Vous êtes sur la page 1sur 60

Preliminary Study on the Use of Computational Fluid Dynamics to

Determine the Frictional Resistance of a Trimaran Ship

Author: Alexander W. Gray

Thesis Mentor: Dr. Andrei Smirnov

Submitted To: Submitted By:

West Virginia University, The Honors College Alexander W. Gray
248 Stalnaker Hall 3909 Autumn Dr.
P.O. Box 6635 Huron, OH 44839
Morgantown, WV 26506 phone: 419-626-1388
phone: 304-293-2100 Major: Mechanical and Aerospace Engineering
Minor: Mathematics

Date of Submission:
May 4, 2007

Submitted as partial completion of the Requirements

for graduation as a University Honors Scholar

Preliminary Study on the Use of Computational Fluid Dynamics to
Determine the Frictional Resistance of a Trimaran Ship

Alexander W. Gray
Majors: Mechanical Engineering
Aerospace Engineering
Minor: Mathematics
School Address: West Virginia University
College of Engineering and Mineral Resources
P.O. Box 6070
Morgantown, WV 26506-6070

Thesis Mentor: Dr. Andrei Smirnov

The purpose of this study is to assess the use of Computational Fluid Dynamics
(CFD) for the calculation of frictional resistance. To make this assessment, CFD
simulations were implemented on a trimaran hull design; a trimaran is a three hulled ship.
The trimaran ship was modeled from the design waterline down to the keel for the
purposes of eliminating air/water interactions and wave formation that causes wave drag.
The CFD simulations were run at speeds of 2, 5, 7, 10, 14, 17, and 20 m/s and the drag
coefficient was recorded for each simulation and then used to calculate the frictional
resistance of the trimaran. The frictional resistance values from the CFD simulations
were then compared to theoretically correct frictional resistance values calculated by use
of analytical equations.
The study shows that the use of CFD to calculate the frictional resistance proves
to be more complex and requires more time than that required by use of the analytical
method. In addition, the CFD results of this study proved to be inaccurate and produced
a high percent error when compared to the theoretically correct values. The poor results
are believed to be a consequence of either an insufficient finite element mesh for the
trimaran model or an error in the setup of the CFD simulation; i.e. the choice of the
turbulent flow model. Based on the time and knowledge required to properly implement

and run CFD simulations, it is concluded that CFD is an inefficient method to use for the
calculation of the frictional resistance of a trimaran ship.

Computational Fluid Dynamics, abbreviated CFD, is defined as “a computational
technology that enables you to study the dynamics of things that flow” (FLUENT Inc.,
2007). In broader terms, CFD is a type of computer software that is designed to enable a
user to simulate the flow of fluid matter around or through a computer-generated model.
CFD is utilized in a wide variety of fields, including engineering, biology, physiology,
and meteorology to name just a few. Applications of CFD range from simple 2-D models
of flow over an airfoil shape, to complex 3-D models of dust particles entering a person’s
lungs, flow past a boat sail, airflow in a tornado, or more complex models of entire planes
and ships, to name a few examples. In this study, CFD was used to model the flow of
water passed a trimaran ship hull.
When designing any marine vessel it is necessary to determine the ships total
resistance in order to calculate the ship’s maximum speed, and to determine what size
engine is required to reach a desired cruising speed. The size of the engine will affect the
weight of the ship and the amount of fuel the ship consumes. As the cost of fuel
continues to rise, greater attention must be paid to the fuel economy of a ship. This is
where CFD comes into the picture. CFD allows ship designers to create a computer-
generated model of a ship and then test the ship at various speeds in a simulated
environment. The results from the CFD simulations can be analyzed to determine the
total resistance of the ship at each speed tested. This allows designers to determine if the
total resistance of the ship is at an acceptable level from a financial standpoint as well as
a physical standpoint. The financial perspective relates to the cost of the engine and the
fuel that the engine consumes in order to meet the ship’s mission requirements. The
physical viewpoint is in reference to the fact that most monohull ships have a maximum
achievable speed of around 30 knots, approximately 15m/s (Zhang 1997). Based on the
results of a CFD simulation, a designer can choose to rethink the ship design or proceed
to a scaled model test.

Scaled model testing in a water-filled towing tank is still considered the most
accurate method of determining the total resistance of a ship. However, scaled model
testing is a costly and time-consuming process. Therefore, ship designers utilize CFD in
order to save time and money. A designer can change the shape of a computer model
much quicker than a physical model, allowing for a faster and cheaper iterative design
process where a ship design is tested, corrected, and then re-tested using CFD until
achieving the desired results. Via CFD analysis, a manufacturer can lower costs of
building their ships by reducing the amount of time and money spent on towing tank
tests, thereby providing their buyers with optimized designs. This optimized design
process can reduce drag and fuel consumption, saving the buyers on the purchase of
costly fuel; here in lies the importance and many benefits of CFD.
The purpose of this thesis is to discuss the use of Computational Fluid Dynamics
to simulate the fluid flow around a trimaran ship hull. The issues discussed in this thesis
are relevant to any field of study that may apply CFD, but specifically to the areas of
hydrodynamics, fluid dynamics, mechanical and aerospace engineering, and marine
engineering. This thesis discusses some of the numerous methods for use in setting up
and running a CFD simulation. This thesis study will also comment on the steps taken to
create the three-dimensional Computer Aided Design (CAD), finite element mesh, and
boundary conditions of the model. In addition, the study will convey the author’s
insights into CFD simulation. These insights may aid in avoiding beginner’s mistakes
and save precious time.

Computational Fluid Dynamics can be thought of as an approach to studying the
fields of fluid dynamics, aerodynamics, and hydrodynamics. CFD is a relatively new
product of two much older approaches to the study of fluid dynamics and aerodynamics,
namely the experimental and theoretical approaches (Anderson 1995). With time, as
computers became capable of more and more complex calculations, researchers began to
analyze their experimental data and theoretical equations by use of self-written computer
programs. Soon, enough researchers were writing computer programs to solve simple
flow models and CFD was born. Now, with the advances in computer hardware and

software, CFD is constantly reaching new levels of modeling capabilities and is being
used in hundreds of academic and commercial fields ranging from biology and
engineering to car and swimsuit design.
In the past, relatively simple two-dimensional flow simulations could be coded
and the results would appear as a graph or data output. Today’s CFD programs such as
FLUENT and CFX are capable of simulating complex three-dimensional flows and the
results can be obtained in numerous outputs such as graphs, contour plots, flowlines, and
vector plots. Recent advances in CFD have enabled researchers to simulate turbulent
flows and/or mixed flows; mixed flow involves more than one fluid type. Turbulent and
mixed flow types involve numerous complex, non-linear equations to be solved
The process by which a CFD program works is beyond the scope of this paper.
However, the basic process starts by creating a computer-generated model of an object
and the desired flow domain, i.e. an airplane wing, air duct, pipe, or boat, using a
Computer Aided Design (CAD) software package. The model’s flow domain is the area
or volume through which the fluid is moving. Using the example of a pipe, the flow
domain is the entire inside of the pipe. For an airfoil (an external flow), the flow domain
consists of the region surrounding the airfoil, and the domain is as large as the user
creates it to be. After creating the model and flow domain, the CAD drawing is then
loaded into another computer program where a grid of points, also called a mesh, is
created across the model and throughout the flow domain. The grid spacing can be of
various shapes and sizes, uniform or non-uniform, structured (Figure1) or unstructured
(Figure 2), in order to properly line up with the geometry of the model.

Figure 1: F-16 Airplane with 3-D Structured Grid (Anderson)

Figure 2: Multi-element Airfoil and Flow Domain with

2-D Unstructured Grid (Anderson)

It is desired to have many grid points on and near the surface of the geometry in order to
create a proper representation of the model geometry and flow properties within the CFD
program. Figure 3 shows an airfoil with a uniform rectangular grid; Figure 4 shows an
airfoil with a stretched grid to properly represent the airfoil geometry.

