Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug.

2003
www.Ommi.co.uk

A review of the finite element analysis of repaired welds under creep conditions

T. H. H YDE*, J. A. WILLIAMS**, A. A. BECKER* and W. SUN*

*School of Mechanical, Materials, Manufacturing Engineering & Management,

UNIVERSITY OF NOTTINGHAM, Nottingham NG7 2RD, UK
**Independent Consultant, East Leake, Leicester LE12 6LJ, UK

Profe
Pro Professor Tom Hyde is the Head of School of Mechanical, Materials, Manufacturing
Engineering and Management and Director of the Rolls Royce UTC in gas turbine
transmission systems at the University of Nottingham. He has around 30 years
experience in the application of analysis to the high temperature deformation and
creep of materials and welds/weld repair and his work has close links with the UK
power industry. (Thomas.Hyde@nottingham.ac.uk)

Eur Ing Dr Adrian Williams has been an independent consultant since 1992, acting
for ETD and other companies. In addition, he is an Industrial Fellow at the School
of Mechanical, Materials, Manufacturing Engineering and Management at the
University of Nottingham. He has 27 years experience in the UK power industry,
up to 1992, in the fields of high temperature materials, including similar and
dissimilar weld performance, component testing, life assessment and weld repair.
(jaw@globalnet.co.uk)

Dr Wei Sun is a Senior Research Fellow in the Structural Integrity and Dynamics
Research Group at the University of Nottingham. He has been working in the field
of high temperature creep since 1993, with emphasis on creep testing, material
properties generation and failure assessment of components (including welds) at
elevated temperature, using the finite element method. (w.sun@nottingham.ac.uk)

Professor A.A. Becker is Professor of Mechanical Engineering and Head of the

Structural Integrity and Dynamics Research Group in the School of Mechanical,
Materials, Manufacturing Engineering and Management at the University of
Nottingham. He has worked in the field of theoretical and computational
mechanics for 24 years and has supervised many research projects in non-linear
computational mechanics, high temperature analysis, fracture and structural
integrity, funded by EPSRC, EU and industry. He has published 127 papers in the
open literature and two textbooks, one on Boundary Element Methods (McGraw-
Hill, 1992) and one on Non-linear Finite Element Analysis (Nafems, 2001).
(a.a.becker@nottingham.ac.uk)

This Paper was first presented at the 3rd International 'HIDA & Integrity Conference, 16-18
September 2002, Lisbon, Portugal.

Abstract

This paper briefly reviews work related to the finite element (FE) analyses of repaired welds, under
creep conditions; this has been mainly focused on the pressurised pipe weldments made of CrMoV
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 2

steels. Weld repair models, creep and damage constitutive equations and methods used for
determining the material properties are briefly described. Typical results for the stress analysis and
life prediction of the repaired welds are presented, covering the effects of material mismatch, repair
profile/dimension, system loading etc., on the stress distributions, damage accumulation and life
assessment. The potential uses and limitations of FE modelling for predicting the performance of
weld repairs are assessed and discussed.

1. Introduction

It is often necessary to carry out weld repairs on components operating at elevated temperature in
power plant in order to ensure their long term integrity [1 ]. In-service inspections of the main steam
piping systems in power plant can often reveal cavitation and cracking, in or in the vicinity of the
weldments. If damage is severe, the components may be replaced. However, in many
circumstances, damage is local and therefore weld repairs can be made in order to reduce the cost
whilst maintaining safety. Experience from the performance of repaired welds has shown that creep
damage is often found in or close to the repair earlier than would have been expected [2]. Extensive
effort has been made to optimise weld repair techniques and to improve the high temperature
performance of weld repairs [e.g. 3, 4].

A typical weld repair procedure includes mechanical removal of the damaged material and the
restoration of the component geometry by welding followed by a post weld heat treatment, which is
intended to locally temper the hard heat-affected zone (HAZ) structure and to reduce the residual
stresses caused by welding. The process of repair welding will create new HAZs, some of which will
be in the original parent, in the weld and in the HAZ materials. Two types of weld repair are
commonly used, i.e. full repair and partial repair.

The major factors which affect the creep behaviour of repaired welds are the relative material
properties for the various zones, repair profile/dimension, and the loading system. Due to the
complex nature of the problem, the finite element (FE) method is usually used to assess the effects
due to these factors and to predict the failure lives of the repaired welds [2,5 ]. Two main types of
material behaviour models are often used in the FE modelling, i.e. simple power law creep
constitutive laws and continuum damage mechanics constitutive equations. Before accurate
modelling can be performed, the material properties for each material zone must be determined.

This paper presents a review of creep and damage FE analyses of repaired welds. The work has
mainly focused on repaired welds in pressurised circumferentially welded pipes made of CrMoV
steels using conventional excavations. Typical results on the stress analyses and failure predictions
are described, covering a range of analyses related to the effects of material properties, repair
profile/dimension, system loading etc, on the creep behaviour of repaired welds. For completeness,
some work on the stress analysis of weld repairs on branch connections is also included. The
potential uses and limitations of FE modelling on weld repair performance assessment are discussed.

