Accepted Manuscript Materials properties for engineering critical assessment: Background to the advice given in BS 7910:2013 Isabel Hadley, Henryk Pisarski PII:
S03080161(18)301741
DOI:
https://doi.org/10.1016/j.ijpvp.2018.10.016
Reference:
IPVP 3780
To appear in:
International Journal of Pressure Vessels and Piping
Received Date: 21 May 2018 Revised Date:
12 October 2018
Accepted Date: 19 October 2018
Please cite this article as: Hadley I, Pisarski H, Materials properties for engineering critical assessment: Background to the advice given in BS 7910:2013, International Journal of Pressure Vessels and Piping (2018), doi: https://doi.org/10.1016/j.ijpvp.2018.10.016. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Materials properties for Engineering Critical Assessment: background to the advice given in BS 7910:2013 Isabel Hadley* and Henryk Pisarski** * Technology Fellow, TWI Ltd, Great Abington, Cambridge, UK, CB21 6AL (
[email protected]) ** Independent consultant
RI PT
ABSTRACT The UK procedure for fracture mechanicsbased assessment of flaws (BS 7910) was extensively updated in late 2013. This paper outlines how and why the clauses relating to materials properties (in particular those relating to brittle and ductile fracture) have been expanded, and explains the origin and extent of validation of the information. Some worked examples of the treatment of fracture toughness are given, based on results from weld metal tests.
SC
INTRODUCTION
M AN U
BS 7910 ‘Guide to methods for assessing the acceptability of flaws in metallic structures’ (1) is a procedure for the assessment of flaws in metallic structures, covering failure by fracture, fatigue, corrosion and creep. It is used to make decisions on serviceability of safetycritical plant and structures, but also in failure investigation, life extension and design. It has been in continuous use since 1999, when it superseded the earlier UK flaw assessment procedure PD 6493. The most recent edition of the procedure was published in 2013 and amended in 2015. It evolved (see Figure 1) from a mixture of sources, including previous editions of BS 7910/PD 6493, the UK nuclear safety assessment procedure R6 (2) and the European SINTAP and FITNET documents (note that the two European procedures (3) are no longer maintained).
TE D
One of the features of BS 7910:2013 is a significant expansion of the clauses on materials properties, particularly those relating to brittle and ductile fracture of ferritic steels. Whilst several publications have already addressed the background to various aspects of the procedure ((4)(6)), the information currently in the public domain relating to materials properties is rather limited. The purpose of this paper is therefore to document the origin and extent of validation of the current clauses relating to materials properties used for fracture assessment, and to identify gaps where future work may be required.
EP
MATERIALS PROPERTY INFORMATION IN BS 7910:2013
AC C
Materials properties information is vital to all three elements of the BS 7910 Failure Assessment Diagram (FAD): the Lr axis (showing proximity to failure by plastic collapse), the Kr axis (representing the proximity to failure by fracture) and the Failure Assessment Line or FAL (showing the relationship between the two). However, it must be recognised that the amount and type of materials property information available to the analyst varies enormously depending on the objective of the analysis. In some cases (for example, assessing the fitnessforservice of a newlyfabricated asset which has been found not to comply in some way with the intended build quality) the analyst will have access to detailed materials certificates and weld procedure qualification records, along with inspection reports and detailed stress analysis, and will be in a position to carry out a ‘known flaw assessment’ as per BS 7910. In general, the style of writing of BS 7910 assumes this is the case, ie that a single known flaw is located and sized, an allowance is made for sizing error, and that it is then assessed on the basis of lowerbound values of materials properties (fracture toughness and tensile properties) and upperbound values of applied and residual stress. At the opposite end of the spectrum, there may be a need to analyse ageing equipment or a new design concept for which little or no materials data exist. In the case of ageing equipment, original construction records may have been lost, and extraction of material from the structure itself for testing may be impossible. Sitebased techniques such as metallographic replication, portable hardness testing and spectrographic analysis may need to be used, combined with use of archive information, destructive testing of a ‘twin’ structure, or manufacture of a mockup, depending on the nature of the problem. Whilst in the case of new design only basic information such as defined by standards or 1
ACCEPTED MANUSCRIPT databases will be available and an iterative method is used to either define flaw acceptance criteria or fracture toughness requirements. The materials property clauses of BS 7910:2013 have therefore been written to cover a range of scenarios; typically, however, the user will reap increased benefit (in terms of larger tolerable flaw size or lower fracture toughness requirement) by carrying out more (and more complex) tests.
RI PT
As a standard for assessment rather than testing, BS 7910 does not include test methods within the document – rather, the materials properties clauses refer out to current testing standards, and provide information (for example, number of tests, areas to be sampled) that are not available in (or not appropriate to) testing standards.
tensile properties (clause 7.1.3), fracture toughness in general (7.1.4 and 7.1.5), fracture toughness of weldments (Annex L), Charpy/fracture toughness correlations (Annex J), Significance of strength mismatch (Annex I), Probabilistic assessment (Annex K).
These will be discussed in turn below.
M AN U
• • • • • •
SC
The development and maintenance of BS 7910 is managed via the BSI WEE/37 committee and 11 associated subcommittees or ‘panels’ (see Figure 2), each of which assumes responsibility for the drafting of particular clauses or annexes, with the main committee providing coordination and oversight. The main clauses concerned with materials properties are:
Treatment of tensile properties (clause 7.1.3)
• •
within a batch of material, due to normal statistical variation, within a given weldment (typically, the intention is that weld metal strength should exceed that of the parent material, ie should ‘overmatch’), as a function of temperature (but properties are typically determined at room temperature only), as a function of the test method employed and the parameter calculated (for example, upper/ lower yield strength, offset yield strength).
