Paper presented at ISOPE 2007, Seventeenth International Offshore and Polar Engineering Conference, Lisbon, 1-7 July 2007.
Numerical finite element analyses (FE) have been undertaken of circumferential flaws in the girth weld of a pipeline subjected to plastic straining simulating installation by reeling, and J driving force curves derived. These were then compared with J driving force curves derived using the guidelines described in DNV-RP-F108 intended for installation methods involving repeated plastic straining. The assessment procedure is based on BS 7910 FAD methods. Strains up to 2% were considered. Analyses were undertaken for surface breaking weld root flaws located at the weld fusion boundary. The pipe was nominally to API 5L Grade X65 strength, with a diameter of 323.9mm (12¾in) and wall thickness of 20.6mm. Conclusions are drawn concerning the conservatism in flaw assessment procedures conducted to DNV-RP-F108 and the treatment of welding residual stress.
The definition of rational flaw acceptance criteria for girth welds in pipelines subjected to axial straining within the context of existing codified fracture mechanics based assessment procedures is problematic since these are essentially stress based. Although there is no fundamental problem in using such procedures, the solutions provided are not always the most suitable for strain based assessments. Nevertheless, with appropriate modifications,assessments based on BS 7910  procedures have been used successfully for a number of years to set acceptance criteria for pipeline installation methods involving plastic straining such as by pipe reeling. The flaw acceptance criteria provided by these methods have, in many cases, enabled larger flaws to be accepted than those based on workmanship standards, such as BS 4515 and API 1104. The benefits to the industry have included reduced repair rates without loss of integrity and increase in lay rates. The demonstration of integrity by means of an existing and well established assessment procedure is considered important since it enables independent third party verification to be undertaken.
Another benefit to industry in using fracture mechanics based assessment methods is that information provided by automated ultrasonic testing on flaw size (height and length) can be assessed properly, since codes such as API 1104and BS 4515 provide workmanship acceptance criteria based on flaw length not height. Nevertheless, for certain situations involving plastic straining, flaw sizes predicted by fracture mechanics procedures can be smaller than those based on workmanship standards. Although it could be argued that there is nothing inherently wrong with such a conclusion because the workmanship criteria are intended for installation methods not involving plastic straining, industry experience indicates that flaw tolerance is better than predicted by fracture mechanics analyses.
In order to address these concerns, a joint industry programme was conducted by DNV-TWI-SINTEF to provide guidelines for fracture control of sub-sea pipelines installation methods involving cyclic plastic straining. The guideline was issued to sponsors in 2003 and this document  provided the basis for a DNV Recommended Practice which was published in 2006.  The flaw assessment procedure, referred to hereafter as the reeling procedure, is based on BS 7910 but with adjustments to make it suitable for plastic straining conditions. Novel features of the procedure include the use of single edge notched tension specimens to define fracture toughness and use of sub-scale segment specimens to validate, by experiment, the procedure used to generate acceptable flaw sizes. The procedure has been used successfully inmany pipeline installation projects and also applied to in-service assessment of pipelines subjected to plastic straining, e.g. due to lateral bucking and ground movement.
The purpose of this paper is to establish margins against possible failure when using the procedure as currently formulated in DNV-RP-F108. This is done by comparing the J driving force curves predicted by the current procedure, fora given pipe size and range of flaw sizes, with those obtained by numerical finite element analyses (FE).
Pipe case modelled
Analyses conducted are representative of pipe that would be would typically be installed by reeling. Although this would involve a series of repeated plastic strains during the installation phase, the case considered here is the application of a single monotonic strain when the pipe is first wound onto the reel. The pipe has an outside diameter of 323.9mm (12¾in) and wall thickness of 20.6mm and nominally to API 5L Grade X65 strength. The 0.2% offset yield strength of the parent pipe in the longitudinal direction considered in these analyses is 485MPa and the tensile strength is 594MPa. The pipe contains a girth weld which has a 0.2% offset yield and tensile strengths, which overmatch the corresponding parent pipe properties by approximately 20%. For the purposes of these analyses, the work hardening rates of both materials are assumed to be the same. The engineering stress-strain curves are shown in Fig.1.