2. Geometric and Material Models for Repaired Welds

2.1 Weld Repair Models

Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 3

Schematic diagrams of some typical weld repair geometries are shown in Fig. 1. A typical weld in a
component consists of parent material (PM), heat-affected zone (HAZ) and weld metal (WM), Fig.
1(a). The material models for a new and an aged weld are shown in Figs. 1(b) and 1(c).

w 7 w=80 7

PM WM PM 2
h T
damage 4 4

HAZ
D (d) full repair
(a) V-weld with damage
3 w=46 3 7 w1=80 7

2 9 2 8
b

1 1 4 5 6 4
6

(b) new weld (e) partial repair i

6 w=46 6 7 w2=40 8

9 2
5 b

4 4 4 6 5 6 4

(c) aged weld (f) partial repair ii

Fig. 1: Typical geometric models: material zones and repair profiles [19]. [ 1 - new PM; 2 - new WM; 3 - new
HAZ (in new weld); 4 - aged PM; 5 - aged WM; 6 - aged HAZ (in aged weld); 7 - new HAZ in aged PM; 8 -
new HAZ in aged WM; 9 - new HAZ in aged HAZ] (Dimensions (mm): D = 355.6, T = 63.5, h = 4, = 15o
and b = T/2)

Figs. 1(d) to 1(f) show two possible weld repair techniques used to remove a typical HAZ crack
running across the wall thickness. In one case, Fig. 1(d), the whole of the original weld and
surrounding parent material is removed (full repair). In the other cases, Figs. 1(e), prw1, and 1(f),
prw2, only part of the original weld is removed (partial repair). The partial weld replacement
technique has been generally used in the past but the full weld repair technique is now used in the
great majority of cases in the UK [6].

It is important to note that when a partial repair of the prw2 type is used, this will generate a new
HAZ, in the previously service exposed weld metal. A similar effect is found if the excavation is
completely within the weld metal. In all other cases given here, any new HAZ is formed in the service
exposed parent pipe material. Furthermore, the full weld repair, which involves the removal of the
weld metal, is equivalent, in terms of geometry, to a virgin weld but with a wider weld. This is
important and allows the results of stress analysis on conventional welds to be used to give some
guidance for weld repair assessment for some practical geometries.

2.2 Material Behaviour Models

Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 4

Two main types of material behaviour models are often used in repaired weld modelling, i.e. power
law creep laws and continuum damage constitutive equations. Continuum damage equations of the
following type [7]:
3 eq S ij m
n

& = A
c
t (1)
2 1 eq
ij

M r
and & =
tm (2)
(1 + )(1 )

where r = 1 + (1 - ) eq (3)

have been used in the FE modelling of repaired welds [e.g. 5 ], in which eq and 1 are the equivalent
and maximum principal stresses, respectively, and S ij is the deviatoric stress. is the damage
variable which varies from 0 (no initial damage) to 1 (failure) and A, m, n, M, , and are
material constants. Since precise material properties, especially for the HAZ material, and also for
the FE damage codes, are not widely available, in many cases, the results of steady-state analyses
are used as a simplified life prediction approach. Steady-state creep solutions, using Nortons creep
law, i.e.

3
& cij = A' neq' 1 Sij (4)
2
can be obtained using commercial FE codes. The failure life can be estimated using the steady-state
peak rupture stress, rp, and the appropriate creep rupture material properties [e.g. 8 ], i.e.
1 /(1+m )
1+ m
tf = p
(5).
M ( r )

There are other forms of constitutive laws which can be used to characterise the full creep curves,
including tertiary creep. Typical examples of these are the two or more damage parameter forms of
Hayhurst and co-workers, [9] and the THETA approaches developed by Wilshire and Evans, [10].
Although these have been used for the analysis of specific homogeneous and weld situations, the
published data have not generally been applied to repair welds. For this reason, this review is limited
to the use of steady state and a single parameter continuum damage law.

3. Determination of Material Properties

In FE modelling, a weld is usually assumed to consist of a number of distinct material zones, each of
which has constant material properties. One of the main difficulties in weld or repaired weld
modelling is the determination of the material properties for the various zones involved. Novel, non-
standard test techniques, such as impression creep [11 ], or small punch disc [12] tests, can be
employed to determine some of the HAZ properties, the latter also giving some rupture data. An
alternative method, using simulated HAZ material, has been used for conventional uniaxial creep
testing [e.g. 13 ].

For the one state variable damage equations, Equations (1-3), the material constants A, n, m, B,
and for parent and weld materials can be determined from conventional uniaxial creep test data,
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 5

while the tri-axial parameter for parent and weld materials can be determined from the data of bi-
axial tests or by matching the results from FE damage modelling with data from notched bar rupture
tests, using the constants A, n, m, B, and obtained from uniaxial tests data [14-16]. For HAZ
materials, constants A, n and m can be obtained from the results of impression creep tests, while
constants M, , and can be determined by matching the results of FE damage modelling of
cross-weld rupture tests [14 ], with the corresponding experimental data, using the A, n and m values
obtained from impression creep tests. Similar, though more complex, processes exist to define the
constants in any of the alternative constitutive law forms, for example see [10,17 ].

