[Skip to content]

.

Comparison of API 1104 Appendix A and BS 7910 Procedures for the assessment of girth weld flaws

Sarah E Smith and Henryk G Pisarski

TWI Ltd

Paper presented at 5th International Pipeline Technology Conference, Ostend, Belgium, 12 - 14 October 2009.

Abstract

The 2007 revision to API 1104 Appendix A for assessing flaws in pipeline girth welds is compared with BS 7910 Level 2 assessment procedures. Results from full-scale pipe bend tests and wide plate tests are used to plot assessment points for both procedures. Both procedures predicted failure but by widely ranging margins of safety. Generally, there was little difference between the two procedures, although assessments to BS 7910 were slightly closer to the assessment curve than those to API 1104. Assessments to API 1104 were more likely to predict failure by plastic collapse than those to BS 7910. Example cases were undertaken to compare predicted tolerable flaw sizes using each procedure.

Introduction

The US Department of Transport and PRCI funded a programme to update API 1104 Appendix A, details of which were published at the IPC 2006 conference.[Wang et al, 2006] The procedure was published by API in July 2007, replacing Appendix A of the 2005 edition of API 1104. It describes a three-tier approach with Options 1-3 assessments for fracture mechanics-based analysis of pipeline girth welds. Option 1 is a graphical approach, while Option 2 is based on a failure assessment curve (FAC) which is similar to the failure assessment diagram (FAD) in BS 7910. The approach has been validated using historical data, including full-scale tests.[Wang et al, 2006] Option 3 is applicable when the pipeline is subjected to fatigue. It does not describe an assessment method but refers to BS 7910.

The new API 1104 rules[API, 2007] are compared with the previous rules [API, 2007] and BS 7910.[BSI, 2005] Validation data for the new acceptance criteria against full-scale pipe bend tests on girth welds is examined.[Coote et al, 1986] The acceptance criteria are also evaluated against wide plate test results.[Denys et al, 2000] Examples showing the effects of including and not including pipe misalignment (hi-lo) and welding residual stresses at girth welds on flaw assessments are given.

Background

The fitness-for-purpose criteria in Appendix A of API 1104[API 1104, 2005] have been revised to account for the actual crack tip opening displacement (CTOD) of the material, and the applied stress and material strength. There are now three options for fitness-for-purpose assessments.[API 1104, 2007] The simpler Option 1 is a graphical method. For a ratio of applied stress to flow stress, referred to as 'load level' Pr, for the case under consideration the ratios of flaw height to wall thickness (WT) and flaw length to outside diameter (OD) are determined. There are two sets of curves, each for a specific CTOD value. The actual CTOD of the material must be equal to or higher than the CTOD for the curve used. Although this method is the simpler of the two, it does not consider the benefits of a higher CTOD than 0.25mm or the consequences of a lower CTOD than 0.1mm. Option 2 is a FAD-based method. The toughness ratio Kr is plotted against the stress ratio Lr, and the point is acceptable if it lies inside the FAD. Although this method is more complex, it allows the CTOD determined from the material to be used. Options 1 and 2 are specifically designed for assessing surface flaws, but embedded flaws are assessed as surface flaws of the same height. Option 3 is applicable where crack growth by fatigue is expected to be significant, and recommends the use of validated fitness-for-purpose procedures, like BS 7910, to develop acceptance criteria.

Full scale pipe bend tests

Test data were obtained from 38 full-scale pipe tests performed on girth welds and pipes containing circumferential flaws.[Pick et al, 1980, and Coote et al, 1986] Details of the test specimens and assessment input details, including yield strength and fracture toughness in terms of CTOD are presented in Table 1. It can be seen that in some cases the CTOD of the material was below 0.05mm, the minimum specified by API 1104. A bending load was applied to a pipe with an artificial flaw located at the position of maximum tensile stress. The strain at failure was measured and the failure stress was calculated from the applied bending moment. Although stress and strain at failure were reported, the bending moment was not. Since the stress was derived from the bending moment, failure stresses greater than the yield strength could be overestimated. Since the flaw size was known, and the aim of the calculations was to predict failure stress and the position of the failure on the FAD the safety factor required by the new API 1104 Appendix A was not applied in the calculations. When estimating tolerable flaw sizes, API 1004 Appendix A Option 2 requires a 'safety factor' of 1.5 to be applied to flaw length. Assessments to BS 7910 (Level 2A FAD) were conducted using TWI software Crackwise 4. The reference stress, used to calculate the plastic collapse, Lr, axis of the FAD was derived from the Kastner equation. For comparison purposes the assessments were conducted assuming zero residual stress; the API 1104 procedure does not include a method for incorporating residual stresses.


