Engineering Failure Analysis 22 (2012) 128–140

Contents lists available at SciVerse ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Analysis of burst tests and long-term hydrostatic tests in produced

water pipelines
H.S. da Costa Mattos ⇑, L.M. Paim, J.M.L. Reis
Laboratório de Mecânica Teórica e Aplicada, Programa de Engenharia Mecânica, Universidade Federal Fluminense, Rua Passo da Pátria 156, 24210-240 Niterói,
RJ, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: The present work is concerned with the analysis of special hydrostatic tests performed in
Received 4 April 2011 metallic pipelines used to convey produced water in offshore oil and gas platforms. The
Received in revised form 31 January 2012 goal is to evaluate the strength of a pipeline with through-thickness corrosion damage
Accepted 31 January 2012
when repaired with special epoxy systems. Burst tests and long-term hydrostatic tests at
Available online 9 February 2012
constant pressure levels performed at the operation temperature (between 353 K and
363 K) are generally recommended for assessing the structural integrity of these repaired
Keywords:
pipelines. Elastoplasticity theory and thermodynamics are used in the modelling of these
Pipeline failures
Corrosion
tests. The goal is to propose phenomenological models, as simple as possible, able to per-
Damage assessment form a physically realistic description of the non-linear phenomena. In the case of hydro-
Failure analysis static tests, it is shown that that both yield and burst pressure can be related with the yield
Plastic deformation stress and ultimate stress obtained in a tensile test. In the case of long-time tests, special
attention is given to the explanation of why huge pressure peaks may be induced by a very
small temperature variation if an inadequate pressure control system is adopted. Model
predictions are compared with experimental results showing a good agreement.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction

In a recent paper [1] it was presented a new methodology to repair localised corrosion damage in metallic pipelines with
epoxy resins. The main motivation for the use of such kind of repair system is corrosion defects in produced water pipelines
used in offshore oil exploitation [2]. Since offshore platforms are hydrocarbon atmospheres, any repair method using equip-
ment that may produce heat and/or sparkling is forbidden. Although the operation pressure of these pipelines is not very
high, the water temperature is between 353 K and 363 K, which can be a major shortcoming for the use of polymeric mate-
rial as repair systems.
The petroleum as found in the nature is actually a mixture, basically composed of oil, gas and water. Oil reservoirs fre-
quently contain large volumes of water, while gas reservoirs tend to produce only small quantities. At the surface, produced
water is separated from the hydrocarbons, treated to remove as much oil as possible, and then either discharged into the sea
or injected back into the wells. In addition, some installations are able to inject produced water into other suitable geological
formations. After treatment, produced water still contains traces of oil and, thus, discharge into the sea is strictly controlled
by legislation. Produced water contains various chemicals such as heavy metals, polynuclear aromatic hydrocarbon, and
radionuclides, is one of the most significant wastewater discharges in the offshore oil industry.
The present paper is concerned with the analysis of some special tests developed for analysing the structural integrity of
metallic pipelines used convey produced water in offshore oil and gas platforms: (i) yield and burst test performed to check

⇑ Corresponding author. Tel./fax: +55 21 2629 5585.

E-mail address: heraldo@mec.uff.br (H.S. da Costa Mattos).

1350-6307/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2012.01.011
 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140 129

the strength of undamaged specimens, and (ii) long-term tests performed at constant pressure (generally above 3 MPa) with
water temperature between 353 K and 363 K in damaged specimens to check the durability some kind of epoxy repair
system.
Hydrostatic tests are normally conducted under industry and/or customer requirements or specifications. They are usu-
ally performed taking as specimen cylindrical vessels closed at the ends and are often used to assess information about the
mechanical strength of a given pipeline (or about the effectiveness of a given repair or reinforcement system in a damaged
pipeline). The material, diameter and wall thickness of the specimen are normally the same than the pipeline. The cylinder is
closed at the extremities using a welded cap or a bolted pressure flange. Nevertheless, since the real pipeline is not a closed
vessel, mistaken conclusions can be made if the perturbation caused by the closed ends of the specimen is not taken into
account.
In the present work the problem is modelled in the context of elasto-plasticity as a thin-walled cylinder closed at the
extremities under pressure. An analytical solution is obtained and it allows correlating both yield and burst pressure with
the yield stress and ultimate stress obtained in a tensile stress.
Real pipelines are long and the effect of axial stresses in straight lines is almost negligible, what is not the case of the
specimens for hydrostatic testing. Since both yield and burst pressures obtained for a closed-ended cylinder are approxi-
mately 2/3 (1.155) higher than the ones obtained for an open-ended cylinder, despite the scatter of the experimental results,
it may be interesting to take into account this theoretical fact in the integrity analysis of a real line. The predictions of burst
pressures of different specimens using the elasto-plastic model are in good agreement with experimental results and are also
compared to those obtained using the ASME Boiler and Pressure Vessel Code [3].
In the case of long-term tests (over 1000 h) under higher fluid temperatures, wrong predictions may occur depending on
how the testing system is prepared. In these tests, specimens with through-wall corrosion defects repaired with some epoxy
system are submitted to a constant pressure (generally above 3 MPa) under temperatures between 60 °C and 90 °C. The idea
is to check the durability of the repair. Normally, firstly temperature is raised up to a given fixed level and then pressure is
applied. After the desired pressure limit is reached, the pressure connection is closed using a valve and, since the pressure is
supposed to be fixed, the only active control system aims to assure that temperature remains constant. Generally these con-
trol systems allow very small oscillations around the temperature reference. Nevertheless, it is verified experimentally that,
if the specimen is completely closed and the mass of fluid is constant inside it, a very small temperature variation (even 1 K)
may induce a huge pressure variation, eventually leading the repair to failure. In the present study, a simple constitutive
equation for water systems at very high pressures is proposed to explain this phenomenon. It is a very simple state law that
extends the Tait equation of state for barotropic fluids (or called the Murnaghan equation of state in the context of an elastic
solid [4–6]) to account for small temperature variations. The predictions of a huge pressure peak in such systems induced by
a very small increase in temperature are in agreement with experimental results. Therefore, long-term hydrostatic tests
must be performed with two active control systems (for temperature and for pressure at the same time). The use of a pres-
sure relief system as a part of the pressure control system is recommended in order to avoid the very fast high pressure peaks
caused by small temperature variations. Without an effective pressure control system, the durability and effectiveness of the
repair system in a long-term hydrostatic test may be underestimated due to a failure caused by undesirable and unexpected
pressure peaks induced by very small temperature variations admitted by the temperature control system.

