Welds, their Quality and Inspection Capability for High Integrity Structures and Components
R E Dolby, I J Munns, C R A Schneider and R H Leggatt TWI, Cambridge
Presented at 1999 TAGSI Seminar - 'Fracture, Plastic Flow and Structural Integrity' (dedicated to Sir Alan Cottrell in the year of his Eightieth Birthday), Held at TWI, Cambridge, UK, 29 April 1999.
The paper examines three main effects of welding, namely flaws, residual stresses, and macro/microstructural changes, but pays particular attention to aspects of flaws and their detectability in view of Sir Alan Cottrell's special interest in this topic. The factors governing the incidence of large flaws and their size distribution are discussed, together with the main measures for their avoidance. Detectability of large flaws by radiography is reviewed and a recently proposed new index for predicting the probability of detection is explained. Recent progress in the detectability of flaws by ultrasonic inspection is then discussed. The paper reviews recent developments in residual stress determination and the benefits of standardised profiles, and finally discusses the main effects of welding on properties and microstructures of joints, particularly heterogeneity issues.
In the 40 years since Sir Alan Cottrell came to the University of Cambridge as Head of the Department of Metallurgy, there have been very many important achievements in the field of structural integrity (SI). The purpose of this paper is to show why welding is crucial to the safety and reliability of high integrity plant. The message will be simply - welds matter in most engineering critical assessments.
During Sir Alan's first years at Cambridge, very few welding processes were available, principally stick electrodes and submerged arc welding; metallurgists were worrying mainly about hydrogen induced heat affected zone (HAZ) cracking; and weld quality was assessed by Charpy V and hardness tests, together with radiography of completed welds.
Today, the joining processes are not too different for high integrity plant such as nuclear constructions, but the welding power sources, consumables and quality assurance have dramatically improved. The effects of irradiation are under much closer study; dissimilar metal joints are always of concern and, in the SI field, fracture mechanics is in routine use with risk assessments often being made using probabilistic approaches. The detection of potentially hazardous flaws by radiography and ultrasonics is now on a much firmer theoretical and practical footing.
In the context of plant safety and reliability, the effects of welding are simply put:-
- - Flaws will have been created
- - Stress concentrations and residual stresses are certain to be present
- - New macro/microstructures will have formed with different properties to the parent steel.
This paper will look at the present position for these effects, identifying current research and recent achievements. The main emphasis will be on flaws and their detectability in view of Sir Alan's special role in emphasising the importance of inspection validation in the early 1980s and his continued interest in this technical area.
2. Flaws in welds arising from fabrication
In the 1950s/60s, welding metallurgists were focussing attention on several forms of cracking that could be created during fabrication; hydrogen induced HAZ cracking in low alloy steels, solidification cracking in stick electrode deposits, small liquation cracks in some C-Mn and stainless steels and post weld heat treatment cracking in some high C stainless grades notably Type 347 for chemical plant.
Today these forms of cracking are well understood, but three additional crack types were encountered during fabrication of high integrity plant in the 1970s/80s which were of major concern, principally because they may be large, say greater than 15mm in height. These were:-
- - Weld metal hydrogen induced cracks
- - Lamellar tearing in the HAZ and/or parent material
- - Reheat cracks in ferritic steel weld metal or HAZ, occurring during PWHT.
No weld can, or should, be assumed to be free of crack type flaws or, indeed, of any other flaw types, some of which may be welder related, e.g. lack of fusion. But, when fabricating nuclear plant today, we know the cracking and other flaw types that could be encountered; we know where they are likely to be found, and we know their likely orientation and probable size. We also know how to specify steels, consumables and welding procedures to minimise their occurrence.
A current preoccupation, however, is extending the life of older plant where the knowledge base at the time of construction was much less good and the incidence of flaws higher. Fortunately we are helped in the analysis of ageing plant by today's more comprehensive knowledge base.
2.2 Flaw size distributions
Given that all welds will contain some flaws, a key question is how many and what size are they? In any probabilistic fracture mechanics approach used for engineering critical assessments, a distribution of flaw sizes in a given component is required and particular consideration needs to be given to the possible presence of large defects. It is the large defects which are likely to be significant in terms of structural integrity and which need attention in terms of evaluation and sizing.
