OTC-28394-MS

Preliminary Design Analysis of Deepwater Production Risers

Arun Dev, Newcastle University in Singapore

Copyright 2018, Offshore Technology Conference

This paper was prepared for presentation at the Offshore Technology Conference Asia held in Kuala Lumpur, Malaysia, 20-23 March 2018.

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
Steel Catenary Risers (SCRs) and Hybrid Risers (HRs) are the most common production riser systems
used in deepwater. These concepts are field proven. SCRs and Steel Lazy Wave Risers (SLWRs), a specific
configuration of SCR, have been widely used in Gulf of Mexico (GoM). HRs have also recently gained
popularity in West of Africa, as the concept proves to be advantageous in harsh environments. This paper
addresses the global structural response of various production risers attached to a Semi-submersible in GoM.
The stress analysis was conducted based on the fundamental thoughts of SCR and HR. Thus, 26 different
models of risers (HRs, SCRs, and SLWRs) were analyzed to investigate the impacts of the various critical
considerations on the riser structure. Details of riser modeling, environmental data, and vessel particulars are
described. A primary vortex induced vibration analysis was also conducted to assess the systems further. The
studies stated above were evaluated using a dynamic analysis software, Orcaflex. An initial cost calculation
was also done for a holistic approach for selection of riser. Finally, the results obtained from the analyses
were assessed. The key findings and conclusions were described. As such, a suitable riser was determined.

Introduction
The term "deepwater" is always in flux. While previously depths of 500m was considered deepwater, the
industry has a large number of operational fields in waters deeper than 2000m. The need for resources like
oil and gas has dramatically increased due to globalization, population growth, and energy demand. The
industry has to resort to deepwater for hydrocarbons, the resources available onshore and shallow water
deplete,
It is common knowledge that with deeper seas, vessels and systems related to drilling and production
of hydrocarbons are exposed to harsh environments, and this poses various technical and economic
impediment. Furthermore, production vessels like semi-submersible and Floating Production Storage and
Offloading (FPSO) vessels used in deepwater are motion prone. Hence, the systems developed for these
conditions have to be substantially more robust, and the critical design considerations have to be extensively
evaluated. While all systems are essential for the success of the floating production platform, the production
riser, as the only connection between the production vessel and the seabed for the transport of fluids, plays
an increasingly important role.
2 OTC-28394-MS

A myriad of riser solutions is available to the industry. However, the most common production riser
solutions deployed by companies are Steel Lazy Wave Risers (SLWRs), a different configuration of SCRs,
and HRs (Bai, 2005).

Background
Numerous studies have been conducted on SCRs and HRs, but most fall short of analyzing the riser
comprehensively, to understand the impact of the various components in the risers. Studies have also not
compared both HRs and SCRs in the same environment with similar operating conditions. Thus, failing to
show the differences and comparisons of the behavior of both the risers to make a better, well-informed
choice of riser systems for ultra-deepwater.
Karunakaran et al. (2015), Cheng et al. (2013) and Yue et al. (2011) agree that the vessel offset and
buoyancy modules are vital attentions in designing SCRs and SLWRs. However, they have not shown the
structural impact of the critical considerations, on the risers. In Cascade & Chinook development, the world’s
deepest HR system, the configuration and position of the Buoyancy Can (BC) and the vessel offset were
the governing parameters in the global design (Song and Streit, 2011). This statement is similar to Webster
et al. (2011), who agreed in their paper that the BC and vessel offset are essential. However, they have not
demonstrated the impact of the parameters on the HRs.

Aim and objectives

The aim of this paper is to analyze and compare the impact of the design aspects of the various types of
production risers. Stress and vortex induced vibration (VIV) analyses will be carried out, and the results
will be used to compare and evaluate the production riser. In addition to the technical aspects, a necessary
cost comparison will be made. Hence, a suitable riser can be determined.