The pre-existing flaws considered were internal surface flaws located at the weld root region of the weld fusion boundary. The nominal weld width chosen for this location was 8mm and is representative of the narrow weld preparations typically employed for these applications. The HAZ strength properties were not specifically modelled. Two series of analyses were undertaken; the first assumed a homogeneous material characterized by the parent pipe tensile properties, viz the weld metal had the same stress-strain curve as the pipe. The second included the girth weld with a higher strength so that the effects of strength overmatching were modelled.
Straining of the pipe when it is bent onto the reel ('reeling-on') was modelled by restraining the leading pipe and applying a bending moment to the trailing pipe which contains the girth weld and flaw. In the model, rotation is applied to a master surface in the axial-vertical plane. This is consistent with a pure bending moment when deformation is purely elastic. This is illustrated in Fig.2. It is acknowledged that the assumption that plane sections remain plane becomes less accurate as plasticity develops. However, it is considered that this error is small for the low strains employed in these analyses, which are typically up to 2%. ABAQUS Version 6.6 was used to carry out the analyses using a small strain formulation and 20-noded, quadratic three dimensional elements (ABAQUS element C3D20R). The J-integral was calculated in a 1mmsquare 'box' of focused mesh incorporating 12 rings of elements surrounding each node on the crack border. In weld strength mismatch cases, the J-integral calculations were taken along contours that were entirely within one material to avoid contours crossing material boundaries; in this case the crack was located just inside the weld metal very close to the fusion boundary. The analyses showed an increasing path-dependence for J as plasticity developed. However, ignoring the innermost 4 rings of elements, the J-integral results from the outermost 5th to 12th rings were path-independent to within 8% of the maximum values of J. The reported values of J refer to results from the 10th contour. The applied strain was defined at one diameter away from the flaw.
The weld profile was treated as parallel sided with the flaw located at the interface/fusion boundary. The surface breaking flaws considered were semi-elliptical in shape with the following heights and lengths: 3x50, 6x25, 6x50 and9x90mm.
Analyses according to DNV-RP-F108
Analyses of the conditions described above were conducted according to DNV-RP-F108 to generate J driving force curve as a function of applied strain. As already explained, the assessment procedure is based on BS 7910 and Level 2B/3BFAD based method was used which employs the parent material stress-strain curve. The method assumes that the weld metal strength overmatches that of the parent pipe but no advantage of this is taken in the analysis. Linear elastic stress intensity factor solutions based on the Newman and Raju equations provided in BS 7910 were employed and plastic collapse was assessed using a reference stress solution based on the Kastner model.  This is recommended in BS 7910 for circumferential flaws in pipes and includes a modification to enable membrane and through-wall bending stresses to be included. Analyses were conducted for a series of strains. Since strain is not a direct input into BS 7910:2005 assessments, the corresponding stress was taken from the parent material stress-strain curve, as recommended in the reeling procedure.  However, in these analyses the true stress true strain curve was used instead of the engineering curve. For small strains less than about 1.5% differences in the stresses between the two curves are relatively small. The stress obtained was assumed to be constant through the pipe wall.
For each of the flaws considered two J driving force curves were obtained. In order to make a direct comparison with the results from the numerical analyses, the first curve was estimated with no allowance for welding residuals tresses. The second curve was obtained assuming that an initial welding residual stress, of yield magnitude, exists in the region containing the flaw. The second curve represents what would be calculated, according to BS 7910 and the reeling procedure, when the residual stress is included in the assessment as a secondary stress (i.e. contributing to fracture but not plastic collapse) in addition to the applied stress. The initial residual stress was allowed to relax, as the applied stress increased, according to a simple rule recommended in BS 7910. This rule does not allow the residual stresses to relax to zero; the minimum value is 40% of yield. This is highly likely to be a conservative estimate, since straining to around 2% would relax the residual stresses transverse to the girth weld to low levels. Nevertheless, the inclusion of residual stresses as described above can be considered to provide a factor of safety,the significance of which will be explained later.
