Flaw Tolerance of Pipelines Containing Circumferential Flaws Subjected to Axial Straining and Internal Pressure 0 Tests and Analyses
H. Pisarski, S. Smith and T. London
Paper presented at ISOPE Conference July 2013, Anchorage, Alaska, USA
A number of codified assessment procedures can be applied to assess the significance of circumferential flaws in pipes but these are generally stress based. Efforts have been made to extend these so that they are applicable when the pipe is subject to axial plastic straining with and without internal pressure. In this paper results are presented for two full–scale tests which were axially loaded beyond yield. They were conducted on parent pipe to API 5L PSL 2 Grade X65, 273mm OD x 18.4mm WT, and contained circumferential surface notches. In the first test the pipe was axially strained until failure and in the second test the pipe was first internally pressurised and then axially strained until a failure condition was reached. In both tests failure was ductile. The full–scale tests were accompanied by small-scale tests which included SENT tests to derive fracture toughness resistance curves. For the materials investigated, the SENT specimens with EDM notches produced almost identical resistance curves as those with fatigue pre-cracks. The behaviour of the pipes in terms of CTOD versus applied strain was compared with finite element analyses and failure analysis diagram (FAD) methods described in BS 7910. It is shown that a modification of the material specific FAD enables it to be extended up to 3% strain.
There are a number of recently developed methods for assessing the strain capacity of pipelines containing circumferential surface flaws, although these are not yet codified (Ostby (2007), Fairchild (2011, Wang (2012)). They have been developed through a combination of numerical analyses and full-scale testing. Failure analysis diagram (FAD) based methods have received less attention but general approaches have been suggested (Budden (2009), Smith (2012)). This paper describes some initial results from a project whose primary objective is to develop a strain-based flaw assessment procedure which quantifies the most important variables that influence flaw tolerance of pipeline girth welds subjected to axial plastic straining with and without internal pressure. If it can be shown to be possible, the intention is to extend failure analysis diagram (FAD) methods, such as described in BS7910, beyond yield into the plastic straining regime. This objective is being achieved through a series of numerical analyses of pipes containing circumferential flaws, material characterisation, especially fracture toughness, full-scale pipe testing and analytical work. This paper describes the results and analyses conducted on the first two full-scale pipe tests. Tests were conducted on seamless pipe to API 5L PSL2 Grade X65 with an average outside diameter of 273.3mm and wall thickness of 18.4mm. Four notches were introduced by electro-discharge machining (EDM) into the pipe outside diameter in the circumferential direction with one at each of the cardinal points. The first pipe was plastically loaded in tension until failure. The second pipe was first internally pressurised to 620barg with water to produce a hoop stress of approximately 87% of parent pipe yield strength. Subsequently, the pipe was axially loaded in tension until through pipe wall tearing (and a leak) occurred from one of the notches. Each pipe was instrumented to record pressure, applied force, local strain, overall strain, and crack mouth opening of each of the notches. The results were analysed to provide the relation between stress and strain, local strain and remote strain, and CTOD and strain. In addition, the CTOD predicted from finite element analyses with and without correction for ductile crack extension were compared with the experimental data. Finally, a comparison is made between J-integral (derived from CTOD) in the pipe with J derived from a conventional (BS 7910 Level 2b) stress-based and a new strain-based FAD.
Pipe Material and Properties
The project used seamless pipe to API 5L PSL2 Grade X65 with an outside diameter of 273.3mm and wall thickness of 18.4mm. Testing showed that the average yield strength (RP0.5%) was 512MPa and tensile strength (Rm) of 597MPa in the longitudinal direction. The engineering stress-strain curves obtained from a series of specimens taken around the circumference of the pipe are shown in Figure 1. In the circumferential direction tensile properties were determined using 8mm round bar specimens which resulted in yield and tensile strengths of 522 and 608MPa, respectively: these differ by less 2% of the longitudinal values. The parent pipe specimens exhibited an upper yield and a Lüders plateau extending for about 1.9% strain before work hardening started at a strain of 2.3%. The strain at the tensile strength was approximately 11%. In subsequent analyses, the upper yield was ignored and the mean stress-strain curve shown in Figure 1 was used.
