SNAME Transactions, Vol. 101, 1993, pp. 609-635

A Review and Comparison of Ship Maneuvering Simulation Methods Roderick A. Barr, Member, Hydronautics Research Inc., Fulton, Maryland

like current methods, used hydrodynamic coefficients based on test data from captive model tests conducted in a towing tank and/or on a rotating arm.

ABSTRACT A brief review of ship maneuvering and shiphandling simulation methods and models is given and the attributes of two representative models are illustrated. A critical review of 14 published simulation models applied to simulation of ESSO OSAKA maneuvering is presented. This review includes a comparison of damping and added mass coefficients for both deep and shallow water and a comparison of total sway and yaw damping forces for sway and yaw rates (velocities) representative of those occurring during typical defmitive maneuvers. It is concluded that while there is a reasonable state of agreement between many of the deep and shallow water simulation models, some models appear suspect and there may exist significant scale effects for the smaller model lengths employed in both captive and free-running model tests. 1. INTRODUCTION The development of high-speed computers has revolutionized both the methods used to predict the behavior of complex systems and phenomena and our resulting understanding of such systems and phenomena. One important predictional technique made feasible by high-speed computers is time-domain simulation, in which behavior of a system is determined by numerical integration of a set of equations of motions. This numerical solution provides a description of the time history of system behavior in response to one or more time-dependent disturbances or forcing functions. In the marine field time-domain simulation has found its widest application in the prediction of controllability and maneuverability of all types of craft and marine systems. Computer-based, time-domain simulation, which is now the most widely used method for evaluating controllability and maneuvering performance of conventional surface ships, submarines and other marine craft, first came into use in the late 1950's. Initial ship simulation methods reflected existing time-domain simulation techniques for submarines and aircraft, many of which used analog computers. These initial methods,

APPLICATIONS OF MANEUVERING AND SHIPHANDLING SIMULATION The increasing importance of simulation is reflected in the recent and current ship maneuvering simulation projects of the Marine Board of the National Research Council. Simulation is now widely used and accepted as a tool for: .

2. 3. .

Ship design and research; Ship equipment design and selection Design of and research on waterways and harbors. Training of ship operating personnel or pilots;

Simulation is also used, albeit sometimes with lower confidence, to predict or confirm nmneuvering performance during ship design. The first three of these applications are typically of greatest interest to most naval architects. The first includes evaluation of the ability of a particular ship design in the areas of: °

2. 3.

Routine and emergency maneuvering; Coursekeeping, controllabilityand stationkeeping Behavior under emergency conditions Such as loss of power or rudder control.

The second application may include selection and sizing of special maneuvering devices, such as thrusters or rotating propellers, required for special operations such as low speed coursekeeping or stationkeeping. The first of these applications is of particular current relevance in view of ongoing activities of the International Maritime Organization (IMO) in the area of ship maneuvering performance standards. The IMO, after almost 20 years of deliberations, appears ready to adopt a set of maneuvering performance standards. If these standards are adopted by the IMO, countriessuch as the U.S.A. will have a clear mandate to prevent both the

A Review and Comparison of Ship Maneuvering Simulation Methods

609

flagging of U.S. ship and the entry into all U.S. ports of any ships that do not meet these standards. If such standards abe adopted, maneuvering and controllability may cease to be the most neglected aspect of ship design. 3. M E T H O D S USED T O PREDICT SHIP MANEUVERING AND SHIPHANDLING

contribution to these coefficients of the bare hull, the appendages (skeg, rudder, etc.) and the propulsor(s). These methods typically include also a corresponding set of equations of motion required to simulate maneuvering performance. Methods which are readily available and which are of particular interest include: 1.

Before the availability of high-speed computers, ship maneuverability was assessed using either freerunning models or empirical techniques for predicting steady turning performance, as described in the First Edition of the Principles of Naval Architecture (PNA). Today the following techniques are available for predicting ship controllability: .

2. .

4.

Free-running model tests; Empirical methods based on available model test data bases; Large scale, manned models; Simulation coupled with captive model tests.

The relative merits of these methods have been long argued, perhaps most recently by Asinovsky (1983). Asinovsky proposed, for ship design purposes, to base evaluation of controllability and maneuverability on a Diagram of Steering, which is in many ways analogous to the results of a spiral maneuver. This Diagram can be constructed using steady-state equations of motion and sway and yaw damping coefficients derived empirically from series test data or from captive model tests of the proposed design. The primary limitation of this method is that it cannot provides information on what is typically the most crucial aspect of maneuvering performance, initial response to commands. The free-running model test was the first reliable method available for assessing ship maneuvering performance. Today free-running model tests are still extensively used to determine performance in standard or definitive maneuvers such as constant RPM (power) turns or zig-zag maneuvers which can be conducted for models of suitable scale within the sometimes limited dimensions of maneuvering basins or wide towing tanks. Such tests are not now as widely used as in the past, primarily because of difficulties in introduce the effects of environment (wind, current and/or waves) and of properly introducing effects of a restricted waterway. Introduction of time-varying control forces, such as rudder deflections and thruster forces can require elaborate radio or other control links between the model and a "shore-based" controller (typically a c o m p u t e r ) o r the use of a preprogrammed control computer on the model A variety of empirical methods have been developed for predicting the hydrodynamic coefficients of typical fully appended ship or for predicting the

610

2.

3.

4.

Data and formulas for deep water hydrodynamic coefficients for bare hulls and appendages given in PNA (Comstock, 1967), which were developed primarily by Jacobs (1964) and which are based in part on the Series 60 model rotating arm test data (Eda and Crane, 1965); Data and interpolation formulas for deep and shallow water (the latter for four hulls only) for fully appended, full form hulls of the MARAD series (Roseman, 1987); Methods developed by Norrbin (1971) which are available as part of a commercial maneuvering simulation computer program Empirical formulas for linear hydrodynamic coefficients developed by Clarke (1983).

A current SNAME T&R Project to develop a modular maneuvering model should also be noted. This project was initiated following the failure of an earlier, internationally funded effort to make significant progress in development of such a modular model. The proposed modular model, which will be concerned primarily with hydrodynamic forces, will treat separately the contributions to hydrodynamic forces of hull, rudder and propulsor and the interactions of these components. In addition, a number of empirical methods have been developed for predicting different aspects of maneuverability or controllability. Many of these methods, which include regression formulas for calculating turning circle characteristics (Lyster and Knights, 1979) and methods for predicting typical maneuvering performance based on statistical analysis of trials data (Barr, 1987), are discussed in the new Panel H10 Maneuvering Design Workbook, to be published shortly by this Society. While not a simulation method, Asinovky's analytical/empirical method for constructing a Diagram of Steering to aid ship design (Asinovsky, 1983) is of some interest here. This method uses the modular approach to predict the required velocity dependent hydrodynamic coefficients, and requires access to data for bare hull, appendage and propulsor hydrodynamic data and interactions which is comparable to the steady-state data required for simulation. Large, manned models have been used with success at Grenoble, France as a method for training ship's officers and pilots in the handling of ships such as tankers in restricted waterways and harbors. However, no other facilities of this type exist, probably due to the

A Review and Comparison of Ship Maneuvering Simulation Methods

large cost of this facility and the large, fully-controlled models. In addition, this technique appears to be primarily, if not exclusively, useful for training purposes, and thus it is not a direct competitor of the other techniques described. Simulation offers several significant advantages over competing techniques, such as free-running model tests, for assessing vessel controllability and maneuvering performance, particularly for realistic operating environments such as those required for operator training or for waterway design. Advantages of simulation include: 1.

2.

3.

