You are on page 1of 13

Drag Force on Cylinders Oscillating

A. R. Anaturk1
at Small Amplitude: A New Mode
P. S. Tromans
A new semi-empirical model is derived for calculations of hydrodynamic damping
H. C. van Hazendonk (or drag) forces on smooth or rough cylinders oscillating at small amplitude and
high frequency. This model covers the attached flow regime where the conven-
C. M. Sluis tional Morison's equation fails to simulate the drag forces acting on circular
cylinders. It involves the calculation of the energy dissipated in the boundary layer
A. Otter on the cylinder surface. The empirical input data for the model are amplitude and
phase of the wall shear stresses on the cylinder. The model is verified against fluid
Koniklijke/ Shell Exploratie en Produktie Laboratorium,
force measurements from a test tank and published data. These experiments were
Rijswijk, The Netherlands; and
carried out at small-amplitude / high-frequency with smooth and rough circular
Technical University of Delft,
Civil Engineering Department,
cylinders. Data from the literature is also included for verification. Results from
Delft, The Netherlands the present work can be used to estimate hydrodynamic damping forces for the
analysis of jack-up response around resonant frequency. Other applications are
dynamic behavior of jackets during installation, SALMs, TLPs and risers.

1 Introduction in which
The response of an offshore structure may be classified as KM = 0.25cmpivD2 (la)
quasi-static or dynamic, depending on the relationship between
the natural vibration frequencies of the structure and the excita- C = 0.5pDcd (lb)
tion frequencies of the waves. Frequently, the designer is forced In Eqs. (1), (la) and (lb), D is the diameter of the cylinder
to tolerate some degree of resonant-type dynamic response at the (roughness may be added in case of marine growth), p is the
natural frequencies of the structure where the dynamic amplifi- water density, and cd and cm are the drag and inertia coefficients
cation of the response is large and directly controlled by damping that are obtained from the experiments. X is the displacement,
forces. Extensive use of jack-ups in the North Sea and interest in dX J dt is the velocity and d Xj dt2 is the acceleration of the
their application in water of up to 100 m depth has raised cylinder. The negative sign in Eq. (1) indicates that the force is
questions about the dynamic behavior of these structures. acting contrary to the motion of the cylinder. It is helpful to list
In general, a structure can oscillate as a result of the wave and the parameters governing this phenomenon. In the rest of this
current load. However, very little is known about the dynamic report we will consider a circular cylinder that is oscillating in
interaction between fluid and structure and its impact on the still water with a displacement
response of the structure. In the present design practice fluid
forces are calculated by using the conventional Morison equa- X = Asm(ut) (2)
tion, which consists of drag and inertia components. The validity
of Morison's equation has not been substantiated, however, Therefore, its velocity and acceleration are
when small-amplitude, high-frequency oscillations are involved. dX
In the present study, we shall concentrate on the fluid forces = A to cosfcoO (3)
v
F(t) when the structure oscillates at a small amplitude and high dt '
frequency, while the surrounding fluid is otherwise still. This and
may correspond to the lower part of the structure where the
d2X
structure is excited by the surface waves that penetrate only to a
—— = -Aorsimot) (4)
limited water depth. According to Morison, the fluid force acting
on a unit length of a circular cylinder then becomes
in which co is equal to 2TT/ and A is the amplitude of the
2 oscillation.
dX dX dX
F(t)/L=-KA -C The fluid force, F, acting on a cylinder oscillating in a still
dt 2
~dt -Z
It (0 fluid, is a function of time (/)> period of the oscillation (T),
diameter of the cylinder (D), maximum cylinder velocity (Au>),
water density (p), kinematic viscosity of water (u) and rough-
Currently, NAM, Velsen-N, The Netherlands ness of the cylinder wall (/•*).
The dimensionless parameters involved can be summarized: K
Contributed by the OMAE Division for publication in the JOURNAL OF
OFFSHORE MECHANICS AND ARCTIC ENGINEERING. Manuscript received by the is the Keulegan-Carpenter number (or 2-wA/D), RE is the
OMAE Division, May 29, 1990; revised manuscript received June 11, 1991. Reynolds number and ks/D is the dimensionless roughness.

