Graham Edwards BSc (Hons), DMS, MInStNDT, MITD is a Principal Research Engineer in the NDT Research Department. As well as leading the specialist site services team he also works on NDT of non-metals and in expert systems.
The pressure-vessel manufacturing industry still provides a demand for traditional welding and non-destructive testing (NDT) methods. It is a conservative industry, which justly prides itself on the quality of its product. Yet economic pressures dictate that it must introduce new technologies in a drive towards greater production efficiency. For welding this may involve new processes and for NDT it may involve replacing radiography with ultrasonics. Graham Edwards looks into the subject of accept-reject criteria.
Most welds contain flaws, but whether these are serious enough to constitute a risk of failure during the service life of the pressure vessel and are therefore what might properly be termed defects, is of utmost importance. Consequently, accept-reject criteria have always erred on the side of caution, balanced by the fact that inspection and repair costs must not become uneconomic.
In Britain, BS 5500, the specification for the manufacture of unfired pressure vessels, has always maintained a high level of quality. It contains a table of criteria which can be used to accept or reject weld flaws, and which has been used successfully for many years - and is indeed found in many other manufacturing specifications. The criteria may sometimes appear rather arbitrary, but experience has shown that flaws that cause an eventual failure will not be accepted. At the same time, the reject criteria do not give rise to excessive repair rates.
However, the criteria used in the Table cause one serious difficulty for volumetric inspection: they are couched in radiographic terms which cannot be applied to ultrasonic inspection. The result has been either a rather vague attempt at applying the criteria to the results of ultrasonic tests, or adopting alternative specifications, such as ASME VIII, which give ultrasonic accept-reject criteria.
With the use of ultrasonics in volumetric inspection of pressure-vessel welds increasing, the BS 5500 committee saw that the situation could not be allowed to continue. TWI was therefore approached to set up a research project, sponsored by its members with an interest in pressure-vessel manufacture, with support from the Department of Trade and Industry. The project has been running for over 12 months and is now almost complete. It has provided not only data which can be used to set new ultrasonic accept-reject criteria for flaws, but also an insight into the practical application of BS 3923, which is called upon by BS 5500 for manual ultrasonic weld testing.
Table 5.7 of BS 5500
The acceptance levels for flaws in BS 5500 are set out in Table 5.7. It uses criteria which cannot be applied to the results of ultrasonic tests. The interpretation of signals from an ultrasonic test is far more subjective than the interpretation of images in a radiograph and, although the ultrasonic method is more sensitive to certain types of flaw, the use of such definitive terms as lack-of-interrun fusion or linear porosity is possible only in radiography.
When sentencing (making an accept-reject decision on) ultrasonic test results, the greatest problems can arise in the criteria used for cavities.
The limit set for localised porosity, for example, is 2% by area. Clearly, this measurement cannot be made ultrasonically and, even in a radiographic image, only an estimate of the density of pores can be made, using a comparator. Even for the simple case of isolated pores or worm-holes, the accuracy of ultrasonic size measurements cannot be relied upon to identify (for example) a 3.0mm diameter pore in a 25mm plate.
Ultrasonics can never distinguish copper or tungsten inclusions from slag inclusions. Copper inclusions are dealt with much more severely than other inclusions by the Standard, and can be distinguished in a radiograph by their greater contrast.
Of course, not all of the Table can be applied to radiography either. Profile defects such as undercut can be identified on a radiograph, but depth measurements have to be made physically. There are many more discrepancies, however, which make Table 5.7 unsuitable for ultrasonic tests - indeed, the only recognition of the requirements of ultrasonic testing is in a footnote to the Table, which allows any indication with a through-thickness dimension of 1.5mm or less to be ignored. The through-thickness dimension cannot, of course, be measured on a radiograph, but can be assessed - though with a high degree of error - ultrasonically.
Table 31 of BS 3923
Table 31 in the most recent edition of BS 3923 (issued in 1986) recommends accept-reject criteria for flaws or imperfections to be used with its test procedures. It recognises ten types of imperfection in six categories: point, threadlike, volumetric, planar, multiple and root profile.