Any maneuver or vessel operation can be rapidly assessed once a suitable simulation model is developed and validated - additional tests are not required for assessing a new maneuver or operation; Training requires a man-in-the-loop, and for greatest validity required, that human decisions be made in real (full-scale) time which is not possible for manned scale models such as used at Grenoble, France, and; Waterway design typically requires a man-in-theloop and real-time decision-making, as well as modeling of the full physical environment which cannot generally be accurately provided in freerunning model experiments.

Perhaps the most important advantage of simulation is that once initial model tests are conducted and a simulation model generated, almost any maneuver or ship operation can be simulated without a need for additional model tests; the simulation model can be readily and economically modified to determine the effect of changes such as increasing rudder size, adding or removing a bow thruster or changing in maneuvering speed. Maneuvering simulation can be either of the fasttime or real-time type. Fast-time simulations are conducted using autopilot or pre-programmed control and simulation of definitive maneuvers are of this type. The time required to conduct such simulations depends only on the complexity of the simulation model and the power of the computer. The simulation on a 386 PC of a typical definitive maneuver will require 10 percent or less of real (ship) time. Fast-time simulations provide a cost effective means for evaluating ship maneuverability, controllability and course- and station-keeping. They can also be used in waterway design studies as a supplement to real-time simulations (Webster, 1992). Real-time or man-in-the-loop simulations are conducted using a human operator or ship-handler and some level of simulator which provided ship state data (position, heading and/or velocities) required to make

meaningful helm and engine orders and a means for implementing these orders. The real-time decision making provided by the ship-handler is essential for shiphandler training and for waterway design studies. The need to assure the validity or a real-time simulation will typically impose demands well beyond those of the fasttime simulation (Webster, 1992). 4. C H A R A C T E R I S T I C S OF M A N E U V E R I N G S I M U L A T I O N M E T H O D S AND M O D E L S This paper is concerned with simulation and the applicability and validity of simulation models and techniques, and not with the broader subject of maneuvering simulators. Simulators, which are widely used in many fields, including the marine field, for training and waterway design, are devices which permit integration of: 1.

2.

3.

A human operator; Numerical feedbacks and displays which a human operator uses to perceive, and to control ship response; A mathematical simulation which determines the response of the vessel to human (or autopilot) control commands.

Figure 1 shows a photograph of a typical bridge simulator and Figure 2 indicates the basic elements of the simulator, and the relationship of the mathematical simulation model (shown within the dotted lines), to the other elements of the simulator.

¸

A Review and Comparison of Ship Maneuvering Simulation Methods

611

m

I j~

_

I 'ENVIRONMENTAL PROPERTIES

ENVRONMENTAL DYNAMICS

--I i I

t

.

_[ Jt "~

~

I I

I

t

~

l

~ "OPERT'E'I

RESPONSE LOOP

] SHIP ""O"ER"ES

~ I Pilot Auto- Irl tHuaaer, I~IDYNAMIC ~ STATE P'IDYNAMIC ~ SHIITE ::,.." I Pilot [ ]Prop, etc ) I IRESPONSEI I ] [RESPONStEJ I~"IST' ~_~_~L A t

",~___~

/

L

,,

.i AUTOPILOT

:

I

!,

r- - ~ . . . . . . . . . . . . . . . . .

~. . . .

I

I

l

I

PILOT

--

~

v

I

. ~sr~NSOR-~

]

[ FEEDBACK LOOP

' I

i I

J Figure 2 Schematic Diagram of Typical Shiphandling Simulator All ship maneuvering or shiphandling simulation models will includes at least: °

2. 3.

Coupled equations of motion (most typically surge, sway and yaw only) Deep water hydrodynamic coefficients Model for rudder(s), propulsor(s) and/or other control device dynamics.

Many simulation models, including those used in most simulators, will also include some or all of the following components: .

S.

6.

.

8.

Shallow water hydrodynamic coefficients or shallow water corrections to deep water coefficients; A model for wind forces A model for second-order wave drift forces (first-order, wave frequency forces are almost always neglected in maneuvering simulations) A model for continuous or interrupted banks A model for ship-to-ship interactions

The recent Marine Board report on Shiphandling Simulation (Webster, 1992) provides a good introduction to and assessment of the current state of simulator technology, and Chapter 5 of that report discusses mathematical models in some detail.

612

4.1 Equations of Motion and H y d r o d y n a m i c Forces

The equations of motion are the central element of any simulation model because they determine the capability and the potential validity of the simulation. Most maneuvering simulations use couple surge-sway-yaw equations of motion. Simulations used for high speed ships, such as naval combatants, use coupled surge-swayroll-yaw equations to determine the magnitude and to account for the effects of the large heel (roll) that can occur during high speed turning. Maneuvering simulation equations contain terms which define hydrodynamic forces arising from the velocity and acceleration of the ship, the deflection of the rudder(s) and the operating conditions (RPM, pitch, etc.) of all propulsors or other force producing devices. Tables 1 and 2 present two representative sets of coupled, non-linear, surge-sway-yaw equations of motions or maneuvering simulation equations. These equations are typical of those used to predict ship maneuvering or handling in their inclusion of terms due to the vessel's inertia and its state variables, which include its accelerations (hydrodynamic or added mass and moment of inertia), velocities (hydrodynamic damping), rudder deflection or thruster state, propulsor RPM, pitch or thrust, any forces arising from interactions of the hull, rudders, thrusters and propellers or other propulsors. The primary non-linearities associated with maneuvering are hydrodynamic damping forces produced by ship sway velocity (sideslip angle) and yaw velocity or

A Review and Comparison of Ship Maneuvering Simulation Methods

F(v)

Table 1 Simulation Equations used by Hydronautics, Incorporated

- ~ ='((x:,%.

+ (x

'1¢o, m=l-x;,'l*t-

(',

"..'+," +

=hi) v=,=)) + ~ t ' .(xT'

bt

r=*)

b c + +: °..+,:'.I

+•t

't: +,"[-'+' £ (~+" +-3 %-"]'.' o

~IUD

F(v)

llllXill

'IWI

- ~ l i l l l 11 + ;IU+ • ¥ 1 1 1 . l ' I Y l l ' )

vtu =

- Iv.

U'ItUe

" e~t + (v R + r x X=) t

+

Xe - I N l t ' I t l = '

+ +,' (*.'+.-, +'.,.,'+,,+,, +'..".'. .+~

Pp

a4 v4 +

.....

-

alv

+ a3v3.

[1]

F(v) -- a,v + a~vlvl + ?

= ( ~ ¢ + ; o ; ) - ~ t" (+;'+ + +,1,1%1,=1) + ~ ='(+;++)

4. ~ t I

a3v 3 +

The second approacu assumes that damping terms must properly describe quadratic damping occurring at large sideslip angles ( = tanl(vL/U) - 90°), and hence that:

r= x I.t/u~u 0 .

+112 V2 +

XIgO' l l l ~l~i I IFI

ukmra, XR+l ' +.

alv

where v is a velocity of interest and F(v) is the force due to that velocity. Ship symmetry requires that F(v) and v have the same sign and thus that only terms of odd power are valid. Neglecting terms of order higher than v3:

A X~LI++_,OyCK+ m(,;-w-,¢+')

-

:)). ,

.,,,.,..,..(+-O

|

tl)

. +m

•

[2]

where the "?" indicates the possible need for an additional higher order (cubic) term. Some organizations using this second approach have found it necessary to add certain cubic terms. Experience has shown that similar forces can be obtained over a wide range of non-dimensional velocities using equations [1] and [2].

|

Table 2 Simulation Equations Used by HSVA/VBD 'thuD'

" YI,' + ;tu=

Wl, l+['

up I

- (dUM + m l m

+fOln l) V h l : l U - flop IJ~. q - lltepUlllea micro - ( n l u t ) l (n¢lu l)

~1

~ (3vUIU n + bv~U[u = + %Din t)

X = m ( ~ , - v r ' x or21 Y = In (~ + ur + x G ÷)

YA',Il ~ I1011E~

N = Izz I~ + m x G (v + ur} + ~ t=

+ ~ t ' (.

