Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Inspection for Root Flaws in Riser Girth Welds


Reliability of Inspection for Root Flaws in Riser Girth Welds

J R Rudlin, C R A Schneider and G R Razmjoo

Paper presented at 23rd International Conference on Offshore Mechanics and Arctic Engineering OMAE, Vancouver, Canada 20-25 June 2004
Paper 51523


The fatigue loading on deep water risers leads to a requirement for the detection of small root flaws during manufacturing inspection. Mechanised welds for offshore pipelines are also subject to extreme loads during laying, leading to a similar requirement. Automated Ultrasonic Testing using zonal methods have been adopted as the inspection method of choice for these inspections, but there is little information in the public domain regarding the expected reliability of the various systems available. Extensive individual inspection qualifications are carried out for each installation. The extent of these could be reduced by the availability of such background information.

This paper reviews data from joint industry projects in the area carried out by TWI, and compares results from these with such data as is available in the public domain. An analysis of future requirements and capability of currently available theoretical models for extending the range of qualifications is also given.


AUT: Automated Ultrasonic Testing by zonal discrimination method
CEA: Commissariat a l'Energie Atomique
CIVA: Proprietary name for software by CEA (France)
ECA: Engineering Critical Assessment
FBH: Flat-bottomed hole
IPLOCA: International Pipe Line and Offshore Contractors Association
JIP: Joint Industry Project
P-Scan: Proprietary name for ultrasonic scanning system manufactured by Force Technology, Denmark
PANI: Programme for the Assessment Non-destructive testing in Industry (UK Health and Safety Executive Research project)
RTD: Rontgen Technische Dienst b.v.(Inspection Company)
TOFD: Time of Flight Diffraction
UT: Ultrasonic Testing


Risers are dynamic fatigue sensitive structures that are usually regarded as the most challenging part of a deep water development. They are subjected to fatigue loads imparted by vessel motion and by wave and current actions. The welds in risers are more critical than those in conventional jacket structures as riser girth welds have no redundancy and failure of one weld represents failure of the whole riser, with serious environmental safety and cost implications. The restriction to single-sided welding, in conjunction with the less stringent control inherent in seamless pipe manufacture, makes the welding of risers more difficult and less reliable. TWI has carried out a series of experiments in a fatigue programme (TWI JIP 5680 Fatigue Performance of Girth Welds Made from One Side) which commenced in 1996. The initial results of this project have shown that failure of the risers in fatigue mode is most likely due to crack initiation at the weld root. This will occur much earlier if manufacturing flaws are present. Therefore detection of flaws in the weld root is an essential part of ensuring the integrity of the riser. Since access directly to the weld root is not an option, the inspection must be carried out from the outside. Automated Ultrasonic Testing by zonal discrimination (AUT) was developed as a high speed inspection method for inspection of pipeline production welds and has been suggested as a sufficiently sensitive means of inspecting single-sided girth welds.

This paper describes work carried out by TWI to investigate the capability of AUT to perform this task.

General review of ultrasonic methods for weld inspection

Ultrasonic testing (UT) as a method of detecting flaws in welded structures has been around for some 40 years. The technology has developed from a simple manual technique to more sophisticated automated techniques. In the 1970sautomated methods of scanning the weld and displaying the data were introduced, with a new sizing technique called time-of-flight diffraction introduced towards the end of the same decade. Multi probe AUT systems were introduced in the1980s, and phased array systems as recently as the late 1990s. The principle of each of these systems is described in more detail below.

Manual Ultrasonic Testing

Figure 1 shows the basic principles of manual UT of a butt weld. The display, known as an A-scan, is a plot of signal amplitude against time. The procedure requires that the operator uses ultrasonic beams with different angles to the surface to build up a picture of any flaws in the interior of the weld. The use of many different angles is considered essential to detect badly oriented flaws. Manual UT has considerable flexibility for different geometries,but its reliability is limited by the possibility that an operator will fail to scan 100% of the area or will be unable to satisfactorily interpret the A-scan signals. Sizing of flaws is typically carried out by measuring the probe movement for the presence of the signal until it falls by a fixed amount (usually 6dB or 20dB), as the probe is traversed across the flaw. Another method is called the Maximum Amplitude technique ('Max Amp'). In this case the probe is moved to maximise the amplitude of the first and last peak signals from an A-Scan. This should pick up the tip-diffracted signals from the flaw extremities.


