You are on page 1of 11

SPE-197958-MS

New Corrosion Growth Model Based on Pipeline Intelligent Pigging


Data - Correlating Corrosion Defects Numbers with their Depth

Mohamed Zouheir Trojette, Anouar Zebibi, Abdallah Hammadi, and Khalid Hosani, ADNOC Offshore

Copyright 2019, Society of Petroleum Engineers

This paper was prepared for presentation at the Abu Dhabi International Petroleum Exhibition & Conference held in Abu Dhabi, UAE, 11-14 November 2019.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
As part of their integrity management system, Oil and Gas operators carry out internal inspection of their
pipelines by intelligent pigging. State of the art MFL and UT inspections are used to detect and accurately
size the defects, which are present in the pipeline. The predominant type of defects reported is due to internal
corrosion.
It is well established that corrosion is a naturally occurring phenomena. When the conditions are right for
corrosion to develop, it starts by a single defect or very few defects which are shallow. Then as the pipeline
is operated and corrosion further develops the defects increase in size and numbers.
This paper review several intelligent pigging reports data, and analyze the reported defects in terms of
numbers and depth for several pipelines, in order to establish a correlation (mathematical model) between
the number of internal corrosion defects and their depth. Defects counts will be made and equations will be
developed for several pipelines. These equations will basically establish the number of defects as a function
of their depth or vice versa.
More over when multiple intelligent pigging runs on same line are available, these derived equations
will be compared with the objective to establish a novel model to determine corrosion growth rate in a non-
conventional manner. In fact the models (# of defects and their depth equation) established for different
inspections will be compared and a corrosion rate model establishing the increase in number of features and
their number over time will be thereafter derived.

INTRODUCTION
As part of their integrity management system, oil and gas operators usually carry out inline inspections
(ILI) of their network of pipelines. The inspection referred to as intelligent pigging (IP), or sometimes smart
pigging, allows the operator to check the condition of the pipelines for multiple types of defects. Particularly
corrosion, which is a major threat for pipelines throughout the industry.
It is well established that corrosion is a naturally occurring phenomenon when the conditions are right for
corrosion to develop. For newly commissioned pipelines, corrosion usually starts with a single (or very few)
shallow defect(s) then grows in size. As other shallower defects grow, the entire population is increased
with an associated increase in corrosion defects size.
2 SPE-197958-MS

This paper reviews several intelligent pigging reports data and analyzes the reported defects in terms of
numbers and depth for several pipelines, in order to establish a correlation (mathematical model) between
the number of internal corrosion defects and their depth. Defects counts will be made and equations will be
developed for several pipelines. These equations will basically establish the number of defects as a function
of their depth or vice versa.
More over when multiple intelligent pigging runs on same line are available, these derived equations
will be compared with the objective to establish a novel model to determine corrosion growth rate in a non-
conventional manner. In fact the models (# of defects and their depth equation) established for different
inspections will be compared and a corrosion rate model establishing the increase in number of features and
their number over time will be thereafter derived.

BACKGROUND
Inspection data from pipelines operating within the same oilfield (under comparable operating conditions
in terms of temperature and pressure) were considered for this study. In addition, only inspections that were
carried out by one IP vendor using the same Magnetic Flux Leakage (MFL) technology will be studied. It
is assumed that these conditions will reduce uncertainties.
For each pipeline, the inspection results provide the number of defects, size of each defect (depth, length,
and width), as well as its position along the line (referred to as absolute distance) and its orientation around
the circumference (referred to as clock position). Moreover, the inspection tool has the capability to classify
both the surface location of the defect (internal or external) and its type (corrosion or non-corrosion defect).
The IP reporting requirement, results presentation and format, as well as other subjects related to IP (e.g.
tool performance specifications and inspection run acceptance criteria) are well accessible in a joint industry
guideline referred to as the Pipeline Operators Forum (POF), which is widely used in the oil and gas industry
(Pipeline Operators Forum n.d.). For instance, the software used to analyze the data obtained by the IP tool
classifies the metal loss anomalies according to POF, as seen in Figure 1.