4. Typical Results of FE Assessment of Repaired Welds

There are only a small number of papers which include FE modelling results on the creep analysis of
repaired welds. The work reviewed here has focused on circumferential pipe welds [2,5,18-20]
made of CrMoV steels. These studies have mainly involved stress and damage analyses, and failure
prediction, in which the effects of material mismatch, repair dimensions/profiles, system load etc., on
the predicted stresses and failure times, are investigated. However, for the case of a full weld repair,
which has a similar geometry to that of a virgin weld, it is possible to use analysis results which have
been obtained for virgin welds.

In all cases, the repair welds are assumed to be post weld heat treated and hence should contain low
residual stresses. It is accepted that advances in welding technology and procedure control have
allowed the use of repair welds in the as-welded state to be considered as a viable possibility.
However, the magnitudes and distributions of residual stress are not known in detail at this stage. For
this reason, the residual stress effects are not considered here although some work has been
published [e.g. 21 ].

4.1 Steady-State Creep

4.1.1 Effect of material properties

As with a normal circumferential weld, the hoop stress distributions within a repaired weld, in a
pressurised pipe, usually show off-loading behaviour, [22 ], i.e. the lower stress occurs in the weaker
material zones and vice versa, where the degree by which a material is stronger or weaker than
another material is defined by the ratio of the secondary creep rates of the relevant materials [e.g. 2].
Therefore, the representative stresses, such as the peak rupture stresses in each of the material
zones, which are used for life estimations, Equation (5), are functions of relative material creep
properties. Generally, in practice, the repair weld metal will be the same as that which was used for
the primary weld fabrication. Hence, in the weld repair state, where both the original parent and weld
metal have been exposed to the service conditions of pressure and temperature, the repair weld
metal will generally be stronger, in creep, than the exposed parent material and weld metal.

In the work by Samuelson et al [2], the peak equivalent and maximum principal stresses within
partially repaired welds (case i, see Fig. 1), in a 1/2CrMoV pipes, with different relative material
properties, were investigated. Both the original and the new WM were 2 1/4Cr1Mo ferritic steels.
Their results have shown that for an originally creep-soft weld with a creep-hard repair, the peak
stresses occur in the new WM, while for an originally creep-hard weld with a creep-soft repair, the
peak stresses occur in the old WM, where creep-soft or creep-hard refer to the relative strain rates
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 6

of the WM to that of the PM. Similar behaviour was observed from analyses by Hyde et al [ 5 ] and
Sun et al [19], for a number of new, aged and repaired welds, in a thick walled 1/2Cr1/2Mo1/4V
pipe, at 640 o C, with an internal pressure pi = 16.55 MPa; the wall thickness, T, and outer diameter,
D, of the pipe are 63.5 mm and 355 mm, respectively. Both the original and the repair weld metals
are 2 1/4Cr1Mo steel. It has been shown that both the peak stress and its position are sensitive to
the relative material properties. For a more particularly partial repair case (case-ii, Fig. 1(f)), simply
changing the relative properties of specific zones, by modifying the A ' value in Equation (4), which
can be thought of as incorporating a reduction in the ductility, following Samuelson, [2 ], can lead to
changes in both the magnitude and position of the failure dominant stress, as shown in Table 1. For
example, the effect of property changes in zone 7, the HAZ in the service exposed parent material,
Fig. 1(f), relative to zone 8, the HAZ in the service exposed weld metal, are considered. It is clear
that as the strain rate ratio, of zone 7, relative to zone 8, is reduced, the peak stresses focus into zone
8. Such behaviour has been identified in practice, [23 ], and in full size pressure vessel tests, [24 ].

Table 1 Effect of A ' value for zone (8), on the failure dominant peak rupture stress, r , and its
p

position for a partially repaired weld (case-ii), see Fig. 1 [19] (internal pressure only; other
material constants for zone (8) and zone (7) are the same)

A ' (zone 8) rp/pi Position zone (8) compared with zone (7)
(zone 7)
= A' 1.82 HAZ (7 or 8) same
> A ' (zone 7) 1.82 HAZ (7) higher in failure ductility
= 0.5 A ' (zone 7) 2.01 HAZ (8) lower in failure ductility
= 0.2 A ' (zone 7) 2.28 HAZ (8) lower in failure ductility

For the particular case of a full repair weld, frw, the main geometry difference is the weld width,
which would typically be an extra 10 mm per side. It has been shown that there is only a small effect
of weld width, within limits, on the local stresses in the weld, for a given set of material properties,
[25-28 ]. Thus, it is possible to incorporate data from more general weld analysis programmes in
repair weld studies, although there may be differences in the way that the results are presented. Using
the creep-hard, creep-soft definitions given earlier, the general concepts of hoop stress off-loading
and no off-loading, within limits, of the axial stress components, for pure pressure loading, are
confirmed, even for more complex geometries. The effect of additional axial loading is confirmed,
[20,27,28 ], as being important for both the magnitude of the stresses and the position at which the
maximum stress occurs. In general, the position of maximum stress moves out to the HAZ/PM
interface with increasing additional axial load.