EP
• •
TE D
Clause 7 of BS 7910 (‘Assessment for fracture resistance’) is the main source of information on materials properties, of which the most important are tensile properties and fracture toughness. The clause addressing tensile properties in BS 7910 has expanded from one short paragraph (in the 2005 edition) to several pages of advice in the 2013 edition, reflecting an increased understanding of the importance of tensile properties in fracture assessment. First, it should be recognised that tensile properties (in particular, the yield strength and UTS) may vary:
AC C
Most fracture assessments are based on the tensile properties of a single homogeneous material, typically the parent metal (assuming weld metal overmatching). An important distinction is drawn in BS 7910:2013 between steels that yield in a continuous manner (stress rises monotonically as a function of strain) and those showing discontinuous yielding (either a yield plateau or, in some cases, the occurrence of an upper yield followed by a yield drop). Examples of these behaviours are shown in Figure 3 and Figure 4. The relationship between load and cracktip Jintegral (and hence the form of the FAL, shown in equation [1]) depends critically on this behaviour: =
+
.
for
=
[1]
,
where is the true strain at the true stress = , E is Young’s modulus and is the yield strength (usually defined, for continuously yielding materials, as the stress corresponding to a 0.2% offset to the linear portion of the curve). Equation [1] describes the socalled Option 2 FAL, which requires the user to have a full stressstrain curve for the material (typically the parent material, assuming weld strength overmatching). There are, however, numerous occasions on which this information will be unavailable; for example, in the early stages of design (material grade may be known, but not the actual source and detailed properties of the material) and in the analysis of ageing structures. For these scenarios, BS 7910 provides a so2
ACCEPTED MANUSCRIPT
RI PT
called Option 1 FAL (Figure 5), which is further subdivided depending on whether the material shows (or is expected to show) continuous or discontinuous yielding. Naturally, this raises the question ‘how do I know what the yield behaviour is like if I don’t have a full stressstrain curve?’ Table 4 of BS 7910 gives some guidance on this issue for a range of structural steels, depending on their grade, composition and processing route. This is not exhaustive, however, and an additional recommendation is to carry out a sensitivity assessment, ie to change the assumption about yield behaviour and see how much the result changes. It should be noted that the main purpose of Table 4 is to provide guidance on steels that are likely to show a yield discontinuity when the only tensile information is that given by the steel specification standard. In European standards yield strength is defined as the upper yield (ReL in Fig. 3). However, for the purpose of ECA, yield strength is defined as 0.95ReL. From Figure 5, it is apparent that the yield behaviour is relatively unimportant in the fracturecontrolled part of the FAD (‘Zone 1’), whereas it could affect results significantly in the ‘knee’ (Zone 2) and collapsecontrolled (Zone 3) regions.
SC
For cases when even the basic tensile properties (yield strength and UTS) are not available, BS 7910 includes information on the relationship between hardness and tensile properties (both σY and σUTS) of CMn steels and weldments. These equations are identical to those in the fracture mechanics test standard for weldments, BS EN ISO 15653.
M AN U
Statistical variation in tensile properties is cited in clause 7.1.3.2; the recommended value of lower bound tensile properties for use in assessments is two standard deviations below the mean (a logical cutoff lower bound value would be the specified minimum yield (or ultimate) strength, if this proves to be higher). However, mean values of tensile properties are recommended for the definition of the FAL and of the cutoff on the Lr axis, Lr,max. Additional newlyincluded information on tensile properties is summarised in Table 1, along with a record of its origin. Table 1 Source of information on tensile properties of steels in BS 7910:2013
Table 3
Source
Comment
Statistical variation in tensile properties
FITNET project and panel discussion
Elastic modulus
Literature review
BS EN ISO 15653
Broadly similar to Annex F of API 5791/ASME FFS1 (7), although derived and expressed in different ways Temperaturedependence of E for both ferritic and austenitic steels is given, in the temperature range 200 o to 650 C (see Figure 6). Cites a paper by Irwin
Committee discussions
Informal information from UK steel industry
DNVOSF101 (8)
Covers CMn and duplex stainless steels at o temperatures up to 200 C
Modification of the RambergOsgood equation
Covers CMn steels and a lower bound estimate of strain at the tensile strength
SINTAP/FITNET
See (9)
TE D
7.1.3.3
Content
EP
7.1.3.2
Other ref. (in BS 7910) Table 2
AC C
Clause no.
7.1.3.4
Equation (2)
7.1.3.4
Equation (3)
7.1.3.4
Figure 13
7.1.3.5
Equations (4)(6)
7.1.3.5
Equation
Temperaturedependence of σY below room temperature Temperaturedependence of σUTS below room temperature Temperaturedependence of σY and σUTS above room temperature Idealised stressstrain curve and estimation of strainhardening exponent, n Estimation of strain
3
ACCEPTED MANUSCRIPT Clause no.
Other ref. (in BS 7910) (7) Table 4
Content
Source
Comment
hardening exponent, n projects Guidance on distinction SINTAP/FITNET See (9) between continuously projects and discontinuously yield behaviour 7.1.3.7 Equations Tensile properties from BS EN ISO 15653 (9)(12) hardness 7.1.3.8 Equations Derivation of true Standard (13)(14) stress/strain from engineering texts engineering stress/strain Note: information in italics refers to the equation/Table number in BS 7910, not in this paper
RI PT
7.1.3.6
M AN U
SC
The modulus of elasticity is required to define the FAL (equation (1)) and in the conversion of CTOD and J values into a common fracture toughness value defined as Kmat. It can be difficult to measure, especially at subzero and elevated temperatures. Recognising this problem, a literature review was undertaken to establish the variation in the modulus of elasticity with temperature for ferritic and austenitic steels. The best fit curves to these data were used to generate Table 3 of the standard. The data obtained for ferritic steels is shown in Figure 6. Treatment of fracture toughness (clause 7.1.4)
The clauses relating to fracture toughness (7.1.4 and 7.1.5) have, like those addressing tensile properties, been extensively revised and expanded in the 2013 edition of BS 7910. Attention is drawn to the importance of:
• •
TE D
• • •
testing materials to a recognised standard (eg (10)(12)), ensuring that the crack tip is located in the required microstructure, testing in sets of at least three similar specimens, and ensuring that additional testing is carried out in cases where there is excessive scatter and/or different types of specimen behaviour, replicating in the test specimen the crack orientation observed (or expected) in the structure, matching the strain rate during testing to that of the structure, testing at the appropriate temperature (especially for ferritic steels, because of the ductile/brittle transition), matching specimen thickness to structural thickness (where possible), considering possible environmental effects on fracture toughness (for example, due to the presence of hydrogen).
