CHÎEF
Leb. y.Tcchthce
oscliod
REPORT ON FRICTIONM4 RESISTAIWE RESEARCH
by
Dr. E. V. Telfer
L An examination of available plank data has been made and it has been found that turbulent data can be simply expressed as
1,2
3ko
(I +
L ) ( ) ) 1/3200B
VLin which B is the draught, when the p1nk has freeboard, or the half depth when the plank is fully submerge (as in the Froude
tests).
®
Is the specific resistance x loi.The above formula satisfactorily interpolates Froudes
var-nished plank tests which are fairly certain to have been made at
a temperature not exceedIng 100 C. which was likely to be the
water temperature at Torquay at the end of April 1872.
Froude's paraffin tests were most probably made during July and August 1872 when the tank water would have a fairly steady temperature of about 180 C0 This higher temperature appears to
be the complete and sufficient explanation of why the paraffin results shoed less resistance than the varnish,
Froudes tests aro thus brought in line with the extrapolator
and by their confirmation of the 3ko value give the minimum tur-bulent friction line without edge effect.
The formula given in (i) allows of a new concept of Froude's
extrapolation from plank to ship. Froude ws merely trying to
extrapolate the resistance of a 19-inch depth plank, to ship
lengths.
In other words he had a strake of platins 19 inches deep,
having two edges, and of ship length. He therefore vas deter-mining not the extrapolated resistance of a smooth ship, but,
granted his extrapoltion was in itself correct, the resistance
of a plated ship having flush butts and strakes 19 inches wide (i.e., one edge in 19 square inches of surface per inch run)0
Froude did not, however, make a ccrrect extrapolation0 He did, nevertheless, use at least one guiding principle0 He
ac-cepted that the total resistance at constant speed plotted to a base of total length LnUSt always be concave to the base line0
An alternative principle, that beyond 50 feet the resistance curve can be replaced by the tangent at 50 feet was referred to
by Froude as an upper limit0 The precise method used by Froude
does not appear to be on record, only the results of the method
are, of course, available0
:i
Let us consider these at a ship Reynolds of i0. For 400 feet the Froúde specific resistance Is 1.88, This value is that
obtained from our formula for a plate 400 feet long and 2 feet draught, I.e., by a plate having 4 square feet of wetted surface
per foot run per edge. Thus the actual value of the Froude
spe-cific resistance can be accurately accepted as applying to a
plated ship having strakes 4 feet wide and flush butts and rIvets No wonder the Froude ship frictional coefficients have resisted
attack for nearly 80 years
Actually, the average plate width for a ship 400 feet long
is now about 6 feet, although long after Froude's time 4 feet
would have been about right For 6 feet width the corresponding
Sspecific
the higher figure of Froude 1.88 (or of unity roughnessresistance reduces to 1.766. Thus, the acceptance ofas recom-mended in my1949
paper) should be a good approximation to the modern plated riveted surface ofships0
Thus, unity roughness would appear to correspond to a ship
plated with strakes one hundredth of the ship length in width.
A complete analysis has been made of temperature correction to model resistance in the light of the extrapolator for turbulent
and mixed flow. The fact that many tanks use too small a
cor-rection is fairly definite proof of the presence of mixed flow in
their normal routine testing. With stimulated turbulence becoming an agreed tank routine, temperature corrections will need to be
completely overhauled.
Previous (and in 1935 Internationally agreed) corrections do not distinguish between model and ship size. A single term
re-sistance fornula gives the same correction for ship and model,
Sbut
experience clearly suggests a much smaller correction frthe shIp.
Further, the 1935
correction does not distinguishbetween up and down to the standard yet the difference is
con-siderable0
Tables of kinematic viscosity and resistance change have been prepared over a wide temperature range for intervals of one tenth
degree centigrade0
The recent work of Allan, Conn and Emerson, and of those who contributed to the discussion on their papers, particularly
Lindbiand and Dawson,li are attempting to assess diff er-ences in terms of V/L , waterline angle and so on, when clearly mixed-flow conditions first demand consIderation of specific
resistance in terms of Reynolds number. Thus for mixed flow
K
It is
the
value of K we want to knov. For example, the PrandtlK value for Gebers 1908 tests was 1700. If the R difference
be-tween the unstimulated and the stimulated model at each Reynolds number divided by the Reynolds number be noted, i.e., the K value obtained, the average K value over the whole model speed range is
a much more fundamental factor (and likely to be independent of
size of módel) than is the plotting above referred to, The
ratio of the K values, say to the Prandtl 1700 standard (or to some
other appropriately assessed standard) may be called X , then I
-X
is a measure of the efficiency cf the turbulence stimulation.The value of K can be shown to equal
®T - ®L]x R0 where
.is
the turbulent specIfic resistance, F the laminar, bot at the critical Reynolds R at which turbulence begins. As Fand ® are calculable once
Re is known it follows that it is R0 which is the vital factor in the K determinatIon. R0should
therfore be known for every tank by testing a standard smooth plan and measuring Rc as has recently been done in Washington by
Couch. The corresponding K value incidently becomes 2k60. R0 thus obtained is a measure of the tank turbulence and will clearly be governed by the interval between consecutive runs. The maximum
repeatable R0 should be known for every tank.
6. Some work has been done on the better understanding
of pnt
or sand roughness in relation to the extrapolator0
Thus, since a rough plank has a constant specific resistance and a smooth plank is given by
= a + b ( ) 1/3
VL
we can sy that the rough resistance is given by
a + o (f
) x 1/3
where f
(4)
x (.)
