1epot Na
327Septsmber 1971
LABORATORIUM VOOR
SCHEEPSBOUWKUNDE
TECHNISCHE HOGESCHOOL DELFT
r
1
Sorni Notiq on the
to.Pi1ot
Of an Unatable Ship
by
P. ICoyaina
s,
Contents
Suamay
List of figures
1
Fundamental conaiderationa
(i) The root locus method
(2) The Nyquist Criterion
2
Some commentn about the firnt order model
An example of the auto-pilot of an unstable ship
Conclusion
List of figuree :
The bloolc digraan of the proportional control
Root bai
L) - plan
(.].+kGJ - planeInspioting circuit
6,
NvqUit
plot of a tabl* shipNyquist plot of an unstable ship Root locus of the first order model Nyquiet plot of the first order model
Diff.rnoe between th. 1st-order model and the 2nd-order mode]
Phase oharacteristios of the ntabl and unstable let-order model
Open loop oharaoteriatios at full load and normal speed condition
Rats control filter
Closed loop characteristics at full load and normal speed condition Gain characteristics of disturbance to ruddar angle
16, Open loop char oteristios at full load and half aped oondition
Closed loop oharnot.ristios at full load and. half spud condition Open loop ohiracteristios at ballast and normal speed condition 19, Closed loop oharaot.ritios at ballast and. normal speed condition
The objective of this note in to mike the problems of the deBirçn
of auto-pilot of an unstable hip clear. The main conclusion io
that, though it in true
that th.
more the ship becomes unitabXe,th. mon, difficult it is to design an auto'pilot, there in no fundamental difference betw..n a stable and unstable ship in
designing an auto-pilot.
Aft.r acm. general and fundamental considerations the difference
between & stabl and unstable ship, an
exemple of the auto-pilot design of an unstable super tanker will be given end th. requirementsfor the auto-pilot of that kind of ships will be dtouaned in the
section 3.
s
1. Fundamental ooped.ratione
In this ieotion, we will oonsider the proportion.], control to see if ther, i. some fundamental differenee between atable and unstable chipa from the viewpoint of control.
The transfer fiastion of th.
ship, G
(ätoçb),
La given as follows s
-i (T1sl) (T2.+1) (i)
In this report, w. use the coordinate system in whioh th. starboard
rudd.r angle (positive) introduces th. starboard turn (positiv.) in
order to make our ooneideration on the control eyRtam eauiaz'..So, the parameters of Eq(l) have the earn. eigne as they had in the original
Nomoto
paper (1) , while K of a stable ship is uaually taken negative.Thsorsticai.ly it is possible to have negative T2 or T3, but fortunately we did not m..t such a Rtrange ship. Consequently, we eons ider them
positive and acuse further that is larger than T2.
Th block diagram of th.
proportional control 'ystem la given in Pig.],
where k is a proportional ratio, e.g.
81kk(4i1..()
(2)The transfer function of that closed loopsystem is k G (s)
(3)
1+G5(o)
stable ship
unstable ship
K +
T1 +
-T2 + +
s
3-(5)
and. thi obaractarietic eQuation is
1 + kG5 (e) O
(4)
If the r.a.l parts of all characteristic roots are negativ., the
8y8'tem L etabi., and if any of them i. positive, the system is
unstable.
There ars ssvera.l methods to s.. whether .
(4)
baa & root with positive real part or not without solving the oharaoteristio equation,Vi will
25e
the root locus method and the Nyquiet Criterion whioh will give us the best appx.oiation of the diff.reno. between the stable and.unstable ship.
(1) Tb. Root Locus Method
This method treats the closed loop iyatem ea a function of the
proportion.]. ratio k, and plote the
roots
in the complex plane when the value of k i. varied from O to oO (this plotting can b. done withoutsolving the characteristic equation). So ve can sa explicity whether the real parte of the characteristic roots are positiv, or not. In
addition
information about the damping ond quickness of .'eeponai
characteristicsof th. system is also obtained.
