Run Comparisons: Using In-Line Inspection Data For The Assessment of Pipelines

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Pipeline Technology 2006 Conference

Run Comparisons: Using in-line Inspection Data for the


Assessment of Pipelines
K. Reber; M. Beller; A.O. Barbian
NDT Systems & Services AG
Germany

Abstract
In the past In-line Inspection (ILI) was mainly done for the purpose of quickly finding defects
that had to be repaired in order to restore the pipeline integrity. Currently this task shifts to a
continuous monitoring of the pipeline's condition, which allows to derive new conclusions
like the assessment of corrosion growth.
The paper will discuss how defect growth can be estimated from continuous inspections and
what implications the advancement in inspection technology has. Several models of corrosion
growth assessment are discussed. Their applicability depends on the condition of the pipeline,
i.e. the density of defects and the available data. A stepwise process can be defined in which
more detailed information will allow to use more accurate methods of corrosion growth
assessment.
In a final stage the data of high-resolution ultrasonic inspection tools can be used to compare
defects on a basis of wall thickness C-Scans. This will generate more precise conclusions
about corrosion growth on single defects, which was not possible with the traditional
statistical approach. In particular it is possible to assess corrosion growth on selected sites
even in an early stage of corrosion or if the number of corrosion sites is small.

1 Introduction
In many countries of the world pipeline regulations not only demand a check of pipeline
integrity in case of doubt or after incidents resulting in loss of property or even life. Instead a
continuous process of constant monitoring of pipeline integrity is required. Often in-line
inspection is the method of choice for these measures. Many regulations demand or
recommend the use of intelligent pigs [1,2].
This is one reason why in-line inspection is nowadays a less exceptional event in the
operation of pipelines, but for many operators has become a process of every day life. While
the inspection results used to be information that was heeded right after delivery and then
archived (unless action was required) it is today used in many circumstances even years after
the actual inspection has taken place.
It is in the benefit of both parties, the ILI-operator and the pipeline operator, to ensure that the
information is delivered such that it can easily be exploited in the future. Naturally the
inspection technology is advancing, which results in a desirable improvement in detection
levels and reliability. In several cases this has left the impression that the comparability of
results with earlier inspections is compromised. However, the reason for systematic
differences in the detection results should not be blamed on the advancement of the
technology, but rather on the limitations of the previous inspection technology. The notion "If
I missed the defect the last time, I better miss it this time, too, so I am at least consistent" is
definitely short-sighted.
Pipeline Technology 2006 Conference

This paper will focus on the issue of repeated ultrasonic inspection. Many aspects, however,
are also applicable to inspections with magnetic flux leakage (MFL). Corrosion growth
studies are a major task in repeated in-line inspection. Several methods have been developed
to pin-point potential growth sites based on ultrasonic data.

2 Procedures
2.1 Matching the lists
In a first step the run comparison is based on the features list, i.e. the result list of findings
after an in-line inspection. The POF document describes what the features list should contain
as a minimum. Whether the old list meets these requirements is not always clear. A minimum
requirement is a list of features with the following items
• Defect size in width, length and depth
• Defect location in distance and orientation
If the defect location is given as the position of the deepest point (ultrasonic inspection UT) or
the maximum magnetic field amplitude (MFL), there is an uncertainty about where exactly
the defect starts. This point is not necessarily in the middle of the whole length of the defect.
A method describing features by call-boxes is preferable. The location must be given with
reference to a pipe tally. Absolute distances cannot be compared in two pig runs. They rather
have a sorting function. A pipe tally with distances corresponding to the defect distances has
to be supplied as well. Both lists have to be in an electronic format like an Excel-Sheet. This
should be especially emphasized to operators. In the past it has become necessary, in some
instances, that a print-out was to be read into an Excel-list via a scanner and a word pattern
recognition system. Although these tools become better and better, the procedure is still error
prone and time consuming. It should be self-evident that record keeping needs to be in
electronic format nowadays.
These lists then need to be put into a database table. Either a proprietary format is used or a
standard model. At NDT Systens & Services AG a software has been set up that will first
match the two pipe tallies. This is not always trivial and may require some manual
verification. Then features are matched joint by joint locating them with respect to girth
welds. Some tolerances can be applied, because deviations of up to 10 cm (4 inch) in axial
distance can be found. If the angle parameter is not given the tolerance in orientation can be
set to 360°. This will find all features matching by distance alone. If the feature density is
high there is a chance that some matches are incorrect. This is a potential problem for older
lists, where feature orientation was not always given.
Figure 1 shows a screenshot of the RunComparison function of NDT’s Analysis Software
PIXUS. Correlating metal loss features can be identified.
With a comparison based on the features list alone the following information can be derived:
• A list of corresponding metal loss features. This list could be used for a step one
corrosion growth analysis.
• A list of discrepancies in feature classification. This can be very important
information. If a feature was assumed to be an inclusion it was not considered a threat
to the integrity of the pipeline. However, if in the second run it is revealed as a pitting
corrosion, things are different. Although a thorough data analysis should minimise
Pipeline Technology 2006 Conference

