You are on page 1of 8

Proceedings of the ASME 2022 14th International Pipeline Conference

IPC2022
September 26-30, 2022, Calgary, Alberta, Canada

IPC2022-89297

MANAGING ROOFTOPPING: AN IN-LINE INSPECTION BASED APPROACH

Brett Conrad Dr. Gerhard Kopp Ross Adamson Tommy Mikalson


TC Energy NDT Global TC Energy TC Energy
Calgary, Canada Stutensee, Germany Calgary, Canada Calgary, Canada

Roger Lai
TC Energy
Calgary, Canada

ABSTRACT however, the ability to detect rooftopping allows for a better


Variation in longitudinal seam weld geometries can pose a understanding of the pipeline system and adds another level of
challenge for liquid operator’s crack management programs. data integration which TC Energy has used to improve their
Both, radial misalignment and angular misalignment (also liquids crack management program.
known as “peaking” or “rooftopping”) contribute to what is Keywords: seam weld, rooftopping, peaking, misalignment,
termed an “anomalous” seam weld geometry. These anomalous in-line inspection, validation
geometries create two challenges for an operator’s crack
management program: they can reduce the fatigue life of the NOMENCLATURE
seam weld due to an increased stress concentration; and they
create an unexpected geometry which challenges crack in-line DSAW double submerged arc weld
inspection tools. The combination of these challenges can create EDM electrical discharge machined
a situation where a seam weld flaw can propagate at a higher UCx ultrasonic inspection crack
rate without being detected by in-line inspection technologies, UMp ultrasonic inspection metal loss pitting
increasing the risk of failure. It is therefore prudent for operators UTCD ultrasonic testing crack detection
to understand where they may have anomalous seam weld UTWM ultrasonic testing wall measurement
geometries and adjust their crack management plan accordingly. h rooftopping height
This paper outlines the approach taken by TC Energy to  rooftopping angle
leverage existing datasets to quantify anomalous seam weld R pipe radius
geometries, and more specifically rooftopping, on pipe following
an incident. Several different in-line inspection technologies 1. INTRODUCTION
were investigated to quantify the presence of rooftopping, with Despite the best efforts in the pipe mill and depending on
various levels of success. One of the most promising approaches the manufacturing and quality control processes in place,
involved sensor stand-off data of an ultrasonic wall thickness occasionally the longitudinal seam weld geometry has variations
measurement technology. The technology was initially pulled in it. These variations can present as radial misalignment,
through a calibration spool in which rooftopping was artificially angular misalignment, or a combination of both (see Figure 1).
introduced. Through collaboration between the TC Energy and While these variations can exist in various longitudinal seam
NDT Global, an algorithm was developed to not only detect, but weld application types, the focus of this work was double
also size the levels of rooftopping within the pipeline. The results submerged arc weld (DSAW) longitudinal seam welds.
of the excavation program found that the technology could
adequately detect rooftopping. The absolute sizing of
rooftopping is an area of focus for future improvements;

1 © 2022 by ASME
Figure 2: PHOTO OF RADIAL MISALIGNMENT GAUGE

Figure 1: DIAGRAM OF ANGULAR MISALIGNMENT


(PEAKING) AND CENTERLINE (RADIAL) MISALIGNMENT

These variations (also described as anomalous seam weld


geometries) increase the stress concentration of the longitudinal
seam weld, reducing its fatigue life. They also potentially impact
the performance of crack in-line inspection technologies. These
impacts to pipe integrity have led to failures within the industry
[1].
Following a release where an anomalous seam weld
geometry was identified as a causal factor, TC Energy began to
investigate what could be done to detect and quantify where else
this seam weld geometry might exist. Investigations into various Figure 3: PHOTO OF CURVATURE TEMPLATE
ILI technologies, including Caliper, ultrasonic testing wall
measurement (UTWM), and ultrasonic testing crack detection As the program matured, the in-ditch measurement moved
(UTCD) were undertaken. Based on discussions with ILI away from these instruments to laser scanning. The laser scan
vendors, other operators and regulators it was determined that it produced a higher resolution data set for the entire longitudinal
was highly unlikely that radial misalignment would be able to be seam, where rooftopping angle and height was measured at 0.5
reliably detected by any ILI technology on the market. This was inch (12.7 mm) increments. To determine the rooftopping height,
due to numerous potential reasons, including tool capabilities, the tangential angle of the pipe adjacent to both sides of the seam
the circumferential resolution of sensors, as well as the nature of weld were projected to an intersection point over the center of
how various sensors interact with the longitudinal seam weld the weld, as shown in Figure 4. The equation used to calculate
during an inspection. TC Energy progressed forward with the rooftopping height is shown in Eqn (1). An example of the
developing an ILI program focused on peaking (i.e. rooftopping) seam weld geometry measurements obtained from the laser scan
to help inform its crack management program in liquid pipelines. analysis is shown in Figure 5; yellow line is outer pipe surface
from laser scan, blue lines are created from analysis of the laser
2. ROOFTOPPING ILI DEVELOPMENT & VALIDATION scan data.

