Structural integrity of corroded girth welds in vintage steel pipelines

Engineering Structures 124 (2016) 429–441

Contents lists available at ScienceDirect

Engineering Structures journal homepage: www.elsevier.com/locate/engstruct

Structural integrity of corroded girth welds in vintage steel pipelines Stijn Hertelé a,⇑, Andrew Cosham b, Paul Roovers c a

Ghent University, Soete Laboratory, Technologiepark Zwijnaarde 903, 9052 Zwijnaarde, Belgium Ninth Planet Engineering Limited, Newcastle Upon Tyne, United Kingdom c Fluxys Belgium SA, Brussels, Belgium b

a r t i c l e

i n f o

a b s t r a c t

Article history: Received 16 February 2016 Revised 15 June 2016 Accepted 27 June 2016

Girth welds of old steel pipelines and their surrounding heat affected zones are susceptible to corrosion attack. The resulting reduction in wall thickness may reduce the axial load or internal pressure bearing capacity to an unsafe level. Since standards provide limited guidance on weld corrosion assessment, the authors have executed an extensive experimental program to evaluate the axial load bearing capacity of corroded girth welds. To this end, curved wide plate tests have been executed and were analyzed by means of 3D digital image correlation. This paper discusses key influence factors related to weld geometry and material (strength and toughness). Then, the results are used to develop an assessment approach, based on Annex G of BS 7910:2013 and modified to account for the elastic–plastic stress–strain concentration resulting from weld misalignment. Ó 2016 Elsevier Ltd. All rights reserved.

Keywords: Pipeline Weld Corrosion Plastic collapse Weld strength mismatch Toughness Misalignment

1. Introduction When ‘vintage’ (say, older than 40 years) steel pipelines for fossil fuel transmission are inspected, circumferential metal loss due to corrosion may be detected in girth welds and their adjacent heat affected zones (HAZs). Such metal loss follows from the potentially suboptimal application of field coatings which were, at their time of installation, not considered as a critical factor. In addition, corrosion may be triggered by sensitive microstructures and/or chemistries associated with the weldment [1,2]. Local metal loss reduces the load bearing capacity of a pipeline. On the one hand, when the predominant loading component is internal pressure, structural integrity depends on the depth a and axial length L of metal loss (Fig. 1). There are well established procedures to assess the severity of corrosion damage in the body of vintage pipes, for instance the ‘modified ASME B31G’ equation [3]. This semi-empirical equation expresses burst pressure pmax or maximum hoop stress rh,max as follows (D and B representing pipe outer diameter and wall thickness, respectively):

1  0:85 Ba D ¼ rh;max ¼ rf pmax a 2B 1  0:85 M B

!

⇑ Corresponding author. E-mail address: [email protected] (S. Hertelé). http://dx.doi.org/10.1016/j.engstruct.2016.06.045 0141-0296/Ó 2016 Elsevier Ltd. All rights reserved.

ð1Þ

with rf a so-called flow stress equal to SMYS + 69 MPa, SMYS being the pipe steel’s specified minimum yield strength. The addition of 69 MPa conservatively accounts for the beneficial effect of strain hardening on load bearing capacity. Using an extensive test database, Leis et al. [4] have shown that a more objective – but potentially non-conservative – value for flow stress would be the ultimate tensile, rather than the yield strength of the pipe metal. Coming back to Eq. (1), 0.85 is an empirical correction factor for non-rectangular metal loss geometry and M is a dimensionless ‘Folias factor’ accounting for stress concentrations in the presence of a notch (M P 1) due to bulging [5]. M reflects an effect of finite corrosion length L. Eq. (1) is valid for a/B 6 0.8 (an empirical limit) and allows for the potential presence of an axial stress ra equal to rh/2, which could be induced by internal pressure due to end cap effects. Martin et al. [6] observed that the collapse based modified ASME B31G equation can be applied to (blunt) corrosion in brittle materials. In their test database, pipeline steels having a Charpy transition temperature of as high as +40 °C were covered. On the other hand, when the predominant loading component is axial stress (resulting from external factors such as ground movement), the severity of metal loss is governed by its depth a and circumferential arc length 2c. In such case, the axial plastic collapse stress is commonly predicted using a criterion developed by Kastner et al. [7]:

ra;max ¼ rf



   1  Ba p  2ca DB    1  Ba p þ 2 Ba sin 2c D

ð2Þ

430

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

Fig. 1. Definition of symbols related to geometry (pipe, corrosion damage) and load state.

This equation has been included in standards such as BS 7910, the British Standard on assessment of flaws in metallic structures [8], both as a reference stress solution for crack-like flaws and for corrosion. Notwithstanding its national character, this standard has established a worldwide reputation and its adoption exceeds British boundaries. Eqs. (1) and (2) have been developed with the aim to assess metal loss in pipe steel, remote from girth welds. It is noted that their primary focus was on crack-like defects, but their application on uniform pipe metal loss is justified as the equations are based on plastic collapse failure (as opposed to toughness driven failure). A non-brittle homogeneous material within a perfectly cylindrical geometry is assumed. Girth welds, however, may encompass brittle microstructures, exhibit heterogeneous material properties, show geometrical imperfections such as misalignment and may house residual stresses. Guidance on corrosion assessments in the vicinity of a weld is vague. For instance, ASME B31G can be adopted for girth weld corrosion in a pressurized pipeline ‘‘provided that the welds are of sound quality, have ductile characteristics and provided workmanship flaws are not present in sufficiently close proximity to interact with the metal loss” [3]. Since these requirements are not quantified, current practice tends to treat girth weld corrosion in vintage pipeline with extreme care. This often results in a large number of unnecessary and expensive pipeline excavation works. Also, ASME B31G cannot be used to assess the effect of the circumferential extent of the corrosion. Recognizing the abovementioned limitation of standards’ advice, a recent literature survey has collected published experimental data on the load bearing capacity of (potentially low toughness) welds showing metal loss [9]. This study, however, concludes that the number of tests performed so far is insufficiently exhaustive to unambiguously judge on the acceptability of weld metal loss. Moreover, not all potential influence factors have been covered with equal detail. In particular, the number of published tests on misaligned welds is very limited. Finally, the literature review

attempts to propose a workmanship criterion by suggesting that metal loss extending up to 20% of the structure’s wall thickness is acceptable irrespective of toughness (in the absence of sharp defects). However, no attempts are made to predict the actual load bearing capacity of corroded welds within the philosophy of an engineering critical assessment. In an attempt to better understand the effect of girth weld corrosion on the structural integrity (i.e., load bearing capacity) of vintage pipelines, the authors have carried out a destructive test program on sample welds extracted from the Belgian gas transmission pipeline grid, operated by Fluxys Belgium SA. This paper reports on the results of this program and evaluates an assessment method. It is structured as follows. Section 2 describes the materials and methods used. Section 3 discusses the experimental results. Attention goes to effects of weld specific features (potentially beneficial or adverse) such as weld strength mismatch, toughness and misalignment. Then, Section 4 evaluates an approach for girth weld corrosion assessment, based on Annex G of BS 7910 and supported by the experimental results. Conclusions are provided in Section 5. 2. Materials and methods Section 2.1 describes the tested materials. Sections 2.2 and 2.3 explain the experimental program, respectively focusing on component and small scale testing. 2.1. Origin of girth welds Ten girth welds (‘W1’ to ‘W10’) were extracted from the Belgian gas pipeline grid (Table 1). Their corresponding pipelines were constructed between 1967 and 1973 and – having successfully operated for at least 40 years – can be categorized as vintage. Different pipe types were covered: seamless, longitudinally seam welded and spirally seam welded. Two API 5L [10] pipe grades

