You are on page 1of 19

OTC-29528-MS

Correlation of Finite Element Analysis FEA Predicted Residual Strength of


Degraded Offshore Mooring Chains with Test Data

Gary H. Farrow, Andrew E. Potts, and Simon Dimopoulos, AMOG Consulting; Andrew A. Kilner, AMOG Inc.

Copyright 2019, Offshore Technology Conference

This paper was prepared for presentation at the Offshore Technology Conference held in Houston, Texas, USA, 6 – 9 May 2019.

This paper was selected for presentation by an OTC 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 Offshore Technology Conference and are subject to correction by the author(s). The material does not necessarily reflect any
position of the Offshore Technology Conference, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Offshore Technology Conference 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 OTC copyright.

Abstract
The first phase of the Chain FEARS (Finite Element Analysis of Residual Strength) Joint Industry Project
(JIP) aimed to develop guidance for the determination of a rational discard criteria for mooring chains
subject to severe pitting corrosion which, based on current code requirements, would otherwise require
immediate removal and replacement.
Critical to the ability to establish rational discard criteria, is the ability to accurately predict the residual
strength of degraded chain, and to have as a benchmark for loss in strength, an accurate estimate of the chain
in its as-new condition. With a correlated FEA method for residual strength prediction and a benchmark for
as-new condition capacity, it would then be possible to establish a theoretical relationship between different
types of degradation and mooring chain capacity loss, from which rational discard criteria would be derived.
To this end the Chain FEARS JIP first developed a Finite Element Analyses (FEA) residual capacity
assessment method to accurately predict the residual strength of degraded chains. A number of assessments
were carried out to establish the sensitivity of the Predicted Break Load (PBL) to both engineering
parameters such as friction coefficient, and numerical modelling techniques. The developed method was
validated by the correlation of the PBL against a number of physical break tests.
This paper presents a review of the break strength test data of pitting corrosion degraded chain links. The
FEA modelling methodology based Predicted Break Load (PBL) are compared with the test data Actual
Break Load (ABL) along with the sensitivity of engineering parameters and numerical model modelling
techniques on predictions. The developed FEA method accurately predicts the location of the ‘failure’ within
the chain string and the ductile necking failure mode, determined to be the prevalent mode of failure for
the chain links samples considered in this study. The degree of correlation between PBL and ABL confirms
that accurate prediction of the effects of corrosion degradation consequent on uniform and large pitting
corrosion can be accurately predicted by use of the Finite Element Method.
The developed FEA method was also employed to establish a benchmark for the strength capacity
of as-new condition links as presented in [1], the basis for assessing the relationship between corrosion
degradation and residual chain link capacity [2] and a basis for a multi-axial fatigue assessment method to
establish the fatigue capacity of as-new and degraded chain links [3,4,5].
2 OTC-29528-MS

Introduction
Deterioration of chain sections of mooring systems can be more rapid than expected, and has in the
past caused serious incidents, including loss of station. In the majority of systems, the actual life of the
components is less than that of the desired life of the installation. This requires a policy of life prediction,
inspection and replacement. To address this issue the Chain FEARS JIP sought to develop guidance for
the determination of a rational discard criteria for both strength and fatigue of mooring chains subject to
severe degradation, such as deep pitting associated with Microbial Induced Corrosion (MIC), extensive
asymmetric wear, and/or advanced loss of metallic section over the entire chain link due to general corrosion
which would otherwise require immediate removal and replacement.
Critical to the ability to establish rational discard criteria, is the ability to accurately predict the residual
strength of degraded chain, and to have as a benchmark for loss in strength, an accurate estimate of the
chain in its as-new condition. With a correlated Finite Element Analysis (FEA) method for residual strength
prediction and a benchmark for as-new condition capacity, it would then be possible to establish a theoretical
relationship between different types of degradation and mooring chain capacity loss, from which rational
discard criteria would be derived.
FEA residual capacity assessment method was developed and correlated against available break strength
test data of corrosion degraded links as part of the JIP. This investigation also required that a number of
assessments be carried out to establish the sensitivity of the Predicted Break Load (PBL) to both engineering
parameters such as friction coefficient, and numerical modelling techniques including element type and
size. The developed method was validated by the correlation of the PBL against a number of Actual Break
Load (ABL) as determined from physical break tests.
FEA were also conducted of as-new chain links in accordance with this validated method to establish the
PBL of as-new links to provide a benchmark for loss of strength associated with degradation. The findings of
this benchmarking work is presented in [1]. Further investigations were conducted to derive the theoretical
relationship between uniform corrosion and capacity loss, and location and extent of mega pit corrosion and
capacity loss, the findings of which are presented in [2].

Background
Ensuring mooring integrity is a challenge for ageing offshore structures and this will become even more
evident as time passes and mooring chains are exposed for longer periods. Understanding better the effective
strength and remaining life of damaged chain links and its possible correlation between local defects
(corrosion pitting, local wear) is an identified need from the mooring industry.
Analyses of recovered chain links show various types of corrosion including:

• severe pitting corrosion;

• severe uniform corrosion;

• severe wear in the inter-link area;

• stress corrosion cracking;

• or nearly no corrosion, i.e. in the as new configuration;

which most of the time can be detected / measured underwater.


