See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/300076655

Slamming Load Prediction Using Overset CFD Methods

Conference Paper · May 2016

DOI: 10.4043/27254-MS

CITATIONS READS
11 3,631

5 authors, including:

Zhirong Shen
American Bureau of Shipping
19 PUBLICATIONS 481 CITATIONS

SEE PROFILE

All content following this page was uploaded by Zhirong Shen on 21 June 2016.

The user has requested enhancement of the downloaded file.

OTC-27254-MS

Slamming Load Prediction Using Overset CFD Methods

Zhirong Shen, Yi-Fang Hsieh, Zhongfu Ge, Richard Korpus, and James Huan, American Bureau of Shipping

Copyright 2016, Offshore Technology Conference

This paper was prepared for presentation at the Offshore Technology Conference held in Houston, Texas, USA, 2–5 May 2016.

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
Slamming pressures provide one of the most important load sets for the design and classification of ships
and offshore structures. These loads are highly localized and nonlinear, and therefore difficult to predict.
Computational Fluid Dynamics (CFD) is a viable tool for their estimation, but validation must be
demonstrated before its use is acceptable in standard practice. This paper addresses validation of
slamming pressure predictions for two popular software packages: OpenFOAM and Star-CCM⫹. Both
2D and 3D tests are undertaken, and the predicted pressures compared to available validation-quality
experimental data. The 2D cases involve tests of generic ship sections dropped from different heights.
Three-dimensional simulations utilize a scale model 10K TEU container ship at different speeds operating
over a range of incident wave conditions. The dynamic overset grid technique is applied to handle
large-amplitude motions. Good agreement between CFD and experimental data is achieved in the bow
area, along the bottom plating, and in the topsides flare area. A few of the comparisons (e.g. on the flat
deck) highlight areas where additional work is justified. Overall, the results demonstrate the accuracy,
flexibility and robustness of CFD performed using the overset grid technique.
Keywords: CFD · slamming · overset · OpenFOAM · Star-CCMⴙ

Introduction
Ships and offshore platforms used in the exploration, production, storage and transportation of oil all
exhibit slamming when subject to extreme weather conditions. This is true both for stationary units (such
as operating FPSOs), and moving vessels (tankers or in-transit MODUs). Unfortunately, predicting the
design loads due to slamming remains a difficult task. Model tests are expensive and simpler engineering
methods sometimes prove unreliable. Computational Fluid Dynamics (CFD) offers a first principals
alternative, and sufficient validation data now exists to quantify the expected level of accuracy. Figure 1
demonstrates a sample prediction of bow-flare slamming by CFD.
Slamming is characterized by localized high-pressure loads acting over short time periods. It can arise
from either wave impact on stationary vessels (such as column run up or topsides impact), or due to
relative motion (such as the bottom plating under a heaving bow). Either type of event is difficult to
simulate numerically. Overset grid methods are an enabling technology essential for the simultaneous
resolution of large incident waves and large relative body motions. Star-CCM⫹ and the version of
OpenFOAM applied herein both utilize the overset method.
2 OTC-27254-MS

Figure 1—Snapshots of a Slamming Event from CFD (OpenFOAM)

