22σ 22 + F66τ 122 − F11 F22 σ 1σ 2 ≤ 1 (3.169)
where:
1 1
F1 = ( + ) (3.170)
s 1Τ s 1C
1 1
F2 = ( + ) (3.171)
s 2Τ s 2C
1
F11 = − (3.172)
s 1Τs 1c
1
F2 = − (3.173)
s 2Τs 2c
1
F66 = ( F ) 2 (3.174)
τ 12
∧
the allowable axial tension Ν will be obtained. Since this procedure yields differ-
∧
ent values of Ν , we have to adopt the minimum one.
3.6 Combination of Bending and Axial Tension

Superposition of the stresses σx, σy, τxy due to axial tension given by:
 h 
σ xN  α11 + 2 b11 
 N   �
 h N X
σ y  = Q ij (θ )  a12 + b12  (3.175)
 N   2  π D
τ xy   h 
a16 + 2 b16 
 
with the corresponding stresses due to bending

 1 
σ xΜ   S ( µ6 C1 + µ7 C2 + µ8 Α 
 Μ   33 
σ y  =  µ 4 C1 + µ5C2 + 3µ3 Α  (3.176)
 Μ  0 
τ xy   
 
Indicate that the following formulas for the principal stresses can be obtained:
σ 1 = (σ xΜ + σ xN ) cos 2 θ + (σ yΜ + σ yN ) sin 2 θ (3.177)
σ 2 = (σ xΜ + σ xN ) sin 2 θ + (σ yΜ + σ yN ) cos 2 θ (3.178)

τ 12 = (σ xΜ + σ xN )(− cos θ sin θ ) + (σ yΜ + σ yN ) cos θ sin θ + (τ xyΜ + τ xyN )(cos 2 θ − sin 2 θ )
(3.179)
It should be noticed that equation (3.176) provides the stresses of the exterior
lamina at the location φ = π/2 (location of the maximum bending stresses). In
equation (3.175), h denotes the thickness of the lamina, D is the exterior diameter
of the pipe, and N is the axial force.
Since the principal stresses σ1, σ2, τ12 are known, the Tsai-Wu failure criterion
can be used for derivation of the diagrams Nα = Να(M), providing the allowable
combinations of the axial force Να and bending moment M.

∧ ∧
When the combined bending moment M y and axial tension N x are acting
∧
on a pipe as in Figure 3.17a, both compressive (due to M y ) and tensile (due to
∧
N x ) stresses are placed upon the narrow strip AB. In cases where the value of the
∧
compressive stress due to M y is higher than the value of the tensile stress due to
∧
N x , a check for buckling has to be performed. To carry out such an inspection, the
concept followed for the approximation of a critical value of the bending moment
for the case of pure bending will again be adopted. However, the value of the local
strain ε ξο at the point A located at the center of the cross-section of the strip AB
∧ ∧
(Fig. 3.17a) is correlated with both bending moment M y and axial tension N x
through the formula:
Combination of Bending and Axial Tension 93
Figure 3.17 (a) Buckling of the narrow strip AB due to combined bending and tension,
(b) Buckling of the narrow strip AB due to axial load.
ο
M y D α11 ∧
ε =
ξ − Nx (3.180)
EI yy 2 π D
where [11]:
π
 1 D2 1 D
EI yy = 2 ∫  cos 2 a + cos 2 a  dθ (3.181)
α 4
0  11
d11 2
∧
In case ε ξο ≤ 0 , the tensile strain due to N x is greater than the compressive stain
∧
due to M y ; therefore, the check for local buckling is not needed. On the other
hand, we recall that the critical value of ε ξο , namely ε ξοc , is:
ε ξοc = a11λcr (3.182)
Therefore, combining equations (3.180) and (3.182), the boundaries of the allow-
∧ ∧
a
able values of M y and N xa in order to avoid local buckling, can be estimated by
the following relation:
ο
M y D α11 ∧
ε =
ξ − N x = a11λcr (3.183)
EI yy 2 π D
under the condition ε ξ0 > 0 , or, according to eq.(3.180):
∧
M y D a11 ∧
> Nx (3.184)
EI yy 2 πD
We recall that the value of λcr should be determined by the following equation
[11]:
 O L   M 0 M n  O L 
T
 Φ ⋅ [ J ] Φ 2 ⋅ [ J ] 
det   −λ  1   = 0 (3.185)
  L O   M n   


M 0   L O Φ 2 ⋅ [ J ] Φ1 ⋅ [ J ]  
where the matrices [O], [L], [Mo], [Mn], [J] and the functions Φ1, Φ2 are given in
section 3.3.
It is important to remind that equation (3.185) contains four unknown con-
stants a, b, c1, c2. For different combinations of these constants, eq. (3.185) yields
different values for λ. The lowest value of λ is the appropriate parameter λcr to be
used in eq. (3.183).
3.7 Combination of External Pressure and Axial

Tension
The composite pipe shown in Fig. 3.18(a) has a mean diameter D. Its wall is
composed of NP number of layers with the stacking sequence [±θ ] . This pipe is
∧
subjected to combined axial tension N x in the x-direction and external pressure
p. The axial load per unit length of the perimeter is:
∧
N x = N x / πD (3.186)
According to section 3.4, external pressure causes in-plane loading in the y-direction
(Fig. 3.18b) given by:
1
Ny = pD (3.187)
2
Combination of External Pressure and Axial Tension 95
Because of the absence of torsion and bending moments, i.e. Nxy = 0, Mx = 0,

My = 0, Mxy = 0, the following matrix equation can be derived:

ª D D D E E E − º 1[ ½ ½

«D ° ° ° °
« D D E E E − »» ° 1 \ ° ° °
« D D D E E E − » ° 1 [\ ° ° °
« » ° ° ° °
« E E E G G G − » ° 0 [ ° ° °
« E E E G G G − » ° 0 \ ° ° °
« » ° ° ° °
« E E E G G G −» ° 0 [\ ° ° °
« ⋅® ¾ = ® ¾
» ° ε [ ° ° 1Ö [ π ' °
« »
» ° ε \ ° ° S' °

«
« ° ° ° °
» ° γ [\ ° ° °

« »
» ° N[ ° ° °

«
« » °
° ° °
« » ° N\ ° ° °
«¬ »¼ °¯ N [\ ¿° °¯ °¿

(3.188)
Solution of above equation yields the strain components:
ε x0 , ε y0 , γ xy0 , k x0 , k y0 , k xy0
∧
Nx pD
ε x0 = a11 + a12 (3.189)
πD 2
∧
0 N pD
ε = a12 x + a22
y (3.190)
πD 2
∧
0 N pD
γ xy = a16 x + a26 (3.191)
πD 2
∧
0 N pD
k = b11 x + b12
x (3.192)
πD 2
∧
0 N pD
k = b12 x + b22
y (3.193)
πD 2
∧
0 N pD
k = b16 x + b26
xy (3.194)
πD 2
Therefore:
ε x = ε x0 + zk x0 (3.195)
ε y = ε y0 + zk y0 (3.196)
γ xy = γ xy0 + zk xy0 (3.197)
With the aid of equations (3.189)–(3.197), the stress components σx, σy, τxy, can be
obtained for every lamina using the well-known relation:
σ x  ε x 
   
 
σ y  = Q ij (θ )  ε y  (3.198)
   
τ xy  γ xy 
Using the matrix [ Τ(θ ) ] the principal stresses σ1, σ2, τ12 can be obtained by the
following equation:
Torsion 97
σ 1  σ x 
    
σ 2  = T (θ )  σ y  (3.199)
τ    
 12  τ xy 
With the aid of eq. (3.199), the Tsai-Wu failure criterion yields the areas of the
∧
allowable combinations of the values of axial force N x and external pressure p
for avoiding failure for every ply.
The failure analysis described above is not sufficient for design purposes. The
derived geometry of the pipeline must be additionally checked by the buckling
model described in section 3.3.2. Acceptable pipe diameter (Dia), fiber orienta-
tion [±θ], number of plies (NP), thickness of lamina (h) are ones that satisfy both
the failure criterion and the buckling model.
3.8 Torsion
When a torque M �
x (see Fig. 19a) is applied to a long composite pipe, a re-
sultant Nξ n is acting on the cross-section of the wall (see Fig. 19b). Taking into
account the equilibrium between Nξ n and M �
x it can be written as:
�
M
Nξ n = x
(3.200)
2π R 2
where R is the radius of the pipe (R = D/2).
Taking into account that:
Nξ = 0 (because of absence of axial force) (3.201)
N n = 0 (because of absence of external pressure) (3.202)
M ξ = M n = M ξ n = 0 (because of absence of moments) (3.203)
The corresponding strains ε ξo , ε no , γ ξon , kξo , kno , kξon can be obtained by the following
matrix equation:
Figure 3.19 Geometry of a composite pipe subjected to torsion.
ª D D D E E E − º 1ξ ½ ½

«D » °1 ° ° °
« D D E E E − » ° Q ° ° °
« D D D E E E − » ° 1ξ Q ° ° °
« » ° ° ° °
« E E E G G G − » ° 0ξ ° ° °
« E E E G G G − » ° 0 Q ° ° °
« » ° ° ° °
« E E E G G G 0
−» ° ξ Q ° ° °
« ⋅ ® ¾ = ® ¾ (3.204)
» ° εξ ° °

°
« »
» ° εQ ° ° °

«
« » ° ° ° °
° γ ξ Q ° ° 0Ö [ π 5 °
« »
» ° Nξ ° ° °

«
« » ° N ° ° °
« » ° Q ° ° °
«¬ »¼ °¯ Nξ Q °¿ °¯ °

¿
The solution of the above equation yields:
�
a16 M
ε ξo = x
(3.205)
2π R 2
a M �
ε no = 26 2x (3.206)
2π R
a M �
γ ξon = 66 2x (3.207)
2π R
Torsion 99
�
b61 M
kξo = x
(3.208)
2π R 2
�
b62 M
kno = x
(3.209)
2 πR 2
b M �
kξon = 66 2x (3.210)
2π R
Therefore
 ε ξ   ε ξo   kξo 
   o  o
 εn  =  εn  + ζ  kn  (3.211)
γ  γ o  k o 
 ξn   ξn   ξn 
For symmetric lay-up of fibers, the maximum shear stresses τ ξ n take place in
the exterior layers of the pipe. Therefore, for ζ = h / 2 equations (3.205)–(3.211)
yield:
 h 
 a16 + 2 b16 
 εξ   
  �
M h 
x 
 εn  = 2  a26 + b62  (3.212)
γ  2π R  2 
 ξn   h 
 a66 + 2 b66 
 
Taking into account the above equation, the stress-strain relation for the exterior
lamina with fiber orientation θ can be written as:
σ ξ   εξ 
   
σ n  = Qij (θ )  ⋅  ε n  (3.213)
τ  γ 
 ξn   ξn 
Therefore, the principal stresses of the exterior lamina can be obtained by the fol-
lowing well-known formula:
σ 1  σ ξ 
   
T 
σ 2  =  (θ )  ⋅ σ n  (3.214)
τ  τ 
 12   ξn 
� a can be obtained by
With the aid of the above equation, the allowable torque M x
the Tsai-Wu criterion:
F1σ 1 + F2σ 2 + F11σ 12 + F22σ 22 + F66τ 122 − F11 F22 σ 1σ 2 ≤ 1 (3.215)
where
F1 = 1 / σ 1T + 1 / σ 1C , F2 = 1 / σ 2T + 1 / σ 2C , F66 = (1 / τ 12F ) 2 , F11 = −1 / σ 1T σ 1C , F2 = −1 / σ 2T σ 2C .

A long laminated pipe of diameter D, made from a multilayered composite
wall with the stacking sequence [ ±φ ] is subjected to pure torsion. The target now
� x that causes elastic buckling.
is the estimation of the critical torque M
When one considers:
(a) the geometric relations between the displacements u, v, w along the axes x,
y, z respectively;
(b) the equilibrium of the resultants acting in the wall, with torque Mx
(c) the equilibrium of exterior shear force per unit length T applied at the edges
of the pipe, with torque M� x , i.e. 2 πR 2T = M � x / 2 πR 2
� x or T = M
(d) the fact that approximation 1+ ( z / R) ≈ 1 yields M xy = M yx and N xy = N yx
[9] the following relations based on [9] can be written as:
∂N x ∂N xy M� x ∂ 2u
R2 +R − =0 (3.216)
∂x ∂θ πR ∂x∂θ
∂N y ∂N xy ∂M y ∂M xy � x  ∂ 2 v ∂w 
M
R + R2 − −R −  + =0 (3.217)
∂θ ∂x ∂θ ∂x πR  ∂x∂θ ∂x 
∂2 M y ∂ 2 M xy ∂2 M x � x  ∂v ∂ 2 w 
M
2
+ 2R + 2
+ RN y −  − =0 (3.218)
∂θ ∂x∂θ ∂x πR  ∂x ∂x∂θ 
Taking into account equations (3.134) as well as the following type of the
solutions:
 λx 
u = Α sin  + mθ (3.219)
 R 
Torsion 101
Figure 3.20 Coordinate system of a pipe under torsion.
 λx 
v = Βsin  + mθ (3.220)
 R 
 λx 
w = C cos  + mθ (3.221)
 R 
n πR
where λ = and n, m integers, the following system of equations can be
derived: L
 G11 G12 G13   Α

G  
G22 G23  Β  = 0 (3.222)
 21  
 31 G32
G G33  C 
where:
G11 = − Α66 m 2 − Α11 λ2 (3.223)
mλ �
G12 = − Α12 m λ − Α66 m λ + Mx (3.224)
πR 2
Β12 2 2Β Β
G13 = m λ + 66 m 2 λ + 11 λ3 (3.225)
R R R
Β21 Β
G21 = − Α12 m λ − Α66 m λ + m λ + 66 m λ (3.226)
R R
Β 22 2 m λ � Β
G22 = − Α 22 m 2 + m + 2
M x − Α66 λ2 + 66 λ2 (3.227)
R πR R
D22 3 Β22 3 λ � D D
G23 = − 2
m + m − 2
M x − 122 m λ2 − 2 662 m λ2 +
R R πR R R (3.228)
Β12 2 Β
+ m λ2 + 66 m λ2
R R
Β21 2 2Β Β
G31 = Α12 λ − m λ − 66 m 2 λ − 113 λ3 (3.229)
R R R
Β22 3 λ � Β 2Β
G32 = Α 22 m − m − 2
M x − 123 m λ2 − 66 m λ2 (3.230)
R πR R R
D22 4 Β 22 2 m λ � D D
G33 = 2
m − m − 2
M x + 124 m 2 λ2 + 122 m 2 λ2 +
R R πR R R
(3.231)
4 D66 2 2 Β12 2 D11 4
+ 2 m λ − λ + 4 λ
R R R
The homogeneous matrix equation (3.222) can yield non-zero solutions for A, B,
C only if the following determinant is zero:
 G11 G12 G13 

 
det  G21 G22 G23  = 0 (3.232)
G G33 
 31 G32
Therefore, the above equation yields an expression for Mx versus the parameters
m and λ. It should be noted that parameter m is an integer while λ is a real number.
For m = 1, it can be seen from equations (3.219)–(3.222) that a cross-section of
the pipe remains circular and moves, during buckling, only in its plane. For m = 1,
in order to calculate the critical value of Mx that causes elastic buckling, we have
to find the value of λ that minimizes the expression for Mx. Normally the minimi-
zation of the function Mx cannot be achieved by an analytical method. However,
for existing commercial software like Mathematica, numerical minimization of
Mx is an easy task.
References
[1] Pavlou D.G., “Computational and experimental analysis of damaged mate-
rials,″ Transworld Research Network, 2007.
References 103
[2] Guedes R.M., Creep and fatigue in polymer matrix composites, Woodhead
publishing, 2011.
[3] Harris B., Fatigue in composites, CRC press, 2003.
[4] Berenberg B., http://composite.about.com
[5] Lekhnitskii S.G., Theory of elasticity of an anisotropic elastic body, Holden-
Day, 1963.
[6] Xia M., Takayanagi H., and Kemmochi K., “Bending behavior of filament-
wound fiber-reinforced sandwich pipes,” Composite Structures 56, 201–
210, 2002.
[7] Reddy J.N., Mechanics of laminated composite plates and shells, CRC
press, 2004.
[8] Tsai S.W., and Wu E.M., “A general theory of strength for anisotropic ma-
terials,″ Journal of Composite Materials 5, 58–80, 1971.
[9] Flügge W., Stresses in shells, Springer-Verlag, 1973.
[10] Timoshenko S.P., and Gere J.M., Theory of elastic stability, Dover publica-
tions 2009.
[11] Kollár L.P., and Springer G.S., Mechanics of composite structures,
Cambridge University Press, 2003.
[12] Rasheed H.A., and Karamanos S.A., “Stability of tubes and pipelines,” in:
Buckling and Postbuckling Structures, edited by Falzon B.G. and Aliabadi
M.H., Imperial College Press, 2008.
Chapter 4
Dynamic Stability of Composite Pipelines
The first known study of the dynamic analysis of pipes conveying liquids ap-
peared in 1950 [1]. It represented an attempt to explain vibrations observed in
the Trans-Arabian pipeline. Today the most complete analysis for flow-induced
vibrations of isotropic pipes (e.g., steel) can be found in the works of Païdoussis
[2,3]. In this chapter, analysis of dynamic stability will be performed for multilay-
ered filament-wound composite pipes.
4.1 Free Vibration of Composite Pipes

Since composite pipes are thin-walled, lightweight structures, they can be ana-
lyzed as slender beams. We consider a composite pipe of length L, flow area A
and mass per unit length m, conveying fluid of mass per unit length M with mean
axial flow velocity U. Due to axial flow, small lateral motions w(x, t) of a long
wavelength (compared to the diameter) will be observed. The aim of the present
section is to estimate the eigenfrequencies and eigenmodes, taking into account
the structural characteristics of the multilayered composite pipe and its interaction
with the fluid flow.
4.1.1 Structural characteristics of composite pipes

When lateral motions w(x, t) are induced due to bending, the main structural
parameter controlling vibration is the stiffness EI� yy (with respect to axis y) of the
composite pipe.
For a wall element such as that illustrated in Figure 4.1a located at (z, a) in the
cross-section of a multilayered pipe (Figure 4.1b), the local strain and curvatures
in its middle plane are:
105
106 Dynamic Stability of Composite Pipelines
Figure 4.1 (a) System of coordinates of a wall element of a multilayered composite pipe,
(b) System of coordinates of a multilayered composite pipe.
1
ε ξo = z (4.1)
ρy
1
κ ξo = cos a (4.2)
ρy
where 1/ ρ y is the curvature of the pipe due to its bending.

On the other hand, the relationships between the strains and curvatures with
the forces and moments for a symmetrical laminate are given by the following
well-known [4] equations:
Free Vibration of Composite Pipes 107
 ε ξo   a11 a12 a16   Nξ 

 o   
 ε n  =  a12 a22 a26   N n  (4.3)
γ o   a a26 a66   Nξ n 
 ξ n   16
 κ ξo   d11 d12 d16   M ξ 

 o   
 κ n  =  d12 d 22 d 26   M n  (4.4)
κ o   d d 26 d 66   M ξ n 
 ξ n   16
Taking into account that for the wall element of Fig. 4.1a is N n = 0, Nξ n = 0, M n = 0,
M ξ n = 0, eqs (4.3), (4.4) above yield:
1 o
Nξ = εξ (4.5)
a11
1 o
Mξ = κξ (4.6)
d11
An elementary bending moment with respect to the axis of bending y is:
� y = ( N dn) z + ( M dn) cos a

dM (4.7)
ξ ξ
With the aid of equations (4.5) and (4.6), the above equation yields:
� y= 1 o 1 o
dM εξ ⋅ zdn + κ ξ cos a ⋅ dn (4.8)
a11 d11
Taking into account equations (4.1), (4.2), the above equation can now be written
as:
� y= 1 1 2 1 1
dM z dn + cos 2 adn (4.9)
ρ y a11 ρ y d11
� y acting in the cross-section of the composite

Therefore, the bending moment M
pipe is:
� y= 1  1 1 
M
ρy ∫  a
(s) 11
z2 +
d11
cos 2 a  dn

(4.10)
where a11 , d11 are members of the compliance matrix and
D
z= cos a (4.11)
2
D
dn = da (4.12)
2
With the aid of eqs. (4.11), (4.12), eq. (4.10) yields:
� y = 1  D + 1  D cos 2 ada
π
ρ y  2a11 d11  ∫o
M   (4.13)
or
πD  D 1 
 + 
� y= 2 2
 11a d11 
M (4.14)
ρy
� yy of a beam subjected to bending (with respect to

It is known that the stiffness EI
the y-axis) is given by the following equation:
� yy = M
EI � y⋅ρ (4.15)
y
Combining eqs. (4.14), (4.15), it can be concluded that the stiffness of a multilay-
ered composite pipe is given by:
� yy = πD  D + 1 
EI (4.16)
2  2a11 d11 
It should be recalled that the parameters a11 , d11 can be derived by the inversion
of the stiffness matrix of the laminated wall, i.e.,
−1
 a11 a12 a16 b11 b12 b16   A11 A12 A16 B11 B12 B16 
a b26   A12 
 12 a22 a26 b21 b22 A22 A26 B12 B22 B26 
 a16 a26 a66 b61 b62 b66   A16 A26 A66 B16 B26 B66 
 = 
 b11 b21 b61 d11 d12 d16   B11 B12 B16 D11 D12 D16 
b b b d12 d 22 d 26   B12 B22 B26 D12 D22 D26 
 12 22 62   
 16 26 b66
b b d16 d 26 d 66   B16 B26 B66 D16 D26 D66 
(4.17)
where
N
Aij = ∑ Q ijk ( zk − zk −1 )
k =1
1 N
Bij = ∑ Q ijk ( zk2 − zk2−1 ) (4.18)
2 k =1
1 N
Dij = ∑ Q ijk ( zk3 − zk3−1 )
3 k =1
In the above equation, zk , zk −1 are the distances of corresponding layers from the
middle plane of the cross-section of the laminate in Figure 4.2a.
(b) Plan form View
figure 4.2 (a) cross-section and (b) plan form view of the laminate [5].
The transformed reduced stiffnesses Q ij are given by the following equations:
Q11 = Q11m 4 + 2 ( Q12 + 2Q66 ) n 2 m 2 + Q22 n 4 


Q12 = ( Q11 + Q22 − 4Q66 ) n m + Q12 ( n + m )
2 2 4 4


Q16 = ( Q11 − Q12 − 2Q66 ) nm3 + ( Q12 − Q22 + 2Q66 ) n3 m 
 (4.19)
Q 22 = Q11n 4 + 2 ( Q12 + 2Q66 ) n 2 m 2 + Q22 m 4 
3
Q 26 = ( Q11 − Q12 − 2Q66 ) n m + ( Q12 − Q22 + 2Q66 ) nm 
3

Q 66 = ( Q11 + Q22 − 2Q12 − 2Q66 ) n 2 m 2 + Q66 ( n 4 + m 4 ) 
where
n = sin θ (4.20)
m = cos θ (4.21)
E1
Q11 = (4.23)
1 −ν 12ν 21
v12 E2 v E
Q12 = = 21 1 (4.24)
1 − v12 v21 1 − v12 v21
E2
Q22 = (4.25)
1 − v12 v21
Q66 = G12 (4.26)
In the above equations (4.20)–(4.26), θ is the fiber direction of each ply, while
E1 , E2 , v12 , v21 , G12 are the modulus of elasticity (E), Poisson’s ratio (ν) and
shear modulus (G) in the principal directions (i.e., the directions along and normal
to the directions of fibers) of a ply. The above procedure for the derivation of the
stiffness of a composite pipe can be summarized in the following diagram:
Figure 4.3 Schematic procedure for the derivation of the stiffness of multilayered com-
posite pipe.
4.1.2 Forces and bending moments acting on a composite pipe element

Each span of the composite pipe depicted in Figure 4.4 has length L, inter-
nal diameter D and is fixed by supports at both its ends. During fluid flow, the
forces and moments acting on an elementary section of pipe are demonstrated in
Figure 4.5.
Figure 4.4 Geometry of a composite pipe conveying fluid.
Figure 4.5 Forces and moments acting on an elementary pipe section.
Acting in the length direction of the deformed pipe are: (a) shear stresses q
due to the friction between the fluid and the interior cylindrical surface πDdx, and
(b) longitudinal tension T.
Acting in the direction normal to the deformed pipe are: (a) shear forces Q,
(b) reaction forces F between the fluid and the interior surface of the pipe, verti-
cal gravity forces mgdx due to the weight of the material of the corresponding
pipe element, and vertical dynamic forces mazp due to the vertical motion of the
elementary mass m of the pipe ( azp is the vertical acceleration). Acting in the y-
direction are the bending moments M � y . Taking into account that the pipe element
has a slope ∂w / ∂x (with respect to the x-axis) due to bending and applying the
equilibrium conditions of the forces projected on the axes x, z, y, the following
can be determined.
4.1.2.1 equilibrium of forces projected in the x-direction
∂T ∂w
+ πDq − F =0 (4.27)
∂x ∂x
4.1.2.2 equilibrium of forces projected in the z-direction
∂Q ∂  ∂w  ∂w
+ F + T  + πDq + mg = mazp (4.28)
∂x ∂x  ∂x  ∂x
4.1.2.3 equilibrium of bending moments in the y-direction
� y
∂M
= −Q (4.29)
∂x
Taking into account equation (4.15), the equation above can be written as:
� yy ∂  1
EI

 = Q (4.30)
∂x  ρ y 
where EI� yy is given by eq. (4.16).

From geometry it is well known that the curvature (1/ ρ y ) is associated with
the deflection w as expressed in the following relation:
1 ∂2 w
= (4.31)
ρ y ∂x 2
Therefore, with the aid of eq. (4.31), eq. (4.30) yields:
� yy ∂ w
3
Q = − EI (4.32)
∂x 3
In the direction along the axis of the deformed pipe, the forces acting on a fluid
element shown in Figure 4.6 are: (a) forces due to pressure p acting on the cross-
section area A = πD 2 / 4 ; (b) shear forces q due to friction of the fluid with the
interior cylindrical surface of the pipe; and (c) horizontal dynamic forces due to
the motion of the fluid in the x and z directions. In the direction normal to the
axis of deformed pipe, the forces acting on the fluid element are: (a) reaction
forces F between the fluid and the interior surface of the pipe; (b) the vertical
gravity forces Mgdx due to the weight of the fluid element; and (c) vertical
dynamic forces due to the motion of the fluid (in the x and z directions).
4.1.2.4 equilibrium of forces projected in the x-direction

∂p ∂w
−A − q πD + F = Maxf (4.33)
∂x ∂x
In the above equation, axf is the horizontal acceleration of the fluid due to its
flowing motion in the x-direction, and in the z-direction because of the vertical
motion of the pipe.
4.1.2.5 equilibrium of forces projected in the z-direction

∂  ∂w  ∂w
−F − A  p  − q πD − Mg = Mazf (4.34)
∂x ∂x ∂x
where azf is the vertical acceleration of the fluid due to its motion in the x-direc-
tion due to flow, and in the z-direction due to the vertical motion of the pipe.
4.2 Accelerations of the Fluid and Pipe Elements

We consider that the motion of the pipe element takes place only in the vertical
direction. Therefore its acceleration in the horizontal direction is:
axp = 0 (4.35)
Figure 4.6 Forces acting on a fluid element.

Accelerations of the Fluid and Pipe Elements 115
As the vertical velocity of the pipe element is ∂w / ∂t , its vertical acceleration

will be:
∂2 w
azp = (4.36)
∂t 2
The definition of the acceleration of the fluid element is more complicated since
this motion takes place in two directions, i.e., in the direction along the pipe due
to its velocity U, and in the vertical direction z, which is due to the vertical motion
of the pipe (see Figure 4.7).
As the slope of the pipe ∂w / ∂x is very small, we can use the following
approximations:
cos(∂w / ∂x) ≈ 1 (4.37)
sin(∂w / ∂x) ≈ ∂w / ∂x (4.38)
Therefore, the projection of the velocity U along the axes x and z has the follow-
ing values:
Ux = U (4.39)
∂w
Uz = U (4.40)
∂x
Figure 4.7 Vectors of velocities of the fluid element in the z-direction.

Taking into account that the vertical velocity of the pipe is ∂w / ∂t , the vertical
velocity of the fluid element can be written as:
∂w ∂w
U zf = +U (4.41)
∂t ∂x
while the horizontal velocity of the fluid element is
U xf ≈ U (4.42)
rf ur f
Taking into account that a = ∂U / ∂t , the accelerations of the fluid element in
the vertical and horizontal directions can be obtained [2]:
∂2 w ∂2 w ∂ 2 w ∂U ∂w
azf = + 2U +U 2 2 + (4.43)
∂t 2
∂x∂t ∂x ∂t ∂x
and
∂U xf ∂U
axf = = (4.44)
∂t ∂t
4.3 Equation of Motion

Neglecting quantities associated with tensioning, pressurization effects and
gravity, which are not important for motion, and considering a constant velocity
U, the equilibrium equations of the pipe element (4.27), (4.28), (4.32) and the
fluid element (4.33), (4.34) can be simplified as follows:
∂w
F = πDq (4.45)
∂x
∂Q ∂w ∂2 w
+ F + πDq =m 2 (4.46)
∂x ∂x ∂t
� yy ∂ w
3
Q = − EI (4.47)
∂x 3
∂w
− q πD + F =0 (4.48)
∂x
∂w  ∂2 w ∂2 w 2 ∂ w
2
− F − q πD =M  2 + 2U + U  (4.49)
∂x  ∂t ∂x∂t ∂x 2 
Equation of Motion 117
From eq. (4.46) it can be obtained:
∂w ∂Q ∂2 w
− F − πDq = −m 2 (4.50)
∂x ∂x ∂t
The combination of equations (4.49) and (4.50) yields :
∂Q ∂2 w ∂2 w ∂2 w ∂2 w
− m 2 = M 2 + 2 MU + MU 2 2 (4.51)
∂x ∂t ∂t ∂x∂t ∂x
Taking into account eq. (4.47), the above equation can now be written as:
4 2 2 2
� y y ∂ w + MU 2 ∂ w + 2 MU ∂ w + ( M + m) ∂ w = 0
EI (4.52)
∂x 4 ∂x 2 ∂x∂t ∂t 2
� yy is given by eq. (4.16) and can be derived by the procedure illustrated
where EI
in Figure 4.3.
The homogeneous partial differential equation above is the equation of motion
describing the free vibration of the composite pipe under consideration.
4.3.1 Solution of equation of motion

The aim of this section is to calculate the eigenfrequencies ω and critical values
Ucr of fluid velocity that cause instability of a composite pipe conveying fluid. The
corresponding boundary-value problem consists of the homogeneous fourth-order
partial differential equation (4.52) and the boundary and initial conditions of the
problem. Assuming that the motion w(x,t) starts at t > 0, the initial conditions at
t = 0 can be written as:
w( x, t ) t =0 = ws ( x)
dws ( x)
w '( x, t ) t =0 =
dx
d 2 ws ( x) (4.53)
w "( x, t ) t = 0 =
dx 2
3
d ws ( x)
w "'( x, t ) t = 0 =
dx 3
where ws ( x) is the static deflection curve because of the uniform self-weight
q = (M + m). Taking into account the expression for obtaining the stiffness of a
composite pipe, i.e., equations (4.16)–(4.26), the function ws ( x) for several types
of support can be approximated as follows:
( M + m) x 2 2
ws ( x) = ( x + 6 L2 − 4 Lx)
�
24 EI yy for cantilever pipe
2
( M + m) x
ws ( x) = ( L − x) 2 for fixed-fixed pipe
� yy
24 EI
(4.54)
( M + m) L  L − x
4
( L − x )3 ( L − x) 4  for fixed-pinned
ws ( x) =  − 3 + 2 
48 EI  L L3 L4  pipe
( M + m) x 3 for pinned-pinned
ws ( x) = ( L − 2 Lx 2 + x 3 )
� yy
24 EI pipe
where L is the length of the composite pipe.

The boundary conditions depend on the type of supports. Table 4.1 summa-
rizes the boundary conditions for different types of support.
Table 4.1
Boundary conditions for several types of support.
Type Support Equation
Pinned
w(0, t ) = 0 (4.55)
∂ 2 w(0, t )
=0 (4.56)
∂x 2
Fixed
w(0, t ) = 0 (4.57)
∂w(0, t )
=0 (4.58)
∂x
Free
∂ 2 w(0, t )
=0 (4.59)
∂x 2
∂ 3 w(0, t )
=0 (4.60)
∂x 3
Type Support Equation
Deflected
spring ∂ 2 w(0, t )
=0 (4.61)
∂x 2
� yy ∂ 3 w(0, t )
− EI = K D w(0, t ) (4.62)
∂x 3
Torsion
spring
w(0, t ) = 0 (4.63)
2
� yy ∂ w(0, t ) = K ∂w(0, t )
EI (4.64)
T
∂x 2 ∂x
Mass
∂ 2 w(0, t )
=0 (4.65)
∂x 2
� yy ∂ 3 w(0, t ) ∂ 2 w(0, t )
− EI 3
= me (4.66)
∂x ∂t 2
Dashpot
∂ 2 w(0, t )
=0 (4.67)
∂x 2
� yy ∂ 3 w(0, t ) ∂w(0, t )
− EI 3
=c (4.68)
∂x ∂t
For the boundary value problem under consideration, we can use solutions of
the form:
w( x, t ) = Re[ f ( x)eiωt ] (4.69)
where f ( x) is a geometric function and ω is the circular frequency to be deter-

mined. Generally the circular frequency ω can be a complex number. The system
will be stable when Im[ω ] > 0 and unstable when Im[ω ] < 0 . In cases where ω is
a real number, the system has neutral stability.
Substitution of the solution (4.69) into the partial differential equation (4.52)
yields:
4 2
� yy d f ( x) + MU 2 d f ( x) + 2 MU ωi df ( x) − ( M + m)ω 2 f ( x) = 0 (4.70)
EI
dx 4 dx 2 dx
For the above equation we can try the solution
f ( x) = Ceiβ x (4.71)
With the aid of eq. (4.71), eq. (4.70) yields:
� yy β 4 − MU 2 β 2 − 2 MU ωβ − ( M + m)ω 2 = 0
EI (4.72)
The algebraic equation above has four solutions:
β1 = −κ 7 − κ 8 (4.73)
β 2 = −κ 7 + κ 8 (4.74)
β 3 = +κ 7 − κ 8 (4.75)
β 4 = +κ 7 + κ 8 (4.76)
where:
1 4 MU ω
κ8 = 2κ 0 − κ 6 − (4.77)
2 2κ 7 EI y y
1
κ7 = κ0 + κ6 (4.78)
2
κ1 3 2 κ5
κ6 = + (4.79)
3EI y yκ 5 EI y y 3 54
κ 5 = (κ 2 + κ 4 )1/ 3 (4.80)
κ 4 = −4κ13 + κ 32 (4.81)
κ 3 = −2 M 3U 6 + 108 EI y y M 2U 2ω 2 − 72 EI y y M (m + M )U 2ω 2 (4.82)
κ 2 = −2 M 3U 6 − 72 EI y y mMU 2ω 2 + 36 EI y y M 2U 2ω 2 (4.83)
κ1 = M 2U 4 − 12 EI y y (m + M )ω 2 (4.84)
2 MU 2
κ0 = (4.85)
3EI y y
Application of the general solution equations (4.69), (4.71), and (4.73-4.85) into
the boundary conditions for several combinations of supports (Table 4.1) yields
the following:
4.3.1.1 cantilever composite pipe

For this type of pipe the ends are fixed-free. Therefore the boundary conditions
are given by equations (4.57), (4.58), and (4.59), (4.60). With the aid of equations
(4.69), (4.71), and (4.73–4.85) the boundary conditions above can be written as:
C1 + C2 + C3 + C4 = 0 (4.86)
β1C1 + β 2 C2 + β 3C3 + β 4 C4 = 0 (4.87)
β12 eiβ1L C1 + β 2 2 eiβ2 L C2 + β 32 eiβ3 L C3 + β 4 2 eiβ4 L C4 = 0 (4.88)
β13 eiβ1L C1 + β 23 eiβ2 L C2 + β 33 eiβ3 L C3 + β 43 eiβ4 L C4 = 0 (4.89)
For a nontrivial solution (with respect to C1 , C2 , C3 , C4 ) of the above equations

(4.86)-(4.89), the corresponding determinant of the matrix of coefficients must
vanish:
 1 1 1 1 
 β β2 β3 β 4 
det  2 1iβ1L =0 (4.90)
 β1 e β 2 2 ei β 2 L β 3 2 ei β3 L β 4 2 ei β 4 L 
 3 iβ1L 
 β1 e β 2 3 ei β 2 L β 33 e i β 3 L β 4 3 ei β 4 L 
Numerical solutions for the eigen-frequencies ωi can be obtained from eq. (4.90)
by taking into account eqs. (4.73)–(4.85).
By applying the concept above, the following equations for obtaining the eigen-
frequencies ωi can be derived for several types of supports:
4.3.1.2 fixed-fixed composite pipe
 1 1 1 1 
 β β2 β3 β 4 
det  iβ11L =0 (4.91)
 e ei β 2 L ei β3 L ei β 4 L 
 iβ1L 
 β1e β 2 ei β 2 L β 3 ei β3 L β 4 ei β 4 L 
4.3.1.3 fixed-pinned composite pipe
 1 1 1 1 
 β β2 β3 β 4 
det  iβ11L =0 (4.92)
 2 iβ1L 
 β1 e β 2 2 ei β 2 L β 3 2 ei β3 L β 4 2 ei β 4 L 
4.3.1.4 pinned-pinned composite pipe
 1 1 1 1 
 β2 β 22 β 32 β 42 
det  iβ11L =0 (4.93)
 2 iβ1L 
 β1 e β 2 2 ei β 2 L β 3 2 ei β3 L β 4 2 ei β 4 L 
As already mentioned, a pipe conveying fluid is stable when Im[ωi ] > 0 and un-
stable when Im[ωi ] < 0 . Therefore, the critical value of the fluid flow Ucr for
passing from stability to instability can be estimated by the following equation:
Im[ωi ] = 0 (4.94)
4.3.2 Types of instability

The equation of motion (4.52) contains the superposition of four terms. The
meaning of these terms is:
� yy d w represents the elastic flexural restoring force.

4
• EI
dx 4
d 2w
• MU 2 represents the centrifugal force of the fluid flowing with constant
dx 2
speed for a curved portion of the pipe (because actually the derivative
d 2 w / dx 2 represents the value 1/ R , where R is the local radius of curvature).
∂2 w
• 2 MU represents the Coriolis force.
∂x∂t
∂2 w
• ( M + m) represents the inertia effects of the masses of fluid and pipe.
∂t 2
In the above four terms, only the second (centrifugal force) and third (Coriolis
force) are influenced by the flow velocity U. For small values of U, the dynamic
behavior of pipe is dominated by the Coriolis force that is proportional to U.
In that case, the system is subjected to flow-induced damping [3] because the
Coriolis forces in the fluid react to the pipe in a direction opposite the motion.
For higher values of U, the centrifugal force (which is proportional to U2) might
overcome the Coriolis damping effect, and the system can lose stability.
Instability can be introduced by either divergence (a static form of instability)
or flutter (a dynamic form of instability) [3]. Divergence may occur if both ends
of the pipe are supported (e.g., fixed-fixed, pinned-pinned, fixed-pinned). In the
case of cantilever pipe (where the one end is free), flutter may occur in the pipe.
For the special case that the direction of the flow is from the free end towards the
fixed one, the cantilever pipe will become unstable (due to flutter) for very small
values of U, and then be stabilized for larger values of U as first pointed out by
Païdoussis and Luu [6]. If we neglect the terms of Coriolis (causing damping in
low flow velocity) and inertial forces, eq. (4.52) yields:
4 2
� yy d w + MU 2 d w = 0
EI (4.95)
4
dx dx 2
From the above equation it is clear that the centrifugal force acts in the same
manner as a compressive axial force in a straight column. Therefore, when U
increases, the “compressive effect” of the centrifugal force can overcome the flex-
ural resistance of the pipe, which causes divergence. Since divergence is a static
rather than a dynamic form of instability, the critical flow U cr causing divergence
may be examined by considering only the time-independent terms of eq. (4.52),
so effectively eq. (4.95). From an engineering point of view, the solution of eq.
(4.52) can yield the critical values of U that cause instability for both cases of
pipes, i.e., cantilever or supported on both ends. However, the value of U that does
cause divergence instability in the second case, can be adequately obtained from
eq. (4.95). In cases where eq. (4.52) is used, obtaining the eigenvalues by solving
equations (4.90)–(4.93) is a very difficult task, even when advanced numerical
methods are implemented. On the other hand, eq. (4.95) has a very simple solu-
tion—but for a uniform pipe. However, for pipes where certain properties may
vary gradually along its length or for cases where long pipes are supported period-
ically by more than two supports, a direct solution is also quite difficult. For these
cases, approximate methods like the Galerkin or FEM can be effectively used for
obtaining the critical values of the velocity U cr that causes instability. In the au-
thor’s opinion, of existing methods for solving equations (4.52) or (4.95) that ap-
ply to cases involving additional supports or temperature gradients or additional
point masses or elastic foundation along the pipe’s length, the most advantageous
is the Transfer Matrices Method (TMM). The Transfer Matrices Method is es-
sentially an analytic method that reduces the higher-order differential equation
(here fourth-order) into a first-order matrix differential equation. The Transfer
Matrices Method provides simpler algebraic equations than eqs. (4.90)–(4.93),
which facilitates the derivation of the critical velocity U cr by using conventional
numerical methods.
4.4 Transfer Matrices Method (TMM)

Problems in engineering are often formulated in terms of a boundary-value prob-
lem where the unknown function f(x) of the Ordinary Differential Equation (ODE)
n
dm
∑a
m=0
m
dx m
f ( x) = q ( x), an ≠ 0, x ∈ [ x1 , xn ] (4.96)
is continuous in the space x1 ≤ x ≤ xn . In typical structural problems, the values

of the unknown function f(x) (or of some of its derivatives) are prescribed at two
distinct boundary points: x1 and xn.
The Ordinary Differential Equation given by eq. (4.96) is considered to be ac-
companied by the following boundary conditions at the boundary points x1 and xn
respectively:
[ f ( x1 ), f '( x1 ),..., f ( k ) ( x1 )] = [e1 , e2 ,..., ek +1 ] (4.97)
[ f ( k +1) ( xn ),..., f ( n −1) ( xn )] = [ek + 2 ,..., en ] (4.98)
where k is an arbitrary order of differentiation ( 1 ≤ k ≤ n ) and e1, e2,…en are

known values. The proposed method presupposes the existence of the n unknown
boundary values u1,u2,…,un:
[ f ( k +1) ( x1 ), f ( k + 2 ) ( x1 ),..., f ( n −1) ( x1 )] = [uk + 2 , uk + 3 ,..., un ] (4.99)
[ f ( xn ), f '( xn ),..., f ( k ) ( xn )] = [u1 , u2 ,..., uk +1 ] (4.100)
which should be determined.