OTC-29528-MS 3

Figure 1—Examples of corroded chain links exhibiting (a) pitting


corrosion, (b) uniform corrosion and (c) Stress Corrosion Cracking.

Pitting and uniform corrosion of chain links can be detected during inspection by cleaning a few chain
links in specific areas of the line. The severity of corrosion (pitting or uniform) can be measured by means
of underwater pictures and scanning methods or by taking a casting / imprint of the pits or by direct gauge /
calliper measurements.
Measurements taken during detailed inspection of a limited number of chain links on a mooring line can
be used in conjunction with a corrosion model and existing pit statistical distribution data to estimate the
severity of corrosion (uniform or pitting). Corrosion models and statistics on pit populations have become
available as a result of the SCORCH JIP [6] and from other Oil Companies initiatives to analyse recovered
chain links.
Based on this information, the next step is to be able to provide some advice in the form of qualitative and
quantitative criteria that can be used by the Mooring Integrity Engineers to decide if (part of) the mooring
line has to be replaced and if so when with regard to its current state and projected state trending the rate
of degradation.
The Chain FEARS JIP built upon the work that has been conducted in the SCORCH JIP quantifying the
corrosion rates that were observed in the field on existing facilities. The SCORCH JIP database of corroded
links consisted of 51 chain links from West Africa and Indonesia, all provided by Chevron. As part of the
SCORCH JIP, FEA was used to correlate the break strength of one degraded link with a physical break test
demonstrating promising levels of break strength correlation.
It was proposed to use these two experimental databases to provide answers to questions such as:

• Can we accurately predict the remaining strength of corroded chain links using FEA as seems to
be the case from the single example that has been considered?
• How important is it to know the exact geometry of the pits, and if not, can we use a simplified
method to assess the remaining strength?
• Such a simplified method would then be translated into either a practical numerical tool, or a set
of practical guidances that can be effectively used by operators as discard criteria for replacing
(part of) the mooring line.

Available Break Test Data


A subset of the SCORCH JIP scanned links were subject to break tests to determine their residual strength
providing a database, include 42 chain links scanned both pre and post break testing, measured or Actual
Breaking Load (ABL), material specimen tensile test data, video footage of break tests and elongation
measurements, although the availability of data was not consistent for all break tests conducted. In all, 6
4 OTC-29528-MS

tests were carried out on links from a West Africa FPU installation and a further 6 test from an Indonesia
FPU installation.
A review of the break tests database was conducted to identify a subset of tests to be used for development
of a calibrated FEA technique to be used for reliably estimating the ABL of degraded chain links. The data
considered included the following:

• Chain link type (assessment of both stud and studless links).

• Location and nature of failure.

• Nature, location and severity of corrosion.

• Percentage reduction in Break Load (BL).

• Availability of associated material data.

• Availability of measured elongation data.

It is evident from the break tests shown in Figure 2 that the load-elongation responses differ between the
break tests. This variation is due to a combination of factors including:

• Different make up of the chain string, including the number of degraded links which are subject to
relatively high plastic elongation, the extent of degradation of the links, and the inclusion within
the string of larger make-up links to achieve the required length for testing.
• The level of load-plateauing of the string. Most strings reaching their capacity, considered to be
limited by bar section necking, as indicated by a plateauing of the load-elongation curve, while
two strings (Sample # 2 which catastrophically failed, and Sample # 6 which appears to have local
material rupture) appear to reach their capacity prior to reaching the load plateau.

Figure 2—West Africa Break Tests - Load vs Elongation Curves

The subset of the break test samples used for correlation and some of their key attributes are presented
in Table 1. It is noted that the capacity loss is with respect to the code MBL and not the actual capacity of
an as-new link, which is demonstrated in [1] to be greater than the MBL.
OTC-29528-MS 5

Table 1—Summary of simulated chain link break test data employed for correlation

Local
Nominal
MBL1 ABL Capacity Area
Origin Sample Type Grade Diameter Failure Location
(kN) (kN) Loss2 Reduction
(mm)
%

West 3 Studlink ORQ 76 4621 4022 13 Necking, Link 2, Crown 26


Africa adjacent to Link 1

West 4 Studlink ORQ 76 4621 4287 7.2 Necking, Link 1, Crown 27


Africa adjacent to Kenter

West 5 Studlink ORQ 76 4621 4365 5.5 Necking, Link 3, Crown –


Africa adjacent to Link 4

Indonesia F8SZ Studless ORQ 70 3970 3390 14.6 Opposite flash butt weld 39

Notes:
1 Minimum Break Strength (MBL) under represents the as-new link capacity
2 With respect to Minimum Break Strength (MBL)

FEA Model Development


Model Geometry Development
Many of the links were 3D laser scanned both pre- and post-break test. The laser scanned point cloud data
surface representations were then converted into 3D CAD nurbed surface representations. These nurbed
surface representations, defined in STEP file format, were then imported into the proprietary BETA CAE
Systems ANSA v14.2.3 pre-processing software [7] for mesh and model generation. A schematic of these
geometric representations is shown in Figure 3.