Methodology and Numerical Works

In this section of the paper, the methodology is described in detail. Initially, MOSES (Bentley, 2015) is
used to carry out the hydrodynamic analysis of the semi-submersible, without any mooring lines or risers
attached. The hydrodynamic study will provide the necessary data outputs like added mass and damping
coefficients, Response Amplitude Operators (RAOs), panel wave frequencies and wave drift forces. The
obtained data is then imported into Orcaflex (Orcina Ltd, 2012).
The vessel model designed in Orcaflex does not influence the results as Orcaflex only calculates vessel
movement and behaviors using the input from MOSES. However, a semi-submersible is modeled for visual
accuracy. In Orcaflex, the mooring lines and risers are attached to perform a coupled analysis. The mooring
configuration is essential as the vessel movements have a significant impact on the stresses experienced by
the riser. While eliminating the vessel movements is impossible, it is crucial that every attempt is made to
reduce the movement of the vessel. The ideal mooring configuration can be found using calculations and
simulations in Orcaflex. The process is iterative, as the mooring lines have to be continuously modified and
assessed until an acceptable vessel movement is achieved.
Fig.1 shows the design flowchart followed for designing the risers. Hoop stress formula provided by
API RP 2RD is used to obtain the wall thickness (American Petroleum Institute, 1998). The wall thickness
found using the method is valid for static analysis. However, the calculated wall thickness has to be changed
slightly through iteration when the dynamic analysis is conducted. Both types of risers were modeled in
Orcaflex as lines, with 20m segments and nodes at each end of the segment. Lines are linear elements used
to model risers and similar items. Details on modeling and specification of the various configurations of
risers are described in sub-section 2.1.
OTC-28394-MS 3

Figure 1—Design flowchart for production risers

Once the dynamic analysis has been carried out on the various types of risers, the results are tabulated
in Microsoft Excel and API RP 2RD is used to calculate von Mises stress as per regulations stated. The
calculated von Mises stress and sufficient tension, where applicable, are applied to evaluate the risers.

Design basis
In order to obtain accurate results, using accurate boundary conditions are extremely crucial. Hence, the
data used for were obtained from available literature.
Design environment. Seabed coefficients are extremely important to ensure accurate depiction of the
seabed as the Soil-Riser interaction is one of the crucial design considerations for a steel catenary riser
(Karunakaran et al., 2015). Table 1 shows the respective values used to represent the seabed condition. The
extreme wave conditions are defined in Table 2.

Table 1—Soil Properties

Parameter Values

Soil stiffness, Vertical (kN/m2) 150

Seabed stiffness, Horizontal (kN/m2) 250

Table 2—Environments Properties

Parameter Values

Wave Spectra ISSC

Significant Wave Height (m) 15.8
Wave Peak Period (s) 15.4
Current Speed at sea level (m/s) 2.4
Wind Speed (m/s) 48
4 OTC-28394-MS

Vessel modelling. The semi-submersible is modeled in MOSES. It takes the design of a typical semi-
submersible with two pontoons and six columns as shown in Fig. 2. The vessel dimensions are shown in
Table 3. To ensure that the wind forces were taken into account, derrick and cranes were modeled into the
semi-submersible as well. It is essential that the semi-submersible is designed as accurately as possible, as
any mistakes made at this point will affect the accuracy and reliability of the results obtained in Orcaflex. It
is also essential to note that the units in Orcaflex and MOSES have to be consistent with each other or the
necessary changes have to be made to the hydrodynamic data before importing. Otherwise, the imported
data will lead to inaccurate and meaningless results.

Figure 2—Semi-submersible model in MOSES

Table 3—Vessel Dimensions

Particulars Values

Length from Aft to Stern (m) 125

Length between Pontoons(m) 50
Length between columns (m) 50
Length of square columns (m) 14

Riser modelling. All risers are designed as a single line anchored to the seabed and connected to the
vessel with a flexible joint. Except for HRs, where the top end of the riser is connected to the buoyancy can
from which the flexible jumper is connected to the vessel. The azimuth of the risers is normal to the wave
direction, to ensure vessel motion has the maximum effect on the riser.
For the analysis of vessel offset on SCR and SLWR, the arc lengths (3195m), as calculated by Orcaflex, of
all the risers are the same, and all the SLWRs have a fixed buoyant length of 800m. The buoyancy modules
of the SLWR are connected at the nodes of the lines. Fig.3 and Fig.4 show the free hanging SCR and SLWR
models attached to a semi-submersible in Orcaflex.

