Chapter 6

Improved Maneuvering-Based
Mathematical Model for Free-Running
Ship Motions in Following Waves Using
High-Fidelity CFD Results
and System-Identification Technique

Motoki Araki, Hamid Sadat-Hosseini, Yugo Sanada, Naoya Umeda

and Frederick Stern

Abstract Predicting maneuverability and stability of a free running ship in

following and quartering waves are one of the most important topics to prevent
broaching; however current mathematical models show quantitative errors with the
experimental data while high-fidelity CFD simulations show quantitative agreement,
which provides the opportunity to improve the mathematical models for free run-
ning ship dynamics in waves. In this study, both maneuvering coefficients and wave
model in the mathematical model are improved utilizing system identification tech-
nique and CFD free running outputs. From turning circle and zigzag calm water CFD
free running data, the maneuvering coefficients are estimated. The wave correction
parameters are introduced to improve the wave model, which are found from a few
forced and free running CFD simulations in waves. The mathematical model with the
improved parameters shows much better agreement with experiments in both calm
water and waves than the original mathematical model. The original mathematical
model was based on the maneuvering coefficients estimated from several captive
tests and wave forces calculated from linear Froude-Krylov forces and diffraction
forces based on a slender ship theory.

M. Araki · Y. Sanada · N. Umeda

Department of Naval Architecture & Ocean Engineering, Osaka University, Osaka, Japan
e-mail: [email protected]
H. Sadat-Hosseini · Y. Sanada · F. Stern (B)
IIHR—Hydroscience & Engineering, The University of Iowa, Iowa City, IA, USA
e-mail: [email protected]
H. Sadat-Hosseini
e-mail: [email protected]
M. Araki
Ocean Engineering Department, National Maritime Research Institutes, Tokyo, Japan
H. Sadat-Hosseini
Department of Mechanical and Energy Engineering, University of North Texas, Denton, TX, USA

© Springer Nature Switzerland AG 2019 91

V. L. Belenky et al. (eds.), Contemporary Ideas on Ship Stability, Fluid Mechanics
and Its Applications 119, https://doi.org/10.1007/978-3-030-00516-0_6
92 M. Araki et al.

Keywords System identification · Constrained least square method

Maneuvering coefficients · Wave model · Free-running ship

List of Symbols

a1,2,3,4 Tuning parameter for wave forces amplitude in surge

aH Rudder and hull hydrodynamic interaction coefficient in sway
B Ship breadth
b1,2,3,4 Tuning parameter for wave forces amplitude in sway
c1,2,3,4 Tuning parameter for wave forces amplitude in roll
CX Nondimensionalized drift wave force in surge
CY Nondimensionalized drift wave force in sway
CN Nondimensionalized drift wave moment in yaw
d Ship draft
d1,2,3,4 Tuning parameter for wave forces amplitude in yaw
Fr Froude number
g Gravitational acceleration
GZ Restoring arm in roll
Ix x Moment of inertia in roll
Izz Moment of inertia in yaw
Jx x Added moment of inertia in roll
Jzz Added moment of inertia in yaw
k Wave number
Kp Derivative of roll moment with roll rate
Kr Derivative of roll moment with yaw rate
KR Rudder force in roll
Ks Rotational index in nonlinear first order Nomoto’s model
Kv Derivative of roll moment with sway velocity
Kw Wave moment in roll
Di f
Kw Diffraction wave moment in roll
FK
Kw Froude-Krylov wave moment in roll
K rrr Derivative of roll moment with cubed yaw rate
K rr v Derivative of roll moment with squared yaw rate and sway velocity
K r vv Derivative of roll moment with squared sway velocity and yaw rate
K vvv Derivative of roll moment with cubed sway velocity
Kφ Derivative of roll moment with roll angle
L Ship length
lR Longitudinal position of rudder center from center of ship gravity
m Ship mass
mW D Shape parameter of Weibull distribution for wave drift forces/moment
mx Added mass in surge
my Added mass in sway
Nr Derivative of yaw moment with yaw rate
6 Improved Maneuvering-Based Mathematical Model … 93

NR Rudder force in yaw

Ns Nonlinear index in nonlinear first order Nomoto’s model
Nv Derivative of yaw moment with sway velocity
Nw Wave moment in yaw
Di f
Nw Diffraction wave moment in yaw
FK
Nw Froude-Krylov wave moment in yaw
Nrrr Derivative of yaw moment with cubed yaw rate
Nrr v Derivative of yaw moment with squared yaw rate and sway velocity
Nr vv Derivative of yaw moment with squared sway velocity and yaw rate
Nvvv Derivative of yaw moment with cubed sway velocity
Nφ Derivative of yaw moment with roll angle
p Roll rate
r Yaw rate
R Ship resistance
T Propeller thrust in surge
tR Rudder and hull hydrodynamic interaction coefficient in surge
Ts Time constant index in nonlinear first order Nomoto’s model
Tw Wave period
u Surge velocity
uw Wave particle velocity in surge
v Sway velocity
vw Wave particle velocity in sway
WC N Drift wave moment in yaw
WC X Drift wave force in surge
WCY Drift wave force in sway
xH Rudder and hull hydrodynamic interaction coefficient in yaw
XR Rudder force in surge
X rr Derivative of surge force with squared yaw rate
Xrv Derivative of surge force with yaw rate and sway velocity
X rr Derivative of surge force with squared sway velocity
Xw Wave force in surge
Di f
Xw Diffraction wave force in surge
FK
Xw Froude-Krylov wave force in surge
Yr Derivative of sway force with yaw rate
YR Rudder force in sway
Yv Derivative of sway force with sway velocity
Yw Wave force in sway
YwF K Froude-Krylov wave force in sway
Di f
Yw Diffraction wave force in sway
Yrrr Derivative of sway force with cubed yaw rate
Yrr v Derivative of sway force with squared yaw rate and sway velocity
Yr vv Derivative of sway force with squared sway velocity and yaw rate
Yvvv Derivative of sway force with cubed sway velocity
Yφ Derivative of sway force with roll angle
94 M. Araki et al.

