Inspecting fusion welds - a standards evaluation

TWI Bulletin, January/February 1995

George Georgiou graduated from Imperial College in Mathematics in 1972 and stayed on for a further year as a research student studying theoretical fluid dynamics. He was a full-time Mathematics lecturer at Tottenham College of Technology until 1983. In the meantime, he gained his PhD, working as a part-time student at City University London on theoretical fluid dynamics in 1982. Since 1990 he has worked at TWI on a variety of NDT-related problems. In particular he is working on the NDT of plastics and adhesives and is currently involved with BSI and CEN Committees who are drafting standards for the ultrasonic inspection of welds. He is currently a member of The British Institute of Non-Destructive Testing Technical and Research Committees. At TWI he is Manager of the Numerical Modelling Section in the Structural Integrity Department.

Peter is currently a Vice-President of The British Institute of Non-Destructive Testing and also sits on the Institute's Council, Finance and General Purposes, President's Advisory and Technical Committees. He joined the Institute in 1982 and was elected a Fellow in 1993.

In this work George Georgiou and Peter Mudge bring together recent analysis carried out by TWI to evaluate proposed European standards for ultrasonic non-destructive testing (NDT) of welds in ferritic steels.

This has involved comparing the performance of the European (CEN) proposals with other established NDT standards on a large number of defect test specimens. Principally, three CEN documents have been evaluated: the inspection document, now approved as a prEN (draft European standard), plus the acceptance criteria and the flaw characterisation documents which, at the time of writing, were being prepared for prEN approval.

In carrying out this work, a large database of ultrasonic responses from flaws has been generated and dedicated software written to analyse it. This software includes features such as automatic checking for database entries, routines for checking the sentencing of the flaws, and other general filtering and computational routines.

This paper reports the findings and views of TWI on this comparative evaluation study.

NDT standards used in this study

Three proposed CEN ultrasonic testing standards were originally developed by the ultrasonics working group of the technical committee responsible for the NDT of welds, i.e. TC 121/SC5B/WG 2. These were committee drafts on restricted circulation, but are referenced here as:

TC121 N97:

Ultrasonic Examination of Welded Joints

TC121/SC5B/WG 2 N70:

Ultrasonic Examination of Welded Joints in Steel, Acceptance Levels

TC 121/SC5B/WG 2 N64 Rev.:

Characterisation of Defects in Welds, Planar or Non-Planar Defects

The public comment stage for the first document (N97) has been overdue for some time. The original target date set by TC 121 was August 1991, with a revised target date of September 1993. N97 specifies the method for manual ultrasonic examination of full penetration ferritic welds (>8mm thick). Seven joint types are catered for, including butt joints in plates and pipes, which were the focus of this study. Three examination levels were specified which correspond to 'moderate', 'high' and 'highest' probability of detection respectively.

The second document (N70) has been finalised and submitted for prEN approval. Two acceptance levels are identified, B and C, compatible with quality levels B and C respectively, in EN 25817, ^[1]   in which flaws are accepted or rejected principally on the basis of their ultrasonic echo amplitude and their length. The same acceptance criteria apply to both quality levels B and C, but examination levels 2 and 1 respectively are specified (from document N97) for weld inspection. Other features are that 2 or 4MHz probes may be used, depending on weld thickness, the evaluation level is 33% DAC relative to a 3mm side-drilled hole (SDH), or equivalent DGS, and that characterisation of flaws is not required (unless agreed between contracting parties) if they are less than 2 x weld thickness in length.

The third document (N64 Rev.) has been ratified by WG2, but has not progressed further. This document is based on earlier research work by the Institut de Soudure ^[2]   on buried flaws, i.e. at least 5mm below a non-ground surface. To characterise a flaw, a flow chart (or 'cascade') with five stages is followed, with each stage having a specific aim: first, avoid classifying low amplitude flaws; second, classify high echo amplitude flaws as 'planar'; third, to classify lack of fusion flaws; fourth, to classify inclusions; and fifth, to classify cracks.

The performance of the above proposed CEN standards were compared with the following established standards (procedure followed by the relevant acceptance criteria in italics):
Radiography: BS 2600; Parts 1 and 2,
BS 5500: 1994 Table 5.7(1)
Ultrasonics: ASME V, Article V,
ASME VIII
Ultrasonics: BS 3923, Part 1, Level IIB,
BS 5500: 1994 Table 5.7(2)

Weld specimens used in the study

A total of 41 butt weld defect test specimens, i.e. 12m of weld, were inspected and sentenced according to the NDT standards above. The majority of the defect specimens were either off-cuts from production welds or were donated by industry. Only seven were manufactured defect specimens. The distribution of the 41 test specimens across the thickness range 8-100mm, is illustrated in the Figure.

