You are on page 1of 7

Integrity Assessment of Pipelines in Operation

Hagbart S. Alsos, Stig Olav Kvarme and Sindre Bruaseth


REINERTSEN AS, Division Engineering
Trondheim, Norway

Bala Krishna Penmatcha, Faisal Reza, Frank Haugen
STATOIL ASA, Transport Net
Krst, Norway














ABSTRACT

The paper discusses integrity management of pipelines in operation.
Typically this involves identification of pipeline response and the
potential risk for failure, e.g. excessive feed-in to lateral buckles,
upheaval buckling, and dynamic behavior of free-spans. Identification
and analysis of such response is not trivial, as large data sets from
external surveys and operational data for large networks of pipelines in
operation need to be processed. Often, history effects from previous
surveys as well as input from Design need to be taken into account to
obtain the full picture of the pipeline behavior. In this work, the
integrity management engineer relies on the set of procedures for
performing integrity assessment, combined with understanding of
pipeline physics.

In order to improve pipeline integrity assessment process, an analysis
tool for performing screening assessment of pipeline response is
developed. This is a combined effort from Reinertsen AS and Statoil
Transport Net. The software tool is developed in order to perform
integrity assessment of pipelines in operation more effective, but
without compromising the confidence of the results. The software
processes spatial input from pipeline surveys i.e. cross-profile data,
along with pipeline design data. This allows pipeline response to be
assessed with reference in design or as-laid configuration. Changes in
response may be monitored over time by comparing the response of
multiple surveys in one analysis. After identification of potentially
critical pipeline response, the pipeline load utilization can be assessed
directly without performing time consuming FE analyses. Instead strain
levels in both vertical and lateral planes are established from processing



the survey data directly. This produces an instant evaluation of the
bending utilization of the pipeline, and is referred to as strain screening.

The paper presents the theoretical background for the strain estimation
methods. The application of the tool is illustrated though a case study
for a 36" gas pipeline on the Norwegian shelf. Further, the paper
discusses the confidence of the various strain screening methods and
their range of application.

KEY WORDS

Pipeline, Subsea, Integrity Management, Finite Element, Local
buckling, Free-spans

NOMENCLATURE

DFI Design Fabrication and Installation
DNV Det Norske Veritas
FE(A) Finite Element (Analysis)
FFT Fast Fourier Transform
GUI Graphical User Interface
HP High Pressure
HT High Temperature
IM Integrity Management
KP Kilometer Point
ROV Remote Operated Vehicle
VIV Vortex Induced Vibrations

420
Proceedings of the Twenty-third (2013) I nternational Offshore and Polar Engineering
Anchorage, Alaska, USA, J une 30J uly 5, 2013
Copyright 2013 by the I nternational Society of Offshore and Polar Engineers (I SOPE)
I SBN 978-1-880653-999 (Set); I SSN 1098-6189 (Set)
www.isope.org

INTRODUCTION AND BACKGROUND

Design of subsea pipelines are performed according to standards such
as DNV-OS-F101 and Companys technical requirements. The design
engineering shall cover all stages of the life span of the pipeline,
including installation operation and other design condition. It is worth
noting that the same requirements which are applied for design, also
applies for integrity assessment during operation. This implies that the
operational history must be followed up to ensure that the pressure and
temperature loads conform with design criteria, and that the pipeline is
responding as expected, e.g. lateral buckles, free-spans and upheaval
buckles. Any change in operational conditions requires further design
assessments and re-qualifications of the pipeline system.

Guidelines for integrity management are presented in DNV-OS-F101
and DNV-RP-F116. The overall integrity management rationale
follows a cyclic procedure consisting of risk assessment and integrity
management (IM) planning. This is followed by inspection and
monitoring, e.g. by performing surveys. After inspection, the integrity
of the pipeline is assessed. If required from the integrity assessment
process, mitigation intervention actions shall be initiated, see Figure 1.


