ARCHIEF
Lab.
y.
Scheepsbouwkunde
Technische Hogeschool
Deift
DETERNATION
OFWAVE BENDING MONTS FOR SHIP DESIGN
by Rutger Bennet
August 1963
(Paper to be presented to the Scandinavian Ship Technical Conference, Abo, Finland, Lth October, 1963)
INTRODUCTION
According to the rules of the Classification Societies,
the longitudinal bending moment is decisive for the midship
section modulus, and hence also to a great extent for the
steel weight This bending moment consists of several rather
independent parts, only one of which - the still water moment
-can be calculated exactly. However, this part often has only
a minor influence on the design. The remaining parts are
often given the common name wave bending moment, which
con-tains both the moment really caused by the waves, and the
added effect of local loads of various kind: slamming,
vibra-tions, temperature variations and others.
For design, part of this moment is calculated according
to a method which in principle has been standardised for a
long time, and the rest of the load is taken care of by a
nominal permitted stress. The notion about what is covered by respectively the bending moment and the stress is vague, but the ratio between them is given a value, the section
modulus, which has been found adequate from experience. It is probable that the scantlings, determined in this way, are
on the whole suitable for ships of those types and sizes
where we have sufficient experience. The difficulties have
started now we begin to build ships outside this experience,
2
In order to learn the real loads on a ship hull in waves,
a great number of measurements have been carried out, both
full scale and on models. For model tests to have any meaning,
it must be possible to convert their results to the
corre-sponding ship. The greatest problems in the comparison between
model and ship are the ratio of bending moment to stress, and
the generation of a realistic wave system in the tank, which
can be directly compared with the conditions during the full
scale tests.
In the ships the stress variation is usually measured
at a point of the hull, where there are no stress
concentra-tions or other disturbances. Calibration of the hull at the
gage 10-cation, i.e. recording of the stress caused by a known
change of static bending moment, has been made in several
cases, and generally the agreement with theory has been
satisfactory. Even if considerable doubts have been expressed
by some investigators, it is believed that the so called
simple beam theory will give a sufficiently accurate value
of the moment from the measured stress. At least this is so
as long as the stress is caused by a static or semistatic
moment, varying only at the slow rate of the waves. The
action of high frequency, dynamic loads is more doubtful,
which also depends on the fact that the instruments used often are too slow to record such variations. At the Swedish
tests this has been considered rather an advantage, as the
primary purpose has been to investigate the 7pui'e,
seniista-tic wave bending moment. The statisseniista-tical pattern which has
been suggested for the variation cannot be expected to be
valid for the vibrations, caused by slamming or 7whipping.
These stresses must be treated separately and added to the design afterwrds. They are not mentioned in this paper.
u
3
The next problem is the wave system. The statistical wave theories have advanced very fast, and many equations have been proposed for wave spectra. During the second half
of 1962 a project was initiated by W. Pierson at New York
University, which is aimed at creating order in this confusion.
A great number of wave records, gathered during several years
on the British weather ships by means of the Tucker Wave
Re-corder, are treated in a computer. Two reports are already
published, containing about 2+00 such spectra, together with
information on corresponding wind speed, /1/e It is thus
possible to find out what wind generated spectra really look
like, and a comparison with one of the available theoretica.1
equations, which was recently performed at Webb Institute of
Naval Architecture, N.Y., has given interesting results which
will be mentioned later in this paper, /2/.
With a great many results of full scale stress
measure-ments available, several model test series, and a number of
recorded wave spectra, it is now essential to find methods of
using all this knowledge0 The difficulty of getting agreement
between model and full scale tests is mainly caused by the
problem of creating comparable wave conditions. Even if the
superposition principle is valid, and typical spectra are known, such a comparison is difficult because so few full
scale tests have been accompanied by wave measurements. For
the model both the moment and the wave system are known, for
the ship only the moment and quite generally which wave
spectra may be expected, but not simultaneous values of these two variables. The only possibility must therefore be a
sta-tistical treatment, similar for ship and model, and then the results can be compared on equal probability levels in both
-
2-cases, Such calculations have been performed independently
at the Swedish Shipbuilding Research Foundation - Chalmers
Technical University, Göteborg, and at Webb Institute of
Naval Architecture, Glen Cove,
N.Y.
