Fracture Mechanics Assessment of Flaws in Pipeline Girth Welds
Henryk Pisarski, Guang Xu and Simon Smith
Proceedings of HSLP-IAP2006: International Seminar on Application of High Strength Line Pipe And Integrity Assessment of Pipeline 2006, June 15-16, 2006, Xi'an, China
The application of conventional fracture mechanics assessment methods to high strength steel pipeline girth welds is challenging when there is a need to assess plastic axial straining combined with internal pressure. To ensure integrity, recognition needs to be made that the steels may have limited work hardening capacity and strength mismatch between the weld metal and parent pipe may be important. This paper explores some of the limitations of conventional fracture mechanics assessments when applied to X100 pipeline girth welds. These are revealed by comparison of J driving force based on modifications to the basic BS 7910 assessment procedures with elastic-plastic finite element analyses. Tensile properties from a representative X100 pipeline steel are used to demonstrate the effects of girth weld strength mismatch and weld width on driving force under the combined effects of internal pressure and axial plastic straining with a weld flaw present. The effects of girth weld strength under and over-matching and parent material strain hardening behaviour (Y/T) on J driving force are explored.
The use of very high strength steel land pipelines for transporting gas poses particular challenges for the design of the steel and girth welds. Depending on location, the pipeline must be able to withstand low temperatures and abnormal loading whilst maintaining structural integrity and at the same time be economically viable. For transporting high-pressure gas over long distances, the use of high strength steels minimizes material, fabrication and welding costs because thinner wall pipes can be employed. Steel strengths currently being used have strengths up to the equivalent to X80 (SMYS=550N/mm 2 ). Higher strength steels of up X100 and X120 (SMYS=690 and 830N/mm 2 ) are being considered, but as yet neither have been used in large quantities. However, much development has been carried out on X100 and it is a possible candidate for transporting gas from the Alaskan NorthSlope into Alberta, Canada.  Although the primary design requirement of the pipelines is to hold pressure safely, environmental conditions such ground movement can impose axial plastic straining on the pipelines. This places severe demands on the integrity of the girth welds. They must have adequate strength and be tolerant to welding flaws. Fracture mechanics methods have proved to be useful in helping to define the level of flaw tolerance required. At the same time application of such methods has helped to minimize costs and delays by reducing the need to carry out unnecessary weld repairs (i.e. repairs of welding flaws that do not compromise the integrity of the pipeline). However, application of conventional fracture mechanics assessment methods, such as those in BS 7910  , to high strength steel pipeline girth welds is challenging. To ensure integrity, there is a need to assess plastic axial straining combined with internal pressure in steels with limited work hardening capacity and strength mismatch between the weld metal and parent pipe with a flaw present in the girth weld. This paper explores some of the limitations of conventional fracture mechanics assessments when applied to X100 pipeline girth welds. These are revealed by comparison of J driving force based on modifications to the basic BS 7910 assessment procedures with elastic-plastic finite element analyses. Tensile properties from a representative X100 pipeline steel are used to demonstrate the effects of girth weld strength mismatch and weld width on driving force under the combined effects of internal pressure and axial plastic straining with a weld flaw present. The effects of girth weld strength under and over-matching and parent material strain hardening capacity (Y/T) on J driving force are explored.
Engineering stress-strain curves were obtained from a Grade X100 pipe with an outside diameter of 915mm and wall thickness of 19mm using 8.9mm diameter specimens. Tensile properties were obtained transverse to the pipe and in the longitudinal direction. Higher strength and a higher yield to tensile ratio (Y/T) were obtained from the former compared with the latter. For the purposes of this work, the transverse properties are referred to as Pipe 1 (Y/T=0.944)and the longitudinal properties as Pipe 2 (Y/T=0.832); the tensile test results are summarized in Table 1 and the engineering stress versus strain curves are shown in Fig.1. All weld metal stress-strain curves were obtained from a weld made in the Grade X100 pipe in a V preparation with 60° included angle. The weld was made using a GMAW process with 95% Ar 5% CO2 shielding gas at aheat input of 0.94J/mm. The consumables and welding procedure were designed to provide tensile properties that closely matched the SMYS of Pipe 1.  However, the result was weld metal that under-matched the actual yield strength of the pipe because the pipe yield strength exceeded SMYS by a significant margin. An additional hypothetical case of 10% over-matching was artificially created for the purposes of analysis by assuming the weld metal to have the same stress-strain curve as Pipe 1 but uplifted by 10%. For the purposes of analysis, the girth welds considered in Pipe 2 were assumed to have the same strength as Pipe 1. In this paper the mismatch ratio M is defined as:
Where σ Yw or σ Yb are the stresses corresponding to the same amount of plastic strain of the weld metal and base metal, respectively. The subscript x in M defines the plastic strain at which mismatch is defined.