Fig. 1. Basic principle of manual ultrasonics - probes are scanned to cover whole volume of weld with different angles

While manual UT is very flexible in individual applications, for routine manufacturing inspection on a riser or pipe containing many single-sided girth welds, it is both very slow and can suffer from reliability problems due to lack of scanning control and operator fatigue.

Automated Ultrasonic Testing

Automating the manual UT process is carried out by means of scanning frames, which move a single probe over the surface in a fixed raster pattern. The probe angle is changed by using different probes, all of which need to complete the full raster, to give all the necessary angles for inspection. Use of this fixed raster enables a comprehensive plot of the position of reflections to be made. This results in a series of projections of a weld showing the flaw positions called B-scan (view from end of weld, or cross section), C-scan (view looking through weld from cap side, or plan view) and D-scan (view from side, showing weld length and height). Figure 2 shows these different views possible with automated inspection.


Fig. 2. Views from an automated inspection


The results from different angle probes are again used to help build up the picture of a detected flaw. Flaw sizing is carried out in a similar manner as manual UT (by probe movement).

A possible limitation for flaw analysis is the fact that the probes cannot be skewed to optimise a flaw signal as is possible in a manual inspection.

This method of inspection is generally more reliable than manual UT, but is still too slow for production application.

Time-of-Flight Diffraction

The time-of-flight diffraction (TOFD) method ( Fig.3) of inspection was introduced as an improved means of sizing flaws. Two probes with broad beams are placed on either side of a weld, one acting as a transmitter, the other as a receiver. The time of flight of any reflection from one probe to the other is monitored. It was shown that an ultrasonic beam generally produces a diffracted signal from the extremities of a flaw as well as a reflected one. The diffracted signals from the extremities therefore appear as signals arriving at different times at the receiver. By carrying out the geometrical calculations an estimation of the through thickness dimension of a flaw can be obtained.


Fig. 3. Time-of-flight ultrasonics

The data obtained are displayed as a scan along the length of a weld, showing its thickness (similar to the D-Scan above). A strong reflected backwall signal shows the wall thickness, and a 'lateral' wave is transmitted along the surface, which can also be seen.

TOFD is a very rapid method of inspecting whole volumes. It is restricted where the time differences are small and the different signals cannot be resolved in time (for example, for small flaws, flaws close to the backwall and flaws close to the inspection surface). The amplitude of the diffracted signals can also be quite low and dependent on the flaw tip geometry.


AUT is now the commonly used acronym for automated ultrasonic testing using the method described below. Figure 4 shows the principle of AUT systems. The concept of AUT was introduced as a means of very rapid inspection of pipe butt welds. This is in contrast to the automated ultrasonic system. In this case the ultrasonic beams are restricted (focussed) to examine only a small area of the weld, and many probes are used, each directed at individual parts of the weld (zones). The zones will include those specifically focussed and angled to detect lack of sidewall and root fusion, together with other wide-angle beam(s) to detect flaws in the weld volume. By using this method only one circumferential scan of a weld is necessary for complete coverage.


Fig. 4. Inspection of weld by single probes inspecting particular weld zones (only three zones shown)

The length of a flaw along a weld is estimated by using the probe movement. There are a number of strategies for dealing with the through thickness dimension. Some operators define a signal detected in a zone as equal to the zone size, but others make estimates depending on the degree of overlap of signals between zones. TOFD is often added to the scanning system and this is available simultaneously.

The use of highly directed beams means that the calibration of AUT systems has to be carried out on material exactly similar to that which is going to be used in practice. The AUT method is given in more detail in ASTM E 1961 [1]

Phased Array Systems

Ultrasonic phased array systems have been available for medical ultrasonics for some time but are a recent introduction to industrial inspection. Phased array probes have the capability for focussing and directing a beam electronically. Figure 5 shows how a phased array probe produces a beam that can be focussed and directed.


Fig. 5. How a phased array probe generates a focus

This means that one phased array probe can replace several conventional focussed beam probes by suitable electronic switching. When used in the same way as AUT, the frame containing the probe is therefore much more compact than the multi-probe system mentioned above. The methods of flaw assessment are the same as AUT.

