TWI Ltd, Granta Park Great Abington, Cambridge, CB21 6AL, UK
Paper presented at Proceedings of the ASME 2014 33rd International Conference on Ocean, Offshore and Arctic Engineering OMAE2014 June 8-13, 2014, San Francisco, California, USA
Fracture toughness test standards require specimens to be fatigue precracked to generate conservative values of fracture toughness. Nonetheless, it is believed that for ductile steels on the upper-shelf, whether or not the electro-discharge machined (EDM) notch is subsequently fatigue precracked does not affect the value of fracture toughness obtained. Avoiding fatigue precracking during single edge notched tension (SENT) specimen preparation would reduce the testing time, and improve notch placement accuracy and straightness. However, there are circumstances when using EDM notches causes the fracture toughness to be overestimated. It is important to know when fatigue precracking affects the fracture toughness and when it does not. In the work presented here, SENT testing was performed on EDM notched specimens, and on specimens which were subsequently fatigue precracked. Tests were conducted at +20°C, -20°C and -80°C to compare ductile and brittle behaviour. The full tearing resistance curves (R-curves) were reasonably independent of the fatigue precracking, but the initiation value of δ0.2 was higher when EDM notches were used. At lower test temperatures, the difference in fracture behaviour between both notch types was more significant. EDM notches can therefore be most justified for the assessment of fracture toughness determined from the maximum load in the load-displacement curve, i.e. upper shelf behaviour. The upper shelf can be determined from standard fracture toughness testing or estimated using Charpy data, without performing additional fracture toughness tests. Several Charpy-based criteria for determining the temperature of the upper shelf were also evaluated in this work.
There is an accepted assumption that fatigue precracking of large or small-scale fracture toughness specimens does not materially affect fracture behaviour if the material is a ductile modern steel on the upper shelf (while not necessarily holding true when brittle behaviour is possible). However, fatigue precracking is required by all fracture toughness test standards, in order to generate conservative values of fracture toughness, regardless of whether the fracture behaviour is ductile or brittle.
Engineering critical assessment (ECA) of pipelines, as described in DNV RP F108 and DNV OS F101 Appendix A, now typically requires the generation of SENT R-curves, and presuppose that the material is on the upper shelf. Testing schedules could be significantly accelerated, and costs reduced, by omitting the use of fatigue precracking during the preparation of test specimens. Being able to carry out fracture toughness testing on specimens without fatigue precracking allows a significant reduction in the overall time for testing. Specimens can be tested as soon as they are machined, instead of requiring an additional precracking stage. The measurement of the precrack shape and the number of validity checks required after testing on the fatigue precrack shape also takes considerable time. Also, without fatigue precracking there is less risk of specimens (particularly weld or HAZ specimens) failing to fully qualify to the notch shape validity limits, where residual stresses can interfere the stable growth of the fatigue precrack and result in irregular shaped cracks.
Fracture toughness tends to decrease with increasing notch acuity (sharpness). To ensure the sharpest crack is tested, fracture toughness test standards normally use fatigue precracked specimens. Fatigue precracking has been the practice since the early days of fracture toughness testing, and in the 1970s the ASTM standards for fatigue precracking were developed to ensure that appreciable plastic material damage to the specimen does not occur during fatigue cracking. The precracking recommendations in the British standards were mostly based on the results of work carried out by ASTM with validation for CTOD testing , allowing a harmonised approach to precracking of single edge notched bend (SENB) fracture toughness test specimens since then.
The intention of fatigue precracking fracture toughness specimens is so that the crack in the test sample imitates the worst (sharpest) potential crack in the material. A notch is machined into the fracture toughness specimen, from which a fatigue crack is grown by applying cyclic loading to the specimen. The process of inserting a fatigue precrack can present difficulties, as it is important that the fatigue crack grows in a uniform fashion. If precracking is performed at too high a stress intensity range then a large plastic zone can form at the precrack tip which may mask the behaviour of the actual material during the test. The fracture mechanics test standards include many checks and restrictions on the fatigue precrack size, position and shape, together with limitations on the maximum allowable fatigue load. Many of these checks can only be performed after testing, and cannot always be met, especially in weld specimens where residual stresses can affect the growth of the precrack. Local compression is therefore often used for weld specimens, where a round platen is used to compress the ligament ahead of the machined notch on either side of the specimen, which relaxes any residual stresses in the specimen prior to fatigue precracking. Guidance on this is given in BS EN ISO 15653 . These standard test methods for SENB specimens, and compact tension specimens, can be used to obtain fracture toughness in both the brittle and ductile regimes.