Figure 3: Uniform Structured Rectangular Grid on Airfoil
and Flow Domain (Anderson)

Figure 4: Structured Non-Uniform Grid on Airfoil

and Flow Domain (Anderson)

In the last few steps of the procedure, the model is loaded into the CFD program
where the user enters information about the flow that they are simulating, such as fluid
type, temperature, and speed. The CFD program then takes the information entered by
the user and applies it to each of the model’s grid points. The final step involves an
iterative process where the governing flow equations are solved at each grid point
repeatedly, until the computer obtains a solution that is constant from one iteration to the
next; when this happens the solution is said to have converged. This is the reason for
having many grid points on or near the surface of the model geometry, the grid points are
where the flow information is being derived during the CFD simulation. Hence, the more
grid points, the more accurate the simulation and the more representative the results are
to the physical flow.
Among its many applications, CFD has been used in the fields of marine and
ocean engineering for quite some time. Because of the demand for fast ships with low
resistance, ship designers have turned to non-conventional designs to try to meet the

market demands. Conventional designs are thought of as monohull displacement and
planing vessels. A displacement ship is one that constantly displaces an amount of water
equal to its weight while in motion. A planing ship is one that moves atop the water
surface while in motion. Non-conventional designs are typically multi-hull, Small
Waterplane Area Twin Hull (SWATH) vessels, and other unique ship designs.
The vessel being studied in this thesis is a multi-hulled ship called a trimaran.
The typical trimaran design consists of three slender hulls, one long main hull and two
shorter outriggers, also called wing hulls or side hulls, as shown in Figure 5.

Figure 5: Sail Powered Trimaran Showing Location of Main Hull

and Outriggers (Latitude 38 Publishing Co., Inc.)

The trimaran’s outriggers are attached to the main hull on either side and provide the
trimaran with its excellent sea-keeping and stability characteristics (Kang, et. al. 2001).
Sea-keeping refers to the stability of a ship in a wavy environment, while at rest or in
motion. A trimaran vessel can be driven by wind, as in Figure 5, or mechanically driven
by means of propeller or water-jet.
The unique design of a trimaran ship provides the vessel with many admirable
characteristics that make it attractive for use in high speed marine applications. For
instance, the slender hulls of the trimaran design yield a small beam to length ratio and,
therefore, a decreased amount of wave making resistance (Armstrong 2004). Wave
making resistance is defined as the resistance due to the loss of energy to the formation of
waves (Harvald 1983). It has been shown by (Xu, H. and Zou, Z. 2001) and (Kang, Kuk-

Jin et. al., 2004) that the wave making resistance can also be decreased by the proper
placement of the outriggers with respect to the main hull. The lower wave making
resistance at high speeds is the biggest advantage to the use of the trimaran design and
has been well documented in numerous studies.
The main drawback of the trimaran design is the increase in wetted surface area,
S, which is the surface area of the ship hull that is in direct contact with the water
(McComas 2007). The increased wetted surface area of the trimaran creates a large body
on which the frictional resistance acts. At lower speeds, the frictional resistance is
greater than the wave making resistance causing the trimaran design to be
disadvantageous at low speeds. It is not until the high speed range, during which the
wave making resistance is greater than the frictional resistance, that the trimaran design
becomes advantageous over the monohull design. For proof of this phenomenon see
Figures 6 and 7, which show the effective power required to overcome the total resistance
on the ship and achieve a given speed.

Figure 6: Predicted Effective Power Requirements for 5000 Tonne Trimaran

and Equivalent Monohull Vessel (Skarda & Walker 2004)

Figure 7: Predicted Effective Power Requirements for 9500 Tonne Trimaran
and Equivalent Monohull Vessel (Skarda & Walker)

Although the trimaran design produces greater resistance than a monohull vessel
at slower speeds, the increased stability provided by the trimaran’s outriggers, in addition
to its lower resistance at high speeds, has allowed the trimaran to be utilized for a wide
variety of nautical applications. Trimarans are frequently used for racing, mass transit,
and recreation due to their high speed capability and excellent sea-keeping stability. The
trimaran design is also favorable for military and research applications because of the
wide deck space for placement of armaments and equipment. In addition, trimarans have
a very shallow draft; which is defined as the distance the boat extends below the water
surface. The lower draft would allow a trimaran to navigate waters and ports that are
much too shallow for an equivalent monohull ship. Several studies, (Skarda & Walker)
and (Professional Engineering 1997), have shown that the trimaran design provides added
protection from torpedo attacks and, hence, improved survivability characteristics, in
addition to the fact that the, “sleek design would shrink its radar signature” (Popular
Mechanics 1996). Due to the trimaran’s numerous military benefits, a joint study was
conducted by the U.S. Navy and the British Defense Evaluation and Research Agency
(DERA) on a two-thirds scale trimaran ship called the RV Triton (Figure 8); RV stands
for research vehicle.

Figure 8: RV Triton Test Platform for the Feasibility of the
Trimaran Design (Global Security.org 2007)

The purpose of the study was to demonstrate the feasibility of the trimaran design.
The RV Triton had a “length overall” (LOA) of 95 m, weight of 800 tonnes, and draft of
only 3 m. The results from the study on the RV Triton conducted by the hydrodynamics
department of the defense contractor QinetiQ “confirmed that the trimaran concept offers
some significant hydrodynamic benefits over a conventional monohull, with no major
drawbacks” (Renilson, et. al. 2004).
As seen, the advantages and disadvantages of the trimaran design due to frictional
resistance and wave making resistance are well known. A great deal of research has been
done on the calculation of the wave making resistance using theoretical equations and
numerical analysis. However, the frictional resistance of the trimaran is typically
obtained by theoretical calculations or by performing towing tank tests to determine the
total resistance, and then subtracting the wave making resistance to obtain an estimate of
the frictional resistance. Since, the frictional resistance of a trimaran is of such great
importance at a wide range of speeds, it was desired to devise additional methods of
calculating the frictional resistance of the trimaran. CFD analysis of the trimaran hull
form can be utilized in order to calculate the frictional resistance of the ship.
The formal objective of this study is to determine the frictional resistance of a
trimaran hull form using Computational Fluid Dynamics analysis. In doing so, the use of
CFD as a tool for the calculation of the frictional resistance will be evaluated. The
trimaran design, hull form, and dimensions were adopted from a previous study
conducted entitled, “A Preliminary Study of Trimarans” (Gray 2006).

Resistance of a marine vessel is defined as, “the fluid force(s) acting on the ship
in such a way as to oppose its motion” (Harvald). The total resistance of a ship, RT, is
divided into two main subcategories of resistance, the residuary resistance, RR, and the
frictional resistance, RF. The residuary resistance is equal to the total resistance minus
the frictional resistance (Munro-Smith 1973),
RR  RT  RF . (1)

The residuary resistance can be further broken down into numerous individual
components of resistance, with the main component being the wave making resistance,
Rw. The focus of this study however, is on the calculation of the frictional resistance.
The frictional resistance is defined as, “the component of resistance obtained by
integrating the tangential stresses over the wetted surface of the ship in the direction of
motion” (Harvald). The wetted surface area of a ship, S, is the area of the ship hull that is
in direct contact with the water. In short, the wetted surface area is the volumetric
displacement of the vessel. The wetted surface area can be calculated directly using
integral techniques, or it can be estimated using analytical approximations, two such
approximations are given in equations 2 and 3,
S  LBTC B , (2)

S  C L . (3)
In Equation 2, L is the length of the ship at the ship’s design waterline, B is the breadth of
the ship at the widest location, T is the depth (draft) of the ship, and CB is the block
coefficient. The block coefficient can be calculated using equation 4,
 (4)
CB  ,
L pp BT

where “The block coefficient is the ratio of the vessel’s volume of displacement, to the
volume of a rectangular block whose sides are equal to the breadth extreme B, the mean
draught T, and the length between perpendiculars Lpp” (Figure 9).