It should be noted however that phased array AUT has ultrasonic beams that are electronically focussed by means of the array only in the through-thickness plane (although some mechanical focussing may be present in the transverse plane). This is likely to mean that the length resolution is less than is possible for a beam that is also focussed in the circumferential direction.

Automated Ultrasonic Display on Information

Figure 6 shows the basic idea of the display from AUT systems. The information is arranged to appear in an electronic strip chart to show:

  • Each channel as an amplitude signal, with a marker when this exceeds a certain threshold.
  • An indication (usually colour coded) of the timing of a signal within a given gated time. This enables more accurate positioning of the flaw.
  • Additional channels, which may include TOFD or scans for the weld body.

Previous Literature on Inspection Reliability of UT

Several studies have been carried out on the inspection reliability of ultrasonic testing. Primarily these have been aimed at the nuclear industry. A summary of the capability of conventional UT has been given by Chapman and Bowker. [2] These typically have been determined by modelling and experimental studies and are associated with thick sections and probe movement methods of sizing. Chapman and Bowker state that for ranges less than 150mm, a flaw of 15mmlong by 3mm height is reliably detectable, provided that the tilt is less than 20° and the skew less than 3°.


Fig. 6. Indications from a typical AUT zonal inspection (different systems may vary)

A study carried out by AEA Technology, the Central Electricity Generating Board, and TWI [3]showed that the detected size can be much less than the actual size.

The PANI project [4] was carried out for the Health and Safety Executive by AEA Technology, and managed by an industrial steering committee including representatives from TWI, BP, BG Technology, Innogy, Esso, SaFed, Shell and Mitsui Babcock.

These trials consisted of manual ultrasonic inspection of welds by 16 inspectors and TOFD inspections. Two sets of artificially produced root flaws were included.

The results showed firstly that the detectability of root flaws appears to be greatly enhanced if the root is ground flush. If the root is left in place then the reliably detectable size is greater than 3mm through-wall extent (but this is based on only 2 flaws). In the PANI report the signal from these flaws is shown to be below the reporting threshold.

The results of these projects carried out on manual ultrasonic inspection indicate that the detectability of root flaws begins to become reliable at around 3mm through-wall, although this has not been quantified in terms of probability of detection. It should be noted that none of these studies specifically used focussed probes or phased array systems. Therefore there is very limited public information on the performance of these latter technologies.

For AUT systems some early work was carried out by TWI for the UK Department of Energy, [5] who compared the performance of P-Scan automated UT, the RTD Rotoscan and radiography. Their inspections seem to have been carried out at too high a sensitivity level (for the P-Scan it was set using a manual API standard) andfor Rotoscan it was set at 12dB above a 1/10 wall thickness through-wall slot. Many indications were reported which were not confirmed by sectioning ( Tables 9-13 of Ref [5] ).

The IPLOCA (International Pipeline and Offshore Contractors Association) organisation sponsored some trials through the University of Ghent, and the results of some of these were reported by Førli. [6] These results showed that 90% probability of detection was achieved for a through-wall extent of 1.2mm at a threshold setting of 40% full screen height (FSH). This was with the signal from a 3mm flat-bottomed hole (FBH) set at50% FSH. In this case no distinction has been made between different flaw types or locations. TOFD was reported as having a sizing capability better than ±0.9mm (with 95% confidence). Smaller flaws appeared to be better sized(although it is not clear how flaws which are less than 2mm through-wall were sized).

Bowers and Warren [7] gave some additional results from the same experiments ( Fig.9 in Ref [7] ) and showed considerable over sizing in addition to Førli's quoted result. Another result given was for pulse echo testing, with a calibration set at 80% full screen height from a 3mm diameter FBH. If a reporting level of40% had been used then the largest flaw missed would have been 2.9mm through wall. At this sensitivity one flaw of 2.1mm though wall gave a 10% signal height, which is likely to have not been reported even at higher sensitivity levels.