In contrast to the history of SENB specimens, SENT specimens were developed to determine R-curves for ductile pipe girth weld materials, which has led to the requirement for fatigue precracking to be questioned as potentially unnecessary for specimens where brittle fracture is not a concern. SENT test specimens exhibit lower constraint than the equivalent thickness SENB specimen, which means they measure higher values of fracture toughness, all else being equal. However, SENT specimens much more closely match the constraint of a flaw in an axially loaded pipe girth weld, giving SENT test specimens an advantage when used in the fitness for service assessment of such pipes. Exxonmobil have developed their own testing protocol for pipes prequalified for high strain conditions, for the measurement of CTOD R-curves in SENT specimens , which permits both fatigue precracked and electro-discharge machined (EDM) notches. The document states that for SENT specimens the EDM notch must be made with a wire of diameter 0.15mm or smaller, at least in the final third of the notch. Research on SENT R-curves by Østby et al  also used EDM notched SENT specimens without further precracking, and assumed this method was acceptable for ductile materials when generating R-curves.
A draft British Standard on SENT test procedure to be published as BS 8571  requires fatigue pre-cracking after EDM notching of specimens. There is consideration by the BSI committee to include an appendix in future versions of the standard to allow EDM notches if sufficient research can show it is justified. The purpose of this research is therefore a comparison between fatigue precracked (‘sharp’) notches and EDM cut (‘blunt’) notches in the performance of SENT tests to help with this justification.
3. Testing Programme
3.1 Previous Data on EDM notched SENB Specimens
TWI has carried out a number of previous projects that has involved the testing of blunt notched fracture toughness specimens. A set of data from a TWI internal exploratory project in 2010 tested four SENB specimens in Grade 50D steel with different notch acuity from a fatigue pre-crack to 2mm radius. The tests carried out at 0°C produced fully ductile behaviour in all specimens, but the measured fracture toughness increased as the notch tip radius increased (Figure 1). The designation of the result of a fracture toughness test when the load-displacement trace shows a maximum load is given as δm or Jm; when the result gives at least 0.2mm of stable tearing but does not reach maximum load the result is given a subscript ‘u’ (i.e. δu), whereas if there is less than 0.2mm of stable tearing the result is described as δc.
The results of these SENB tests challenge the assumption that the upper shelf fracture toughness is independent of notch acuity. Follow-on work at TWI that year performed further twelve SENB tests with different notch radii on 35mm thick quenched and tempered structural steel with specified yield stress of 690MPa, collated together with 60 similar tests performed by Anthony Horn for his PhD thesis  so make a set of 72 results altogether. The fracture toughness tests were performed at 0°C, to obtain a range of results from brittle failures (δ/Jc) for the sharpest notches, to unloading results for the bluntest notches. The results confirm that the measured fracture toughness increases, and fracture behaviour appears more ductile as the notch tip radius increased (Figure 1).
Figure 1. Historical SENB fracture toughness test data from TWI, and Anthony Horn  from steel specimens with different notch radii, tested at 0°C.