2.1 Field Measurement of Longitudinal Seam Weld


Geometry
One of the key aspects of developing an ILI program is
understanding how the “truth” data is generated in the field.
Early in the excavation program following the release, the radial
misalignment was measured using a gauge provided by the
manufacturer of the susceptible pipe (see Figure 2). An
aluminum curvature template matching the nominal outer
diameter was used to determine if any rooftopping was present
(see Figure 3).

2 © 2022 by ASME
rooftopping may exist; however, no indications were found in
the field. Caliper data from a traditional caliper tool with
mechanical sensor arms was analyzed in-house at TC Energy.
The analysis methodology looked at the change in measurements
over a given joint and four candidate joints were selected for
excavation. None of these excavations had any rooftopping or
radial misalignment that was beyond the allowable
measurements outlined in the 46th Edition of API 5L [2].
Once the existing data sets had been exhausted, NDT
Global’s Atlas UG caliper technology was selected to gather new
data with the purpose of identifying rooftopping, based on
discussions with various ILI vendors, industry experts and the
regulator. The Atlas UG inspection technology was added to a
scheduled EVO UMp (UTWM) inspection and completed in Q2
Figure 4: CONVERSION FROM ROOFTOPPING ANGLE TO of 2021. Following the review of both the Atlas UG and EVO
ROOFTOPPING HEIGHT UMp data, it was determined that the UMp’s increased
circumferential resolution provided the best opportunity to detect
ℎ 𝑅 1 (1) and quantify rooftopping. Following the initial discussions, it
was decided to proceed with the UMp rooftopping algorithm
development.

2.3 UMp Algorithm Development


A direct measurement of the rooftopping height in the ILI
data is problematic for three main reasons. Firstly, the maximum
rooftopping height is expected in the center of the long seam, but
at this location the weld cap masks the real rooftopping height.
Furthermore, the weld cap is often a bad reflector and thus the
data quality may not be optimal directly in the center of the weld.
A third problem is a possible movement of the sensor plate into
the rooftopping which results in an underestimation of the real
rooftopping height, see Figure 6.

Figure 6: MOVEMENT OF A SENSOR PLATE (GREY) INTO A


ROOFTOPPING. IN THE LEFT FIGURE THE SENSOR PLATE IS
IN ITS NOMINAL POSITION, THE REAL ROOFTOPPING
HEIGHT CAN BE MEASURED. IN THE RIGHT FIGURE THE
SENSOR PLATE MOVED INTO THE ROOFTOPPNG AND A
DIRECT MEASUREMENT UNDERESTIMATES THE REAL
ROOFTOPPING HEIGHT. BUT THE MEASURED ROOFTOPPING
Figure 5: EXAMPLE OF SEAM WELD GEOMETRY ANGLE WILL STILL BE CORRECT.
MEASUREMENTS OBTAINED FROM LASER SCAN ANALYSIS
Further possible problems are the limited resolution of the
2.2 Investigation of Candidate Technologies ILI tools and possible coincident ovalities. All these possible
To identify and quantify rootopping in the system, TC error sources led to the conclusion that measurement of the
Energy began looking at existing data sets on susceptible pipe. rooftopping angle leads to a more robust data set compared with
During the initial review of available data, TC Energy looked at the rooftopping height. This is because no measurements directly
several different technologies to identify rooftopping. Initially, at the weld cap are necessary [3][4].
the stand-off sensor of a UTCD technology was investigated. It For measuring the rooftopping height, reference points are
was determined that the circumferential resolution of this selected on both sides of the long seam and the rooftopping angle
technology was not high enough to reliably detect any is calculated from the measured slopes of the two reference
rooftopping, and the performance of the technology would be arrays. It is important that none of the data points on the
dependent on the chance that the sensor was in the correct reference arrays hit the weld cap since in this case the calculated
orientation relative to the longitudinal seam weld. Two slope would be incorrect. Therefore, some data points in the
excavations were performed where the technology identified that center of the long seam are not used for the calculations.

