43

(昭和54年11月 日本 造 船 学 会秋 季 講 演 会 に お い て 講 演)

A Method for Predicting Ship Wake from Model Wake

by Tetsuo Nagamatsu, Member*

Summary
This paper presents a method to predict ship wake from measured model wake
by applying the theory of two-dimensional turbulent wake. From basic investiga-
tions on two-dimensional wing, the following physical properties of turbulent wake
were made clear with respect to the scale effect on viscous wake.
(1) The width of wake is in proportion to the momentum thickness and not to
the scale of model and ship.
(2) Velocity distribution in wake varies downstream as a function of non-
dimensional distance divided by momentum thickness.
(3) The value of the wake peak is affected significantly by the difference of the
velocity profile at the stern frame between model and ship.
In the well-known Sasajima-Tanaka's method only the first item of above is taken
into account for scaling the viscous wake.
The method for predicting ship wake was developed in view of the above pro-
perties. Applicability of the method to the wake of ship was examined by use of
the geosim models of a cargo ship and it was shown that the method presented here
provides reasonable results and can be applied as a practical means for predicting
ship wake.

bility of theoretical treatment.

1. Introduction Then the experiments were made on two geosim
Unsteady propeller cavitation and propeller ship models and it was shown that the method for
vibratory forces depend strongly on velocity dis- predicting ship wake from model wake provides
tribution at the propeller position. Therefore, reasonable results including wake peak.
cavitation tests of a model propeller and estima-
2.
Theory of Two-dimensional
tion of propeller vibratory forces are made gener-
Turbulent Wake
ally for wake similar to that of the ship.
The distribution of the ship wake is ordinarily Let us consider steady-state turbulent wake
predicted from the wake measured on a geometri- behind a two-dimensional body.
cally similar model. There have been a number The origin of (X, Y) coordinate system is
of proposals for the methods of prediction.1)2)3)4) located at the after end of the body. X is distance
Sasajima-Tanaka's method1) is widely known, in in the direction of main flow and Y is distance
which width of wake is reduced in proportion to normal to it. U and V are the velocity com-
frictional resistance coefficient with the wake peak ponents in the X and Y directions, respectively.
assumed un-varied. The overscore denotes the time-mean values and
On the other hand, wake peak plays an impor- the prime the fluctuating quantities, viz.
tant role for unsteady cavitation phenomena of
(1)
propeller blades and propeller-induced pressure
fluctuation5). Nevertheless scale effect on wake
The two-dimensional Reynolds equation of
peak has not been made clear. motion in the X-direction can be written as
In the present study, attempts were made to
follows.
incorporate scale effect on wake peak in predict-
ing ship wake from model wake. To this end basic
investigations were carried out on two-dimen-
sional problems to grasp the physical properties
of turbulent wake and to examine the applica- (2)

* Nagasaki Experimental Tank , MHI. where p, it and P are the density, viscosity and
 44 日本 造 船 学 会 論 文 集 第146号

pressure, respectively. If the values of f (t) is not so large, the right

If Reynolds number is sufficiently large, the hand side of eq. (12) may be put equal to zero
molecular viscous terms are negligibly small as from practical viewpoint. (see Appendix.)
compared with the turbulent viscous terms. Then eq. (12) is reduced to
Then, taking the continuity condition into
account, eq. (2) is reduced to (14)

If the boundary condition and the initial

(3)
values are given as
Here, we introduce Reichardt's momentum at
transfer law6) (15)
at

(4) the solution of the partial differential equation

(14) at t is obtained as follows.
where ă, which might be termed a momentum

transfer length, has the dimension of length and

must be determined empirically.

Further, approximating as (16)

3. Experimental Investigations on
and substituting eq. (4) into eq. (3), the following Two-dimensional Wing
equation is obtained; 3.1 Wing Models and Test Conditions
In order to examine the applicability of the
(5)
theoretical treatment described above, velocity
Now, we assume that pressure in wake is nearly measurements in wake behind two-dimensional
constant in the Y-direction and the following wings8) were carried out in Nagasaki Experimen-
relation is applicable. tal Tank.
Two rectangular wings of geometrically similar
(6) sections with different chord length of 1.42 m
and 3.50 m were used for the experiments. The
where Ue is the velocity at the edge of the wake.
section shape of the wings is similar to the blade
Let us introduce non-dimensional quantities
thickness distribution of the ITTC standard
defined as follows.
propeller with zero camber as shown in Fig. 1.
Furthermore, experiments of the larger wing
with its surface roughened by wire meshes were
(7) conducted to investigate the effect of the profile
where Į0 is twice the momentum thickness of of velocity distribution in wake on the momen-
boundary layer at the after end of the body and tum transfer length.
U•‡ is the velocity of the undisturbed oncoming Cylindrical studs were mounted at 0.05L from
stream. the leading edge of each wing as turbulent stimu-
From eqs. (6) and (7), eq. (5) is reduced to lator.