2. Modelling the burst pressure of elasto-plastic cylinders

The axial stress induced by the pressure applied at the extremities of the specimen can be important in the case of burst
tests of pressurised cylinders. A variety of papers concerned with this subject have been developed within the framework of
modern continuum plasticity [7–12]. Chistopher et al. [8] have examined several predictive equations with the failure data
of different steel vessels and concluded that there is no unique failure criterion which can predict accurately the failure
pressure of cylindrical vessels. One of the main motivations for carrying experimental investigations on the burst tests
of pressurised cylinders is the evaluation of the failure behaviour of steel rocket motor cases under internal pressure
[13,14].
The goal of this section is to propose a theoretical analysis restricted to thin-walled pipes and considering a particular set
of elasto-plastic constitutive equations in order to obtain a simple expression useful for failure pressure estimates. The idea
is to obtain analytical expressions, as simple as possible, showing that both yield and burst pressures of a thin-walled elasto-
plastic cylinder under a monotonically increasing internal pressure can be estimated through a simple tensile test, reducing
the necessity of performing a large number of hydrostatic tests, saving time and reducing experimental costs.

2.1. Summary of the elasto-plastic constitutive equations

The following set of elasto-plastic constitutive equations is a particular case of the constitutive equations discussed in
[15] but restricted to isotropic hardening. These equations are adequate to model the monotonic inelastic behaviour of
metallic material undergoing a quasi-static and isothermal process at room temperature.
In the framework of small deformations and isothermal processes, besides de stress tensor r and the strain tensor
e ¼ 12 ½ru þ ðruÞT  (u is the displacement at a given material point), in this theory it is considered the following auxiliary vari-
130 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140

ables: the plastic strain tensor ep , the cumulated plastic strain p and another variable Y, related to the isotropic hardening. A
complete set of elasto-plastic constitutive equations is given by:
mE E ð1 þ mÞ m
r¼ trðe  ep Þ1 þ ðe  ep Þ ) ðe  ep Þ ¼ r  trðrÞ1 ð1Þ
ð1 þ mÞð1  2mÞ ð1 þ mÞ E E

3
e_ p ¼ Sp_ ð2Þ
2J

Y ¼ ry þ v 1 ½1  expðv 2 pÞ ð3Þ

p_ P 0; f ¼ ðJ  YÞ 6 0; _ ¼0
pf ð4Þ
with
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u 3 3
3 u3 X X
J¼ ðS : SÞ ¼ t ðSij Þ2 ð5Þ
2 2 i¼1 j¼1

where E is the young modulus, m the Poisson’s ratio and ry, v1, v2 are positive constants that characterise the plastic behav-
iour of the material. 1 is the identity tensor, r is the stress tensor and S is the deviatoric stress tensor given by
     
1
S¼ r tr r 1 ð6Þ
3
J is the von Mises equivalent stress. Y is an auxiliary variable related with the isotropic hardening. Noting the eingenvalues of
S by {S1, S2, S3}, the elastic domain can be represented in the space of the principal directions of the deviatoric stress as a
pffiffiffiffiffiffiffiffi
sphere centred at the origin with radius R ¼ 2=3Y (see Fig. 1). p is usually called the accumulated plastic strain and p_
can be interpreted as Lagrange multiplier associated to the constraint f 6 0. Function f characterises the elasticity domain
and the plastic yielding surface. From the constraint pf _ ¼ 0; it is possible to conclude that p_ ¼ 0 if f < 0 and hence e_ p ¼ 0
_
(see Eq. (2)). If p–0; _ ¼ 0 it comes that necessarily f = 0. Besides, from Eqs. (2) and (3) it is possible
from the constraint pf
p _
to verify that, in this case, e_ –0 and Y–0, Therefore, the elasto-plastic material is characterised by an elastic domain in
p
the stress space where yielding doesn’t occur (e_ ¼ 0; p_ ¼ Y_ ¼ 0 if f < 0). Generally the following initial conditions are used
for a ‘‘virgin’’ material
pðt ¼ 0Þ ¼ 0; ep ðt ¼ 0Þ ¼ 0 ð7Þ

From now on, initial conditions (7) are assumed to hold in the analysis. It is also important to remark that the evolution
law (2) with boundary condition (7) and definition (6) imply that the principal directions the stress tensor, of the deviatoric
stress tensor and of the plastic strain tensor are the same. From evolution law (2) and considering initial conditions (7), it is
possible to verify that the following relation always holds
p
Si ei
¼ 8ði; j ¼ 1; 2; or 3Þ ð8Þ
Sj epj

With Si(i = 1, 2 or 3) and epi ði ¼ 1; 2 or 3Þ being the principal components (eigenvalues) respectively of S and ep .

