AUTOMATED ULTRASONIC PHASED ARRAY INSPECTION OF

FATIGUE SENSITIVE RISER GIRTH WELDS WITH A WELD OVERLAY

LAYER OF CORROSIVE RESISTANT ALLOY (CRA)

Philip Ducharme1; Sbastien Rigault2; Ingmar Strijdonk3; Nicolas Feuilly4; Olivier Diligent4; Pascal
Pich5; Frdric Jacques1

KEYWORDS: pipeline, riser, FPSO, Corrosion-resistant alloy, CRA, weld overlay, girth weld,
phased array, automated ultrasonic testing, AUT, NDT, ultrasound, inspection, austenitic,
carbon steel, stainless, SS316, Inconel, Inconel, Inconel 625, Inconel 825, chrome, Duplex,
Super Duplex

Abstract
Fatigue sensitive pipelines present a challenge for nondestructive examination (NDE) of girth
welds. The fatigue to which the risers are subject requires that all remaining defects in the
welds be of a very small sizeas small as 0.5 mm x 12 mm in the surface regions of the
weld.
Detection and accurate sizing of such imperfections are already a significant challenge on
Carbon Steel (CS) risers and CRA metallurgically bonded pipe. This paper presents a greater
challenge, that is the presence of a CRA weld overlay as opposed to other types of CRA
overlay for the purpose of providing fatigue and corrosion resistant performances to the
risers.
To satisfactorily address these challenges, a substantial technological leap in AUT techniques
was required for the inspection of pipe-to-pipe girth welds. This paper reviews how these
challenges were eventually overcome through the development of a reliable, robust, and
efficient phased array AUT technique. This paper will also address the performances
achieved during validation.

ABSOLUTE NDE, 765 rue de lglise, St-Romuald, Qc, Canada, G6W 5M6; Philip.Ducharme@AbsoluTeNDE.com
Olympus NDT, Qubec, Canada
3
ABSOLUTE NDE, Rio de Janeiro, Brasil
4
SAIPEM SA, Paris, France
5
ABSOLUTE NDE, Salies de Barn, France
2

Introduction
Over the last decade, the oil and gas industry has significantly increased its use of pipelines with
an inner layer of corrosion resistant alloy (CRA), due to its ability to enhance fatigue corrosion
resistance. (Geveneux, 2010)
This CRA layer is relatively thin (within the 3 to 5 mm range), and is frequently used to line the
inner surface of carbon steel pipes in order to improve their chemical and mechanical properties.
CRA is either Inconel 625 or 825, Stainless Steel 316, Duplex or Super Duplex alloy.
Two methods are typically used to bond the anticorrosion alloy to the inner surface of the
external carbon steel layer.
The first method consists of hot rolling a CRA layer on a carbon steel plate. A thin layer made of
a third material is generally placed between the CRA and the carbon steel in order to improve the
metallurgical characteristics of the bond.
The second method, which is referred to as weld overlay clad, consists of creating a layer by
welding successive passes of CRA on the inner surface of a carbon steel pipe, as shown in Figure
1A.
Ultrasonic techniques for carbon steel weld examination are well known and widely used.
(ASTM, 1998) (R/D Tech, 2007) These ultrasonic examination techniques, which use the
standard shear wave full skip technique, do not provide satisfactory results when applied on
bimetallic welds with CRA weld overlay clad pipes. Material microstructure at grain scale is
heterogeneous and anisotropic, which generates scattering of the ultrasonic wave, attenuation,
and beam skewing (American Welding Society, 1986) (Institut de Soudure, 2009) (Ogilvy, 1986)
(IIW Commision V, 2008).
Hot-rolled clad produces a flat and uniform steel-to-clad interface. Difficulties in UT
interpretation are therefore strictly related to the heterogeneous and anisotropic weld
microstructure, and to the clad interface between two distinct materials.
The weld overlay clad creates a wavy interface with an anisotropic coarse-grained microstructure
that does not permit use of the full skip ultrasonic technique. As such, both the coarse-grained
microstructure and the uneven clad interface must be taken into account.
In order to improve ultrasonic examination of weld overlay clad pipes, a novel phased array AUT
technique was developed that also provides greater flexibility for various inspection
configurations in comparison with conventional AUT systems.
This document describes an innovative ultrasonic phased array technique successfully used for
fatigue sensitive weld inspection of CRA weld overlay clad pipes during a FPSO construction
project with underwater manifolds and clustered wells located at a maximum depth of 2,500
meters.
The following sections describe the problem, and explain the technical aspect of the innovation,
followed by an analysis of the performances achieved during the projects validation phase.