The tensile properties used for the analyses are summarized in Table 1 and stress-strain curves are compared in Fig.1. Based on yield strength, the levels of mismatch achieved were M = 0.83 and 1.1 in Pipe 1 and 0.97 and 1.29 in Pipe 2.
Table 1 Tensile properties
|Yield strength, MPa
|Tensile strength, MPa
|Ref Pipe 1
|Ref Pipe 2
|Parent pipe 1
|Parent pipe 2
|OM weld metal, WM 2
|GMAW weld metal, WM 1
Fig.1. Engineering stress-strain curves for parent Pipes 1 & 2 and weld metal
Fracture assessment - based on codified methods
The procedures are based on the engineering critical assessment methods described in BS 7190  but modified for strength mismatch effects in accordance with R6. These modifications affect the shape of the failure assessment diagram (FAD) and the definition of L r which accounts for plasticity effects. For simplicity, the material specific (Level 2B) FAD was employed which requires single values of fracture  toughness. The procedure is also applicable to the Level 3B FAD, which requires input of a fracture toughness resistance curve and is appropriate when material is on the upper shelf and there is no risk of brittle fracture. The main modification to the Level 2B FAD is the introduction of an 'equivalent' stress-strain curve to generate the FAD and definition of yield strength for the L r axis. The equivalent stress-strain curve is derived from a weighted average of the weld metal and parent pipe stress-strain curves. The weighting is provided by the ratio of the mismatch limit load to the limit load for homogeneous material (i.e. no strength mismatch).
Since the weld metal and base metal will, in general, work harden differently, the degree of mismatch will vary with plastic strain.
The equivalent stress-strain curve is as follows:
F Lmis is the mismatch limit load (see below). F Lb is the limit load for a homogeneous component made of material with yield stress σ Yb . The ratio ( F Lmis / F Lb ) is defined for each value of mismatch ratio M x , at different values of plastic strain ( ε p ).
The equivalent yield strength at 0.2% offset strain is given by:
An 'equivalent' flow strength,
, for the mismatched geometry is defined as:
is the flow strength of the weaker material. Here, flow strength is defined as the average of the yield stress and ultimate tensile strength.
Mismatch limit loads are given in R6  for fully circumferential internal cracks in cylinders; these are taken from work by Schwalbe et al.  (As far as is known, there are currently no general mismatch limit load solutions for finite length flaws in girth welds in the public domain.) The mismatch limit load solutions consider parallel-sided welds and are for weldmetal centreline flaws. Further details of the analyses are given by Pisarski et al. 
In order to allow for biaxially due to internal pressure, the strength mismatch corrected stress-strain curves were further modified for the increase in yield strength associated with biaxial loading. This was accomplished by inputting the pipe dimensions, internal pressure and remote axial strain using special pipe elements (called elbow elements) in the FE program ABAQUS. (This analysis did not consider flaws in the weld). In the model, design pressure was applied first and then axial strain. The results were used to generate an axial stress-strain response for the material. The plastic component of this response curve was taken to be representative of the stress-strain curve for the material. For simplicity, the elastic stress-strain curve, up to the elevated yield strength, was assumed to be the same as the uniaxial stress-strain curve (i.e. based on Young's modulus). This strength mismatch and biaxially corrected stress-strain curve provided the basis for subsequent fracture mechanics analyses based on BS 7910 assessment procedures.