Recent advances in the overset grid technique have demonstrated its flexibility and efficiency for
marine and offshore applications. Carrica et al. (2008) investigated the broaching of ships in high sea
states (SS8) using the dynamic overset grid method. The authors apply the SUGGAR software package
(Noack, 2005) to handle dynamic motions of each overset grid. Chen (2010) used the Finite-Analytic
Navier-Stokes (FANS) and chimera grid approach to investigate wave loads on a fixed offshore platform.
Li et al (2015) applied the dynamic overset method coupled with multi-body dynamics to perform
aero-elastic simulations of an offshore wind turbine named NREL 5-MW. Shen et al. (2014; 2014; 2015)
coupled SUGGAR and OpenFOAM, and performed a series of CFD validations for ship self-propulsion,
seakeeping and maneuvering.
In this paper, the overset approach is applied to simulate a series of 2D and 3D model-scale
experiments for which validation-quality pressure and force measurements exist. Two-dimensional cases
include a range of ship-like sections dropped free fall from various heights into a basin of calm water.
Each section is instrumented to measure trajectory, pressure, and surface force. Three-dimensional
simulations include a model-scale ship progressing into regular waves with a number of pressure and force
measurements taken over the bow, topsides flare, deck, and stern regions. Comparisons between simulated
and experimental data are provided and demonstrate reasonable levels of accuracy.
Computational methods
Two widely used CFD packages are applied in this study: a version of OpenFOAM modified to
incorporate overset grids; and Star-CCM⫹.
OpenFOAM
OpenFOAM is an open source software package for solving computational physics problems. The version
applied here is 2.4.0 and maintained by the OpenFOAM Foundation. It provides solutions to the unsteady
incompressible Reynolds-Averaged Navier-Stokes (RANS) equations discretized using the finite volume
method (FVM). Pressure-velocity coupling is provided by the PIMPLE algorithm. The free surface
interface between air and water is resolved using the Volume of Fluid (VOF) method, and a k - ␻ SST
model provides for turbulence closure.
In order to handle large-amplitude motions of ships and platforms in big waves it was necessary to add
a dynamic overset grid capability. This method uses a stand-alone overset grid assembler to cut holes in
the grid around solid bodies, and to compute domain connectivity information (DCI) between indepen-
dently moving overset grid blocks. The code is written using the new C⫹⫹11 standard. The assembler
calls OpenFOAM to fetch grid information for fast searches and DCI computation. An overlap minimi-
zation procedure is included to find the best overlapping donor-receptor pairs and reduce overlapping
regions to the smallest allowable level for a given order of accuracy.
Each block of overset grid cells can move relative to all other pieces. Motion can be either prescribed,
or computed from a six degree-of-freedom summation of forces within that block. The overset grid
assembler can currently handle up to two levels of relative motion in a hierarchical manner. For example,
a first level of relative motion might be the ship moving through a seaway relative to an Earth-fixed
background grid. The second level of relative motion could be between the propeller and the ship and/or
OTC-27254-MS 3

the rudder and the ship. This two-level hierarchy is sufficient for most of the ship and offshore
hydrodynamic applications, but a third level could be easily added (e.g. a trim tab moving relative to the
rudder). The assembler is capable of processing grids with arbitrary polyhedral cells and thereby allows
grid importation from a wide range of mesh generations.
The described OpenFOAM version with dynamic overset grids has been validated against benchmark
cases from the Tokyo 2015 CFD Workshop (Shen and Korpus, 2015). The validation cases include
free-running simulations of ship self-propulsion, and course keeping with rotating propellers and moving
rudders. Both calm water and regular wave simulations are included and good agreement with benchmark
experimental data is obtained.

Star-CCMⴙ
Star-CCM⫹ is a commercial CFD software package developed by CD-adapco (2015). This study utilizes
version 10.02, which contains a multi-level overset mesh capability similar to that described for
OpenFOAM. Star-CCM⫹ uses the finite volume method to discretize field equations and a SIMPLE
algorithm to provide pressure-velocity coupling. The VOF method is used to capture the free surface, and
various two-equation models can be used for turbulence closure. Rigid body motion can be either
specified (as in the case of rotating propellers), or computed using the 6-DOF Dynamic Fluid Body
Interaction method (as for ship motion in waves). Hexahedra grids are generated using the Trimmer mesh
technique, and wall prism layers added to resolve boundary layers to acceptable levels for the wall
function approach.