Figure 9: Ship Dimensional Variables (Rawson 1976)

Note, in this paper Lpp is also referred to as L, the total length of the ship at the design
waterline. The design waterline of a marine vessel is the location of the water surface
with respect to the vessel’s hull. In Equation 3, C is a correction coefficient, generally
about 2.58,  is the displacement of the vessel in tonnes, and L is the length in meters
(Munro-Smith). Using either Equation 2 or Equation 3 will give a reasonable
approximation of the wetted surface area of the ship.
One advantage to the use of Computational Fluid Dynamics is that many CFD
programs are capable of calculating several physical properties of the model’s geometry.
These properties include, for example, the integral calculation of the ship’s entire surface
area or volume. Calculating the surface area of the ship in this manner provides a more
precise answer than can be obtained by the approximations of Equations 2 or 3, since the
area is obtained by direct analysis of surface integrals. After calculating the vessel’s
wetted surface area, the value for the ship’s coefficient of friction must be obtained. The
coefficient of friction is a function of the Reynolds number and can be estimated by using
the “ITTC 1957 Model-Ship Correlation Line” (Harvald), which is produced by means
0.075 (5)
CF  .
log10 R n 22
Once the value of the ship’s coefficient of friction is established, the frictional resistance
may be calculated.

To ascertain the efficiency of using Computational Fluid Dynamics to calculate
the frictional resistance, several properties of the trimaran vessel are determined by
means of CFD simulations. These values include the wetted surface area, S, and the
coefficient of friction, CF. The frictional resistance, RF, is then calculated utilizing the
data resulting from the CFD simulations of the trimaran run at various speeds. The
processes carried out to develop solutions to the frictional resistance of the trimaran hull
form involved the use of several commercial computer software programs. The program
ProE, was utilized for the creation of the Computer Aided Design (CAD) three-
dimensional trimaran model, GAMBIT was then used to create the finite volume mesh of
the model, and lastly the CFD analysis was carried out with the use of FLUENT. The
results of the CFD simulations were examined and compared to analytically calculated
trimaran data and towing tank test data of a scaled trimaran model in order to determine
the efficiency and reliability of the application of CFD to calculate the frictional
resistance for a trimaran ship design.
In order to obtain a CFD calculation of only the frictional resistance of the
trimaran, one must take into consideration that the resistance of a ship consists of several
components of resistance. The total resistance of a ship can be broken down into at least
a dozen different components of resistance other than those already mentioned, i.e.
viscous, pressure, viscous pressure (form), appendage, and air resistances (Harvald).
Care must be taken when deciding how to model the trimaran design in order to maintain
focus on the goal of obtaining only the frictional resistance component of the trimaran.
The two main components of the total resistance are the frictional and wave making
resistances, which are summed to produce the total resistance,
RT  RR  RF . (6)
In Equation 6, RR is the residuary resistance, of which the greatest component is the wave
making resistance, RW. It is therefore desirable to model the trimaran in such a way as to
eliminate the wave making resistance, thereby isolating the frictional resistance. A novel
approach to the trimaran model was taken to achieve the elimination of the wave making
resistance when running CFD simulations. This approach involved the modeling of the
trimaran’s hulls from only the design waterline down to the keel. Figure 10 illustrates the
meaning of the design waterline and keel of a trimaran.



Figure 10: Trimaran Ship Design Waterline and Keel (Clarke 1975)
Modeling the trimaran in this manner eliminates the need to run a multiphase (i.e. more
than one fluid type) CFD simulation. By removing the air/water interface and setting the
design waterline as the top of the control volume domain, the ability to create waves, and
hence wave making resistance, is eliminated.
For this study, the trimaran model was created in the manner as just described,
from the design waterline to the keel. Station coordinates, found in Tables 1 and 2,
describing the trimaran’s planforms, which may be found in Appendix A, were used to
create a three-dimensional model of the trimaran, using CAD software.
Table 1: Main Hull Trimaran Station Coordinates (mm)

Table 2: Outrigger Trimaran Station Coordinates (mm)

The station coordinates that are given in units of mm, were converted over to (x, y, z)
coordinates with units of meters for the purpose of creating the trimaran model at full
scale in three-dimensional space. The resulting values may be found in Tables 3 and 4.

Table 3: Station Coordinates for Main Hull Converted to (x,y,z) Coordinates (m)

Table 4: Station Coordinates for Outriggers Converted to (x,y,z) Coordinates (m)

With the station coordinates for the trimaran’s main hull and outriggers converted to
(x,y,z) coordinates with units of m, the resulting dimensions for the main hull and
outriggers were as tabulated in Table 5.
Table 5: Trimaran Dimensions

Figure 11 shows a schematic representation of the dimensions referred to in Table 5. The

dimensions govern the size of the trimaran hulls, as well as the location of the wing hulls
with respect to the main hull.

Figure 11: Description of Trimaran Main and Wing Hull Dimensions (DUT 2007)

The remaining dimensions for the trimaran, as obtained from (Gray 2000), result
in the outriggers being placed at a distance b=5 m, in the y-direction from the center, o,
of the main hull and a distance of a=0 m, in the x-direction from the origin, o. The CAD
package ProE was utilized for the creation of the three-dimensional trimaran model. A
detailed description of ProE drawing process may be found in Appendix B.
Once the ProE trimaran model was complete, the fluid flow domain for the CFD
simulation was then established. The creation of the flow domain can be done using the
any CAD program, such as ProE. For this study, the trimaran drawing was first imported
into the meshing program GAMBIT. After importing the drawing, the flow domain and
finite volume mesh were created using tools within GAMBIT, the details of which may be
found in Appendix D. In an attempt to test for grid independence, which is obtained
when the results from consecutive CFD simulations change a negligible amount after
increasing the number of finite elements in the model and then retesting, the same
trimaran model was meshed two separate times. The resulting meshed trimaran models
are referred to as Trimaran5 and Trimaran7. Again, Appendix D details the meshing
process of both trimarans. Trimaran5 was meshed with a total of 219,923 finite elements
and Trimaran7, as desired, was meshed with a greater number of finite elements,
278,258. Comments will be made later on this study’s attempt to test for grid
The CFD program FLUENT, created by ANSYS Inc., was used for the simulation
of the fluid flow around the trimaran hull for the purpose of calculating the frictional
resistance of the trimaran’s hulls. A detailed listing of the steps taken to set up and
perform the FLUENT CFD analyses can be found in Appendix E. Using FLUENT, the
CFD simulations were setup and run for speeds of 2, 5, 7, 10 14, 17, and 20 m/s. The
purpose of running at multiple speeds was to gather frictional resistance data across a
spectrum of the trimaran’s possible operating speeds. A point of note need be made here;
during the first attempt at performing simulations on Trimaran5, an error was made in the
input of the value for the turbulence kinetic energy, TKE (m2/s2). The error stems from
assuming the value to be constant during initial simulations of Trimaran5. However, the
TKE is actually a function of the velocity, as noted in Equation 7 (Smirnov 2006), and
therefore not constant from simulation to simulation as the velocity changes,

TKE  0.01 * V  .
2 (7)

The simulations for the CFD model Trimaran5 were consequently run twice to
correct for the error made during the first set of simulations. The two sets of simulations
are referred to as Trimaran5 Old and Trimaran5 Correct, or simply Tri5_Old and
Trimaran5. Later the results are compared to show the effect of the TKE value on the
simulations. In addition, the results from Trimaran5 Corrected and Trimaran7 are also
compared in an attempt to prove grid independence.
Several approaches were taken in order to assess the effectiveness of using CFD
for the calculation of the frictional resistance of a trimaran. The CFD data was compared
to analytically determined values of the frictional resistance of the full-scale trimaran, as
well as data obtained from scaled model tests of the trimaran design in a water towing
Calculation of the trimaran’s frictional resistance, was used as display in Equation

 
R F  C F 1 2 V 2 S . (8)

In order to use Equation 8, one must first determine both the coefficient of friction, CF,
and the wetted surface area, S. For the analytical results, the frictional resistance was
determined using the “ITTC 1957 Model-Ship Correlation Line” (Harvald), and Equation
5, reproduced here for convenience,
0.075 (5)
CF  ,
log10 R n 22
to calculate the coefficient of friction, and Equation 8 to then calculate RF. When using
Equation 8, the wetted surface area, S, was found by use of FLUENT. Equation 5 is
dependent on the Reynolds number, Re, calculated using (Young, et. al 2004),
VL (9)
Rn  .