Fig.1 Parent pipe stress-strain curves obtained from longitudinal specimens at various positions around the circumference. The thick solid line represents the mean to the data, ignoring the upper yield.
Fracture toughness was determined using single edge notch tension (SENT) specimens based on the design and testing method described in DNV RP F108 (2006). The specimen had an over-square cross-section (2BxB, where B is 16mm; this corresponds to the pipe wall thickness after machining to remove curvature), with crack depth to width ratio (a/W) of approximately 0.3. The specimens were notched from the pipe OD by electro discharge machining (EDM) and then fatigue pre-cracked to achieve the final crack depth. In addition, a set of tests was conducted using EDM notches located in the weld metal of a girth weld which was tested for a subsequent phase of the project. The radius of the notch tip was 0.12mm and similar to the radius employed to make the notches in the full-scale pipe tests. The purpose of these tests was to determine whether notch root radius influenced fracture toughness resistance curve in material (weld metal in this case) with lower resistance than the parent pipe.
The CTOD resistance curves (R-curves) were determined using a multiple specimen procedure. CTOD was obtained using a double clip gauge method in which the clip gauge readings obtained above the notch mouth are extrapolated to the crack tip. The equation used is:
ao = initial crack height
V1 = crack mouth opening measured by clip gauge at a height of z1 above the surface
V2 = crack mouth opening measured by clip gauge at the height of z2 above the surface.
The CTOD R-curves obtained in this way are shown in in Figure 2 and are described by the Equations 2, 3 and 4 (the units are in mm):
Parent pipe (fatigue precracked):
δ = 1.917 Δa0.704 (2)
Weld metal (fatigue precracked):
δ = 1.248 Δa0.6 (3)
Weld metal with EDM notch (0.12mm tip radius):
δ = 1.303 Δa0.549 (4)
The weld metal R-curves are almost identical indicating that, in this case, insensitivity to crack tip radius relative to the standard fatigue precracked specimens. Although these tests were conducted on weld metal, the same insensitivity to crack tip radius is expected in the tougher parent pipe.
The pipes were cut to a length of 2000mm and marked with arbitrary cardinal points of 12, 3, 6 and 9 o’clock. EDM notches were made at the mid-length of the pipe in the circumferential direction at each of the cardinal points. EDM notching was conducted in two stages, using a thicker blade to start. The final notch size was achieved with a nominally 0.2mm thick blade to provide a notch tip radius similar to that achieved in the SENT specimens. The notch shape was canoe shaped with a constant depth and semi-circular ends. A pair of knife edges was positioned to each side of the notch via steel shims spot welded to the pipe surface close to the notch mouth. The knife edges were instrumented with a pair of clip gauges for determination of CTOD. CTOD was estimated from the clip gauges using the same formula as used for the SENT tests (Equation 1). Table 1 provides details of the actual crack depths and lengths that were measured after completing the tests. The nominal notch sizes were 5x100mm, 6x50mm, 3x100mm and 4x50mm and are referred to as such in the paper.
Heavy steel flanges were welded to the ends of the pipes to enable them to be bolted to the face plates of the tension test machine. The pipes were instrumented with strain gauges (5mm gauge length) located between the notches and on the same plane as the notches. Additional strain gauges were located remote from any influence that the notch might have on the strain field at a distance of 300mm either side of the notch. Overall strain was also measured over a gauge length of 700mm using a pair of displacement transducers positioned on each side of the pipe. Internal pressure and axial force applied to the pipe was also recorded using appropriate transducers.
Fig.2 SENT CTOD R-curves for parent pipe and weld metal using fatigue precracked specimens except where indicated.
The pipes were tested in a machine employing a pair of pistons, each capable of exerting 1000 tonne axial force, which reacted against the face plates to which the test pipes were bolted via the flange arrangement. The whole set-up was located in TWI’s dedicated 8m x 3m pressure test facility which is designed to safely contain any release of content or fragmentation products (should that arise). The test set-up is shown in Figure 3.