Table 1 Overview of tested girth welds. Weld

Year of installation

Pipe seam type

API 5L [10] pipe grade

Nominal outer diameter D (mm)

Nominal wall thickness B (mm)

D/B (–)

W1 W2 W3 W4 W5 W6 W7 W8 W9 W10

1967 1971 1971 1968 1967 1971 1969 1969 1973 1973

Seamless Spiral Longitudinal Longitudinal Seamless Spiral Longitudinal Longitudinal Longitudinal Longitudinal

X46 X60 X60 X60 X46 X60 X60 X60 X60 X60

350 500 400 500 350 500 914 914 914 914

6.4 5.6 6.5 7.2 6.4 5.6 10.2 12.2 10.2 12.2

55 89 62 69 55 89 90 75 90 75

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

were tested: X46 (SMYS = 317 MPa) and X60 (SMYS = 414 MPa). Nominal outer diameter D varied between 350 mm and 914 mm and nominal wall thickness B ranged from 5.6 mm to 12.2 mm, D/B-ratio taking values between 55 and 90. The 914 mm diameter samples (W7 to W10) are further referred to as ‘large diameter’ welds, whereas the other pipes (diameter 500 mm or less; W1 to W6) are said to have a small diameter. This distinction will facilitate the description of the experimental program in Sections 2.2 and 2.3. 2.2. (Medium) curved wide plate testing All ten girth welds have been curved wide plate tested. A curved wide plate test is an intermediate scale tensile test on an unflattened (hence, curved), axially oriented sample of a pipeline section containing a girth weld at mid-length. Compared with full scale pipe testing, multiple tests can be performed for each girth weld. Additionally, material properties can be characterized by surrounding small scale tests (Section 2.3), thus enabling proper interpretations of the curved wide plate test results. In accordance with Fig. 1, the axial stress in a curved wide plate specimen is symbolized as ra. It is recognized that uniaxial loading does not entirely reflect the stress state in pressurized full pipes, where a biaxial stress state may occur. Hence, a translation of curved wide plate test results into full pipe scenarios is necessary and is treated further elsewhere in this paper (Section 4). Specimen dimensions are summarized in Fig. 2. In comparison with traditional curved wide plate testing practice where the prismatic section width W is between 200 mm and 400 mm [11], the specimens adopted for this study were smaller (W = 120 or 150 mm) to allow for more test specimens to be extracted. They are therefore referred to as ‘medium’ curved wide plate or MWP specimens. Specimens from the small diameter welds (W1 to W6; Fig. 2a) were smaller than those of the large diameter welds (W7 to W10; Fig. 2b) to avoid excessive plate curvature. In the majority of specimens, corrosion metal loss was simulated by machining off part of the girth weld from the cap side. A 25 mm diameter spherical mill was used in sequential passes longitudinal to the girth weld direction, covering the entire weld cap and at least 3 mm at either side of the weld cap. As a result, not only the welds themselves but also their adjacent HAZs have been

431

reduced in wall thickness. The profile of each machined weld cap was measured over three paths transverse to the weld, equally divided over the damaged area having a width 2c (Fig. 3a). The quantification of defect depth in terms of a unique value a proved challenging because of two reasons: the presence of weld misalignment (also referred to as ‘hi-lo’) and the potential variation of such misalignment over the specimen width. For simplicity, a has been defined as the machining depth with respect to the lower pipe (average of all three profile measurements). The unambiguous nature of a introduces conservatism as can be understood from the following. If a represents the largest wall thickness reduction within the weld, it is an objective measure of the damage severity. Alternatively, if a does not represent the largest wall thickness reduction within the weld, it underestimates the severity of damage. This is a safe approach for developing damage acceptance criteria as allowable wall thickness reductions will accordingly be underestimated. The weld cap profiles were further used to obtain the average weld cap ‘hi-lo’ ecap (Fig. 3a). Combined with measurements of plate thickness at either sides of the weld (B1 and B2), it was possible to calculate the weld misalignment e (Fig. 3b). In the remainder of this paper, the average plate thickness 0.5 (B1 + B2) is denoted as B. As regards instrumentation, full-field surface strain distributions were obtained by means of digital image correlation (DIC), an optical technique which is thoroughly explained in handbooks e.g. Ref. [12]. Images were taken using a stereo vision system, thus also yielding the three-dimensional profile of the investigated area of interest (‘3D DIC’). Images were obtained using two synchronized monochromatic 14 bit cameras, each having a resolution of 2452 by 2054 pixels. The images were analyzed using the commercially available VIC3D software of Correlated Solutions Inc. A proper DIC analysis requires the application of a non-uniform, high-contrast speckle pattern on the area of interest. Such patterns were achieved by painting a uniform white layer on a ground and degreased area of interest, followed by projecting black paint droplets upon the dried white layer. A more detailed description of 3D DIC measurements on medium wide plate specimens is provided in [11]. Fig. 4 summarizes the test configuration (subfigure a) and shows an MWP specimen, mounted in Soete Laboratory’s 2.5 MN universal test rig (subfigure b). The speckled area of interest covers a large portion of the prismatic part of the specimen, including the damaged girth weld which is located at mid-length. Tests were performed at room temperature. 3D DIC facilitated a proper interpretation of MWP test results by tracking the following information:  The evolution of the 3D girth weld profile. In particular, the evolution of weld misalignment as tensile load increases receives attention in Section 3.3.  Full-field in-plane strain distributions, providing a clear image on the structural response of the specimen.  An average ‘weld strain’ in the direction of applied load. This strain was defined as the displacement between two points traversing the corroded weld at mid-width of the specimen, divided by their initial gauge length (30 mm). It gives an indication of average plasticity in the weld and, thus, the nature of failure which can be either toughness driven (brittle; little or no plasticity) or plastic collapse driven (pronounced plasticity).

Fig. 2. Geometry of medium curved wide plate specimens: (a) welds W1-W6; (b) welds W7-W10.