Since, the coefficient of friction is based on the Reynolds number, which in turn is
based on the length of the ship, the Reynolds number, coefficient of friction, and the
frictional resistance components must be calculated separately for both the main and side
hulls of the trimaran. The separate components are then added to obtain an analytical

value for the total frictional resistance of the trimaran. The numerical results for the
frictional resistance are calculated using only Equation 8 with the values for CF and S
over the entire trimaran ship being found directly from the CFD simulations using

The purpose of this preliminary study was to determine the effectiveness of using
Computational Fluid Dynamics for the calculation of the frictional resistance of the
trimaran hull form. In order to assess the effectiveness, the programs ProE, GAMBIT,
and FLUENT were utilized for the creation of the three-dimensional trimaran drawing,
meshing, and CFD simulations, respectively. The numerical results obtained from
FLUENT, in conjunction with analytical formulas, were used to calculate the frictional
resistance of the trimaran. The results obtained from these calculations were then
compared to results of a towing tank test of a scaled model trimaran. The scaled model
test data was obtained courtesy of the Naval Architecture Department at Dalian
University of Technology in Dalian, China.
The resulting three-dimensional ProE drawings can be seen in Appendix F
followed by pictures of the GAMBIT generated control volume, Trimaran5, and
Trimaran7 meshes in Appendix G. Figure 12 shows the resulting final dimensions of the
GAMBIT model that was later meshed for use in the CFD simulations.

Figure 12: Dimensions of Flow Domain for CFD Simulations
After meshing the trimaran and control volume, the CFD simulations were run
using FLUENT and the results for the drag coefficient were compiled. As mentioned
earlier, because of the way in which the trimaran and flow domain were drawn, the
resulting drag coefficient values obtained from the CFD simulations are composed of
only the frictional resistance component and a much smaller pressure drag coefficient
value. Table 6 shows the values obtained for the frictional resistance coefficient at each
speed tested for Trimaran5 Old, Trimaran5, and Trimaran7.
Table 6: Coefficient of Friction for all Trimaran CFD Simulations

After obtaining the values for the frictional resistance coefficient, the value of the
frictional resistance was then determined using Equation 8, reproduced here for

 
R F  C F 1 2 V 2 S . (8)

In order to determine the frictional resistance, the value for the wetted surface area was
first obtained from FLUENT. The resulting value for the wetted surface area was
determined to be S=233.4694 m2. The frictional resistance values for the three sets of
trimaran simulations were then compiled in Table 7.
Table 7: Calculated Frictional Resistance Values for CFD Trimaran Simulations

In order to verify that the CFD simulations are accurate, the data was compared to
analytically calculated values of the frictional resistance; the analytical values were
assumed to be theoretically correct. Using the ITTC 1957 Model-Ship Correlation Line,
Equation 5, the analytical frictional resistance coefficient values were calculated,
followed by the calculation of the frictional resistance values for the trimaran itself using
Equation 8. The theoretical value of the coefficient of friction for the side and main hull
are as presented in Table 8.
Table 8: Analytical Values for Coefficient of Friction

From the friction coefficient values for the side and main hulls, the frictional resistance of
the side hull, main hull, and total trimaran were calculated and tabulated in Table 9.

Table 9: Analytical Values for Frictional Resistance of Trimaran

After completing the numerical and analytical calculations for the coefficient of friction,
as well as the frictional resistance for the trimaran design, it is appropriate to compare the
analytical and numerical results to determine the accuracy of the CFD simulations.

The purpose of this study was to assess the use of Computational Fluid Dynamics
for the determination of the frictional resistance of the trimaran hull form. In order to
assess the use of CFD for determining the frictional resistance of the trimaran, CFD
simulations were compared to analytical calculations of the frictional resistance and
towing tank test data obtained from a scaled trimaran model. To compare the scaled
model towing tank test data to the CFD data, the results from the towing tank tests were
re-scaled to equivalent full-scale values. The data is referred to as the “Converted Model
Data.” The resulting comparisons of the numerical CFD data, the analytical values, and
the Converted Model Data can be seen in Table 10 and Figure 13.

Table 10: Frictional Resistance Data for the Trimaran Design

Figure 13: Results for Trimaran Frictional Resistance Calculations vs. Speed
Table 10 and Figure 13 clearly show that the CFD results for the trimaran
frictional resistance do not agree with the analytical results. Table 10 and Figure 13 also
show that the difference in keeping the turbulence kinetic energy constant versus
changing the value based on the velocity, as in Equation 7, yields a negligible difference
in the results from for the frictional resistance. The differences in the CFD and analytical
results yield large percent errors, as may be observed in Table 11.

Table 11: Percent Error for CFD Frictional Resistance Data Based on Analytical Data
Speed (m/s) Tri5_Old Trimaran5 Trimaran7
2 1.650E+08 1.644E+08 1.511E+08
5 1.171E+09 1.171E+09 1.081E+09
7 2.402E+09 2.403E+09 2.223E+09
10 5.139E+09 5.143E+09 4.759E+09
14 1.052E+10 1.053E+10 9.758E+09
17 1.591E+10 1.593E+10 1.476E+10
20 2.248E+10 2.250E+10 2.087E+10

A check was made to ensure that the analytical values were correctly calculated. Two
analytical resistance values at speeds of 5 m/s and 6.5 m/s were compared to the
Converted Model Data for the corresponding speeds and shown in Table 12.

Table 12: Percent Error for Analytical Frictional Resistance Data

Based on Converted Scale-Model Test Data

Speed (m/s) Analytical

5 1.53
6.5 50.84

The percent error for the speed of 5 m/s proved to be a reasonably small value.
However, the percent error for the speed at 7 m/s was higher than desired. One reason for
the high percent error may be due to the difference in calculated wetted surface area
between the converted model data and the CFD data. The wetted surface area of the
scaled model was S=2.302 m2, which converts to S=230.2 m2, whereas the wetted
surface area calculated for the CFD trimaran model was S=233.4696 m2. The differences
in wetted surface area could be due to small differences in the trimaran hull shape for the
scaled model versus the CFD model. Taking into consideration the differences in the hull
wetted surface areas between the scaled and CFD models, it is still reasonable to assume
that the analytical results were correctly calculated. Since the analytical values are based
off of the International Towing Tank Conference 1957 Model-Ship Correlation line,
which is a proven method of calculation, these analytical values are used as the
theoretically correct values for the comparison of the frictional resistance of the trimaran
Knowing that the analytical results were correctly calculated, using Equations 5
and 8, the CFD results are revisited. The percent error between the analytical results and
the CFD results is extremely high. This extreme difference indicates that the CFD results
inappropriately represent the frictional resistance on the trimaran ship. Due to the
complexity of setting up a CFD analysis, there are many opportunities for error. One
possible source of error is an inappropriate finite volume mesh of the model. To test this
theory, an initial effort was made to check for grid independence of the CFD model. If
grid independence is obtained then the finite volume mesh is appropriate, if it is not

obtained this would indicate a need to refine the mesh by increasing the number of
elements in the mesh. The efforts to check for grid independence resulted in the
simulations Trimaran5 Corrected and Trimaran7 with respective mesh sizes of 219,923
and 278,258 finite elements. The CFD simulations were run at several speeds between 2
m/s up to 20 m/s, and the results are shown in Figure 14.