Innovative Inspection Solution for CRA Weld Overlay Clad

Pipe
As previously highlighted, the difficulties encountered during AUT inspection of welds on CRA
weld overlay clad pipes result from the internal CRA layer, which induces a wavy interface
between the carbon steel and clad, and an anis
anisotropic coarse-grained
grained microstructure, as shown on
Figure 1.. Due to the phenomena observed above in relation to welded clad, the standard AUT
inspectionn techniques using the full skip path (ASTM, 1998) are inefficient for girth weld
inspection of CRA weld overlay clad pipes
pipes. (Institut de Soudure, 2009)

B
A
Figure 1: A) Macrograph of weld overlay clad interface with carbon steel
steel.. The interface is not
flat, but rather erratically wavy; B) Columnar dendrites oriented in the through
through-thickness
thickness
direction of the weld overlay clad.
Because skipping is not possible on tthe pipes inner surface,, inspection is exclusively restricted to
the first ultrasonic leg.
The initial rational for inspecting girth welds on CRA overlay clad pipes or any bimetallic girth
weld was to inspect several zones of the weld using a series of ul
ultrasonic
trasonic beams placed at
different angles. One problem with this technique is that high angle longitudinal beams (over
70) are very difficult to control, calibrate, and analyze, which gives rise to a blind zone in the
upper part of the weld.

Technical Innovations
To cover this blind zone, a special wedge was developed that enables the probe to ride the weld
bead, as shown on Figure 2B. This new wedge fits the AUT standard inspection head.

CRA weld caps typically have a very low profile (less than 3 mm), and are very smooth, which
reduces mechanical stress raisers. The special wedge was designed to fit a weld bead up to 5 mm
high.
Use of this special wedge, with its phased array probe, in combination with other phased array
probes and creeping wave probes positioned on each side of the weld, provides coverage over the
entire thickness of the weld on the CRA overlay clad pipe.

Inspection configuration and coverage

The methodology, which is used to inspect the weld overlay clad joint, is illustrated in Figure 2.
The first few millimeters under the surface are inspected using creeping wave probes (Figure
2A). Creeping waves can detect 0.5 mm high buried and surface-breaking defects in the first
4 mm of the welds depth. Figure 3A shows the responses from three surface-breaking notches
with heights of 0.5 mm, 1 mm, and 2 mm, respectively. The signal-to-noise ratio for the 0.5 mm
notch is higher than 18 dB, which demonstrates how easy it is to detect such defects.Starting
2.5 mm deep below the surface, the phased array probe mounted on the special wedge covers the
entire volume and fusion line of the weld, in addition to the heat affected zone (HAZ). Efficiency
is optimum up to 2 mm above the root (Figure 2B).
Finally, two phased array channels (groups of beams) are produced by the twin phased array
probes located on each side of the weld. One channel covers the weld from about 10 mm deep to
the root (Figure 2C), while the other is dedicated exclusively to the root area (Figure 2D). The
latter is used to accurately discriminate and size 0.5 mm high defects. Figure 3B shows a signal
from a 0.5 x 5 mm notch machined in a 4.5 mm thick Inconel weld overlay cladding. The
signal-to-noise ratio on that notch is around 12 dB, which demonstrates the excellent flaw
detection sensitivity in the root area. Generally speaking, defects higher than 0.5 mm in the root
area are not permitted for CRA pipeline inspection. This inspection technique complies with that
requirement.