Two-dimensional ship section drop test

Case conditions
The 2D ship section drop tests derive from experiments performed under the Wave Induced Loads on
Ships (WILS) Program. The experiments were carried out by KRISO (2014), and three different section
shapes were tested as shown in Figure 2. Models I and II are chosen for the CFD validation study
described below. The width and length of the model are 718 mm and 800 mm, respectively. The height
from the keel to the top of the model is 541 mm. The net weight of each section is different, so weights
are added to bring all test configurations to one uniform mass. The final total weight of each model is
128.54 kg. Three force sensors (F1, F2, and F3) and three pressure sensors (P1, P2, and P3) are distributed
on the model to measure local impact force and pressure (Figure 2). Impact force is measured through a
force sensor plate measuring 50 mm ⫻ 50 mm. The models are dropped from four different heights: 17
cm, 24 cm, 30 cm and 35 cm, and each height is defined based on the reference line shown in Figure 2.
In both experiment and CFD the section is modeled as a free falling body with one degree of freedom.
The only difference is that the experiment includes a spring-mounted stop to prevent the section from
dropping to the bottom of the tank. CFD simulations do not model the stop, but it should be noted that
the spring activates only well after the maximum impact pressure is recorded.
4 OTC-27254-MS

Figure 2—Simplified 2D Ship Sections and Locations of Pressure and Force Sensors (a) Side View and (b) Top View

CFD model setup

OpenFOAM The 2D computational domain has a width of 2.8 m and a height of 1.62 m for Model I.
The height of the domain increases for Model II to maintain a water depth of 1 m. The Hexpress gridder
(available as part of the FINE/Marine software package from Numeca) is used to create a computational
mesh. The grid consists of two overset components: the body-fitted part shown in red in Figure 3 (a); and
the background far field block. The same mesh cell sizes are set in the overlapping region between body
grid and background grid to maximize overset interpolation accuracy. The mesh size nearest the test
geometry is about 1 mm, and increases to 8 mm near the interface between the two grids. The background
grid has additional refinement zones near the free surface and along the path of the test section in the
course of its fall. The total mesh size for Model I is about 83,000 cells -- 25,000 cells for the body grid
and 58,000 cells for the background. The total mesh size for Models II is larger at 120,000 cells. Both
grids are two-dimensional (i.e. one cell wide). The time step is 0.0005 seconds for all runs.

Figure 3—Mesh for 2D Drop Test (Model I) (a) OpenFOAM, and (b) Star-CCMⴙ (Left Global View, Right Detail)
OTC-27254-MS 5

Star-CCMⴙ Unlike the OpenFOAM setup, the computational domain chosen for Star-CCM⫹ is
three-dimensional. The domain has a width of 2.4 m, a height of 1.68 m and a thickness of 0.2 m. The
thickness of the domain is only 25% of the actual model length (800 mm), but is wide enough to capture
the three-dimensional nature of the force sensor plates. The realizable k-␧ turbulence model is applied.
The motion of the section is modeled as a free-fall object using the DFBI algorithm. Two overset grid
blocks (body and background) are used. Figure 3 (b) displays the side view of the body grid (the dashed
region) and the background grid for Model I. The mesh is generated using StarCCM⫹’s Trimmer Mesh
method. Minimum cell size on the wall surface is about 0.5 mm, and expands to 8 mm near the overset
interface. Two refinements zones are added to the background grid to resolve the free surface and the
dropping body trajectory. The mesh size for Model I and II are about 1.2M and 1.5M, respectively. Time
step used for the simulations is 0.0005 seconds.
Results and discussions
Due to space limitations, only a subset of results will be presented. Figure 4 shows the time histories from
Model I dropped at a height of 35 cm. The top row shows pressure transducer signatures; the second row
shows force transducer signatures; and the bottom row shows trajectory measurements. Good agreements
are achieved between OpenFOAM and Star-CCM⫹ even though the former is 2D and the later 3D. Notice
that the predicted pressure peaks are generally higher than the measurements. This is a reverse from the
usual trend of CFD-to-EFD comparisons, so an investigation was undertaken to identify probable cause.
It was found that the experimental report (KRISO, 2014) expresses some concern about the performance
of the pressure sensors. Given that the force sensors do not show any questionable trending compared to
other CFD predictions, and do not have their performance brought into question, we will base any
conclusions about prediction accuracy on those. In this case, the CFD predictions seem quite accurate and
correctly capture both the trend and magnitude right up until the spring stop starts to act (around 0.4
seconds). The motions of the falling ship section are also well predicted by both CFD solvers. Figure 5
shows a similar set of results for Model II dropping from the same height (35 cm). A similar conclusion
can be drawn for this case -- pressure peaks are over-predicted by both CFD solvers, but force peaks and
motion trajectories are reasonably well predicted.