A recent TAGSI report examined the available data for thick butt welds in both nuclear and non-nuclear components  . It concluded that large defects of greater than 15mm in height are rare but because existing published flaw distributions have been obtained by fitting an assumed probability density function to flaws with relatively small heights, they are likely to be in error at large flaw heights. In addition, because there are a variety of metallurgical mechanisms giving rise to flaws in welds, each flaw type will have its own height distribution and it is unlikely that a single probabilistic density function will exist. Overall there is very little published data on the incidence of large defects, this being partly explained by the fact that weld repairs are usually undertaken on large defects and this biases the information on distributions to small sizes.
Figure 1 shows a measured probability density function for flaws in ducting and boiler welds in Magnox reactor circuits, and its associated extrapolation, together with a predicted function derived by knowledge elicitation from welding engineers at Rolls Royce and Associates  . Whilst there is good agreement at small flaw sizes, predicting the presence of large flaws is the main source of uncertainty so it is clear that the main way forward is:
- to have a thorough knowledge of the circumstances in which large flaws can form, and then design these out by correct choice of materials, weld detail designs, welding process and welding procedure, and
- to use inspection techniques that maximise the probability of detection of large flaws.
The next section will summarise the factors leading to the larger flaw types, these being our biggest concern.
Fig.1 Duct/boiler welds (MMA and SA) 25mm thickness ( Ref.2 )
2.3 Large flaw types
There are basically three forms of cracking in ferritic steels which could have heights greater than 15mm arising from fabrication, namely weld metal hydrogen cracks, lamellar tears and reheat cracks. All these types can run across several weld beads or heat affected zones. Other crack types, such as liquidation cracks, solidification cracking and HAZ hydrogen cracks, are almost invariably size constrained to one weld bead although, occasionally, the latter can extend into adjacent weld metal in some joint configurations and result in cracks of more significant height. Lastly, lack of sidewall fusion is a further flaw type that could exceed 15mm in height in welds of more modern nuclear plant where narrow gap welding processes have become the norm for vessel manufacture. The main flaws of concern are, therefore:-
|Weld metal hydrogen cracks
|Weld configuration at risk:
|Welding processes at risk:
||submerged arc with basic fluxes, MMA
||clean SA wires
consumables of low H 2 potential
dry fluxes and coatings
adequate preheat, interpass, post heat
||HAZ and parent steel
|Weld configurations at risk:
||T-butts, heavy fillets
|Welding processes at risk:
||MMA, submerged arc, FCAW
||using parent material with low inclusion content and >20% STRA
|Reheat cracks (occurring during PWHT)
||HAZ, close to fusion boundary
|Weld configurations at risk:
||thick butts, heavy fillets
|Processes at risk:
||MMA, submerged arc, FCAW
||parent metal selection
low heat inputs
|Lack of sidewall fusion
||Fusion line between weld metal and HAZ
|Weld configuration at risk:
||narrow gap butt welds
|Processes at risk:
||SA or GMA
For all of these large flaw types, very little new in our qualitative understanding of their formation has turned up over the last decade. The welding procedural factors and the steel and consumable chemistries and processing routes which influence their occurrence are well known. Whether the flaws arise in practice depends crucially:
- on the steel and consumable purchasing specifications, and
- on the quality assurance (QA) procedures used in the fabrication shop.
With today's modern steels and consumables, and with good 'housekeeping' and QA, large flaws should be completely absent. What is required is assurance from inspection methods that in the unlikely event that large flaws have formed, then the probability of detecting such flaws is at the highest possible level. This aspect is discussed in the following section.
3. Reliability of inspection
Not all sizes of flaw are detectable by non-destructive testing (NDT). Reliable flaw detection depends, not only on the size of the flaw, but also on its orientation, morphology and the particular NDT method employed. To gain an insight into the capability of NDT a number of significant studies have been carried out, many of these funded by the nuclear industry.
One of the most significant contributions to the understanding of radiographic detection capability was the work carried out by Pollitt and Halmshaw in the early 1960s. This work led to the development of a theoretical model  , often referred to as the Pollitt model, which can be used to predict the detectability of flaws under specific radiographic conditions. To use the Pollitt model the following inputs are required:
- knowledge of the radiographic inspection procedure
- a measurement of the achieved radiographic image quality (usually measured directly from the radiograph)
- an estimate of flaw size, orientation and gape
Using this information, the model can then predict whether or not a flaw of the size and orientation specified is likely to be detected under the prescribed radiographic conditions.