2.2. Tensile test in elasto-plastic specimens

In this section it is considered the tensile test of an elasto-plastic tensile specimen with gauge length L0 and cross section
A0. Assuming a quasi-static and isothermal process, the stress tensor r and the deviatoric stress S in the useful portion of the
specimen are given by

Fig. 1. Elasticity domain in the space of the principal directions of the deviatoric stress tensor.
 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140 131

0 1 0 1
rðtÞ ¼ ðf ðtÞ=Ao Þ 2
3
rðtÞ
B C
r¼B
@ 0
C
A; S¼B
@  r3ðtÞ C
A ð9Þ
0 rðtÞ
 3

where F(t) is the external tensile force and d(t) the prescribed elongation. Using Eq. (9) and definition (5), it is possible to
conclude that J ¼ jrj ¼ j AFo j and therefore, in a tensile test with monotonically increasing prescribed force the constitutive
Eqs. (1)–(4) can be reduced to
r ¼ Eðe  ep Þ ð10Þ

e_ p ¼ p_ ð11Þ

p_ P 0 f ¼ r  Y 6 0; _ ¼0
pf ð12Þ
where, in this case

e ¼ ðd=Lo Þ; ep ¼ ðdp =Lo Þ; d ¼ de þ dp ð13Þ

with de being the elastic or reversible part of d and dp the irreversible or plastic parcel of the useful length elongation d. Using
the definition of the hardening variable Y in Eq. (3), it is possible to obtain the following relation:
r ¼ ½ry þ v 1 ½1  expðv 2 ep Þ if r > ry ð14Þ
p
The maximum admissible axial stress component in a quasi-static test is obtained taking the limit of r as e ? 1. Hence

rmax ¼ lim
p
½ry þ v 1 ½1  expðv 2 ep Þ ¼ ðry þ v 1 Þ ð15Þ
e !1

Besides, the following analytic expression can be obtained

  

r ðry þ v 1  rÞ
e¼  ln ð16Þ
E v1
With r ¼ AFo and hxi = max{0, x}.
The proposed one-dimensional constitutive equations are valid for quasi-static and isothermal processes for a monoton-
ically increasing prescribed external force. Rupture in a test performed with prescribed monotonically increasing force is
quite different from the rupture in a test performed with a prescribed monotonically increasing strain. Strain softening that
occurs in a tensile test with prescribed strain is not considered by this model. It is important to remark that, due to the
particular nature of the loading history (monotonically increasing prescribed force), these constitutive equations are ade-
quate to identify the ‘‘limit axial stress’’ beyond which the hypothesis of quasi-static process is invalid and the dynamic
must be accounted since the acceleration field is no longer negligible. From the engineering point of view such stress
can be taken as the ultimate stress: the rupture process is considered instantaneous after this stress level is attained
(see Fig. 2).

2.3. Thin-walled elasto-plastic cylinders under internal pressure

In this section it is considered an elasto-plastic cylinder with internal radius ri thickness e submitted to an internal pres-
sure P. The internal radius ri and the thickness e are such that . (ri/e) > 10. The components of the stress tensor r in cylindrical
coordinates for a thin-walled cylinder undergoing a quasi-static and isothermal process are classically approximated in the
framework of membranes theory by the following expressions:
2 3
rr 0 0 (
Pri 0; for open  ended cylinders
r¼6
40 rh 7
0 5 with rr ¼ 0; rh ¼ rz ¼ ð17Þ
e Pri
2e
¼ r2h ; for closed  ended cylinders
0 0 rz

Fig. 2. Dynamic rupture in a tensile test with monotonically increasing prescribed force.
132 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140

rr is the radial stress component, rh the circumferential stress component and rz the axial stress component. All other com-
ponents are considered to be equal to zero. Since the circumferential and axial components are not independent in the case
of closed-ended pipes (rz = rh/2), additional simplifications can be obtained.
From Eq. (6), it is possible to conclude that the radial, circumferential and axial components Sr, Sh, Sz of the deviatoric
stress tensor S are its only non-zero components: Sr ¼ ðrr  r3h Þ ¼  r3h Sh ¼ ðrh  r3h Þ ¼  2r3 h ; Sz ¼ ðrz  r3h Þ ¼  r3h , for open-
ended cylinders and Sr ¼ 0  r2h ¼  r2h ;Sh ¼ rh  r2h ¼ r2h ; Sz ¼ r2h  r2h ¼ 0, for closed-ended cylinders. Using definition (5) of
pffiffi pffiffi
the von Mises equivalent stress we obtain: J ¼ jrh j ¼ Pre i for open-ended cylinders and J ¼ 23 jrh j ¼ 23 Pre i for closed-ended
cylinders.
It is important to observe that the principal components of the deviatoric stress and of the plastic deformation are not
independent, what allows introducing additional simplifications in the equations. From Eq. (9), it comes that
ðSr =Sh Þ ¼ ðepr =eph Þ and the following relations always hold: epr ¼ epz ¼  12 eph for open-ended cylinders and epr ¼ eph ; epz ¼ 0, for
closed-ended cylinders.
Using the previous results, it is possible to obtain the following set of elasto-plastic constitutive equations for thin-walled
cylinders
1
ðeh  eph Þ ¼ Pr i ; for open-ended cylinders ð18:1Þ
Ee