Figure 3—SCR model in Orcaflex

OTC-28394-MS 5

Figure 4—SLWR model in Orcaflex

All SLWRs with vertical Buoyancy modules and change in buoyant length have the same offset of 2500m.
The four SLWRs with vertical buoyancy have different lengths of vertical buoyancy from 50m to 125m but
have similar arc length and overall length as the SLWR with 2500m offset. The four SLWRs with varying
buoyant length from 800m to 1500m have increased overall length proportionate to the increase in buoyant
length, but arc length is kept constant. The overall length, includes flowlines that connect the riser to the
wellhead which is not the concern in this paper. Hence, the focus will be on the arc length which is kept
constant to reduce introducing other variables, like weight, from affecting the comparison.
All HR models, apart from the four models with varying BC volume, have BC specifications as stated in
Table 5. The length of the flexible jumper is kept constant for all models except the four models with varying
vessel offset and the four models with varying BC depth. The BC is designed as a free-floating 6DOF buoy
with the rigid riser and flexible jumper attached to it. Fig. 5 depicts an example of the top section of the riser.

Figure 5—Top section of the Hybrid riser

Table 4 shows the detailed parameters of the production risers. Orcaflex calculates other values about
the properties of steel pipes based on the Poisson ratio, Young's modulus and technical data available. The
material and specifications used are similar to SCR to limit the differences so that the comparison will be
more accurate. Hence, the rigid riser specification will follow Table 4.
6 OTC-28394-MS

Table 4—Properties of SCR/SLWR/HR

Property Values

Material grade X70

Poisson’s ratio 0.29
Young's modulus(GPa) 200
Outer diameter(m) 0.274
Wall thickness (m) 0.024
Coating thickness (mm) 60
Coating density (kg/m3) 950
0.7 to
Flex joint + TJS (m)
0.274
Buoyancy module length(m) 3
Buoyancy modules diameter (m) 1.3
Buoyancy module weight (kg/m3) 395
Design pressure (psi) 10000
Seabed friction coefficient for line 0.5

Table 5—Properties of HR

Property Values

Buoyancy can diameter (m) 6

Buoyancy can length (m) 40
Buoyancy can depth (m) 200
Length of flexible jumper (m) 450

In Table 6, the mooring properties are shown. As stated previously, apart from manual calculations,
multiple iterations were required to obtain optimum mooring configuration. However, it should be noted
that semi-submersibles often use dynamic positioning in deepwater drilling and workover/completions.

Table 6—Properties of mooring lines

Property Values

Number of mooring lines 12

Mooring line length (m) 2775
Chain-
Type of mooring line Wire-
Chain
Chain length(m), top & bottom 150m
6x19
Wire type w/Fibre
Core

Results and Discussion

This section will cover the results and discussion of the analyses conducted. All stresses used are calculated
according to API RP 2RD (American Petroleum Institute, 1998). Due to the differences in length for some
risers due to the configuration, instead of comparing across the length, it will be split into three segments:
OTC-28394-MS 7

hang off point, intermediate length and touchdown point (TDP). Since the stresses are always higher at the
hang off and TDP, the intermediate length will thus not be considered.

Semi-submersible
The RAO of the semi-submersible is shown in Fig.6. It clearly shows that due to the direction of the wave,
0 degree, surge and heave are the predominate motions. The vessel could have been optimised by having
better pontoon and column designs.

Figure 6—RAO of the Semi-submersible

Floater offset on SCR and SLWR

Fig.7 shows the max stress obtained from the Orcaflex dynamic analysis. It is clear that the stresses increase
at the hang off point as the offset increases. It is due to the increased weight of the SCR. It also shows
the SLWR has almost 15% lesser stress compared to free hanging SCR and all the SLWR are within the
allowable limit. It can also be observed that when the offset increases, the stress at TDP decreases for the
SLWR but the opposite happens for SCR. It demonstrates the effectiveness of the SLWR configuration, in
negating the stresses at TDP. While it is apparent that the TDP stress for SLWR with 1500m offset is higher
than SCRs, the dynamic behavior of the riser at TDP will change due to the configuration. Hence, there are
other factors that play a part in the structural integrity of the riser. Given that TDP experiences the most
fatigue damage and is of higher priority for most SCR designs, it can be concluded that a more substantial
offset is favorable for SLWR but detrimental for SCR.