A number of workers, [27,28 , for example], have carried out parametric studies to examine the
effects of specific variables on the peak stresses and these will have generally involved a large
number of separate finite element analyses. An alternative method, proposed by Hyde et al, [29],
which incorporates a multi-material formulation, using a Norton law of the form & / &o = ( / nom )n ,
can be used for the parametric stress analyses of welded components, to evaluate the effects of
variations in material properties. At a position of interest in material i, of a p-material component, the
general form for the stress, i, is given by
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 7

1 ni

p
& nj
i n j
f j (n 1 , n 2 ,..., n p ,[ dim]j )
oi
= 1 (6)
j =1

&
oj nom

where & o1 , & o 2 , , & op and n1, n2, , np, are material constants and nom is a conveniently defined
nominal stress. f1, f2, ...., fp are functions of stress indices, n1, n2, , np, and non-dimensional
functions of dimensions, dim. It can be seen that the effects of & oi / & o1 , & oi / & o 2 , ..., & oi / & op are
explicitly defined. Knowing the fi values for a particular n-set, the corresponding i values can be
determined from Equation (6). Such relationships have been directly applied to simplify the
parametric analysis of welded components, which would otherwise require a huge number of FE
calculations to assess the effects of material property variations on the stresses. An example, showing
the effect of material properties on the equivalent stress, in a three-material pipe weld, is presented in
Fig. 2. Details on this approach and its applications have been published [ 29,30 ].

Fig. 2: Variations of eq/pi in the type IV zone, at the outer surface, with & o 2 / & 01 , for a range of
& o 2 / & 03 , for n = 5. (D = 355.6, T = 63.5, h = 4, w = 46 (mm), = 15 [30])
o

There is a marked effect of the relative creep properties across the repair weld, on the stresses in the
weld and these, if the multiaxial failure criteria are known, can be used to obtain failure life
predictions. Ideally, a perfect match between the creep properties in all of the zones would be
created but this would be impractical to achieve in practice, since the creep properties of the service
exposed parent material is a function of the service time, temperature and stress.

4.1.2 Effect of repair profiles and weld width

The final geometry of a repaired weld depends on the repair profile and dimension, where the
excavation width, for example, w1 in Fig. 1(d), for a full repair, is an important parameter to ensure
the full removal of damaged material and to minimise preparation and welding costs. Results [5] for a
number of CrMoV repaired welds have shown that the peak stresses are slightly reduced (< 10%)
with increasing w1 values within a practical range. Also, there is a limited effect of the repair profiles,
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 8

i.e. full repair or partial repairs, Fig. 1, on the peak stresses. The results have shown that the
magnitudes of the peak rupture stresses for the full, partial-i and partial-ii cases are in the ratio 1 :
1.03 : 1.04, indicating that there may not be any significant benefit in choosing a particular repair
profile, with regard to the peak stress and the resulting failure life. As noted above, similar small
effects, for a specific set of material properties in the weld, were also found by Fidler, [25 ], Hyde et
al, [26], Law and Payton, [27], and Vazda, [28 ].

In practice, the included angle of the preparation will not exceed 45 o unless there is a specific access
problem for the welder. Again, although such an effect has not explicitly been studied for the prw
type repair geometries, the effect has been studied for conventional welds which will give guidance
for the full weld repair form, [31 ]. Again the effect is relatively small, for fixed property differences
across the weld. Any effects have been found to be larger, albeit still small, when additional loading is
higher and for included angles in excess of 45o. Vazda, [28 ], in particular, considered his results in
terms of stress concentration factors and it was shown that these can be quite high. Other workers
have generally considered only stress effects for application to rupture studies. However, to all
intents and purposes, it would appear that the effect of weld preparation angle can be ignored unless
additional loading and weld angles are large.

4.1.3 Effect of system load

The effects of system loading on the stresses within repaired welds in pipes have been investigated
[2,5,20] and [27,28 ]; the system loads were applied by an addition of constant axial loads applied to
the pipe ends. The results obtained by Hyde et al [5,20 ] have shown that a non-linear increase of the
peak rupture stresses occurs as the axial load is increased. In this case, for which a weaker HAZ
was used, with low or moderate end load, stresses were found to vary insignificantly across the
HAZ. However, the variations are more significant when the axial load is high, and the peak
stresses, within the HAZ, occur near the HAZ/PM boundary. An example of the variations of the
rupture stress with axial loading, in a full repaired CrMoV weld, is shown in Fig. 3. The observed
non-linear increase of the peak stresses with axial load (Fig. 3) was also obtained by Samuelson et al
[2 ], for an originally creep-hard weld with a creep-soft repair, and by Law and Payton, [27 ], for
both a creep-hard and a creep-soft case of a frw geometry.

Fig. 3: Variation of steady-state rupture stresses across the HAZ, along a line near
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 9

the outer surface of a fully repaired weld, with ? ax/? mh, for pi = 16.55 MPa [5].

4.1.4. Other weld repair geometries

The behaviour of circumferential butt weld geometry has been highlighted here. However, there are
also results obtained from limited examinations of alternative, but no less important components,
typically branch connections and castings/forgings which contain defects resulting from the original
fabrication. The latter can be realistically modelled using existing approaches, as long as the materials
data are available for castings/forgings. The former, however, is a more difficult problem and it is
only more recently that these geometries have been studied, [32-35]. Li, [3 5], has specifically
examined the repair of branch connections for correlation with the results of a major weld repair
programme on test vessels, [36 ]. A 3-D, two material model of the repair weld, as illustrated in Fig.
4, was used to determine the differences between results for matched and under matched weld
metals and the position of failure. The lowest stresses and the longest life estimates were shown to
occur with under matched weld metal.