zH Height of hydrodynamic sway force application point from center of ship

gravity
zH R Height of rudder force application point in roll
αX Tuning parameter for drift wave force in surge
αY Tuning parameter for drift wave force in sway
αN Tuning parameter for drift wave moment in yaw
β1 Tuning parameter for wave particle velocity in surge
β2 Tuning parameter for wave particle velocity in sway
γR Flow straitening coefficient
δ Rudder angle
ε Rudder effectiveness coefficient
εa1,2,3,4 Tuning parameter for wave forces phase lag in surge
εb1,2,3,4 Tuning parameter for wave forces phase lag in sway
εc1,2,3,4 Tuning parameter for wave forces phase lag in roll
εd1,2,3,4 Tuning parameter for wave forces phase lag in yaw
ηW D Scale parameter of Weibull distribution for wave drift forces/moment
ζw Wave amplitudes
λ Wave length
ξG Longitudinal position of center of ship gravity from a wave trough
ρ Water density
φ Roll angle
ψ Yaw angle
ψ0 Tuning parameter for wave drift wave force phase lag in surge
ω Wave frequency

6.1 Introduction

Maneuverability and stability of a free running ship in waves are one of the most
important topics considered in the ship design. Especially, in severe following and
quartering waves, the ship is very likely to broach and capsize.
Mathematical models (MM) and recently computational fluid dynamic (CFD),
numerically solving Navier-Stokes equation including viscous effects, are used
to predict ship stability and maneuverability in calm water and waves. The MM
approach in this paper means an approach consisting of two layered sub systems.
In the lower layer, hydrodynamic forces mainly due to potential flow are calcu-
lated by solving partial differential equations of potential flow and hydrodynamic
forces mainly due to viscosity flow are estimated with captive model experiments or
empirical formulas. In the upper layer, ship motions are calculated by solving ordi-
nary differential equations with initial conditions. Since short computational time is
required to sweep out dangerous maneuvering and wave conditions from a huge num-
ber of suspect conditions, the MM method shows superior ability to the CFD; MM
needs less than a minute for one free running simulation using a personal computer
while CFD needs a few weeks or a month using a very expensive supercomputer.
6 Improved Maneuvering-Based Mathematical Model … 95

Meanwhile, it is also very expensive and time-consuming to predict maneuvering

and rudder coefficients from captive model tests which are necessary for MM method
while CFD just needs ship geometry and propeller characteristics.
The MM free running simulations in calm water showed that MM was very sen-
sitive to the accuracy of the maneuvering coefficients such that the scatter in the MM
predictions were substantial for MM with coefficients estimated from different cap-
tive tests (Stern et al. 2011). Also, MM showed only qualitative agreement with model
experiment (ME) free running results in following and quartering waves while CFD
shows quantitative agreement (Sadat-Hosseini et al. 2011). Since CFD free running
simulation can provide not only ship motion but also total forces/moments acting
on the ship which are unknown during ME free running, it could give a chance to
modify and tune MM to reduce the disagreement with ME free running results in
calm water and waves.
Several mathematical models have been developed. Abkowitz (1964) developed a
mathematical model to describe the hydrodynamic forces/moments acting on the ship
with polynomial expressions using Taylor expansion on state variables. Christensen
and Blanke (1986) developed 2nd order modulus expansions, which represent the
hydrodynamic forces at angle of incidence: cross-flow drag. Recently a new maneu-
vering model was developed from first principles of low aspect-ratio aerodynamic
theory and Lagrangian mechanics (Ross et al. 2007). Meanwhile, the Maneuvering
Mathematical Modeling Group (MMG) (Ogawa and Kasai 1978; MMG 1980) devel-
oped a mathematical model, which explicitly includes the individual open water char-
acteristics of the hull/propeller/rudder and their interactions. Issues for improvement
include both wave terms in the mathematical model and methods for obtaining the
wave maneuvering coefficients. Usually, the maneuvering coefficients are assumed
constant, which could be realistic for high encounter frequency wave conditions. In
contrast, Son and Nomoto (1982) and Araki et al. (2010) showed a large variation
of maneuvering coefficients in following waves in which the encounter frequency is
very low.
System identification (SI) techniques are developed in control engineering to build
mathematical models for dynamical systems by estimating maneuvering coefficients.
The least square (LS) is the one of the simplest and the extended Kalman filtering
(EKF) (Lewis 1986) is one of the most widely used methods in engineering. Nonaka
et al. (1972) employed LS to estimate maneuvering coefficients from experimen-
tal free-running data with random rudder motions and the Abkowitz mathematical
model. However, the estimated maneuvering coefficients were not accurate, which
was due to some of the derivatives drifting to the wrong values known as the simulta-
neous drift problem (Kang et al. 1984). The constrained least square (CLS) method
using the generalized reduced gradient algorithm developed by Lasdon et al. (1978)
can help avoiding the simultaneous drift problem. EKF using full-scale trial data and
the Abkowitz mathematical model was employed by Abkowitz (1980). The zigzag
and turning simulation using estimated maneuvering coefficients showed reasonable
agreement with the data. Rhee and Kim (1999) employed EKF for free-running trial
data (zigzag, turning circle, large angle zigzag tests, etc.) and the MMG mathematical
model to find the best trial type for system identification. The maneuvering coeffi-
96 M. Araki et al.

