Proceedings of the ASME 2022 14th International Pipeline Conference

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

IPC2022-87320

IMPLEMENTATION OF API 1183 RECOMMENDED PRACTICE FOR RELIABILITY-BASED

ASSESSMENT OF DENTS IN LIQUID PIPELINES

Muntaseer Kainat1, Amandeep Virk1, Nader Yoosef-Ghodsi1, Steven Bott1

1
Enbridge Liquids Pipelines, Edmonton, AB, Canada

ABSTRACT 1183 recommendations have proven to be effective, but

Prior to the publication of API 1183 (Recommended inefficient due to being overly conservative. Efforts have been
Practice for Assessment and Management of Pipeline Dents) in made to validate, and where possible, to calibrate the techniques
2020, there was no industry consensus on one method to evaluate through comparison to experimental results, field findings, and
the Fitness for Purpose for dents to be implemented in integrity historical failures. These efforts have enabled Enbridge to tackle
management programs. Regulations in Canada and the United the over-conservatism of the models for certain combinations
States regarding the repair of dents are primarily based on depth and ranges of operating parameters through novel techniques,
and interaction with stress risers. API 1183 has put forth specific which are described in this paper.
methodologies for screening and detailed assessment of dents Keywords: API 1183, Dent Assessment, Dent Integrity
which consider both strain-based and fatigue-based failure Management, Reliability, Probabilistic Analysis, Dent Strain,
mechanisms. Dent Fatigue
Enbridge Liquid Pipelines had previously presented a
framework to support systemwide dent assessment with an NOMENCLATURE
efficient reliability-based approach. Following the publication of API American Petroleum Institute
API 1183, this approach has been further modified to comply SQuAD Semi-Quantitative Analysis of Dents
with the API recommendations for dent assessment. Both the QuAD Quantitative Analysis of Dents
screening and detailed analyses within this framework account SEP Safe Excavation Pressure
for the properties of the pipe, dent, and interacting features, the FEA Finite Element Analysis
operating condition and history of the line, restraint condition, PRCI Pipeline Research Council International
and associated uncertainties. These analysis techniques combine ILI Inline Inspection
inline inspection results and engineering analysis with their ASME American Society of Mechanical Engineers
uncertainties, providing a means for quantitative assessment of
dents. This paper demonstrates the alignment of Enbridge’s dent 1. INTRODUCTION
management framework with API 1183 recommendations, and The need for enhancements to existing dent management
discusses the modifications made for probabilistic assessment of strategies has been thoroughly identified within the industry,
dents. In the absence of specific guidelines for probabilistic leading to substantial research and development in this area [1,
assessment in API 1183, Enbridge relied on relevant publications 2, 3]. Recently, the American Petroleum Institute (API) has
and industry best practices for considering uncertainties within published a landmark recommended best practice (API RP 1183)
the probabilistic assessment. This framework has been for incorporating available enhancements into industry approved
implemented for systemwide analysis with over 5,000 geometric dent assessment processes [4]. Enbridge LP has leveraged the
anomalies assessed to date. From this implementation research studies in the industry to develop a reliability-based
experience, the challenges with probabilistic analysis and integrity management framework for dents and geometric
potential areas of further improvement have been identified and anomalies1, which includes three main assessment models:
discussed in detail in this paper. In particular, the • The Semi-Quantitative Analysis of Dents (SQuAD)
recommendations in API 1183 regarding dent fatigue model [5, 6, 7], which was developed as a screening
assessment, and the fatigue life reduction factor due to weld assessment tool for strain and S-N based fatigue failure
interaction are observed to be overly conservative. Overall, the modes of dent features.
reliability-based dent management framework following API

1
Enbridge defines geometric anomalies as any deformation reported by
Inline Inspection tools with a maximum depth less than 2% of the nominal
outside diameter of the pipe.