3.1.1 Validation of Pollitt model
At the time the Pollitt model was developed, some experimental work was carried out to validate the theoretical approach used  . This work involved radiographing a steel test block containing parallel-sided, planar slots of different depths and orientations. The agreement observed between Pollitt theory and the practical results was sufficiently close to enable the limits of detectability of these 'artificial' defects to be predicted with considerable confidence. This is not surprising since Pollitt theory itself considers flaws to be of relatively simple shape. For example, a crack or lack of sidewall fusion flaw is modelled as a smooth, parallel-sided slot; identical in character to those flaws examined experimentally. Because of this it has been unclear whether Pollitt predictions are valid for the more complex morphologies of real welding flaws, which may be rough and of variable gape (opening). This gap in knowledge has persisted until relatively recently, when a large-scale investigation into the capability of radiography was funded by the UK nuclear licensees.
The core of this work was carried out by TWI and focused on the detectability of large planar manufacturing defects, at least 15mm in through-wall size, in ferritic steel butt welds 50-114mm thick. Radiographic detectability was assessed using seven specimens containing 19 deliberately induced, but metallurgically realistic, planar welding defects. The defects considered were: centreline solidification cracking, lack of sidewall fusion and hydrogen cracking (both in the weld metal and heat affected zone).
These specimens were subject to over 130 different exposure conditions, using both X-ray and Co-60 sources, to give over 300 flaw/radiograph combinations for evaluation. The exposures included angled shots to simulate radiography of different weld geometries, and some involved the use of spacer plates to increase the thickness of steel radiographed. Each radiograph was interpreted 'blind' by two qualified radiographers, who were asked to classify any flaws they detected as being either 'easily visible' (EV) or 'barely visible' (BV) on the radiograph. After radiography, the specimens were sectioned, and detailed measurements of the size, gape, orientation and roughness of the defects were made.
The key findings from the main study were that:
- Detectability of large planar flaws (>15mm in through-wall size) generally decreases with increasing thickness of steel radiographed and, for most flaw types, is noticeably worse at thicknesses of 100mm or above.
- Detectability generally decreases as the angle of misorientation between the flaw and the radiographic beam is increased. This reduction in detectability is more pronounced for relatively smooth, tight flaws such as the transverseweld metal hydrogen cracks.
- Detectability improves with increasing gape (i.e. higher angles of misorientation can be tolerated for flaws of larger gape). This effect is shown in Fig 2.
Fig.2 Radiographic detectability versus flaw gape and misorientation angle
In any assessment of radiographic performance it is important to acknowledge the risk of misinterpretation. Serious defects may be detected but, in some cases, these may be mischaracterised and passed as acceptable, e.g. small cracks may be misinterpreted as less serious 'thread-like' flaws. In the main study, reported here, all of the flaws were relatively large (i.e. ≥15mm in through-wall size) and all were correctly identified as planar.
Detailed sectioning data from the main study has enabled the experimental detectability of each flaw/radiograph to be compared with simple Pollitt theory. In over 89% of the cases considered, simple Pollitt theory correctly predicted the response of one or more of the radiographers. In the few cases where predictions made by Pollitt disagreed with experiment, there was a tendency for Pollitt theory to behave conservatively, predicting that flaws should not have been detected where in fact they were. The inherent conservatism of simple Pollitt theory is not entirely unexpected, since the model strictly only applies to smooth planar slots of uniform gape. Real metallurgical flaws are often rough, wavy and of variable gape - all characteristics which may enhance their radiographic detectability. A more rigorous theoretical model, which can accommodate these additional factors, has been developed by BAM, Berlin, and an initial investigation has given encouraging results.
3.1.2 Theoretical index of detectability
More recently, the simple Pollitt model of radiographic detectability has been extended to include the concept of an 'index of detectability', which quantifies the theoretical detectability of a particular flaw under specified radiographic conditions 
. In other words, it is now possible to estimate the margin of flaw detectability rather than stating only whether the flaw is detectable or not.