In total, 34 MWP tests have been conducted, with relative wall thickness reduction a/B ranging between 5.5% and 35.8%. Of these, 33 specimens contained machined metal loss to simulate corrosion damage. For 23 specimens, machined metal loss extended over the entire prismatic specimen width. Ten other specimens were damaged over roughly two thirds of the specimen width to investigate

432

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

4

W

Path i Path ii Path iii

3

Profile (mm)

Metal loss

Variation of weld misalignment

2c

Three paths (i, ii, iii)

Average hi-lo ecap

2

L 1

Damage depth a 0 0

10

20

30

40

50

Position in direction transverse to weld (mm)

(a) ecap B1

e

B2

(b) Fig. 3. (a) Damage profile measurements over three paths at different positions in an MWP specimen, illustrating the definitions of machining damage depth a and (average) hi-lo ecap. (b) Definition of weld misalignment, taking into account pipe thickness variations.

Fig. 4. Soete Laboratory’s setup for medium curved wide plate (MWP) testing: (a) overview and 3D DIC setup; (b) mounted MWP specimen, showing the DIC area of interest (covered with black speckles on a white background).

the effect of surrounding undamaged weld metal. Although less relevant for axial tensile loading, it is worth noting that length L of machined metal loss varied between 19 and 29 mm. The remaining specimen (extracted from weld W5) was naturally corroded as shown in Fig. 5, extracted from a 3D DIC profile measurement. It contained narrow corrosion channels at the cap side of both HAZs, covering the entire specimen width and locally reducing the wall thickness up to 19%. Table A.1 (App. A) summarizes the MWP test database.

2.3. Small scale mechanical testing Steel of all involved pipes has been tensile tested at room temperature according to ISO 6892-1 [13] using 25 mm wide full thickness prismatic specimens, extracted in the axial direction of the pipe. Yield strength was characterized on the basis of Rt0.5 (stress at 0.5% total strain) for continuously yielding steels, and ReL (lower yield strength) for steels exhibiting a yield plateau. Note that the choice for Rt0.5 stems from API 5L [10], which prefers this value

433

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

1.1 mm

isometric view (true aspect ratio)

side view (vertical scale exaggerated)

Fig. 5. 3D profile measurement of the MWP specimen sampling areas of natural HAZ corrosion (test 15).

strength properties. Mismatch values above and below unity represent strength overmatching and undermatching welds, respectively. Both MY and MT could be readily obtained for welds W7 to W10 as their weld metal was tensile tested. Based on the knowledge that ultimate tensile strength is closely related to Vickers macro hardness, MT has been estimated for the small diameter welds (W1 to W6) on the basis of average hardness values of weld and pipe metal. To this end, hardness maps of weld macrographs have been constructed from large numbers of Vickers indentations with a proof load of 5 kgf (‘HV5’; Fig. 6). Table 2 reveals a wide range of covered weld strength mismatch values, MT for instance ranging between 0.963 (slightly undermatching) and 1.281 (strongly overmatching). All but two of the welds were overmatching with respect to the weaker pipe. Note the simplified nature of weld strength mismatch quantification on the basis of these numbers, as hardness maps revealed highly heterogeneous pipe and weld metal properties for some welds. For instance, weld W4 shows three distinct hardness bands which are clearly linked to the weld bead deposition sequence (Fig. 6). It connects two heterogeneous pipes, as the pipe steel left to the weld is significantly harder than that on the right side. Finally, each girth weld has been Charpy V notch tested, thus providing an indication of notch toughness. The test procedure conformed to ISO 148-1 [15]. A striker with a radius of 2 mm was used. Three repeat tests were performed for different combinations of test temperature (0 °C or 20 °C) and notch location (either at the weld metal center or targeting the HAZ/fusion line;

over the more common 0.2% proof stress Rp0.2, acknowledging the fact that both are similar for line pipe steels and that Rt0.5 is easier to obtain. Yield-to-tensile ratio (further simply denoted as Y/T) was defined as either Rt0.5/Rm or ReL/Rm, Rm being the ultimate tensile strength. The tested pipe steels were found to be considerably work hardening, Y/T ranging between 0.67 and 0.84. Linked to this is the highly ductile nature of the steels, uniform elongation (engineering strain at Rm) ranging between 8% and 17%. The observations of strong work hardening and ductile behaviour are to be expected for vintage lower grade pipe steel (as opposed to contemporary high-strength pipe steels [14]). For the large diameter girth welds (W7 to W10), weld metal stress–strain properties were obtained from ‘all weld metal tensile tests’ of circumferentially oriented round bar specimens having a test section diameter of 6 mm. Here, Y/T ranged between 0.74 and 0.87 and uniform elongation ranged between 7% and 11%, indicating less work hardening and ductility than the pipe steels. Testing the small diameter welds (W1 to W6) in this manner has not been performed, as their thin-walled and narrow nature would have yielded impractically small tensile test specimens. An important parameter of girth weld integrity is the weld strength mismatch, defined as the dimensionless ratio of weld metal strength to pipe steel strength. It can be based on either yield strengths (MY) or ultimate tensile strengths (MT). The unavoidable scatter in stress–strain properties results in girth welds connecting a ‘weaker’ pipe (having lower strength characteristics) with a ‘stronger’ pipe. MY and MT have been based on the weaker pipe’s

HV5 150 216

5 mm

Fig. 6. Vickers hardness maps (shown here for weld W4) reveal significant heterogeneity patterns in some tested girth welds.

Table 2 Weld strength mismatch values based on yield strength (MY) and ultimate tensile strength (MT). For welds W1 to W6, MT was estimated on the basis of Vickers hardness maps. Weld

W1

W2

W3

W4

W5

W6

W7

W8

W9

W10

MY (–) MT (–)

– 0.975

– 1.207

– 1.269

– 1.054

– 0.963

– 1.132

1.170 1.088

1.435 1.169

1.277 1.140

1.494 1.281

434

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

imens were characterized by an extreme weld misalignment e (around 30% of the average wall thickness) which tends to reduce the axial failure stress as further discussed (Section 3.3).  One might expect higher failure stresses for specimens that were not damaged over their entire width (2c/W = 60%, open markers in Fig. 7). Such trend, however, is not observed in a consistent manner. This observation is briefly touched in Section 3.3 and further investigated in Section 4.2. There, it will be confirmed that width is secondary to other factors.

notch in through-thickness direction). Subsize specimens were tested for welds W1 to W6 (5  10 mm2 cross section) and W7 (7.5  10 mm2 cross section) due to limited wall thickness and/or severe weld misalignment. At 20 °C (the MWP testing temperature), full size equivalent Charpy values down to 32 J were observed. Five welds (W3, W5, W7, W8, W9) show transitional behaviour at 20 °C (less than 100% shear area in the fracture surface). Four of these were close to fully ductile (shear area percentages above 85%).

Notably, post mortem fractography did not reveal any presence of natural weld flaws such as porosities or slag inclusions, which would negatively influence the load bearing capacity.

3. Results and discussion 3.1. Overall trends

3.2. The role of toughness

Fig. 7 plots the axial stress at failure (defined as the occurrence of maximum load prior to either collapse, or unstable fracture) as a function of relative wall thickness reduction. The stress values are normalized against the yield strength of the weakest pipe steel in the connection. This representation makes sense as values above 100% indicate the attainment of remote (pipe steel) yielding prior to failure. In such cases, the corrosion in the weld is definitively acceptable as remote yielding is considered to be a limit state in a stress based design philosophy. For instance, the commonly used EPRG Tier 2 guidelines for the assessment of sharp girth weld defects [16] adopt this limit state. A first inspection of Fig. 7 reveals that the normalized axial stress at failure tends to drop with increasing relative damage depth a/B, as can be expected. Some observations deserve particular attention.