Naturally these fewattempts cannot be regarded as any final answer to all
questions, but the results are certainly promising. It ought
to be possible very soon to predict from model tests the
wave bending moments acting on a ship during its service
life. This may in the first place have importance for ships
of unusual form or size, where experience is now lacking. It may, however, later become a normal procedure for every new ship, just as it is to determine the required horse po-wer from model tests. A similar analysis can also be applied
to systematic investigations of the influence of various
hull parameters on bending moment, motions or speed in waves. Regarding the design of the structure, it iaust of course
be pointed out that everything is far from finished with the
still water and wave bending moment. First, a number of other loads have to be added, and further it is not clear what is really decisive for the structural strength. On the other hand
the wave bending moment is certainly a major part of the
va-rying load, and the question of the importance of fatigue can
never be solved unless the long term distribution of this moment is known /3/.,
ANALYSIS OF FULL SCALE TESTS
Stress measurements at sea have been statistically
ana-lysed since Jasperas classical paper to SNAME in 1956. As in
so many other cases, Jasper had a predecessor in Ltcmdr Roop, USN, who more or less intuitively suggested a statistical
5
approach already in l9LO /L/. Roop found empirically a
distri-bution of wave stresses, which later proved to be very close
to the Rayleigh distribution. This has now been found as a
result of the assumption that the waves can be regarded as
a normal, stochastic process, and it is the generally
accep-ted basis of all statistical analysis of ship response to
waves . Incidentally, Roop even went a step further and indi-cated what long term distribution might be expected, if a
great number of short term distributions were added.
The agreement between different investigators in this
area seems to end with the Rayleigh distribution, and there is great uncertainty as to how to continue the analysis of
data. One reason for this is uncertainty as to what is the
actual purpose. Many authors consider it necessary to find
the greatest value of wave stress, or bending moment, which can ever occur. Attempts have been made to define the most
severe condition for a certain ship length and form, then
to extrapolate test results to this condition. It is, however,
not certain this is possible, or even desirable. The wave
moment is depending on so many different factors, all of
which have a wide statistical scatter, that it is practically impossible to find the combination giving the absolute
ex-treme value. An important factor, for instance, is the total
time the ship will encounter every such combination, In
gene-ral it must be assumed that this time is shorter, the more
severe the wave condition becomes, which means a lower
proba-bility of high bending moments. These are two factors of
directly opposite influence. Further, there is the very little
known influence of how the ship is handled, what somebody has called the captain's spectrum. It is undoubtedly a fact,
6
that the greatest wave stress recorded in many ships has
occurred in 7- Beau'ort, which is not so very extreme0
If the wave system is represented by a spectrum, this
has three important parameters: general shape, area f'o-quency range, especielly the fref'o-quency of the peak0 At least
for higher wind spueds, the shape does not seem to vary too
much. The area: which determines the wave heights2 may, how-ever, vary between very de limits, even for constant wind
speed, and this is also the case for the peak frequency. The
result is that the wave bending moment may be of widely
different order of magnitude at different occasions in wave
systems, which are visually judged to be identical, either
they are defined by wind speed or by sorne observed wave
height0 To be sure of having measured the highest possible
value ±n each condition so defined, it may be necessary to
carry cn the tests for the enti:'e serioo life of the ship0
The only possibility of getting any information regarding
extreme values within a reasonable length of time, is by
statistical analysis It is then of no importance how the
wave systems are defined, onJy a statistical distribution of the moment variation is determined for each such s:rstem
according to the chosen definition.
There are many advantages in expressing the wave
con-ditions by the Beaufort number0 Ships' officers have much
experience of this procedure, and there is generally a fairly good agreei'ient with corresponding wind speeds. The
available statistics on wind speeds is still much better
than on wave heights, and it is therefore easier to
deter-mine the probability of a certain wind speed than of a
7
of a given bending moment it is essential to know the
per-centage of the total time at sea every wave condition is
expected to prevail.