Figure 14: Frictional Resistance Plot for Trimaran5 Correct and Trimaran7
Figure 14 indicates that as the speed increases the values for the frictional
resistance for Trimaran5 Corrected and Trimaran7 diverge. This leads to the conclusion
that grid independence was never reached for the CFD simulations. If there are not
enough grid points in the finite element mesh of the model, then the CFD program will
not see the appropriate geometry when running simulations. Figure 15 shows an
explanation of the correlation between the finite grid representation and the CFD
interpretation of the grid.

Figure 15: Interpretation of Finite Grid in CFD Program
In Figure 15, the black dots represent finite elements. The shape with the greater amount
of finite elements retains a better representation of its true geometric shape when
interpreted by the CFD program. The lack of grid independence leads to the conclusion
that the true shape of the trimaran ship is not being accurately depicted within FLUENT.
It is recommended that the finite element mesh of the three-dimensional trimaran
model continue to be refined until achieving grid independence. Once grid independence
has been achieved, the new CFD results could then be compared to the analytical results
in order to re-assess the accuracy of using CFD to determine the frictional resistance of
the trimaran ship.
Another source of error could come from the interpretation of the FLUENT drag
coefficient value itself. For the numerical calculations, it was assumed that the drag
coefficient value from FLUENT was a result of only the frictional resistance, having
eliminated the wave making resistance, and ignoring the contribution of the pressure or
form drag component. It is possible that the form drag component is not of a negligible
amount and, therefore, the assumption to ignore its contribution would be inappropriate.
However, this is an unlikely source of error since the form drag of a ship is typically less
then the frictional resistance, and it would certainly not be as large as the numerically
calculated resistance values.
As it stands, the results for the frictional resistance of the trimaran ship as
determined by use of CFD are inaccurate. The inaccuracy of the results means that the
CFD values cannot be used to calculate any additional information about the ship’s

hydrodynamic properties, such as the power requirements. Although the CFD results are
inaccurate, the analytically calculated results are correct and could be utilized to calculate
additional hydrodynamic properties of the ship. This leads to the assessment that CFD is
not the most efficient method of calculating the frictional resistance of the trimaran ship.
There are several reasons for this conclusion. The main reason is the significant increase
in time needed to setup and perform the CFD calculations. The time necessary to draw
and mesh the trimaran model, and run the simulations, is much greater than the time
required to perform the analytical calculations. Increasing the number of finite elements
until achieving grid independence will only increase the time required for computations
and convergence of the governing flow equations.
In addition to the increased time, the solutions for the CFD results were
inaccurate. However, it should be noted that the poor results could be due to an incorrect
user input, when the simulation reaches a boundary condition value, for example.
Because of the numerous opportunities to make an input error when setting up a CFD
simulation, it can be asserted that CFD is a difficult tool for beginners to properly use.
This is not to say that CFD should not be used or cannot be used to properly simulate the
flow around the trimaran hull for the purposes of calculating the frictional resistance of a
trimaran ship. However, if one takes into consideration the difficulty of properly setting
up the CFD simulations and the time required to design and mesh the trimaran model, to
obtain an approximate value of the frictional resistance, it is simpler and quicker to use
the analytical formulas instead of Computational Fluid Dynamics.
In order to properly implement Computational Fluid Dynamics it is important that
one have a good understanding of the governing flow physics. If CFD is implemented
without a proper understanding of the flow physics, the results are often incorrect or
misinterpreted. The analytical formulas for calculating the frictional resistance are much
simpler and do not require an understanding of the flow physics. Even with a
knowledgeable understanding of the flow physics, there is a vast array of settings within
the program FLUENT for which the user can define flow properties, boundary conditions,
flow models, etc. that it is difficult for a novice to use the CFD program to properly
model anything other than simple flows. For this reason, it is recommended that
analytical equations be used for the calculation of the frictional resistance of the trimaran.

Appendix A
Trimaran Planform

Appendix B

Detailed Description of ProE

Trimaran Drawing Process

The (x,y,z) coordinates that form the trimaran’s XY and YZ planforms (Appendix
A) were typed up in Microsoft Notepad to form separate *.ibl files describing the
contours of the trimaran’s hulls in the horizontal and vertical directions. An example
*.ibl file for the main hull’s vertical contours is included in Appendix C. Note, in order
to save the Microsoft Notepad file as an *.ibl file, it is important to change the “file type”
to “All files” when saving the file. After changing the file type, be sure to enter the
desired file name along with the .ibl ending.
The *.ibl files for the vertical and horizontal directions are easily loaded into
ProE to create datum curves. The combination of the horizontal and vertical datum
curves yield the three-dimensional wire frame trimaran model. The final steps then
involve connecting the vertical and horizontal datum curves and creating surface areas
from the wireframe model. These steps are rather basic and are not described here.
Note, for this study the main hull and outriggers were created as separate “parts”
in ProE and then added together to form a complete “assembly” drawing. The outriggers
were placed at a distance b=5 m, in the y-direction from the center, o, of the main hull
and a distance of a=0 m, in the x-direction from the origin, o.

Appendix C

Sample *.ibl File for the Vertical (YZ)

Trimaran Datum Curves

open Begin curve!14
arclength -14 1.7410 1.2 Begin curve!27
Begin section! -14 1.7820 1.6 -8 1.7000 0.8
Begin curve!1 Begin curve!15 -8 1.7410 1.2
-20 1.7000 0.8 -14 1.7820 1.6 Begin curve!28
-20 1.7410 1.2 -14 1.7920 1.7 -8 1.7410 1.2
Begin curve!2 Begin curve!16 -8 1.7820 1.6
-20 1.7410 1.2 -12 0.2430 0.4 Begin curve!29
-20 1.7820 1.6 -12 1.7000 0.8 -8 1.7820 1.6
Begin curve!3 Begin curve!17 -8 1.7920 1.7
-20 1.7820 1.6 -12 1.7000 0.8 Begin curve!30
-20 1.7920 1.7 -12 1.7410 1.2 -6 0.0500 0.28
Begin curve!4 Begin curve!18 -6 0.7640 0.4
-18 0.0500 0.7 -12 1.7410 1.2 Begin curve!31
-18 1.7000 0.8 -12 1.7820 1.6 -6 0.7640 0.4
Begin curve!5 Begin curve!19 -6 1.6940 0.8
-18 1.7000 0.8 -12 1.7820 1.6 Begin curve!32
-18 1.7410 1.2 -12 1.7920 1.7 -6 1.6940 0.8
Begin curve!6 -6 1.7380 1.2
-18 1.7410 1.2 Begin curve!20 Begin curve!33
-18 1.7820 1.6 -10 0.0500 0.36 -6 1.7380 1.2
Begin curve!7 -10 0.4960 0.4 -6 1.7790 1.6
-18 1.7820 1.6 Begin curve!21 Begin curve!34
-18 1.7920 1.7 -10 0.4960 0.4 -6 1.7790 1.6
Begin curve!8 -10 1.7000 0.8 -6 1.7890 1.7
-16 0.0500 0.6 Begin curve!22 Begin curve!35
-16 1.7000 0.8 -10 1.7000 0.8 -4 0.0500 0.24
Begin curve!9 -10 1.7410 1.2 -4 0.8290 0.4
-16 1.7000 0.8 Begin curve!23 Begin curve!36
-16 1.7410 1.2 -10 1.7410 1.2 -4 0.8290 0.4
Begin curve!10 -10 1.7820 1.6 -4 1.6620 0.8
-16 1.7410 1.2 Begin curve!24 Begin curve!37
-16 1.7820 1.6 -10 1.7820 1.6 -4 1.6620 0.8
Begin curve!11 -10 1.7920 1.7 -4 1.7250 1.2
-16 1.7820 1.6 Begin curve!25 Begin curve!38
-16 1.7920 1.7 -8 0.0500 0.32 -4 1.7250 1.2
Begin curve!12 -8 0.6550 0.4 -4 1.7660 1.6
-14 0.0500 0.5 Begin curve!26 Begin curve!39
-14 1.7000 0.8 -8 0.6550 0.4 -4 1.7660 1.6
Begin curve!13 -8 1.7000 0.8 -4 1.7770 1.7
-14 1.7000 0.8
-14 1.7410 1.2