Assuming the transformations
y = f ( x) (4.101)
Transfer Matrices Method (TMM) 125
y1 = y ' = f '( x) (4.102)
and
ym = y 'm −1 = f ( m ) ( x) m = 2, 3,..., n (4.103)
the n known and the n unknown boundary conditions given by equations (4.97)–
(4.100) can be represented by the following matrix equations:
[ I ]nxn ⋅ [ X 1]nx1 = [ EU ]nx1 (4.104)
[ I ]nxn ⋅ [ XN ]nx1 = [UE ]nx1 (4.105)
where [I]nxn is the identity matrix
1 0 L 0
0 1 L 0 
[ I ]nxn = (4.106)
M M O 0
 
0 0 L 1  nxn
and [X1]nx1, [XN]nx1 are the state vectors of the boundary points x1, xn respectively
[ X 1]nx1 = [ y ( x1 ), y1 ( x1 ),..., yk ( x1 ), yk +1 ( x1 ),..., yn −1 ( x1 )]T (4.107)
[ XN ]nx1 = [ y ( xn ), y1 ( xn ),..., yk ( xn ), yk +1 ( xn ),..., yn −1 ( xn )]T (4.108)
while [EU]nx1, [UE]nx1 are the vectors containing the following values of the
(known and unknown) boundary conditions:
[ EU ]nx1 = [e1 , e2 ,..., ek , uk +1 ,..., un −1 ]T (4.109)
[UE ]nx1 = [u1 , u2 ,..., uk , ek +1 ,..., en −1 ]T (4.110)
With the aid of transformations given by equations (4.101)–(4.103), the Ordinary

Differential Equation (ODE) (4.96) can be written:
1 a a a a
yn = y 'n −1 = f ( n ) ( x) = q ( x) − n −1 yn −1 − n − 2 yn − 2 −  − 1 y1 − 0 y (4.111)
an an an an an
Then the system of equations (4.101)–(4.103) and (4.111) can be written in the
following matrix form:
 0 1 0 0 L 0   0 
 y     y   
 y   0 0 1 0 L 0 
y   0 
 1   0 0 0 1 L 0 
1 
 0 
d  y2     y2   
 = 0 0 0 0 L 0  + 0 
dx  y3   y
M 
3 
 M M M M O  M 
 M    M   
   a0 a1 a2 a a   q( x) 
 yn −1   − − − − 3 L − n −1   yn −1   
 an an an an an   an 
(4.112)
Equation (4.112) has the form
[ X ]' = [ A][ X ] + [ F ] (4.113)
where
[ X ] = [ y ( x), y3 ( x),  yn −1 ( x) ]
T
y1 ( x), y2 ( x), (4.114)
T
 q( x) 
[ F ] = 0, 0, 0, 0,   (4.115)
 an 
 1 for j = i + 1
 a j −1

[ A] = { Aij } =  − for i = n and j = 1, 2,..., n (4.116)
 an
0 for the other values i, j
For the end point x = xn the matrix equation (4.112) can be written as:
[ XN ]' = [ AN ][ XN ] + [ FN ] (4.117)
where [XN] is the state vector of the end point xn, given by eq. (4.108), and
T
 q ( xn ) 
[ FN ] = 0, 0, 0, 0,   (4.118)
 an 
Transfer Matrices Method (TMM) 127
 1 for j = i + 1
 a j −1

[ AN ] = { AN ij } =  − for i = n and j = 1, 2,..., n (4.119)
 an
Then, the solution of the matrix differential equation (4.117) can be obtained [e.g. 7]:
xn
[ XN ] = ∫ e[ AN ]( xn − s ) [ F ( s )]ds + e[ AN ]( xn − x1 ) [ X 1] (4.120)
x1
where [X1] is the state vector of the initial point x1, given by eq. (4.107), and
T
 q( s) 
[ F ( s )] = 0, 0, 0, 0,   (4.121)
 an 
The eq. (4.120) is a matrix equation correlating the state vector [XN] of the end-
point xn with the state vector [X1] of the initial point x1. For the case that q(x) = 0
(e.g., for the case of free motion), the matrix e[ AN ]( xn − x1 ) correlating the state vec-
tors between two points i and n is called Transfer Matrix. Eq. (4.120) can be
considered as an algebraic linear system of n equations with 2n unknown val-
ues, which are incorporated in the boundary state vectors [X1] and [XN], i.e.,
[ y( x1 ), y1 ( x1 ), y2 ( x1 ), y3 ( x1 ),  yn −1 ( x1 ), y( xn ), y1 ( xn ),  yn −1 ( xn )].
To solve this (n × 2n) linear algebraic system, n additional linear equations are
required in order to yield a complete (2n × 2n) system. These n additional equa-
tions can be obtained by the n known boundary values e1, e2, …,ep, …,en which are
incorporated in matrix equations (4.104)–(4.105). Considering equations (4.104),
(4.105), (4.120) and taking into account the transformations given by equations
(4.101)–(4.103), the following linear system can be obtained:
[ A11 ] [ A12 ]   [ X 1]  [ B1 ] 
[ A ] [ A ] ⋅ [ XN ] = [ B ] (4.122)
 21 22     2 
where the sub-matrices [A11], [A12], [A21], [A22], [B1], [B2] are given by the
following matrix equations:
[ A11 ]nxn = e[ AN ]( xn − x1 ) (4.123)
[ A12 ]nxn = −[ I ]nxn (4.124)

[ I ] [ 0] 
[ A21 ] =  kxk
[0]
(4.125)
 [0]
[0] [0] 
[ A22 ] =   (4.126)
[ 0 ] [ I ](n−k ) x(n−k ) 
xn
[ B1 ]nx1 = − ∫ e[ AN ]( xn − s ) F ( s )ds (4.127)

x1
[ B2 ]nx1 = [e1 , e2 ,..., en ]T (4.128)
From the solution of the linear algebraic system (4.122), the unknown bound-
ary values [u1, …, un] of the boundary points x1 and xn (see equations (4.104),
(4.105)) can be obtained. Therefore, e.g., at the point x1, the value of the function
f(x1) = y(x1) and the values of its derivatives y1(x1), y2(x1), …,yk(x1), …,yn-1(x1), i.e.,
the vector [EU]nx1, are now known. Then, the solution
[ X Ξ] = [ y (ξ ), y3 (ξ ),  yn −1 (ξ ) ]
T
y1 (ξ ), y2 (ξ ), (4.129)
of each distinct point ξ ( x1 ≤ ξ ≤ xn ) can be obtained by the following linear

algebraic system:
[ AA11 ] [ AA12 ]   [ X 1]  [ BB1 ] 

[ AA ] [ AA ] ⋅ [ X Ξ] = [ BB ] (4.130)
 21 22     2 
where the sub-matrices [AA11], [AA12], [AA21], [AA22], [BB1], [BB2] are given by the
following matrix equations:
[ AA11 ]nxn = e[ AΞ ](ξ − x1 ) (4.131)
[ AA12 ]nxn = −[ I ]nxn (4.132)
[ AA21 ] = [ I ]nxn (4.133)
[ AA22 ] = [ 0]nxn (4.134)
[ BB1 ]nx1 = − ∫ e[ AΞ ](ξ − s ) F ( s )ds (4.135)

x1
Estimation of Critical Velocity for Composite Pipes Conveying Fluid 129
[ BB2 ]nx1 = [ EU ]nx1 (4.136)
 1 for j = i + 1
 a j −1

[ AΞ] = { AΞ ij } =  − for i = n and j = 1, 2,..., n (4.137)
 an
4.5 Estimation of Critical Velocity for

Composite Pipes Conveying Fluid
(General case: solution of equation (4.52))
Taking into account the theory described in the previous section, eq. (4.122)
has to be formulated for the case of a composite pipe conveying fluid. Assuming
that the length of the pipe is L, the matrix [AN] should be derived by drawing on
the coefficients a0 , a1 , a2 , a3 of the differential equation (4.70). For n = 4, a com-
parison of eqs. (4.70) and (4.96) yields:
a0 = −( M + m)ω 2 (4.138)
a1 = 2 MU ω i (4.139)
a2 = MU 2 (4.140)
a3 = 0 (4.141)
� yy
a4 = EI (4.142)
Therefore, according to eq. (4.119), the matrix [AN] can be written as:
 0 1 0 0
 0 0 1 0

[ ]
AN =  0 0 0 1 (4.143)
 2 
 ( M + m) ω −2 MU ω i − MU 2
0
 EI � yy � EI yy � yy
EI 
 
Taking into account the above equation, as well as equations (4.123)–(4.128),

eq. (4.122) yields:
 y ( 0)   0 
 y ( 0)   0 
 1   
 y 2 ( 0)   0 
   
 e[ AN ].L 
  [ − I 4 x 4 ]  y3 (0)   0 
 =  (4.144)
 [ A21 ]
 [ A22 ]   y ( L)   e1 
 y1 ( L)  e2 
   
 y2 ( L)   e3 
 y ( L )  e 
 3   4
In above matrix equation, the matrices [ A21 ] , [ A22 ] , {e1 , e2 , e3 , e4 } depend on the
T
boundary conditions at the end points x = 0 and x = L of the pipe. For the cases of
cantilever, fixed-fixed, pinned-pinned, and fixed-pinned pipes, the above matrices
have the following forms.
4.5.1 Cantilever pipe
1 0 0 0
0 1 0 0 
[ A21 ] =  (4.145)
0 0 0 0
 
0 0 0 0
0 0 0 0
0 0 0 0 
[ A22 ] =  (4.146)
0 0 1 0
 
0 0 0 1
4.5.2 Fixed-fixed pipe
1 0 0 0
0 1 0 0 
[ A21 ] =  (4.147)
0 0 0 0
 
0 0 0 0
Estimation of Critical Velocity for Composite Pipes Conveying Fluid 131
0 0 0 0
0 0 0 0 
[ A22 ] =  (4.148)
1 0 0 0
 
0 1 0 0
4.5.3 Pinned-pinned pipe
1 0 0 0
0 0 1 0
[ A21 ] =  (4.149)
0 0 0 0
 
0 0 0 0
0 0 0 0
0 0 0 0 
[ A22 ] =  (4.150)
1 0 0 0
 
0 0 1 0
4.5.4 Fixed-pinned pipe
1 0 0 0
0 1 0 0 
[ A21 ] =  (4.151)
0 0 0 0
 
0 0 0 0
0 0 0 0
0 0 0 0 
[ A22 ] =  (4.152)
1 0 0 0
 
0 0 1 0
For all above cases of support, the vector {e1 , e2 , e3 , e4 } has the form:
T
{e1 , e2 , e3 , e4 } = {0, 0, 0, 0}
T T
(4.153)
The condition for non-trivial solutions of matrix equation (4.144) is
 e[ AN ]⋅ L  [ − I 4 x 4 ]
det    =0 (4.154)
 [ A21 ] [ A22 ] 

For given values of EI� yy , M , m, the equation above yields a correlation of ω

with U. For any value of U, the value of ω is a complex number. As was already
mentioned, instability occurs for the value of U cr yielding Im [ω ] < 0 .
For several values of U, equation (4.154) can be solved numerically using a
standard commercial software, e.g., Mathematica. The obtained results Re [ω ]
and Im [ω ] can be presented in an Argand diagram ( Re [ω ] vs Im [ω ]) showing
the evolution of the system with increasing U, from stability (U < U cr ) to insta-
bility (U ≥ U cr ) .
4.6 Effect of Temperature (Thermal Load)

In long pipes the temperature can vary over certain portions of their length. The
effect of thermal loads yields longitudinal strain εx. On the other hand, for pipes
where both ends are supported (e.g., fixed-fixed, pinned-pinned, fixed-pinned), no
longitudinal movement is allowed. Therefore, the condition
L
∫ ε dx = 0
0
x
yields existence of axial force T (tension or compression). Adding equations

(4.27) and (4.33) yields:
∂T ∂P
−A = maxf (4.155)
∂x ∂x
Assuming a constant flow velocity U, and assuming that the pressure is indepen-
dent from x, the above equation can be written as:
∂T
=0 (4.156)
∂x
From eq. (4.156) it can be concluded that the tension T has a constant value caus-
ing constant longitudinal stress. For a multi-layered composite pipe consisting of
NP plies with symmetric fiber orientation +θ or –θ, the longitudinal stress of each
ply has the value
Effect of Temperature (Thermal Load) 133
(T / NP )
σx = (4.157)
π Dh
where h is the thickness of each ply and D is the mean diameter of the pipe. Since
the pipe is subjected to internal pressure P, the stress s y for the multi-layered
long pipe is
D
s y = ( P / NP ) (4.158)
2h
We recall that for a lamina with fiber orientation θ, it can be written:
ε x − ax ( ∆Τ ) = S11σ x + S12σ y (4.159)
or
ε x = S11σ x + S12σ y + ax ( ∆Τ ) (4.160)
where ΔΤ is the temperature difference and ax is the coefficient of thermal ex-

pansion in the x-direction. According to Chapter 1, the coefficient ax can be cor-
related with the coefficients of thermal expansion in the principal directions of a
lamina a1 , a2 by the equation:
ax ( ∆Τ ) = a1 cos 2 θ ( ∆Τ ) + a2 sin 2 θ ( ∆Τ ) (4.161)
With the aid of the above equation, eq. (4.160) can now be written:
ε x = S11σ x + S12σ y + ( a1 cos 2 θ + a2 sin 2 θ ) ∆Τ (4.162)
where:
cos 4 θ  2v12 1  2 2 sin 4 θ

S11 = +− +  cos θ sin θ + (4.163)
E1  E1 G12  E2
 1 1 1  2 v12
 cos θ sin θ − ( cos θ + sin θ )
2 4 4
S12 =  + − (4.164)
E
 1 E 2 G12  E 1
Taking into account equations (4.157), (4.158) and (4.162)–(4.164), the condition
of no longitudinal movement at the pipe ends, which is
L
∫ ε x dx = 0 (4.165)
0
yields:
πD  D ⋅ P ⋅ S12 
T=− 
S11  2
( )
+ NP ⋅ h ⋅ a1 cos 2 θ + a2 sin 2 θ ∆Τ  (4.166)

If we do not neglect the tensioning from eq. (4.28), eq. (4.46) can be written:
∂Q  ∂w  ∂  ∂w  ∂2 w
+  F + πDq  +  T  = m (4.167)
∂x  ∂x  ∂x  ∂x  ∂t 2
Combining the above equation with eq. (4.49) yields:
∂Q  ∂2 w ∂2 w ∂ 2 w  ∂T ∂w ∂2 w ∂2 w
−M  2 + 2U +U 2 2  + + T 2 = m 2 (4.168)
∂x  ∂t ∂x∂t ∂x  ∂x ∂x ∂x ∂t
Since the axial force T has the same value for every point x, the derivative ∂T / ∂x
has the value
∂T
=0 (4.169)
∂x
With the aid of equations (4.47) and (4.169), eq. (4.168) yields:
4 2 2 2
� yy ∂ w + ( MU 2 − T ) ∂ w + 2 MU ∂ w + ( M + m ) ∂ w = 0
EI (4.170)
∂x 4 ∂x 2 ∂x∂t ∂t 2
Taking into account eq. (4.166), the above equation can now be written as:
� yy ∂ w +  MU 2 + π D  D ⋅ P ⋅ S12 + NP ⋅ h ( a cos 2 θ + a sin 2 θ ) ∆Τ   ∂ w +

4 2
EI 4   1 2  2
∂x  S11  2   ∂x

∂2 w ∂2 w
2 MU + ( M + m) 2 = 0
∂x∂t ∂t
(4.171)
Using the notation
π D  D ⋅ P ⋅ S12 
H (U , ∆Τ ) = MU 2 +  + NP ⋅ h ( a1 cos 2 θ + a2 sin 2 θ ) ∆Τ 
S11  2 
(4.172)
Effect of Additional Mass 135
the matrix [ AN ] of eq. (4.143) now has the form:
 0 1 0 0
 0 0 1 0 

[ AN ] =  0 0 0 1 (4.173)
 
 ( M + m)ω − H (U , ∆Τ )
2
−2 MU ω i 
 0
� yy
EI � EI yy � yy
EI
 
The matrix above can be used in the Transfer Matrix Method (TMM) to determine
the critical flow velocity U cr for a multilayered composite pipe under the influ-
ence of temperature change.
4.7 Effect of Additional Mass

Because the specific weight of composite materials is relatively low (almost
25% of the specific weight of steel) and their stiffness is very high, composite
pipes can be characterized as lightweight structures. During the installation of a
composite pipe of diameter D = 1.0 m and thickness 4.0 mm into the sea, a buoy-
ancy reaction of 0.78 tn/m acts on the pipe in a direction opposite to its weight.
Therefore, in order to install hollow composite pipe underwater, additional mass
must be added to offset the buoyancy forces. Such additional weights may be equal-
ly spaced metallic collars located, for example, on the joints of the pipe (Fig. 4.8a).
However, the mass and the additional stiffness of these collars alter the gravita-
tional forces of the pipe and therefore influence its dynamic behavior. In order to
estimate the critical velocity U cr for a composite pipe containing additional mass,
the local transfer matrix of each pipe segment equipped with a collar needs to be
derived. Taking into account equation (4.143), the matrix [ AN ] now has the form:
(a) (b)
figure 4.8 (a) Installation of a composite pipe into the sea by using metallic collars, right
(b) Collar-pipe cross section.
 0 1 0 0
 0 0 1 0 

[ AN c ] =  0 0 0 1 (4.174)
 
 ( M + m + mc ) ω
2 2
−2 MU ω i − MU 
 0
 � c
EI � EI c � c
EI 
where mc is the mass per unit length of the collar and EI � c is the stiffness of the
collar-pipe cross-section.
Therefore, the matrix equation correlating the state vectors { y, y1 , y2 y3 } of the
T
end points A and B of the pipe segment equipped with a collar is:
y y
y  y 
 1  1
 
y
= [TM c ] ⋅   (4.175)
 2  y2 
 y3   y3 
B A
where [TM c ] is the transfer matrix of the collar given by:
[TM c ] = e[ AN ]⋅L
c c
(4.176)
In the above equation, Lc is the length of the collar. In order to calculate the
� c we shall consider the collar-pipe system as itself a composite pipe
stiffness EI
consisting of a multi-layered anisotropic pipe (with internal radius r1 and external
radius r2 = r1 + NP ⋅ h ), as well as an isotropic steel pipe (with internal radius r2
and external radius r3 ) perfectly bonded (Fig. 4.8b). In such a case, the elements
of the ABD matrix should be calculated by the following formulas:
c
Aijc = Aij + Q ij ( r3 − r2 ) (4.177)
Bijc = Bij + Q ij ( r32 − r22 )

c
(4.178)
Dijc = Dij + Q ij ( r33 − r23 )

c
(4.179)
In the above equation, Aij , Bij , Dij correspond to the multi-layered pipe, while
c
the parameters Q ij correspond to the steel collar. Since steel can be considered
as an isotropic material (n = 0, m = 1 because [T ] = I ) where E1 = E2 = E3 = E ,
c
ν 23 = ν 13 = ν 12 = ν , G23 = G13 = G12 = G , the parameters Q ij can be derived by
the following equations:
c E
Q11 = (4.180)
1 −ν 2
c νE
Q12 = (4.181)
1 −ν 2
c
Q16 = 0 (4.182)
c E
Q 22 = (4.183)
1 −ν 2
c
Q 26 = 0 (4.184)
c
Q 66 = G (4.185)
where E, G, ν are, in order, the elasticity modulus, shear modulus and Poisson’s
ratio of steel.
With the aid of equations (4.177)–(4.179), the stiffness of the collar-pipe seg-
ment can be calculated by following the procedure outlined in Figure 4.3 and
using equation (4.16).
However, when the stiffness of a collar has a value that is much higher than
the value of the stiffness of the pipe, the inversion of the ABD matrix may yield
numerical errors. In such cases, since the stiffness of collar is dominant, the stiff-
ness of the pipe in the transfer matrix of the collar-pipe segment can be neglected.
For a better understanding of the above procedure, the critical velocity U cr will
be estimated for a cantilever pipe (Fig. 4.9) containing an additional mass located
at its end.
Figure 4.9 A model of cantilever composite pipe containing a point mass.

This model may represent the fallacious patent [8] for preventing torsional
buckling of drill-strings by the use of a floating drill-bit for oil exploration, rotat-
ing as a turbine under the action of the flow [2]. For this example the following
data will be used:
Material of the pipe: S Glass/Epoxy

Thickness of ply: h = 0.15 mm
Number of plies: NP = 10
Orientation of fibers: ± ( π / 4)
Mechanical properties of ply:
E1 = 39 × 109 N / m 2
E2 = 8.6 × 109 N / m 2
ν 12 = 0.28
G12 = 3.8 × 109 N / m 2
Density of pipe: ρ p = 2.1× 103 Kg / m3
Density of fluid: ρ f = 103 Kg / m3
Inner diameter of pipe: 2r1 = 0.1 m
Exterior diameter of pipe: 2r2 = 2 ( r1 + NP ⋅ h ) = 0.103 m
Material of additional mass (collar): Steel
Mechanical properties of steel:
Es = 196 × 109 N / m 2
Gs = 73 × 109 N / m 2
ν s = 0.33
Density of steel: ρ s = 7800 Kg / m3
Exterior diameter of collar:
Length of the pipe: L12 = 1000 m
Length of the collar (in order to overcome the buoyancy force): L23 = 3.0 m
According to the procedure found in Fig. 4.3 and eq. (4.16), the calculation of the
value of the stiffness of the pipe (segment 1-2) yields:
� 12 =� EI yy = 39157 N ⋅ m 2
EI
The stiffness of the collar is given by the well-known expression:
s
4
3 ( 2 )
� c = E π r 4 − r 4 � 2.31 × 109 N ⋅ m 2
EI
� yy /� EI c ≈ 0.00001 , along the segment 2-3 the stiffness of the collar dom-
Since EI
inates the bending behavior. Therefore, EI � 23 ≈� EI c .
4.7.1 Transfer matrix of the segment 1-2 (pipe)
Mass per unit length of the pipe: m = 2 πr1 NP h ρ p = 0.9896 Kg / m
Mass per unit length of the fluid: M = π r12 ⋅ ρ f = 7.854 Kg / m
According to eq. (4.143), the matrix [ AN ] for the pipe yields
 0 1 0 0
 0 0 1 0 
[ AN ]12 =
 0 0 0 1
 −4 
 2.26 × 10 ⋅ ω
2
−4.01× 10 ⋅ U ω i −2.00 × 10−4 ⋅ U 2
−4
0
[ AN ]12 ⋅ L12
Therefore the transfer matrix for the segment 1-2 is [TM ]12 = e
4.7.2 Transfer matrix of the segment 2-3 (collar-pipe)
2 2
( )
Mass per unit length of the collar: mc = π r3 − r2 ⋅ ρs = 11747 Kg / m .
According to eq. (4.174), the matrix [ AN ] for the segment 2-3 is
 0 1 0 0
 0 0 1 0 
[ AN ]23 =
 0 0 0 1
 −6 
5.09 × 10 ⋅ ω
2
−6.8 × 10 ⋅ U ω i −3.4 × 10−9 ⋅ U 2
−9
0
[ AN ]23 ⋅ L23
and the corresponding transfer matrix is [TM ]23 = e
4.7.3 Global transfer matrix

From the definition of a transfer matrix (equation (4.120)), the following can
be written:
{ XN3 } = [TM 23 ]{ XN 2 }
{ XN 2 } = [TM 12 ]{ XN1}
where { XN j } is the state vector { y, y1 , y2 , y3 } of a point j.

T
The above equations yield:
{ XN3 } = [TM 23 ][TM 12 ]{ XN1}

Therefore, the global transfer matrix [GTM ] of the system correlating the state
vectors of the end points 1 and 3 is always the product of the transfer matrices of
the segments, i.e., [GTM ] = [TM 23 ] ⋅ [TM 12 ] .
Since [ A] ⋅ [ B ] ≠ [ B ].[ A] , it is important to mention that the multiplication of
the transfer matrices should be started from the transfer matrix of the last segment.
We recall that the algebraic equation yielding ω versus the fluid flow U is
eq. (4.154). In this equation, the transfer matrix should now be replaced by the
[GTM ] derived above, i.e.,
[GTM ] [ − I 4 x 4 ] = 0
det 
[ A22 ] 

 [ A21 ]
where [ A21 ] and [ A22 ] are the transfer matrices incorporating the boundary condi-
tions for a cantilever pipe, given by equations (4.145), (4.146).
Because only a numerical solution for the above equation is possible, the soft-
ware Mathematica has been used to derive the value of ω for incremental values
of U. The critical value of U is the lowest value of U yielding a shift of Im {ω}
from a positive to a negative value. In the attached Mathematica code we started
by calculating the ω from a small initial value of U = 0.1 m/s. The value of ω is
calculated incrementally for the values U = 0.1, 0.11, 0.12, 0.13, 0.14, 0.15, 0.16,
0.17, 0.18, 0.19, 1.00. We found that Im {ω} moves from positive to negative
values where U cr = 0.9 m/s. This value corresponds to a critical fluid flow
Q = 3600 πr12U cr = 25 m3 / s .
Effects of an Elastic Foundation 141
4.8 Effects of an Elastic Foundation

Very often, long composite pipes are embedded in soil or lie on the bottom of
the sea. The additional support of the soil yields a stiffer system. In order to simu-
late the dynamic behavior of a composite pipe resting on an elastic (Winkler-type)
foundation, the equation of motion (eq. 4.52) should be adapted [2] by taking into
account its elastic constant k, i.e.:
4 2 2 2
� yy ∂ w + MU 2 ∂ w + 2 MU ∂ w + ( M + m) ∂ w + k w = 0
EI (4.186)
4 2
∂x ∂x ∂x∂t ∂t 2
Trying the solution given in eq. (4.69), the above equation yields:
� yy d f ( x ) + MU 2 d f ( x ) + 2 MU ω i df ( x ) +  k − ( M + m ) ω 2  f ( x ) = 0
4 2
EI 4  
dx dx 2 dx
(4.187)
Following the procedure for deriving the corresponding transfer matrix, the ma-
trix [ AN ] given in eq. (4.143) now takes the following form:
 0 1 0 0
 0 0 1 0 

 AN  = 
K
0 0 0 1 (4.188)
 
 ( M + m)ω − k
2 2
−2 MU ω i − MU 
 0
� yy
EI � EI yy � yy
EI
 
Therefore, the transfer matrix correlating the state vectors between two points i
and j of the pipe can now be written as:
TM ij  = e   ij 
 AN K  ⋅ L
(4.189)
 
where Lij is the length between the nodes i and j. The equation providing the value
of ω versus U can now be written (see eq. (4.154)) as:
 e  AN K  ⋅ Lij 
[ − I 4 x 4 ]
det    =0 (4.190)
 
 [ A21 ] [ A22 ] 
where the matrices [ A21 ] and [ A22 ] corresponding to the boundary conditions can
be obtained by eqs. (4.145)-(4.152). Here it should be recalled that critical flow
speed U cr is the lowest value of U yielding the shift of Im {ω} from a positive sign
to a negative one.
4.9 Effect of Additional Supports

Long composite pipes may be supported on more than two supports. In most
cases, the intermediate supports are simple, and the spans between them are
equally spaced as depicted in Figure 4.10.
The state vectors { y, y1 , y2 , y3 } of two successive nodes i, j are correlated by
T
the well-known transfer matrix given by:
TM ij  = e[ AN ]⋅ L  (4.191)

 
where L is the length of each span and [ AN ] is the matrix given by eq. (4.143).
Therefore:
y y
y  y 
 1
 
y
= [ ]ij  y1 
TM j >i (4.192)
 2  2
 y3   y3 
j i
The equation above can be written in the following form:
 y  
  
  y1  
  y2  
  
  y3 i 
[TM ] [ − I 4 x 4 ]  y  = [0] (4.193)

  
ij
  y1  
  
  y2  
 y  
 3  j 
where [ I 4 x 4 ] is the unit matrix.

Apart from the spans, a similar formulation can be derived for the cross-sections
just before (L) and just after (R), an intermediate support (Fig. 4.11).
Effect of Additional Supports 143
Figure 4.10 Periodically and simply supported composite pipe.
Figure 4.11 Equilibrium of an intermediate support.
• •
Using the symbols ( L ) and ( R ) for the state variables (i.e., deflection w,
slope w’, bending moment M, shearing force Q) located just left and just right of
the node, the following compatibility equations and equilibrium conditions can
be written:
wL = 0 (4.194)
wR = 0 (4.195)
wL' = wR' (4.196)
ML = MR (4.197)
Taking into account that M = EIw '' , the above equations can be written in the
following matrix form:
  wL  
  
  w 'L  
  w ''L  
1 0 0 0 0 0 0 0
  
0 0 0 0 1 0 0 0    L i 
 w '''
'
   = [ 0] (4.198)
0 1 0 0 0 −1 0 0   wR  
    w 'R  
0 0 1 0 0 0 −1 0
  
  w ''R  
  w '''  
 R  j 
With the aid of equations (4.69) and (4.101)–(4.103), the above equation yields:
 y  
  
  y1  
1 0 0 0 0 0 0 0   y2  
0   
0 0 0 1 0 0 0    y3  L 
 = [ 0]


0 1 0 0 0 −1 0 0  (4.199)
   y  
0 0 1 0 0 0 −1 0   y1  
  
  y2  
 y  
 3 R 
Taking into account the boundary conditions of the supports at the end points 1
and 5, the following matrix equations can be written:
y 
 
1 0 0 0   y1 
0 1 0 0  ⋅  y  = 0 (4.200)
   2
 y3 
1
y 
 
1 0 0 0   y1 
0 1 0 0  ⋅  y  = 0 (4.201)
   2
 y3 
5
Effect of Additional Supports 145
Applying equation (4.193) for all spans (i-j) namely (1-2), (2-3), (3-4), (4-5), and
equation (4.199) for all intermediate supports namely 2, 3, 4 and taking into ac-
count the boundary conditions (4.200), (4.201) of the end supports, the following
matrix equation can be derived:
[G32×32 ] ⋅ {Y32×1} = [0] (4.202)
where
{Y32×1} = {{Y }1 , {Y }2 , {Y }2 , {Y }3 , {Y3 } , {Y4 } , {Y4 } , {Y }5 }

L R L R L R T
(4.203)
{Y }1 = { y, y1 , y2 , y3 }1
T
(4.204)
= { y L , y1L , y2L , y3L } for j = 2, 3, 4 (4.205)

T
{Y } j
L

j
{Y } j = { y R , y1R , y2R , y R T
}
R
3 j for j = 2, 3, 4 (4.206)
{Y }5 = { y, y1 , y3 , y4 }5 (4.207)
and
ϭ
ϭ

[70 ] [ − , × ]

ϭ
ϭ
ϭ Ͳϭ
ϭ Ͳϭ

[70 ] [ − , × ]

ϭ
[ × ]
* = ϭ
ϭ Ͳϭ
ϭ Ͳϭ

[70 ] [ − , × ]

ϭ
ϭ
ϭ Ͳϭ
ϭ Ͳϭ

[70 ] [ − , × ]

ϭ
ϭ
(4.208)
The condition for a non-trivial solution of eq. (4.202) is
det [G32×32 ] = 0 (4.209)

The above equation can be solved numerically yielding the values of ω versus
the values of U. It should be again noted that the value of U yielding the shift of
Im {ω} from a positive to a negative sign is critical.
4.9.1 Example
The fixed-fixed pipe containing three intermediate supports shown in Fig. 4.10
is modeled using the above procedure. The spans of this pipe are equidistant from
one another with length L12 = L23 = L34 = L45 = 10m . The interior diameter of the
pipe is 2r1 = 0.10m . The pipe wall is composed of Np = 50 layers of S-Glass/
Epoxy fiber reinforced composite. The thickness of each layer is h = 0.150 mm,
while the fiber orientation is θ = ±(π / 4) . The density of the composite material
9 3 3
is ρ p = 2.1× 10 kg / m , while the density of the fluid is ρ f = 1000kg / m . The
material properties in the principal directions of each ply are E1 = 39 × 10 N / m 2 ,
9
E2 = 8.6 × 109 N / m 2 , ν 12 = 0.28 , G12 = 3.8 × 109 N / m 2 . Following the procedure

of Fig. 4.3 and using eq. (4.16) the value EI � yy = 196224 N ⋅ m 2 has been obtained
for the stiffness of the pipe. Using the above data, the mass per unit length for
the pipe and the fluid have the values m = ρ p (2π r1 )( NP ⋅ h) = 4.948kg / m and
M = ρ f (π r12 ) = 7.854kg / m . Using the derived values EI � yy , m ,
M , the follow-
ing formula for the matrix [ AN ] has been obtained:
 0 1 0 0
 0 0 1 0 
[ AN ] = 
0 0 0 1
 −5 2 
6.524 × 10 ω −8.0051× 10 ωUi −4.0025 × 10−5 U 2
−5
0
Taking into account eq. (4.191), the transfer matrix [TM ]ij can now be derived
for every span. Using eq. (4.208) the matrix [G32×32 ] can be obtained yielding
eq. (4.209). Starting from a small initial value U = 1.0 m / s , which is changed
incrementally with step 1.0 m / s , the eigen-frequency ω has been calculat-
ed numerically by solving eq. (4.209) using Mathematica (see attached code).
The results indicated that for the values of U ≤ 70 m / s , Im {ω} ≈ 0 . For
U = 71 m / s the value of Im {ω} was positive (Im {ω} = +2.83007) , while for
the next increment of velocity (i.e. U = 72 m / s ) the Im {ω} became nega-
tive (Im {ω} = −4.02455) . Therefore the critical flow velocity of the sample
multi-supported pipe is U cr = 72 m / s . If the intermediate supports of the pipe
are removed, eq. (4.154) yields U cr = 28 m / s (see the attached computer code
in Mathematica). Therefore, the three intermediate supports have stabilized the
composite pipe by increasing the critical velocity almost 2.6 times.
Estimation of Critical Flow Velocity in Relation to Divergence 147
4.10 Estimation of Critical Flow Velocity in

Relation to Divergence
As mentioned, divergence may occur if both ends of a pipe are supported (for
cantilevered pipes this instability is unlikely to happen). For sufficiently high-flow
velocities the bending due to divergence instability can be so extreme that the pipe
can fail. Therefore, the estimation of the critical flow velocity where divergence
instability can be encountered is an important consideration in the design of slen-
der pipelines that carry fluids. We recall that the eigenvalue problem associated
with divergence instability is described by the following differential equation:
4 2
� yy d w + MU 2 d w = 0
EI (4.210)
dx 4 dx 2
The above equation can be modified in order to take into account thermal loads
(ΔΤ) or effect of elastic foundation, i.e.:
4.10.1 Thermal load effect
4 2
� yy d w + H (U , ∆Τ) d w = 0
EI (4.211)
4
dx dx 2
where
π ⋅ D ⋅ NP ⋅ h
H (U , ∆Τ) = MU 2 + (a1 cos 2 θ + a2 sin 2 θ )∆Τ (4.212)
S11
4.10.2 Elastic foundation effect
4 2
� yy d w + MU 2 d w + k w = 0
EI (4.213)
4
dx dx 2
where k is the elastic constant of the foundation.
In cases where the effects of thermal loads and the elastic foundation occur
simultaneously, the combination of eqs. (4.211) and (4.213) yields:
4.10.3 Thermal load and elastic foundation effects
4 2
� yy d w + H (U , ∆Τ) d w + k w = 0
EI (4.214)
4
dx dx 2
Following the method of transfer matrices, the following form for the matrix [AN]
can be obtained for the case of divergence instability:
 0 1 0 0
 0 0 1 0 

[ AN D ] =  0 0 0 1 (4.215)
 
 −k 0
− H (U , ∆Τ)
0
 EI
� �
EI 
 yy yy 
Thus the transfer matrix correlating the state vectors {w,w’,w’’,w’’’}T of two suc-
cessive points i and j has the form:
TM ijD  = e[ 

D
AN ]⋅ Lij
  (4.216)
where Lij is the length of the span i-j.

As it is already known, the eigenvalues ω versus the fluid speed U for a pipe
supported to its both ends can be obtained by the following equation:
 TM ijD  [ − I 4 x 4 ]
det     = [ 0] (4.217)
 [ A21 ] [ A22 ] 
where [ I 4 x 4 ] is the 4 × 4 unit matrix and [ A21 ] , [ A22 ] are correlated with the
boundary conditions. For fixed-fixed, pinned-pinned and fixed-pinned pipe, the
matrices [ A21 ] , [ A22 ] are given by eqs. (4.146)–(4.152).
If the thermal load and elastic foundation effects are neglected, the solution of
eq. (4.217) for the cases of fixed-fixed, pinned-pinned and fixed-pinned pipes is
simple, i.e:
4.10.4 Fixed-fixed pipe
MU 2 2 κπ
= (4.218)
� yy
EI L
where k is an integer.
The lowest value of U corresponds to the lowest non-zero value of k, i.e.,
κ = 1. Therefore, for fixed-fixed pipe, the critical value of U is
Hydraulic Hammer 149
4π � yy
EI
U cr = (4.219)
L M
4.10.5 Pinned-pinned pipe
MU 2 κπ
= (4.220)
� yy
EI L
yielding
π � yy
EI
U cr = (4.221)
L M
4.10.6 Fixed-pinned pipe
MU 2 4.49
� (4.222)
� yy
EI L
yielding
� yy
4.49 EI
U cr = (4.223)
L M
4.11 Hydraulic Hammer

A sudden change of discharge in a composite pipe conveying liquid may re-
sult in stresses of sufficient magnitude to exceed the design stresses. The reason
for this phenomenon is the created impulse force, which is a result of Newton’s
impulse-momentum equation. Actually the impulse force is a pressure shock that
travels with high velocity in the upstream and downstream directions. The cre-
ation of pressure shock in a fluid due to a sudden change of discharge is called
hydraulic hammer and must be taken into account as an important parameter for
pipelines designed with composite materials.
In order to estimate the pressure change due to hydraulic hammer we shall

initially consider the evolution of the liquid volume element ABCD shown in
Fig. 4.12.
Before a sudden stoppage of flow, the diameter of the liquid element was r
and its length was (ΔL +ΔL0). After the sudden stoppage of flow, the length is
decreased by ΔL and the diameter increased by Δr, which causes elastic dilatation
of the pipe’s walls. The volume of the liquid element before and after this change
is given by the following known formulas:
Vbefore = πr 2 ( ∆L + ∆L0 ) (4.224)
Vafter = π(r + ∆r ) 2 ⋅ ∆L0 (4.225)
Assuming that ∆r 2 ≈ 0 the change of the volume ΔV is
∆V = Vafter − Vbefore = πr 2 ⋅ ∆L − 2 πr ⋅ ∆r ⋅ ∆L0 (4.226)
The bulk modulus of elasticity E of the liquid can be derived by the relation
∆V
P=E (4.227)
Vbefore
where P is the pressure that causes the dilatation ΔV. With the aid of eqs. (4.224),
(4.225), the last equation yields
P(∆L + ∆L0 )
E= (4.228)
 ∆r 
∆L − 2   ∆L0
 r 
Figure 4.12 Evolution of a liquid volume element after a sudden stoppage of flow.
As ∆L � ∆L0 , the simplification ∆L + ∆L0 ≈ ∆L0 can be adopted, yielding:
P ⋅ ∆L0
E= (4.229)
 ∆r 
∆L − 2   ∆L0
 r 
The pressure P causes a force F = P(πr2) in an opposite direction to the direction

of the movement of the center of gravity of the liquid element, which causes re-
tardation a. According to Newton’s law, the retardation a can be estimated by the
following formula:
F = m⋅a (4.230)
or
P(π r 2 ) = ρVa (4.231)
where ρ is the density of the liquid and V is the volume of the liquid element:
V = πr 2 ( ∆L + ∆L0 ) (4.232)
With the aid of eq. (4.232), eq. (4.231) yields:
P
α= (4.233)
ρ (∆L + ∆L0 )
During the retardation, the center of gravity of the liquid element has traveled a
distance (ΔL/2). This distance is given by the well-known rule:
∆L U 2
= (4.234)
2 2a
where U is the liquid velocity. With the aid of eq. (4.232), the last equation yields:
U 2 ρ ( ∆L + ∆L0 )
∆L = (4.235)
P
Assuming that ∆L = U ∆t and ∆L + ∆L0 ≈ ∆L0 , the above equation can be written
as:
∆L0
P = ρU (4.236)
∆t
Taking into account the elastic properties of an anisotropic pipe’s wall, the pres-
sure P can be correlated with the dilatation ∆r / r . Considering the multi-layered
pipe of Fig. 4.13, the following known relations can be written:
1
Ny = PD (4.237)
2
ε y0 = a22 N y (4.238)
In the above equations, N y is the force per unit length acting in the y-direction
of the pipe’s wall, and ε y0 is the strain in the same direction in the middle of the
wall’s thickness. The parameter a 22 can be derived by the inversion of the ABD
matrix.
As the length of the perimeter of the pipe is 2πr , the strain ε y0 can also be
given by the formula:
∆ ( 2π r )
ε y0 = (4.239)
2π r
or
∆r
ε y0 = (4.2340)
r
Figure 4.13 Part of a filament-wound multi-layered pipe.