Journal of Offshore Mechanics and Arctic Engineering MAY 1992, Vol. 1 1 4 / 91


Copyright © 1992 by ASME
Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
The Keulegan-Carpenter number K and the Reynolds number RE in which / is 1/ T. /3 is proportional to the square of the ratio of
can be combined to give the oscillatory Reynolds number cylinder diameter to laminar boundary layer thickness.
According to existing experimental data, different flow regimes
may occur around the oscillating cylinder depending on the
RE JD2 parameters (see Fig. 1). The most important change in the flow
(5)
characteristics occurs when the flow changes from the attached to

SEPARATION LINES BASED


10'
Vortex shedding - ON MINIMUM c d

Present Exps (Smooth)

Re g =500
Present Exps (Rough)
Separated flow r z

Sarpkaya 86 (Smooth)

:• 1 0 1 -

.. Re x =100
Offshore
Engineering
Interest

^r*^ a : a "br,
5 10l

Laminar flow region


ion — • -"*{• —)
10"

5-10" I I I I | 1 1 1—I I I II | I I ~]—i—l i i i


3 4
10' 10 10 10 = 10<
Oscillating Reynolds Number
Fig. 1 Flow around oscillating cylinders

Nomenclature.

layer (cylinder surface) (m)


A = amplitude of cylinder dis-
placement (m) (N/m 2 ) 5 = boundary-layer thickness
a = local amplitude of fluid dis- vr,vg = radial and azimuthal velocity
placement just outside bound- in boundary layer (m/s) h = finite distance from cylinder
ary layer, relative to cylinder u, v = horizontal and vertical veloci- surface (m)
(m) ties in x, y coordinates (m/s) v = kinetic viscosity (m 2 /s)
T = period of oscillation (s) Fd{t) = drag force (or damping force) p = mass density of water
/ = frequency of oscillation (Hz) per unit length (N/m) (kg/m 3 )
io = angular frequency (2 -K / T) Fd = amplitude of drag (or damp- TW = wall shear stress (N/m 2 )
(rad/s) ing) force per unit length \j/ = phase shift between wall
D = cylinder diameter (m) ~ (N/m) shear stress and velocity out-
dX j dt = cylinder velocity (m/s) Fd = amplitude of drag (or damp- side boundary layer (deg)
d2XJdt2 = cylinder acceleration (m/s 2 ) ing force) in nondimensional /3 = oscillatory Reynolds no.
U = local velocity peak just out- form (fD2/v)
side boundary-layer, relative W = rate of work done per unit K = Keulegan-Carpenter no.
to cylinder motion (m/s) time and unit area (W/m 2 ) (2TTA/D)
/ = time (s) W* = mean work done per unit time Re = local Reynolds no. (Ua/v)
r,6 = cylinder coordinates, radial, and cylinder length (W/m) Re6 = local Reynolds no. based on
axial directions fw = friction factor; fw/(l/2PU2) boundary layer thickness
x, v = local coordinate system at- r* = roughness size (m) (U5/v)
tached to cylinder surface ks = Nikuradse roughness (about RE = global Reynolds no.
dp/dx = pressure drop along boundary twice of roughness thickness) (AwD/v)

92 / V o l . 114, MAY 1992 Transactions of the ASME

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
the separated type. In fact, Morison's equation was developed to using the theory developed by Stokes. The force on a sinu-
match the peak forces of a separated flow. When the amplitude soidally oscillating cylinder is then expressed by
of oscillation becomes very small,,flow around the cylinder may
be attached and the forces cannot be described by the classical F{t)/L= - 0 .25TT pD2o)(Aw)(Fdcos(wt) - ^sin(co^))
Morison's equation. Within the attached flow regime, the bound-
ary layer may be laminar, transitional or turbulent. (6)
At small amplitudes of oscillation, flow around the cylinder is in which Fd and F,- are the dimensionless drag and iner-
laminar. With increasing amplitude, and hence local flow veloc- tia coefficients [1]. Stokes [2] and later Wang [3] gave these
ity, the initially laminar boundary layer becomes turbulent. It is coefficients for (3 > 1 and K < 1 as
probable that different parts of the boundary layer on the cylinder
surface may be in different flow regimes. This is illustrated in ra=4(7r/3) ' +0(p-i)+0(K) (7)
Fig. 2. The region a around 6 = 0 can be turbulent-smooth. The
size of this region strongly depends on the ratio between local and
oscillation amplitude and roughness size. Region b is turbulent 1/2 [
rough and c is laminar. Region c contributes least to overall drag /7.= 1+4(TT/3) + <9(/3) + 0(K) (8)
forces. Further increase in oscillation amplitude eventually re- in which O(-) represents "of order."
sults in flow separating from the cylinder surface. Separation
may start in regions a, b or c.
A typical velocity profile based on flat plate oscillations is
shown in Fig. 3. When the boundary layer around the cylinder is
laminar and attached, the forces on the cylinder can be calculated

