|TABLE OF CONTENTS|
2. EXPERIMENTAL PROCEDURE FOR PHASE I TRIALS
3. SUMMARY OF RADIOGRAPHIC RESULTS FOR PHASES I AND II
4. DISCUSSION OF RESULTS
5. DEVELOPMENT OF THEORETICAL MODEL FOR RADIOGRAPHY OF REAL METALLURGICAL DEFECTS
6. SUMMARY OF RECENT REVIEW OF DEFECT CHARACTERISTICS
This paper describes an extensive series of experimental studies of radiographic capability, for material thicknesses in the range 50-114 mm. These studies have been supported by surveys of the relevant parameters of real metallurgical defects to confirm the realism of the defects used. The results have been interpreted using a well-established, albeit simplified, theoretical model of the radiographic process, but further work on a more comprehensive theoretical model is in progress to provide more precise comparisons of theoretical and experimental results.
Considerable care has been taken to produce planar defects which realistically simulate those which might conceivably occur during welding of thick-section ferritic steel pressure vessels. One key feature is the orientation of the defect relative to the radiographic beam, and this can be controlled reasonably precisely when inducing defects in test specimens. Another crucial parameter for radiographic detection is the crack face separation (gape), which can only be measured by sectioning.
The results demonstrate the capability of radiography to detect a wide range of manufacturing defects of a size which might be of structural concern, even when they are substantially misoriented to the beam. Once sectioning has been completed (currently underway), the results will provide quantitative evidence of the combinations of defect size, tightness and orientation which can be detected.
The practical assessments of radiographic capability have been performed by TWI. Funding for this has been provided by the Industry Management Committee (IMC) which is supported by the nuclear licensees in Britain.
The radiographic capability for detection of planar defects is particularly difficult to assess because, apart from the defect's size, its orientation relative to the beam and the separation of the defect faces (the defect gape) are both crucial. A simple model (Pollitt, 1962) for predicting radiographic capability for smooth, parallel-sided, planar defects has existed for over 30 years and the predictions of the model confirm the trends shown experimentally. A fuller predictive capability should be possible when recent studies of the characteristics of real cracks (especially gapes) are combined with an improved mathematical modelling of radiography which is being developed in collaboration with BAM, Berlin.
The experimental work described in this paper provides direct evidence of the capability of radiography. It focuses on the radiography performed on the welds of the steel Magnox reactor pressure vessels (RPVs) during their construction in the 1 950s and 1 960s. However the improved understanding of the critical characteristics of defects, together with the theoretical modelling studies, should allow extrapolation to other situations without undue difficulty.
The main experimental trials described in this paper were undertaken at TWI and are Phases I and 11 of a programme funded by the Industry Management Committee (IMC) which is supported by the nuclear licensees in Britain. The review of the characteristics of real manufacturing defects was performed on behalf of Magnox Electric plc.
SUMMARY OF DEFECTS IN SPECIMENS A-D
|Flaw Type||Ultrasonic Testing/ |
|A||1||Longitudinal HAZ hydrogen crack||21.5||109||+ 18|
|2||Lack of sidewall fusion||22.5||73||-22|
|3||Longitudinal HAZ hydrogen crack||33.0||123||+ 17|
|4||Lack of sidewall fusion||17.5||83||-22|
|B||5||Lack of sidewall fusion||33.5||82||+ 23|
|6||Solidification crack||59.5||92||+4|| |
|7||Longitudinal HAZ hydrogen crack||50.0||132||- 19|
|8||Lack of sidewall fusion||40.0||71||+22|| |
|9||Solidification crack||41.0||85||-6|| |
|10||Longitudinal HAZ hydrogen crack||39.5||109||- 24|
|C||11||Solidification crack||21.0||87||- 20 t|
|12||Longitl. weld metal hydrogen crack||35.5||123||+30|
|D||14||Transverse weld metal hydrogen crack||20.7*||20||3*||0.029*|
|15||Transverse weld metal hydrogen crack||21.1 *||22||0*||0.015 *|
|16||Transverse weld metal hydrogen crack||21.3*||21||0*||0.002*|
The +- signs for the tilts indicate tilts in opposite senses in the block.|
* By sectioning
t Recent results indicate that this defect may have a tilt closer to 0°.