1 © 2019 by ASME
 • The Quantitative Analysis of Dents (QuAD) model [8, 9,
10], which was developed as detailed assessment tool for 3. SQuAD OVERVIEW
strain and S-N based fatigue failure modes of dents using Enbridge started the development of reliability-based
finite element analysis (FEA). screening model for dent assessment in early 2016. The early
• The Safe Excavation Pressure (SEP) model [11, 12], version employed strain limit state which semi-quantitatively
which was developed to determine the safe excavation accounted for coincident feature interaction and qualitatively
pressure of a dent feature using a fracture mechanics considered pressure cycling severity [5, 6]. The model was
based approach within reliability context. further enhanced through the inclusion of fatigue limit state and
adjusted to achieve the best possible alignment with the draft
Development of reliability-based models for analyzing version of API 1183 [7, 14]. Following the publication of API
dents and geometric anomalies was initiated in 2015. The initial 1183 in November 2020, additional modifications were made to
models only incorporated strain limit state. As the PRCI ensure complete alignment with the recommended practice.
Mechanical Damage projects MD-4-2 and MD-4-9 were near Critical components of the current version of SQuAD are listed
completion at that time, Enbridge was able to leverage and below:
incorporate the findings from these projects into the analysis ▪ Dent geometry characterization is performed in
models. Particularly, in 2017, when the fatigue limit state was accordance with Section 6.1 and 6.2 of API 1183.
being developed, the full-scale test results from MD-4-2 and Caliper data (including the presence of pipe cross
fatigue assessment methods put forth in MD-4-9 were heavily section ovalization, if any) is processed using low pass
leveraged [2, 3]. At the same time, the safe excavation pressure filter to reduce concomitant systematic noise, followed
model development was initiated which relied heavily on the by gaussian filter to reduce random noise (technical
findings from MD-5 [13]. Between 2017 and 2020, these models details can be found in [14]).
were tested, validated, and implemented while continuous ▪ Restraint condition is determined using both clock
improvements were performed in parallel. By the end of 2020, position and restraint parameter calculation in
when API 1183 was published, these screening and detailed accordance with Section 6.4.1 and 6.4.2 of API 1183.
assessment models were already closely aligned with the Clock position takes precedence when restraint
recommended practice. Since 2021, Enbridge has completed parameter calculation is invalid due to complex dent
further adjustments to the models and the dent management shape. Un-restrained condition is assumed when clock
framework to ensure complete alignment with API 1183 [4]. position and restraint parameter contradict each other.
Both dents and geometric anomalies are currently being assessed ▪ Coincident features and interacting defects are
using these models within the scope of Enbridge’s dent integrity identified is accordance with Section 6.5 of API 1183.
management framework. This includes the criteria for girth weld, longitudinal
seam weld and spiral seam weld interaction (Section
2. DENT INTEGRITY MANAGEMENT FRAMEWORK 6.5.1), corrosion interaction (Section 6.5.2), and
API 1183 [4] states that a dent integrity management process multiple dent interaction (Section 6.5.6).
should comprise of seven key elements as shown in Figure 1. ▪ Operating condition is characterized in accordance with
The Enbridge dent integrity management framework aligns with Section 6.6 of API 1183. This includes the estimation
the elements by utilizing pre-existing processes for deterministic of maximum pressure at dent location (Section 6.6.1)
analysis and implementing reliability-based assessments for and use of pressure time history (Section 6.6.2.1) to
screening (Element 4) and fitness for service (Element 5). The perform rainflow counting (Section 6.6.3.1). Linear
use of reliability-based assessment facilitates the quantification interpolation is performed between the discharge and
of risk associated with each feature considering the likelihood of suction locations to scale the pressure ranges for the
failure and corresponding consequences. Mitigation efforts can fatigue assessment of each individual dent feature or
thus be prioritized based on risk ranking. These elements of the geometric anomaly. When necessary, location specific
management framework and the alignment between API 1183 pressure spectrum is determined as per Section 8.1.3 of
and Enbridge dent integrity management framework are API 1176 [15] to facilitate more accurate and less
described in more detail throughout this paper. conservative fatigue assessment.
▪ Stress analysis and fatigue assessment in SQuAD are
Remove Non-
Engineering Critical Assessment
Non-Injurious Features Removed from Assessment
performed using FEA models in accordance with
Dent Features (Continue Monitoring)
Section 7.5 of API 1183, where the deformed pipe
Element 1:
In-Line Inspection/In-
Element 2:
Identify Dent
Element 3:
Coincident Feature and
Element 4:
Screening
Element 5:
Fitness for Service
Element 6:
Element 7:
Remediation as
geometry due to the presence of a dent or geometric
Decision
Ditch Observation Features Operations Data Assessment Assessment Required
anomaly is created using the ILI reported three-
Features Requiring Immediate Action (No ECA performed)
dimensional (3D) caliper data and linear elastic
FIGURE 1: ELEMENTS OF DENT INTEGRITY MANAGEMENT property is assigned to the pipe material. These
FRAMEWORK
simplified FEA models do not capture the residual
stresses developed during dent formation and are not
able to mimic the non-linear response of dents subjected

2 © 2019 by ASME
 to pressure load. However, due to the use of linear indenting object and application of pressure. Using
elastic material properties, these models typically these dimensions may lead to incorrect approximation
produce conservative stress results. Additionally, of strain, since the major portion of plastic strain
restrained dents and geometric anomalies behave in an accumulation occurs towards the end of indentation. In
un-restrained manner in this approach, leading to overly SQuAD, the maximum depth and the profile of un-
conservative results. restrained dents and geometric anomalies are scaled to
During the development of SQuAD, several areas of higher depths to better capture the plastic strain
improvement were identified and deemed necessary prior to accumulation. Currently, a constant scaling factor of
implementation of the model as part of the dent integrity 1.3 is used for strain estimation which was based on
management framework. Following are the key components of recommended best practice [17] and determined
SQuAD model that were incorporated in an attempt to enhance through comparison and calibration against FEA
the screening assessment beyond API 1183 recommendation models. For stress analysis using deformed shape
based on Enbridge’s experience: (devoid of residual stresses) and linear elastic material,
▪ In SQuAD, strain is calculated as per API 1183 using a scaling factor of 1.2 was observed to be sufficient
the equations for calculating the combined strain from a similar calibration study for un-restrained dents
presented in ASME B 31.8 non-mandatory appendix R and geometric anomalies.
[16]. Due to the ambiguity in ASME B31.8 regarding ▪ While API 1183 does not recommend any additional
dent length definition, Enbridge has developed an strain assessment criteria in case of corrosion
approach calibrated to finite element analysis results for interaction, SQuAD employs strain capacity derating
defining the length of dents and geometric anomalies in factors when a dent’s most severe point (MSP) resides
the longitudinal direction. In this approach, the length within the extent of a corrosion feature. This was
is measured at a certain depth, which is a function of implemented based on the observations of FEA results.
pipe outside diameter, diameter to wall thickness ratio, This is an artifact of the loading sequence and contact
ILI reported maximum depth, and restraint condition. algorithm used for simulating the indentation load in
For restrained dents and geometric anomalies, this FEA and generally leads to conservative strain
depth varies between 50% and 95% of the maximum estimations. It was also observed from the NDE reports
depth while for un-restrained cases it varies between of features which meet this condition (i.e. the dent MSP
75% and 95% of the maximum depth as shown in located within corrosion), where they may sometimes
Figure 2 below. For shallower depths and lower D/t be categorized as injurious.
ratio pipes, the length is calculated closer to the dent ▪ In the case of weld interaction, two separate fatigue
apex, thus the longitudinal membrane strain is captured assessments are carried out: one for the dent or
in a more conservative way. geometric anomaly using the maximum principal stress
from simplified FEA within the whole dented region,
100% and a second one for interacting weld fatigue using the
Maximum
maximum principal stress within the proximity of the
90% weld.
Depth for Length Calculation