Figure 7—Stresses for SCR and SLWR with different offsets from vessel

The effective tension for the SCR was not considered as it has been shown that for deepwater application
from Fig. 7, that SLWR is more efficient and effective.
The sufficient tension is lesser for SLWR with lower vessel offset due to the weight. However, it is clear
that at around 1700m, the point where buoyancy modules are attached to the SLWR, the tension starts
to rise until the end of the buoyant section. This is due to the forced curvature of the riser pipe from the
buoyancy. After which, as shown in Fig.8, the tension reduces, The SLWR with the longer offset has, the
higher tension, which is to be expected, given the added weight due to longer lengths.
8 OTC-28394-MS

Figure 8—Effective Tension for SLWR with different offsets from vessel

SLWR with vertical buoyancy module

Fig.9 shows that the stresses for SLWR with vertical buoyancy modules have little impact on the riser
stresses, 267 to 270 MPa. Fig.10 shows a slight decrease in sufficient tension at the hang off, but eventually,
all the different risers have the same tension. Hence, the purpose of using a vertical buoyancy would be
to alleviate the hang off tension but not to reduce the overall stresses faced by the riser. However, it is
crucial to consider that with too low tension, compression might occur due to the dynamic nature of the
risers which would indefinitely lead to failure. However, it could be possible that the vertical buoyancy can
be used to mitigate fatigue damage at the hang off location, but that would require further research to be
done in the future.

Figure 9—Stresses for varying vertical buoyancy

Figure 10—Effective tension for varying vertical buoyancy

SLWR with varying vertical buoyant length

Similar to the Vertical Buoyancy method, the base riser is SLWR with the 2500m offset from the floater.
Fig.11 shows the stresses obtained. It is apparent that there is a reduction in stresses at the hang off point
as the length increases. However, when the buoyant length increases from 1000 to 1500m the stress at TDP
increases from 219 to 224 MPa. It is due to the excessive buoyancy, which causes the angle at which the riser
encounters the seabed to be too steep. It can be summarised that while the increasing the buoyant length,
lower the stresses significantly at the hang off location, it causes the stresses at the TDP to be higher. It
OTC-28394-MS 9

shows that for a given SLWR, there is an optimum buoyant length which causes the riser to experience the
least stresses. For this particular riser, the length is higher than 800m but lower than 1200m. Comparing
Fig.11 and Fig.12, it is clear that the proportion of buoyant length to the overall length of the riser plays a
vital role in the stresses experienced at TDP.

Figure 11—Stresses for different buoyant lengths

Figure 12—Effective tension for different buoyant lengths

Fig.12 shows the effective tension of the riser; it can be seen that a longer buoyant length offers lower
hang off tension but leads to a higher spike in tension at the end of the buoyant length. It also shows a
small spike, caused by excess buoyancy, at the beginning of the buoyant length before the riser. This spike
is more present for risers with buoyant length higher than 1200m. It is due to the sharp curvature at the
beginning of the buoyant length.

HR with varying offset

Table 7 shows the maximum API stress experienced by the jumper with the respective offset. While there
is a small increase in stress, the stress experienced by the jumper is far lesser than the yield strength of the
jumper. Thus, it is insignificant.

Table 7—Jumper maximum API stress for vessel offset

Vessel offset Stress (MPa)

250m 120.81
300m 120.89
350m 120.97
400m 121.11

Fig.13 shows the stresses experienced by the rigid riser, 326 to 331 MPa. Similar to the stresses of the
jumper, the vessel offset barely affected the stress experienced by the riser. It is due to the active decoupling
of the riser from the floater and is one of HRs’ advantages, especially in harsh environments.
10 OTC-28394-MS

Figure 13—Stresses for varying offsets

HR with varying depth of buoyancy can

Fig.14 shows the stresses of HR with varying depth of BC; it can be seen that the stresses increase from 315
to 328 MPa when the depth increase from 150m to 300m. Table 8 shows that the jumper stresses increase
as the depth of the BC increases, similar to the rigid riser. Even though the increase in stress is not drastic
until other analyses such as vortex forces and bending moments are carried out, it is not possible to neglect
the depth of the BC. The current present also has a major role in choosing the depth of the BC.