The 'oot loot of a stable
and unstable ship nr. oompared in Fig. 2.
In both oases, thesystem has three roots. One of them ntarts from
and ends at (_i/T,.o) a. the
proportional ratto, k,
increasesits valu, from O too0. The other two are starting from (4/P1,o) and
(0,0), meeting inch other at a certain point between those points and are going to infinity making a pair of oonugats roots.
As 4./P1 is positive in case of an unstable ship, those roots have
positive real parts
while k remains in a certain .maU region. Thit
whenwe mak, k large enough, the real parta of those roots are pulled in the negative direction affected by the position of .'l/T3. The asymptotic
line of the bei is $
(-4
-
+)
so, if -lIT1 is not so large in comparison with the difference of 1/P2
end l/T3, the system can b. made atable by proportional control only
It in very int.reiting to see that ther, io no fundamental difference
in the pattern of root bei of atable and unstable ships. $o,in fact
there is no significant differenc, in designing an auto-pilot of slightly
stable or slightly unstable ships.
(2) The l4yiuist Criterion
The phenomena w have obaerved on th. root bous method can aleo be
explained by the Nyquist Criterion.
W. are int.r.t.d, to know whether the characteristic equation ha.. a zero
with real part or . If it has a zero in the right half plane of the complex plane, Eq.
(5)
hau a. pole in the right half plane, and, th. system ii unstable.Let u. assume that zeros
and
polea of Eq. (4) are &tøtributed as shown in Pig. 3 and ocnsider th. mapping of oloued oirouits, Q1 and Q, in thte 3 -
ptan into the (].+kÇ) plane (Pig. 4).Au we can easily see, the closed circuit,
Ql
which turns around a. zero in. cloak vi., direction i. mapped into the 11+kGsJ - plane so sa to turn around the origin in clock wine direction Q.,', because the argument ischanged with
an
amount of 21t. and Q2, turning around a pole, is mappedinto Q which turns around. the origin in counter obook wise direotlon.
When the closed circuit contains two zeros, the mapping of that circuit turn, around the origin twics, and
80
on.Therefore,
if we ohose the closed circuit in the
Cu) - plane
in. such a way as shown in Pig. 5, and make R infinity io as to cover th. wholeright halt plame, w. can se. the difference of number of zeros and poles
in the right half plan., e.g.
I- Z - P
or Z-N+P
(6)where i
N z number of turn, in oboek wise
Z z number of zeros in R4H.P.
P z number of Doles in RILPI
-5
s
Usually we know P, because the poles of the characteristic equation are the name as those of tho open loop transfer function, kG, so, we canfind Z by counting the number of
turns, N,
in the [1+kG5) - plane.If we consider the mapping into the
L1cGI - plane
instead of the[1+kGJ
plan, we ehould count th. number of turns around the point (-i,o)
instead of the origin.As th. mapping of A-SB (s o'.joo) of Fig. 5 into the
tk%3 - plane
coincides with the open loop frequency respons. of the system,
and
B-'C -D is mapped into the origin, and the mapping of D-3 A is anmetric
with that of A-+ B, we only need to consider the mapping of A-+B in order
to know N. It means that we may design the auto-pilot of an unstable
ship in the usual way with the Bode Diagram, e g., if the phase lag at the
frequency where1kGfl.
i
je less than 180°, the system is stable Let us return to the problem of auto-pilot.The vsotor loou (frequency response ) of a stable system is shown in Pig.6
In that figure N = P 0, so Z = O, and, the system is alwars stable.
In caes of an unstable ahip, we have a negative time constant, T1, so the
characteristic eqw3tion han a pole, -itP1, in the right half plane (=i).
The vector locus turns around (-i, o) one time in clock wise direction
when k is not large (rig.
7 (a)),
so, N - 1. Therefore, we get Z - 2 byEq.(6); it means that we have two zeros in the
right half
plan. and the system is unstable. But when we make the proportional ratio, k, largeenough, the vector locus turns around (-1,0) one time in counter clock
wies direction (N.-1). Therefore, Z - N + P - O in this oase,and the system is st9bl.. The above is the alternative explanation of what we
observed on th. root locus method.