cients reconstructed from the large angle zigzag test showed the smallest error with
the original coefficients. Zhang and Zou (2011) employed support vector machine,
one of the artificial intelligence methods, for zigzag test and the Abkowitz mathemat-
ical model for which the reconstructed coefficients showed close agreement with the
original maneuvering coefficients. Several other researchers (for instance Shi et al.
2009) have employed EKF to estimate ship maneuvering coefficients.
Most of the studies were conducted for calm water and experimental data were
used to improve the mathematical model by utilizing a system identification tech-
nique. The authors used CFD outputs to improve the mathematical model predictions
in calm water (Araki et al. 2012). Hydrodynamic and rudder maneuvering coeffi-
cients included in MM were estimated from turning circle and zigzag CFD free
running simulations trial data. The MM simulations using the predicted coefficients
showed much better agreement with ME free running than those using coefficients
estimated from captive model experiments and empirical prediction.
For maneuvering prediction in waves, the mathematical models often use the
hydrodynamic maneuvering coefficients estimated from the experimental captive
test in calm water. Also, the wave forces are considered as the summation of Froude-
Krylov and diffraction forces. These result in differences between MM predictions
and experimental free running data since the maneuvering coefficients variations due
to waves and wave drift forces are important for MM prediction in waves (Son and
Hamamoto 1982).
The objective of the present work is to employ the system identification technique
with CFD outputs to improve MM predictions in following and quartering waves by
tuning the maneuvering coefficients and wave forces. Since the ultimate goal of this
study is to predict broaching with MM modified with SI, the 4DOF MM (Umeda
et al. 2008) which is popular for broaching prediction model is applied as a basic
MM for this study. The CLS system identification technique was used to estimate
the manoeuvring coefficients and the wave forces using CFD outputs. Herein, only
moderate wave conditions are considered which do not cause broaching. The wave
forces/effects are found from CFD simulations. First CFD free running simulations
in waves are executed. Second, CFD forced motion simulations in calm water are
performed with imposing exactly same motions as the free running simulation. The
wave forces/effects are estimated as the difference between the total force of the
first and second simulations. The CFD wave forces/effects are compared with the
conventional MM wave model based on slender body theory and used to tune MM
wave forces/effects by the system identification technique. The improved MM and
CFD free running simulation results are compared with that of ME. Here it should
be pointed out that the MM and CFD simulations are done before the ME data are
available.
6 Improved Maneuvering-Based Mathematical Model … 97

Table 6.1 Principal particulars of the ONRTH

Model scale
Length (L) 3.147 m
Breadth (B) 0.384 m
Depth (D) 0.266 m
Draft (d) 0.112 m
Displacement (W) 72.6 kg
Metacentric height (GM) 0.0424 m
Natural roll period (Tφ) 1.644 s
Rudder area (AR) 0.012 m2 × 2
Block coefficient (Cb) 0.535
Vertical position of CoG from waterline −0.392 × d
(downward positive) (OG)
Radius of gyration in pitch (κyy) 0.25 × L
Maximum rudder angle (δmax ) ±35°

Fig. 6.1 Body plan of the

ONRTH model

WL

6.2 ME, CFD, and MM Methods

6.2.1 Subject Ship

The 1/49 scaled model of ONR tumblehome (ONRTH), was developed at Naval
Surface Warfare Center Carderock Division (Bishop et al. 2005), appended with
skeg, bilge keels, rudders, shafts with propeller shaft brackets and twin propellers
was used for the free running experiments. The main particulars of the ONRTH ship
are listed in Table 6.1. The details of the body plan and the model are shown in
Figs. 6.1 and 6.2.
98 M. Araki et al.

Fig. 6.2 Bow and stern of the ONRTH model: a bow; b stern

6.2.2 ME Method

All ME free running data was acquired in IIHR wave basin. The wave basin has
dimensions of 40 × 20 m2 with 3 m water depth and is designed to test captive or
radio-controlled model scale ships.
The model launch system enables specification and replication of the free running
trial initial conditions. Roll, pitch, and yaw angles of the model ship were measured by
a fiber optical gyroscope. Meanwhile, the plane trajectory of the model was recorded
by the tracking system, which uses two-camera vision. The tracking cameras capture
two LED lights placed on the deck of the model. In order to increase the reliability and
accuracy of the 5DOF (Degree of Freedom) measurement and to enable measurement
for all 6DOF of the free running model, i.e. the heave motion, a 6DOF visual motion
capture system was added to the tracking system. A detailed description of the wave
basin and wavemakers, carriage model tracking, 6DOF visual motion capture and
free running 6DOF systems, model geometry and ballasting, and free running trials
tests in calm water and waves is provided by Sanada et al. (2012).
The experimental procedure was as follows. First, the model ship was fixed on
the launch system by electromagnetics while heave, roll, and pitch are free. After the
propeller starts to rotate, the model was accelerated by the launch system to reach
the target speed. Since the towing system acts as the hard spring there would be
small oscillations for the surge motion of the towed ship. After the ship was at the
target speed the model was towed for more distance until the bow was located on
the wave crest. Then the model was released and the rudder controller was activated
after few seconds to start maneuvering. The propeller rate was kept constant during
free running. The ME and CFD trial conditions are shown in Table 6.2.
6 Improved Maneuvering-Based Mathematical Model … 99

Table 6.2 ME and CFD free running test matrice

Test Nominal δ (deg) ψ *c (deg) H/λ λ/Lpp
Fr
ME Calm Turning 0.1, 0.2 25, 35 NA
water circle
Zigzag 0.1, 0.2 10, 20, 35 10, 20,
35
Large 0.2 35 90
angle
zigzag
Wave Zigzag 0.1, 0.2 20 20 0.02, 1.0
0.03
Straight 0.1, 0.2 NA NA 0.02, 1.0
running 0.03
Course 0.1, 0.2 Proportional 20 0.02, 1.0
keeping control, P  1 0.03
CFD Calm Turning 0.2 25 NA
water circle
Zigzag 0.2 20 20
Large 0.2 35 90
angle
zigzag
Wave Zigzag 0.2 20 20 0.02, 1.0
0.03
Straight 0.2 NA NA 0.02, 1.0
running 0.03
Course 0.2 Proportional 20 0.02, 1.0
keeping control, P  1 0.03
ψ *c Target yaw angle

6.2.3 CFD Method

The code CFDShip-Iowa v4 (Carrica et al. 2010) is used for the CFD computa-
tions. The CFDShip-Iowa is an overset, block structured CFD solver designed for
ship applications using either absolute or relative inertial non-orthogonal curvilinear
coordinate system for arbitrary moving but non-deforming control volumes. Tur-
bulence models include blended k-ε/k-ω based isotropic and anisotropic Raynolds
Averaged Navier Stokes (RANS), and (detached eddy simulations) DES approaches
with near-wall or wall functions. The discretized geometries of the twin propellers
were not included in the simulations. Instead, a simplified body force model is used
for the propeller which prescribes axisymmetric body force with axial and tangential
components.
The propeller model requires the experimental open water curves and advance
coefficients as input and provides the torque and thrust forces. The open water curves
100 M. Araki et al.