Begin curve!40 Begin curve!54 Begin curve!68
-2 0.0500 0.20 2 1.6560 1.6 8 1.2190 1.2
-2 0.8470 0.4 2 1.6700 1.7 8 1.3840 1.6
Begin curve!41 Begin curve!55 Begin curve!69
-2 0.8470 0.4 4 0.0500 0.08 8 1.3840 1.6
-2 1.5940 0.8 4 0.6990 0.4 8 1.4190 1.7
Begin curve!42 Begin curve!56 Begin curve!70
-2 1.5940 0.8 4 0.6990 0.4 10 0.0490 0.0
-2 1.6970 1.2 4 1.2280 0.8 10 0.3930 0.4
Begin curve!43 Begin curve!67 Begin curve!71
-2 1.6970 1.2 4 1.2280 0.8 10 0.3930 0.4
-2 1.7410 1.6 4 1.5120 1.2 10 0.7280 0.8
Begin curve!44 Begin curve!58 Begin curve!72
-2 1.7410 1.6 4 1.5120 1.2 10 0.7280 0.8
-2 1.7520 1.7 4 1.6000 1.6 10 0.9990 1.2
Begin curve!45 Begin curve!59 Begin curve!73
0 0.0500 0.16 4 1.6000 1.6 10 0.9990 1.2
0 0.8230 0.4 4 1.6200 1.7 10 1.2050 1.6
Begin curve!46 Begin curve!60 Begin curve!74
0 0.8230 0.4 6 0.0500 0.04 10 1.2050 1.6
0 1.4960 0.8 6 0.6130 0.4 10 1.2470 1.7
Begin curve!47 Begin curve!61 Begin curve!75
0 1.4960 0.8 6 0.6130 0.4 12 0.0440 0.0
0 1.6500 1.2 6 1.0750 0.8 12 0.3010 0.4
Begin curve!48 Begin curve!62 Begin curve!76
0 1.6500 1.2 6 1.0750 0.8 12 0.3010 0.4
0 1.6990 1.6 6 1.4000 1.2 12 0.5680 0.8
Begin curve!49 Begin curve!63 Begin curve!77
0 1.6990 1.6 6 1.4000 1.2 12 0.5680 0.8
0 1.7110 1.7 6 1.5150 1.6 12 0.7910 1.2
Begin curve!50 Begin curve!64 Begin curve!78
2 0.0500 0.12 6 1.5150 1.6 12 0.7910 1.2
2 0.7710 0.4 6 1.5410 1.7 12 0.9830 1.6
Begin curve!51 Begin curve!65 Begin curve!79
2 0.7710 0.4 8 0.0500 0.0 12 0.9830 1.6
2 1.3710 0.8 8 0.4990 0.4 12 1.0250 1.7
Begin curve!52 Begin curve!66 Begin curve!80
2 1.3710 0.8 8 0.4990 0.4 14 0.0320 0.0
2 1.5930 1.2 8 0.8990 0.8 14 0.2100 0.4
Begin curve!53 Begin curve!67 Begin curve!81
2 1.5930 1.2 8 0.8990 0.8 14 0.2100 0.4
2 1.6560 1.6 8 1.2190 1.2 14 0.4130 0.8
Begin curve!82

14 0.4130 0.8
14 0.5870 1.2
Begin curve!83
14 0.5870 1.2
14 0.7380 1.6
Begin curve!84
14 0.7380 1.6
14 0.7740 1.7
Begin curve!85
16 0.0160 0.0
16 0.1230 0.4
Begin curve!86
16 0.1230 0.4
16 0.2570 0.8
Begin curve!87
16 0.2570 0.8
16 0.3800 1.2
Begin curve!88
16 0.3800 1.2
16 0.4960 1.6
Begin curve!89
16 0.4960 1.6
16 0.5250 1.7
Begin curve!90
18 0.0000 0.0
18 0.0470 0.4
Begin curve!91
18 0.0470 0.4
18 0.1090 0.8
Begin curve!92
18 0.1090 0.8
18 0.1750 1.2
Begin curve!93
18 0.1750 1.2
18 0.2530 1.6
Begin curve!94
18 0.2530 1.6
18 0.2710 1.7

Appendix D

Detailed Description of Modeling Processes

Using GAMBIT for Trimaran5 and Trimaran

In order to import the ProE trimaran model into GAMBIT, the drawing was saved
as a *.igs file. An *.igs file is a type of CAD drawing file that is compatible with
GAMBIT. The program GAMBIT can be used to draw the CFD model, however, its main
function is to generate the mesh on a two or three-dimensional model. After importing
the *.igs trimaran model, there are several quick procedures that can be done to ensure
that the model is not “dirty,” meaning free of holes, gaps, and excess geometry such as
lines and areas. Some of these steps include: a) the “Make Tolerant” option, which
improves geometric connectivity of lines, b) “Heal Geometry” option, which is an
alternate method to improve geometric connectivity of lines and areas, c) “Cleanup
Duplicate Faces,” used to eliminate any instance of a repeated geometric area, and d) the
“Cleanup Holes” function, used to locate holes in the geometry of the model (FLUENT
Inc. 2006).
The dimensions of the control volume for this study are a length of x=125 m, a
width of y=50 m, and a height of z=10 m. The top of the control volume is matched up to
the top of the trimaran model, which is the design waterline of the trimaran, and the
trimaran is set back 22.5 m from the front of the flow domain (Figure D-1).

Figure D-1: Dimensions of Flow Domain and Placement of Trimaran in Domain

The dimensions were set to these particular values in order to ensure that the size of the
control volume would not have an effect on the flow development during CFD
simulations. To complete the model, it is necessary to join the trimaran volume to the
control volume so that there is a connection between the two components. The
connection is necessary for the CFD program to properly understand the boundaries of
the model. GAMBIT was chosen for the creation of the control volume, because it
provides simple steps for creating and connecting the trimaran and flow domain volumes.
Several steps are taken in GAMBIT in order to join the two volumes together. The
following description of the GAMBIT processes assumes a general familiarity with the
program. In general, these steps involve (FLUENT Inc. 2006):
o Creating a “brick” with the desired dimensions of the control volume
o Properly aligning the trimaran within the control volume
o Deleting the “High Level Brick Geometry” and retaining its “Lower Geometry”
o Connect Faces (Note: areas are referred to as “faces” in GAMBIT)
- Pick top faces of trimaran hulls and top face of control volume
- Select “Real and Virtual (tolerance)” option
- Select “T-Junctions” option and apply
o Create the flow domain by “Stitching Faces”
- Select any one of the faces of the “brick”
- Retain the “Number: Single Volume” option
- Select “Type: Virtual” option and apply
With the trimaran and flow domain volumes joined together, the entire geometry
of the model was completed. GAMBIT was then utilized to generate the finite element
mesh on the trimaran areas, and the finite volume mesh throughout the control volume.
To create the finite element mesh of the trimaran, a “size function” was used in
GAMBIT. The “size function” was attached to the faces of the trimaran and the variables
used were as follows: angle=10, growth rate=1.2, size limit=10. After creating the size
function, the faces of the trimaran were meshed using triangular paved elements. Finally,
the entire model volume was meshed with tetrahedral/hybrid elements of the “TGrid”
type. The resulting mesh generated for the CFD model was of the unstructured type.