Control volume (see Fig. 3)


r

t0cos 8

Flow Regimes:
a) Turbulent smooth
b) Turbulent rough
c) Laminar

Direction of motion
Acocos cot
Fig. 2 Direction of the wall shear stress and its components in the direction of motion

Control volume

8P
8x

Boundary layer
profile

-*• X
a d6

Fig. 3 Control volume showing the flow up to y - h


Journal of Offshore Mechanics and Arctic Engineering MAY 1992, Vol. 1 1 4 / 93

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Brouwers [4] showed by a perturbation expansion in powers of circular cylinders. The essence of the model is that the mean rate
K that the solution is still valid for even higher order solutions of of the work done by drag or damping forces must be equal to the
K. He concluded that higher order solutions of the flow field do rate of energy dissipation in the flow. For an attached flow at a
not introduce any additional terms in the description of the practical Reynolds number virtually all the energy dissipation is
hydrodynamic force as obtained from Stokes formulas. In any in the thin boundary layer on the cylinder surface.
case, the K-/3 region in which the boundary-layer is laminar and The model proposed here is based on linear-momentum equa-
Stokes solution may be valid, especially in the large jS-region, is tions for a uniform flow over a flat bed. The effects of the
very limited. Therefore, Stokes' solution is not of much use for various terms such as convective and centrifugal terms become
practical purposes. negligible in the linearization procedure. The effect of the insta-
Since we are mainly interested in the damping (or drag) bility described by Honji [10] and Hall [5] is neglected. Using
forces, Eqs. (6) and (7) may be combined to give the peak drag these equations, a simple expression can be obtained for the
force per unit length of the cylinder as effect of wall-shear stresses. From this, one may easily obtain
the drag forces.
Fd = irpDw2A v /(2u/co) (9)
2.1 New Drag-Force Model. The model is based on the
in which a) is 2 irf. following assumption: for a cylinder oscillating at high Reynolds
Very little work has been performed so far to identify the numbers and attached flow, the boundary layer thickness (5) is
region to which Stokes' solution is restricted. In any case, the very thin (typically less than 1 mm at high /3); and therefore, so
critical point where transition begins also defines the limit for the long as the boundary layer remains attached, the flow outside of
laminar flow solutions, since the governing equations are no the boundary layer can be assumed to be represented by unsteady
longer valid above this limit. A careful analysis of the governing potential flow. The boundary layer may be laminar, transitional
equations shows that increasing centrifugal forces may cause or turbulent, depending on the time or the location around the
instabilities in the boundary layer. Such an analysis has been cylinder. We will neglect the transition zone since this occurs in
made by Hall [5], who analyzed the stability of the Stokes layer a relatively small region. Since the boundary layer is relatively
on a cylinder for very small amplitudes of oscillation (K -> 0) and thin, any surface roughness even that of a nominally smooth
a very large oscillating Reynolds number, j3. He concluded that surface may be sufficient for the boundary layer to become
the flow becomes unstable to small perturbations when K ex- turbulent. As a result, the nature of the boundary-layer flow may
ceeds a critical value that is a function of /3 given by be controlled by the roughness and the Reynolds number.
We may use linearized equations to describe the boundary
K cr = 5.775/3"'/ 4 (l + 0 . 2 0 5 / r 1 / 4 ) (10) layer. Thus, the momentum equation is
For a flat plate, shear flow instability may also occur. In that du 1 dp 1 drx
case the critical line is given by Kamphuis [6] as (12)
It p dx p dy
Re*
K cr - just outside the boundary layer this reduces to
fltpl (11)
2VP dUe 1 dp
in which the flow becomes unstable [7] at Re6 = 90 (86-100) (13)
dt p dx
and becomes turbulent at Re6 « 500.
For an oscillating cylinder it is possible that instability is due Work done in the boundary layer per unit area of cylinder
to centrifugal or shear-flow effects; this depends on the position surface per unit time is
in the K-/3 diagram, Fig. 1.
bp_
Depending on the ft parameter, the flow may separate from
the surface of the cylinder even if the boundary layer is still
laminar. This point may correspond to a minimum in cd, the
W
-I' dx
udy (14)