Fig. 6.3 CFD overset grids

for ONRTH hull and
appendages

are defined as a second order polynomial fit of the experimental K T (J) and K Q (J)
curves. The advance coefficient is computed using ship speed with neglecting the
wake effects. Herein, two PID controllers are used. The heading controller acting on
the rudders are responsible to turn the rudders to keep the ship in the desired direction.
The speed controller acting on the body force propeller model is responsible to rotate
the propellers at appropriate propeller rate to keep the ship at the desired speed. The
heading controller uses P  1 for the proportional gain and zero for both the integral
and derivative gains mimicking the experimental setup which uses a proportional
heading control.
The CFD initial condition is different with ME in several ways. The CFD model
was accelerated with infinite rate to the target speed unlike ME. Then the model
was towed at target speed which was constant while the model was only free to
heave and pitch and not roll until the wave trough was located at midship. After
that, the model was released and rudder controller was activated immediately to
start maneuvering. The differences between ME and CFD setup might cause some
discrepancies between ME and CFD results.
The free model is appended with skeg, bilge keels, superstructure, rudders, rudder
roots, shafts, and propeller brackets same as the ME model but not appended with
actual propellers. The computational grids are overset with independent grids for the
hull, superstructure, appendages, refinement, and background, and then assembled
together to generate the total grid. The total number of grid points is 12.1 M for free
model simulations. Details of the grids are shown in Table 6.3 and Fig. 6.3. The free
running in waves and calm water verification studies have been done (Sadat-Hosseini
et al. 2011; Araki et al. 2012) which showed quantitative agreement with ME results.

6.2.4 MM Method

4DOF maneuvering mathematical model was used for the MM simulations as shown
in Eqs. (6.1)–(6.5). The low encounter frequency model (Umeda et al. 2008) is 4DOF
surge-sway-roll-yaw model and is modified especially for surge equation and rudder
6 Improved Maneuvering-Based Mathematical Model … 101

Table 6.3 Grids for free model simulations

Name Size (grid points) # of procs Type
Hull S/Pa 199 × 61 × 104 12 (×2) Double O
(1.26 M ×2)
Skeg S/P 61 × 49 × 40 1 (×2) O
(0.12 M ×2)
Bilge keel S/P 99 × 45 × 50 2 (×2) H
(0.23 M ×2)
Rudder root collar S/P 121 × 35 × 28 1 (×2) O
(0.12 M ×2)
Rudder root gap S/P 121 × 51 × 19 2 (×2) Conformal to collar
(0.12 M ×2)
Rudder outer S/P 61 × 36 × 55 1 (×2) Double O
(0.12 M ×2)
Rudder inner S/P 61 × 36 × 55 1 (×2) Double O
(0.12 M ×2)
Rudder gap S/P 121 × 51 × 19 2 (×2) Conformal to inner
(0.12 M ×2) and outer
Shaft collar S/P 39 × 50 × 57 1 (×2) O
(0.11 M ×2)
Shaft proper S/P 74 × 41 × 37 1 (×2) O
(0.11 M ×2)
Shaft tip S/P 110 × 117 × 100 12 (×2) O with end pole
(1.29 M ×2)
Strut outer S/P 69 × 34 × 50 1 (×2) O
(0.12 M ×2)
Strut inner S/P 69 × 34 × 50 1 (×2) O
(0.12 M ×2)
Superstructure 165 × 61 × 85 8 Wrap
(0.86 M)
Refinement 145 × 81 × 113 12 Cartesian
(1.33 M)
Background 213 × 84 × 113 20 O
(2.02 M)
Total (12.1 M) 116
a S/P Starboard/Port

model (Araki et al. 2012). The model is developed in horizontal body axes which
are shown in Fig. 6.4.

(m + m x )u̇ − (m + m y )vr  T (u; n) − R(u; n) + X vv (u)v 2 + X vr (u)vr

+ X rr (u)r 2 + X R (δ, u, v, r ) + X W (6.1)
102 M. Araki et al.

Fig. 6.4 Coordinate system for 4DOF MM

(m + m y )v̇ + (m + m x )ur  Yv (u)v + Yr (u)r + Yφ (u)φ

+ Yvvv (u)v 3 + Yvvr (u)v 2 r + Yvrr (u)r 2 v
+ Yrrr (u)r 3 + Y R (δ, u, v, r ) + YW (6.2)
(Ix x + Jx x ) ṗ  m x z H ur + K v (u)v + K r (u)r + K p (u) p + K φ (u)φ
− mgG Z (φ) + K vvv (u)v 3 + K vvr (u)v 2 r + K vrr (u)r 2 v
+ K rrr (u)r 3 + K R (δ, u, v, r ) + K W (6.3)
(Izz + Jzz )ṙ  Nv (u)v + Nr (u)r + Nφ (u)φ + Nvvv (u)v + Nvvr (u)v r
3 2

+ Nvrr (u)r 2 v + Nrrr (u)r 3 + N R (δ, φ, u, v, r ) + N W (6.4)

Here

[K v K r K vvv K vvr K vrr K rrr ]T  z H [Yv Yr Yvvv Yvvr Yvrr Yrrr ]T (6.5)

In the mathematical model, resistance is estimated from a captive model exper-

iment and the thrust is estimated from propeller open water tests in calm water as
described in Umeda et al. (2008). Roll restoring moment (mgGZ) is estimated from
hydrostatic calculations in calm water. Maneuvering coefficients including heel-
induced hydrodynamic derivatives are estimated from calm water captive model
experiments (Hashimoto et al. 2008). Roll damping is estimated from roll decay
model tests (Umeda et al. 2008). For the ONRTH, the values of correction coef-
ficients for rudder are empirically developed from other model experiments (Kose
et al. 1981). The empirical values are also used for the interaction force coefficients
induced on the hull by rudder nominal force. All maneuvering and rudder coeffi-
cients are listed in Table 6.4. In wave cases, the wave forces X w , Y w , K w , and N w are
calculated from linear Froude-Krylov forces and diffraction forces based on slender
ship theory for zero encounter frequency. It is known that the wave particle velocity
affects the rudder and propeller inflow speed which is important for the maneuver-
ability in waves. Therefore the effects of wave particle velocity to propeller thrust
and rudder force are taken into account.
6 Improved Maneuvering-Based Mathematical Model … 103

