I Hadley and J B Wintle - TWI
Paper presented at Developments in pressure equipment - where to next? I Mech E, London, 23-24 November 2004.
The future of fitness for service (FFS) assessment is secure because of the significant safety and economic benefits that arise from the application of FFS techniques. This paper reviews the evolution of FFS assessment procedures for fracture assessment. It traces the development of the current procedures to the present day and highlights the new areas of development. The factors that are driving these developments are also discussed.
This paper reviews the evolution of FFS procedures for fracture assessment. It traces the development of the current procedures to the present day and highlights the new areas of development. The factors that are driving these development are discussed.
In cases where defects (cracks, fabrication flaws, in-service flaws) are (or could be) present in safety-critical welded structures such as pressure vessels, it is both prudent and efficient to develop standardised methods to evaluate them. These methods, know as fitness for purpose (FFP), fitness for service (FFS) or Engineering Critical Assessment (ECA), can be used for various applications such as demonstrating a safety case, extending operational life or calculating defect acceptance criteria for welded fabrication of a new structure. FFS procedures cover a variety of failure/damage modes, including fatigue, corrosion and creep; this paper, however, considers only failure under static loading, by fracture and/or plastic collapse.
2. Development within the nuclear industry
It is now almost 30 years since the publication of the first UK procedure for the assessment of the integrity of structures containing defects. This document, known as R6, was first published in 1976 by the UK Central Electricity Generating Board (CEGB). Its main aim was to ensure that safety cases relating to structural integrity were carried out in a reproducible and consistent fashion. R6 was initially aimed principally at analysis of nuclear pressure vessels and piping, and has influenced other nuclear standards and procedures such as ASME XI and the GE-EPRI procedure.
Although applicable in principle to all types of welded or fabricated structures, it gained its early reputation through its application to the analysis of thick-walled pressure vessels, often in the stress-relieved condition. The development of R6 was driven by the need to demonstrate the safety of welded pressure systems in all types of plant operated by CEGB where the limitations of linear elastic fracture mechanics were apparent. The relevant components tended to be stress relieved and made from ductile materials. New ultrasonic defect detection and sizing techniques that were being more widely used, often for the first time on some equipment, required methods for setting reporting levels and assessing any planar defects found during inspections.
R6 introduced the concept of the failure assessment diagram (FAD), an example of which is shown in Fig.1. The vertical axis of the FAD (Kr) represents the ratio of applied stress intensity factor to the fracture toughness of the component material, whilst the horizontal axis (Lr) represents the ratio of the applied load to the load required to cause plastic collapse of the flawed section. The interaction between the two failure modes is represented by the failure assessment line, based on the tensile properties of the component material. Consequently, values of Kr and Lr can be calculated independently for flawed components and the point plotted on the FAD; points lying inside the FAD are deemed safe, those outside are potentially unsafe.
The FAD concept marked a significant advance in improving the accessibility and application of fracture assessment. It provided a visual and graphical means to see the proximity to failure and the sensitivity to changing the values of the key parameters according to their degree of uncertainty. The complex mathematical background of fracture mechanics was hidden, yet the method allowed reasoned judgements to be made by more broadly based engineers and gave confidence to open up the field.
R6 has continued to be maintained and developed, first by the CEGB, later by the successor organisation Nuclear Electric and now by an industrial collaboration managed by British Energy. It is now in its forth major revision. The current revision incorporates several changes to ensure consistency with the SINTAP fracture assessment procedure ('Structural Integrity Assessment Procedures for European Industry'), developed as part of a European BRITE-Euram programme. Although widely used outside the nuclear industry, R6 is particularly suited to the type of fracture assessments likely to arise in nuclear plant, for example, in its treatment of thermal and residual stresses and its expectation that detailed analysis and material data will be available.
3. Development in other industries
In parallel with the development of the R6 procedures, researchers at the British Welding Research Association (later The Welding Institute/TWI), were working on the concept of evaluating material toughness by measuring the displacement of crack faces prior to brittle failure. This proved to be particularly important in cases where significant crack tip plastic deformation precedes failure and measurement of plane strain fracture toughness is not possible. The relationship between this measurement of toughness, now known as CTOD or δmat (Crack Tip Opening Displacement), and the crack driving force was incorporated into the so-called CTOD design curve, a semi-empirical curve used to assess defect tolerance in welded joints.
