TECHNISCHE HOGESCHOOL DELFT
AFDELING DER SCHEEPSBOUW- EN SCHEEPVAARTKUNDEDELFT UNIVERSITY OF TECHNOLOGY
Department of Shipbuilding and Shipping
Ship. Structures Laboratory
Report No.
SSL 231a
Paper prepared J'or the ISO/TC1O8/SC2/WG2
A PROPOSAL FOR A CRITERION FOR MULTIPLE FREQUENCY
VIBRATION ACCEPTABILITY ABOARD SHIPS
by
R. Wereldsrna
A PROPOSAL FOR A CRITERION FOR MULTIPLE FRESUENCY VIBRATIONACCEPTAB1ILITY ABOARD SHIPS
by R. Wereldsma
Introduction
-1-It is very difficult in nature to find a pure harmonic time dependent phenomenon. Most of the vibrations encountered are of random nature.
As a matter of fact the pure harmonic function is man-made and is a good
help in understanding the physical problems encountered. The randomness of ship vibrations asks for a spectral analysis, where less approximations
are made with .respect to the time function.
Local resonance phenomena may influence the randomness in such a manner
that a harmonic function is closely approximated. This condition is.
ex-ploited when considerations are made concerning the 'maximum repetitive amplitude', a definition that generates problems when interpretations are
to be made. For that reason a more sophisticated approach is necessary to
avoid problems of interpretation.. This approach is found iñ the spectral
analysis of the signais encountered /1/.
Further interpretation of the power density spectrum of narrow band
rdom signals
In the appendix of this paper an outline. isgiven of the interpretation of power density spectra and the approximations by d:iscrete frequency signals, to which is referred.
For a narrow band random signai having one dominant freq.uency (to which the ISO/TC1O8/SC2/WG2 has restricted itself for the time being) the fol-lowing interpretation can be made (see Fig. I).
Since the original signal has one dominant frequency the total power
distribution over the frequency range is not interesting and a simple
'root mean square' instrument can read the amplitude of the harmonic function of the single frequency approximation and the frequency itself
is determined by the character of the signal (half the number of zero
crossings pe.r time unit). This 'root mean square' value has to be
con-verted to the maximum repetitive amplitude by means of a 'cres.t fac.tor'
and an evaluation of the amplitude against the developed single frequency criterion can take place. For pure sinusoidal signals the crest factor
amounts .to /2.
Since we refer to the maximum repetitive amplitude a .higher crest factor (e.. g.. 2 or 3 times /2) may be applied (a value which we have to agree. upon).
Multiple f requency criterion
For signals having more dominant frequencies a more complex analysis is necessary, see Fig. 2.
In the first place it is necessary to determine the shape of the power
density spectrum.
This can be done by a .tracking filter having a sufficient small bandwidth
(see appendix). In the case of Fig. 2 this bandwidth eq:uals half the
dis-tance between. two adjacent dominant frequencies. If w2 and w3 happen
to be blade freq.uencies and their multiples (e.g 10, 20 and 30 cps) the
required filter characteristic shows a bandwidth of say 5-8 cps, which is a strong requirement, not so easy to be. obtained with passive filtering.
Having this filter, the hatched area in Fig. 2 can be sensed by, a real
'mean square' instrument (Watt-meter), giving the square of the amplitude A
of signal S2 in Fig. 2.
r( \2 (s2\
ç)
tc)
< IM=
overn
nIf there should be no frequency dependency in the single frequency criterion
a more simple rule could be developed,. because M1, M2 and M3 should be equal.
in that case the signais S2 and 53 can first be summed before evaluated, which means that also the original signals can be added together without
filtering.
Our criterion can then be replaced by a somewhat stronger criterion but more simple to handle. If we assume the single frequency criterion to be constant valued, with maximum allowable single fréquency amplitude M, the 3-component signai, measured with a broad band 'root mean square' meter and factorized with the crest factor can simply be evaluated against M. This procedure is
proposed to do in the first place. When the acceptability is discussable a more advanced method needs to be applied for a precize analysis.
Finally, it is suggested, when the multiple frequency criterion developed on the basis of spectral analysis becomes feasible, to transform the single. frequency criterion based on 'maximum repetitive amplitudes' to a 'mean
square' criteriOn, which is more adap.ted for spectral evaluation, and avoids the problematic 'crest factor'.
References
Similarly the strength of S3 and S3 can be obtained, so that the discrete frequency approximation can be made and an evaluation against the developed criterion is possible.
The strength of the three signals S1, S2 and S3 has to be converted to.
'maximum repetitive amlitudes' by means of a crest factor (see page 1). in this way we obtain SI, S2 and S3.
Now we have to design a method to evaluate the 3 simultaneously occurring signals against the single frequency criterion as developed.
Under the assumption that squared linear superposition is applicable (which has to be verified; in the first place it is assumed that no cross correla-.
tion exists between the various frequencies of the signal, in the second place the human sensitivity needs to allow this superposition /2/), the
followingrule can be developed (see Fig. 3).
