Full-Scale Testing for Validation of Fracture Assessments

Paper presented at FITNET 2006. International Conference on Fitness-for-Service, 17-19 May 2006, Amsterdam, The Netherlands. Paper FITNET 06-018

Abstract

Around 300 full-scale and large-scale fracture test results have been examined and interpreted in terms of the new European fitness-for-service procedure FITNET and/or the UK procedure BS7910. The cases covered a range of:

• materials (structural steels, line pipe steels, stainless steels, aluminium alloys),
• component types (plates, cylinders, welded details),
• flaw geometries (through-thickness, surface, embedded), flaw types (natural and artificial),
• sources of loading (pressure, tension, bending, residual stress).

Various FITNET options ('basic' option 0, 'standard' options 1 and 3, and 'advanced' option 5) were used, depending on the type of data available. In all cases examined, failure occurred outside the failure analysis line, as would be expected when representative but conservative input data are used.

This paper highlights three series of tests (64 test results) from the 300 tests examined, to illustrate various aspects of the FITNET procedure.

Introduction

Part of the work of the FITNET consortium (Work Package 5, led by VTT, Finland) was to collate a number of failure case studies for fracture, fatigue, creep and corrosion damage, and to analyse them in terms of existing procedures, in order to provide:

• Validation of the existing FFS procedures.
• An indication of areas where current procedures might need to be modified.
• A set of test data for validation of future amendments to the procedure.
• A teaching aid based on real laboratory data and structural failures.

TWI has access to a wide range of data from large-scale fracture tests, predominantly wide plate tests and pressure tests, some of which go back over 20 years. It also has details of many failure case histories. Some of these data have already been used for validation of fitness-for-service (FFS) assessments, namely the British procedure PD6493^[1] and its successor BS 7910^[2].Results of the validation studies were published, first in the form of TWI Members' Reports,^[3,4] and later as open publications. ^[5]

As part of its contribution to FITNET, TWI updated the case studies presented in ^[3] using current BS 7910 procedures and software, and augmented them by considering additional case studies carried out subsequently. Full analyses, comprising around 300 test results, are given in section 13 of the FITNET procedure and in reference. ^[4]

Note that, at the time of carrying out the study, the FITNET fracture assessment procedures were still under development. By contrast, the BS7910 procedures were in the public domain, and appropriate commercial software was available. Consequently, the analyses were carried out initially using BS7910, and subsequently re-interpreted in terms of the draft FITNET procedures. The two methods are similar in principle, with a hierarchy of possible approaches('Levels' in BS7910, 'Options' in FITNET). Table 1 shows the approximate equivalence between the two fracture assessment procedures.

Table 1 Approximate equivalents between BS7910 and FITNET fracture assessment levels

This paper presents just three sets of case studies from the full report. The first concerns a series of wide plate tests (unaxial and biaxial) carried out over a range of temperatures on a pressure vessel steel (ASTM A533B). The second addresses a programme of wide plate tests on plain structural steel, plus additional tests on welded structural details made from the same material. The third set of tests comprised unaxial and biaxial wide plate tests carried out on a pressure vessel steel, with selected specimens subjected to preloads in order to investigate load history effects.

A533B Test programme

At various times over the past 20 years, TWI carried out a series of wide plate tests^[6] on A533B pressure vessel steel, mainly on behalf of the Nuclear Installations Inspectorate (NII) and the Central Electricity Generating Board (CEGB). These programmes were concerned with various aspects of the structural integrity of steels, including the effects of biaxial loading and warm pre-stressing on the fracture behaviour ofASTM A533B pressure vessel steel.

The material was characterised through a series of small-scale tests:

• Charpy tests in the L-T and L-S orientation,
• Standard fracture mechanics testing (SENB specimens), to generate single-point values of toughness (in terms of K_Ic, critical J and critical CTOD) at a range of temperatures
• Tearing resistance curves (R-curves) in terms of J and CTOD, at temperatures of 100, -70 and +70°C.
• Low-constraint fracture mechanics tests, using shallow-notch SENB specimens
• Full stress-strain curves, generated at +70°C, 70°C and -96°C.