Table 1 Full-scale test details from [Coote et al, 1986]

Test
No
Dimensions
OD, in x WT,
mm
Yield strength,
MPa
CTOD,
mm
a,
mm
2c,
mm
Stress estimated
from bending
moment at failure
MPa
Strain at
failure,
%
Notes
4 36 x 11.1 531 0.1 5.9 63.5 >723 >0.5 1
5 36 x 11.1 531 0.1 5.5 69.8 >723 >0.5  
6 36 x 11.1 531 0.03 7.8 68.6 591 0.33 1,2
7 36 x 11.1 531 0.03 5.3 61 570 0.31 2
8 36 x 11.1 531 0.03 10.1 76.5 270 0.14 1,2
9 36 x 11.1 531 0.1 8.8 81.8 655 0.48 1
10 36 x 11.1 531 0.03 6.4 59.3 470 0.23 1,2
11 36 x 11.1 531 0.1 9.3 79 655 0.49 1
12 36 x 11.1 531 0.1 6.3 63.5 >755 >0.5 1
13 36 x 11.1 531 0.1 6.1 59.6 612 0.36 1
14 36 x 11.1 531 0.1 5.5 64.8 >726 >0.51  
15 36 x 11.1 531 0.1 5.5 60.3 683 0.47  
16 36 x 10.28 689 0.1 4.1 300 >690 >0.78  
17 36 x 10.28 689 0.1 3.6 300 >690 >0.71  
18 36 x 11.1 531 0.1 3.3 265 569 0.3  
19 36 x 11.1 531 0.1 3.2 278 635 0.37  
21 36 x 11.1 466 0.1 3.9 279 411 0.2  
22 36 x 11.1 466 0.1 3.7 331 390 0.2  
23 36 x 11.1 466 0.1 3.5 75 655 0.49  
24 42 x 15 496 0.1 0.9 14 >606 >0.7  
25 42 x 15 496 0.1 3 38 >606 >0.8  
26 42 x 15 496 0.1 8 70 >606 >0.62 1
27 36 x 11.1 441 0.23 3.7 315 527 0.27  
28 36 x 11.1 466 0.1 3.1 282 470 0.32  
29 36 x 11.72 470 0.1 2.9 280 533 0.35  
30 36 x 11.72 470 0.1 3.7 134 461 0.31  
31 36 x 11.72 470 0.1 2.2 116 656 0.65  
32 24 x 6.76 532 0.08 3.1 100 487 0.195  
33 24 x 6.76 532 0.08 2.8 199 427 0.201  
34 24 x 6.76 532 0.08 3.1 51 542 0.275  
35 24 x 6.76 532 0.08 3.9 107 513 0.165 1
46 36 x 11.7 460 0.1 2 112 635 >0.75  
47 36 x 11.7 460 0.1 3.9 141 629 0.35  
48 36 x 11.7 460 0.1 3.5 300 462 0.26  
49 30 x 19 472 0.04 3.48 105 566 0.38 2
50 30 x 19 472 0.06 3.73 139 595 0.46  
51 30 x 19 472 0.09 5 125 520 0.45  
52 28 x 24.4 470 0.08 10.9 127 586 0.6  

Notes:
1 Flaw larger than API allows (height greater than 0.5 WT).
2 CTOD below 0.05mm (the minimum specified by Option 2 of the new API 1104).

The results are shown plotted on the FAD in Figure 1 for assessments performed according to both the new API 1104 Appendix A and to BS 7910. All of the assessment points are outside the FAD, including those with low CTOD, and most are in the elastic-plastic area, nearer the plastic collapse region, ie Lr≥1. These include several cases where the original crack size was greater than the limits imposed by API 1104, and several cases where the CTOD was below the 0.05mm limit. Comparing the results from the API 1104 Appendix A procedure with those from BS 7910, both show wide scatter. The points closest to the line were obtained from the BS 7910 procedure, but the points farthest from the line were also obtained using the BS 7910 procedure. The API 1104 results also imply in most cases that the failure will be less brittle and more controlled by plastic collapse than that predicted by BS 7910.