Figure 2: Ultrasonic coverage:

A) From surface to 4 mm deep with creeping wave.
B) 2.5 mm deep from the OD to 2 mm above the internal surface, with the special phased array
probe.
C) 10 mm deep from the OD to the internal surface with phased array probes.
D) Root area/internal surface with pha
phased array probes.

Figure 3: Signal from surface-breaking notches

A) (OD) 5 mm long notches with heights of 2 mm, 1mm, and 0.5 mm, respectively, seen
with creeping waves; signal-to-noise ratio is higher than 18 dB;
B) (ID) (root zone) 5 mm long notches with heights of 2 mm, 1 mm, and 0.5 mm,
respectively, seen with Root channels. The 0.5 mm notch has a signal-to-noise ratio of
approximately 12 dB through a 4.5 mm thick clad.

Sizing method and back diffraction

Because the welds scan plan is not divided into zones, the AUT standard sizing methodology
(referred to as zone discrimination) cannot be applied using this innovative technique. As
such, another technique called back-diffraction is preferred for sizing the height of all
indications detected with phased array probes (see Figure 4). The amplitude-based sizing
technique is still used in some cases, for example, with creeping wave data, or when the backdiffraction technique cannot be used.
For AUT data interpretation, this back-diffraction sizing technique requires powerful software
with three-dimensional displays, combined with specific sizing cursors and tools (Figure 4).
(Olympus NDT, 2009)
This sizing technique is based on measurement of the relative time-of-flight (TOF) difference
between the diffracted signals of a flaw. (Jacques, et al., 2003) This is similar to traditional
TOFD, except that the ultrasonic signal is emitted and received back using the same probe. The
height of an indication is given by Equation 1:

Equation 1

H = (UT2 - UT1) / cos

Where H is the defect height (vertical extension), UT1 and UT2 are the UT paths for the lower
and upper tip, respectively, and is the ultrasonic refracted angle in the material. (Jacques, et al.,
2003)

Figure 4: Sizing software

A) Example of a 3D representation of an indication measured with specific cursors.
B) Diffraction signal from the upper tip, and reflection signal from the corner trap of the crack.

Results and discussion

A validation program to prove the detection and sizing reliability of this technique was performed
on Inconel CRA weld overlay clad pipes on artificially induced reflectors that are depicted in
the DNV document DNV-OS-F101 App. E. The quantity of the artificial reflectors provided was
insufficient to fully comply with the criteria laid out in the afore mentioned DNV document.
This program was completed prior to a construction project involving inspection of fatigue welds
on FPSO with underwater manifolds and clustered wells located at a maximum depth of 2,500
meters.
Five girth welds with a wall thickness of 32.2 mm, including a clad layer of Inconel 625
(27.7 mm WT carbon steel + 4.5 mm CRA weld overlay) with natural defects, were prepared for

this AUT validation. They were scanned with the pipe axis in the horizontal position, (5G) and
indications were evaluated as per the pending approval procedure. This project specification
required a system capable of detecting and accurately sizing surface-breaking and embedded
flaws as small as 0.5 mm x 12 mm, and 1.5 mm x 22 mm, respectively.
From these five (5) defective welds, 55 imperfections and one (1) clear area were assessed
through destructive testing (DT) for comparison with the AUT assessments. These 56 indications
led to about 620 macro sections.
The assessment indications were chosen in order to provide uniform distribution over the entire
weld thickness (see Table 1). Note that the population in the root area is over exaggerated to
provide more information about this critical zone.
Table 1: Summary table of defects distribution over the thickness of the five (5) welds
AUT Defect
Positioning
within the
Thickness
Cap
0 - 5mm
5 - 10 mm
10 - 15 mm
15 - 20 mm
20 - 25 mm
25 - 30 mm
Root
Clear area
Total