(8)

Assuming that K is constant in wake and taking

(9)
as was done by Toyoda-Hirayama7) and putting

(10)
where

(11)

eq. (8) can be rewritten as

Fig. 1 Profile of tested wing
(12)

where The velocity measurements were made by use

of a 5-hole pitot tube. The test conditions are
(13) summarized in Table 1.
 A Method for Predicting Ship Wake from Model Wake 45

Table 1 Test conditions of two-dimensional wings

3.2 Estimation of Momentum Transfer Length

Since eq. (16) is expressed in terms of t instead
of x, let us first determine K-non-dimensional
momentum transfer length-by comparing the
variation of wake peak with t as obtained by
eq. (16) and the measured wake peak expressed
as a function of x. If we denote the velocity at
v= 0-centerline velocity-by do, this is ex-
pressed by

(17)

For the three wings with different conditions

as described in Table 1, uo were calculated by Fig. 3 Variation of calculated centerline

eq. (17), in which to was taken at the trailing velocity with t
edge and q(to,ā) was obtained from the velocity

Fig. 2 Velocity distributions measured at Fig. 4 Variation of measured centerline

trailing edge of wings velocity with x

Table 2 Results of measurement at trailing edge of wings

 46 日本 造 船 学 会 論 文 集 第146号

Fig. 5 Relationship between t and x

Fig. 7 Longitudinal variation of centerline
velocity
distribution measured there as shown in Fig. 2

and Table 2. The calculated results are compared with the

The calculated results are plotted against t measured ones in Figs. 6 and 7, in which the
in Fig. 3 and the measured ones are plotted
following points are observed.
against x in Fig. 4.
(1) The calculated values are in good agree-
Comparing Figs. 3 and 4, relationship between
ment with those measured by experiments,
t and x is obtained and plotted in Fig. 5. It is
therefore the velocity distribution in wake of
found that t is approximately proportional to x
the two-dimensional wings can be calculated by
and there is no remarkable difference due to size
eq. (16) with ƒÈ=0.06.
and surface roughness of the wings. Therefore
(2) The velocity distributions of the two
it is considered that the non-dimensional momen-
wings, when plotted to the base of Y/Į0, coincide
tum transfer length ƒÈ is scarcely affected by the
nearly with each other except the vicinity of
scale of wing and the velocity profile on it.
y =0, i.e. the inner region.
Considering eq. (7), ƒÈ is approximately deter-
This means that scale effect on the velocity
mined from Fig. 5 as
distribution in the outer region of wake can be

(18) predicted by non-dimensionalizing Y with the

3.3 Velocity Distribution in Wake momentum thickness Į0. Furthermore, assuming

With the above value of ƒÈ, velocity distribu- (19)

tions in wake at X/L = 0.01 were calculated for
where C, is viscous resistance coefficient, it can
both smaller and larger wings with smooth surface
be said as proposed by Sasajima-Tanakau that
by use of eq. (16), taking the velocity distribu-
the width of wake in the outer region varies in
tions measured at the trailing edge as the initial
proportion to the viscous resistance coefficient.
values. Longitudinal variation of centerline
(3) On the other hand, there exists an ap-
velocity was also calculated for both wings.
preciable difference of velocity between the two
wings in the inner region as shown in Fig. 6.

We can also see different values of centerline

velocity among the tested wings as compared in

Fig. 7 at the same relative distance X/L. Such

variation of wake is discussed further in the

following section.

3.4 Variation of Velocity Distribution near

Wake Peak

From the above mentioned investigations, it is

concluded that the velocity distribution in wake

of the two-dimensional wings can be calculated

by eq. (16) with ƒÈ=0.06. Eq. (16) indicates that

the velocity distribution in wake depends on

q(to,v) and (t-to), which correspond to the initial

velocity distribution and the distance from the

initial position, respectively.