▪ Currently, dents associated with gouges or cracks are

(% Maximum Depth)

80% not assessed and are selected for excavation and repair.
Minimum for Un-restrained Condition
API 1183 presents a general high-level description of the
70% requirements and challenges associated with probabilistic
assessment in Section 8.5 but does not provide any detailed
60%
guidelines for such assessment within its scope. Following are
the key elements of SQuAD pertaining to reliability-based
Minimum for Restrained Condition assessment of dents and geometric anomalies:
50%
D/t=106.72 ▪ Measurement uncertainties associated with caliper ILI
D/t=65.55
D/t=49.16 data is considered by generating random and plausible
40%
0.0% 1.0% 2.0% 3.0% 4.0% 5.0%
dent profiles as a function of the vendor specified tool
Maximum Depth (% OD) accuracy [7, 14].
FIGURE 2: CALIBRATED APPROACH FOR DENT LENGTH ▪ The uncertainty associated with the strain capacity of
CALCULATION AT CERTAIN DEPTHS FOR ASME STRAIN pipe material is considered by use of a general strain
EVALUATION capacity distribution based on available tensile test data
[18], maximum elongation limit from API 5L [19], and
▪ The shapes and dimensions of un-restrained dents and industry recommendations for uncertainty [20].
geometric anomalies reported by caliper inline Probabilistic derating factors are used on the capacity
inspection tools correspond to those after rebounding distribution in the cases of corrosion interaction and/or
and re-rounding of the pipe, following removal of the weld interaction.

3 © 2019 by ASME
 ▪ Potential uncertainties associated with the stresses element analysis is performed within QuAD, which accounts for
obtained from simplified linear elastic FEA models are geometric and material non-linearity. The FEA modeling
accounted for by defining the stress demand as a normal approach has been validated against full-scale indentation test
distribution with a coefficient of variation (CoV) of available in literature [28] and has been published previously
10%. [29]. The general modeling approach and consideration of
▪ Uncertainties associated with S-N parameter is adopted different sequences of indentation and pressure loads have been
from BS 7608 [21]. API 1183 recommends the use of presented and studied in [29]. The FEA approach in QuAD
BS 7608 Class D mean – 1 Standard Deviation curve aligns with the recommendations in API 1183 Section 8.3.5, as
for deterministic fatigue assessment of plain dents. For well as API 579-1 [30].
probabilistic analysis, the Class D mean S-N curve and Reliability based analysis of strain limit state is performed
its corresponding standard deviation (as specified in BS within QuAD by use of FEA models and response surface
7608) is used within SQuAD. method. In this approach, the input parameters to the FEA
▪ Recommendations in API 1183 regarding fatigue models (such as stress-strain curve, pipe wall thickness, dent
damage capacity is consistent with the Palmgren-Miner depth) are varied within their specified statistical distributions
rule [22], which states that fatigue failure will occur and the outputs are recorded from multiple FEA runs. A response
when the cumulative fatigue damage approaches or surface is then developed by performing regression analysis on
exceeds a value of 1.0. Due to the random nature of the input and FEA output values. The response surface acts as a
fatigue damage, this value may deviate systematically closed form solution to the specific problem and becomes a
and randomly from unity, and can be treated as a “surrogate model” which is used for reliability analysis. The
random variable that accounts for the shortcomings of methodology has been presented in previous publications [8, 9,
Miner’s rule [23]. Wirsching (1984) suggested that 10], and is not described in detail within the scope of this paper.
instead of unity, the capacity can be modelled as a Reliability analysis of fatigue limit state follows the same
lognormal variable with a mean of unity and a approach as in SQuAD [7], with the following differences:
coefficient of variation of 0.3 [24]. In SQuAD, the ▪ The principal stress range due to the pressure cycling
fatigue damage capacity is modeled after [24]. obtained from detailed FEA model is used
However, the CoV is defined as a function of the age of deterministically for fatigue reliability analysis.
a line and the duration of available pressure data such ▪ In case of girth weld interaction, the axial component of
that the CoV scales down when the available pressure the stress range is used instead of the principal
data covers a greater portion of the line’s age. component.
▪ API 1183 recommends a 10 times reduction factor on The general flow of reliability-based fatigue analysis within
the deterministic remaining life of dents interacting QuAD is shown schematically in Figure 3, and typical
with welds. For probabilistic assessment within probability of failure (PoF) output from QuAD is shown in
SQuAD, a reduction factor distribution has been Figure 4.
developed using full scale test results [25, 26], industry
recommendations [27], and subject matter expert’s Pressure Spectrum Rainflow Counting Numerical Modelling
review. A statistical analysis carried out on 32 full-scale
fatigue test results [25] showed that the reduction factor
distribution has a mean value of 1.97, a standard
deviation of 1.2739, and can be best fitted by an
exponential distribution. Due to the small data set used
in this analysis (32 data points cannot be considered
statistically significant), conservative assumptions were S-N Curve Linear Relation
incorporated in defining a probabilistic reduction
factor. The reduction factor is defined using a normal Output