,' v t rl'lgll)

X = X.-,'~ d:

~j t*(llvq'UltVp +- " . I - I

'VlLlVtl)

+

=

b|

bt

++"°.' [-'.(+- + ++ +h*r*. .it m

.

(+

P4 I) + T I ' " . ' . . ' .

+ ~ t~(ll++UllVl + " . I . I ' + = I ' '

+ Xdd'dd

+ Xrr'rr

+

Xvd'Vd

+ Xrv'rV

+ Xduu'duu

+

Xddu'ddu

+ Y~"~"

+

Y0

+ Yv'v

+ Yvvv'VVV

+

Yu'U

÷ Yd'd

÷ Yddd'ddd

+

Ydu'du

+ Yr'r

+ Yrrr'rrr

+

Yvdd'Vdd

Y = Y~'v

+ +=o n+4,1,l'

* V;.i,i

•

Vm )

Ulll - (dU|bol + l ~ l [ I m + IOfn l)

v.® - Iva+ ~x l,~/u~

+ Ywd'VVd ~pl . I u m l

Xvv'VV

+ +'.'11

,,[+,,' ,-'#, +,,.' ,•,]

lr+o " l + , ' . p ' ( + ;

+

et

. - ,+

Xuu'UU

m +m

I ("""" >!::i':"'"r >J

++ ,.'., ,xr-~-u--

÷

+ Yrw'rvy

+ Yvrr'Vrr

+ bwmlU~O + evO let

x . . - i++' (x~rleo= II

N = N~'v

+

N;.'~-

+

NO

+ Nv'V

+

Nvvv'WV

+

Nu'u

- x' l=t. I I ) v~

rate. Traditionally there have been two approaches to describing damping forces, a mathematical or perturbation approach and an empirical or physical approach. The first approach assumes that damping can be described by:

+ Nd'd

+

Nddd'ddd

+

Ndu'du

+ Nr'r

+

Nrrr'rrr

+

Nvdd'Vdd

Nrvv'rVv

+

Nvrr'Vrr

+ Nvvd'VVd • +

A Review and Comparison of Ship Maneuvering Simulation Methods

613

The Table 1 equations (Miller, 1980), from Hydronautics, Inc. (now Hydronautics Research) include quadratic and cubic damping terms, while the HSVA/VBD equations (Gronarz 1988) of Table 2 include only cubic damping terms.

4.2 Treatment of Machinery Dynamics The dynamic behavior and response to commands of main propulsion machinery can have a large effect on highly dynamic maneuvers and operations such as crash stops and stationkeeping, and on shiphandling in restricted and complex waterways where frequent changes in engine orders occur. Simulation models of varying sophistication have been developed for different types of prime movers and propulsors, (Lewis, 1966, Rubis and Harper, 1972, Oltmann and Sharma, 1985 ). Such simulations have been developed for fixed-and controllable-pitch propellers and for Voith-Schneider propellers (Van Dyke and Wendel, 1981); it is not known if machinery simulations for steerable propellers exist. Some type of machinery simulation has been integrated into ship maneuvering simulations used by major simulator facilities. For many shiphandling and maneuvering operations rapid changes in propulsor RPM, pitch and/or power will not be required and rather simplifies algorithms can be used to predict time-varying propulsor thrust. Simulation of standard definitive maneuvers are typically conducted assuming constant propulsor RPM and pitch. Figure 3 shows a simplified block diagram for a CODOG machinery simulation. The more comprehensive diagram for a gas turbine/CRP propeller plant given by Rubis and Harper(1972), illustrates the complexity of a thermodynamically based simulation. ~'~P

I

OWER COt4TROL/I LEVER

/~

~ ~

~

"

,--DIESEL ENGINE /~ ~ _ MODEL

--

~

~

~

lz_

~

.----:-==-1"---

~ -

-

~

~

•

"x

~r~

"

Yz/

Figure 4, from this same paper, illustrates the rapid changes in fuel rate (throttle setting) and engine torque that can occur during maneuvers such as crash stops. Because of the high attainable time rates of change of fuel flow rate, a much smaller numerical integration time step is required for at least some parts of the machinery than for the ship. Typically, the integration time step used to calculate fuel flow rate is 10 percent or less of that used for integration of ship equations of motion.

I

70

----

--~-

!

m-"--r SHIP ERASHBACX - - - -

:

,?o

%

tO 0

T

R

A

L°ri°s I -

-

-

-

-

-

TAM

~

-II

~

59~r

50 40

o

50 - - - - -

~

IO 0 -I0

0

500

i000

1500

2000

E N G I N E SPEED

2500

3000

3500

NE, rpm

Figure 4 Simulated Gas Turbine Power Plant Response During Crash Stop Standard steering machinery is typically designed to provide a fixed rate of rudder deflection (typically 2.3 degrees per second for larger commercial ships), and there is usually little need to simulate the dynamics of such steering systems. However, it can be important to include any significant time lag that will occur between helm command and initiation of rudder motion. Most special maneuvering devices, such as bow thrusters, can provide rather rapid time rates of change of RPM or pitch and applied force. The dynamics of such devices are therefore often modeled with simple algorithms based on manufacturer supplied information and time constants (T seconds from full port to full starboard). The more critical problem for such devices is the effect on their hydrodynamic performance of ship motions and interactions with the hull or other devices. Manufacturers are often not a good source of this type of performance data. The effect of ship forward speed and device-device interactions have been presented by various investigators (Chislett, 1966 and Ankudinov, et al, 1986). 4.3 Treatment of Environmental Effects

Figure 3 Simplified Block Diagram for CODOG Machinery Simulation Model

614

Ship maneuvering and handling are effected by the operating environment which can include wind, current and waves. An adequate treatment of these effects must generally be incorporated into any simulation used

A Review and Comparison of Ship Maneuvering Simulation Methods

for applications other than the evaluation of ship design or operations in highly protected environments. The effect of current velocity is accounted for by using a relative velocity through the water or the vector difference of the ship and current velocities. This approach is universally accepted for cases where current velocity in uniform and where underkeel clearance is not too small. For very small underkeel clearances this approach may not be appropriate, but it cannot in fact be demonstrated that the simulation as a whole is valid for shallow water (depth-to-draft ratios of 1.2 or less). Temporal variations of current velocity are typically small or of very low frequency and are therefore usually neglected. Spatial variations of current velocity can be large, particularly in restricted waterways. Important vertical and horizontal variations in current velocity typically cannot be neglected. Various methods are used to account for vertical and horizontal current profiles. Typically, length- and/or depth-averaged longitudinal and transverse components are used, and a rotational components of current velocity is determined by integrating the transverse component of current velocity over the ship length. The effect of wind is typically calculated using a relative air velocity, aerodynamic force coefficients determined from wind tunnel tests of ships of similar type and above-water profile and windage areas and centers of pressure. Resulting forces will be sufficiently accurate if suitable wind force coefficient data are available. Account is usually taken of the effect of wind near-surface boundary layer, while horizontal variations of wind velocity are usually neglected. Some waterway studies have included wind shadows of large building. Temporal variations of wind velocity and direction are important for, and are sometimes included in, simulations of operations such as stationkeeping. The effect of wave drift forces can be important for open water operations such as stationkeeping. A variety of largely proprietary empirical and analytical methods are used to calculate such forces. The validity of such methods is generally unknown, and the treatment of wave drift forces in ship handling simulations must be considered to be at least in part "black art." Figure 5 indicates the simulated effect on turning of the ESSO OSAKA from an initial speed of five knots of a 26 knot wind and a 10 foot significant height wave (conditions typical of the environment at the Louisiana Offshore Oil Port or Loop).