Fig. 3 Pipe welded to end flanges and installed in tension machine ready for testing
The pipe tests were conducted at ambient temperature which was 15° C. When internal pressure was employed, the pipe was internally pressurised to 620barg (a hoop stress corresponding to 87% of the yield strength) before axial straining commenced. During straining, pipe volume changes required manual adjustments to maintain constant pressure; typical variations were ± 10barg of the nominal value.
Table 1 EDM notch sizes in pipes and results at maximum load together with ductile crack extension (Da)
a x 2c,
Strain at maximum load, %
TWI 2– 1 (Axial loading)
4.6 x 92.3
4.6 x 92.4
5.8 x 46.5
5.7 x 45.8
TWI 2– 2a (Pressure and axial loading)
2.5 x 98.7
2.9 x 98.4
3.7 x 50.0
3.9 x 50.0
** Not measured as specimen fractured by ductile instability
An attempt was made to estimate ductile tearing from each notch by periodically partially unloading the pipe and then re‑loading. By recording the change in force and crack mouth opening displacement, unloading compliance was used to estimate crack depth and the extent of ductile tearing. This information was then used to generate a CTOD resistance curve for each notch. These results are not discussed in this paper.
Results and Discussions
Pipe under Axial Loading
The results are summarised in Table 1. Where complete pipe separation did not take place, the notched/cracked region was subsequently cut from the pipe, and broken open at low temperature to induce cleavage and reveal the fracture surfaces for examination.
Fig. 4 Stress versus strain measured over a gauge length of 700mm at two positions (12 and 6 o’clock) in pipe tension test TWI 2-1 (no imposed internal pressure).
The first pipe test (TWI 2-1) was conducted on pipe material under axial loading only. Failure occurred at an overall strain of 5.14% when complete separation of the pipe occurred from the notches. However, the strain and CTOD at maximum load was 3.8% and 3.8mm respectively. The stress-strain curve is shown in Figure 4 where overall average strain was measured over the 700mm gauge length along the 12 and 6 o’clock positions. Local strain, measured by strain gauges, relative to overall strain shows the spread of plasticity along the pipe, as illustrated in Figure 5. The development of plasticity after yield strain was complex in the body of the pipe reflecting the Lüders plateau phase. The first region to plastically strain was material between the notches (SG1-4). The next region to plastically strain was material up to 300mm from the notch plane (SG5-8), but this was discontinuous during the Lüders phase, with each location straining differently relative to the overall strain. At overall strains between 2.4 and 3.7%, all locations strained uniformly. At overall strains above 3.7%, local strains between the notches increased rapidly up to final failure at an overall strain of 5.14%.
Fig. 5 Local strains measured using strain gauges versus average overall strain measured over a gauge length of 700mm.
Fig. 6 CTOD versus average overall strain compared with FEA modelling with and without ductile tearing for cracks at 3 and 9 o’clock.
The CTOD versus strain plots shown in Figures 6 and 7 indicate that at 3 and 9 o’clock both 6x50mm notches exhibited near identical behaviour. However, at 12 and 6 o’clock, containing 5x100mm notches higher CTOD values were obtained at 12 o’clock compared with 6 o’clock for the same overall strain. There appears to be zero error in CTOD at 6 o’clock but this could be possible local crack closure resulting from straightening of the pipe during initial elastic loading. Nevertheless, the differences in CTOD were significantly reduced at strains exceeding 3.8%. During the Lüders plateau phase, CTOD remained relatively constant and only began to increase when work hardening started (at approximately 2.5% strain).
Fig. 7 CTOD versus average overall strain compared with FEA modelling with and without ductile tearing for cracks at 12 and 6 o’clock.
Visual inspection of notches just prior to final failure showed that through pipe wall crack extension by tearing had taken place. A larger crack gape was shown by the notch at 6 o’clock compared with 12 o’clock indicated that final fast fracture, when it did take place, which was ductile (a ductile instability), probably started from the 6 o’clock notch. Final fracture, which was a ductile instability, resulted in complete separation of the pipe. The flat, grey region below each EDM notch indicates the ductile crack extension during axial loading, as shown in Figure 8.