Figure 1: Integrity management cycle, as presented in DNV-RP-F116

Assessment of pipeline structural integrity is typically established by
analyzing external survey data and comparing the results with criteria
from design. The integrity assessment is closely linked to the risk
identification aspects. This provides indications about potential failure
modes, which may take place if one or more design or operation
parameters are violated. The failure modes may be associated with
response such as local buckling, fatigue due to VIV, significant
pipeline movement and bursting of the pipeline. The type of failure
mode that may be relevant for the pipeline defines response to be
analyzed, e.g. fatigue evaluation for in free-spans and evaluation of
global buckling utilization for HP/HT subsea pipelines. For pipelines
designed with corrosion allowance, wall thickness reduction needs to
be monitored. This is, however, not addressed in this paper.

One challenge for engineers involved in IM processes is to identify
response that may lead to failure, and estimate deformations and loads
from survey and process data. During design engineering, detailed FE
analyses are performed in the planning pipeline in-place configuration.
Potential failure and risk aspect is then reflected by the various
guidelines and design analyses. Applying the same level of details for
integrity assessment of pipelines for each new survey would surely be
time consuming, if not unnecessary. After all, all design and
operational aspect is covered by the engineering phase. The relevant
scope of work for integrity assessment is therefore to assess the
potential change in behavior of the pipeline relative to that defined in
the design process. Assessment of pipeline integrity therefore requires
other and more efficient methods to identify and quantify operational
pipeline behavior. It is, however, important that such methods are able
to produce reliable results, which exceed the precision of engineering
judgment. In order to improve the IM process, a software tool for
processing pipeline surveys and estimating pipeline response is
presented. The analysis methods are based on analysis methodology for
Pipeline Integrity Assessment Tool presented by Kvarme et. al
(2012). This paper focuses on the application of the survey processing
methods, and presents comparative results from FE analyses. The
screening and FE analyses are performed for as-laid and hot surveys of
a newly installed 36 pipeline running from an offshore platform to an
onshore terminal in western Norway.


SYSTEM FOR STRUCTURAL INTEGRITY ASSESSMENT

Integrity assessment procedure
The complexity of integrity assessment for pipelines in operation is
typically reflected by the risk for failure and expected response. As
presented by Kvarme et. al (2012), Nes et al (2008) and Etterdal et al
(2008), the integrity assessment can be broken down into three analysis
levels. Each level involves assessment procedures where the
complexity of the analysis is increased with the level value. For a level
1 analysis, the pipeline response is identified. This involves
identification of global buckles, large free-spans, axial movement and
any change in response over time. If the pipeline configuration is
unchanged, no further analyses are required. The level 2 assessment is a
continuation of the level 1 assessment. At level 2, the pipeline response
is typically quantified through screening analysis. This may for
instance involve assessment of strains in lateral buckles and over rough
terrain. Free-spans may furthermore be assessed based on criteria from
analysis databases generated during design engineering. If anomalies
are detected, the integrity assessment proceeds with a detailed level 3
analysis. This involves detailed FE analysis and is typically performed
if the pipeline is subjected to unstable response not predicted during
design engineering, or if there is a plan for a system upgrade. If
anomalies are detected for both level 2 and 3, mitigation actions are
required as indicated in Figure 1.

Software toolbox
The software tool presented in this paper is developed by Reinertsen
AS together with Statoil Transport Net. The tool contains and a set of
computerized procedures and analysis tools for detecting and analysing
pipeline response from survey data. Focus is mainly placed on
improving the efficiency of level 1 and 2 assessments. The tool
contains a wide range of visualization possibilities, and may be applied
to compare pipeline response obtained from several surveys
simultaneously. By establishing both on-bottom configuration and
bending loads at the same time and combined with design analysis data,
results may be visualized together with the survey configurations. This
feature enables users to efficiently identify changes in pipeline response
over time, and to compare the response observed from survey with
assumptions/predictions made during the design engineering.

The software is compiled from the Python package to a single
executable, and has a user friendly graphical user interface. Typical
functionalities for the integrity assessment software are; evaluation of
pipeline configuration based on multiple surveys, identification of
lateral buckles and large free-spans, estimation of pipeline strains,
estimation of free-span fatigue and analytical assessment of axial forces
and feed-in into lateral buckles. If design criteria are unavailable for the
integrity management engineer, the softwares analytical module may
421

be applied to establish simplified criteria for analysis. The layout is
illustrated in figure 2.