Fig. 6 Comparison of velocity distributions Reminding eqs. (7) and (9), and choosing the

in wake of wings trailing edge of wing as initial position, it is

 A Method for Predicting Ship Wake from Model Wake 47

noticed that the value of (t-to) at a certain

position of X/L varies in proportion to relative

momentum thickness Į0/L. The values of (t-to)

at X/L =0.01 of the two wings with smooth

surface, for example, are as follows.

for the larger wing

t-to=0.111 for the smaller wing

The variation of velocity distribution with t

for the larger wing is shown in Fig. 8, which

indicates that wake peak varies with t remark-

ably but the velocity distribution in the outer

region varies little.

Let us consider next the initial velocity distri-

bution q(to,ā) at the trailing edge. The measured

velocity distributions at the trailing edge of the Fig. 10 Velocity distributions calculated at
two wings, when plotted to the base of Y/00, t= 0.136
are considered to coincide nearly with each other,

as shown in Fig. 9, except near the wall.

However, the two velocity distributions calcu- region is very sensitive to the initial velocity
lated at t-to= 0.136 by use of the initial velocity distribution and the distance from the initial
distributions in Fig. 9 are, as compared in Fig . position.
10, considerably different in the inner region. In The appreciable difference of wake peak shown
other words the velocity distribution in the inner in Figs. 6 and 7 is due to such effects of initial

velocity distribution and X/Į0 distance in rela-

tion to momentum thickness of the wake.

4. Further Investigations on Ship

Models

4.1 Models and Test Results

In order to examine the applicability of the

method of calculation for two-dimensional wings,

further investigations were carried out for cargo

ship models and the results were compared with

the measured ones. Two geosim models of cargo

ship, 4.5 m and 7.5 m in length, were used in

the experiments.

The velocity measurements for each model were

done at several vertical sections including the

Fig. 8 Variation of velocity distribution position of propeller and stern frame. The meas-
with t urements were conducted by use of a 5-hole

pitot tube which was traversed at four different

depths. The measured velocity distributions at

the position of stern frame and propeller are

shown in Figs. 11 and 12 respectively.

4.2 Calculation of Velocity Distribution in

Wake

Based on the values measured at the position

of stern frame as shown in Fig. 11 the velocity

distributions at the propeller position were

calculated by use of eq. (16) with ƒÈ= 0.06. The

results are also shown in Fig. 12.

Agreement between the calculated and meas-

ured values is not so good. This implies that it is

difficult to apply the simple relations (16) obtain-

ed for two-dimensional flow to such a three-

dimensional wake of ship model.

Fig. 9 Velocity distributions measured at However, since the three-dimensional character
trailing edge of wings of the flow is considered to become less in wake
48 日本 造 船 学 会 論 文 集 第146号

Fig. 11 Velocity distributions measured at

stern frame of cargo ship models Fig. 12 Comparison between measured and
calculated velocity distributions at
propeller position

than at the aft end of model, further calculation

was made using the velocity distribution in wake
at midway between the propeller position and stern
frame instead of those at the position of stern
frame. The results of calculation for the propeller
position are shown in Fig. 13 together with the
measured velocity distributions.
In this case, good agreement can be seen be-
tween the calculated and measured results, and
it may be said that above mentioned relations
(16) and (18) give fairly good approximation to
the wake of ship models.
5. Prediction of Ship Wake from
Model Wake
5.1 Principle of the Prediction Method
Based on the above mentioned investigations,
let us study a method for predicting ship wake
from measured model wake by use of eq. (16).
In the proceeding cases of two-dimensional
wings and ship models, attempts were made to
calculate the wake when the velocity distribu-
Fig. 13 Comparison between measured and tion was given at the aft end of the body or
calculated velocity distributions at slightly down-stream of it.
propeller position Wake measurements on model ships are, how-
 A Method for Predicting Ship Wake from Model Wake 49

ever, mostly conducted at the position of propel-

ler, and therefore let us consider a method for (28)

predicting ship wake at the position of propeller
from the model wake measured at the position
and P is Gamma function.
of propeller.
Then, we define ƒÁ as follows
Let XƒÏ denote the distance from stern frame

to propeller position and for simplicity, let sub- (29)

scribts s and ni express hereafter ship and model

respectively. From numerical calculations of ƒÁ for various

Considering the relation ( 19) , xƒÏs is derived as values of n, it is found that variation of (4t1)P

and „C(ƒÀ+ 1) is very small. Therefore we can put

(20) approximately

Therefore, (30)