Weld strain at maximum load (%)

Significant plasticity occurred in the weld region for all specimens. The lowest observed axial weld strain (defined in Section 2.2) at maximum load is as much as 3.2%, which is far beyond the linear elasticity (strains of 0.5% and below). There is no correlation with Charpy V notch energy (Fig. 8). The diamond markers in the figure

 For the tests performed, wall thickness reductions up to 20% did not inhibit the occurrence of remote yielding prior to failure. This observation supports the empirical workmanship criterion proposed in [9], as discussed in the introductory section.  Given a certain damage depth a/B, higher axial stresses tend to be allowable for welds with a higher strength mismatch. Indeed, the upper part of the scatter band is formed by welds for which MT > 1.1 (circular markers). Five tests may be ruled out in this comparison. Three of these collapsed in the weakest pipe steel due to the presence of pipe corrosion damage remote from the girth weld. As a consequence, these specimens could not attain the actual failure stress of the weld. Two other spec-

min

ave max

15

10

5

0 30

40

50

60

70

2c/W = 100% 60%

Failure in weakest pipe steel

140

MT < 1.1 Remote yielding

Axial stress at failure / yield strength of weakest pipe steel (%)

Machined welds

130 120 110

> 1.1 Naturally corroded weld Severe misalignment

100 90 80 0

10

20

90

100

Fig. 8. Weld strain at maximum load is not related to Charpy V-notch energy for the tests performed.

160 150

80

CVN energy (J) (full size equivalent, 20 °C, weld metal centre and HAZ)

30

40

Relative wall thickness reduction a/B (%) Fig. 7. Summary of all tests: normalized axial stress at failure as a function of relative wall thickness reduction.

435

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

represent average values of six Charpy tests (two notch locations, three tests per location – see Section 2.3) at 20 °C, the temperature at which the MWP tests were performed. Error bars indicate minimum and maximum values of these six tests. Energy values of sub-size specimens (welds W1 to W7) have been converted into full size equivalents. The high weld strains and their independence of Charpy energy indicate that failures were dominated by plastic collapse rather than toughness within the conditions tested (i.e., minimum Charpy values exceeding 32 J, average Charpy values exceeding 44 J). This observation agrees with numerical analyses reported in [4], indicating that toughness controlled fracture of corroded welds may only occur when Charpy V notch energy drops below 27 J. In the same study, failure is even suggested to be plastic collapse driven regardless of the Charpy level for a/B < 20%. A second confirmation can be found in the EPRG Tier 2 guidelines for the assessment of girth weld defects [16], which require minimum and average Charpy V notch energies of 30 J and 40 J for collapse driven failure to occur. As these guidelines were developed for sharp defects, the corresponding toughness criteria can be safely applied for the assessment of blunt corrosion damage, too.

Axial stress / yield strength of weakest pipe steel (%)

140

Test 32 (W10, a/B = 0.20, MT = 1.281, (Y/T)pipe = 0.71, e/B = 0.153)

120

remote yielding

100 80

Test 12 (W4, a/B = 0.23, MT = 1.054, (Y/T)pipe = 0.83, e/B = 0.047)

60 40 20 0 0

1

2

3

4

5

6

7

8

9

10

Weld strain (%) Fig. 9. There is a strong influence of weld strength mismatch MT and pipe yield-totensile ratio (Y/T)pipe.

3.3. Influence factors This section focuses on effects of constitutive properties (strain hardening, weld strength mismatch) and geometry (weld misalignment) on the observed failure stresses. First, the effects of strain hardening (expressed as Y/T) and weld strength mismatch can be understood from the abovementioned finding of Leis et al. [4] that the ultimate tensile strength of the damaged material is an objective value for flow stress in Eq. (1). In this study, it would imply that the weld metal’s ultimate tensile strength Rm,weld governs the plastic collapse load. Historically, the (empirical) definition of flow stress rf used in ASME B31G is based upon the yield strength of the pipe steel. The resulting conservativeness is indicated by the ratio between Rm,weld and the yield strength of the pipe steel, and this ratio is equal to MT/(Y/T)pipe. In other words, high weld ultimate tensile strength overmatch and high pipe strain hardening (low Y/T) are beneficial with respect to load bearing capacity. Noteworthy, in case the weld itself is strongly heterogeneous as for instance in Fig. 6, Rm,weld should represent average properties. The above is experimentally confirmed. For instance, Fig. 9 compares axial stress – weld strain plots of specimens from welds that strongly differ in terms of both MT and (Y/T)pipe (W4 and W10). Weld W10 has a significantly higher MT/(Y/T)pipe ratio than weld W4 (1.80 versus 1.27, respectively). The severity of metal loss is comparable. Clearly, the specimen from W10 showed a significantly higher failure stress and attained remote yielding. In contrast, the specimen from W4 failed at 90% of the remote yielding level. Notably, the adverse effect of weld misalignment on failure stress (discussed in the paragraph below) does not bias the interpretation of Fig. 9 as test specimen 32 showed a more severe misalignment than test specimen 12. Second, weld misalignment tends to reduce the failure stress. This follows from local bending upon the application of an axial load, as revealed in Fig. 10 for a specimen with a severely misaligned weld (e/B = 31%). At a normalized axial stress level of 110% (near maximum load), misalignment has completely flattened out due to localized bending, as reflected in strong strain concentrations near the ‘low’ side of the weld. These concentrations are expected to facilitate the onset of plastic collapse. Interestingly, the undamaged weld caps at either sides of the damaged region do not inhibit the strain concentration band in the corroded weld, which is simply deflected into the specimen sides of the ‘lower’

Test 9 • • • •

Tensile load

Weld W3 a/B = 18% e/B = 31% 2c/W = 58%

Prior to test

Weld damage

‘Low’ side

During test (axial stress = 110% of yield strength of weaker pipe steel)

Misalignment has flattened out

‘High’ side Strain concentration

Weld cap

Axial true strain 0%

13%

Fig. 10. 3D DIC explains the adverse effect of weld misalignment on the axial load bearing capacity of a girth weld having metal loss.