The largest difference between the stable and unstable ships is the phase characteristic of the G. The phase lag at the very low
frequency is _900
for a stable shipand -
270e for an unstable ship, and tends to - 380°when the frequency goen to large values in both oases. It is sure that the
difference in the phas. lag oharaot.ristio makes it more difrioult to
design art auto-pilot for an unstable ahip than for a stable one, but we
2.
Some Comments about the First Order r':odel
Usually1 the manoeuvrability of ships is orten detoribed by the first order system model. But, from the point of view of the
deign
of en auto-pilot, the first order system cannot provide us with.noug'h information. The root locus and the vector locus of the proportional control of the unstable first order ayat.m are ahowii
in the Figs. 8 and
9.
In Fig. 9, P is always 1(-l/T, iko) end N is always 1, so Z is alwaya 2. It means
the unstable first
order system cannotbe made
stable bythe proportional control only. Certinly,
the
same thing can be seen by the root locus method too.Sometimes it is aid that we can not use the first order system for the
design of n auto-pilot. But, it is not a big problem to aiow whether
it io possible to get a stable system by proportional control only.
Because we
ned s rather
big amount of rate control in both oasesanyway to have a satinfaotory..damping of the system. Therefore, to use the first order model means that we will
over-.etimate the necessary
amount of rate control; this can be shown clearly by the Bode Diagram
(Fig. io).
Although the difference of the phase charaoteriatic; are not small,
Lt in not correct to say that the first order system does not give
any information for the deiign of an auto-pilot; we can get rather
significant information from
the first order model even if it is notenough.
In spit. of the above discussion, the parameters of the linearized
first order model, which are determined by
fitting
the results of a Zig-Zag trial with large rudder angle, ta of no use for the designof an auto-pilot. Because the unstable ship could be described as a
stable ship by th. non-linear characteristics. In that ease, the
phase
characteristic is fatally mistanken as shown in Fig. U.3.
n examl. of the Auto-Pilot of an Unstable Shin
Por example, let us cnaider the auto-pilot of the 2O0.000DW1
tanker, the teering characteristics of which are íiven in
2J
full load
ballast
-
0.0434/seo 0.0471/seo- 269.3
eso
60 seo
T2
9.3
sec 3 seoT3
20.0
seo
6 seo
Table I The steering oharacteristlea of
a 200.000 DW tanker
(these values are recalculated from 23
according
to thedifference
ofderinition)
The
Bad. Diagram
of thefull load condition i. shown in
Pig.
3.2.
As it is known, the open loop oharlcteristjos must have
a
phase marginof 4flv
6° to
make the damping of thc closed system satisfactory,
So, we ha
to introduce rate control to improve the phase
owracteritic.
That abe done easily by the following filter.
Tb. Bad. diagram of that filter is shown In Pig. 13(a)
Eep.si.j]y wh.Ti we design auto-pilots of ships,
we haveto be careful
of the followl4g points
t
(i) Not to maki $1v. rudder
aij1. large against the disturbano, of
waves,
b.oaus. th.
itIon of
'&.r at high frequency
isof no use to
4p4y the pirforinanes
of the syst.aand introduce rather larga
Increase ot resUtae.
3)As w. ean se, in Fig. 15, the rate control
has a tena.noy toino'ers. the ruddr angle at high
frequency, so wehad bettör to out it ott by a low pass filter if possible.
(2) The ohørot.ri.io of ship. change according to the loading
condition and ship apead. So, it the filter of rate oontrol La
fixed to fit the normal spied and full load condition, it may
not b. suitable for th. other conditions Therefore, e hav, to detçinnin. if it is necessary to change the settings of an auto.. pilot aecording to tb. condition1
This
problem will be moresevera for the unstable ship because of ita delicate phase
oharaot.rirt io..