The method was first published as a BSI PD (Published Document) 6493 in 1980. It was enthusiastically adopted by the UK offshore industry in particular for the assessment of offshore structures, pressure vessels and pipelines. As these tended to be in the as-welded condition, the CTOD approach to characterising fracture was appropriate without undue conservatism. The initial driver was the need to reduce repair rates during fabrication through the application of fracture assessment at the design stage, although it later became clear that fracture assessment was also important in justifying life of thick walled structures subject to fatigue loading.
The differences between R6 and PD6493 were largely those of terminology and presentation and, while the underlying fracture mechanics principles were the same, this was not recognised for a number of years. It is to the credit of those on the WEE 37 committee in the 1980's that two approaches were brought closer together. The second edition of PD6493, published in 1991, adopted the Failure Analysis Diagram (FAD) approach used in R6. This process continued in 1999/2000 with the issue of BS 7910 in 1999, which has superseded PD6493 and now has the status of a British Standard Guide.
4. Current issues
In spite of the many years of successful application of FFS methods, some problems remain with the acceptance, use and interpretation of the procedures. For example, both R6 and BS 7910 are written in a way that requires the inputs to the analysis to be conservative - upper bound values of 'applied' variables such as stress and flaw size, plus lower bound values of 'resistance' variables such as fracture toughness and tensile properties. It may be for this reason that there is a perception amongst some users that the methods are 'too conservative'.
Experience from validation studies shows that the locus of failure of a test component containing a flaw that has been analysed using an FFS method is likely to be well outside the FAD, not on the assessment line. An example of this can be seen in Fig.2, which summarises the results of a large number of wide plate tests and other full-scale tests carried out on well-characterised materials in which the size of flaw and applied stresses at failure can be very accurately measured. Analysis of the results was carried out using standard fracture assessment procedures, e.g. using handbook stress intensity factor and plastic collapse solutions.
The difference between the failure assessment line and the locus of failure is due to a number of inherent safety factors in the procedures. For example, the 'secondary' stresses arising from welding may not be known accurately, but estimated from simplified, conservative, methods. The difference between the analysis point and the failure assessment line can, however, be minimised by application of a range of advanced structural integrity techniques, some of which are described in appendices to the main fracture assessment procedures. Examples include:
- Calculation of crack driving force using Finite Element Analysis (FEA) of the cracked body (rather than use of a standard K-solution),
- Evaluation of materials properties using statistical techniques,
- Calculation and/or measurement of the residual stress field at the crack tip,
- Analysis of crack tip constraint,
- Analysis of the effects of weld metal overmatching.
5. Constraint based fracture
An example of the use of more advanced procedures is shown in Fig.3, which shows the results of a uniaxially loaded wide plate test analysed using standard procedures, then re-analysed using the constraint-based appendices of R6 and SINTAP. The results of the analysis move from outside the failure analysis line to a point close to or on the line when constraint correction is applied. Obviously a price is paid for this increasing accuracy in terms of the costs of increased testing and analysis.
Constraint based procedures take greater account of the effect on the flaw of the distribution of stresses within the context of the structure, and this may impact both the measurement of fracture toughness and the calculation of crack driving force. The matching of constraint between fracture toughness test specimens and real structures is becoming more accepted, one example being the use of the single edge notched tension specimen (SENT) for assessing flaws in pipe girth welds. Work is also in progress to try to relax the need for using local collapse in structures by moving to a global collapse criteria in structures where there is clearly structural redundancy surrounding the flaw. In the future, this will improve the assessment of buried flaws, short flaws and flaws near nozzles.
6. Flaws in mismatched welds
A second example of the benefits of a more detailed analysis of a fracture assessment problem is shown in Fig.4 with regard to overmatching and undermatching welds (i.e. those where the weld tensile properties are higher/lower than those of the parent material). This shows the results of fracture analyses (in practice, analysis of plastic collapse conditions) on girth welds in grade X100 pipelines with different 'overmatch ratios' (M0.2), where M0.2<1 denotes undermatched welds and M0.2>1 corresponds to overmatched welds. Results are shown in terms of a normalised maximum tolerable load at failure, Pm/SMYS (Pm is the predicted failure load and SMYS the Specified Minimum Yield Strength). The predicted failure condition is shown as a function of M0.2 and 2H, calculated either from a 'standard' BS 7910 approach or using the mismatch limit load solutions of SINTAP and R6. The influence of weld width (2H) is also considered.