3 © 2022 by ASME
Furthermore, it is important to project the measured stand- Peaking Angle
25
Peaking Height
5

off data to the real pipe geometry. Otherwise the method would 20 4

ignore the pipe curvature and the real rooftopping angle is 15 3

underestimated. 10 2

measured peaking height [mm]
Figure 7 shows in the upper graph the measured stand-off

measured peaking angle [°]
5 1

data in the second test spool (see next chapter). In the center of 0 0

the data the weld cap can easily be identified. The blue markers
‐25 ‐20 ‐15 ‐10 ‐5 0 5 10 15 20 25 ‐5 ‐4 ‐3 ‐2 ‐1 0 1 2 3 4 5

‐5 ‐1

show the edges of the sensor plates of the UMp tool. In the lower ‐10 ‐2

figure the same data but now projected to the real pipe geometry ‐15 ‐3

is shown. The reference arrays on both sides (red dots) exclude ‐20 ‐4

the long seam zone with the weld cap. The slopes of the reference ‐25
peaking angle [°]
‐5
peaking height [mm]

arrays in the projected data can be calculated by a linear


Figure 8: SENSITIVITY CURVES FOR THE ROOFTOPPING
regression and the rooftopping angle can be derived from these ANGLE (LEFT) AND THE ROOFTOPPING HEIGHT (RIGHT). THE
slopes. Once the rooftopping angle is determined, the INSENSITIVE ZONE RANGES FROM -10° TO +10° OR -1 MM TO
rooftopping height can be calculated using Eqn (1). +1 MM, RESPECTIVELY.

Movement of the sensor plates under a rooftopping anomaly


(see Figure 6) is also reflected in the measured stand-off data.
Figure 9 shows an example of movement of the sensor plate
which results in a step in the measured data on the left side. A
linear regression will return an incorrect slope since it cannot
properly detect and handle outliers (dashed line). In contrast, a
robust regression identifies and removes outliers in the
Figure 7: MEASURED STAND-OFF VALUES AROUND incremental slopes and the final result is correct (solid line).
THE LONG SEAM (UPPER FIGURE). THE PROJECTED
DATA AND THE REFERENCE ARRAYS (RED DOTS) ARE
SHOWN IN THE LOWER FIGURE. THE BLUE MARKERS
SHOW THE EDGES OF THE SENSOR PLATES OF THE ILI
TOOL
While not the focus of TC Energy’s program, inward
peaking can also be detected and sized by this algorithm. The
same approach can be used for inward peaking if a Kelvin
transform is applied to the data. The Kelvin transform inverts the Figure 9: THE MOVEMENT OF THE SENSOR PLATE RESULTS
space with the nominal pipe contour as fixed points. The IN A SIGNIFICANT STEP IN THE DATA ON THE LEFT SIDE AND
magnitude of the angles is preserved by this operation but their THE LINEAR REGRESSION CANNOT CALCULATE THE
signs are reversed. PROPER SLOPE (DASHED LINE IN THE LOWER FIGURE). A
ROBUST REGRESSION CAN BETTER HANDLE STEPS IN THE
Since the measurement of the slope is not taken directly at DATA AND THE CALCULATED SLOPE REFLECTS THE REAL
the long seam, the angle-based measurement is not sensitive for SLOPE OF THE INNER PIPE SURFACE (SOLID LINE).
very small rooftopping heights. The sensitivity for a 30" pipeline
and a long seam weld zone of seven tracks (about 28 mm or 1.1") A more detailed discussion of the angle-based approach and
can be assessed using model data, see Figure 8. The insensitive possible error sources can be found in [3] and [4].
zone for this setup ranges from about -10° to +10° for the
rooftopping angle or -1 mm to +1 mm (±0.04”) for the 2.4 Initial Validation
rooftopping height, respectively. The reporting threshold for the Prior to the inspection in Q2 of 2021, TC Energy had
rooftopping height was defined as 30% of the wall thickness for developed two different correlation spools with multiple seam
this project, and it therefore can be concluded that for a nominal welds. One correlation spool contained EDM notches and fatigue
wall thickness of 10 mm (0.4”) this method should be adequately cracks in the original and artificial seam welds, where one
sensitive. artificial seam weld contained radial misalignment and the other
two artificial seam welds contained non-symmetrical
rooftopping. The second correlation spool had two artificial
seam welds that only contained symmetrical rooftopping of
various heights, as shown in Figure 10 of a mock-up test of the
fabrication process.

