Andrew R. Malpas, Philippa L. Moore, and Henryk G. Pisarski
Paper presented at ISOPE-2012. The 22nd International Ocean and Polar Engineering Conference, Rodos Palace Hotel, Rhodes (Rodos), Greece, June 17-22, 2012.
Single edge notched tension (SENT) specimens prepared in accordance with DNV RP F108 are usually surface-notched with an 'over-square' or 2BxB geometry. Since the crack extends over a crack front equal to the specimen dimension of 2B, growing a uniform, straight-fronted fatigue pre-crack is difficult, especially when testing weld metal or heat affected zone (HAZ). There are no standardised crack front straightness requirements for SENT specimens, so equivalent requirements for SENB specimens are applied, but not often met. TWI has reviewed its SENT test data to compare crack front straightness requirements from several testing standards. The qualification pass rate for parent metal specimens ranges between 13.8% and 100% depending on the standard used. Recommendations are made on suitable crack shape validity requirements for SENT specimens notched into parent, welds and HAZs.
KEY WORDS: SENT; SE(T); fatigue; pre-crack; straightness; qualification; fracture; toughness.
||Initial crack length
||Maximum crack length (of 9 initial crack measurements)
||Minimum crack length (of 9 initial crack measurements)
||Maximum crack length (of 7 inner crack measurements)
||Minimum crack length (of 7 inner crack measurements)
||Specimen 'thickness', dimension perpendicular to the width (see Fig.1)
||Specimen 'width', the dimension in the direction of crack propagation (see Fig.1)
The established methods for carrying out fracture toughness testing using lower constraint single edge notched tension (SENT) specimens employs surface-notching of an 'over square' or 2BxB geometry specimen, Fig.1. This wide pre-crack front with relatively little specimen width to grow and stabilise the fatigue pre-crack means that achieving a uniformly straight fatigue pre-crack can be difficult, especially in specimens notched into weld metal or heat affected zone (HAZ). It has been observed that SENT specimens in thicker material are more likely to meet the validity requirement than those from thinner material. [Pisarski, 2010]
The only current specification for SENT testing, DNV-RP-F108,[DNV, 2006] specifies the use of over-square specimens but does not give explicit validity requirements for crack front straightness. However, many test houses apply the equivalent requirements for SENB weld specimens such as those given in BS 7448 Part 2,[BSI 1997a] and have struggled to meet them. In 2010, BS 7448 Part 2 was superseded by BS EN ISO 15653 [ISO, 2010] which applies slightly different criteria for valid crack front straightness.
Fig.1 An 'over-square' SENT test specimen showing B (equal to twice the material thickness) and W (equal to the material thickness).
TWI has carried out a review of its existing SENT test data to compare the crack front straightness requirements given in a number of testing standards, and to quantify the effect of specimen thickness on crack front straightness in SENT specimens. The findings will help form part of a future British Standard for SENT testing.
Calculating crack curvature
A total of 412 SENT test results were reviewed, consisting of 70 parent metal specimens, 130 weld metal specimens and 212 fusion line (HAZ) specimens. All the specimens were of 2BxB design (that is, the specimen thickness dimension being twice the material thickness and the specimen width being equal to the specimen thickness, shown in Fig.1), with material thickness of between 6 and 28mm (and therefore the SENT specimen thickness was between 12 and 56mm). None of the specimens were side-grooved, as this is not a requirement of DNV RP F108. The fatigue pre-crack measurements reported for each test specimen were tabulated and the average crack length determined in accordance with the appropriate standard. The data was then used to perform the crack shape qualification checks as detailed in standards including BS 7448 Parts 1, 2 and 4,[BSI, 1991, 1997a and 1997b] ISO 12135 [ISO, 2002], BS EN ISO 15653 [ISO, 2010], ASTM 1820 and ASTM 1290.[ASTM, 2008 and 2009] Of these, only BS7448 Part 2 and BS EN ISO 15653 specifically address specimens notched into weld and HAZs; the others are intended for plain, or parent, material. It should be noted that the standards listed above do not include the 'over-square' 2BxB design that is used for SENT tests. The crack shape criteria given in these standards is intended for BxB or Bx2B SENB bend specimens instead.