The figure shows the advantage in both overmatched and undermatched welds of more detailed analysis over approaches to fracture assessment where the minimum tensile properties are assumed throughout. The benefit of using the more advanced analysis, especially for the case of narrow undermatched welds, is clear.
7. Treatment of input data
The R6 and PD6493 procedures allow the user to analyse the problem at various levels depending on the quality of the input data - an answer that is 'too conservative' is often indicative of a lack of input data rather than a weakness in the fracture assessment procedure itself. If the inputs to the problem are sufficiently well-defined, as is likely to be the case with laboratory tests on well-characterised materials, the failure locus can be accurately predicted. Conversely, if a fracture assessment of a 50-year-old vessel with little or no supporting design/materials data is required, conservative assumptions have to be made regarding the input data, and it may be more difficult to make a case for fitness-for-service.
The advent of finite element analysis has without doubt made stress analysis of structures more accurate for a given defined loading The benefit of this refinement for fracture assessment is, however, limited in many cases, since the greatest uncertainties arise in relation to the loading, materials data and the sizing, shape, position and orientation of the flaw. In respect of the latter, the new capabilities of finite element analysis to incorporate the real size and shape flaw within the model will bring about a better estimation of the crack driving force, and consequently more realistic assessments. This development needs to be matched by a similar improvement in the sizing capabilities of NDT procedures.
Variabilities of materials data and loading have traditionally been dealt with using lower bound and best estimate values within sensitivity analyses. Probabilistic fracture assessment offers a means to incorporate these variabilities if the probability density distributions can be estimated statistically, and the software now exists for computing multiple variables. In terms of procedures, a full probabilistic assessment is unlikely to be generally acceptable and work is required to simplify such assessments to allow a certain probability of fracture to be obtained by using partial safety factors representative of the degree of variability on best estimate values.
8. Strain based approaches
There is now a trend towards fracture assessment procedures for dealing with problems specific to certain industries. For example, in the marine pipeline industry, pipeline reeling is an efficient and widely-used technique for installing small-diameter marine pipelines, since it allows the welding and inspection of girth welds (and repair, where necessary) to be carried out onshore. During installation by reeling, the pipeline may be subjected to total strains of around 2% across pipeline girth welds. Consequently, there is a risk of extension of small defects during installation, and a need to demonstrate that the inspection techniques used are sufficiently sensitive to detect and reject flaws that could extend by unstable ductile tearing during installation.
Current procedures such as BS 7910 and R6, which assume elastic and small-scale yielding fracture behaviour, do not adequately address this large strain situation. They are much too conservative. Hence, strain-based procedures have recently been developed and validated. The current procedure entails measurement of the tearing resistance of the weld using low-constraint specimens, such as the single edge notched tension specimen (SENT), and the use of the material stress-strain curve to derive 'equivalent' elastic stresses corresponding to the appropriate applied strain. The predicted amount of tearing during installation is finally validated by use of a surface-notched strip specimen.
Similar approaches can be adopted for other cases of strain-controlled loading, such as may arise during seismic loading or as a result of landslip. Strain based approaches can also improve the assessment of defects in high residual stress fields, and will become more important with the tendency to avoid post weld heat treatment in welded structures. Elastic-plastic finite element analysis now provides a means for determining the stresses and strains in structures subject to combined pressure, thermal and system loads, and removes the need to categorise loading as primary or secondary, often a cause of difficulty and uncertainty. In order to utilise these results, strain based approaches are required, and there is scope for considerably more development in the area outside the pipe reeling arena.
9. Ductile fracture of pipelines
Since the 1970s, the gas industry in the UK and the US has tended to develop its own procedures for dealing with fracture of pipelines. These have paralleled the developments elsewhere, but concentrated on ductile fracture of crack-like and other types of flaw in pipes. With more oil and gas being transported in long distance pipelines, there is a drive to use higher strength steels where the possibility and consequences of long running ductile fractures may be greater than previously.