Fig.8 Fracture appearance of axial pipe straining test (Test TWI 2-1).
Pipe under Combined Internal Pressure and Axial Loading
The second pipe test (TWI 2-2a) was conducted with constant internal pressure which was applied before axial straining.
Fig. 9 Axial and circumferential (hoop) strains (in microstrain) measured using strain gauges 300mm from crack versus time (note, internal pressure is applied before axial straining.
Figure 9 shows internal pressurisation preceding external axial straining and the stability of the pressure during the test. After yield, strain increased rapidly during the Lüders plateau phase. Above 2% strain (20,000 microstrain), the increase in strain with time was more uniform. The development of a Lüders plateau phase is also shown by the stress versus overall strain curve shown in Figure 10 and the discontinuous nature of this phase is indicated by the local longitudinal strain versus overall strain plots in Figure 11. As was evident in the test conducted with axial straining alone, the axial plastic strain between the notches was initially higher than strain elsewhere in the pipe, see Figure 11. Uniform straining at all locations was achieved at strains above 1.9%.
Fig. 10 Applied axial stress versus strains measured over a gauge length of 700mm. Leak from the 3x100mm notch located at 6o’clock occurred at 4.08% strain
Fig. 11 Local strains measured using strain gauges between the cracks and 300mm from them versus average strain measured over a gauge length of 700mm
The development of CTOD with increase in strain is shown for the two notch sizes in Figures 12 and 13. Although the behaviour of the two 4x50mm notches at 9 and 3 o’clock was similar, the 3x100mm notches at 12 and 6 o’clock differed significantly. Initially, CTOD was lower at 6 o’clock than at 12 o’clock, but when overall strain exceeded 3.5%, the reverse took place and CTOD increased rapidly until a leak occurred at 6 o’clock at 4.1% strain. As observed in the pipe test conducted axial straining alone, CTOD remained essentially constant during the Lüders phase. CTOD began to increase at around 1.9% strain; this was approximately 0.6% strain lower than in the previous test. The maximum load condition was reached at a strain of 3.47% when the CTOD was 2.83mm. When the leak occurred there was a rapid de-pressurisation of the pipe and the test was discontinued. At all notches, crack extension was by ductile tearing and the fracture face at the location of the notch which caused the leak (at 6 o’clock) is shown in Figure 14.
Fig. 12 CTOD versus average remote strain compared with FEA modelling with and without ductile tearing for cracks at 12 and 6 o’clock.
Fig. 13 CTOD versus average remote strain compared with FEA modelling with and without ductile tearing for cracks at 3 and 9 o’clock.
Non-linear (finite strain) finite element analyses (FEA) of the pipe were conducted using the mean stress-strain curve obtained from the parent pipe (see Figure 1). CTOD was defined using the 45° intercept method. The predicted CTOD versus strain (over a 700mm gauge length) is compared with the experimental results in Figures 6 and 7 and 12 and 13. Two sets of results are given, the first is for a stationary crack of the initial size and the second is an analysis with a correction for ductile crack extension. This was obtained by conducting a series of FEAs for different crack depths. The CTOD driving force curves obtained in this way were compared with the SENT CTOD R-curve. From the intersection of these curves the relationship between CTOD and strain and the corresponding ductile crack extension was established. Tangency indicated a ductile instability condition (i.e. rapid ductile failure). Although there are differences between the experimental and FEA predicted CTOD versus applied strain curves, the trends are generally similar. Furthermore, there is better agreement when the CTOD predicted by FEA were corrected for ductile crack extension.
Fig. 14 Appearance of notch at 6 o’clock (causing leak) after breaking open section of pipe cold.