ð2  mÞ
ðeh  eph Þ ¼ Pri ; for closed-ended cylinders ð18:2Þ
2Ee

e_ ph ¼ p;
_ for open-ended cylinders ð19:1Þ
pffiffiffi
3
e_ ph ¼ _
p; for closed-ended cylinders ð19:2Þ
2

p_  0; f  0; _ ¼ 0 with
pf ð20Þ

Pr i
f ¼  Y  0; for open-ended cylinders ð21:1Þ
e
pffiffiffi
3 Pr i
f ¼  Y  0; for closed-ended cylinders ð21:2Þ
2 e

Y ¼ ry þ v 1 ½1  expðv 2 pÞ ð22Þ

where, in this case
eh ¼ ðd=ri Þ; eph ¼ ðdp =ri Þ; d ¼ de þ dp ð23Þ
with de being the elastic or reversible part of d and dp the irreversible or plastic parcel of the internal radius variation d.
For a monotonic loading (for instance, P = at, a > 0), the behaviour of the cylinder is elastic (eph ¼ 0) while f < 0 and, in this
case, the pressure P is related with the circumferential plastic strain component eph ¼ 0 by the following relation:
e Pri
P¼ ½ry þ v 1 ½1  expðv 2 eph Þ; if P ry ; for open-ended cylinders ð24:1Þ
ri e
pffiffiffi
2e 3 Pr i
P ¼ pffiffiffi ½ry þ v 1 ½1  expðv 2 eph Þ if P ry ; for closed-ended cylinders ð24:2Þ
3r i 2 e
The yield pressure Py is obtained taking eph ¼ 0
e
P y ¼ ry ; for open-ended cylinders ð25:1Þ
ri

2 e
Py ¼ pffiffiffi ry ; for closed-ended cylinders ð25:2Þ
3 ri
Therefore, once the geometric parameters e and ri of the cylinder are known, it can be easily verified that the yield pres-
sure Py can be obtained from the axial yield stress ry. The maximum pressure Pmax is obtained by taking the limit of P as.
eph ! 1. Hence
 
e  e
Pmax ¼ lim ry þ v 1 ½1  expðv 2 eph Þ ¼ ðry þ v 1 Þ; for open-ended cylinders ð26:1Þ
p
eh !1 ri ri
 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140 133

 
2e  p
 2 e
Pmax ¼ lim p ffiffiffi r y þ v 1 ½1  expðv 2 eh Þ  ¼ pffiffiffi ðry þ v 1 Þ; for closed-ended cylinders ð26:2Þ
eph !1 3r i 3 ri
From Eqs. (26) and (15) it can be easily verified that the maximum pressure Pmax can be related with the maximum axial
stress rmax = (ry + v1) obtained in a tensile test. Besides, the following analytic expression can be obtained
  

rh ðry þ v 1  rh Þ
eh ¼  ln ; for open-ended cylinders ð27:1Þ
E v1
2 * 2 pffiffi 3+3
ð2  mÞ ry þ v 1  2
3
rh
eh ¼4 rh  ln 4 5 5; for closed-ended cylinders ð27:2Þ
2E v1
with rh ¼ Pre i and hxi = max{0, x}.
The set of Eqs. (18)–(22) models the quasi-static and isothermal behaviour of an elasto-plastic cylinder (close-ended or
open-ended) submitted to a monotonically increasing pressure. It is important to remark that, due to the particular nature of
the pressure history, similarly to the case of a tensile test, these constitutive equations are adequate to identify the ‘‘limit
pressure’’ Pmax beyond which the hypothesis of quasi-static process is invalid and the dynamic must be accounted, since
the acceleration field is no longer negligible. From the engineering point of view such pressure can be taken as the limit pres-
sure (or rupture pressure), beyond which there is not enough time to make any repair procedure: the rupture process is con-
sidered brutal and instantaneous after this pressure level is attained (see Fig. 3). Such reasoning is very similar to the one
adopted in fracture mechanics in order to define the critical load in a cracked medium.
The proof of this fact can be obtained within a thermodynamic framework summarised in the following. In order to pres-
ent the formal theoretical arguments, it is considered as a system an arbitrary part (with thickness e, length L and internal
radius ri) of a thin-walled pipe submitted to an internal pressure P is taken as mechanical system. L is ‘‘small enough’’ to be
reasonable assuming that the internal radius variation d is the same along the length of the system.
Noting U(t), K(t), and Pext(t) and Ph(t), respectively, the internal energy, the kinetic energy, the power of the external forces
of the system at instant t, the first law of thermodynamics (FLT) for the system undergoing isothermal processes can be ex-
pressed as an energy balance:
dK dU
FLT ) ¼ P ext  ð28Þ
dt dt
Assuming that the internal energy can be decomposed in a purely elastic (or reversible) part W and an irreversible part D,
related with dissipative mechanisms (U = W + D), it comes that
dK dW dD
FLT ) ¼ Pext   ð29Þ
dt dt dt
Quasi-static processes are those in which the term dK
dt
is very small compared to ðPext  dU
dt
Þ and the FLT can be simplified to
dW dD
FLT ) Pext   ¼0 ð30Þ
dt dt
For the particular case of the pipe system under pressure, the power of the external forces has the following form:
dd dd
Pext ¼ ð2pr i LÞP ¼ ð2pLri ÞP ð31Þ
dt dt
Besides, from the constitutive equations, it is possible to verify that, for a monotonically increasing pressure, the irrevers-
ible part D of the free energy U is such that
dD dd
D ! ð2pr i LÞcd; when P ! Pmax : Hence ! ð2pr i LÞc ; when P ! Pmax ð32Þ
dt dt
where c is a positive constant that depends on the material properties. It is also possible to conclude that the elastic part W of
the free energy U is such that

Fig. 3. Dynamic rupture in a hydrostatic test with monotonically increasing pressure.