Combining eqs. (4.237), (4.238), (4.240) the following can be obtained:
∆r
= Pa22 r (4.241)
r
By writing eq. (4.228) in the form
 ∆r  P∆L0
∆L − 2   ∆L0 = (4.242)
 r  E
and using eq. (4.241), the following formula can be written:
P∆L0
∆L − 2 Pa22 r ∆L0 = (4.243)
E
With the aid of eq. (4.236), the above equation yields:
∆L0 ρU ∆L
∆L − ρUa22 D∆L0 = ∆L0 0 (4.244)
∆t E ∆t
Dividing both parts of the above equation by ∆t the following is obtained:
2 2
∆L  ∆L  ρU  ∆L0 
− ρUa22 D  0  =
E  ∆t 
(4.245)
∆t  ∆ t 
Since ∆L / ∆t = U , the above equation yields:
∆L0 E/ρ
= (4.246)
∆t 1 + a22 DE
In the above equation, E / ρ represents the velocity c∗ of pressure shocks in

fluids of infinite extent, which is called “celerity,” while ∆L0 / ∆t is the celerity c
of the shock wave in the composite pipe. Therefore, eq. (4.246) can now be writ-
ten in the following form:
c 1
= (4.247)
c∗ 1 + a22 DE
It should be noted that the parameters c∗ and E are constants with known values
for most of the liquids (e.g., Table 4.2).
Table 4.2
Values of c∗ and E for widely used liquids [9].
Liquid c∗ ( m / s ) E ( N / m2 )
Water 1509 219 × 107
Benzene 1070 102 × 107
Crude oil 1402 175 × 107
Carbon tetrachloride 933 141× 107
Combining eqs. (4.236), (4.245), (4.246), the pressure change due to hydraulic
hammer can be estimated by using the following formula:
ρUc∗
P∗ = (4.248)
1 + a22 DE
Therefore, the total liquid pressure in the composite pipe will oscillate within the
range of:
Ptot = P0 ± P∗ (4.249)
The required time period T for the shock pressure to travel from, and back to, the
point where sudden stoppage of the flow occurred is
2L
T= (4.250)
c
where L is the length of the pipe to the point where a valve stopped the flow.
In cases where the time of closure of the valve is not zero but Tν , and in
specific case where Tv > T, the maximum overpressure can be determined by the
formula:
2L
P∗ = ρU (4.251)
Tν
In a case where Tv < T, rapid closure is considered as equivalent to instanta-

neous closure, and the shock pressure will reach its maximum value P∗ given by
eq. (4.248).
Wave Propagation Due to Hydraulic Hammer 155
4.11.1 Shock pressure in a branched pipe
When a main pipe 1 is branched into two branches, namely 2 and 3, the pressure
shock (due to hydraulic hammer) in each branch can be estimated by applying
the following formulas:
P2∗ = SP1∗ (4.252)
P3∗ = SP1∗
where
A1
2
C1
S= (4.253)
A1 A2 A3
+ +
C1 C2 C3
In the above equation, A1 , A2 , A3 are the areas of the cross-sections of the pipes 1,
2, 3, and C1 , C2 , C3 are the corresponding pressure speeds given by eq. (4.247).
4.12 Wave Propagation Due to Hydraulic Hammer

In this section, the effect of moving pressure shock on the deflection of a pipe’s
wall will be analyzed. The high velocity of pressure shock causes intensive vibra-
tion of the wall, which leads to dynamic stresses.
For the derivation of the dynamic model of the radial deflection w ( x, t ) cor-
responding to a point of the mid-plane of the pipe’s cross-section located at a
distance x where the equilibrium equations of a wall element (Fig. 4.14) will be
initially considered [e.g. 4]:
∂N x
=0 (4.254)
∂x
∂

∂x
( RN xϕ + M xϕ ) = 0 (4.255)
Νϕ ∂2 M x ∂2 w
− − q z + ρ h =0 (4.256)
R ∂x 2 ∂t 2
Figure 4.14 Coordinate system for a composite pipe subjected to moving pressure shock
due to hydraulic hammer.
∂M x
Qx = (4.257)
∂x
∂M xϕ
Qϕ = (4.258)
∂x
In the above equations, ρ is the density of the composite material, and h is the
total thickness of the pipe’s wall.
The relationships between the displacements u , v, w on the directions
x, ϕ , z with the corresponding strains ε x0 , ε ϕ , γ xϕ and curvatures k x , kϕ , k xϕ in
0 0
the reference (middle) surface are given by [e.g. 4]:
∂u
ε x0 = (4.259)
∂x
w
ε ϕ0 = (4.260)
R
∂υ 0
γ x0ϕ = (4.261)
∂x
∂2 w
kx = − (4.262)
∂x 2
w
kϕ = − (4.263)
R2
2 ∂υ
k xϕ = − (4.264)
R ∂x
We recall that the above strains and curvatures are correlated with the stress resul-
tants N x , Nϕ , N xϕ , M x , M ϕ , M xϕ by the well-known relation:
Β16  ε x 
0
 N x   Α11 Α12 Α16 Β11 Β12
    
 Nϕ   Α12 Β 26  ε ϕ 
0
Α 22 Α 26 Β12 Β22
N  Α  
 xϕ  16 Α 26 Α66 Β16 Β 26 Β66  γ x0ϕ 
 =   (4.265)
M
 x   11 Β Β12 Β16 D11 D12 D16  k x 
 M ϕ   Β12 Β 22 Β 26 D12 D22 D26  kϕ 
    
 M xϕ   Β16 Β 26 Β66 D16 D26 D66  k 
 xϕ 
where the members Aij , Bij , Dij , i = 1, 2, 6, j = 1, 2, 6 of the stiffness matrix are
given in Chapter 1.
Combining eqs. (4.254)–(4.265) and taking into account the symmetry
condition
υ =0 (4.266)
and the loading conditions
� x =0
N ( axial force ) (4.267)
T� = 0 ( torque ) (4.268)
the following partial differential equation can be derived:
∂4 w ∂2 w ∂2 w
f1 4
+ f 2 2 + f 3 w + ρ h 2 = qz (4.269)
∂x ∂x ∂t
In the above equation, the coefficients f1 , f 2 , f 3 are given by [4]:
f1 = H 22 (4.270)
1
f2 = ( H 21 + H12 ) (4.271)
R
H11
f3 = (4.272)
R2
where
 H11 H12 
= [ a1 ] − [ a4 ][ a3 ] [ a2 ]
−1
H 
H 22 
(4.273)
 21
 B22 
 A22 − R B12 
[ a1 ] =   (4.274)
 B − D12 D11 
 12 R 
 B12 
 A12 − R B11 
[ a2 ] =   (4.275)
 A − D26 D
B16 + 16 
 26 R 2 R 
 2 B16 
 A11 A16 −
R 
[ a3 ] =   (4.276)
 A + B16 B 2D
A66 − 66 − 266 
 16 R R R 
 2 B26 
 A12 A26 −
R 
[ a4 ] =   (4.277)
B 2D
 11 B16 − 16 
R 
The loading parameter qz represents the pressure acting on the wall’s surface.
Since the pressure shock due to hydraulic hammer is travelling with a constant
velocity c, its value in an arbitrary location x is given by the following equation:
qz = δ ( x − ct ) P∗ (4.278)
where δ is the Dirac delta function, t is time and P∗ is the value of the pressure
shock.
Taking into account eq. (4.278), eq. (4.269) can now be written:
∂4 w ∂2 w ∂2 w
f1 4
+ f 2 2 + f 3 w + ρ h 2 = δ ( x − ct ) P∗ (4.279)
∂x ∂x ∂t
Assuming that the ends of the shell are simply supported on their perimeters, the
above equation is associated with the following boundary conditions:
w ( 0, t ) = 0 (4.280)
∂ 2 w ( x, t )
=0 (4.281)
∂x 2 x =0
w ( L, t ) = 0 (4.282)
∂ 2 w ( x, t )
=0 (4.283)
∂x 2 x=L
Considering that the pipe is non-deformed at t = 0 , the following initial condi-

tions can be derived:
w ( x, 0 ) = 0 (4.284)
∂w ( x, t )
=0 (4.285)
∂t t =0
The partial differential equation (4.279) with the associated boundary and ini-
tial conditions given by eqs. (4.280)–(4.285) will be solved by using integral
transforms.
Taking into account the following relations of finite sine Fourier transform
[10]:
L
j πx
V ( j , t ) = ∫ w ( x, t ) sin dx j = 1, 2, 3,..... (4.286)
0
L
2 ∞ j πx
w ( x, t ) = ∑ V (i, t ) sin L (4.287)
L j =1
as well as the following property of the Dirac delta function [11]:
L
j πx j πct L
∫ δ ( x − ct ) P
∗
sin dx = P∗ sin for c〈 (4.288)
0
L L t
eq. (4.279) can be written in the following form:
j 4π 4 j 2π 2 jπ ct
4
f1V ( j , t ) − 2
f 2V ( j , t ) + f 3V ( j , t ) + ρ hV ( j , t ) = P∗ sin (4.289)
L L L
Using the following notations:
j 4π 4 f1
ω(2j ) = (4.290)
L4 ρ h
ρ h f3
Ω(2 j ) = ω(2j ) − ω( j ) f 2 + (4.291)
f1 ρ h
πc
ω= (4.292)
L
eq. (4.289) can be written:
P∗
V ( j , t ) + Ω(2 j )V ( j , t ) = sin ( jωt ) (4.293)
ρh
Inserting the Laplace transform
V ∗ ( j , ξ ) = L {V ( j , t ) ; ξ } (4.294)
into eq. (4.293) yields:
P∗ jω ξ 1
V ∗ ( j, ξ ) = (4.295)
(
ρ h ( ξ 2 + j 2ω 2 ) ξ 2 + Ω 2 j
( ) )
Taking the inverse Laplace transform [10] of the above equation, the following
solution can be derived:
jπ x
∞ sin
w ( x, t ) = w0 ∑ L  j 2 ( j 2 − a 2 ) sin jωt − ja ( j 2 − a 2 ) sin Ω( j ) t  (4.296)
2 2  
j =1 j ( j −a )
2 2
where w0 is the deflection with a static pressure ( c = 0 ) acting at the middle

x = L / 2 of the pipe and
ω
a= (4.297)
Ω(1)
Using the notation

Ω(1) L
= ccr (4.298)
π
and taking into account the period T(1) of the first free vibration
2π
T(1) = (4.299)
Ω(1)
as well as the time T for the shock pressure to travel over the pipe
L
T= (4.300)
c
the following relations can be obtained:
π
ω= (4.301)
T
2π
Ω(1) = (4.302)
T(1)
With the aid of eqs. (4.298)–(4.302), eq. (4.297) can be written in the following
form:
T(1) 2 π / Ω(1) c
a= = =
(Ω )
(4.303)
2T 2L / c
(1) L π
yielding
c
a= (4.304)
ccr
4.12.1 Example
For a multi-layered pipe made by E-glass/epoxy we shall derive the curves
w ( x, t ) for t = 0,1, 2,.....,15 sec.
Taking into account the mechanical properties of the composite material
E1 = 39 × 109 N m2
E2 = 8.6 × 109 N m2

ν 12 = 0.28
G12 = 3.8 × 109 N m 2
θ = π 12 (fiber orientation)
NP = 10 (number of plies)
h = 0.150 × 10−3 m (thickness of ply)
we calculate the following material parameters:
Si , j ( i, j = 11, 12, 21, 22, 66 )

Qi , j ( i, j = 11, 12, 21, 22, 66 )
Qi , j ( i, j = 11,, 12, 16, 22, 26, 66 )
 A11 A12 A16 B11 B12 B16 

A A22 A26 B12 B22 B26 
 12
A A26 A66 B16 B26 B66 
[ ABD ] =  B16 B12 B16 D11 D12 D16 

 11
 B12 B22 B26 D12 D22 D26 
 
 B16 B26 B66 D16 D26 D66 
 a11 a12 a16 b11 b12 b16 

a a22 a26 b21 b22 b26 
 12
a a26 a66 b61 b62 b66 
[ abd ] =  b16  = [ ABD ]
−1
 11 b21 b61 d11 d12 d16 

 b12 b22 b62 d12 d 22 d 26 
 
 b16 b26 b66 d16 d 26 d 66 
Taking into account the member a22 = 3.02085 × 10−7 of the above derived matrix
[ abd ] , and assuming the bulk modulus of elasticity∗ E = 219 × 107 N/m2 for water,
the pipe’s diameter D = 0.10 m and the celerity c = 1509 m s of the water, the
velocity of the pressure shock can be obtained by the equation:
c∗
c= = 184 m s
1 + a22 DE
For fluid velocity U = 3 m s , the pressure shock for instantaneous closure of the
valve is P∗ = ρUc = 1000 × 3 × 184 = 552000 Pa above the static pressure.
Using the members of the matrix [ abd ] , the matrices [ a1 ] , [ a2 ] , [ a3 ] , [ a4 ] can

now be obtained (eqs. 4.274–4.277):
 2.20976 × 107 20804 

[ a1 ] =  
 20388 62.7832 
 2.73226 × 107 62783.2 

[ a2 ] =  
 −217 993.966 
8.37109 × 107 −38600.6 

[ a3 ] =  
 19300.3 7.89715 × 106 
 2.77387 × 107 −14466.6 

[ a4 ] =  
 62783.2 907.115 
With the aid of the results above, the following parameters of the governing equa-
tion of wave propagation (eq. 4.279) can be derived:
 H11 H12 
= [ a1 ] − [ a4 ][ a3 ] [ a2 ]
−1
H 
H 22 
 21
or
H H12  1.30438 × 107 0.207852 

 11 = 
 H 21 H 22   −103.248 15.5797 
f1 = H 22 = 15.5797
1
f2 = ( H 21 + H12 ) = 2060.8
R
H11
f3 = = 5.21754 × 109
R2
Taking into account the coefficients f1 , f 2 , f 3 , ρ , h = NP × 0.150 × 10−3 , c, P∗ ,

the normalized deformation w ( x, t ) / w0 for a pipe with length L = 1000 m and
a = 0.8 can be obtained from eq. (4.296).
The following diagrams display the results for t = 0, 0.1, 0.2, ...., 1.0 sec.
,
,
,
,
References
[1] Ashley H., and Haviland G., “Bending vibrations of a pipeline containing
flowing fluid,” Journal of Applied Mechanics 17, 229–232, 1950.
[2] Païdoussis M.P., Fluid-Structure Interactions, Slender Structures and Axial
Flow, Vol. 1, Academic Press, 1998.
[3] Païdoussis M.P., Price S.J., and de Langre E., Fluid-Structure Interactions,
Cross-Flow-Induced instabilities, Cambridge University Press, 2011.
[4] Kollár L.P., and Springer G.S., Mechanics of Composite Structures,
[5] Hyer M.W., Stress analysis of fiber-reinforced composite materials,
DEStech Publications, 2009.
[6] Païdoussis M.P., and Luu T.P., “Dynamic of a pipe aspirating fluid such
as might be used in ocean mining.” ASME Journal of Energy Resources
Technology 107, 250–255, 1985.
[7] Deif A.S., Advanced matrix theory for scientists and engineers, Abacus
Press, 1991.
[8] Den Hartog J.P., Mechanical Vibrations, 4th edition, Mc Graw Hill, N.Y.
1956.
[9] Simon A.L., Hydraulics, John Wiley & Sons, 1986.
[10] Frýba L., Vibration of solids and structures under moving loads,
ThomasTelford, 1999.
[11] Hoskins R.F., Delta Functions, Horwood Publishing, 1999.
Chapter 5
Connection and Supports of Composite Pipelines
5.1 Joining of Composite Pipelines

Numerous joining methods are employed in composite pipelines depending on
the manufacturer. However, the standard method for joining filament-wound FRP
pipes is the butt joint (Fig. 5.1). This conventional joint is made by wrapping lay-
ers of fiber impregnated with a catalyst resin over the butted joint. The butt joint is
economical, permanent and very satisfactory [1]. Another widely used connection
method based on joining pipe pieces by adhesive layer is the “socket adhesive
joint” (Fig.5.2). In order for the joint to be effective, the adhesive layer must be
able to carry the developed shear stresses.
Figure 5.1 Butt joint.
169
170 Connection and Supports of Composite Pipelines
Figure 5.2 Socket adhesive joint.
Accurate determination of the stresses within the adhesive would require com-
plex elasto-plastic numerical modeling or experimental investigation. Since such
procedures are not efficient for engineering design [2], analytical models for es-
timating the allowable axial force and bending or torsional moment can be used
for design purposes.
5.1.1 Approximate mechanical model for axial loading

During axial loading of two filament-wound FRP pipes connected by an
adhesive joint (Fig. 5.1 or Fig. 5.2), the normal stresses in each pipe segment
are uniformly distributed along the periphery of the cross-section of each pipe.
Therefore, the mechanical model simulating the loading of the joint can be ap-
proximately represented by the configuration of two thin strips and an adhesive
layer, as depicted in Figure 5.3.
In this model, the center of coordinate x is located in the middle of L. The dis-
tribution of shear stresses along the length L of adhesive is not uniform. Therefore,
in order to estimate the maximum shear stress, the deformation of an element of
the adhesive layer will first be analyzed (Fig. 5.4).
Figure 5.3 Mechanical model of an adhesive joint.

Joining of Composite Pipelines 171
Figure 5.4 Deformation of an adhesive element.
The length of the initial non-deformed adhesive element ΑΒΓΔ is dx, while its
width is equal to the thickness ta of the adhesive layer. The axial force N acting
upon the strips 1 and 2 produces shear strain γ and normal mean strains ε1o and
ε 2o in the cross- section of each thin strip. Due to shear strain γ, the points Δ and Γ
will be moved to the new locations Δ1 and Γ1 respectively. Therefore, the element
ΑΒΓΔ will be transformed into the deformed element ΑΒΓ1Δ1 and the segment
Γ1∆1 will keep its length dx. Following the shear deformation, the normal strains
ε1o and ε 2o are assumed to act on the segments ∆1Γ1 and AB respectively, yield-
ing the corresponding elongation ε1dx and ε 2 dx . Therefore, the final location of
the points Γ and B will be Γ4B1 respectively.
If we draw the fictitious lines Β1Γ 2 and Β1Γ3 , the following geometric prop-
erties can be written:
Γ2Γ4
γ + dγ = (5.1)
Β1Γ 2
or
∆Γ 4 − ΑΒ1
γ + dγ = (5.2)
Β1Γ 2
Taking into account the definition of strain, the above equation can be written as:
(γ ta + dx + ε1dx ) − ( dx + ε 2 dx )
γ + dγ = (5.3)
ta
or
ε1 − ε 2
dγ = dx (5.4)
ta
which leads to:

dγ 1
= ( ε1 − ε 2 ) (5.5)
dx ta
Taking the derivative with respect to x in both sides of the above equation, it can
be determined that:
d 2γ 1  d ε1 d ε 2 
− − =0
dx 2 ta  dx dx 
(5.6)
Denoting by Ga the shear modulus of the adhesive, and using Hooke’s law
τ = Ga γ (5.7)
equation (5.6) can now be written in the following form:
d 2τ Ga  d ε1 d ε 2 
− − =0
dx 2 ta  dx dx 
(5.8)
Let us now investigate the equilibrium of an element of strip 1 (Fig. 5.5).
Figure 5.5 Equilibrium of a strip element.

Denoting by Ex the extensional modulus of elasticity in the x-direction [e.g.,

3] of a lamina given by:
E1
Ex = (5.9)
 E  E
cos 4 θ +  1 − 2ν 12  sin 2 θ cos 2 θ + 1 sin 4 θ
 G12  E2
where θ is the fiber orientation and E1 , E2 , G12 , ν 12 are the elasticity properties
along the principal axes, the stress in the x-direction (of each lamina) can be given
by Hooke’s law in a global coordinate system:
σ x = Ex εx (5.10)
Because of the symmetric winding (with the fibers orientation ±θ ) of a multi-

layered filament-wound pipe, the above Hooke’s law for a pipe’s laminated wall
consisting of NP layers can be written as:
σ x = NP Ex εx (5.11)
Therefore, the equilibrium equation of the strip element 1 (Fig. 5.5) can be writ-
ten as:
Ex ( ε1 + d ε1 ) NP s t − τ s dx − Ex ε1 NP s t = 0 (5.12)
which gives
d ε1 1
= τ (5.13)
dx Ex NP t
The corresponding result for strip 2 has a similar form:
dε 2 1
=− τ (5.14)
dx Ex NP t
With the aid of equations (5.13) and (5.14), equation (5.8) can now be written:
d 2τ
− λ 2τ = 0 (5.15)
dx 2
where
2 Ga
λ2 = (5.16)
Ex NP t ta
The general solution of equation (5.15) is:
τ ( x) = A1 cosh(λ x) + A2 sinh(λ x) (5.17)
where A1 , A2 are constants to be determined by the following conditions at

x = L / 2 and x = 0
ε 2 ( L / 2) = 0 (5.18)
σx
ε 1 ( L / 2) = (5.19)
Ex NP
ε1 (0) = ε 2 (0) (due to symmetry) (5.20)
Since
∧
N
σx = (5.21)
π Dt
∧
where N is the axial force acting on a pipeline, eq. (5.19) can be written as:
∧
N
ε 1 ( L / 2) = (5.22)
π D t E X NP
With the aid of eq. (5.20), equation (5.5) yields:
dγ
=0 (5.23)
dx x = 0
Combining the above equation with equation (5.7) it can be determined that:
dτ
=0 (5.24)
dx x = 0
Therefore, eq. (5.17) with the aid of the equation above yields:
A2 = 0 (5.25)
Thus, the general solution for shear stress distribution along the adhesive can now
be simplified:
τ ( x) = Α1 cosh(λ x) (5.26)
Combining equation (5.5) with equations (5.18) and (5.22) can be written as:
 ∧


dγ
=
1 N − 0 (5.27)
dx x=L/2 ta  π D t Ex NP 
 
The above equation with the aid of eq. (5.7) yields

dτ G
= a⋅ N (5.28)
dx x=L/2 ta π D t Ex NP
With the help of the above equation, the unknown constant A1 can now be deter-
mined by eq. (5.26):
∧
Ga N
A1 = (5.29)
ta π D t Ex NP λ sinh(λ L / 2)
Therefore, the distribution of shear stress along the length of the adhesive can be
approximated by the formula:
∧
Ga N
τ ( x) = cosh(λ x) (5.30)
ta π D t Ex NP λ sinh(λ L / 2)
A graphical representation of the above distribution is shown in Fig. 5.6. From

this figure it can be concluded that the maximum values of τ ( x) occur at the ends
x = ± L / 2 of the adhesive layer. Therefore,
τ max = τ ( L / 2) (5.31)
Figure 5.6 Shear stress distribution along the adhesive layer.
or with the aid of eq. (5.30)
∧
Ga N
τ max = (5.32)
ta π D t Ex NP λ tanh(λ L / 2)
τ max has a considerably higher value than the mean value:
∧
N
τm = (5.33)
π DL
Combining eqs. (5.32), (5.33), the stress concentration factor K h at the ends of
the adhesive layer is
τ max
Kh = (5.34)
τm
or
Ga L
Kh = (5.35)
ta t Ex NP λ tanh(λ L / 2)
Assuming that adhesive failure will occur when:
τ max = τ α (5.36)
where τ α is the allowable shear stress of the adhesive, the allowable axial force
∧
N a for the joint can be approximated by combining equations (5.32) and (5.36):
∧
τα
Na = ta π D t Ex NP λ tanh(λ L / 2) (5.37)
Ga
Schematic graphical representation of the above equation is shown in Fig.5.7.

According to this figure it can be concluded that for large values of L, the capacity
of the joint to carry axial force is independent of the adhesive length.
Since for large values of L
tan(λ L / 2) → 1 (5.38)
the allowable axial force tends asymptotically to the value:
τα
Fmax → ta π D t Ex NP λ (5.39)
Ga
Results for the allowable axial force of butt joints for a wide range of multi-
layered filament wound pipes made from E-glass/epoxy and S-glass/epoxy are
presented in Chapter 10.
Figure 5.7 Capacity of a joint to bear axial force versus adhesive length.
5.1.2 Approximate mechanical model for bending

For estimating the allowable bending moment for a butted joint, we shall again
use the value of an imaginary axial force N that causes the same maximum strain
ε ξo with the actual bending moment M (see Chapter 3).
According to this concept (Fig. 5.8), the axial force N yielding equivalent max-
imum strain with the bending moment M can be approximated from the following
formula:
�
N M� D
α11 = (5.40)
π D EI
� 2
where
π
� = 2  D + 1  cos 2 θ D dθ
2
EI ∫0  4a11 d11  2
(5.41)
Therefore,
� = 4 πD 2 d11 �
N M (5.42)
d11 D 2 + 4a11
With the aid of the above equation, eq. (5.37) provides the approximate value of
the allowable bending moment M �
a :
Figure 5.8 Approximation of the maximum local strain caused by a bending moment M,
with the local strain caused by an axial force N.
Above-Ground Pipes 179
 D a11  τ α
�
M a = +  ta t Ex NP λ tanh ( λ L 2 ) (5.43)
 4 d11 D  Ga
Results for the allowable bending moment of butt joints for a wide range of mul-
tilayered filament-wound pipes made from E-glass/epoxy and S-glass/epoxy are
displayed in the nomographs in Chapter 10.
5.2 Above-Ground Pipes

When a pipeline is installed above the ground, the types of supports shown in
Figures 5.9-5.11 are required.
Pipelines above ground can be suspended (Fig. 5.9), clamped on the ground
(Fig. 5.10), or placed in a shoe (Fig. 5.11).
Important parameters to be derived for pinned-supported pipes are: (a) the
maximum spacing between supports, (b) the minimum hanger widths, and (c) the
allowable deflection of expansion loops for anchored pipelines.
Figure 5.9 Support for a suspended pipe.

Figure 5.10 Pipeline clamped on the ground.
Figure 5.11 Pipeline supported by a shoe.
5.2.1 Maximum spacing between supports

In order to estimate the maximum space L between supports of a long pinned-
supported pipeline, the static model of a continuous beam (Fig. 5.12) will be ad-
opted to simulate the pipeline.
One criterion for estimation of L is the maximum bending moment M max given
by [4]:
M max = λ wL2 (5.44)

Figure 5.12 Bending moments’ distribution along a pinned-supported pipeline.
where λ is a coefficient given in Table 5.1, and w is the uniformly distributed load
per longitudinal unit. w depends on the specific gravities γ f , γ p and material
volumes V f , V p of the fluid and pipe respectively. Therefore,
π D2
w=π Dt γp + γf (5.45)
4
where t and D are the thickness of the pipe’s wall and the mean diameter,
respectively.
Combining equations (5.44) and (5.45), the maximum length between supports
can be estimated by the following rule:
M max
L≤ (5.46)
λπ D ( tγ p + 0.25 Dγ f )
In the above equation, the value of M max should be the minimum value between
the allowable bending moment M a (for avoiding failure) and the critical moment
M cr (for avoiding buckling) derived in Chapter 3.
M max = min { M a , M cr } (5.47)
According to Table 5.1 it can be shown that the value of λ tends to be stable for
more than six supports. The maximum value λ = 0.125 occurs in the span of a pipe
supported by two supports or over the intermediate support for a pipe resting on
three supports. In practice, it is prudent and safe to adopt this maximum value for
every case of pinned-supported pipeline. Taking into account equations (5.46),
(5.47) and the diagrams demonstrating the allowable and the critical value of the
bending moment for pipes under bending made of E-glass/epoxy and S-glass/
epoxy, results have been calculated that estimate the maximum spacing Lmax be-
tween supports for a wide range of multi-layered filament-wound pipelines. These
Table 5.1
Values of coefficient λ for continuous pipeline over equal spans (Fig.5.12).
Number of supports Coefficient λ corresponding Coefficient λ corresponding
to maximum bending to maximum bending
moment over support moment in span
2 0 0.125
3 0.125 0.703
4 0.100 0.080
5 0.107 0.077
6 0.105 0.078
7 0.105 0.078
8 0.105 0.078
results are presented in Chapter 10. For the derivation of the above results, the
4 3
usual values γ p = 104 N / m3 and γ f = 1.3 x 10 N / m were adopted for the pipes’
material and the fluid, respectively.
5.2.2 Minimum hanger widths

In order to estimate the minimum hanger width B, the segment of pipe within
the hanger will be simulated using a thin-walled ring (Fig. 13a) of mean radius
R = D/2.
The contact force Q� transferred from the hanger to the pipe (point C) causes
[5] a bending moment M �
c to the lower area of the ring (Fig. 13b).
�
M �
c ≈ 0.24QR (5.48)
�
and a compressive force N c
� ≈ 0.24Q
N � (5.49)
c
Moreover, compressive forces N �

ξ act in the x direction due to bending of the
�
pipeline by the bending moment M y (Fig. 5.13a). We recall that the above com-
pressive forces can be estimated by eq. (5.42).
Figure 5.13 (a) Simulation of a pipe segment inside a hanger using a thin-walled ring,
�
(b) Bending moment M �
c and axial force N c due to Q.
Figure 5.14 shows the loading conditions of a pipe’s wall element located at
point C due to shear force Q � and bending moment M �
y .
Taking into account that the width of the element is denoted by B, the unit
forces M c , N c , Nξ can be obtained with the help of eqs. (5.48), (5.49), (5.42),
and (5.44):
�
0.24QR
Mc = (5.50)
B
�
0.24Q
Nc = (5.51)
B
4 Dd11 M� y
Nξ = 2
(5.52)
d11 D + 4a11 B
In the above equation the value of M � y can be substituted by the value obtained
�
by eq. (5.44). The shear force Q for every pipe-cross section, left or right of each
support in a continuous pipeline, as shown in Figure 5.12, is given by the formula:
� = µwL
Q (5.53)
Figure 5.14 Loading conditions of a pipe’s wall element located at point C (see Figure 5.13).
where the coefficient µ can be obtained from Table 5.2.

According to Table 5.2 it can be shown that the value of µ tends to be sta-
ble for more than six supports. The maximum value µ = 0.605 can be adopted,
yielding:
0.145wLR
Mc = (5.54)
B
0.145wL
Nc = (5.55)
B
Table 5.2
Values of the coefficient µ for a continuous pipeline over
equal spans, as depicted in Figure 5.12.
Number of supports Coefficient µ corresponding to maximum shear force
on each side of supports L =left, R =right
L R
2 0 0.5
3 0.625 0.625
4 0.600 0.500
5 0.607 0.536
6 0.605 0.526
7 0.605 0.029
8 0.605 0.529
With the aid of eq. (5.44) and taking into account the value λ = 0.125, eq. (5.52)
yields:
0.5 Dd11 wL2

Nξ = (5.56)
(d 11 D 2 + 4a11 ) B
Combining the laminate compliance matrix (“abd” matrix) with eqs. (5.54),
(5.55), and (5.56) the following matrix equation can be obtained:
ª D D D E E E − º 1 [ ½ ½

«D ° ° ° °
« D D E E E − »» ° 1 \ ° ° °

« D D D E E E − » ° 1 [\ ° ° °
« » ° ° ° °
« E E E G G G − » ° 0 [ ° ° °
« E E E G G G − » ° 0 \ ° ° °
« » ° ° ° °
« E E E G G G −» ° 0 [\ ° ° °
« ® ¾ = ® ¾
» ° ε [ ° ° − 1ξ °
« »
» ° ε \ ° °− 1 F °

«
« ° ° ° °
» ° γ [\ ° ° °
« »
» ° N [ ° ° °

«
« » °
° ° °
« » ° N \ ° °− 0 F °
¬« ¼» ¯° N [\ ¿° °¯ °¿

(5.57)
Solutions of the above equation provide the values of strain components

ε xo , ε yo , γ xyo , k xo , k yo , k xyo :
ε x0 
 0 b12 a12 a11 
ε
 y  b a12 
 22 a22
 0 M c 
γ
 xy b62 a26 a16   
 0  = −   Nc  (5.58)
k x   d12 b21 b11  N 
k 0  d b22 b12   ξ
 y   22 
k 0   d 26 b26 b16 
 xy 
Therefore, on the exterior layers of the laminate, the strain components εx, εy, γxy
can be obtained from the following equation:
ε x  ε x0  k x0 
   0    0 
ε y  = ε y  ± z k y  (5.59)
   0  0
γ xy  γ xy  k xy 
where z is the half thickness of the laminate.

With the aid of eq. (5.59), the stress components σ x , σ y , τ xy can be obtained
for each lamina using the following well-known relation:
σ x  εx 
    
σ y  = Q ij (θ )   ε y  (5.60)
τ  γ 
 xy   xy 
Using the matrix [ Τ(θ ) ] , the principal stresses σ 1 , σ 2 ,τ 12 in the exterior layers
can be determined as follows:
σ 1  σ x 
   
σ 2  = T (θ )  σ y  (5.61)
τ  τ 
 12   xy 
With the aid of the above equation, the Tsai-Wu failure criterion can provide an
estimation of the minimum hanger width B. Results of the hanger width B for a
wide range of multi-layered filament wound pipes made of E-glass/epoxy and
S-glass/epoxy are presented in Chapter 10.
5.2.3 Sizing of expansion loops

Pipelines with free ends experience changes in dimensions as a result of tem-
perature variation. When the ends are anchored, the pipeline will be placed in
a condition of stress and will exert reactive forces and moments at its ends. To
avoid the catastrophic consequences of possible buckling due to the above condi-
tions, expansion loops (Fig. 5.15) are installed, especially for longer lines. The
basic problems in analyzing temperature change effects in a configuration such as
that seen in Fig. 5.15 are to estimate the magnitude of such stresses and to check
whether or not they are tolerable. For these problems, the mechanical model of
Fig. 5.16 will be used to simulate the system pipeline-expansion loop shown
in Fig. 5.15. Assuming that an expansion loop is located at the midpoint of the
distance between two anchors, we can analyze the half structure because of its
symmetry.
Figure 5.15 Design configuration of an expansion loop.
Figure 5.16 Mechanical model of the configuration of an expansion loop.
In the model shown in Figure 5.16, the left anchor (point A) and the point
located in the middle of the horizontal part of the expansion loop (point D) are
simulated by fixed supports. The point B joining the pipeline with the expansion
loop is guided horizontally. The aim of this analysis is to estimate the values of
the compressive axial force F in the pipe AB and the bending moment Μ Β at the
point B (critical point) due to a temperature increase of ΔΤ. These values will be
used for checking the tolerance of the structure to failure and buckling. In order
to estimate the axial force F, the L-shaped part of the extension loop (part BCD)
will be simulated by a linear spring (Fig. 5.17).
The equivalent spring constant k can be estimated by the ratio of the horizontal
force acting on point B of the expansion loop (Fig. 5.16) over its corresponding
horizontal movement, thus: ∆xB
F
k= (5.62)
∆xB
Figure 5.17 Simulation of the expansion loop by a linear spring. (a) Structure when
∆Τ = 0 ; (b) Structure after temperature increase ∆Τ ≠ 0 ; (c) Deflection when the end
B is free.
According to [5], the horizontal movement ∆xB of the point B in the L-shaped
frame BCD (Fig. 5.16) can be obtained by the formula:
∆xB = CHV VB + CHM M B − LFH (5.63)
where VB , M B are the vertical reaction and bending moment acting in the support
B (Fig. 5.16). The parameters CHV , CHM , LFH are given by [5]:
ab 2
CHV = (5.64)
2 EI
a2 ab
CHM = + (5.65)
2 EI EI
LFH = FCHH (5.66)
where
a3 a 2b
CHH = + (5.67)
3EI EI
In the above formulas, E can be approximately substituted with the extensional

modulus of elasticity Ex given by equation (5.9), and I is the moment of inertia
of a thin-walled pipe given by:
3
 D
I = π  t (5.68)
 2
where t, D are the thickness and the diameter of the pipe.
The reactions VB , M B can be obtained also from [5]:
LFV ⋅ CMM − LFM ⋅ CVM

VB = (5.69)
CVV ⋅ CMM − CVM
2
LFM ⋅ CVV − LFV ⋅ CVM

MB = (5.70)
CVV ⋅ CMM − CVM
2
where
LFV = F ⋅ CVH (5.71)
LFM = F ⋅ CMH (5.72)
ab 2
CVH = (5.73)
2 EI
a2 ab
CMH = + (5.74)
2 EI EI
a+b
CMM = (5.75)
EI
b2
CVM = (5.76)
2 EI
b3
CVV = (5.77)
3EI
Combining equations (5.69)–(5.72), one can arrive at:
CVH CMM − CMH CVM

VB = F (5.78)
CVV CMM − CVM
2
CΜH CVV − CVH CVM

ΜB = F (5.79)
CVV CMM − CVM2
With the aid of eqs. (5.66), (5.78), (5.79), eq. (5.63) yields:
CVH CMM − CMH CVM C C − CVH CVM

∆xB = [CHV 2
+ CHM MH VV − CHH ]F (5.80)
CVV CMM − CVM CVV CMM − CV2M
Combining the above equation with eq. (5.62), the following formula, which pro-
vides the spring constant k, can be obtained:
CVH CMM − CMH CVM C C − CVH CVM

k = [CHV 2
+ CHM MH VV 2
− CHH ]−1 (5.81)
CVV CMM − CVM CVV CMM − CVM
According to Fig. 5.17c, for the case of free expansion of pipe due to a temperature
rise ΔΤ, the movement e1 + e2 of the end B can be approximated by the formula:
e1 + e2 = a x L ∆Τ (5.82)
where a x is the longitudinal (i.e., along the pipe’s axis) coefficient of thermal
deformation. a x can be approximated by the following formula:
α x = α1 cos 2 θ + a2 sin 2 θ (5.83)
In the above formula a1 , a 2 are the thermal deformation coefficients in principal

coordinates 1 and 2, and θ is the fiber orientation in a filament wound pipe with
the stacking sequence [±θ ] .
Since the spring (Fig. 17b) is compressed by a deflection:
e1 = ∆xB (5.84)
according to eq. (5.62), it can be written that:
F = k ⋅ e1 (5.85)
where the constant k is given by the eq. (5.81). On the other hand, due to the axial
force F, the longitudinal deflection of the pipe (Fig. 17b) can be approximated by
the formula:
Ε⋅Α
F= e2 (5.86)
L
where E can be approximately substituted with Ex, which is given by eq. (5.9), and
A is the pipe wall’s cross-section:
Α = πDt (5.87)
Combining eqs. (5.85) and (5.86) yields:
Lk
e2 = e1 (5.88)
EA
With the help of the above equation, eq. (5.82) can now be written:
E ⋅ A ⋅ α x ⋅ L ⋅ ∆Τ
e1 = (5.89)
E⋅ A+ k⋅L
Taking into account the above equation, the axial force F can now be estimated
by eq. (5.85)
E ⋅ A ⋅ α x ⋅ L ⋅ ∆Τ
F=k (5.90)
E⋅ A+ k⋅L
Combining the above equation with eq. (5.79), the value of bending moment MB
of the point B joining the pipe with the expansion loop can now be estimated as:
E A ax L ∆Τ CMH CVV − CVH CVM

MB = k ⋅ (5.91)
E A+k L CVV CMM − CVM2
Using the values F, MB, the following conditions must be satisfied for an extension
loop to be acceptable:
F < Na (5.92)
F < λcr (5.93)
MB < M a (5.94)
M B < M cr (5.95)
In the above equations, N a is the allowable axial force to avoid failure (Chapter 3),
λcr is the minimum eigenvalue (Chapter 3) in order to avoid buckling, M a is the
allowable bending moment to avoid failure (Chapter 3) and M cr is the critical
value of a bending moment to avoid buckling.
5.3 Underground Pipelines

Locating composite pipelines under roads and railroads is a common practice.
In these cases the pipeline should be installed at a great enough depth, which is the
main design parameter. For estimating the minimum value of installation depth,
the soil load and wheel load should be accounted for. Due to these loads, any
soil element located at a distance z below the surface is subjected to compressive
stresses s y , s z (see Fig. 5.18).
Since during maintenance the pipeline is empty, the above compressive stresses
cannot be equilibrated by internal pressure. Therefore, internal pressure should be
ignored and only s y , s z due to wheel and soil load should be taken into account
for dimensioning the installation of an underground pipeline. From Figure 5.18 it
can be shown that s y vanishes rapidly versus depth. Therefore, the estimation of
the installation depth by taking into account only s z should be a safe scenario.
The distribution of s z versus z below the contact surface can be obtained from
the following solution of Boussinesq [6] concerning semi-infinite elastic, homo-
geneous, isotropic soil:
3F
σz = (5.96)
2π z 2
Figure 5.18 Compressive stresses versus distance below surface.