Comparison of pipe test with FAD assessment
The flaw assessment methods described in BS 7910 (1997) are based on a failure analysis diagram (FAD) which was originally developed by Ainsworth (1984). The relationship between fracture and local plastic collapse is obtained from a J-based assessment line. An important feature of the FAD is that the elastic-plastic J can be determined from the elastic J via the stress-strain curve. The material specific assessment line (levels 2b and 3b in BS7910) is defined by Equation (5):
δr = δI/δmat
Jr = JI/Jmat
δI and JI = elastic applied δ or J
Jmat and δmat = material fracture toughness
σref = reference true stress (which depends on flaw size and component geometry)
εref = reference true strain (the strain at the reference stress derived from the material stress-strain curve)
Since the assessment line uses the material stress-strain curve, it should be possible to apply the FAD to strain-based assessments since there is correspondence between stress and strain via the stress-strain curve. BS 7910 (2007) is normally employed when the nominal applied stress is below yield. Numerous finite element analysis and large scale test have shown that the approach is conservative in that the fracture driving force curve (J versus stress) predicted by the assessment line is greater than that obtained by FEA or experiment. However, when the remotely applied stress exceeds yield, the J driving force curve becomes non-conservative. This has been discussed by Pisarski and Cheaitani (2007) in the context flaw assessments carried in accordance with DNV RP F108 (2006). However, it should be noted they showed that, in practice, the non-conservatism is reduced because of the conservative treatment of residual stresses and that the beneficial effect of weld strength overmatching, which is not normally taken into account when assessing flaws in welds.
With respect to the current tests on plain pipe, the non-conservatism in the DNV/BS7910 approach is illustrated in Figure 15. Here the J driving force is derived from the level 2B FAD in BS 7910 (2007) for the 6x50mm crack in pipe test TWI 2-1 (axial loading only). This is compared with pipe the test. In this comparison the experimentally measured CTOD was converted into J. J and CTOD were derived from FEA which provided an estimate for ‘m’ in Equation 6:
Above yield (σY), ‘m’ varied with applied strain but on average it was around 1.42 (for strains <3%). Using this value J was estimated from the experimentally measured CTOD in test TWI 2-1 using Equation (7):
J = 1.42 σY δexperimental (7)
Fig. 15 J versus applied strain derived from BS 7910 Level 2b FAD, strain based modification to FAD proposed by Smith (SS_SB_FAD) compared with J estimate from plain pipe test TWI 2_1 (axial strain only. J derived from CTOD using an “m” value of 1.42 (derived from FEA). No allowance for ductile tearing is included in the FAD-based methods.
Recently, Smith (2012) pointed out that for power law hardening material a modification to the FAD described by Ainsworth (1984) was necessary, otherwise the material specific FAD would underestimate driving force J and CTOD by a factor of up to two in low strain hardening materials when the applied strain exceeds yield. He introduced a parameter “x” to correct for this; it changes from unity to two as applied load increases (increasing Lr). The modified material specific FAD is given in Equation 8:
x = 0.5 (3+ tanh(c1(Lr-c2)))
c1 and c2 are tuning contants and have values of 5 and 0.5 here.
The effect of using this equation is shown in Figure 15 by the curve “SS_SB_FAD 6x50”. For strains below yield the predicted J is identical to that obtained from the conventional Level 2b FAD to BS7910. However, at higher strains, J is increased significantly (by a factor of 2, as implied the modified FAD equation) and is similar to that estimated from the pipe test. The procedure has merit in that all the stress intensity factor and reference stress solutions in BS 7910 can be used without modification, thus simplifying strain-based assessments. If these results are confirmed by additional tests and analyses, then with a small modification to the material specific FAD, assessment procedures based on BS7910 (2007) can be safely extended beyond yield (their current limit) and to strains of at least up to 3%.
It may be noted that in developing a strain-based FAD, Budden (2009) suggested that the reference strain should correspond to the applied strain and that in the fully plastic limit a factor of 2 on strain is applied to avoid non-conservative assessments. The modified FAD equation (Equation 8) proposed by Smith (2012) was adopted here because it is less arbitrary and consistent with BS 7910 procedures.
