The Performance of Automated Ultrasonic Testing

(AUT) of Mechanised Pipeline Girth Welds

Lewis Morgan,
Advantica Technologies,
United Kingdom.
[email protected]
Paper presented at the 8th ECNDT, Barcelona, June 2002

ABSTRACT

The use of ultrasonics to test pipeline girth welds during pipeline construction has
increased substantially around the world since equipment first became available in
the 1980’s. This has been driven in part by a substantial improvement in the
ultrasonic technology and in part by the safety and environmental benefits of
avoiding the use of ionising radiation for radiography.

This paper describes work conducted at Advantica Technologies to support a

Transco Pipeline construction programme. Two types of weld were studied.
Samples of both were prepared containing a range of defects. Seven AUT
contracting companies then tested the sample welds in blind trials. Their results
were then compared with the best of several radiographic inspections. Some of
the defects were then cut open for a definitive appraisal.

Particular attention was paid to the types of defect associated with mechanized
welding such as copper pick-up, cold laps and other lack of fusion defects.
Cracking, transverse to the direction of welding, was also simulated.

In general, results were obtained which were comparable to radiography in all

areas except for isolated pores and inclusions. Copper cracking was generally
misclassified and transverse defects were not detected reliably.

The novelty and complexity of the AUT data places a renewed demand on the
skills of company representatives in terms of overseeing the application of AUT
in the field. Measures, including training, need to be considered in order to
guarantee the quality of contractor performance and this would best be addressed
as an industry-wide activity. It is also important that users get the best economic
benefit from the new technology and this means using the capability of ultrasonics
to define the size of any defects that are found. This would be in order to leave
small defects unrepaired where they are shown to be safe through a fitness-for-
purpose analysis. The issues surrounding the accuracy of size evaluation and
further work needed in this area is discussed.

Introduction
 Traditionally, pipeline girth welds made during pipeline construction in the UK
have been manufactured using the Manual Metal Arc (MMA) welding process
and inspected using 100% X-radiography. There is some degree of radiation
hazard to the public and operators in using x-rays and Transco have considered
whether this can be reduced or removed.

In 1981 work was conducted by British Gas to assess five ultrasonic scanner
systems to replace x-rays but they were all found to have deficiencies, especially
in discriminating different defect types. Since then there has been several
technical developments so that, today, the use of automated ultrasonic testing
(AUT) of girth welds is becoming increasingly accepted as a reliable and
beneficial alternative to radiographic testing (RT). Confidence has been generated
by an extensive list of pipeline construction projects where it has been used
successfully. While the basic approach is similar from most suppliers, there are
significant differences as well as new developments in response to the demands of
the marketplace. It was decided to conduct a trial of several AUT systems to
establish the capabilities. It was carried out by Advantica Technologies on behalf
of Transco.

AUT developments

The early AUT systems used probes to focus beams of ultrasound onto zones
along the weld fusion line. They were set at appropriate angles to detect reflected
energy from any lack of fusion at the focal areas. A set of probes was therefore
needed to cover the whole of the weld thickness, one set on each side. This is
illustrated in figure.1.

Fig 1: Schematic of pulse-echo reflection beam paths.

Since then further probes have been added to inspect the volume of the weld,
another pair to give diffracted data and two more pairs for transverse defect
detection. The evolution of the equipment has thus led to a complex set of data
requiring analysis. Further details of these elements can be found in references 1-
3.

The plan of work

 Nine butt welds were fabricated in 48 inch (1220mm) diameter, X80 grade, pipes
using two mechanised welding systems. These welds contained a range of
deliberately made defects. Seven AUT contracting companies were invited to
bring their equipment to inspect the welds at a Transco site and report on the
defects they found.

Each company inspected the welds independently, under the oversight of

Advantica staff. Data collected was interpreted off-site following each test and the
reports were then sent to Advantica for comparison against radiographic
inspection, which was taken as the standard. Some of the defects were cut open to
confirm their nature and size.

The objective was to establish the effectiveness of AUT as an adequate alternative

to radiography for pipeline construction when using mechanised welding. It was
also intended that the evaluation should prepare for the drafting of a specification
for AUT use, which could then be applied to future construction projects.

COMPANIES INVOLVED

In alphabetical order, the AUT companies who took part in the trials were: -

 Institut de Soudure, France

 OIS, Aberdeen
 RTD, Rotterdam
 Servicios de Control e Inspección SA, (SCI), Madrid
 Shaw QED, Great Yarmouth
 Solus Schall , Oceaneering, Aberdeen
 Weldsonix, Canada

Several other companies were invited to take part in the trials but, for a variety of
reasons, were unable to do so.