Results and discussion
Comparison of J driving force curves
Figures 3-6 show the relationship between the J driving force and remote applied strain predicted by numerical analysis (labeled as 'J (FE)') and according to the reeling procedure (labeled as 'DNV-RP-F108') for the four flawsizes considered. Results are shown from numerical analyses obtained on homogenous material (or an 'even strength matched' condition where the tensile properties of the weld metal are the same as those of the parent pipe) and, for the3x50 and 6x50mm flaws, on models where the tensile properties of the weld metal overmatch those of the parent pipe by approximately 20%. Both sets of numerical analyses were conducted without including welding residual stresses. Two sets of results based on the reeling procedure, obtained with and without including welding residual stresses as specified above, are shown (both assume homogeneous material).
Focusing first on results obtained assuming homogeneous material without including welding residual stresses, it can be seen that when the flaw height is less than 40% of the wall thickness ( Figs 3-5), the reeling procedure consistently underestimates the J driving force for applied strains greater than about 0.75%. When the strains are less than 0.75%, there is good agreement in predicted J values for the two analysis methods. For a deep (>40% of wall thickness) long flaw (90mm), the reeling analysis over-predicts the driving force J ( Fig.6).
Generally, the reeling procedure is used to justify acceptance of relatively small flaws, certainly less than half way through the wall thickness, therefore, the non-conservative nature of the reeling analyses when the pipe is subject to plastic straining is of concern. However, there are mitigating factors that reduce the possibility of obtaining non-conserve assessments and these are discussed next.
Factors affecting the reeling procedure driving force curves
One of the requirements of the reeling procedure  is that the weld metal strength must overmatch that of the parent pipe. This is important because overmatching does confer a benefit by 'protecting' flaws located in both the weld metal and fusion boundary. This 'protection' can be observed in the J driving force curves predicted by numerical analyses when the girth weld is included (in this case weld metal yield strength over matches that of the parent pipe by approximately 20%). Figures 3-4 show that for fusion boundary flaws, the predicted J driving force for plastic straining conditions is reduced significantly and below that predicted by the reeling procedure. In this case the weld width was 8mm.If the weld width were increased, the J driving force would decrease. Weld metal strength under matching will significantly increase J and is not permitted by the reeling guidelines. 
Another requirement of the reeling procedure is to allow for the effects of welding residual stresses. Two alternatives for incorporating these are permitted by the reeling procedure. The first is based on recommendations in BS7910and the second is based on adding yield strain to the applied strain. When the assessment is to BS 7910, as adopted in this paper, an initial welding residual stress, of yield magnitude, is assumed to exist in the region containing the flaw. As the applied stress is increased, the residual stress is allowed to relax down to 40% of the yield strength of the parent metal. The maximum relaxation occurs when the reference stress is equal to the flow strength (average ofyield and tensile strengths). However, as explained previously, a lower level of residual stress is likely to be present in the actual weld at applied strains above yield. The residual stress is treated as a secondary stress and therefore does not contribute to plastic collapse but does affect the fracture axis of the FAD. The outcome is that the residual stress will increase the J driving force for the reeling analysis as illustrated in Figs 3-6. These show that the reeling analysis curves including the effects of welding residual stress (labeled 'DNV-RP-F108 homogeneous material, with residual stress') lie well above all other curves including the numerical analysis curves obtained assuming homogeneous material and the reeling analysis curves obtained without allowance for welding residual stresses.