Figure 4 —Time Histories of the Impact Pressure and Force for Model I from Drop Height of 35 cm
6 OTC-27254-MS

Figure 5—Time Histories of Impact Pressure and Force for Model II from Drop Height of 35 cm

A summary including pressure and force peaks for the complete set of simulations is shown in Figures
6 and 7. For Model I both OpenFOAM and Star-CCM⫹ over-predict force peaks by 10 to 15%.
Star-CCM⫹ shows slightly better consistency than OpenFOAM, but it cannot be concluded whether this
is due to code differences or mesh dimensionality. Given the three-dimensionality of the transducers, one
is tempted to conclude 3D simulations are more reliable. For the Model II predictions in Figure 7 the
OpenFOAM results follow the Star-CCM⫹ prediction more consistently, but both predictions still
over-predict the experimental peaks by 10 to 15%.

Figure 6 —Force and Pressure Peaks at Different Locations and Drop Heights for Model I
OTC-27254-MS 7

Figure 7—Force and Pressure Peaks at Different Locations and Drop Heights for Model II

Three-dimensional ship slamming in regular waves

Case conditions
The second validation study addresses slamming loads on a 10K TEU container ship. Figure 8 shows the
ship geometry used for both CFD and model experiments, and Table 1 lists its principal dimensions. Note
that the tests involve a towed model in the Ocean Engineering wave basin at KRISO and the tests were
performed without rudder or propeller. Details of the tests can be found in KRISO (2014).

Figure 8 —Geometry of the WILS 10K TEU Containership

8 OTC-27254-MS

Table 1—Main Particulars of the Study Ship

Unit Full scale Model scale

Scale - 1/60
LOA m 336.641 5.611
LPP m 321.000 5.350
Breadth m 48.400 0.807
Depth m 27.200 0.453
Draft m 15.000 0.250
Displacement 143,741,920 ton 665.472 kg
KM 23.296 0.388
GM 2.000 0.033
KG 21.296 0.355
LCG (from AP fwd⫹) m 152.495 2.542
kxx 19.073 0.318
kyy 77.228 1.287
kzz 77.228 1.287

Fifteen force sensors (FB1~FB15) were placed on the model around the bow and forward topsides flare
area. One pressure sensor (PB) is located on the flat bottom just aft of the bow. Five force sensors
(FS1~FS5) are distributed over the stern area well aft and close to the centerline skeg. Two force sensors
are placed on the deck (FD1 and FD2) just aft of the bow, and two more mounted vertically on a
breakwater placed near the bow (FBW1 and FBW2). Locations are shown in more detail in Figure 9.
Notice that FBW1 and FBW2 are installed symmetrically on the same breakwater, and each of them
accounts for half of the total impact force on the breakwater.