Wide plate tests were then carried out, at a range of temperatures between +70°C and -163°C, covering the full ductile to brittle transition temperature curve. Biaxial load ratios of k=0 (ie uniaxial loading), k=0.5 and k=1 (ie equibiaxial loading) were included in the test programme, where k represents the ratio of the load parallel to the flaw to that perpendicular to the flaw. Both semi-elliptical surface breaking and through-thickness flaws were introduced, ie SCT (surface-cracked tension) and CCT (centre-cracked tension) specimens. Plates were either tested in the original 50mm thick condition, or machined down to 25mm thick plate. The full wide plate test matrix is shown inTableable 2.

Table 2 Details of wide plate tests on A533B grade steel

The wide plate tests were analysed in accordance with BS7910 level 2a or 3a, approximately equivalent to FITNET Option 1 (initiation route and tearing route), as follows:

• The membrane stress applied perpendicular to the flaw at the point of failure was treated as the primary stress, P_m.
• The fracture toughness K_mat used in the analysis was taken as the lowest of three SENB test results or equivalent (ie Minimum of Three Equivalent or MOTE) at the appropriate temperature and section thickness.
• Analyses were carried out using a BS 7910 Level 2a FAD. This is very close to the FITNET Option 1 FAD for continuous yielding.
• The appropriate tensile properties for each test temperature were estimated from the properties measured at +70°C (this simplification allows all results to be plotted on a single FAD). Note that in practice the material showed discontinuous yielding, which is reflected in the shape of the FAD for L_r>1.

The results of the analysis are given in Figure 1. Numbers adjacent to each data point indicate the test ID as shown in Table 2. It can be seen that all the specimens failed in the 'unacceptable' region of the failure assessment diagram (FAD), as expected. Results fall in the fracture-dominated (ID 2, 8, 15, 16, 40), 'knee' (ID 14, 41) and collapse-dominated (all other data points) regions of the FAD, depending on the test temperature.

Option 1 of FITNET gives the user the choice of an analysis based on initiation of fracture only, or a full tearing analysis. All of the single points in Figure 1 correspond to initiation of fracture (analysis to BS7910 level2a).

The results of test ID 1 have also been interpreted using an Option 1 tearing analysis (BS7910 level 3a); the result can be seen in Figure 1 as a locus of points corresponding to various levels of tear length between 0.2and 5mm. At low levels of tear length, Δa, the initiation and tearing analyses are virtually coincident.

Selected low-temperature tests (ID 12, 40 and 41) were analysed next using the FITNET Option 3 (BS7910 Level 2b) FAD. Results of the analysis are shown in Figure 2. Note the slightly different shape of the FAD (solid line, based on the actual stress-strain curve at -96°C) compared with the Option 1 analysis (dashed line). Since the Option 1 failure line lies wholly inside the Option 3 line, there is a slight benefit in using Option 3, although this is more apparent for test ID 12 than for the other tests.

The CCT tests on 50mm thick plate at -100°C (ID 40 and 41) proved to be particularly amenable to analysis using the more advanced options in FITNET. Numerical analysis^[7] of the CCT test configuration had predicted a strong effect of biaxiality on the effective toughness of through-cracked plates tested in the lower transition, with uniaxially loaded specimens showing an effective toughness some four times higher than that of SENB specimens. Conversely, equibiaxially loaded specimens (k=1) were predicted to have an effective toughness approximately the same as that of an SENB specimen. This trend, initially predicted by numerical analysis, was confirmed by wide plate testing. The effect can be seen in Figure 2 by comparing the positions of the analysis points 40 (equibiaxial loading) and 41 (uniaxial loading). The latter point lies further from the failure analysis line, ie the safety factor associated with the standard analysis route is higher.

Later, further analysis of the CCT wide plate test results was carried out under the European SINTAP ('Structural Integrity procedures for European Industry'^[8]) programme.^[9,10] A series of low-constraint small-scale fracture mechanics tests was carried out using shallow-notched SENB specimens as part of the validation of Appendix 3 of the SINTAP procedure, which addresses the treatment of crack tip constraint. The uniaxial wide plate test (ID 41) was then re-analysed using the SINTAP 'constraint matching' option, equivalent to FITNET Option 5. The equivalent analysis was not carried out with test ID 40, since the similar constraint conditions in deeply-notched SENB specimens and equibiaxially loaded CCT plates would produce similar results to those of an Option 3 analysis.