Kopp et al [8] stated that the accuracy of flaw sizing near the weld inner and outer surface had been shown to be '0.2-3mm' (it is not clear from the paper whether this means '0.2-0.3mm' or '0.2-3.0mm', but the text suggests the former). Alsothey state that the sizing capability for embedded flaws was of the order of ± 0.8mm through wall height.

Ginzel [9] suggested that sizing accuracies claimed for amplitude methods were exaggerated, and an improved method of flaw sizing might be to use an increased number of zones. The same author [10] also reported that the signal from root flaws from high angle probes is more likely to be a reflected signal than a direct signal. This has an impact on the setting up of the focal distance, although he indicated that this didnot affect the inspection reliability.

In a later paper Gross et al [11] compared the sizes of signals from various notches and flat-bottomed holes and ultimately decided that flaws could be grouped by amplitude into 4 categories: the lowest one had sizes from 0.5 to 1.5mm. This corresponded to anamplitude range of 5 to 30% FSH compared with the response from a 2mm diameter FBH.

More recently Morgan et al [12] reported results of a substantial qualification exercise involving seven teams using AUT. He concluded that Kopp's result (above) was optimistic, but that a maximum sizing underestimate of 1.5mm bounds most data points.

Reasons for Variability in Ultrasonic Inspection

Although ultrasonic systems can be very accurately set up and calibrated, they can never achieve the capability one might expect because of the variability of flaws and the frequency content of the ultrasonic signals. Suppose, forexample, that a flaw is a planar reflector perpendicular to the beam, but of a dimension similar to a small number of wavelengths (if the frequency content is 2-8MHz the wavelength will be 1.6-0.4mm for shear waves in steel). A flawwill not reflect the beam in a consistent manner due to interference between reflected and diffracted signals. [13] Even where only parts of a flaw (e.g. the ends) have dimensions of this order, the signals may become irregular. The resolution of the size of a flaw cannot be better than a wavelength even in ideal conditions. The use of probemovement for sizing is limited by the profile at different frequencies within the beam. Even a point flaw may appear to have the width of the beam by this method.

Misorientation of the flaw to the beam in any direction will reduce the reflected signal, as will curvature of the flaw.

Other possible effects that may cause variability arise from coupling variations, or slight changes due to play in the mechanical system, possibly due to pipe surface variations, wall thickness changes in the pipe and surfacecondition.

Particular Problems for Inspection of Weld Roots

The weld root has specific difficulties for ultrasonic inspection. This is primarily due to the fact that the root geometry itself can be a source of ultrasonic signals. An angled ultrasonic beam, when transmitted into a typicalweld root, will be reflected from the surface of the weld bead, particularly from the side opposite to the applied beam. The signal from this surface could easily be greater than that from a small flaw and the latter may be obscured.AUT systems do try to minimise its effect by time gating the signals to pick up only those from the part of the weld root adjacent to the incident beam, but this may be difficult especially where the weld root is small.

There is also a problem with sizing of surface breaking flaws by ultrasonic testing. One of the strongest signals obtained in ultrasonic testing is that from a corner. Once this captures the beam then signal amplitude will notincrease significantly with flaw height. These effects are shown in Fig.7.

Another difficulty occurs when there is high/low between the pipes (due to thickness variation and misalignment. This can mean small flaws are hidden or there is a potential for false calls.


Fig. 7. (Upper) Full reflection of beam from corner. Increasing the size of the corner does not affect the amplitude of the returned signal.
(Lower) Partial reflection. Amplitude of size of reflection is dependent on defect size, but this is complicated by diffracted signals. Part of the beam continues past the corner.

AUT trials carried out by TWI

TWI has carried out a number of trials for AUT and has also conducted analysis and qualification exercises. The details of these are given below.

TWI Joint industry Project


For the JIP two groups of pipe samples were used. The first group contained 10 samples of 24 inch diameter with 31mm wall thickness. These samples included 40 intentional lack of penetration flaws produced in the same way as wouldoccur for flaws during manufacture. The flaw lengths were obtained by comparison of radiography and visual inspection, the flaw heights by the Alternating Current Potential Drop (ACPD) technique and by replication.