134 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140

dW
! 0 when P ! Pmax ð33Þ
dt
From Eqs. (30)–(33), it comes that the following condition must hold in a quasi-static process
dd
ðP  cÞ ! 0 when P ! pmax ð34Þ
dt
Since the term dddt
is greater than zero, the previous relation is impossible to be obtained for a monotonically increasing
pressure. Therefore the quasi-static version (30) of the FLT is not valid when P ? Pmax since the variation of the kinetic en-
ergy cannot be neglected. In other words, this means that it is impossible to have a quasi-static process when P ? Pmax. For a
monotonically increasing loading history, Pmax is then the ‘‘limit pressure’’ beyond which the hypothesis of quasi-static pro-
cess is invalid and the dynamic must be accounted since the acceleration field is no longer negligible. From the engineering
point of view it is reasonable to take such limit as the onset of pipe rupture, and hence, as the ultimate pressure.

3. Modelling the effect of small temperature variations of the water in long-term hydrostatic tests

The rehabilitation of corroded pipelines with epoxy repair systems is becoming a well accepted engineering practice and
an interesting alternative to the classical repair methods for metallic pipes, mainly in the oil industry, saving time and allow-
ing safer operations.
In these repair systems, a piping or vessel segment is reinforced by wrapping it with concentric coils of composite mate-
rial after the application of epoxy filler in the corrosion defect. Corroded pipelines with through-walls defects can also be
repaired with different epoxy systems (see [1], for instance). In this case, the challenge is to assure a satisfactory level of
structural integrity and to assure that the pipe wont leak after repair.
In the case of through-walls corrosion defects in produced water pipelines used in offshore oil exploitation, the water
temperature is a major shortcoming for the use of polymeric material as repair systems. In this case, it is usual to perform
long-term hydrostatic tests (between 1000 and 2000 h) with constant internal pressure and temperature (between 353 K
and 363 K) to check the repair effectiveness to avoid leaking.
Long-term hydrostatic tests at this temperature range may be performed whether using a pool with hot water in which
the specimen in immersed or a system especially designed for this procedure (see [1]) in which the whole system (including
the electrical resistance) is threaded at one extremity of the specimen.
Normally, firstly temperature is raised up to a given fixed level and then pressure is applied. After the desired pressure
limit is reached, the pressure connection is closed using a valve and, since the pressure is supposed to be fixed, the only ac-
tive control system aims to assure that temperature remains constant. Generally these control systems allow very small
oscillations around the temperature reference. No matter the long-term hydrostatic test apparatus adopted, to accurately
control temperature without extensive operator involvement, a temperature control system relies upon a controller, which
accepts a temperature sensor such as a thermocouple or RTD as input. It compares the actual temperature to the desired
control temperature, or set point, and provides an output to a control element. The controller is one part of the entire control
system, and the whole system should be analysed in selecting the proper controller. There are a few different types of con-
trollers but the simplest and most usual in this kind of testing is the ‘‘on-off’’ controller. Generally the temperature oscilla-
tions are small compared to the control temperature (set point). Nevertheless, it is verified experimentally that, in a closed
vessel (the mass of fluid is constant inside it) with liquid water at high pressure, a small variation of the temperature above
the desired control temperature may cause a huge pressure variation. In the case of a long-term test in a damaged pipe re-
paired with an epoxy system, very small variations of temperature due to the control systems may cause a pressure peak
leading to failure. Generally this kind of the pressure peak is very fast and eventually is not recorded, what may cause mis-
taken conclusions about the durability of the epoxy repair system. The failure of the repair may be caused by pressure var-
iation and not because of the temperature effect on the polymer behaviour.

3.1. Summary of the equation of state for liquid water at high pressures

The goal of this section is to present an equation of state for liquid water at high pressures. The proposed equation is a
generalisation of Tait equation of state to include very small temperature variations. Tait equation of state for compressible
liquids (or called the Murnaghan equation of state in the context of an elastic solid [4–6]) models a liquid such as water as a
compressible, barotropic liquid whose bulk modulus is an affine function of pressure. Hence, this equation of state involves
only the density and pressure variables. However, it is a highly non-linear equation of the form
 c  ð1=cÞ
q PþB
P ¼ ðP o þ BÞ  B ) q ¼ qo ð35Þ
qo Po þ B
where P and q denote the pressure and the density, respectively. Po and qo are the pressure and density at a reference state. B
and c are positive parameters.
Two alternative constitutive equations are suggested to extend the state law (35) in order to account very small temper-
ature variations from a reference temperature ho
 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140 135

 ð1=cÞ  c
q PþB q
¼ ½1  gðh  ho Þ ) P ¼ ðP o þ BÞ ½1  gðh  ho Þ1 B ð36:1Þ
qo Po þ B qo
 ð1=cÞ  c
q PþB q
¼  gðh  ho Þ ) P ¼ ðPo þ BÞ þ gðh  ho Þ  B ð36:2Þ
qo Po þ B qo
With g being a positive pressure dependent parameter.