Underground Pipelines 193
According to Burland et al. [7], the above solution accurately describes the stress
s z for most soil conditions.
Assuming that the pipe segment influenced by the surface load F is a ring of
radius R and width b, the mechanical model shown in Fig. 5.19 can be safely
used for estimating the minimum installation depth zmin . In this model q is the
unit load given by:
q = bs z (5.97)
With the aid of eq. (5.96) the above equation can be written as:
3bF
q= (5.98)
2 πz 2
Figure 5.19 Mechanical model of the loading of an underground pipe embedded in instal-
lation depth z.
The bending moment M � c and axial force N

� c acting on the point C (conserva-
tive scenario) are given by the following formulas:
� c = 1 q R2
M (5.99)
4
� c =q R
N (5.100)
The combination of equations (5.98)–(5.100) yields the unit bending moment

My = M� c b and the unit axial force N = N
� c b acting on the pipe’s cross-sec-
y
tion (Fig. 5.20) at location C.
3 R2
My = F (5.101)
8π z 2
3 R
Ny = F (5.102)
2π z 2
Combining the laminate compliance matrix with equations (5.101) and (5.102),
the following matrix equation can be obtained:
½
ª D D D E E E − º 1[ ½ ° °
«D » °1 ° ° °
D D E E E − » ° \° ° °
«
« D D D E E E − » ° 1 [\ ° °
°
« » ° ° ° °
« E E E G G G − » ° 0[ ° ° °
«E E E G G G − » °0\ ° °
°
« » ° ° ° °
« E E E G G G −» ° 0 [\ ° ° °
« » ® ¾= ® ¾
«

» ° ε[ ° ° 5 °
« » ° ε \ ° ° − π ] ) °
« ° ° ° °
» ° γ [\ ° ° °
« »
« » ° N [ ° ° °
« » ° ° ° °
« » ° N\ ° ° 5
°
° ° ° − ) °
«¬ ¼ » N
¯ [\ ¿ ° π ]
°
¯ ¿
(5.103)
Underground Pipelines 195
Figure 5.20 Bending moment and axial force acting on the longitudinal cross-section of
a composite pipe.
Solution of the above equation yields:
 ε x0   b12 a12 
 0 b
ε y   22 a22   3 R2 
γ xy0   b62 a26   F 
8 z2 
 0  = −    (5.104)
k
 x  d12 b21   3 R F
 ky 0  d 22 b22   2π z 2 
 0  
 k xy   d 26 b26 
Using eq. (5.104) the strain components ε x , ε y , γ xy of the exterior layer of the
composite wall can be obtained by the following equation:
ε x  ε y  k x0 
0
   0  h  0 
ε y  = ε y  ± k y  (5.105)
   0  2  0
γ xy  γ xy  k y 
where h is the thickness of the laminate. Therefore, the stress components

σ x , σ y ,τ xy of the exterior lamina are
σ x  ε x 
   
σ y  = Q ij (ϑ )  ε y  (5.106)
   
τ xy  γ xy 
and the principal stresses σ 1 , σ 2 ,τ 12 can be obtained with the aid of matrix
Τ (ϑ ) 
σ 1  σ x 
   
σ 2  = Τ (ϑ )  σ y  (5.107)
τ   
 12  τ xy 
Using the above equation, the Tsai-Wu failure criterion can provide an estimation of
the minimum installation depth zmin versus the wheel and the soil load F , the pipe’s
diameter Dia , the fiber orientation ϑ , and the number of plies NP . Results for
a wide range of multilayered filament-wound pipes made of E-glass/epoxy and
S-glass/epoxy are displayed in Chapter 10.
References
[1] Mallinson J.H. Corrosion-resistant plastic composites in chemical plant de-
sign, Marcel Dekker, 1988.
[2] Wang C.H., and Rose L.R.F., “Stress analysis and failure assessment of lap
joints” in: Recent Advances in Structural Joints and Repairs for Composite
Materials, edited by Tong L., and Soutis C., Kluwer Academic Publishers,
2010.
[3] Kollár L.P., and Springer G.S., Mechanics of Composite Structures,
[4] Vidosic J.P., “Mechanics of materials” in: Marks’ Standard Handbook for
Mechanical Engineers, 8th edition, edited by Baumeister Th. et al., McGraw-
Hill, 1979.
[5] Young W.C., ROARK’S Formulas for Stress & Strain, McGraw-Hill, 1989.
[6] Boussinesq J., Application des potentials_lY_tude de lY_quilibre et du
movement des solids elastiques, Gauthier-Villars, Paris, 1885.
[7] Burland J.B., Broms B.B., and de Mello V.F.B., “Behaviour of founda-
tions and structures,” Proc. 9th Int. Conf. Soil Mechanics and Foundation
Engineering, Tokyo, 1978.
Chapter 6
Creep Design of Piping Applications Using

Composite Materials
Rui Miranda Guedes1 and Hugo Faria2

1
Department of Mechanical Engineering (DEMec), Faculty of Engineering of the
University of Porto (FEUP), Rua Dr. Roberto Frias s/n, 4200-465 Porto, Portugal.
2
INEGI-Instituto de Engenharia Mecânica e Gestão Industrial, Rua Dr. Roberto Frias,
400, 4200-465 Porto, Portugal
6.1 Introduction
The application of glass fiber-reinforced thermoset matrix composites as pri-
mary structures has been impeded by a number of problems. Ochoa and Salama
[1] pointed out five basic reasons, which include two very difficult challenges:
the need to have databases for long-term damage mechanisms for lifetime predic-
tion and the need for nondestructive evaluation (NDE) and in-service monitoring.
Both are interrelated in terms of simulating service loads, manufacturing, and
installation procedures. The proper environmental and loading conditions used to
accelerate long-term testing are not fully known in advance. To be representative,
they must generate damage evolution and failure modes similar those occurring in
real conditions over 20 to 30 years. The heterogeneity and anisotropy introduced
by the fiber and matrix make such an analysis very complicated. Moreover any
change in the manufacturing procedure, matrix or fiber, produces a different mate-
rial system, which in turn invalidates the use of previous experimental databases.
Pipelines, risers and piping systems are examples of primary applications of
filament-wound composites in the oil and gas industry. The thermosetting resins
commonly used are epoxy and polyester resin systems. The advantages are well
known: high stiffness-to-weight ratio and corrosion resistance.
The design of these structures is highly demanding, since they are expected
to remain in service for more than 20 years. During service, pipe structures may
197
198 Creep Design of Piping Applications Using Composite Materials
be subjected to both permanent and cycling loads. Although they do offer high
mechanical performance, over time their strength and stiffness may decay sig-
nificantly, a consequence of the viscoelastic nature of the matrix, damage accu-
mulation within the matrix and fiber breakage. One serious consequence of these
events is premature failure, usually catastrophic.
The lack of full understanding of the fundamental parameters controlling long-
term materials’ performance necessarily leads to over-design. In this context, the
lifetime prediction of these structures is an important issue not completely solved.
Standards for certification of glass reinforced plastic (GRP) pipes require at least
10,000 hours of testing and a high number of specimens. Even though these strin-
gent requirements may be seen as reasonable in terms of safety, they severely re-
strict the improvement and innovation of new products, which in turn may inhibit
greater commercial penetration in the markets.
The present chapter will present the following argument. The second sec-
tion reviews selected theoretical approaches for long-term failure criteria. Time-
dependent failure criteria are presented and developed in a manner useful for
practical engineering applications. The third section attempts to couple damage
with creep and fatigue effects on a long-term basis, based on safety factors. It
starts with a brief theoretical description of damage initiation and propagation
during static loading. Long-term sustained and cycling loads usually produce
damage similar to that originated under quasi-static loading. Experimental creep
failure data under sustained hydrostatic pressure are presented, followed by an
example of preliminary design for long-term creep. The fourth section is devoted
to fatigue failure. Experimental data published for filament-wound pipe failure
under cycling hydrostatic pressure are presented and discussed. The fifth section
describes and compares the standards for the design and qualification of pipes.
Finally, the sixth section presents a worked example with preliminary calculations
of qualification pressure for a filament-wound pipe, using known material proper-
ties, based on ISO 14692.
6.2 Creep Damage Accumulation Mechanisms in

Composite Materials
In engineering, creep designates the gradual increase in strain that occurs in a
material when it is subjected to a sustained load over a certain period of time. Even
at room temperature polymers and polymer-based composites can undergo creep at
relatively low stress levels. Moreover, given enough time, creep will lead to rupture.
Usually the creep strain evolution may be divided into three distinct regions:
I) first-stage or primary creep, which starts at a rapid rate and slows with time; II)
second-stage (secondary) creep, where creep has a relatively uniform rate (mini-
mum gradient); III) third-stage (tertiary), where creep manifests an accelerating
creep rate and terminates with failure of the material. This is illustrated schemati-
cally in Figure 6.1.
Creep Damage Accumulation Mechanisms in Composite Materials 199
Failure
Creep strain Region III

Region II
Region I
Time
figure 6.1 Typical creep strain evolution divided into three regions.
General strength theories do not include the creep-to-yield or creep-to-rupture

process. Since the stress-strain analysis is based on continuum mechanics, a dif-
ficulty presents itself when attempts are made to predict failure in general and
creep failure in particular of polymers and polymer matrix based composites.
Fracture mechanics and damage mechanics include the distribution of defects into
continuum models, which allow time-dependent failure prediction. Energy-based
failure criteria provide another possible approach. An example is establishing that
the energy accumulated in the springs of the viscoelastic mechanical model, des-
ignated as free energy, has a limit value. This limit can be or not be constant.
Earlier approaches to the prediction of time-dependent failures provided explicit
elementary equations to predict behavior over a lifetime. Following this line, a
global and homogeneous analysis was chosen because it is more convenient for
practical engineering applications.
In this context it becomes important to predict the effects of long-term load-
ing on deformation and failure behavior. A reliable prediction requires the use of
models that capture accurately the stress/strain time-dependent evolution of the
filament-wound cylindrical pipes. Most plastics, composites and other synthetic
materials exhibit strong hereditary type non-linear viscoelastic behaviors, which
render any stress-strain analysis of these materials a significant challenge. Single-
integral formulations, such as the Schapery nonlinear viscoelastic theory [2], have
proven to be very effective in representing the creep strain of filament-wound
pipes [3]:
t d ( g 2σ (τ ) )
ε ( t ) = J 0 g 0σ ( t ) + g1 ∫ ∆J (ψ −ψ ′ ) dτ (6.1)
0 dτ
with the correspondent reduced times Ψ and Ψ ′ given by:
tdτ ′ τ dτ ′
Ψ=∫ , Ψ′ = ∫ (6.2)
0 a 0 aσ
σ
where g 0 , g1 , g 2 , as are stress-dependent nonlinearizing parameters. The kernel

∆J ( t ) is the time-dependent compliance. The compliance is usually expressed
by a power law [2–4]:
∆J ( t ) = J1t n (6.3)
where J1 , n are linear viscoelastic parameters.

Ghorbel [3] simplified the Schapery formulation since he did not perform all
the necessary creep and creep-recovery tests to obtain the full material stress-
dependent nonlinearizing parameters. Furthermore, it was found that g 0 = 1 . The
simplified expression for strain creep under sustained load was
ε ( t ) = ε ( 0 ) + g (σ ) t n (6.4)
where g (σ ) = βσ and β and m are material properties.

m
It must be emphasized that long-term strain creep measurement for filament-

wound cylinders is not usually done; only the time-to-failure is needed to qualify
pipes.
In the case of multi-directional filament-wound pipes, it is possible to perform
elastic-viscoelastic analyses using an approach based on classical laminated plate
theory (CLPT) as exemplified in [5]. That specific methodology includes a time-
dependent failure criterion. However for thick cylinders the CLPT approach does
not work. Hence a method to calculate the time-dependent stress-strain state in
nonlinear viscoelastic thick multilayered composite cylinders was proposed and
implemented [6]. This work demonstrated the importance of the viscoelasticity
effect over the time-dependent internal stress distribution through thick laminated
cylinders.
One of the first theoretical attempts to include time in a material strength for-
mulation for viscoelastic materials was developed by Reiner and Weissenberg [7].
The Reiner-Weissenberg criterion [7] states that the work done during the loading
by external forces on a viscoelastic material is converted into a stored portion
(potential energy) and a dissipated portion (loss energy). Hence the criterion as-
sumes that the instant of failure depends on a conjunction between distortional
free energy and dissipated energy, with the threshold value of the distortional
energy being the governing quantity.
Let us assume that the unidirectional strain response of a linear viscoelastic
material under arbitrary stress s(t), is given by the power law as:
t  t − τ  ∂σ (τ )
n
ε ( t ) = D0σ ( t ) + D1 ∫   dτ (6.5)
 τ 0  ∂τ
0
where D0, D1, n are material constants and t0 represents the time unity (equal to
1 second or 1 hour or 1 day, etc.).
The free stored energy, using the Hunter [8] formulation, is given by:
1 t t  2t − τ 1 − τ 2  ∂σ (τ 1 ) ∂σ (τ 2 )
n
1
Ws ( t ) = ε ( t ) σ ( t ) − D0σ ( t ) − ∫ ∫ D1  dτ 1dτ 2
2

2 2 0 0  τ0  ∂τ 1 ∂τ 2
. (6.6)
The total energy is defined as:
t ∂ε (τ )
Wt ( t ) = ∫ σ (τ ) dτ . (6.7)
0 ∂τ
Accordingly, these time-dependent failure criteria [9] predict the lifetime under
constant load as a function of the applied load s0 and the strength under an instan-
taneous condition sR:
D
The Reiner-Weissenberg Criterion (R-W) states that Ws ( t ) ≤ 0 s R 2 ,
2
1 1
 t f   1  n  D0  n  1  1
 =   − 1
n
n  
. (6.8)
 τ 0   2 − 2   D1   γ 
D0 2
The Maximum Work Stress Criterion (MWS) states that Wt ( t ) ≤ sR ,
2
1
 tf   1 D0  n  1  1
=   − 1
n
 . (6.9)
τ0   2 D1   γ 
The Maximum Strain Criterion (MS) states that ε ( t ) ≤ D0σ R ,
1
 t f   D0  n  1  1
 =   − 1 n
. (6.10)

 τ 0   D1   γ 
The modified Reiner-Weissenberg Criterion (MR-W) states that
D 
Wt ( t ) ≤  0 s R  s ( t )
 2  ,
1 1
 tf   1  n  D0  n  1  1
=   − 1 n
 n   
. (6.11)
τ0   2 − 2   D1   γ 
where γ = σ 0 2 σ R 2 .
In short, these approaches establish a relationship between time to failure, vis-
coelastic properties and static stress failure throughout a stored elastic energy
limit concept. As an approximation, it is not difficult to conclude that we can take
t~σ−2/n for the R–W and the MSW criteria and t~σ−1/n for the MS and the MR-W
criteria. Similar results were obtained using fracture mechanics concepts [10]. In
fact, these concepts established a relationship between time-to-failure, viscoelas-
tic properties and strength properties [11-12], which are similar to the previous
approach. The main difference in these failure criteria is the interpretation of the
physical constants. According to Song et al. [13], there are three major phenom-
ena, which frequently occur simultaneously, responsible for the creep failure of
viscoelastic materials: (1) time-dependent constitutive equations; (2) time to the
formation of overstressed polymer chains in a localized plastic area, i.e., a fracture
initiation mechanism; and (3) the kinetics of molecular flow and bond rupture
of the overstressed polymer chains. The fracture mechanics approach assumes
the existence of defects from the start and develops a theory about the kinetic
crack growth, i.e., the fracture initiation process is neglected. In the previous ap-
proach, the stored energy in the material, i.e., the energy stored by all springs in
the viscoelastic model, can be compared with the energy necessary to stretch the
polymer chains and promote their bond rupture. In fact, it is possible to visualize
the polymer chains as linear springs acting as energy accumulators. Nevertheless,
these energy accumulators have a limited capacity above which bond rupture oc-
curs. Therefore the stored energy limit, called critical energy, can be related to the
energy involved in all microscale bond ruptures that lead to creep-rupture. Most
probably this critical energy depends on the internal state. In reality, there are ex-
perimental indications that the critical energy is temperature- and strain-rate- de-
pendent [14], at least for temperatures lower than the glass transition temperature
Tg (or for shorter times). This means that the R–W criterion is not universal. On
the other hand, there is experimental evidence, for polymers and composite poly-
mers, that change in the fracture mode is a result of change in critical energy with
temperature and strain rate [14]. Finally, it is not difficult to accept that creep-
rupture is strongly related to creep compliance or relaxation modulus. This rela-
tionship emerges naturally from theoretical approaches like fracture mechanics
and energy criteria. Furthermore, creep-rupture and relaxation modulus variations

over time, measured experimentally, resemble one another in an extraordinary
manner. Most probably this signifies that a change in the relaxation modulus cor-
responds to a change in the strength.
The rate theory of fracture is based on a molecular approach, i.e., on the kinet-
ics of molecular flow and the bond rupture of the polymer chains. Drawing on
these approaches Zhurkov [15] presented one of the first models for predicting a
material’s lifetime tf (except for very small stresses) in terms of a constant stress
level s,
t f = t0 exp (U 0 − γσ ) kT  , (6.12)
where t0 is a constant on the order of the molecular oscillation period of 10-13s, k

is the Boltzmann constant, T is the absolute temperature, U0 is a constant for each
material regardless its structure and treatment, and g depends on the previous
treatments of the material and varies over a wide range for different materials.
Griffith et al. [16] applied a modified version of the Zurkov equation:
U 
t f = t0 exp  0 − γσ  , (6.13)
 kT 
known as the modified rate equation, for predicting the time to rupture of continu-
ous fiber-reinforced plastics with reasonable success.
The fracture mechanics analysis was extended to viscoelastic media to predict
the time-dependent growth of flaws or cracks. Several authors produced extensive
work in this area [11,12,17–20]. Schapery [19,20] developed a theory of crack
growth that was used to predict the crack speed and lifetime for an elastomer
under uniaxial and biaxial stress states. For a centrally cracked viscoelastic plate
with a creep compliance given by eq. (6.5) under constant load, Schapery [11]
deduced, after some simplifications, a simple relationship between stress and time
to failure:
 tf  −2 (1+1 n )
  = ( Bσ 0 ) , (6.14)
τ0 
where n is the exponent of the creep compliance power law and B a parameter
that depends on the geometry and properties of the material. Leon and Weitsman
[21] and Corum et al. [22] used this approach where B was considered an experi-
mental constant, to fit creep-rupture data with considerable success. Christensen
[23] developed a kinetic crack formulation to predict the creep rupture lifetime
for polymers. The lifetime was determined from the time required for an initial
crack to grow sufficiently large to cause failure. The method assumed quasi-static
conditions and applies only to the central crack problem. The polymeric material
was taken to be in the glassy elastic state, as would be normal in most applica-
tions. For general stress, s ( t ) we have:
(ατ 0 )
1 − (σ ( t ) σ R ) (σ (τ ) σ )
1m tf 1 m +1
=∫ R dτ , (6.15)
0
where γ = σ 0 2 σ R 2 , m is the exponent of the power law relaxation function and a

is a parameter governed by the geometry and viscoelastic properties. For constant
stress, s = s 0 , the lifetime is given as:
 tf  α  1 
 =  1 m − 1 . (6.16)
τ0  γ  γ 

A classical approach for considering the degradation of mechanical properties

is provided by the method of continuum damage mechanics (CDM). Following
the original ideas of Kachanov [24], the net stress, defined as the remaining load-
bearing cross-section of the material is given [25] by:
σ
σ = , (6.17)
1− ω
where 0 ≤ ω ≤ 1 is the damage variable. At rupture, no load-bearing area remains,
and the net stress tends to infinity when ω → 1.
Kachanov [24] assumes the following damage growth law:
ν
 σ (t ) 
ω ( t ) = C   , (6.18)
 1 − ω (t ) 
where C and ν are material constants. This equation leads to a separable differ-
ential equation for ω ( t ) , assuming ω(0) = 0
(1 − ω ( t ) ) ω ( t ) = Cσ ν ( t ) ⇒ 1 − (1 − ω ( t ) )
ν 1+ν t
= C (1 + ν ) ∫ σ ν (τ ) dτ .
0
(6.19)
The damage growth law is given as:
ω ( t ) = 1 − 1 − C (1 + ν ) ∫ σ ν (τ ) dτ  . (6.20)
t 1+ν

 0 
Assuming failure when ω = 1, then the following expression is obtained:
t
C (1 + ν ) ∫ σ ν (τ ) dτ = 1 . (6.21)
0
From the previous relationship, the time to failure for creep is readily obtained,
assuming s ( t ) = s 0 ,
1
tc = . (6.22)
C (1 + ν ) σ 0ν
Clearly, this result is equivalent to the one obtained previously by using the
Schapery theory [11]. Therefore, the creep lifetime expressions obtained for both
theoretical approaches are directly comparable and are, in fact, equivalent, even
though the parameters have distinct physical interpretations.
Damage evolution does depend strongly on a number of different factors acting
simultaneously, i.e., temperature, moisture, stress, viscoelasticity, viscoplasticity,
etc. Each of these factors is time-dependent. In practice, the influence of any one
on long-term failure is measured independently, i.e., under constant conditions.
Hence, a further methodology is needed to account for their combined effects.
One crucial question remaining to be solved completely is how to predict dam-
age accumulation, or the remaining strength, after a fatigue or creep cycle at mul-
tiple stress levels, based on the fatigue and creep master curves. Miner’s Rule
[26] is an example of a simple way to account for damage accumulation due to
different fatigue cycles. This damage fraction rule is also designated as the Linear
Cumulative Damage law (LCD). For fatigue, it states that failure occurs when the
following condition is satisfied:
N ∆ni (s i )
∑ n (s )
i
= 1 , (6.23)
f i
where n f is the number of cycles to failure at stress level s i and ∆ni is the num-
ber of cycles applied at each stress level s i of the fatigue cycle. Hence eq. (6.23)
provides a failure criterion for fatigue. The corresponding form for creep condi-
tions is given by:
N ∆ti (s i )
∑ t (s )
i
= 1 , (6.24)
c i
where tc is the creep rupture lifetime at stress level s i and ∆ti is the time ap-
plied at each stress level s i . Once more, equations (6.23) and (6.24) specify a
criterion for a lifetime at multiple stress levels. Later, Bowman and Barker [27]
suggested a combination of both damage fraction rules to analyze experimental
data for thermoplastics tested until failure under a trapezoidal loading profile,
which combined fatigue with creep. Although the Miner’s rule can predict accu-
rately the failure of fiber-reinforced polymers under certain combined stress lev-
els, it proved to be inadequate in many other cases. However, due to its simplicity,
it is still used in engineering design.
A cumulative damage theory developed to address various applied problems in
which time, temperature, and cyclic loading are given explicitly, was developed
by Reifsnider et al. [28–30].
The basic form of the strength evolution equation calculates the remaining
strength Fr
z  σ (τ )   j −1
Fr = 1 − ∫ 1 − Fa    jτ dτ , (6.25)
 X (τ )  
0 

where z = t τ , t is the time variable, τ is a characteristic time associated with the

process, Fa is the normalized failure function that applies to a specific controlling
failure mode, and j is a material parameter. This material parameter does influence
damage progression. If j < 1, the rate of degradation is greatest at the beginning,
but if j > 1, the rate of degradation increases as a function of time. However, if
j = 1, there is no explicit time dependence in the rate of degradation. The failure
criterion is given by Fr = Fa . This approach has been used successfully for more
than 20 years by Reifsnider et al. [28–30] to solve various applied problems where
time, temperature, and cyclic loading are explicit influences.
6.3 Short and Long-Term Static Failure of

Composite Pipes
6.3.1 Damage modeling
The application of damage mechanics enables one to model the matrix degra-
dation resulting from cracking in filament-wound pipes [31]. This damage is quan-
tified in terms of crack density, i.e., the reciprocal of crack spacing. The model
used by Roberts et al. [32], based on the previous work of Zang and Gudmundson
[33], applies to microcracks with crack surfaces parallel to the fiber direction and
perpendicular to the lamina plane. In the end, the stiffness matrix of each ply is
obtained by subtracting the damage tensor from the ply stiffness as:
Qij total = Qij elastic − Qij damage , (6.26)

Short and Long-Term Static Failure of Composite Pipes 207
 E1 E2 
1 − ν ν ν 12 0 
1 − ν 12ν 21
 12 21 
 E2 
where Qij elastic = 0 
 1 −ν 12ν 21 
 Sym. G12 
 
 
  E1 
2
 E1   E2  
 β 2 ν 12  β 2 ν 12   0 
  1 −ν 12ν 21   1 −ν 12ν 21   1 −ν 12ν 21  
 2 
damage   E2  
and Qij =ρ β2   0 
  1 −ν 12ν 21  
 Sym. β1G12 2 
 
 
 
where E1 is the longitudinal modulus (in the fiber direction), E2 is the transverse
modulus (perpendicular to the fiber direction), ν 12 is the major in-plane Poisson’s
ratio (i.e., the Poisson’s ratio that corresponds to a contraction in the transverse
direction when an extension is applied in the longitudinal direction and is related
to the minor in-plane Poisson’s ratio by ν 21 E2 = ν 12 E1 ), and G12 is the in-plane
shear modulus. The dimensionless or normalized crack density in the ply is de-
fined by the ratio of ply thickness and crack spacing ρ = t s . The coefficient β1
relates to crack-face displacement in Mode III anti-plane strain, and the coefficient
β 2 relates to a Mode I crack opening. These two coefficients are expressed by:
2 π 
β1 = ln cosh  ρ 
π G12 ρ 2
2 
, (6.27)
π  1 −ν 12ν 21  10 an
β2 =  ∑
2  E2  n =1 (1 + ρ )n
where the constants an are tabulated in [32]. For a ply with a free surface, β1 is
2.0 times, and β 2 2.51 times, their corresponding values [34].
The crack density effect on elastic constants is shown in Figure 6.2, with the
results normalized by its initial undamaged value, which is obtained from classical
laminated plate theory (CLPT) coupled with equation (22.6). The main conclu-
sion for angle-ply laminates is that the axial modulus decreases most rapidly with
increasing crack density. Frost [31] suggested an approximation using a linear
1.20
Normalized change in elastic constant
1.00 Poisson's ratio
0.80 Shear modulus
0.60
Hoop modulus
0.40
Axial modulus
0.20
0.00
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Normalized crack density
figure 6.2 Normalized change in elastic constants versus crack density for ±55º pipe
laminate, calculated from the model of Roberts at al. [32] (lines) and using the linear ap-
proximation given by Frost [31] (points).
relationship for elastic properties that decay with crack normalized density. The
simplified approach proposed by Frost [31] captures the initial changes in elastic
properties, as depicted in Figure 6.2. The observed differences arise from the fact
that the elastic material properties are not exactly the same. Hence, this simplified
approach does predict relevant decay in the pipe’s elastic properties with good
accuracy for the important crack density range prior to failure.
The relation between the crack density and the applied stress, for each ply, was
approximated as:
1+ k 2ρ 2 = f (σ 2 ,τ 12 ) [31], (6.28)
( E1 + E2 ) G12 LIVE GRAPH

with k2 = Click here to view
E1 E2
2 2
 σ2   τ 12  σ2 τ 12
and f (σ 2 ,τ 12 ) =   +   −
σ
 2, failure   τ 12, failure  σ τ
2 , failure 12 , failure
where s 2 and τ 12 represent the ply transverse and shear stresses and s 2, failure ,
and τ 12, failure the respective failure stresses. The function in (6.24) gives the nor-
malized crack density as a function of a second-order polynomial that depends
on both the shear and transverse stresses, which are the stresses contributing to
crack formation. Since the stiffness in (6.22) depends on crack density, a nonlin-
ear stress–strain relation is obtained. This was applied by Roberts et al. [32] to the
deformation of internally pressurized pipes with promising results.
The methodology is easy to implement, and some calculations were performed
using data for E-glass/epoxy ±55º filament-wound pipes provided by Soden et al.
[35–36]. In Figures 6.3 and 6.4, predictions and experimental data are compared
for stress-strain response under two different modes: hydrostatic pressure with
closed and open ends. The theory provides a good approximation to the stress-
strain experimental curves.
In general, the failure parameter is given as:
f (σ 2 ,τ 12 ) = C , (6.29)
For short-term and non-aged samples, C = 1 . In cases of long-term analysis,

the failure parameter becomes dependent on time, static or cyclic loading and
temperature:
C = Acreep ATemperature Afatigue , (6.30)
LIVE GRAPH
Click here to view
300
250 Experimental
Axial strain
Hoop Stress (MPa9
Hoop strain
200 Calculated
150
100
σhoop/σaxial=2
50
0
0.000 0.005 0.010 0.015 0.020
Strain
figure 6.3 Comparison between experimental data and predictions of the stress-strain
behavior of an E-glass/epoxy ±55º filament-wound pipe under hydrostatic pressure with
closed ends.
LIVE GRAPH
Click here to view
300
250
Hoop strain
Hoop Stress (MPa9
Axial strain
200
150
Experimental
100 Calculated
σaxial/σhoop=0
50
0
-0.050 -0.030 -0.010 0.010 0.030 0.050
Strain
figure 6.4 Comparison between experimental data and predictions of the stress-strain
behavior of an E-glass/epoxy ±55º filament-wound pipe under hydrostatic pressure with
open ends.
According to Frost [31], for long-term sustained loading the factor Acreep can be
set to 0.5 for 20-year conservative design. Using this assumption and the previous
theoretical approach, it is possible to predict the failure envelope for short-term
and long-term cases under different loading modes, as shown in Figure 6.5, for
an E-glass/epoxy ±55º filament-wound pipe. Failure was assumed to occur when
the normalized crack density reaches the value of 0.5 (2.2mm for a ply thickness
of 0.25mm).
LIVE GRAPH
Click here to view
180
160
140
Axial Stress (MPa)
120
100
80
60
40 Short-term prediction
20 Long-term prediction
0
0 50 100 150 200 250 300 350
Hoop Stress (MPa)
figure 6.5 Predicted failure envelope for short-term and long-term sustained loadings on
an E-glass/epoxy ±55º filament-wound pipe.
In general, the failure mechanism of most glass-reinforced plastics, like fil-

ament-wound pipes, does not change from short to long term. Based on this as-
sumption and using the Paris law, Frost [31] showed that the creep factor may be
given as:
1
Acreep = , (6.31)
ta
where a is equal to the time exponent obtained by regression of the creep failure
curves.
For cyclic loads exceeding 7000 cycles over the design life, fatigue effects
must be considered. These effects are more severe than creep effects. The stan-
dard ISO 14692-3:2002 proposes the following factor:
    N − 7000  
Afatigue =  R 2 + (1 − R 2 )  exp (1 − R 2 ) 1 −
1
 , (6.32)
108  

 16   
where R is the ratio between the minimum and maximum loads (or stresses) of
the load (or stress) cycle and N is the total number of cycles during service life.
For illustration purposes an example of static (creep) and cyclic pipe test re-
sults at 65ºC can be cited from [37]. In Figure 6.6, the creep failures curves are
shown, and in Figure 6.7 the experimental data and the predicted curves are rep-
resented. The fatigue curves were predicted from creep failure curves using the
factor Afatigue from ISO 14692-3:2002, with R = 0. The predictions are very close
to experimental data.
300
Water, 65ºC
250
Hoop Stress (MPa)
200
y = 445.7633x-0.0624
150 R2 = 0.9970
100 Experimental
Curve Fit
50
0
1.E+00 1.E+02 1.E+04 1.E+06 1.E+08 1.E+10
Time to Failure (s)
figure 6.6 Experimental creep failure of pipes at 65ºC.

LIVE GRAPH
Click here to view
300
Water, 65ºC
250
Hoop Stress (MPa)
Experimental
200
150
ISO 14692
100
50
0
1.E+00 1.E+02 1.E+04 1.E+06 1.E+08 1.E+10
Cycles to Failure
figure 6.7 Experimental and predictions using creep data (ISO) for fatigue failure of
pipes at 65ºC.
Hale et al. [38] provide a good account of hygrothermal effects on filament-

wound glass fiber-reinforced pipes that were subjected to prolonged exposure to
high- temperature water. Two materials systems in common use in offshore piping
applications were used, ±55º filament-wound E-glass/phenolic and ±55º filament-
wound E-glass/epoxy pipes. The loading modes were hydrostatic pressure with
closed and open ends and axial loading. The last mode was produced by applying
the same pressure to the interior and exterior of the pipe to eliminate the hoop
stress.
Biaxial failure envelopes were obtained for both types of pipe materials in
the water-saturated state at a range of temperatures from 20º to 160ºC. It was ob-
served that the strength of the epoxy pipes was significantly reduced as tempera-
ture increases, especially in the matrix-dominated modes close to pure hoop and
pure axial loading. In contrast, the phenolic pipes were unaffected by temperature
for all tested modes.
Concerning the application of failure theories to this type of filament-wound
pipe, we must note, as a consequence of the wwfe (world-wide-failure-exercise)
[39], it is recommended that special care be taken wherever large deformations
may be involved. This is especially relevant for ±55º laminates, since none of the
failure theories coped very well with these cases.
6.3.2 Creep rupture

This section offers an overview of long-term creep failure data for internally
pressurized filament-wound pipes. Unfortunately, the available experimental data
on creep and creep failure of composite pipes is quite limited.
Mieras [40] described the creep behavior of internally pressurized filament-

wound glass-reinforced epoxy and polyester pipes. It was shown how important
the resin is for the long-term behavior of the pipes. Creep occurred at all stress
levels down to 40 MPa. At higher stresses, creep leads to weeping. In Figure 6.8,
long-term weeping stresses are plotted.
The failure mechanism for weepage in filament-wound pipes was described in
a great detail by Jones and Hull [41] for two distinct modes, hydrostatic pressure
with open and closed ends. The interlaminar cracks associated with transverse
matrix cracking tend to grow and intersect, providing paths through the pipe wall
for weeping to occur.
Since creep loading at higher stress levels leads to weeping and eventually to
total failure, it must be concluded that creep does promote damage comparable to
the short-term cracking that was observed for the UEWS (Ultimate Elastic Wall
Stress) tests [40] as depicted in Figure 6.9.
Nevertheless, the results reported by Mieras [40] on hoop stress/hoop strain
curves were not observed by others, especially by Hull et al. [42]. For mode 2 (hy-
drostatic pressure with closed ends) the sharp bend in the hoop stress/hoop strain
curve was not observed. Hull et al. [42] advanced an explanation, suggesting the
phenomenon originated from the step-loading used by Mieras [40].
The UEWS test is used to identify a stress level beyond which permanent de-
formations are obtained and the material creep rate increases substantially. The
UEWS test applies increasing pressure levels based on groups of 10 one-minute
hydrostatic pressure cycles [40,43–45]. The procedure consists of performing a
series of loading cycles in successive increasingly higher loading steps (10 equal
cycles at each level), while measuring the maximum hoop or axial strain obtained

\ [
5
+RRS6WUHVV03D

\ [ (SR[\
5 3RO\HVWHU

7LPHWR)DLOXUHKRXU
figure 6.8 Long-term weeping stresses in epoxy and polyester filament-wound pipes (af-
ter Mieras [40]).
LIVE GRAPH
Click here to view
120
100
Hoop Stress (MPa)
80
60
after 1 min
40
after 1000 hour
20 after UEWS test
0
0 0.002 0.004 0.006 0.008 0.01
Hoop Strain
figure 6.9 Isochronous stress/strain curves for epoxy pipes compared with that of the pipe
after Ultimate Elastic Wall Stress (UEWS) tests.
LIVE GRAPH
Click here to view
in the first and tenth cycle of each step. The difference between these two strains
is used to verify whether elastic limits have been reached. Initially, the pipe speci-
mens are loaded using a hydrostatic pressure up to 10% of the expected UEWS
level with the hydrostatic pressure maintained for 1 minute. Then the hydrostatic
pressure is released, and the pipe remains unloaded for another minute. This load-
ing-unloading pattern is repeated 10 times with the strains being measured at the
end of the first and tenth one-minute load cycle. Subsequently, similar 10-cycle
series are successively repeated with the hydrostatic pressure in each step being
increased by 10% of the expected UEWS hydrostatic pressure. The test is sched-
uled to continue until a certain level of permanent deformation is reached (5%).
The UEWS test is a simple procedure that takes a few hours to perform. As
Gibson et al. [37] maintain, the UEWS can become an accepted alternative for
re-confirming pipe qualification whenever a minor product change is made. The
actual procedure for re-confirmation also mandates survival tests with samples
held under hydrostatic pressure for 1000 h. Survival in this test leads to a conclu-
sion: the samples tested are at least as good as those originally qualified.
Ghorbel et al. [3, 46] investigated the creep and damage of filament-wound
pipes reinforced with E-glass fibers wound at ±55º for two different resin systems:
polyester and vinylester. The response of the preconditioned pipes in water at
60ºC under sustained hydrostatic pressure with closed-ends was nonlinear visco-
elastic. It was observed that creep under hygrothermal conditions induces inter-
laminar cracking. The time-dependent failure under sustained loading followed a
trend similar to that depicted in Figures 6.10 and 6.11.
LIVE GRAPH
Click here to view
15.00
E glass/polyester
Water at 60ºC
Pressure (MPa)
10.00
5.00
y = 7.90603x-0.07181
R2 = 0.68232
0.00
0.01 0.1 1 10 100 1000 10000
Time to Failure (hour)
figure 6.10 Long-term creep failure of E glass/polymers pipes wound at ±55ºC immersed
in water at 60ºC [3].
20.00
ECR glass/vinylester
15.00 Water at 60ºC

Pressure (MPa)
10.00
y = 16.40839x-0.04707
5.00 R2 = 0.99351
0.00
0.01 0.1 1 10 100 1000 10000
figure 6.11 Long-term creep failure of E glass/vinylester pipes wound at ±55ºC immersed
in water at 60ºC [3].
LIVE GRAPH
Click here to view
It was concluded that the failure mechanisms and time-dependent failure de-
pend strongly on the resin, as can be observed in Figures 6.10–6.11. In contrast
to high pressure levels, at low pressure levels the effect of environment must be
assessed, since it significantly affects the time-dependent failure.
The determination of the long-term resistance to internal pressure, based on

the standard EN1447, which is accomplished by imposing hydrostatic internal
pressure on the specimens using axial constrained-free-end sealing devices, was
described by Faria and Guedes [47–48]. Figure 6.12 depicts the testing apparatus.
GFRP pipes of three different construction types, from major European manu-
facturers, were used. The properties of the series of specimens used are displayed
in Table 6.1.
Figures 6.13 to 6.16 depict the creep failure curves for each pipe type. It is clear
that the pipes’ construction and material composition have a strong influence on their
long-term behavior, especially when comparing a matrix of epoxy resin and polyester
resin (UP). Despite considerable scatter in the tests results, which is usual, statistical
regression analysis determined no relevant differences between using complete data
(including tests up to 10,000 h) or using only data from shorter tests (up to 1000 h).
These long-term hydrostatic tests will be used later in this chapter in a discussion
of pipe qualification.
Figure 6.12 Apparatus used in the EN1447 test procedures.

Table 6.1
GFRP pipe specimens used in the experimental test series.
*)53SLSHW\SH
$ % & '
+\EULG±
ILODPHQW
ILODPHQW ILODPHQWZLQGLQJFHQWULIXJDO
&RQVWUXFWLRQ7\SH ZLQGLQJ
ZLQGLQJ83 PDW FDVWLQJ83
HSR[\
GHSRVLWLRQ83
1RPLQDO'LDPHWHUPP

1RPLQDO6WLIIQHVV1P
1RPLQDO3UHVVXUHEDU
6SHFLPHQ/HQJWKPP
LIVE GRAPH
Click here to view
9.0
8.0 Pipe type A
Internal Pressure (MPa)
7.0
6.0
5.0
4.0
3.0 y = 5.786961x-0.006811
2.0 R² = 0.571338
1.0
0.0
0.1 1 10 100 1000 10000 100000
Figure 6.13 Creep failure results for type A pipe.
LIVE GRAPH
Click here to view
9.0
8.0 Pipe type B
7.0
6.0
5.0
4.0
3.0 y = 7.828656x-0.035688
2.0 R² = 0.741599
1.0
0.0
0.1 1 10 100 1000 10000 100000
figure 6.14 Creep failure results for type B pipes.