Figure 9 —Locations of Impact Sensors on (a) Bow, (b) Deck and Breakwater, and (c) Stern

Four scenarios from the WILS program were tested: Cases 107, 218, 224 and 315. Each case specifies
a different wave condition and ship speed as shown in Table 2. The selected speeds range from 5 to 18
knots and wave heights from 9 to 18 m full scale. The maximum wave steepness (H/␭) is 0.07009 in
Case-224. Notice that the simulation and experiment are carried out at model scale but the results are
presented in full scale.
OTC-27254-MS 9

Table 2—Summary of the Slamming Scenarios (Full Scale Values)

Case # 107 218 224 315

Ship speed 18 kn 15 kn 5 kn 5 kn
Heading angle 180° 180° 180° 180°
Wave Length (␭/LPP) 0.9 0.8 0.8 1.0
Wave height (H) 12 m 15 m 18 m 9m
Wave Steepness (H/LPP) 0.04154 0.05841 0.07009 0.02804
Encounter period 9.47 s 9.26 s 11.37 s 12.86 s

CFD model setup

The computational domain for Star-CCM⫹ extends from 1.5 LPP upstream of the forward perpendicular
to 2.5 LPP downstream of the aft perpendicular. The domain width extends 2.0 LPP from the center plane.
The domain depth extends from 1.5 LPP below the calm waterline to 0.5 LPP above it. The domain used
by OpenFOAM is slightly different. The inlet boundary is 1.0 LPP upstream of the ship and the outlet is
2.0 LPP downstream. The distance from center plane to sidewall is 1.2 LPP. The top and bottom are 1.0
LPP above and below the free surface.
The overset grids for 3D slamming are illustrated in Figure 10. The left column shows OpenFOAM
grids and right column Star-CCM⫹ grids. The first row depicts the surface mesh on the hull, including
details of the deck and forces sensors. During mesh generation, each force sensor is treated as a separate
region from the hull so that flow-induced forces can be isolated. The second and third rows of Figure 10
show cross sections of the overset grids, and show that both codes use essentially the same strategy. Both
grids consist of two overset component layers. The first component is the hull grid, resolving the near field
and boundary layer flow on the moving ship. The second one is the Earth-fixed background grid, which
resolves the far-field flow including inflow and outflow boundary conditions, wave inputs, wave
diffraction, and wave radiation. For OpenFOAM, each of the component grid was generated using
Hexpress and then assembled using the overset grid assembler developed for OpenFOAM. For Star-
CCM⫹, the grids are generated using its internal Trimmer Mesh method and assembled by its native
overset grid assembler. Both grids consist of hex-dominant polyhedral cells.
10 OTC-27254-MS

Figure 10 —CFD Computational Grids for Three-Dimensional Ship Slamming (WILS Case-224): (a) OpenFOAM Grids from Hexpress and
(b) Star-CCMⴙ Trimmer Mesh Grids

A summary of resulting total mesh size is presented in Table 3. OpenFOAM grids for the different tests
were generated by replacing only the background grids with resolutions sufficient for each wave condition
(i.e. the hull-fixed block is recycled). For Star-CCM⫹ the complete grids are rebuilt from scratch. The
time steps are also listed in Table 3. OpenFOAM uses the same time step (⌬t ⫽ 0.002 s in model scale)
for all cases. Star-CCM⫹ applies ⌬t ⫽ 0.001 s in the high-speed conditions (Case-107 and 218) ⌬t ⫽
0.002 s and in low-speed conditions (Case-224 and 315). In the CFD simulations, the heave and pitch
motions are free to move while surge, sway, roll and yaw are fixed.
OTC-27254-MS 11

Table 3—Mesh Size and Time Step for the CFD Simulations
Case # 107 218 224 315

OpenFOAM
Hull 1,397,556 1,397,556 1,397,556 1,397,556
Background 1,572,352 1,493,976 1,493,976 1,572,352
Total 2,969,908 2,891,532 2,891,532 2,969,908
Time step 0.002s 0.002s 0.002s 0.002s
Wall clock time per time step 32 s 36 s 32 s 33 s
Star-CCMⴙ
Hull 2,396,795 1,314,960 1,280,135 2,251,781
Background 1,505,809 974,475 1,071,923 1,248,509
Total 3,902,604 2,289,435 2,352,058 3,500,290
Time step 0.001s 0.001s 0.002s 0.002s
Wall clock time per time step 42 s 26 s 30 s 39 s

Generation of input waves in StarCCM⫹ is handled using oscillatory boundary conditions imposed at
the upstream boundary. Wave reflection at the outflow and side boundaries is avoided using damping from
a numerical beach. OpenFOAM uses a slightly different approach wherein a relaxation technique is
applied to blend the far field CFD solution with a known potential flow behavior outside the viscous flow
domain. The waves2Foam library (Jacobsen et al., 2012) is used to implement the method. The result is
demonstrated by Figure 11 that shows blending zones both at inlet (for wave generation and outlet (for
wave damping).