Fig.1. FAD for full-scale results pipe tests assessed using API 1104 Appendix A 2007, Option 2 and BS 7910 Level 2A procedures
Fig.1. FAD for full-scale results pipe tests assessed using API 1104 Appendix A 2007, Option 2 and BS 7910 Level 2A procedures


Figure 2 shows the failure stress predicted by API 1104 and BS 7910 compared with the actual failure stress in each case. As would be expected, in all cases the predicted failure stress was below the actual failure stress. There is also slightly more scatter in predictions with BS 7910 compared with API 1104. In several cases the failure stress predicted by BS 7910 was close to the measured failure stress, as would be expected since several points were close to the line in the FAD. Since the safety factor required by the API 1104 Appendix A procedure was not applied in the calculations, the conservatism of the assessment would be increased had the safety factor been included.

Fig.2. Predicted failure stress versus actual failure stress for full-scale pipe tests
Fig.2. Predicted failure stress versus actual failure stress for full-scale pipe tests


Wide plate tests

Wide plate test data from 15 tests were obtained from a paper by [Denys et al (2000)]. Girth welds in pipe (Grades X70 and X65) of 42 and 48in diameter were tested. Details of the test specimens including yield strength, CTOD and initial crack dimensions are given in Table 2. Because the flaw size was known, and the aim of the calculations, in this case, was not to predict tolerable flaw sizes, the safety factor required by Option 2 of the new API 1104 Appendix A was not applied. Assessments to BS 7910 (Level 2A FAD) were conducted using the flat plate solutions for the estimates of Kr and Lr axes in the FAD and assuming zero residual stresses.


Table 2 Test specimen details from [Denys et al, 2000]

Test Pipe details Yield strength,
MPa
CTOD
mm
a
mm
2c
mm
Stress at failure,
MPa
Strain at failure,
%
CA1 X70 48" x 16.9mm 498 0.093 5 133 552 1.29
CA2 4.25 430 535 0.83
CA3 8.5 86 556 1.55
CB1 X70 48" x 16.9mm 498 0.131 5 133 524 0.88
CB2 4.25 132 524 1.01
CB3 8.5 86 526 1.09
CC1 X70 48" x 16.9mm 498 0.448 5 133 547 1.15
CC2 4.25 430 519 0.76
CC3 8.5 86 546 1.29
CD1 X65 42" x 30.5mm 470 0.412 5 226 548 2.85
CD2 7.65 241 517 1.09
CD3 7.65 253 522 1.46
CE1 X65 42" x 30.5mm 470 0.382 5 226 542 2.25
CE2 7.7 241 504 1.10
CE3 7.65 253 517 1.33


The FAD with assessment points calculated according to the new API 1104 (Option 2 assessment) and BS 7910 (Level 2 assessment) is shown in Figure 3 for the X70 and X65 materials. In all cases, the assessment points are outside the FAD.

Fig.3. FAD for wide plate tests on X70 and X65 material, assessed according to API 1104 Appendix A 2007, Option 2 and BS 7910 Level 2A procedures
Fig.3. FAD for wide plate tests on X70 and X65 material, assessed according to API 1104 Appendix A 2007, Option 2 and BS 7910 Level 2A procedures

 

As noted previously, overall the new API 1104 assessments tend to cluster in the plastic collapse region of the FAD (Lr>1) more than the BS 7910 assessments. Again the BS 7910 assessments are more scattered than the API 1104 but this is marginal. From Figure 3, the BS 7910 assessment points for X70 material appear in some cases to be more conservative, ie further away from the FAD assessment line, than those of API 1104. On the other hand, for the X65 pipe, the BS 7910 results seem slightly less conservative, nearer the FAD.

The failure stress in each case predicted by the new API 1104 Appendix A procedure and by BS 7910 is shown plotted against the measured stress at failure in Figure 4. The predicted failure stress for both procedures was below the measured failure stress in all cases. Again, slightly more scatter is observed in the BS 7910 results, especially for the higher strength. However, as noted previously, the API 1104 Appendix A assessments exclude the safety factor on crack length so the results could be more conservative than indicated in Figures 3 and 4, had the safety factor been included.