The flaw case considered was an internal circumferential flaw located in a girth weld in a Grade X100 (SMYS in the circumferential direction is assumed to be 690N/mm 2 ) pipe with a diameter of 915mm and wall thickness of 19mm. The parent pipe tensile properties used for these analyses were those provided by Pipe 1 (see Table 1 and Fig.1). The hoop stress due to internal pressure was assumed to be 80% SMYS. For the purposes of assessment, the girth weld was assumed to be parallel sided and weld widths (2H) of 5, 10 and 20mm were considered. A flaw, 3mmhigh (intended to represent a welding flaw equivalent to a typical weld bead height) was located in the centre of the weld. Some subsequent analyses were undertaken assuming a finite length surface flaw, 3mm high and 100mm long. These analyses were undertaken assuming no strength mismatch (M=1.0 in Pipe 1). With this information, material specific FADs to BS 7910  (Level 2B) were generated for each mismatch case. The L rmax cut-off was extended beyond that recommended in BS 7910 to avoid limiting the analyses. The analyses were conducted using TWI software Crackwise 4 that automates many of the procedures in BS 7910. For the pipe geometry and loading condition the stress intensity factor was derived from a flat plate solution and the reference stress was based on the Kastner equation. The analyses were conducted to establish the relationship between the driving force or required fracture toughness in terms of J and applied stresses. Axial strains were derived from the strength mismatch and biaxially modified stress-strain curves, where appropriate. For simplicity, the analyses were conducted assuming that no residual stresses are present.
Fracture assessment - FEA
The pipe and flaw geometry considered in the codified assessment was also used in the finite element model, as shown in Fig.2. However, in this model only half the weld width, H=5mm, was considered. FEA was conducted using ABAQUS and an axi-symmetric model. The internal crack was assumed to be at the middle of the pipe length (2L=5OD). Only one half of the pipe cross-section was included in the FE model due to symmetry in geometry and loading. Internal pressure (to 80% SMYS) was applied first, but the pressure on the crack surface was ignored in the model since it was insignificant compared with the principal stresses. Uniform axial displacement was then applied at the pipe end, after the pressure loading. Appropriate displacement boundary conditions were imposed on the crack ligament to represent the effect of the remaining half of the pipe length not modelled. Crack driving force J and global axial strain were obtained from the FE analysis. The crack driving force J was calculated using the J-integral scheme in ABAQUS and the global strains at various axial displacements were taken from an element near the pipe end. Analyses were carried using the stress-strain curves for Pipe 1, weld metals 1 and 2 to model the two strength mismatch levels and enable comparisons to be made of the J driving force curves from the codified procedure. Further analyses were conducted using the tensile properties from Pipe 2, weld metals 1 and 2 to assess the effect of pipe work hardening behaviour on Jdriving force.
Fig.2. Finite element axi-symmetric model (not to scale)
Results and discussion
The effect of strength mismatch on the J driving force curve for a 3mm deep fully circumferential surface crack derived from the modified BS 7910 assessment procedure but subjected to axial strain only, is shown in Fig.3 as a function of applied axial stress normalised by SMYS. It should be noted that SMYS in this case is 690N/mm2 which is specified for the hoop direction. In this case, yield strength of Pipe 1 in the axial direction exceeds SMYS. Fig.3 shows that weld metal strength under-matching (Pipe 1, M=0.83) results in a significant increase in crack driving force J compared with the over-matched condition (Pipe 1, M=1.1). Furthermore, the driving force is influenced by weld width. For the under-matched condition, decreases in weld width from 20mm to 5mm reduce J driving force. This is because the development of crack tip plasticity becomes more influenced by the stronger parent pipe as weld width is reduced. For the over-matched condition, the opposite occurs. In this case, the beneficial effect is because the stronger material is inhibiting the development of crack tip plasticity. In addition, the analyses show an abrupt increase in driving force at stresses well below the actual longitudinal yield strength of the parent material in Pipe 1 (see Table 1). This is an indication that in Pipe 1, local yielding conditions are controlling the behaviour because the flaw height is significant with respect to the pipe wall thickness. Indeed local failure at the weld is predicted before plasticity can develop axially in the pipe.