Figure 11—Relaxation Zones for Wave Maker and Damping in OpenFOAM

Table 3 also includes run time benchmarks. It should be noted that the timings for OpenFOAM and
Star-CCM⫹ derive from different clusters. OpenFOAM simulations are performed using 16 cores (Intel
E5-2667 v2, 3.30GHz) on a single node, and Star-CCM⫹ simulations are running with 24 cores (Intel
E5-4650, 2.70GHz) on two nodes with InfiniBand (12 cores on each node). Despite the different CPUs
involved, OpenFOAM and Star-CCM⫹ show roughly similar run time performance when the total cell
and thread counts are taken into consideration.
Results and discussions
Predicted heave and pitch motions for the four simulations are shown in Figure 12. For the high-speed
conditions (Case-107 and 218), both OpenFOAM and Star-CCM⫹ obtain reasonable results compared
with measurements. The heave motion in Case-218 by Star-CCM⫹ is slightly under-predicted. For
low-speed conditions (Case-224 and 315), both solvers predict the pitch motion well. However, a
12 OTC-27254-MS

disturbance is observed in the time history of heave predicted Star-CCM⫹. It is believed this is due to
reflected waves from the far-field boundaries. In the low-speed condition, transient startup waves move
faster than the ship speed. The reflected wave is able to catch up to the ship and create a disturbance in
the ship motion.

Figure 12—Time Histories of Heave and Pitch for (a) Case-107, (b) Case-218, (c) Case-224 and (d) Case-315

As with the 2D results, only a partial set of comparisons are presented. To demonstrate validation of
bow-flare slamming force and pressure, two of the four cases (107 and 224) are selected for discussion.
Case-107 has the highest speed (18 knots full scale) and Case-224 has the largest wave height (18 m full
scale). The time history of slamming force and pressure on selected sensors over the bow (specifically
OTC-27254-MS 13

FB8, FB11, FB15 and PB) are shown in Figure 13. The pressure shown in the figures is the dynamic
pressure referenced to the initial water surface. Both OpenFOAM and Star-CCM⫹ predict force and
pressure well for most sensors. Only sensor FB11 of Case-107 under-predicts the pressure peak.

Figure 13—Time Histories of Simulated and Measured Bow-Flare Slamming Force and Pressure from Selected Sensors (a) Case-107
and (b) Case-224

Case-107 and 224 are selected to investigate the impact force from green water on deck. Figure 14
illustrates the time history of impact force on the deck (FD1) and breakwater (FBW1). The impact force
on deck does not show the same level of correlation to experiment as the bow-flare locations. While good
agreements are found for FD1 in Case-224, the same transducer in Case-107 shows that both OpenFOAM
and Star-CCM⫹ under-predict peak force. It is possible that surge motion cause this discrepancy because
the experimental motions are constrained by a spring and the CFD is not.
14 OTC-27254-MS

Figure 14 —Time Histories of Simulated and Measured Forces on Deck from Selected Sensors (a) Case-107 and (b) Case-224

Case-107 and Case-315 are selected for the investigation of stern slamming, and Figure 14 presents
time histories from the stern transducers FS1 and FS5. For Case-107 OpenFOAM shows good agreements
with Star-CCM⫹. However, both codes fail to capture the higher frequency oscillations shown in the
measurements. It is believed these high-frequency oscillations are due to whipping events induced by
slamming, and could only be properly captured in CFD if the hull were treated as a flexible beam. In
Case-315 for example, extraordinarily large force peaks are observed in the measurements but not the
CFD. Interestingly, these large force peaks do not occur at other sensor locations (e.g. FS5). Good
agreement between CFD and EFD is achieved at that sensor location. The reasons for this are unclear and
require further investigation.