Fig.4. Predicted failure stress versus actual failure stress for wide plate tests
Fig.4. Predicted failure stress versus actual failure stress for wide plate tests

 

Effects of welding residual stress and girth weld misalignment - example case

In order to compare the API 1104 Option 2 assessment method with BS 7910 Level 2A assessment procedures, a theoretical case was considered. This was a 42in OD pipe with 22mm WT with the specified properties of Grade X65. A CTOD of 0.5mm was assumed, and an applied axial membrane stress of 85% SMYS. Several assessments were carried out to BS 7910: one case assuming residual stress to be at initially at yield but allowed to relax depending on the applied stress, in accordance with BS 7910 procedures; one case assumed zero residual stress. Two further cases were considered, similar to the first two but with 1mm misalignment. The stress concentration factor caused by the misalignment was 1.15.

Figure 5 shows the results in terms of curves of flaw height against predicted tolerable flaw length. Calculations to BS 7910 Level 2A assuming zero residual stresses predicted the largest tolerable flaw dimensions in all cases. When residual stresses were considered the BS 7910 procedure was more conservative than the new API 1104 methods for deeper flaws (with depth from about 7.5 to 11mm, the maximum flaw height considered here). For shallower flaws however, the new API 1104 predicted more conservative tolerable flaw lengths. Applying the required safety factor means that results for flaws with height less than 7.5mm are more conservative than BS 7910; without the safety factor, flaws with height less than about 7mm are more conservative.

Fig.5. Comparison of predicted maximum tolerable surface flaw sizes for in a 42in OD x 22mm WT, X65 pipe using different assessment procedures
Fig.5. Comparison of predicted maximum tolerable surface flaw sizes for in a 42in OD x 22mm WT, X65 pipe using different assessment procedures


When girth weld misalignment was considered the BS 7910 procedures reduced the predicted tolerable flaw sizes, as expected. This was most noticeable when yield strength magnitude residual stresses were also assumed. In this case smaller flaw sizes for flaw height up to about 5.5mm were predicted using BS 7910 than with API 1104. When residual stresses were assumed to be zero (with weld misalignment of 1mm) then, as can be seen in Figure 5, larger flaws were predicted to be tolerable, in comparison to the case where residual welding stresses were included. This means that the point where the new API procedure (with the safety factor applied) predicted smaller tolerable flaw lengths for a given flaw height was at flaw heights of around 8mm. However it should be noted that the API 1104 procedure does not make any allowance for misalignment or residual stresses.

The old API 1104 method was most conservative, except for longer flaws, where the limits on flaw length applied by the new API 1104 were less than those for the previous API 1104.

Discussion

Results from 38 full-scale bend tests were evaluated using Option 2 of the new Appendix A to API 1104. These included some specimens where the original flaw size was larger than the limits specified in API 1104 and some specimens with fracture toughness below the specified 0.05mm limit. All of the assessment points for these tests were outside the FAD. The API 1104 assessment points were nearer the plastic collapse region of the FAD than those from BS 7910. Since these tests were assessed without applying the safety factor and were outside the FAD, this implies that in cases where the safety factor is applied, the conservatism of the new API 1104 Appendix A procedure will be increased (and the accuracy reduced).

Similar observations were made for wide plate test data for X70 material. All were outside the FAD, with the API 1104 results nearer the plastic collapse region. The wide plate tests for X65 material were also outside the FAD, with the results of the API 1104 assessments in the plastic collapse region while the BS 7910 Level 2 assessment points were at the knee of the FAD. For the X65 material however, the API 1104 results were more conservative. These tests were also assessed without applying the safety factor required by the new version of API 1104, meaning that predicted flaw dimensions with the safety factor applied should be even more conservative.

A theoretical case was considered, using the specified properties of X65 material, to compare tolerable flaw lengths predicted using the old and new versions of API 1104, and using BS 7910 Level 2. For shallower flaws API 1104 Option 2 predicted smaller tolerable flaw lengths than BS 7910. For deeper flaws, BS 7910 predicted smaller tolerable flaw lengths. The new API 1104 Appendix A assessment predicted larger tolerable flaw lengths than the old version for deeper flaws. When the effects of residual stresses were ignored, BS 7910 predicted larger tolerable flaw dimensions in all cases. The API 1104 procedure has no means of including the effects of residual stress. Since tensile residuals are known to contribute to the risk of brittle fracture, ignoring them could result in unsafe predictions of acceptable flaw size. Although API 1104 claims to avoid this by specifying a minimum CTOD of 0.05mm, the approach seems rather arbitrary especially when applied to high strength steels, and large wall thickness, where tensile residual stresses could be significant.