The definition of initial crack length, a0, is consistent across all these standards. The initial crack length, a0, is obtained by firstly averaging the two measurements at the outer points and then averaging this value with the seven inner points. It should be noted that BS 7448 (all parts) and ISO 12135 (ISO 15653 refers the user to ISO 12135) state that the two outer measurements should be taken 0.01B from the surface; whilst, by contrast, ASTM E1820 and E1290 state that the two outer measurements should be taken 0.005W from the surface. Whilst these two statements are identical for Bx2B specimens, they differ for BxB specimens or specimens of W/B not equal to 2. All the standards define the dimensions B and W in accordance with the orientations given in Fig.1, it is only the use of either B or W to define other dimensions that differs. The different methods used by each standard to determine the fatigue pre-crack front straightness are summarised below.
- BS 7448: Part 1[BSI, 1991] states that the difference between any two of the nine crack length measurements shall not exceed 10% of the average initial crack length, a0. [clause 8.7.2b]
amax - amin ? 0.1 a0 ���� (1)
- BS 7448: Part 2 [BSI, 1997a] is intended for specimens notched into the HAZ or weld metal and states that the difference between any two of the seven inner crack length measurements shall not exceed 20% of the average initial crack length, a0. The wording used in Part 2 explicitly states that the validation check is purely a relaxation (to 20%) of the criterion given in Part 1, however the definition has been changed from 'any two of the nine measurements' to 'any two of the seven inner measurements'. [clause 12.4.1]
amax7 - amin7 ? 0.2 a0(2)
- BS 7448: Part 4 [BSI, 1997b] states that the difference between a0 and any of the nine crack length measurements contributing to a0 shall not exceed 10% a0 for parent metal specimens, and permits a relaxation to 20% a0 for weld or HAZ specimens. [clause 9.9.2b]
Max[(a0 - amin),(amax - a0)] ? 0.1 a0(3)
It should be noted that Part 4 is used to define a fracture toughness resistance curve and the SENB and CT specimens employed are usually side-grooved, so crack front straightness requirements are applied to the fatigue pre-crack front after side-grooving.
Part 3 of BS 7448 [BSI, 2005] for dynamic fracture toughness testing states that for parent material, the difference between any two of the nine crack length measurements shall not exceed 10% of a0 (which is the same definition as BS 7448 Part 1), and for weld/HAZ material the difference between any two of the nine crack length measurements shall not exceed 20 % of a0. [clause 9.7.2b]
- ISO 12135  states that none of the seven interior initial crack length measurements should differ by more than 0.1a0 from the nine-point average initial crack length. [clause 8.3.6h]
Max[(a0 - amin7),(amax7 - a0)] ? 0.1 a0(4)
- ISO 15653  states that for CTOD and J tests using bend specimens, the fatigue crack front straightness requirement defined in ISO 12135 may be broadened to 0.2 a0. However, for compact specimens the requirement may not be relaxed, and all K1C tests must conform entirely to ISO 12135. [clause 12.4.3].
- ASTM E1820  states that none of the nine physical measurements of initial crack size shall differ by more than 0.05B from the average a0, where B is the specimen thickness. [clause 220.127.116.11]
Max[(a0 - amin),(amax - a0)] ? 0.05 B ���� (5)
- ASTM E1290  states that the difference between the maximum and minimum of all nine crack length measurements of the fatigue crack does not exceed 0.1 of the original crack size a0. This is consistent with BS 7448: Part 1 (Eqn 1). [clause 18.104.22.168]
As is shown, nearly every standard (and succinct parts within the standards) are subtly different. The results from the analysis work presented here goes someway to highlight these differences on the qualification of test results.
Crack shape validity limits
Figs 1 to 5 show the crack curvature against specimen size, expressed as the B dimension, for the various standards. Trendlines have been plotted to show the relationship between curvature and thickness. In addition the validity limits of 10% and 20% of a0 for parent and welds respectively are shown where appropriate. The full set of 412 data points are analysed to each standard, but the data points corresponding to parent, weld and fusion line are separately identified, since not all are applicable to every standard.
Fig.1 Percentage crack front curvature assessed to BS 7448 Part 1 or ASTM E1290, with the parent metal 10% limit line shown
Fig.2 Percentage crack front curvature assessed to BS 7448 Part 2, with the weld metal and HAZ 20% limit line shown
Fig.3 Percentage crack front curvature assessed to BS 7448 Part 4, with the parent metal 10% limit line, and weld metal and HAZ 20% limit line shown
Fig.4 Percentage crack front curvature assessed to ISO 12135 and ISO 15653, with the parent metal 10% limit line, and weld metal and HAZ 20% limit line shown
Fig.5 Percentage crack front curvature assessed ASTM E1820, with the parent metal limit line shown
Table 1. Mean values of crack front curvature (as a % of a0, or for ASTM E1820, as a % of B) calculated according to validation criteria in standards.
|Part 1||Part 2||Part 4|
Table 1 shows the mean values of crack front curvature calculated to the validation criterion for each standard, while Table 2 shows the percentage of specimens which qualified to each validation criterion. These results are given separately for parent, weld and fusion line specimens.
