You are on page 1of 9

Proceedings of the 2022 14th International Pipeline Conference

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

IPC2022-86934

NOT ALL DATA IS GOOD DATA: THE CHALLENGES OF USING MACHINE LEARNING
WITH ILI.

Adrian Belanger Dane Burden Paul Dalfonso


T.D. Williamson T.D. Williamson T.D. Williamson
Houston, TX, USA Salt Lake City, UT, USA Salt Lake City, UT, USA

ABSTRACT are correlated to the predicted geometries of interacting ILI


Machine learning (ML) has grown extensively in most signatures that do not always align one-to-one. The estimated
industries, with learning models driving state of the art position on the pipe is made from onboard ILI tool instruments
performance in a variety of tasks. The in-line inspection (ILI) and is subject to measurement accuracy. The small errors
industry is no exception: applications of machine learning contained in each measurement method can make precisely
techniques have provided promising results for a wide range of matching the NDE measurements to the ILI measurements a
needs. Metal loss anomaly sizing, fitting classification, and tedious effort and minimizing the compounding errors requires
identification of interacting threats have all benefited from diligence.
different forms of learning models. The success of any learning The one-to-many and direct-to-indirect measurement
model requires detailed attention at all stages of the process, as relationships between NDE and ILI make generalizing input
small nuances often manifest misleading results. features and labeling training data a challenging task. In
The ability of an ILI data analyst to accurately identify addition to overcoming the difficulties associated with
pipeline anomalies, most importantly anomalies that affect the correlating and labeling training data, significant attention must
integrity of a pipeline, is based on experience. Experience can also be given to the distribution of features being represented.
be sub-divided into two categories, observations and truth, This includes metal loss geometry as well as the engineered
which allow a data analyst to identify patterns and make model input features. With dozens of dimensions included in the
predictions. Supervised learning models mimic this process by input vector of supervised learning models, representing the
using a mapped set of inputs (observations) to outputs (truth) to possible permutations can be overwhelming. Using NDE
develop a mathematical function that can be applied to new external laser scans to maximize training data and
examples. The input parameters, referred to as training data, are dimensionality reduction techniques can help, but do not remove
a vector of engineered features relating to the desired output. all underlying contributions.
Like a data analyst’s experience, the quantity, quality and After curating, developing, and training a model, it is
representation of the training data directly influence the possible to assess the associated performance, but this should be
performance of predictions made by the model. approached with cautious optimism. To accomplish this, a
Data curation, a process that includes the collection, general machine learning practice is to hold out a portion of the
analysis and labeling of data used to train the mathematical training data to test model predictions while an iterative
model, is critical and time-consuming. When curating data for development cycle refines the engineered features, model
the development of a supervised model, two main obstacles must architecture and training process. Since no two pipelines or
be overcome. First, data from two measurement methods, for inspections are identical, validation against blind data is crucial.
example, magnetic flux leakage (MFL) and in-the-ditch non- Although the standard training hold-out practice may make data
destructive examination (NDE), must be precisely correlated. appear to be blind, it often includes highly correlated features
Second, the training data must represent the population of future and the iterative development cycle may inadvertently lead to
predictions and minimize coverage error by containing enough bias. Attention to the distribution of blind hold-out data helps
appropriate examples. Metal loss anomalies examined by NDE