Figure 3—3D Geometric Data Processing

The reported accuracy of the 3D laser scanned point cloud data to that of the scanned surface is ±0.05
mm, and that of the surfaces applied to the point cloud data is between ±0.1 and ±0.2 mm. The accumulative
tolerance between the 3D CAD representation of the link and the scanned surface is thus ±0.25 mm on
bar radius. The associated maximum error on cross section area of the 3D CAD model link representation,
assuming the maximum radial error is applied uniformly around the entire cross section circumference of
6 OTC-29528-MS

the bar, is ±1.3 to ±1.4% for the 70 mm and 76 mm break test links respectively. A systematic error resulting
in both the point cloud data and the generated 3D surface being in phase and uniform around the entire
circumference is unlikely and as such the actual error associated with the 3D CAD data representation of
the bar cross section is considered to be significantly less than these values.
A number of FE models with different extent of links comprising the modelled chain string (i.e single
link, double link and triple link) were employed to assist in correlation of tether elongation under load,
and to establish any sensitivity to end condition effects on the capacity prediction. In all cases the models
included, in addition to the full links, partial ‘load’ links to better represent the actual test interlink contact
conditions on the full link to give insight into effects of interlink contact on the chain capacity. An example
of a two link FE model is shown in Figure 4.

Figure 4—Two link model with two dummy ‘load’ links (one a Kenter shackle)

Finite Element Mesh


The finite element models were generated using ANSA pre-processing software. Individual 3D CAD surface
representations of the chain string links were imported into ANSA, oriented and aligned with respect to each
other based on photographic evidence of the chain string, and the inter-link contact surfaces brought into
close proximity to each other. While the ABAQUS [8] solution can resolve gaps between contact surfaces
to form contact, it was found to be beneficial in respect of numerical stability that the adjoining surfaces be
within 1 mm of each other as a starting point for the analysis.
Not all individual links were scanned with the same reference system and thus care was required to orient
and align the links with respect to each other in representative and compatible configuration to achieve
good seating of the inter-link contact surfaces and ensure that bending moments were not induced through
contact surface and link axis misalignment.
Due to the irregular nature of the chain surface the development of a swept, or auto volume hexahedral
element mesh of the degraded chain link volume, as is typically used for chain link assessment was not
considered feasible and instead auto meshing techniques were employed to generate a high order tetrahedral
mesh (C3D10 elements as defined in ABAQUS). A general surface mesh density of 15 mm was employed
to keep the analysis run times manageable, with local interlink contact zone and corroded pit location
refinement of element length to 5mm or less. This was found to provide increased mesh quality, better
representation of irregular surface geometry, improved contact response and increased convergence of
predicted stresses in regions of high stress gradient. Sensitivity studies were conducted employing different
element types and mesh refinement to determine the effects on capacity prediction. An example of mesh
alignment and refinement to represent the edge of a pit is shown in Figure 5.
OTC-29528-MS 7

Figure 5—Example of mesh refinement in interlink contact and high stress gradient regions

Restraints, Loads and Contact


Rigid ‘spider’ elements with a single master node located on the chain axis were applied at the cut-planes
of the ‘dummy’ end links to allow for a single point of load application and a minimum restraint system
necessary to prevent rigid bosy motion of the chain string. Tensile loading was applied through the use of
enforced displacement applied to the rigid ‘spider’ element master node on the dummy link at one end of
the chain string, while translational restraints were applied to the rigid ‘spider’ element master node on
the dummy link at the opposite end of the chain string. Numerical instability is often encountered using
the enforced displacement method if the model contains more than two link volumes and the inter-link
contact surfaces cannot be aligned sufficiently close to each other when assembling the individual links
within the model. In order to overcome this instability, an initial small force control step using Numerical
Stabilisation to initiate contact between all volumes prior to application of the enforced displacement loading
was demonstrated to be effective.
Interlink contact was modelled as ‘surface-to-surface’ contact pairs between discrete sets defined to
include only those element surfaces that may become in-contact at the inter-link contact region as shown
in Figure 6. These proved to be more robust than ‘node-to-surface’ contact pairs and less computationally
intensive and error prone than defining contact between element sets containing all element surfaces. A
‘Hard’ Contact-Pressure over-closure treatment was applied and the default slip tolerance of 0.005 was
employed for the contact.
8 OTC-29528-MS

Figure 6—Example of mesh refinement in interlink contact and high stress gradient regions

The use of offsetting nodes on the adjacent contact surfaces at the commencement of the analysis to
artificially create initial contact between adjacent contact surfaces was avoided since it results in a non-
smooth contact surface which affects stress distribution and contact surface interaction under higher loads.
An assumed friction coefficient of μ = 0.35 was employed at the inter-link contact region for the baseline
condition. A standard Coulomb friction model was applied and friction coefficients were varied as part of
a sensitivity study to determine the relationship between PBL to ABL correlation and friction coefficient.