Figure 1—Graphical presentation of metal-loss anomalies (POF, 2016). A=wall thickness or 10mm, whichever values is greater.
SPE-197958-MS 3

SCOPE
The scope of this paper is to:

• First, establish any correlation between the number of internal corrosion defects [N] and their
associated depth [d (%ML)]. It is expected that as the number of defects increases, the sizes
also increase (only one size parameter is considered in this study: the defect depth). The type of
relationship between N and d is of interest, i.e. is it linear, exponential, etc., and is it similar for
the considered lines?
• Then, develop a corrosion growth rate model for each considered pipeline and test that model by
predicting future correlations between N and d to determine the viability of such a model.

Pipelines studied
Three pipelines made of Grade X65 carbon steel were selected. These lines are operating in same offshore
oil field with similar operating conditions in terms of temperature and pressure. Moreover, corrosion
mitigations in place are also similar. Some key data regarding the selected lines are listed in Table 1.

Table 1—Pipeline candidates considered for the study.

Installation Thickness Inspection Year


Pipelines Size Service Length (m)
Year (mm) 1 st 2nd 3rd 4th

A 12" Oil 1981 5825 12.7 Oct 2008 Oct 2010 Jul 2014 Apr 2018

B 12" Oil 1981 1644 12.7 Jul 2008 Oct 2010 Jun 2014 Mar 2018

C 8" Oil 1980 4371 11.13 Jun 2008 Mar 2011 Nov 2014 No inspection

The three pipelines were inspected at different time intervals. Each inspection resulted in an increase in
the total number of reported features with an associated increase in depths. Figure 2 summarize the findings.

Figure 2—Graphical representation of the summary of inspection results from pipelines A, B, and C.

As expected, there is a clear trend of increase in the total number of defects due to corrosion. Moreover,
the increase per year is not constant either for any particular pipeline. For instance, the increase per year
of Pipeline A between 2008 and 2010 inspection is 176 defect per year, whereby the rate decreases to 137
defect per year between 2010 and 2014 inspection. This change in number of defects per year for pipeline
A is shown in Figure 3.
4 SPE-197958-MS

Figure 3—Graph showing the change of rate of defect increase in Pipeline A.

METHODOLOGY
Subsequent to the above holistic approach, the increase in the number of features for each defect depth
(starting from the ILI contractor's reporting threshold of 10%) has been made based on the following
methodology.

Modeling the data


A count of the number of internal corrosion defects for each depth (above 10%) was made for each pipeline
after the baseline inspection (first inspection). Due to the limited number of reported defects above 30%,
the defects depth has been truncated to a maximum depth of 30%. This allows for meaningful sample size
for each depth, which would also dictate the type of function associated with the curve more accurately. A
sample of the data for Pipeline A can be seen in Figure 4.
The data evidently seems to be following a continuous curve. Excel was used to determine the best-fit
curve to help in modeling the data. It is noted from the plot that as d approaches zero, N gets larger. In
addition, as d increases, N approaches zero. These are characteristics of exponential and power functions,
which were considered for the model. The two functions will be tested using Excel to plot them as trendlines
on the data sets. The functions’ coefficients of determination (R2) will then be compared for each inspection.
For the sake of simplicity, Figure 5 gives an example of this comparison where Pipeline B 2008 IP data was
used to evaluate exponential function versus power function.
SPE-197958-MS 5

Figure 4—Data used to evaluate Pipeline B, as an example of the study data.


The chart plots the number of features against the depth of the features.

Figure 5—Trendline comparison between using an exponential function or a power function.

The above comparison test results are summarized in Table 2.

Table 2—Pipeline B best-fit curve test

Curve function Equation Coefficient of Determination (R2) Remarks

Exponential N = 3290e-23.33d 0.984 Best-fit

Power N = 0.025d -4.299 0.964 2nd best-fit


6 SPE-197958-MS

In the above example, the exponential function fits the data best and the power function still gives a good
approximation to the data set. When the same exercise was carried out on the three pipelines, 7 out of 11 of
the inspections showed that the exponential function is the best fit, and is therefore the selected correlation.
Based on the above, the following statement can be made:
If a pipeline has been inspected using a standard MFL Tool and N is the number of internal corrosion
defects reported and d is the reported depth (%ML), then:

Where A and α are two constant coefficients specific to the pipeline inspection.
In order to establish the corrosion growth rate model, the defect depth (d) is of importance. The depth
can be obtained by plotting a d vs. N curve (rather than N vs. d). Similarly, solving a previously obtained
N equation for d would yield the same result. Figure 6 shows the earlier used Pipeline B data plotted in
this manner.