4 © 2022 by ASME
Figure 10: MOCK-UP TEST OF SYMMETRICAL
ROOFTOPPING

The UMp technology was used to inspect the second Figure 12: ROOFTOPPING HEIGHT COMPARISON OF JOINT
correlation spool with symmetrical rooftopping to validate the #1
algorithm. In addition to the correlation spool, the data for four
joints was also requested from NDT Global where the
longitudinal seam weld had been laser scanned during previous
excavations. The UMp data was then compared to the five data
sets as part of the initial validation. Once the UMp data had been
provided, the associated laser scan data was provided to NDT
Global, and the measurements were compared. The comparison
is shown in Figure 11 - Figure 15. Based on the sensitivity
analysis outlined in Section 2.3, the plots were limited to
calculated rooftopping heights greater than 1 mm (0.04”). The
laser scan height is the rooftopping height as calculated from the
laser scan rooftopping angle measurement. The “Vertical Height
NDT” data is the robust regression calculated height, and the
“Vertical Height (Fit) NDT” data is the linear regression
calculated height. Figure 13: ROOFTOPPING HEIGHT COMPARISON OF JOINT
#2

Figure 11: ROOFTOPPING HEIGHT COMPARISON OF


CORRELATION SPOOL
Figure 14: ROOFTOPPING HEIGHT COMPARISON FOR JOINT
#3

5 © 2022 by ASME
algorithm. A total of 24 rooftopping anomalies with heights
greater than 3 mm (0.12”) were identified within the pump
section. These 24 anomalies were reported across 14 joints, with
one of the joints being previously excavated with no cracking
features identified. The length and maximum heights of all the
reported anomalies were provided by NDT Global. Based on
these measurements, TC Energy requested the rooftopping
height profiles for the entirety of ten of the joints. In addition to
these profiles, NDT Global also reviewed all 14 joints in their
UCx Eclipse data from an inspection completed the previous
year to identify any below specification reflectors that may be
present. Only the previously excavated joint had a coincident
UCx Eclipse reflector, which correlated to a lack of fusion
Figure 15: ROOFTOPPING HEIGHT COMPARISON FOR JOINT feature that was found in the ditch to be smaller than the
#4 minimum detection threshold of the technology. There were no
other rooftopping anomalies coincident with a UCx Eclipse
In general, the UMp appears to identify the general trends of reflector.
rooftopping within the joint. For Joint #1 (Figure 12), there TC Energy reviewed the profiles, it became apparent that
appears to be a drop off of the rooftopping height. This is some of the shorter indications were most likely due to data noise
explained by the “dead zone” explained in 2.3. For Joint #2 rather than any realistic manufacturing mechanism that could
(Figure 13), the UMp appears to undercall the rooftopping height produce such localized rooftopping. These types of profiles were
for the beginning of the joint. This occurred due to the labelled as lower confidence calls, with profiles showing gradual
suboptimal sensor position relative to the longitudinal seam changes in the rooftopping height being labelled as higher
weld. The sensor position relative to the seam weld is shown in confidence calls.
Figure 16. In this pipe joint the edge of two sensor plates (blue With the information provided, TC Energy developed an
markers) is located at the weld cap which might have resulted in initial response criteria looking at the following information:
some tilting of the sensor plates. This is the worst-case scenario  Coincident reflector in UCx Eclipse data;
for the angle-based approach to measure the rooftopping height.
 Rooftopping height;
 Rooftopping length;
 Quality of Profile
Based on this review, three joints were selected for
excavation:
 Joint #5: Appeared to have rooftopping along the
entire joint and had a good, consistent profile;
Figure 16: UMP SENSOR PLATE POSITION RELATIVE TO  Joint #6: Contained the anomaly with the highest
LONGSEAM IN JOINT #2 rooftopping measurement. The profile and short
length reduce the confidence of the call; however,
2.5 Algorithm Improvements Based on Validation due to the severity was selected out of an
The initial validation efforts showed relatively good abundance of caution; and
performance of the UMp algorithm; however, several possible  Joint #7: Contained two anomalies, one of which
improvements were identified by NDT Global. was the longest reported anomaly. Also had a good
In the original algorithm five data points around the long consistent profile.
seam were excluded for the calculation of the rooftopping height.
After reviewing the findings, NDT Global determined that 4. EXCAVATION PROGRAM RESULTS
excluding seven sensors rather than five provided better results The three excavations were completed in Q4 of 2021. None
and made the adjustments accordingly. of the joints contained any linear indications in the longitudinal
In addition to excluding more sensors adjacent to the seam weld. The results of the comparison between the laser scan
longseam weld, the algorithm for the robust regression was and NDT Global measurements are shown in Figure 17 - Figure
revised. In the first version a simple median-based outlier 19. Note that the ILI call boxes outlined in the figures were based
detection was used. The improved version is based on a on the original algorithm, while the improved algorithm was
statistical analysis of the incremental slopes in the stand-off data provided afterwards.
on both sides of the long seam.