Combining the first and the second step, wake

(21)
peak of ship can be predicted by
and
(31)
(22)
where uo1 is the value of uo obtained in the first
where t and to express the propeller position of
step, i.e. before correction for velocity distribu-
ship and model, respectively.
tion has been introduced.
With the velocity distribution q(to,ā) at the
Now the value of wake peak and the velocity
propeller position of model and (t-to) given by
distribution in the outer region have been
eq. (22), the velocity distribution q(t, y) in wake
determined but there remains a problem how to
is calculated by eq. (16). After that, width of
estimate the velocity distribution in the inner
wake is corrected according to the ratio of
region between the wake peak and outer region.
viscous resistance coefficient as follows.
This could be approximated by a type of error
function.
(23)
(32)
Nextly, the difference of the initial velocity So that a and b may be determined for example
distribution between ship and model is taken into to satisfy the value of (1-17os) and 50% of it, viz.,
account since this effect is considerably large as
described in 3.4. (33)
For simplicity, let us approximate the velocity
profile in the boundary layer at body end by where ysi/2 is the value of y at (1 - us)=1/2(1-
power law, i.e.
fi0s).
(24) 5.2 Simplified Prediction Method
As described in 3.4 the velocity distribution in
where 8 is the boundary layer thickness. the outer region of ship wake may be directly
Then, predicted from that of model wake by use of the
ratio of viscous resistance coefficient.
(25) Here, let us consider a simplified method of
where predicting the wake peak of ship without numeri-
cal calculation of eq. (16) described in the previ-
ous section. First, we approximate the velocity
(26) distribution measured at propeller position of
model by a type of error function as applied in
the previous section.
and H is shape factor. (34)
Substituting eq. (25) into eq. (17), we get a and b are determined in similar manner and
substituting eq. (34) into eq. (17), not at the
propeller position of ship is obtained as

(27)
(35)
where
 50 日本 造 船 学会 論 文 集 第146号

After that, correction of ƒÁ or ƒÁ' is made the In the case of two-dimensional wing, the
same as previous section. smaller wing was regarded as model and the
5.3 Examination of the Prediction Method velocity distribution ft of the larger wing at X/L =
The method for predicting ship wake men- 0.01 was predicted. The results agree well with
tioned above is applied to the two-dimensional the measured results as shown in Fig. 14. It can
wing and two geosim models of cargo ship and be also seen that the both methods predict
its applicability is discussed. almost the same values.
The wake distributions at the propeller posi-
tion of the larger model of cargo ship were
predicted by means of both prediction methods
from the velocity distribution measured at the
propeller position of the smaller model. The
results are in good agreement with the measured
ones of the larger model as shown in Fig. 15,
where for the velocity at the edge of wake Ue
the measured values were used.
From these examinations, it can be said that
the methods for predicting ship wake presented,
here provide reasonable results and can be ap-
plied for practical use.
5.4 Prediction of Ship Wake
As an example of the prediction method de-
Fig. 14 Measured and predicted velocity
scribed above, calculation was made on ship
distributions in wake of two—dimen-
wake from the wake of the larger model of cargo
sional wing
ship used in this study under the following condi-
tions.

Fig. 15 Measured and predicted wake distri-

butions at propeller position of Fig. 16 Predicted wake distributions at
cargo ship model propeller position of full scale ship
 A Method for Predicting Ship Wake from Model Wake 51