The second group consisted of thirty five 12 inch diameter samples, some of which contained fatigue cracks. These were initially sized by ACPD and Magnetic Particle Inspection (MPI). Some of the samples were constructed from failedfatigue samples and used fatigue cracks initiated during a fatigue test. The cause of failure had been removed from the sample, therefore it was necessary to make a complete circumference so that the AUT systems could be applied. Thiswas done by first screening the samples to avoid cutting a cracked area, then cutting matching pairs to complete a circumference and finally tack welding the parts together.

The use of fatigue cracks was agreed by the project steering committee on the basis that, although these were not exactly typical of flaws likely to be found during manufacture, they were likely to be the most difficult type of flawto detect. Furthermore, they can provide valuable data for any future attempt to carry out in-service inspection.


Four inspection companies carried out the trials, which were conducted at TWI. The 24 inch diameter samples were inspected horizontally as shown in Fig.8. The 12 inch diameter samples were mounted vertically as shown in Fig.9. The sample under test was mounted between two 'pup' pieces, one of which formed a base to hold the sample, while the other provided sufficient extra length for the probe carriage scanner band.


Fig. 8. Multi-probe equipment mounted horizontally on 24-inch diameter sample


Fig. 9. Arrangement for inspection of 12-inch diameter pipes (shows phased array equipment)

The teams chose to operate from their vehicles to avoid moving electronic equipment. Electricity for the equipment and a water supply for the couplant were provided. The latter was collected in trays below the samples.


It was recognised that calibration samples were required to match each different type of sample to be used in the trials. It was agreed that TWI should provide the calibration samples. This was done for three companies, who provided details of the required reflectors in the pipes. The other company preferred to manufacture their own reflectors and pipe was supplied to them for this purpose.


It was considered that the replication method combined with an ACPD survey provided adequate accuracy for the lack of penetration flaws. Sectioning was carried out of all known fatigue cracks that were missed by the inspection teams. Sectioning was also carried out where three or more inspection teams reported an indication where no known flaw was present (suspected false calls). No additional flaws were found in this process. However it should be noted that there were false calls by each operator to varying degrees.

Qualification Exercises

TWI has carried out a number of qualification exercises for various clients, and while the detail must remain confidential, permission has been given to use this data in a non-attributable way.


The third group contains qualification exercises in which root flaws were included. These were manual welds, containing mainly lack of root fusion flaws manufactured at TWI by the bridge method, [14] but they also included a few lack of penetration flaws.

The diameter of the pipes varied from 12 to 18 inch and the thickness from 14 to 25mm.

In the fourth group of results, TWI has carried out statistical assessments of qualification results, where the inspection and sectioning have been carried out by others.

Inspection Teams

In the third group two inspection teams were used. In the fourth group each set of samples was inspected by one or two teams.


For both groups each company manufactured its own calibration samples and used its own inspection procedure and analysis method for the trials.


In the third group, trials were carried out at the premises of the inspection teams, with a TWI observer present.


The samples in the third group were sectioned by freeze breaking. In the fourth group salami sectioning was generally used.

Results and discussion

Schneider and Rudlin [15] described the methods of producing probability of detection (POD) curves. There are two main options: (1) to calculate POD by the use of 'hit/miss' data and (2) to use signal amplitude data compared with the known threshold(response versus size method). In the result given below the 'hit/miss' method was used for the data from the group sponsored project (first two groups) and the third group described above, whereas the 'response versus size' method has been used for the fourth group. The decision on which method to use is based on the data available.

There are various ways of presenting the POD curve. This is largely a matter of choice given the data available; we have chosen to use the method described in MIL-HDBK 1823 as this is the closest document to a standard for these purposes. [16] In this case, both the hit/miss data and the 'response versus size' data are fitted to a cumulative log normal model for the POD. This was shown to fit the data well.

POD data essentially depends on the procedure, the specimens and the operators. In addition it must be recognised that the qualification procedure may also have an influence on the results (for example the sectioning may have been carried out at only one location in a flaw, thereby running the risk of an optimistic result because the maximum flaw size has been missed).

However there are basic similarities between the trials carried out, including the calibration method (usually a slot in the root), which are broadly consistent with the guidance in ASTM E 1961.