such discrepancies, the evolvement of inspection technology and analysis procedures


will inevitably lead to discrepancies.
• A list of features that have been missed in the first inspection. Either they have
developed in the meantime or these defects have been missed due to inferior
inspection technology.
• A list of features that have been missed in the second inspection. If the second
inspection is carried out independent of the results of the first inspection this is still
possible although not likely. In some cases the second inspection may find a flaw to be
shallower than before. If it falls short of the analysis threshold it will not be listed.

Figure 1: A sample screenshot of the RunComparison Function of NDT’s analysis Software. The blue
arrows on the right shows a corresponding metal loss. The left arrow indicates that channeling corrosion
has newly developped.

2.2 Statistical Analysis


If a run comparison was carried out for a kind of feature that is associated with a depth (like
corrosion) it will become important to check for changes in depth. Even for a single defect
this has a statistical character, because the measurement itself is associated with a degree of
uncertainty. In UT-inspection it is common to take the deepest reading of a box region and
report this as the depth of the defect. Thus the depth is an extreme value potentially affected
by false readings. It has been proposed to use an average depth value for corrosion growth
detection. Although by definition the corrosion growth rate is the evolvement of the deepest
point, an average depth value may bring evidence on the existence of active corrosion with a
higher certainty. In UT-inspection it is preferable to use the measurement of the remaining
ligament as a basis for comparison. The depth is really derived from the difference between
remaining ligament and local wall thickness.
Albeit, when calculating a corrosion growth rate for every single defect, the measurement
error can be dominating. If the measurement is repeated many times the effect of
Pipeline Technology 2006 Conference

measurement error can be minimized. For In-Line inspection a repetition of the inspection is
not an option. If, on the other hand, the assumption is made, that the corrosion rate is the same
for most defects, the changes in depth can be used as several measurements of the same
corrosion rate. The relevance of the result will thus depend on the number of pairs and the
accuracy of the two tools. A sample histogram of changes in depth is shown in Figure 2.
25
Deviation from zero
20
number of features

15

10

0
-2 -1,8 -1,6 -1,4 -1,2 -1 -0,8 -0,6 -0,4 -0,2 0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8 2
Increase in depth [mm]

Figure 2: The increase in depth for 104 sample pairs of metal loss features as a histogram. There is a
deviation of the mean from zero, that cannot be explained by the pig measurement uncertainty alone.

It is an oversimplification to assign a single growth rate to a pipeline altogether. Areas of


active corrosion would be hidden if the average is taken over all parts of the line that are
potentially not affected. The ambient conditions are rarely unchanged over the whole length
of the line. To account for these problems a certain subset of features should be selected for
assessment of corrosion growth. Several solutions to this problem have been proposed.
Data Segmentation
The data has to be segmented into subsets for which the evidence of active corrosion is
possibly higher than for the set as a whole. For n values of change in depth the number of
possible subsets would be 2n. Obviously this number is too large. Important constraints are
that the subsets should not mix internal and external defects and that they should be in close
proximity. A decision tree method has been proposed in [3] to find these subsets
automatically.
Running Average
Another solution for choosing the right subset of metal loss sites for the corrosion growth
study has been proposed in [4]. The pipeline is divided into regions that constitutes a natural
segmentation of the total distance. Separators could be transitions from above ground to
below ground. It could be envisaged that soil conditions or other information typical for DA1
methods are used as a separator as well. No region should be longer than 1 mile. A running