Material Properties
An elasto-plastic material model was used to represent the material behaviour for the chain string. An
elastic modulus of 175 GPa derived based on the average linear gradient from two material tensile tests was
assumed. The true stress versus plastic strain curve for the three West Africa material sample tests were
derived from engineering stress-strain curves extracted from tensile test measurements for adjoining links
in the respective chain tethers.
Due to the limited extent of chain certification data available for the Indonesia chain, the
representativeness of the data to the break tested chain sample could not be confirmed. The minimum test
derived yield and UTS values and typical elongation at failure data and reduction of area data extracted for
ORQ grade bar stock was used to derive the true stress versus plastic strain curve using the Ramberg-Osgood
model. The Yield/UTS ratio for this material test was, consistent with that of the West Africa material tests,
at the lower end of the range of that for ORQ chain grade material. Due to the uncertainty associated with
the representativeness of the Indonesia material properties chain Sample F8SZ was assessed with both the
Indonesia and the West Africa Sample # 4 material flow curves.
The derived true stress-strain flow curves were validated by simulating the material tensile test from
which the engineering stress-strain curves were extracted for comparison with the measured data.
The ‘dummy’ end links were either defined as i) being rigid, ii) with linear-elastic material properties or
iii) with an increased yield strength over that of the chain string to ensure that they did not reach capacity at
loads lower than the mid-string failed link ABL to avoid modelled restraint mechanism specific results in the
OTC-29528-MS 9

end links. This simplified approach for the ‘dummy’ end links was taken as the function of these links was
only to represent with reasonable accuracy an inter-link contact surface for the chain string to react against.

FE Analysis and Post Processing


The simulations of the break tests were conducted using ABAQUS/Standard v6.12 software [8]. An implicit
numerical solution was conducted incorporating non-linear material properties, non-linear geometry (gap/
contact) and allowing for large deformations (i.e update of stiffness matrix). The rate of the break tests
was low enough to consider the result to be independent of inertial effects or material strain rate effects.
Post processing of the analyses was conducted using ABAQUS CAE, with time step results exported to a
spreadsheet for data reduction and the generation of force-elongation graphs.

Correlation with Break Test Data


Comparison of measured and predicted data was made where possible as in Table 2. The Indonesia chain
links were tested to destruction and thus the failed link was evident. The West Africa links were loaded only
until their maximum capacity was reached and the resistance across the chain string began to reduce from
its peak value thus catastrophic failure of the link was avoided. For these tests the link deemed to have the
lowest strength, and hence as having ‘failed’ was determined by identifying the link with the largest residual
longitudinal elongation. The mode of failure of the failed links were also determined from review of the
nature of load-elongation curve as shown in Figure 2 and from the 3D scanned data of the post-failure links.

Table 2—Summary of available chain link break test data

Break Test Overall Tether Link Deformed


Origin Break Load Failure Location Failure Mode
Sample # Elongation1 Elongation Shape

West Africa 1 X X3 X X X2 X

West Africa 4 X X3 X X X2 X

West Africa 5 X X3 X X X2 X

Indonesia F8SZ X X X X

Notes:
1 Gauge length included make-up chain links including kenter shackles
2 Measured crown contact surface to crown contact surface
3 Assumed to be the link with the largest residual elongation

Predicted Break Load (PBL) was compared to the recorded Actual Break Load (ABL). A nominal friction
coefficient of μ = 0.35 was assumed for the purposes of initial comparison, however, the effect of friction
variation was assessed to determine what friction coefficient achieved the best correlation and how this
compares with likely coefficient range for the break test conditions.
While overall tether load-elongation curves were available for the West Africa break tests the gauge
length, which includes make-up links and kenter shackles, was not known or represented in the FE
models. As such direct explicit comparison measured elongation with predictions was not possible, however
comment on the trends and overall magnitudes are made where possible. To this end, the deformed shapes
of the West Africa links were compared on as consistent a basis as possible whereby the links which were
still intack but exhibited gross plastic elongation were therby deemed to have ‘failed’. This was done to
establish the level of consistency between the break tests and the predicted deformations and to support
correlation in respect of the mode of failure. The elongation and deformed shape for the Indonesia link
could not be compared, as no load elongation data was recorded and the link failed catastrophically such
that no link elongation measurements were made.
10 OTC-29528-MS

Break Load Correlation


The comparison of PBL and ABL is presented in Table 3.

Table 3—Summary of break load correlation

Break Test
Origin ABL (kN) PBL (kN) Difference % Failure Location Failure Mode
Sample #

West Africa 1 4022 39211 –2.5% Crown of Link 2 adjacent to Link 1 Ductile Necking

<42872 <1% Capacity of 2nd weakest link (Link 2) Ductile Necking


– upper bound capacity of failed link
West Africa 4 4287
>41343 >-4% Unconverged lower bound Ductile Necking
capacity of failed link

West Africa 5 4365 4333 –0.7% Link 3 Crown adjacent to Link 4 Ductile Necking

3322 4 –2% Straight section adjacent to flash butt weld Ductile Necking
Indonesia F8SZ 3390
3439 5 1.4% Straight section adjacent to flash butt weld Ductile Necking

Notes:
1 Both one link and three link models
2 Model did not incorporate Kenter shackle attached to failed Link 1
3 Model incorporated dummy Kenter shackle attached to Link 1
4 West Africa chain ORQ properties
5 Indonesia chain properties

The two simulations of Sample #1were made, one with a single link and the other a three link model.
Both predicted failure to occur on Link 2 consistent with evidence, and both gave a PBL 2.5% less than
the ABL indicating the assessment to be independent of the number of links and associated difference in
boundary conditions (Figure 7).