1 © 2022 by ASME
verify the generalization of a model and more accurately the standard reported features of depth, length and width.
represents future performance. Regression models have been used as far back as Gauss in his
This paper describes the development of an ILI machine modeling of the heavenly bodies [6] and are a staple of
learning model at each stage of the process and contrasts the empirically derived engineering models [7].
performance of correctly utilizing data in training and testing to As computational speeds and memory advanced, so did the
that of incorrectly utilized data. methodologies that used them. AI got its real start when
mathematician and computer scientist Alan Turning asked the
NOMENCLATURE question, “Can machines think?” [8] Symbolic AI was the first
AI Artificial Intelligence attempt at modeling how to think using predefined rules, much
AUT Automated Ultrasonic Technology as humans do with logic, mathematics and language.
DEF Internal Geometry or Deformation Symbolic AI developed considerably over the course of the
Measurements 1960s, but due to the a priori knowledge that had to be
DVP Dig Verification Program programmed into these systems they began to be supplanted by
EMAT Electromagnetic Acoustic Testing neural networks. Neural networks are a form of AI called
GPS Global Positioning System machine learning designed to learn from data that is presented to
GPU Graphical Processing Unit it. There is no need for machine learning to be programmed with
ILI In-line Inspection explicit sets of instructions and symbols.
KNN K-Nearest Neighbors Classification Model Neural networks were used on many ILI problems from
LFM Low Field Magnetic Flux Leakage classifying features to determining the severity of mechanical
MDS Multiple Dataset damage [9].Neural networks derive their name from the original
MFL Magnetic Flux Leakage idea to mimic the connections of neurons in the brain. However,
NDE Non-Destructive Examination even today these methods are grossly oversimplified when
PAUT Phased Array Ultrasonic Technology compared to the human brain. Neural networks comprise layers,
SMFL Spiral Magnetic Flux Leakage each one containing a node (see Figure 1). The node is basically
UT Ultrasonic Testing a function that takes in inputs, processes them and then sends out
outputs that may be the inputs to another layer. These nodes
1. INTRODUCTION have numerical coefficients associated with them and they
Inspection of steel pipelines with magnetic flux leakage “learn” by adjusting these coefficients to match their outputs to
(MFL) tools has been around since the 1960s. At first, the data the expected outputs they are being trained on. The data set used
from these tools was analog with coarse circumferential and for this learning step is called the training set. After being
axial resolutions measured by induction coil sensors. The sizing trained, the model outputs are compared to data it has not been
algorithms used were mostly based on the amplitude of the trained on, called the validation set.
recorded signals with a lot of analyst interpretation. Since then
things have changed considerably. During the 1990’s two
revolutions occurred in the in-line inspection (ILI) community:
advances in measurement technologies with the application of
hall sensors and ultrasonic testing (UT), and the computer
revolution that ushered in powerful desktop digital processing.
The latter continued into the 21st century with data processing
power growing exponentially and breakthroughs in artificial
intelligence (AI) including machine learning and deep learning.
Now, in the 2020s, we are seeing the impact of these
advancements over a range of ILI technologies used in pipeline
integrity [1]‒[4].
Machine learning tools are powerful but “with great power
there must also come great responsibility.” [5] There are two
regimes that must be addressed in building a reliable machine
learning algorithm: the type of machine learning methods to use
and the proper curation of truth data.

2. A CURSORY WALK-THROUGH OF LEARNING


MODELS
In early days of ILI, depth prediction algorithms were
mainly some combination of regression models. Manufactured FIGURE 1: SCHEMATIC OF NEURAL NETWORK
anomalies were put into pipe with very controlled dimensions,
which allowed for the construction of regression curves based on

2 © 2022 by ASME
There have been many adaptations and variations on the generalize well to new examples. This is generally overcome by
regression fitting of nodes, including trees, forest, boosting. and, introducing additional training examples that more
most recently, deep learning. comprehensively represent the prediction space. The quantity of
Trees and forests use nodes but instead of assembling them training data may determine how sophisticated the model can be.
in layers they structure them in branches and leaves. Boosting A small data set may be more prone to overfitting when using
guides the learning process by combing the outputs of many deep learning techniques, whereas it may be more appropriate to
weak learners to obtain an overall strong model. Deep learning, apply random forest or gradient boosting [10].
the next generation of neural nets, exploits the ability to mine
large amounts of data using the computational power of
specifically engineered processing units including graphical
processing units (GPU). Deep learning models have enormous
amounts of nodes in many layers densely connected to each
other.
Another technique worth mentioning are support vector
machines, which belong to a category of pattern recognition
algorithms called kernels. This technique takes in an array of
features for each data point and operates on the distances
between these data points, treating the feature array as a
coordinate in a higher dimensional space. Support vectors are
(a) PIPELINE (A) METAL LOSS (b) PIPELINE (A)
very good at grouping similar data points and separating groups
DEPTH DISTRIBUTION RESIDUAL ERRORS
into classes.