drag coefficient, as suggested by Sarpkaya [8]. At high Reynolds in which h is a finite distance from the cylinder surface (see Fig.
numbers a turbulent boundary layer separation is more likely to 3). In (12), (13), (14) u, p and rxy are ensemble-averaged
occur. terms such that turbulent fluctuations are excluded. Substitute
In Fig. 1, separation lines (where cd in minimum) from (13) into (12) and (14)
Sarpkaya's [8] and Justesen's [9] experiments have also been dUe rh
included. Hall's transition line given by Eq. (10) is shown in W=p- < / udy (15)
Fig. 1 to indicate possible regions where flow characteristics dt 7 n
may change. For a smooth cylinder, the boundary layer around and
the cylinder is more likely to be turbulent at high Reynolds
numbers and the separation is therefore turbulent. For a rough du dUe 1 dr
(16)
cylinder separation is expected to be turbulent for most of the Yt dt p dy
region.
When the boundary layer around the cylinder is (partly) Integrate (16) with respect to y from 0 to h, and with respect to
turbulent and attached, the flow field around the cylinder cannot time
be determined purely theoretically. Therefore, the objective of
this study is to develop a semi-empirical model to calculate udy = hUe ,dt (17)
p J
forces acting on an oscillating circular cylinder when the flow is
attached and the boundary layer is perhaps partly turbulent and where rw is the shear stress at the wall.
partly laminar (see Figs. 2, 3) depending on the local flow Substitute (17) into (15),
parameters such as K, /3 and roughness.

2 A New Drag Force Model for The Attached


W=p-
dt
h-Ue frw dt (18a)
Flow Regime
Since the product of dUe/dt and Ue must have a long-term
In this section we will describe a new drag force model mean of zero, it cannot contribute to the mean rate of energy
developed for small-amplitude oscillations of smooth and rough dissipation. The overall mean rate of dissipation may be obtained

94 / V o l . 114, MAY 1 9 9 2 Transactions of the ASME

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
by integrating the remainder of the equation over time and the in which Fd is the peak force when the cylinder oscillates with
cylinder surface Ucos(o)t). Therefore, the work (per unit length) done by this
force can be given as
D fT f2irrdUe f
dd-dt (186)
W* = FdUcos2 (oit)dt (31)
where T is the oscillation period and 6 azimuthal angle (see Fig.
2). When Ue and TW are pure harmonics of frequency w or after integration
Ue = t/cos u>t (19) (32)
T„ = f COs(<J)T + t^) (20) This work should be equal to the work done by the viscous
and (18) becomes forces acting on the cylinder surface expressed in (28). There-
fore, the drag force per unit length for a linear relation becomes

wd*- = DT u(e)-f(e)cos^de (21) 2Wd*


Fd = •• 8 DpU2f* (33)
U
where tp is the phase angle between Ue and TW.
For laminar flow in which / * can be obtained from (29).
It is convenient to make this force dimensionless with the
^ = 7T/4 (22) coefficient of Eq. (6); i.e. divide (33) by 0.25irpD2w( Aco) to
get a nondimensional force
f/ = v 2 pvU/8. (23)
- 32 A
For turbulent flow i/< and f must be obtained from measure- (34)
ments. For a flat plate it is usual to write •K D