distribution (based on subject matter expertise) with a = f( , , N, , t)

mean of 5.5 (consistent with [27]) and a standard
deviation of 1.2739 (consistent with experimentally
observed variation) and is truncated between 1 and 10. FIGURE 3: DIFFERENT STAGES OF FATIGUE ASSESSMENT
A complete description of the SQuAD model is not possible
within the scope of this paper but is available in previous
publications [5, 6, 7].

4. QuAD OVERVIEW
Dents and geometric anomalies that are assessed and
identified to be injurious using SQuAD are selected for detailed
fitness for service assessment using QuAD. A detailed finite

4 © 2019 by ASME
 1.E+00 is determined using strain limit damage approach. The worst case
of these three scenarios is then selected to determine SEP.
1.E-01

1.E-02
6. IMPLEMENTATION EXPERIENCE
Fatigue PoF (Dent)

1.E-03
The dent integrity management framework has been tested
on 79 previously excavated features. This includes quantitative
1.E-04 assessment by subject matter experts, screening assessment
using SQuAD, and fitness for service assessment using QuAD.
Current Age
1.E-05
Reinspection + 1 Year Overall, the program demonstrated positive performance 85% of
Fatigue PoF Vs Age the time, and was conservative 15% of the time. No false
1.E-06
0 20 40 60 80 100 120 140 160 negatives (i.e. no non-conservative cases) were identified from
Age (Year)
this exercise (Figure 5).
FIGURE 4: TYPICAL OUTPUT FROM RELIABILITY-BASED
FATIGUE ANALYSIS IN QuAD True Positive
16 Features,
20%
5. SEP OVERVIEW
In rare cases where deterministic SEP has significant effect
True Negative
on throughput, or a line becomes inoperable due to SEP being 51 Features,
lower than operability limits, a safe excavation pressure study is 65%
performed using a reliability-based SEP model [11, 12]. A
feature-specific and fracture mechanics based assessment
approach was developed which incorporated both ILI data and
finite element analysis (FEA) to determine the SEP. This
assessment accounts for the uncertainties associated with False Positive
12 Features,
material properties and ILI tool measurement. Since crack ILI 15%
tools may be unable to detect and/or size cracks within dents, this
approach also considers the probability of an un-identified crack
within a dent or geometric anomaly. The probability of an un-
identified crack in dent or geometric anomaly can be determined
through statistical analysis of previous NDE findings. FIGURE 5: PERFORMANCE OF DENT INTEGRITY
MANAGEMENT FRAMEWORK AS COMPARED AGAINST NDE
The assessment method incorporates the API 579-1 [30]
FINDINGS
Failure Assessment Diagram (FAD) method on an uncracked
FEA model as opposed to explicitly incorporating the crack Prior to implementation, SQuAD was validated by
geometry in the FEA model. A validation study was performed
comparing against QuAD using a set of more than 100 unique
against ten full scale burst tests available in literature [31]. The
dent features within Enbridge system as was previously
study showed good agreement of the burst pressure of dent-crack published in [7]. This validation study demonstrated that SQuAD
defects predicted by FEA models with those observed in the full- performed conservatively 55% of the time in terms of strain, and
scale tests. This comparison of the results of FEA in conjunction 46% of the time in terms of fatigue. Following implementation,
with FAD with the full-scale tests ensured the accuracy and over 5000 dents and geometric anomalies in Enbridge system has
conservatism of burst pressure prediction. A reliability-based
been assessed with SQuAD to date. Of these, approximately 200
approach was then designed which accounts for the uncertainties
features have been further assessed for fatigue and 100 features
associated with the analysis. have been assessed for strain using QuAD. Figure 6 shows the
Three different scenarios are investigated within each comparison of SQuAD and QuAD PoF corresponding to strain
feature specific assessment. First scenario considers that a crack and fatigue limit states. For ease of visualization, the minimum
is present at the location of the maximum principal stress when PoF value is truncated at 1.0E-06 in the plots.
the pipe is being excavated at pressure. The probability of an un-
identified crack in dent and the crack size is determined based on
previous NDE findings. In the second scenario, a 10% of wall
thickness deep crack with a length ten times its depth is assumed
to coincide with the location of maximum stress range due to
pressure cycling. The crack is assumed to be present since
construction and is grown using Paris Law crack growth
equation to obtain the final size at the time of excavation. In the
third scenario, the size and location of a crack due to indentation