In the first place, we will cone idar the rate control filter for the normal speed and rul3. load condition. The filter should have a
ph&ai lead of about 4Q0 and high frequency ut oharacteristios.
Let us try the filter of Pig. 13b that has 450 of phase liad and the relations of TD,
'
L2 '
T» - 82 TL2
Combining that filter (T»I4OOao) with th. ship, we get the open loop characteristics of Pig. 12 (broken line). Prom that figure we can ice that the phane margin i more than 40° if the value of the
proportional ratio varied between lk4.
Oerbainly, the rutar we used
here can
not be said the optimal one,but if we make T» smaller, we need to increase k and the ration of the rudder angle and the disturbance of waves (S/n) will be inorsaed, and
if we choose a larger T», the phase margin is not enough. that filter oar' be said rathergood.
The closed loop eheraot.risito. are shown in Fig. 14 according to the
various proportional ratio and the 8/n ratio is not so larg. at the
frequency range of wave.. (hg. i).
Next, let us oonnidar the situation when th. ship speed is reduced to
helf th. normal speed.
If we aawne that nondimansional parameter, are not changed, the time
constante ars increased by a faotor two and the proportionality constant )C is reduced to ht1f.
The open ioop
ohernoteristios ere shown in Fig.
16 with the earne rateThe yatem has enoigb phase margin for 2 <k <12. It means that the same filter can be used in r'ther wide speed range if we
Ret k:
between 2 arid 4, or we need to increase sliitly the netting of proportional ratiowhen the spead of ship is riduced. The olosed loop characteristics are
shown in Pig 17
The open loop oharaoteri.tioe of the ballast condition with the sama
filter are shown in Pig. 18 and the olosed loop, in Pig 19. The suitabLe range of k is lees than 1.6 by Pig. 18. This valus is not sufficient,
because a too small k-va1u soastimse introduoss a rather large constant error of course against the constant disturbano. like winds.
It insane that th. filter for the ful]. load condition is no mor appropriate for the ballast condition, end we have to change not only the proportional ratio but also the setting of rate control filter.
4. Cono lue ion
From the ab ve discussion, we may iay that if it is po8ible to use
the rate control filter with a lar'e time constant (TD), it ja not so
difficult to design an auto-pilot for an unstable ship as long as as the value of i/T1 remains small. When -1/Ti is not small, we will meet
the difficulty of larg. rudder angles in rough seas because th. frequency
range of large rate control will coincide with that of waves.
To obtain the purely linear oharacterietioB of ships is the moat important
feature when we design an auto-pilot. A linearized medel of the dominant
non-linear characteristics may work
weil for the prediction of the othernenoeuvrea but does not work at ail for auto-pilot design.
-References
Nomoto, K.
"a1ysis of K.mpf'n Standard Manoeuvre Test and Proposed Steering
Quality Indicei"
First Symposium on :;hip Mano.uvrability, 1960
Glansdorp, C4C.
"Simulation of Pull-Scale Renulti of Manosuvring Trials of a
200.000 tons Tanker with a Simple Mathematical Model"
Report No 301, Shipbuilding Laboratory of P.R. Deift, March 1971
Koyema, P.
"On the Optimum Automatic Control ystom of ships at Sea"
J.SIN.A. Japan, Vol. 122,1967/ Selected pnper from J.,N.A. Japan
Ô
Gs(s
Fig. I
The block diagram o
the proportional
control
(a)
stable ship
(b) unstable
ship
Fig. 5
Inepeotiug circuit
(a)
(b)
Fig. 6
Nyquist plot of a atable ship
(a)
smaLL k(b)
Large kW-bO
k=oø
k=0
=0
0
-1/1
k=00
Fig. 8
Root locus of
the first order model
0db
-180°
-270°
phase
required phase margin
Fig. 9 Nyquist plot of
the first order model
Fig. 10
Difference between the 1-st order
model
and. the 2-nd order model
gain
ist -order
g
.
_90ç
Logu
Fig. 11
Phase characteristics or the atable
and,
unstable 1.-et order model
go i n
0db
00