Database software

A total of 490 flaws were examined. The details of each of these have been entered into a database, with up to 38 fields of information for each flaw. Dedicated software has been written which checks for rogue field entries, checks the sentencing, whenever appropriate, and enables the user to carry out a wide range of filters and computations for tabular and graphical output.

Results of the comparative study

Detection (CEN/TC 121 N97)

When working to N97, a significant number of flaws were not evaluated ( i.e. not detected) compared to BS 3923 Part 1, Level IIB and ASME V (Article V). Table 1 gives the levels of detection in % terms, and shows that N97 detected about 35% less than the other two standards. The evaluation level for the CEN procedure is currently 33% DAC relative to a 3mm SDH ( i.e. DAC -10dB). If a lower evaluation level is used ( i.e. 20% DAC or DAC -14dB) then about 10% more flaws will be detected (shown in brackets in Table 1). This low detection rate is attributed predominantly to the number of scans in examination Level 2 of N97, and not to the evaluation level. The principal differences between N97 and BS 3923 Part 1, Level IIB are given in Table 2 for longitudinal defects in butt welds. The number of scans can differ by as much as a factor of four.

Table 1 Levels of detection in % (values in brackets for N97 correspond to a DAC -14dB evaluation level)

Table 2 Principal differences in N97 (Level 2) and BS 3923 Part 1, Level IIB, for longitudinal defects in butt welds

Accept/reject performance (CEN/TC 121/SC 5B/WG 2 N70)

In N70, recommendations are made for probe frequency, evaluation levels, and proximity rules for individual flaws close enough to be considered as one. All these can potentially affect the flaw reject rate, and the database was analysed to assess precisely what effect they have. The probe frequency has relatively little effect on the overall reject rate. With regard to the evaluation level, 50 flaws were found to have echo amplitudes between DAC -10dB and DAC -14dB ( i.e. accepted by default). Table 3 shows that while the majority of these are accepted by the other criteria, the ones rejected include cracks lack of sidewall and lack of root fusion.

Table 3 Sentencing of 50 flaws accepted by default by N70( i.e. DAC -14dB < echo height < DAC -10dB)

Table 4 Accept/reject rates (%) for the whole flaw population (490 flaws). The bracketed values represent the accept/reject rates with flaw characterisation according to N64 Rev.

Table 5 Accept/reject rates (%) for 45 flaws; 7 cracks, 1 lack of penetration, 1 lack of root fusion and 36 lack of sidewall fusion as detected by radiography (CEN values are with characterisation according to N64 Rev.)

The most significant result is the relatively low reject rate using N70 over the whole flaw population, in comparison with the other acceptance criteria. Table 4 gives the % reject rate using N70 for testing carried out with a 2MHz probe, a 4MHz probe and a combination of 2MHz and 4MHz probes. The bracketed results in Table 4 correspond to the reject rates when flaw characterisation is carried out according to N64 Rev. It is important to note that the comparatively low reject rate according to radiography, is a direct consequence of a large number of linear inclusion type flaws in the thicker specimens ( i.e. >50mm), see the Fig., which were undetected by radiography and hence accepted by default. This also highlights the potential danger of using radiography as a benchmark for ultrasonic results when comparing NDT of thick weld specimens.

Another significant result is that the reject rate for serious flaws using N70 is also relatively low, even with flaw characterisation. The serious flaws here are those that radiography rejects outright, irrespective of the flaw size. Table 5 illustrates the reject rates using ultrasonic testing for 45 such flaws.

Characterisation (CEN/TC 121/SC 5B/WG 2 N64 Rev.)

When one looks into the cascade procedure in more detail, the terms planar and non-planar are misnomers. For example, one of the criteria for a planar flaw is that it has a high echo amplitude ( i.e. > DAC +6dB). On this basis, there will be a number of flaw types which fall into the planar category which are not notionally planar. This study has shown this to be the case. Further, a number of seemingly innocuous flaws were placed into the planar category, which then become immediately rejectable. This is one reason for the significant impact which flaw characterisation has on the reject rate, illustrated by the bracketed values in Table 4.

Another finding is that the cascade procedure is probe frequency dependent. That is, flaws that are characterised as being a particular type using a 2MHz probe, will be characterised differently using a 4MHz probe. However, the impact of probe frequency on the reject rate is not significant, as already shown in Table 4.

Conclusions and recommendations

• Level 3 and Level 2 examinations in N97 are used for weld qualities B and C respectively.
• Flaw characterisation is extended to all flaws and includes more categories than just 'planar' or 'non-planar',
• Flaw acceptance thresholds for quality levels B and C are re-assessed in the light of these results,
• Any response during the public comment period for each of the CEN proposals should consider them collectively, rather than in isolation.

References