Professional Documents
Culture Documents
By F. H. T o d d 1
The principal method of predicting ship horsepower is still by the use of scale models in a
towing tank. However, the problem of the extrapolation of the data from model to ship
is an extremely difficult one which is still not satisfactorily solved. This paper discusses
the types of scale effect which arise in the interpretation of model resistance and propulsion
data. It suggests that although satisfactory engineering solutions can be obtained by the
use of gross overall correlation coefficients derived from a comparison of ship trial results
with predictions from model tests, yet many unknown areas remain in our basic knowledge
of how correctly to extrapolate model data. The need is emphasized for further experi-
mental and theoretical work in the model field, to clear up such questions as blockage
effects, turbulence stimulation, and the division of resistance into its several components,
and for full-scale trials in which all the quantities measured on the model will also be
measured on the ship. Such trials would give for the first time information on the scale
effects on wake, thrust deduction, propeller open efficiency, and overall propulsive effi-
ciency. Until we know the extent of such effects, we lack guidance as to where to look
for further improvement in ship performance.
ThE prediction of ship power from model experiments about the true method of prediction from model to ship,
remains one of the primary problems in our business. and had greater confidence in the resulting estimntes of
I t lies at the very foundation of all our work; its solu- ship power, these allowances could be reduced and a lower
tion is extremely difficult and it should certainly be a powered engine employed.
m a t t e r of first-class importance to the American Towing I t m a y be objected t h a t today the machinery is
T a n k Conference. chosen from an existing array of engines and that small
While our present methods, combined with the use of reductions in estimated power are not likely to influence
experience and correlation allowances derived from the choice. This m a y be true in m a n y cases, but in
model-ship comparisons, m a y enable us to get adequate others such a reduction m a y well enable the next size
engineering answers, scientifically our knowledge is most smaller engine to be used, with beneficial effects on dead-
inadequate. The ability to obtain such answers has per- weight, first cost, running costs, and maintenance. As
haps induced too complacent tm attitude toward the need an example, in one design, as a result of model tests, the
for more basic information in the whole field of model-ship original intention to fit a 4-cylinder diesel was changed
correlation. A proper solution of the problem would, in- to the use of the next larger 5-cylinder type. Ou trial,
deed, also have important practical applications. Until the ship could have quite easily attained her designed
we know exactly how the total resistance of a model is speed with the original 4-cylinder engine. This in-
divided among its components, it is obviously more dicates how ~:~ more complete understanding of the
difficult to discover how to reduce the total resistance. model-ship correlation problem could be of practical
The step from the resistance to the shaft horsepower also interest to the builder and operator.
involves a number of factors which m a y be subject to A fairly complete review of this subject has been given
greater or lesser amounts of scale effect. At present, the elsewhere [112 Some of the more important points and
shaft horsepower installed in a ship includes a "service some new knowledge are summarized in this ps,per.
allowance" which is designed to enable the ship to main-
M o d e l Reslstance
rain speed in bad weather or when the hull is rough
through corrosion or fouling. In general, such power As is well known, the components of model[ resistance
cannot be used to maintain sea speed in any really bad interact one with the other and are not clearly separable.
weather, because the speed is then governed by other T h e y cannot, be dealt with as a single entity, for the
considerations, such as excess motions of pitching, heav- viscous and wave-making components follow diff'erent
ing, and rolling. I t is believed t h a t if we knew more natural laws, typified by the Reynolds n u m b e r R~ =
VL/~, and the Froude number F,~ = V/(gL) V~', respec-
Scientific Adviser to the Technical Director, David Taylor
ModeI Basin, Department of the Navy, Washington, D. C. tively.
Presented at the American Towing Tank Conference, Webb Froude in 1868 proposed to divide the total model
Institute of Naval Architecture, Glen Cove, Long Island, N. Y.,
September 1965. Numbers in brackets designate References at end of paper.
4
i.e., at the same value of V / L v' would have the same
wflue for model and ship. The R,,~ for the model was
taken to be the resistance of an "equivalent plank,"
having the same wetted area and length as the model.