5 © 2019 by ASME
 1.0E+00 was achieved by performing additional problem specific analysis
and modifications which is described later in Section 8. No non-
conservative results have been observed to date.
1.0E-01
7. OBSERVATIONS ON MODEL PERFORMANCE
Conservative True Positive
1.0E-02 62.4% 1.0% With the alignment with API 1183, model performance
SQuAD Strain PoF

remains generally acceptable for most use cases. There are some
challenges with implementation that have been identified as
1.0E-03
follows:
▪ The limitation regarding dent length definition has been
minimized by a calibrated method as described in
Section 3. A holistic approach was taken for this
1.0E-04
purpose, where the calibration was performed by
True Negative Non-Conservative comparing SQuAD and QuAD strain PoF outputs
36.6% 0.0%
instead of deterministic strain output. This was
1.0E-05
necessary since the uncertainty corresponding to caliper
measurement accuracy is handled very differently in the
two models. The effectiveness of the calibration is
1.0E-06
1.0E-06 1.0E-05 1.0E-04 1.0E-03 1.0E-02 1.0E-01 1.0E+00
evident from Figure 6(a) where SQuAD strain PoFs are
(a) observed to be consistently conservative, and false
QuAD Strain PoF
negatives are non-existent (compared to QuAD). When
1.0E+00 deterministic strains from SQuAD and QuAD are
Conservative compared, a reasonably good agreement is still
28.1% observed (Figure 7). However, further calibration
1.0E-01 and/or modification may be necessary for solely
deterministic assessment. This specific issue is
currently being investigated within the scope of PRCI
1.0E-02 project MD-5-2 “Improve Dent/Cracking Assessment
SQuAD Fatigue PoF

Methods” (PR-214-203806-R01) [32]. The preliminary

True Positive
7.8% results in the draft report appear to be comparable to the
1.0E-03
trend shown in Figure 7, considering the fact that the
strains shown in Figure 7 were calculated using ILI
reported field dent shapes subject to measurement
1.0E-04
inaccuracy, while the PRCI approach primarily
employs idealized dents (shapes obtained from FEA)
Non-Conservative
and full-scale test data.
0.0%
▪ Un-restrained dents and geometric anomalies, as
1.0E-05
mentioned previously, are scaled by a factor of 1.3 in
True Negative
64.1% SQuAD strain assessment, and by a factor of 1.2 for
SQuAD stress assessment. From implementation
1.0E-06
1.0E-06 1.0E-05 1.0E-04 1.0E-03 1.0E-02 1.0E-01 1.0E+00
experience, it has been observed that a fixed scaling
(b) factor of 1.3 sometimes leads to inconsistent
QuAD Fatigue PoF
conservatism when deterministic strain outputs from
SQuAD are compared with QuAD outputs (Figure 7).
FIGURE 6: SQuAD VS QuAD POF CORRESPONDING TO (a) Although this limitation does not affect the
STRAIN LIMIT STATE AND (b) FATIGUE LIMIT STATE effectiveness of the probabilistic assessment, (due to the
nature of calibration, and as shown in Figure 6(a)), this
For this comparison and for the purpose of demonstration, is still a shortcoming of the model and is partially
an arbitrary permissible PoF of 1.0E-03 has been selected. For responsible for the over-conservatism observed in strain
an individual feature, if both SQuAD and QuAD PoFs are greater results. The stresses using a fixed scaling of 1.2 have
than or equal to 1.0E-03, that feature is considered a “true been observed to compare well with QuAD stresses
positive” and so on. The comparison shows that SQuAD has (Figure 8). Any non-conservatism observed in the
performed conservatively 62.4% of the time and 28.1% of the deterministic results do not have any adverse effects on
time in terms of strain and fatigue, respectively, when compared the final PoF due to the fact that in SQuAD fatigue
to QuAD. The reduced conservatism in the fatigue assessment analysis, the stress is considered as a random variable

6 © 2019 by ASME
 with a CoV of 10% which follows a normal distribution. ▪ Restraint parameter calculation has proven to be
Similar to the strain issue, this limitation does not affect challenging using ILI reported data. Of the 5423 field
the effectiveness of the probabilistic fatigue assessment dents and geometric anomalies analyzed with SQuAD
(as shown in Figure 6(b)), but this is still a shortcoming to date, valid restraint parameter could not be calculated
of the model and is partially responsible for the over- for 3437 dent features (63% of total population). These
conservatism observed in fatigue results. Regardless, 3437 features primarily consist of multiple peak dents,
this is considered a minor limitation of the model which and a minor portion consists of dents with insufficient
will need to be addressed. data provided by ILI vendor. Restraint parameter
calculated following API 1183 equations have shown
40% some inconsistencies in predicting dent restraint
condition correctly. Out of 730 top-side dents (above
35%
the 4 o’clock and 8 o’clock positions), 122 dents had a
calculated restraint parameter value greater than 20,
(Combined Strain using ASME B31.8 Appendix R)

indicating restrained condition, which potentially

30%
would have led to non-conservative results. In a handful
of cases, where the calculated restraint parameter
25% indicated un-restrained condition for bottom-side dents,
the ILI reported profile could not be matched through
SQuAD Strain

20% many trials in FEA until the load sequence was changed
to simulate restrained condition. On the other hand, in
some cases, bottom-side dents with restraint parameter
15%
value less than 20 (indicating un-restrained condition)
could only be matched in FEA when simulated as un-
10% restrained. Overall, the restraint parameter did not
prove to be as useful as was previously expected. This
5% can be attributed to the fact that the restraint parameter
Un-Restrained approach was developed based on FEA results of
Restrained idealized dents [25], therefore has limitations when
0%
0% 5% 10% 15% 20% 25% 30% 35% 40% applied to ILI reported dent shapes which are
QuAD Strain significantly more random in nature. Additionally,
(Equivalent Plastic Strain form Detailed FEA) majority of this inconsistency has been observed for
FIGURE 7: SQuAD VS QuAD STRAIN geometric anomalies with depths less than 2% of OD,
where the threshold value of 20 may be inappropriate
▪ As previously mentioned, the stress analysis within
1400 SQuAD follows the recommendations in section 7.5 of
API 1183. Restrained dents and geometric anomalies
Maximum Principal Stress Range from SQuAD