Number of Defects
Weld No. 1

Weld No. 2

Weld No. 3

Weld No. 4

Weld No. 5

Total

1
0
2
1
0
0
1
6
0
11

1
2
1
2
1
0
0
4
0
11

1
0
4
0
1
1
2
2
0
11

2
2
0
0
1
2
1
2
1
11

1
2
1
0
2
1
0
5
0
12

6
6
8
3
5
4
4
19
1
56

Length sizing

AUT assessments
Two different operators conducted the AUT data interpretation. The minute number of
differences between the two operators AUT interpretations, in addition to the negligible
difference between two sizing values for a single defect, demonstrate the consistency of the two
operators AUT results. The AUT procedure is therefore not operator-dependent for detection and
flaw sizing.

AUT versus RT
As a preliminary study, a comparison was carried out between this AUT technique and that
employed for radiography testing (RT). As expected, the results demonstrated that AUT is
capable of detecting more indications than RT.
For these five welds, RT was only capable of detecting defects that were a minimum of three
times as long and high as the critical defect dimensions.

AUT versus DT
Comparisons of defect length and height assessments between AUT and destructive testing (DT)
assessments produced the following results.

Height sizing comparisons

A direct comparison between AUT assessments and macrosectioning is shown in Figure 5.
With the exception of three (3) oversized defects, all of the 52 remaining flaws assessed during
the validation process were within a 1 mm sizing accuracy range. When the study is restricted to
the ID and OD connected flaws, the range increases to an accuracy of 0.5 mm. Height sizing
accuracy is therefore compliant with the strictest standards of the oil and gas industry.
Overall, the probability of undersizing a flaw height by more than 1 mm has proven to be less
than 2.5%.

AUT assessments (mm)

6
5

Oversizing

4
3
2

Undersizing
1
0
0

2
3
4
Macro measurements (mm)

Figure 5: Comparison between AUT assessments and DT measurements

Ideal sizing;
(
Surface flaw scatter;

+/- 0.5 mm;

+/- 1 mm;
Embedded flaw scatter)

Safety limit against undersizing;

Table 2: AUT vs. DT, including data from the two operators

Number of Samples Assessed

RMS Vertical Sizing Error
Std Vertical Sizing Error

Surface
(mm)
50
0,4
0,3

Additional Information
Total Samples Assessed
False Calls
True Negative
Cap Flaw Assessments
Embedded Flaw Assessments
Root Flaw Assessments

Embedded
(mm)
60
0,7
0,7

112
0
2
12
60
38

As shown in Table 2, based exclusively on the data for 60 embedded defect assessments, a
standard vertical sizing error of 0.7 mm was found. For the surface-breaking flaws on the cap or
root, the 50 AUT and DT flaw assessments produced a standard vertical sizing error of 0.3 mm.
The safety limit against undersizing, defined as Equation 2, was also calculated. (Nordtest, 1998)
Equation 2

H 95 = Average( H i ) 1.64* RMS( H i )

The 95% safety limit for embedded flaws was found to be -1.1 mm, and increases to -0.4 mm for
surface flaws.
It is important to note that no defects were missed, and no false positive alerts were made during
this validation process.
Probability of Detection
The standard reference document for probability of detection (POD) assessments is the Nordtest
TECHN Report 394. However, since no convergence with this method was possible due to the
fact that every single indication was detected and sized, the vs. a (circumflex A vs. normal
A) method was employed. (RESEARCH REPORT 454: Jacobi Consulting Limited, 2006)
Probability of detection is dependent on the flaw size needed to exceed a certain threshold for
evaluation. POD is also dependent on several parameters, including:
Flaw orientation
Sensitivity
Flaw type/reflectors
Operator skills
Material
POD is to be measured against defect severity (height) to allow for direct use of AUT results. An
estimate of the POD values based on qualifications with a limited number of defects has an
uncertainty, which must be taken into account when using or comparing POD values, for
instance, a confidence level.
POD Curve
A schematic representation of a POD curve is provided in Figure 6A which shows two curves,
the average POD (normal line) and the lower confidence limit (dashed line).
When building a POD curve, some welds with deliberately induced defects are inspected, and all
indications above a certain amplitude threshold are reported. The welds are then destructively
tested to determine the actual defect heights of all the indications.
Destructive testing involves slicing through the weld at predefined locations, and performing a
comparison with AUT assessments in order to determine where a detection may or may not have
occurred.