--O b . WIND ONLY

II. NO WIND OR WAVES V = 26 KNOTS

c. WAVES ONLY d . WIND AND WAVES I~l~,,l~ 10 FEET

Figure 5 Simulated Effect of Wind and W a v e s on Turning of the ESSO O S A K A treated by introducing quasi-steady sway forces and yaw moments which depend on ship and waterway geometry and on instantaneous ship position relative to waterway boundaries. These methods, which are typically proprietary, are primarily based on data from several large studies of forces on models moving through a channel (Norrbin, 1978, Dand 1982 and Hatton, et al, 1984) and on various experimental and analytical studies of forces acting on a ship as it passes a short or interrupted bank (Norrbin, 1974) or a moving or stopped ship on a parallel course. While it is difficult to validate or even evaluate methods used to simulate such interactions, it should be noted that good agreement has been found between simulated and actual ship behavior for many ship operations in the Panama Canal and in other restricted waterways. However, such good agreement may be achieved only after ship and waterway model are empirically adjusted or "fine-tuned." It is not clear if existing simulation models can, without modifications that cause them to become unsuitable for general use, adequately reproduce highly interactive maneuvers such as the special passing technique, or "Texas Chicken," which is routinely used in the Houston Ship Canal.

4.4 Treatment of Interaction Effects 4.5 Solution of the Simulation Equations Ship handling can be greatly effected by proximity to, and hydrodynamic interaction with, banks or channel boundaries, nearby structures such as finger piers or nearby ships. Such interactions are typically

A wide array of numerical tools are available for solution of the equations of motion. Ship simulation equations are typically solved using a simple Eulerian

A Review and Comparison of Ship Maneuvering Simulation Methods

615

solution of the equations of motion. Ship simulation equations are typically solved using a simple Eulerian integration scheme; the power of current computers (including PC's) generally allow the use of a sufficiently small time step (one second for large ships) to achieve the required computational accuracy. There is rarely any benefit to using more sophisticated integration schemes such as Fourth Order Runge-Kutta for solving the ship equations. However, such methods may be appropriate, and can be essential, for complex machinery simulations that could require Eulerian integration time steps of 0.01 to 0.1 seconds. It is not known if anyone has looked at conditions under which maneuvering simulation equations can produce a classical "chaotic" behavior. Observed behavior of some highly directionally unstable ships might be considered to be chaotic! 5.0 ACCURACY AND VALIDATION OF SIMULATION M E T H O D S AND MODELS The recent Marine Board Committee on Shiphandling Simulators considered the question of the accuracy and validation of shiphandling simulation techniques and mathematical modeling, and its conclusions are summarized by Webster (1992). While that Committee was primarily concerned with validation of a complete simulator, and with associated concerns about the validity of the man-machine interfaces, it did conclude that even the validation of the underlying simulation model was problematic. The Committee identified a lack of relevant data, and any basis for validating simulator-predicted handling, for ships operating with very small underkeel clearances. Inputs from experienced pilots and shiphandlers have been the primary means available for validating results obtained using man-in-the-loop ship maneuvering or shiphandling simulators. Similar subjective validations by experienced shiphandlers have also been used t o validate results of fast-time or autopilot controlled simulations. The primary means used to validate mathematical simulation models has been to compare results obtained using these models with maneuvering trials data for one or more representative ships. While this means of validation is useful, it has a number of potential or actual shortcomings, including: .

.

616

Trials are almost always conducted in deep, unrestricted water, whereas critical shiphandling invariably occurs in shallow and/or restricted water; For many ships trials are only conducted in ballast conditions, although the most critical handling requirements often occur at full load

3.

4.

condition; The development of the simulation model, and particularly the derivation of the required hydrodynamic coefficients from test data, can be as much art and experience as rigorous science, and coefficients are routinely adjusted to achieve satisfactory between simulated and measured results for key maneuvers such as a 35 degree rudder turn; Trials data are often of limited accuracy and may not adequately reflect the influences of the physical environment during the trials.

For these reasons it has been difficult to establish, using trials data, the general validity of simulation techniques or particular simulation models. However, such trials data have made it possible demonstrate that simulation as a technology, and at least some specific simulation models, are valid for certain limited ranges of operating conditions, such as operation in water of moderate or great depth. Several attempt were made to use data from carefully conducted maneuvering trials to validate simulation methods. As a result of an International Towing Tank Conference (ITI~C) initiative, a number of organizations tested models and conducted maneuvering simulations for a modified Mariner (the U.S.N. Compass Island). Some of the results for the Compass Island are presented and discussed by Hagen (1983). In the early 1980's the U. S. Coast Guard conducted a very complete set of trials of the WMEC 270 cutter, in part to provide a means for assessing the validity of various methods for developing maneuvering simulation models. This effort included development of simulation models and maneuvering predictions using both: .

.

An extensive set of deep and shallow water captive model tests and maneuvering simulations at Hydronautics, Inc. (Horwitz, et al, 1983); Use of the MARAD developed MARSIS instrumentation package by SCI to develop a parameter identification based simulation model (Trankle, 1987).

The captive model test data based simulation results (Horwitz, et al, 1983) were in generally good agreement with the trials data The relative worse agreement of the parameter identification based results were attributed to shortcomings of the MARSIS instrumentation package rather than shortcomings of the parameter identification methods employed. In 1978 a comprehensive set of trials of the VLCC ESSO OSAKA was conducted in deep and shallow water. Two purposes of these trials were to better

A Review and Comparison of Ship Maneuvering Simulation Methods

characterize the handling qualities of a typical VLCC and to provide trials data of the scope and quality required to adequately validate maneuvering prediction methods. The use of the ESSO Osaka results for validation of simulation methods is discussed in the nest section. 6. C O M P A R I S O N OF ESSO OSAKA S I M U L A T I O N M O D E L S AND RESULTS

organizations that were known to have carried out model tests and maneuvering simulations for the ESSO OSAKA. This table lists the length of the model used and the types of tests that were conducted for deep and shallow water. Table 6 lists the organizations for which simulation models were available for the present study; JAMP is a maneuvering performance analysis committee of the Society of Naval Architects of Japan (SNAJ, 1985).

The full scale maneuvering trials of the ESSO OSAKA, carried out in 1978 under the direction of the late C. Lincoln Crane, (Crane, 1979) offer a truly unique basis for evaluating the state-of-the-art and the potential accuracy of maneuvering simulation methods. The E s s o OSAKA provides this unique basis because: 1.

2.

.

.

Table 3 Principal Characteristics of the ESSO OSAKA Length, LWL Length, LBP Beam (Mid) Draft (Mid) Displacement Trim

A large number of standard and non-standard maneuvers were executed; All trials were conducted with unusual care and attention to correction of results for environmental factors such as ocean current; Trials were conducted in deep water and in two water depths for which important bottom effects occur (underkeel clearances of 20 and 50 percent of static draft); At least 19 organizations developed model test data based simulation models, with 11 of these including both deep and shallow water models.