Table 2. Percentage count of the number of specimens qualified to each validation criterion.
|�||% a0||BS 7448||ISO|
|Part 1||Part 2||Part 4|
Fig.1 shows the crack front straightness versus specimen thickness for crack front straightness as defined from BS 7448 Part 1 (and ASTM E1290). The linear trend lines show that there is a relationship between crack front straightness and specimen thickness. The results indicate that crack front straightness improves with increasing specimen thickness. Crack front straightness improves by approximately 5% when a 45mm thick specimen is compared with a 25mm thick specimen, for parent material.
The average crack front straightness for parent material in all thicknesses calculated to BS7448: Part 1 is 17.7%. The crack front straightness must be less than 10% of a0 to fully qualify to BS 7448 Part 1, and only 13.8% of the parent metal specimens in the data set pass this criterion.
Fig.2 shows the crack front straightness versus specimen thickness for straightness as defined from BS 7448 Part 2. This differs from Part 1 in that the qualification check only considers the inner seven initial crack length measurements and the criterion limit is increased to 20% for weld or HAZ notched specimens. Whilst Figure 2 shows similar trends to Figure 1, it is clear that all of the points have shifted down the ordinate, such that more specimens meet the qualification requirements. For parent metal specimens, the average crack front curvature reduces to 9.2% according to the definition in BS 7448 Part 2 compared to Part 1, meaning that the percentage of specimens in the data set which have straightness variation less than 10% of a0 increases to 70.7%. For weld and HAZ specimens the average straightness variation is 18 to 20% meaning that between 50 and 65% of specimens are qualified with variation less than 20% of a0.
Fig.3 shows the crack front straightness versus specimen thickness, for straightness as defined from BS 7448 Part 4. The data points are grouped a lot closer than in Figures 1 and 2, and the effect of specimen thickness on crack front straightness is reduced. If straightness values of fatigue pre-cracks for parent metal specimens are calculated to the criteria given in Part 4, the average crack front straightness is 12.4% and the percentage of specimens in the data set which qualify is 34.5%. 72 to 84% of Weld and HAZ specimens meet the crack straightness requirements in Part 4.
Figs.4 and 5 show charts of crack front straightness versus specimen thickness for crack front straightness defined as per ISO 12135 and BS EN ISO 15653 (for welds and HAZs), and ASTM E1820 respectively. The crack front straightness calculated to ISO 12135 would result in an average of 5.5% and a pass rate of 89.7% for parent metal specimens. The weld and HAZ specimens have an average pre-crack straightness of 10%, and when permitted up to 20% crack straightness in BS EN ISO 15653 means that 96% of these specimens are qualified. ASTM E1820 defines crack front straightness using specimen thickness and sets a 5% criterion limit. Under this regime, 100% of parent metal specimens in the data set would qualify, and over 92% of weld and HAZ specimens.
The relationship between ASTM E1820 crack front straightness and specimen thickness for parent metal specimens in the data set is actually the opposite of that exhibited under the other standards. The data shown in Figure 5 suggest that crack front straightness of parent material gets slightly worse with increasing specimen thickness.
Effect of crack curvature on fracture toughness
The various methods to determine crack curvature given in the different standards give a corresponding wide range of qualification rates, but does it matter which test standard method is used? The objective of setting limits within a testing standard is to avoid validating test results from specimens with crack curvature that is sufficiently excessive to affect the fracture toughness result. Before a particular method can be recommended for a testing standard for SENT testing, it is necessary to make sure that any curvature limits are neither too lenient nor over-restrictive. A reasonable level of crack curvature should not cause the fracture toughness to vary by more than 10% from a perfectly straight crack front, based on other validity limits in Standards being set to give a similar level of error. To quantify the effect of crack curvature on the fracture toughness a series of numerical models were carried out for SENT specimens of similar dimensions and average crack depth, varying only in the amount of crack curvature.
Finite element models of SENT specimens were generated in Abaqus CAE version 6.11-2, and an example of the finite element model and mesh for crack case 2 is shown in Fig.6. A quarter of the specimen was modelled, with symmetry defined on the (x,y)- and (z,y)-planes. Three solid, deformable parts were assembled with tie constraints to allow for a finer mesh near the crack tip where the highest stress concentrations were expected and a coarser mesh further away near the grips.