I t was determined from the results of plank experiments
run by Froude which were assumed to apply to new, g v~ i= v~
freshly painted steel surfaces. 4-~ 4Vs ~
On the Froude assmnption, we write VL
C>, = C,,,, + C,,.,, (1)
d
where C~,~,, is measured in the model tests, C~,,,~ is ob- c Fo~l
tained from phmk data, and C~e.vis tile difference (Cr_~ --
C,,.,,~). Fig. 1 Extrapolation from m o d e l to ship using form factor
For the ship, at the same speed-length ratio, V,s/L,s '/~ method
= VM/LM V~
of it is certainly of viscous origin. The method proposed
C~,,,~is obtained from extrapolated plank data, and CRs = by Hughes assumes th'-~t the total viscous resistance
Ct~v according to Froude's law of comparison. Hence should scale according to Reynolds Law, i.e., bear a fixed
ratio to tile resistance of the equivalent two-dimensional
CT,~ = C~,,~ -- (C~,,, -- C,~,~) (2) plank, the same ratio for nmdel and for ship. He intro-
This principle ]s still in use today in ahnost all towing duced a form factor (1 + k), writing the total viscous
tanks, the only difference in practice being the use of resistance as
different plank friction formulas by which (CF_~ -- C~,s)
C~ = (1 + lc) C~,o = r C~o
is estimated, such as the A T T C 1947 and I T T C 1957
lines, which apply to smooth surfaces, a where (1 + k) depends upon the aspect ratio of a plank
I t is seen from equation (2) thai, the extrapolation de- or the hull shape of a model. I t is assumed to have the
pends only on the slope of the extrapolation line used same value for all geosim models and for the correspond-
between model and ship; i.e., on the factor (Cur -- ing ship.
The extrapolation procedure can now be written in the
More recently, Hughes has done much experimentM form
work in an effort to determine a two-dimensional tur-
bulent flow line, free from the aspect ratio, edge, and
surface effects present in earlier experiments. His re- CTs = Cuos (1 + lc) + C,r~
sulting line where Cw~ = Cws is tile coefficient of wave-malting
resistance, Fig. 1. Hence
0.066
C~o = (3)
CTs = C,,,,~ -- (C,~o,,, -- C,~o,~) (1 + /~) (4)
2.03) ~
(log~0 R,~ -
at the same wflue of V / L '/~ I t is seen t h a t in using Hughes' proposed method both
where Cro is the resistance coefficient for two-dimensionM the slope and the position of the CFo line are important,
turbulent flow, gives values of Cu considerably lower t h a n since the factor (1 + ]c) depends on the ordinates Cr.u
the A T T C line at ship v'dues [2]. and CFov. The value of (1 + It) is to be deternfined
Assuming such a two-dimensional line were established, from the "run-in" point H, where the curve of the model
the correct extrapolation method would still remain to be total resistance coefficient Cr.~f becomes approximately
found. At present all the resistance in excess of the parallel to the CFo curve, below which point the wave-
equiwdent plank resistance is assumed to scale ae- making resistance coefficient C~r.~ is assumed to be
cording to Froude's Law, i.e., as the displacement or Xa negligible.
at the same Froude number, X being the linear scale be- This is a difficult point to determine, calling for ex-
tween model and ship. This cannot be correct, for some treme care and precision in model tests, and a t t e m p t s
ATTC = Americ~m Towing Tank Conference, ITTC = have been made to express (1 + ]c) in terms of hull co-
InternationM Towing Tank Conference. efficients and proportions.