Table 6.4 Values of original and SI-calm maneuvering and rudder coefficients used in 4-DOF
nonlinear MM
Coef. Original SI-calm Coef. Original SI-calm
ε 1.0 0.75 Y vrr −0.80 0.32
γR 0.70 0.55 Y rrr 0.174 0.080
l R /L −1.00 −0.95 Yφ −5.1E−04 −6.5E−04
tR 0.30 0.10 J xx 4.1E−05 0.0001
aH 0.25 0.23 zH 0.852 1.08
zHR /d 0.854 0.802 Kp −0.243 −0.203
x H /L −0.45 −0.52 Kφ 6.3E−04 1.0E−03
mx 0.0131 0.0 J zz 0.0079 0.0059
X vv −0.0858 −0.070 Nv −0.0932 −0.0851
X vr 0.0522 0.065 Nr −0.0549 −0.0395
X rr −0.0213 −0.025 N vvv −0.532 −0.492
my 0.109 −0.070 N vvr −0.629 −0.805
Yv −0.30 −0.20 N vrr −0.139 −0.121
Yr −0.0832 0.07 N rrr −0.00446 −0.0065
Y vvv −1.77 −2.0 Nφ −0.00511 −0.00989
Y vvr 0.262 0.32

However, the MM simulations using these coefficients and wave model show
some discrepancy with the ME free running in calm water and waves. The authors
tuned the calm water maneuvering and rudder coefficients values using the SI with
CFD free running results in calm water. The results showed better agreement with the
ME calm water data (Araki et al. 2012). The values of tuned coefficients are shown
in Table 6.4. Herein, the wave model is tuned using the SI technique with CFD free
running and captive results in waves.

6.3 SI Method and the Results

A constrained least square (CLS) method using generalized reduced gradient algo-
rithm (Lasdon et al. 1978) is used for SI. The CLS method provided reasonable
maneuvering coefficients from CFD calm water maneuvers (Araki et al. 2012). To
predict wave forces/effects, it is necessary to extract the wave forces/effects from
total hydrodynamic forces. To achieve this purpose, first 6DOF CFD free running
simulations in waves are executed. Second, CFD forced motion simulations in calm
water are performed with imposing exactly same motions as the free running simu-
lation. Thus the wave forces/effects are estimated as the difference between the total
force of the first and second simulations.
104 M. Araki et al.

Fig. 6.5 Wave forces during

ψC  20° course keeping towed free
maneuver in following waves
with nominal Fr  0.20, wave
steepness 1/50 and wave
length to ship length ratio 1.0

Figure 6.5 shows the extracted CFD wave forces/effects during ψC  20° course
keeping maneuver in quartering waves with nominal Fr  0.20, wave steepness 1/50
and wave length to ship length ratio 1.0. During the free running, the model is
imposed with constant forward speed Fr  0.20 and constant yaw angle 20° until
8.02 s then released to start course keeping maneuvers. Here the “MM” is the wave
forces computed by MM wave model.
It is shown that MM overestimates surge wave force and underestimates sway,
roll, and yaw wave forces. Although definitions of wave forces are different, these
tendencies are found in previous research by Hashimoto et al. (2011) as well where
they compared MM and captive ME wave exciting forces for ONRTH. Moreover the
MM waveforms seem to be different from CFD waveforms especially after the model
is released. This could because that MM wave model merely includes the Froude-
Krylov and diffraction forces while CFD wave force includes all wave effects not
just Froude-Krylov and diffraction forces but also wave drift forces and maneuvering
6 Improved Maneuvering-Based Mathematical Model … 105

coefficients variations due to waves. According to CFD extracted wave forces/effects,

the wave correction parameters are estimated by SI methods to tune MM wave model
based on Eqs. (6.6)–(6.8).

X W  a1 X WFK
+ kζw a2 sin(kξG + εa2 ) · X vv · v 2

+a3 sin(kξG + εa3 ) · X vr · vr + a4 sin(kξG + εa4 ) · X rr · r 2 + WC X
Di f
YW  b1 YWF K + b2 YW + kζw (b3 sin(kξG + εb3 ) · Yv · v
+b4 sin(kξG + εb4 ) · Yr · r ) + WCY
Di f
K W  c1 K WF K + c2 K W + kζw (c3 sin(kξG + εc3 ) · K v · v + c4 sin(kξG + εc4 ) · K r · r )
Di f
NW  d1 N WF K + d2 N W + kζw (d3 sin(kξG + εd3 ) · Nv · v
+d4 sin(kξG + εd4 ) · Nr · r ) + WC N (6.6)

where
 
WC X  ρgζw2 B 2 /L · sin(ψ/2 + ψ0 ) · 1 − e−10Fr · C X (Tw )
WCY  ρgζw2 B 2 /L · sin ψ · CY (Tw )
WC N  ρgζw2 B 2 · sin ψ · C N (Tw ) (6.7)
      
mW D Tw m−1 Tw m
C X,Y,N (Tw )  α X,Y,N · exp − (6.8)
ηW D ηW D ηW D
u w  β1 · ζw ω cos ψe−kz cos(kξG + kx cos ψ)
vw  −β2 · ζw ω sin ψe−kz cos(kξG + kx cos ψ) (6.9)

The new wave model includes the major maneuvering coefficients variations and
wave drift forces. For the simplification the wave drift coefficients shown in Eq. (6.8)
are expressed as the Weibull distribution respect to wave periods. The shape and
scale parameter mWD , ηWD of Eq. (6.8) are determined from the Yasukawa’s research
(2006). Also the normal force of CFD rudder is used to tune the wave particle velocity
effects to propeller and rudder as shown in Eq. (6.9).
Here a, b, c, d, ε in Eq. (6.6), ψ 0 in Eq. (6.7), α in Eq. (6.8), and β in Eq. (6.9)
are tuned by the SI. These SI procedures are repeated to the other cases: straight
running and 20/20 zigzag with nominal Fr  0.20, wave steepness 1/50 and wave
length to ship length ratio 1.0. The original wave correction coefficients and the
tuned coefficients are shown in Table 6.5.
106 M. Araki et al.