1
Direct Assessment, A method to assess the integrity of a pipeline based on various sources of information. See
for instance NACE RP 0502.
Pipeline Technology 2006 Conference

average pit depth is calculated of a pipeline and compares it with a previous measurement. In
this case it is not necessary to identify every feature with its counterpart in the older report.
However, the metal loss features have to be abundant to efficiently use this method. This is
another reason, why ,especially for corrosion growth assessment, the pigging vendor should
supply the feature information as detailed as possible. Metal loss defects should not be
grouped into large clusters and even shallow defects can later on contribute to deliver
evidence of corrosion growth.

2.3 Run Comparison with UT-data


Because ultrasonic data has no intrinsic ambiguity, it is also much more valuable long-term.
Archived MFL data that is older than 10 years is often not used for run comparisons. While
for UT-data, advancements in technology have also improved the overall value of the data,
even very old data is still directly comparable to data of very recent UT pigs. Figure 3 shows
a metal loss defect recorded in two inspections. The lower part shows the result of an earlier
inspection that is about 12 years older. The upper part shows the recent results as they were
obtained by NDT's LineExplorer UM inspection tool. The upper diagram in each box is the C-
Scan, the lower one a B-Scan at the deepest point location.
The data of the earlier inspection was governed by echo loss. It is shown in green (dark grey
in black and white reprints). Except for some spots in the weld echo loss is missing in the
recent data. The effect is also seen in the B-Scans. In the earlier measurement the profile is
basically flat (a remaining ligament of 0 mm is the conventional way of indicating echo loss).
Only a few points with a wall thickness deviating from the nominal wall thickness can be
observed. In the recent measurement the full profile is revealed.
The presence of echo loss has an effect on the reported depth. In the first inspection the defect
was reported with a depth of 2 mm. The deepest point is now measured as 3 mm. Based on a
simple feature to feature comparison one may be tempted to conclude that active corrosion
has been present (or still is). The analysis of the actual data, however, reveals that it is much
more likely that the deepest point was masked by echo loss, thus leading to a shallower depth.
The conclusion that the old tool should be run again, in order to have the same masking of the
defect and hence the same depth is not permissible. The distribution of echo loss is a matter of
probability. To account for this uncertainty in a corrosion growth assessment the assumed
accuracy of the tools should be altered. While the recent measurement is likely to fulfil it
stated accuracy level of ± 0.4 mm (0.016 Inch), the old measurement falls short of the
accuracy of ± 0.5 mm (0.02 Inch). Instead a lower level of accuracy, like for instance ± 1 mm
(0.04 Inch) should be assumed. This way also a feature by feature comparison can be carried
out, which would then reveal, whether the data quality is sufficient or not. Here the result
would not be the presence (or absence) of active corrosion, but the lack of evidence.
Pipeline Technology 2006 Conference

Figure 3 A defect seen by two inspections with different UT-tools. The lower inspection data is 12 years
older. The upper picture show the defect as it was seen recently by an NDT pig.

If the investigation is then extended to a pixel-by-pixel comparison, some treatment of the


matrix of wall thickness data is needed to account for the following discrepancies:
• The sensor spacing may be different. Thus the number of lines in the matrix
corresponding to the same circumferential range is different. The matrix with the
lower number of lines can be interpolated.
• The same applies for the axial sampling density. In traditional tools the sampling was
controlled by frequency. Speed variation would thus change the spacing in between
data points.
• The cut-out of the feature box from the C-Scan will always vary. So a means to move
the boxes with respect to each other is required.
• The wall thickness of the nominal wall next to the defect should be compared. If it
doesn't give the same value, a correction should be done.
• Echo loss is the most problematic adaptation. Echo loss can also be interpolated, but
this generates defect profiles that have not really been measured. It would be better to
neglect areas with echo loss altogether. In the example in Figure 3 only the shape of
the metal loss area can be compared.
Figure 4 shows a sample comparison of two matrices of wall thickness values. Both data
values are displayed with the same colour code. The sizes have been adapted to show the
same area of the pipe surface. In the lower part the river bottom profile is shown for the two
measurements. There is a relevant deviation between the two indications – an increase in
metal loss depth.
Pipeline Technology 2006 Conference

Figure 4: Pixel-by-pixel comparison of a metal loss defect. There is an obvious change in the profile of the
flaw. The depth of the metal loss has increased.