2.1 Representing the Prediction Space


When choosing an machine learning model, it is very
important to understand the nature of the data that will provide
the inputs into the model and what the model is expected to
predict. Are you trying to predict a feature type such as a mill
anomaly from an internal pit? Are you trying to estimate the
dimensions of a corrosion pit? Or are you more interested in the
estimated severity of a mechanical damage event?
It is important to define the prediction space, that is, the set
of all expected predictions of the model, because overly (d) PIPELINE (B) METAL
generalizing the space can miss critical exceptions and trying to (c) COMPARISON OF
PIPELINE (A) & (B) LOSS ERRORS USING
model every possible prediction can make the problem METAL LOSS DEPTH MODEL TRAINED ON
intractable. DISTRIBUTION PIPELINE (A)
The type of prediction is also important. There are three
basic types of predictions: classification, clustering and
FIGURE 2: COMPARISON OF A MACHINE LEARNING MODEL
regression. Classification matches the data to known classes
TRAINED ON A SPECIFIC PIPELINE APPLIED TO ANOTHER.
based on how well those features match data known to belong to
that class. Clustering matches how well data points match each The choice of model is important, but the true strength of
other and makes distinctions between different populations. machine learning is the data used to train the model. The
Regression makes quantitative predictions based on the prediction space needs to effectively represent the expected
relationships of previously measured data matched to known predicted values that the model may return. Using large amounts
predictions. of data over a small subset of the whole prediction space will not
In addition, there are supervised and unsupervised learning. produce accurate results for data outside of that training set.
If the training data has known properties, then the model can be Training data can also have unknown biases that do not show up
trained to predict known data; this is called supervised learning. until the model is presented with new data that has different
Unsupervised learning is a method of finding patterns and parameters. Figure 2 shows results of training a model using only
features in data that do not have known properties. data from one pipeline and validating it on a completely different
Choosing an machine learning model can be dictated by the pipeline. The training set used to build the model (figure 2a)
data available for training and the complexity of the problem to shows excellent residual errors (figure 2b) with 80% of the
solve. In addition to the type of predictions expected by the populations within ±0.05t. However, when this model is applied
model, the quantity and quality of data points used in training to another pipeline with different conditions, such as grade of
and validating are critical factors. If the problem is very complex pipe and distribution of corrosion types, it performs sub-
or there is high correlation between the sets of data, then the optimally with a bias to overcall depths (figure 2d). The solution
prediction space will be too narrow, and a model will not to this is either to train the model to be pipeline independent or

3 © 2022 by ASME
to detect and tune to pipeline differences by training it on data a digital gage to the pipe with a magnetic bridging bar can ensure
across all the cases to be inspected. a more level mounting on the pipe surface, reducing errors due
to rocking or tilting of the gage. Both gages are limited, however,
3. DATA CORRELATION by the width of the point extending into the metal loss.
ILI uses various technologies to gather data on the condition Because measuring only the deepest point can be difficult
of a pipeline while traveling internally through it. Technologies and it ignores other zones in the area, aligning the appropriate
include magnetic flux leakage (MFL), ultrasonic testing (UT), ILI signal associated with the actual deepest metal loss region
electromagnetic acoustic testing (EMAT) and internal geometry may not be feasible. Since areas of corrosion often manifest
measurements (DEF). Inspections can be done with separate multiple ILI signal responses there can be a one-to-many
tools; however, their results must be aligned for comprehensive relationship of NDE-to-ILI. This can lead to correlation errors
integrity analysis. Aligning data requires matching the position between ILI and NDE by assigning an ILI training sample
and orientation of each inspection and accounting for the errors (signal) to the incorrect target NDE value (measured depth).
associated with each tools’ method of measuring axial and
circumferential orientation. Pattern matching techniques of the
datasets can be used to refine alignment, but measurement
uncertainty will always be present, which complicates utilizing
separate inspections. The error resulting from matching ILI to
ILI or NDE to ILI will be referred to as correlation error. One
way to minimize ILI to ILI correlation error is to run multiple
technologies on a single inspection tool.
When an ILI is completed using multiple datasets, a more (a) MANUAL GAGE (b) DIGITAL GAGE
comprehensive view of what is happening on the pipeline can be FIGURE 3: EXTERNAL PIT GAGES
visualized.[11] This allows for the development of more
comprehensive machine learning model inputs. By capturing
NDE data collected at precisely known locations on the pipe, an
alignment can also be made with each of the ILI datasets,
building up the prediction space and reducing the error
associated with correlation.