EP
• • •
AC C
The main change relative to the 2005 edition is the introduction of a single fracture toughness parameter Kmat, which can be derived from fracture toughness tests using any of the standard test parameters: KIc, Jintegral or CTOD. As outlined in an earlier publication (4), previous editions of BS 7910 and PD6493 presented the user with the option of calculating crack driving force in terms of either elastic stress intensity (KI) or CTOD driving force (δI), and elasticplastic materials toughness in terms of either critical Jintegral (Jmat) or critical CTOD (δmat). The two assessment routes reflect the history of the procedure, which can be traced mainly to the nuclear industry (the R6 procedure has long been based on determination of toughness in terms of J) and the offshore industry (which used CTOD to determine toughness, and PD 6493 to analyse fitnessforservice). The calculation of KI was based on LEFM, whilst that of δI was derived from KI via the partlyempirical CTOD design curve. Experience over the years has shown that the margins against failure are in general higher when CTOD is used as the parameter for both driving force and materials toughness. The reason for this would appear to lie in the definition of δI in BS 7910:2005: !" =
#$ %
[2]
&
where X is (according to BS 7910:2005) ‘a factor (generally of value between 1 and 2) influenced by crack tip constraint and geometric constraint and the work hardening capacity of the material’. 4
ACCEPTED MANUSCRIPT Unfortunately, BS 7910:2005 does not give clear advice on how to determine X other than by conducting numerical analysis of the structural component to derive applied values of KI and δI (which is somewhat inconsistent with the notion of a standard based entirely on the use of parametric equations). It implies (based on work carried out under the SINTAP and FITNET projects) that the same value X can also be derived from the materials fracture toughness, calculating it directly from a comparison of Jmat and δmat, ie: '=
()*+ ,)*+  . %
[3]
RI PT
where ν is Poisson’s ratio.
In practice, a default value X=1 in equation [2] was recommended in BS 7910:2005 for cases in which the user had only CTOD test data available, leading to the situation in which the same structure could be assessed in terms of Jintegral and CTOD (derived from the same test specimens), but different (higher) safety margins would be calculated for calculations based on CTOD.
()*+
M AN U
/( = 0
SC
In the 2013 revision of BS 7910, the potential contradiction between K/Jbased and CTODbased calculations is resolved by calculating crack driving force in terms of KI only, and materials resistance in terms of Kmat, which can be derived from KIc, J (KJ) or CTOD (KCTOD) as appropriate. KJ is then calculated directly from J as:  1%
and from CTOD as: /2345 = 0
,)*+
 1%
[4]
[5]
7 = 1.517 ;
<=>
?
[email protected]
TE D
The coefficient m can then be related to the tensile properties, but essentially strain hardening capacity, of the material (at the same temperature as that of the fracture toughness test) as follows: [6]
AC C
EP
where m=1.5 is used if equation [6] can not be applied. Equation [6] was derived from fracture toughness tests conducted by TWI using deeply notched bend specimens and is applicable to a wide range of steels with yield/tensile strength ratios in the range 0.3<(σY/σU)<0.98, and has been validated against 665 sets of data, as shown in Figure 7. It should be noted that use of equation [6] is intended principally for cases where only CTOD data are available, for example analyses based on historical data; it is envisaged that the Jintegral (or KIc, if appropriate) should be determined directly in any future tests carried out as part of an analysis to BS 7910 (it should also be noted that testing procedures to derive J and CTOD are essentially identical and that only the calculation methods differ). In removing one anomaly (the different safety margins associated with Jbased and CTODbased analyses), have we introduced another one, ie the ‘demotion’ of the concept of CTOD, which is now firmly embedded in the oil & gas and civil construction industries in particular? This was certainly not the intention. Testing standards such as BS EN 15653, BS ISO 12135 (12) and ASTM E1820 recognise methods for determining both CTOD and Jintegral from a single specimen and are likely to remain ‘bilingual’, both because of their familiarity to users and because several offshorerelated codes and standards (eg API 2Z (13), BS EN 10225 (14)) cite CTOD (only). Moreover, CTOD is considered to be a more stable parameter in the context of the important, and growing, field of strainbased assessment (not currently covered by BS 7910, but due for inclusion in future revision). Statistical treatment of fracture toughness (clause 7.1.5) Use of the minimum of three fracture toughness results in BS 7910 analyses (clause 7.1.4.10) as a definition of characteristic lower bound fracture toughness has been historically justified by reference to longterm user experience and validation exercises; it is equivalent to defining the median th toughness with 87.5% confidence, or the 20 percentile with 50% confidence. 5
ACCEPTED MANUSCRIPT
SC
RI PT
If three tests are carried out and the scatter in results is high (for example, an individual value of Kmat falls below 70% of the mean or above 140% of the mean of three results), additional tests are required, aimed at producing similar confidence to that implied by the use of the minimum of three specimens. Clause 7.1.5.2 addresses the situation in which the user has carried out more than three (but typically fewer than 15) fracture mechanics tests, perhaps because of excess scatter in the results, and wishes to define a characteristic lowerbound toughness, termed the Minimum of Three Equivalent (MOTE). Table 5 of BS 7910 defines the MOTE as the secondlowest value of m results if 6≤m≤10, and the thirdlowest if 11≤m≤15 (the same information was present in BS PD6493:1991 and in the 2005 edition of BS 7910, but it has now been moved from an Annex to the main fracture clauses). The numbers are based on consideration of the binomial distribution but modified by analyses of CTOD and wide plate tests conducted on BS 4360 Grade 50D steel plate. Jutla and Garwood (15) were able to show that using the CTOD design curve, the mean factor of safety on flaw size ranged from 2.3 to 3 for up to 15 results and then decreased thereafter. It is unclear how the safety factor would change if the analyses were repeated using Kmat, instead of CTODmat, data and the current Option 2 FAD instead of the CTOD design curve. Since the current assessment procedures are considered for accurate and lower inherent safety factors, it is suspected that the safety factor would be smaller. th