LIVE GRAPH
Click here to view
9.0
Internal Pressure (MPa) 8.0 Pipe type C
7.0
6.0
y = 3.785206x-0.059767
5.0 R² = 0.425245
4.0
3.0
2.0
1.0
0.0
0.1 1 10 100 1000 10000 100000
Figure 6.15 Creep failure results for type C pipes.
LIVE GRAPH
Click here to view
9.0
8.0 Pipe type D
7.0
6.0
y = 3.675653x-0.005417
5.0 R² = 0.125838
4.0
3.0
2.0
1.0
0.0
0.1 1 10 100 1000 10000 100000
figure 6.16 Creep failure results for type D pipes.
6.3.3 An example of preliminary design for the long-term

The netting analysis is useful for preliminary calculation of filament-wound
pipes under hydrostatic pressure. These structures are primarily loaded in mem-
brane, and that may be considered the simplest case for design. The netting analy-
sis assumes that the stresses induced in the structure are borne entirely by the fi-
bers, and the strength of the resin is neglected. Moreover, it assumes that the fibers
possess no bending or shearing stiffness, and carry only the axial tensile loads.
The netting analysis is described in detail in [49–50]. Evans and Gibson [51]
obtained analytical expressions for the stress–strain relations from classical
laminated plate theory (CLPT) to show the discrepancy between CLPT and the
netting theory and reveal significant factors for design. Verchery’s [52] approach
to the same problem proved to be more effective for deriving explicit design for-
mulas applicable to any state of stress and stacking sequence.
As an example of long-term preliminary analysis, let us suppose we have a
±55º glass/epoxy filament wound pipe with an internal diameter of 100mm and
a wall thickness of 5mm, sustaining a hydrostatic pressure of 0.6 MPa. The hoop
stress is
Pi D 0.6 ⋅100
s hoop = = = 6MPa , (6.33)
2t 2⋅5
and axial stress is s axial = 0.5 ⋅ s hoop

The netting analysis for this condition, with an optimal winding angle of
arctan 2 ≅ 54.74º , assumes that the stress normal to the fibers is null and the
stress in the fibers direction is calculated as [53]:
1
sf = s hoop ≅ 9.0MPa , (6.34)
sin 2 ( 54.74º )
Now we can calculate the fiber strength, which remains after degradation on a
long-term basis [53], i.e., the allowable fiber design strength for the pipe as:
σ α = σ v Pt Psc Ps , (6.35)
where σ α is the virgin fiber strength, Pt is the thermal degradation factor, Psc is
the stress concentration factor and Ps is the factor for long-term static loading.
Let us assume that for the glass fiber we have σ α = 157MPa . The strength reduc-
tion due to thermal degradation is assumed as Pt = 0.8 . After the fiber is placed
in the strand, the strength tensile reduction is about 25%, i.e., Psc = 0.75 , due to
localized stress concentrations, fiber crossovers and residual stress, among others.
Since glass fibers are particularly susceptible to static fatigue effects (creep), in
many cases the strength reduction due to long-term static loading is very high,
i.e., Ps ∈ [ 0.1, 0.2] [53]. Assuming in this case Ps = 0.1 , the allowable fiber design
strength becomes:
σ α = 157 ( 0.8 )( 0.75 )( 0.1) ≅ 9.4MPa , (6.36)
This residual strength is slightly larger than the fiber stress calculated from the
netting analysis, which indicates a reasonable long-term preliminary design.
6.4 Lifetime of Composites Pipes Under Cyclic

Loading
Polymers and polymer-based composites fail, given enough time, when sub-
mitted to cyclic loads at stresses well below their static failure loads. This phe-
nomenon is called fatigue. The typical approach to fatigue is to develop fatigue
curves, i.e., applied stress (S) against the number of cycles until failure (N). These
curves are usually designated as S-N curves and resemble a sigmoid function,
displaying a stress limit for large cycles, suggesting a fatigue limit. Moreover the
S-N curves for polymers and polymer based composites are extremely frequency
dependent.
As many research works have shown, cyclic loading can significantly degrade
the stiffness and strength of pipes. Frost [54] and Frost and Cervenka [43] ob-
tained the S-N (fatigue) curves of ±55º glass-fiber/epoxy filament wound pipes
with 100mm internal diameter (14 MPa rated pressure). The curves were fitted
using the power law expressed as:
 t 
log( P) = A − B log   (6.37)
τ0 
where P is the pressure (or hoop stress or axial stress), t the time, and τ 0 the
reference time. The curves are extrapolated for 20 years to define the long-term
hydrostatic pressure (LTHP). The resulting value is multiplied by a number of
factors (typically 0.5 [43]), resulting in a pressure known as the hydrostatic design
basis (HDB) [44].
Tarakcioglu et al. [55] tested ±55º glass-fiber/epoxy filament-wound pipes un-
der internal pressure. The pipes had four layers with 1.6 mm in thickness, 300mm
in length and an inner diameter of 72mm. Stress levels were 30% (121.5 MPa),
35% (141.7 MPa), 40% (162 MPa), 50% (202.5 MPa), 60% (243 MPa) and 70%
(283.5 MPa) of the static strength of the specimen (405 MPa). Sinusoidal stress
levels were applied at 0.42 Hz for R = 0.05 stress ratio (R = Maximum stress/
Minimum stress). Fatigue results were recorded for three different damage stages,
namely, whitening, leakage and final failure. The three stages of whitening (fiber/
matrix interface debonding and delamination), leakage initiation, and final failure
occurred sequentially. Micrographs from an SEM observation proved that there
is an analogy between the macro-damage stages and the micro-damage mecha-
nisms. It is possible to fit equation (6.37) to the three damage stages shown in
Figures 6.17–6.19 with a very good correlation. For all stress levels, ultimate
failure occurred almost immediately after leakage initiation.
Another study of E-glass/epoxy filament-wound pipes of four layers with a
±75º winding angle, using the same test conditions, led to similar conclusions
[56]. In both cases no evidence of a fatigue limit was found.
LIVE GRAPH
Click here to view
Lifetime of Composites Pipes Under Cyclic Loading 221
350
300
Hoop Stress (MPa)
250
200
150
100 y = 314.22x-0.1029
50 R2 = 0.9577
0
1 10 100 1000 10000 100000 1000000
Cycles to Whitening Initiation
figure 6.17 S–N curve for whitening initiation of ±55º glass-fiber/epoxy filament-wound
pipes.
LIVE GRAPH
Click here to view
300
250
Hoop Stress (MPa)
200
150
100
y = 620.51x-0.1555
50 R2 = 0.9694
0
1 10 100 1000 10000 100000 1000000
Cycles to Leakage Initiation
figure 6.18 S–N curve for leakage initiation of ±55º glass-fiber/epoxy filament-wound
pipes.
Fatigue under biaxial loading was obtained by Perreux and Joseph [57] for
E-glass/epoxy pipes with four layers with ±55º winding angle. The pipes had a
diameter of 60mm and a length of 2800mm and were cut into 350mm lengths.
The fatigue tests were done under three different modes: hydrostatic pressure with
closed and free ends and axial tensile loading.
LIVE GRAPH
Click here to view
300
250
Hoop Stress (MPa)
200
150
100
y = 625.14x-0.1561
50 R2 = 0.9698
0
1 10 100 1000 10000 100000 1000000
Cycles to Failure
figure 6.19 S–N curve for final failure of ±55º glass-fiber/epoxy filament-wound pipes.
LIVE GRAPH
The results are summarized in Figures 6.20–6.22. Click here to view
700
600 Internal Pressure with free ends
Hoop Stress (MPa)
500
400
300
y = 585.58x-0.1622
200 R2 = 0.9402
100
0
0.1 1 10 100 1000 10000 100000 1000000
Cycles to Failure
figure 6.20 S–N curve for final failure of ±55º E-glass/epoxy filament-wound pipes under
hydrostatic pressure with free ends for a frequency of 0.2 Hz.
The previous results can be put in the form of an isonumber of cycles to failure,
depicted in Figure 6.23. This provides an approximate idea of the evolution of
failure envelopes with cycle fatigue. Hence, in this case, the failure envelope for
106 cycles can be obtained from the static failure envelope by applying a scaling
factor of 0.1.
Lifetime of Composites Pipes Under Cyclic Loading 223
700
600 Internal Pressure with closed ends
Hoop Stress (MPa)
500
400
300 y = 590.96x-0.1546
200 R2 = 0.9758
100
0
0.1 1 10 100 1000 10000 100000 1000000
Cycles to Failure
LIVE GRAPH
Click here to view
hydrostatic pressure with closed ends for a frequency of 0.2 Hz.
90
80 Tensile
70
Axial Stress (MPa)
60
50
40 y = 80.894x-0.0771
30 R2 = 0.888
20
10
0
0.1 1 10 100 1000 10000 100000
Cycles to Failure
LIVE GRAPH
Click here to view
axial tensile load for a frequency of 0.2 Hz.
The effect of frequency on fatigue lifetime was also studied by Perreux and
Joseph [57]. It was concluded that fatigue lifetime does increase with frequency
from 0.2 to 1Hz, i.e., 2~3 fold higher. Later, Perreux and Thiebaud [58] concluded
there are two concurrent phenomena that influence fatigue failure: (1) interaction
between creep and fatigue at low frequency, which increases the lifetime when the
350
Scaling factor Static
300
Axial Stress (MPa)
250
200
150
102 cycles
100
50 104
106
0
0 100 200 300 400 500 600 700
Hoop Stress (MPa)
figure 6.23 Failure envelopes for ±55º E-glass/epoxy filament-wound pipes under biaxial
tensile load, for multiple fatigue cycles.
LIVE GRAPH
Click here to view
frequency is increased, and (2) a thermal effect due to viscoplastic dissipation at
a higher frequency, which reduces the lifetime when the frequency is increased.
Thus, all situations can be observed, depending on which phenomena are most
intense. It should be noted that there is no effect of the frequency on the lifetime if
the phenomena have similar intensity. However, this result shows that a material’s
lifetime as measured in fatigue can depend on the shape of the specimen as well
as on thermal dissipation.
Ellyin and Martens [59] investigated the multi-axial fatigue of pipes, showing
that there are two stages in failure: a functional one and final structure rupture.
Kaynak and Mat [60] studied tensile fatigue and showed that damage has three
stages: crack initiation, crack growth and concentration along the fibers’ direction,
and fiber failure. They also studied the effect of frequency on fatigue lifetime
and concluded, as Joseph and Perreux [57] had, that the tensile fatigue lifetime
increases with frequency.
In conclusion, the effect of cycling loading on damage initiation and propaga-
tion is more severe than sustained loading, i.e., creep loading. The UEWS tests,
previously mentioned and described, are strongly linked to fatigue tests as was
demonstrated by Gibson et al. [37]. A simple simulation of a UEWS test can be
done, using creep and fatigue failure curves from [37], to compute the cumulative
damage that originated separately from creep and fatigue. This was accounted for
using Miner’s law as suggested in [37]. The calculations are depicted in Figure
6.24. It is quite obvious that in the UEWS test the fatigue effect on damage is
much more pronounced.
Applicable Standards 225
1.0
UEWS
0.8
Damage factor
0.6 Fatigue (Miner's Law)

Creep (Miner's Law)
0.4
0.2
0.0
0 50 100 150 200 250 300
Hoop Stress (MPa)
figure 6.24 Computed Miner’s law sum, for creep and fatigue, at each hoop stress level
in the UEWS test.
6.5 Applicable Standards LIVE GRAPH

Click here to view
6.5.1 Identification and comparison of main standards
Among the increasing range of applications where glass-fiber-reinforced plas-
tic (GRP) pipes are used, oil/gas and sewage transportation are the most relevant.
For the former, internal pressure is considered to be the representative and gov-
erning loading condition. Therefore, the product and test standards specify de-
sign and qualification criteria based on this loading case. Sewage pipe are treated
mostly as a low-pressure application, and thus reduced qualification criteria may
be adopted. Also, whether the pipe is above‑ or underground determines the type
of external supports and the conditions to which the pipes are subjected during
their in‑service life time. This also can lead to changes in design and qualification
procedures.
Despite the fact that highly specific requirements may be needed for each in-
dividual piping application, the major standardization organizations worldwide
have developed product and test standards that strive to cover the most com-
mon applications. Namely, ASTM International (formerly known as the American
Society for Testing and Materials), AWWA—American Water Works Association,
CEN—the European Committee for Standardization and ISO—International
Organization for Standardization have all published multiple standards compris-
ing guidelines for both manufacturers and end-users of GRP piping systems.
During the last 25 years, such normative references have evolved from the initial,
empirically-based conservative approaches to a combination of experimental and
statistical analyses, which nonetheless are also empirical. Hence, these standards
have established a baseline for the design, manufacture, qualification, installation

and operation of GRP piping systems (i.e., pipes, fittings and related accessories).
In Table 6.2, a comparison of the required/preferred procedures of the main
GRP pipe product standards is presented. The standards for the product-types
mentioned below [54–59] are compared:
• ASTM D3262—Standard Specification for “Fiberglass” (Glass-Fiber-

Reinforced Thermosetting-Resin) Sewer Pipe.
Reinforced Thermosetting-Resin) Pressure Pipe.
Reinforced Thermosetting-Resin) Sewer and Industrial Pressure Pipe.
• ASTM D2997—Standard Specification for Centrifugally Cast “Fiberglass”
(Glass-Fiber-Reinforced Thermosetting-Resin) Pipe.
• AWWA C950—Standard for Fiberglass Pressure Pipe.
• AWWA M45—Fiberglass Pipe Design Manual.
• EN 1796—Plastics piping systems for water supply with or without pres-
sure—Glass‑reinforced thermosetting plastics (GRP) based on unsaturated
polyester resin (UP).
• EN 14364—Plastics piping systems for drainage and sewerage with or with-
out pressure—Glass‑reinforced thermosetting plastics (GRP) based on un-
saturated polyester resin (UP)—Specifications for pipes, fittings and joints.
• ISO 10467—Plastics piping systems for pressure and non‑pressure drain-
age and sewerage—Glass‑reinforced thermosetting plastics (GRP) systems
based on unsaturated polyester (UP) resin,
• ISO 10639—Plastics piping systems for pressure and non‑pressure water
supply—Glass‑reinforced thermosetting plastics (GRP) systems based on
unsaturated polyester (UP) resin.
• ISO 14692—Petroleum and natural gas industries—Glass-reinforced plas-
tics (GRP) piping.
These standards are cited because they form the current references for industry
and end-users worldwide. Each group of product standards (ASTM, AWWA, EN
and ISO) covers: pressure and non-pressure, aboveground and underground, off-
shore and onshore applications of GRP pipes. Although they are suitable for GRP
pipes manufactured by different processes (filament winding, centrifugal casting,
hand lay‑up etc.), and with different matrix materials (polyester, vinyl‑ester and
epoxy resins), each refers to a specific manufacturing process and material con-
figuration as the basis for the design and qualification methodology stated in it.
ASTM D3517, AWWA C950, EN 1796 and ISO 10639 are alternatively used
mainly for water supply piping systems with in‑service internal pressure loading
conditions. ASTM D3262, ASTM D3754, EN 14364 and ISO 10467 are alterna-
tively used mainly for sewage transportation piping systems where no internal
Table 6.2
Identification and comparison of GRP pipe standards.
Test/Parameter Standard
ASTM ASTM ASTM ASTM AWWA AWWA EN 1796 EN ISO ISO ISO
D3262 D3517 D3754 D2997 C950 M45 14364 10467 10639 14692
Initial circumferen- n/a ASTM ASTM ASTM ASTM ASTM EN 1394 EN 1394 ISO 8521 ISO 8521 ASTM
tial tensile strength D1599 D1599 D1599 D1599 D1599 D1599
(failure pressure)
Long-term circum- n/a ASTM ASTM ASTM ASTM ASTM EN 1447 EN 1447 ISO 7509 ISO 7509 ASTM
ferential tensile D2992 D2992 D2992 D2992 D2992 D2992
strength (failure
pressure)
Cyclic pressure n/a n/a n/a ASTM ASTM ASTM EN 1638 EN 1638 ISO ISO ASTM
strength D2143 D2143 D2143 15306 15306 D2143
Initial specific ring
227
ASTM ASTM ASTM ASTM ASTM ASTM EN 1228 EN 1228 ISO 7685 ISO 7685 ASTM
stiffness D2412 D2412 D2412 D2412 D2412 D2412 D2412
Long-term specific n/a n/a n/a n/a n/a n/a ISO ISO ISO ISO n/a
ring stiffness 10468 10468 10468 10468
Initial ring deflection ASTM ASTM ASTM n/a ASTM ASTM ISO ISO ISO ISO ASTM
D2412 D2412 D2412 D2412 D2412 10466 10466 10466 10466 D2412
Long-term ring n/a n/a n/a n/a n/a n/a ISO ISO ISO ISO n/a
deflection 10471 10471 10471 10471
Longitudinal tensile ASTM ASTM ASTM ASTM ASTM ASTM EN 1393 EN 1393 ISO 8513 ISO 8513 ASTM
strength D638 D638 D638 D2105 D2105 D2105 D2105
Methods for regres- n/a (ASTM (ASTM (ASTM (ASTM (ASTM ISO ISO ISO ISO (ISO
sion analysis of test D2992) D2992) D2992) D2992) D2992) 10928 10928 10928 10928 14692)
data
pressure is considered. ISO 14692 is broadly used in the oil and gas industries,
mainly for off‑shore applications. Since it covers the design and qualification of
suspended pipe systems, AWWA M45 is often complimentarily used for under-
ground pipe systems within the same industries.
The main differences between ISO 14692 and the other ISO and CEN product
standards for pressure piping applications are that:
• the regression qualification procedure is based on a design life of 20 years

(approx. 175,400 hours), instead of 50 years (approx. 438,000 hours);
• the full regression qualification procedure requires only 18 specimens to be
tested, instead of 23 specimens;
• it provides default values for the slope of the regression line for preliminary
design calculations and/or determination of test conditions (in cases where
there is no data);
• it allows a breakdown of a product family into family representatives, prod-
uct sectors, component variants and product sector representatives in order
to reduce the overall quantity and duration of the qualification tests; in this
way, only the family representatives are required to pass through the full
qualification procedure while the other variants have to pass only a 1000 h
survival test (ASTM D1598);
• it establishes a normative procedure to determine the factored failure enve-
lope, considering a biaxial (hoop plus axial) stress state (Figure 6.25);
• the reference test temperature is 65°C, instead of room temperature;
• it establishes a limited (less stringent) qualification procedure for low‑pres-
sure water applications.
As can readily be observed in Table 6.2, the standard ISO 14692 adopted the
ASTM methodology and thus allows for design and qualification procedures
that are less expensive and time‑consuming than other ISO and CEN standards.
Nevertheless, the ISO 14692 procedure retains a conservative quality by: (1) re-
quiring higher test temperatures (thus aging the specimens during the tests) and
(2) using a lower confidence-limit level to determine the long-term hydrostatic
pressure (instead of directly extrapolating from the regression line).
In order to assess the regression line that permits the prediction of the long‑term
hydrostatic pressure or hoop stress level that leads to a lifetime of 20 years (ISO
14692) or 50 years (EN 1796), the critical procedure lies within the standards tests
ASTM D2992 or EN 1447, respectively. These standards establish the procedures
for obtaining the hydrostatic design basis (HDB) or the pressure-design basis
(PDB) for GRP piping products, by evaluating strength‑regression data derived
from pipe or fittings tests. The data obtained from these test methods is plotted as
hoop stress or internal pressure versus time‑to‑failure relationships at the selected
temperatures that simulate actual anticipated product end-use conditions. This
practice defines a hydrostatic design basis (HDB) for material in straight, hollow
Figure 6.25 Idealized envelopes for a single‑wound angle-ply GRP pipe with winding
angles ranging from 45º to 75º. (1) long‑term design envelope, (2) idealized long‑term
envelope, (3) idealized short‑term envelope, (4) schematic representation of the short‑term
failure envelope (after ISO 14692:2002).
cylindrical shapes where hoop stress can be easily calculated, and a pressure-
design basis (PDB) for fittings and joints where stresses are more complex.
ASTM D2992, in particular, includes two test procedures, accounting for two
typical in‑service loading cases (not limited to internal pressure):
• Procedure A—a minimum of 18 specimens of pipe (or fittings) are placed

under cyclic internal pressures at a cycle rate of 25 cycles per minute, at
several different pressures. The stress or pressure values for the test are
selected to obtain a proper distribution of failure points in the time decades
(in log scale). The cyclic long‑term hydrostatic strength (LTHS) of a pipe
is obtained by an extrapolation of a log-log plot of the linear regression line
for hoop stress versus cycles to failure.
• Procedure B—a minimum of 18 specimens of pipe (or fittings) are placed
under constant internal pressures at different pressure levels in a controlled
temperature environment. The time‑to‑failure for each pressure level is
recorded. The stress or pressure values for the test are selected to obtain a
proper distribution of failure points along the time decades (in log scale).
EN 1447 is truly identical to ASTM D2992, differing only in the long‑term

time objective (50 years instead of 20 years) and thereby requiring a greater num-
ber of specimens so that the extrapolation can be done with the same statistical
confidence level. Both standards methods implicitly assume that the mechanisms
responsible for long‑term failure are the same at different levels of load and from
short‑ to long‑term. Although this limitation is not explicitly addressed in the stan-
dards, the experimental evidence of the whole failure phenomenon and the lack
of adequate information were the main reason for extending the test periods to
over 10,000 hours. In a logarithmic time scale, 10,000 hours is only 1.2 and 1.6
decades distant from 20 years and 50 years, respectively. This makes the existing
test and prediction methods seem reasonable.
In the following section a practical case is presented where the regression
analysis procedure for long‑term hydrostatic pressure is directly applied and the
results are discussed.
6.5.2 Long-term qualification tests of four different types of GRP

Two typical in-service load conditions for GRP piping systems are internal pres-
sure and ring deflection. For these, empirical test methods have been developed
and are described in the standards ISO 7509:2000 (equivalent to EN 1447:2009)
and ISO 10471:2003 (replacing EN 1227:1997), respectively.
The lack of completely understood failure mechanisms and long‑term materi-
als performance knowledge necessarily leads to conservative over‑estimation and
consequently to over‑design of the pipe structures. The existence of various types
of GRP pipe construction, namely filament wound, centrifugal cast, and hybrid
ones, also inhibits any relaxation of test specifications.
Experimental results from standard test procedures, according to EN 1447:2009
and ISO 10928:2009 (which replaced EN 705:1994), conducted on GRP pipes of
four different types and respective data analysis are presented below.
The test procedure specified in EN1447 relates to the creep behavior of GRP
pipes and requires constant internal pressure loading conditions to be imposed at
different levels and during different periods of time on each specimen. The time
periods range from a few minutes up to 10,000 hours. Unlike other normative
procedures (such as ISO 14692), the test temperature is room temperature. The
long-term resistance to internal pressure is determined by imposing hydrostatic
internal pressure on the specimens using axial constrained-free end sealing devic-
es. Failure must occur within a determined valid failure zone (in the middle of the
specimen) in order to be validated. Time‑to‑failure is registered. In Figure 6.12
the testing apparatus used in the experimental tests for internal pressure property
analysis is depicted.
In this test series, GRP pipes of four different construction types, from four
main European manufacturers, were used. The properties of the series of speci-
mens used are indicated in Table 6.1, in section 6.3.
All tests showed considerable scatter in results, which is typical given the ma-
terial variability of these composite constructions. Different levels of admissible
loading and different trends in the results were also observed for GRP pipes that
varied in their construction.
The statistical treatment of the data according to ISO 10928:2009 confirmed

that only data from campaigns A and B were suitable for extrapolation. However,
the experimental procedure showed the data to be applicable to all of the different
types, if a few more specimens had been tested in campaigns C and D. Graphic
representation of the data is shown in Figure 6.26, accomplished by plotting pres-
sure loading versus time‑to‑failure in log‑log scales. The regression trend lines
are also plotted.
From these results several conclusions may be drawn:
• filament‑wound pipes are more resistant than the other construction types to
structural degradation when subjected to constant internal pressure. This is
observed from the smaller slope of the regression lines of pipes A and B;
• pipes of type A, made with epoxy resin, displayed very consistent behavior
for all test periods, suggesting that this resin system might be less vulner-
able to moisture absorption and softening by resin‑fiber de-cohesion and
general degradation;
• pipes of type C, manufactured by a hybrid combination of filament winding
(continuous fibers) and mat deposition (short fibers), with relevant inclu-
sions of silica, showed a very steep slope, thus displaying an overall degra-
dation rate much higher than the others. The lack of a continuous reinforce-
ment surely increases the creep factor of the structure, since only the resin
permanently sustains circumferential stress;
A
B
1,8 C
D
Pressure [log bar]
1,6
1,4
1,2
1
-2 -1 0 1 2 3 4 5
Time to Failure [log h]
figure 6.26 Results of the test series conducted accordingly to EN 1447:2009 – pressure
versus time‑to‑failure in log-log scale [59].
• pipes A, C and D show very similar short‑term results, which agrees with
their equal nominal specifications, namely, initial stiffness and strength,
since they are classified for the same nominal service pressure and ring stiff-
ness. Specimens of type B seem clearly to be over‑designed for the specific
application of internal pressure.
6.6 Practical Design: A Case Study

In this section, a practical design case is addressed in accordance with ISO
14692:2002 [54]. In the example, material and experimental short‑term failure
test data are used to estimate the long‑term failure envelope of a specific GRP
pipe type. The nominal pressure rating of such a pipe system is then determined
by following the design procedure established in the referenced standard for a
design life of 20 years.
In order to correlate this practical example with real applications within the
oil & gas fields, as well as with the previous experience gathered from the ex-
perimental tests summarized in Figure 6.26, a glass‑fiber reinforced epoxy (GRE)
pipe system was selected. The properties of a filament‑wound GRE pipe with
two ±55º layers, characterized by Soden et al. [35], were considered. The type of
construction and geometry of the pipes are presented in Table 6.3. The mechanical
properties for each of the four single-angle laminae are summarized in Table 6.4.
Since no specific long‑term test data is available for the selected pipe speci-
mens, the test data obtained for the pipe specimens of type A in Figure 6.26 has
been used, in order to establish a reasonable long‑term hydrostatic pressure, pLTHP,
as a reference for the design procedure.
Table 6.3
Main specifications of the GRE pipe specimens considered
in the practical design case [35].
Construction Type Filament Winding
fibre type Silenka E-glass 1200TEX
resin type MY750/HY917/DY063 epoxy
winding angle [º] ±55
number of ± plies 2
internal diameter (DN) [mm] 100
average wall thickness [mm] 1
fibre volume fraction [%] 60
post‑curing cycle 2 h @ 90ºC + 1.5 h @130ºC + 2 h @ 150ºC
Practical Design: A Case Study 233
Table 6.4
Mechanical properties of each single-angle (+55º or –55º) lamina of GRE
pipe specimens considered in the practical design case [35].
axial modulus [GPa] 39.4
circumferential modulus [GPa] 46.6
axial tensile strength [MPa] 767
axial compressive strength [MPa] 578
circumferential tensile strength [MPa] 1071
circumferential compressive strength [MPa] 739
In fact, type A pipe specimens are similar to the ones selected here, both in
terms of the applied materials (E‑glass fiber and epoxy resin) and manufactur-
ing technique (filament winding). Hence, the absolute value of the slope/gradi-
ent, G = 0.0069 , of the regression line observed in the tests mentioned above
(Fig. 6.26, etc.) can constitute a reference for the expected degradation of the
properties of the pipes under consideration. However, since the tests presented in
Figure 6.26 were conducted at room temperature (EN 1447 [59]) instead of at the
higher temperature specified in ISO 14692 [54], the gradient considered hereafter
is five (5) times higher than that one. Thus, the absolute value of the gradient of
the idealized regression line is conservatively given by:
G = 5 × 0.0069 = 0.0345 (6.38)
which corresponds to an estimated degradation of the overall properties of ap-

proximately 34% during the 20 years of its useful life. The expected minimum
long‑term hydrostatic pressure, pLTHP , for the selected pipe is given by:
2 t (s shu × 0.66 )
pLTHP = = 14.137 MPa = 141.37 bar (6.39)
D
where s shu is the short‑term hoop (circumferential) ultimate strength of the pipe
specimen according to the material property data provided in Table 6.4, and t and
D are the mean wall thickness and diameter of the pipe specimens, respectively.
For the sake of simplicity, the qualifying pressure, pq , is herewith assumed to
be equal to pLTHP .
The pressure rating to provide in the product literature is given by:
pNPR = f 2 f 3 pq (6.40)
where f 2 is defined as the part load factor and f 3 is a factor accounting for the
limited axial load capability of GRP/GRE pipes. Default values for f 2 are given
in ISO 14692 [54] depending on the load-type assessment that is to be considered
for the application. These values are 0.67 for the case of sustained loads (pressure,
mass, etc.) excluding thermal effects and 0.89 for the combination of sustained
loads and occasional loads (water hammer, blast, etc.). Since our assessment is
based on sustained internal pressure (with eventual minor induced axial stress), a
value of f 2 = 0.67 is used.
For further assessments, let us assume that the biaxial stress ratio, circumferen-
tial stress/axial stress, of the application and installation for which this pipe is be-
ing designed is 1:1 and that the non‑pressure‑induced axial stress is s ab = 5 MPa .
For the calculation of f 3 , the biaxial strength ratio, r , must be assessed using
equation (6.40):
s sa ( 0:1)
r=2 (6.41)
s sh( 2:1)
where s sa ( 0:1) is the short‑term axial strength under axial loading only and s sh( 2:1)
is the short‑term hoop (circumferential) strength under 2:1 stress conditions.
Within the scope of the World Wide Failure Exercise, Soden et al. [36] presented
extensive results for the failure of GRP/GRE pipes under biaxial stress conditions.
Table 6.5 lists the failure data points for the ±55º GRE pipe considered in this
practical example. These pairs of hoop/axial stress at failure data points define the
short‑term failure envelope as shown later.
From these results one can easily calculate r by using the averaged values of
s sa ( 0:1) = 68 MPa and s sh( 2:1) = 736 MPa that these authors found for the specific
pipe under consideration. Hence, the value of the ratio r is
s sa ( 0:1) 68
r=2 =2 ≈ 0.185 << 1 (6.42)
s sh( 2:1) 736
Since r < 1 , then f 3 is given by:
2 s ab
f3 = 1 − (6.43)
r f 2 s fs
where s ab is the non‑pressure‑induced axial stress and s fs is the qualified stress,

which is related to the factored qualified pressure, pqf , by the following relation:
pqf D
s fs = (6.44)
2t
Table 6.5
Biaxial failure test results of thin ±55º GRE pipes subjected to combined internal pressure and axial loading [36].
Hoop Stress Axial Stress Stress Ratio Hoop Stress Axial Stress Stress Ratio Hoop Stress Axial Stress Stress Ratio
[MPa] [MPa] [hoop:axial] [MPa] [MPa] [hoop:axial] [MPa] [MPa] [hoop:axial]
0 74 0:1 741 370 2:1 761 138 11:2
0 62 0:1 717 358 2:1 676 67 10:1
107 143 3:4 750 375 2:1 516 0 1:0
198 198 1:1 835 334 5:2 594 0 1:0
235
331 280 6:5 803 321 5:2 638 0 1:0

374 288 13:10 914 305 3:1 622 0 1:0
525 332 3:2 939 283 13:4 544 0 1:0
599 349 7:4 867 262 13:4 492 -27
723 365 2:1 921 263 7:2 256 -65
736 368 2:1 817 148 11:2
The factored qualified pressure, pqf , is related to the qualified pressure, pq , as

follows:
pqf = A1 A2 A3 pq (6.45)
where A1 , A2 and A3 are, respectively, the partial factors for temperature, chemi-
cal resistance, and cyclic service. These additional factors account for specific ser-
vice conditions that eventually cannot be considered in the qualification program.
Since the operating temperature that one posits for the present example is less
than or equal to 65ºC, A1 = 1 . It is also assumed that the resin is chemically com-
patible with the operating fluid and therefore A2 = 1 . Given the nature of the pres-
ent study and the typical applications of such GRP pipes, it is supposed that the
predicted number of pressure or other cycles is less than 7,000 over the design life
(20 years), and so the service is considered static. Under these conditions, A3 = 1 .
The factored qualified pressure is equal to the non‑factored qualified pressure
and, therefore, the qualified stress, s fs , is determined as follows:
pqf D pq D 14.137 × 100

s fs = = = = 706.85 MPa (6.46)
2t 2t 2 ×1
Finally, f 3 can be accessed from equation (6.36), as follows:
2 s ab 2×5
f3 = 1 − = 1− ≈ 0.886 (6.47)
r f 2 s fs 0.185 × 0.67 × 706.85
and pressure rating, pNPR , to be provided in the product literature is determined

as:
pNPR = f 2 f 3 pq = 0.67 × 0.886 × 14.137 = 8.39 MPa = 83.9 bar (6.48)
The typical nominal pressure rating of this specific GRE pipe ( D = 100 mm ,
t = 1 mm , ϕ = ±55º ) would thus be PN = 80 bar .
In order to establish the long‑term failure envelope for this specific GRE pipe,
which defines its domain of allowable operating conditions under combined biax-
ial stress states, one must first determine the idealized long‑term failure envelope.
This envelope is geometrically identical to the short‑term envelope (experimental
data in Table 6.5), with the three main data points scaled through f scale , which is
given by:
s qs s fs 706.85
f scale = = = ≈ 0.96 (6.49)
s sh( 2:1) s sh( 2:1) 736
Conclusions 237
Then, the non‑factored long‑term design envelope is based on this idealized

long‑term envelope multiplied by f 2 . In this particular case, the factored long‑term
design envelope would be equal to the non‑factored design envelope, since all the
partial factors ( A1 , A2 and A3 ) are equal to unity.
In Figure 6.27 the short‑term and long‑term envelopes are depicted, based on
the short‑term experimental data [36] and the factors just calculated above.
6.7 Conclusions
The aim of the present chapter was to provide a broad description of the pres-
ent state-of-the-art concerning qualification and preliminary long-term design of
filament-wound pipes. The worked examples were mainly focused on angle-ply
stacking sequences, especially on the ±55º winding angle. Unfortunately, there is
a lack of published experimental data reporting long-term creep failure on fila-
ment-wound pipes. Moreover, since qualification standards only mandate one to
register the time or cycles to failure, experimental data on creep strain of filament-
wound pipes is even scarcer. For obvious reasons, companies are reluctant to pub-
licize important information derived from millions of creep and fatigue testing
Figure 6.27 Failure envelopes for the selected/designed GRE pipe, calculated in accor-
dance with ISO 14692. Dots are the short‑term experimental data. The solid lines repre-
sent idealized short‑term (thinnest lines), idealized long‑term (medium lines) and design
long‑term (thickest lines) failure envelopes.
hours. Yet fatigue curves for various filament-wound pipes, available in the tech-
nical literature, are more substantial. It should be pointed out that damage under
cyclic loading is more severe than damage under sustained loading. The standards
used for design and qualification of filament-wound pipes, which cover different
types of applications, were presented and compared. In the specific case of pipes
for oil and gas transportation, one worldwide standard accepted by industry is ISO
14692. The present chapter closed with a worked example based on ISO 14692,
which discussed the safety factors used in durability design.
6.8 Acknowledgements
Some test series presented here were conducted within research supported
by the Growth Programme of the European Commission and the Portuguese
Foundation for Science and Technology through projects G6RD-CT2000-00259
and POCTI/EME/47734/2002, respectively. The authors also acknowledge Dr.
Catherine Hervé (CETIM) for the data compilation of hydrostatic pressure tests
from project G6RD-CT2000-00259 partners.
References
[1] Ochoa Ozden O., Salama Mamdouh M. “Offshore composites: Transition
barriers to an enabling technology,” Composites Science and Technology
65:2588–2596, 2005.
[2] Schapery, R. A., (1969), “On the Characterization of Nonlinear Viscolelastic
Materials,” Polymer Engineering and Science 9(4):295–310, 1969.
[3] Ghorbel I., “Durability of closed-end pressurized GRP filament-wound
pipes under hygrothermal aging conditions. Part II: creep tests.” Journal of
Composite Materials 30(14): 1581–1595, 1996.
[4] Scott David W., Lai James S. and Zureick Abdul-Hamid. “Creep Behavior
of Fiber-Reinforced Polymeric Composites: A Review of the Technical
Literature. “Journal of Reinforced Plastics and Composites 14: 588–617,
1995.
[5] Guedes R. M. “An energy criterion to predict delayed failure of multi-
directional polymer matrix composites based on a non-linear viscoelastic
model,” Composites Part A 35(5): 559–571, 2004.
[6] Guedes R.M. “Nonlinear viscoelastic analysis of thick-walled cylindri-
cal composite pipes.” International Journal of Mechanical Sciences 52
(8):1064–1073, 2010.
References 239
[7] Reiner M., Weissenberg K. “A thermodynamic theory of the strength of the

materials.” Rheol Leaflet 10:12–20, 1939.
[8] Hunter S. C. “Tentative Equations for the Propagation of Stress, Strain and
Temperature Fields in Viscoelastic Solids,” J Mech Phys Solids 9(1):39–51,
1961.
[9] Guedes R.M. “Mathematical analysis of energies for viscoelastic materials
and energy based failure criteria for creep loading.” Mechanics of Time-
Dependent Materials 8(2):169–192, 2004.
[10] Guedes RM. “Lifetime predictions of polymer matrix composites under
constant or monotonic load.” Composites Part A 37 (5): 703–715, 2006.
[11] Schapery R. A. “Theory of crack initiation and growth in viscoelastic me-
dia. 3. Analysis of continuous growth.” Internat. J. Fracture 11(4):549–
562, 1975.
[12] Christensen R. M. “Lifetime predictions for polymers and composites un-
der constant load.” J. Rheology 25:517–528, 1981.
[13] Song M.S., Hua G.X. and Hub L.J., “Prediction of long-term mechanical
behaviour and lifetime of polymeric materials.” Polymer Testing 17:311–
332, 1998.
[14] Raghavan J., Meshii M. “Creep Rupture of Polymer Composites.”
Composite Science and Technology 57:375–388, 1997.
[15] Zhurkov S.N. “Kinetic concept of the strength of solids.” Int J Fract Mech
1(4):311–323, 1965.
[16] Griffith W. I. Morris D.H. and Brinson H.F. “The accelerated character-
ization of viscoelastic composite materials.” Virginia Polytechnic Institute
Report VPI-E-80-15: Blacksburg, VA, 1980.
[17] Knauss W.G. “Delayed failure-the Griffith problem for linearly viscoelastic
materials.” Internat. J. Fracture 6:7–20, 1970.
[18] Kostrov B.V. and Nikitin L.V. “Some general problems of mechanics of
brittle fracture.” Arch. Mech. Stos. (English version) 22:749–776, 1970.
[19] Schapery R.A. “Theory of crack initiation and growth in viscoelastic media.
1. Theoretical development.” Int J Fract 11(1):141–59, 1975.
[20] Schapery R.A. “Theory of crack initiation and growth in viscoelastic media.
2. Approximate methods of analysis.” Int J Fract 11(3):369–88, 1975.
[21] Leon R., Weitsman YJ. “Time-to-failure of randomly reinforced glass
strand/urethane matrix composites: data, statistical analysis and theoretical
prediction.” Mechanics of Materials 33(3):127–137, 2001.
[22] Ren W. “Creep behavior of a continuous strand, swirl mat reinforced poly-
meric composite in simulated automotive environments for durability in-
vestigation; Part I: experimental development and creep-rupture,” Materials
Science and Engineering A 334(1–2):312–319, 2002.
[23] Christensen R.M., “An evaluation of linear cumulative damage (Miner’s
Law) using kinetic crack growth theory.” Mech Time-Dependent Mater
6(4):363–77, 2002.
[24] Kachanov, L. M., “On the Time to Failure Under Creep Conditions.” Izv.
Akad. Nauk SSR, Otd. Tekhn. Nauk 8:26–31, 1958.
[25] Stigh U. “Continuum Damage Mechanics and the Life-Fraction Rule.”
Journal of Applied Mechanics 73(4):702–704, 2006.
[26] Miner M.A. “Cumulative damage in fatigue.” J. Appl. Mech. 12:A159–
A164, 1945.
[27] Bowman J., Barker M.B.A. “Methodology for describing creep–fatigue in-
teractions in thermoplastic components.” Polym Eng Sci 26(22):1582–90,
1986.
[28] Reifsnider K.L., Stinchcomb W.W. “A critical element model of the residual
strength and life of fatigue-loaded composite coupons.” In: Hahn HT, editor.
Composite materials: fatigue and fracture (ASTM STP 907). Philadelphia
(PA): American Society for Testing and Materials; p. 298–313, 1986.
[29] Reifsnider K., Case S., Duthoi J. “The mechanics of composite strength
evolution.” Composite Sci Technol 60:2539–46, 2000.
[30] Reifsnider K., Case S. Damage tolerance and durability in material sys-
tems. Wiley-Interscience; 2002.
[31] Frost S.R. “Ageing of composite in oil and gas applications,” in Ageing of
composites Edited by R Martin, Woodhead Publishing Limited, Cambridge,
2008.
[32] Roberts S.J., Evans J.T., Gibson A.G., Frost S.R. “The effect of matrix mi-
crocracks on the stress-strain relationship in fiber composite tubes.” Journal
of Composite Materials 37 (17):1509–1523, 2003.
[33] Zang W. and Gudmundson P. “An Analytical Model for the Thermoelastic
Properties of Composite Laminates Containing Transverse Matrix Cracks,”
International Journal of Solids and Structures, 30: 3211–3231, 1993.
[34] Gudmundson Peter and Östlund Sören.“Numerical Verification of a
Procedure for Calculation of Elastic Constants in Microcracking Composite
Laminates.” Journal of Composite Materials 26(17):2480–2492, 1992.
References 241
[35] Soden P. D., Hinton M. J. and Kaddour A. S. “Lamina properties, lay-up con-
figurations and loading conditions for a range of fibre-reinforced composite
laminates.” Composites Science and Technology 58 (1998) 1011–1022.
[36] Soden P.D., Hinton M.J., Kaddour A.S. “Biaxial test results for strength
and deformation of a range of E-glass and carbon fibre reinforced compos-
ite laminates: failure exercise benchmark data.” Composites Science and
Technology 62:1489–1514, 2002.
[37] Gibson A.G., Abdul Majid M.S., Assaleh T.A., Hale J.M., Fahrer A.,
Rookus C.A.P., Hekman M. “Qualification and lifetime modelling of fibre-
glass pipe.” Plastics, Rubber and Composites 40 (2):80–85, 2011.
[38] Hale J.M., Gibson A.G., Speake S.D. “Biaxial failure envelope and creep
testing of fibre reinforced plastic pipes in high temperature aqueous envi-
ronments.” Journal of Composite Materials 36 (3):257–270, 2002.
[39] Soden P.D., Kaddour A.S., Hinton M.J. “Recommendations for designers
and researchers resulting from the world-wide failure exercise.” Composites
Science and Technology 64 (3–4):589–604, 2004.
[40] Mieras H.J.M.A. “Irreversible creep of filament-wound glass-reinforced
resin pipes,” Plastics & Polymers 41:84–88, 1973.
[41] Jones M. L. C., Hull D. “Microscopy of failure mechanisms in filament-
wound pipe.” Journal of Materials Science 14:165–174, 1979.
[42] Hull D., Legg M.J., Spencer B. “Failure of glass/polyester filament wound
pipe,” Composites 9 (1):17–24, 1978.
[43] Frost S. R. and Cervenka A. “Glass fiber-reinforced epoxy matrix filament-
wound pipes for use in the oil industry.” Compos. Manuf. 5 (2):73–81, 1994.
[44] Gibson A.G., Dodds N., Frost S.R., Sheldrake T. “Novel approach to quali-
fication of non-metallic pipe systems - As applied to reinforced thermoplas-
tic pipe.” Plastics, Rubber and Composites, 34 (7):301–304, 2005.
[45] Abdul Majid M.S., Assaleh T.A., Gibson A.G., Hale J.M., Fahrer A., Rookus
C.A.P., Hekman M. “Ultimate elastic wall stress (UEWS) test of glass fibre
reinforced epoxy (GRE) pipe.” Composites: Part A 42:1500–1508, 2011.
[46] Ghorbel D. Valentin, Yrieix M.-C., Spiteri P. “The effect of resin flexibil-
ity on the creep behaviour of filament-wound glass-reinforced resin pipes.”
Developments in the science and technology of composite materials: Fourth
European Conference on Composite Materials (ECCM4), September 25–
28, 1990, Stuttgart, F.R.G. J. Füller, European Economic Community.
[47] Faria H., Guedes R.M. “Long-term behaviour of GFRP pipes: Reducing the
prediction test duration.” Polymer Testing 29 (3):337–345, 2010.
[48] Guedes R.M., Sá A., Faria H. “On the prediction of long-term creep-failure
of GRP pipes in aqueous environment,” Polymer Composites 31 (6):1047–
1055, 2010.
[49] Gdoutos E. E., Pilakoutas K., Rodopoulos C. A. Failure analysis of indus-
trial composite materials. McGraw-Hill Professional; first edition, 2000.
[50] Eckold G., “Mechanics of materials behaviour,” Chapter 3 in Design and
Manufacture of Composite Structures, Woodhead Publishing, Cambridge,
1994 pages 86–88. ISBN 1-85573-051-0.
[51] Evans J.T. and Gibson A.G. “Composite angle ply laminates and netting
analysis.” Proceedings of the Royal Society, Part A: Mathematical, Physical
and Engineering Sciences 458:3079-3088, 2002.
[52] Verchery G. “Design of laminates for in-plane loading.” ICCM International
Conferences on Composite Materials, 2009.
[53] Tew B.W. “Preliminary design of tubular composite structures using net-
ting theory and composite degradation factors.” Journal of Pressure Vessel
Technology, Transactions of the ASME 117 (4):390–394, 1995.
[54] ISO 14692:2002—Petroleum and natural gas industries Glass reinforced
plastics (GRP) piping, International Organization for Standardization,
Geneva, 2002.
[55] AWWA M45:2005—Fiberglass Pipe Design, 2nd edition, American Water
Works Association, Denver, 2005.
[56] Laney Patrick. Use of Composite Pipe Materials in the Transportation of
Natural Gas, Idaho National Engineering and Environmental Laboratory,
Idaho, 2002.
[57] HOBAS CC-GRP Pipes—Features, Tests and Benefits, Hobas Engineering
GmbH, Klagenfurt, 2010.
[58] ISO 10928:2009 Plastics piping systems—Glass-reinforced thermosetting
plastics (GRP) pipes and fittings—Methods for regression analysis and
their use, International Organization for Standardization, Geneva, 2009.
[59] EN 1447:2009 Plastics piping systems Glass reinforced thermosetting plas-
tics (GRP) pipes Determination of long term resistance to internal pressure,
European Committee for Standardization, Brussels, 2009.
Chapter 7
Flow Capacity of Composite Pipelines
7.1 Gas Transmission