Table 6.5 Original wave Coef. Orig SI-wave Coef. Orig SI-wave
correction coefficients and the
coefficients estimated by SI a1 1.0 0.758 c1 1.0 1.26
using extracted CFD wave a2 0.0 16.33 c2 1.0 0.355
forces/effects data a3 0.0 0.855 c3 0.0 0.510
a4 0.0 0.132 c4 0.0 0.195
εa2 0.0 0.391 εc3 0.0 −0.99
εa3 0.0 3.21 εc4 0.0 1.03
εa4 0.0 0.0 d1 1.0 1.43
αX 0.0 −46.3 d2 1.0 0.403
b1 1.0 1.35 d3 0.0 1.01
b2 1.0 0.423 d4 0.0 0.213
b3 0.0 2.18 εd3 0.0 0.982
b4 0.0 0.496 εd4 0.0 −0.99
εb3 0.0 −0.552 αN 0.0 −2.50
εb4 0.0 0.810 ψ0 0.0 0.301
αY 0.0 −106.2 β1 1.0 0.643
β2 1.0 0.425

6.4 Comparison Between EFD, CFD, and MM Free

Runnings

6.4.1 Maneuvering in Calm Water

Comparison between ME, CFD, “MM-Orig.”, and “MM-SI-calm” trajectories are

shown in Fig. 6.6. Here “MM-Orig.” indicates the MM simulation using the maneu-
vering coefficients estimated from captive model test and rudder coefficients pre-
dicted from empirical charts (Kose et al. 1981). “MM-SI-calm” indicates MM sim-
ulation using maneuvering and rudder coefficients estimated by SI using CFD calm
water maneuvering data (Araki et al. 2012). Figure 6.6a shows the trajectories of
turning circle tests (δ  25°, Fr  0.20), Fig. 6.6b shows zigzag tests (ψc /δ  20/20,
Fr  0.20) trajectories, and Fig. 6.6c shows large angle zigzag tests (ψc /δ  90/35,
Fr  0.20).
The steady state variables and turning parameters for turning circle and 1st, 2nd
overshoot angles and K s , T s , N s steering quality indices for zigzag cases (ABS 2006)
for both CFD and MM predictions are compared against ME ones and the overall
errors are plotted in Fig. 6.7. The steering quality indices are computed from nonlinear
first order Nomoto’s model (Norrbin 1963) shown in Eq. (6.10). As shown in Figs. 6.6
and 6.7, the errors of MM-SI-calm are much smaller than that of MM-Orig.

Ts ṙ + Ns r 3 + r  K s δ (6.10)
6 Improved Maneuvering-Based Mathematical Model … 107

Fig. 6.6 Trajectories of ME,

CFD, MM-Orig., and
MM-SI free running in calm
water: a δ  25° turning
circle; b 20/20 zigzag; c
90/35 large angle zigzag
108 M. Araki et al.

Fig. 6.7 Errors from ME 35

free running results: “T25”
30
δ  25° turning circle; “Z20”
20/20 zigzag; “Z90” 90/35 25
zigzag; “Global Av.” average
20 CFD

E%D
of T25, Z20, and Z90 errors
15 MM-SI-calm
MM-Orig.
10

0
T25 Z20 Z90 Global Av.

6.4.2 Straight Running in Following Waves

Figure 6.8 shows the comparison between CFD, MM-SI-calm and MM-SI-wave
straight running in following waves with nominal Fr  0.20, wave steepness 1/50 and
wave length to ship length ratio 1.0. Here “MM-SI-calm” indicates MM simulation
using maneuvering and rudder coefficients estimated by SI from CFD calm water
maneuvering data (Araki et al. 2012) with original wave model. “MM-SI-wave”
indicates MM simulations using same maneuvering and rudder coefficients with
“MM-SI-calm” but with new wave model shown in Eq. (6.6) which is the improved
wave forces using CFD wave forces/effects data.
In Fig. 6.8, CFD shows remarkable agreement with ME especially for the surge
and pitch motions. It is showing possibility to replace ME free running test with CFD
simulations even in wave conditions. Here heave and pitch motions for the 4DOF
(surge-sway-yaw-roll) MM are assumed to be the same as the static equilibrium
positions of the ship in waves. MM shows larger heave and pitch motion than those
of CFD and ME which indicates that 6DOF model could be desirable. In surge
motion, CFD successfully reproduce the nominal speed loss due to waves. The MM-
SI-calm fails to express the nominal speed loss and the surging amplitudes are larger
than that of CFD and ME. The MM-SI-calm cannot represent nominal speed loss
because the wave drift terms are not included in the MM-SI-calm model. Meanwhile
MM-SI-wave successfully predicts nominal speed loss and surging amplitude within
high degree of accuracy.

6.4.3 Course Keeping and Zigzag in Following

and Quartering Waves

Course keeping and zigzag simulations in wave conditions are shown in this section.
The simulation procedure is as follows. First the model is accelerated to the target
ship speed with 2DOF (heave and pitch). After the model reaches to the speed, the
6 Improved Maneuvering-Based Mathematical Model … 109

Fr=0.20

θ [deg]

Fig. 6.8 Straight running in following waves with nominal Fr  0.20, wave steepness 1/50, and
wave length to ship length ratio 1.0

model is towed with constant speed for a while and released when the bow is located
on the wave crest. The rudder control starts just after the model is released. In the
ME, it should be noted that the towing time was very short because of the limitation
of the facility’s size. Moreover it was 3DOF (heave, pitch and roll) during towing in
ME.
Figure 6.9 shows the comparison between CFD and MM-SI-calm ψC  20° course
keeping in quartering waves with nominal Fr  0.20, wave steepness 1/50 and wave
length to ship length ratio 1.0. Here the ME and CFD rudder control start just after
the model is released at a wave trough.
In the trajectory, CFD course deviation shows good agreement with ME which
indicates that CFD well predicts the wave drift force. Although the ME shows wobbly
trajectory compared to CFD due to large oscillations for sway motions. Due to the
sway motion error, the roll motions show some difference between ME and CFD
while the error is much smaller than that of sway motion. However CFD successfully
predicts the surge and yaw motions in quartering waves. Paying attention to CFD and
MM results, MM-SI-calm shows small course deviation compared to that of CFD.
110 M. Araki et al.