  PREPARATIONS

Pipe material:
In order that the AUT contractors were able to set up their equipment effectively,
a coupon of pipe material was sent to each one, to be made into calibration pieces,
as specified in the ASTM standard (Ref.4.). This was modified to comply with the
requirement to detect transverse defects. There is a considerable time and cost
involved in this preparatory work.

Weld profile
Companies who offer mechanised welding services use various types of welding
preparations for the profile of the ends of the pipe. There are two types in
common use; the ‘CRC-type’ of preparation and the ‘J’-type, illustrated below:
 a) The CRC-type weld has a root deposited
from an internal head. b) The J-prep is
deposited from the outside against an
internal copper backing ring
Fig 2: Weld Profiles.

The different welding processes can result in some differences in the types of
defect that can occur. Consequently, both types of weld were used for the trial.

Since the AUT systems are set up with the focussed beams of sound directed at
different regions of the fusion faces of the weld (as described previously), it
follows that the precise weld preparation profile must be specified to each
prospective inspection contractor for the systems to be set up appropriately. This
information was provided along with the calibration coupons to each company.

WELD DEFECTS

A wide range of defect types and condition was specified for specialised welding
contractors to insert into the welds during their fabrication. These included both
simple defects such as lack of fusion (LoF) both in the root and the sidewall (at
and between different weld passes) as well as lack of penetration (LoP) at the
root. Areas of porosity were required, as this was believed to be a feature that
could be a problem for AUT detection. ‘Combination defects’ were also
requested, so that some features would extend across more than one weld pass at,
or near, the same circumferential position. The possibility of copper being picked
up from the copper backing (as used in the J-preparation weld) was of some
concern. It was intended to simulate the inclusion of copper, both as copper
pickup from the backing bar and also where the welding head copper tip might
make contact with the joint. Transverse cracking was also included in the
specification. Such cracking could not readily be induced during welding so it
was decided to machine narrow slots into the welds after fabrication. To be
consistent with the API 5L specification (Ref.5.) for axial weld cracking in
linepipe, the slot depths were made around 10% of wall thickness i.e.
approximating to an N10 notch.

After being welded, the welds were inspected using panoramic X-radiography.
These radiographs were analysed by four independent, qualified interpreters.
They were also re-radiographed and digitised both for the record and as a further
check. This degree of rigor was warranted since the RT results were to be used as
 the baseline to evaluate the reliability of the AUT results. The interpretation of the
radiographs demonstrated that the welding companies had produced a good set of
defects in both types of weld and the defects were well distributed around the
weld circumferences. Additionally, two slots were ground externally, one through
the weld cap of each type of weld and two more were ground internally, through
the weld root of each type. The locations of these slots were randomly located
away from other defects.

The distribution of the types of defect that were produced is illustrated in the table
of fig.3. A plot of the RT lengths of the defects, also shown in figure 3,
demonstrates that they were deliberately made to be short and so were more
difficult to detect by the AUT. This had the added advantage of allowing a lot of
defects to be included. Nearly 90 defects were used; eight inspections (from seven
companies) thus gave a reasonable statistical set of data.

Type of defect Number

Porosity 15
Lack of Fusion 44
Lack of Root
17
Fusion/Pen.
Transverse 7
Other 14

Fig 3: The distribution of length and types of defect

included in the tests.

The nine welds amounted to 35 metres of weldment. This means there was a
considerable length of non-defective weld. If any company had attempted to
achieve good detection by over-reporting this would give spurious indications that
would be evident. (In the event this didn’t happen; the spurious indications were
consistently low).

TRIALS
 The two pipe spools were located with good access and good facilities in terms of
comfort – a dry working environment with light, power and water. Adequate time
was given to each AUT company to ensure that they would not be pressured by
time. This was because it was intended to test only the capability of the equipment
and interpretation. A standard briefing was given to each company. The weld caps
were covered with tape to prevent sight of the external defects and the pipe ends
were covered to prevent access to see the insides. No accept/reject level was
defined – all reflectors considered significant were to be reported.

RESULTS

The first analysis of results was to compare how the AUT indications lined up
with the RT results. The accuracy of characterisation was evaluated subsequently
on a subset of defects by metallographic section of all the defects in one weld,
selected with a good selection of defect types.

The results confirmed that very good detection performance is available from
AUT systems generically. All linear features longer than about 10 mm were
detected by some of the systems tested, as well as most clusters of porosity above
10 mm. Overall the Lack-of-Fusion and Lack-of-Root-Fusion defects were
detected in 94% and 90% of the cases, while porosity was detected only 75% of
the time. The transverse defects were detected only in about 70% of the cases.
Scattered or isolated features, including pores, cavities and inclusions (metallic
and non-metallic), were generally not detected. Neither were geometric features,
such as ‘low cap’, ‘underflush’ or ‘flush root’, reported.