It was found that BS 7910 was more conservative than the new API 1104 Appendix A procedure when misalignment was considered. In this case only one level of misalignment (of 1mm) was considered and for cases where higher levels of misalignment are present the difference between the procedures would increase. As there is no way of accounting for misalignment in an assessment to API 1104, this suggests it could potentially be non-conservative where high levels of misalignment are present.

The effect of misalignment at the girth weld was considered in further detail by comparing the driving force CTOD versus applied axial strain curves derived from a 3-D finite element analysis with those from the API 1104 Appendix A and BS 7910 procedures, see Figure 6.The analyses were conducted for a 400mm OD x 20mm WT pipe in bending with a circumferential surface crack 3 x 50mm on the OD. The pipe had a yield strength (at 0.5% strain) of 465MPa and tensile strength of 531MPa. The analyses were conducted assuming a weld width of 10mm, even strength mismatch with respect to the parent pipe and misalignment (e) of 0 and 1.5mm. In these analyses, strain for a given applied stress was derived from the pipe material stress-strain curve.

Figure 6 shows that misalignment increases the CTOD driving force and the difference increases significantly as yield strain is approached and exceeded. Comparison of the finite element results with predictions made using the BS 7910 Level 2B (material specific) FAD and API 1104 FAC assessment methods all show similar behaviour up to about 0.5% strain for zero misalignment. The versatility of the BS 7910 procedure enables higher strains to be analysed. For the case analysed, the BS 7910 procedure starts to become unconservative for strains above approximately 0.5%.

 Fig.6. Comparison of CTOD driving force curves for a 3 x 50mm circumferential surface crack in a 400mm OD x 20mm WT pipe with misalignment (e) of 0 and 1.5mm in a girth weld of width 10mm
Fig.6. Comparison of CTOD driving force curves for a 3 x 50mm circumferential surface crack in a 400mm OD x 20mm WT pipe with misalignment (e) of 0 and 1.5mm in a girth weld of width 10mm

However, by including the stress concentration caused by misalignment in the BS 7910 procedure, the predicted driving force CTOD is increased significantly and is conservative with respect to the finite element analysis for strains up to at least 1.1%. The stress concentration was treated as a local bending stress of which 15% was a primary stress (ie contributing to both fracture and plastic collapse axes of the FAD) whilst 85% was treated as a secondary stress (ie contributing to the fracture axis of the FAD only). The background to this approach is described elsewhere.[Cheaitani, 2009] Again, the versatility of the BS 7910 approach enables the effects of misalignment to be captured; these become increasingly important at installation/service strains approaching and exceeding yield. In contrast, the API 1104 procedure ignores misalignment and becomes increasingly non-conservative at higher strains, in cases above approximately 0.2% strain or 90% of yield strength. API 1104 Appendix A states that its use is restricted to conditions where the maximum axial design stress is no greater than the SMYS and the maximum axial design strain is no greater than 0.5%.

The new API 1104 method has several advantages over the old method. The actual fracture toughness of the material is considered, as is the material strength. The possibility of failure by plastic collapse is also taken into consideration. However, some disadvantages remain: misalignment at the weld is not considered, nor are other variations in pipe geometry, and there is no way to use the specific stress-strain curve of the material, or to allow for any yield discontinuity. This could result in unsafe predictions of allowable flaw size in materials which exhibit a discontinuous yielding or a Lüders plateau, such as seamless pipe. In addition, while API 1104 Appendix A states that residual stresses are accounted for by specifying minimum CTOD and Charpy energy values, no mention is made of the magnitude of residual stresses, or of the possibility of relaxation.

These place severe limitations on the versatility of the API 1104 approach which are not present with the BS 7910 procedure. Indeed, the limitations have the potential of missing important features about the condition of the girth weld which could lead to unsafe estimates of acceptable flaw size to be made.