Further analyses of these tests and other full-scale pipe tests which contain flaws in girth welds are being undertaken. These will help to extend and validate the modified FAD based assessment procedure described here. The results of that work will be published in due course.
Although the validity of J, as originally defined, at large plastic strains can be questioned, results of tests and analyses, such as those shown in Figure 15, indicate that it has merit. Nevertheless, CTOD appears to be less affected by the theoretical limitations applicable to J. Furthermore, transferability of results from small-scale to large-scale tests is more easily verified with CTOD compared with J. Future work will investigate the extension of the FAD approach based on CTOD for high strain applications.
From the full-scale pipe tests and analyses conducted the following conclusions are drawn.
- Tensile tests conducted on the parent pipe exhibited a yield discontinuity in the form of a Lüders plateau, this behaviour was reflected by the plastic straining behaviour of the full-scale pipe tests (in both the stress versus strain and CTOD versus strain curves) with and without internal pressure.
- Beyond yield strain, the initial development of plasticity in the pipe was complex owing to the Lüders plateau but CTOD remained essentially constant. However, once work hardening had started plastic strains developed uniformly in the body of the pipe and CTOD increased as axial loading was increased.
- In both pipe tests final failure was ductile; a ductile instability occurred (after stable tearing) in the pipe that was axially loaded, whilst a leak occurred from a notch which had extended by stable ductile tearing in the pipe that was internally pressurised and axially strained.
- For the materials investigated, the SENT specimens with EDM notches produced almost identical resistance cures as those with fatigue pre-cracks.
- Finite element analyses were able to predict the trends exhibited by the full scale tests; better modelling was achieved by including allowance for stable crack extension by ductile tearing.
- A modification of the material specific failure analysis diagram in BS7910, which has the effect increasing the driving force when the applied strain exceeds yield, was found to provide a good description of the driving force (in this case J) versus applied strain behaviour of the pipe test subjected to axial loading.
The authors gratefully acknowledge the financial support provided by the sponsors to this project which were: ExxonMobil, Petrobras, Saipem and Subsea 7.
- Ainsworth R.A. (1984). ‘The assessment of defects in structures of strain hardening material’, Engineering Fracture Mechanics 19, 633-642.
- BS 7910:2005 (Incorporating amendment 1) (2007). ‘Guide to methods for assessing the acceptability of flaws in metallic structures’, BSI Standards, London, UK.
- Budden P.J. (2009). ‘Numerical validation of a strain-based failure assessment diagram approach to fracture’, Proc. ASME Pressure Vessels and Piping conference, PVP2009-77377.
- DNV–RP-F108 (2006). Fracture control for pipeline installation methods introducing cyclic plastic strain. Det Norske Veritas, Hovik, Norway.
- Fairchild D.P., Macia M.L., Kibey J., Wang S., Krishnan V.R., Bardi F., Tang H. and Cheng W. (2011). ‘A multi-tiered procedure for engineering critical assessment of strain- based pipelines’, Proceedings of 21st International Offshore and Polar Energy (ISOPE) Conference.
- Ostby E. and Hellesvik A. (2007). ‘Fracture control of offshore pipelines JIP, results from large scale testing of the effects of biaxial loading on the strain capacity of pipes with defects’. Proceedings of 17th International Offshore and Polar Energy (ISOPE) Conference.
- Pisarski H.G. and Cheaitani M. (2008). ‘Development of girth weld flaw assessment procedures for pipelines subjected to plastic straining’, Int. Journal of Offshore and Polar Engineering. 18, 3, 183-187.
- Smith S. (2012). ‘Development of the BS 7910 failure assessment diagram for strain based design with application to pipelines’, Proc. ASME 32st Int. Conf. on Ocean, Offshore and Arctic Engineering, Paper OMAE2012-83527.
- Wang Y-Y, Liu M. Song Y. and Horsley D. (2012). ‘Tensile strain models for strain-based design of pipelines’, Proc. ASME 32st Int. Conf. on Ocean, Offshore and Arctic Engineering, Paper OMAE2012-84241