1200 are inherently modeled as un-restrained in this

approach, leading to highly conservative stress results
1000
in the majority of cases. A comparison of the maximum
principal stress ranges of restrained dents from SQuAD
– simplified FEA models and QuAD – detailed FEA
(MPa)

800
models is shown in Figure 9.
▪ In general, fatigue assessment has proven to be overly
600
conservative for both SQuAD and QuAD. QuAD FEA
approach is able to reduce the conservatism associated
400 with stress by considering geometric and material non-
linearity, but the overall conservatism is driven by the
200 S-N parameters and fatigue life reduction factor for
Un-Restrained weld interaction. Additionally, for reliability-based
0 fatigue assessment, the uncertainties associated with S-
0 200 400 600 800 1000 1200 1400 N parameter is considered, which significantly drives
Maximum Principal Stress Range from QuAD
(MPa) the estimated fatigue PoFs. The S-N curve
FIGURE 8: SQuAD VS QuAD MAXIMUM PRINCIPAL recommended in API 1183 is based on full scale dent
STRESS RANGE FOR UN-RESTRAINED DENTS fatigue tests and corresponding FE analyses, where the
Class D mean – 1 standard deviation curve was
observed to fit the fatigue test data conservatively. This

7 © 2019 by ASME
 S-N curve was not developed for pipeline dents, and potential for losing the deformation shape while
therefore the uncertainties corresponding to this curve reducing noise.
specified in BS 7608 does not necessarily apply to dent
fatigue behaviour. The mean value of the S-N intercept 8. CONTINUOUS IMPROVEMENTS
(Log(C)) is unknown for dent fatigue application, and In the previous sections, the challenges with implementing
the specified standard deviation does not represent the API 1183 for reliability-based dent assessment and resulting
actual scatter that would be observed from dent fatigue conservatisms have been discussed. As dents and geometric
tests, if a sufficiently large number of such test results anomalies are primarily defined by their shapes and restraint
were available. This is believed to be one of the reasons condition, the variation in their characteristics and behaviour is
behind the overly conservative nature of reliability- extremely wide. A pipeline system as large and diverse as the
based dent fatigue assessment. The reduction factor for Enbridge system only adds more complexity to this problem.
weld interaction has been incorporated probabilistically Therefore, one general model or approach for screening
within fatigue assessment in an attempt to reduce assessment becomes infeasible, and line or operating condition
conservatism but has not shown any significant specific modifications are required for effective and efficient
improvement. These limitations lead to unrealistically dent integrity management. Enbridge has attempted to perform
high number of dents requiring detailed analysis line specific calibration of SQuAD using detailed FEA results
(QuAD) or mitigation. Excavation results indicate that from QuAD in cases where the number of features requiring
the method is overly conservative, especially for dents detailed assessment is too large and infeasible.
interacting with welds. However, the reduction factor For shallow dents on small diameter and low D/t ratio pipes,
distribution does not account for the type of weld or the strain PoFs from SQuAD are observed to be unrealistic. In
weld quality, and it is possible that the reduction factor these cases, when profiles are randomly generated to consider
distribution is not conservative for certain types and tool measurement uncertainly, the strains calculated for
vintages of welds. shallower depths become artificially high. For these cases, a
trending analysis is performed. Select features are first analyzed
1400 using QuAD, and the SQuAD strain uncertainties are updated to
align better with QuAD results. The updated approach is then
Maximum Principal Stress Range from SQuAD

1200 implemented on all features previously flagged by SQuAD to

obtain more representative results.
1000
The screening FEA model in SQuAD is extremely
conservative for restrained dents, as described previously. For
these cases, select features flagged for detailed assessment are
(MPa)

800
assessed using QuAD. These features are then re-assessed using
SQuAD methodology with non-linear material properties (as
600 opposed to linear elastic material). The SQuAD and QuAD
stresses are compared to ensure no non-conservative results
400 exist. If the comparison confirms that non-linear material can be
used within the screening FEA model for a particular segment of
200 a line, then all restrained dents are re-assessed using non-linear
Restrained material properties. This has proven to be a very effective
0 approach for many old lines.
0 200 400 600 800 1000 1200 1400 In addition to the aforementioned methods, previous NDE
Maximum Principal Stress Range from QuAD
(MPa)
results are also leveraged in some cases, when useful data is
FIGURE 9: SQuAD VS QuAD MAXIMUM PRINCIPAL available. For example, the reduction factor distribution for weld
STRESS RANGE FOR RESTRAINED DENTS interaction has been modified to reduce conservatism for certain
lines by calibrating against NDE results. In these cases,
▪ Enbridge has opted for reporting dents down to the ILI previously excavated dents with any potential indication of
detection threshold. Consequently, a very large fatigue damage due to weld interaction are first assessed using
population of geometric anomalies with less than 1% of data from the ILI inspection that triggered excavation. The
OD depth are being reported by more recent tool runs. reduction factor is then gradually modified while ensuring true
These geometric anomalies pose challenges with positives do not become false negatives.
assessing them as the signal noise in many cases
approach the detection threshold value, and the 9. POTENTIAL LONG-TERM SOLUTIONS
deformation shape signature cannot be isolated Based on Enbridge’s experience, a few areas of
properly. A tentative solution to this problem is improvement that can be investigated to advance the efficiency
increasing the filter strength, however, there is a and accuracy of the dent management framework are discussed
below.