In reference /5/ was shown, how a long term
distribu-tion of bending moments in each eaufort number, or small
group of such numbers, can be calculated by convolution of
two separate distributions: the Rayleigh distribution in
every single wave spectrum, and the distribution of its
parameter r (root mean spuare) in the different spectra with-in each group, A similar c1cu.lation hs b2en made for an
Ameriáan C 4-ship, fig. l If then the probability of each
group is known, which is the same as the time the ship will meet corresponding wave conditions during a specified period,
the expected ma:circium value is given. In fig. 1 is
illustra-ted, how the maximum for this ship is very nearly constant
in all conditions more severe than about 7 aft. It is
per-haps more correct to y, that the probability of a certain
high value of bending moment is nearly constant in all wave
conditions above a certain limit. This also agrees with the
experience that the highest value sometimes has been measured
in rather moderate weather. Above that limit it is quite
fortuitous in which of all possible conditions the highest
value happens to be measured. This limit may not be the
same for all ships, among other things it is probably
de-pending on ship length. The point on the weather scale
where the curve flattens out is moving towards more severe
conditions with increasing ship length. This can be calcu-lated from model tests
As was pointed out above, not only the extreme values are of interest, but all amplitudes which are expected to occur at least up to io6 times during a normal service
period. All the weather groups must then be added, each
weighted according to its own probability of occurrence,
fig. 2. The principle of this calculation was demonstrated
in ref. /5/.
This method of regarding the final long term
distribu-tion as a sum of a. great number of short term distribudistribu-tions,
added after being weighted according to weather statistics,
leads naturally to a similar method for analysis of model test results4 A direct comparison between model and ship is
then possible, and consequently also general conclusions can
be drawn from model tests0
ANALYSIS OF MODEL TESTS
Model tests in regular waves have no meaning if they
cannot be evaluated in a way to make them valid in real,
irregular wave systems0 The only possibility for this is given by the principle of superposition, which assumes that
the response to each frequency in the wave spectrum can be
determined by a test in a regular wave with the same
fre-quency, or wave length.
Rather different results have been obtained regarding the validity of this principle0 Most available test results
show that the linear approach is goDd enough for practical
use, but there are exceptions and the question may not be
quite satisfactorily solved.
Model tests may also be carried out in irregular waves,
either with a spectrum directly programmed to agree with an
actually recorded one, or with a quite arbitrary spectrum.
As it is practically impossible to run tests in all the
superposi-
-9-tion is still necessary. If the model spectrum is
approxi-mately of the same shape as the real one, the influence of
any possible unlinearity will be small, and that may be a
reason for trying to find good approximations to wave
spec-tra.
Under a cont.ìct with the American Bureau of Shiping,
an investigation is being made at Webb Institute of Naval
Architecture by EV. Lewis and others, about how model
tests might be made to agree with full scale tests, and how
such results are to be interpreted. VosserTh tests in
regu-lar waves with models of Series 60 /6/ were used as a basis
for calculations. A certain number of the spectra published
in ref. /1/ were also investigated, in order to find ave-rages for their shape and area at various wind speeds, i'2/.
At first, bending moment spectra were calculated in three
typical wave spectra, recorded in raspectively 37, 52 and 62 knots winde The results indicate the influence of length,
block coefficient and speed, fig. 3. Preliminary studies
showed the necessity of taking the short-crestedness of the
waves into account, which is an effect of the angular spread of the wave energy. A good general agreement with measured three dimensional spectra /7/ was obtained by letting the
energy vary as cos2a, where a is the deviation from the
main direction. The ordinate of the unidirectional spectrum
was thus multiplied by the factor ? cos2a, which makes the
integral between plus and minus 900 equal to unity.
If the angular spread of the energy is to be taken
into account, model tests in many different headings must
be available. For these calculations were used 0, 20, 2+0, 60, 0 and 90°. It must be emphasized that such a calculation
10
-The trends with length and block, shown in fig. 3, have
also been found at full scale tests. The nondimensional
mo-ment is decreasing with increasing length, and increasing with the block coefficient. Comparing L = 15O ft, C = .60
with L = 600 ft, 0B = the short, slender ship has a
greater relative moment in wind speeds below about 50 knots, while the longer and fuller ship has the greatest moment in more severe weather. The curve for the former ship is
consi-derably more flat in its upper part, and the moment is only
slightly increasing above 50 knots. This is a result of the
frequency ranges of the spectra, and especially of the
fre-quency for the maximum energy, in relation to the ship
length.