This author prepared for the recent Marine Board Committee on "Shiphandling Simulation: Application to Waterway Design," a limited, critical comparison of a number of the available ESSO OSAKA mathematical maneuvering simulation models and published comparisons of actual trials results and simulation results. This initial comparison, which was too extensive to include in the report of the Marine Board Committee, is available on a rather limited basis as a Marine Board background paper. The comparison made for this paper has been greatly increased in scope to more fully evaluate the value of the ESSO OSAKA data as a validation tool and to assist simulation users in their assessment of the validity and applicability of maneuvering and shiphandling

simulation. Table 3 presents the principal characteristics of the ESSO OSAKA. Table 4 lists the maneuvers conducted in deep water and at water depths equal to 1.5 and 1.2 times the ship at-rest draft during the special trials of the ESSO OSAKA. It can be seen from Table 4 that a number of non-standard maneuvers (accelerating and coasting turns, coasting zig-zags and various crash stopping maneuvers), as well as more conventional definitive maneuvers were conducted. Table 5, which was prepared by the Maneuvering Committee of the International Towing Tank Conference (ITTC), lists the

CB

CM Cp Propellers Rudders Rudder Area

1125 feet (343 meters) 1066 feet (325 meters) 173.9 feet (53 meters) 71.5 feet (21.8 meters) 314,410 long tons Level 0.830 0.998 0.830 One One 1290 square feet (119.8 Square meters)

Table 4 Summary of ESSO OSAKA Maneuvering Trials

TTFE OF PJHEUVEROR CALIBRATION RUN

,. POUI[U't'ERS T......... 35°~,udd., .......a e ebd. Turn e l e r35°, a t l n &rudd.~ 35d R rudder r......... ,~ - 35° , rudder •Z maneuver, ......... ~0/20 20120 z maneuver ..... ,.~ 10110

Sr£EO OF AFFrlOACIITO I~II£U1/ERS, KNOTS

0,~,,,0,,, 1.2 RIIIII.f~W 5.7 5. 7

0,~,~,,,T 1 li h~nlllN T 7

0........ '~ r 7. 1,

al ? nEaP

O~

O*

5

5

5

T

T

T

5

5 T

5 7

T

. , . . u d z ~. . . . . .

T

T

T

Sp~r-,

T ]-5

7

73.5

3.5

5.5

3.5 3-5

Stop, 35°I. rudder Stop. 350R rudder Stop . . . . t r o l l l d headJ.n& Stop, s t e e r l ' . g f o r -

].S

....... b ,..,~

2.

CALIBRAt'IOH

II~S

,~,,~ ..... t . o , o . .r~,.,,~,,,.p,o~ .'

S,.d~,~.

~r^~ .0.2

t.,.,

3.5

~.5. S,

3, T.~

7, "

0.5

1,

A Review and Comparison of Ship Maneuvering Simulation Methods

12

~

617

Table 5 Organizations Predicting ESSO OSAKA Maneuvering

I B o d e l Size ( L p p i n metre)

Name ot O [ g a n t s e L l o n

HodeL Tests

Water Depth

Captive Flee Cspt[ve Free Cap,lye Flee Captive Captive Captive Ceptlve

Deep

neep Oeep Shallow ShallOW

Britlsh Harltlme Technology, Feltham g u l g s r L s n Ship H y d r o d y n a m i c C e n t r e , Veins llamburglsche S c h i ~ f b l u - Y e r J u c h s Install ni¢oshlla University H i t a c h i Zomen, A k a s h i nydgonau~lcs Ine.e L a u r e l , H d . lshlka~sJima-Rarima heavy I n d u s .

3.536 1.425 3.411 6.125 5.000 7.222 3.000 3.000 T.aS7 4.OOH

I n s t [ ~ u ~ o Naztonale peg S t u d I e d [sFerienze dL g c c h l t e t t u [ I HIvals, Rome R r y l o v S h l p b u i l d l n ; ~eaeacch I n s t i t u t e s Leningrad Xgulh~ U n i v e r s i t y , rukuoka

4.514

rglm

Deep

2.17 6.51 2.500

Captive fret Captive

Deep

I I l t s u b l s h l Heavy I n d u s t r i e s , Nagasaki Hltsui Engineering & Shipbuilding Co. L t d . , ^ k l s h l m a Nippon ~okan Co. L t d . , T e u - C l t y

4.600

Captive T[as Captive

Shallow

Deep

osaka University Ship Hesel[ch 1 n s ¢ i L u t e , Tokyo

1 . 0 0 0 41 4.000

Cap~ive rtee F~le CIptlve r~ee Captive Captive

Shellc,¢

1)

2} 3) 4)

Shallow Shlllov Deep

Shailov

Tree

Stevens I n s t i t u t e Hoboken. 8 . 3 .

2 . 5 0 0 51

~.OOO

o [ Technology#

~umitomo MeaTy I n d u s t r l s s e Sanlgaws Tokyo O n i v e r s i t T H i H e ¢ c s n t l l e Marina Yerluchelnetslt lflt B|nnenechit£bau, D~llhutg

nsmatkl!

1}.

2is

3),

1.425 1| 2.500 Sl 3 . 0 0 0 31 S.OOO 2)

Free Captive

Captl*e

Shallow

Deep Shallow

Deep

Shsllov Shallov

rcee

4) a n d $1 i n d i c a t e t h a t a e l n q t e a o d s l yam used by d l [ [ s t u n t organ|zaelons.

Table 6 Code Identifying Sources of ESSO OSAKA Simulations Organization

Ankudinov and Miller; 1979, Miller, 1980; Dand and Hood, 1983; Abkowitz, 1984; SNAJ, 1985; Bogdonov, et al, 1987; Gronarz and Muller, 1988). In most cases comparisons were made only for typical maneuvers such as 35 degree rudder turns and 10-10 or 20-20 zig-zag maneuvers. The agreement between actual and simulated results for these maneuvers range from fair to good. Figures 6 and 7 and Figures 8 and 9 present comparisons which illustrate better and poorer agreement, respectively. The published comparisons indicate that while some simulation methods provide satisfactory results for a wide range of maneuvers (Miller, 1980), other methods appear deficient even for standard maneuvers such as a 35 degree rudder turn. This disparity in results may be due to differences in the basic model, accuracy or scope of the tests and data, possible model scale effects, skill of the researcher in interpreting test results and extent to which coefficients were adjusted to enhance agreement. Only one organization, Hydronautics, Inc., is known to have made comparisons for all deep and shallow maneuvers (Miller, 1980). The generally good agreement found for ALL trials maneuvers, as illustrated by the few results in Figures 6 and 7, lead to a relatively higher level of confidence in, and degree of validation of, this model. Confidence in this model is enhanced by results obtained in pre-trial simulations for a number of maneuvers made using the same equations of motion and a set of empirically estimated hydrodynamic coefficients. The level of validation of other models which can be obtained from the ESSO OSAKA results decreases with decreasing number of maneuvers for which good agreement of simulated and trials results was achieved.

Code

6.2 Comparison of Published Simulation Models Hydronautics, Inc. (Post Test) " " (Pre-Test) NMI (BMT) (Large Model) " " (Small Model) Bulgarian Ship Hydro. Center Davidson Laboratory HSVA-Hamburg/VBD-Duisburg MIT (Abkowitz) JAMP (2.5 meter model) JAMP (3.0 meter model) JAMP (4.0 meter model) JAMP (4.6 meter model) JAMP (6.0 meter model) JAMP (2.5 meter model)

1 1A 2 2A 3 4 5 6 7 8 9 10 11 12

6.1 C o m p a r i s o n of Published Simulation Results A large number of comparisons of trials and simulated maneuvers have been published (Eda, 1979;

618

It is not feasible to reproduce here each of the published ESSO OSAKA simulation models. However, it is possible to illustrate the nature of these models by comparing the more important coefficients used in each model. Tables 7 and 8 compare sway force and yaw moment damping coefficients used in the 14 published deep water and eight published shallow water simulation models, respectively, where organization numbers are defined in Table 6. The role of these coefficients are illustrated in the representative equations of motion presented in Tables 1 and 2. Rather dramatic differences exist in these published coefficients for the different models. However, significant differences in simulation models cannot be deduced from differences in individual coefficients. One meaningful means for comparing different models is to compare predicted forces for representative values of sway and yaw velocity. The differences in the coefficients in Tables 7 and 8 reflect both inherent differences in the model test data and simulation models

A Review and Comparison of Ship Maneuvering Simulation Methods

and differences in the allocation of measured forces between various linear and non-linear terms.