M AN U
It can be argued that using the 20 percentile with only 50% confidence is a rather ‘lenient’ definition of lower bound fracture toughness compared with, say, the equivalent limits stipulated for tensile properties, especially when viewed against the potentially very severe consequences of brittle fracture. It may be that the historical acceptability of the minimum of three results is linked to the fact that use of a standard deeplynotched (highconstraint) fracture specimen tends to include a safety margin due to cracktip constraint when applied to real welded structures. This will not generally be the case, however (particularly when methods such as Annex N are used to allow for constraint differences between specimen and structure), so BS 7910 also gives other, more rigorous, approaches for determining lower bound fracture toughness.
C
B = /
C
− DF
.G
EP
/
TE D
Clause 7.1.5.3 presents a statistical method of treating fracture toughness data in order to define a lower bound estimate of Kmat with high confidence (as stated above, the MOTE approach gives a high th (87.5%) confidence in defining the median Kmat, but only a 50% confidence in defining the lower 20 percentile). The clause can be used for datasets where 3≤m≤20. It assumes Kmat to be normally or B C ) and standard deviation (D) from the data and lognormally distributed: the user calculates mean / uses standard statistical tables (included in Table 6 of BS 7910) to calculate the lower bound fracture th toughness (here defined as the lower 20 percentile) at 90% confidence. [7]
where F .G is given as a function of m in Table 6 of BS 7910:2013 and represents the lower 20 percentile for a onesided tolerance limit in a normal distribution.
th
AC C
Analysis of fracture toughness and wide plate data on weld HAZs conducted by Pisarski and Jutla th (16) provided support for the use of the 20 percentile (with 90% confidence) from a lognormal distribution. A plastic collapse modified CTOD strip yield model (similar to the level 2 FAD in BS PD6493:1991) was employed. This showed that the factor of safety on flaw size was approximately 2 when there were results from at least 12 specimens nominally notched into the HAZ. When the specimens were subjected to posttest metallography, to confirm that the most brittle regions of the HAZ were tested (GCHAZ and ICGCHAZ) and at least 9 test results were available then the median th or 50 percentile value (again with 90% confidence) provided the same factor of safety on crack size. The FITNET procedure suggests that when safety critical structures are being assessed and failure th could result in loss of life then a lower percentile should be employed, say the 5 percentile. This is still a subject of discussion. Where safety critical structures are being assessed consideration should be given to the use of probabilistic rather than deterministic analyses or, as a first step, use of partial safety factors as described in Annex K to achieve the desired target reliability. The choice of fracture toughness value to use in an assessment is a particular problem that bedevils deterministic analyses. The usual approach is to “calibrate” the choice using wide plate data, as described above. However, this not completely satisfactory as it is difficult (and expensive) to have a sufficient number of wide plate tests to make a statistically homogeneous data set. Furthermore, in the past insufficient recognition and allowance was made for the differences in cracktip constraint 6
ACCEPTED MANUSCRIPT between the fracture toughness specimen, usually representing high constraint, and the lower constraint wide plate test. This would account for some of the apparent safety factor and blurred the choice of the appropriate fracture toughness value. An advantage of a probabilistic approach is that there is no need for a single value of fracture toughness; the whole fracture toughness distribution is used to estimate the probability of failure. This aspect is dealt with in Annex K. The Master Curve concept (Clause 7 and Annex L)
/
C
M AN U
SC
RI PT
Ideally, fracture toughness tests are carried out on specimens that match the structure being analysed in terms of section thickness, temperature and strain rate. However, it is recognised that there are situations in which it may be necessary to use subsize specimens, or to use data obtained at some temperature other than the minimum design temperature of the structure. The concept of the Master Curve (referred to in various sections of BS 7910, including parts of clause 7, and Annexes J, K and L) is central to the treatment of fracture toughness in BS 7910. It assumes that the fracture toughness of ferritic steel in the lower transition region can be expressed in terms of a single indexing parameter, T0, which represents the temperature at which the median fracture toughness of a 25mm thick specimen is 100MPa√m. Fracture toughness is assumed to follow a weak link, three parameter Weibull distribution in which shape and shift parameters are fixed taking the values of 0.25 (exponent) and 20MPa√m, respectively. The lower part transition curve is defined by an exponential curve which is independent of the material (hence the Master Curve). Different materials are characterised by a shift in the curve which is defined by T0. It was developed for ferritic and bainitic steels and the history of its development is described by Wallin (17). The concept has been extensively used by the nuclear pressure vessel industry. The influence of specimen thickness, temperature (when reasonably close to T0) and the required degree of confidence in the result can all be expressed through the use of a single equation: = 20 + J11 + 77expN0.019 P − PQ RS T V U
o
with units of MPa√m, mm and C.