In so doing, the Standard has removed from the ultrasonic test operator the responsibility of identifying the type of weld flaw. The identification is made on the basis of the echo dynamic pattern given by the imperfection in response to probe movements. The information needed from each type of imperfection includes maximum echo height, length and, for more critical examinations, the through-thickness height. The Standard recommends the information to be recorded, but does not give the limits which can be used as accept-reject criteria for imperfections; these will depend on the application.
The research project at TWI has therefore been the first systematic exercise to arrive at accept-reject criteria according to the requirements of BS 3923.
Ultrasonics versus radiography
The results of ultrasonic tests cannot be equated with radiography for two reasons. The first has already been mentioned: whereas a radiograph provides an image of a flaw which is often similar to its visual appearance, the A-scan display of ultrasonic test signals provides information which can be interpreted as a flaw only after careful consideration of the position, type and amplitude of the signal. The ultrasonic technique used in industry is far less sophisticated than that used clinically where very realistic images are obtained.
The second reason is that the two methods have different sensitivities to different types of flaw. Generally, radiography is more sensitive to volumetric flaws such as inclusions and porosity, while ultrasonics is more sensitive to planar flaws such as lack of fusion and cracks.
Radiography is therefore more suitable as a quality-control tool. A radiograph will easily reveal the slag inclusions and porosity indicative of poor workmanship. Ultrasonics is more suited to critical fitness-for-purpose examinations, where volumetric flaws may sometimes be ignored and greater emphasis placed on sizing planar flaws.
So why replace radiography with ultrasonics in the quality control of pressure vessels? First, there are the health hazards of radiation to be taken into account, and the consequent downtime incurred while radiography is carried out. Second, radiography is more expensive than ultrasonics if long lengths of weld are to be tested. In addition, ultrasonic procedures have improved and there is an increasing interest in fingerprinting welds ultrasonically so that they can be re-examined after they have been in service. There are, moreover, complex weld geometries - nozzle and branch welds for example - which cannot be tested radiographically.
Work at TWI
The representatives from the pressure-vessel industry in the sponsor group set the principal objective of the work as derivation of accept-reject criteria for ultrasonic tests which, if applied to normal production welds, would not lead to an increase in the repair rate above that which would have been attained with radiography and Table 5.7. Moreover, the ultrasonic test procedures had to be simple enough not to increase the inspection costs.
This objective was therefore straightforward and somewhat different from that of most research projects involving comparisons of ultrasonics and radiography. It was accepted that ultrasonics was going to detect a different population of flaws from radiography. There was no need for any detailed appraisal of this discrepancy; the main concern was how the accept-reject criteria, along with the ultrasonic test procedure, would deal with the vast majority of flaws which are innocuous.
The programme of work involved testing several tens of metres of weld with radiography followed by ultrasonics to both BS 3923, Levels 2B and 3, and ASME.
In the initial phase of the work to prove the test procedures one of the problems was to achieve the correct sensitivity with the procedure. For example, it is not widely recognised that adequate sensitivity from ASME procedures can be achieved only by using 2.25MHz probes, and not the 5MHz probes more widely adopted in industry. Since low-frequency probes are rarely used in the UK it was decided to discontinue the ASME procedures and concentrate on the BS 3923 procedures.
To evaluate the sensitivity of the technique, that is to say that part of the procedure which governs probe-angle selection and scanning directions, data were collected from each probe scan at half and full skip and all test surfaces during the preliminary scans. This is not a normal requirement of the test procedure. The sensitivity of different probe angles to the same defect, at half and full skip positions and from different sides of the weld, could thus be evaluated before more detailed scans were used to locate and measure the flaw.
These data have provided useful information for improving the test procedure.
With the final test data, the measurements of signal amplitude, length and through height could be compared for a large population of flaw types including transverse cracks, lack of fusion, porosity and slag inclusions.
The accept-reject criteria applied to this population could then be altered, for example by raising the amplitude threshold level for rejecting planar flaws from 50% to 100% distance amplitude correction (DAC) level, and the consequent effect on repair rate determined.
European standards
The forthcoming harmonisation of European standards in 1992 put a new emphasis on the project. The present division in Europe between those ultrasonic test procedures which use flat-bottomed holes as reference reflectors and those which use side-drilled holes has made it very difficult to set common accept-reject criteria. The results of the project should have a strong influence on the decision as to which method is adopted.