4. Results and discussion

4.1. Burst tests

Eqs. (25.1), (25.2), (26.1), and (26.2) show that, in order to obtain a preliminary estimative of both yield pressure and
burst pressure for a given cylinder with internal radius ri and wall thickness e (with ri P 10e), it is only necessary to know
the yield stress ry and the ultimate stress rmax obtained experimentally in a tensile test of the pipe material. Besides, Eqs.
(25.1), (25.2), (26.1), and (26.2) show that both the yield pressure and burst pressure in a close-ended cylinder are p2ffiffi3 times
higher (1.155 times – 15.5%) than in an open-ended cylinder.
Long straight thin-walled pipelines are reasonably modelled by the open-ended cylinder model. Hydrostatic specimens
are better modelled by the closed-ended cylinder model. Hence, both yield pressure and burst pressure of a metallic spec-
imen measured in a laboratory test tend to be higher than the ones of a long straight thin-walled pipeline with same mate-
rial, diameter and wall thickness. Despite the scatter of the experimental results, it may be interesting to take into account
this theoretical fact in the integrity analysis of a real structure.
This section aims at comparing the model prediction of the burst pressure of close-ended cylindrical pressure vessels with
some experimental results. The following expression to compute the burst pressure is found at the ASME Boiler and Pressure
Vessel Code [3]
xe
Pmax ¼ ðrmax Þ ð37Þ
r i þ 0:6e
where rmax is the rupture stress obtained in a tensile test; x (0 < x 6 1) is the efficiency of the welded joint = 1 (for seamless
pipe); ri is the internal radius (in) and e the wall thickness (in). From now on, the burst pressure of a specimen will be
approximated by using two different approaches: (i) multiply the rupture stress obtained in tensile test by a factor ri þ0:6e xe
2x effiffi
(ASME standard) or (ii) multiply the rupture stress obtained in tensile test by a factor r 3 (proposed elasto-plastic model).
p
i
Hydrostatic tests performed by Loureiro [16] in seamless pipes were chosen as a preliminary step to evaluate the ade-
quacy of the model prediction of the burst pressure. The internal radius and wall thickness of the pipe were, respectively,
ri = 64.01 mm and e = 6.55 mm and pipe material was an aluminium alloy SB-241 6063-T5.
The average value for rmax was 177.62 MPa, obtained after five tensile tests performed according to ASTM B557 standard
[17]. In these burst tests, pressure was gradually raised until rupture. Table 2 presents the experimental and predicted burst
pressures. The burst pressure for this seamless pipe predicted using ASME standard Eq. (37) is, about 82.5% of the pressure
measured experimentally (respectively 17.13 MPa and 20.76 MPa). The burst pressure prediction using the elasto-plasticity
model for an open-ended cylinder (Eq. (26.1)) is 18.18 MPa, about 87.6% of the pressure measured experimentally. The burst
pressure prediction using the elasto-plasticity model for a closed-ended cylinder (Eq. (26.2)) is 20.99 MPa (1.0% above than
experimental). The ASME code predictions of the burst pressure using Eq. (37) and the elasto-platic prediction considering an
open-ended cylinder (Eq. (26.1)) are more conservative than using Eq. (26.2) (closed-ended cylinder). Predictions using
expression provided by the ASME code are more adequate for long lines which can be modelled as open-ended pipes.
Due to the scatter of experimental results, the burst pressures obtained using the ASME code and Eq. (26.2) may eventually
be used, respectively, as lower and upper limits for designers.
Burst tests of rocket motor cases made performed by Beena et al. [13] were also used to check the adequacy of the model
prediction. The internal radius wall thickness of the pipe were ri = 103.3 mm and e = 2.6 mm and pipe material was a 15CDV6
steel with rmax = 1010.0 MPa. The name 15CDV6 is a French designation in which the first number is equal to 100 times the
concentration of carbon. The letters which follow indicate the other elements present, in the decreasing order of concentra-
tion. The last digit is equal to four times the concentration of chromium. In this French notation, C stands for chromium, D for
molybdenum, and V for vanadium. It therefore follows that, in 15CDV6 steel, the concentration of carbon is 0.15%, while that
of chromium is 1.5%, and the concentrations of molybdenum and vanadium are less than 1.5% each. Hence, it is a low-alloy
steel, in which the proportion by weight of all the alloying elements combined is less than 5%.
The experimental burst pressures obtained in two tests were 28.86 MPa and 29.23 MPa (average 29.05 MPa). The burst
pressure predicted using ASME code (Eq. (37)) is 25.48 MPa. The predicted burst pressure using the elasto-plasticity model
for an open-ended cylinder (Eq. (26.1)) is 25.42 MPa. The burst pressure prediction using the elasto-plasticity model for a
closed-ended cylinder (Eq. (26.2)) is 29.39 MPa (1.15 times the ASME standard) which is in good agreement with the two
burst test results of the rocket motor cases. The experimental 0.2% yield stress measured in a tensile test was
915.27 MPa. The experimental yield pressure was 26.49 MPa. The predicted yield pressure using the elasto-plasticity theory
using Eq. (25.2) is 26.63 MPa.
136 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140

As it can be verified, once again ASME predictions of the burst pressures are closer to the ones obtained using the elasto-
plasticity theory for open-ended cylinders. ASME predictions are more adequate for long lines that can better be modelled as
open-ended cylinders.
In the analysis it was adopted and efficiency x = 1, but the cases are not seamless pipes and the authors in [13] affirm that
the weld efficiency is 93.5–98% (0.935 6 x 6 0.98). In this specific case it is recommended, in analogy with the ASME code, a
correction factor due to the welded joint.