Branch
y

x
Original Weld

z
Weld Repair

Main Pipe

Axial centre line of the pipe x

Fig. 4: Schematic diagram of a weld repair at a T-joint.

4.2 Failure Life Assessment

A simple approach to life estimation, using Equation (5), involves the use of the steady-state peak
rupture stress and the relevant uniaxial rupture properties; the failure life of the weld is taken to be the
minimum of the failure times predicted for each of material zones, and the failure site is the position
where the dominant peak rupture stress occurs [ 8 ]. A more realistic value can be obtained by taking
account of the complete creep curve, incorporating the tertiary as well as secondary creep regions.
Generally, this is done through the use of continuum damage laws, see Equations (1-3), and failure at
a particular point is defined as occurring when the accumulated damage, , reaches unity.
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 10

4.2.1 Failure assessment based on steady-state stresses

The steady state approach has been used in conjunction with either a maximum principal stress or a
maximum equivalent stress criterion for defining the rupture time. Generally, better estimates for life
involve a multiaxial failure criterion, based on either a stress or strain controlled criterion, although the
stress approach is more usual, [7].

Table 2 Typical example of results obtained from steady state analyses, using a maximum
principal stress or equivalent stress rupture criterion, [ 2].

Model Stress Parent life, h Original weld metal Weld repair life, h HAZ life, h
life, h
Original weld, C-S P only 6564 11796
(8892) (14346)
" P + axial 6155 5620
load (5263) (6346)
Original weld, C- P only 7008 24484 3697
S; (9210) (21876) (5717)
Weld repair, C-H
" P + axial 6356 7836 3500
load (5777) (6346) (3500)
Original weld, C- P only 8587 4377
H (11892) (5066)
" P + axial 8295 3800
load (7008) (3406)
Original weld, C- P only 4952 2830 12993
H; (8892) (3800) (17620)
Weld repair, C-S
" P + axial 5599 2431 5193
load (5263) (2371) (9351)
Original weld, C- P only 5599 2556 9785 10750
H; Weld repair C- (6564) (3315) (15088) (27588)
S + HAZ
" P + axial 3138 2049 5193 4734
load (2974) (2099) (5193) (6031)
Note: xxxx is the life under principal stress control and (xxxx) is that under equivalent stress control (P
Pressure)

Results presented by Samuelson et al [2] for a 1/2CrMoV: 2 1/4Cr1Mo repaired weld are a good
illustration. For an originally creep-soft weld with a creep-hard repair, failure is controlled by the
peak stresses in the new WM, while for an originally creep-hard weld with a creep-soft repair, the
failure is controlled by the peak stresses in the old WM, as illustrated in Table 2. The results show
the differences that can occur due to the choice of multi-axial stress criterion, but they also clearly
illustrate that the failure of the repaired weld is always predicted to be in the weld metal, either for the
original or the weld repair case and the predicted life is unaffected by the insertion of an HAZ.

4.2.2 Failure assessment based on damage

Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 11

The continuum damage laws, Equations (1-3), define material failure as occurring when the damage
parameter, , equals unity. In practice, for a pipe with a weld, failure was assumed to have occurred
when the failure damage zone, i.e. the zone in which 1, was found to extend through a
significant part of the wall thickness.

Damage analyses of the repaired welds were performed by Hyde et al [18,19 ], to investigate the
effects of weld dimensions, material properties and end loads etc., for the same welds under the
same loading conditions as for those for which steady-state analyses have been performed. The
results for a number of CrMoV pipe welds/repaired welds where the HAZs were weak, [5,19],
showed that steady-state predictions, using the multiaxial relationship in Equation (3), are
conservative, compared to the damage predictions and the failure occurred in the HAZ, initially
starting from the outer surface and moving inwards. A comparison is given in Table 3, from which the
effects of material property, repair profile and end load, on the failure life, can be directly evaluated.
For these particular cases, the life predictions using steady-state stresses were found to
underestimate the failure life by about 30 to 40% under the closed-end condition (ax/mh = 0.306)
and by about 20 to 30% when an additional axial load, up to the value of the mean diameter hoop
stress, was applied. However, the failure positions, predicted by continuum damage and steady-state
analyses, have been found to be consistent, confirming the validity of the technique, as a simple,
alternative approach to the use of continuum damage analyses.

Table 3 Life prediction (h) for a number of CrMoV welds, obtained from steady-state and
damage analyses, with p i = 16.55 MPa, for a range of axial load [19] ( ax and mh are the mean
axial end stress and the mean diameter hoop stress).

New weld Aged weld Full repair Partial repair -i

ax / mh Damage S-S Damage S-S Damage S-S Damage S-S
0.306 21,018 14,795 15,640 8,942 10,803 7,077 9,892 6,467
0.5 16,266 12,815 12,962 8,182 9,305 6,525 8520 6,082
0.75 8,274 6,484 7,197 5,415 6,206 5,185 5,632 4,530
1.0 4,186 3,265 3,966 2,895 3,751 2,971 3,307 2,678

In addition, an estimate can be made for the initiation and growth of Type IV cracks of the same
CrMoV welds [37] and the crack growth rate by monitoring the damage accumulation along the
HAZ. Fig. 5 shows typical damage variations with normalised distance along the HAZ close to the
type IV region, at different times, for the partially repaired weld (case-i).
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 12

Fig. 5: Damage variations in and along the HAZ, near the type IV region of partially
repaired weld-i, at different times, with pi = 16.55 MPa, under a closed-end condition [18].