3.2 SENT Testing
These previous projects on notch acuity used SENB specimens rather than SENTs to compare the effect of the notch on fracture toughness. For this paper, there was a need to generate equivalent data looking at notch acuity for SENT specimens, both for ductile and brittle behaviour. Therefore twenty four 2BxB SENT specimens of size 34x17mm in cross section were machined from 20mm thick X70 GMAW girth welded pipe material, and surface notched into the weld metal from the OD. The material yield strength was 478MPa and ultimate tensile strength (UTS) was 584MPa, and the weld metal yield strength was 604MPa and UTS was 714MPa at room temperature. Twelve of the SENT specimens were EDM notched to a notch depth to specimen width (a/W) ratio of 0.4, and the other twelve were EDM notched and then fatigue precracked to try and achieve the equivalent a/W ratio (although the actual a/W ratios were between 0.41 and 0.45). The EDM wire had a diameter of 0.25mm, creating a notch width of 0.3mm. The ‘blunt’ EDM notch specimens were identified as W01-01 to W01-12, and the ‘sharp’ notch fatigue pre-cracked specimens were W01-13 to W01-24. The first six specimens of each set were used to generate a multiple specimen R-curve at room temperature; the second set of six were tested at low temperatures to compare the relative behaviour when less ductile behaviour is seen.
Initially -20°C was used to test two specimens from each set, but the results at this temperature were still fully ductile, so the remaining four specimens from each set were tested at -80°C. This was the lowest feasible test temperature to be able to maintain a stable temperature during soaking and testing of the SENT specimens. Although it was not cold enough to see fully brittle δ/Jc results (where fracture occurs with less than 0.2mm of stable tearing), it was sufficiently cold to achieve δ/Ju behaviour (fracture before reaching maximum load, but at more than 0.2mm of ductile tearing). These results were sufficient to show significant differences in the fracture behaviour between the two notch types.
The remaining weld material was used to generate a Charpy transition curve from specimens notched into the weld metal.
The multiple-specimen J and CTOD R-curves are shown in Figures 2 and 3. The R-curve generated from ‘blunt’ EDM notched specimens showed a slightly higher R-curve than the one generated from ‘sharp’ fatigue precracked specimens, but with a shallower slope so that both R-curves converged at just beyond 1mm of tearing. There was more scatter in the individual data points from the fatigue precracked specimens, possibly reflecting the variations in the exact notch depth and shape.
The hypothesis tested by this paper is that the apparent fracture toughness measured from tests using EDM notches would be similar to that from fatigue precracked specimens for ductile material on the upper shelf. The results from the R-curve tests performed in this work cast doubt on the robustness of this assumption. The difference in the R-curves for blunt and sharp notched specimens differed most at the initiation of the R-curve, defined as the toughness at 0.2mm of tearing, either interpolated or extrapolated from the experimental data. The R-curve from the EDM notched specimens gave a J0.2 of 348kJ/m2 compared to 231kJ/m2 for the fatigue pre-cracked specimens. In terms of CTOD, the R-curve from the EDM notched specimens gave δ0.2 of 0.53mm compared to 0.33mm for the fatigue precracked specimens. This suggests that even when ductile behaviour is observed, the notch acuity does affect the fracture performance at the initiation of ductile tearing.
Figure 2. J multiple-specimen R-curves generated from SENT specimens with blunt and sharp notches. The position of the initiation fracture toughness is indicated, along with the position above which specimens reached or exceeded maximum load behaviour.
Figure 3. CTOD multiple-specimen R-curves generated from SENT specimens with blunt and sharp notches. The position of the initiation fracture toughness is indicated, along with the position above which specimens reached or exceeded maximum load behaviour.
When the difference in toughness was compared for the specimens which were tested to beyond maximum load in the load-displacement trace (at Δa of ≥0.9mm for these tests), then the results from the EDM notched specimens are similar to the fatigue pre-cracked specimens. This suggests that in order for EDM notched specimens to be equivalent to fatigue-precracked specimens, the appropriate single point value of fracture toughness to determine is that based on the maximum load rather than the initiation of the tearing resistance curve.
4.2 Low Temperature Tests
At a temperature of -20°C both the EDM notched specimens and the pre-cracked specimens showed very similar fully ductile behaviour, with load-displacement traces extending to beyond maximum load (δm behaviour). However, for the tests performed at -80°C the pre-cracked specimens were showing transition zone behaviour with fracture toughness as low as 0.03mm in one instance, and all of the tests showing either δc or δu behaviour. This was significantly different to the EDM notched specimens tested at -80°C which still mainly showed δm fully ductile behaviour and a minimum CTOD of 0.74mm. The fracture toughness results in terms of CTOD plotted against temperature are shown in Figure 4. The values of fracture toughness plotted for room temperature are those for the specimens from the R-curve which were loaded beyond maximum load, and then re assessed at the maximum load position to determine the toughness, in an equivalent way to the single specimens tested at low temperature.