The index ( I) is derived using the following simple formula:
I = ln( Δ x flaw ) - ln( Δ x IQI )
In the equation above, Pollitt theory is used to express the flaw as a step-change in thickness ( Δ x flaw ), which theoretically produces the same change in density on the radiograph as if it were examined using the same radiographic exposure conditions. This theoretical step-change in thickness is then compared with a measured value, which indicates the smallest step-change in thickness actually detectable on the radiograph, ( Δ x IQI ). This value is measured directly from the radiograph using an Image Quality Indicator (IQI). The index is then simply the ratio of these two thickness values, presented on a log scale. In this way, flaws that produce positive indices are judged as being detectable, and flaws with negative indices as being undetectable. It follows from this that an index of zero is the threshold of detection and that, as the index becomes more positive, the more reliable flaw detection becomes. This concept leads naturally to the derivation of Probability of Detection (POD) curves.
A theoretical POD curve, fitted to data generated in the main study, is shown in Fig 3. Clearly, this fitted curve applies only to flaws of a similar type to those examined experimentally; caution needs to be exercised in extrapolating the results to other flaws. In Fig 3, we can see evidence of the pessimism inherent in the Pollitt model, with a proportion of flaws still detectable at negative indices.
Fig.3 Typical POD fitted to curve for the radiography of thick section welds, containing varying types of flaws
3.1.3 Application of theoretical index of detectability
The ability to derive POD data for specific flaws examined under specific radiographic conditions, provides the engineer with a powerful tool which can be used to improve plant safety in a number of different ways. For example:
- If an Engineering Critical Assessment (ECA) is used to define the critical size of flaw which can be tolerated by plant operating under known loading conditions, then the index can be used to determine how reliably flaws of this size can be detected by radiography.
- Alternatively, prior knowledge of the likely types, sizes and orientations of flaw which might occur in a particular structure could be used in conjunction with the index, to define the minimum performance which must be achieved by radiography if a radiographic inspection regime is to be effective.
In both of the above cases, the quality of the POD data generated depends critically on how accurately flaw parameters, such as through-wall size, orientation and gape, can be defined. In many instances it is possible to define a 'worst-case' flaw orientation angle based on knowledge of the weld prep, and an upper limit to flaw through-wall size based on ECA data or knowledge of the weld bead size. This leaves flaw gape, which is perhaps the most difficult parameter to define.
Extensive gape measurements on sectioned flaws have shown that it is virtually impossible to describe a single flaw using a single gape value. In Fig 4 for example, the two flaws presented are each characterised by a range of measured gapes. If these results are supplemented with data from other similar sized flaws, then it is possible to assign flaws of a particular type and size a characteristic gape distribution. These gape distributions can then be used to calculate a range of detection indices associated with that particular type and size of flaw.
Fig.4 Gape variation for different flaw types
Figure 5 shows how this knowledge could be used to define the minimum performance which must be achieved by radiography for it to be an effective inspection tool. This figure shows a series of POD curves associated with different quality radiographic procedures. The quality of the procedure is indicated by the number of IQI wires visible on the radiograph. A wire IQI comprises a series of seven wires of decreasing diameter. Image quality is measured by recording the number of wires that can be seen on the radiograph. The more wires visible, the higher the sensitivity of the technique and, theoretically, the better the radiographic detection performance. In practice, it is envisaged that design engineers and metallurgists will be able to define the likely types, sizes and orientations of flaw which might occur in a particular structure, in order to enable a range of detection parameters to be calculated and located on the POD curves. This is illustrated by the shaded region in Fig 5. Knowing this, it is now possible to estimate the performance of any new radiographic technique, directly from the IQI sensitivity achieved. For example, if a particular technique produces radiographs where only 3 wires are visible, then the majority of flaws defined by the shaded area in Fig 5 will be missed. However, if the technique is improved (by using a better quality film, a smaller source size, etc.), so that 5 wires are now visible, then detectability also improves and the vast majority of the same flaws will be detected.
Fig.5 Predicting the effectiveness of radiographic inspection
Overall, this extensive programme of work has shown that radiography is capable of detecting a wide range of planar flaws, particularly if they are extensive in both length and height. However, flaw types such as certain hydrogen-induced cracks can exhibit unfavourable combinations of gape and orientation, which may make even large flaws undetectable. Nevertheless this work has shown that the capability of radiography to detect large planar flaws is surprisingly high and better than simple predictive modelling would suggest. The introduction of an 'index' of detectability enables, for the first time, POD to be predicted for different radiographic techniques.