Figure 2: Pipeline structural assessment tool.

An example of on-bottom integrity assessment using the survey
processing functionalities is illustrated in figure 3. In this case the
survey of a pipeline is processed for evaluation of bending strains in the
vertical plane along with an assessment of free-spans compared with
free-span criteria from the DFI documentation.




Figure 3: Assessment of free-spans and on-bottom loads based on
survey input and DFI resume data. A: showing vertical pipeline
configuration, B: showing bending strain distribution in the vertical
plane, C: free-span gap with gap criteria, D: Free-span length versus
length criteria. Gray vertical columns in figure B indicate rock covers.
In the following sections, pipeline buckling and level 2-3 estimation of
buckle utilization are described in further details. The technical
background for the strain estimation and criteria algorithms are
presented.


ASSESSMENT OF GLOBAL BUCKLES

Pipeline surveys provide information about the position of the pipeline
and identified anomalies, e.g. debris near the pipeline, coating damage,
excessive anode consumption, long free-spans and large lateral
displacements. Based on this information it is possible to assess the
pipeline response that takes place. Examples on unexpected response
may be formation of new lateral buckles, excessive feed-in into one
single buckle, or lateral pullout from interference with third party
actions. This may be assessed by simply comparing the pipeline
position from several surveys. Establishing the loads from response is
however not straight forward. As surveys are typically performed using
ROV based pipeline tracking, inaccuracy in positional accuracy may
emerge from interfering noise from poor visibility, insufficient
measuring accuracy, errors from the marine contractor, or harmonic
motions from the pipe tracker, as illustrated in figure 4. In order to
better assess the true deformation, FE approaches are often applied for
re-calculation of pipeline response. However, this is time consuming
and requires detailed information about seabed conditions, pipeline
parameters and conditions, and operational conditions.



Figure 4: Example on extreme survey noise from pipe tracking

As the FE approach is time consuming, and in many cases, unnecessary
for the level of details required for screening, Reinertsen AS together
with Statoil Transport Net have implemented analysis algorithms for
handling survey noise. Such methods involve smoothing and filtering
methods to remove noise and establish pipeline loads directly from
survey data, see Kvarme et al (2012). The methods have proved
consistent and display reliable results compared with estimates from FE
analyses. The methods are further investigated in the following.

Strain Estimation Methods
Bending strains obtained from the coordinates of a pipeline survey can
ideally be established from the curvature description of its slope. This
is described by the second derivative of the pipeline route.

=
d
2
ds
2
(s) (1)

wheres denotes the slope of the route and f is the survey function. This
may be derived further into an expression in the Cartesian system to
describe the curvature as a function of x, and y coordinates.
A
B
C
D
422


=
d
2
](x)
dx
2
_1+j
d](x)
dx
[
2
]
32
(2)

As indicated above, the problem with such a description is survey
noise. By taking the derivative to a function subjected to noise, the
noise itself will be magnified. If survey errors are sufficiently
pronounced, the second derivative of that function may produce results
where the noise is dominant and the physical response may be difficult
to interpret. In order to overcome this problem two survey processing
methods are applied:

1. Survey smoothing by polynomial regression
2. Curvature filtering

The polynomial fitting method is illustrated in figure 5. The method
first requires the longitudinal position (often referred to as Kilometer
Point, KP) of a lateral buckle to be identified (buckle apex). Thereafter
a polynomial fit is made for the global buckle using a higher order
polynomial expression. The length and order of the polynomial fit is
essential for the outcome of the results. Typically, the polynomial curve
fitting shall be established for a section between the zero moment
points in a buckle. After having established an analytical model for the
buckle apex, the curvature may be established as given in equation 2.

Note that the accuracy of the method depends on the order of the
polynomial expression and the fitting length. Typically, polynomials of
fourth to sixth orders are applied. The polynomial order shall not
reduce below the power of four. Lower orders will produce linear and
constant strain distributions and are not able to model the buckle apex
area sufficiently. The validity of the method is previously described by
Kvarme et al (2012).