Consequently, there is renewed interest in ductile fracture and arrest of long dynamic through-wall cracks in order to select pipeline parent materials and seam welds with appropriate fracture properties and for the design of crack arresters. The search is on for a different fracture criterion from J for assessing continued propagation or arrest. There is now much attention given to the use of the cohesive zone model, an early approach to describing fracture based on the crack face separation and energy densities ahead of the crack. Further developments in this area can be expected.
10. Training and certification
The development of standardised fracture assessment procedures has created a requirement for the training of practitioners. Fracture assessment can be a safety critical activity, yet the procedures need interpretation and require a degree of engineering judgement if they are to be properly applied. The ready availability of user-friendly software makes training for its appropriate use more important.
A number of organisations, including British Energy and TWI and some universities, offer training to those wishing to undertake fracture assessment. There is, however, a current lack of defined level of competence or certification, so it is difficult for companies to know the level of expertise of their staff or contractors. Benefit would be obtained from having a common certification in fracture assessment based on an agreed syllabus and examination.
11. European standardisation
There is now considerable convergence between the two UK fracture assessment procedures although, in view of the continued difference in user communities, each continues to be separately maintained and validated. Meanwhile, development of FFS procedures is continuing via the European thermatic network FITNET, the ultimate aim of which is to produce a European consensus on FFS analysis. This will take input from R6, BS 7910, SINTAP and other procedures along with industry experience in order to produce a method that is both technically rigorous and user-friendly, and that meets the needs of a range of different industries.
The eventual aim is to publish the output of FITNET in the form of a CEN document via a CEN Workshop Agreement (CWA). This is a mechanism that allows for relatively rapid publication of an agreed procedure, and is particularly appropriate for an FFS procedure, which can be seen as a framework rather than a set of rigid rules. Moreover, publication through a CWA would not cause standstill on existing national procedures such as BS 7910, as would be the case for a CEN standard.
12. Future research
Whilst current fracture assessment procedures have already brought about enormous benefits for fabricators and operators in terms of cost savings and better understanding of structural safety, we have by no means reached the end of the road so far as further development of procedures is concerned. A better understanding is clearly required in the areas such as strain-based approaches, the link between defect tolerance and the reliability of NDT techniques, and the analysis of dynamic fracture. While fracture assessment procedures are established for steel and aluminium, their application to other metallic materials is less common, and their validity to non-metallic materials such as ceramics, plastics and composites is virtually untested.
Fracture assessment development is active in the UK at TWI, British Energy, British Nuclear Group, Serco Assurance and some universities. There are also several research institutes in mainland Europe that are developing new understanding and procedures, notably GKSS in Germany and VTT in Finland. Whist the British Energy managed R6 programme can feed directly into the R6 procedure, bodies such as FITNET and the R6 and BS 7910 committees can only really collate information that is already published. What would be particularly advantageous is a research programme co-ordinated at a European level.
- 'Assessment of the integrity of structures containing defects', R6 Rev. 4, British Energy Generation Ltd/BEG(UK)L, 2001.
- SINTAP, 'Structural integrity procedures for European Industry', 1999, BRITE-EURAM contract No. BRPR-CT95-0024.
- BS 7910: 'Guide on methods for assessing the acceptability of flaws in fusion welded structures' (incorporating Amendment 1), BSI, 1999.
- H G Pisarski and C M Wignall: 'Fracture toughness estimation for pipeline girth welds', International Pipeline Conference, IPC 2002, Calgary Alberta, Canada, ASME, IPC 2002-27094, 29 September - 3 October2002.
- H G Pisarski, Y Tkach and M Quintana: 'Evaluation of weld metal strength mismatch in X100 pipeline girth welds, Paper IPC 04-0232, Proceedings of IPC 2004 International Pipeline Conference, October 4 - 8, 2004,Calgary, Alberta, Canada.
- A Cornec, I Scheider and K-H Schwalbe: 'On the practical application of the cohesive model', Engineering Fracture Mechanics, Volume 70, Issue 14, Pages 1963-1987, 2003.