f, = pfwU2/2 (24) Quadratic Form (Morison). In this case one may formulate
the forces by using a quadratic variation with the cylinder
where fw is an experimentally obtained friction factor [11, 12, velocity
13] that depends on the local roughness and Reynolds number
FdM = 0.5pDcdU2\cos(ut)\cos(ut) (35)
R e = Ua/v (25)
per unit cylinder length. Again, the work done by this force can
where U and a are local particle velocity and amplitude. also be calculated via
The distribution of the peak velocity U in the potential flow
just outside the boundary layer is given by
W£ = - ( 0.5pDcdU2\cos(o>t)\cos(ut)dt (36a)
U=2Aoscosd (26) T Jr.
Substitution of (24) and (26) into (21) gives
Wd*={2/3ir)pDcdU2 (36b)
Wd* = 4Dp(Aci)3 r fw(d)cos3d cos ^ d6 (27)
•'o and equating it to Eq. (28) one obtains a drag coefficient

Thus, the mean rate of energy dissipation is c r f =67r/* (37)

Wf = 4Dp(Aa)3f* (28) We note that the drag coefficient cd and dimensionless damp-
ing force Fd can be related by equating (36) to (32) to obtain
where
Fd=(S/3*3)Kcd (38)
/*= /* fw(d)cos3eCostd6 (29)
2.3 Wall Shear Stress: Friction Factor and Phase Angle.
We note that fw(6) and \p are functions of 6 and depend on local To calculate / * , one needs the friction factor fw and the phase
flow conditions, defined by the local Reynolds number, Re. angle \p between the wall shear stress and the local external
velocity. We mentioned earlier that it is possible to neglect
2.2 Linear and Quadratic Force Derivation. The mean curvature and centrifugal effects under certain assumptions. If
rate of energy dissipation expressed by Eq. (28) must equal the these assumptions hold, one can adopt fiat-plate wall shear stress
mean rate of work obtained from the drag or damping force on such as (24). In (24), fw, the friction factor, must be obtained
the cylinder. from the experiments. i/s the phase angle between wall shear
As we discussed in Section 1, there are various ways of stress and external velocity, can also be extracted from the flat
representing the fluid forces acting on oscillating structures. One plate measurements. In this section, we will discuss the available
way is simply to assume a linear relationship between the drag data and their use in the present drag-force model.
force and the velocity of the cylinder. Another way is to use the
quadratic relation of Morison to represent the drag forces acting 2.3.1 Friction Factors. In general, the friction factor fw,
on the oscillating cylinder. In the following, we will attempt to for a uniform, oscillating flow over a flat bed, can be defined by
derive the drag force both ways.
TH, = 0 . 5 / W P C / 2 (39)
Linear Force. In this case the drag forces may be assumed
to be linearly related to the cylinder velocity The variation of fw over a cycle is shown in Fig. 4.
For a laminar boundary-layer flow, an analytical expression
F
d = Fdcos(0lt) (30) for the wall shear stress can be derived by using Stokes solu-

Journal of Offshore Mechanics and Arctic Engineering MAY 1992, Vol. 1 1 4 / 9 5


Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
0.016 is exceeded by 10 percent of the grain size. In the present study,
ks will be used as twice of the roughness size, (/•*).
For smooth beds, the friction factor proposed by Jonnson may
be adopted [14]
0.008
/ w , = 0.09*Re- (41)

Swart [15] proposed the following form for the rough bed
turbulent flow region

/ l v r = 0.00251*exp 5.21" for a/ks> 1.57

-0.008
(42 a)
fwr = 0.3; for a/ks< 1.57 (426)