As stated in the inspection procedure, an initial evaluation threshold at 20% FSH was used, and
all defects were found. Accordingly, POD cu
curves
rves could not be determined because the calculation
could not converge. However, by varying the evaluation threshold to 30% (defects below the
threshold are considered not found, whereas defects above the threshold are considered to be
found), some flaws were intentionally missed in order to enable the calculation to converge.
The POD as per DNV rules is characterized by one single value, i.e. the h90/95, which represents
the defect height at which the probability of detection is 90% with a confidence of 95%, as
indicated in Figure 6A.
Figure B shows the POD curve calculated fro
from the complete data set.. The solid line constructed
using the analytical vs. a method indicates that a POD of 90% is achieved for flaws with a
vertical extent greater than 0.28 mm. This POD curve is also considered the average POD.
The confidence level
evel is an indication of the reliability of the results (lower curve). A 95%
confidence indicates that if the test were run 100 times, 95 times out of 100 tests, the POD curve
would not be lower than the curve indicated by the 95% confidence curve.
The 95%
% confidence level is achieved for flaws 0.68 mm high. This result is achieved using an
evaluation threshold of 30% full screen height (FSH).
Because there is no AUT system capable of detecting 100% of all flaws, a study with more data
would normally provide more realistic results. However, the results obtained with the 30%
threshold are highly satisfactory in comparison with standard criteria required in the oil and gas
industry.
The overall results show a significant improvement in the performance achi
achieved
eved by an AUT
system on an austenitic weld overlay pipeline girth weld inspection.

(A)

100%
90%
80%

60%
50%
40%

vs a POD 30 % FSH

30%

Lower Confidence Limit 95 %

0.283 mm

20%
10%
0%
0

0.2

0.676 mm

Probability

70%

0.4

0.6

0.8

1.2

1.4

1.6

1.8

Height (mm)

(B)
Figure 6
A) Illustration of detection curve including the lower confidence limit and the definition of the h0
and h90/95 (ML: Maximum Likelihood)
B) POD curve taking all flaws into account

Conclusion
An innovative automated ultrasonic inspection technique was developed for pipeline girth welds
with weld overlay cladding of austenitic material. This technique covers 100% of the weld
volume using phased array ultrasound, in addition to creeping wave probes. The novelty of the
discussed technique is that a phased array probe runs on top of the weld. This probe is mounted
on a wedge with a resilient contact surface, which enables the inspection to be carried out even
when the weld cap is up to 5 mm high.
This technique has been qualified on 12.75 in. OD x 32.2 mm thick pipes (including a 4.5 mm
Inconel 625 weld overlay layer), and could be qualified on virtually any bevel configuration or
thickness.
By using the vs. a statistical method and comparing 55 defects assessed by two (2)
independent operators (for a total of 110 samples), a 90% probability of detection with a
confidence level of 95% (h90/95) was achieved for defects as small as 0.68 mm in height.
The system sizes the vertical extent of the defects with a systematic deviation of 0.3 mm for
surface-breaking flaws, and 0.7 mm for embedded flaws. The probability of undersizing by
more than one millimeter (1 mm) is less than 2.5%.
Owing to its proven capability, this new technique meets the stringent requirements of the oil and
gas industry for fatigue sensitive weld inspection, and even surpasses them.