Ratio of viscous Wake Characteristics Based on Model

resistance coefficient CƒÒs/CƒÒm= 0.60
Wake Survey," Symposium on High Power-
Estimated power law n = 6 for ship ed Propulsion of Large Ships, Wageningen,
n=5 for model
1974.
The exponents of power law were estimated by 3) Dyne, G., "A Study of the Scale Effect on
referring to the measured data of ship9) and con- Wake, Propeller Cavitation and Vibratory
sidering the scale of model. Pressure at Hull of Two Tanker Models,"
Furthermore, ue of ship wake is assumed to be Trans. SNAME, Vol. 82, 1974.
equal to that of model wake. 4) Rader, H. P., "Die Bedeutung des Nach-
The predicted wake distributions of ship are stroms fiir die Wechselwirkungen zwischen
shown in Fig. 16. The velocity distributions Schiffsrumpf und Propeller," Hamburgi-
estimated by the ratio of CƒÒs/CƒÒm according to sche Shiffbau-Versuchsanstalt G.m.b.H.
Sasajima-Tanaka1) are also shown for reference. Bericht Nr. 1506, 1976.
It is to be noted in this figure that the velocity 5) Huse, E., "Cavitation Induced Hull Pres-
distribution in the outer region is close to that sures, some recent developments of model
estimated by CƒÒs/CƒÒm but peak value of ship is testing techniques," Symposium on High.
considerably lower (about 10% in the present Powered Propulsion of Large Ships, Wage-
case) than the model values. ningen, 1974.
6) Schlichting, H., "Boundary Layer Theory"
6. Concluding Remarks
McGraw-Hill, 1960.
A method is presented to predict wake distribu- 7) Toyoda, K. and Hirayama, N., "Turbulent
tion of ship from model wake measured at the Near Wake of a Flat Plate (Part 1, In-
propeller position. The method has been develop- compressible Flow)," Trans. of the Japan
ed based on theoretical and experimental investi- Society of Mechanical Engineers, Vol. 39,
gations of two-dimensional turbulent wake No. 326, 1973 (in Japanese).
referring to Reichardt's hypothesis on momentum 8) Taniguchi, K. and Sasajima, T., "Scale and
transfer law. Roughness Effects of Propeller Section
Applicability of the method was examined by Drag.," Journal of the Society of Naval
experiments for the two-dimensional wings and Architects of Japan, Vol. 133, 1973.
the two geosim models of cargo ship and it was 9) Taniguchi, K. and Fujita, T., "Comparison
found that the method would give reasonable of Velocity Distribution in the Boundary
results for practical use. Layer between Ship and Model," Journal
For further development of the method, how- of the Society of Naval Architects of Japan,
ever, more detailed investigations are needed Vol. 127, 1970.
with respect to the characteristics of momentum 10) Kodaira, Y., "Mathematics for phsics"
transfer length and to the more reasonable estima- Iwanami, 1933 (in Japanese).
tion of velocity at outer edge of wake and shape 11) Himeno, Y. and et. al., "Methods for
factor of ship wake. Predicting Ship Wake Development and
Furthermore, study on scale effect on tangential Its Displacement Effect Using Boundary
wake and behavior of bilge vortices are necessary. Layer Approximation," Journal of the
Kansai Society of Naval Architects, Japan,
Acknowledgement
Vol. 163, 1976 (in Japanese).
The author wishes to express his deep gratitude

to Mr. Tanibayashi, Manager of Resistance and

Appendix
Propulsion Research Laboratory of Nagasaki Now, assuming
Technical Institute, for his continuing guidance
(A . 1)
and useful discussions in carrying out the present
eq. (12) is reduced to
study.

The author also wishes to express his apprecia-

(A . 2)
tion to Mr. Tokunaga and all members of Naga-

saki Experimental Tank for their cooperation in If the boundary condition and the initial
carrying out this investigation. values are given as follows

Reference q=0 at y•}B

(A . 3)
1) Sasajima, H. and Tanaka, I., "On the q=g(y) at t=0

Estimation of Wake of Ships," Proceedings where B expresses the width of wake, the solu-
of the Eleventh ITTC, Tokyo, 1966. tion of the partial differential equation (A.2) is
2) Hoekstra, M., "Prediction of Full Scale obtained as10)
 52 日本 造船 学 会 論 文 集 第146号

(A .4)
where

(A.5)
Fig. A.2 Variation of velocity at edge of
On the other hand, neglecting the pressure wake
gradient, eq. (A.2) is reduced to

(A .6)

The solution of eq. (A.6) is obtained under the

same condition of eq. (A.3) as follows:

(A . 7)

Numerical calculations of eq.s (A.4) and (A.7)

were performed by use of the measured velocity
profile of the larger wing with smooth surface
as initial condition. Fig. A.3 Calculated velocity profiles with
The value of C was determined from the and without pressure gradient

measured pressure gradient as shown in Fig. A.1.

cited from reference (11).
(A .8)
Then the variation of ue is also shown in Fig. A.2.
The calculated velocity profiles with and
without pressure gradient are compared in Fig.
A.3. It is observed that the difference between
Fig. A.1 Pressure distribution along water them is very small. This means that the effect of
line, measurements by Himeno, Y. pressure gradient on the velocity distribution fi
and et. al. (11) can be regarded to be negligible for practical use.