-0.016 These friction factors ((40)-(42)) are shown in Fig. 5. It is


90° 180° also shown there how the friction factor used in the present
cot analysis varies around the cylinder depending on the local ampli-
Fig. 4 Variation of the shear stress during a cycle (see Sleath [17]) tude. The dark line in Fig. 5 actually shows a typical calculation
where flow type changes with d. The regions mentioned there
correspond to the regions shown in Fig. 2.
tion. Therefore, in this case, the friction factor can be defined It is worthwhile to remark that above we have tried to select
analytically as formulations for the friction factor that fit the experimental data
available. Considerably more effort would be required to use
experimental data directly for every a/ks and Re number
J iW (40) and the consequent improvement in the results would not be
Re A cos 6 substantial.
in which Re is local Reynolds number defined by (25). 2.3.2 The Phase Angle Between Shear Stress and Exter-
For a turbulent boundary-layer flow however, there is no nal Velocity. The Stokes analysis shows that, in a laminar
analytical description of the wall shear stress. Fortunately, there flow, the shear stress on the cylinder wall has a phase lead of 45
are some models available that describe the velocity-profile in deg relative to the external velocity. According to Batchelor [16]
the boundary layer under certain assumptions from which fric- this is a direct result of the effect of the pressure gradient on the
tion factors can be derived. Furthermore, there are friction work done by the forces in phase with the velocity.
factors that are obtained from flat-plate experiments. In a turbulent flow, the phase shift depends on the relative
For a turbulent boundary-layer flow, different relations de- roughness a/ks and on the Reynolds number. According to
scribing flow over smooth or rough beds, depending on the ratio Sleath [13], for Reynolds numbers larger than 700, the phase
a J' ks (in which ks is the Nikuradse roughness of the surface) shift \j/ depends only on a/ks. His experiments showed that \j/
are available. Note that ks is defined by 2D90, the diameter that decreases with a until a/ks reaches about 70, and for a/ks>

a/ks =2

a/kg = 10
Friction coefficients used for different 6 angles (see Fig. 2)
!
a/ks = 100

0.01 a/ks = 1000


a/ks =5000

0.001
10.000 100.000 1.000.000 10.000.000 100.000.000
Reynolds number (Re)
Fig. 5 Friction coefficients for different flow regimes

96 / V o l . 114, MAY 1 9 9 2 Transactions of the ASME

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
70, it is roughly 25 deg. Recent data at high Reynolds number For straightforward calculations, \p is assumed to be constant
[17] show that in the case of rough beds for a / ks varying from for the turbulent part of the boundary-layer flow, although a
5 to 150, \f/ varies from 19.5 to 24 deg. variation with 8 could be used. In a later section, we will
Jonsson suggests that \j/ = 25 deg for a j ks= 100 and \p = 11 investigate the influence of \p by using Q, 15, and 25 deg.
deg for a/ks= 1000. Kajiura [18] presents two graphs for the
phase shift; one for smooth beds and one for rough beds. For 2.4 Test of the Drag Force Model for Fully Attached
smooth beds, \f/ decreases from \j/ = 45 deg at Reynolds = 104 to Laminar Flow. If the model is correct it should at least
\j/ = 12 deg at Reynolds = 108. For rough beds, i/' = 45 deg for reproduce the drag force result obtained by the Stokes solution
values of a / ks up to 50 and then decreases till i/- = 18 deg for for fully laminar attached flow.
ajks 5000. For laminar flow, \// = ir /4 and the friction factor is given in
Jensen [19] presented results for smooth beds similar to those Eq. (40). If one substitutes these into Eq. (29), then
of Kajiura. All phase-shift data mentioned above are plotted in
two graphs: Fig. 6 shows the data for smooth beds and Fig. 7
shows the data for rough beds. Unfortunately it is impossible to /,* = (43)
obtain precise values for \p from this information. It is clear,
however, that the phase shift in turbulent flow in less than 45
deg. For turbulent flow on a smooth bed, a phase shift of 10 to By substituting (43) into (33), one obtains the drag force as
15 deg seems to agree quite well with all the experimental
investigations, for local Reynolds numbers larger than 106. For Fdl=-wpDu2A ^(2v/u) (44)
turbulent flow on a rough bed, \j/ depends strongly on a/ks and
fluctuates around 25 deg. which is the same as equation (9) derived by Stokes.