436

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

Test 1 (weld W1, MT = 0.975) Metal loss, machined over entire specimen width a/B = 21.2%, e/B = 4.3%

Test 15 (weld W5, MT = 0.963) Natural corrosion a/B = 19.0%, e/B = 8.0% 0.00

Axial true strain (-)

0.05

Fig. 11. Top view of axial true strain distributions at the onset of remote yielding in welds with (top) machined and (bottom) natural metal loss, and further similar conditions.

pipe steel. This supports the third observation of Section 3.1 regarding the limited effect of damage width. 3.4. Analysis of specimen with naturally corroded girth weld As mentioned above, one specimen contained a naturally corroded girth weld. It failed by plastic collapse of the weld, preceded by significant yielding in the base metal. DIC analysis revealed that the strain concentration in the girth weld was significantly smaller than that measured in specimens with machined corrosion damage (and further comparable influence parameters). This is illustrated in Fig. 11, showing distributions of axial true strain of the naturally corroded specimen (test 15) and a specimen of weld W1 with similar weld strength mismatch, wall thickness reduction and weld misalignment (test 1). The conservative nature of milling the weldment is logical as it creates a region of uniform metal loss, whereas natural corrosion is characterized by an irregular profile having regions of limited damage. It is concluded that the practice of milling the weldment to simulate natural corrosion is conservative and allows to develop safe weld assessment procedures. 4. Development of assessment criteria This section is concerned with the formulation and evaluation of an assessment methodology, supported by the experimental database. Five factors deserve consideration: (relative) dimensions of metal loss, Charpy V notch energy, weld strength mismatch, strain hardening and weld misalignment. Section 4.1 provides the proposed general methodology to do so. The approach is then evaluated and fine-tuned on the basis of the axially loaded MWP test results (Section 4.2). Finally, the observations are extrapolated to full pipe geometries with axial loading or internal pressure dominated scenarios (Section 4.3). 4.1. General methodology The proposed methodology is based on the most recent revision (2013, incorporating corrigendum no. 1) of BS 7910 [8], and it is also informed by ASME B31G [3]. It utilizes the reference stress concept, which addresses plastic collapse as the mechanism governing failure based on original analyses by Ainsworth [17]. Reference stress rref is defined as follows:

rref rf ¼ ra ra;max

ð3Þ

where ra is the remotely applied stress (e.g., axial stress in the case of MWP testing), ra,max its critical value corresponding with plastic collapse, and rf an assumed flow stress. In words, the applied stress reaches its maximum allowable value when reference stress attains the flow stress. The availability of a reference stress solution rref/ra

as a function of parameters describing the structure (geometry, loading conditions) allows to estimate the allowable stress ra,max = rf(rref/ra)1. As regards the definition of flow stress, the ‘modified ASME B31G’ definition of SMYS + 69 MPa deserves consideration and its conservatism will be evaluated against the experimental MWP database. On the other hand, taking Rm,weld as a flow stress will also be examined, since this definition is expected to produce the closest agreement between predicted and observed load bearing capacities [4]. Note that Rm,weld is an actual value (of the ultimate tensile strength of the weld metal), whereas SMYS represents a specified minimum value (of the yield strength of the base metal). It is noted that other flow stress definitions exist. Two important examples are BS7910, which defines rf as the average of minimum specified yield and ultimate tensile strength, and CSA Z662 [18], which defines rf between 1.03 SMYS and 1.10 SMYS (the exact factor depending on pipe grade). It turns out that these alternative definitions yield lower flow stress values than the modified ASME B31G definition and, hence, are more conservative. In other words, conservatism of the modified ASME B31G definition (which will be shown further in this paper) implies an equal statement for the BS7910 and CSA Z662 definitions of flow stress. At first sight, Eq. (3) appears to assume that the remotely applied stress ra is the only stress value contributing to failure. However, weld misalignment gives rise to a local bending stress (having a maximum value ±rb at the outer regions of the cross section) which might be expected to facilitate plastic collapse. BS7910 provides reference stress solutions that account for misalignment by introducing rb into the equation. Hereby, the equilibrium of bending moment implies a constant ratio rb/ra, being a function of the misaligned geometry. Solutions for rb/ra are provided in BS7910. These solutions can be included into the reference stress approach (Eq. (3)) as will be illustrated in the following sections. It is noted that all misalignment bending solutions within the standard are based on linear elasticity. Notably, residual stresses are neglected in standardized reference stress equations. These are believed not to contribute to plastic collapse as such stresses redistribute upon the increasing development of plasticity. Hence residual stresses are not further considered. 4.2. Axially loaded plates The MWP test database allows the evaluation of the methodology explained above. To this end, the abovementioned procedure is elaborated for the case of axially tension loaded plates containing a corroded weld perpendicular to the direction of loading (Fig. 3). The plate has a thickness B and width W; the metal loss has a depth a and width 2c. Weld misalignment is symbolized by e. The slight curvature of the MWP specimens is neglected as it is believed to be

437

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

of minor relevance to the collapse load of the plate. In other words, flat plate solutions are adopted. This choice is motivated by earlier studies on the plastic collapse of curved plates with sharp defects [19], which confirmed that neglecting the effect of plate curvature on limit load is a sound simplification. Axial tension is symbolized by means of ra, the remotely applied membrane stress. BS7910 advises the following solution for locally induced bending stress rb in axially loaded misaligned plates:

rb 3e ¼ ra B

ð4Þ

As noted in Section 4.1, this solution assumes small, linear elastic deformations. The resulting value can be substituted into the following reference stress solution for flat plates, developed by Willoughby and Davey [20]:

rref ¼ ra

rb ra

þ

  2 rb ra

þ 9ð1  a00 Þ2

0:5 ð5Þ

3ð1  a00 Þ2

where a00 is given by:

a00 ¼

8 a=B > < 1þB=c

W P 2ðc þ BÞ

2ac > BW

W < 2ðc þ BÞ

:

ð6Þ

The special case where corrosion covers the entire plate width (2c = W) is characterized by a00 = a/B. Combining Eqs. (3)–(6) leads to the following predicted failure stress:

ra;max ¼ rf

ð1  a00 Þ2 h  i0:5 2 e þ Be þ ð1  a00 Þ2 B

ð7Þ

Finally, in the hypothetical absence of misalignment effects, the failure criterion simply reduces to:

ra;max ¼ rf ð1  a00 Þ

ð8Þ

f= SMYS + 69 MPa Rm,weld

600

500

400

300

 Whereas taking flow stress as SMYS + 69 MPa (crossed markers) indeed provides conservative failure stress predictions, the obtained factor of safety may be undesirably excessive. For one specimen that failed at 435 MPa applied stress, the predicted failure stress was as low as 235 MPa which is merely 55% of the actual value.  Contrary to expectations, taking Rm,weld as a flow stress value mostly yields conservative failure stress predictions as the majority of all data points (diamond markers) fall below the 1:1 line. This indicates that Eq. (7) does not accurately describe all geometrical effects on collapse resistance. A closer investigation reveals an overprediction of the effect of weld misalignment, since the ratio of actual to predicted failure stress increases as e/B increases (Fig. 13a). Hereby, for non-misaligned welds, accurate predictions are made using rf = Rm,weld as the linear regression line in the graph approaches unity towards e/B = 0. This is a confirmation of Leis et al.’s statement regarding the use of actual ultimate tensile strength as a flow stress [4]. Compared with the above, neglecting weld misalignment appears to provide more objective failure stress predictions for rf = Rm,weld as the trend of corresponding data points in Fig. 12(b) (diamond markers) is closer to the 1:1 line than in Fig. 12(a). Using this flow stress definition, there is a slight tendency towards failure stress overestimations as weld misalignment increases (Fig. 13b). Hence, there is an observed effect of weld misalignment (even though it is overpredicted by Eq. (7)). This, however, does not produce unconservative predictions for the MWP test database when adopting the modified ASME B31G definition of SMYS + 69 MPa for flow stress (Fig. 12b, crossed markers). As a final observation to Figs. 12 and 13, the failure stress of the naturally corroded specimen is predicted with, compared to the other test results, a high degree of conservatism (owing to its irregular profile). This agrees with the discussion of Fig. 11. The overprediction of the effect of weld misalignment can be understood from the assumption of linear elasticity to produce Eq. (4). Whereas this theory assumes small deformations, the severe plasticity preceding collapse causes weld misalignment to flatten out as observed in Fig. 10. As a result, the corresponding