If residual stress is treated as a yield strain added to the applied strain, the J driving force curves for the reeling analyses in Figs 3-5 are shifted to the left by 0.24% (the parent pipe yield strain). Although this increases the J driving force for a given applied strain, it is insufficient to correct for its non-conservatism, for flaw heights less than 40% of the pipe wall thickness, in comparison with the results from numerical analysis at strains above approximately 1%. Clearly, the alternatives offered by the reeling procedure are not equivalent and on the basis of the results presented here, the recommendation is to treat residual stress in accordance with BS7910. Treatment of residual as an addition to the applied strain can lead to the prediction of a non-conservative J driving force when the applied strain exceeds approximately 1%.
A third factor, which is more difficult to quantify, concerns the impact of the materials' resistance to fracture, or fracture toughness, on the driving force curve. The reeling procedure recommends that single edge notch tension(SENT) specimens are employed to determine fracture toughness. These specimens are designed to be more representative of the crack tip constraint conditions of flaws in the pipe girth weld than conventional, deeply notched single edge notched bend (SENB) specimens. Analyses which have compared crack tip constraint in the SENT specimen and flaw in pipe subjected to plastic straining [5-7] indicate that crack tip constraint (in terms of the Q stress) in the specimen is higher than in the pipe for the range of conditions examined. If the fracture toughness specimen were designed to match the crack tip constraint inthe pipe, the outcome would be that higher fracture toughness would be measured compared with SENT design currently recommended by the reeling procedure.  Consequently, the design of the SENT specimens can provide a degree of conservatism by providing a pessimistic value of fracture toughness, as long as crack tip constraint is higher than that for a flaw in the pipe.
These three factors, weld strength overmatching, welding residual stresses and fracture toughness, mean that the apparent non-conservatism in the reeling analyses at strains above 0.75% is reduced. As shown in Fig.3-6, the requirements for weld strength overmatching and the inclusion of welding residual stresses as a secondary stress in the analysis in accordance with BS710 (even at high strains) ensured that the J driving force estimated according to the reeling procedure  was above the driving force estimated by numerical analysis. This conclusion would appear to be supported by satisfactory field experience when these methods have been used to set flaw acceptance criteria for girth welds. In addition, the reeling procedure  does require validation by mean of sub-scale segment tests. These specimens, which are taken from the girth weld and contain representative flaws, are strained to simulate reeling. Ductile crack extension observed from the flawin the specimen is compared with that predicted from the analysis; the procedure is considered satisfactory, or validated, provided that the predicted crack extension is greater than that observed. This procedure is a safeguard that helps to minimise the risk of unsafe predictions of flaw size being made.
Nevertheless, despite the comments made above, the assessment procedure should provide a more accurate modelling of the appropriate driving force curve derived from numerical analysis. There are a number of possibilities including:1) modifying the reference stress solution (it can be shown that the apparent non-conservatism of the J curves, obtained according to the reeling procedure for the 3x50, 6x50 and 6x25mm flaws without allowing for residual stresses, canbe eliminated by increasing the reference stress for these flaws by 3 to 5%); 2) allowing for strength mismatch (this requires mismatch limit load solutions for circumferential finite length flaws in cylindrical sections, which do not exist at present); 3) conducting the analyses on the basis of strain rather than stress, possibly within the context of a strain-based FAD (a number of possible approaches have recently emerged but these are yet to be fully developed and validated); 4) the use of more refined and potentially more accurate models to allow for the effects of welding residual stresses.
One approach that is currently being developed at TWI, which shows promising results, is to use J-based reference stress/strain solutions. The procedure is broadly consistent with the FAD based assessment framework provided by BS7910 but can be expressed in terms of applied strain.  Examples of the results from this approach are shown in Fig.7 for homogeneous material. For these cases it can be seen that the predicted J driving force curve accurately tracks that predicted by numerical analysis.
