Modelling and Measurements for the Assessment of a Full Scale Pipe Bend Test
Simon D Smith 1 , Henryk G Pisarski 1 and Cosmas Vlattas 2
1 TWI Ltd, Cambridge UK
2 Saipem UK Ltd, Surrey UK
Paper presented at ISOPE 2007, Seventeenth International Offshore and Polar Engineering Conference, Lisbon, 1-7 July 2007.
There is a significant need for well documented full scale tests to provide verification results for Strain Based Design (SBD) methods. The current paper presents some results from a full scale pipe bend test. The pipe contained a girth butt weld with a surface crack located at the edge of the weld. A nominal strain of around 1% was applied. Strain and mouth opening measurements were made. The crack was broken open after the test to reveal the extent of ductile crack growth. The full scale test was analysed using FEA and SBD methods.
KEY WORDS: FEA; Full Scale testing; Tearing; Strain Based Design; Pipe bending; Crack.
Sub sea pipelines are frequently laid in a manner that imposes significant bending strains on the pipe. In some cases this has lead to failure of the pipe even before any operational loads are applied to it. Such failures could have been prevented if accurate methods of assessment were adopted. These are presently being developed for pipe-lay.
There has been a significant amount of research into assessment methods for structures experiencing nominal stresses that are below yield strength. Codes have been developed to assess flaw tolerance in the weakest parts of the structure, like welds for example. Welds can contain flaws and regions of lower fracture toughness than the parent materials. The codes have recognised this and provided rules for flaw assessment. However, most of the codes are not suitable for significant applied strains (above the nominal material yield strain).
The current paper describes a test conducted on a typical pipe length with a weld at the centre and a surface-breaking defect in the weld. The weld was a typical pipe-line girth butt weld, and the applied load levels caused nominal strains at the crack to reach 1%, or well over the nominal yield strain of 0.2%. The pipe did not fail, so fractography was undertaken to measure the amount of ductile tearing that occurred during the test. The result of the fractography was compared with the expected tearing based upon separate small scale test results. Further synthesis using finite element analysis (FEA) was done to provide information about the deformations of the pipe during the test.
Pipe bend test
The pipe was 24in diameter with a wall thickness of 24.5mm. The specimen was 6m long with the weld located at the centre of its length. The weld was made with the Passo, automatic GMAW process. CO 2 shielding was used on the root pass, and 70%CO 2 , 30%Ar for the filling and capping passes. The weld was made into a steep sided machined V preparation in approximately 8 passes, using Redaelli RMS wire. All passes were made with a heat input in the range 0.6-0.9kJ/mm except the root where the heat input was 0.4-0.5kJ/mm.
A misalignment of the pipe axes was deliberately made across the weld. The misalignment was greatest at a location designated 6 o'clock on the pipe, and a flaw up to 3.3mm deep by 162mm surface length was introduced at this location off the weld centre line as shown in Fig.1. The crack was manufactured by use of a 0.15mm wide slitting wheel. The average misalignment at the crack was 1.7mm. A peak misalignment of 3mm was measured. The crack location, misalignment and the weld geometry are shown schematically in Fig.1.
Fig.1. Geometry of the assumed crack in the FEA model
The tensile properties of the pipes and weld were measured on material from the pipe test taken from the neutral axis of the pipe. This meant that the actual material could be tested in a relatively virgin form. The tensile test results are shown in Table 1. The ductile tearing resistance of the weld metal was determined from samples taken from the neutral axis of the pipe. The weld metal R-curve ( Fig.2) was determined using deeply notched (a/w=0.5) square sectioned bend specimens tested in accordance with BS7448 Part 4.