Before proceeding to the CFD program, the surfaces of the completed model must
be labeled according to the boundary types and conditions that they will represent in the
CFD simulation. The front of the control volume was labeled “INLET,” and set to be a
“velocity inlet,” meaning that the velocity is specified at this location when running the
CFD simulation. The back surface of the control volume was labeled “OUTLET” and set
as a “pressure outlet,” meaning that the pressure is a known input for the CFD simulation.
The trimaran was labeled as “HULL” and set to a “wall,” which tells the CFD simulation
that these surfaces should be treated as a solid wall. The remaining surfaces were set as
walls by default. Specification of the boundary types and conditions was the first step in
the process of setting up the actual CFD simulation and the last step performed in
GAMBIT. This model resulted in a total of 219,923 finite elements and is referred to as
Trimaran5, since it was the fifth attempt at properly creating and meshing the trimaran
CFD model.
The Trimaran7 CFD model was created following all the same steps as were used
to create the Trimaran5 model except for the meshing process. Trimaran7 was meshed
with an increased number of finite elements, 278,256 total elements, consequently
increasing the detail of the model’s geometry. As mentioned in the methodology section
of this document, the intended function of the two models was to show grid independence
for the results of the CFD simulations.
The increase in finite elements for Trimaran7 was achieved by ensuring that the
bow edge and keel edges of the trimaran’s hulls would have a greater number of finite
elements representing their geometry. To do so, first the edges that form the bow edge of
the trimaran’s main hull (Figure D-2 shows the edges) were meshed while specifying a
meshing “ratio” equal to 1.16 and “Type: successive ratio” with an “interval size” of

Figure D-2: Bow Edge of Trimaran Model

The main hull keel edges (Figure D-3) were then meshed using all default values and
changing “Type: successive ratio” to a ratio of 1.0 and an “interval size” of 0.5. The
wing hulls were then meshed in a similar manner as the main hull.

Figure D-3: Bottom View of Trimaran Model Showing Keel Edges

The edges of the wing hulls were meshed using the same settings as the main hull
bow edge, but with different values. When meshing the wing hull bow edges, a “ratio” of
1.0 and “interval size” of 0.504 was used. When meshing the keel edges, the “interval
size” was changed to 0.25. All the faces of the trimaran were then meshed using the
“Tri” elements of “type: paved”, the “default spacing” value, and an “interval size” of
0.25. The “default spacing” value follows the mesh already in place on the keel and bow
edges of the trimaran model. Lastly, the control volume of the model was meshed using
the tetrahedral/hybrid elements again, but with an “interval size” of 4.0. The finished,
ready-for-CFD-analysis, model was saved as Trimaran7.

Appendix E

Steps Followed for Setup of CFD Simulations for Tri5_Old,

Trimaran5, and Trimaran7 Using FLUENT

The main steps in running a CFD simulation involve preprocessing, running the
simulation, and post-processing. During preprocessing, the user inputs all relevant
physical properties that pertain to the simulation. The running of the simulation involves
solving iterations of the governing flow equations until the simulation converges on a
solution. Lastly, post-processing involves using CFD tools, in conjunction with one’s
knowledge of aerodynamics or hydrodynamics, to analyze the data from the convergence
of the simulation.
The first step in launching the program FLUENT was to chose the “3ddp” solver
control. “3ddp” stands for three-dimensional double precision solver, and the results
from the double precision solver are generally more accurate then the single precision
solver. The drawback of the double precision solver is that it does require longer running
times. After selecting the solver precision, click the “browse” button and search for the
trimaran file. Trimaran5 was chosen first since it posses a smaller mesh size. Once the
mesh is loaded into FLUENT, it is common practice to perform a grid “check” and a grid
“smooth/swap.” The main purpose of performing a grid check is to ensure that the
minimum volume is a non-negative value. Using the grid “smooth/swap” FLUENT visits
all the nodes in the mesh and checks instances where the skewed nodes can be made
smoother or swapped out and replaced by a virtual node. Skewness is defined as “the
difference between the shape of the cell and the shape of an equilateral cell of equivalent
volume” (FLUENT Inc. 2006). The grid swap function is repeatedly utilized until zero
nodes are swapped. This is an important step in preprocessing since, “Highly skewed
cells can decrease accuracy and destabilize the (CFD) solution” (FLUENT Inc.). The
next step would be to set the scale of the geometric units, however the scaled units of
FLUENT are SI units of length equal to meters (m), mass in kilograms (kg), etc., which
are the same units used to create the full-scale trimaran model. Thus, the scale was left at
the default values.
The next step in preprocessing is to define the “Model” properties for the
simulation. The steps taken are as follows:
1) Define → Models → Solver, default values were used
2) Define → Models → Viscous, the k-epsilon model was chosen since it is
considered to be robust and an industry standard (FLUENT Inc.)

3) Define → Models → Energy, turn on by selecting
4) Create new material: Define → Materials
a) Set name = water
b) Density  = 1000 kg/m3, specific heat capacity Cp = 4216 J/kg*K,
thermal conductivity k = 0.677 W/m*K, viscosity u = 8*10-4kg/m*s
c) Then click “change create” and be sure to click “No” when asked if
you want to overwrite air. Note: you can also choose the values for
water from the FLUENT database if desired.
5) Define → Boundary Conditions, the following names are based on the names
that were used to label the model in GAMBIT:
a) Set fluid type: zone = fluid, type = fluid, click “set” button, Specify
Material = water, click “Ok”
b) Set inlet conditions: zone = Inlet, type = velocity_inlet, click “set”,
under “Velocity Specification Method” choose “components” and set
Velocity = ____ x-dir., y-dir. = 0, z-dir. = 0, Temp. = 293 K, under
“Turbulence Specification Method” choose “k-”, the value for
Turbulence Kinetic Energy, TKE is based on Equation 7, and the
Turbulence Dissipation Rate, TDR = 0 m2/s3 (Smirnov 2006). Set
outlet conditions: zone = outlet, type = pressure_outlet, click “set,”
enter values gage pressure = 0 Pa, backflow temp. = 300 K, under
“Turbulence Specification Method” choose “k-”, set the backflow
TKE = Equation 6 value based on velocity, and backflow TDR = 0
m2/s3 (Smirnov 2006)
c) Keep the defaults for Wall = type: wall, and Hull = type: wall
6) Define → Operating Conditions, keep defaults

In step 5-b, the velocity is entered at the inlet. This step tells the CFD program
that water is going to be flowing in through this face of the control volume. The velocity
is varied at the inlet and several simulations are run, each with a different inlet velocity,
in order to make an assessment of the frictional resistance of the trimaran over a range of
operational speeds. At this point in the description, it is pertinent to point out that during

initial simulations a critical error was made in neglecting to account for the changes in
TKE value for each change in velocity, as noted in Equation E-1,
TKE  0.01 * V  .
2 (E-1)