Other activities at TWI include numerical analyses leading to the derivation of crack driving force curves in terms of CTOD (and J) vs. applied strain for both surface and embedded flaws. These analyses include the use of large-strain large-displacement formulation. Findings from these activities will enable the production of improved guidance on assessing the significance of girth weld flaws subjected to significant plastic straining, which will address the weaknesses and limitations of existing guidance including the reeling procedure considered above. In the meantime and until better advice can be provided, it would be prudent to apply a cautious approach when conducting flaw assessments when the applied strain is significantly above 0.75%, which includes consideration of the effects of welding residual stresses explicitly, as already described.
Numerical analyses (FE) have been undertaken of circumferential flaws in the girth weld of a pipeline subjected to plastic straining simulating installation by reeling and J driving force curves derived. These were then compared with J driving force curves derived using guidelines described in DNV-RP-F108 intended for installation methods involving significant plastic straining. The assessment procedure is based on BS 7910 FAD methods. Strains up to 2% were considered. Analyses were undertaken for surface breaking weld root flaws located at the weld fusion boundary. The pipe was nominally to API 5L Grade X65 strength, with a diameter of 323.9mm (12¾in) and wall thickness of 20.6mm.The following conclusions are drawn.
- When the weld has the same tensile properties as the parent pipe ('even matched' strength condition) and welding residual stress is ignored, the J driving force predicted by numerical analysis for 3x50, 6x25 and 6x50mm flaws is higher than that predicted by the reeling procedure based on the BS 7910 assessment method for strains exceeding 0.75%.
- When the weld metal with a yield strength 20% higher than the parent pipe is included in the numerical analysis model, the J driving force is reduced relative to the even matched strength condition and below that predicted by the reeling procedure based on the BS 7910 assessment method for strains up to at least 2%.
- When welding residual stress, in accordance with BS7910, is included in the assessments based on the reeling procedure, as a secondary stress (i.e. contributing to fracture but not plastic collapse) in addition to the applied stress, the resulting J driving force curves lie well above all other curves including the numerical analysis curves obtained assuming homogeneous material.
- The inclusion of welding residual stresses in the analysis and the requirement for weld strength overmatching mean that J driving force calculated using the reeling procedure (DNV, 2006) based on BS 7910 should be above the driving force estimated by numerical analysis.
The authors are pleased to acknowledge the assistance provided by Dr Martin Goldthorpe in conducting the numerical analyses.
- BS, 2005. 'Guide for assessing the significance of flaws in metallic structures' BS 7910:2005, British Standards, London.
- Cheaitani MJ (2008). 'ECA of circumferential girth weld flaws', to be published, 2008.
- DNV (2006). 'Fracture control for pipeline installation methods introducing cyclic plastic strain' DNV-RP-F108.
- Kastner, W, Rohrich E, Schmitt W and Steinbach R (1981). 'Critical crack sizes in ductile piping'. Int. J. Pres and Piping 9 (3), 197-219.
- Nyhus B, Polanaco ML and Öjasæther O (2003). 'SENT specimens an alternative to SENB specimens for fracture mechanics testing of pipelines'. Proceedings of 22nd International Conference on Offshore Mechanics and Arctic Engineering, 8-13 June 2003, Cancun, Mexico, paper OMAE2003-37370.
- Nyhus B et al (2005). 'Fracture control - offshore pipelines. Experimental studies on the effect of crack depth and asymmetric geometries on ductile tearing resistance'. Proceedings of 24th International Conference on Offshore Mechanics and Arctic Engineering, 12-17 June 2005, Halkidiki, Greece, paper OMAE2005-67532.
- Pisarski HG and Wignall CM (2002). 'Fracture toughness estimation for pipeline girth welds'. Proceedings of International Pipeline Conference, 28 September-3 October 2002, Calgary, Canada, paper IPC02-27094.
- Wästberg, Pisarski HG and Nyhus B (2004). 'Guidelines for engineering critical assessments for pipeline installation methods introducing cyclic plastic strain'. Proceedings of 23rd International Conference on Offshore Mechanics and Arctic Engineering, 20-25 June 2004, Vancouver, Canada, paper OMAE2004-51061.