Great care was taken to select methods of introducing the defects which would give a natural metallurgical morphology while achieving the large sizes required (ideally at least 75 mm long by 15 mm through-wall): see Wooldridge et al (1997) for further details. Table I summarises the defects introduced into the four specimens. Through-wall extents, lengths and tilts are given, generally as measured by ultrasonic inspection, but where indicated by sectioning for Specimen D (where gape measurements were also made as shown). Sectioning of Specimens A to C, and manufacture of a new Specimen E containing further hydrogen cracks, are currently underway.
The radiographic techniques were designed to represent those used during construction of the Magnox pressure vessels and covered X-ray and Co-60 techniques. Spacer plates were used to simulate additional thicknesses of material, and angled shots allowed the range of defect orientations relative to the radiographic beam to be varied.
In all, 64 radiographs were taken of the four specimens (25 X-ray and 39 gamma), giving 263 defect/radiograph combinations. Each radiograph was independently examined by two interpreters qualified to PCN Level 11 radiography (welds). The interpreters had no prior knowledge of the defects, and the radiographs were mixed up before interpretation to reduce the chance of the interpreters being presented with consecutive radiographs of the same specimen. Each interpreter recorded the position and location of each detected defect, and classified its visibility as either easily visible (EV) or barely visible (BV).
Some of the key results are:
|(a)||Detectability generally decreased with increasing penetrated thickness or increasing misorientation angle.|
|(b)||The solidification cracks were always detected, even in those cases where the angle of misorientation exceeded 40°.|
|(c)||The LOSWF defects were always detected for those cases where the angle of misorientation was about 40° or less. The defects were generally not detected in those cases where the angle was about 60°, but these were nearly all for a large penetrated thickness of about 125 mm.|
|(d)||The long hydrogen cracks in the weld metal or HAZ (Specimens A to C), introduced using hardenable buttering and cellulosic consumables, were generally still detected at misorientation angles of 40° or more. The shorter transverse weld metal hydrogen cracks in Specimen D proved most difficult to detect, with several non-detections for penetrated thicknesses of 100 mm or more, even at small misorientation angles. (A non-detection at 75 mm was due to obscuration of the image of the transverse crack by an IQI wire.)|
Radiographic detectability of all planar defect types in the
Phase I study, with increasing defect through-wall size:
results from Interpreter 1 (above) and 2 (below).
The poorer detectability of the hydrogen cracks in Specimen D is almost certainly due to their smaller gape (crack face separation). Gape is known to be an important factor affecting detectability, as discussed in the next Section. The gapes of the transverse weld metal hydrogen cracks in Specimen D have been measured to be typically 30 am or less (see Table I). The gapes of the defects in the other specimens will be determined when the specimens are sectioned, and it will then be possible to examine detectability quantitatively as a function of gape.
Additional data on gape and other defect parameters, and on radiographic capability, has been generated in an auxiliary Phase II study. This study was based mainly on an existing TWI database of NDT responses, augmented by additional radiography of specimens containing hydrogen cracks and by sectioning of some of the specimens. The results are generally consistent with the main study described above (see Section 7 below for more details), though many of the defects were too small to be directly relevant to the Magnox RPV application.
The effect of defect TWE on detectability, for the Phase I study, is shown in Figure 1. Although performance appears to be poorer at the smaller TWEs (15-25 mm), most of the non-detections there are in fact of the three transverse weld metal hydrogen cracks in Specimen D. These cracks are different in several ways from all the other defects: they are transverse, they are only about 20 mm long (Table I), compared with at least 70 mm for the other defects, and they were made in a different way in different material. We cannot therefore conclude that the nondetections of these cracks are due solely to their relatively small TWE. The results in Figure 1 thus suggest at most only a very modest improvement in radiographic detectability with increasing TWE for the defects in this study (all larger than 15 mm TWE).
Figure 2 shows Pollitt predictions for the maximum angle of misorientation which slots of various gapes and TWEs can reach before they become undetectable by radiography. Results are given for two material thicknesses of 75 and 100 mm, and assume radiography at a BWRA step-hole IQI sensitivity of 1 .66% (as specified for most of the Magnox RPV radiography). A fixed total unsharpness of 0.6 mm is assumed for simplicity; in practice this will have varied in both the Magnox RPV and the TWI radiography, for example because various source/focal spot sizes and source/focus-to-film distances were used, thus varying the geometrical unsharpness.