4. USING NDE VALIDATION DATA


In the early days of ILI, NDE validation digs were done to
assess the performance of ILI tools and recalibrate an ILI run if
the results were not within the tool stated specifications. Since
then, the accuracy of both ILI tools and NDE examination
techniques has continued to improve, generally through
technology enhancements. However, improvements to ILI
predictions also rely heavily on the accuracy of NDE
FIGURE 4: MANUAL UT PROBE
measurements, as these are the target data for model training and
testing. Although some NDE technologies are extremely Internal measurements have similar limitations as a hand-
accurate, the contributions from in-the-ditch environments, held UT probe (Figure 4) is used to find and size the anomaly
human factors, time constraints and feature correlations present and is subject to errors in positioning and coupling to the surface
opportunities for additional error [12] to be included in the model of the pipe. In addition, probe measurements are typically taken
development. Therefore, careful data curation is a crucial step on a grid and are limited by the grid resolution. In a conventional
when developing any predictive model. Curation is the process UT scan the probe used is a straight (or normal) beam which
of aligning ILI and NDE data sets and evaluating the quality of reflects off the bottom surface of internal corrosion, returning a
both to determine the applicability of labeling target data. measurement of the remaining pipe wall. This method can miss
pinhole pitting, narrow mill anomalies or even the peak depth
4.1 Manual Tools that may be detected by MFL. Generally, only the maximum
Analog and digital pit gages (Figures 3a & 3b) are measured depths are recorded in the dig reports, with little else
commonly used in a standard excavation dig to measure metal about the morphologies or sizes of the anomalies. This limits the
loss due to external corrosion. Typically, only the deepest point available data from which to develop length and width models.
in an area of general corrosion will be measured or . Drawbacks
to using either type of pit gage exist [13]. For example, analog 4.2 Automated Tools
pit gages rest on the outer surface of the pipe, so if corrosion is Since the late 1990s, automated systems have been used to
extensive, the gage may not sit level with the nominal surface of map pipe surfaces. Laser and coherent light profilometry is used
the pipe and will not accurately measure the pit depth. Securing

4 © 2022 by ASME
to map the external surface [13] and automated ultrasonic multiple dataset (MDS) tool in a joint of test pipe. Standard ILI
technology (AUT) [14] maps the internal surface. and NDE reports use boxes to identify and correlate metal loss
Figure 5 shows the use of a handheld profilometry unit in features, with NDE generally only providing information on the
the ditch and the associated digital output of the scanned area. deepest feature in the area. This is a poor utilization of the true
The pipe is prepared by removing a small area of coating and power of using profilometry data to validate ILI results, as
placing some magnetic markers for reference calibration. The evident in Figure 7, where far more comparisons could be made
unit is manually positioned and the software automatically if signal to signal matching occurred. In cases where corrosion
calculates the surface profile of the pipe at resolutions much finer is complex, for example, matching boxes can lead to
than can be achieved with pit gages. The data is stored in digital misalignments as overlapping areas can include pits and signals
formats that can easily be converted into usable formats by ILI not used in the calculation of the boxes. Using actual depth
providers and integrity consultants. Figure 6 shows an AUT measurements at each axial and circumferential position allows
setup that can provide resolution and digital outputs similar to metal loss to be mapped directly to the ILI response location.
that of laser profilometry. The features from the ILI data can then be mapped directly to the
NDE data, minimizing the correlation error, maximizing the
number of correlated points and creating a larger model
prediction space to be developed.

FIGURE 5: OPTICAL PROFILOMETRY

FIGURE 7: ALIGMENT OF ILI DATA AND PROFILOMETRY

5. The Dig Verification Program

Dig verification is a process where an operator exposes the


pipe and performs NDE to validate the calls made using an ILI
tool.
Traditionally, the dig verification process has been
executed in the manner shown in Figure 8. The vendor performs
ILI runs using information collected from a pipeline
questionnaire completed by the operator that includes details
FIGURE 6: PROFILE OF INTERNAL CORROSION USING AUT
about navigability, product and pressures used during the run.
The vendor then issues a preliminary report providing
Figure 7 shows an alignment using high resolution information the operator needs before performing verification
profilometry data (Creaform HandySCAN) and the digs. The verification digs from the preliminary report can be
corresponding MFL dataset from a T.D. Williamson (TDW) used to assess the performance of the inspection and determine

5 © 2022 by ASME
if any analysis adjustments may be warranted. Any applicable planning of the Plan step. This transition from Act to Plan is a
adjustments can then be implemented into the final report. More decision process made by operators, vendors and third-party
verification digs can be done after the final report, and the cycle consultants. Planning for NDE at this stage will direct future
of assessing performance and implementing applicable attention of ILI data analysis toward an operator’s most pressing
adjustments and refinements can be repeated until a stable set of integrity threats, assist in the selection of appropriate NDE
accurate and actionable information is achieved. methods and facilitate the logistics of selecting dig sites. The
Check section is modified so as not to explicitly state digs
because the integrity planning is done in the Plan stage.
Use of the continuous improvement cycle of providing
successive feedback allows digs to be augmented with historical
and previous ILI data. The key in this process is the input from
Act into Plan, which provides a means to leveraging historical
data and continually improving high-resolution ILI. This
minimizes future reliance on validation digs and provides greater
confidence in using ILI data in integrity programs.