Figure 3 compares the results of a 'standard' analysis to BS 7910 (FITNET Option 3) with a more advanced analysis (FITNET Option 5) using the constraint matching approach. The datapoint associated with 'constraint-corrected' data lies much closer to the failure analysis line, indicating the benefits to be gained from using constraint correction.

A further option available in FITNET, but recommended only in cases where no fracture mechanics data are available, is the so-called Option 0. Here fracture toughness is estimated from Charpy energy using lower-bound correlations. An Option 0 analysis has been carried out for illustration on three tests from this series: ID 12, 40 and 41. The fracture toughness was estimated from the appropriate T27J temperature using the FITNET MasterCurve correlations. An analysis was then carried out using the Option 1 FAD, leading to values of K_r =2.79, 3.39 and 3.63 for tests 12, 40 and 41 respectively (L_r values remain as shown in Figure 2).

In order to compare the various assessment levels, the associated safety factors were calculated, as shown in Figure 4. If O represents the origin of the FAD and B represents the analysis point associated with the test result, the line OB intersects the failure analysis line at A. The ratio OB/OA represents the safety factor associated with the analysis.

Table 3 shows the safety factors for various levels of analysis for test ID 12, 40 and 41. A trend of decreasing safety factor with increasing Option no. is evident; in particular there is a large benefit in using Option 1instead of Option 0. The benefits of moving from Option 1 to Option 3 could be significant in the 'knee' region of the FAD (see Figure 2) but have little influence on the fracture-dominated cases ID 40 and 41. Moving up from Option 3 to Option 5 is beneficial in the case of test ID 41, but would be unlikely to have much effect in the case of test ID 40, where the analysis point lies close to the FAD even at Option 3.

Table 3 Safety factors associated with analysis of selected cases using FITNET Options 0-5

Structural steel test programme

The second set of case studies concerns a large collaborative fracture toughness testing programme, conducted in 1984 by twelve European laboratories, including TWI.^[11] This programme was selected for the FITNET validation task because of the existence of a large database of underlying fracture toughness data (fracture toughness was determined on many hundreds of specimens, at several European laboratories). Small-scale fracture mechanics test data were available throughout the transition range, using standard (deeply-notched) specimens. [Some 'shallow notch' testing was also carried out but has not been considered in this work, since the results cannot readily be interpreted in terms of modern test standards.]

The material was 52mm thick 'node quality' BS 4360 Grade 50D steel (low-sulphur steel with guaranteed through-thickness elongation) for offshore applications.

Four European laboratories, one of which was TWI, carried out wide plate tests, and results for all laboratories are analysed here. In all, 30 parent material wide plates (plate widths ranging from 150mm to 700mm) were carried out,at test temperatures of both -65 (in the transition region) and 120°C (in the lower shelf/lower transition). Test geometries included SCT, CCT and ESCT (extended surface crack), as summarised in Table 4. An additional test result (ID 234) from outside the original test programme has been included; this was a CCT test carried out on a 900mm wide plate of parent material, with the central crack encircled by a through-thickness circular weld made by electron beam welding. The purpose of the ring weld was purely to introduce welding residual stresses in the crack tip area, whilst not affecting the microstructure sampled by the crack tip.

Table 4 Structural steel fracture toughness testing programme 

Results of all 31 tests are shown in terms of a BS 7910 Level 2a (FITNET Option 1) FAD in Figure 5. The value of fracture toughness used in the calculation is based on (µ-1SD), ie one standard deviation below the mean value of fracture toughness, measured on a large sample of specimens in terms of critical J. A cut-off of L_r,max=1.0 has been used, since the material has a discontinuity (a Lüder's extension) in the stress-strain curve. Note that results corresponding to parent materials tested in the transition region (ID 201-226) are clustered in the 'knee' area of the FAD, whilst the four tests carried out at -120°C (ID 227-230) and the single result for the ring-welded specimen (ID 234) lie in the fracture-dominated part of the diagram.