Fig.3. Driving force J versus non-dimensionalised stress in Pipe 1 for different weld widths and strength mismatch (SMYS =690N/mm 2 ): uniaxial loading; derived from a modified BS 7910 procedure
The analyses were then repeated for the biaxial loading condition (i.e. pressure plus axial loading). Figure 4 shows the results for a fixed weld width of 10mm and mismatch levels of 0.83, 1.0 and 1.1 in Pipe 1. Although the relative effects of strength mismatch on the J driving force curves are retained under biaxial loading as they were under uniaxial loading, internal pressure appears to reduce the driving force J for a given axial stress. This is apparently inconsistent with findings reported by Jayadevan et al  and Østby  who showed that driving force CTOD was increased by biaxial loading (internal pressure plus axial tension).
Fig.4. Driving force J in Pipe 1 for biaxial and uniaxial loading for a 3mm high flaw in a weld 10mm wide; derived from a modified BS 7910 procedure
However, in that work driving force was plotted against strain whilst Fig.4 is plotted against stress. Since biaxiality elevates the stress-strain curve, a given applied stress will correspond to a fairly constant diving force but the strain will be higher under uniaxial compared with biaxial loading. So when driving force is plotted against strain, biaxial loading will result in a higher driving force for a given strain.
Further analyses were conducted for the homogeneous material condition (Pipe 1, M=1) in order to assess the effect of biaxially on finite length circumferential flaws. A surface flaw 3mm deep and 100mm long was chosen for these analyses so that remote plastic axial strains could be applied. The results are presented in Fig.5 and show that when the pipe is plastically straining axially, internal pressure increases the J driving force relative to the uniaxial loading case. This is consistent with published findings. [6,7] The analyses show that for a fully circumferential flaw 3mm high, local failure will occur at the weld before plasticity can develop in the parent material of Pipe 1. This arises because of the very rapid increase in J driving force for a small increase in remote strain.
Fig.5. Driving force J versus applied strain for continuous and 100mm long circumferential cracks 3mm high under uniaxial and biaxial loading in Pipe 1, derived from a modified BS 7910 procedure
The results from the FE analyses are presented in Fig.6, where driving force J is plotted against applied axial strain. These analyses fully account for the effects of biaxially on crack tip conditions as well as the effect on tensile properties.
Figure 6 shows that there is a rapid increase in driving force J for small increases in strain. This indicates that for the flaw case analysed (a fully circumferential flaw 3mm high), plastic straining will be localized at the flaw before axial plastic straining can develop in the parent pipe, as indicated by the simplified analyses. An illustration of the localised plastic occurring ahead of the crack is provided by the deformed mesh from the FE analysis, see Fig.7. A shear band develops, which ignores the overmatched weld, and results in 'necking' on the back face.
Fig.6. J driving force curves predicted by FE analysis for biaxial loading (with strength mismatch) and uniaxial loading, Pipe 1
The advantage of weld metal strength over-matching, by reducing the driving force J, and the disadvantage of under-matching, by increasing driving force J, are apparent and the relative ratings are the same as shown by the simplified analyses.
Fig.7. Deformed mesh shape in overmatched weld at ~0.32% axial strain (crack on the left)
Figure 8 compares the results from the FE analyses with those from the simplified fracture mechanics analyses. When uniaxial loading is applied (no internal pressure) and there is no strength mismatch (Pipe 1, M=1.0), there isa good agreement between the driving force J's predicted by the BS 7910 procedure and FE analyses for strains up to 0.32%. At higher strains, driving force J increases rapidly but more so with the BS 7910 procedure. This implies that the BS 7910 procedure is more conservative than the FE analysis. With biaxial loading the divergence occurs earlier, at about 0.2% strain, and the BS 7910 procedure now significantly underestimates the driving force J compared with FE analyses. This apparent non-conservatism is of concern and may be related to the derivation of the equivalent stress-strain curve to account for strength mismatch. However, further work is necessary to resolve this.
Fig.8. Comparison of J driving force from FE analysis and BS 7910 based procedures (shown as dotted lines and open symbols) for Pipe 1
Both the simplified and FE based fracture mechanics analyses show that the circumferential welds in the X100 material represented by Pipe 1 (Y/T=0.944) are not tolerant to fully circumferential flaws 3mm high when subject to the combination of internal pressure and significant axial strain. However, the analyses indicate that improved flaw tolerance is possible for 3mm high flaws if the length is limited to a small part of the circumference (see Fig.5). It has been shown that flaw tolerance is a critical factor affecting the integrity of girth welds in X100 pipelines. Weld strength over-matching coupled with a need to control flaw size are necessary requirements to ensure integrity.