Fig. 6.9 ψC  20° course

keeping maneuver in
quartering waves with
nominal Fr  0.20, wave
steepness 1/50, and wave
length to ship length ratio 1.0

ψ [deg]
δ [deg]
6 Improved Maneuvering-Based Mathematical Model … 111

From the state variables comparisons, it is clear that MM-SI-calm has some dis-
crepancy on the wave forces and wave drift effects compared to CFD and ME.
MM-SI-calm’s wave model overestimates the surge wave force and underestimates
the sway, roll, and yaw wave forces. The MM-SI-wave shows better agreement with
CFD than MM-SI-calm for state variables and the trajectory. The wave drift effects
can be seen in sway motion prediction which improves the prediction of the course
deviation.
Figure 6.10 shows the comparison between ME, CFD and MM 20/20 zigzag in
following and quartering waves with nominal Fr  0.20, wave steepness 1/50 and
wave length to ship length ratio 1.0. The CFD results show good agreement with
shifted ME for trajectory, surge, and yaw motions. In sway motion, CFD seems
underestimating the wave force compared to ME which could explain the discrepancy
of the roll motions. MM-SI-calm shows qualitative agreement with ME maneuver but
not quantitative. MM-SI-calm overestimates surge wave force and underestimates
sway, and yaw wave forces. The MM-SI-calm prediction of the zigzag trajectory is
very close to the one predicted in calm water shown in Fig. 6.6b. This is due to the
fact that the maneuvering coefficients oscillations and drift forces induced by waves
are neglected in MM-SI-calm. The MM-SI-wave improves the prediction as it shows
the oscillations on the state variables induced by the waves. Also, the speed loss is
predicted well in MM-SI-wave such that the trajectory shows good agreement with
CFD and ME ones.

6.5 Conclusions

System identification method using CFD free running data is shown to be an efficient
approach for estimating maneuvering, rudder, and wave correction coefficients in the
MM. Araki et al. (2012) show the reasonable maneuvering and rudder coefficients
can be obtained from a few CFD free running data in calm water. However, the MM
still shows some error predicting the ship motion in waves. The original MM includes
the Froude-Krylov and diffraction forces as the wave forces and the wave particle
velocity as the wave effect on the propeller and rudder which clearly fails to predict
the oscillation amplitudes and the wave drift effects. Therefore the MM wave model
is improved by adding correction parameters for Froude-Krylov, diffraction forces,
and wave particle velocity. Moreover, effects of wave drift forces and maneuver-
ing coefficient variations due to waves are taken into account and these correction
parameters are predicted by CLS using the extracted CFD wave forces/effects data.
The extracted CFD wave forces/effects data are generated from the CFD free running
data in waves and CFD forced motion data in calm water. The MM simulations using
the new wave model and estimated wave correction coefficients show much better
agreement with CFD than the MM simulations using the original wave model. The
CFD simulations are validated with ME free running results; CFD mostly shows
quantitative agreement with ME which shows the possibility of replacing ME free
running trials with CFD simulations.
112 M. Araki et al.

Fig. 6.10 20/20 zigzag in

quartering waves with
nominal Fr  0.20, wave
steepness 1/50, and wave
length to ship length ratio 1.0 wave

δ
6 Improved Maneuvering-Based Mathematical Model … 113

Several follow-up studies were conducted subsequent to this research to improve

the mathematical model. Araki et al. (2013) improved the mathematical models for
the wave forces, maneuvering coefficients variations, and rudder forces. However,
the wave model provided too stable motions in severe waves and is unable to pre-
dict the instabilities such as broaching. Yoneda et al. (2017) and Mizumoto et al.
(2018) conducted more studies to improve the wave model for severe condition. In
future, more experiments and simulations will be conducted in IIHR wave basin and
National Research Institute of Fisheries Engineering (Japan) to facilitate the mathe-
matical model improvement for conditions with higher ship speed and tougher waves
including irregular waves near to broaching conditions. The CFD and experimental
studies will be used to include not only the wave drift force but also the nonlinear
wave effects including the memory effect functions (Mikami and Kashiwagi 2009)
and the rudder and propeller exposing effects in the mathematical model.

Acknowledgements This research was sponsored by the Office of Naval Research Grant N000141-
21-05-6-8 and NICOP Grant N00014-09-1-1089 under the administration Dr. Patrick Purtell. The
CFD simulations were conducted utilizing DoD HPC. The authors are grateful to Mr. K. Tanimoto
and Ms. K. Takagi of Osaka University and Mr. A. Hanaoka of The University of Iowa, IIHR for
assistance conducting the experiments.

References

Abkowitz, M.A., 1964, Lecture of ship hydrodynamics—steering and maneuverability, Hydro-and

Aerodynamic Laboratory, Report No. Hy-5.
Abkowitz M. A., 1980, Measurement of Hydrodynamic Characteristics from Ship Maneuvering
Trials by System Identification, SNAME Transactions, 88, 283–318.
American Bureau of Shipping, 2006, Guide for Vessel Maneuverability, Houston, TX.
Araki, M., Umeda, N., Hashimoto, H., Matsuda, A., 2010. “Broaching prediction using an improved
system-based approach”. Proceedings of 28th Symposium on Naval Hydrodynamics, pp. 56–68.
Araki, M., Sadat-Hosseini, H., Sanada, Y. Tanimoto, K., Umeda, N., and Stern, F., 2012, Estimat-
ing Maneuvering Coefficients Using System identification Methods with Experimental, System-
based, and CFD Free-running Trial Data, Ocean Engineering, 51:63–84.
Araki, M., Sadat-Hosseini H., Sanada, Y., Umeda, N., and Stern, F., 2013, “System Identification
using CFD Captive and Free Running Tests in Severe Stern Waves,” Proceedings 13th Interna-
tional Ship Stability Workshop, Brest, France, September 23 – 26.
Bishop, R., Belknap, W., Turner, C., Simon, B., Kim, J., 2005, Parametric investigation on the
influence of GM, roll damping, and above-water form on the roll response of model 5613, Report
NSWCCD-50-TR-2005/027.
Carrica, P. M., Huang, J., Noack, R., Kaushik, D., Smith, B., Stern, F., 2010, “Large-scale DES com-
putations of the forward speed diffraction and pitch and heave problems for a surface combatant”,
Computer & Fluids, 39(7):1095–1111.
Christensen, A., Blanke, M., 1986. A Linearized state-space model in steering and roll of high-speed
container ship. Technical Report 86-D-574, Servolaboratoriet, Technical University of Denmark.
Hashimoto, H., Umeda, N., and Matsuda, A., 2011, “Broaching prediction of a wave-piercing
tumblehome vessel with twin screw and twin rudder”, Journal of Marine Science and Technology,
16(4):448–461.
114 M. Araki et al.