Figure 6—Logarithmic function obtained when plotting d vs. N.

Table 3 shows how the data from Table 2 (Pipeline B 2008 IP) was used to derive the depth equation.

Table 3—Solving previously obtained N equation for d.

Previous equation Solving for d Coefficient of Determination (R2)

N = 3290e-23.33d d = −0.042ln(N) + 0.345 0.984

Accordingly, the following equation is derived:


SPE-197958-MS 7

Where and

Corrosion growth rate model


When two inspection data are available for the same defect then a corrosion growth rate, in its most simplistic
form, can be defined as the rate of change of depth over the time interval between two inspections. This is
the API 570 (Piping Inspection code) definition (American Petroleum Institute 2018).
In an attempt to develop a corrosion growth rate (CGR) model, the above definition was applied on two
data sets from the previously mentioned inspections. The data from pipeline A inspections will be used
to show how the CGR model can be obtained. Figure 7. Pipeline A, first two inspections data. Figure 7
shows this data.

Figure 7—Pipeline A, first two inspections data.

It is observed that most of the data from the 2010 inspection is to the right of the data from the 2008
inspection, indicating an increase in the number of defects for a specific depth. The equations for the depth
of defects as a function of the number of defects can be used to acquire a CGR model as a function of the
depth of defects (d) and time (ΔT). From equation 3, the time interval (ΔT) between Pipeline A 2008 and
2010 inspections is 2 years.
The function of CGR in a specific time interval can be expressed as:

Where: i is the first inspection's year and j is the second inspection's year
 d is the depth of defects function
8 SPE-197958-MS

ΔT is the time interval between the first and second inspection


Once the above is applied to pipeline A, then the CGR model between 2008 and 2010 inspections is as
shown in Table 4.

Table 4—Pipeline A example of CGR model obtained using 2008 and 2010 inspection data.

Function Equation Unit

d08 -0.045ln(N) + 0.320 %ML

d10 -0.039ln(N) + 0.311 %ML

ΔT08,10 2 year

CGR08,10 (d10 – d08)/ΔT08,10 = 0.003ln(N) – 0.005 %ML/year

So, the corrosion growth rate model based on 2008 and 2010 inspections data is as follows:

Corrosion prediction
If the CGR function is assumed to be constant over time, then it can be used to predict the pipeline condition
in the future. Basically, if three inspection data are available for a given pipeline, the CGR can be obtained
using the first two inspections data, then the CGR can be applied to the second inspection to determine the
pipeline's condition at the date of the third inspection. Finally the actual third inspection is used to test how
close the predictions are to the actual inspection data.
If a pipeline has been inspected in year i, then inspected again in year j (where j>i), and CGRi,j is
established as above mentioned, then for any year k, the predicted depth function at year k can be expressed
using the following equation:

Where dk' is the predicted depth at (third) year k


 dj is the depth function of the second year inspection
ΔTj,k is the time interval between the second and third inspection
CGRi,j is the corrosion growth rate model between the first and second inspection data
By comparing dk' and dk then we can establish how accurate the prediction of third inspection is compared
to the actual results.
Table 5 shows Pipeline A 2014 inspection data prediction using the above equation.

Table 5—Pipeline A example for prediction of 2014 inspection data.

Function Equation Unit

d10 −0.045ln(N) + 0.320 %ML

ΔT10,14 3.75 year

CGR08,10 0.003ln(N) – 0.005 %ML/year

d'14 d10+ΔT10,14·CGR08,10 = – 0.028 ln(N) + 0.294 %ML

Therefore, the predicted defect depth equation for Pipeline A at 2014 is as follows:
SPE-197958-MS 9

This is to be compared with the actual defect depth equation for the same line and year, which is:

The above equations are plotted in Figure 8.