3.2 Ultrasonic Inspection
The PISC I trials in the 1970s  were one of the first investigations of the capability of ultrasonic inspection techniques to detect and size defects of concern in nuclear components. A series of test pieces containing implanted flaws were inspected by national nuclear inspection teams who had no previous knowledge of the flaws. Surprisingly, the results showed that certain procedures in use at the time gave levels of performance much lower than that expected, and that, if not properly controlled, ultrasonics could give an unacceptably low level of performance.
This was recognised by Sir Alan Cottrell and Sir Walter Marshall who, in the second Marshall Study Group report on reactor pressure vessel integrity  , recommended, in 1982, independent validation 'to ensure that the Licensing Authority can assess the adequacy of the chosen inspection procedures'. Following this recommendation and support from the UK regulatory body (the Nuclear Installations Inspectorate) the CEGB proposed, as part of the safety case for the UK's first PWR at Sizewell B, setting up an Independent Validation Centre. This was to qualify the inspection of safety related components, including the reactor pressure vessel and steam generators.
The experience gained from the UK of independent validation for Sizewell B components and the results from the PISC II and PISC III inspection trials led to qualification becoming recognised at a European level as having benefit for nuclear safety. A European Network for Inspection Qualification (ENIQ) was set up to develop a process for inspection qualification that could be generally applied. The result was the publication in 1997 of the European Methodology for Qualification  . Inspection qualification has also been incorporated into Appendix VIII of the ASME XI pressure vessel code, where it is called 'performance demonstration'  .
The European Methodology involves a combination of practical assessment (blind or non-blind trials), together with a paper-based technical justification. The technical justification can include previous experience of the procedure, laboratory studies and theoretical modelling. In particular, there are well-established theoretical models available for the ultrasonic inspection of planar flaws in ferritic steel, which have been extensively validated against experiment (e.g. Ref.10 ).
A large amount of NDT reliability data from blind trials such as PISC has recently been compiled  as part of a European collaborative project called SINTAP (Structural INTegrity Assessment Procedure). For ultrasonic inspection, the SINTAP results show a wide variation in the performance of different teams, mainly due to the different procedures used (see Fig 6). The SINTAP document identifies reasonably practicable qualification targets for inspection procedures in terms of Flaw Detection Probability (FDP) and Correct Rejection Probability (CRP), based on the reliability data collected. Values are given for flaws having through-wall extents of 5%, 10%, 40% and 100%.
Fig.6 Effectiveness of ultrasonic testing of thick section welds
The SINTAP document takes a deliberately 'broad brush' approach to the presentation of inspection reliability data. The document quantifies inspection performance by averaging over flaw parameters such as orientation, roughness, length and through-wall position. Such parameters can have a profound effect - see, for instance Fig 7 (reproduced from Toft  ). The SINTAP document is therefore of limited value for specific applications. However, it does provide an insight into the overall level of NDT performance that can be achieved in practice, and thus provides reasonably practicable qualification targets for specific inspection procedures. The SINTAP data relate mainly to flaws of a planar type that are roughly perpendicular to the surface (defect tilts up to 30°); these flaws are likely to include hydrogen cracks, reheat cracks and lack of sidewall fusion flaws.
Fig.7 Typical pulse-echo responses from 25mm diameter flaws in thick-section ferritic steel. ( Ref.12 )
One of the categories of components considered in the SINTAP report is that of a ferritic steel pressure vessel or pipe having diameter greater than 250mm and wall thickness t greater than 30mm. The SINTAP report gives the best attainable reliability of qualified ultrasonic testing of flaws with a through-wall extent (TWE) of 0.1 t to be FDP=95% and CRP=90%; these values appear to apply to defects longer than about 20mm. The corresponding reliability of 'good practice' radiography is given as FDP=90% and CRP=80%. Thus, in broad terms, the SINTAP results show that the inspection of thick section ferritic welds can be qualified to give highly reliable detection of flaws that extend through 10% of the wall thickness.