Table 1 Tensile test results for weld and parent material taken from the neutral axis of specimen TWI 05 after testing
|Material||Specimen cross section,|
|0.2% proof strength,|
|Parent pipe AQM 6467
||38 x 25.4
|Parent pipe AQM 6472
||38 x 25.4
|Parent pipe AQM 6472
|Parent pipe AQM 6472
|Weld metal (TWI 05)
Average 0.2% yield = 527.5MPa, Overmatch ratio at 0.2% strain = 1.15
Fig.2. R-curve for the weld metal measured on a SENB specimen. A further curve showing a possible curve for tensile loading based upon three times the toughness is also shown
The pipe was loaded in 4 point bending with inner loading points separated by 1220mm and the outer supports positioned 5600mm apart ( Fig.3). Strain gauges were attached to the pipe at the weld near the crack and also at 90° intervals around the pipe circumference ( Fig.4). The crack mouth opening was measured using a double clip gauge and the average strain over a 200mm length spanning the flaw and the weld was measured with a displacement transducer.
Fig.3. Actuators and test rig of the full scale pipe bend test. The pipe is shown in the rig
Fig.4. Strain gauge (SG) locations around the flaw in the full scale pipe bend test. Pipe 6467 is on the left hand side
Finite element analysis
A 3D brick element model of the test specimen was generated. The pipes on either side of the weld were assumed to be perfect cylinders of 607mm OD and 24.5mm wall thickness. The cross section of the weld was based upon a macrosection. The weld was assumed to be 7.5mm wide. The weld was also skewed as shown in Fig.1 to mimic axial misalignment present in the test pipe. Axial lines on both pipes were displaced by 2.5mm at the centre of the crack.
The initial and final crack dimensions were measured after the test. The initial surface length of the crack was 162mm and the initial maximum depth of the crack was 3.3mm, but, at top dead centre the depth was 2.9mm. The crack was cut using a 0.15mm wide slitting wheel. The crack shape was idealised in the model. A uniform crack depth of 3.3mm was assumed and the crack ends were assumed to be circular of radius 3.3mm ( Fig.1). The width of the crack was assumed to be slightly greater than the width of the slitting wheel. A width of 0.2mm was chosen. The tip of the crack was semi-circular and had diameter of 0.2mm.
Clip gauges were attached to the specimen at the centre of the crack at 2mm and 12mm above the crack mouth respectively. The FE displacements at these locations were calculated using beam elements extending from the face of the crack to the correct heights above the mouth of the crack.
R-curve type analyses require the driving force as a function of load and crack length. Further analyses were performed with crack depths of 3.0mm and 3.6mm for this purpose.
There was a small difference between the tensile data for the parent pipe material on either side of the weld. The appropriate tensile data was applied to each side of the weld.
A further FE analysis was run to investigate weld metal undermatch. The weld metal tensile data was kept the same. The weld undermatch was achieved by increasing all points on the AQM 6467 curve by the same increment in stress whilst keeping Youngs modulus unchanged. The new curve was used for both pipes. It had a yield strength of 660MPa, giving a weld metal undermatch of 10%.
The plane of bending was also assumed to be a plane of symmetry. Half of the specimen was meshed, and symmetry boundary conditions were applied to the symmetry face.
Nodes at the outer support points of the test were fixed in the vertical directions. The loading was applied using vertical applied displacements at one node at each of the load jacking points. Detailed modelling of the support and jacking points was not undertaken because these were not important locations in the test. The FE results showed that point loading created artificially high deformations at these points. The material at the supports was therefore restricted to deform elastically to prevent this. The FE results were not plotted against the applied displacement to avoid inaccuracies associated with local deformation at the supports and jacks. The average strain measured at gauges 250mm from the weld was used as a loading parameter instead. The average strain gives a measure of deformation that is more closely related to the loading to be defined for pipe lay ( i.e. the average strain near to the weld).