It is generally considered that AUT is better than RT in detecting planar defects.

This is confirmed by these results. The start of a defective area was shown to be
detected earlier by the AUT in many of the defects that were sectioned (e.g.
Figure 4 indicates that a Lack of Side Wall Fusion of even 0.5 mm through-wall
height has been detected .)
 Fig 4: The LoF defect on the left fusion face, at 5 mm from the outer surface, was
found by all but one AUT system. It was not visible at all on the radiographs. (Wall
thickness = 16mm) (1 mm bar on enlargement.).

Some variation between inspection companies was apparent, indicating a lack of

consistency. This is illustrated in the table below.

Missed by 1 or Missed by 2 or Missed by 3 or Missed by 4 or

Defect Type more more more more
inspections inspections inspections inspections
Porosity 58% 41% 41% 25%
Lack of Fusion 38% 23% 13% 8%
Lack of Root
46% 40% 27% 13%
Fusion/Penetration
Transverse 100% 60% 40% 20%
Other 88% 75% 62% 50%
Table 1: Breakdown of missed defects.

While porosity is confirmed as problematic for defection, it is worth noting that

half the inspections found 75% of these small areas of porosity.

The transverse defects were not detected well, especially the external slots.

The ‘other’ defects include geometric features (such as ‘low cap’, ‘underflush’ or
‘flush root’), which are not good ultrasonic reflectors, accounting for the low
detection rate.

An important factor is likely to be the human operator’s individual interpretation

of the data, when the complexity of the data involved is considered. In order to
get good performance in field operations with AUT systems, this suggests that
attention to the competence of the operator is essential. An on-the-job test of
operator competence has been proposed for Transco operations, much as is done
for welders.

Three of the eight inspections were conducted using phased array equipment. No
trend was evident to suggest that either better or worse results were obtained with
the phased arrays.

When considering the deficiencies of AUT, it should be remembered that RT is

by no means a perfect inspection system to compare it with. However, the
approach does allow a good breadth of coverage in the comparison.

The Detection of Copper.

Several areas of copper, inserted into the weld, were available for study. No pick-
up of copper from the internal backing plate was evident.

The most easily detected area (called ‘craze cracks’ by RT) was detected by all
the AUT systems, but was generally characterized as ‘porosity’. A section,
(Figure 5), shows this form of cracking to be much like liquation cracking and is
distributed throughout the weld so that its misclassification as porosity is
understandable.

Fig 5: Defects associated with copper in the

weld.

The second area of copper was only detected by RT as a small ‘transverse defect’;
most of the feature was invisible on the radiographs, even on careful re-
interpretation. When it was sectioned it proved to be more complex than expected
and extended along the weld length (>12 mm ). Only one company detected it as a
transverse feature, though two others also reported a ‘planar’ element. Generally
it was reported as ‘porosity’.

Since these defects seemed to tax the interpreters’ capabilities, an explanation was
sought. The two features were included into a set of other, real areas of porosity
and the list fed back to each AUT contractor. They were asked if they could
distinguish the two areas of copper-related defect and invited to say what, in the
signals, made them think it might be due to copper. Generally, only one of the two
areas was picked out, but three companies picked out neither area. More
experience is clearly needed. This is important since the significant size (the
accept/reject boundary) is very different for the two types of feature.
Misclassifying a planar defect as volumetric might effectively be the same as
missing it altogether, since it might not be investigated or repaired.

Transverse defects
The transverse slots that were cut across the welds were generally detectable, but
with some notable ‘misses’. Each of the internal slots was eventually detected
across the board (one company did so only on a second attempt). The detection of
the external slots was more variable. About half the AUT companies missed the
external slot in the J-prep. weld but only two missed that in the CRC-type weld.

Each of the contractors set up their equipment with transverse slots in the
calibration coupons but these are made of plane pipe with no weld cap or root
bead. This geometry difference may account to some degree for the missed
defects. A false cap may be worth including with the calibration or set-up block to
reduce timing errors.

The detection of these defects is not covered in standards and an industry

initiative to improve this aspect may be warranted.

Seam weld region defects

The region where the pipes’ seam welds run into the girth weld is also a region
needing special attention. The ultrasonic properties of the seam weld material are
unknown, so it cannot be assumed that the ultrasonic beams are appropriately
targeted. In practice, any defect that extends into such an area might need to be
further examined, manually, for any extension that would take it into the ‘reject’
class. Alternatively, such a defect could be assessed with an additional length
added equal to the width of the pipe seam weld i.e. assume that the defect extends
right across the seam weld area. This does not appear to be done generally and
needs to be included in specifications. The location of the seam weld is generally
visible on the TOFD trace.