For each of the components a single frequency evaluation can be made.
If the strength of one of the components happens to be the maximum allowable value, the other components are not acceptable (the total 'volume of
accept-ance' is consumed)
The 'volume of acceptance' determines the maximum contribution of each of the components relative to each other, which results in the following rule:
or in generai:
/1/ Wereldsmna, R. 'Ship vibration evaluation through narrow band spectral
analysis'. SSL-report No. 231, Sept. 1980.
/2/ Fothergill, L.C. and Griffin, M.J. 'The evaluation of discomfort pro-duced by multiple frequency whole-body vibration',
PO.S. of narrow
band random
signal with one
dominant
frequency
Wdidealized I.PO.S.
A
single frequency
representation
having the same
"mean square value"
as the original
narrow band
random signal,
w
Fi:g..1 Narrow band random signal with
one dominant frequency
apprdximated by a singLe
s
s
Li w V)PD.S. of multiple
narrow band random
signal having dominant
frequencies w
I- w- w
2 5idealized I.PO.S.
w.
discrete frequency
approximation
of the
PD.S.
Fig.2 Multiple narrow band random
signal approximation
SI
single frequency criterion for
maximum repetitive amplitudes.
maximum acceptabLe repetitive
amplitudes for single frequency
signaLs
criterion for acceptance
w3
w-maximum repetitive amplitudes
of the three composing discrete
sinusoidal signal components.
2 2
M1
M3
=
al
-Appendix
THE POWER DENSITY SPECTRUM AND THE BEAM LOAD ANALOGY
The power density spectrum reflects the distribution of the 'power' located in an infinitely small frequency range.
. . r (amplitude)2
The dimension of the. spectrum equals i .
. [Unit of frequency
A similar presentation can be found in the distributed loading of a
canti-levered beam. .
r
-IThe load of the beam is expressed as I . orce and can be interpreted
.
L'"
of leng.thjas a load density along the beam length (normally. named 'distributed loadt).
In Fig. AI.a a powe.r density spectrum (P.D.S.) and beam load distribution
q(x) are shown. To obtain the shearing force in the beam the distributed load has to be integrated along the beam axis.
The shearforce in the beam equals the integrated distributed load from the free end to the point of observation.
Similarly the power density spectrum can be integrated along the frequency axis and results in the 'power spectrum'. (This word may lead to
misunder-standing becaus,e in the professional language 'power spectrum' often means
'power density spectrum'. Therefore the wording 'integrated power density spectrum' is used in order to avoid misunderstanding (I.P.D.S.)).
This function represents the total 'power' A2 which is located in the fre-quency range from O Hz through the frefre-quency of observation. In Fig.
ALb
both integrated functions are presented. The shearing force is the analogyfor the power.
The integration over the full beam length gives us the total shearing force felt by the support of the beam. The integration of the total frequency range gives us thetotal 'power' of the signal as dissipated in a 1Q
re-sistor.
It is this integrated power density spectrum that gives us the possibility to make interpretations with deterministic pure harmonic signals, similarly as the integrated load distribution of the beam gives us the possibility to make interpretations with deterministic forces.
The best single force approximation for the beam loading has a magnitude
that generates the same shea-ring force at the beam support, and is located
in the centre of gravity of the,distributed beam load.
The best single frequency approximation for the signal under consideration
has an amplitude gene-rating the same power in a IQ resistor and has a f
re-quency that is located in the 'centre of- gravity' of the power density
spec-trum, see Fig. Al.c. The amplitude of the single frequency approximation is known as the 'root-mean-square-value' of the total signal.. Going back to the presentation in densities we obtain the patterns as shown in Fig. AI.d.
In the beam load analogy the load distribution of the single force approxima-tion results in an impulse of infinite height having an area equal to the force F, being the integration of the original distributed load q(x). In the spectral presentation the single frequency approximation results in an impulsive spectrum, having infinite height and infinitely small bandwidth having an area equal to the total power of the.harmonic signal.
It can be concluded that not very much resemblance exists between the original power density spectrum and the approximated single frequency power density
spectrum. A better approximation an be obtained by 3 harmonic deterministic
signals.
I
In Fig. A2.a the resulting integrated power densit spectrum and the power density spectrum are shown as well as for the vibration signal as well as fo-r
the beam analogy. Comparison with the original power density spectrum and load distribution respectively is not very satisfactory either.
a2
-Amuitiple harmonic approximation is shown in Fig. A2.b. It can be concluded that the approximation is improved. For an infinite number of harmonic sig-nais having an infinitely small amplitude the apprOximation of the integrated power density spectrum and the power density spectrum itself is accurate and
in the limit exact.
An alternative for the approximation of a broad banded random signal is the
breakdown into a number of narrow band random signals (N..B.R.S.).
In Fig. A3.a an illustration is given of this approach and it may be concluded that a closer approximation is possible than for a multiple pure harmonic
ap-proximation. S
In the limit again an infinite large number of infinite small banded random signals having an infinite small power gives an exact representation of the original broad banded random signal, (see Fig. A3.b). .