Pipe test results
The pipe bend was increased until the 200mm displacement transducer had measured 2mm or an average of 1% strain over the crack. The peak strains measured at the gauges were less than 1% because of the additional compliance of the specimen due to the crack. Peak tensile bending strains of around 0.8% were measured and the lowest compressive bending strain measured was -0.6% ( Fig.5). The average strain measured with the strain gauges at the end of loading was 0.75%. The finite element analysis prediction of the bending strains is also shown in Fig.5. The FEA results are in good agreement with experiment at lower levels of bending strain, but slightly under predict the measured values at higher loading levels.
Fig.5. Strain measured at 250mm from the crack. Gauges 6 and 12 are in line with the crack, whilst Gauge 2 is diametrically opposite the crack. Finite element results are also shown
The FEA results were used to calculate the distribution of strain across the weld adjacent to the crack. It was found that the strain changed by a factor of around 4 across the weld because of the misalignment ( Fig.6).
Fig.6. FEA predicted distribution of strain across the weld. The x-axis refers to the distance from A towards B shown in the detail
A comparison of the measured CTOD with values predicted by FEA and other methods is presented in Fig.7. The measured CTOD reached a peak value of about 0.5mm. It was therefore anticipated that the crack growth would be around 1.3mm based upon the SENB R-curve shown in Fig.2. The crack was sectioned after the test and the amount of crack growth measured. It was found that the crack growth was about 0.3mm, significantly less than anticipated from the small scale test results. The small scale test R-curve was measured on a deeply notched bend test in which the crack tip constraint is relatively high. The pipe test was conducted in bending as well, but the crack experienced mostly tensile loading, so it is thought that the crack tip constraint was significantly less than the small scale test. The difference between pipe and small scale test tearing was probably due to the difference in crack tip constraint between the two specimens. There is some experimental evidence [Wang et al 2004] that the difference in toughness between SENB and tensile wide plate specimens is significant. A factor of three is shown in Fig.2 as an example. The predicted tearing from the 'tensile' curve for a CTOD of about 0.5mm ( Fig.7) would be around 0.5mm ( Fig.2). This is closer to the measured value.
Fig.7. Comparison of the predicted and measured CTOD plotted against the remote average strain from a 200mm gauge length over the crack
The FEA predicted CTOD using the same method as the experiment (i.e. measurement of the crack mouth opening and extrapolation to the crack tip) gave an accurate prediction of the experimental CTOD values - Fig.7. Further estimates of CTOD were made using formulae from the literature ( [Schwalbe et al 1998, BSI 1980] ). The GKSS/ETM expression is given by Eq 1 and the CTOD design curve of BSI PD 6493 1980 is given by Eq 2.
The crack depth, a in Eq.1 is assumed to be the correct maximum depth of the actual crack. PD 6493:1980 provides a procedure for the calculation of the equivalent crack size Symbol.1. for Eq.2. A value of 4.9mm was calculated for the crack shown in Fig.1.
The applied strain, ε, assumed in the above equations was the nominal value measured across the crack in the test, multiplied by a strain concentration factor ( εCF) associated with the pipe misalignment. A value of 2.7 was chosen from the strain distribution across the weld near the crack, Fig.6 (the nominal strain was 0.7% and the strain at the location of the crack was about 1.9%). It appears that the PD 6493:1980 formula is generally conservative, at least until a strain of about 0.7%, but the ETM method is more accurate at lower strains. The effect of residual stresses has been ignored because the work is presently concentrating on the results at large strains, when residual stresses will have been redistributed by mechanical loading. It is not difficult to suggest how residual stresses and residual stress redistribution could be incorporated into this method.
Fig.8. FEA predicted distribution of strain across a 10% undermatched weld. The x-axis refers to the distance from A towards B shown in the detail of Fig.6.
The εCF was provided by FEA and includes the effect of misalignment and mismatch. It is anticipated that the future use of methods like this would rely on tabulated strain concentration factors for practical pipe geometries and material properties. For example, the strain distribution across an under matched weld is shown in Fig.8.