For a better view of the extreme values, the results
in the 62 knots spectrum are shown in fig. L1., as function
of ship length. The scale of moments here represent the
highest expected value of 10,000, which corresponds to
about 20 hours in this spectrum. In the fig. are also shown
estimated maximum values from full scale tests according to
ref. /5/.
There is certainly at least a qualitative agreement
between models and ships in fig. L1., but the comparison is
still not satisfactory. One single spectrum cannot represent
the most severe conditions during the entire service life
of a ship. For a true comparison, the model tests must be
analysed statistically with the same method as the full
scale tests. It is then necessary to have a statistical
distribution of spectra.
An investigation of a great number of the spectra
suggested by Darbyshire in ref. // was found to agree very
well with most of the records as to shape and peak frequencies,
but not so well as to area, i.e. wave heights. To find an expressionf' the variation of the heights, the measured
spectra were grouped according to wind speed, and the
distri-bution of the root mean square heights was determined in
each group. The mean values were compared with observations,
given by Roll /9/, with the surprising result that the wave height, visually observed on the weather ships is on the average equal to the root mean square of all heights in the
spectrum, the r-value. This indicates that spectra are
about L.O% higher than previously assumed, when the observed
height was taken to be the significant value. An average
spectrum at a certain wind speed consequently seems to have
Darbyshire's shape and frequencies, and area according to
visual observations.
On the basis of this investigation a series of average
spectra in various wind speeds were calculated, which may be
assumed to represent typical North Atlantic conditions. By an approximate method the scatter of the measured spectra about this average was included in the calculation, and some account was also taken of the fact that the relative heading to the dominant wave direction is varying. For each spectrum the corresponding moment spectrum was calculated, which
directly gave the parameter r of the Rayleigh distribution.
Finally all the Rayleigh short term distributions were added, with regard to the probability of each spectrum from weather statistics /9/. The resulting long term distribution
is directly comparable to the one calculated from full scale tests according to the method given in ref. /5/. The model
12
-At the same time a similar investigation was conducted
at SSF-CTH in Gothenbu.rg, with somewhat different assumptions regarding the wave spectra /13/. The same Darbyshire equation
was used, but it was modified to make both heights and
fre-quencies agree with visual observations In refG /9/ are
given distributions of simultaneously observed heights and
wave lengths. A spectrum was fitted to each of those combi-nations, and it was assumed to be valid during a length of
time corresponding to the relative number of observations.
The observed height was taken to be the significant height,
and the observed wave length to be the significant
apparent
period. According to the Webb investigation the visual
height seems to agree better with the root
mean square than with the significant value, and this method therefore gives
too low bending moments. On the other hand no account was
taken of the
energy spread, nor of the fact that the ship
has various headings relative to the dominant wave direction. These two factors both tend to make the calculated moments
too high. Results for two ships are given in fig. 6.
DETERMINATION OF WAVE BENDING MOMENT FOR A NEW SHIP A calculation of expected wave bending moments in a
ship of a certain length and hull form requires the follow-ing conditions to be fulfilled:
The amplitude operator is completely determined. The superposition method of calculating the moment
spectrum from amplitude operator and wave spectrum is valid.
Every wave spectrum the ship will ever encounter is
13
-L) The Rayleigh distribution is valid for the morent
amplitudes in each of these spectra0
These four points obviously contain all necessary
in-formation for a complete calculation of the probability of any arbitrary bending moment, and such a calculation would
be exact in a statistical sense0
The amplitude operator, or the wave moment as a
func-tion of wave lengths, can be determined in regular or in
irregular waves. An irregular system with a spectrum
approxi-mately similar to the real sea spectrum in shape and fre-quency range is bound to give the quickest result, as the complete function is obtained by a small number of runs. A transformation from one spectrum to another one, even
with very different degree of severity, has been shown to
agree very well at tests in the Davidson Laboratory /10/.