8 © 2019 by ASME
 ▪ For strain estimation of un-restrained dents, the When API 1183 was published, the screening and detailed
importance of proper scaling has been discussed assessment models were already closely aligned with the
previously. It has been observed from FE analysis that recommended practice. Since 2021, Enbridge has completed
both depth and sharpness can change significantly when further adjustments to the models and the dent management
an un-restrained dent rebounds due to removal of framework to ensure complete alignment with API 1183.
indenter and application of pressure. Therefore, the The dent integrity management framework has been tested
scaling must be applied in all three directions (radial, on 79 previously excavated features, demonstrating positive
longitudinal, and circumferential). These factors can performance 85% of the time, conservative performance 15% of
potentially be functions of the D/t ratio of the pipe, the time, and producing no false negatives (i.e. no non-
maximum pressure experienced by the dent, pressure at conservative cases). The screening versus detailed model
the time of inspection, and reported maximum depth. comparison also did not elicit any non-conservative screening
Carefully designed and elaborate parametric study can model performance.
be carried out to establish these relationships. For Some of the challenges with implementation of the dent
screening level stress assessment of un-restrained dents, assessment models include: lack of clarity in the dent length
a similar study can be carried out. definition in ASME B31.8 strain formulas, lack of guidance on
▪ The conservatism within screening level stress the dent scaling required for strain or fatigue analysis of un-
assessment of restrained dents can be reduced by restrained dents, inadequate performance of the restraint
carefully defining restrained conditions in the parameter, highly conservative stress and fatigue analysis when
simplified FEA. The extent of rebounding can be using the screening tool’s simplified FEA for restrained dents,
limited by use of boundary conditions at the dent MSP, highly conservative S-N curve and the reduction factor for weld
although this may lead to some stress concentrations. interaction, and probabilistic assessment of very shallow
An alternative approach is to define partial restraint geometric anomalies. Some solutions have been offered for
using spring elements. These elements can be used to these challenges in this paper, but more studies are required to
connect the dent MSP and a reference point. A provide more effective scaling factors for strain or fatigue
parametric study will be required to determine proper analysis of un-restrained dents, to reduce the conservatism
stiffness values and/or other spring properties such that within screening level stress evaluation of restrained dents by
the stress outputs from simplified FEA agree with those carefully defining restrained conditions in the simplified FEA, to
from detailed FEA. reduce the conservatism of the S-N curve and the reduction
▪ The overly conservative nature of the fatigue factor for weld interaction, and to provide further guidance for
assessment tools is due mainly to the use of very probabilistic analysis of dents.
conservative S-N curve (namely, Log(C)) and the
reduction factor for weld interaction. More fatigue test ACKNOWLEDGEMENTS
data can result in modifying the above parameters to The authors would like to acknowledge Enbridge for
reduce the conservatism for both deterministic and supporting publication of the work, and Stantec for logistic
probabilistic analyses as well as better definition of supports. The authors would also like to thank Mr. Doug Langer
these parameters’ distributions for probabilistic and Mr. Neil Hodson for their contributions and invaluable input.
analysis. More fatigue test data can also result in tighter
distributions for the parameters, which would make the DISCLAIMER
probabilistic analysis more accurate and further reduce Any information or data pertaining to Enbridge Liquids
the over-conservatism. Pipelines or its affiliates, contained in this paper was provided to
▪ Further guidance for probabilistic analysis of dents the authors with the express permission of Enbridge Liquids
(beyond section 8.5 API 1183) will provide a standard Pipelines, or its affiliates. However, this paper is the work and
approach for the industry in utilizing a reliability-based opinion of the authors and is not to be interpreted as Enbridge
methodology, which can facilitate risk evaluations and Liquids Pipelines, or its affiliates’, position or procedure
mitigations. regarding matters referred to in this paper. Enbridge Liquids
Pipelines and its affiliates and their respective employees,
10. CONCLUSION officers, director and agents shall not be liable for any claims for
Enbridge LP has leveraged the research studies in the loss, damage or costs, of any kind whatsoever, arising from the
industry to develop a reliability-based dent integrity errors, inaccuracies or incompleteness of the information and
management framework, which includes screening and detailed data contained in this paper or for any loss, damage or costs that
assessment tools for strain and fatigue failure modes of dents and may arise from the use or interpretation of this paper.
geometric anomalies as well as a tool to determine the safe
excavation pressure of dents. The use of reliability-based REFERENCES
assessment facilitates the quantification of risk, and thus, a risk- [1] Gao, M. and Krishnamurthy, R. “Mechanical Damage in
based prioritization of mitigation efforts. Pipelines: A Review of the Methods and Improvements in
Characterization, Evaluation, and Mitigation.” Oil and Gas