3 Run Comparison with Crack Inspection Data


Crack inspection using ILI-tools has also become a widely used inspection task in the pipeline
industry in the recent years. Although not applied on a routine basis, many pipelines are not
only inspected in the case of actual threat, but also in a precautionary manner. However, the
experience in run comparison of subsequent crack inspections is very limited. This has
different reasons.
• Most of the crack inspection is done with the angled beam ultrasonic technique [5].
Other technologies have also been applied like Elastic Wave [6] and Transverse MFL
[7]. Apart from the angled beam UT, none of the technologies ever gained widespread
support. Because of the difficulties in technology, a direct comparison between
different technologies does not seam reasonable.
• The accuracy of the depth measurement is not as reliable as for ultrasonic wall
thickness inspection. Cracks can grow in length and in depth. From a defect
assessment point of view the depth is a crucial measure. Comparing the depth based
on the features list is not very revealing.
• Usually the crack inspection technology is applied if the pipeline is susceptible to
stress corrosion cracking. These types of cracks are typically found in colonies. Crack
colonies are easily detected with the angled beam tools. However, the resolution of
single cracks is often not possible. This would be necessary if the change in size of
single cracks was to be evaluated. SCC is a crack type that potentially grows quite
rapidly. If cracks of this type are found, the affected pipe sections would most often be
replaced and thus are no longer found in a second inspection later on.
Pipeline Technology 2006 Conference

•For corrosion the corrosion rate will immediately allow to calculate the remaining life
of a pipeline and give information on reasonable reinspection intervals. The corrosion
rate is easily found with two inspection runs, assuming the time interval to be
sufficiently large. For cracks the growth rate is governed by fracture mechanical laws.
Both, for SCC and for fatigue cracks the growth rate is given by crack size, stress
levels and time or number of cycles. The remaining life can thus be calculated based
on theoretical considerations. Theoretical models also exist for corrosion growth rates
in pipelines [8], but have rarely been employed, because the required input data is not
available.
Nevertheless the question of comparing results between crack inspections will arise some day
and even here some conclusive results are anticipated. Other advancements in the field of
crack detection will support this development.

4 Conclusion
The data originating from an in-line inspection is nowadays not only used at the time the
report is issued, but will be put in context with other inspection data. This can be especially
valuable in the case of ultrasonic inspection. As pigging vendors and inspection technologies
change, pipeline operators should ensure that inspection data and results are made available at
the time of delivery and are still available to third parties many years later. The advancement
of inspection technology does not diminish the continuity in inspection but allows to draw
conclusions even with pigging results that have been archived for a long time.

References
[1] Technische Richtlinien für Rohrfernleitungen Chapter 12.3.4.2 (Pipeline regulation in
Germany)
[2] Code of Federal Regulation DOT 49 CFR Part 195.452 (j)
[3] Approach to Assessment of Corrosion Growth in Pipelines, B. Gu, R. Kania, S.
Sharma and M. Gao, IPC2002 27243 Proceedings of the IPC’02 4th International
Pipeline Conference 2002, Calgary
[4] S. W. Rust, E.R. Johnson, Method identifies active corrosion, Pipeline and Gas
Technology, June 2002, p.16-22
[5] H.H. Willems, O.A. Barbian, and N.I. Uzelac, "Internal Inspection Device for
Detection of Longitudinal Cracks in Oil and Gas Pipelines - Results from an
Operational Experience", ASME International Pipeline Conference, Calgary, June 9 -
14, 1996.
[6] G.H. Wilkie, T.J. Elm, D.L. Engen, Comparison of crack detection in-line inspection
tools IPC 2002, Proceedings, 4th International Pipeline Conference 2002
[7] Field Tests demonstrate TFI detects long seam weld defects, P. Mundell, K. Grimes,
PipeLine&Gas Industry, 82(6), (1999)
[8] Norsok Standard M-506, CO2 Corrosion rate calculation model

This paper is an updated version of a contribution the 7th International Conference and Exhibition on Pipeline
Rehabilitation & Maintenance 2005, Bahrain

You might also like