2w e
Pmax ¼ pffiffiffi ðrmax Þ ð38Þ
3 ri
With this correction, the burst pressure predicted using the elasto-plastic model is even closer than the obtained
experimentally.
Finally, results of burst tests in maraging steel rocket motor cases performed in [18] and presented in [14] were consid-
ered. The internal radius of the pipe was ri = 45 mm and the maximum strength rmax = 2155 MPa. The tests results for five
different values of the wall thickness are presented in Table 1 (each column corresponds to a different test).
The difference between the predicted results from elasto-plasticity theory and experimental is higher. However it is
important to explain that the steel pipes used to perform those tests were not seamless and the efficiency x of the welded
joint is lower than 1. In this specific case it is suggested, in analogy with the ASME code, a correction of approximately
x = 0.96 (efficiency of 96%).
Table 1 presents a summary of the results (experimental and predicted burst pressures) presented in this section.

4.2. Long-term hydrostatic tests

Long-term hydrostatic tests were performed with a constant internal pressure P0 = 10.34 MPa [1500 psi] at 353.15 K. An
experimental set-up was conceived to check the effectiveness of the methodology, trying to approximate a real repair oper-
ation. 50.8 mm (200 ) diameter with 5.54 mm wall thickness API 5L grade B steel pipes, normally used in offshore platform for
produced water, with a circular hole of 10 mm were used to build the specimens for hydrostatic tests. The material proper-
ties are: E = 182.00 GPa; ry = 449.00 MPa, v1 = 177.20 MPa; v1 = 100.00 MPa. Using Eq. (26.1), it is possible to obtain that, for
an undamaged pipe Pmax > 100 MPa while the mean failure pressure for the repair is Pf = 17.2 MPa. Burst of a repairing pipe
leads to lower burst pressure due to mismatch effect between repairing and pipe materials. Typical local failure modes of the
epoxy repair are shown in Fig. 4.
The epoxy repair system is described in details in [1]. A commercial fast curing polymer-based system (called System II in
[1]) consisting of a mixture of epoxy resin and aluminium powder was used to perform the repair of the through-thickness
damage in each specimen. The whole system to control water temperature (including the electrical resistance) is threaded at
one extremity of the specimen. Figs. 5 and 6 show the basic set-up apparatus and temperature control system. The control
system allows a 1 K temperature variation. Besides the pressure control system, a pressure relief valve relieves liquid pres-
sure above a pre-set limit. Therefore, the mass of fluid inside the specimen may vary in order to assure a constant pressure.
Fig. 7 shows a curve pressure versus time in a typical test in a closed system (fixed mass of fluid). It can be verified peaks
of pressure even with such a limited temperature variation. Rupture has occurred after 6 days, when a pressure peak much
above the static strength (Pf = 17.2 MPa) of the repair occurred. The pressure peaks can be explained using constitutive Eqs.
(36.1) and (36.2) that are proposed for small variations of temperature of liquid water at high pressures. Considering the ini-
tial state: Po = 10.34 MPa; qo = 958 kg m3; ho = 353.15 K, a reasonable approximation for the parameters B and c that appear
in Tait equation [5] is B = 2.9  108 kg m1 s2 and c = 7.15. In order to identify the additional parameter g that appears in
Eqs. (36.1) and (36.2) it is interesting to observe the curve presented in Fig. 8, showing how density varies with pressure for a
constant pressure of 0.1 MPa (1 atm). Although curve q  h shown in Fig. 8 presents a non-linear behaviour, it can be approx-
imated by a linear function in the neighbourhood of the reference temperature ho (see Fig. 9). g is a pressure dependent
parameter. Hence, assuming small temperature variations from the reference temperature ho, it is reasonable to assume a
constant value g = 6.48  104 K1.

Table 1
Experimental and predicted burst pressures.

Reference Ref. [16] Ref. [13] Ref. [14]

Alloy Aluminium SB-241 6063-T5 15CDV6 steel Maraging steel
Internal radius (mm) 64.01 103.3 45
Wall thickness (mm 6.55 2.6 1.630 1.735 1.756 1.763 1.793
Experimental burst pressure (MPa) 20.76 29.05 86.62 92.30 94.50 94.00 94.00
Model prediction (MPa). 20.99 29.39 90.13 95.94 97.10 97.49 99.14
Closed-ended cylinder – Eq. (26.2)
Model prediction (MPa) 18.18 25.42 78.06 83.09 84.09 84.43 85.86
Open-ended cylinder – Eq. (26.1)
ASME prediction (MPa) – Eq. (37) 17.13 25.48 76.72 81.28 82.23 82.56 83.93
 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140 137

Fig. 4. Initial through-the-thickness axial crack.

Fig. 5. Basic test set-up apparatus.

In the case of a closed vessel (constant mass of fluid), the density q tends to increase when a higher pressure is applied:
ðqq Þ > 1 Table 2 presents the pressure variation obtained using Eqs. (36.1) and (36.2) and assuming. ðqq Þ  1 The pressure
o o
predicted using Eqs. (36.1) and (36.2) are very close, what means that, for a very small temperature variation from the ref-
138 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140

Fig. 6. Temperature control system. 1 is the pressure water machine connection, 2 is the temperature control thermostat and 3 is the electrical resistance.

Fig. 7. Typical pressure versus time curve at average pressure of 10.34 MPa [1500 psi] and temperature of 80 °C. The temperature control system allows a
1 °C variation. Test performed in a closed system.

Fig. 8. Curve q  h. Water at a constant pressure of 0.1 MPa (1 atm).

 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140 139

q
Fig. 9. Detail of the curve qo  ðh  ho Þ obtained from Fig. 6. ho = 80 °C.

Table 2
Pressure variation with temperature from the reference state, using Eq. (36).