Figure 7—Sample # 1 predicted plastic strain distribution and load-elongation curve

The initial three link model simulation of Sample #4 incorporated a dummy end link (and not a Kenter
shackle as employed in the actual break test) predicted failure of Link 2 with a PBL 1% greater than ABL
(Figure 8).
OTC-29528-MS 11

Figure 8—Sample # 4 predicted plastic strain distribution and load-elongation curves

A revised model two link model was assessed that incorporated a ‘dummy’ kenter shackle representation
attached to Link 1 to better represent the as-tested chain string with an increased contact zone mesh
refinement. Based on chain elongation this model correctly predicted the failure location to be at the crown
of Link 1 adjacent to the kenter shackle, albeit that the model did not run to convergence. The change
in failure location is due to the change in boundary condition of the end link (Link 1), brought about by
replacing the ‘dummy’ end link with a ‘dummy’ kenter shackle which has a larger diameter bar section at the
crown contact location. This change in failure location indicates the sensitivity to and the importance of the
representativeness of the boundary conditions applied to the model. A further investigation was conducted
to establish the sensitivity of the outcomes to the assumed kenter shackle material properties. Both elastic
and elasto-plastic properties were employed with no appreciable effect on predicted outcomes.
The two link model did not converge sufficiently to reach the load-elongation plateau, however, the
peak load reached, which was 4% less than the ABL, represents the lower bound predicted capacity of
the failed link and that had the analysis converged to the load plateau, a PBL of between −4% and 1% of
ABL would have been achieved. The numerical instability associated with the model was considered to be
associated with the high local strains present at the edge of the pit located on the interface between Link
1 and the revised boundary condition affected by the kenter shackle. This numerical instability highlights
the difficulty encountered when severe pitting corrosion is present at precisely the interface between two
adjacent contact surfaces.
Simulation of Sample # 5 predicted the failure of Link 3 to be be consistent with the evidenced break
test data and the PBL within 1% of the ABL (Figure 9).

Figure 9—Sample # 5 predicted plastic strain distribution and load-elongation curves


12 OTC-29528-MS

Simulations of Sample # F8SZ predict the location of necking onset of Link 1 and hence the location
of failure within the link, consistent with the actual break test data, to be on the straight section opposite
the flash butt weld (Figure 10). Simulations were conducted usuing material flowcurves based on both the
Indonesia material data and the measuredWest Africa (Sample # 4) engineering stress-strain curves. The
result shows that the predicted breaking load for both assumed material properties, associated with differing
section necking, is within 2% of the actual breaking load from the physical test (i.e. between −2% and
1.4%). This difference between the PBLs for the two different material flow curves is consistent with the
difference in yield strength and UTS of 4% and 2% respectively.

Figure 10—Sample # F8SZ comparison of predicted and actual location of necking

Link Elongation Correlation


A qualitative comparison of the deformed link shapes indicates that the general deformation trends from
the FE analysis are visually consistent with those from the scanned link geometry, both of which indicate
thinning in the pitted crown region typically associated with the occurrence of a ductile material response
(and on the straight section for Sample #F8SZ). The magnitude of deformation and hence strain levels,
however, are over predicted in the FE analysis, although the extent of over prediction is reduced somewhat
if elastic recovery of elongation is accounted for.
A comparison of the Sample # 1 predicted load-elongation curve (three links and two half links) and
measured load-elongation curve for the entire chain string (for unknown gauge length) is shown in Figure 11:

• The initial linear response of the break test string is less stiff than the predicted stiffness. This is
likely due to the recoverable elastic extensibility imparted by the longer break test string including
make–up links and a kenter shackle.
• The commencement of the non-linear response (apparent yield) of the FE model occurs at a lower
load than that of the measured break test response. This is likely to be as a consequence of the
proof load previously applied to the break test chains during manufacture, which will result in the
commencement of yielding again only upon application of loads close to that of the proof load.
• If the elongations that occurs below the proof load magnitude are discarded and only the post-
proof load curves considered, the difference between measured and predicted elongation levels are
smaller by around 25 mm indicating improved correlation. Despite this adjustment, the predicted
elongations remain greater than measured.
OTC-29528-MS 13

Figure 11—Comparison of Predicted and Measured elongation

Strain levels at the commencement of the necking induced load plateaus are typically less than
the assumed strain at rupture (typically around 105% for offshore mooring grade chain steel under
predominantly tensile loading). Notwithstanding the over-prediction of deformation and associated over
prediction of strain levels evidenced in the model, the magnitude of the predicted strains indicate that the
capacity of the chain link is limited by a ductile necking type failure mode, which manifests itself in the load-
elongation plateau. The predicted ductile necking type failure mode is consistent with both the evidenced
lack of material rupture and the local thinning of the crown section on the ‘failed’ link post-failure, and the
evidenced load-elongation plateau.