4. Residual stresses
Residual stresses can be a significant or dominant component of the driving force for crack initiation and propagation in high quality plant. Over the last decade or so excellent progress has been made and new analytical and experimental methods have been developed for quantifying the residual stresses in welded joints. Sophisticated numerical modelling techniques are now being widely used  , particularly in nuclear applications, to simulate the welding process and model the development of residual stresses during welding, after PWHT, after proof testing and in service under normal and abnormal operating conditions  . The neutron diffraction  and the deep hole methods  for measuring internal residual stresses have been developed, and the X-ray diffraction  and hole drilling methods  for measuring near-surface residual stresses have been standardised and widely applied.
The approximate depth of penetration in steel for the most commonly used measurement methods is shown in Fig.8 (from Ref.17 ). As an example of the data produced, Fig 9 shows residual stress measurements at a repair weld  in a 2 1 / 4 Cr-Mo steel with a yield strength of 455N/mm 2 . In this example, the maximum transverse residual stresses in the repair weld were 320N/mm 2 after welding, and 80N/mm 2 after 1680 hours at a representative service temperature of 540°C. These data could be used in engineering critical assessments of defects in the repair weld, and would provide much more accurate and less conservative results than the conventional assumption of yield magnitude residual stresses.
Fig.8 Residual stress measurement techniques
Fig.9 Through-wall distribution of transverse residual stress at 10mm deep repair weld in 76mm 2 1 / 4 Cr-1Mo steel
The use of these analytical and experimental techniques has led to a greater knowledge and understanding of the effect of component geometry, restraint, welding procedure, thermal properties, mechanical properties, phase changes and transformation plasticity on the magnitudes and distributions of residual stresses in welded joints  . This knowledge has led to more accurate analysis of the role of residual stresses in failure mechanisms, to the development of techniques for reducing residual stresses in sensitive locations, and to the preparation of standardised residual stress profiles for use in assessing the acceptability of defects in welded structures [21-23] .
Figure 10 shows a standardised profile for a T butt weld, which is conservatively defined relative to actual residual stress measurements.
Fig.10 Through-wall distribution of transverse residual stress at toe of T-butt weld
5. New macro/microstructures and associated properties
5.1 Macro Effects
In any welds in high integrity plant, there can be a major mismatch in both strength and toughness between the weld regions in the parent steel 
. Strength mismatch is more than likely, even after PWHT, the weld metal usually having the highest tensile properties when welding C-Mn and low alloy steels. The degree of overmatch will depend on the welding process and consumables, as well as the welding parameters used.
In addition, some variation in weld properties through the thickness of the joint should always be assumed, unless shown to be absent  . When making the joint, regions near the weld root will have a higher level of parent plate dilution, a different chemistry and a different thermal and strain history compared to weld passes nearer the joint surface. Higher tensile properties can be expected in weld metal near the root as a result, and these may still be present after PWHT. Specific details relating to the welding procedure are also important here, such as whether a different process was used to make the root passes compared to subsequent ones, or whether the joint was chipped or back gouged to allow easy access to the root when completing the second side of a joint. The way the plates are restrained during the welding procedure also matters as this can affect the strain history in root areas in partially completed joints.
The above discussion is also highly relevant to weld toughness properties, which must be assumed to be different from parent plate levels unless shown otherwise. Toughness variations through the joint thickness are more than likely, root regions often showing a lower cleavage resistance than near surface regions. Figure 11 shows some typical data in which these effects are evident in welded C-Mn steels, the root regions showing the highest tensile properties and lowest toughness.
Fig.11 Effect of PWHT on 40J Charpy V transition temperature of a MMA weld in C-Mn steel (Ref.25)
A higher weld metal strength can be beneficial of course. It can protect weld flaws from applied strains, depending on the direction of the applied strain and the orientation of the flaws, and this effect can be quantified in terms of the probability of fracture from such flaws . Plate tensile properties in a given structure have a minimum specified level but will have a mean and distribution just as for the associated weld deposits and any probabilistic approach must recognise this. However, the major concern in any case of a high level of strength mismatch, is that other properties may have been significantly changed, such as toughness, and this needs to be carefully followed up in any structural integrity argument.