The influence of gas properties and pipeline structural parameters on flow effi-
ciency can be analyzed by using the flow equation. For a compressible fluid (e.g.,
natural gas) with density ρ and pressure P flowing with a mean velocity u through
an inclined pipe (Fig. 7.1) with cross-section area A, the flow is controlled by the
Bernoulli equation:
A u A
⋅ du + A ⋅ dP + sin a ⋅ dy + π Ddy ⋅τ = 0 (7.1)
gc υ u
where:
υ is the gas specific volume

2
g c is a proportionality factor ( g c = 32.2 lbm × ft / lb f × sec )
a is the slope of the pipe (Fig. 7.1)
dy is an elementary length of the pipe
τ is the shear stress acting on the inner surface of the pipe.
243
244 Flow Capacity of Composite Pipelines
Figure 7.1 Element of a pipe conveying gas.
Taking into account the following Fanning equation providing the energy losses
due to friction
2 fu 2
dF = dL (7.2)
gc D
where f is the friction factor and L the pipeline length, the Bernoulli equation
can be written in the following form:
1 u dP sin a 2 f u2
du + + dy + dL = 0 (7.3)
gc υ 2 υ υ2 gc D υ 2
In the above equation, the ratio u / υ expresses the mass flow rate m through a
cross-section A, i.e., u / υ = m / A , and can be considered as a constant c :
u
=c (7.4)
υ
In eq. (7.3) every member of the summation expresses energy. Table 7.1 summa-
rizes the meaning of each energy term:
Gas Transmission 245
Table 7.1
Physical meaning of the terms of Bernoulli equation.
Energy terms Physical meaning
Kinetic energy
1 u
du = K .E.
gc υ 2
dP / υ = PR.E. Pressure energy
Potential energy
sin a
dy = P.E.
υ2
Energy losses due to friction
2 f u2
dL = E.L.
gc D υ 2
Integrating each energy term of eq. (7.3) along a finite pipe element lying be-
tween the cross-sections 1 and 2 (see Fig. 7.1), it can be determined that:
(a) Kinetic energy

2
1 u
K .E. = ∫ du (7.5)
1
gc υ 2
Taking into account eq. (7.4), the above equation can be written as:
2
c2 du
K .E. =
gc ∫
1
u
(7.6)
or
c 2 u2
K .E. = ln (7.7)
g c u1
where u1 , u2 are the flow velocities in cross-sections 1 and 2 respectively.
(b) Pressure energy

2
dP
PR.E = ∫ (7.8)
1
υ
or
2
PR.E = ∫ ρ dP (7.9)
1
Taking into account the law of real gases,
PV = nZRT (7.10)
where Z is the compressibility factor, R is the gas constant, and
m
n= (7.11)
M
m
ρ= (7.12)
V
where M is the average molecular weight of the gas, equation (7.9) can be writ-
ten as:
2
M ⋅P
PR.E = ∫ dP (7.13)
1
Z ⋅ R ⋅T
Using the notation,

T1 + T2
Tm = (7.14)
2
equation (7.13) yields
M P2 2 − P12
PR.E = (7.15)
Z m RTm 2
where Z m is the average value of the compressibility factor of the gas.

According to Kay’s rule [7], the compressibility factor Z m can be obtained by
using the values of the average pseudo-critical temperature Tc' and the average
pseudo-critical pressure Pc' given by the following equations:
Tc' = Tc1 y1 + Tc 2 y2 + ..... + TcN y N (7.16)
Pc' = Pc1 y1 + Pc 2 y2 + ..... + PcN y N (7.17)
In the foregoing equations, Tci , Pci (i = 1, 2, ..., N ) are the critical tempera-
ture and pressure respectively of each component of the gas (Table 7.2), and
yi (i = 1, 2, ..., N ) is the corresponding mole fraction.
Table 7.2
Critical temperature and pressure of constituents of natural gas
(Campbell Petroleum Series).
Gas Molecular Tci Pci
Component i Weight (K) (MPa)
C1 16.043 191 4.60
C2 30.070 305 4.88
C3 44.097 370 4.25
iC4 58.124 408 3.65
nC4 58.124 425 3.80
iC5 72.151 460 3.39
nC5 72.151 470 3.37
nC6 86.178 507 3.01
nC7 100.205 540 2.74
N2 28.016 126 3.40
CO2 44.010 304 7.38
H 2S 34.076 373 8.96
O2 32.000 155 5.04
H2 2.016 33 1.30
H2O 18.015 647 22.06
Air 28.960 132 3.77
He 4.000 5 0.23
Using equations (7.16), (7.17), the pseudo-critical pressure Pr' and tempera-
ture Tr' can be obtained from the following equations:
Pm
Pr' = (7.18)
Pc'
Tm
Tr' = (7.19)
Tc'
where Tm is given by the eq. (7.14) and Pm by the following equation:
∫ P dP
2
2 P1 ⋅ P2 
Pm = 1
2
=  P1 + P2 −  (7.20)
3 P1 + P2 
∫ PdP
1
Having obtained Pr' and Tr' , the average compressibility factor can be estimated
by the ref. [1].
(c) Potential Energy
2 2
sin a
P.E. = ∫ 2
dy = ∫ ρ 2 sin a dy (7.21)
1 υ 1
Taking into account equations (7.10)–(7.12), the above equation yields:
2 2
 PM 
P.E. = ∫   sin a dy (7.22)
1
ZRT 
For a gas containing N constituents, the parameters P, Z, T should be replaced by

the parameters Pm, Zm, Tm derived in the previous chapter.
Furthermore, since
2
∫ 1
sin ady = ∆Η (7.23)
where ΔH is the relative elevation of the cross section 2 with respect to cross sec-
tion 1, the potential energy term can be expressed by the following formula:
Pm2 M 2
P.E. = ∆Η (7.24)
Z m2 R 2Tm2
(d) Energy losses due to friction

According to Table 7.1 and eq. (7.4), the term of energy losses due to friction
can be written as follows:
2 fC 2 2
E.L. =
gc D ∫ 1
dL (7.25)
or
2 fC 2
E.L. = L (7.26)
gc D
7.1.1 Estimation of gas flow rate

With the aid of equations (7.7), (7.15), (7.24), (7.26), the integration of equa-
tion (7.3) yields:
C 2 u2 M P22 − P12 P2 M 2 2 fC 2
ln + + 2m 2 2 ∆Η + L =0
g c u1 Z m RTm 2 Z RT g D (7.27)
1424 3 1442443 1m42m43 12 4c 4 3
Kinetic Pressure Potential Energy Losses
Energy Energy Energy Due to Friction
For high-pressure gas transmission pipelines, the contribution of the terms of

kinetic and potential energy compared to the other terms is insignificant and can
be neglected. According to this assumption, equation (7.27) can be simplified,
yielding:
M P22 − P12 2 fC 2
+ L=0 (7.28)
Z m RTm 2 gc D
We recall that
•
m
C= (7.29)
A
where A is the cross-section area
πD 2
A= (7.30)
4
On the other hand, the gas law at standard conditions is
• •
Pb V b = nb Z b RTb (7.31)
•
where V b , Pb , Tb , Z b are the volumetric gas flow rate, pressure, temperature
and compressibility factor respectively, at standard conditions (Pb = 14.7 psia,
Tb = 520 °R, Zb ≈ 1).
Since
•
m •
n= (7.32)
M
equation (7.29), with the aid of eqs. (7.30)–(7.32), can be written as:
•
4 V b M Pb2
C= (7.33)
πRTb Z b D 2
Using the following definition of gas gravity G:
M
G= (7.34)
M air
 can be
and combining equations (7.28), (7.33), the volumetric gas flow Vb
obtained:
• P12 − P22 1 52
Vb = A D (7.35)
Z mTm GL f
where A is a constant parameter, taking the following value:
g c R Z bTb
A= π (7.36)
1.856 Pb
Equation (7.35) provides the gas flow rate at standard conditions of a pipe of di-
ameter D and length L conveying gas with inlet pressure P1 and exit pressure P2
for laminar, partially turbulent or fully turbulent flow and high pressure.
It is important to mention that the gas flow rate depends on the parameters
1 f , D 5 2 , 1 Tm .Therefore, the following conclusions can be determined:
1. The friction factor f is an important parameter controlling the ability of a

pipeline to convey gas. Pipes with a smooth interior surface and small val-
ues of f can considerably increase the gas flow rate.
2. Since the exponent of D is 5/2, an increase of diameter by a factor 2 yields
an increase in the gas flow rate of 5.6 times.
3. The dependence of the gas flow rate on the parameter 1 Tm means that
lower temperature values will increase the flow capacity of the pipeline.
7.1.1.1 friction factor for turbulent and partially turbulent flow regimes
Equation (7.35) requires a numerical value for the friction factor f. For rough
material surfaces and high values of fluid velocity (or flow rate), the flow is char-
acterized as turbulent, while for smooth material surfaces and low values of fluid
velocity, the flow is stable (or laminar).
A criterion for flow characterization is the value of the well-known Reynolds
dimensionless number Re:
ρ Du
Re = (7.37)
µ
where ρ is the fluid density and μ is the fluid viscosity. For Re > 2000 the flow is
turbulent, while for Re < 2000 the flow is stable or laminar.
For high-pressure gas pipelines, only fully and partially turbulent flow can be
observed. For partially turbulent flow, the friction coefficient can be correlated
with the Reynolds number by the following Prandtl-Von Karman equation:
1  Re 
= 4 log10  − 0.6 (7.38)
f  1 f 
 
In the case of fully turbulent flow, the friction factor f can be estimated by the
Nikuradse equation:
1  D
= 4 log10  3.7  (7.39)
f  e
where e is the surface roughness.

From eq. (7.39) it can be concluded that low values of surface roughness yield
low values of e, resulting in a higher flow capacity of the transmission line.
Since the interior surface of pipelines made from composite materials is
smoother and less corroded than the surface of steel pipes, their roughness e has
lower values (Table 7.3), which has advantages.
7.1.1.1.1 Example
Two pipelines with the same length and same diameter D = 1.0 m are fabricated
from different materials, with pipeline A constructed from glass fiber-reinforced
polymer and pipeline B from steel. Both pipelines carry the same gas with fully
turbulent flow under the same pressure and temperature conditions. The ratio of
the flow capacity V b of the two pipelines can be estimated by using eq. (7.35).
Table 7.3
Pipe roughness.
Material e (inches)
Glass-fiber reinforced pipe 0.0007869
Steel corroded 0.019688
Steel non-corroded 0.001966
Dividing the flow capacities of the two pipelines, the following formula can be
obtained:
• A
Vb fB
= (7.40)
• B fA
Vb
where VbA , f A are the flow capacity and the friction factor of the glass-fiber
reinforced pipeline, while VbB , f B are the corresponding parameters of the steel
pipeline.
Taking into account equation (7.39), the ratio of the flow capacities can be
written as:
 D
•
A log10  3.7 A 
V  e 
b
= (7.41)
•
B  D
V b log10  3.7 B 
 e 
With the aid of Table 7.3, the above equation yields:
• •
VbA ≈ 1.60 VbB for corroded steel (7.42)
• •
VbA ≈ 1.12 VbB for non-corroded steel (7.43)
The above results show that after a period of years, when the steel pipeline of
D = 1.0 m becomes corroded, the pipeline made from glass-fiber reinforced plas-
tic can supply 60% more gas flow. However, even when both pipelines are new,
the gas flow capacity of the composite pipeline is 12% higher than that of the steel
pipeline.
7.2 Liquid Transmission

7.2.1 Flow capacity for laminar liquid flow
For a pipe of length L (Fig. 7.1) conveying liquid, the Bernoulli equation yields:
ρ Lu 2
P1 − P2 = f − ∆P∆Η (7.44)
2D
where u is the mean velocity of liquid, ρ is the liquid density and ∆P∆H is the pres-
sure loss due to the elevation ΔΗ (Fig. 7.1).
Liquid Transmission 253
For laminar liquid flow (Re < 2000) the friction factor f is independent of sur-
face roughness and depends only on the Reynolds number:
64
f = (7.45)
Re
Taking into account the definition of the Reynolds number given by equation
(7.37), the combination of equation (7.44) and (7.45) yields:
u=
[ P1 − P2 + ∆P∆Η ] D 2 (7.46)
32 µL
Since the flow is given by:
πD 2
Q= u (7.47)
4
the above equation with the aid of eq. (7.46) can be written as:
π D 4 [ P1 − P2 + ∆P∆Η ]
Q= (7.48)
128µ L
When the difference of the elevation of the cross-sections 1 and 2 (see Fig. 7.1) is
ΔΗ, the parameter ΔPΔΗ is given by the following equation:
∆P∆Η = ws ∆Η (7.49)
where ws is the specific weight of the liquid.

Therefore, the flow capacity of a pipe conveying liquid in laminar flow condi-
tions can be estimated by the following equation:
π D 4 [ P1 − P2 + ws ∆H ]
Q= (7.50)
128µ L
From the above equation it can be seen that the flow capacity depends on the pa-
rameters D4 and 1/ µ .Therefore, an increase of diameter by a factor 2 yields an
increase of liquid flow rate of 16 times. On the other hand, transmission of a liquid
with double viscosity μ decreases the flow by 50%.
7.2.2 Flow capacity for turbulent flow

When Re > 2000 the liquid flow is turbulent. In that case, the friction factor f
of the eq. (7.44) cannot be obtained by eq. (7.45) because it depends on interior
surface roughness e. The dependence of f on e for turbulent flow of liquids can be
given by the Nikuradse equation:
1 D
= 1.14 + 2 log10 (7.51)
f e
Combining the above equation with equations (7.44) and (7.49) yields:
π D  2 ( P1 − P2 + ws ∆Η ) 2.5
Q= 1.14 + 2 log  D (7.52)
4 e ρL
It can be seen that the liquid flow rate in fully turbulent conditions depends on
D 2.5 . Therefore, an increase of diameter by a factor 2 will increase the liquid flow
rate by a factor of 5.66.
It should be noted that the estimation of f leads to uncertain results [9] in the tran-
sitional range when the flow changes from laminar to turbulent conditions. However,
since almost all designs concern flow in the fully turbulent range, this consequence
is rather immaterial. A complete plot of friction factor f versus Reynolds number
and roughness is given in Moody’s diagram (Fig. 7.2) for all flow regimes.
Figure 7.2 Moody’s diagram.

Multiphase Flow 255
7.3 Multiphase Flow

Pipelines often carry one or more fluids in both a gas and liquid phase (e.g.,
gas-condensate or gas-oil-water flow). These kinds of flow are called “multiphase
flows” and are controlled by the densities and viscosities of the liquid and gas
constituents as well as their velocities, volume fractions, and interfacial shear
stresses. Because of density differences among the fluids, the distribution of the
different phases within the pipeline creates different flow regimes, which are gov-
erned by different hydrodynamic equations. One of the most important data for
the mechanical design of a pipeline is pressure drop. However, in multiphase
flow, different flow regimes may induce different pressure drops. Therefore, for
successful pipeline design the flow regimes must be predicted and the relevant
flow model applied.
7.3.1 Multiphase flow regimes for inclined pipelines

7.3.1.1 phenomenology of flow regimes
The classification of multiphase flow regimes is based on experimental ob-
servations. From a literature search [e.g. 3, 4], the following cases have been ac-
cepted as the most widely used definitions for multiphase regimes:
• Stratified smooth flow: In stratified smooth flow the gas flows on the top of
the liquid and their interface is smooth (Fig. 7.3a).This regime takes place
with low gas and liquid velocities.
• Stratified wavy flow: In stratified wavy flow, the gas again flows on the
top of the liquid, but when the gas and/or liquid velocity increases, their
interface becomes wavy (Fig. 7.3b).
• Slug flow: In slug flow, the waves of liquid flow are large enough to block
the gas flow (Fig. 7.3c). This regime occurs with higher values of gas flow.
The waves in such cases are called liquid slugs. Very often liquid slugs may
contain gas bubbles.
• Annular flow: With annular flow, the gas flows as a core in the center of the
pipeline and the liquid flow surrounds the gas flow (Fig. 7.3d). This regime
comes into play when gas velocity values are high. The gas core may con-
tain small liquid droplets.
• Dispersed bubble flow: Dispersed bubble flow is characterized by a con-
tinuous liquid phase containing a high density of discrete gas bubbles with
variable size and shape (Fig. 7.3e). This regime happens in conditions of
very low values of gas velocity and high values of liquid velocity.
Taking into account the superficial liquid velocity U sl , the superficial gas ve-
locity U sg , the pressure gradient for single-phase gas flow ( dP dx ) g , the
Figure 7.3 Multiphase flow regimes: (a) stratified smooth flow, (b) stratified wavy flow,
(c) slug flow, (d) annular flow, (e) dispersed bubble flow.
pressure gradient for single-phase liquid flow ( dP dx )l , the densities of the liq-
uid and gas ρl , ρ g respectively, the liquid kinematic viscosity ν l , the pipeline
inclination angle θ, the inner diameter of the pipe D and the gravitational accelera-
2
tion g = 9.81 m / s , Taitel and Duckler [5] have developed a flow regime map
(Fig. 7.4a) based upon mechanistic models. The coordinate system for the curves
A, B, D is shown in Figs 7.4b and 7.4c.
Multiphase Flow 257
Figure 7.4 Multiphase flow regime map: (a) schematic representation of regimes [5];
(b) coordinate system for the curves A, B ; (c) coordinate system of the curve D.
7.3.1.2 criteria for flow regime transitions

The elevation of the points along a pipeline often changes, causing alterations
in flow parameters (velocities, pressures etc.).The changes in flow parameters re-
sult in spatial variations in flow regimes. To predict the regime transitions, several
quantitative criteria have been published. These can assist the pipeline designer
in correctly modeling a multiphase flow. For gas-liquid flow the following are
widely used:
7.3.1.2.1 Transition from stratified to slug flow

Criterion of Taitel and Duckler [5]:

 h
u g ≥ 1 − l 
( ρ − ρ ) A g cos θ + u
l g g
(7.53)
 D ( dA
l dhl ) ρg
w
where:
u g � is the gas velocity on the flat liquid surface
uw is the velocity of the solitary wave uw = ghl( )

hl is the surface liquid height (from the bottom of the pipe) at the area of
flat gas-liquid interface
Ag is the cross-sectional area occupied by gas at the area of flat gas-liquid
interface
Al is the cross-sectional area occupied by liquid at the area of flat gas-liquid
interface
D is the diameter of the pipe
θ is the slope of the pipe
ρl , ρ g are the densities of the liquid and gas respectively
g = 9.81 m / s 2 is the gravitational acceleration
7.3.1.2.2 Transition from slug to annular flow

Criterion of Barnea et al. [6]:
hl
≥ 0.35 (7.54)
D
Multiphase Flow 259
7.3.1.2.3 Transition from stratified smooth to wavy flow

Criterion of Taitel and Duckler [5]
4vl ( ρl − ρ g ) g cos θ
ug ≥ (7.55)
Sc ρ g ul
where:
vl is the liquid kinematic viscosity

ul is the average liquid velocity
Sc ≈ 0.01 (sheltering coefficient [5])
7.3.1.2.4 Transition from slug to dispersed bubble flow

Criterion of Taitel and Duckler [5]:
4 Ag g cos θ ρl − ρ g
ul ≥ (7.56)
Si fl ρl
where :
Si is the perimeter of gas-liquid interface (Fig. 7.6)

f l � is the liquid friction factor
7.3.1.2.5 Pressure drop in multiphase flow

Considering that all the flow parameters are independent of time, the pressure
drop in steady-state conditions may be estimated by the following models:
7.3.1.3 stratified flow model

In a stratified flow regime, the multiphase flow of a gas–liquid fluid can be
considered as two separate flows. Taking into account the equilibrium conditions
of a fluid volume element (Fig. 7.5), the following model can be obtained:
dp
p ⋅ Al − ( p + ∆x) Al + τ i si ∆x − τ l sl ∆x = 0 (7.57)
dx
Figure 7.5 Equilibrium of fluid volume element.
where p is the gas pressure, ∆x is the length of the fluid volume element, τ i and
τ l are the shear stresses at liquid-gas and liquid–pipe interfaces and si , sl are
the cross-sectional lengths at the gas-liquid interface and at the wetted periphery,
respectively (Fig. 7.6).
After several algebraic operations, equation (7.57) can be written as:
dp
− Al + τ i si − τ l sl = 0 (7.58)
dx
Figure 7.6 Geometry of the cross-section of a pipe in a stratified smooth flow regime
(gas-liquid flow).
Multiphase Flow 261
Equilibrium conditions of the gas phase yield a similar equation:
dp
− Ag + τ i si − τ g sg = 0 (7.59)
dx
where S g is the periphery of pipe in contact with gas (Fig. 7.7).
Combining equations (7.58) and (7.59) leads to:
sg sl 1 1
τg −τl + τ i si ( + ) = 0 (7.60)
Ag Al Ag Al
The shear stresses τ g , τ l ,τ i at the gas-wall, liquid-wall and gas–liquid interfaces

can be evaluated by the following well-known equations:
1
τ g = fg ρ g u g2 (7.61)
2
1
τ l = fl ρl ul2 (7.62)
2
1
τ i = fi ρ g (u g − ul ) 2 (7.63)
2
where f g , f l , f i are the corresponding friction factors.
Figure 7.7 Slug flow in an inclined pipe.

We recall that for laminar and partially turbulent flow the friction factor de-
pends on the Reynolds number given by equation (7.37). Especially for gases, the
Reynolds number should preferably be expressed in terms of the gas flow rate. For
this reason, Kennedy [7] has proposed the following equation:
0.7105 ⋅ Pb ⋅ γ ⋅ Qg
Re = (7.64)
Tb ⋅ µ ⋅ D
where Pb is base pressure in psia, Tb is base temperature in o R , γ is gas specific

gravity (dimensionless), μ is gas viscosity in cp, Qg is gas flow rate in f t 3 / day
and D is pipe diameter. In the case of a liquid-gas flow, instead of the pipe diam-
eter D, the corresponding hydraulic diameters should be used, which are given by:
Al
Dl = 4 (7.65)
sl
Ag
Dg = 4 (7.66)
sg + si
where Sl, Sg, Si are shown in Fig. 7.6.

Taking into account: (a) the already mentioned equations for computing the fric-
tion factor for gas or liquid flow, and (b) eqs. (7.64)–(7.66), the formulas summarized
in Table 7.4 can be used for two-phase flow (gas-liquid) for the three flow regimes:
For the interface between the liquid and gas one can use:
fi = f g (7.67)
Table 7.4
Friction factors for gas–liquid flow in a laminar,
partially turbulent, and fully turbulent regime.
GAS LIQUID
Laminar 360.31Tb µ g Ag 16 µl sl
Re < 2000 fg = fl =
Pbγ g Qg ( sg + si ) ρl Al ul
Partially
1 0.177 Pbγ g Qg ( sg + si ) f g 1 4ρ A u f
Turbulent = 4 log10 ( ) − 0.6 = 4 log10 ( l l l l ) − 0.6
2000 < Re < 3000 fg Tb µ g Ag fl µl sl
Fully turbulent  14.8 Ag 

1 1  14.8 Al 
= 4 log10  = 1.14 + 2 log10 
 e( s + s )  
Re > 3000
fg  g i  fl  e ⋅ sl 
Multiphase Flow 263
With the aid of the equations given in Table 7.4 and equations (7.61)–(7.63)
and (7.67), eq. (7.60) yields the following flow model for gas-liquid stratified flow
in laminar, partially turbulent and fully turbulent conditions:
f g ρ g u g2 sg 2
f l ρl ul2 sl f i ρ g (u g − ul ) si ( Al + Ag )
− + =0 (7.68)
2 Ag 2 Al 2 Al Ag
It should be noted that for the partially turbulent regime, the friction factor
f must be determined iteratively from the corresponding non-linear equation
(Table 7.4). The gas and liquid velocities are given by the ratios ug = Qg/Ag and
ul = Ql/Al , where Qg, Ql are the gas and liquid volumetric flow rates and Ag , Al
are the pipeline cross–sectional area occupied by gas and liquid, respectively.
Since the parameters Si, Sg, Sl, Ag, Al can be expressed in terms of liquid depth hl
(Fig. 7.6), the parameter hl can be obtained from equation (7.68). Once the liquid
depth hl and the correlated parameters Si, Sg, Sl, Ag, Al are known, the pressure drop
can be calculated using either eq. (7.58) or eq. (7.59).
7.3.1.3.1 Evaluation of parameters si , sg , sl , Ag , Al as a function of hl

7.3.1.3.1.1 Evaluation of Si
From the Pythagorean theorem one can write:
si = 2 R 2 − (hl − R) 2 (7.69)
or
D2 D
si = 2 − (hl − ) 2 (7.70)
4 2
7.3.1.3.1.2 Evaluation of S g
Using the geometric definition of angle θ, one can determine:
sg = θ R (7.71)
Since
θ = π −ϕ (7.72)
and
hl − R
cos ϕ = (7.73)
si / 2
eq. (7.71) yields:

  h − R 
sg =  π − arc cos  l R (7.74)
  si / 2  
With the aid of eq. (7.70), the above equation can now be written as:
  hl − D / 2  D
sg =  π − arc cos   (7.75)
  D 2 / 4 − (h − D / 2) 2   2
 l 
where D = 2R
7.3.1.3.1.3 Evaluation of S l
From Fig. 7.6 it can be shown that
sl + sg = πD (7.76)
Therefore, with the aid of equation (7.75) it can be obtained:
D  hl − D / 2 
sl =  π + arc cos   (7.77)
2   D 2 / 4 − (h − D / 2) 2  
l 
7.3.1.3.1.4 Evaluation of Ag
Taking into account the symbols shown in Fig. 7.6, the following property of
the geometry can be written [e.g., 8]:
D − hl
Ag =
6 si
( 3( D − hl )2 + 4si2 ) (7.78)
With the aid of eq. (7.70), the above equation yields:
D − hl
Ag =
2
12 D / 4 − (hl − D / 2) 2
(3( D − h )
l
2
+ 16 ( D 2 / 4 − (hl − D / 2) 2 ) )
(7.79)
Multiphase Flow 265
7.3.1.3.1.5 Evaluation of Al
Since
D2
Al + Ag = π (7.80)
4
a combination of the above equation with eq. (7.79) yields:
πD 2 D − hl
Al =
4
−
12 D / 4 − (hl − D / 2) 2
2
( (
3( D − hl ) 2 + 16 D 2 / 4 − (hl − D / 2) 2 ))
(7.81)
Equation (7.69)–(7.81) indicate that eq. (7.68) is a non-linear algebraic equation

with respect to liquid depth hl .Therefore, this equation should be solved numeri-
cally. The solution is more difficult for the case of partially turbulent flow, because
the friction factors (given in Table 7.4) must be derived iteratively.
7.3.1.4 slug flow model

As shown in Fig. 7.7, a slug flow is constituted by: (a) the region consisting
of liquid film at the bottom and the gas volume above it, and (b) the slug region
consisting of the slug body containing gas bubbles. During steady-state slug flow,
fluids picked up from the front equal the fluids being sloughed off from the back.
Therefore, the slug front velocity is higher than the mean value of the slug veloc-
ity. This slug front velocity, uT , is called “translational velocity.”
The film region, LF , together with the slug region, LS , is called the “slug unit”
[9]. Slug flow can be considered as a continuous (unit to unit) flow.
Considering both gas volume and liquid film within the segment LF as the
control volume, the equilibrium of pressure forces acting at the left and right
boundaries of the liquid film with the shear forces acting at the liquid–gas and
liquid–pipe interfaces yields:
∆p ρl (uT − uF )(us − uF ) τ i si − τ F sF
= + − ρl g sin θ (7.82)
LF LF H lF A
where Δp is the pressure difference in the left and right boundaries of the liquid
film, A is the pipe’s cross-section area, uT is the slug translational velocity, us is
the slug velocity that equals the mixture velocity, uF is the liquid film velocity,
and H lF is the liquid holdup inside the liquid film given by H lF = Vl / V , where
Vl is the volume of the liquid film within the segment LF and V is the whole
pipeline segment volume. Similarly, the equilibrium equation for the gas volume
lying above the liquid film can be written as:
∆p ρ g (uT − uc )(us − uc ) τ i si + τ c sc
= − − ρ g g sin θ (7.83)
LF LF (1 − H lF ) A
where uc is the gas core velocity.

Combining equations (7.82) and (7.83) leads to:
τ F sF LF
ρl (uT − uF )(us − uF ) − ρ g (uT − uc )(us − uc ) − +
H lF A
τ c sc LF 1
+ τ i si LF − ( ρl − ρ g ) g sin θ = 0 (7.84)
(1 − H lF ) A H lF (1 − H lF ) A
In the above equation, the shear stresses τ F , τ c , τ i can be given in the following
formulas [9]:
1
τF = f l ρl uF2 (7.85)
2
1
τc = f g ρ g uc2 (7.86)
2
1
τi = f g ρ g (uc − uF )(uc − uF ) (7.87)
2
where the friction factors can be obtained from Table 7.4.
The geometric parameters sF , si , sc are given by Zhang et al. [10] as:
sF = πDΘ l (7.88)
 D2  sin(2 π Θ l )  
sF   π Θ l −  − H lf A + H lF AD sin( π Θ l )
 4 2 
si = (7.89)
D2  sin(2 π Θ l ) 
4  π Θ l − 
2
πD 2
sc = − sF (7.90)
4
Multiphase Flow 267
where [10]:
0.25 0.8
σ ρg 1  ρlU sl2 D   U sg2 
Θl = Θlo ( water )0.15 +     (7.91)
σ ρl − ρ g cos θ  σ  2
 (1 − H lF ) gD 
In the above equation, Θl , Θlo are the pipe wall fraction wetted by liquid with
curved and flat gas/liquid interfaces, respectively; s water and s are the water and
liquid surface tension. The symbols U sl , U sg represent the liquid and gas super-
ficial velocities given by the following relations:
Ql
U sl = (7.92)
A
Qg
U sg = (7.93)
A
where Ql , Qg are the liquid and gas volumetric flow rates.
With the aid of equations (7.85)–(7.93) and the equations of Table 7.4, equation
(7.84) can provide the value of the liquid holdup inside the liquid film H lF . Once
the H lF and the correlated parameters sF , si , sc are known, the pressure drop can
be calculated by either eq. (7.82) or eq. (7.83).
7.3.1.5 annular flow model

Taking into account Fig. 7.3d, the equilibrium equation is applied to both the
liquid film and the gas core, which yields:
dp
− Af + τ i si − τ f s f − Af ρl g sin θ = 0 (7.94)
dx
and
dp
− Ac − τ i si − Ac ρ g g sin θ = 0 (7.95)
dx
where � A� f is the cross-sectional area of the liquid film, � Ac is the cross-sectional
area of the gas core, and si , s f are the perimeters of the gas–liquid and liquid–
pipe interfaces given by:
si = π ( D − 2δ ) (7.96)
s f = πD (7.97)
In equation (7.96), δ denotes the average film thickness.