Because the POD result was obtained by increasing the detection threshold to enable
mathematical convergence, it can be assumed that the actual capability of the system is superior
to the critical vertical height of 0.68 mm. Then, the accuracy of the statistical calculations could
be further enhanced by increasing the number of samples used for evaluation.

Abbreviations
AUT
CS
CRA
DT
FPSO
FSH
ID
OD
PA
RT
TOF

Automated Ultrasonic Testing

Carbon Steel
Corrosive Resistant Alloy
Destructive Testing
Floating Production, Storage and Offloading
Full Screen Height
Internal Diameter
Outside Diameter
Phased Array
Radiographic Testing
Time of Flight

Bibliography
American Welding Society. 1986. Handbook on the Ultrasonic Examination of Austenitic
Welds. 1986. ISBN 9997643585.
ASTM. 1998. E 1962-98, Standard Practice for Mechanized Ultrasonic Examination of Girth
Welds Using Zonal Discrimination with Focused Search Units. 1998.
Det Norsk Veritas. 2007. DNV OS-F101, Submarine Pipeline Systems, App D, Non-Destructive
Testing (NDT). 2007.
Erha and Erha North Development: Automatic Ultrasonic Inspection of Corrosion Resistant
Alloy Clad Steel Catenary Risers. Sutherland, J.B. et al. 2006. Houston, Texas : Curran
Associates, inc., 2006. Offshore Technology Conference.
Geveneux, H. 2010. AKPO Project - The Successful Implementation of Steel Catenary Risers.
2010. OTC 20994.
Hellier, C. 2001. Handbook of Nondestructive Evaluation. s.l. : McGraw-Hill, 2001. ISBN
0070281211.
Hugdell, R. J. and Gray, B. S. 1985. The ultrasonic inspection of austenitic materials - State of
the art report. Northern Division Report, United Kingdom Atomic Energy Authority. Risley
Warrington : Risley Nuclear Power Development Laboratories, 1985. ND-R-1201(R) / CSNI
Report No. 94.
IIW Commision V. 2008. Handbook on the Ultrasonic Examination of Austenitic and Dissimilar
Welds. s.l. : DVS Media, 2008. ISBN: 978-3-87155-969-3.
Institut de Soudure. 2009. Technical Advices on Ultrasonic Testing of Butt Welds in API 5L
X60 Steel Pipes Internally Clad with Inconel 625. Paris : s.n., 2009. p. 12.
Jacques, F., Moreau, F. and Ginzel, E. 2003. Ultrasonic backscatter sizing using phased array developments in tip diffraction flaw sizing. Insight - Non-Destructive Testing and Condition
Monitoring. November, 2003, Vol. 45, 11, pp. 724-728.
Nordtest. 1998. Guidelines for NDE Reliability and Descriptions, NT Techn Report 394. 1998.
Ogilvy, J A. 1986. Ultrasonic Beam Profiles and Beam Propagation in Austenitic Weld using a
Theoretical Ray Tracing Model. Ultrasonics. Vol. 24, #6 1986, p. 337.
Olympus NDT. 2009. Advanced Calculator: Users Manual. Waltham, MA : Olympus NDT,
2009.
. 2009. TomoView: Users Manual. Waltham, MA : Olympus NDT, 2009. DMTA079-01EN.
Palanichamy, P., et al. 1995. Ultrasonic velocity measurements for estimation of grain size in
austenitic stainless steel. NDT & E International. 1995, Vol. 28, 3, pp. 179-185.
R/D Tech. 2007. Introduction to Phased Array Ultrasonic Technology Application. Qubec :
s.n., 2007. 9735933.
RESEARCH REPORT 454: Jacobi Consulting Limited. 2006. Probability of Detection (PoD)
curves; Derivation, applications and limitations. London : Health and Safety Executive, 2006.
Ultrasonic Phased Array Inspection of Welded Pipes Using Wave Mode-Converted at the Inner
Surface of the Pipe. Long, R., Cawley, P. and Russell, J. 2009. 2009, JRC-NDE 2009.