Concluding remarks

Results from full-scale bend tests and wide plate tests from girth welds in X65 and X70 pipe were evaluated using Option 2 of the new Appendix A and BS 7910 Level 2A procedures assuming zero residual stresses. Both procedures predicted failure but by widely ranging margins of safety. Generally, there was little difference between the two procedures, although assessments to BS 7910 were slightly closer to the assessment curve than those to API 1104. Assessments to API 1104 were more likely to predict failure by plastic collapse than those to BS 7910, although these calculations did not include the safety factor on crack length required by the new API 1104.

Example cases were considered to compare tolerable flaw lengths predicted using the old and new versions of API 1104 and BS 7910 Level 2. For shallower flaws both versions of API 1104 predicted smaller tolerable flaw lengths than BS 7910. For deeper flaws, BS 7910 predicted smaller tolerable flaw lengths. When residual stresses were assumed to be zero, BS 7910 predicted larger tolerable flaw dimensions. However, in thick section welds and welds in high strength steels, significant tensile residual stresses are likely to be present which will increase the risk of fracture. Ignoring them could result in unsafe predictions of tolerable flaw size.

Since BS 7910 is able to specifically include stress concentrations arising from misalignment (and other deviations from intended geometry), smaller tolerable flaws were predicted than both versions of API 1104. As there is no way of accounting for misalignment in an assessment to API 1104, this indicates that it could be potentially non-conservative where misalignment or other deviations from intended geometry are present. The effect of misalignment in increasing the CTOD driving force (or CTOD requirement to avoid failure) was illustrated for a circumferential crack in pipe using the BS 7910 and API 1104 Appendix A procedures, and comparison with finite element analyses (see Figure 6). For the case considered (1.5mm misalignment) the API 1104 procedure became non-conservative at strains exceeding 0.2% or 90% of yield strength. The degree of non-conservatism will increase as misalignment is increased. On the other hand, the BS 7910 procedure had the ability to provide conservative but realistic estimates of the CTOD driving force curve to at least 1.1% strain.

Although the new API 1104 method has advantages over the old some significant disadvantages remain: misalignment of pipes at girth welds is ignored, there is inadequate treatment of residual stresses and the shape of the stress-strain curve, but especially yield discontinuity, is not considered. These factors limit the versatility of the procedure but can unwittingly result in unsafe predictions of acceptable flaw size. These are not limitations with the BS 7910 procedure. Consequently for critical applications it is recommended that BS 7910 assessment procedures are considered instead.

References

API 1104, 2005: 'Welding of pipelines and related facilities', 20th edition, American Petroleum Institute.

API 1104 2007: 'Welding of pipelines and related facilities', 20th edition, amended July 2007, American Petroleum Institute.

BS 7910, 2005: Incorporating Amendment No 1, 'Guide to methods for assessing the acceptability of flaws in metallic structures', British Standards Institution.

Coote R I, Glover A G, Pick R J and Burns D J, 1986: 'Alternative girth weld acceptance standards in the Canadian gas pipeline code', Third International Conference, Welding and Performance of Pipelines, November, Volume 1, Paper 21, pp21-33.

Cheaitani M J, 2009: TWI, to be published.

Denys R M, Lefevre A A, De Jaeger C and Claessens S, 2000: 'Failure characteristics and defect tolerance levels of girth welds in large diameter X65/X70 steel linepipes: experimental verification through wide plate testing and comparison with ECA prediction models', Pipeline Technology, May, Vol.1.

FITNET, 2008: 'FITNET Fitness-for-service procedure'. Revision MK8. FITNET Thematic Network.

Glover A G, Coote R I and Pick R J, 1981: 'Engineering critical assessment of pipeline girth welds', Proceedings of Conference on Fitness-for-Purpose Validation of Welded Constructions, The Welding Institute, Paper 30, 17-19 November.

Pick R J, Glover A G and Coote R I, 1980: 'Full-scale testing of large diameter pipelines', Proceedings of Conference on Pipeline and Energy Plant Piping: Design and Technology, Welding Institute of Canada.

Wang Y-Y, Liu M, Horsley D and Bauman G, 2006: 'A tiered approach to girth weld defect acceptance criteria for stress-based design of pipelines', Proceedings of IPC2006, 6th International Conference.