Sensitivity Assessment
Chain Surface Representation
It is anticipated that the representativeness of the FE model definition of the chain link surface will influence
the accuracy of the predicted capacity of the chain. Three sources of error associated with the definition of
the FE model surface are identified as follows:

• 3D laser scanned point cloud data accuracy with respect to the chain surface.

• 3D CAD surface generation from point cloud data.

• Finite element mesh generation from 3D CAD surface.

The accuracy on radial measurement associated with 3D laser scanning conducted on the break test links
is ±0.05 mm, and that in respect of 3D CAD surface generation from the point cloud data is between 0.1
mm and 0.2 mm. The capacity of the links is proportional to link cross sectional area whereby the error
in capacity associated with the definition of the geometry can be estimated by the error tolerance on the
cross sectional area.
The combined accuracy of the scanned links is estimated on this basis to be approximately 1.3%, however,
this represent an unrealistic upperbound as a systematic error resulting in both the point cloud data and
the generated 3D CAD surface being in phase and uniform around the entire circumference is unlikely.
Further, an estimate of the error in cross sectional area associated with element faceting for 5 mm elements
(as employed in the critical regions and interlink contact area) is approximately 0.3%. In this regard,
14 OTC-29528-MS

requirements to achieve adequate convergence of the solution is the main driver for mesh refinement and
not the error associated with surface faceting.
Such levels of surface modeling error on cross sectional area, even at their maximum value, are small
in comparison with other influencing factors, such as interlink friction coefficient, and are not considered
likely to have affected the underlying correlation levels achieved between the ABL and PBL magnitudes.
The stated surface definition accuracy will be influenced by the choice of CAD software, the skill of the
software user and the density of the point cloud data from which the 3D CAD surface is generated. Gaps
within the point cloud data will require that the 3D CAD surface be interpolated between the points and the
loss of accuracy through this interpolation will depend on the level of irregularity of the scanned surface and
hence the level of degradation. It is anticipated that interpolation, in particular longitudinally along a straight
section of bar, will require lower point cloud density than that required for circumferential definition, or for
the definition of an irregular pit. For this reason, a required cross sectional area accuracy cannot be explicitly
determined with respect to the point cloud data without consideration of the CAD software, the techniques
employed by the software user or the density of the point cloud data.
While neither the accuracy or the density of the laser scanning is considered to be limiting for onshore
application of currently available commercial equipment (assuming comparable to that provided for the
presented study), these factors may be limiting for emerging subsea scanning technologies. Consideration
needs to be given to these aspects of subsea photogrammetry to establish the accuracy of the scanning
method and the predicted capacity. A minimum cross sectional area accuracy of around 3-5% may likely
be reasonable in the context of determining the remaining link capacity of in-situ degraded chain links.

Element Type and Formulation


Analyses were undertaken to investigate the influence of element type selection and order. While it was not
possible to generate swept hexahedral mesh within the irregular corroded link volume, auto hex meshing
of the volume was possible. The irregularity of the volume did however, result in a high proportion of first
(low) order tetrahedral fill-in elements for this element type.
It was evident that the use of low order hexahedral elements significantly reduces the analysis time,
however, this benefit would need to be offset against the accuracy of the solution associated with
the presence of first order tetrahedral elements and possible increased meshing time associated with
minimisation of the first order tetrahedral element count.
It was evident, as shown in Figure 12 that the predicted force versus elongation curve for the standard
(C3D10) and modified (C3D10M) tetrahedral formulations are consistent, although the modified elements
had difficulty converging. The hexahedral element types, which also all produced a consistent elongation
curve, resulted in a stiffer force versus elongation curve than the tetrahedral elements and over predicted
the chain capacity by 2 to 3%. The increased stiffness of the hexahedral element models may be associated
with the high proportion of first order tetrahedral fill-in elements within the meshed volume.
OTC-29528-MS 15

Figure 12—Comparison of predicted elongation with respect to element type

First order tetrahedral elements are known to be over stiff and accordingly are not recommended for use
in structural assessments. A comparison was, however, conducted to determine the difference in PBL and
predicted elongation between second (high) order and first (low) order tetrahedral elements, and the effect
of friction coefficient on first order tetrahedral elements. The relative force displacement curves are shown
in Figure 13.