-

-

TRIAl.

. . . .

tlYDRONAUTICS,

INC.

SIMULATION TRIAL . . . .

IIYORONAUTIC5*

IHC.

0.'5? o

SIMULATION

2O'

.
0.86 design speed + SWAY - YAW Stable solution - SWAY + YAW Stable solution + SWAY + YAW Unstable solution as can be seen in Fig. 19 herewith. Three conditions must occur for a pitchfork bifilrcation to exist. These are: 1. The system, before bifilrcation, when slightly perturbed returns to a steady state; the closer the system approaches the point of bifi~rcation the longer it takes for it to return to steady state.

\ Fig. 19

2. The system when slightly perturbed after the point of bifi~rcation will return to a steady state other than that of before the bifilrcation point. Or 3. The system when not perturbed will not stay at steady state. The resulting shape looks like a pitchfork with the handle and two outer prongs being stable while the middle prong is unstable. This can also be seen in Fig. 19. Although the existence of multiple solutions does not mean the existence of chaos in itself, and further investigation is necessary, it does bring about an interesting fact that nonlinear differential equations do have multiple solutions. Depending on the initial conditions, you will be attracted to one solution or the other.

Lewis Molter, Member [The views expressed herein are the opinions of the discusser and not necessarily those of the Department of Defense or the Deparhnent of the Navy.] My compliments to the author for a most interesting and complete summary of surface ship maneuvering prediction. The author certainly has a good grasp of both sides of many of the issues facing accurate maneuvering simulation. The accuracy of ship maneuvering characteristics has intrigued me for years. The first conclusion in this paper is that "Existing simulation models based on suitable captive model test data appear capable of predicting with suitable accuracy the maneuvering or performance of ships . . . . " Yet there are major differences between simulation models of the Esso Osaka created by different organizations. How does the industry go about improving or proving the accuracy? The trne time history of the motion of a ship during a maneuver is impossible to determine precisely by any of the three experiment techniques, i.e., free model experiments, captive model experiments, or ship trials. Free nmning models suffer from the possible effects of scale, including control surface boundary layers, nonscalable hull resistance effects that effect propeller speed and rndder forces, and propulsion system dynamics to mention a few. Techniques have been devised to "correct" these deficiencies but their effectiveness is questionable. Captive model experiments, including both on a horizontal planar motion mechanism or a rotating arm suffer control surface boundary layer effects, and hull resistance effects to a lesser extent than free models. Techniques are available to minimize these effects. However, a very complicated data analysis is required to obtain the actual hull forces and an even more sophisticated motion prediction procedure is required to obtain maneuvering predictions. There is an enormous opportunity for simple errors to occur. Who has the best math model? This question is a mystery to me because the form of the equation should be determined from the forces and moments obtained for each ship. The measured forces and moments are fit to an equation or processed through a regression analysis. The resulting coefficients and the equations are incorporated into the final simulation equations. There are laws of physics that should be used to guide the form of the equations. Unfortunately, the form of the final equation is not unique. If the same set of model test results is analyzed by several highly trained scientists, several simulation equations will result. The predicted ship motions from each simulation model are likely to be different. Frequently more than interaction effects are "'empirically adjusted" or "fine tuned" to achieve reasonable results.

A Review and Comparison of Ship Maneuvering Simulation Methods

631

A bigger question is, How many degrees of freedom should be modeled? There is evidence that adding the effects of heel to the turn simulation can change tactical diameters by as nmch as 20%.

Do ship trials give the accurate ship maneuvering characteristics? Of course, if you make proper allowance for uncontrollable environmental conditions, such as wind, waves, and current, and for ship ballast conditions which were known to some level of uncertainty when the ship left the dock and continually change during the trials, etc. The trial results are accurate until the hull fouls or the ship changes displacement or trim. If maneuvering performance estimates from all measurement and prediction techniques would include statistical error bounds, the industry could start to approach the problem of accuracy. This is not an easy task, but it must be done.

Nils H. Norrbin Member The author merits compliments for an interesting review of current techniques in prediction and simulation of ship maneuvers, and, in particular, for a critical comparison of some of the results from the Esso Osaka post-trial cooperative program. In this discussion I will add some observations on the validation of the degree of dynamic stability of the ship suggested by the various results documented by the author. In commenting on the numerical approximation of hydrodynamic damping the author rightly points out that cubic or abs-square approaches (with a possible inclusion of additional higher-order terms) have shown to produce similar forces over a wide range of sideslip or turning ratios. The more serious observation, however, is that the first-order derivatives may then come out widely different, a fact which may often be hidden by manipulations made to ascertain the validation of the highly nonlinear trial maneuvers. I have chosen here to examine this problem by focusing on the dynamic stability lever

l'r - l'~ = (mxc m - Y~

N,~) : L

as it may be calculated from the attthor's Table 7. It may be recalled that the original Esso Osaka spiral tests gave evidence of a "marginally stable to slightly positively stable" ship, i.e., with l~" -lv" in the range of 0 - 0,10, say. In Fig. 20 accompanying this discussion, lr" (filled) and lo" (open symbols) are plotted to a base of scale model length. Results from essentially cubic approximations are indicated by triangles, those from essentially abs-square approximations by squares. For clearness, results inferring a positive dynamic stability have been marked by a bar below the number of the source organization also shown. It is reasonable to assume the expected values to scatter within 0,35 < 1~' < 0,45 and 0,45 < lr" < 0,55. The appearance of the diagram then invites the following comments: • The results for 2A and 4 m a y suffer from small-model scale effects. • The //-values of 2 and 5 should be reconsidered, as they m a y suffer from errors introduced in the measurements, in the analysis or in the "tuning" of the mathematical model; alternatively, there m a y be printing errors or the discusser m a y have misinterpreted Table 7! (He has, admittedly, ventured to change some signs, as he does not believe in positive N(r) damping moments!) • Which one of the models 1 and 1A was used for the simulations included in Figs. 6 and 7? Are these two models based on the same captive model tests, differing only in the way of curve fitting or tuning?

632

Eugene R. Miller, Jr., Member First, I would like to thank the author for his efforts. For those who are involved in ship maneuvering studies, port and waterway development, and related applications, the validity and applicability of the simulation models used is an issue of critical importance. The type of information presented in this paper is rarely seen and important to know. Further, the development of this information is tedious and I don't think the author had much help from graduate students. I would like to cover two areas addressed by the author in this paper, the comparison of model test results and associated scale effects and the evolution of maneuvering simulation methods.