This is equivalent to assuming a constant value for TKE during each simulation, even
though the simulations were run for speeds of 2, 5, 7, 10, 14, 17, and 20 m/s and TKE is
dependent upon the velocity. This gave the results that are referred to as Trimaran5 Old
or simply Tri5_Old. After the realization of this error, the Trimaran5 model was retested
for each speed while taking into account the change in TKE for each change in velocity,
the results are referred to as Trimaran5 Corrected or simply Trimaran5.
With the model controls, boundary conditions, and operating conditions set, the
last few steps before iterating involve setting up the monitoring of residuals and forces
during the iterations. Residuals are defined as “the small imbalance that is created during
the course of the iterative solution algorithm. A small non-zero value that typically
decreases as the solution converges” (FLUENT Inc.). The residuals can be thought of as
the truncation error of the Taylor Series developed, finite difference representations, of
the governing flow equations. In order to enable plotting of the residuals for the purpose
of monitoring convergence of the simulation, a short procedure was followed:
o Solve → Monitors → Residual…
- select “Plot” and “Print” under options
- set the convergence value to 10-6 for each residual
The residuals monitored and plotted for this study include continuity, x-velocity, y-
velocity, z-velocity, energy, k value, and epsilon ().
The “Plot” option produces an XY plot of the monitored values at every iteration.
The “Print” option produces a list of each residual monitored at every iteration. In a
similar fashion the drag force, static pressure, total pressure, and velocity of the trimaran
hull were monitored during the iterations. To monitor the force on the hull the following
procedure was performed:
o Solve → Monitors → Force…
o select “Plot” and “Print”
o under “Wall Zones” select “hull”
o under “Coeff:” select “drag” then click apply

Three surface properties of the hull were monitored during the iterations, namely
the total pressure, static pressure, and velocity magnitude. The velocity magnitude was
monitored for the sole purpose of ensuring that the boundary conditions were correctly
applied to the hull and that it had no relative velocity. In order to set up the surface
monitors the following procedure was followed:
o Solve → Monitors → Surface…
o Under “Surface Monitors” enter a value of 3
o The “Names” used for this study for each of the monitors were “stat-press,
total-press, and vel-mag”
o Click the “Plot” and “Print” options
o For each monitor click “Define”
- On the next screen under the dropdown menus, chose:
1) Pressure then static pressure
2) Pressure then total pressure
3) Velocity then magnitude
The reader is encouraged to keep in mind that the residuals, forces, and monitors
can be altered, and the user is free to monitor whatever values they desire, within the
capabilities of FLUENT. Here, the author is strictly detailing the monitors used for
current study. Also, note that the author has learned that if you wish to have access to the
exact value of each residual, force, or surface monitors anytime after running the
simulations, be certain to also select “write to file,” which creates a file of each exact
value throughout the iterations. With the monitors for the simulation set, the next step is
to begin the iterations.
To start the iteration process, the simulation must first be initialized. Initializing
the simulation tells the CFD program what the starting, known or guessed, values are.
The starting values are used by the CFD program to begin the iteration process of solving
the governing flow equations for the model. To initialize the simulation, the following
procedure was performed:
o Solve → Initialize
o From the dropdown menu choose “Inlet” under “compute from”

o Click “Init” (short for initialize)
o Close the panel
The final step before beginning the iterations is saving the case file. FLUENT saves the
preprocessing settings as a *.cas file. This is a very important step and will save a lot of
time. Otherwise, it is quite possible that a CFD simulation will crash the first few times it
is run, due to some minor discrepancies in the preprocessing setup, such as boundary or
operating conditions or due to a computer hardware problem, such as not enough free
computer memory. After saving the case file, the next step is to start the actual CFD
iterations, using the following procedure:
o Solve → Iterate
o Enter the number of desired iterations. Choose a large number to ensure
that the CFD program will iterate until all the residuals have converged.
o Click “iterate”
At this point, FLUENT will begin solving the governing flow equations by iterating until
reaching convergence of all residuals, or until the input number of iterations is reached.

Appendix F

ProE Trimaran Drawings

Figure F-1: Front View of Trimaran

Figure F-2: Skew View of Trimaran

Figure F-3: Skew View of Trimaran

Figure F-4: Side and Bottom Views of Trimaran Model

Appendix G

GAMBIT Model and Mesh Pictures

Figure G-1: Four Isometric Views of Trimaran Model and Control Volume

Figure G-2: Trimaran5 Meshed Hulls

Figure G-3: Mesh of Trimaran5 Control Volume

Figure G-4: Close-up View of Mesh Surrounding the Trimaran5 Model

Figure G-5: Finite Volume Mesh Surrounding Trimaran5 Model

Figure G-6: Trimaran7 Meshed Hulls

Figure G-7: Mesh of Trimaran7 Control Volume

Figure G-8: Close-up View of Mesh Surrounding Trimaran7

Figure G-9: Finite Volume Mesh Surrounding Trimaran7
Notice the increase in the number of finite elements in the Trimaran7 mesh versus the
Trimaran5 mesh. Additionally, notice the increased clustering of the elements in the
immediate vicinity of the Trimaran7 model. This increase provides better detail of the
flow properties near the trimaran.

Armstrong, NA. “Coming Soon to a Port Near You – The 126 Metre Austal Trimaran,”
International Conference: Design and Operation of Trimaran Ships, April 2004,
London, U.K. 2004.

Anderson, John D. Jr., Computational Fluid Dynamics: The Basics With Applications,
McGraw-Hill, Inc., New York, NY. 1995.

Clarke, D. H., Trimarans: An Introduction, Fletcher and Son Ltd., Norwich, Great
Britain. 1975.

FLUENT, Fluent Inc., date visited: 02/22/07, FLUENT Inc., Subsidiary of ANSYS Inc.,

Gray, Alexander, “A Preliminary Study of Trimarans,” Research Experience for

Undergraduates: Marine Science and Engineering China 2006, Report No. 06-13,
Department of Civil and Environmental Engineering, Clarkson University
Potsdam, NY, 2006

Harvald, SV. AA., Resistance and Propulsion of Ships, John Wiley & Sons, Inc., New
York, NY. 1983.

Kang, Kun-Jin, Lee, Chun-Ju, et. al., “Design and Hydrodynamic Performance of a
Frigate Class Trimaran,” International Conference: Design and Operation of
Trimaran Ships, April 2004, London, U.K. 2004 pgs 184-194

Munro-Smith, R. Ships and Naval Architecture. Institute of Marine Engineers, St.

Stephen’s Bristol Press Ltd., Filton, Bristol. 1973.

Rawson, K.J. Basic Ship Theory, Vol. 1. Logman Inc., New York, 1976

Renilson, Martin, Scrace, Bob, et. al., “Trials To Measure The Hydrodynamic
Performance of the RV Triton,” International Conference: Design and Operation
of Trimaran Ships, April 2004, London, U.K. 2004 pg 5 & 17.

SeaKayak, Hank McComas, date visited: 03/01/07,


Skarda, RK, & Walker, M., “Merrits of the Monohull and Trimaran Against the
Requirement for Future Surface Combatant,” International Conference: Design
and Operation of Trimaran Ships, April 2004, London, U.K. 2004 pg 132.

Smirnov, Andrei. “Re: Dr. Smirnov,” Email to Alexander Gray 13 November 2006
“Three hulls are better than one,” Professional Engineering, June 1997, Vol. 10,
Issue 12, 1997 pg 18

“Three hulls for one warship,” Popular Mechanics, February 1996, Vol. 173, Issue 2,
1996 pg 26.

Young, Donald F., Bruce R. Munson, & Theodore H. Okiishi, Fluid Mechanics, John
Wiley & Sons, Inc., U.S.A. 2004.

Zhang, Junwu, Design and Hydrodynamic Performance of Trimaran Displacement Ships,

PhD Thesis, Department of Mechanical Engineering, University College London,