Size measurement
Two size measures of a defect are significant – the length along the line of the
weld and the length through the thickness of the wall i.e. its ‘height’. Only the
first measure is available from radiography but both can be estimated from the
ultrasonic data.

The length along the weld tends to be longer when measured ultrasonically.
Measuring the through-wall height of a defect by ultrasonics is conventional, but
is prone to errors. This is even truer of AUT systems, where only amplitude data
at a fixed range are available, though Time Of Flight Diffraction (TOFD) probes
are also now used. TOFD size measures have been proven to be good for buried
and deep defects (Ref.6.) but are limited in accuracy when used on shallow, near-
 surface features. Conventional, zone-based sizing can give an error band in the
range from 0-6mm. A comparison of amplitude based AUT size values with
destructive measures (Ref.7.) has shown that these are generally, but not
exclusively, conservative. The defects in this Advantica study have been analysed
in a similar way and have led to a similar conclusion. Details of this will be
included in a future publication.

An improved method, using the special options of phased array data, has been
demonstrated on calibration defects (Ref.8.). An accuracy of better than 1 mm has
been cited but it is unlikely to be as good for real defects. Neither is it yet a viable
method of sizing defects in field operations where rapid processing is required.

When defect size values are put into ECA assessments a substantial conservatism
becomes apparent. Real defects are rarely of the simple parabolic shape assumed
by the ECA. The sensitivity of AUT may indicate a long tail either side of a short
feature, but it will then be assessed as having the depth of the short section and
the length of the whole indication, including the tails; it is therefore more likely to
fail assessment and to be repaired, perhaps unnecessarily. The fracture mechanics
method needs to be able to handle complex shaped defects to reduce such over-
conservatism.

Field experience

Following the positive results of this work the practical aspects of using AUT
have been evaluated during construction of a new pipeline in Scotland in 2001.
For the first 26 km of the line both RT and AUT were carried out and the results
compared. Every indication noted on both records was entered onto a database, of
a substantial size. This allows the impact of different inspection criteria to be
studied. For example, with the same specification the repair rates using AUT or
RT were shown to be comparable. If an engineering critical assessment (ECA)
specification were to be used, however, the repair rate was shown to drop
substantially, with the potential for significant cost savings.

A similarly low repair rate of 1% was reported (Ref.7.) for the Alliance pipeline
construction, coupled with a very high productivity rate, which AUT can support.
This has also been experience in a recent British Gas offshore pipe-lay, where
even lower repair rates were achieved with an API1104 based ECA (Ref .9.)

Conclusions

The AUT systems on trial have produced inspection results that compare well
with radiography, with good detection except for isolated pores and inclusions.
Clustered porosity is frequently found, especially larger clusters. Copper cracking
is generally misclassified when it is detected and so may be left unrepaired.
Transverse defects are not detected reliably. Further work is needed to improve
the detection of both copper and transverse defects. Some variations are apparent
 in the performance of different operators and companies and an on-the-job testing
of operator competence has been proposed as this is such a crucial element of the
operation.

References

1. J.A.de Raad, High Speed Ultrasonic Inspection of Field Girth Welds During
Pipeline Construction, Pipeline Technology Conference, Ostende, Belgium, 1990
2. E.Ginzel & R.Ginzel, B.Gross, M.Hoff, P.Manuel, Developments in Ultrasonic
Inspection for Total Inspection of Pipeline Girth Welds, 8th Symposium on
Pipeline Research, Houston, Texas, August 1993.
3. E.Ginzel, "http://www.ndt.net/article/ginz_pl/ginz_pl.htm">Further
Developments In Ultrasonic Inspection of Pipeline Girth Welds, NDT.net
(http://wwww.ndt.net) - June 1996, Vol.1 No.06.
4. ASTM ‘Standard Practice for Mechanised Ultrasonic Examination of Girth welds
Using Zonal Discrimination with Focused search Units’, E-1961-98.
5. API Specification 5L. ‘Specification for Line Pipe’.
6. J.Bowers, and E. Warren, EMC Ltd., "The Application of Automatic Ultrasonic
(AUT) Inspection For Subsea Pipelines", 24th Offshore Pipeline Technology
Conference, Amsterdam, Feb.2001.
7. B. Gross, T. Connelly, H. van Dijk and A. G-Scott, "Flaw sizing using
mechanised ultrasonic inspection on pipeline girth welds", - NDT.net- July 2001,
Vol.6, No.7.
8. Michael Moles, Pascal Piché and Noël Dubé, "Application of Phased Array
Ultrasonics for Girth Welds", Final Report 1999 PRCI Project Number PR-270-
9813.
9. API Standard 1104, ‘Welding of Pipelines and Related Facilities’, Nineteenth
Edition, September 1999.