.
WXY Z

 [
\]
.
[8]
TE D
Here, T and B refer to the temperature and section thickness of interest, and Pf is the probability of Kmat being lower than estimated from equation [8]. Hence, for T=T0, B=25mm and Pf=0.5, Kmat=100MPa√m, whereas at the lower end of the transition curve and for low values of Pf, Kmat approaches a lower bound of 20MPa√m. Figure 8 shows examples, with section thicknesses of 10, 25 and 100mm and Pf values of 0.05 and 0.5 (the former is typically used for assessment purposes, whilst the latter could be useful in design/failure analysis studies).
AC C
EP
The Master Curve model is implicit in clauses 7.1.4.77.1.4.9 of BS 7910, which address treatment of of subsized specimens, and includes equations to ensure that fracture occurs under smallscale yielding conditions. It is also referenced in clause 7.1.5.6, which refers to the more advanced methods of analysis given by Annex L. Advanced treatment for statistical analysis of fracture toughness (Annex L) Annex L of BS 7910 (‘Fracture toughness determination for welds’) contains general advice on test philosophy and procedures, plus a more advanced treatment for the statistical analysis of fracture toughness data (clause L.9). Two scenarios are envisaged: for homogeneous materials such as parent metals and weldments in the PWHT condition, the Master Curve method as described in ASTM E1921 (18) is appropriate, whereas for inhomogeneous materials such as weldments in the aswelded condition, the MML (maximum likelihood) procedure is adopted, taken from the SINTAP and FITNET procedures. The steps in the procedure, described in detail in clause L.9.3, are as follows: • •
Censor the dataset to ensure that individual values are below a limiting value KJc,lim which ensures smallscale yielding in the data fitted (those which do not qualify are assigned a dummy value of fracture toughness). If the specimen thickness is 25mm, use data as they are. 7
ACCEPTED MANUSCRIPT • • • • •
RI PT
•
If not (ie if less than or greater than 25mm thick), correct each value of fracture toughness to what it would be in the equivalent 25mm specimen. Estimate a trial value of T0, ie the temperature at which the median fracture toughness of a 25mm thick specimen is 100MPa√m. Iterate on the value of T0 to minimise the maximum likelihood function. From T0, and the actual section thickness of the specimens, estimate a value of fracture toughness from equation [8] at the desired level of confidence. If the dataset is homogeneous, the analysis is complete, and is equivalent to the method given in ASTM E1921. If not, the user proceeds to Stage 2 (and if necessary, Stage 3) to estimate the properties of the lower tail of an inhomogeneous distribution.
Annex L also describes two further aspects of the treatment of fracture toughness data: the application of the value of Kmat (whether calculated from Annex L or by some other means) to structures containing long flaws (clause L.9.5) and the effects of strain rate (clause L.9.6).
/
C ^
= 20 + _/
C
− 20` T V ^
.
M AN U
SC
The implication of clause L.9.5 (‘Value of Kmat for use in integrity assessments’) is that, since the Master Curve method is based on weak link statistics, the fracture toughness to be used in an integrity assessment (Kmat(l)) should reflect the length of the crack in the structure, not only its length in the test specimen (where it is typically limited to the specimen thickness, B). A correction to Kmat is given in terms of the value of fracture toughness for a 25mm thick specimen (Kmat(25)) and the crack front length, l: [9]
TE D
Note that l represents the length of the crack front over which the stress intensity is constant, so is not exactly the same as either the crack length, 2c, or the total length of the crack front. A limit of l=2B in equation [9] is stipulated for very long cracks. Clearly, this clause is out of step with the remainder of the standard, in which it is assumed that matching the specimen thickness with the structural thickness will be sufficient to determine an appropriate level of fracture toughness. Use of Clause L.9.5 is thus potentially problematic for calculation of (for example) maximum tolerable flaw size; fracture toughness is assumed to be a function of flaw length and thus an iterative calculation is required.
Annex J
EP
Clause L.9.6 gives recommendations for adjusting the Master Curve to take account of strain rate effects in the ductilebrittle transition. A shift, ∆T0, in the value of T0 as determined from static tests, is proposed as a function of Krate (rate of increase of stress intensity, /a ) and the (static) yield strength of the material.
AC C
Annex J (‘Use of Charpy Vnotch impact tests to estimate fracture toughness’) describes a number of approaches to estimating a lowerbound value of Kmat for ferritic steels, depending on whether the Charpy tests were carried out on the lower shelf, in the lower transition, or on the upper shelf, and the temperature for which Kmat is required. The use of subsize Charpy data (ie specimens with thickness <10mm) is also described and a flowchart included to guide the firsttime user in particular. The equations are largely taken from work undertaken in the SINTAP and FITNET projects (19). Lower shelf/lower transition behaviour is covered by clause J.2.1, which is intended for the situation in which the user has standard (10x10mm Vnotch) Charpy energy results available at a temperature T and wishes to estimate a value of Kmat at the same temperature. A simple equation originating from the Swedish INSTA procedure (20) is recommended:
/
C
= b_12cde − 20 ` Z
25 \ f
.
[10]
g + 20
where Cv represents the Charpy Vnotch energy (in J) for a 10mm thick specimen. Other units are mm and MPa√m . 8
ACCEPTED MANUSCRIPT An alternative to the use of equation [10] is to combine the Master Curve concept described earlier with an empirical correlation between T0 and the Charpy transition temperature (typically the temperature required to obtain 27J energy absorption for lowerstrength, or 40J for higherstrength, steels) to derive a method for estimating Kmat from T27J or T40J as follows: P = P h( − 18Q d or P = Pj
(
− 24Q d
[11] o
/
C
= 20 + J11 + 77expN0.019 P − PQ − P# RS T V U
.