The expected benefits of employing pipe material with a lower yield strength but lower Y/T (improved work hardening capacity) were explored by FEA using the tensile properties from Pipe 2 (yield strength = 680MPa, Y/T=0.832). The weld metal properties were the same as in Pipe 1. The results from these analyses are shown in Fig.9 where they are compared with those from Pipe 1.
Fig.9. J driving force curves from FEA for Pipes 1 & 2 with Y/T=0.944 & 0.832, respectively
The benefit of higher strain hardening capacity (lower Y/T) in generally reducing the J driving force curves for the Pipe 2 compared with Pipe 1 are immediately apparent. For example, at an axial strain of 0.33% the driving force Jis reduced by 80% in Pipe 2 compared with Pipe 1 for the even matched condition (M=1.0). Furthermore, localised yielding around the flaw is no longer the dominant failure condition, even for the slightly under-matched weld (Pipe 2,M=0.97). The parent material in Pipe 2 can now contain axial plastic strains when a fully circumferential flaw is present in the weld metal and the weld metal slightly under-matches the strength of the pipe. Weld strength overmatching(Pipe 2, M=1.29) has, in this case, a significant effect on reducing the driving force on the flaw. For example, at 0.5% axial strain the driving force is reduced by almost 40% compared with the even match condition (Pipe 2,M=1.0).
Both the simplified and FE based fracture mechanics analyses conducted on the grade X100 pipe steel, represented by Pipe 1 (Y/T=0.944) in this study, and girth welds show that for a fully circumferential flaw 3mm high, local plasticity effects control behaviour to such an extent that significant axial strains cannot be achieved in the parent pipe. Weld strength over-matching combined with increased weld width is shown to be beneficial but, in this case, the benefit is small because of the strong effect of local plasticity at the crack tip. Biaxial loading (internal pressure plus external axial loading) increases the driving force J in comparison with uniaxial loading. The discrepancy between the effects of biaxiality predicted by the simplified analyses and the FE analysis is attributed to the limitations/conservatism associated with the reference stress solution used in the former. New mismatch reference stress(limit load) solutions for circumferential cracks in pipes could improve the situation. Changing the parent pipe to one with a lower yield strength but higher work hardening capacity (yield strength =680MPa, Y/T=0.832), as represented by Pipe 2, improves the axial strain capacity of the pipe, even in the presence of a girth weld flaw. In this case, a small degree of weld metal strength under-matching can be tolerated (M=0.97) at moderate axial strains without a rapid increase in diving force J. The lower Y/T also reduces significantly the driving force J at moderate plastic axial strains.
- Howard R: 'Alaska Gas - an X100 opportunity: Project overview' X-100 Forum, CSM, Pula, Sardinia, Italy, 6 April 2006.
- BS 7910:2005: 'Guide on methods for assessing the acceptability of flaws in metallic structures', British Standards Institute, London, July 2005.
- Pisarski H G., Tkach Y. and Quintana M: 'Evaluation of weld metal strength mismatch in X100 pipeline girth welds', International Pipeline Conference, Calgary Alberta , Canada, IPC 04-0232, October 2004.
- R6 Revision 4, 2001: 'Assessment of the integrity of structures containing defects', British Energy Ltd, April 2001.
- Schwalbe K-H., Kim J J, Hao S, Cornec A and Koçak M: 1997, 'EFAM ETM-MM96 - The ETM method for assessing the significance of crack-like defects in joints with mechanical heterogeneity (strength mismatch)', GKSS 97/E/9, GKSS, Geesthacht, Germany.
- Jayadevan K R, Østby E and Thaulow C: 'Fracture response of pipelines subjected to large plastic deformation under tension', International Journal of Pressure Vessels and Piping, 81, 771-783, 2004.
- Østby E. 'Fracture control - Offshore pipelines. New strain-based fracture mechanics equations including the effects of biaxial loading, mismatch and misalignment' 24th Int. Conference on Offshore Mechanics and Arctic Engineering, Halkidki, Greece, June 2005, paper OMAE2005-67518, ASME.