Figure 13—Comparison of predicted elongation with respect to tetrahedral element order

The findings of this study are that the the first order tetrahedral element model predict a PBL 20%
greater than the second order tetrahedral elements, and over predict the ABL by around 18%. A reduction in
assumed friction coefficient from μ = 0.35 to an unrealistic lowerbound of 0.1 reduces this ABL overestimate
to about 15%. Further, the elongation associated with the two first order tetrahedral models is significantly
less than that of the second order tetrahedral model. While it is possible to demonstrate correlation with
test data through use of stiffer low order tetrahedral elements offset by a reduced friction coefficient, such
correlation does not support validation of the method and should be avoided.
16 OTC-29528-MS

Mesh Refinement
The influence of element size in the contact zones was investigated by refining the mesh in the contact
area and failure region from the baseline size of 5 mm to 1 mm. The force-elongation and the PBL results
demonstrate that predicted behaviour is insensitive to the local refinement at the contact zone indicating
that the baseline mesh is of adequate density for convergence of the predicted capacity.
A further assessment with an increase in global mesh size of from 15 mm to 20 mm in the non-failing links
indicated the predicted force-elongation response and PBL to be insensitive to this change with the benefit
of decreased run times, although convergene issues were sometimes encountered due to overly distorted
elements. This suggests that coarsening, to the extent implemented in this assessment, can reasonably be
used to reduce analysis run-time without affecting the force versus elongation behaviour if convergence of
the solution can be achieved with the coarsened mesh.
The study has demonstrated that improved numerical stability and hence convergence to higher applied
loads was achieved with increased mesh refinement, particularly in the case of severely corroded pits located
at, or adjacent, to the inter-link contact area. The main contributor to non-convergence was excessive plastic
deformation within models employing element meshes that were either poorly formed and/or used elements
that were too large. Accordingly, numerical stability requirements on mesh refinement would appear to be
more onerous than the requirements for convergence of break strength prediction. For this reason, while an
element size of around 5 mm in the contact region and critical pitted areas is shown to be acceptable for
convergence of break strength, the nature of the pit surface irregularity may require that greater local mesh
refinement be applied for convergence to achieve the load plateau.

Stabalisation
Numerical stabilisation introduces virtual loading into the model and has been shown in other FE analyses
to result in erroneous predictions. Consequently, sensitivity analyses were undertaken to investigate the
effect that numerical stabilisation had on the predicted load–elongation curve. It was found that the effective
stiffness of the model is influenced significantly by stabilisation for loads below 60% of the PBL, however,
the force versus elongation curves converge, regardless of the level of stabilisation, for loads above
approximately 60% of the PBL. Further analyses were thus undertaken in which stabilisation was applied
only during the first 100 kN of load application where it was found that while stabilisation is active, the
effective stiffness is reduced, but a discrete change in stiffness is noted after the removal of stabilization. The
use of initial stabalisation does not significantly affect the load versus elongation curve once stabilisation
has been removed. Consequently, it is considered prudent that stabilisation should be avoided, or its use
limited where possible, and preferably forced displacement boundary conditions used instead.

Material Modelling
Sensitivity assessments were conducted to determine the relationship between PBL and different material
curves. Minor differences in elongation were evident, while the difference in PBL was consistent with the
difference in UTS.

Proof Loading
The baseline methodology employed in the analyses presented in this investigation begins with scanned link
geometry, which inherently includes the as-manufactured plastic deformation generated by proof loading.
The already physically deformed geometry is then loaded up to its peak capacity well in excess of the
proof load. The presence of residual stresses that would have been generated from the proof loading was
investigated to gauge what if any effect it had on the PBL values.
To this end analyses were undertaken to investigate the influence of applying a proof loading cycle on
both the residual capacity of chain links, and on elongation of the chain to determine whether neglecting
proof loading may contribute to the poor correlation of chain elongation at failure. Two different material
OTC-29528-MS 17

hardening models (i.e. isotropic and combined) were used to establish the effects that these may have on
predicted elongation. These analyses were undertaken based on idealised studless chain link geometry.
The following treatments were applied to investigate the influence of proof loading and hardening
models:
1. No Proof Load – The idealised as-new chain link was loaded until it approached its peak capacity.
This is defined as the baseline analysis.
2. Proof Load and Unload – The idealised as-new chain link was proof loaded, unloaded and then
reloaded until it approached its peak capacity.
Chain links previously subject to the proof/unload cycle were predicted to exhibit a stiffer initial load up
and to have an increased apparent yield point when compared to a chain link not previously subject to proof
loading and an offset in the overall level of elongation.
Both these offsets are introduced due to the material hardening that occurs during the first load increase
to the proof load. As such a link previously subject to proof loading is predicted to exhibit reduced overall
elongation levels for a given applied load and an increased apparent yield level. This elongation and
yield offset do not, however, fully account for the apparent discrepancy between measured and predicted
elongation data.
The overall force-elongation behavior (slope of curve) and the PBL at higher loads beyond the yield point
for both methods were found to be essentially the same, whereby the application of a proof/unload cycle is
not considered essential to accurately determine the PBL of the chain.
The general elongation levels and PBL determined for the two different numerical material hardening
models are consistent. This is consistent with the presence of very little, if any, reverse yielding of material
during the unload cycle, to the extent that either method would be considered representative.

Effects of Friction
A range of interlink contact friction coefficients (μ) were assessed to determine the sensitivity of PBL to
variation in assumed friction coefficient. The force-elongation curves for each of the friction analyses are
presents in Figure 14.