WXY Z

 [
\]
.
RI PT
Then, allowing for some scatter (a standard deviation of ±15 C) in the relationship between T0 and the Charpy transition temperature), equation [8] can be rewritten as: [12]
o
SC
where TK is a temperature term to take account of the scatter in equation [11]. BS 7910 recommends o the value TK=+25 C unless sufficient information is available to justify a lower value. In addition, Pf, the probability of Kmat being less than estimated, is recommended to be 0.05, in other words there is a 95% probability that the Kmat will be higher than estimated. Use of a higher value of Pf is permitted if experimental data supports it, for example by conducting calibration tests. o
M AN U
So, for a material with T27J=32 C (T0=50 C, based on equation [11]), a lower bound value of Kmat=48MPa√m would be calculated from equation [12] for Pf=0.05, even though the median fracture toughness is 100MPa√m. The user therefore pays a heavy penalty for deriving Kmat from Charpy energy, primarily because of the inevitable scatter in materials properties.
TE D
Equations [10] and [12] are intended to cover the same region of the ductilebrittle transition curve, ie the lower shelf/lower transition region, but the latter is clearly more versatile in terms of its ability to take into account the fact that Charpy test temperature may well be different from (typically lower than) the minimum design temperature, and its inclusion of a confidence factor Pf. According to BS 7910, equation [10] tends to give lower estimates of Kmat than does [12] (at Pf=0.05) below the Charpy transition temperature (T27J or T40J).
/
C
= 0.54de + 55
EP
Although they are cited as being appropriate to the lower shelf/lower transition region, neither [10] nor [12] have explicit validity limits; as seen from Figure 8 and equation [8], the Master Curve model assumes an exponential function, so if T>>T0, there is no upper limit to Kmat. In practice, of course, Kmat is bounded by the toughness observed in the upper shelf region, which is addressed in clauses J.2.4 and J.2.5. Essentially, J.2.4 gives an upper limit to the value of Kmat that can be assumed, based on the Charpy energy at the temperature of interest: [13]
AC C
with units of J and MPa√m.
Equation [13] was derived for ferritic steels showing upper shelf behaviour in the Charpy specimens, with yield strength up to 480MPa and with relatively poor resistance to ductile tearing, such as might be associated with steel with high levels of sulphur and carbon. The derivation of the equation is not known but it has been in use since the publication of BS PD 6493 since 1980. Checks against experimental data confirm that it does indeed provide a safe lower bound estimate of fracture toughness for older steels, hence its retention in BS7910:2013. An alternative equation (introduced to BS 7910 for the first time in the 2013 revision) is given for the initiation toughness associated with cleaner, more modern, steels. Equation [14] is included in the FITNET procedure and was incorporated into BS 7910 because Equation [13] was too conservative for modern pressure vessel and structural steels. /
C
t.%uv
. A l 0.53dnop 0.2
[email protected]@2qrs = 1000 1 − w
[14]
where dnop is the upper shelf Charpy energy (in J), E and w are the elastic constants and Kmat is in MPa√m. 9
ACCEPTED MANUSCRIPT
RI PT
Annex J is perhaps the most commonly misused part of BS 7910; users speak of ‘converting’ their Charpy data to fracture toughness data, or reverseengineer the equations to calculate Charpy requirements without appreciating the uncertainties and safety margins involved in the process, potentially resulting in unachievable specifications. The first key point to note is that BS 7910, as written, is intended to allow estimation of a lower bound value of Kmat from Charpy energy, rather than viceversa. The second is that the methods assume that the microstructure sampled by the Charpy specimen is the same as the one that would have been sampled had a fracture toughness test been carried out. However, it is wellknown that a fracture toughness specimen, with its very fine fatigue crack tip, is capable of sampling a very specific region of a weldment (such as the graincoarsened area of a HAZ), and can therefore be used to identify (and characterise the toughness of) socalled local brittle zones (LBZs). The relatively blunt notch in the Charpy specimen will sample a larger area and may therefore be insensitive to LBZs. Third, the different crack tip/notch tip strain rates associated with the two different types of test may result in anomalies in attempting to correlate them with each other. Annex I
M AN U
SC
Annex I (‘Significance of strength mismatch on the fracture behaviour of welded joints’) gives advice on construction of the FAL for cases in which the weld metal overmatches (or undermatches) the parent metal in terms of strength. Since both the parent material and the weld metal can show either continuous or discontinuous yielding, there are four possible scenarios to take account of in constructing a FAL even at the simplest level, ie Option 1: both components continuous, both discontinuous, or one of each. Advice is also given on constructing an Option 2 FAL when full stressstrain curves are available for both weld metal and parent metal. Annex I is used in conjunction with Annex P, which gives the limit loads (and hence Lr) for strengthmismatched joints. Analysis based on strength mismatch is currently not widely used, in part because of the relatively small number of geometries for which a limit load for mismatched material is available. Annex K
EP
FUTURE DEVELOPMENTS
TE D
Clause K.6.3 of BS 7910 also addresses certain aspects of the statistical analysis of fracture toughness data, emphasising the high uncertainty that can be associated with fracture toughness. Advice is given on the likely distribution of fracture toughness results (normal, lognormal or Weibull), but the detailed treatment of fracture toughness is given elsewhere (clause 7 and Annex L) as described earlier.