Fig.6. An example of the quarter-specimen mesh from one of the SENT specimen numerical models (crack case 2).
The model was meshed with 20-node quadratic brick, reduced integration elements (element type C3D20R in Abaqus). Incremental plasticity theory was used with the elastic-plastic material data taken from smoothed and averaged experimental data as shown in Fig.7. The Young's modulus was taken to be 203GPa with a Poisson's ratio of 0.3.True stress and true plastic strain data were then calculated based on the yield point of the material at 523MPa. A sharp crack was defined with propagation direction normal to the crack front because of its curved profile. The J-integral was calculated using a domain integral technique over 11 contours, with the radius of the outermost contour approximately equal to 2.1mm. Convergence of J was observed by the 11th contour and thus path-independence of J was achieved. To simulate the SENT test, the faces that would be attached to the 'mechanical grip' were kinematically coupled to a reference point, which was displaced in the x-direction with the reaction force output at each increment. A single point was constrained in the y-direction to ensure no rigid body motion.
The crack shapes assessed using numerical modelling are shown in Figure 8. All the SENT specimens were 2BxB of 20mm by 40mm in size, notched to an average depth, a0 of 6.6mm, giving an a/W ratio of 0.33 for each model case. The specimens were assumed to be parent metal with yield strength of 523MPa and tensile strength of 610MPa.
Fig.7 Stress-strain experimental data and smoothed data used in FEA
Fig.8 Crack shapes modelled to assess the effect of crack curvature on the fracture toughness
Table 3. Numerical modelling crack shape cases (shown in Fig.8) and their percentage of crack curvature to various standards. Also listed are the crack curvature percentage limits for each standard.
|Model case||BS 7448|
|ISO 12135||ASTM E1820|
Case 1 is a perfectly straight crack front for reference, while Case 2 is a curved crack front which is sufficiently straight to be permissible by all the standards (BS 7448 Parts 1 and 4, ISO 12135 and ASTM E1820). Case 3 is not permissible to BS 7448 Part 1, while case 4 is not permissible to BS 7448 Parts 1 or 4. Case 5 is not permissible by any of the UK standards for parent metals and is only permissible to ASTM E1820. The percentage curvature for the crack front shapes shown in Fig.6 assessed to various standards is summarized in Table 3.
Table 4 shows the predicted values of J and CTOD for each model case, and the percentage error for each case relative to a perfectly straight crack front. The J and CTOD values correspond to the first attainment of a maximum force plateau in the simulated SENT test, as established at the specimen mid-plane (i.e. maximum crack length). All the analyses assume a stationary crack. It can be seen that all the crack shapes modelled had CTOD results that were within 10% of the result for a perfectly straight crack front, whereas the scatter in the values of J is slightly higher, up to 13% in model case 5.
Table 4. Values of J and CTOD predicted using numerical modelling, from specimens with a straight crack front (case 1) to the most curved pre-crack shape (case 5). The percentage error in fracture toughness relative to the straight crack front are also given.
|Model case||Max load|
Effect of thickness
The results from this extensive data set show that there is a weak correlation between crack front straightness and specimen thickness, indicating that crack front straightness does improve with increasing specimen thickness. This improvement is understood to be because the crack length is physically longer and therefore has a greater opportunity to straighten up during pre-cracking. An exception to this observation is the parent data assessed to ASTM E1820, but this may be related to the inclusion of specimen thickness in the derivation of curvature. It can also be noted that because of this definition, crack front straightness is more easily achieved to ASTM E1820 for the over-square specimens in this study, compared to other standards.
Linear regression was used to plot the best fit lines for Figs.1 to 5 and gave coefficients of determination or R-squared (R2) values to indicate how well the trendline matched the original data points. An R2 of 1.0 is a perfect fit, while an R2 of zero shows no relationship at all. None of the trendlines showed strong correlations; all the R2 values were below 0.3, with more than half showing correlations of less than 0.1. This means that the effect of specimen thickness on the SENT specimen crack front curvature is minor.
Effect of testing standard
It has been shown that for a given set of parent metal SENT specimens, the crack front straightness qualification rate ranges between 13.8% (BS7448: Part 1) and 100% (ASTM E1820) depending on the standard used. This clearly highlights the large differences between how different standards quantify and asses crack front straightness.