If the operator is obtained in regular or irregular
waves, it is in either case necessary to take into account the effect of the angular spread of wave energy, usually
called the short-crestedness of the sea. Results from test
runs in different headings must then be available, and the
energy spreading function must be known. After a sufficient
number of such complete tests, it might be possible to find general correction factors, by means of which results in
only one direction could be made valid in a three dimensional
spectrum. It is, however, evident that the spread of energy
has an essential influence on both moment and other responses.
During 25 years a ship may encounter about 100,000 dif-ferent spectra, and it is hardly conceivable even with elec-trafic computers to calculate that number of Rayleigh
to one single long term distribution. Moreover, it is not
possible to know all the spectra, because each ship will
only meet a random sample of the infinite number of spectra
that may occur at sea. A statistical approximation is thus simply necessary.
In the chapter on Analysis of Model Tests, two methods
were described, which have been used to obtain a long term
distribution of spectra. A continued analysis of the lOO
spectra published in ref. /1/ (which probably will be
fol-lowed by more) will undoubtedly give a much better basis for this type of calculations. It might, for instance, be
possi-ble directly to determine distributions for the ordinates of
all spectra at every frequency (= wave lengths), and in that
way obtain a statistical mean spectrum covering a long
period of time, together with the actual scatter about this mean.
The method of superposition has in many cases been
shown to be applicable in practice, even if it is not
strict-ly proved, and the Rayleigh short term distribution has been
ccnfirmed by so many investigations that it may be adopted
with great confidence. By means of an amplitude operator
and a distribution of spectra, a long term distribution of
bending moment amplitudes can consequently be calculated.
Incidentally, this method is not only applicable to the
bending moments, but to all ship responses to waves: motions,
accelerations, speed, or power. It should be possible to
find by integration an average speed during, say ten years
from such results as are given by Gerritsma and others in
15
-A true wave statistics may be obtained in two ways:
continued work on actual wave records, and analysis accord-ing to the proposed method of all model test results now
available. By comparison with the full scale tests, also
available to a great number, empirical spectrum fami1ies
can be obtained, which give the best result for the greatest
number of models, within the confidence limits that are calculated /1L1/. For normal ships, results for the most similar model in a test series may be used as an approximate
basis for design; for ships of unusual size or form a
speci-al model test has to be done. It is probable that it will
not be long before it becomes part of normal routine to test
models of all new ships also in waves, with measurements
including bending moment and shear force.
This paper is to a great extent based on work done
during a visit to Webb Institute of Naval Architecture,
N.Y., as part of a contract with American Bureau of Shipping.
The author is indebted to these institutions and to Professor
E.V. Lewis for this opportunity, and for permission to
pub-lish the results. A travel grant from OECD, allocated through
the Swedish State Technical Research Council is also
REFERENCES
/i/
Moskowitz, L., Pierson, W.J,, Mehr, E.
Wave Spectra Estimated
from Wave Records Obtained
by the
OWS WEATHER EXPLORER and the OWS WEATHER REPORTERO
New York University.
Department of Meteorology
and Oceanography.
Part I, Nov. 1962, Part
II, March
1963,/2/
Goodrich, G,H., Bonnet, R0, Lewis, E.V.
The Use of Model Test
Data to Predict Statistical
Trends
of Wave Bending Marnent0
Prcgress
Report No0
3of Project for American
Bureau of
Shipping0 (Unpublished)
Webb Institute of Naval Architecture.
April
1963,
/3/
Nibbering, J.JOW.
Vermoeiing van scheepsconstructies.
Schip en Werf
30(1963) 10, s. 275-25.
/L./
Roop, W.P,
Service Strain Tests,
Technique and Procedure0
US Experimental Model Basin.
Report No,
46e, 1940,/5/
Bennet, R., Ivarson, A.,
Nordenström, N,
Results from Full Scale
Measurements and Predictions of
Wave Bending Moments
Acting on Ships0
Stiftelsen för Skeppsbyggnadsteknisk
For skning, Got eborg0
Report No, 32, 1962,
/6/
Vossers, G,, Swaan, W.A0,
Rijken, H.
Experiments with Series 60
Models in Waves.
Trans, Soc. Nay0 Arch,
Mar, Eng.