AC C
As described above, substantial changes to the materials properties clauses of BS 7910:2005 were made in order to produce the current BS 7910:2013. The changes proposed for the next revision of BS 7910 (scheduled for late 2019) will include corrections and clarifications, in particular with respect to the definition of tensile properties and the treatment of fracture toughness data. It has become apparent that, when the more advanced features of BS 7910 (such as implementation of Annex N ‘Allowance for constraint effects’) are used, a more rigorous definition of lower bound fracture toughness than MOTE (Minimum of three Equivalent) may be needed; for cases such as these, the statistical analysis methods given in Annex L may be appropriate. The case study presented below has also highlighted potential discrepancies between the different ways of defining lower bound fracture toughness in brittle materials. Finally, the title of Annex L (‘Fracture toughness determination for welds’) may need to change to reflect its general applicability; ‘Determination of fracture toughness and statistical treatment of fracture toughness data’ might be a more accurate description. CASE STUDY An example of application of the method given in clause 7.1.5.3 of BS 7910 is shown in Figure 9. It shows Kmat, based on a sample of 20 weld metal fracture toughness results from a European round robin exercise (21). The results had been collated from an original batch of 59 CTOD test results, by censoring any that did not meet all of the qualification requirements of the testing standard (an early draft of what is now BS EN ISO 15653) and calculating CTOD on the basis of a common value of o yield strength. Tests had been carried out at 60 C, in the lower transition region of the ductilebrittle transition curve. Although the resulting dataset was not perfectly statistically homogeneous (some residual interlaboratory variation was noted), it is sufficient to illustrate the method. For the purposes 10
ACCEPTED MANUSCRIPT of this study, the results in terms of CTOD were converted into individual Kmat results, using equation [5]. Statistical analysis showed that the best fit was obtained assuming a lognormal distribution (although other distributions such as a normal distribution could not be excluded on the basis of the relatively small dataset). Using all 20 results: • •
RI PT
•
the mean value of ln(Kmat) was 4.61, ie mean Kmat=exp(4.61)=100.2MPa√m if a lognormal distribution is assumed, B C , was 107.6MPa√m if a normal the arithmetic mean value of Kmat (based on all the data), ie / distribution is assumed, Median value of Kmat was 97.0MPa√m.
standard deviation, S, is 43MPa√m, k=1.24 (ie the characteristic lower bound is 1.24S below the mean), Kmat=54.4MPa√m.
M AN U
• • •
SC
Sets of between three and 19 Kmat results were then chosen at random from the set of 20 available. For the set of three, a check on homogeneity was made as per BS 7910:2013, ie the dataset was used only if the maximum value of Kmat was less than 140% of the mean, and minimum more than 70% of the mean (there is no advice in BS 7910 as to how to conduct homogeneity checks on larger sample sizes). The characteristic value Kmat was then calculated as per equation [7] for all datasets with 3≤m≤20, assuming both a normal and a lognormal distribution. From Figure 9, it can be seen that the value Kmat oscillates considerably around the best estimate (assumed to be that based on the entire batch, ie m=20), stabilising at around m=15. For m=20, ie using all available data and assuming a normal distribution:
TE D
B C , in view of the relatively large standard deviation. If a logand is roughly half of the raw mean, / normal distribution is used, a slightly high value of Kmat=62.2MPa√m is calculated. Results from the MOTE analysis lie consistently above those from the statistical analyses, in spite of the fact that the implication of the current wording of the standard is that MOTE and statistical analysis would be expected to give broadly similar results. This trend is consistent with that reported for another batch of material examined by one of the authors (22). It would be prudent to revisit this implication in future, drawing on additional datasets where possible.
EP
Of course, the extent of scatter in fracture toughness in ferritic steels will depend on the test temperature relative to the ductilebrittle transition curve, with very high scatter expected in the transition region and relatively low scatter on the upper shelf. It may also depend on microstructural factors, although the particular example shown here relates to a throughthickness notch in weld metal, in which the crack front samples a range of different microstructures, so specimentospecimen variability in microstructure is likely to have been small.
AC C
The use of Annex L is also illustrated below, using the same 20 datapoints as those discussed above, o ie 50mm thick fracture toughness data from weld metal tested at 60 C. o
From the ASTM E1921 procedure (Stage 1), T0=85.8 C. From the MML procedure (Stage 2), o T0=66.2 C. The difference between the values of T0 at Stage 1 and Stage 2 is substantial and o indicative of data inhomogeneity, as noted in the original work. From T0=66.2 C, Kmat can be o calculated for a 50mm thick specimen and a temperature of 60 C at various levels of confidence from the Master Curve equation [8]. For Pf=0.05, Kmat=59.1MPa√m, and for Pf=0.5, Kmat=94.9MPa√m. Note that Stage 3 of the MML procedure (which applies a further correction when there are fewer than 10 results in the dataset) was not needed in this case. Results from the MOTE method, clause 7.1.5 and Annex L applied to the same set of 20 weld metal fracture toughness results are summarised in Table 2. Table 2 Summary of results from analysis of fracture toughness data Approach
Normal distribution
Mean toughness, MPa√m 107.6
Lower bound toughness, MPa√m
Comment
54.4
All data 11
ACCEPTED MANUSCRIPT Lognormal distribution Absolute minimum (3≤m≤5) MOTE (6≤m≤15)
100.2
62.2
All data

Between 53.3 and 75.3

Between 63.7 and 117.4 Between 34.1 and 81.2 59.1
Three results chosen at random from 20; those with excessive scatter ignored See Figure 9
Clause 7.1.5 94.9
RI PT
Annex L
See Figure 9
CONCLUDING REMARKS
M AN U
SC
The materials properties clauses of BS 7910:2013 have been substantially expanded relative to the earlier (2005) edition. They now include improved information on the determination of tensile properties and a hierarchy of methods for determining fracture toughness (particularly in materials showing brittle behaviour) ranging from MOTE (Minimum of three Equivalent) to the detailed statistical analysis given in Annex L. Application of the three different methods is illustrated through a case study, based on a set of fracture toughness results from a weld metal. This has identified potential discrepancies between the definition of Kmat derived from MOTE and those derived from more advanced methods; examination of additional datasets is recommended in future revisions of BS 7910. ACKNOWLEDGEMENTS
The contribution of all members of the materials properties panel is gratefully acknowledged. The financial support given to TWI for standardsrelated work through its industrial membership is also greatly appreciated. REFERENCES
TE D
1. BS 7910:2013+A1:2015 incorporating Amendment 1 and Corrigenda 12: ‘Guide to methods for assessing the acceptability of flaws in metallic structures’, BSI, 2015. 2. R6  Assessment of the Integrity of Structures Containing Defects: Revision 4, 2000, as amended. 3. http://www.eurofitnet.org/ and http://www.eurofitnet.org/sintap_index.html.