Axial failure stress, predicted by Eq. (8) (MPa)

Axial failure stress, predicted by Eq. (7) (MPa)

Eqs. (7) and (8) have been evaluated against the experimentally observed failure stresses (Fig. 12). Note that the three test results corresponding to failure in the pipe steel (recall Fig. 7) have been omitted since these do not represent weld collapse. In this and following figures, data corresponding with the naturally corroded MWP specimen are indicated by arrows. Both abovementioned expressions for rf are considered: SMYS + 69 MPa and Rm,weld. The former should be conservative (data points below the 1:1 line),

the latter should yield accurate predictions (data points around the 1:1 line). Comparing Fig. 12(a) and (b), it is evident that taking into account weld misalignment has a notable effect on the predicted failure stress. When misalignment is included (Fig. 12a), the following is observed:

Unsafe 1:1

f= SMYS + 69 MPa Rm,weld

600

500

400

300

Unsafe 1:1

Safe

Safe

200

200 200

300

400

500

600

Axial failure stress, experimental (MPa)

(a) Eq. (7); misalignment is included

200

300

400

500

600

Axial failure stress, experimental (MPa)

(b) Eq. (8), misalignment is neglected

Fig. 12. Comparisons of experimental failure stresses with analytical predictions, (a) including misalignment (Eq. (7)), (b) neglecting misalignment (Eq. (8)). Arrows indicate data corresponding with the naturally corroded MWP specimen (test 15).

438

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

Eq. (7), f = Rm,weld Linear (regression line Lineair )

Eq. (8), f = Rm,weld Linear (regression line Lineair )

1.4

Actual failure stress / predicted value (-)

Actual failure stress / predicted value (-)

1.4

1.2

1.0

0.8

0.6

1.2

1.0

0.8

0.6 0

15

30

0

15

30

Relative misalignment e/B (%)

Relative misalignment e/B (%)

(a) Taking into account misalignment

(b) Neglecting misalignment

Fig. 13. The effect of weld misalignment is (a) overpredicted by Eq. (7), and (b) underpredicted by Eq. (8). Arrows indicate data corresponding with the naturally corroded MWP specimen (test 15).

f= SMYS + 69 MPa Rm,weld

600

e kt  1 ¼ eeff kt;eff  1

400 Unsafe 1:1

ð9Þ

The exact value of this reduction factor is dependent on the strain hardening properties of the material, the amount of actual weld misalignment and the applied axial stress ra. The following case study considers a strength evenmatching (i.e., homogeneous) weldment where ra equals the yield strength of the material, thus representing the onset of remote yielding. Power-law strain hard-

Eq. (7), f = Rm,weld Linear (regression line Lineair )

1.4

500

300

aligned welds in the DNV offshore standard for submarine pipeline systems, DNV-OS-F101 [22], states that an elastic–plastic concentration of the product of stress and strain is constant. The concentration factor is equal to kt2, where kt represents the linear-elastic stress concentration factor. The actual (‘‘effective”) elastic–plastic stress concentration kt,eff is then obtained from the intersection between the hyperbolic stress–strain concentration locus and the material’s stress–strain response (Fig. 15). It is noted that, albeit advised in DNV-OS-F101, the proposed use of Neuber’s rule may be considered fairly unconventional for the investigated problem, as its current application is mainly oriented towards the fatigue analysis of notched structures. In BS7910, the locally increased stress ktra is said to be the sum of ra + rb, the maximum tensile stress occurring in the misaligned cross section. Hence, rb/ra equals kt – 1. Since this stress ratio is proportional to e in the collapse assessment procedure (Eq. (4)), the reduction factor between actual and effective weld misalignment is given by:

Actual failure stress / predicted value (-)

Axial failure stress, predicted by Eq. (7) (MPa)

bending stress rb is strongly reduced with respect to its value predicted under elastic conditions. In an attempt to account for this effect, the concept of an ‘‘effective” weld misalignment eeff is introduced. This is a virtual reduced misalignment value that acknowledges the occurrence of flattening upon plastic deformation and gives rise to proper misalignment effects when used for Eq. (7) instead of the actual misalignment e. From an empirical point of view, defining eeff as 1/3 of its original value e in the undeformed state (i.e., 1/3 of the value shown in Fig. 3b) appears to remove the overprediction of misalignment effects in the MWP test database. Indeed, the resulting regression line has become near to flat and approaches unity regardless of misalignment (Fig. 14b). Fig. 14(a) evaluates the performance of failure stress predictions when using the effective misalignment concept. Similar to Figs. 12 and 13, the arrows indicate data corresponding with the naturally corroded specimen. Compared with Fig. 12, rf = Rm,weld now yields objective predictions (as could be expected from Fig. 14b) and the conservatism for rf = SMYS + 69 MPa has reduced in comparison to Fig. 12(a), predicted collapse stresses ranging between 63% and 89% of the corresponding actual values, with an average of 73.5% and a standard deviation of 6.9%. The boundary between safe and unsafe predictions exceeds the average prediction-to-actual ratio of failure stress by (100–73.5)/6.9 = 3.8 times its standard deviation. The empirical observation that eeff = e/3 is motivated to be sound by using Neuber’s rule for elastic–plastic stress–strain concentrations [21]. This rule, advised for the flaw assessment of mis-

1.2

1.0

0.8

Safe 200

0.6 200

300

400

500

600

0

5

10

Axial failure stress, experimental (MPa)

Relative effective misalignment eeff/B (%)

(a)

(b)

Fig. 14. Taking the ‘‘effective” misalignment value as 1/3 of the original value removes the overprediction of misalignment effects: (a) comparison of failure stresses and their predicted values; (b) conservatism as a function of effective misalignment.