APRIL 1 9 6 6 153
Table 1 Prediction of Ship Resistance From Model Results
Model -Large (22.9 ft) Small (5.36 %)
ATTC ITTC Froude Hughes' ATTC ITTC Froude Hughes'
Method used line line coefficients form factor line line coefficients form factor
CTs X 19 a 2.510 2. 455 2. 910 2.14 2.60 2. 365 3. 270 1. 970
Percent difference from
ATTC prediction --2.2 -}-16.0 --14.7 ... --9.0 +25.8 --24.2
Nomenclature
t = ~hrust deduction fraction ~rt = hull etiieiency
= (T - R ) / T = (1 -- R/T), where R is resistance of ship when = (1 - t)/(1 - w)
towed and T propeller thrust required to propel it al; the ~R = relative rotative eiI%iency, ratio of propeller efIiciency
same speed behind model to that in open water when delivering same
w = wake fraction thrust at same revolutions per unit time
= (V - VA)/V, where V is speed of ship and VA average inflow r/D = overall or quasi-propulsive eoefIicient
velocity to propeller, called speed of advance = ~Tu X "Oo X ~e
r/o = ef[icieney of propeller in open water, without a model hull = effeeUve or tow-rope horsepower/delivered horsepower at,
ahead of it propeller
0,2.~
0,~.i /-J
/ i /
0,2o /
f
/
o1~- / / /
/
~o~EL
Lo~. 2~
Fig. 2 Variation of thrust-deduction coefficient with scale
Table 2 Propulsion Factors From Victory Geosim Tests Table 3 Values of Propulsive Coefficient 7D for Tanker,
Model scale 1/23 1/6 From Reference [6]
Model length 19.3 ft 74.0 ft Sc~de 1/55 1/45 1/35 1/25
t 0.22 0.27 log R,, 6.4 6.6 6.8 7.0
w 0.35 0.25 n9 for ship SP point O.62 O. 67 O. 72 O.75
~H 1.19 0.98 Length of model LM ft 10.05 12.27 15.76 22.10
~o O. 63 O. 65
~R 1.04 1.05
~D 0.78 0.67 model lengths being from 10.05 to 73.7 ft. I n addition to
resistance tests, propulsion tests were made at propeller
loadings corresponding to model and ship propulsion
out at the Netherlands T a n k on the Victory ship design points and at one intermediate one.
[4]. T h e y covered models from 7.4 ft to 74 ft in length. The resistance tests showed t h a t the Hughes' method
If we confine ourselves to the results of the 19.3-ft model, gave less evidence of scale effect in the E H P predicted
which is a typical length employed in commercial testing, from the various scale models, but t h a t the E H P was
and the 74-ft "ship," the values of t, w, 70, 7~, ~ , and ~D considerably less than t h a t predicted by using the I T T C
are as shown in Table 2. line.
The value of t shows considerable increase in going The open propeller tests showed considerable scale
from a length of 19.3 to one of 74 ft, and this is only a effect on open efficiency ~o--from 0.50 at 1 X 105 R~ for
continuation of the trend from the 7.4 to the 19.3-ft the 1/55 scale to 0.62 at 3 X 1@ R,~ for the 1/25 scale
modeh The only full-scale measurements we have are model (Rn measured at 0.7 radius, J = 0.6). There were
those carried out on the Albacore, which show the same no open results for a 1/7.5 scale propeller, but the curves
quantitative trend in t to a Reynolds-number base, Fig. of 7o indicated t h a t little further increase in 7o could be
2 [5]. expected for larger propellers.
The wake decreases with increasing length of model, The analysis wake showed predominantly a reduction
as might be expected due to the reduction in specific fric- with increasing Rn, the effect of propeller loading being
tional resistance. negligible. The nominal wake was greater than the
Because of these changes in t and w, the reduction in analysis wake (0.40 as compared with 0.35). Also,
hull efficiency 7~ is from 1.19 to 0.98, with a correspond- the wake distribution changed with increasing scale--the
ing fall in propulsive efficiency from 0.78 to 0.67, the average wake was less and the peaks sharper. This could
latter being mitigated to some extent by an increase in affect harmonic analyses used for the calculation of torque
~o from 0.63 to 0.65. and thrust fluctuations and vibration forces.
I t is evident t h a t if these trends continued to the ship, The thrust-deduction coefficient t did not v a r y much
the propulsive efficiency would be much lower than that in the four smaller models (R~ = 106 to 107), but did v a r y
of the models, even that with a length of 74 ft. with propeller loading. The mean value was about 0.19.
I t is curious t h a t of all the model results those from For the model boat, of 1/7.5 scale, t was 0.26, which
the largest (74 ft) seem to be most open to doubt. change is comparable with t h a t found in the N S M B
Further information on the scale effect on propulsion Victory tests (t = 0.22 to 0.27, Table 2) and the Albacore.
