Author:
Title:
Rap or t Date:
Sumarv:
DELFI UNIVERSITY OF TECHNOLOGY
LABORATORY FOR
MEASUREMENT AND CONTROL
Ir. W. Veldhuyzen
Modelling the helmsman of a super-tanker.
N 93
March 1973.
A supertanker can he considered as a non-linear system which
responds very slowly to changes in the rudder posision. Moreover, this type of ships is often unstable in loaded condition, i.e. it has a tendency to start
turning
either to the left or to the right.These properties make the supertanker very hard to handle.
A number of tests is performed using a ship maneuvering simulator. The trained subjects had to steer a 200.000 tons tanker along a varying course. The records of this tests were used to define a mathematical model, describing the behavior of a helmsman. The mathematical model obtained in this way is fairly good. A discus-sion of the further research is given.
The research mentioned in this report was sponsored by the Deift University Foundation (Delfts Hogeschool Fonds).
MODELLING THE HELMSMAN OF A SUPERTANKER.
Introduction
Most of the investigations concerned with the behavior of the human operator as a controller have been executed with reference to the pilot of an aircraft or spacecraft. Some work in this field has been done on the control of submarines. The human operator as a controller of surface ships, however, did not get very much atten-tion until recently.
In the Netherlands about two years ago the Institute TNO for Mechanical Construction (TNO-IWECO) at Deift built a ship
maneu-vering simulator [i] in order:
To study ship maneuverability.
o To obtain data for ship and harbor design. To establish criteria for ships and harbors. To execute traffic control studies.
s To study nautical instrument design and automatic control. To train officers and pilots.
The simulator design was based on data and experiences obtained with a previously built experimental simulator, designed by TNO-IWECO in cooperation with the TNO Institute for Perception (TNO-IZF) at Soesterberg; another partner in the simulator project was the Shipbuilding Laboratory of the Delft University of Techno-logy.
Earlier Stuurman [2] executed a number of trials on the experimen-tal simulator in which he showed that for small ships the control behavior of the helmsman could very well be approximated by means of a describing function model. He also found evidence that for larger ships a nonlinear model probably would give a more realistic description of the helmsman's behavior.
in consult with TNO-IWECO it was decided to continue the work of Stuurman as a joint activity of the Shipbuilding Laboratory and the Man-Machine Systems Group. Special emphasis was to be laid on modelling the helmsman of a supertanker with the following goals
2
in mind:
o To provide data on which the maneuvers of this type of ship under human control can be predicted in a number of situations. o To enable an evaluation of the employment of a human pilot
versus the use of an autopilot.
o To investigate in more detail 'the advantages of predictive dis-plays in supertanker control.
Some introductory experiments have been executed using the simu-lator as a supertanker moving at constant speed on an almost straight course. A first attempt has been made to describe the behavior of the helmsman by means, of a nonlinear model using four
trainees of the School of Navigation at Rotterdam as subjects.
2 Ship dynamics
The dynamics of a ship depend not only on the properties of the ship itself like shape, dimension, mass and engine power, but also on the topology of the surrounding water. The motions of a ship in the horizontal plane can be described by a set of nonlinear differ-ential equations. These equations describe the translations of the ship in a direction corresponding to the longitudinal axis of the ship and in a direction perpendicular to this axis as well as the rotation about a vertical axis through the center of gravity.
Fig. 1 gives an indication of the variables concerned.
In 1957, Nornoto [3] showed that if it is assumed thai the ship is sailing at a constant speed, then the relation between the
rudder angle and the rate of turn r can be described by means
of a second order linear differential equation. For most of the
maller ships this equation gives an adequate description of the ship's behavior in a number of standard maneuvers. For a
super-tanker, however, it was found that the behavior was essentially
nonlinear. Based on full scale trials }3ech [;4] proposed to extend
Nomoto's equation with a nonlinear term. This leads to the follow-ing relation [6]
V Yo V, centor of grevity-ships trajectory u -X0 FIGURE 2:
The quantities invoLved in the desc?'iption of the
ship's maneuvers.
where r(t) d(t)/dt is the rate of turn, t) is the heading
angle, t(t) is the rudder angle, and where the quantities a1, a2, T, T2, T3 and K are constants. Here, it should be noted that all constants in this equation are dependent on the hydrodynamic
proper-ties of the ship, which are related to, among other things, its speed, ìts load condition and the possible restrictions in the surrounding water.