The results of initiation time and remaining life are presented in Table 4. It can be seen that under
closed-end conditions, the percentages of the remaining times compared with the total creep lives
(failure lives), for the new, aged and repaired welds, are in a similar range, i.e. about 7-10%.
However, when ax/mh = 1, the ratio is about 20% for the fully repaired weld and is about 10% for
the partially repaired weld.

Table 4 Failure life, tf, initiation time, t o, and the remaining life, tf to, estimated from the results
of damage analyses, for ax/ mh = 0.306 and ax/ mh = 1 [37].

ax/mh t (h) New Aged Full Partial-i

tf 21,018 15,640 10,803 9,892
0.306 to 19,000 14,500 10,000 9,150
(closed-end) (t f - t o)/t f (%) 9.60 7.28 7.43 7.50
tf ---- ---- 3,751 3,307
1.0 to ---- ---- 3,000 3,000
(t f - t o)/t f (%) ---- ---- 20.02 9.28

5. Discussion

This review is intended to illustrate the current status of the analysis of repair welds under creep
conditions, using specific illustrations, where possible, from existing literature. The approach used for
repaired welds is identical to that used for virgin welds, although the geometries and material zones
are different. In addition, the repair welds are assumed to be fully post weld heat treated, PWHT, so
that the residual stresses in the weld can be assumed to have very low values, at the service
temperature, and therefore can be ignored. For a full weld repair, which is similar to the geometry of
a virgin weld but wider, use is also made of analyses which consider general weld geometry effects.

Real welds exhibit distinct metallurgical structures in certain zones; these are a result of the maximum
temperature and cooling rates experienced by the specific material during welding fabrication. These
structures, although complex, are reproducible with welding control. Furthermore, the properties of
the materials in these regions have a direct effect on the damage initiation and growth sites. Most
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 13

current analysis methods model the weld structure, for simplicity, by considering the weld metal,
parent metal and HAZ as three separate structures with distinct properties. The differences between
a new and an aged weld materials are mainly in the properties, while the difference between an un-
repaired weld and a repaired weld are in both the properties and geometries. Results obtained have
shown that, in general, the effect of repair profile/dimension on the failure life is relatively small, and
therefore the material property difference introduced by weld repair, appears to be the major factor
controlling the life of a repaired weld.

The results by Hyde et al [5 ] have shown that the lives of fully and partially repaired welds are
similar, with fully repaired weld showing a slightly longer life (about 10%), which is mainly due to the
difference in geometry. For the partial repair cases, the life difference due to the two different repair
profiles is insignificant, with case-i giving a slightly longer life (< 3%). The excavation width, for the
full repair, in a practical range, will not make a major difference in life. In addition, a weld repair
could lead to a significant decrease in life, compared to that of a un-repaired weld, this being highly
dependent upon the relative material properties of the constituents of the original and the repaired
welds. The implications of these results could have practical significance for the development of weld
repair strategies.

In general, predictions indicate that failure will occur within the zone for which the material properties
are the weakest and the failure site is directly affected by additional system loading. For instance, for
repaired welds in CrMoV pipe [5,18], failure occurs in the weaker material, i.e. the HAZ, and the
failure sites move from HAZ/WM interface to the Type IV position, when excessive axial loading is
applied. This is consistent with laboratory and plant experience.

An important result, obtained from the comparison of the results of the steady state and the damage
analysis methods, concerns the conservatism of steady-state prediction, compared with damage
prediction. The steady state approach underestimated the failure life by about 30-40% under pure
pressure conditions, and by about 20-30% when additional axial loading is significant. Also, the
failure positions predicted by the two approaches were found to be consistent. Although only proven
for the CrMoV repair weld studies, where the HAZ was assumed to have properties equivalent to
the Type IV material in this case, there is no reason to believe that similar conclusions cannot be
drawn for other weld material combinations.

A significant result, which is confirmed by plant and repair welded pressure vessel experience, is that
a partial weld repair, in which a HAZ is formed in the service exposed weld metal, produces
relatively weaker creep properties in this region and premature damage initiation and failure at that
position.

The FE analyses reviewed here have been primarily concerned with the estimation of weld life and
ways of determining the general factors influencing the life of repaired welds and with predicting Type
IV cracking. More sophisticated methods in which more advanced constitutive laws are utilised are
possible, following the work of Hayhurst and co-workers, [9]; these involve the use of two or three
parameter damage models and the use of other methods such as the THETA Projection approach,
[10]. Use of these methods requires specific modification to more generally used finite element
programmes. However, use of such approaches could result in improvements to the accuracy of
extrapolation methods. It is expected that future work will focus more strongly on the more detailed
factors that have been practically identified as important, such as detailed microstructural control, the
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 14

use of a strain-based approach, the effect of residual stresses etc., and to incorporate the features
that have been shown to be important.

For practical applications, it is important to realise that components requiring repair will not have
been subjected to identical service exposure conditions and hence, any finite element approaches
must be able to incorporate a basic understanding of the prime factors affecting life and then, through
a sensitivity analysis, allow the results to be generalised and applied in a safe, conservative and
economic manner.