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

Fig. 15. Graphical representation of Neuber’s rule for calculation of an elastic– plastic stress concentration factor.

e/eeff using Neuber's rule ( -)

ening is assumed. A published relation [14] between the strain hardening exponent and Y/T-ratio was used to cover steels having a Y/T-ratio between 0.67 and 0.87, the range of values observed in the experimental study. Three actual misalignment levels were considered, in line with the experimental MWP database: e/B = 10%, 20%, 30%. Fig. 16 plots the resulting misalignment reduction factors to obtain eeff. For strongly strain hardening materials (low Y/T), Neuber’s rule predicts misalignment reduction factors similar to the empirical value. The reduction factor increases as strain hardening decreases (Y/T increases), reaching a value of 8.5 for Y/T = 0.87 and e/B = 30%. This case study indicates that reducing the actual misalignment by a factor 3 is not unsound and is conservative with respect to a rigorous analysis using Neuber’s rule. As a final point to this section, it is interesting to check if the definition of a00 (Eq. (6)) allows the analytical prediction of the limited effect of 2c/W on failure stress, as observed in the experimental program and intuitively explained in the discussion of Fig. 7. The effect of damage width 2c/W is reflected in rref/ra through the term a00 . In the following reasoning, weld misalignment is neglected and Eq. (8) applies. Fig. 17 plots the evolution of ra,max/rf (or, in other words, 1 – a00 ) as a function of 2c/W for a case study (a/B = 0.2, W/B = 20). These values are realistic with respect to a fair number of the per-

10 9 8 7 6

e/B = 30% 20% 10%

5 4 3

Empirical value 2

439

Fig. 17. A case study illustrates the limited effect of 2c/W on predicted failure stress for 2c/W P 60%.

formed MWP tests. Notable is the kink around 2c/W = 0.9, reflecting the piecewise nature of the equation. For 2c/W exceeding this value, the relation between 2c/W and a00 is linear. For 2c/W immediately below this value, however, there is a very limited influence of flaw width on a00 . As a result, the difference between relative failure stress for 2c/W = 1.0 (representing 23 of 33 MWP tests with machined corrosion) and for 2c/W around 0.6 (representing the other 10 MWP tests with machined damage, and indicated by the dashed line in Fig. 17) is a mere 3%. This observation agrees with the limited role of 2c/W observed in the experimental database. Fig. 17 predicts an increasing sensitivity to damage width for decreasing values of 2c/W, becoming significant only for cases which are excluded from the test matrix. The sound applicability of Eq. (7), assuming SMYS + 69 MPa for flow stress, requires a set of prerequisites related to material properties. Firstly, failure should be driven by plastic collapse. To this end, the Charpy criterion of the EPRG Tier 2 guidelines for girth weld defect assessment [16] are proposed (at least 30 J and 40 J as minimum and average, full size equivalent, values from a set of three tests). This requirement holds for sharp defect assessments, where the potential influence of toughness is much more apparent. Moreover, it is validated by the current test program as some Charpy characterizations are very close to the 30 J/40 J limits. The criterion should be met for both weld metal center and HAZ notched Charpy specimens. Recalling the observations of Martin et al. [6] (see Section 1), further testing on lower toughness welds is expected to broaden the applicability range. The weld should at least be strength evenmatching in terms of ultimate tensile strength (i.e., MT P 1). SMYS + 69 MPa is considered to be conservative when applied to vintage pipelines, for which ample strain hardening (low Y/T-ratio) can be expected. In contemporary high-strength line pipe steels, the average of the specified minimum yield and ultimate tensile strength might be more appropriate, see ASME B31G-2012. 4.3. Girth welded pipes under axial load and/or internal pressure

1 0.65

0.70

0.75

0.80

0.85

0.90

Y/T (-) Fig. 16. Misalignment reduction factors calculated using Neuber’s rule, compared with the empirically advised value. The empirical value is safe with respect to the calculated values.

The previous section has shown that standardized plastic collapse assessments allow the conservative prediction of the failure stresses observed in the MWP test program, and describe the role of weld and metal loss properties factors (metal loss depth and width, weld strength, strain hardening, weld misalignment). Applicability criteria in terms of weld toughness and weld strength mismatch have been identified.

440

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

It is suggested that this approach can be extrapolated to girth welded pipes withstanding an axial tension and/or an internal pressure (similar to Fig. 1, but containing a girth weld). The combination of these loads is likely to give rise to a predominantly biaxial load state, described by an axial stress ra and a hoop stress rh (both of tensile nature). Recall that rh uniquely results from internal pressure, whereas ra may be attributed to the internal pressure and to an externally applied load. In such case, BS7910 [8] suggests to perform separate assessments for the two hypothetical scenarios in which only one of the stress components (either ra or rh) is present. Both assessments should yield that plastic collapse is not attained to consider the damage as safely allowable. Focussing first on the scenario where ra > 0 and rh = 0, significant damage dimensions are a and 2c (L plays no role). BS7910 [8] provides the following reference stress solution for external surface flaws characterized by these dimensions:





experimental nature of this investigation. The nature of simulated damage (machined metal loss) has proven to be conservative with respect to a test on a girth weld containing actual weld corrosion. In conclusion, plastic collapse equations can be applied to assess the acceptability of girth weld metal loss in vintage pipelines. The equations cover the effects of pipe and flaw geometry and of weld misalignment. The effect of the latter is overpredicted by the linear-elastic equations in BS7910 [8]. An empirical concept of ‘effective misalignment’ is introduced and is supported by analyses using Neuber’s rule for elastic–plastic stress–strain concentrations. The analyses reveal a beneficial effect of high strain hardening and strength overmatching welds. The girth welds should be at least evenmatching in terms of ultimate tensile strength, and the minimum/average impact energy from a set of three Charpy V notch tests should be 30/40 J (full size equivalent).

 

1  Ba p þ 2 Ba sin 2c rref 2 rb   D þ ¼ ra 1  Ba p  2ca 3ð1  a00 Þ2 ra DB

ð10Þ

where similar to Eq. (6), a00 is a given dimensionless function of pipe and damage dimensions and rb/ra expresses the effect of misalignment. In absence of misalignment, Eq. (10) reduces to the Kastner equation (Eq. (2)). The advised contribution of misalignment (rb/ra) is again based on linear elasticity and, based on the MWP test results, it is expected that this correction can be relaxed by using the ‘effective misalignment’ concept. It is recognized that a full pipe may respond differently to weld misalignment due to the different constraint of surrounding material. Nonetheless, BS7910 [8] treats a pipe configuration with misaligned girth welds fairly similarly to a flat plate configuration: rb/ra = 3e/(B(1  m2)). We consider the lower resistance against flattening in a plate configuration (as observed in Fig. 10) to be a conservative aspect of wide plate testing over full pipe testing. Secondly, the scenario where rh > 0 and ra = 0 may be readily assessed using the ASME B31G [3] equation (Eq. (1)) (note that a similar expression is provided in BS7910 [8]). In this case, the governing damage parameters are a and L (not 2c). BS7910 would advise that there is no effect of weld misalignment in this assessment, as it is oriented in the plane perpendicular to the hoop direction. By inference, the applicability criteria for girth welds in axially loaded plates are also proposed for the assessment of girth welded pipes:

Acknowledgements The authors would like to acknowledge the financial support of Fluxys Belgium and the BOF (Special Research Fund – Ghent University, grant nr. BOF12/PDO/049). The assistance of the technical staff of Soete Laboratory is appreciated. The authors are grateful to Dr. Brian Leis for helpful discussions. Appendix A. Summary of MWP test results See Table A.1.