Figure 14—Stresses of varying depth of buoyancy can

Table 8—Jumper maximum API stress for BC depth

Depth of BC Stress (MPa)

150m 130.51
200m 130.62
250m 130.71
300m 132.75

HR with varying volume of buoyancy can

The volume of BC does not affect the stresses experienced by the jumper as shown in Table 8. While
changing the volume of BC gives rise to a higher top tension, it does not affect the jumper in any way. Fig.15
shows a large increase in stress contrary to the jumper stresses as the excess top tension enormously affects
the rigid riser. It shows that while buoyancy is necessary for the riser to be upright, too much is detrimental
to the system. Of all the considerations of HR, the volume of BC affects the stresses most. Therefore, extra
care has to go to the design and specification of the BC.
OTC-28394-MS 11

Table 9—Jumper maximum API stress for volume of BC

Stress
Diameter of Buoyancy Can
(MPa)

6m 120.86
7m 120.87
8m 120.89
9m 120.89

Figure 15—Stresses for change in BC volume

Vortex Induced Vibration

Vortex induced vibration (VIV) is a critical consideration in assessing a riser. Given the slender body
of the riser, it is more susceptible to VIV and damages caused by it. Table 10 shows the vortex forces
experienced by the rigid riser. While there is VIV present, given the low magnitude, it can easily be mitigated
by the addition of vortex suppressors. Fig.16 and Fig.17 show the in-line and transverse forces on the
flexible jumper that ranges from -2.4 to 8.1 kN/m and -8.6 to 8.9 kN/m respectively. While these effects are
significant, with strakes or other VIV suppressors, these can quickly be reduced. The in-line and transverse
forces on the SCR range from -1.1 to 24.1 kN/m and -15 to 11 kN/m respectively, as shown in Fig.18
and 19. The vortex forces experienced by SLWR are approximately 50% lesser than SCR, and there is
little to no vortex forces present after the buoyant length starts as shown in Fig.20 and 21. As previously
discussed in sub-section 3.3, vertical buoyancy modules do not lower the stresses. However, based on this
VIV analysis, it is apparent that the buoyancy modules help in reducing the VIV forces. While this is a
preliminary VIV analysis, it is conclusive that HR performs better as compared to SCR and SLWR. The
VIV forces experienced by both components of HR are far lesser.

Table 10—Rigid Riser In-line and Transverse Force

Property Values

Maximum In-line Force (kN/m) 0.0419

Minimum In-line Force (kN/m) -0.0267
Maximum Transverse Force (kN/m) 0.0243
Minimum Transverse Force (kN/m) -0.0234
12 OTC-28394-MS

Figure 16—Flexible jumper in-line force

Figure 17—Flexible jumper transverse force

Figure 18—SCR in-line force

Figure 19—SCR transverse force

OTC-28394-MS 13

Figure 20—SLWR in-line force

Figure 21—SLWR transverse force

Preliminary cost comparison of HR and SLWR

Table 11 shows the cost calculation of the risers. The costs were obtained by extrapolating the data provided,
using Microsoft Excel, by 2H Offshore to suit the risers in this paper. It is essential that the cost comparison
is made. If the cost is too high, regardless of the performance of the riser, it would not be feasible. The
table shows the SCR cost as the base cost and SCRS are still very much considered for deepwater floaters.
Further, the HR cost is almost 50% more expensive than the SLWR. Most of the cost is due to the transport,
installation, and engineering difficulty that comes with HR due to the complex components like BC and
gooseneck assemblies.

Table 11—Cost per riser

SCR SLWR
HR (M
(M (M
USD)
USD) USD)

Material and
31.25 37.5 42.5
Fabrication
Transport and
9.7 13.2 29
Installation
Engineering and
4.2 4.2 10.75
PM

Total cost 45.2 54.9 82.25

Fig.22 shows the lowest, maximum stresses experienced by the risers from the analyses above. It is clear
that the stress difference between the risers is not drastic. The results were checked according to API RP
2RD, which states that the calculated von Mises stress must be lesser than 2/3 of the yield strength of steel,
298 MPa. Thus, as shown in Fig.7 and Fig.22, all configurations of SLWR satisfies the API standard at
the maximum utilization of 0.89 of the allowable stress, which is at the hang off point of the SLWR with
2500m offset. However, with changes in the configuration of HR, it would be possible to meet the API
requirements as it only failed to meet the demand by 7%.
14 OTC-28394-MS