A Strain Based Design (SBD) assessment is based on applied strain, material toughness and the correct prediction of the crack tip driving force. The full scale test described here has addressed each these components of SBD. The CTOD crack tip driving force was measured during the test. The measured driving force can therefore be directly compared with the material toughness as measured by small scale test specimens. The comparison permits a direct prediction of the crack growth. The crack growth prediction based upon small scale tests was significantly larger than the actual crack growth in the full scale test. The difference was attributed to the higher constraint that exists in the small scale test specimen. An improved prediction was made by conversion of the high constraint bend result into a low constraint toughness curve using the methods similar to those discussed by [Wang et al, 2004] . Clearly there is a significant need to match the constraint of the small scale test specimens with the constraint that exists at cracks in welds in pipes. For installation of subsea pipelines involving plastic straining, current practice [DNV 2006] is to use single edge notch tension (SENT) specimens to derive fracture toughness since this is better at replicating the crack tip constraint conditions for a flaw in a pipe compared with the deeply notched bend specimen [Pisarski, 2002] . The bending loads applied to pipe line during laying operations and during in-service movement on the seabed tend to apply predominantly tensile stresses in the whole section containing the flaw. Membrane loads produce low constraint loading of cracks. However, higher constraint could exist if the crack was located in an under matched weld metal at the weld root. Also, biaxial loading conditions when the pipe is subjected to axial strain combined with internal pressure or severe misalignment can cause high constraint. The SBD procedures should therefore consider the full likely loading of a flaw when recommendations of toughness testing procedures are made.
The crack tip driving force in the full scale test was assessed in three ways. The value was measured and also predicted by FEA and estimated from the applied level of strain. It was found that the crack driving force could be accurately predict using a strain based method provided that the strain distribution at the weld was correctly included. A misalignment was deliberately made at the weld, and this misalignment caused a strain concentration at the location of the crack. The crack driving force was accurately predicted when the nominal strain and a strain concentration factor were used to determine the local strain at the defect. The driving force was then determined using some extremely simple formulae (see Eq.1 and Eq.2). The simplicity of these formulae is attractive and suggests that SBD for pipe lay could be straightforward. Methods of strain based defect assessment already exist [BSI 1980 and Schwalbe 1998] and these have been proven to be satisfactory for the present test specimen. The method does not need a stress versus strain curve. The predicted crack driving force is a linear function of the applied strain in the high strain regime. This simplifies the process of crack driving force calculation. The simple equations mean that accurate defect assessments could easily be done without the need for advanced software tools to determine the elastic-plastic crack tip loading. More work is needed to confirm this assessment and to develop an overall procedure. For example, a plastic collapse calculation would be needed.
The full scale test has shown that pipe line flaw assessment methods should use toughnesses that are representative of the membrane tension that is applied on the tensile side of a pipe in bending.
Simple crack driving force prediction methods have been assessed and shown to be suitable when the results were compared with the full scale test. The method is based upon the local strain at the flaw, including any local strain concentrations. The results suggest that existing strain based crack driving force procedures could provide the basis for SBD.
BSI: 'Guidance on some methods for the derivation of acceptance levels for defects in fusion welded joints', BSI PD 6493:1980.
DNV-RP-F101, January 2006. Fracture control for pipeline installation methods introducing cyclic plastic strain.
Pisarski H G and Wignall C M (2002): 'Fracture toughness estimation for pipeline girth welds', Proc of IPC 2002, Calgary, Canada, Oct 2002, paper IPC 02-27094, ASME.
Schwalbe K-H, Zerbst U, Kim Y-J et al (1998): 'EFAM ETM 97 - The ETM method for assessing the significance of crack-like defects in engineering structures, comprising versions ETM 97/1 and ETM 97/2', GKSS 98/E/6, 1998.
Wang Y Y, Horsley D J, Cheng W, Glover A, McLamb M and Zhou J (2004): 'Tensile strain limits of girth welds with surface breaking defects, Part 2: Experimental correlation and validation', Pipeline Technology Conference, May 2004, Ostend Belgium