The striking difference in the fracture toughness measured at -80°C between the fatigue precracked specimens and the EDM notched specimens highlights the risks in masking the true potentially brittle behaviour of a material when using EDM notched specimens which could even show maximum load behaviour. Using inappropriate fracture toughness data in this way would lead to large overestimates of the critical weld flaw sizes, meaning that structures to could be assessed as safe when there is a chance of brittle failure occurring.
Figure 4. Fracture behaviour of EDM and fatigue precracked SENT specimens over a range of temperatures
Table 1 Assessment of notch and precrack shapes based on curvature defined from 
|W01-01 to 12
Figure 5. A selection of the fracture faces from SENT specimens tested for the multiple specimen R-curves, notched with EDM alone (top) and with fatigue precracking (below). All specimens had nominal a/W ratio of 0.4.
5.1 Notch Shape & Validity
One benefit of notching solely with EDM is that the notch front is perfectly straight, and there are no concerns about irregular notch shapes which can be an issue when fatigue precracking weld or HAZ specimens. The fatigue precracked specimens tested in this work had precrack shapes assessed according to BS 7448 Part 4 , as the maximum difference between any one of the nine crack front measurement and the average crack length, and had curvatures from 7% and 17% (Table 1). These are within the tolerances justified from recent TWI research , which recommended that less than 20% curvature to this criterion means that specimens are valid. The variation in the fatigue pre-crack shapes can be seen in Figure 5, against the perfectly straight EDM notched specimens for the R-curve, which then show the stable tearing during the test very clearly. However, none of the fatigue precrack shapes in the SENT specimens tested in this work was invalid, suggesting that using EDM as a way to ensure acceptable notch shapes was not necessary for these surface notched weld metal specimens.
5.2 The Upper Shelf
The hypothesis that precracking is not necessary for SENT specimens which exhibit ductile behaviour, i.e. on the upper shelf, depends upon having a reliable method to ensure that the material is indeed on the upper shelf prior to SENT testing. Since the blunt notch SENT test could mask potentially brittle behaviour (as seen in Figure 4), it in itself cannot be relied upon to demonstrate that the material is on the upper shelf.
The most reliable method to confirm upper shelf behaviour would be to carry out triplicate tests on precracked SENB specimens at the intended test temperature to demonstrate ductile behaviour. This method would give absolute proof of fracture behaviour on the upper shelf, but would not assist much in shortening the overall duration of the SENT tests, nor reduce the costs, since it would involve an additional stage of fracture toughness testing either preceding or concurrent with the SENT testing. However, these validation tests need not be R-curve tests; it would be sufficient for the SENB specimen to give a load versus displacement trace which extends beyond the maximum load in order to show sufficient ductility.
A potential alternative to carrying out this additional fracture toughness testing is to make an estimation of upper shelf behaviour based on Charpy tests, which are simpler and quicker to perform. There are some published methods to determine the temperature above which upper shelf behaviour will occur in fracture toughness tests, based on analysis of the Charpy ductile to brittle transition curve.
If neither Charpy-based methods nor other fracture toughness data can show that the material is on the upper shelf then fatigue precracked SENT specimens should be used in the absence of any other evidence of upper shelf behaviour.
5.3 Charpy Criteria for the Upper Shelf
The use of Charpy data to predict upper shelf fracture toughness behaviour is not new. In 1984 Denys and Lin  discussed notch tip acuity in SENB specimens, and performed their own programme of Charpy tests to accompany their precracked and sawn notch SENB tests. Their data was fairly scattered, but suggested that a 27J Charpy impact energy was sufficient to ensure enough ductility in order for the notch acuity to be irrelevant. Although this approach seems sensible, the limited data upon which it is based suggests that a more cautious method would be necessary to be able to make a more general recommendation for SENT specimens.