Hashimoto, H., Stern, F., and Sadat-Hosseini, H., 2008, An application of CFD for advanced broach-
ing prediction (2ndReport), Conference Proceedings of the Japan Society of Naval Architects and
Ocean Engineers, Vol. 6, pp. 237–240.
Kang, C.G., Seo, S.H., Kim, J.S., 1984. “Maneuverability analysis of ship by system identification
technique”. SNAK, 21, No.4.
Kose, K., Yumuro, A., Yoshimura, Y., 1981, “Concrete of mathematical model for ship maneuver-
ability”, Proceedings of 3rd Symposium on Ship Maneuverability, Society of Naval Architects of
Japan, pp. 27–80, (in Japanese).
Lasdon, L. S., Waren, A. D., Jain, A., Ratner, M., 1978, Design and testing of a generalized reduced
gradient code for nonlinear programming, Journal of ACM Transactions on Mathematical Soft-
ware, Volume 4 Issue 1.
Lewis, F. L., 1986, Optimal Estimation –with an introduction to stochastic control theory, John
Wiley & Sons, Inc.
Mikami, T., Kashiwagi, M., 2009, “Time-domain strip method with memory-effect function con-
sidering body nonlinear wave-body interactions (2nd report)”, Journal of Marin Science and
Technology, 14(2) 185–199.
Mizumoto, K., Stern, F., Araki, M., Umeda, N., 2018, “CFD-based system identification for improv-
ing a system-based simulation model of broaching in stern quartering waves (tentative title)”,
Proceeding of 13th International Conference on Stability of Ships and Ocean Vehicles STAB
2018, Kobe, Japan, 92–98.
MMG, 1980. MMG report V. Bulletin of Society of Naval Architects of Japan. 616, 565–576.
Nonaka, K., Mori, M., Matsumoto, N., 1972. “Estimating Maneuvering Coefficients from Free-
Running Trials”. Proceedings of 20th Meetings of Ship Research Institute.
Norrbin, N. H., 1963, “On the design and analysis of zig-zag test on base of quasi linear fre-
quency response”, Technical Report B104-3, The Swedish State Shipbuilding Experimental Tank,
Gothenburg, Sweden.
Ogawa, A., Kasai, H., 1978. “On the mathematical model of maneuvering motion of ship”. Inter-
national Shipbuilding Progress 25 (292), 306–319.
Rhee, K. P., Kim, K., 1999, A new sea trial method for estimating hydrodynamic derivatives, Journal
of Ship & Ocean Technology 3, (3), 25–44.
Ross, A.,.T., Perez, and T., Fossen, 2007. A novel maneuvering model based on lowaspect-ratio lift
theory and Lagrangian mechanics. Proceedings of the IFAC Conference on Control Applications
in Marine System (CAMS).
Sadat-Hosseini, H., Carrica, M. P., Stern, F., Umeda, N., Hashimoto, H., Yamamura, S., Mastuda,
A., 2011, “CFD, system-based and ME study of ship dynamic instability events: surf-riding,
periodic motion, and broaching”, Ocean Engineering, Vol. 38, Issue 1, pp. 88–110.
Sanada, Y., Tanimoto, K., Takagi, K., Sano, M., Yeo, D.J., Toda, Y., Stern, F., 2012, “Trajectories
of local flow field measurement around ONR tumblehome in maneuvering motion”, Proceedings
of 29th Symposium on Naval Hydrodynamics, Gothenburg, Sweden.
Shi, C., Zhao, D., Peng, J., Shen, C., 2009. “Identification of ship maneuvering model using extended
kalman filtering”. International Journal Marine Navigation and Safety of Sea Transportation 3
(1), 105–110.
Son K. H., Nomoto K., 1982, Combined behavior of manoeuvring and roll motion in following
wave, Journal of the Society of Naval Architects of Japan, Vol.152, pp. 207–218.
Stern, F., Agdrup, K., Kim, S. Y., Hochbaum, A. C., Rhee, K. P., Quadvlieg, F., Perdon, P., Hino,
T., Broglia, R., and Gorski, J., 2011, “Experience from 2008 – the first workshop on verification
and validation of ship maneuvering simulation methods,” Journal of Ship Research, vol. 55, No.
2, pp. 135–147.
Umeda, N., Matsuda, A., Hashimoto, H., Yamamura, S., and Maki, A., 2008, “Model experiments
on extreme motions of a wave-piercing Tumblehome vessel in following and quartering waves”,
Journal of the Japan Society of Naval Architects and Ocean Engineers, Vol. 8, pp. 123–129.
Yasukawa, H., 2006, Simulation of Ship Maneuvering in Waves (1st report: turning motion), Journal
of the Japan Society of Naval Architects and Ocean Engineers, Vol. 4, pp. 127–136, (in Japanese).
6 Improved Maneuvering-Based Mathematical Model … 115

Yoneda, S., Hashimoto, H., Matsuda, A., Tahara, Y., Terada, D., Stern, F., 2017, “Investigation on the
improvement of estimation accuracy of wave-exciting forces acting on ships in stern quartering
waves”, Conference Proceedings of the Japan Society of Naval Architects and Ocean Engineers,
Vol. 24. (in Japanese).
Zhang, X.G., Zou, Z.J., 2011. “Identification of Abkowitz model for ship maneuvering motion using
epsilon-support vector regression”, Journal of Hydrodynamics, Ser. B 23 (3), 353–360.