Figure 5: Polynomial fit to survey data in a lateral buckle

As an alternative to the polynomial fit model, a filter approach may be
applied directly for curvature data from survey samples. This is
referred to as the curvature filtering procedure. The method is based on
data processing using the Fast Fourier Transform or alternatively the
short time FFT to reduce high frequent noise. In all essence this
produces a low pass filter, where curvature data is transformed into the
frequency domain and undesired frequencies are removed, before
returning the processed data back into the spatial domain. This low pass
filter removes undesired noise according to a predefined criterion,
which was described by Kvarme et al (2012). The filtering process is
illustrated in Figure 6.



Figure 6: Low pass filtering of survey data (curvature data)
Applications and validity of methods

As the polynomial and curvature filter methods have different
approaches for strain estimation, they may be combined for quality
assurance. The polynomial regression and filtering methods does
however have some similarities. For instance, according to Hobbs
(1981), an elastic lateral buckle may be expressed in terms of a
combined cosine and second order polynomial function, i.e.

y =
m
n
4
j-
cos (nx)
cos (nL2)
-
n
2
x
2
2
+
n
2
L
2
8
+ 1[ (3)

where n and m are constant values, (y, x) are pipeline coordinates and L
is the buckle length. In terms of the polynomial fit method, the cosine
term of the Hobbs equation can be expressed through Taylor expansion,
as given in equation (4). The remaining terms of equation (4) may then
simply be added to the expanded cosine series. This yields an extended
Maclaurin series, which is a pure polynomial expression.
(4)

In order to sufficiently represent the cosine function between the points
of zero moment, the polynomial must at least be a fourth order
function. Thus, lower criterion for polynomial order is the power of
four. This representation is illustrated for one half wave length of a
cosine curve in Figure 7.

If the same rationale is followed for evaluating the buckle curvature, eq
(1), it is observed that the second order polynomial term in the Hobbs
function (equation 3) is reduced to cosine function plus a constant. This
function may well be represented by a simple two term Fourier series.
This implies that both the polynomial expression and curvature filter
representation can represent the Hobbs function using the same length
criterion and provide identical apex strain, as illustrated in Figure 7.



Figure 7: Fourth order Maclaurin series compared with a Cosine
function.


= r
L
crit
M =0
M =0
423


As both polynomial functions between order 4-6 and trigonometrically
functions applied in Fourier sums correlate within the given window of
figure 7, the maximum length for the smoothing criteria coincide also
for analyzing strains from survey data. This implies that the same upper
bound length threshold can be applied for both strain estimation
methods to represent the apex curvature of a buckle. This length tends
to be in the order of 50-100 x OD, depending on the magnitude of
strains. I.e. pipelines subjected to plastic deformations will require
shorter threshold lengths than pipelines subjected to elastic
deformation. In other words increased smoothing resolution is required
to capture strain concentrations from plastic deformations in the buckle
apex.

Verification of strain assessment is performed for pipeline data with
survey noise. Typically the survey noise is high frequent compared to
the response of buckles and general pipeline response. Thus the strain
estimations shall be able to predict strain from a survey within a given
filtering range. This range stretches from the domain of the survey
noise to the minimum wave length required to represent the lateral
buckle. This length may vary depending on the pipeline behavior, i.e.
fully elastic or subjected to plastic deformation. A good indication of
successful strain estimation is correlation between strain estimation
using different filter lengths and characteristic lengths for polynomial
regression, i.e. producing converging results. The range of validity for
survey smoothing is illustrated in figure 8, where the strain was
estimated using a large range of evaluation length criteria (threshold
values). If survey noise start to interfere with expected pipeline
response, such as global buckles, it may be difficult to establish a good
strain estimate. In such cases assistance from FE analyses are required.


Figure 8: Sensitivity of applying various thresholding criteria using the
curvature filtering approach (low pass filter)


Assessment criteria
In pipeline design, load utilization is normally given by load based
criteria, Ref DNV-OS-F101. Screening, on the other hand is performed
by displaying strains. A transformation from the bending moment
criterion to a bending strain (or curvature) criterion is therefore needed.
This can be performed by evaluating the bending of a pipe element
subjected to pressure and axial force using an FE approach. In this case
a one-element FE solver with an included code check is developed for
fast assessment of critical bending strains and moments for pressurized
pipelines. By analyzing the intersection point between the moment-
strain curve and critical moment from the code check, a conjugate
strain criterion can be established. This criterion may be applied
directly as an upper strain level for strain distributions from surveys.