By combining equations (7.94)–(7.97), the following formula can be obtained:
πD  1 1 
τf − τ iπ ( D − 2δ )  +  + ( ρl − ρ g ) g sin θ = 0 (7.98)
Af A 
 f Ac 
The cross-sectional areas Af , Ac can be correlated with the liquid film thickness δ:
Af = π ( D − δ )δ (7.99)
D
2

Ac = π  − δ  (7.100)
 2 
Moreover, the shear stresses τ i , τ f can be calculated by the following formulas:
1
τi = f g ρ g (u g − ul ) 2 (7.101)
2
1
τf = f l ρl ul 2 (7.102)
2
where the friction factors f g , f l can be obtained from Table 7.4 depending on
the flow regime. When the volumetric flow rates Qg � � and � Ql of the gas and liquid
respectively are known, the corresponding gas and liquid velocities can be esti-
mated by:
Qg
ug = (7.103)
Ac
Ql
ul = (7.104)
Af
Therefore, with the aid of equations (7.99)–(7.104), equation (7.98) can provide
the liquid film thickness δ. Consequently, when the thickness δ is known, the pa-
rameters si , s f , Ac , Af , τ i , τ f can be obtained by equations (7.96), (7.97), (7.99),
(7.100), and (7.101–7.104). Using the above results, the pressure drop for the
annular gas–liquid flow model can be estimated by either eq. (7.94) or eq. (7.95).
7.3.1.6 dispersed bubble flow model

Assuming a uniform distribution of bubbles within the liquid phase, the dis-
persed bubble flow can be treated as a pseudo-single-flow case. Such a flow type
can be modeled by the theory of liquid transmission.
References 269
References
[1] Katz et al., Handbook of gas engineering, McGraw-Hill, 1959.
[2] Simon A.L., Hydraulics, John Wiley & Sons, 1986.
[3] Song S.H., Characterization and metering of multiphase mixtures from
deep subsea wells, Ph.D. dissertation, The University of Texas at Austin,
1994.
[4] Bergles A.E., Collier J.G., Delhaye J.M., Hewitt G.F., and Mayinger F.,
Two-phase flow and heat transfer in the power and process industries,
McGraw-Hill, 1981.
[5] Taitel Y., and Dukler A.E., “A model for predicting flow regime transitions
in horizontal and near horizontal gas-liquid flow,” AIChE Journal, 22, 1,
1976.
[6] Barnea D., Shoham O., and Taitel Y., “Flow pattern transition for vertical
downward inclined two-phase flow: Horizontal to Vertical,” Chem. Eng.
Sci., 37, 1982.
[7] Kennedy J.L., Oil and gas pipeline fundamentals, Pennwell, Tulsa, 1993.
[8] Gieck K., and Gieck R., Engineering formulas, McGraw-Hill, 2006.
[9] Guo B., Song S., Chacko J., and Ghalambor A., Offshore pipelines, Elsevier,
2005.
[10] Zhang H.Q., Wang Q., Sarica C., and Brill J.P., “Unified model for gas-
liquid pipe flow via slug dynamics-Part 1: Model Development,” Journal of
Energy Resources Technology, 125, 4, 2003.
Chapter 8
Optimization of Material Cost
From the previous discussions regarding the mechanical design of composite

pipes, it can be determined that the parameters affecting the allowable loads, i.e.,
axial tensile force, external pressure, bending moment and torsional moment are:
(a) the elastic properties of the material;

(b) the strength of the material under tensile, compressive, and shear stress
conditions;
(c) the diameter of the pipe, D;
(d) the number of plies, NP, composing the pipe’s wall;
(e) the fiber orientation.
Therefore, for any specific material, only parameters (c), (d), (e) can control
the amount of the composite material needed for manufacturing a safe pipeline—
and, thus, eventually, the costs. Among these parameters, the diameter is cor-
related with the required flow capacity of the pipeline, and its estimation will be
based on a supply/demand forecast. On the other hand, the quantity of the material
is linearly dependent on the number of plies. Therefore, the present discussion
concerning the optimization of the material cost will focus on fiber orientation, θ.
8.1 Fiber Orientation and Loading Forces

In Chapter 10, one can find results of the allowable forces in axial tension,
external pressure, bending and torsion for multilayered filament-wound pipes
made from two widely used materials, E-glass/epoxy and S-glass/epoxy. The de-
rived diagrams correspond to pipes with the following dimensions: diameters,
D = 0.2–1.2m; number of plies, NP = 10–50; and fiber orientation θ = ±15°,
271
272 Optimization of Material Cost
±30°, ±45°, ±60°, ±75°. From the foregoing values of θ, the values presented in
Table 8.1 maximize the corresponding allowable forces and, therefore, minimize
the material quantity.
When a combination of the pure loading cases is used, the optimum fiber ori-
entation will vary between the limiting values that correspond to the pure (uncom-
bined) loading cases. For a pipe with diameter Dia = 0.4 m consisting of NP =
50 plies that is subjected to combined bending moment and external pressure, the
allowable bending moment M (Nm) for four values of external pressure, namely p
= 100, 150, 200, 250 kPa, is shown below in Figure 8. 1.
From this figure it can be concluded that an increase in the external pressure
drives the value of the optimum fiber orientation (θ corresponding to maximum
M) towards the optimum fiber orientation of the pure pressure. Figure 8.2 cor-
relates the optimum fiber orientation θ shown in Fig. 8.1 with the value of the
external pressure.
For loading combinations, the value of the optimum fiber orientation can be
estimated with the aid of the values of the optimum fiber orientation of the pure
loads. Using linear interpolation, the following procedures for three types of load-
ing combinations can be derived.
8.1.1 Optimum fiber orientation for the combination of axial tension and
external pressure
Since the optimum fiber orientations for pure axial tension and pure external
pressure are 0° and 90° respectively, the optimum fiber orientation θop for any
combination of the above loading cases will have a value between 0° and 90°. It is
obvious that for high values of external pressure and low values of axial tension,
θop will have a value near 90°. For loading combinations where the axial tension
is predominant, θop will be a value near 0°. For any pipe diameter and a certain
number of plies that constitute the pipe’s wall, the allowable pure axial tension
Nα for a fiber orientation 0° and the allowable pure external pressure for a fiber
Table 8.1
Optimum fiber orientation θ (deg).
LOADING FAILURE BUCKLING
TYPE E-Glass/Epoxy S-Glass/Epoxy E-Glass/Epoxy S-Glass/Epoxy
Axial Tension 0° 0° — —
Extrnal Pressure 90° 90° 0° 0°
Bending 30° 30° 0° 0°
Torsion 45° 45° 60° 60°
Fiber Orientation and Loading Forces 273
Figure 8.1 Allowable bending moment versus fiber orientation of a pipe with diameter
Dia = 0.4 m and NP = 50 plies subjected to external pressure p = 100, 150, 200, 250 KPa.
orientation of 90° can create an envelope (Fig. 8.3), which indicates the optimum
fiber orientation for any combination (N, p). Assuming that the hypotenuse of the
triangle of Fig. 8.3 represents the variation of the optimum fiber orientation from
0° (optimum θ for pure axial tension) to 90° (optimum θ for pure external pres-
sure), then the intersection of the direction OC, corresponding to a combination
(N, p) with the hypotenuse, can provide an approximation of θ°P for the combined
loading of axial loading and external pressure (N, p).
In order to derive an analytic formula for the estimation of optimum fiber ori-
entation for a loading combination (N, p), the geometric properties of the triangle
OHD, shown in Fig. 8.4, will be used. According to this figure the following geo-
metric properties can be written as:
GB CA
= (8.1)
OB OA
and
tan ϕ 2 = cot ϕ1 (8.2)

Figure 8.2 Correlation of the optimum fiber orientation with the value of the external
pressure for the loading case of Figure 8.1.
Figure 8.3 Envelope for definition of the optimum fiber orientation for any combination
of axial tension and external pressure (N, p).
Figure 8.4 Geometric model for definition of optimum fiber orientation θop for any com-
bination of axial tension N with external pressure p.
Using the loading quantities corresponding to the sides of the triangles, the
above equations can be written as:
Ν1 Ν
= (8.3)
p1 p
Ν1 Ν a − Ν1
= (8.4)
pa − p1 p1
The system of equations (8.3), (8.4) yields
pa Ν a Ν
Ν1 = (8.5)
p a Ν + pN a
and
pa Ν a p
p1 = (8.6)
p a Ν + pN a
With the aid of equations (8.5), (8.6), the segments α, β in Figure 8.4 can now be
determined:
β = ( Να − Ν1 ) 2 + p12 (8.7)
a = ( pa − p1 ) 2 + Ν12 (8.8)
Taking into account equations (8.5)–(8.8), the optimum fiber orientation for a
loading combination (N, p) can now be estimated:
β
θ op = 90 (8.9)
α +β
8.1.2 Optimum fiber orientation for the combination of bending and axial
tension
The envelope shown in Figure 8.5 corresponds to the combined bending mo-
ment M with an axial tensile force N.
Adhering to the procedure of the previous section, we obtain the following
formulas:
Μa Νa Ν
Ν1 = (8.10)
Μ a Ν + ΜN a
Μa Νa Μ
Μ1 = (8.11)
Μ a Ν + ΜN a
β = ( Να − Ν1 ) 2 + Μ12 (8.12)
a = (Μa − Μ1 ) 2 + Ν12 (8.13)
β
θ op = 30 (8.14)
α +β
It should be noted that for the case of the (N, M) combination, θop has values
between 0° (optimum fiber orientation of the pure axial load) and 30° (optimum
fiber orientation of pure bending).
Figure 8.5 Definition of optimum fiber orientation θop for a combination of bending and
axial tension.
8.1.3 Optimum fiber orientation for the combination of bending and external
pressure
The corresponding envelope for the combination of bending moment M with
external pressure p is shown in Figure 8.6.
Figure 8.6 Definition of optimum fiber orientation θop for a combination of bending and
external pressure.
Using the concepts explained in the previous sections, the following formulas
can be used for the estimation of θop for a combination of bending moment and
external pressure.
pa Μ a Μ
Μ1 = (8.15)
p a Μ + pΜ a
pa Μ a p
p1 = (8.16)
p a Μ + pΜ a
β = (Μα − Μ1 ) 2 + p12 (8.17)
a = ( pa − p1 ) 2 + Μ12 (8.18)
β
θ op = 30 + (90 − 30 ) (8.19)
α +β
With reference to Figure 8.6, it should be mentioned that for the loading combina-
tion (M, p), θop has values between 30° (optimum fiber orientation of pure bend-
ing) and 90° (optimum fiber orientation of pure pressure).
Chapter 9
Quality Control of Composite Pipe Systems
9.1 Test Methods and Material Characterization

Since composites are anisotropic materials, the conventional test methods for
characterizing homogeneous isotropic materials are not suitable. Therefore, com-
mittee D30 of ASTM and SACMA (Suppliers of Advanced Composite Materials
Association) have provided standards in order to compare, validate, and docu-
ment composite material properties. Since the field of composite materials’ char-
acterization is very wide, and therefore beyond the scope of the present book, only
a brief presentation of the main methods will be presented below. However, the
reader can find an excellent detailed overview in the refs. [1-3].
9.1.1 Thermal analysis DSC (Differential Scanning Calorimetry)

The DSC method is based on monitoring the difference in the heat flow be-
tween a sample and a reference material. From the results of this method, glass
transition and melting temperature, the degree of crystallinity for thermoplastics,
the residual heat of reaction for thermosets and the curing characteristics of ther-
mosets can be obtained, in order to evaluate the material for quality, process op-
timization and curing simulations. The DSC method is important where matrix-
dominated failures are likely to occur [4].
279
280 Quality Control of Composite Pipe Systems
9.1.2 Measurement of residual stresses

As a result of the non-homogeneity of composite materials, the different mate-
rial properties (e.g., thermal expansion coefficients, modulus of elasticity etc.) of
the constituents yield residual stresses, i.e., left-over internal stresses existing in
the absence of external loads. Residual stresses can also be generated during pro-
cessing (e.g., curing). Since residual stresses may reduce the strength of a material
or change the shape of a structural part, determining what they are is very impor-
tant. The main techniques for measuring residual stress are: hole drilling, X-ray
diffraction, neutron diffraction, the compliance method, Raman spectroscopy etc.
The drilling method is based on residual stress relaxation that results from
drilling itself. Changes in the strains around the hole, measured by a special strain
gage rosette, permit one to calculate the principal residual stresses.
The X-ray diffraction method is based on irradiation of the composite mate-
rial with high-energy X-rays. The radiation penetrates the surface of the crystal
planes and diffracts some of the X-rays. Using diffractometer techniques, mea-
surement of the changes in the inner planar spacing allows the derived elastic
strains calculation.
The neutron diffraction method (NDM) is similar to that of X-ray diffraction.
However, since the penetration depth of neutrons is greater, the NDM method is
advantageous because it can provide large quantities of data over the entire sur-
face and through the depth of the sample [3]. This method is applied mainly to
metal matrix composites, since it requires crystalline samples.
The compliance method focuses on measurement of deformations after a slot
is incrementally machined into the sample. From the obtained strain versus slot
depth results, the required values of residual stresses are calculated. The measure-
ment of strains in the compliance method is carried out by using a strain gage
rosette mounted on the specimen’s surface.
Raman spectroscopy is based on the shift to lower wave length numbers of the
Raman band of the spectrum as fibers undergo tensile deformation. Therefore, this
method can provide a map of the strain along the fibers.
9.1.3 Creep strain and creep rupture tests

Since FRPs are composed of fibers (which exhibit elastic behavior) and a
polymer matrix (which is a viscoelastic material), they exhibit a combination of
the behavior of their constituents. Therefore, the deformation of FRPs is time-
dependent even under constant loading conditions. This behavior is called creep.
During creep, damage accumulation takes place within the material’s structure
and can lead to rupture. In order to determine the creep behavior of composites,
two types of tests are usually used: (i) creep compliance tests that aim to measure
strain as a function of time for a constant stress, and (ii) creep rupture tests that
aim to measure time to failure. The duration of creep tests varies from a few hours
Test Methods and Material Characterization 281
to decades. Although creep testing can include any kind of loading, uniaxial or
bi-axial, the most common type is the tensile uniaxial creep test. To overcome the
disadvantage of the long-term testing requirements to complete a creep test, much
research is carried out with the objective of achieving accelerated characteriza-
tion of creep in composite materials. Since creep testing is the most common
testing for composites, detailed specifications (e.g., ASTM D 2990) are available
for comparison and validation of material properties. A representation of creep
testing results is shown in Fig. 9.1ab.
The configuration of straight-sided specimens used for creep testing is the same
as the configuration of the specimens for the static tensile test (ASTM D 3039).
However, for compressive creep specimens, special stiffening guides should be
placed on both flat sides of specimens in order to avoid buckling. The usual di-
mensions for uniaxial creep specimens are 150 mm for length, 13 mm for width
and 3 mm for thickness. Suitable support devices for holding the specimen’s ends
during creep testing are ones containing vise-like end tabs with inclined surfaces
to increase grip pressure and facilitate specimen alignment (Fig. 9.2).
Computerized servo-hydraulic machines are currently the best choice for per-
forming displacement-controlled or load-controlled creep tests. A typical creep-
testing machine is equipped with an environmental damper and heater for control-
ling the test conditions. The load capacity of creep machines varies from 1000 lbf
to 50000 lbf. The load acting on the specimen is measured by a pressure sensor
placed in a hydraulic network. The sensor is calibrated to provide a digital output
of the instant value of the load. Since conventional extensometers have the dis-
advantage of measuring a specimen’s deformation in only one direction, during
creep testing, strain is frequently measured by strain gages, which are a low-cost
solution and provide the capacity of multidirectional strain measurements.
Figure 9.1 Typical creep testing results for a certain temperature: (a) creep strain versus
time, (b) applied stress vs. time to failure.
Figure 9.2 Clamps for a creep test.
9.1.4 Impact testing

Composite structures in service often suffer from impact damage (e.g., a
pipe struck by foreign objects during transportation, installation or operation).
Therefore, impact resistance is an important mechanical parameter. The two main
types of testing are: (a) low-velocity impact, and (b) ballistic impact. Since ballis-
tic impact is important mainly for assessing projectiles striking composite armors,
and thus beyond the scope of this book, we shall focus on low-velocity impact.
Impact tests mainly measure energy absorption resulting from a specimen’s frac-
ture and the resistance of a specimen versus time. The two standard low-velocity
impact tests are: a) drop weight impact testing, and b) Charpy impact testing. The
drop weight impact test is the most common for composite materials. The test
machine is equipped with a weight raised to a prescribed height and released to
strike the specimen, which is a composite plate clamped along its edges according
to ASTM D 5628.
A schematic representation of the values of energy and resistance versus time,
obtained by an impact test, is demonstrated in Fig. 9.3.
Once the curve of resistance R = R(t) is measured, the corresponding accelera-
tion (or retardation) γ(t) can be obtained by Newton’s law:
R(t ) = m γ (t ) + m g
(9.1)
where m is the mass of the impactor and g is the gravitational acceleration

(g = 9.81 m/sec2).
Therefore, the velocity U(t) can be obtained by the following integration:
U (t ) = ∫ γ (t )dt (9.2)
On the other hand, energy is given by the following known integral:
E (t ) = ∫ R(t )dx (9.3)

Figure 9.3 Schematic representation of impact testing results: (a) absorbed energy versus
time, (b) specimen resistance versus time.
where x is the displacement of the contact point.

Using the definition of the velocity:
dx
U (t ) = (9.4)
dt
eq. (9.3) yields:
dx
E (t ) = ∫ R(t ) dt (9.5)
dt
or
E (t ) = ∫ R(t )U (t )dt (9.6)
With the aid of equations (9.1), (9.2), the above equation yields:
 R(t ) 
E (t ) = ∫ R(t )  − g  dt (9.7)
 m 
The Charpy impact test is older than the drop-weight one. It was used for
the experimental determination of the fracture toughness of metals. Due to its
simplicity, this method has been adopted for composites testing too. The Charpy
machine (Fig. 9.4) is equipped with a heavy pendulum that is raised to a known
height and released, impacting a prismatic specimen. The difference between the
initial and final heights H and h respectively is proportional to the amount of the
absorbed energy due to damage accumulation in the specimen. Therefore, the
total absorbed energy can be estimated by the formula:
Etot = G ( H − h) (9.8)
where G is the weight of the impactor.

Modern Charpy machines are equipped with an electronic impactor that is able
to record the specimen’s resistance R(t) versus time. Typical dimensions for a
fiber- reinforced polymer Charpy specimen are: 126 mm length, 12.7 width and
3.0-12.7 mm thickness. Typical fiber orientation angles for the specimens are: 0o,
10o, 22.5o, 30o, 45o, 67.5o and 90o. To obtain reliable results it is important to test
specimens with different fiber directions. Sometimes, post-impact analysis using
ultrasound or other non-destructive test methods is needed to fully investigate the
failure type.
9.1.5 Fatigue testing

Variable loading conditions cause damage accumulation within a composite ma-
terials’ microstructure. From a micromechanical view point, such damage accumu-
lation is the result of fiber fractures, fiber-matrix debonding, matrix cracking etc.
Figure 9.4 Standard Charpy machine.

Macroscopically, damage accumulation during repetitive loading reduces the

strength and the remaining life of the structural part, a phenomenon referred to as
fatigue. To obtain design data concerning the fatigue behavior of composites and
to compare them under standard conditions, fatigue test methods have been devel-
oped and standardized. Existing standards are related to coupon fatigue testing (e.g.,
ASTM D 3479, ASTM D 6115, EN ISO 13003) or to component fatigue testing
(e.g., ISO 14269 regarding GFRP pipes for use in the offshore petroleum industry,
EN 12245 for fiber wrapped gas cylinders, etc.). The fatigue loading can be har-
monic (i.e., consisting of well-distinguished loading cycles) or irregular (Fig. 5ab).
In the harmonic loading history shown in Fig. 9.5a, σa and σm are the values of
stress amplitude and mean stress respectively, while σmax and σmin are the maximum
and minimum stress levels. The relation of σα and σm to σmax and σmin is given by:
s max − s min
sa = (9.9)
2
s + s min
sm = (9.10)
2
In engineering practice, graduated variable amplitude loading histories (Fig. 9.6)
are often introduced.
Figure 9.5 Fatigue loading types: (a) harmonic, and (b) irregular.
Figure 9.6 Graduated variable amplitude fatigue loading.
The test specimens (Fig. 9.7) may contain holes or notches or be smooth.
Standard dimensions for fatigue specimens are 195 mm for length and 12.5–
25 mm for effective width.
The aim of fatigue testing is the derivation of curves that describe the fatigue
life, i.e., the number of loading cycles up to failure versus the stress amplitude
of fatigue loading. These curves are called S-N or Woehler curves. A schematic
representation of the procedure for the derivation of an S-N curve is shown in
Fig. 9.8.
Figure 9.7 Types of fatigue specimens.

International Standards for Composite Pipes 287
Figure 9.8 Derivation of an S-N curve from fatigue tests with differing stress amplitudes.
For performing fatigue tests, computer controlled servo-hydraulic machines

are used. These machines offer the ability to perform load- or strain-controlled
fatigue tests with harmonic, block or irregular loading histories. Since scatter in
the fatigue results is normal, statistical analysis is required to provide reliable data
for engineering design.
9.2 International Standards for Composite Pipes

Due to the wide applications of composite pipes, their testing, design and in-
stallation are standardized. Currently, United States organizations covering the
composite pipes’ technology and applications are: ASTM (American Society for
Testing and Materials), ASME (American Society of Mechanical Engineers),
API (American Petroleum Institute) and AWWA (American Water Works
Association). In addition, international standards such as DIN (Deutsches Institut
fuer Normierung), BS (British Standards), AFNOR (L’ Association Francaise de
Normalisation), as well as Japanese standards JIS (Japanese Industrial Standards)
and International Standards ISO (International Standards Organization) are used
and offer important tools for developing composite pipeline networks for oil, gas
and water applications. Since the analysis of all standards is impossible, only a
sample of the main American, European, Japanese and International standards is
summarized in Table 9.1.
Table 9.1
Standards for composite pipes.
Code Subject
ASTM D 2996 Standard Specification for Filament-Wound Reinforced Thermosetting
Resin Pipe, 1–16 inches (25-400 mm) in diameter.
ASTM D 2517 Standard Specification for Reinforced Thermosetting Resin Pipe.
ASTM D 2997 Standard Specification for Centrifugally Cast, Reinforced Thermosetting
Resin Pipe.
ASTM D 3262 Standard Specification for Reinforced Plastic Mortar Sewer Pipe.
ASTM D 3517 Standard Specification for Fiberglass Pressure Pipe.
ASTM D 3754 Standard Specification for Fiberglass Sewer and Industrial Pressure Pipe,
8-144 inches in diameter (200–3600 mm).
ASTM D 4024 Standard Specification for Reinforced Thermosetting Resin Flanges.
ASTM D 4161 Standard Specification for Fiberglass Pipe Joints using Flexible Elastomeric
Seals.
ASTM D 5686 Standard Specification for ‘Fiberglass’ (Glass Fiber-Reinforced
Thermosetting Resin) Pipes and Fittings, Adhesive Bonded Joint Type
Epoxy Resin, for Condensate Return Lines.
ASTM D 3567 Standard Practice for Determining Dimensions of Reinforced
AMERICAN STANDARDS
Thermosetting Resin Pipe and Fittings.

ASTM D 2563 Standard Practice for Classifying Visual Defects in Glass Reinforced Plastic
Laminate Parts.
ASTM D 3839 Standard Practice for Underground Installation of Flexible Reinforced
Thermosetting Resin Pipe and Reinforced Plastic Mortar Pipe.
AWWA C 950 Appendix C-Installation.
API RP 15 L4 Recommended Practice for Care and Use of Reinforced Thermosetting
Resin Line Pipe.
API RP 15 A4 Recommended Practice for Care and Use of Reinforced Thermosetting
Resin Casing and Tubing.
ASTM D 638 Standard Test Methods for Tensile Properties of Plastics.
ASTM D 1599 Short-Term Hydraulic Failure Pressure of Plastic Pipe, Tubing and Fittings.
ASTM D 2290 Apparent Tensile Strength of Ring or Tubular Plastics and Reinforced
Plastic Pipe and Tube by Split Disk Method.
ASTM D 2105 Longitudinal Tensile Properties of Reinforced Thermosetting Plastic Pipe
and Tube.
ASTM D 695 Standard Test Methods for Compressive Properties of Rigid Plastics.
ASTM D 790 Standard Test Methods for Flexural Properties of Unreinforced and
Reinforced Plastics and Electrical Insulating Materials.
ASTM D Measuring Beam Deflection of Reinforced Thermosetting Plastic Pipe
12925 under Full Bore Flow.
ASTM D1598 Time to Failure of Plastic Pipe under Constant Internal Pressure.
ASTM D 2143 Cyclic Pressure Strength of Reinforced Thermosetting Plastic Pipe.
Code Subject
ASTM D 2992 Obtaining Hydrostatic Design Basis for Reinforced Thermosetting Resin
Pipe and Fittings, Procedure A, Cyclic/Procedure B, Static.
ASTM D 2412 External Loading Characteristics of Plastic Pipe by Parallel Plate Loading.
AMERICAN STANDARDS (CONT’D)
ASTM D 2924 External Pressure Resistance of Reinforced Thermosetting Resin Pipe.

ASTM C 581 Standard Practice for Determining Chemical Resistance of Thermosetting
Resins used in Glass Reinforced Structures intended for Liquid Service.
ASTM D 3615 Chemical Resistance of Thermoset Molding compounds used in the
Manufacture of Molded Fittings.
ASTM D 3681 Chemical Resistance of Reinforced Thermosetting Resin Pipe in a
Deflected Condition.
AWWA C 950 Standard for Fiberglass Pressure Pipe for Water Service, 1–144 inches in
diameter (25–3600mm).
API 15 LR Specification for Low Pressure Fiberglass Line Pipe, 2-16 inches in diam-
eter, up to 1000 psi (cyclic)/25–300 mm, 70 bar.
API 15 HR Specification for High Pressure Fiberglass Line Pipe, 1-8 inches in diam-
eter, above 1000 psi (cyclic)/25–300 mm, 70 bar.
API 15 AR Specification for Fiberglass Tubing.
Code Subject
BS 3974 Specification for Pipe Supports. Part 1: Pipe hangers, slider and roller-type sup-
ports. Part 2: Pipe clamps, cages, cantilevers and attachments to beams. Part 3:
Large-bore, high-temperature, marine and other applications.
BRITISH STANDARDS
BS 5350 Method of Test for Adhesives. Part C5: Determination of bond strength in longi-
tudinal shear.
BS 5480 Specification for Glass Fiber-Reinforced Plastics (FRP) Pipes and Fittings for Use
for Water Supply of Sewage. Part 1: Dimensions, materials and classifications.
Part 2: Design and performance requirements.
BS 6464 Specification for Reinforced Plastics Pipes, Fittings and Joints for Process Plants.
BS 7159 Design and Construction of Glass Reinforced Plastics (FRP) Piping Systems for
Individual Plants of Sites.
BS 8010 Code of Practice for Pipeline. Section 2.5. Glass reinforced thermosetting
plastics.
Code Subject
DIN 16 867 Glass fiber-reinforced polyester resin (UP-GF) Pipes. Fittings and joints for
chemical pipelines. Technical delivery conditions.
DIN 16 868 Glass fiber-reinforced unsaturated polyester resin (UP-GF) Pipes. Part 1:
Wound, filled, dimensions. Part 2: Wound, filled. General quality.
DIN 16 869 Centrifugally cast filled fiber-reinforced unsaturated polyester resin
(UP-GF) Pipes. Part 1: Dimensions. Part 2: General quality requirements,
testing.
DIN 16 870-1 Wound glass fiber-reinforced epoxy pipes; dimensions.
DIN 16 871 Centrifugally cast glass fiber-reinforced epoxy pipe; dimensions.
DIN 16 964 Wound glass fiber-reinforced polyester resin (UP-GF) pipes, general quality
requirements. Testing.
DIN 16 965 Parts 1, 2, 4 and 5: Wound glass fiber-reinforced polyester resin pipes, types
GERMAN STANDARDS
A, B, D and E; dimensions.
DIN16 966 Part 1 – Glass fiber-reinforced polyester resin pipes. Fittings and joints.
General Quality requirements, testing.
Part 2- Elbows. Dimensions.
Part 4- Tees and nozzles. Dimensions.
Part 5 - Reducers. ngs and joints, bushings, flanges, flanged and butted
joints. General quality Requirements, testing.
Part 8 –Laminated joints. Dimensions.
DIN 19565-1 Centrifugally cast and filled polyester resin glass fiber-reinforced (UP-GF)
Pipes and fittings for buried drains and sewers, dimensions and technical
delivery conditions.
DIN 53 769 Part 1—Testing of glass fiber-reinforced plastic pipelines, determination of
the adhesive shear strength of type B pipeline components.
Part 2—Testing of glass fiber-reinforced plastic pipes; long-term hydro-
static pressure test.
Part 3—Testing of glass fiber-reinforced plastic pipes; short-term flattening
test and flattening endurance.
Part 6—Testing of glass fiber-reinforced plastic pipes; testing of pipes and
fittings under pulsating conditions.
DIN 54 815 Pipes of filled polyester resin molding materials. Part 1: Dimensions, mate-
rials, designation. Part 2: Requirements, testing.
Code Subject
T57 200 Pipes and Fittings of composite glass thermosetting materials. General review.
Description. Classification. Characteristics.
T57 201 Pipes and Fittings in FRP. Test to determine the hoop rigidity.
T57 202 Reinforced plastic pipes. Sealing ring type joints for installation under pressure
FRENCH STANDARDS (AFNOR)
or not. Suitability for use.

T57 203 Glass fiber-reinforced plastic pipes. Dimensions.
T57 205 Glass fiber-reinforced plastic pipes. Test method for short-term resistance, under
pressure, to rupture.
T57-206 Glass fiber-reinforced plastic pipes. Glass epoxy resin pipes for the transport of
hot water under pressure. Characteristics and test methods.
T57-207 Glass fiber-reinforced plastic pipes. Collection of basic data for the dimensional
calculation of pipes and fittings under constant internal pressure. Test method.
T57-208 Fiber-reinforced plastic pipes. Design and dimensioning of cemented socket
assemblies.
T57-209 Fiber-reinforced plastic pipes. Underground installation of flexible pipelines
with or without pressure.
T57-213 Fiber-reinforced plastic pipes. Resistance determination under cyclic internal
test pressure. Test method.
Code Subject
JIS K 7013 Fiber-reinforced plastic pipes.
JIS K 7014 Fittings and joints for fiber-reinforced plastic pipes.
JIS K 7020 Glass reinforced thermosetting plastics (GRP) pipes and fittings. Methods for
regression analysis and their use.
JIS K 7030 Pipes and fittings made of glass fiber reinforced plastics (GRP). Definitions
of terms relating pressure, including relationship between them, and terms for
installation and jointing.
JAPANESE STANDARDS
JIS K 7031 Plastics piping systems. Glass reinforced thermosetting plastics (GRP) pipe
and fittings. Test methods to prove the watertightness of the wall under short-
term internal pressure.
JIS K 7032 Plastic piping systems. Glass reinforced thermosetting plastics (GRP) pipes.
Determination of initial specific ring stiffness.
Determination of initial tensile properties.
JIS K 7034 Plastics piping systems. Pipes made of glass reinforced thermosetting plastic
(GRP). Determination of the resistance to chemical attack for the inside of a
section in a deflected condition.
Determination of the creep factor under wet conditions and calculation of the
long-term specific ring stiffness.
JIS K 7036 Plastics piping systems. Glass reinforced thermosetting plastics (GRP) pipes
and fittings. Test methods to prove the design of bolted flanged joints.
Code Subject
JAPANSESE (CONT’D)
JIS K 7037 Plastics piping systems. Glass reinforced thermosetting plastics (GRP) pipes.
Determination of the apparent initial circumferential tensile strength.
Test method to prove the resistance to initial ring deflection.
Determination of the long-term ultimate bending strain and calculation of the
long-term ultimate relative ring deflection, both under wet conditions.
JIS K 7040 Plastics piping systems. Glass reinforced thermosetting plastics (GRP) pipes
and fittings. Test methods to prove the design of cemented or wrapped joints.
Code Subject
EN NF DIN 705 Plastic piping systems. Glass reinforced thermosetting plastics (GRP)
pipes and fittings. Methods for regression analyses and their use.
pipes. Determination of the creep factor under dry conditions.
EN NF DIN 1115 Plastic piping systems for underground drainage and sewerage under
pressure. Glass reinforced thermosetting plastics (GRP) based on poly-
ester resin.
Part 1: General.
Part 2: Pipes with flexible, reduced articulation or rigid joints.
Part 3: Fittings.
Part 4: Ancillary equipment.
EUROPEAN STANDARDS (EN)
Part 5: Fitness for purpose of the system.

Part 6: Recommended practice for installation.
EN NF DIN 1119 Plastic piping systems. Joints for glass reinforced thermosetting plastics
(GRP) pipes and fittings. Test methods for watertightness and resistance
to damage of flexible and reduced articulation joints.
pipes and fittings. Determination of the resistance to chemical attack
from the inside of a section in a deflected condition.
pipes and fittings. Determination of the creep factor under wet conditions
and calculation of the long-term specific ring stiffness.
pipes. Test method to prove the resistance to initial ring deflection.
pipes. Determination of the long-term ultimate relative ring deflection
under wet conditions.
pipes. Determination of initial specific ring stiffness.
pipes. Test methods to prove the watertightness of the wall under shot-
term internal pressure.
Code Subject
pipes. Determination of initial longitudinal tensile properties.
pipes. Determination of the apparent initial circumferential tensile
strength.
pipes. Determination of a long-term resistance to internal pressure.
components. Test methods to prove the design of rigid locked socket and
spigot joints with elastomeric seals.
EUROPEAN STANDARDS (CONT’D)
components. Test methods to prove the design of cemented socket and
spigot joints.
components. Test methods to prove the design of bolted flanged joints.
EN NF DIN 1636 Plastic piping systems for non-pressure drainage sewerage. Glass rein-
forced thermosetting plastics (GRP).
Part 1: General.
Part 2: Pipes with flexible reduced articulation or rigid joints.
pipes. Test methods for the effects of cyclic internal pressure.
EN NF DIN 1796 Plastic piping systems for water supply with or without pressure. Glass
reinforced thermosetting plastics (GRP) based on polyester resin (UP).
Part 1: General.
Part 2: Pipes with flexible, reduced articulation or rigid joints.
Part 5: Fitness for purpose of the system.
Part 6: Recommended practice for installation.
pipes. Determination of the relative flexural creep factor following expo-
sure to a chemical environment.
Code Subject
ISO DIS 7370 Glass fiber-reinforced thermosetting plastics (GRP) pipes and fittings.
Nominal diameters, specified diameters and standard lengths.
INTERNATIONAL (ISO)
ISO DIS 7509 Glass reinforced thermosetting plastics (GRP) pipes. Determination of
time to failure under sustained internal pressure.
ISO 7510 Plastics piping systems. Glass reinforced thermosetting plastics (GRP)
pipes and fittings. Test methods to prove the watertightness of the wall
under short-term internal pressure.
ISO 7511 Glass fiber-reinforced thermosetting plastic (GRP) pipes and fittings. Test
methods to prove the watertightness of the wall under short-term internal
pressure.
Code Subject
pipes. Determination of the creep factor under dry conditions.
ISO 7685 Plastics piping systems. Glass reinforced thermosetting plastics (GRP).
Determination of initial specific ring stiffness.
ISO DIS 8483 Glass reinforced thermosetting plastics (GRP) pipes and fittings. Test
methods to prove the design of bolted flanged joints.
ISO DIS 8513 Glass reinforced thermosetting plastics (GRP) pipes. Determination of
initial longitudinal properties.
ISO 8521 Plastic piping systems. Glass reinforced thermosetting plastics (GRP)
pipes. Determination of the apparent initial circumferential tensile
strength.
ISO DIS 8533 Glass reinforced thermosetting plastics (GRP) pipes and fittings. Test
method to prove the design of cemented wrapped joints.
ISO 8572 Pipes and fittings made of glass reinforced thermosetting plastics (GRP).
Definitions of terms relating to pressure, including relationships between
INTERNATIONAL STANDARDS (ISO) (CONT’D)
them, and terms for installation and jointing.

ISO 8795 Plastic pipe systems for the conveyance of water intended for human con-
sumption. Migration assessment. Determination of migration values for
plastic pipes. (Note: valid for thermosetting and thermoplastic materials.)
ISO 10465 Underground installation of flexible glass reinforced thermosetting resin
(GRP) pipes.
Part 1: Installation.
Part 2: Comparison of static calculation methods.
Part 3: Installation parameters and application limits.
pipes. Test methods to prove the resistance to initial ring deflection.
ISO DIS 10467 Plastic piping system for pressure and non-pressure drainage and sewer-
age. Glass reinforced thermosetting plastics (GRP) based on unsaturated
polyester (UP) resins.
ISO DIS 10468 Plastic piping systems. Glass reinforced thermosetting plastics (GRP)
pipes. Determination of the long-term specific ring creep stiffness under
wet conditions and calculation of the wet creep factor.
ISO DIS 10471 Plastic piping systems. Glass reinforced thermosetting plastics (GRP)
pipes. Determination of the long-term ultimate bending strain and the
long-term ultimate relative ring deflection under wet conditions.
ISO DIS 10639 Plastic piping systems for water supply with or without pressure. Glass
reinforced thermosetting plastics (GRP) based on unsaturated polyester
(UP) resins.
ISO 10928 Plastic pipe systems. Glass reinforced thermosetting plastics (GRP) pipes
and fittings. Methods for regression analysis and their use.
ISO DIS 10952 Plastic pipe systems. Glass fiber-reinforced thermosetting plastics (GRP)
pipes and fittings. Determination of the resistance to chemical attack from
the inside of a section in a deflected condition.
ISO DIS 14828 Plastic pipe systems. Glass reinforced thermosetting plastics (GRP) pipes.
Determination of the long-term specific ring relaxation stiffness under wet
conditions and calculation of the wet relaxation factor.
Detection of Defects and Structural Health Monitoring 295
9.3 Detection of Defects and Structural Health

Monitoring
FRP composite pipelines usually do not exhibit visible fiber cracking, matrix
cracking, debonding or delamination prior to failure. Therefore, non-destructive
methods such as ultrasonography, infrared thermography, acoustic emission, elec-
tromagnetic infrared technique etc. have been applied to assess their structural
health. Acoustic emission is based on the detection (using piezoelectric sensors)
of transient elastic waves created by the rapid release of energy during loading-
induced damage accumulation. Acoustic emission can quantitatively detect the
location of defects; however, quantification of material damage is almost impos-
sible. Electromagnetic infrared thermography is predicated on the detection of
defects-induced local perturbations of an electromagnetic field, which are trans-
mitted through the composite material. However, the method is very expensive.
The most common method for detection of defects in composites is ultrasonog-
raphy. Due to the importance of this method, it will be explained in section 9.7.
Since pipe removal for testing entails downtime of the installation and very
high labor costs, in-situ structural health monitoring based on bonded or embed-
ded sensors within the composite material is now routinely used. Though numer-
ous methods, for example, electric impedance and electromagnetic response have
been developed, piezoelectric- and optical fiber-based systems are most common-
ly used for real-time monitoring of structural health.
9.3.1 Piezoelectric techniques

Piezoelectric techniques utilize piezoelectric sensors made of ceramic or poly-
meric materials that produce electrical signals in response to strain vibrations. Strain
waves propagate through the wall of composite pipes and interact with the piezo-
electric device. Since the structural defects in the composite material change its
stiffness and since the stiffness changes yield changes in the wave frequency and
amplitude (e.g., in Lamb waves) special software is used to treat the changes in the
waveforms and thereby provide information regarding the existence of damage, its
location, type and severity. Piezoelectric sensors are low-cost devices and do not
need an external power source for their operation. They can either be bonded to, or
embedded within, the wall of a composite pipe and thus provide real-time monitor-
ing of the materials’ damage accumulation. Piezoelectrics are well suited for local-
ized damage detection, except in pipes conveying high-temperature fluids.
9.3.2 Optical fiber-based techniques

Optical fiber-based techniques were first used in 1979 by Langley Research
Center of NASA in order to monitor strain in composite panels operating at low
temperature. These techniques are based on the fact that when a beam of light
enters one end of a fiber optic cable, the light is completely reflected within the
cable, even in the presence of curvature. Changes in strain due to localized dam-
age affect the transmittance behavior of the optical fiber. Detection of the changes
in the intensity or frequency of the light signal provides information regarding the
type and severity of damage. EFPI (Extrinsic Fabry-Perot Interferometric) and
FBG (Fiber Bragg Grating)-type sensors are used for structural health monitoring.
The EFPI type is based on detection of the change of displacement between the
two ends of a cleaved optical fiber. The spacing of the two ends is on the order of
100μm. One fiber end transmits multiple frequency light signals, while the other
end acts as a receiver. The damage-induced strain changes the displacement be-
tween the two ends, resulting in phase differences between the reflected waves.
The FBG type sensor is based on strain-induced shifting in the light wavelengths
(corresponding to Bragg wavelength [5]). The shift in the Bragg wavelength is
detected by opto-electrical sensors and translated to strain.
9.3.3 Ultrasonic testing

Ultrasonic testing is the most common non-destructive method for flaw detec-
tion, crack evaluation, material characterization, dimensional measurements etc.
It is based on the defect-induced reflection of an ultrasound wave, which is trans-
formed into an electrical signal that provides information concerning defect loca-
tion, orientation and size. The physical properties used for the description of how
the signal propagates through the anisotropic material are wave frequency and
acoustic impedance Z = ρc, where c is the acoustic velocity and ρ is the density of
the material. Two ultrasonic testing configurations can be used: pulse-echo (PE)
and through-transmission (TT). In the PE configuration the transducer for trans-
mission and reception of the signal is placed on the external side of the pipeline,
while in the TT configuration the transducer producing the signal is placed on the
one side and the transducer receiving the signal is located on the opposite side of
the composite wall. Since the TT configuration is impractical for testing pipelines,
only the PE technique is usually used for quality control.
Due to the high impedance mismatch of air, a liquid coupling medium, such as
water or an oil-based gel is needed between the transducer and the surface of the
composite material to be tested. However, because the use of liquid couplers is
impractical, recent advanced air-coupled ultrasonic transducers have been devel-
oped for pipeline testing. These highly sensitive transducers can be piezoelectric
or MEM (Micro-Electro-Mechanical). MEM transducers provide better acoustic
impedance than piezoelectric ones. Frequency and bandwidth are the main pa-
rameters to be considered for transducer data, to achieve the optimum sensitivity
and resolution of the system.
References 297
References
[1] Adams D., Carlsson L., and Pipes B., Experimental characterization of ad-
vanced composite materials, CRC Press, 2003.
[2] Jenkins C.H., editor, Manual on experimental methods for mechanical test-
ing of composites, The Fairmont Press, Inc. Lilburn, GA, 1998.
[3] Kessler M., Advanced topics in characterization of composites, Trafford,
2004.
[4] Dallas G., “Thermal analysis,” ASM Handbook-Composites, ASM
International, 2001.
[5] Hare D., and Moore T.C., “Characteristics of extrinsic Fabry-Perot inter-
ferometric (EFPI) fiber-optic strain gages,” NASA Technical Publication,
NASA/TP-2000-210639, 2000.
Chapter 10
Case Studies
Introduction
Chapter 10 presents a collection of nomographs for the direct mechanical de-
sign of GFRP composite pipes under a wide range of pure and combined load-
ing conditions. These diagrams contain data that are based on model behavior
of composite materials, mainly E-glass/epoxy and S-glass/epoxy, given multiple
loading parameters. Data are also provided on parameters for the construction of
pipelines, hanger supports, and appropriate depths for underground pipe.
10.1 Axial Tension

10.1.1 Results of failure model for axial tension
Taking into account the derived model given in Chapter 3, the allowable ten-
∧
sile force Ν has been estimated for pipes made from the materials: (a) E-glass/
epoxy and (b) S-glass/epoxy. The following diagrams illustrate the allowable val-
∧
ues Ν for pipes of diameters Dia = 0.1–1.2 m constituted by plies of thickness
0.150 mm, fiber orientation θ = ±15°, ±30°, ±45°, ±60°, ±75° and with the number
of plies NP = 10, 20, 30, 40, 50.
299
300 Case Studies
10.1.1.1 multilayered filament-wound E-glass/epoxy pipes subjected to axial

force
Axial Tension 301
302 Case Studies
Axial Tension 303
304 Case Studies
Axial Tension 305
10.1.1.2 multilayered filament-wound S-glass/epoxy pipes subjected to axial

force
306 Case Studies
Axial Tension 307
308 Case Studies
Axial Tension 309
310 Case Studies
10.2 Pure Bending

10.2.1 Results of failure model for pure bending
Taking into account the model described in Chapter 3, the allowable values
of bending moment M have been estimated for pipes made from the materials:
(a) E-glass/epoxy, and (b) S-glass/epoxy. The calculations were carried out by a
computer code titled “BENDING,” which was developed with the standard soft-
ware “Mathematica” (see enclosed CD-ROM). By using several values of M (last
command), the program can calculate the value of the Tsai-Wu expression. The
value of M that yields unit value for the Tsai-Wu expression should be adopted.
The diagrams below, derived from the software, present the allowable values M
for pipes of diameters Dia = 0.10 m, ,1.2 m consisting of plies with a thickness
0.150 mm and fiber orientation θ = ±15°, ±30°, …, ±75° for the number of plies
NP = 10, , 50.
10.2.1.1 multilayered filament-wound E-glass/epoxy pipes subjected to pure

bending
Pure Bending 311
312 Case Studies
Pure Bending 313
10.2.1.2 multilayered filament-wound S-glass/epoxy pipes subjected to pure

bending
314 Case Studies
Pure Bending 315
316 Case Studies
10.2.2 Results of buckling model for pure bending