In this study a particular ship viz, a 220.000 tons deadweight tanker in loaded condition has been chosen. Table 1 shows the principdl data of the ship. The constants in
q.
C i) for thisship have been determined by Clansdorp [5,6] in a series of full
scale trials; they are given in Table 2. 1f a stationary situ-ation is considered, that is, i(t) = 0, '(t) 0, t) 0, then
Eq. ( 1) changes into:
TABLE 2:
Principal data of the 2003000 tdw tanker i fully loaded condition.
TABLE 2:
Constants in the equation describing the relation between
the rudder angle and the rate
of
turn for the supertanker3in deep and still water; the speed considered is 7.72 rn/sec. Length Breadth Depth Draft D i s p 1 a c e m e n t Froude number 310.00 rn 47.16 rn 24.50 m 18.90 F) 238.000 n 0. 14 4
Fig. 2 represents Eq. ( 2); this static characteristic
is given
for the ship in fully loaded and bailasted condition., Tine ship is
course unstable in loaded condition; it has a natural tendency to deviate from the straight course and to start turning either in one direction or the other.
3 The maneuvering simulator
The simulator consists of a wheelhouse which has the same appearance
as that of a real sea-going vessel. The fore part of the ship, the
sea and a coastline are displayed on a screen in front of the
wheel-constants dimension
numerical values fully baden ballasted 1 2 2 -1 a2 sec /rad 80,000 16,200 T sec 250 80 T sec 10 3 T3 sec 20 6 K
sec1
-0.0434 -0. 0471starboard -16 -2 --03 - 0.4 rOte of Q turn r [o ort 8 12 rudcJr
I
5 starboa rd - -12 -03 01 -OEt -Q2 - 03 rateof turn r fa ¡sec port 0 4 8 12 -+-s-rudder crrgle 6[] FIGURE 2:Re lation be tween the rudder angle and the rate of turn r in the stationary state for the ship considered in this
investigation as found by Glansdorp [5] (a: loaded condition;
b: ballasted condition).
house. The total angle of vision of the helmsman is120°. The image of the fore part of the ship is static; it is produced by two slide projectors which have a fixed position. The coast line is generated by means of a point light source arid a movable model with three degrees of freedom simulating two translations and one
rotation in the horizontal plane. Fig. 3 shows the simulator
during a simulated approach of a harbor. A blockdiagram of the system, including the helmsman, is given in Fig. LI..
On an analog computer the dynamics of the ship to be simulated, including the characteristics of thrust engine and rudder engine, can he programmed. The computer yields the signals which control the environmental display system and also the instruments such as
compass, rudder position indicator, log, etc. The helmsman has the same controls at his disposal for maneuvering as on a real ship viz, the wheel, which gives the input to the rudder engine, and
prcj
LJ
FIGURE 3:
The Ship Research and Maneuvering 3imulator
of
theInstiute TNO for Mechanical- Constructions at Del-ft.
L
::::1i
.--- i'---J desir.d t s ate II Cii
FIGURE 4:Blcckdiagrarn
of
the TNO simulator.d stu r b a r e s 6
r
---i
L.
.. I d yn.j
I :1 Sfl*A t k;naWwr.a'-..,-wheet'
UCAG .rnp-tr :1.1 i- j
7
the telegraph to the engine, which
governs the speed of the
pro-peller'. External disturbances simulating
the effects
ofwind, waves
and currents can also be introduced into the model on tne analog
computer.
4 The experiments [7; 8; 9]
In the experiments described
here, the simulator
has beenused as
a supertanker at full sea
in fully loaded condition
as well as inballasted condition, moving at a constant speed. The analog com-puter
was programmed according
to Eq.( 1) based on the constants
given by Glansdorp as indicated in Table 2. The rudder engine
was also included in the simulation. Its dynamics have been
chosen
according to the Eqs
C 3) and C LI.).T4 (S(t) (t)
(
t)
I M,
where 6d(t) is the position of the steering wheel or the desired
rudder angle; where T4 is a time constant of 1 sec and where
the
quantity M is the maximal value of the rotation speed of
the rudder (0.045 rad/sec).
The subjects were four trainees of the School of Navigation
at
Rotterdam. They were studying for the rank of first or second mate
after having been at sea for several years. All four
trainees were experienced in steering conventional cargo ships; only one of them
(subject A) had sailed on tankers up to 90,000 tons dead
weight. Their task consisted of following a straight course for about half an hour, or of following a preprogrammed but unpredictable
course which changed between +20
and
_20 around a certain nominalcourse
for
about fourty minutes. This prescribed course acted as a testsignal
for
the man-supertanker system.Because it is
not a
real-istic situation
to follow acontinuously changing
course, a binary
signal was chosen. The construction
of the binary test signal
wasbased on the idea that with the experiments to be executed also linear models of the helmsman's behavior might be investigated.
Therefore the input signai (desired course) was constructed in such a way that about 70 percent of the input's energy was concen-trated at only four frequencies. The amplitude of the binary signal was equal to the two degrees earlier mentioned. The four sinusoidal
components were the third, fifth, eight and thirteenth harmonics of a sinusoid with a period of forty minutes; the phase at the initial point was chosen at random.
TABLE
3: S:nimary of ti
tests performed.
* With disturbances due to waves.
**- With rate of turn indicator.
--- Because of a short age of time this run could not be performed.
)ATE 1972 April 25 1972 Anni 26 1972 May 2 1972 May 3 7.00pm 8.00pm 0.00pm 10.00pm Subject A loaded cand. course chang. Subject c loaded cond. course chang. Subject A 13 ball. cond. course chang. Subject D loaded cond. course chang. Subject B 2 loaded cand. coursa keep, Subject D 8 loaded cand, course chang. Subject B loaded cand. course chang. Subject c 20 loaded cond. course chang. Subject A 3 loaded cand. course keep. 9 Subject C loaded cand, course keep. 15 Gubject A loaded cond. course keep. 21 Subject D loaded cand. course keep. Subject B loaded cand. course chang. Subject D 10 loaded cond. course keep. Subject B loaded cand. course keep. 22 Subject C loaded cond. course keep. Subject A 5 ball, cand, course keep. Subject C 11 loaded cand. course keep. Subject A 17 loaded cond. course keep. Subject 23 loaded cond. course keep. Subject 3 6 ball. cond. course keep. Subject D 12 4* loaded cand. course chang. Subject B * loaded cond. course keep. Subject C 24 * loaded cand. course keep.
g
1n Table 3 a summary of the tests is given. During the experi-ments the compass as well as the rudder angle indicator were used.
Three tests have been executed in which also the rate or turn indicator was used. With the exception of two trials no external disturbances simulating ship motions originating from wind, waves and currents were introduced; these external disturbances were generated by means of a digital computer and consisted of the sum
of 23 sinusoids simulating the motions of the fully loaded tanker in a following sea with long crested waves. 'These waves were consid-ered to originate from a wind with a force of eight to nine
Beaufort. The spectral density function of these motions have been calculated in reference [io']. During the tests the following sig-nals were recorded on magnetic tape:
e The desired course
The course of the ship t).
The desired rudder angle
s The rudder angle (t).
The rate of turn r(t)
6.5 Modelling the helmsman's behavior
In the Figs 5 through 7 some examples are given of the time
histories of the desired course Yd(t); the course of the ship '4.t(t);
the desired rudder angle d(t); the rudder angle 5(t), and finally
the rate of turn r(t) as recorded during the tests. In all cases the records show that the helmsman generates a rudder angle t)
as output which consists of discrete steps. Hence the records indicate that a linear model to describe the helmsman's behavior will not fit the data very well. To check this presumption the ratio between the energy not located at the four frequencies
men-tioned in paragraph 14 ard the total energy of the signal
consid-ered has been calculated. The table 14 shows the results for the
four subjects, each having performed two runs of forty minutes. From the table it can be concluded that the remnant energy, in particular of the desired rudder angle 5d(t), is that high in most of the runs observed, that it does not seem appropriate to focus the attention on linear descriptions of the helmsman's behavior
TABLE
4:
Survey of the remnant energy in the signais observed
for eight tests with four' subjects: Loaded condition;
without rate of turn indicator; without disturbances.
any longer. It suggests that the helmsman bases his dicisions on when to move and how much to move the wheel on some criterion which
is, for instance, a function of the heading angle (t); the rate of turn r(t); the position of the rudder t), and, of course, the desired angle
By the compass the course t) is displayed to the helmsman; while in the case that no rate of turn indicator is used,it is assumed that the helmsman estimates the rate of turn r(t) directly from the time history of the course (t). Based on the records obtained by preliminary tests as well as on the records shown in the Figs
5, 6 and 7 a model for the helmsman's behavior is proposed which states that if the absolute value of a quantity s(t) which is defined as:
(t) [(t) - d(t)l + c1q'(t) ( 5)
exceeds a certain threshold value d1,
the
helmsman moves thesteering wheel into a position Sd (t) according to the Eq. C 6): *
(t)
a(t) + bF(t)
±c[(t)]3
d sign e(t) ( 5)In this equation the quantity Je(t) (t) 4Jd(t) and the
quan-tities tp(t), d(t) and 'ut) are the values of these variables at the moment that the threshold value is exceeded. As a result of
this action, the quantity s(t) wil decrease; if now Is(t)
becomes less than a second threshold value d2, it is assumed that
10 Subject A B C f) Run 1 2 1 2 1 2 1 2
Yd(t)
28.0
28.6
23.3
28.0 28.0 28.6 28.3 28.8t)
10.3
21.6
7.8
27.7 50.1 19.5 1.726.6
td(t)
88.7
88.0
90.9
9'4.3 78.285.9
78.1 87.96(t)
86.2
86.0
88.8
91.3
77.6
Bb.7
77.1
86.5
r(t)
48.3
47.9
49.552.9
70.0
53.2
35.0
56.1
rI
[rad/ sec2 o -10--20 L -10 -20 0.02 o - 0.02 1f! [degr.] -2 10 o { degr. j 10 o [dgr.] -10 0.1 r=1/J [ radi sec] -0.1 i 0.02 o [ rad/sec2] -0.02 o s 10 15 E'IGUJ?E 5: t [mia]Example
of
the time histories of the desired course (t); thecourse of the ship (t); the desired rudder angle 6(); the rudder
anale (t); the rate of turn r(t), and the derivatiie of the rate
of
turn with respect to the time '(t): subject A; without rate of turn indicator; without disturbances; loaded condition.L
uJJ
L
rifl,i
1.] I [degr.] -2 [degr.] -2 5d A 20 [degr.] 10 0.1 0 - 01r:1//
4 {radi sec J o [egr. J 20 10 o 10 15 t [mia] o2 [ degr] -2
V'A
2 o [ deç.r -2 20 [ dgr.J 10 o -10 - 20 5 A 20 [ degr.]i:
r1jf
0.1 O [ rad/sec ] -0.1rrV'
A 0.02 [ raisec 1 -0.02jrA2
[d2gr.] I -2 O [degr.] -10 0.1 O[ ruìsecl
; = 0.02 o21!
[rad/secj
-0.02 s 10 - t [mm] M4ArrwT&rJ_
15 12 10 15 FIGURE6:
-
t[rni]
Example
of
the time historiesof
the desired course » ,(t); thecourse of the ship (t); the desired rudder angie 6(); the rudder
angle (t); the rate
of
turn r(t), and the derivatie of the rateof
turn with respect to the time '(t): Subject B; without rateof
turn indicator; without disturbances; loaded condition. -1pJ
tIT
¡
ypiJ
I
. t{degr.] li, [degr.] degrj I (5 [degr. ]
rr
[rad/sec];-:;
2 rad/sec [ degr.] [ degr.] l'IGURE7:
Example
of
the timcourse of
the ship angle (t); the raof
turn with respeturn indicatori Wi (5 [degr.] [rad/sec] 2L o -2 20 o -lo -20 0.1 o -0.1 L 0.02 o -0.02 2 10 0.02 [rad/sec2] -0.02
i-y
A sw
LLr
10r.
15 »_ t mirie histories of the desired course
p(t); the desired rudder angle 5 ();
te of turn r(t), and the deri?'atie
of
et to the time i'(t): Subject D; without thout disturbances; loaded condition.13 ; the the rudder the rate rate
of
L-
-r 10r
2 L o -2r
1 o -2 o -lo lo o -10 0.1 o -0.1 O s 10 isthe helmsman moves the rudder again into zero position.
Parameter Estimation
The parameters a, h, c, d, d1 and d2 can be estimated by
minimizing a quantity J1 defined as:
J1
+ J
Tc( t)] dt, C 7)where the signai c(t) represents the dif[erence between the human
operator output d(t) and the model output 6d*(t). The minimal
value of the quantity J1 can he found by partial differentiation of this quantity with respect to each of the unknown parameters and by setting the result equal to zero. This yields as many equa-tions as there are unknown parameters. 9ecause there does not exist a simple analytical relation between the input and the output of
the model of the helmsman it is not possible to solve the equations just-mentioned in a simple way. Furthermore, it should he noted that if a model with given parameters should be inserted into the control loop instead of the helmsman, this would lead to another
time history for the quantities 5d(t), i(t) and r(t). Go, in order
to get an unbiased estimate of the parameters in the human operator model, a comparison should be made between the output of the helms-man and the output of the model, where this model is also part of a closed loop system with the ship's model (See Fig. 8). The parameters can be found by means of a direct search method mini-mizing the quantity J1 with respect to the model parameters. With
the help of Eq. ( 7) the quantity E1 can be defined:
T T f[c(t)]2dt E o T 2 *
Q[dt]2dt
Id(t)l dt
o ( 8)The quantity E1 indicates how well the model output approximates the actual output of the helmsman.
(t) he (msmon porarneters modo) of
ödit)
helmsrnejn direc t scrc h rn o t h o d t) FIGURE 8:Application of an error criterion in such a way that unbiased parameters can be obtained for a model of a system in a closed loop.
The correspondence between the time histories of the
actual course
of the ship t) steered by the helmsman and those of the ship
model '0(t) steered by the model of the helrnsmarYs behavior can be expressed by the quantity E2, defined as:
T
f[t)-t)]2dt
n E ( 2 Tf [(
t)] 2dt o 7 ResultsBased on the data of one test, in which the subject had
been
instructed to steer a ship in loaded condition along
a straight course for a period of 25 minutes, the parameters of a series of
simple models have been estimated [7; 8; g] according
to the
para-meter estimation method just-mentioned. The first model was a
linear one; later models were based on the Eqs
( 5) and ( 6),
s)
cri erion E(t)
JI
model of V'(t) ship
15
16
he it that nonlinear terms in the Eq. ( 6) were neglected. The
results with these simple models, however, were very poor, and did not give any indication how to modify the models used in order to
get a better result. Therefore, the task of the helmsman, which during these tests was to keep a ship in loaded condition on a straight course, was changed into the task of following a certain unpredictable but well defined course. In this case, again the parameters were estimated, and now the results were much better; they even lead to a direction in which the structure of the model had to be changed. Finally, the model as given by the Eqs ( 5)
and ( 6) was obtained.
The table
6.5
shows the valuesof
the model parameters as well asthe values of the quantities E1 and E2 calculated from four tests
TABLE 5:
Summary of the pararrie ters of the mode i given by
the
Eqs (
5) and
(6) and the quantities E1 and E2.
Subject A B C D Date 1972 May 2 1972 May 2 1912 May 3 1972 May 3 Test nr 15 1t 20 21 a 3.5 1.9 F.2 2.0 i [sec.] 176.9
155.0
210.0 230.5 c [degr.2] 0.0 0.1 -0.1 0.1 d [degr.] 1.7 1.0 1.0 1.5 C1 [sec.] 20.3 30.0 22.5 30.0 djdegr.] -5.0 0.2 0.3 0.2 d2[degr.] -5.0 0.1 0.2 0.0 40.1 70.9 43.4 32.8 E9[-c] 8.7 -- -- 3.117
executed by the four subjects A, B, C and D (See table 3). The Figs 9 and 10 show some typical time histories of the actual
signals d(t) 1'(t) and d(t) compared with the model outputs
(t)and d*(t). Some remarks with respect to the results obtained
can be made.
Although a remarkable difference in the steering behavior of the subjects exists (See the Figs 5, 5 and 7), the difference of the parameters between the different subjects is rather small. o In paticular for the subjects A and D the model fits the test
results fairly well.
o Although the output of the model of the helmsman's behavior does notaiways resemble the actual output very weil the course of the
ship generated by the model experiment fits extremely well. o In a few cases, for instance the test with subject C, the
opti-mization of the parameters by minimizing the quantity
'1 was
difficult. The problem was to find the absolute minimum of the quantity J1 in stead of a local minimum.
S Further research
Until now, only the parametérs of the model have been estimated e
for four subjects steering a fully loaded tanker along a prescribed course changing at certain moments from +2 degrees to -2 degrees and back. Also some tests were performed with a tanker in hallasted condition and with disturbances (simulating for instance the motions originating from seawaves) acting on the ship [io], but at this
moment no parameter is established. Some of these tests, in particu-lar those with the ship sailing in open sea with rather high waves, showed that fairly dangerous situations can occur when the helmsman is not well trained. Therefore, there is a need to work out the re-sults of the other test runs very carefully.
Sorne points which certainly have to be further explored will be: A study of the influence of the test signal used (the desired course) on the structure of the model and on the model parameters. A study of the influence of the ship dynamics on the structure of the helmsman model and on the model parameters.
\tf 2 [degrees] L [degrees] b' 1G U R E V1d [ degrees ]
Vt
11it) X. turn indicator.9: Typical time histories
of
the actual signals Pd(t)J
p(t) and
d(t) eompard with
the model outputs
(t) and
d*(t): Subject A; loaded condition; without rate of
a 0 5 10 15 20 25 30 35 . t [rnmn] FIGURE
20: Typica'.
time histor-tesof
the actual signals Jd(thj
i(t) and Sd(t) compared with
the model outputs
*(t) and
(t): Subject D; loaded condition; without rate
of
turn indicator. 35 30 t [mm] 1S 10 5 20 25 ô [dtgrees ] ifin
A study of the influence of the disturbances acting on the ship on the structure of the model and on he model parameters.
G The meaning of the strong difference between the rather weak
resemblance between the output of the model of the helmsman's behavior with the actual helmsman's output on the one side and the excellent agreement of the ship's course and that of the output of the ship's model on the other.
The physiological and psychological interpretatLon of the model parameters, as well as the cybernetic aspects.
A c k n o w ledge men t s
The authors gratefully acknowledge the contribution of C.C. Glans-dorp of the Shipbuilding Laboratory of the Delft University of Tehnologv. They are alsc greatly indebted to TNO-IWECO for provi-ding the facilities todo the experiments. The special thanks go to the staff of the TNO-IWECO simulator group for their contribu-tions in the preparation as well as in the execution of the trials. Finally, the authors would like to express their gratitude to the subjects from the School of Navigation at Rotterdam for their wholehearted cooperation.
10 References
i. Brummer, G.M..A.; van Wijk, W.R.
The Ship Manoeuvring and Research Simulator of the Institute TNO for Mechanical ConstructiOn, Deift, Inst. TNO for Noch. Constr.
Report Nr. 8l33/ (1970), pp. 1-32.
2. Stuurman, A.M., Modelling the helmsman: A study to
define a mathematical model describing the behaviour of a helmsman steering a ship along a straight course, Deift, Inst. TNO for Mech. Constr.,
20
3. Nomoto, K.; Taguchi, T ; Honda, K. ; Hirano., S.,
On the steering qualities of ships, International Shipbuilding Progress,
Vol. , Nr. 35 (1957), pp. 35-t-370.
l4 Bech, M.; Wagner Smitt, L.,
Analogue Simu1ation of Ship Manoeu-vres, Lynghy (Denmark), HyA Report
Uy 1 (1969), pp. 1-2'4.
h. Clansdorp, C.C.; Buitenhek, M.,
Manoeuvring Trials with a 200,000 tons Tanker, Deift, Shipbuilding Laboratory of the Deift University of Technology, Report Nr. 28 (1969), pp. 1-31.
Glansdorp, C.C.,
Simulation of Full Scale Results of Manceuvring Trials with a 200,000 tons Tanker with a simple Mathemati-cal Model, Deift, ShiDbuildin
Laboratory of the Deift University of Technology, Report Nr. 301 (1971), pp. 1-2.
Veldhuyzen , W .,
Len eerste model van bet regelgedrag van de roerganger van cen supertanker (A preliminary model to describe the behavior of a helmsman of a
super-tanker), ir-thesis B 810, L.v.W.M.R., Delft (1971), pp. 1-70.
V1dhuyzen, W.,
Onderzoek naar de regel techn ische aspekten van bet gedrag van de roer-ganger van een supertanker (An Inves-tigation on the behavior of a helms-man steering a Supertanker),
Report N 87, L.v.W.M.R., Deift (1972), pp. l-47.
21
.9. Veldhuyzen, W.; Luntereri, A. van; Stassen, H.G.,
Model1ingthe Heisman of. a
Super-tanker: Sorne preliminary Experiments,
preprint 8th Annual Conf, on Manual
Control, Ann Harbor (1972), pp. 1-17.
10. Veldhuyzen, W.,
3orekendespektra van de gierbewegin-gen van een 220.000 TDW tanker
(Calculated spectra of the yawing movements of a 220,000 tdw tanker),