A more accurate analysis of the tests carried out at -65°C is given in terms of a Level 2b/Option 3 analysis in Figure 6, which shows failure of several specimens in the so-called 'knee' region, very close to the failure assessment line. This phenomenon (the observation of lower inherent safety factors in the knee area of the FAD) has also been discussed by Muhammed et al^[12] in a separate study of a large number of wide plate tests.

Large-scale tests on bridge details

Three large-scale tests (ID 231-233) were carried out on welded plates representing various bridge details, as part of a later programme^[13] to devise new rules for avoidance of brittle fracture in bridges (BS 5400:Part 3). Note that the Charpy energy of the steel tested (27J at -103°C) greatly exceeded the specification requirement (27J at -30°C). The very low test temperature (-100°C) was therefore intended to provide a critical test of the proposed rules for bridge design, rather than to represent a realistic temperature for a bridge component.

Table 5 Conditions for bridge detail tests

The tests were carried out on surface-notched and edge-notched plates representing various bridge details, as shown in Figure 7.The tests were carried out on surface-notched and edge-notched plates representing various bridge details, as shown in Figure 7. Standard K-solutions for surface-cracked plates, modified to take account of the presence of a welded attachment, were used for the analysis of the first two cases. In other words, the presence of the welded attachment was accounted for by using an M_k correction factor, derived from the attachment length in accordance with BS 7910 using the 2-D M_k solution. The depth of the welding residual stress field was calculated from the welding heat input as per Annex Q of BS 7910:2005. Note that the very bulky cover plate on the second test stiffens the right-hand side of the specimen, and induces a bending stress, even though the loading applied was nominally tensile. This bending stress, determined from strain gauges, was included in the analysis.

The third case represents an edge-welded detail, which is considered to be very detrimental to structural integrity and is thus not permitted by many codes for load-carrying members. This was analysed by calculating a stress concentration factor (SCF=1.6) at the position corresponding to the crack tip, using a finite element model of the uncracked geometry. It was then analysed as an edge crack in a plate.

Note that, in spite of the presence of welded attachments, all crack tips were located in parent material, so that the tensile and fracture toughness properties from the original ECSC programme (see Table 5) could be used. Welding affected only the details such as secondary stress, determination of M_k and local bending in the cover plate detail.

All three assessment points (see Figure 8) lie outside the FAD, with the cover plate and transverse attachment details relatively close to the failure assessment line. The analysis point for the edge-welded detail (ID 233)lies well outside the failure analysis line (K_r=2.2), suggesting that a more advanced analysis (e.g. using Option 4 of FITNET) might be beneficial, although such an analysis has not been carried out at the time of writing.

Proof loading test programme

The third series of tests considered comprised twelve wide plate tests, carried out at TWI as part of a project to evaluate the influence of warm proof loading on low temperature fracture behaviour^[14]. The material for all the tests was BS 1501-224-490B LT50 pressure vessel steel, and the tests are identified as ID 801-812. Wide plate specimens were notched in both the parent metal (PM) and weld metal (WM), and tested under both uniaxial and biaxial tension at below-ambient temperatures. Some plates were proof loaded at room temperature prior to testing as shown in Table 6.

Table 6 Wide plate tests on warm pre-stressed (WPS) pressure vessel steels

Analysis of the parent metal wide plate specimens has been carried out in accordance with BS 7910, with toughness given in terms of K_mat calculated from J. Results are presented in Figure 9. The effects of warm pre-stressing (WPS) can be seen qualitatively by comparing ID 803 (no WPS) with 804 (WPS), and ID 805 (no WPS) with 806 (WPS). Points associated with WPS specimens lie further from the failure line, suggesting a WPS effect (although it is marginal in the case of ID 803/804).

The analysis of the welded specimens takes the effects of WPS into account by using the measured values of residual stresses (a centre hole strain gauge rosette technique was used for surface readings, and block removal, splitting and layering for through-thickness measurements).