Figure 9: Bending-moment/strain configuration.

It is emphasized that when evaluating the bending strain value based on
a given survey, several checks shall be performed e.g. the actual
operating condition at the time of survey as well as any possibilities of
additional bending effects due to third party activity e.g. trawl
interference.


APPLICATION - 36 PIPELINE CASE STUDY

Introduction
The response of a recently installed 36" gas export pipeline is
investigated to illustrate the application of the integrity assessment tool.
The pipeline is routed from an offshore platform to an onshore
receiving terminal. In the following, the hot-section of the pipeline is
investigated. This section has a snaking design to trigger lateral
buckles.

As-laid analysis
After installation, an as-laid survey was conducted in order to ensure
that installation had been performed according to plan. The assessment
of pipeline integrity was performed by first analyzing the on-bottom
integrity in as-laid condition using the integrity assessment screening
tool. Thereafter as-laid analyses for temporary and operation conditions
were performed using FEA to verify lateral buckling behavior. A direct
comparison of in-plane bending strains for as-laid condition using both
directly processed survey data (filtered curvature method) and detailed
FEA is illustrated in figure 10. Both FEA and simplified strain
estimation provide very similar results

Operational analysis and hot-pass assessment
A hot-pass survey is typically performed within a year after startup to
verify that the integrity and response of the pipeline is according to
design and within design criteria stated by DNV-OS-F101. As part of
the final design delivery, the pipeline response was analyzed based on
information obtained from the hot-pass survey. The screening analyses
results were thereafter compared to the detailed FE in-place analyses
for operation having initial configuration from the as-laid survey.











424



Figure 10: In-plane bending strain assessment from as-laid
configuration. A: Lateral pipeline configuration, B: Strain estimation
using filtered curvature method, C: Bending strains from FEA.


The pipeline has a snaking design for the first 15 km, figure 12 A.
Within this hot-zone, eight lateral buckles were planned in order to
relax axial forces within the pipeline system. A 3D plot of one such
buckle is illustrated in Figure 11.




Figure 11. Hot survey configuration SIMVIS representation.

Based on information from both as-laid survey and hot-survey, the
differential configuration is established, i.e.

f(x) = f(x)
hot-pass
- f(x)
as-laid
(5)





where f(x) and x represent the survey data. Based on this representation,
eight lateral buckles had developed, as planned, figure 12 B. By
applying the strain screening methodologies such as curvature filtering
and polynomial smoothing methods, the strain levels in each buckle
was established. This is illustrated in Figure 12 C and D. A typical
threshold length of 100xOD was applied to remove effects from survey
noise. Automatic sensitivity functionalities were then applied to ensure
that strain levels was consistent for variations of threshold lengths.






Figure 12. Strain estimation of lateral buckles. A: Lateral pipeline
configuration, B: difference between Hot-pass relative to as-laid
configuration indicating lateral buckles, C: Strain estimation using the
curvature filter approach, and D: strain distribution at buckle 5.


In addition to analyzing strains using the curvature filtering method,
bending strains spot checks were performed for each buckle using the
polynomial fit method, as seen in Figure 13. Figure 12 C and D, and 13
C and D, both presents bending strains in the buckle apex of KP 8.3.
Both methods consistently indicate strains in the area of 0.13%. It is
furthermore seen that the smoothing criterion (given by the length
between zero bending moments) may be varied for the range between
50xOD to beyond 100xOD without affecting the strain results in the
lateral buckle, Figure 13 D.


A
B
C
A
B
C
D
425



Figure 13. Strain estimation of buckle 5illustrated in figure 12. A:
Lateral configuration of buckle 5, B curvature distribution, C: Strain
distribution between zero bending moments and D: sensitivity curve for
smoothing thresholds (length between zero bending moment at each
side of buckle apex).


It is observed that all strains determined from processing the hot-pass
survey are all in the range of 0.10%-0.15%. This is well below the
critical level of 0.27% established from the code check, figure 9.