6. Concluding Remarks

A review of finite element (FE) analyses of the creep of stress relieved repaired welds has been
presented, focussing on the creep analyses of such welds in pressurised pipe weldments using actual
creep data from ferritic steels and from parametric studies. The potential uses of FE modelling for
weld repair performance assessment have been discussed.

Most current FE models of welds consider three distinct regions; the weld metal, parent metal and
HAZ and there is little information where the HAZ has been further subdivided. Previous work has
shown that the effect of the repair dimension on the creep failure life can be relatively small, and
therefore the material property differences introduced by weld repair is likely to be the major factor
controlling the life of a repaired weld. For full repair welds, the results of analyses have suggested
that the excavation width, for a range of property differences, will not have a major influence on the
rupture life. In general, failure is predicted to occur within the material zone with a lowest life under
the relevant representative rupture strength. The failure site can be directly affected by the addition of
system loading.

Steady-state creep life prediction using FE models is shown to be conservative compared with FE
continuum damage predictions. Future improvements of FE modelling may include the use of more
advanced constitutive laws and the use of sensitivity analyses to generalise the approach.

Acknowledgements

The authors wish to acknowledge EPSRC, British Energy plc., PowerGen plc. and Innogy plc. for
the financial support provided through an EPSRC/ESR21 grant.

References

[1] Viswanathan, R., Weld repair of aged piping - a literature review, 2nd. Int. EPRI Conf. on Welding and Repair
Technology of Power Plants , Daytona Beach, Florida, May 1996.
[2] Samuelson, L. A., Segle, P. and Storesund, J., Life assessment of repaired welds in high temperature
applications, Welding & Repair Technology for Fossil Power Plants , EPRI Int. Conf ., Williamsburg, Virginia,
1994.
[3] Meyer, H. J., Mayer, K. H. and Streatz, A. J. A., Upgrading and retrofitting of earlier steam power plants
up to and beyond the year 2000, FWP Journal, Vol. 29, 1989, 15-25.
[4] Gandy, D. W., Findlan, S. J. and Viswanathan, R., A. comparison of conventional post weld heat
treatment and temperbead weld repair techniques for service-aged 2-1/4Cr-1Mo girth weldments, 2nd. Int.
EPRI Conf. on Welding and Repair Technology of Power Plants, Daytona Beach, Florida, May 1996.
[5] Hyde, T. H., Sun W. and Becker, A. A., Life assessment of weld repairs in 1/2Cr1/2Mo1/4V:2 1/4Cr1Mo
main steam pipes using the finite element method, J. Strain Analysis , Vol. 35, No. 5, 2000, 359-372.
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 15