9 © 2019 by ASME
Pipelines: Integrity and Safety Handbook. John Wiley & Sons, [15] American Petroleum Institute, API RP 1176,
Inc., Hoboken (2015), pp.289-325. Recommended Practice for Assessment and Management of
[2] Pipeline Research Council International. “Fatigue Life Cracking in Pipelines, API, Errata 1, February 2021.
Assessment of Dents with and without Interacting Features, [16] American Society of Mechanical Engineers, Gas
Final Report.” Prepared by BMT, Catalogue Number PR-214- Transmission and Distribution Piping System, ASME B31.8,
114500-R01. 2018. 2020.
[3] Pipeline Research Council International. “Full-Scale [17] American Petroleum Institute, API 579-2/ASME FFS-
Demonstration of the Interaction of Dents with Welds and 2, Fitness-For-Service Example Problem Manual, 2009.
Localized Corrosion Defects.” PRCI Project MD-4-2, Catalogue [18] Lin, M., 2015. Characterization of Tensile and Fracture
Number PR-214-073510. 2018. Properties of X52 Steel Pipes and Their Girth Welds., PhD
[4] American Petroleum Institute, API RP 1183, Thesis, University of Alberta. Edmonton.
Assessment and Management of Dents in Pipelines, API, 2020. [19] American Petroleum Institute, Specifications for Line
[5] Hassanien, S. and Langer, D., 2018, April. A Semi- Pipe, API 5L, 2015.
Quantitative Analysis of Dents Associated with Corrosion: A [20] Canadian Standards Association, Standard Z662: Oil
Reliability-Based Approach. In CORROSION 2018. OnePetro. and Gas Pipeline Systems, 2019.
[6] Langer, D., Hassanien, S. and Woo, J., 2018, [21] BS 7608:2014+A1 - Guide to fatigue design and
September. Semi-Quantitative Reliability-Based Ranking assessment of steel products, 2015.
Method for Assessment of Pipeline Dents with Stress Risers. [22] Hosford, W.F., 2010. Mechanical behavior of
In International Pipeline Conference (Vol. 51876, p. materials. Cambridge university press.
V002T07A003). American Society of Mechanical Engineers. [23] JCSS Probabilistic Model Code., Part 3: Resistance
[7] Virk, A.S., Langer, D., Woo, J., Yoosef-Ghodsi, N. and Models., Joint Committee of Structural Safety (2001).
Kainat, M., 2020, September. Improved Semi-Quantitative [24] Wirsching, P.H., 1984. Fatigue reliability for offshore
Reliability-Based Method for Assessment of Pipeline Dents with structures. Journal of Structural Engineering, 110(10), pp.2340-
Stress Risers. In International Pipeline Conference (Vol. 84447, 2356.
p. V001T03A017). American Society of Mechanical Engineers. [25] Pipeline Research Council International, “Fatigue Life
[8] Hassanien, S., Kainat, M., Adeeb, S. and Langer, D., Assessment of Dents with and without Interacting Features”,
2016, September. On the use of surrogate models in reliability- MD-4-9 PRCI Final Report prepared by BMT, Catalog No. PR-
based analysis of dented pipes. In International Pipeline 214-114500-R01, November 2018.
Conference (Vol. 50266, p. V002T07A018). American Society [26] Full-Scale Demonstration of the Interaction of Dents
of Mechanical Engineers. with Welds and Localized Corrosion Defects, PRCI Project MD-
[9] Langer, D., Kainat, M. and Woo, J., 2017, March. 4-2 (PR-214-073510).
Reliability Assessment of Corrosion Features Interacting with [27] DNVGL-RP-C203, Fatigue design of offshore steel
Pipeline Dents. In CORROSION 2017. OnePetro. structures, April 2016.
[10] Abdelmoety, A.K., Kainat, M., Yoosef-Ghodsi, N., Li, [28] Ghaednia, H., Gerard, K., Bhattacharjee, S. and Das, S.,
Y. and Adeeb, S., 2021. Strain-Based Reliability Analysis of 2014, November. Behavior of NPS30 pipe subject to denting
Dented Pipelines Using a Response Surface Method. Journal of load. In ASME International Mechanical Engineering Congress
Pipeline Science and Engineering. and Exposition (Vol. 46583, p. V009T12A026). American
[11] Kainat, M., Langer D., and Hassanien, S., "Do We Need Society of Mechanical Engineers.
a Safe Excavation Pressure for Dented Pipelines: How Should it [29] Kainat, M., Woo, J., Langer, D., Krausert, T., Cheng,
Be Defined?." 2018 12th International Pipeline Conference. J.R., Hassanien, S. and Adeeb, S., 2019. Effects of loading
American Society of Mechanical Engineers Digital Collection, sequences on remaining life of plain dents in buried liquid
2018. pipelines. Journal of Pipeline Systems Engineering and
[12] Okoloekwe, C., Fowler, M., Virk, A., Yoosef-Ghodsi, Practice, 10(2), p.04019001.
N. and Kainat, M., "Reliability-Based Assessment of Safe [30] American Petroleum Institute, API 579-1/ASME FFS-
Excavation Pressure for Dented Pipelines." 2020 13th 1, Fitness-for-Service, 2016.
International Pipeline Conference. American Society of [31] Ghaednia, H., 2015. Burst Strength of NPS30 Steel
Mechanical Engineers Digital Collection, 2020. Pipes with Dent-Crack Defect., PhD Thesis, University of
[13] Pipeline Research Council International. “Safe Windsor. Windsor.
Inspection Procedures for Dent and Gouge Damage.” PRCI [32] Pipeline Research Council International, "Improve
Project MD-5, Catalogue Number PR-218-063505. 2018. Dent/Cracking Assessment Methods, PRCI Project MD-5-2,
[14] Ergezinger, N., Virk, A.S., Woo, J., Kainat, M. and Catalogue Number PR-214-203806-R01."
Adeeb, S., 2020, September. Application of Noise Filtering
Techniques for the Quantification of Uncertainty in Dent Strain
Calculations. In International Pipeline Conference (Vol. 84447,
p. V001T03A026). American Society of Mechanical Engineers.

10 © 2019 by ASME