Paper presented at IEE colloquium 'New applications in modelling and inversion techniques for NDT', 29 January 1999, London, UK.
1C R A Schneider and I J Munns are with TWI, Abington Hall, Abington, Cambridge, CB1 6AL - Tel: +44 (0)1223 899000
2G A Georgiou (formerly with TWI) is director of Jacobi Consulting Ltd, 57 Ockendon Rd, London, N1 3NL
3R K Chapman and G S Woodcock are with British Energy plc, Barnett Way, Barnwood, Gloucester, GL4 3RS
This paper describes the experimental validation of a simple model of radiography, first published by Pollitt in 1962, which treats flaws as smooth, parallel-sided slots.
Six thick-section specimens were manufactured to represent welds in Magnox pressure vessels (now operated by BNFL Magnox Generation). The welds contain 16 large, realistic, planar defects. They were radiographed under various exposure conditions and evaluated by two interpreters. The specimens were then sectioned to determine defect size, orientation, gape and roughness.
The experimental data show variations in detectability that are strongly correlated with theoretical predictions. In almost all cases, Pollitt theory is either accurate or pessimistic. We have used the experimental data to derive statistical models for the reliability of radiography. Such models can be used to estimate the probability of detecting a defect (with associated confidence limits), from knowledge of its size, orientation and other relevant parameters.
During the construction of the steel Magnox reactor pressure vessels (RPVs) in the 1950s/60s, radiography was the main method of non-destructive testing of the welds. These vessels are now operated by BNFL Magnox Generation. This paper is based on a programme, funded by both the UK nuclear licensees, to establish the reliability of radiographic examination for large planar defects in welds 50-114mm thick.
In the main part of the programme, six test specimens were manufactured containing 16 metallurgically realistic large planar welding defects (>15mm in through-wall extent). The specimens were radiographed under various exposure conditions, including angled shots and the use of spacer plates to simulate thicker specimens. The thickness range was 50 to 114mm. This gave 129 radiographs in all, comprising 308 flaw-radiograph combinations, each of which was evaluated by two interpreters.
TWI sectioned the specimens to determine defect sizes, orientations, gapes and roughness. The performance of the radiography was then analysed as a function of these defect parameters and of radiographic parameters, such as source type, unsharpness and film type.
The experimental data collected by TWI shows variations in detectability with such parameters as defect size, gape and orientation, which are broadly in line with what is expected on simple theoretical arguments (e.g. increasing detectability with increasing gape). However, it is difficult to display the results in a form that enables a more quantitative comparison with theory. This is because of the large number of interrelated parameters that vary from defect to defect or exposure to exposure, making it difficult to isolate the effects of individual parameters.
To help clarify the underlying physical processes governing flaw detection, TWI has developed an 'index of detectability' based on a simple model described by Pollitt 
. The index is strongly correlated with the experimental results, and there is a simple functional relationship between the probability of defect detection and the index, for the data considered in this study. For example, Figure 1
illustrates Interpreter 1's results, together with the fitted relationships.
Logistic regression analysis confirms previous findings that Pollitt theory is broadly pessimistic in predicting radiographic detectability, and enables the extent of this pessimism to be quantified in terms of the probability of defect detection.
The analysis shows that the index has a far stronger influence on the detectability of planar flaws than any other parameter studied. This confirms its value as a derived parameter. However, for a given index of detectability, additional factors due to human performance (the interpreter) and flaw morphology also seem to be important - the variation in detectability with flaw morphology partly manifests itself as an apparent variation with different flaw types. A recent experimental study of human performance has shown, however, that there is little variability between different interpreters in the detectability of large planar defects. This study suggests that the apparent variations of human performance in the earlier study were, in fact, largely due to different qualitative interpretations of defect visibility, rather than being related to defect detectability. Other factors that seem to have an effect on detectability, over and above that predicted by Pollitt theory, are unsharpness, effective source-to-film distance and penetrated thickness. However, these are weak effects and may be due to inaccuracies in the existing form of Pollitt theory. Further work, based on an alternative, more rigorous, theoretical formulation is already in progress.
The models described can be used to estimate how easily a defect can be detected, from knowledge of its size, orientation and other relevant parameters. In principle, the probability of detection and an associated confidence interval can be estimated. However, it is possible that the models may be biased by the choice of defect population in the experimental study, which comprises just 16 large planar flaws, all in thick-section welds; caution needs to be exercised in extrapolating the results to other flaws.
This paper is published with the permission of British Nuclear Fuels Ltd and British Energy plc. The work was funded by the Industry Management Committee (IMC), which represents British Nuclear Fuels Ltd, British Energy plc and the Health and Safety Executive.
- Pollitt C G: 'Radiographic sensitivity'. British Journal of NDT September 1962 4 (3) 71-80.