15 IMJ 1974 Lab.
y. Scheepsbouwkun&
ec nisc Ï HpgesdJ
DIGITAL 'ILTEiiIG TCHiIU 11 TML. A1i'IDelfi
ZIG-ZA MA1OEUWRE PLTS.
y K. 1.omoto.
fi. n
PREFACE : In the analysis of zig-z.g tests it Is often
desirable to eLiminate the noise
involved
i.n therecorded ship motion.
It is particu1arly so at, the
phase-plane analysis to ob.t.ifl the charat.erStic parameters of a ship. in question, includiflg the. non-linear element, since
in such analysis a
numericl diffeientiatiOfl of
thenieasured z-rate ta needed.At the Zagreb meettr
in
97O the writer presented anumerical technique for this urpose, i.e. "seveb points stnoothin
etnp3,oying 3rd. power ppl1n9ialS.
This did apparently wori well
bt later lt beBrrze
Iear that it has a risk of
-introduing high
freq.ueflcy noisè. cC. RINA paper
by C1rke and Glansdor.p, 1972.
The present note relates to a new
nwtéril filtering
teca-nique to smoothenthe measured. yaw-rate and
at the sanie timeto give ita differentiation
with m.inimuìn possible noise. 2. LOW-PA S.S FILTER : First. consider the ideal low-passfilter whose frequency response is:
..1L
i
-J
tamplitude
ratiof
,'=
»
fo
Wo
t,o
-
phase
._foi'
4,1/The wighting function of this
filter, i.e.,
i.tSrepOflBe to an
unit impul8e input applied at. t = O, is
S/ri.,
'c-C
This is shown in Fig. 1. It ShoUld be notedthat the
response
to an impulse at t = O does
exist
even at t < This taImpossible in any physical
filter, inat-her words, the
ideal low-pass filter is not "physically realizable".In computation, however, the ideal low-pass filtering is
posstb..e. To make .a oonvolutiofl integral of
the weighting
function and any given input signa].
-iS
the answer. That is,t
T-
r) dc
where
»t/denotes the given inut.as
a. function of time:
ad
- t)
the filtered signai.
bliotheek y
OnderafdeI . -.-- . eepsbouwkunde
niçh ,Hoaeschoo
A practical difficulty is introduced here from th.e poor ;convergence. (or decay) of &-) , which is remedied b.y modifyiag
the ideal ¿.V) by multiplying with "Lag-window" function
'*
:tL is to be adjusted depending upon situations. According to
.the experience tL= 30 sec. is more than adequate for actual ships with time interval for computation (sampling time) of 0.5
to i sec.; even t= 10 did apparently work very welL.
Next question is what is the best choice with
&c
,. thecut-off frequency. If u)a is too high, much iloise will remain
and if too low, quick change of ship motion may be distorted, which usually happens at rudder execution. k compromise kiss to
be done any way. We will come back to this point later.
3. DIFFREMIAL FILTER : The same principle as the low-pass filter can be used to design a numerical differentiating filter to obtain the yaw acceleration from the measured yaw-rate with minimum possible noise.
The frequency response of the ideal filter in this case is:
F)
for'w
o
"C-,4
DP4//
4-,.T.he resulting weighting function is
Z-'Z
C'OS -ct
S/'v
,Tie lag window,
±(*
is again called for to improve the convergence.
j;L1.. CiWIC IN CUT-OFF FREQUENCY ; Fig. 2. shows the effect of the cut-off frequency of
:the low-pass and differential plus low-pass' filters mentioned previous1y, upon the. phase plane tra jectôry for a yaw-rate zig-zag manoeuvre. This is not an experimental result but a computed ,one and makes it clear how the cut-off frequency d.istort.s the
t.rajectory. The, mathematical model employed is shown in the
:fjgure.
The result tells us that the cut-off frequency less than :;O.l c/s. will introduce an appreciable distortion at-the yaw
acceleration peaks, which results in. an erroneous evaluation of
and
.
On the other bend an experimental result of an actual -tanker as shown on Fig. 3 indi.te6 still some noise with the cut-off frequency of 0.167 c/s. The ship si.z.e is much the same as the mathematical model.
The optimum cut-Off frequency depends considerablY upon the size and speed of a ship in question and should be kept for further survey, but 0.07 to 0.1 c/s. may be a reasonable cholse
for large tankers fully loaied.
Finally an emphasis should e laid. upon the adwantage of the zig-zag manoeuvres of an established limit cycle. For this kind of motion it is possible to get a nice phase-Plane portrait by averaging over a feví cycles and with a relativelY high cut-off frequency. From this point of vie& the yaw-rate zig-zag test as proposed at 13th ITTC is perhaps most suitable for this sort