POD is also a statistical quantity, and its value is strengthened by the amount of data (i.e. the number of flaws). Statistical tests can be used to test whether different data sets are statistically homogeneous, i.e. whether they appear to be drawn from the same population. These tests showed no evidence of statistical inhomogeneity (at the 5% significance level) between the first three groups. The mean POD for these three groups is therefore considered to be representative of each individual group. Moreover, the assumption of statistical homogeneity between the three groups permits confidence limits on the POD to be estimated (see Fig.10).

In the fourth group the number of misses was again small, but amplitude data was available, which provided extra information on which POD curves could be based. In these tests, typically the qualification was taken over a range off laws, and therefore there were relatively few root flaws. However, there appeared to be significant differences (at the 5% level) between the results of the individual trials in this group of data, which may have been due to the particular flaws or procedures used. For the sake of simplicity, the mean POD for this group is given in Fig.10. But it is inappropriate to construct confidence limits in this case, because of the suspected inhomogeneity within this group.


Fig. 10. POD curves obtained

Figure 10 shows that, for groups 1-3, 90% POD is achieved at a through thickness dimension of about 1mm. This estimate is based on 90 flaws. The lower 95% confidence limit reaches a POD of 90% at a through thickness dimension of about 1.5mm. For the fourth group, a mean POD of 90% is achieved at a through thickness dimension of about 1.3mm on the basis of 31 flaws in 5 different trials.

It can be seen that the AUT method appears considerably more reliable for root flaw detection than manual methods, where reliable detection, as mentioned above, is achieved for a through-wall dimension of around 3mm. This may have implications for example, where repairs are carried out, and manual UT is specified as the follow up NDT method.

Clearly, a summary curve such as Fig.10 is not strictly appropriate for use as a performance indicator for the systems in individual applications. The relevance of this curve is limited by the following factors:

  1. A single section of a flaw may not necessarily pass through the largest through thickness dimension. However in many cases three or more sections were made and these were targeted at the maximum ultrasonic amplitude.
  2. No account has been taken of the different calibration systems, different reporting criteria or different equipment or probes involved. The use of ASTM E1961 as a standard does however limit the variability likely to be observed.
  3. No account has been made of flaw shape or length.
  4. The thickness range is limited (from 12mm to 32mm total).

One might also speculate that the AUT operators who carry out the qualification trials for inspection companies are the best operators, and this may influence the results. A certification system (other than those operating within companies) to check operator performance would enable a greater degree of confidence to be placed in the results obtained.

The false call rate noted in the TWI JIP may have arisen because of the trial situation and the expectation of the operators. False calls by operators on site would soon lead to a lack of confidence in the technique, and this appears not to have occurred. However, it is not known whether the apparent difference between the trial and site situations results in a reduction in POD.

Use of data

Inspection qualification for AUT systems

Current practice for AUT qualification is a separate trial for each new task. It is suggested that the POD curve provided here can be used as generic evidence for the detectability of root flaws in qualification exercises. Of course where better inspection reliability is required then additional trials will be needed. The use of modelling should enable the study of individual effects that are difficult to control by experiment, thus enabling more cost-effective qualification.

Use of POD and sizing data in ECA

ECA calculations of the minimum flaw sizes permissible in a given pipe are usually based on a rectangular or semi-elliptical flaw shape. POD curves are generally based on actual defect shapes, and use a peak though thickness dimension. In most cases it seems reasonable to suppose that the POD data obtained from these trials to check that flaws of the required size are detectable is conservative.

Modelling capabilities

Modelling of AUT would be useful in applying the technique to qualification of new situations. Validated models for the interaction of ultrasonic beams with flaws are not common and have mostly been developed for the nuclear industry. These models do not generally address two of the key features of AUT testing, namely focussed probes (including phased array probes) and the response from flaws having a dimension less than an ultrasonic wavelength. Two recent developments have filled this gap to a certain extent, although neither has yet been fully validated:

  • CEA have produced the CIVA modelling suite, which includes models for phased arrays, as well as fixed-focus beams. [16]
  • TWI have shown in initial trials that a finite element model developed for guided wave inspection, where the flaws are generally less than the wavelength, can be adapted to conventional UT.

Future work

TWI is currently involved in a project which will provide additional data for the above generic POD curve, although its primary aim is to produce acceptance criteria, and the effort is therefore concentrated on crack sizing. This project will complete in 2004.