1960./7/
Canham, HOJOSO,
Cartwright, D.E., Goodrich,
G., Hogben, N,
Seakeeping Trials
on OWS WEATHER REPORTEROTrans0 Royal Inst. Nay,
Arch0
1962,
//
Darbyshire, J0
A Further Investigation
of Wind Generated Waves.
Deutsche Hydrogra.phische
Zeitschrift
12(1959) 1.
/9/
Roll, H,U.
Height, Length and Steepness of Seawaves in the North
Atlantic0
Soc0 Naii-, Arch, Mar.
Eng,
Technical and Research
Bulletin 1-19,
l95.
/10/ Dalzell, J0F.
Trends with Speed, Ship Length and Sea Severity
of
Midship Bending Moment of
a Destroyer Type Ship in Head
Seas,
Davidson Lab, Stevens
Inst. of Technology.
Report
R-92, 1962.
/11/
Bennet, R
A Comparison of Measured and Statistically Calculated
Wave Stress Distributions,
Progress Report No0
L- of Project for AmerIcan Bureau of
Shipping0 (Unpublished)
Webb Institute of Naval
Architecture0
Juno
1963,/12/
Zubaly, R.B,
Trends of Bending Moments in
Irregular Seas.
Progress Report No, 2 of Project
for American Bureau of
Shipping, (Unpublished)
Webb Institute of Naval
Architecture0
January
1963,/13/
Ivarson, A,
Ore Carrier Model Tests,
Stiftolson för Skeppsbyggnadsteknisk
Forskning. Göteborg
Report No
35,/14/
Nordonström, N,
On Estimation of Long Term
Distributions of Wave Induced
Midship Bending Moments in Ships,
Chalmers Tekniska 1-lögakola,
Institutionen för Skeppsbyggnadsteknik.
Aug,
1963,/15/
Gorritsma, J,, van den Bosch,
J.J,, Beukelrnan, W.
Propulsion in Regular and Irregular
Waves.
mt, Shipbuilding Progress,
(196l)
2,
(June), s
235-247,A6/
Swaan,
:J.A0, Rijken,
L{,Speed Loss at Sea as a Function of Longitudinal Weight
Distribution0
North East Coast Inst. Eng.
Shipb,
z
FIG. 2. LONG TERM
DISTRIBUTION 0F EFFLCTIVE
WAVE PLIGHT
SAIE SHIP AS FIG. 1.
'DN
I, ß F TESTS CALCULATED FROM(SERIES 60)
MODEL FULL TESTS SCALP/
I LL
N
'Q3r'
io ioo/. no' ¡DFIG. I. EFFECTIVE WAVE HEIGHT EXPECTED
TO BE EXCEEDEDWITH P% PROBABILITY IN VARIOUS BEAUFORT NUI'4BERS
SHIP: S.S. WOLVERINE STATE, L
=¿96'3
0B = 0,65
m-WIND SPEED
KNOTS
FIG. 3. VARIATION OF W.B.M. COEFFICIENT WITH WIND
SPEED.ROOT MEAN SQUARE VALUE OF ni.
(REF. /12/)
[F. [5]. CB 0W - 066
o PEF[53, C.O-l5-D.BO
SHIP LENGTH
I I I I I
FEE T
FIG.
Li.. VARIATION OF W.B.I. COEFFICIE
WITH SHIP LENGTH.
EXPECTED MAX. OF 10,000 IN 62 KNOTS
SPECTRUM.500
ft.. SEC.
-50 KNOTS (Jo)
ttOO 200IjO KNOTS (s-9)
OKNOTS (7)
ZOKNOTS (5)
02. 0,ts 0,6 0B 1,0 W RA 0/8.FIG. 5. MODIFIED DARBYSHIRE
4
-3
-2
-I f
6 a. BULK CARRIER. L
L.6O FEET? 8 7 i$ 5 4 3 2 I 0 72 -IO o 6 z ß(/L/( CAìi!/Ei #100(6 TíT PgZß/Cf(ON f.YE4It CANADA (iN/P C)
6 b. DRY CARGO SHIP. L
265 FEETFIG.
6.
LONG TERM DISTRIBUTIONS FROM MODELAND FULL SCALE TESTS.