5.2 Weld Pass Effects
Within a given joint, separate weld passes create heterogeneity in structure and properties. A given point in a weld will receive perhaps a dozen thermal cycles, but the key thermal cycles are usually those that reheat a region to between the Ac 1 temperature and the melting point. These will either cause significant austenite grain refinement or major austenite grain coarsening. As a result, strength and toughness will be locally affected. What matters to the macro properties of a joint is usually the size of coarse grained areas in weld metal or HAZ and their distribution, since coarse colonies usually have lower cleavage resistance [26,27] . Many welding factors influence this, particularly the welding process, welding position and the welding procedure and parameters, and the joint preparation. Maximum toughness in weld metal comes from small weld passes of low heat input which create the maximum austenite grain refinement. In the HAZ, narrow gap preparations help maximise grain refinement in addition to small weld passes.
Recognising this effect helps in understanding the behaviour of specific macro tests for structural integrity assessment such as those for fracture toughness TWI experience is that specimens notched and pre-cracked from the surface of the weld may show a different mean and distribution of results from those in which the notch and pre-crack are through thickness because the crack tips will sample coarse grained regions of different colony size and distribution. The surface notched specimens tend to show the lowest mean values and lowest 95% confidence limit.
5.3 Microstructural Issues
The importance of developing fine grain microstructures for good toughness in the weld metal and HAZ is well recognised. Ever since the 1970s, when the need to develop high levels of acicular ferrite in weld metals was understood, together with the key role of non metallic inclusions in nucleating the acicular ferrite  , the welding community has been attempting to manipulate the formulations of consumables (stick electrodes, wires, fluxes, gas shields) to achieve this result. There have been many successes in this respect although many questions remain. Today's consumables use Ti and B extensively but elements such as Mn, Ni and Mo have also been optimised and have brought about big improvements to cleavage resistance  .
The role of oxygen is also much clearer. This element is needed to ensure a minimum inclusion content for acicular ferrite nucleation, but the interaction of strong deoxidants in weld pools, e.g. Si, Al, Ti is complex  , and far from understood. What is clear is that moving towards joining processes such as laser and electron beam for high integrity plant creates new metallurgical issues since these are very low oxygen processes and it is not easy to develop the necessary inclusion content required for acicular ferrite nucleation in the fused zone  .
Non-metallic inclusions have a major effect on toughness in the ductile mode, and the volume fraction and particle size distribution present depends on the welding process, being highest for flux processes, such as submerged-arc and MMA welding. Deposit sulphur and oxygen levels both contribute to the non-metallic inclusion volume fraction and are the primary factors to control through appropriate specification of the consumables.
The very large increase in non-metallic inclusions in weld metal compared to parent steel and the HAZ is often forgotten when modelling micromechanical effects. It should be noted that inter-inclusion spacings are typically 1-2µm in say, submerged weld metal and this should be borne in mind when considering the mesh size for finite element studies.
6. Concluding remarks
The achievements of the welding community in relation to high integrity plant have been many and various in the last four decades. This paper has reviewed the main effects of welding, namely: flaws and their detectability, residual stresses and the complex macro-microstructures introduced with associated changes in mechanical properties. Looking back over, say, ten years, then most progress has been made in:
- the probability of flaw detection by radiography and ultrasonics (underpinned by improved theoretical models),
- the modelling and experimental measurement of residual stresses and the interaction of residual stress with primary stress.
The evidence is clear that welds matter to high integrity plant and always will. Looking forward five years, we can expect steady and successful progress in theoretical modelling for predicting the probability of flaws occurring in welds and their likely size distribution. There is also potential for much further progress in determining probability of detection curves for ultrasonic inspection of different flaw types.
More standard residual stress profiles will be generated for varying joints and residual stress modelling will be increasingly used in engineering critical assessments. Finally, steady progress will be made in theoretical modelling of macro/microstructures in weld metals and HAZs, eventually allowing prediction of some key properties of joints.
||TAGSI Report '(P96) 124' 'Defect distributions and the probability of large defects', April 1998.
||Chapman O J V:
||'Reliability & Risk in Pressure Vessels & Piping', PVP Vol. 251, ASME 1993, 81-89.
||Pollitt C G:
||'Radiographic sensitivity', BJNDT 1962, 4 (34), 71-80.
||'Limitations of radiography in detecting crack-like defects in thick sections', BJNDT 1962, 4 (4), 103-119.