factors has recently been published by the Berlin T a n k The propulsive efficiency 7D was mainly affected b y
[6]. Geosim models were run for a 552-ft tanker, the the scale effects in w, t, and 70. Table 3 gives the values
scales varying from 55 to 7.5, the corresponding range of of 7D derived directly from the model results at the ship
self-propulsion point. If the components of r/v are extra- M a n y data have been amassed from shill) trials, and
polated to the ship values, then the calculated value of ~, each towing t a n k continuously carries out such correla-
is about 0.69, as compared with 0.75 derived directly tions to determine values of C.4, using its own particular
from the 22.1-ft model. methods of extrapolation and so gaining and maintaining
The full-scale trials of the tanker showed t h a t the confidence in its own predictions. E v e n so, a great
model-ship correlation allowance CA needed to reconcile m a n y unknown factors remain in individual cases, such
actual and I T T C predicted powers was reasonable--from as the actual roughness of' a particular ship's troll, weather
1.5 to 6.5 percent for 12 to 16 knots speed range. How- effects, and so on. The ship data obtained in this way
ever, this is not a confirmation of the I T T C line, since show a wide scatter, with one class of six sister ships
there are still unknown scale effects on w, t, and so forth, showing as much as 23 percent difference in power.
and perhaps the E H P should be lower if a steeper extra- Sometimes the correlation a]lowance C~ is negative,
polator is necessary to reconcile geosim data. especially in very long, all-welded ships [7 ].
On the other hand, comparisons between the results There are m a n y other items in the correlation picture
for three Series 60 models, 20 ft in length, run at T M B which it is only possib]e to mention briefly. The co-
and those for three geosims, 14 ft long, run in the Univer- efficient of additional resistance due to bossings or struts
sity of Michigan, showed the larger models to have lower and shafts, as measured on the model, is sometimes ap-
values of t. plied directly to the ship, sometimes reduced because of
Long experience would lead to the view t h a t there is the small size of such appendages on the model. These
no place in the model-ship correlation for an overall scale different practices can cause as much as 10 to 12 percent
effect of these magnitudes. But the situation has never difference in the total ship-power prediction. Recent
been satisfactorily cleared up. The original intention work on wave-making resistance has suggested that the
in the D u t c h program was to finish by towing or propel- coefficient Cr~v m a y contain other components, while the
ling by aircraft jet engines a full-sized Victory ship, so effect of viscosity on wave-making resistance m a y intro-
obtaining resistance, and then propelling it to obtain the duce a scale effect as great as 7 percent in going from
thrust, and hence a full-scale value of t. This last step model to ship at a value of V / L v~ = 0.9, with the effect
has never been taken, however, partly for cost reasons, practically disappearing at low and high values [8].
partly because the ships are now so old that their sur- Experiments measuring viscous drag by wake surveys
faces are too poor to use t h e m as correlation ships. behind a model have indicated t h a t perhaps the viscous-
I t will be agreed, perhaps, t h a t the whole state of drag coefficient is a function of both R,, and Fn, and so
our knowledge of scale effect in this area is most un- dependent on shape of hull. Experiments on separation
satisfactory. at the after end of fuller forms have given contrary in-
dications as to scale effect between model and ship.
Comparison of Data Evidence has also been found in experiments with
The resistance of the ship, deduced from the fidl- model propellers, both in open water and behind a
scale trials using either the thrust or horsepower, and model, t h a t appreciable laminar flow m a y occur, suggest-
assuming no scale effect in t or ~D, can be compared with ing t h a t turbulence stimulation m a y be necessary here as
the resistance predicted from the model. This has been well as on the hull model [9 ].
done in Table 4 for the tanker results already used in The correlation between the R P I \ 1 % u n d on the ship
Table 1, and the values of the correlation allowance CA trials and those predicted from the model is not so good,
can be determined. and differs between different towing tanks. In some
I t will be seen that using the large model the wtlues of cases the measured R P M are less than those predicted,
CA X 10 a for the different extrapolation methods v a r y in others greater, and the correlation also varies with the
from --0.200 to -t-0.570, while using the small model roughness of the hull [10,11].
they range from - 0 . 5 6 0 to +0.740.
Concluding Remarks
I t should be noted t h a t these differences have nothing
to do with the ship only one value results from the trials. There are two fields in which further work is neces-
T h e y are entirely due to the methods of extrapolation sa W to improve the present situation. In the model area,
and the difference in size of the models. This i's sufficient there are a n u m b e r of points on which further informa-
to justify the use of the term "model-ship correlation tion is needed: The use of standard models for monitor-
allowance," CA, in place of the old roughness allowance. ing changes in resistance, due perhaps to changes in the