More recently, Wilkowski et al.  performed SENT tests using fatigue precracked specimens over a range of temperatures to define the ductile to brittle transition as part of the validation of their Master Curve of Fracture Initiation Transition Temperature (FITT). The FITT is defined as the temperature at which the SENT test specimen tested under displacement control will give a load versus displacement trace which just reaches a maximum load before the specimen fails (Figure 6). This is also the definition for a fracture toughness test where the result changes from being δu to δm, and is therefore the FITT is a reasonable description of the beginning of the upper shelf.
The Master Curve of Fracture Initiation Transition Temperature (FITT) from Wilkowski et al  is determined based on the temperature at which 85% shear area is obtained from a ductile to brittle Charpy transition curve. The 85% shear area is equivalent to 15% crystallinity, which is how the fracture appearance is sometimes alternatively expressed. The T85%SA temperature can be determined by fitting an S-shaped ‘tanh’ function to the Charpy data tested over a range of temperatures, and using this curve function to define the temperature at 85% shear area. This method means that Charpy specimens can be used to predict the temperature at which SENT specimens would be just at the upper shelf, and may therefore be the most appropriate Charpy-based criterion for this research.
Charpy testing can be subject to significant amounts of scatter, particularly for specimens tested around the transition temperature, and typically ten Charpy tests used to generate the ductile-to-brittle transition curve might fall neatly onto a tanh function, or not. It might not be possible to place as much importance on values of T85% determined from scattered Charpy data than data sets which neatly define a specific transition curve.
Figure 6. Definition of Fracture Initiation Transition Temperature based on just reaching maximum load, as given in .
5.4 Evaluating the Charpy Criteria
The Charpy transition curve for the weld metal notched specimens in this work is shown in Figure 7. The Charpy transition ‘curve’ for this steel was linear rather than showing a defined S-shaped tanh function, particularly for the upper shelf where, with only nine Charpy specimens available to test, there was insufficient data to define an upper shelf plateau. Nonetheless, a best attempt at fitting a tanh function to the lower shelf data was made in order to evaluate the Charpy-based methods to define upper shelf fracture behaviour (Figure 8). The highest Charpy impact test result was 190J at -10°C; the fitted tanh function estimated the upper shelf to be around 180J.
Figure 7. Charpy transition curve for specimens notched into the weld metal, expressed in terms of impact energy and % crystallinity.
Figure 8. Charpy transition curve tanh curve best fit to the experimental data
The first Charpy criterion to define the upper shelf from  is based on 27J Charpy energy, developed for SENB fracture toughness specimens. From the experimental Charpy transition curve, 27J is achieved at -95°C. Based on this criterion, this temperature would be taken as the temperature above which notch acuity is irrelevant, but the experimental SENT data shown in Figure 4 shows that the notch sharpness at a test temperature higher than this (at -80°C) shows a significant effect. Therefore this Charpy criterion is not appropriate for SENT specimens.
Using the Wilkowski method  to determine the Fracture Initiation Transition Temperature (FITT) first requires the 85% shear area to be determined. The Charpy transition curve in Figure 7 gives an 85% shear area (equivalent to 15% crystallinity) at -15°C. Converting this into a prediction of FITT is done using Wilkowski et al’s empirical formulae  in a spreadsheet based on their own experimental data. This calculation was performed assuming an a/W ratio of 0.4 for a pipe wall thickness of 17mm. This gave a FITT prediction of around -80°C. This is one of the temperatures at which the SENT tests were carried out, and it was the temperature at which the significant difference in fracture behaviour between the notched was seen. This method therefore seems to have potential as a Charpy-based criterion to justify using EDM notched SENT specimens. However, to fully validate the Wilkowski method, it is recommended to perform further SENT tests using blunt and sharp notches at temperatures close to the FITT to determine how accurate the prediction of FITT is for ensuring upper shelf behaviour.