Although none of these standards has been written and validated for test specimens with the 2BxB SENT geometry, when SENT testing becomes fully standardised it is important that the crack front straightness qualification requirements are reasonable and can be met by the majority of specimens. A validity limit which intrinsically invalidates, say, more than 25% of the test specimens would be unfeasible and uncommercial for a testing standard. This makes BS 7448 Part 1 and 2, and Part 4 for parent metal specimens, too strict to allow the majority of test specimens to pass their criteria.
Any qualification limit should also not be so permissive so that it allows excessive crack curvature that could affect the fracture toughness result. An error of, say, around 10% in the value of fracture toughness determined would be considered acceptable for a curved crack front limit.
The numerical modeling was carried out to quantify the error in fracture toughness with different levels of crack front curvature. All of the standards reviewed in this work would be acceptable as crack straightness validity criteria for determining CTOD from SENT specimens, since they all gave less than 10% error even for case 5. However, the error in the values of J were slightly higher, giving 10% error for case 4 and 13% error in case 5. Therefore the numerical modelling cases show that the curvature of case 4 would seem acceptable without biasing the fracture toughness results, with perhaps some relaxation closer to case 5 being permissible for welds and HAZs where CTOD is the required fracture toughness parameter.
These results suggest that the 20% maximum curvature limit in BS EN ISO 15653 and the 5%B maximum limit in ASTM E1820 seem rather too generous, and that specimens which are at the maximum of these curvature allowances may show excessive error in the value of fracture toughness obtained, especially when J is to be determined.
The crack straightness requirements defined in BS 7448 Part 4 but modified to allow up to 20% curvature for both parent and weld metal SENT specimens would result in over 70% of weld specimens and over 90% of parent metal specimens in the data set used in this paper meeting the crack straightness requirements, while the potential error in CTOD would be less than 10% and the error in J would be less than 13%. This recommendation seems to be the best compromise between quality and feasibility.
All these analyses are based on test specimens with plain sides. An SENT testing procedure being developed by ExxonMobil[Tang, 2010] uses square section (BxB) specimens with side-grooves machined into the sides after fatigue pre-cracking. This has been done to achieve sufficiently straight fatigue pre-crack fronts and for straight-fronted ductile tearing when determining fracture toughness resistance curves. Clearly, the use of side-grooving would make it easier to qualify tests for crack front straightness and such an option should be included in the development on an SENT testing standard.
It is recommended that for SENT specimens the crack front straightness qualification requirement defined in BS 7448 Part 4 be employed, but changed to allow up to 20% curvature for specimens notched into parent metal, weld or HAZ.
The authors wish to acknowledge the contributions of our TWI colleagues Jerry Godden, David Seaman and Phil Robinson for their work in carrying out the SENT tests described here, and thanks also Emily Hutchison and Tyler London who carried out the numerical modelling for this work.
ASTM (2008). 'Standard test method for crack tip opening displacement (CTOD) fracture toughness measurement', ASTM E1290.
ASTM (2009). 'Standard test method for measurement of fracture toughness', ASTM E1820.
BSI (1991). 'Fracture mechanics toughness tests: Part 1: Method for determination of KIC, critical CTOD and critical J values of metallic materials', BS 7448-1:1991.
BSI (1997a). 'Fracture mechanics toughness tests: Part 2: Method for determination of KIC, critical CTOD and critical J values of welds in metallic materials', BS 7448-2:1997. SUPERSEDED BY BS EN ISO 15653:2010.
BSI (1997b) 'Fracture mechanics toughness tests. Part 4: Method for determination of fracture resistance curves and initiation values for stable crack extension in metallic materials' BS 7448-4:1997.
BSI (2005) 'Fracture mechanics toughness tests. Part 3: Method for determination of fracture toughness of metallic materials at rates of increase in stress intensity factor greater than 3.0 MPa m0.5 s-1' BS 7448-3:2005.
DNV (2006). 'Fracture control for pipeline installation methods introducing cyclic plastic strain' Recommended Practice DNV-RP-F108.
ISO (2002). 'Metallic materials - Unified method of test for the determination of quasistatic fracture toughness' ISO 12135:2002 INC CORR 1.
ISO (2010). 'Metallic materials - Method of test for the determination of quasistatic fracture toughness of welds'. BS EN ISO 15653:2010.
Pisarski, H (2010). 'Determination of pipe girth weld fracture toughness using SENT specimens', in Proc. 8th International Pipeline Conference, IPC2010, Calgary, Alberta, Canada.
Tang, H et al (2010). 'Development of the SENT test for strain-based design of welded pipelines', in Proc. 8th International Pipeline Conference, IPC2010, Calgary, Alberta, Canada.