Figure 14—Comparison of predicted elongation with respect to friction coefficient

The analyses with the higher friction coefficients (μ = 0.6 and 0.7) did not converge to the peak load
capacity of the links and the anticipated converged values are well above the ABL. The friction coefficients
μ = 0.2 and 0.35 underestimate the ABL by 9% and 2.5% respectively while the friction coefficients μ =
0.45 overestimates the ABL by 1.5%.
18 OTC-29528-MS

The relationship between friction coefficient and PBL for the three converged solutions indicates a 4.3%
BL increase for an increase in friction coefficient μ of +0.1, and correlation between the PBL and ABL
would be achieved assuming a friction coefficient μ of +0.41.

Conclusions
The key conclusions drawn from this work package from the Chain FEARS JIP may be summarised as
follow:
1. A Finite Element (FE) method for the determination of degraded chain link capacity has been
developed based on correlation of structural performance with measured data and with support from
a series of analyses exploring the sensitivity of predicted chain link capacity to a range of modelling
input parameters.
2. The FEA method accurately predicts the location of the ‘failure’ within the chain string and the ductile
necking failure mode, determined to be the prevalent mode of failure for the chain links samples
considered in this study.
3. The Predicted Break Load (PBL) consistently underestimates the Actual Break Load (ABL) for the
assumed friction coefficient of μ = 0.35 by around 2-4% for both studlink and studless chain.
4. Interlink friction coefficient (μ) is demonstrated to have a significant effect on the predicted capacity
of the chain links with an increase in inter-link friction coefficient μ of +0.1 predicted to results in
an increase in PBL of around 5%.
5. A friction coefficient of μ = 0.41, which falls within and towards the lower end of the likely range
of friction coefficient of μ = 0.4-0.5 for breaking load tests conducted in air under dry conditions,
demonstrated good correlation between the PBL and ABL of a degraded link.
6. The effective inter-link contact friction under ‘Seawater Immersed’ conditions is expected to be
different to ‘In Air’ test conditions, with installed subsea coefficient generally acknowledged to
be approximately μ = 0.2-0.3. As a consequence of this the seawater immersed break strength is
anticipated to be around 10% less than that for in-air break tests.
7. While element formations (type) were shown to make little difference to the PBL of the links, the
use of first (low) order tetrahedral elements were shown to be over stiff and to over predict the link
capacity by up to around 20% and thus should not be used for capacity assessment.
8. Chain link elongations and strains are over-predicted by the proposed numerical modelling method.
This over-prediction is not considered to influence the correlation and hence the accuracy of the
predicted residual chain capacity (PBL) in cases where the capacity is limited by the prevalent ductile
necking failure mode. It is, however, anticipated that it will result in a conservative under-prediction of
chain residual capacity in cases where the chain capacity is limited by a material rupture failure mode.
9. Increased accuracy in the prediction of elongation and strains would be required to reduce or eliminate
this anticipated conservatism in predicted residual capacity for chains limited by a material rupture
failure mode. It is considered likely that the discrepancy in elongation is due to the limitations of the
material yield model employed and further investigation would be required to establish and correlate
a more representative multi-axial material yield model.

Acknowledgements
This paper reports on core work undertaken within the Chain FEARS JIP, which was funded by the following
companies and whose support the authors gratefully acknowledge:
American Bureau of Shipping (ABS), Asian Star, BureauVeritas, BW Offshore, Chevron, Conoco
Phillips, ExxonMobil, INPEX, OMV, Shell, SOFEC, Vicinay Innovacion, Zheng Mao.
OTC-29528-MS 19

References
[1] Potts A.E., Farrow G.H. et al, (2017) OTC Proceedings, "Investigations into Break Strength of
Offshore Mooring Chain"s, OTC-27678-MS
[2] Farrow G.H., Potts A.E. et al, (2019) OTC Proceedings, "Effects of Uniform and Mega Pitting
Corrosion on Residual Strength of Degraded Offshore Mooring Chai"n, (Manuscript Submitted),
OTC-29402-MS
[3] Farrow G.H., Potts A.E. et al, (2017) OMAE Proceedings, "Investigations into Fatigue
Performance of Offshore Mooring Chain"s, OMAE2017-62218
[4] Farrow G.H., Potts A.E. et al, (2019) OMAE Proceedings, "Review and Comparison of Collated
Offshore Mooring Chain Fatigue Test Dat"a, (Manuscript Submitted), OMAE2019-95875
[5] Farrow G.H., Potts A.E. et al, (2019) OMAE Proceedings, "Development of a New, Correlated
FEA Method of Assessing Mooring Chain Fatigu"e, (Manuscript Submitted), OMAE2019-95882
[6] Rosen J. et al, (2015), OTC Proceedings, "SCORCH JIP – Findings from Investigations into
Mooring Chain and Wire Rope Corrosion in Warm Water"s, OTC-26017
[7] Online Sources: WWW, BETA CAE Solutions, Available: https://www.beta-cae.com/ansa.htm
[8] Online Source: WWW, Dassault Systems, Available: https://www.3ds.com/products-services/
simulia/products/abaqus/

You might also like