It is noted that there is a wealth of data from many qualification trials available in various companies, which of course could not be considered here. It would be a useful exercise to gather more of these data to improve the accuracy of the POD curves available.

TWI are involved in a European Commission funded evaluation of time-of-flight ultrasonics for the detection of flaws in manufacturing of welds. This should report in 2004.

Validation of the use of new modelling software (particularly CIVA) is also being carried out in TWI's core research programme. TWI's finite element models mentioned above will also be further developed and validated. Development of these models could lead to better understanding of the techniques and the results.

Development of techniques for clad and duplex pipes is a special problem not covered by the above work and is currently under way at TWI.


Two probability of detection curves for root flaws in riser welds have been produced from a series of qualification trials in which TWI was involved.

90% POD was achieved at a flaw height of 1mm for one set of data analysed by the 'hit/miss' method. The lower 95% confidence limit reaches a POD of 90% at a through thickness dimension of about 1.5mm. A mean POD of 90% was achieved at a flaw height of 1.3mm for a different set of data analysed by the 'response versus size' method.

The AUT method appears to be much more reliable than manual UT for the inspection of the type of girth welds considered in this paper.


The authors would like to thank the sponsors of the JIP for their support. This included BP Exploration, Marathon Oil, Health and Safety Executive, Statoil, Saipem SpA, Stolt Comex, Coflexip Stena Offshore, Drilquip, Shaw Pipeline Services, RTD, Oceaneering and OIS.


  1. 'Standard practice for mechanised ultrasonic examination of girth welds using zonal discrimination with focussed search units'. ASTM Document E 1961, 1998.
  2. Chapman R K and Bowker K J: 'The production of capability statements for standard NDT procedures'. Insight, 43, Jan 2001, pp36-38.
  3. Final Report 'The characterisation and size measurement of weld defects in ferritic steel by ultrasonic testing: Part 1: Non planar defects', TWI report 3537/4/77.
  4. PANI Project Final Report (1999) from Serco Assurance Ltd, Risley, Warrington, Cheshire UK.
  5. Mudge P J: 'Offshore pipeline girth welds'. Non Destructive Testing Department of Energy Report OTI 88 530 HMSO (1988).
  6. Førli O: 'Automated ultrasonic testing during offshore pipelaying: Acceptance criteria and qualification'. IIW Document V-1144-99.
  7. Bowers J and Warren E: 'The application of automated ultrasonic inspection for subsea pipelines'. 24th Offshore Pipeline Technology Conference, Amsterdam, Feb 2001.
  8. Kopp F, Perkins G, Stevens D and Prentice G: 'High resolution sizing capabilities for automated ultrasonic inspection of offshore risers'. Pipes and Pipelines International May-June 2001, pp 12-21.
  9. Ginzel E A: 'Amplitude sizing and mechanised ultrasonic inspection using linear scanning'. 5 (4). April 2000
  10. Ginzel E A: 'Signal interpretation misconceptions on near surface targets in mechanised UT on pipeline girth welds'., 4, Oct 1999.
  11. Gross B, Connelly T, van Dijk H and Gilroy-Scott A: 'Flaw sizing using mechanised ultrasonic inspection on pipeline girth welds'., 6 (7) July 2001.
  12. Morgan L, Nolan P, Kirkham A and Wilkinson R: 'The use of automated ultrasonic testing (AUT) in pipeline construction'. Insight, 45 (11), November 2003 pp746-753.
  13. Coffey J M and Chapman R K: 'Application of elastic scattering theory for smooth flat cracks to the quantitative prediction of ultrasonic defect detection and sizing'. Nucl. Energy 22, No 5, October 1983, pp319-333.
  14. Lucas W: 'Making defective welds for Sizewell B: 'Welding and Metal Fabrication, March 1992, pp81-86
  15. Schneider C R A and Rudlin J R: 'Review of statistical methods used in quantifying NDT reliability'. Insight, 46 (2), February 2004, pp77-79.
  16. Lhémery A, Calmon P, LecSur-Taïbi I, Raillon R and Paradis L: 'Modeling tools for ultrasonic inspection of welds'. NDT&E International, 33 (2000), pp499-213.

For more information please email: