AFDELING DER SCHEEPSBOUW- EN SCHEEPVAARTKUNDE LABORATORIUM VOOR SCHEEPSHYDROMECHANICA
Rapport No. 454.
ON SHIP AUTOMATION USING COMPUTERS.
M.Hirota
tJI
Deift University of Technology
Ship Hydromechanics Laboratory Mekeiweg 2
Deift 2208
Contents: page nr.
o. Summary
I
I. Position Fix at Sea
J
1.1. Astro-fix, Statistical Test I
1.2. Accuracy of Astronomical Observations at Sea 2
On Ship Automation Using Computers
by
Minoru HirotaX
Ship's Motion under Steering 3
2.1. Static Parameters (PAL-Test) 3
2.2. Dynamics 4
Optimal Steering
3.1. Course-Changing in Minimum Time (SHO-Method)
3.2. Course-Keeping by Minimum Time Loss or Minimum Fuel Consumption
Conclusion
References
Figures
X Kobe University of Mercantile
Marine
4 4 5 6 8 IO
)
)
o
O
o. Summary.
1. Position Fix at Sea.
This report consists mainly of an English translation of the author's work reported in Japanese. It contains (a) position fix at sea by using computers,
( b) automatic control of a ship's course, course-changing and course-keeping.
The report is not only theoretical but also practical, so all experiments are performed by an actual ship (a training ship m.v. Fukae Maru, 360 tons D.W.); the optimal controls are given by a feedback system. In this connection, a new method of the spiral test is described, as well as a practical algorithm for
optimal control, and some data obtained from actual ships, for instance, stan-dard deviation of error of astronomical observations at sea, and the ship's
static parameters and dynamics under steering.
We have many systems and apparatus to fix the ship's position at sea. In this paper, the use of the astronomical system is explained, because it is a typi-cal and basitypi-cal system at sea, and its extension can be applied to Decca,
Loran, Omega or Satelite system.
There are several spherical-trigonometrical formulae for astronomical position
fix which are well known to the navigators. 1, 2, 3 . The reduction of the
motion of celestial bodies is also formulated, so these are omitted in this report, and only statistical treatments are shown below. The accuracy of astronomical observation at sea is considered, because it is necessary to reject accidental errors by a statistical criterion. Some programs have been
used for computerization of the ship's position calculation.
The application of Kalman filter is omitted, because the estimated values of coveriance matrices are difficult to find and there is no theory of rejection
of erroneous data.
1.1. Astro-fix, Statistical Test 4, 5].
When the navigator observes the altitude of a star (a0 ), he measures also the
time of observation ( t0), and prepares a dead-reckoned position. Then he cal-culates a altitude (as) and an azimuth (Z) which are expected in t0 at the
dead-reckoned position withan aid of a navigational almanac. The difference between observed and calculated altitude (a0 - a) is called as an intercept
2
,dL
n-X E5iZCôsZ.
Yi coszZ.
We can estimate a2 as follows:
Si;
Ast
Z2(N-2,
) 2<
X2N2
jq)
where N
,st
s
Degree of freedom = N 2mi : number of position fix, m :
total number of position fixes.
The value of o 1'O is obtained from about 800 observations of 13 ship's
navigators after rejection of some data by statistical test.
It may looks rather large compared with other results, but it is the
ob-servation error of practical navigator's routine work at sea.
n
=-!
IL:CÖsZL
(2)
The most probable position (X0 ,Y ) is obtained by solving egs (2). Let S
be the sum of squares of residuals,
z
S=et2 zfi
-
(x-X0)5ZL -(T-
Y)coZ}
(3)
Then S/a2has a X2-distribution of the degree of freedom n-2, where G
is the standard deviation of the errors of observations. We can reject an unsuitable position fix which contains very large accidental error by the
statistical test. According to my experience, the risk factor of first kind
ç co is 5% for suitable rate of rejection which coincides with navigator's
practice 1
6]
1.2. Accuracy of Astronomical Observations at Sea.
(4')
)
)
I
position near the dead-reckoned position as follows:X 5'vi Z
+ Y
c-osZL = r:
(L=
1,2,
(1)
where -:
: number of observation. The total number of observations (n )
must be larger than 2, in order to get ship's position on the surface of the earth and an evaluation of accuracy of observations.
Cannonical equations are given by atplying the least square method
to egs (1)
lt
,t
t'tX
s
4Z Z ¿+ 'r'
O
In recent years the ship becomes larger, but the number of crew is decreasing. Consequently the importance of automatic control cannot be exaggerated. We must know the character of a ship in order to control its motion. There are many results from model experiments, but full-scale data are rather scare. Some full-scale measurements have been performed to determine drift angle or pivoting point, speed reduction under steady turning, and T , K for dynamics;
then automatic control theory is applied to the actual ship's data. A new
method of spiral test is explained in this connection.
2.1. Static Parameters
7I
The frequency distribution of residuals is shown in Fig. 3. It is not
Gaussian but of the 2-type.
For a detailed discussion the reader is referred to paper [5]
2. Ship's Motion under Steering.
We have no reliable method to measure the drift angle ( ) and the tangent
speed ( V) of an actual ship even under steady turning conditions.
A new method of PAL-test (Parallel Alignment Lines Test) and its full-scale
test are reported.
Suppose we have a transit line TL on the land, and a ship is steered in a steady turning and crosses TL at P1 and P2 as shown in Fig. 4. The bearing
of the
TL is B
, and the perpendicular to TL hasa direction
B + 900.
Then direction of the tangent at P1 isB + 90° - Gt + 90°,
and it is equal to O - , where , is the heading of the ship at P and is the drift angle of observer's position.Similarly we can have the equation at P2
02 _ = B + 90° + + 9O
Therefore we get a measured drift angle by adding two directions of
tangent.
F°2
-2
=: 28 ±(36o°,)x
i2Zn= 0,1
p
=
(e,*o2)/z
-8 ±(/800»
¡p-4-Now there are many pairs of two parallel transit lines, for instance, mile posts
for speed test. If the same steady turning is proceeded crossing these two lines,
we can get many data O, t1,O,t2. . . .., O , t. Applying the least square method,
we can determine many parameters less than n , for instance,
, V and other parameters with correction of wind and current effects. Shown in Figure 5.
If two observers are on a ship, for instance, at the bridge ( B ) and at the stern post ( s ), they can get two drift angles i ,
2 at each observation point.
Then we can easily find the pivoting point (P) and the drift angle at the center
of the gravity ( ), as shown in Figure 6.
Some data obtained from m.v. Fukae Maru are shown in Fig. 7, 8, 9, 10 and Table 1,
Fig. 11 with her particulars. The data shows that the ship
does not have a large nonlinearity in steering motion.
2.2. Dynamics
181.
)
Dynamic qualities of the shipTs steering are obtained by Kempf' s zig-zag test and Nomoto's analysis, as shown in Fig. 12.
3. Optimal Steering
(8,
9, IO, I I )When a ship comes in a narrow channel or a harbour, a quarter-master steers her under the order of a navigator, because automatic steering devices
are mainly
used for a course-keeping. The time optimal control of course-changing is given by a so-called bang-bang control, but the difficulties arise from the constraints of a control and a state variable. The control is usually given as a time function,
and not good for feedback control. The author introduces a
new algorithm
(Successi-vely Heightening Order Method) to solve the control law in a simple form. For
details of the theory the reader is referred to reference L 9}.
As for the course-keeping, it is well known that the rudder action and the centrifu-gal force under yawing can be the cause of considerable speed loss or power loss. A reasonable performance index should be determined by experiments and the optimal
control is applied to it.
3.1. Course-changing in Minimum Time (SHO-method).
Ship's dynamics are:
iw
)
= Tw +/<T
where iji : ship's heading, w : turning rate, 5 : rudder angle, u : control, A : maximum turning rate of rudder.
The optimal control with state variable constraints is given as the function of time, but it is easily contaminated by noise and inaccuracy of parameters in dynamics giving unsatisfactory results. Now we have a control law for this problem and can compose a feedback system as shown in Fig. 13. The
control law is given as follows: Definitions:
6;:=srt(4j,
6'z=srC4'1),
63=5rlc:43')
where :
2
2/2AD
+/KD
if
?2O,
?=
1S2/2
AR/K/A,
? = ?
/A
=
KAT(- 12
e2/T)(wk_
AT)61*?2)/T
if
2>
û,
3=
/A
,
T,13*
3 -
K- !+ e
Z3/T e12'3')/TJ
+(w -kS-
T)e+?2
Control Law: if ifj4S
1ifiIDtd3>0O
Íf34O
otherwise . - . . . - _ j-. (if(ífp-1
2 T -/ otherwise - - -if43=O
if..Off 4:',4o
. if4, ;:.Q
(End of Control) - -- o
Computer simulation and the actual ship's experiments by manual steering were reported, because it was not possible to construct the controller
using a computer at that time. Fig. 14,15,16.
3.2. Course-Keeping by Minimum Time Loss or Minimum Fuel Consumption (12
The performance index of a quadratic form is convenient to get the control
J
cZc'o
owhere (e)2/2
: distance loss caused by course error,: speed loss or fuel consumption loss caused by steering, O : heading error frç expected
course, : drift angle.
Dynamics for a ship are:
0=0)
=Tw
T'S
=T'p
KT'g
Then the control law is:
The experiments have been performed with m.v. Fukae Maru resulting in:
X 4 for speed lOss and X 6 for fuel consumption loss.
She ran along the speed test course under zig-zag manoeuvre to obtain time
inter-val ( T) between two crossings of transit lines. Thenthe correct0 factor
Jç__os
(e-3) dt
is calculated from recorded data to give rriean speed VThe speed loss factor k is given by the equation
V
ho C i-kr),
whereV0 is the speed without steering and X 2k. Some data are s}c in
Fig. 17, 18, 19.
The experiments have been carried out under strict regulation of engine revolution, so the parameter A for speed loss is less than for ari
ordinary ship and the parameters ) for fuel consumption loss nay be larger thar
for an ordi-nary ship.
4. Conclusion.
This reort aims at getting a total control system of a ship's
POSition
and
has provided some theory and technique for this purpose. The work, however, has not been finished completely.
The considered ship is a small training ship, and has a good
stabi1tin
stee-ring. The problem of automatic control of an unstable ship is nowunder consi-deration. The performance index must be improved by additional term
but
the experiments needed is difficult and expensive. In fact a ship
position must be automatically controlled, but the higher order control 5YStem is difficult to solve. Kalman filters must also be studied.
6
íi2T
-
11]
)
The author is grateful to Prof.ir. J. Gerritsma and Mr. W. Beukelman of the Deift University of Technology for having a chance to publish
this summary.
References:
8
.1
M. Hirota: The Logic of Position Fix; T.N.S.J., vol November 1968.
. 40, pp 71 - 78,
M. Hirota: Estimation and Representation of Ship's Position Error; Introduction of Statistical Hypothesis Test;
T.N.S.J.,, vol. 41, pp 97 102, May 1969.
Two-dimensional Joint Probability Density;
T.N.S..J., vol. 42, pp 85 - 90, October 1969rn
The Accuracy of Astro-Fix in Practice at Sea;
T.N.S.J., vol. 43, pp 121 - 130, May 1970.
Working Party : The Accuracy of Astronomical Observations
at Sea;
Journal of Royal Institute of Navigation, vol.X,
,
. 23, July 1957.
M. Hirota: A New Method of Turning Circle Test and its Full-scale Experiments; Transaction of Society of Naval Architects of Japan, vol. 129, pp. 273 - 280, May 1971.
M. Hirota: Time Optimal Control Law to Alter Ship's Course and its Manual Steering Experiments; T.S.N.A.J., vol. 131,
. 65 - 78, May 1972.
)
)
)
ti]
M. Hirota: Application of Digital Computer to Navigation, ii, Great Circle Route by Rhumb-Line; Bulletin of Kobe University ofMercan-tile Marine, Part II, vol. 15, pp 39 48, March 1968.
M. Hirota: A Digital Computer for Training the Astro_fix at Sea; Navigation (Journal of Nautical Society of Japan), no. 32, pp 83 - 87, September 1970.
[31
M. Hirota: Astronomical Navigation with a Digital Computer; Spherical Trigonometry; Transaction of Nautical Society of Japan, vol. 36, pp. 59 - 64, October 1966Fourrier Series Interpolation of Nautical Almanac Data, T.N.S.J., vol. 37, pp 29 34, May 1967.
Chebyshev Polynomical Interpolation of NautiQal Almanac Data, T.N.S.J., vol. 38, pp 87 92, Ocotber 1967.
I9]
M. Hirota: Successively Heightening Order Method for Time Optimal Control Theory and its Application to Optimal Steering of a Ship; Bulletin of Kobe University of Mercantile Marine, Vol. 20, Jan. 1973.Iio]
M. Hirota: Determination of Time Optimal Control in Time Sequence(Successively Heightening Order Method) ; Transaction of Society
of Instruments and Control Engineers of Japan, vol. 7, no. 5,
pp. 427 - 433, October 1971.
EiiI M. Hirota: Time Optimal Control of the Course of a Ship (Successively Heightening Order Method); T.S.I.C.E.J., vol. 8, no. 2, pp 242 - 250,
April 1972.
e
i: 121 M. Hirota: Performance Index for Course-Keeping (Part 1), Economic
I
Fig. 1. A line of position.
Fig. 2. Residuals of observations.
Fig. 3. Frequency distribution of
residuals of astronomical observations.
o 858 total number of observations
. 659 after rejection within group
+. 357 after rejection between groups
55 data are not shown in the figure, because their residuals exceed ±4'.
A curve shows N(-O. I , I .0)
D.R. position Y (North) o
(+ujc')
) X (East) O9.t..
O/)o58t
t
s é i-357 i1ttr--
N(-o.1, s'o) 25' -4.' -3' -2' -1' 0' 1' 2' 3' 4'drift angle. Fig. 5. PAL-test. ( a) approach perpendicular to transit line. (b) approach parallel to transit line. s
Fig. 6. Determination of nivoting
noínt and drift angle at center of gravity. c:' 't
tf\
,;-ífr6= (a)- (+XìA)i
.Ij t., t, t t3 t4 ti, 00 o» e o e e,,;o
a. X e 8j- it, tatì
4 ... ;h aj e2 9 e ... -bjOac& ... a.f
12
-Fig 7 Response of angular velocity to the rudder
8ng es. MS Fukae Maru, V0 10 kuots, arid
engine rovolution WS kept constant. Spiral
test at speed 10. 6' and at speed O 9 . 6'.
Zig-zag test at speed +9. 6'. (These marks are common with Fig. 8 and B
Fig 8 Speed reduction ratio due to steering.
MS Fukae Maru, Vo:i:1O knots, Spiral
test (PAL-test).
Fig 9 udder angle versus turnirg radius of the center of wheel hcuse. MS Fukae Maru,
VO:1O knots, Spiral test (PAL-test).
Fig. O Drift angle
(PALtest).
(a)
o 2O0 tqt
fo' 20'8
pproacii speed s.8'O.2'
(a) at steering room.
(b) at stern post.
o. O. Z 0.3 o 0.1
(b)
M.S FUKAE MARU
GENERAL ARRAtGEMEN1
(a) Side 'iew and top view.
a,
11h::
:L
f
G/
ji
3.00- o. w. L
li
'2. ¿ 5 -.---(b) Body plan.Fig. i i Flans of the experimental ship.
nl
'GT 361.71
750PS/72Orpm (Experi ment condi ti ons)ì. E i L 37.00rn
ìth
2.18 Propeller 285-3ISrpm--tA'
B 7.80m 12.8knots Speed 9. 6-10. 9 knotsJI?
410m IMJ:! 1/35.8*d
2.684m t. 'J.A 0.871m Draft d1 =2. 16, d=2.71 Deadweight 365 tonì<;ít
'D0.61m
Cb 0. 555, C =0. 647 Pitch angle 20 (fixed)Co
Computer
14
-u K' 1.0 00 o -.0-70 tS.69
-S -Ll -30 -20s Zig- zig test
o Zig-zag test ehe se
ç day + Spiral test J s
z
o .
o f _$_Io
20 30 KsT +
Fig.
13. Feedback control loop of ship's course-changing.
w
lo 20
s
Fig. 1'- Trajectories and rudder angle response of the time optimal steering obtained by thgiia computer
simulation and drawn by an XY-plotter. (Steering indices are rounded off.)
is
Figj 12 Steering indices of the experimenta' ship.
MS Fukae Maru, zig-zag test. Nondimenstonal
T=TV/Lpp, K'KLpp/V; where V(8'11') is approaching speed. T' 1.0 0.5 * . s .8 %
;
O . o o ..,,,,,7
,
N \\\
';
/ s
,.. io 15 20"v,;,
40 +S'//
TIHE OPTIMAL 5TEERINr4
INPICES T4 sec- K=o.is/sec ,=4dei/sec DISde&
(0) (b) (C) (d)
Ï
10 20\
K= 0.075. 0.15 , 0.30 (1/sec) -20 -4Ö -60 -O 30 2 , 4 S (dei/sec) D IO, 15. 25 (de) 50'\60
65 4-I
)(. 4- 0Fig.( i 5 The effect of parameter variation on the retponse of rudder angle. (Except one varied parameter,
the remainder is fixed as follows : A = 4
/sec, D15', T
5 sec, K0. 15/sec. Initial conditionsare common o= O , = O '/sec, *.s=-60. )
40
\
60(a) Starboard, Or right turn. (b) Port, or left turn.
Fi 1. Records of hand steering experiments. (Dotted lines are calculated responses. Although experimentalists steered partly by feedback according to the method (ffl), they could meet her on the new course by i
error.)
A
r
/1
=i::
!I:11
t(sec) IO 2O T= 2.5, , IO (sec) ;o'
20o
60 O. 20 -IcFig. i 7 Speed reduction ratio
I
and solid line are obtainedby steady turning,
X
by zig-zag test,A
by modified zig-zag test. These symbols are common in Fig.18 and 19.
Fig. i 8 Fuel cons::mption (Clean bottom)
Fig. i 9 Fuel ccnsumption (Fouled bottor)
300 750 ç7O0 - £SO-0O 750 goo- 750- 700-2cf
i«
o /o' 20 I:
-20°-ib'
Tt c:';tin. cc/3Osac. (Heavy oJ. P
cçrcaç. s;el 9.96 3ft rev. 2)O rp ¡it,:t angle 20.0 Cil temp. 35C r
:.
Gert1etrCeZes
/o
20'
ie1conSptiOfl, cc/3Osec.
\
roah5pe 9.05 ': 5hat rev . 7S itc ig1e 19. oil tp. 35C ted ca1i ApPrc&ch'.çeed 9.02 ;ft . 25 ZlT v;1e :o.c o ternp. 34C :-Ocad 2:0. ./0O 00 /7° O ISpiral test, y Zg-zgtest, difie zig-zag test-to L
-20
A