Conclusions and Recommendations
The following conclusions have been drawn from the results of this work:
- Experimental R-curves generated for sharp and blunt notched SENTs showed that the blunt notched R-curve was slightly higher (possibly due to the slightly shallower notch depth and/or the blunter notch) but that both R-curves converged after a level of tearing greater than around 1mm.
- When comparing the interpolated or extrapolated initiation fracture toughness (δ0.2) alone, the two notch types showed a significant difference in toughness at 20°C, with EDM notched specimens over-estimating the fracture toughness determined from precracked specimens.
- There is little difference between the fracture toughness determined at the maximum load for EDM and fatigue precracked specimens.
- At temperatures where the pre-cracked specimens started to show transition behaviour, at -80°C in this case, there were significant difference in the fracture toughness values measured, since the EDM notched specimens were still showing fully ductile behaviour.
- EDM notched SENT specimens should only be used for the assessment of fracture toughness determined from the maximum load in the load-displacement curve, i.e. upper shelf behaviour.
The recommendations are that:
If it cannot be shown that the material is on the upper shelf then precracking must be used for SENT specimens to avoid over-estimating the fracture toughness.
The situation for which the EDM notch seems most suitable is for generating a full R-curve provided that upper shelf behaviour can be guaranteed, or for the determination of a single point value of toughness from the maximum load of the load-displacement trace instead of the initiation fracture toughness from the R-curve. EDM notching should use a wire up to 0.3mm wide. Further data and validation by comparing precracked and EDM notched SENT tests would strengthen these recommendations.
The most promising Charpy-based criteria for ensuring the upper shelf for SENT specimens is based on using the Charpy transition curve to determine the FITT using the Wilkowski method, which is the temperature at which the load-displacement trace for a SENT test just reaches maximum load.
The confidence in these recommendations would be increased by generating further experimental data on different materials and welds to confirm the conclusions of this work.
- Egan G R, 1971: ‘Some aspects of fracture toughness testing’. Research report, Cambridge: The Welding Institute, 1971.
- BS EN ISO 15653:2010 ‘Metallic materials -- Method of test for the determination of quasistatic fracture toughness of welds’ British Standards Institution 2010.
- ExxonMobil, 2010: ‘Measurement of Crack-Tip Opening Displacement (CTOD) Fracture Resistance Curves Using Single-Edge Notched Tension (SENT) Specimens, ExxonMobil Upstream Research Company, September 20, 2010.
- Østby E, Sandvik A, Levold E, Nyhus B, Thaulow C, 2009, ‘The effects of weld metal mismatch and crack position on the strain capacity in SENT specimens in an X65 material’, in Proceedings ISOPE-2009, The 19th International Offshore (Ocean) and Polar Engineering Conference, Osaka, Japan, June 21−26.
- BS 8571:2014, ‘Method of test for determination of fracture toughness of welds in metallic materials using single edge notched tension (SENT) specimens’. British Standards Institution 2014.
- Horn A, ‘Development of an engineering assessment procedure for predicting cleavage fracture from non-sharp defects using the Failure Assessment Diagram’ PhD Thesis, University of Manchester, 2010.
- BS 7448-4 ‘Fracture mechanics toughness tests – Part 4: method for determination of fracture resistance curves and initiation values for stable crack extension in metallic materials’. British Standards Institution, 1997.
- Malpas A, Moore P and Pisarski H, 2012: ‘Crack front straightness validity in SENT specimens’, In Proceedings ISOPE-2012, The 22nd International Offshore (Ocean) and Polar Engineering Conference, Rhodes, Greece, June 17−22.
- Denys R and Lin Y S, 1984: ‘The influence of notch tip acuity on the CTOD [crack tip opening displacement] fracture toughness’, In Proceedings, Quality and Reliability in Welding., International Conference, Hangzhou (Hangchow), China, 6-8 Sept.1984. Session C. Paper C2.
- Wilkowski G, Rudland D, Shim D-J, Horsley D, 2008: ‘Predicting the brittle-to-ductile transition temperatures for surface cracks in pipeline girth welds – it’s better than you thought’, in Proceedings IPC 2008 7th International Pipeline Conference, Calgary, Canada, 29 Sept -3 Oct.