Comparison of model test results and associated scale eff e c t s ~ A significant part of the paper is devoted to a comparison of the forces and moments reported from a number of different model tests on geosims of the Esso Osaka, in particular, Figs. 10-17 and Appendix A. Unfortunately, I don't think that we can conclude very much from these comparisons except that there is reason for suspicion about some of the resuhs. Consider the data presented in Figs. 10 and 11 for Y' and N' as a function ofv'. These data are results from about the simplest type of tests that can be performed in a towing tank. Sorting through the figures and the data in Appendix A, one can see that the one outlying data curve comes from tests of a 5-m model (near the larger end of the size range). We can also note that two establishments tested the same small model and got results that varied between them by 15% to 20%. This is on the same order as the differences from one test facility on tests of both a large and small model (Table 11). On the other hand, some of the results from tests of the largest models from three different establishments are in good agreement. Part of the differences may be due to the test procedures. For example, some facilities derive Y" and N" versus v" from straightline towing tests over a range of drift angles. Others may derive the same forces and moments by the extrapolation of rotating arm tests over a range of drift angles to an r" of zero. As an aside, one should not put much attention on the data for Y' versus r' (Figs. 14 and 16). These data are derived from measurements that involve small differences of large numbers and a lot of scatter should be expected. Evolution of simulation technology and methods--In his concluding remarks, the author observes that "there is little evidence to suggest a comparably rapid or revolutionary evolution of simulation technology and methods.'" This is trne, but I do expect that there will be a considerable evolution of simulation modeling over the next several years. Specifically, I expect some or all of the following: Physics based models The increase in available computer power makes simulations based on direct calculations of forces, moments, and flow velocities likely. For example, it is no problem to determine nonlinear forces due to crossflow velocities by a direct integration along the hull at each time step. Application of full-scale trials--The availability and use of fidl-scale trial data to develop and/or validate maneuvering simulations will increase. This is due to the development of differential global positioning systems (DGPS) which provide very low cost and widely available high precision position measurements. Six (6) degree-of-freedom modeling--The development of simulators with motion bases, coupled with the availability of computer image visual scene generation systems that can realistically represent three-dimensional waves and ship motions will drive the need for improved time domain six degreeof-freedom modeling. There should be fairly rapid development of applications in this area.

A Review and Comparison of Ship Maneuvering Simulation Methods

1,5

J

1,0

0,5

,_,[

1A

L

f

7

Lm

ZA I~,

0 0

1

2

3

t,

5

6

8 m

Fig. 20 Esso Osaka in deep water~ynamic stability lever and components as calculated from linear force derivatives dedved by various organizations H. Paul Cojeen, Member [The views expressed herein are the opinions of the discusser and not necessarily those of the Department of Transportation or the U.S. Coast Guard.]

I am pleased to have been asked to comment on Dr. Barfs excellent paper. I believe that Dr. Barfs contributions to advances in maneuvering are to be commended. I am appreciative of his kind remarks, and certainly I have been a witness to numerous of those accomplishments reported in this comprehensive paper which will be of value to our technical community for some years. With those words on the paper, I would like to highlight to the reader those aspects of his paper which are particularly noteworthy; I have no substantive criticisms of the paper. To those of us associated with the Marine Board report (Webster, 1992), I was hoping that the author would have taken the opportunity to make more far-reaching conclusions as to simulator quality and limitations than the learned members of the MB Committee were able. However, Dr. Barr has done his colleagues on the MB Committee a service by reporting on the comparisons of the extensive Esso Osaka model tests with fidl scale trials conducted in the late 1970s in the Gulf of Mexico. I believe that the anthor can include Esso Osaka models performed in the late 1980s by Merintech and the Norwegian Institute of Technology; could Tables 5 and 6 be accordingly amended in the author's closure? The final area I would like to discuss pertains to Sections 2 and 3 of the paper on predictions of maneuvering performance, especially the work now being finalized by the IMO. After 20 years, a recommended set of maneuvering performance standards should be approved by the 18th Assembly. The Sub-Committee on Ship Design and Equipment (DE)

has been following a plan of action agreed and approvedby the Sub-Committee in early 1991. Prediction of maneuvering performance in the design stage is important, and has been the subject of many philosophical discussions here and in London. There are available throughout the world technical community over a dozen prediction methods based on regression of empirical performance. It is my belief that for the majority of new designs, the maneuvering performance will be more than adequate, and use by the designer of one of these prediction models is both practical and expeditious. Each of these prediction models have a varying degree of uncertainty based largely on the "database" used to develop it. The point I want to make is, if the new vessel characteristics are outside the ranges of the parent data, then more extensive analyses have to be sought by the owner. Further work on the maneuvering performance standards is not necessary at this time, but I call your attention to the "Explanatory Notes" which the DE Sub-Committee will complete in February (1994), and should be approved by the Maritime Safety Committee in April. Please help us complete this work through assistance to Panel H-10 and through the SOLAS Working Group supporting DE. John C. Daldola, Member In 1981 this discusser made a presentation regarding "Maneuvering (Daidola & Daniel 1981) Considerations in the Ship Design Spiral" to the Metropolitan New York Section of SNAME (Daidola & Daniel 1981). It addressed the dearth of techniques available to a designer to analyze vessel maneuverability as well as the absence of comprehensive criteria on which to judge satisfactory performance. Today, as the author has pointed out, the IMO has suggested criteria for the maneuvering characteristics of ships including

A Review and Comparison of Ship Maneuvering Simulation Methods

633

turning circle, tactical diameter, initial turning ability, over shoot and stopping distance. The situation with maneuvering simulation has also improved. Although by their nature the more comprehensive simulation techniques are private, they are usually made publicly available by their owners in a service form. Of particular use in design is the technique of numerical simulation without model tests but theoretically determined hydrodynamic coefficients, which does not seem to appear in the author's list of techniques. For vessels within the proven bound of the theoretical procedures, which is usually acknowledged with the results of previous model testing, the difference in time and cost to obtain results over techniques requiring testing can be very substantial. Once IMO criteria are adopted it is believed that the feedback of fidl-scale vessel maneuvering data will be rapid. This should filrther aid in calibrating simulation techniques in a similar fashion as is present in the area of resistance and propulsion. The feedback of data should also serve to help tune the criteria. The author has provided a usefid update on the status of maneuvering sinmlation. The technology is definitely moving •ahead but as he has concluded, more has yet to be done. Additional reference Daidola and Daniel, "'Maneuvering Considerations in the Ship Design Spiral," New York, SNAME, March 1981.

Robert Sedat, Member [The views expressed herein are those of the discusser and not necessarily those of the Department of Transportation or the U.S. Coast Guard.] Thanks to Dr. Barr for a very intbrnmtive paper on some of the many sources of uncertainty in maneuvering simulation. His paper concentrates on differences in the hydrodynamic coefficients used to calculate hull tbrces. While other sources of uncertainty are mentioned, the brevity of their discussion seems to imply that they are less significant. On the topic of predicting hull forces, there are fimdamental differences between steady rotating arm tests and unsteady PMM tests which are amplitude and freqnency dependent. Also, different tanks use difl'erent dynamometers, calibration techniques, centripetal force corrections, and regression models. Large test matrices are needed to predict fbrces developed at high drift angles and astern conditions. Scale effects also become more problematic at high drift angles where viscous forces and flow separation effects become significant. Finally, if results are to be usefid for preliminary design, a whole series of systematically varied hull shapes must be tested. Dr. Barr has not discussed the many different models fbr propeller and redder torce prediction. These are complicated by downstream interactions (hull -~ propeller, and hull -~ propeller --~ redder), and upstream interactions (propeller hull, and rudder -~ propeller --) hull). While the concepts of wake fraction and thrust deduction are universal tbr steady straight ahead motion, modeling of analogous interactions is still raider development in the field of ship maneuvering. There are also different opinions on how many degrees of freedom to include in the equations of motion (roll, engine rpm?), which added-mass and damping tenns are negligible, and how to deternune those which are not. Finally, there are differences in the treatment of effects of trim, wind, waves, current, shallow water, ship-ship interactions, and b~mk effects. This is not meant to derogate the sinmlation successes which have been achieved by clever simplifications of the 634

above issues. Rather, it is intended to point out that maneuvering simulation encompasses a whole genre of techniques with a common but difficult goal. Present simulations are usefid for qualitative comparisons, but quantitative results should be viewed with skepticism unless validated by extensive fidlscale trials data. Fortunately, simulation models are constantly being improved, and the development of differential GPS promises to make it much easier to conduct accurate ship trials.