Figure 2 Predictions of Pollitt theory for radiographic detectability of smooth parallel-sided slots in material of thickness T = 75 or 100 mm, assuming a total unsharpness of 0.6 mm and a BWRA IQI sensitivity of 1.66%. The through-wall extent of the slot is measured perpendicular to the component surfaces, not along the face of the slot. |
The results of Figure 2 show that:
|(a)||detectability improves with increasing gape (more tilt can be tolerated at the larger gapes);|
|(b)||detectability decreases with increasing penetrated thickness (less tilt can be tolerated at 100 mm thickness than at 75 mm);|
|(c)||for misoriented defects of TWE greater than about 10 mm, the TWE does not significantly affect detectability (the curves flatten off).|
These qualitative trends are also generally seen in the experimental results. We cannot yet make many quantitative comparisons between Pollitt theory and the experimental results, since most of the defects have not yet been sectioned and we do not have data on gape.
The model has the following elements:
|(a)||A pre-processing module, which generates specific realisations of rough defects of variable gape from a small number of statistical parameters.|
|(b)||A CAD interface, which enables this rough defect to be positioned in a component of possibly complex geometry, in an arbitrary position and orientation.|
|(c)||A radiographic simulation module, which predicts the radiographic image of the defect in the component. The source and film positions and characteristics can be specified by the user. The module is based on a ray tracing algorithm in which rays are traced from each source position to each film position and are attenuated depending on how much material they pass through. Multiple scattering is currently modelled using experimental build up factors, though an alternative mathematical approach is being pursued by BAM and collaborators in Minsk.|
A working prototype of the model is now available. The simulated radiographs which it generates for cracks look realistic, and encouraging agreement with experiment has been obtained for a simple wire IQI and a step-wedge. The model is now being developed further, one problem being to remove dark artefacts which sometimes appear on the simulations. The aim is to compare the model predictions with the real radiographs from the experimental study, using the measured defect statistics from the sectioning of Specimens A-E as input to the model.
Examples of the output of the review are shown in Figures 3 and 4, for centre-line solidification cracks and longitudinal weld metal hydrogen cracks respectively. In each case the typical gapes of the measured cracks are plotted against their through-wall sizes. Distinction is made between artificial and natural defects.
|Figure 3 Review of literature and TWI archives: average gape versus through-wall size for centre-line solidification defects.||Figure 4 Review of literature and TWI archives: average gape versus through-wall size for longitudinal weld metal hydrogen cracks.|
Once the sectioning of Specimens A-E is complete, the gapes of the defects in these specimens will be compared to those found in the review, so that the realism of the defect gapes in Specimens A-E can be assessed.
The simplified theoretical predictions (Figure 2) show that the detectability of large planar defects is determined primarily by the orientation to the radiographic beam and the opening between the crack faces (gape). For defects exceeding about 10 mm in through-wall extent, increasing defect TWE is less significant. All these predictions are consistent with the experimental results - although for many defects we can currently only guess the actual gape from a knowledge of the defect type.
Many large planar defects have significant crack openings (gapes), which aid detection by radiography. Solidification cracks are generally open and were detectable in the Phase I and 11 studies (for through-wall extents exceeding 4 mm) at up to 40° misorientation. Large lack of fusion defects are often open and again were detectable in the Phase I and 11 studies at misorientations up to about 40°. The hydrogen cracks appear to fall into two categories: the short, mainly transverse, cracks created by a high-restraint groove weld and only partly drying the welding rods, and the long defects created using hardenable buttering and cellulosic weld metal. The long defects created by the latter process were generally detectable at angles up to about 40° misorientation, whereas the short cracks made by the former process could be non-detectable even for angles of misorientation less than 5°. Sectioning and further analysis of the characteristics of these defects is required before one can fully assess how well the defects simulate the characteristics of any large, extended hydrogen cracks which might be postulated to occur in practice.
Production of test pieces with realistic defects requires considerable skill if the defect parameters are to be representative of those which might occur in practice. The literature review of defect morphology described in Section 6 provides valuable data against which to compare the results of sectioning the defects in Specimens A-E.
Realistic defects are not easy to describe with a few simple parameters such as through-wall extent and crack gape because the characteristics vary from one place to another along each defect. For this reason their radiographic inspection cannot be accurately modelled theoretically using the Pollitt model for parallel-sided smooth slots. The BAM model described in Section 5 provides a promising tool for modelling realistic defects.
Wooldridge A B. Chapman R K, Woodcock G S. Munns I J and Georgiou G A (1997), Reliability of radiography for detection of planar manufacturing defects in thick section welds. Insight, 39 (3), pp 139-147 (March 1997).
Wooldridge A B (1996), Demonstrating the capability and reliability of NDT inspections, Proc. of 14th World Conf on NDT, Vol 1, pp 169-173, New Delhi, Dec 1996.