||Munns I J and Schneider C R A:
||'The reliability of radiography of thick-section welds', presented at 'Review of Progress in Quantitative NDE', Montreal, 25-30 July 1999. Publ: New York, NY 10013, USA, Plenum Press, 2000; Eds. D.O. Thompson and D.E. Chimenti, Iowa State University. (Awaiting publication)
||'Plate Inspection Steering Committee (PISC),' (full reports), EUR 6371 EN, Vols I-V, European Commission 1979.
||'An assessment of the integrity of PWR pressure vessels', HMSO 1982.
||'European methodology for qualification' (2 nd Issue), EUR 17299 EN, European Commission DG III, 1997.
||'ASME boiler and pressure vessel code; Section XI: Rules for in-service inspection of nuclear power plant components', 1995 edition including 1996 addenda.
||Chapman R K and Toft M W:
||'Modelling the ultrasonic inspection of smooth and rough planar defects', 1999 IEE colloquium on 'New applications in modelling and inversion techniques for NDT'.
||Crutzen S, Fabbri L, Schneider C R A, Stacey A and Visser P:
||'Compilation of component inspection reliability data in the SINTAP project', OMAE 98 Conference, Lisbon.
||Toft M W:
||'Experimental studies of ultrasonic reflection from various types of misoriented defect', NDT-86 Conference, Newcastle, 193-206.
||Smith S D:
||'A review of numerical modelling of fusion welding for the prediction of residual stresses and distortion', TWI Members Report 437/1991, December 1991.
||Leblond J B, Devaux J C, Saillard P:
||'Elastic-plastic analysis of an underclad crack produced by reheat cracking in a PWR vessel'. In: Proc. 4 th International Conference on Numerical Methods in Fracture Mechanics, March 1987, 781-793.
||Webster G A and Ezeilo A N:
||'Principles of the measurement of residual stress by neutron diffraction'. PSI Proc. 96002, ISSN 1019-6447, New Instruments and Science around SINQ (Ed A.Furver), Paul Scherver Istitut, Switzerland, November 1986, 217-234.
||Leggatt R H, Smith D J, Smith S D and Faure F:
||'Development and experimental validation of the deep hole method for residual stress measurement'. Journal of Strain Analysis, 31 (3), 1996, 177-186.
||Handbook of measurement of residual stresses'. Society for Experimental Methods, Inc. Fairmont Press, Inc. 1996.
||'Determining residual stresses by the hole-drilling strain-gauge method'. ASTM standard E 837-92. American Society for Testing and Materials.
||Leggatt R H and Friedman I M:
||'Residual weldment stresses in controlled deposition repairs to 1 1 / 2 Cr- 1 / 2 Mo and 2 1 / 4 Cr-1Mo steels'. ASME PVP Conference , Montreal, Quebec, Canada, 21-26 July 1996.
||Leggatt R H:
||'Welding residual stresses'. In: The 5 th International conference on Residual Stresses (ICRS5), Linköping, Sweden, 16-18 June, 1997. Eds Ericsson T, Oden M, Anderson A, Linköping University, Sweden, Vol.1, 12-25.
||Barthelemy J Y:
||'SINTAP Task 4. Compendium of residual stress profiles. Final report'. Institut de Soudure, May 1999.
||BS 7910:1999: 'Guidance on methods for assessing the acceptability of flaws in fusion welded structures'. British Standards Institution, 4 th Draft after Public comments, April 1999.
||API: 'Recommended practice for fitness-for-service'. Issue 12, September 1999 (Special Release).
||Schwalbe K H et al:
||'Mismatching of interfaces and welds'. Second International Symposium, April 1996 GKSS Geestacht, Germany 1997.
||Robinson J L:
||Welding Research International 8 (6), 1978.
||Pisarski H G:
||'Metallurgy, welding and qualification of microalloyed (HSLA) steel weldments'. International Conference, Houston, Texas, USA, November 1990. American Welding Society.
||Pisarski H G and Slatcher S:
||'Offshore mechanics and arctic engineering'. 12 th International Conference, OMAE 1993, Glasgow.
||Abson D J, Dolby R E and Hart P H M:
||'Trends in steels and consumables for welding'. Proc. International Conference, 1979, London, The Welding Institute.
||Abson D J and Pargeter R J:
||International metals reviews, 1986 31 (4).
||Dolby R E:
||Steel World, 1998, 3 (1).