Figure 22—Comparison of stresses of SLWR and HR

Conclusions and future recommendations

In this paper, various production risers with multiple changes to their principal considerations have been
analyzed. Based on the analyses conducted and the findings of this limited research, it can be concluded
with the following remarks:
1. Vessel offsets have little to no impact on the stresses of HR. However, SLWR and SCR are significantly
affected by the offsets as shown in Fig.7. As such, it is proven that the decoupling ability of HR is
superior.
2. Buoyant length on SLWR is essential in lowering the stress and improving the fatigue life, but when
in excess, is detrimental to the production riser system.
3. Vertical buoyancy modules have little effect on the stress of the riser and should be used only to lower
tension at hang off location; it was also shown that the buoyant length lowers the VIV forces in SLWR.
Hence, Vertical buoyancy modules could be used to reduce VIV at the hang off point.
4. BC volume has the most significant effect on the stress of the riser; BC should be carefully designed
to get optimum performance.
5. Flexible jumper stresses are not affected by changes in HR key design considerations. However, other
factors like bend radius are but are not shown.
6. VIV suppressors are essential for SCR and SLWR. While the rigid riser does not experience many
VIV forces, the flexible jumper would benefit from the suppressors. Hence, it can be established that
HR has better fatigue performance as compared to SCR and SLWR.
7. HR costs the most, followed by SLWR and SCR costs the least. The factors that contribute to the
increase is cost are transport, installation, and engineering. Hence, if more efficient ways to install
and transport HR is established, it will make HR an attractive solution.
8. SLWR cleared the API standard check, with a utilization of 0.89, compared to the 1.07 usage for
the HR. Anything above a value of 1 means that the structure has failed the check. However, slight
modifications to HR system, e.g., increasing of wall thickness, will allow the structure to pass the
API code check as well.
Finally, based on the analyses and comparisons made, it is apparent that SLWR is the ideal riser for
deepwater due to it being $27.3 million cheaper and having passed the API RP 2RD standards.
In future research, different wave headings should be used as they will have different impacts on the riser.
Future studies should also focus on the optimisation of hang off location and hang off systems, as the stresses
are highest at hanging off point. Further investigation could be performed on the effect of configurations
on the fatigue life of the risers using SHEAR7 and VIVA can be utilized for VIV fatigue analysis. Semi-
submersibles are commonly used with Dynamic Positioning (DP). Hence, in future research, it would be
better to assess the riser with DP instead of mooring lines.
OTC-28394-MS 15

Acknowledgements
The author would like to thank Mr Vivekraj Kannalgan who did his UG (BEng with Honors in Offshore
Engineering) final year project (FYP) under the joint supervision of the author and Dr John Preedy. The
author would also like to express his gratitude to Dr John Preedy for his advice and inputs throughout this
project.

References
American Petroleum Institute, 2010. Design of risers for floating production systems (FPSs) and tension-leg platforms
(TLPs). API Publishing services, Washington DC.
Bai, Y., Bai, Q., 2005 Subsea pipelines and risers. Elsevier, London.
Bentley, 2013. Reference Manual for MOSES.
Karunakaran, D., Subramanian, S., Baarholm, R., 2015. Steel Lazy Wave Riser Configuration for Turret Moored FPSO
with disconnectable Turret in Deepwater. Proceedings of ASME 2015 34th International Conference on Ocean,
Offshore and Arctic Engineering.
Orcina Ltd., 2012. Orcaflex Manual. UK
Song, R., Streit, P., 2011. Design of World’s Deepest Hybrid Riser System for the Cascade & Chinook Development.
Proceedings of Offshore Technology Conference.
Webster, W., Kang, Z., Liang, W., Kang, Y., Sun, L., 2011. Bundled hybrid offset riser global strength analysis. Journal
of Marine Science and Application 10(4), 465–470.
Yue, B., Walter, D., Yu, W., Raghavan, K., Thompson, H., 2011. Lazy wave SCR on turret moored FPSO. Proceedings of
The Twenty-first International Offshore and Polar Engineering Conference.