6. MODEL ENGINEERING USING NDE DATA


NDE dig feedback data is generally provided in the form of
a PDF or spreadsheet report. In a manual examination using
either a pit gauge or a handheld probe, the report will contain
pertinent information such as distance to a reference feature, e.g.
distance to an upstream weld, the orientation of the observed
metal loss deemed associated with the reported ILI feature and
FIGURE 8: THE DIG VERIFICATION PROCESS its measured depth. The report may also contain other
information, such as the feature length and width and pictures of
5.1 Plan, Do, Check, Act the area of corrosion as found in the ditch. Occasionally,
The Dig Verification Program (DVP) takes validation a step measurements of adjacent pitting may be included, although
further. In accordance with the API 1160 “Plan, Do, Check, Act” verification and repair digs usually target a specific feature and
(PDCA) [15] process, a system of continuous improvement is much of the surrounding area is not included.
outlined as shown in Figure 9. The flow is similar to Figure 8 An automatic NDE survey, such as laser profilometry or
but there are some subtle differences. AUT, will contain more data. This usually includes a C-scan of
Integrity Planning the area including dimensions of each of the individual
Integrity Feedback anomalies contained within the area. This provides a much more
comprehensive view of the investigated area, allowing for a more
Internally and to Plan accurate ILI correlation.
providers ILI Run Figure 10a provides an example of the C-scan from laser
profilometry with individually identified features with their size
Figure classifications. Figures 10b – 10d show the corresponding
signals of three components of the MFL from an ILI tool using
Remediation Act axial magnetization. A qualitative examination reveals the areas
& Repairs Analysis of the metal loss that have good correlation with MFL signals
Do and the areas where the technology is not as sensitive. Any
model inputs engineered from the MFL signals should have a
correlation to the physical dimensions of the geometry measured
Reassessment Preliminary by the NDE. Model inputs engineered from isolated machined
/ Regrade anomalies oversimplify the geometries of real-world corrosion
Check and thus have fundamental modeling limits. While C-scan
Validation Assessment images from profilometry reports are helpful in correlating NDE
data with ILI, they do not provide actual axial-circumferential
Final Report array of depth values. These values can easily be exported from
the NDE software and are invaluable in advancing the
FIGURE 9: PDCA CYCLE
development of sizing models.
Customer questionnaire data is replaced by the integrity
management data provided from the integrity feedback in the Act
step of the PDCA which is then provided to the integrity

6 © 2022 by ASME
profilometry depth data and the yellow contour lines are the pit
dimensions called by the NDE. The bottom picture is the ILI
MFL data, and the red squares are the pit dimensions called by
the ILI. Looking at the axial-circumferential data displayed in
the image, the MFL signals can be matched to the corresponding
measurements in the NDE data. The discrepancy in cluster
sizing is due to the implementation of interaction criteria.
Aligning both data sets enables a more accurate correlation of
ILI and NDE.