Figure 8—Plot of 2014 predicted data and 2014 IP data.

Corrosion Growth Rate model testing


Comparing the figures obtained from the depth prediction model with the actual data from an inspection
run will determine how accurate the model is. To test the model, the predicted and actual depth data from
Pipeline A 2014 inspection was plotted against each other and a regression analysis was made. The analysis
is shown in Figure 9.
10 SPE-197958-MS

Figure 9—Regression analysis of 2014 predicted data vs 2014 IP data.

The regression statistics show a reasonably good fit between the model and the inspection data as
shown in the above graph and is evident by the Correlation Coefficient (R). R Square, or the coefficient of
determination also shows high goodness of fit.

STUDY ASSUMPTIONS AND LIMITATIONS


Clearly, throughout the above analysis several assumptions were made:
1. The study is only applicable to inspection data obtained using magnetic flux leakage technology.
2. Defect depth used were in the range 10% to 30% only and are all classified as internal corrosion.
3. Intelligent Pigging tool depth tolerances have not been considered. While this could affect the depth
equations, it is not expected to have a perceptible effect on the CGR model.
4. Uniform (constant) corrosion rate between each of the pipeline's inspections. While this is plausible if
the corrosion mitigations are in place and the operating conditions are the same, it may not be the case
if major changes have occurred. Usually changes in operating conditions are very rare and take a long
time to have a significant impact (creeping changes), they can however be abrupt such as changing
the pipeline service (from oil to gas for instance) or shutting down the line for an extended durations.
5. The inspection tool has behaved consistently and within spec for all the inspection runs. This is a
reasonable assumption as long as the runs have been declared successful and field verifications did
not reveal any discrepancy based on API 1163 (American Petroleum Institute 2013).

CONCLUSION
The review of intelligent pigging results based on a standard MFL inspection revealed a clear correlation
between the number of internal corrosion defects and their depth, as shown in the following equation:
SPE-197958-MS 11

Where N is the number of defects


 d is the depth of defect in %ML
A and α are two constant coefficients specific to the pipeline inspection.
Moreover, the depth of internal corrosion defects and their count (number) was also modeled as:

Where and
When two inspections for same pipeline are available, then the corrosion growth rate (CGR) model can
be established based on the difference of depth over time in a similar approach to API 570 provisions for
single defect measurements.
The CGR model is also a logarithmic function that uses the above depth equation and is described as:

Where: i is the first inspection's year and j is the second inspection's year
 d is the depth of defects function
ΔT is the time interval between the first and second inspection
The CGR model can be used to predict the future conditions of the line at any year by applying the CGR
model function to the latest inspection as described in the paper and is represented by the below equation:

Where d'k is the predicted depth at (third) year k


 dj is the depth function of the second year inspection
ΔTj,k is the time interval between the second and third inspection
CGRi,j is the corrosion growth rate model between the first and second inspection data
The model has been tested for several inspections from three different pipelines using regression
analysis. The regression analysis showed good fit between the model and the inspection data evident by the
Correlation Coefficient (R) and the coefficient of determination (R Square).

FUTURE LOOK AHEAD


This subject requires further research by pipeline operators to come up with their specific pipeline models
and test them against actual data. Such models can be made to incorporate additional variables and/or
assumptions made in this study. For instance, data from different inspection vendors can be considered.
Moreover, Ultrasonic technology (UT) inspection data can also be evaluated for use with the model. The
impact of inspection tool tolerances on the developed data would also be of interest.
Once the CGR model is well established, it will enable pipeline operators to predict a pipeline's overall
health at any given year with good confidence without having to carry out intelligent pigging. However,
the CGR model is not meant to substitute intelligent pigging inspections, but only to use its results and
supplement it.

REFERENCES
American Petroleum Institute. API 570 Piping Inspection Code. 2018.
American Petroleum Institute. API Standard 1163 In-line Inspection Systems Qualification. 2013.
Pipeline Operators Forum. "Specifications and requirements for in-line inspection of pipelines." Vers. 2016. n.d.
www.pipelineoperators.org.

You might also like