EP
4. Isabel Hadley and Henryk G Pisarski: ‘Overview of BS 7910:2013’, ESIA12, 12th International Conference on Engineering Structural Integrity Assessment, 28 and 29 May 2013, Manchester, UK.
AC C
5. Isabel Hadley: ‘Progress Towards the Revision of BS 7910’ PVP201157307: ASME Pressure Vessels and Piping Division Conference, Baltimore, Maryland, USA, July 2011. 6. John Sharples, Peter Gill, Liwu Wei and Steve Bate, ‘Revised Guidance on Residual Stresses in BS 7910’, Proceedings of the ASME 2011 Pressure Vessels & Piping Division Conference PVP2011 July 17  21, 2011, Baltimore, Maryland, USA. Paper No. PVP201157071. rd
7. API 5791/ASME FFS1: ‘Fitness for service’, 3 edition, 2016. 8. DNVOSF101, ‘Submarine Pipeline Systems’, October 2013, http://rules.dnvgl.com/docs/pdf/DNV/codes/docs/201310/OSF101.pdf (now superseded by DNVGLSTF101 Submarine pipeline systems, October 2017, https://www.dnvgl.com/oilgas/download/dnvglstf101submarinepipelinesystems.html) 9. AC Bannister, J Ruiz Ocejo and F GutierrezSolana: ‘Implications of the yield stress/tensile stress ratio to the SINTAP failure assessment diagrams for homogeneous materials’, Engineering Fracture Mechanics 67 (2000), 547562. 10. BS EN ISO 15653:2018: ‘Metallic materials. Method of test for the determination of quasistatic fracture toughness of welds’, BSI. 12
ACCEPTED MANUSCRIPT 11. ASTM E1820  17a: ‘Standard Test Method for Measurement of Fracture Toughness’, ASTM. 12. ISO 12135:2016: ‘Metallic materials  Unified method of test for the determination of quasistatic fracture toughness’. 13. API RP 2Z: ‘Recommended Practice for Preproduction Qualification for Steel Plates for Offshore Structures’, 4th Edition, September 1, 2005.
RI PT
14. BS EN 10225:2009: ‘Weldable structural steels for fixed offshore structures. Technical delivery conditions’. 15. Jutla, T and Garwood, SJ: ‘Interpretation of fracture toughness data’ Metal Construction, 19(5), 276281R, (May 1987), ISSN 03077896.
SC
16. HG Pisarski and T Jutla: ‘Statistical correlation of large and small scale fracture toughness tests notched into weld HAZs’, Int. Conf on Weld Failures, London 2124 November 1988, P11 – P113, TWI. 17. Kim Wallin: ‘Fracture toughness of engineering materials: estimation and application’, EMAS Publications, 2011, ISBN 0955299462.
M AN U
18. ASTM E192118: ‘Standard Test Method for Determination of Reference Temperature, T0, for Ferritic Steels in the Transition Range’. 19. E Lucon, K Wallin, P Langenberg and H Pisarski: ‘The use of Charpy/fracture toughness correlations in the FITNET procedure’, OMAE200567569, Offshore Mechanics and Arctic Engineering, Halkidiki, Greece, 1217 June 2005. 20. INSTA technical report 1991: ‘Assessment of structures containing discontinuities, Materials Standards Institution, Stockholm (now superseded).
TE D
21. I Hadley and MG Dawes: 'Fracture toughness testing of weld metal: results of a European round robin', Fatigue and Fracture of Engineering Materials and Structures, 19/8, 963973, 1996.
AC C
EP
22. Pisarski, H: ‘Treatment of fracture toughness data for Engineering Critical Assessment (ECA)’, Welding in the World, 2017, 61(4), 723732.
13
AC C
EP
TE D
Figure 1 Development of BS 7910
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
Figure 2 Committee Structure 14
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
Figure 3 Example of continuous stressstrain curve (xy =400MPa)
Figure 4 Example of a discontinuous stressstrain curve showing various definitions of yield strength
15
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
Figure 5 Example of Option 1 FADs for continuously and discontinuously yielding materials
Figure 6 Variation of the modulus of elasticity for ferritic steels with temperature
16
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
Figure 7 Calculation of the parameter ‘m’ from experimental data
Figure 8 Example of the Master curve for three different section thicknesses – median and lower o bound curves; T0=50 C assumed
17
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
Figure 9 Example of the use of statistical analysis of a dataset with 3≤m≤20
18
ACCEPTED MANUSCRIPT Highlights
RI PT
SC M AN U TE D EP
•
The background to the materials properties clauses of BS 7910:2013 is described. A case study, based on a set of fracture toughness data from weld metal, illustrates three different ways in which lower bound fracture toughness can be calculated: o Minimum of Three Equivalent (MOTE), as per Table 5 of the standard, o statistical analysis using Table 6 of the standard o use of the Master Curve (Annex L). Indications for future changes to the materials properties clauses are given.
AC C
• •