10.2.2.1 buckling model for multilayered filament-wound E-glass/epoxy pipes
subjected to pure bending
Pure Bending 317
Next Page
318 Case Studies

Previous Page
Pure Bending 319

320 Case Studies
Pure Bending 321
10.2.2.2 buckling model for multilayered filament-wound S-glass/epoxy pipes

subjected to pure bending
322 Case Studies
Pure Bending 323
324 Case Studies
Pure Bending 325
326 Case Studies
10.3 External Pressure

10.3.1 Results of failure model for external pressure
Taking into account the model for external pressure exhibited in Chapter 3,
the allowable external pressure Pa has been estimated for pipes made from the
materials: (a) E-glass/epoxy, (b) S-glass/epoxy. The following diagrams present
the allowable values Pa for pipes of diameter: Dia = 0.10–1.20 m constituted by
plies of thickness 0.150 mm, fiber orientation θ = ±15°, ±30°, ±45°, ±60°, ±75°for
number of plies NP = 10–50.
10.3.1.1 multilayered filament wound E-glass/epoxy pipes subjected to

external pressure
328 Case Studies
330 Case Studies
10.3.1.2 multilayered filament-wound S-glass/epoxy pipes subjected to

external pressure

DĂƚĞƌŝĂů 'ĞŶĞƌŝĐ^Ͳ'ůĂƐƐͬƉŽǆǇh
dŚŝĐŬŶĞƐƐŽĨůĂǇĞƌ;ŵŵͿ Ϭ͘ϭϱϬ
&ŝďĞƌǀŽůƵŵĞĨƌĂĐƚŝŽŶ;йͿ Ϭ͘ϱϬ
>ĂŵŝŶĂƚĞĚĞŶƐŝƚǇ;ŐͬĐŵϯͿ Ϯ͘ϬϬ
KƌŝĞŶƚĂƚŝŽŶŽĨĨŝďĞƌƐ;ƌĂĚͿ ʋͬϭϮ

SD N3D

'LDP
332 Case Studies
334 Case Studies
10.3.2 Results of buckling model for external pressure

Using the buckling model for external pressure presented in Chapter 3, the
critical pressure pcr has been estimated for pipes made of: (a) E-glass/epoxy and
(b) S-glass/epoxy.
In the following curves, pcr is demonstrated for pipes of diameters Dia = 0.10
–1.20 m and made from plies with a thickness 0.150 mm, fiber orientation θ = ±30°,
±45°, ±60°, ±75° and number of plies NP = 10, 20, 30, 40, 50.
10.3.2.1 buckling model for multilayered filament-wound E-glass/epoxy pipes

subjected to external pressure
336 Case Studies
10.3.2.2 buckling model for multilayered filament-wound S-glass/epoxy pipes

subjected to external pressure
338 Case Studies
Next Page

Previous Page
340 Case Studies
10.4 Torsion
10.4.1 Results of failure model for torsion
Taking into account the model for torsion explained in Chapter 3, the allowable
torsion moment My has been estimated for pipes made from: (a) E-glass/epoxy
and (b) S-glass/epoxy. The following diagrams present the allowable values My
for pipes of diameter: Dia = 0.10–1.20 m constituted by plies of thickness 0.150
mm, fiber orientation θ = ±15°, ±30°, ±45°, ±60°, ±75° for a number of plies NP
= 10–50.
10.4.1.1 multilayered filament wound E-glass/epoxy pipes subjected to

torsion
Torsion 341
342 Case Studies
Torsion 343
10.4.1.2 multilayered filament wound S-glass/epoxy pipes subjected to

torsion
344 Case Studies
Butt Joints of Multilayered Filament-Wound Pipes 345
10.5 Butt Joints of Multilayered Filament-Wound

Pipes
Allowable axial tensions and bending moments are shown below for E-glass/
epoxy and S-glass epoxy materials with fiber orientation θ = ±45° and different
numbers of layers and ratios L/D (i.e. joint’s length L over its diameter D).
E-glass epoxy material
llowable axial tension
Number of Layers = 10
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
N (N) D (m) N (N) D (m) N (N) D (m) N (N) D (m)
0.1 131868 0.1 102498 0.1 123046 0.1 131868 0.1 135402
0.2 275026 0.2 263637 0.2 273558 0.2 275026 0.2 275241
0.3 412909 0.3 410338 0.3 412773 0.3 412909 0.3 412917
0.4 550556 0.4 550052 0.4 550546 0.4 550556 0.4 550557
0.5 688196 0.5 688103 0.5 688195 0.5 688196 0.5 688196
0.6 825835 0.6 825819 0.6 825835 0.6 825835 0.6 825835
0.7 963474 0.7 963471 0.7 963474 0.7 963474 0.7 963474
1101110 0.8 1101110 0.8 1101110 0.8 1101110 0.8 1101110
0.9 1238750 0.9 1238750 0.9 1238750 0.9 1238750 0.9 1238750
1.0 1376390 1.0 1376390 1.0 1376390 1.0 1376390 1.0 1376390
1.1 1514030 1.1 1514030 1.1 1514030 1.1 1514030 1.1 1514030
1.2 1651670 1.2 1651670 1.2 1651670 1.2 1651670 1.2 1651670
llowable bending moment
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
D (m) M (N * m) D (m) M (N * m) D (m) M (N * m) D (m) M (N * m) D (m) M(
0.1 4893.22 0.1 8159.01 0.1 9794.65 0.1 10496.9 0.1 10778.2
0.2 16314.4 0.2 20989 0.2 21770.7 0.2 21887.5 0.2 21904.7
0.3 29376.1 0.3 32654.7 0.3 32848.5 0.3 32859.4 0.3 32860
0.4 41975.6 0.4 43772.6 0.4 43811.9 0.4 43812.7 0.4 43812.7
0.5 53875.6 0.5 54758.2 0.5 54765.5 0.5 54765.5 0.5 54765.5
0.6 65307.8 0.6 65717.1 0.6 65718.4 0.6 65718.4 0.6 65718.4
0.7 76487.7 0.7 76671.1 0.7 76671.3 0.7 76671.3 0.7 76671.3
0.8 87543.9 0.8 87624.2 0.8 87624.2 0.8 87624.2 0.8 87624.2
0.9 98542.6 0.9 98577.1 0.9 98577.1 0.9 98577.1 0.9 98577.1
1.0 109515 1.0 109530 1.0 109530 1.0 109530 1.0 109530
1.1 120477 1.1 120483 1.1 120483 1.1 120483 1.1 120483
1.2 131433 1.2 131436 1.2 131436 1.2 131436 1.2 131436
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
N (N) D (m) N (N) D (m) N (N) D (m) N (N) D (m) N
0.1 64887.1 0.1 122943 0.1 169940 0.1 204997 0.1 229584
0.2 245886 0.2 409994 0.2 492185 0.2 527472 0.2 541610
0.3 509820 0.3 738277 0.3 804245 0.3 820675 0.3 824611
0.4 819988 0.4 1054940 0.4 1094230 0.4 110010 0.4 1100970
0.5 1147920 0.5 1354020 0.5 1374350 0.5 1376210 0.5 1376370
0.6 1476550 0.6 1641350 0.6 1651090 0.6 1651640 0.6 1651670
0.7 1797980 0.7 1922330 0.7 1926790 0.7 1926940 0.7 1926950
0.8 2109890 0.8 2200210 0.8 2202180 0.8 2202230 0.8 2202230
0.9 2412740 0.9 2476640 0.9 2477490 0.9 2477500 0.9 2477500
1.0 2708050 1.0 2752410 1.0 2752780 1.0 2752780 1.0 2752780
1.1 2997530 1.1 3027910 1.1 3028060 1.1 3028060 1.1 3028060
1.2 3282700 1.2 3303270 1.2 3303340 1.2 3303340 1.2 3303340
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 5169.74 0.1 9795.25 0.1 13539.6 0.1 16332.7 0.1 18291.6
0.2 19572.9 0.2 32636.1 0.2 39178.6 0.2 41987.4 0.2 43112.9
0.3 40575.6 0.3 58758.1 0.3 64008.3 0.3 65316 0.3 65629.2
0.4 65257.4 0.4 83956 0.4 87082.9 0.4 87550 0.4 87618.6
0.5 91353 0.5 107755 0.5 109373 0.5 109520 0.5 109534
0.6 117504 0.6 130619 0.6 131394 0.6 131438 0.6 131440
0.7 143082 0.7 152978 0.7 153333 0.7 153345 0.7 153345
0.8 167903 0.8 175090 0.8 175247 0.8 175251 0.8 175251
0.9 192002 0.9 197087 0.9 197156 0.9 197156 0.9 197156
1.0 215502 1.0 219033 1.0 219062 1.0 219062 1.0 219062
1.1 238538 1.1 240955 1.1 240968 1.1 240968 1.1 240968
1.2 261231 1.2 262868 1.2 262874 1.2 262874 1.2 262874
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 65570.9 0.1 127916 0.1 184415 0.1 233430 0.1 274370
0.2 255832 0.2 466861 0.2 614991 0.2 707587 0.2 761338
0.3 553245 0.3 922486 0.3 1107420 0.3 1186810 0.3 1218620
0.4 933722 0.4 1415170 0.4 1582410 0.4 1632140 0.4 1646220
0.5 1371850 0.5 1903340 0.5 2031040 0.5 2057780 0.5 2063210
0.6 1844970 0.6 2373620 0.6 2462030 0.6 2475230 0.6 2477170
0.7 2335300 0.7 2825920 0.7 2883500 0.7 2889690 0.7 2890340
0.8 2830350 0.8 3264280 0.8 3300310 0.8 3303110 0.8 3303320
0.9 3322250 0.9 3693040 0.9 3714950 0.9 3716180 0.9 3716250
1.0 3806690 1.0 4115560 1.0 4128620 1.0 4129150 1.0 4129170
1.1 4281800 1.1 4534190 1.1 4541860 1.1 4542090 1.1 4542090
1.2 4747240 1.2 4950470 1.2 4954910 1.2 4955010 1.2 4955010
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 5232.05 0.1 10206.7 0.1 14714.9 0.1 18262 0.1 21892.7
0.2 20372.2 0.2 37176.7 0.2 48972.4 0.2 56346 0.2 60626.2
0.3 44039 0.3 73431.1 0.3 88151.8 0.3 94471.8 0.3 97004
0.4 74315.7 0.4 112635 0.4 12594.8 0.4 129904 0.4 131024
0.5 109180 0.5 151480 0.5 161642 0.5 163771 0.5 164203
0.6 146829 0.6 188901 0.6 195964 0.6 196988 0.6 197142
0.7 185848 0.7 224892 0.7 229474 0.7 229967 0.7 230019
0.8 225241 0.8 259774 0.8 262642 0.8 262864 0.8 262881
0.9 264385 0.9 293892 0.9 295636 0.9 295734 0.9 295740
1.0 302935 1.0 327515 1.0 328554 1.0 328596 1.0 328598
1.1 340742 1.1 360828 1.1 361438 1.1 361456 1.1 361456
1.2 377781 1.2 393953 1.2 394307 1.2 394314 1.2 394315
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 65814.3 0.1 129774 0.1 190228 0.1 245886 0.1 295903
0.2 259548 0.2 491773 0.2 679761 0.2 819988 0.2 918337
0.3 570685 0.3 1019640 0.3 1310740 0.3 1476550 0.3 1564110
0.4 983546 0.4 1639980 0.4 1968740 0.4 2109890 0.4 2166440
0.5 1479510 0.5 2295840 0.5 2606850 0.5 2708050 0.5 2739250
0.6 2039280 0.6 2953110 0.6 3216980 0.6 3282700 0.6 3298440
0.7 2644690 0.7 3595960 0.7 3304560 0.7 3844670 0.7 3852180
0.8 3279950 0.8 4219770 0.8 4376640 0.8 4400420 0.8 4403860
0.9 3932220 0.9 4825470 0.9 4939930 0.9 4953270 0.9 4954810
1.0 4591680 1.0 5416100 1.0 5497410 1.0 5504830 1.0 5505500
1.1 5251270 1.1 5995060 1.1 6051760 1.1 6055810 1.1 6056100
1.2 5906220 1.2 6565400 1.2 6604360 1.2 6606550 1.2 6606670
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 5262.48 0.1 10376.7 0.1 15210.5 0.1 19660.9 0.1 23660.2
0.2 20679 0.2 39181 0.2 54158.5 0.2 65330.8 0.2 73166.6
0.3 4543.9 0.3 8113.7 0.3 104361 0.3 117563 0.3 124534
0.4 78291.6 0.4 130544 0.4 156714 0.4 167950 0.4 172451
0.5 117759 0.5 182732 0.5 207486 0.5 215541 0.5 218024
0.6 162303 0.6 235032 0.6 256033 0.6 261264 0.6 262517
0.7 210478 0.7 286186 0.7 302787 0.7 305979 0.7 356577
0.8 261030 0.8 335824 0.8 348332 0.8 350200 0.8 350474
0.9 312935 0.9 384022 0.9 393130 0.9 394192 0.9 394315
1.0 365412 1.0 431020 1.0 437431 1.0 438081 1.0 438135
1.1 417899 1.1 477091 1.1 481603 1.1 481925 1.1 481948
1.2 470018 1.2 522476 1.2 525576 1.2 525750 1.2 525760
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 65927.7 0.1 130656 0.1 193073 0.1 252222 0.1 307358
0.2 261313 0.2 504443 0.2 715940 0.2 889420 0.2 1024980
0.3 579218 0.3 1073910 0.3 1442610 0.3 1690450 0.3 1845690
0.4 1008890 0.4 1778840 0.4 2253930 0.4 2509700 0.4 2637360
0.5 1536790 0.5 2562460 0.5 3076160 0.5 3296700 0.5 3385060
0.6 2147820 0.6 3380890 0.6 3877360 0.6 4048000 0.6 4103380
0.7 2826440 0.7 4205270 0.7 4650080 0.7 4773230 0.7 4805830
0.8 3557680 0.8 5019400 0.8 5397330 0.8 5482120 0.8 5500520
0.9 4327830 0.9 5816040 0.9 6125050 0.9 6181520 0.9 6191590
1.0 5124920 1.0 6593390 1.0 6838960 1.0 6875650 1.0 6881030
1.1 5938940 1.1 7352570 1.1 7543540 1.1 7566940 1.1 7569760
1.2 6761790 1.2 8095990 1.2 8242030 1.2 8256720 1.2 8258190
Allowable bending moment
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 5285.71 0.1 10475.3 0.1 15479.5 0.1 20221.7 0.1 24642.2
0.2 20833.6 0.2 40217.6 0.2 57079.5 0.2 70910.5 0.2 81718.6
0.3 46131.1 0.3 85530.3 0.3 114895 0.3 134634 0.3 146998
0.4 80322.3 0.4 141622 0.4 179446 0.4 199809 0.4 209973
0.5 122331 0.5 203975 0.5 244866 0.5 262422 0.5 269455
0.6 170954 0.6 269099 0.6 308615 0.6 322196 0.6 326604
0.7 224956 0.7 334696 0.7 370098 0.7 379900 0.7 382495
0.8 283144 0.8 399478 0.8 429556 0.8 436305 0.8 437769
0.9 344429 0.9 462869 0.9 487461 0.9 491955 0.9 492757
1.0 407859 1.0 524725 1.0 544268 1.0 547188 1.0 547616
1.1 472635 1.1 585135 1.1 600333 1.1 602195 1.1 602420
1.2 538114 1.2 644292 1.2 655914 1.2 657083 1.2 657200
S-glass/epoxy material
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
N (N) D (m) N (N) D (m) N (N) D (m) N (N) D (m) N
0.1 61972.6 0.1 105114 0.1 128147 0.1 138728 0.1 143251
0.2 210228 0.2 277456 0.2 290251 0.2 292407 0.2 292762
0.3 384441 0.3 435376 0.3 438990 0.3 439231 0.3 439247
0.4 554911 0.4 584814 0.4 585642 0.4 585664 0.4 585664
0.5 716256 0.5 731906 0.5 732079 0.5 732081 0.5 732081
0.6 870752 0.6 878462 0.6 878497 0.6 878497 0.6 878497
0.7 1021240 0.7 1.024910 0.7 1024910 0.7 1.02491 0.7 1024910
0.8 1169630 0.8 1171330 0.8 1171330 0.8 1171330 0.8 1171330
0.9 1316970 0.9 1317740 0.9 1317750 0.9 1317750 0.9 1317750
1.0 1463810 1.0 1464160 1.0 1464160 1.0 1464160 1.0 1464160
1.1 1610420 1.1 1610580 1.1 1610580 1.1 1610580 1.1 1610580
1.2 1756920 1.2 1756990 1.2 1756990 1.2 1756990 1.2 1756990
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 4933.11 0.1 8367.21 0.1 10200.7 0.1 11042.9 0.1 11403
0.2 16730.7 0.2 22080.9 0.2 23099.1 0.2 23270.7 0.2 23299
0.3 30593.9 0.3 34647.3 0.3 34934.9 0.3 34954.1 0.3 34955.3
0.4 44159.3 0.4 46538.9 0.4 46604.7 0.4 46606.5 0.4 46606.6
0.5 56998.5 0.5 58243.9 0.5 58257.7 0.5 58257.8 0.5 58257.8
0.6 69292.8 0.6 69906.4 0.6 69909.1 0.6 69909.1 0.6 69909.1
0.7 81268.3 0.7 81559.9 0.7 81560.5 0.7 81560.5 0.7 81560.5
0.8 93076.4 0.8 93211.7 0.8 93211.8 0.8 93211.8 0.8 93211.8
0.9 104801 0.9 104863 0.9 104863 0.9 104863 0.9 104863
1.0 116487 1.0 116515 1.0 116515 1.0 116515 1.0 116515
1.1 128154 1.1 128166 1.1 128166 1.1 128166 1.1 128166
1.2 139812 1.2 139817 1.2 139817 1.2 139817 1.2 139817
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 65028.8 0.1 123945 0.1 172738 0.1 210228 0.1 237408
0.2 247891 0.2 420456 0.2 512588 0.2 554911 0.2 573005
0.3 518214 0.3 768882 0.3 848849 0.3 870752 0.3 876492
0.4 840911 0.4 1109820 0.4 1161000 0.4 1169630 0.4 1171050
0.5 1187040 0.5 1432510 0.5 1460820 0.5 1463810 0.5 1464120
0.6 1537760 0.6 1741500 0.6 1755960 0.6 1756920 0.6 1756990
0.7 1883230 0.7 2042480 0.7 2049510 0.7 2049810 0.7 2049830
0.8 2219640 0.8 2339250 0.8 2342570 0.8 2342660 0.8 2342660
0.9 2546550 0.9 2633940 0.9 2635460 0.9 2635490 0.9 2635490
1.0 2865020 1.0 2927620 1.0 2928310 1.0 2928320 1.0 2928320
1.1 3176660 1.1 3220840 1.1 3221150 1.1 3221150 1.1 3221150
1.2 3483010 1.2 3513850 1.2 3513990 1.2 3513990 1.2 3513990
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 5181.04 0.1 9875.09 0.1 13762.5 0.1 16749.5 0.1 18915
0.2 19732.4 0.2 33468.8 0.2 40802.7 0.2 44171.7 0.2 45612
0.3 41243.6 0.3 61193.8 0.3 67558.3 0.3 69301 0.3 69758.4
0.4 66922.6 0.4 88323.5 0.4 92396.6 0.4 93082.9 0.4 93196.1
0.5 94466.1 0.5 114001 0.5 116254 0.5 116492 0.5 116517
0.6 122375 0.6 138589 0.6 139739 0.6 139816 0.6 139821
0.7 149867 0.7 162540 0.7 163099 0.7 163123 0.7 163124
0.8 176637 0.8 186155 0.8 186419 0.8 186426 0.8 186426
0.9 202651 0.9 209605 0.9 209727 0.9 209729 0.9 209729
1.0 227994 1.0 232976 1.0 233031 1.0 233031 1.0 233031
1.1 252793 1.1 256309 1.1 256334 1.1 256334 1.1 256334
1.2 277171 1.2 279626 1.2 279636 1.2 279637 1.2 279637
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 65635.4 0.1 128404 0.1 185918 0.1 236590 0.1 279713
0.2 256807 0.2 473179 0.2 630683 0.2 733546 0.2 796045
0.3 557754 0.3 946025 0.3 1153320 0.3 1248550 0.3 1289260
0.4 946359 0.4 1467090 0.4 1664730 0.4 1728810 0.4 1748490
0.5 1398560 0.5 1990110 0.5 2148770 0.5 1754440 0.5 2193880
0.6 1892050 0.6 2497100 0.6 2612260 0.6 2195720 0.6 2634860
0.7 2408470 0.7 2985260 0.7 3063730 0.7 3074720 0.7 3074580
0.8 2934180 0.8 3457620 0.8 3508880 0.8 3513980 0.8 3513950
0.9 3459970 0.9 3918380 0.9 3950910 0.9 3.953230 0.9 3953220
1.0 3980220 1.0 4371240 1.0 4391440 1.0 4392480 1.0 4392480
1.1 4492010 1.1 4818920 1.1 4831260 1.1 4831730 1.1 4831730
1.2 4994200 1.2 5263320 1.2 5270770 1.2 5709800 1.2 5270980
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 5237.2 0.1 10245.6 0.1 14834.8 0.1 18878 0.1 22318.9
0.2 20449.9 0.2 37679.8 0.2 50222.1 0.2 58413.1 0.2 63390
0.3 44398 0.3 75304.9 0.3 91806.1 0.3 99386.3 0.3 102627
0.4 75321.6 0.4 116767 0.4 132498 0.4 137598 0.4 139164
0.5 111306 0.5 158385 0.5 171012 0.5 139638 0.5 174602
0.6 150576 0.6 198728 0.6 207892 0.6 174749 0.6 209691
0.7 191671 0.7 237573 0.7 243817 0.7 244692 0.7 244680
0.8 233505 0.8 275160 0.8 279240 0.8 279646 0.8 279643
0.9 275345 0.9 311826 0.9 314414 0.9 314599 0.9 314598
1.0 316745 1.0 347861 1.0 349469 1.0 349552 1.0 349552
1.1 357471 1.1 383486 1.1 384468 1.1 384506 1.1 384506
1.2 397433 1.2 418850 1.2 419443 1.2 419459 1.2 419459
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 65850.9 0.1 130058 0.1 191136 0.1 247891 0.1 299489
0.2 260115 0.2 495781 0.2 690951 0.2 840911 0.2 949631
0.3 573408 0.3 1036430 0.3 1349950 0.3 1537760 0.3 1642120
0.4 991562 0.4 1681820 0.4 2050350 0.4 2219640 0.4 2292020
0.5 1497440 0.5 2374080 0.5 2736870 0.5 2865020 0.5 2907710
0.6 2072850 0.6 3075530 0.6 3395400 0.6 3483010 0.6 3505970
0.7 2700190 0.7 3766470 0.7 4028790 0.7 4084970 0.7 4096630
0.8 3363640 0.8 4439290 0.8 4644010 0.8 4678510 0.8 4684200
0.9 4049850 0.9 5093090 0.9 5247330 0.9 5267880 0.9 5270570
1.0 4748150 1.0 5730050 1.0 5843280 1.0 5855250 1.0 5856500
1.1 5450480 1.1 6353310 1.1 6434840 1.1 6441690 1.1 6442260
1.2 6151060 1.2 6966020 1.2 7023830 1.2 7027700 1.2 7027960
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 5265.4 0.1 10399.3 0.1 15283.1 0.1 19821.2 0.1 23946.9
0.2 20724.1 0.2 39500.4 0.2 55050.1 0.2 66997.9 0.2 75659.9
0.3 45654.7 0.3 82520.2 0.3 107483 0.3 122437 0.3 130746
0.4 78929.7 0.4 133875 0.4 163211 0.4 176687 0.4 182448
0.5 119186 0.5 188959 0.5 217835 0.5 228035 0.5 231432
0.6 164974 0.6 244775 0.6 270233 0.6 277206 0.6 279033
0.7 214896 0.7 299755 0.7 320633 0.7 325103 0.7 326031
0.8 267690 0.8 353294 0.8 369586 0.8 372332 0.8 372785
0.9 322296 0.9 405320 0.9 417594 0.9 419229 0.9 419444
1.0 377864 1.0 456005 1.0 465016 1.0 465968 1.0 466068
1.1 433753 1.1 505601 1.1 512089 1.1 512633 1.1 512679
1.2 489502 1.2 554356 1.2 558958 1.2 559265 1.2 559286
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 65951.2 0.1 130841 0.1 193674 0.1 253581 0.1 309863
0.2 261681 0.2 507162 0.2 724021 0.2 905662 0.2 1051140
0.3 581021 0.3 1086030 0.3 1474200 0.3 1745290 0.3 1922210
0.4 1014320 0.4 1811320 0.4 2327060 0.4 2620150 0.4 2774560
0.5 1549320 0.5 2627850 0.5 3203680 0.5 3468200 0.5 3581280
0.6 2172060 0.6 3490590 0.6 4065190 0.6 4278980 0.6 4353760
0.7 2867970 0.7 4367880 0.7 4898970 0.7 5059850 0.7 5106210
0.8 3622650 0.8 5240310 0.8 5705310 0.8 5820620 0.8 5848140
0.9 4422590 0.9 6097780 0.9 6489170 0.9 6569020 0.9 6584840
1.0 5255690 1.0 6936390 1.0 7256200 1.0 7310170 1.0 7319060
1.1 6111470 1.1 7755900 1.1 8011520 1.1 8047200 1.1 8052110
1.2 6981180 1.2 8557960 1.2 8758690 1.2 8781960 1.2 8784620
Allowable bending moment
/D = 0.05 L/D = 0.10 L/D = 0.15 L/D = 0.20 L/D = 0.25
0.1 5287.59 0.1 10490.1 0.1 15527.6 0.1 20330.7 0.1 24843.1
0.2 20863 0.2 40434.4 0.2 57723.8 0.2 72205.4 0.2 83803.8
0.3 46274.7 0.3 86495.7 0.3 117411 0.3 139002 0.3 153092
0.4 80755.2 0.4 144208 0.4 185268 0.4 208603 0.4 220896
0.5 123328 0.5 209180 0.5 255017 0.5 276073 0.5 285075
0.6 172883 0.6 277830 0.6 323565 0.6 340581 0.6 346533
0.7 228261 0.7 347638 0.7 389907 0.7 402712 0.7 406401
0.8 288315 0.8 417059 0.8 454067 0.8 463245 0.8 465434
0.9 351972 0.9 485291 0.9 516439 0.9 522794 0.9 524054
1.0 418266 1.0 552022 1.0 577479 1.0 581768 1.0 582476
1.1 486365 1.1 617233 1.1 637576 1.1 640416 1.1 640806
1.2 555573 1.2 681056 1.2 697031 1.2 698882 1.2 699094
366 Case Studies
10.6 Hanger Width

10.6.1 E-glass/epoxy material
θ =π /12 L = 5
θ =π /12 L = 10
Hanger Width 367
θ =π /12 L = 15
10.6.2 S-glass-epoxy material
θ =π /12 L = 5
368 Case Studies
θ =π /12 L = 10
θ =π /12 L = 15
Spaces Between Supports 369
10.7 Spaces Between Supports

10.7.1 E-glass/epoxy material
NP = 50
NP = 40
L (m)
NP = 30
NP = 20
NP = 10
θ = 15° (π/12 rad)
NP = 50
NP = 40
NP = 30
NP = 20
NP = 10
θ = 30° (π/6 rad)

370 Case Studies
NP = 50
NP = 40
NP = 30
NP = 20
NP = 10
θ = 45° (π/4 rad)
NP = 50
NP = 40
NP = 30
NP = 20
NP = 10
θ = 60° (π/3 rad)

NP = 50
NP = 40
NP = 30
θ = 75° (5π/12 rad)
10.7.2 material: S-glass/epoxy
NP = 50
NP = 40
NP = 30
NP = 20
NP = 10
θ = 15° (π/12 rad)

372 Case Studies
NP = 50
NP = 40
NP = 30
NP = 20
NP = 10
θ = 30° (π/6 rad)
NP = 50
NP = 40
NP = 30
NP = 20
NP = 10
θ = 45° (π/4 rad)

NP = 50
NP = 40
NP = 30
NP = 20
NP = 10
θ = 60° (π/3 rad)
NP = 50
NP = 40
NP = 30
NP = 20
NP = 10
θ = 75° (5π/12 rad)

374 Case Studies
10.8 Installation Depth for Underground

Pipelines vs. the Vertical Load F
10.8.1 E-glass-epoxy materials
θ =π /12 F = 75 000 (N)
θ =π /12 F = 150 000 (N)

Installation Depth for Underground Pipelines vs. the Vertical Load F 375
θ =π /12 F = 225 000 (N)
θ =π /12 F = 300 000 (N)

376 Case Studies
θ =π /6 F = 75 000 (N)
θ =π /6 F = 150 000 (N)

θ =π /6 F = 225 000 (N)
θ =π /6 F = 300 000 (N)

378 Case Studies
θ =π /4 F = 75 000 (N)
θ =π /4 F = 150 000 (N)

θ =π /4 F = 225 000 (N)
θ =π /4 F = 300 000 (N)

380 Case Studies
θ =π /3 F = 75 000 (N)
θ =π /3 F = 150 000 (N)

θ =π /3 F = 225 000 (N)
θ =π /3 F = 300 000 (N)

382 Case Studies
θ = 5π /12 F = 75 000 (N)
θ = 5π /12 F = 150 000 (N)

θ = 5π /12 F = 225 000 (N)
θ = 5π /12 F = 300 000 (N)

384 Case Studies
10.8.2 S-glass/epoxy material
θ = π /12 F = 75 000 (N)
θ = π /12 F = 150 000 (N)

θ = π /12 F = 225 000 (N)
θ = π /12 F = 300 000 (N)

386 Case Studies
θ = π /6 F = 75 000 (N)
θ = π /6 F = 150 000 (N)

θ = π /6 F = 225 000 (N)
θ = π /6 F = 300 000 (N)

388 Case Studies
θ = π /4 F = 75 000 (N)
θ = π /4 F = 150 000 (N)

θ = π /4 F = 225 000 (N)
θ = π /4 F = 300 000 (N)

390 Case Studies
θ = π /3 F = 75 000 (N)
θ = π /3 F = 150 000 (N)

θ = π /3 F = 225 000 (N)
θ = π /3 F = 300 000 (N)

392 Case Studies
θ = 5π /12 F = 75 000 (N)
θ = 5π /12 F = 150 000 (N)

θ = 5π /12 F = 225 000 (N)
θ = 5π /12 F = 300 000 (N)

INDEX
Index Terms Links
ABD matrix 35 36 37
acceleration 114 115
Acoustic emission 295
additional mass effect 135
adhesive joint 170
A-Glass 48
allowable shear stress 177
annular flow 255 256 258
267
axial load 84
bending moment 143

bending 55 56 57
81 91 111
Bernoulli equation 243 244 252
Bisphenol fumerate resin 46
Buckling 57 73 80
82 92 100
bulk modulus of elasticity 150 162
buoyancy forces 135
This page has been reformatted by Knovel to provide easier navigation.
Index Terms Links
butt joint 169
cantilever pipe 118 121 123

130 137 147
carbon fibers 45 49
celerity 153 162
centrifugal force 122 123
C-Glass 48
Charpy machine 284
chlorenic resin 46
Chlorenic 47
circular frequency 119
classical lamination theory 24
compatibility equations 68 143
compliance coefficients 42
compliance matrix 6 36 42
185 194
Compliance method 280
compressibility factor 246 248 249
Compressive failure stress 40
compressive fluid 243
compressive load 84
connections 169
continental pipelines 53
coordinate system 13
coordinate transformation 13
coriolis force 123
Index Terms Links
Creep 56 197 280

Creep compliance test 280
Creep rapture test 280
critical temperature 246
critical velocity 117 129 146
147 148
Cross-Ply laminates 36 38
Curing 45
cyclic loading 220
damage accumulation 198 280

damping 123
dashpot 119
deflected spring 119
deflection 143
Differential Scanning Calorimetry 279
Dirac delta function 159
dispersed bubble flow 255 256 268
divergence 123 147
Drilling method 280
drill-string 138
dynamic loads 56
dynamic model 155
dynamic stability 105
dynamic stresses 155

Index Terms Links
E-Glass 48
Eigenfrequencies 117
elastic constant 141 147
elastic dilatation 150 152
elastic foundation 141 147
elevated temperature 56 57
energy losses 245 248
equation of motion 116 117
equibrium equations 67 143
expansion loops 179 186
extensional modulus of elasticity 173
external pressure 55 57 77
81 83 94
failure criteria 39
failure 56
Fanning equation 244
Fatigue testing 284 286
Fatigue 220
Fiber materials 48
fiber orientation 58 173 271
Filament winding 50 51
filament-wound 170 271

Index Terms Links
fixed-fixed pipe 118 122 123

130 131 132
148
flexural force 122
floating drill-bit 138
flow capacity 243 252 253
flow equation 243
fluid hammer 6
fluid viscosity 251
flutter 123
Fourier transform 159
friction factor 244 250 251
252 253
Furane resins 46
Furane 47
Galerkin method 123

gas constant 246
gas flow rate 249 250
gas gravity 250
gas specific volume 243
gas transmission 243
Glass fibers 48
global coordinate system 13
global transfer matrix 140
graphite 45

Index Terms Links
hanger width 182

helical pattern 51
hole drilling 280
Hooke’s law 1 2 9
10 11 12
172 173
hydraulic hammer 149 150 154
155
hydrodynamic forces 56
hydrostatic design basis 220
Impact testing 282

impulse-momentum equation 149
inertia effects 123
Infrared thermography 295
Instability 122 123 124
installation loads 53
integral transform 159
International standards 287
J-Lay 53 56
joining methods 169

Index Terms Links
Kevlar 45 50
kinetic energy 245
Kirchhoff assumption 27 37
Lamina 13
laminar flow 250 251 252
262
laminate nomenclature 24
laminate strains 28
laminate stresses 30
Laplace transform 160
law of real gases 246
Lekhnitskii formalism 59
lifetime prediction 197
Liquid epoxy resin 46
Liquid flow rate 253
liquid transmission 252
long-term hydrostatic pressure 220
Low-velocity impact 282
Mandrel 51
Mass 119
material characterization 279

Index Terms Links
material cost 271

matrix directions 1
Matrix material 45
Maxwell-Betti Reciprocal Theorem 5
mechanical design 53
mechanical strains 6 7 9
13
metallic collars 135 136 138
modulus of elasticity 1 58
moisture expansion coefficients 9 19
moisture strains 8 9 12
19 23 56
57
molecular weight 246
Moody’s diagram 254
multi-layered pipe 67
multiphase flow 255
neutral stability 119

Neutron diffraction 280
Newton’s equation 149 151
Nikuradse equation 251 253
Nylon 48
oil exploration 138

Index Terms Links
operation loads 53 56
Optical fiber-based techniques 295
Optimization 271
optimum fiber orientation 272 273 276
277
Orthophlatic polyester 47
Overpressure 154
partially turbulent flow 250 251 262

payout head 51
periodically supported pipe 143
Phenolic resin 46
Phenolic 47
Piezoelectric techniques 295
pinned-pinned pipe 118 122 123
130 131 132
148
plane stress 10 11 14
Poisson’s ratio 1 58
polar pattern 51
polyacrylonitril 49
Poly-Amide Imide 48
Poly-Benzimidazoles (PBls) 47
Poly-Carbonate (PC) 48
Polyester resin 46
Poly-Ether Imide 48
Poly-Ether-Ether Ketone (PEEK) 47 48
Index Terms Links
Polymer matrices 45
Poly-Phenylene Sulfide (PPS) 47 48
Poly-Propylene (PP) 48
Poly-Sulfone (PS) 48
Polyurethane resin 46
potential energy 245 248
Prandtl-Von Karman equation 251
pressure drop 255
pressure energy 245
pressure shock 149 153 155
158 161 162
principal coordinate system 1 13
principal directions 1 7
proportionality factor 243
pseudo-critical pressure 246 247
pseudo-critical temperature 246 247
pseudo-static loads 55
quality control 279
Raman spectroscopy 280

Rayon 49
reduced compliances 14 15
reduced stiffnesses 11 15 16
residual stresses 280

Index Terms Links
Reynolds number 250 251 253

262
S-Glass 48
shear failure stress 40
shear modulus 172
shock wave 153
single-layered pipe 59
S-Lay 53 56
Slope 143
slug flow 255 256 258
265
socket adhesive joint 169 170
soil-pipe interaction 57
Solid epoxy resin 46
stiffness matrix 6 36 157
stratified smooth flow 255 256 259
stratified wavy flow 255 256 259
stress-strain relations 7 8
Structural health monitoring 295
Symmetric balanced laminates 36 38
Symmetric Cross-Ply laminates 36 38
Symmetric laminates 36 38
temperature gradients 56

Index Terms Links
tensile failure stress 40

tension 55 57 88
91 94
Terephlatic polyester 47
thermal analysis 279
thermal expansion coefficient 133
thermal expansion coefficients 7 12 19
190
thermal load 132 147
thermal strains 6 7 9
12 19 23
thermal stresses 56 57
Thermoplastic resin 45 47 48
Thermosetting resin 45 46
Torsion 56 57 97
torsion spring 119
total strains 6
Towing 53
Transfer Matrices Method 124 135 148
Transfer Matrix 127 139
transformation matrix 14
transformation of engineering properties 16
translational velocity 265
Tsai-Wu Failure Criterion 39 40 41
turbulent flow 250 251 262
twisting curvature 30

Index Terms Links
Ultrasonic testing 296

Ultrasonography 295
underground pipe 56
underground pipelines 192
velocity 115
vibration 56 105 155
Vinyl ester resin 46
Vinyl ester 47
wave propagation 155 163

winding angle 51
Winkler-type foundation 141
X-ray diffraction 280

Composite Materials in Piping Applications PDF

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

Composite Materials in Piping Applications PDF

Uploaded by

Copyright:

Available Formats

Composite Materials in

Dimitrios G. Pavlou, Ph.D.

DEStech Publications, Inc.

DEStech Publications, Inc.

Copyrig ht © 2013 by DEStech Publications, Inc.

No part of this pub licati on may be reprodu ced, sto red in a

Prin ted in the United States of Ameri ca

Library of Congress Contro l Nu mber: 20139 35448

HOW TO ORDER THIS BOOK

DEStech Publications, Inc.

The author wishes to express his appreciation to colleagues who helped in

sheet. Since density is ρ = Mass/V, the allowable force P can be formulated as

UNIT PRICE $/ton

YIELD STRESS Mpa

Taking into account additional reductions in maintenance costs due to better

Chapter 1 provides a brief overview of the mechanical behavior of compos-

1 Mechanical Behavior of Fiber Reinforced

2 Classification, Properties and Production

3 Mechanical Design of Composite Pipelines.................53

3.8.2 Buckling model 100

4 Dynamic Stability of Composite Pipelines..................105

4.12.1 Example 161

5 Connection and Supports of Composite Pipelines....169

6 Creep Design of Piping Applications Using

7 Flow Capacity of Composite Pipelines........................243

7.2 Liquid Transmission 252

8 Optimization of Material Cost......................................271

9 Quality Control of Composite Pipe Systems...............279

10.3 External Pressure 326

Mechanical Behavior of Fiber Reinforced

1.1 Mechanical Behavior of Laminae

Figure 1.1 Principal coordinate system.

the relationship between the stresses and strains in the directions x1 , x2 , x3

 σ1   C11 C12 C13 0 0 0   ε1 

where Ci j {i = 1 − 6 , j = 1 − 6} , are material constants.

Figure 1.2 Stresses acting on the principal directions.

 ε1   S11 S12 S13 0 0 0   σ1 

{σ1 } = [C ] {ε1 } (1.16)

From eqs. (1.16) and (1.17), it is obvious that

According to the Maxwell-Betti Reciprocal Theorem, the following relationships

S21 = S12 (1.22)

C21 = C12 (1.28)

S33 S11 − S132

S12 S13 − S 23 S11

C31 = C13 (1.31)

C32 = C23 (1.32)

1.1.2 Effects of free thermal strains

the assumption that the fiber–matrix system can be treated as a homogeneous

 ε1 − α1 ∆Τ   S11 S12 S13 0 0 0   σ1 

 σ1   C11 C12 C13 0 0 0   ε1 − α1 ∆Τ 

where α1 , α2 , α3 are the thermal expansion coefficients in the principal directions,

Case (a) Stress with no thermal effects:

 ε1   S11 S12 S13 0 0 0   σ1 

Case (c) Stress caused by a temperature change of a fully restrained (against

 σ1   C11 C12 C13 0 0 0   − α1 ∆Τ 

1.1.3 Effects of free moisture strains

be proportional to the amount of liquid absorbed. Therefore, in analogy to the

 ε1 − β1 ∆Μ   S11 S12 S13 0 0 0   σ1 

 σ1   C11 C12 C13 0 0 0   ε1 − β1 ∆Μ 

 ε1 − α1 ∆Τ − β1 ∆Μ   S11 S12 S13 0 0 0   σ1 

 σ1   C11 C12 C13 0 0 0   ε1 − α1 ∆Τ − β1 ∆Μ 

1.1.4 Plane stress constitutive relations

 ε1   S11 S12 S13 0 0 0   σ1 