(a) LASER PROFILOMETRY (b) MFL - AXIAL 7. SIZING MODEL EXAMPLE


This section will demonstrate the relationship between the
quality of the training data with its derived features and the type
of machine learning model used.
The examples shown in Figures 12 and 13 are k-nearest
neighbor (KNN) models. A KNN model is a regression model
that makes predictions based on the closest trained examples
from the feature space of the k-nearest neighbors.[17] machine
learning models typically use splits of the training data, using
some percentage of data for the model training and the remaining
data to test (validate). Splits are created by partitioning the data
(c) MFL - RADIAL (d) MFL - CIRCUMFERENTIAL differently for each split based on statistical methods and then
synthesizing the results in the final model allowing all training
FIGURE 10: COMPARISON OF MFL RESPONSE TO data to be used. It also minimizes overfitting to the data and
MEASURED PROFILE OF GENERAL CORROSION. enables estimation of the model performance on data the model
was not trained on. The schemes used for training and test data
As the resolution of NDE data and ILI data increases, cross- splits are extremely important as test data that reflects an unseen
correlation issues that had traditionally been intractable can now example best represents future prediction performance.
be addressed. Metal loss clustering, for example, is based on Often a randomly sampled 20% is withheld from the training
anomaly dimensions and the distance between adjacent dataset. This provides enough data (80%) to train the model
anomalies. Since ILI and NDE are based on different while leaving a remaining dataset against which to assess how
measurement methods, these calculations often do not match. well the model predicts. However, it is imperative to carefully
However, they can be reconciled if the raw data from both consider if the examples in the test dataset represent the
surveys is compared and correlated [16]. population of future examples. If the test dataset does not
represent future examples and is too similar to the training
dataset, then the model will learn how to perform well
specifically on the test dataset and not generalize to future
examples. Any performance metrics derived from this scenario
will be inflated as performance on new examples that differ from
the test dataset will be worse. The following examples utilize a
unique test dataset sampling scheme designed to significantly
improve the test dataset’s representation of future examples.
The following is three examples of how training data and
model selection affect performance. The validation data is the
same for all three scenarios and is the curated field measurements
for five different test splits.
Figure 12 shows a KNN model that was trained on machined
defects with simple well-defined geometries and its performance
on the validation data. As can be seen, the model performs
FIGURE 11: COMPARISON OF LASER PROFILOMETRY poorly and has a wide error distribution of predicted depths to
DATA (TOP) WITH ILI MFL DATA (BOTTOM) AND THE TWO
actual measured values. This is mostly due to the differences in
OVERLAIN(MIDDLE). RED BOXES ARE ANOMALY
DIMENSIONS CALLED BY ILI AND YELLOW CONTOURS signals from machined anomalies to field data. Though
ARE DIMENSIONS CALLED BY NDE. machined anomalies are a good foundation for analytical models,
machine learning models need to be trained on the data that has
Figure 11 shows an example of discrepancies between ILI features similar to the ones they are expected to predict.
and NDE metal loss calls. The top image is the C-scan of the

7 © 2022 by ASME
Figure 13 shows the KNN model retrained using field data
instead of machine pull test features. As can be seen there is a
dramatic improvement in performance, with a tighter error
distribution. However, the field data contains a population
imbalance that has far more samples in the moderate depth range.
This introduces a bias into the model as the majority of neighbors
are in the moderate depth range and highlights the importance of
having representation over the full range of geometries to be
sized.
For the KNN model, using curated field data results
improves accuracy by relying on nearest neighbors. In densely
populated regions it performs well but struggles in regions more
sparsely populated with feedback. Switching to a gradient boost
(XGB) model allows for a more adaptive fitting using an
ensemble of many small decision trees that each minimize the
residual prediction errors [18]. Figure 14 shows the result of
using the XGB model. The linearity of the unity line has been
captured and the mean error is lowered.
A special note should be made regarding Figure 14. Split 1
shows a subpopulation that significantly differs from the other
points in the validation data (blue box). This subpopulation is
clearly an outlier from the rest of the population and is not
represented well in the training data. By identifying this, a
FIGURE 12: RESULTS OF TRAINING A K NEAREST detailed review of those specific anomalies can be done and new
NEIGHBOR (KNN) MODEL ON MACHINED ANOMALIES AND model inputs can be engineered or changes in the modeling
PREDICTING DEPTHS OF FIELD CORROSION. architecture can be made to improve performance.

FIGURE 13: RESULTS OF TRAINING A K NEAREST FIGURE 14: RESULTS OF TRAINING A XGB MODEL ON
NEIGHBOR (KNN) MODEL ON FIELD DATA AND FIELD DATA AND PREDICTING THEIR DEPTHS. AN OUTLIER
PREDICTING THEIR DEPTHS. SUBPOPULATION IS HIGHLIGHTED IN BLUE.