Finally, the pipeline integrity was assessed for the operational condition
using FEA (in-place assessment with basis in as-laid configuration).
The pipeline was subjected to thermal and pressure loads, as evaluated
and predicted in details design. The pipeline expanded into lateral
buckles as planned. Bending strains were thereafter extracted from the
FE analyses and compared directly towards the screening results from
the hot-survey. Strain values from both FE analyses and screening are
listed in table 1 and correlate well. This indicates that the simple strain
screening approach using both polynomial regression and curvature
filtering is able to predict pipeline strains at sufficient accuracy
provided that the survey data has sufficient quality. Considering the
efficiency of the methods and simple nature, the approach is applicable
for quick check for general response screening during operation.

Table 1: Screening comparison with FEA



CONCLUSIONS

Structural integrity assessment of pipeline response during operation is
addressed and a software tool for level 1 and level 2 evaluations is
presented. The tool enables the user to evaluate the pipeline response
over time. The toolbox allows processing of free-spans, evaluation of
bottom roughness and assessment of lateral buckles. Both qualitative
assessments as well as quantitative analyses may be performed.
Examples on such analyses are estimation of free-span criticality,
pipeline strains and analytical assessment of feed-in and axial forces.

The software toolkit has a user friendly interface and allows large data
volumes to be assessed effectively without compromising the quality of
the assessment. The presented strain estimation methods produce
results that are in-line with those given by FE analysis results.

The procedures and criteria for strain analysis presented above have
been subjected to extensive analyses and comparison with simple FEA,
e.g. as presented by Kvarme et.al. (2012). Further, we see consistency
between the applied techniques and analytical formulations presented
by Hobbs (1984). This proves the accuracy and potential for the
methods. However, it is important to also emphasize the boundaries
and limitations for the strain estimation methods. Engineers applying
such methods should aware of interfering non-physical effects, which
may be part of the survey data set. This may for instance come from
survey noise, which interferes with expected pipeline response.
Alternatively, if lateral buckles are subjected to significant plastic
deformation, e.g. plastic stains above 0.3-0.4%, more detailed
assessment may be required. This emphasizes the importance of
engineering experience and judgment in the integrity assessment
process.

The examples above are based on response in the horizontal plane. The
methods may also be applied for the bending response in the vertical
plane. Further, the system allows free-spans to be assessed and
compared with design criteria. This allows integrity management
engineers to get an overview of the pipeline condition and its
development during operation.


REFERENCES

Det Norske Veritas, Offshore Standard F101 (2010), Submarine Pipeline
Systems, DNV-OS-F101.
Det Norske Veritas, Recommended Practice F116 (2009), Integrity
Management of Submarine Pipeline Systems, DNV-RP-F116.
Etterdal, G, Nes H, Kvarme S O, Svardal S, (2008) Integrity assessment
of HP/HT infield pipelines Experiences with a new methodology
applied in the Norwegian Sea, Paper No OMAE 2008 57353.
Hobbs, R E, (1984), In-Service Buckling of Heated Pipelines, Journal
of Transportation Engineering, vol. 110, No 2.
Hobbs, R E, (1981), Pipeline Buckling Caused by Axial Loads,
Journal of Constructional Steel Research, v 1, n 2, p 2-10.
Kvarme, S O, Nordsve, A C, Alsos H S, (2012): Advanced Structural
Integrity Assessment of Subsea Pipelines in Operation, Proc Intl
Ocean and Polar Eng Conf, Rhodes, Greece, ISOPE, Vol 2.
www.isope.org .
Nes ,H, Etterdal B, Svardal S (2008), Condition Management of HP/HT
pipelines a new approach, Paper No OMAE 2008 57352.

Curve
number
Design Strains
ANSYS
Strains
Hot Survey

MAX
b=0,27%
FE analysis
Curvature
filter
Strain
Utilization[%]
1 0,1255 0,1265 0.47
2 0,1140 0,1170 0.43
3 0,1355 0,1570 0.58
4 0,1260 0,1399 0.52
5 0,1455 0,1347 0.50
6 0,1220 0,1365 0.51
7 0,1305 0,1281 0.47
8 0,1405 0,1262 0.47
A
B
C
D
426

You might also like