Michael Schmiechen, Member [Oral.] After the excursion into universal chaos I would like to draw attention to some more down-to-earth observations. The first concerns models and parameters in general. As a matter of fact any model or set of equations spans a space of representation and the parameters of a system are its coordinates in that space. Evidently the same system may be described in terms of different models and consequently by different sets of parameters. There is nothing wrong with this. We have only to acknowledge that models and parameters are inseparably connected with each other and that parameters cannot be compared unless we have made sure that the models used to identify the parameters are exactly the same. The next observation concerns the omnipresent noise. On the basis of experimental data alone we cannot decide which model is the more "'correct." Further, we have to keep in mind that the values of the parameters depend on the algorithm used for their identification. Unless we are using the same model and the same algorithm for parameter identification we cannot compare parameters. Particular care has to be exercised to provide test data tbr identification which do not exhibit singularity, i.e., which are lacking information concerning the parameters to be identified. The third observation concerns the case of singularity of the system behavior. If a system is marginally stable, i.e., sensitive to initial conditions and chaotic behavior, the parameters cannot be identified uniquely. Again there is nothing wrong with this. There are algorithms which can deal with such situations. Any set of parameters is equivalent and describes the behavior of the system perfectly well, but evidently parameters can no longer be compared directly. The fourth observations concerns higher-order models, where parameters are usually determined in terms of whole matrices. Again, in this case parameters cannot be considered individually but only matrix-wise. The behavior of a system is deternained by the invariants of the matrices, e.g., by the determinants, etc., and not by their individtml elements. Consequently, in sensitivity studies one cannot study the influence of individual matrix elements but has to change the whole nmtrix in a physically meaningfill way.

Author's Closure The author thanks all of the discussers for their vahmble comments and supplementary information. Lt. Lo Sciuto has presented an interesting analysis of the coursekeeping stability of the Esso Osaka. His results indicate that a bi-stable condition occurs for speeds greater than 86% of the design speed, a result that is surprising in view of our traditional belief, based on linear stability theory, that stability, characteristics are speed independent if all coefficients are speed independent. I hope that he can provide references for the methods employed to obtain these results and the source of the speed dependence of the results. Mr. Motter has raised a number of issues affecting the accuracy of manenvering predictions, including limitations on each of the relevant types of model testing. He raises the

A Review and Comparison of Ship Maneuvering Simulation Methods

question of the number of degrees-of-freedom (DOF) required. Many fast-time simulation models use 3DOF, although models used for high-speed ships typically include a fourth DOF, roll. I suspect that the quoted difference of 20% is for a case of high-speed turning. It should also be noted that some large simulator facilities are now using 6DOF simulations primarily to increase the fidelity of the simulation environment, rather than the validity of the simulation ship trajectory. Establishing statistical error bounds is part of the process of establishing the attainable and required accuracy of the simulation model. A carefidly planned and comprehensive research program will be required to establish a usefid quantitative definition of attainable and required simulation accuracy. I wonder who is able or willing to pay the bill for such a program. Dr. Norrbin has provided a vahmble addition to the paper through his analysis of linear coursekeeping stability based on the published data for various Esso Osaka models. It has generally been our practice to use stability derivatives, which can be different from the linear damping coefficients, for evaluation of stability--the procedure helps to avoid the problems associated with specific predictional models described by Dr. Norrbin. The results in the figure provide an additional concern about the possible existence of model test scale effects. Figures 6 and 7 results are based on a simulation model which used coefficients derived from LAHPMM tests conducted at Hydronautics, Inc., after the conduct of the ship trials. Dr. Norrbin has raised a question about possible errors in published data as reproduced in this paper--I suspect that there probably are errors in some published coefficients and it is also possible that I have misinterpreted some of the less well-defined coefficients. Mr. Miller has correctly pointed out that agreement shown in Figs. 10 and 11 is generally good, with a few outlying results. However, I would note that most of the figures and tables show a less satisfactory state of agreement. It is certainly more difficult to determine cross-coupling terms and forces, such as those of Figs. 14 and 16. However, these terms can have a significant effect on predicted maneuvering performance and I do not believe that poor accuracy of such forces can or should be accepted. I doubt that we will, over the next few years, see a significant evolution of simulation technology and methods in the first two areas mentioned by Mr. Miller. Theoretical methods continue to remain more promising than reality for problems such as the prediction of most hydrodynamic forces associated with maneuvering. The general availability of DGPS will certainly increase the potential accuracy of trials data, but this alone cannot ensure that filture trials data will be of the scope of accuracy required for validation and improvement of simulation modeling. The use of 6DOF simulation models at large simulator facilities is already a reality, and rapid developments in this area can be expected. However, the use ofa 6DOF simulation is designed primarily to improve the fidelity of the bridge environment rather than the accuracy of the hydrodynamic simulation model. Mr. Cojeen wishes the paper had provided more broadreaching conclusions on simulator quality and limitations. I felt that such conclusions were not appropriate to a paper on simnlation technology, as distinct from simulators, and were not warranted by the scope or quantity of comparisons presented. I would have been happy to included Esso Osaka

data from Marintech--it was my understanding that such data were not available at the time the source data were originally collected. I am glad that Mr. Cojeen has discussed the current activities of the IMO in the maneuvering performance area. The new IMO maneuvering performance standards will almost certainly lead to a significant increase in the effort devoted to, and the resulting accuracy of, ship maneuvering performance predictions. Dr. Daidola's discussion provides the valuable perspective of a ship.designer. His referenced paper has been an important stimuhts of work in the area of ship maneuverability. Theoretical or empirical-theoretical methods are certainly needed for assessment of maneuverability during early stages of ship design. However, I am not aware of the existence of published methods which are capable of accurately predicting all of the coefficients required for simulation of the maneuvering of any representative ship; this capability is claimed for a number of proprietary methods. The experience of the past ten years does not provide great confidence that there will be a rapid feedback of maneuvering trials data to the IMO, and there will remain a concern about the quality of normal trials data. Mr. Sedat has touched on a number of important issues relating to model test techniques, propeller-hull-redder interactions and required degrees-of-freedom. Choice of number of degrees-of-freedom is addressed in my response to other discussers. The paper notes the importance of hull-propellerrndder interactions and indicates the reasons for not addressing this complex area in greater detail. As noted in the paper, each experimental method now used for determination of hydrodynamic coefficients has advantages and disadvantages. The choice of the method used will depend on practical considerations, such as choice of test facility, and on the intended use of the data. Thus, the planar motion mechanism can be the best choice for developing a fidl maneuvering simulation model, while the rotating arm will usually be the best choice for studying steady turning performance. With any experimental method employed, care must be taken to ensure that accuracy of the results is not compromised by the specific test techniques or equipment employed, the scope of tests, the data analysis methods used or the model size (scale effects). Dr. Schmiechen suggests that it is not the form of the equations, but rather the representational space consisting of the equations and associated parameters or coefficients that is important, and that component forces are therefore not significant. While I certainly agree that it is not meaningfid to separately consider the equations and coefficients, I cannot agree that there is no merit in the comparing component forces, which represent the total forces associated with one or combinations of state variables, predicted by various simulation methods (combinations of equations and coefficients). Mr. Jim White (Member) (comments from the floor) has suggested that valid results can probably be obtained using both small and large models if one remains vigilant during tests and during analysis of test results. While this may be trne, I do not believe that we now have meaningfid quantitative criteria for making such an assessment. The difficult problem of validation becomes even more difficult when additional uncertainties, such as potential model test scale effects, are introduced. It is probably only through systematic model tests that issues such as scale effects can be adequately resolved. Mr. John Allison's (Member) comments (from the floor) on the use and value of simulators are appreciated.

A Review and Comparison of Ship Maneuvering Simulation Methods

635