1/3 must be a constant, being the transversedimension of the roughness. For this to be so, f
( ) must take
the form (iL_) 1/3 and thus
a 4 ( vp ) 1/3 x ( i) ) 1/3 w VL a + o
L4
)l/3
T 3V
From which we see that the relative roughness is the ratio
of the roughness to the length. of the surface. Hence the same
roughness on planks of increasing length will have progressively
decreasing specific resistance just as found by W. Froude,
If the constant specific resistance is presumed to continue without transition to the smooth surface specific resistance then
this equality becomes
and hence
VL = b3 . L
e3
which is the Reynolds number at which the rough plank departs
from smooth plank resistance0 This relation may also be written
Vt b3
2)
which thus gives the critical roughness Reynolds number. This relation can be checked against various data, e, f. Perring I.C.S.T.3.
l9k8 p.
108,
from which V. _ say 125, andtherefore b/c 5. Hence as the zero edge Yalue of b is 31..0, the
value of e is
68.
Thus roughness resistance is given by= a -+
1.2
+6834/a
IL
a lower limit of V can be chosen from Perringts data, say 93.
Thenc75and
il= 1.2 +
These expressions can be checked against the Prandtl-Schlich-ting formula
l03/189 +
1.62 log
L)2.5
by plotting this to a base of . The values given by Von
The new expression is quite close to the Von Karman data0
These latter differ from the Prandtl-Schlichting by a greater
amount thanthe difference between the new formula and the Von
Karman data0
The data also apparently co firm the retention of the 1.2 term as the limiting value of as _- O.
L
An extremely interesting check on the structure of the
fore-going; i.e0, on the use of the L2 limit and the 3,j roughness
parameter is got by analyzing W0 Froude's rough
Tplank tests0
ifl
this case it is not necessary to know the exact value ofIt is sufficient to assume A = unity and to plot the sific re-sistances for the same actual roughness to a base of
34
1 When this is done straight lines can undoubtedly be passed ' throughthe spots and through the limit value of 1.2. Separate lines are,
of course, drawn for each of the roughnesses, fine, medium and
coarse used by Freude0 Later it is proposed to show that these separate lines accurately coalesce when the measured roughnesses
of Froude are included in the relative roughness term0
70 The treatment of the roughness problem in the extrapolator
technique is now somewhat clearer0 If reference is made to ad joining diagram (figure (a)) ft is seen that for every Reynolds number there Is a relative roughness which, when saturated, causes
the specific resistance beyond that point to remain constant. It is thus possible to have a double calibration of the extrapolation diagram base in relative roughness as well as in Reynolds number function.
Now it appears probable from some work by Theodorsen on
roughened rotating cylinders that for lower densities of rough-ness the point where the resistance begins to increase beyond the
smooth surface value will remain unchanged0 Thus depending upon
density we can draw from the point on the smooth extrapolator diagram a series of rays between the horizontal ray for complete
roughness and the extrapolator for complete smoothness, Each of
these rays will correspond to a particular density of roughness. This ray arrangement would seem to be independent of the relative
roughness; i.e., the ray would be the same whatever relative roughness were chosen since in the extrapolator presentation the triangle formed by the completely rough and the completely smooth
lines are similar for all Reynolds numbers0 This facilitates
the generalization, for ve can draw a line parallel to the smooth
extrapolator through the orIgin, (figure b), then from the origin
further rays can be drawn corresponding to definite densities of
roughness (which have of course to be experimentally determined),
large near to the base0 Thus, to determine the roughness
addition for known density and roughness in any particular ship,
all that is required is to draw from that point on the calibrated base line correspondlng to the relative roughness, a line parallel
to the particular density ray0
This line should give the pcint roughness addition to the ship resistance; and is seen to imply an increasing relative addition
with increased Reynolds number0 There is still the edge
rough-ness effects to allow for; and these in principle can be repre-sented as rays below the base line corresponding to various ratios
of plate-width to ship length (figure e). Thus to combine pnt and edge roughnesses, ve proceed as before but note only the
inter-section of the line parallel to the pont roughness density ray
on to the zero ordinate0 Then from this point join to the point
on the edge ray below the relative roughness abscissae0 This line
will thus be the total roughness addition0 From the construction
it follows that the sum of the roughnesses may be practically
eonstant This will be the case in the normal riveted ship0 In the ship with welded butts and riveted seams and welded frames the edge roughness may predominate and the total roughness will become less with increasing Reynolds number0 With entirely
welded ships the roughness will, be practically zero0
One final point may be ent.ioned. A statistical analysis of roughness on actual ships (made available by the late G. S. Baker) shows that as riveted ships increase in size their relative rough-ness rapidly decreases but their roughrough-ness density definitely
in-creases. Similarly the ratio of plate width to ship length
be-comes smaller the longer the
ship0
The precise effect of these dfferenees may be to equalize
the total roughness between small, and large
ships0
The smallerReynolds numbers of small shIps appears to produce a
preponder-ating edge
resIstance which generally result in the small shipalways having a larger total roughness resistance than the smaller,
the net effect being probably to endorse for riveted ships the
plausibility of the unity roughness, ray concept advanced pre viously in 19k90
£ RO t) G ADO CCMBN ED N E - oç-LO W D N S T Y