Figure 15—Time Histories of Simulated and Measured Stern Slamming Forces from Selected Sensors (a) Case-107 and (b) Case-315
OTC-27254-MS 15

These bow slam induced whipping events are also observed in the Case-107 and 218 experimental
measurements. Figure 16 shows force histories for bow and stern sensors side by side using an expanded
the time scale to show detail. The bow shows a very weak whipping signal that is much exaggerated at
the stern. Note that the high frequency part of the stern signal coincides roughly with the predicted
whipping period of 2.5 seconds. It should also be noted that were this high frequency component filtered
out, the CFD predictions would provide a reasonable estimate to the measured time history. CFD solvers
can capture this oscillation, but only if a fluid/structures interaction simulation is conducted to resolve the
flexing hull.

Figure 16 —Time Histories Showing Whipping after a Slamming Event

Figure 17 summarizes the average peak force and pressure from all sensors and all cases. Both
OpenFOAM and Star-CCM⫹ show good comparisons to measured quantities in the bow-flare regions.
For the impact of green water on deck areas, there is a noticeable discrepancy in the breakwater sensors
(FBW1) for Case-107 and 224. Further investigations are underway to improve this comparison. As for
stern slamming, good agreement can be achieved for Case-107 and 218 but not for Case-224 and 315. The
extraordinarily large peak values in EFD measurements (FS1 in Case-224, FS1 and FS4 in Case-315) are
not picked up by CFD methods. This discrepancy is also under further investigation. The two CFD codes
usually show good agreement and consistent trends..
16 OTC-27254-MS

Figure 17—Average Force and Pressure Peaks from all Sensors in Case-107, 218, 224 and 315

Figures 18 and 19 illustrate three snapshots during a slamming event from Case-224 from two different
points of view. OpenFOAM, StarCCM⫹, and model test video are shown at the same instants in time.
From the front-side view (Figure 18), the water spray induced by the slamming events and water column
impacted on deck are clearly observed. Since OpenFOAM and Star-CCM⫹ both apply the VOF method,
OTC-27254-MS 17

they demonstrate good resolution of the strongly non-linear free surface. Notice that the CFD simulations
depict less water spray than the video. While part of this is likely due to excessive dissipation once the
spray sheet leaves a region of high grid resolution, some if the effect is simply due to visualization. The
free surface in CFD pictures is depicted (somewhat arbitrarily) as the location where the volume fraction
equals 0.5. When the free surface disappears in a CFD picture it does not mean that water is lost. It merely
means that the spray has spread out to the point where less than one-half the cell volume is full of water.
From the side view (Figure 19), the ship motions induced by the incident waves accompanying the
bow-flare slamming event are observed. Good correlation of ship motions and wave patterns between
CFD and experiment are shown in the figures.

Figure 18 —Three Snapshots (front-side view) of Bow Flare Slamming for Case-224 (5 knots, 180 degrees, ␭/LPPⴝ0.8, Hⴝ18m)
18 OTC-27254-MS

Figure 19 —Three Snapshots (side view) of Bow Flare Slamming for Case-224 (5 knots, 180 degrees, ␭/LPPⴝ0.8, Hⴝ18m)