Table A.1 Summary of MWP test results.

 Charpy V notch energy should be at least 30 J/40 J (minimum/ average of three tests; full size equivalent values) for weld metal center as well as HAZ notched specimens, at the minimum design temperature.  The girth weld should be at least match the pipe steel in terms of ultimate tensile strength, and should connect vintage pipes characterized by high work hardening (low Y/T) and ductility. Further experimental work may prove useful to explicitly validate the suggested approach for girth welded pipes. 5. Conclusions Assessment equations for pipe corrosion have been considered for application to girth welds showing metal loss in vintage pipelines. To this purpose, an extensive experimental program comprising medium curved wide plate tests and supporting small scale tests has been executed and analyzed. Three girth weld characteristics (weld strength mismatch, misalignment, and Charpy V notch toughness) have been explicitly investigated, whereas a fourth element (weld residual stress) is implicitly covered by the

a

Test

Weld

a/B (%)

2c/W (%)

e/B (%)

Failure stress/yield strength weakest pipe (%)

Failure locationa

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

W1 W1 W2 W2 W2 W2 W3 W3 W3 W4 W4 W4 W4 W4 W5 W5 W6 W6 W7 W7 W7 W7 W8 W8 W8 W8 W9 W9 W9 W9 W10 W10 W10 W10

21.2 14.0 15.4 27.6 17.9 14.5 12.5 17.6 17.7 22.3 25.2 22.6 25.9 21.3 19.0b 10.9 35.8 18.3 11.4 16.5 29.7 27.7 8.7 18.5 27.7 19.3 8.8 18.8 28.5 26.8 5.5 19.5 30.6 31.2

100 60 100 100 60 100 100 100 60 100 100 100 60 60 N/A 60 100 100 100 100 100 60 100 100 100 60 100 100 100 60 100 100 60 100

4.3 22.8 1.4 8.7 8.6 8.8 5.3 25.5 30.5 8.0 4.0 4.7 15.6 19.8 8.0 12.3 18.3 6.2 3.7 5.3 22.2 22.4 6.1 1.6 10.9 10.5 0.2 1.5 0.1 2.1 4.4 15.3 7.5 11.1

114.3 111.8 131.1 110.5 130.3 133.7 116.1 110.4 112.0 91.8 93.7 90.8 95.6 96.5 127.9 132.0 91.8 121.1 133.1 123.2 82.4 110.3 148.1 127.6 102.2 127.6 135.3 118.6 105.9 122.6 144.8 118.0 117.8 107.5

W W W W W W B W W W W W W W W W W W W W W W W W W W B W W W B W W W

W: damaged weld; B: base metal. Naturally corroded weld; value mentioned is peak value recorded over specimen width. b

S. Hertelé et al. / Engineering Structures 124 (2016) 429–441

References [1] Zhang C, Cheng YF. Corrosion of welded X100 pipeline steel in a near-neutral pH solution. J Mater Eng Perform 2010;19:834–40. [2] Eliyan FF, Alfantazi A. Corrosion of the heat-affected zones (HAZs) of API-X100 pipeline steel in dilute bicarbonate solutions at 90 °C – an electrochemical evaluation. Corros Sci 2013;74:297–307. [3] ASME B31G. Manual for determining the remaining strength of corroded pipelines. New York, USA: American Society of Mechanical Engineers; 2012. [4] Leis BN, Clark EB, Zhu XK, Galliher RD. Guidelines for assessing corrosion associated with girth and long-seam welds. Gas Technology Institute, USA, report GRI-04/0119; 2004. p. 100. [5] Folias ES. Failure of pressurized vessels. In: Cherepanov GP, editor. Fracture – a topical encyclopedia of current knowledge. Malabar, Florida, USA: Krieger Publishing Company; 1998. p. 275–88. [6] Martin M, Andrews RM, Chauhan V. The remaining strength of corroded low toughness pipe. GL Industrial Services Ltd., report 9247 issue 3.0 (PHMSA DTPH56-05-T-0003); 2009. [7] Kastner W, Röhrich E, Schmitt W, Steinbuch R. Critical crack sizes in ductile piping. Int J Press Vess Pip 1981;9:197–219. [8] BS 7910:2013+A1:2015. Incorporating corrigendum No. 1. Guide to methods for assessing the acceptability of flaws in metallic structures. London, UK: British Standards Institution. [9] Roovers P, Swankie T, Hertelé S. Assessment of corrosion associated with the girth or long seam weld in vintage pipelines (EPRG project 177/2014). In: Denys R, Hertelé S, Tiratsoo J, editors. Proc 1st Ageing Pipelines Conf, Ostend, Belgium; 2015. [10] API 5L. Specification for line pipe. Washington, USA: American Petroleum Institute; 2012. [11] Hertelé S, De Waele W, Denys R, Verstraete M. Investigation of strain measurements in (curved) wide plate specimens using digital image

[12]

[13]

[14]

[15]

[16]

[17] [18] [19]

[20]

[21] [22]

441

correlation and finite element analysis. J Strain Anal Eng Des 2012;47:276–88. Sutton MA, Orteu JJ, Schreier HW. Image correlation for shape, motion and deformation measurements. Basic concepts, theory and applications. New York: Springer; 2009. ISO 6892-1. Metallic materials – tensile testing – part 1: method of test at room temperature. Geneva, Switzerland: International Organization for Standardization; 2009. Hertelé S, De Waele W, Denys R, Verstraete M. Full-range stress–strain behaviour of contemporary pipeline steels: part I. Model description. Int J Press Vess Pip 2012;92:34–40. ISO 148-1. Metallic materials – Charpy pendulum impact test – part 1: test method. Geneva, Switzerland: International Organization for Standardization; 2009. Andrews RM, Denys RM, Knauf G, Zarea M. EPRG guidelines on the assessment of defects in transmission pipeline girth welds – revision 2014. J Pipeline Eng 2015;14(1):9–21. Ainsworth RA. The assessment of defects in structures of strain hardening material. Eng Fract Mech 1984;19(4):633–42. CSA Z662. Oil and gas pipeline systems. Toronto, Canada: Canadian Standards Association; 2015. Hertelé S, Verstraete M, Denys R, De Waele W. Global collapse and J integral analysis for inner-diameter defected curved plates in tension. Int J Press Vess Pip 2013;111–112:44–53. Willoughby AA, Davey TG. Plastic collapse at part wall flaws in plates. In: Wei RP, editor. Fracture mechanics: perspectives and directions. Proc 20th Nat Symp ASTM STP 1020, Bethlehem, Pennsylvania, USA; 1989. Neuber H. Theory of stress concentration for shear strained prismatical bodies with arbitrary non linear stress strain law. J Appl Mech 1961;28(4):544–50. DNV-OS-F101. Offshore standard – submarine pipeline systems. Høvik, Norway: DNV GL; 2013.