[6] Coleman, M. C. and Miller, D. A., Inspection, assessment and repair strategies for type IV cracking in
power plant components, AWS/EPRI Conf ., Orlando, Florida, 1994, 5-15.
[7] Hayhurst, D.R., The role of creep damage in structural mechanics, Engineering Approaches to High
Temperature, Ed. B. Wilshire and D. R. J. Owen, Pineridge Press, Swansea, 1983.
[8] Hyde, T. H., Sun, W., Becker, A. A., Assessment of the use of finite element creep steady-state stresses
for predicting the creep life of welded pipes, Proc. 4th Int. Conf. on Computational Structures Technology ,
Civil-Comp Press, Edinburgh, Ed. B. H. V. Topping, 247-251, 1998.
[9] Perrin, I. J. and Hayhurst, D. R. A method for transformation of creep constitutive equations, Int. J. Pres.
Ves. & Piping, 68, 1996, 299-309.
[10] Evans R.W. and Wilshire, B., Creep of Metals and Alloys, The Institute of Metals, London, UK, 1985.
[11] Hyde, T. H., Sun, W. and Becker, A. A. Analysis of the impression creep test method using a rectangular
indentor for determining the creep properties in welds. Int. J. Mech. Sci., 38, 1996, 1089-1102.
[12] Parker, J. D. and James, J. D. Creep behaviour of miniature disc specimens of low alloy steel, PVP-Vol.
279, Developments in a Progressing Technology. ASME, 1994, 167-172.
[13] Eggeler, G., Ramteke, A., Coleman, M. C., Chew, B., Peter, G., Burblies, A., Hald. J., Jefferey, C., Rantala,
J., De Witte, M. and Mohrmann, R., Analysis of creep in a welded P91 pressure vessel, Int. J. Press.
Vess. & Piping , 60, 1994, 237-257.
[14] Hyde T. H. and Sun W., Determining high temperature properties of weld materials, JSME Int. J. Solid
Mechanics & Material Engineering , Series A, Vol. 43, No. 4, 2000, 408-414.
[15] Hayhurst, D. R. Creep rupture under multi-axial states of stress. J. Mech. Phys. Solids , 20, 1972, 381-390.
[16] Hyde, T. H., Sun, W. and Williams, J. A., Creep behaviour of parent, weld and HAZ materials of new,
service-aged and repaired 1/2Cr1/2Mo1/4V: 2 1/4Cr1Mo pipe welds at 640o C, Material at High
Temperatures, 16 (3), 1999, 117-129.
[17] Perrin, I. J. and Hayhurst, D. R. Creep constitutive equations for 0.5Cr0.5Mo0.25V ferritic steel in the
temperature range 600-675o C. J. Strain Analysis, 31, 1996, 299-314.
[18] Sun, W., Hyde, T. H, Becker, A. A. and Williams, J. A., Comparison of the creep and damage failure
prediction of the new, service-aged and repaired thick-walled circumferential CrMoV pipe welds using
material properties at 640o C, Int. J. Pres. Ves. & Piping, 77, 2000, 389-398.
[19] Sun W., Hyde T. H., Becker A. A. and Williams J. A., A study of weld repairs in a CrMoV serv ice-
exposed pipe weld, Proc. 9th Int. Conf. on Creep & Fracture of Engineering Materials & Structures, Swansea,
April 2001, 613-622.
[20] Hyde, T. H., Sun, W., Becker, A. A., Effects of end loading on the creep failure behaviour of CrMoV
welds in main steam pipelines, Proc. 6th Int. Conf. on Damage and Fracture Mechanics , Ed. A. P. S.
Selvadurai and C. A. Brebbia, 415-424, WIT Press. May 2000, Montreal, Canada.
[21] Zhang, J., Dong, P. and Bouchard, J., Repair length effects on weld residual stress distributions in a pipe,
ASME PVP -Vol. 410-1, 2000, 43-51.
[22] Walters, D., Stress analysis of cylindrical butt welds under creep conditions, CEGB Report RD/B/N3716,
CEGB, UK, 1976.
[23] Bissell, A., Scully, S. and Small, R., A case history of creep crack growth in a seam welded reheater outlet
header, 2nd. HIDA Conf. on Advances in Defect Assessment of High Temperature Plant, Stuttgart, Germany,
2000.
[24] Rotvel, F., Sampietri, C., Verelst, L., van Wortd, H. and Li, Ying Zhi, Header integrity assessment, Conf.
on Baltice IV, Plant Maintenance for Managing Life and Performance, Espoo, Finland, VTT, 1998.
[25] Fidler, R., The effect of weld width on the performance of narrow gap welds in creep conditions, CEGB
Report TPRD/M/1583/R86, 1986.
[26] Hyde, T.H., Williams, J.A. and Sun, W., Assessment of creep behaviour of narrow gap welds. Int. J. Pres.
Vess. & Piping , 76, 1999, 515-525.
[27] Law, M. and Payten, W., Weld performance under creep using Finite Element modelling, Int. J. Press. Vess.
& Piping, 72, 1997, 45-49.
[28] Vazda, D., On concentration effects in circumferential welds due to dissimilar creep properties, Int. J.
Press. Vess. & Piping , 73, 1997, 119-126.
[29] Hyde, T. H., Sun, W., Tang, A. and Budden, P. J., An inductive procedure for determining the stresses in
multi-material components under steady-state creep, J. Strain Analysis , Vol. 35, No. 5, 2000, 347-358.
[30] Hyde, T. H., Sun, W. and Tang, A., A parametric analysis of stresses in a thick-walled pipe weld during
steady-state creep, Proc. 5th Int. Colloquium on Ageing of Materials and Methods for the Assessment of Lifetimes
of Engineering Plant, Cape Town, April 1999, 231-246.
Review of FE analysis of repaired welds OMMI (Vol. 2, Issue 2) Aug. 2003 16

[31] Hyde, T.H., Sun, W., Becker, A.A. and Williams, J.A., Effect of weld angle and axial load on the creep
failure behaviour of an internally pressurised thick walled CrMoV pipe weld, Int. J. Press. Vess. & Piping ,
78, 2001, 365-372.
[32] Bruck, P. M., Smith, R. E. and Findlan, S. J., Finite element analysis of piping elbows with weld overlay
repair, 2nd. Int. EPRI Conf. on Welding and Repair Technology of Power Plants , Daytona Beach, Florida, May
1996.
[33] Rayner, G., A parametric analysis of stresses in a welded branched pipe under creep conditions, Internal
Report, University of Nottingham, March 2002.
[34] Hayhurst, D. R. Computational continuum damage mechanics: it use in the prediction of creep in
structures past, present and future. Proc. of IUTAM Symposium on Creep in Structures (Eds S. Murakami
and N. Ohno), Japan, 175-188 (Kluwer Academic Publishers), 2000.
[35] Li, Y. Z., Modelling for life assessment of local weld repair of 10CrMo910 T-joints by using matched and
under-matched weld metal, KEMA Report: 9870002-KPG/I&M 98-1142, 1998.
[36] Van Wortel, H., Effectiveness of repair in service exposed components by grinding or by welding
without PWHT, Conf. on Langzeitverhalten Warmfester Stahle und Hochtemperaturwerkstoffe, Nov. 1999,
Dusseldorf, Germany, Vdeh, 1999.
[37] Hyde, T. H., Sun, W. and Becker, A. A., A damage mechanics approach to predicting initiation and
growth of type IV cracks in CrMoV weldments, 2nd Int. HIDA Conf. on Advances in Defect Assessment in
High Temperature Plant , MPA Stuttgart, Germany, October 2000. Also in Int. J. Pres. Ves. & Piping, 78, No.
11-12, 2001, 765-771.