Temperature (°C) Pressure Eq. (36.1) (MPa) Pressure Eq. (36.2) (MPa)
80.0 10.34 10.34
80.5 16.98 16.98
81.0 23.64 23.63

Fig. 10. Curve (P  Po)  (h  ho) obtained using Eq. (36.1) and considering very small temperature variations.

erence temperature, it is almost indifferent to consider any of them. For very small temperature variation, the pressure var-
iation (P  Po) increases almost linearly with (h  ho) as shown in Fig. 10.
The mean failure pressure measured in hydrostatic tests in which a monotonically increasing load is applied until the fail-
ure of the repair is Pf = 10.1 MPa. As it can be verified in the test shown in Fig. 7, the internal pressure in the specimen
reached the limit pressure Pr due to the temperature oscillation admitted by the control system (always smaller than
1 K). Therefore, the prediction of a strong pressure variation made in Table 2 for closed systems is confirmed experimentally
what indicates that a very small temperature variation (around 0.5 K) can eventually lead the repair to failure and that the
use of an adequate pressure control system is essential in order to avoid big pressure oscillations during the testing. All spec-
imens repaired with the same epoxy system and tested under the same pressure and temperature conditions (five speci-
mens) but using the pressure relief valve and active pressure control did not fail after 2000 h.

5. Concluding remarks

The present paper uses elasto-plastic constitutive equations and also a thermodynamic framework to analyse special
hydrostatic tests developed for the evaluation of the structural integrity of produced water pipelines used in offshore oil
exploitation. Besides, simple expressions are proposed to estimate the yield and burst pressures of the specimens submitted
to hydrostatic tests, only requiring the knowledge of the yield stress and the ultimate stress obtained in tensile tests with the
pipe material.
It is also important to remark that wrong conclusions can be made if a closed system is adopted in long-term hydrostatic
tests of epoxy repair systems performed in specimens at temperatures between 353 K and 363 K. It is verified both exper-
imentally and theoretically that, if the specimen is completely closed and the mass of fluid is constant inside it, a very small
temperature variation may induce a huge pressure variation, eventually leading the repair to failure. Generally this kind of
the pressure peak is very fast and eventually is not recorded, what may cause mistaken conclusions about the durability of
the epoxy repair system.
140 H.S. da Costa Mattos et al. / Engineering Failure Analysis 22 (2012) 128–140

References

[1] da Costa Mattos HS, Reis JML, Perrut V, Sampaio RF. An alternative methodology to repair localized corrosion damage in metallic pipelines with epoxy
resins. Mater Des 2009;30:3581–91.
[2] Veil JA, Puder M, Elcock D, Redweik Jr. RA. White paper describing produced water from production of crude oil, natural gas, and coal bed methane,
prepared by Argonne National Laboratory, Argonne, Illinois for the US Department of Energy, National Energy Technology Laboratory; 2004. <http://
www.evs.anl.gov/pub/dsp_detail.cfm?PubID=1715>.
[3] ASME Boiler and Pressure Vessel Code – Section VII – Division 1, 1992 ed. 1995 addenda.
[4] Murnaghan FD. Finite deformation of an elastic solid. New York: Wiley; 1951.
[5] Shyue KM. A fluid-mixture type algorithm for barotropic two-fluid flow problems. J Comput Phys 2004;200:718–48.
[6] Farhat C, Rallu A, Shankaran S. A higher-order generalized ghost fluid method for the poor for the three-dimensional two-phase flow computation of
underwater explosions and implosions. J Comput Phys 2008;227:7674–700.
[7] Durban D. Large strain solution for pressurized elasto-plastic tubes. Trans ASME J Appl Mech 1979;46:228–30.
[8] Chistopher T. A comparative study on failure pressure estimations of unflawed cylindrical vessels. Int J Pres Ves Pip 2002;79:53–66.
[9] Durban D. A finite strain axially symmetric solution for elastic tubes. Int J Solids Struct 1988;24:675–82.
[10] Durban D, Kubi M. A general solution for the pressurized elasto-plastic tube. Trans ASME J Appl Mech 1992;59:20–6.
[11] Law M, Bowie G. Prediction of failure strain and burst pressure in high yield-to-tensile strength ratio linepipe. Int J Pres Ves Pip 2007;84:487–92.
[12] Xian-Kui Z, Leis BN. Evaluation of burst pressure prediction models for line pipes. Int J Pres Ves Pip 2012;89:85–97.
[13] Beena AP, Sundaresan MK, Nageswara Rao B. Destructive stress of 15CDV6 steel rocket motor cases and their application to lightweight design. Int J
Pres Ves Pip 1995;62-3:313–20.
[14] Subhananda Rao A, Venkata Rao G, Nageswara Rao B. Effect of long-seam mismatch on the burst pressure of maraging steel rocket motor cases. Eng Fail
Anal 2005;12-2:325–36.
[15] da Costa Mattos HS, Soares Filho PF. Approximate local elasto-plastic solution for notched plates undergoing cyclic tensile loading. Mater Des
2010;31:4336–47.
[16] Loureiro SMNA. Reinforcement of cylindrical pressure vessels using polymeric composites (in Portuguese). MSc thesis, COPPE – Universidade Federal
do Rio de Janeiro; 1999.
[17] ASTM Standart B557. Standard test methods for tension testing wrought and cast aluminum- and magnesium-alloy products.
[18] Margetson J. Burst pressure predictions of rocket motors. AIAA Paper No.78-1567. AIAA/SAE 14th Joint Propulsion Conference Las Vegas, NV USA;
1978.