Another element in the analogy is found in the determination of the power density spectrum and its sensitivity for the bandwidth of the filters applied. Accoring Fig. 30 and Fig. 31 of report No.. 231 of the Ship Structures
Labor-atory ) there exists a strong sensitivity for the applied bandwidth of the
equipint that measures the power density spectrum.
In the beam analogy this effect can be illustrated easily and is shown in Fig. A4 where for the original beam load distribution 3 types of sensors are used to measure this distribution. It can be noticed that a smoothing effect occurs when the length of the sensor is increased.
The 'characteristic length' of the beam loading and the length of the sensor are related to each other for an accurate measurement. The length of the sensor determines the solvability of the instrument. Strong gradients in the
loading are only properly sensed by 'short' sensors, Similar effects occur when the P.D.S. of a signal is measured with a filter having a variable
band-width.
Therefore, before measuring a power density spectrum it is necessary from a point of view of accuracy to know the character of the spectrum (which is unknown!) and efforts should be paid, in a specific field of applicàtion as we are concerned with, to learn by advanced instrumentation the character of the spectra, generally encountered, so that the required bandwidth for more
.simple equipment can be determined in connection with the requ.ired
solvabil-ity. The bandwidth of the filters applied should be at maximum one half of the
bandwidth of the narrow banded signals,, a subject to be standardized. Finally, the beam analogy will be applied for a 'multiple frequency' signal
composed of e.g. three narrow band random signals. In Fig. A5 the discrete frequency approximation of the original signal is shown together with the beam load analogy.
For these narrow band random signals the approximation by the three discrete frequencies results in an acceptable reproduction of the power density
spec-trum and can be handled simultaneously in a similar way as the single fre-quency approach by the 'maximum repetitive amplitude' method.
R. Wereldsma: 'Ship vibration evaluation through narrow band spectral
fi.
LI w U)fig. Ala
o"centre of frequency"
Po.s.
power density
spectrum
I.PD.s..
integrated power
density spectrum
best single harmonic
approximation
for equal total
power
w
power density
spectrum of
best single
harmonic
approximation
fig.Alc
fig. Aid
oof Length q(x)
load
per unit
411111
Xfqdx
fig. Alb
Fbest single force
approximation
FJq Cx) dx
over beam length
Fig. Al Simple single frequency approximation of continuoUs spectrum
and its beam
analogy.
shearing
force
X
-best single force
approximation
for equal
shearing force
4
Original Pos;
Ipower density
spectrum
tIu,
I I Ii
I i i I I I I I .1 I Ii
I. I I I I II
I1,
A;
original
I. ROS.and 3 step appr.
I
i'
e1,;/J\3I3m0
app
mulliple harmonic
appr. of
J, PO.S.A2..a
ifli
i i iIItI
T I Iliii
i Imultiple
:i
harmonic
l'li
i i1approf
Ii'RO.S.
'i
I I A2. b Fshearing force
and 3 step appc
I I I
ii.
IJI
-
j
3 force appr.
/
i 'iof origiruL
,
N_i
'¼toad
s. s. s.multiple step
appr. of
shearing force
I'
I I I I i iIII
I I .1 I I I i::
i I I i I I i i i iniuttip(e
I i
t_i
i ilit
ii force appr.
I
I'
ti l I
i of original
/
'i
i! beam Load
Fig.A2 Improved multiple frequency approximation
of continuous spectrum
and its beam analogy.
e.
--
originaL
/'
integrated
appr. beam Loading
i i i
yand
(shearing force)its
i I (.°-
I I X-
1!n!
i____oppr.
P.DS i'I
___originaL
POS.
loading
L
original
beamAillili
X-
Xpower
density
spectrum
of
composing
narrow
band
random
signais.
Fig.
A3a
Power
density
spectrum
approximated
bythree
narrow
bond
random
signals
and
its
beamload
analogy.
i
'u
appr.distribution
of
beamloading
j
N.
/
'
original
beam Load inw
u,
power
density
spectrum
approximate
representation
.bynarrow
band
random
signaLs:havin:g
bandwidth
and
power
equaLto
hatched
area
e.g.Fig.
A3
b MuLtipLenarrow
band
random
signal
approximation
of
broadband
random
signaL
beam Load
analogy
X
-f
x.
approximation
of
beam Loadby
constant
distributed
Loadingsranging
oVerx11-x1
and
having
an
integrated
weight
equal
to
originaL
beam load
distribution
X -'
1''
dimensions of 3 types
of sensors measuring
2 '
the beam loading
(comparable to pressure
3 '
pick-ups in the
2-dimensional. case)
sensor moving along the beam axis
and sensing the hatched area,
devided by the sensor Length.
X
-measured beam load,
distorted by sensor length
Fig. A4 Effect of filter bandwidth
A
A2 3
"2
appr. LPD.S.
approximate P.O.S. impulsive of
nature,due to discrete frequency
approximations.
i
I appr. shear
force function
shear force
function
distributed
beam loading
X
-approximate beam loading,
impulsive of nature, due to
-