3. EXCAVATION SELECTION
While the initial validation was being undertaken, NDT
Global analyzed a requested pump section using the UMp

6 © 2022 by ASME
rooftopping height seen in the laser scan data, which helps ensure
that any significant rooftopping can be seen by the UMp
technology.

5. LEVERAGING ROOFTOPPING DATA IN CRACK


MANAGEMENT PROGRAM
The ability of UMp technology to detect and quantify the
presence of rooftopping in a given pipeline segment is valuable
in a liquids pipeline crack management program. The reduced
fatigue life of the longitudinal seam weld and impact on UTCD
performance should be considered at locations where
rooftopping exists.
The reduced fatigue life of longitudinal seam welds has
already been addressed through modification of Paris Law
parameters for susceptible pipe. Previous work [1] has shown
Figure 17: JOINT #5 COMPARISON OF LASER TO UMP
ROOFTOPPING HEIGHTS
that the presence of rooftopping, absent an interacting flaw, does
not impact the burst pressure of the pipe. Considering this, any
reported rooftopping anomalies will be monitored and will be
reviewed in subsequent crack inspections to identify any below-
specification reflectors that may be seen by the technology.
To address the impact to UTCD ILI performance, TC
Energy and NDT Global are working together to determine a
revised Probability of Detection (POD) curve for crack-like
features at various rooftopping heights. This information will
provide a response criterion where if any rooftopping anomalies
are found with a certain height, regardless of interacting features,
this will trigger an excavation in the future to investigate possible
undetected fatigue cracks. Simulation results from NDT Global
have already been used to modify TC Energy’s re-inspection
interval assuming a larger initial crack size.

Figure 18: JOINT #6 COMPARISON OF LASER TO UMP 6. CONCLUSION


ROOFTOPPING HEIGHT Based on the work completed between TC Energy and NDT
Global on managing rooftopping:
 TC Energy looked at the stand-off sensors of a
UTCD technology as well as two different caliper
technologies. All of which were unsuccessful in
identifying rooftopping.
 The UMp algorithm has shown success in
detecting rooftopping in field conditions.
 The inclusion of profile data is extremely
important in considering the confidence of a given
rooftopping anomaly call reported.
 While further work is required to refine the sizing
of reported rooftopping anomalies, the algorithm
appears to be conservative.
 Although a dormant threat in isolation, leveraging
the UMp rooftopping data can help liquid pipe
Figure 19: JOINT #7 COMPARISON OF LASER TO UMP operators improve their crack management
ROOFTOPPING HEIGHT programs and avoid potential failures.
Overall, the comparison shows that the UMp algorithm is  In-ditch collection and subsequent analysis of laser
capable of detecting rooftopping. The correlation also showed scan data was an accurate and reliable method to
that integrating the profile information can help eliminate measure the actual rooftopping angle, which was
potential false calls (as that seen in the case of Joint #6). The used to validate the UMp algorithm.
UMp algorithm also appears to be slightly overcalling the

7 © 2022 by ASME
ACKNOWLEDGEMENTS [2] API Specification 5L Line Pipe; 46th Edition (2018)
The authors would like to acknowledge the support of Terry [3] Kopp, G., Frank, M., Meinzer, T., & Yanez, R. (2022).
Huang, Billy Zhang and Wei Xiang from TC Energy for their Improved Measurement of Peaking Angles and Peaking
support in analyzing the traditional Caliper data; Rapid3D in Heights using ILI Data. Pipeline Pigging and Integrity
their development and execution of an analysis to derive Management Conference. Houston: Clarion Technical
rooftopping angle, height and radial misalignment from laser Conferences and Great Southern Press.
scan measurements and EMC2 (Engineering Mechanics of [4] Kopp, G., Frank, M., Meinzer, T., & Yanez, R. (2022).
Columbus Corporation) for their development and fabrication of Improved Measurement of Peaking Angles and Peaking
the correlation spools. The authors would also like to thank Dr. Heights using ILI Data. 17th Pipeline Technology
Christina Jung for her support in setting up the algorithms and Conference. Berlin: EITEP – Euro Institute for Information
processing the UMp data. and Technology Transfer in Environmental Protection
GmbH.
REFERENCES
[1] Dafea, M; Hopkins, P., Palmer-Jones, R., de Bourayne, P., &
Blin, L. (2012) IPC 2012-90587 Investigation into the Failure
of a 40” Diameter Crude Oil Pipeline. International Pipeline
Conference. Calgary: ASME.

8 © 2022 by ASME

You might also like