Conclusions
This paper presents CFD predictions of slamming loads using dynamic overset grid technology. Both 2D
ship sections and 3D ship cases are simulated, and comparisons made to experimental data for validation.
Two finite volume solvers, OpenFOAM and Star-CCM⫹, are applied. A new dynamic overset grid
technique is developed for OpenFOAM so that both codes use similar technology to handle large-
amplitude motions during slamming.
From the computational comparisons OpenFOAM and Star-CCM⫹ predict similar trends and show
similar capabilities. Both predict bow and bottom plating impact forces well and show good agreement
with experimental data. They have similar weaknesses in some slamming areas such as a 10 to 15%
over-prediction of force in the 2D cases. In 3D they both under-predict green water impact force in the
deck and breakwater areas in. The solvers show similar run time efficiency based on 3D slamming cases.
The dynamic overset grid technique show great flexibility for simulating large-amplitude motions. It
is able to handle the motion of a free fall object and the large pitch and heave motion of a floating object.
Although only one wave direction is considered in this study, the present overset grid technique can easily
be applied to oblique waves or even directional irregular waves. The degrees of freedom can also be
extended from heave and pitch motions to full 6-DoF motions with the overset grid technique.
The results are invaluable for demonstrating the utility of CFD to predict slamming loads. Accuracy
and run time estimates allow other researchers to judge the practicality of the method for their own uses,
and the best practices described herein allow them to transition the knowledge to their own applications.
Although the number of practical examples is small, the results also provide significant data for designers
to estimate the magnitude and extent of slamming loads on similar vessels.

References
Carrica, P. M., Paik, K. J., Hosseini, H. S., and Stern, F. 2008. URANS analysis of a broaching event in irregular quartering
seas. Journal of Marine Science and Technology, 13 (4): 395–407.
CD-adapco. 2015. Star-CCM⫹ User Guide 10.02.
 OTC-27254-MS 19

Chen, H. C. 2010. Time-domain simulation of nonlinear wave impact loads on fixed offshore platform and decks.
Presented at the 20th International Offshore and Polar Engineering Conference (pp. 364 –371). Presented at the 20th
International Offshore and Polar Engineering Conference, Beijing, China.
Jacobsen, N. G., Fuhrman, D. R., and Fredsøe, J. 2012. A wave generation toolbox for the open-source CFD library:
OpenFoam®. International Journal for Numerical Methods in Fluids, 70 (9): 1073–1088.
KRISO. 2014. Wave Induced Loads on Ships -- Joint Industry Project-III. Korea Research Institute of Ships & Ocean
Engineering.
Li, Y., Castro, A. M., Sinokrot, T., Prescott, W., and Carrica, P. M. 2015. Coupled multi-body dynamics and CFD for wind
turbine simulation including explicit wind turbulence. Renewable Energy, 76: 338 –361.
Noack, R. W. 2005. SUGGAR: a general capability for moving body overset grid assembly. Presented at the 17th AIAA
Computational Fluid Dynamics Conference. Toronto, Ontario, Canada. AIAA 2005-5117.
Shen, Z., Carrica, P. M., and Wan, D. 2014. Ship motions of KCS in head waves with rotating propeller using overset grid
method. Presented at the 33rd International Conference on Ocean, Offshore and Arctic Engineering. San Francisco,
California, USA.
Shen, Z. and Korpus, R. 2015. Numerical Simulations of Ship Self-Propulsion and Maneuvering Using Dynamic Overset
Grids in OpenFOAM. Presented at the Tokyo 2015 A Workshop on CFD in Ship Hydrodynamics. Presented at the
Tokyo 2015 A Workshop on CFD in Ship Hydrodynamics, Tokyo, Japan.
Shen, Z., Wan, D., and Carrica, P. M. 2014. RANS Simulations of Free Maneuvers with Moving Rudders and Propellers
Using Overset Grids in OpenFOAM. Presented at the SIMMAN workshop on Verification and Validation of Ship
Maneuvering Simulation Methods. Presented at the SIMMAN workshop on Verification and Validation of Ship
Maneuvering Simulation Methods, Lyngby, Denmark.
Shen, Z., Wan, D., and Carrica, P. M. 2015. Dynamic overset grids in OpenFOAM with application to KCS self-propulsion
and maneuvering. Ocean Engineering, 108: 287–306.

View publication stats