8 © 2022 by ASME
8. CONCLUSION tree induction." Journal of Artificial Intelligence
As demonstrated in this paper, machine learning is a Research (2003): 315–354.
powerful tool, but only when carefully applied. Data used in [11] Kirkwood, M., Burden, D. and Maldonado, M., “The Art of
model development requires diligent curation, precise Looking: An In-Line Inspection Perspective,” Proceedings
correlation and accurate measurements. Subject matter experts of the 2019 Pipeline Pigging and Integrity Management
play a vital role in selecting a model, engineering its inputs and Conference, Houston, TX, USA, February 18-22, 2019, No.
checking the prediction space is well represented. Operators can 10.
aid in the development by utilizing NDE techniques that [12] McNealy, R., McCann, R., Van Hook, M., Stiff, A. and
maximize the number of measurements per dig and provide Kania, R., “In-Line Inspection Performance III, Effect of In-
detailed information on anomaly geometry. Ditch Errors in Determining ILI Performance,” Proceedings
of the 8th International Pipeline Conference, IPC2010-
ACKNOWLEDGEMENTS 31269, pp. 469-473, Calgary, Alberta, Sept 27 – Oct 1, 2010,
Tod Barker, Jeremy Clark, Josh Yunik, Robert Coleman, https://doi.org/10.1115/IPC2010-31269
Chris Goller, Jed Ludlow, James Simek, Andrew Moore, Callie [13] Tomar, M., Fingerhut, M., and Yu, D., “Qualification of ILI
Lamb and Eric South. Performance in Accordance with API 1163 and the Potential
Impact for Management of Pipeline Integrity,” Proceedings
REFERENCES of the IPC2008, 7th International Pipeline Conference,
[1] Burden, D., Dalfonso, P. and Belanger, A., “The Current IPC2008-64469, pp. 693-703, Calgary, Alberta, Sept 29-
Progeny of In-Line Inspection Machine Learning,” Oct3, 2008, https://doi.org/10.1115/IPC200864469
Proceedings of the 2020 Pipeline Pigging and Integrity [14] Feole, R. A., “Automation of Fast Ultrasonic Technique for
Management Conference, Houston, TX, USA, February 17- Pipeline Weld Examinations,” PRCI, 1998, PR-250-9612,
21, 2020, No. 54. Cat. No. L51777.
[2] Conrad, B., Kissel, M., Tse, V., Barlow, J., Chen, J., and [15] American Petroleum Institute. (April 2019). Managing
Kania, R., “Considerations for validating an ILI technology: System Integrity for Hazardous Liquid Pipelines. API
controlled implementation of a new MFL product,” Standard 1160, Section 4.1
Proceedings of the 2020 Pipeline Pigging Management [16] Lamb, C. and Dalfonso, P., “How ILI Feature Grouping Can
Conference, Houston, TX, USA, February 17-21, 2020, No. Impact Apparent Accuracy from NDE,” Proceedings of the
19 2022 Pipeline Pigging and Integrity Management
[3] Chakraborty, I. and Vyvial, B., “Using deep learning to Conference, Houston, TX, USA, Feuruary 2-40 2022, No.
identify the severity of pipeline dents,” Proceedings of the 43.
2020 Pipeline Pigging and Integrity Management [17] Goldberger, J., Rowels, S., Hinton, G. and Salakhutdnov,
Conference, Houston, TX, USA, February 17-21, 2020, No. R., “Neighbourhood Components Analysis,” Advances in
70 Neural Information Processing Systems, Vol. 17, May 2005,
[4] Hurd, G., Elliott, J., Farnie, S., and Sutherland, J., pp 513-520.
“Advancing In-line Inspection technology and pipeline risk [18] Friedman, J. H. 1999, "Greedy Function Approximation: A
management through advanced analytics of big data,” Gradient Boosting Machine" (PDF), IMS 1999 Reitz
Proceedings of the 2020 Pipeline Pigging and Integrity Lecture.
Management Conference, Houston, TX, USA, February 17-
21, 2020, No. 13
[5] Lee, S., 1962, Amazing Fantasy, #15, Marvel.
[6] Gauss, Carl F., 1809, Theory of the Motion of the Heavenly
Bodies Moving about the Sun in Conic Sections, Little,
Brown and Company, Boston, Doi: 10.5962/bhl.title.19023
[7] Box, G., and Draper, N., 1987, Empirical Model-Building
and Response Surfaces, Wiley, New York.
[8] Turing, A., 1950, “Computing Machinery and Intelligence,”
MIND a Quarterly Review of Psychology and Philosophy,
pp 433-460
[9] Bubenik, T. A., Nestleroth, J.B., Davis, R. J., Crouch, A.,
Upda, S., and Afzal, A. K., 2000, “In-line Inspection
Technologies for Mechanical Damage and SCC in
Pipelines: Final Report,” US DOT, OPS, Report No.
DTRS56-96-C-0010.
[10] Weiss, Gary M., and Foster Provost. "Learning when
training data are costly: the effect of class distribution on

9 © 2022 by ASME

You might also like