Note that the equibiaxial test at -70°C (ID 809) led to failure under a membrane stress of 136N/mm². A similar plate was warm prestressed and then re-tested at progressively lower temperatures. The first test (ID 810) did not fail at the load limit of the test machine at -70°C (equivalent to a membrane stress of 345N/mm²), nor again at -90°C (ID 811, P_m=345N/mm²). Failure finally occurred in the third test at -120°C (ID 812, P_m=351N/mm²), indicating a strong effect of WPS.

This series of tests provides an interesting validation case for examining the effects of proof loading and warm pre-stress, of particular interest to the pressure vessel industry. The analysis shown above is based on the main clauses of BS 7910. Re-analysis in terms of Annex O of BS7910 and/or FITNET Section 11 is therefore recommended.

Concluding comments

The tests described in this paper cover only a small percentage of the 300 test cases analysed by TWI, which in turn constitute only a part of the validation package (section 13/Volume 3 of FITNET). Nevertheless, the 64 test results presented above provide some flavour of the overall dataset, a summary of which is shown in Figure 10. All points represent failure of a test specimen or component; when conservative but representative input data were used, as recommended by BS7910 and FITNET, the analysis points fall outside the failure analysis line, confirming the validity of the fracture assessment procedure.

Acknowledgments

This work was jointly funded by the Industrial Members of TWI (as part of the Core Research Programme) and the European Commission's FITNET Thematic Network (5th Framework Programme, Contract No. G1RT-CT-2001-05071). Thanks are also due to Dr Henryk Pisarski for his valuable comments.

References

1. PD6493:1991: 'Guidance on methods for assessing the acceptability of flaws in fusion welded structures', British Standards Institution, London, 1991.
2. BS 7910:2005: 'Guide to methods for assessing the acceptability of flaws in metallic structures', British Standards Institution, 2005.
3. Challenger N V, Phaal R and Garwood S J: 'Appraisal of PD 6493:1991 Fracture assessment procedures: Part I: TWI data, Part II: Published and additional TWI data, and Part III: Assessment of actual failures.' TWI Report 512/1995, June 1995.
4. Hadley, I and Moore P: 'Fracture case studies for the validation of fitness-for-service procedures', TWI members' report 850/2006
5. Wiesner C S, Maddox S J, Xu W, Webster G A, Burdekin F M, Andrews R M and Harrison J D: 'Engineering critical analyses to BS 7910 - The UK guide on methods for assessing the acceptability of flaws in metallic structures', International Journal of Pressure Vessels and Piping 77 (2000) 833-893.
6. Phaal R, Andrews RM and Garwood S J: 'TWI biaxial test program, 1984 to 1994', presented at a seminar 'Constraint and the definition of upper shelf behaviour' TWI, 12 April 1994, published in International Journal of Pressure Vessels and Piping, 64 (1995), iii-iv, pp177-190
7. Andrews RM and Garwood SJ: 'An analysis of fracture under biaxial loading using the non-singular T-stress', ASTM Symposium Forth Worth, Dallas, USA, 17-18 Nov 1993
8. SINTAP: 'Structural integrity procedures for European Industry', final procedure, Nov 1999
9. Hadley I: 'Effect of crack tip constraint on fracture toughness of A533B steel and validation of the SINTAP constraint procedure' (TWI project 88269), Report SINTAP/TWI/012, 1999.
10. Ainsworth R A, Sattari-Far I, Sherry A H and Hadley I: 'Development of methods for including constraint effects within the SINTAP procedures', ECF12, 14-18 September, 1998, Sheffield, UK.
11. Towers O L, Williams S and Harrison J D: 'ECSC Collaborative elastic-plastic fracture toughness testing and assessment methods.' TWI Contract Report 3571/10M/84, June 1984.
12. Muhammed A, Pisarski H G and Stacey A: 'Using wide plate test results to improve predictions from probabilistic fracture mechanics'. 13th European Conference on Fracture, ECF 13, 6-9 September 2000, San Sebastian, Spain.
13. Ogle M H, Burdekin F M and Hadley I: 'Material selection requirements for civil structures', 56th Annual Assembly of the 56th Annual Assembly of the IIW, 6-11 July 2003, Bucharest, Romania.
14. Bell K and Garwood S J: 'Influence of proof loading on component reliability at low temperatures.' TWI Group Sponsored Project Report 5561/22/92, January 1992.