Laminar

i Mini; 1—i M mi] 1—i i i i n i | 1—i i i i n i | 1—i M i n i | 1—i i i n n |


100 1000 10.000 100.000 1.000.000 10.000.000 100.000.000
Reynolds number
Fig. 6 Phase shift between shear stress and outer velocity for a smooth bed

SU - i

40-

3 0 - \ Sleath '84 ^~^~-^^^


—• Kajiura
^^Jonsson

. .-.• Sleath '88


20-

a Jonsson
10-

I I I I I i i i i i
0 100 200 300 400 500 600 700 800 900 1000
Relative roughness a/ks
Fig. 7 Phase shift between shear stress and outer velocity for a rough bed

Journal of Offshore Mechanics and Arctic Engineering MAY 1992, Vol. 1 1 4 / 97

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
2.5 Summary of the New Model. The work done by the low as Hall's line. This is the attached-flow region where the
oscillating flow on the unit length of the cylinder per unit time is model described in Section 2 could be applied. Therefore,
the experimental results of interest are those of Sarpkaya,
Wd* = 4Dp(Au)3f* (28) Rodenbusch and Justesen.
In Fig. 8, we compare drag coefficients obtained from Sarp-
in which / * is kaya's experiments [8] with the new drag-force model. These
experiments were performed in a U-tube in which he used test
cylinders with diameters of 7.4, 8.6 and 24.5 cm, yielding /3 as
/ * = / ' f„(e)cos3e cos yi>de (29)

Integration over Q has not been performed since flow may be


partly laminar or partly turbulent over the circumference. In (29)
/„, and \j/ are functions of 6.
If the force used in practice is a linear relation such as Eq. (6),
then the nondimensional force coefficient is

^ 32 A
(34)
•w D

If a quadratic relation between the cylinder velocity and the


drag force is used (such as in Morison's equation, Eq. (1)), then
the drag force can be estimated via

67T/* (37)
2-10"1 10u 101
Keulegan-Carpenter Number: K
The friction factors, fw in (29) can be calculated from Eqs.
(39) to (41). For the laminar part of the flow (see Fig. 2), we use Fig. 8 Sarpkaya's smooth cylinder experiments in a U-tube (D = 7.4,
8.6, 24.5 cm); ks = 1.5 mm, ^ = 25 deg are used in the work model
7r/4 for \j/ and for the turbulent part \j/ m a v vary around 25 deg.
Limits of the proposed model cannot be determined theoreti-
cally. In our view the model is valid as long as the flow is
attached. More detailed theoretical efforts so far available [20]
are not reliable enough to implement or even to indicate the
limitations of the present method. Experiments seem to be the
only means of determining the range of validity of our approach.

3 Comparison of Theoretical Results With


Experimental Data
In this section we will attempt to compare the new drag force
model derived in Section 2 with the results of experiments with
smooth and rough circular cylinders.
Experimental results are obtained either from literature or
from our own measurements. For a large portion of the K-/3
plane, sufficient published data were not available, particularly in 2-10-2 10-1 ' 10o

the high /3 and low K region. For this reason a set of experi- Keulegan-Carpenter Number: K
ments has been performed to obtain data for both smooth and Fig. 9 Rodenbusch's experiments with a smooth cylinder (D = 1 m)
rough circular cylinders. These experiments are reported in (constant /3 is extracted from constant RE experiments 0 = 440000)
detail elsewhere [21].
3.1 Smooth Circular Cylinders. Smooth circular cylin-
ders are relatively difficult for our model to handle since the 101
: Stokes solution
degree of roughness is needed when the flow is not laminar. In - New drag model
fact roughness may be important even for a nominally smooth
member since the boundary-layer thickness is very small espe-
cially at high |S values. Therefore any industrial pipe or off-
shore-structure member that was assumed to be "smooth" may
actually be "rough" when small-amplitude and high-frequency 10°-
CO

oscillations are considered. In the following comparisons we will "^^ ^ ^


o
always assume some roughness for the turbulent, attached part of O
o
2 o
the flow. The roughness is estimated arbitrarily for different o
° V r*/D
experiments depending on the cylinder diameter used.
^ ^ ^ - 1 5 ° !• 0.003
3.1.1 Data From Literature for Smooth Cylinder. Most «. " . ^25° J
10 " - - 2 5 ° 0.001
of the past experiments were performed at relatively low /3
values 0(10 3 ). The study using the highest Reynolds number is \ ,
5-10"' 1 1 1 1 1 1 If -T-TT-r
Rodenbusch's [22], although he did not include very small 2-10" 2 1CT 1
10° 101
amplitudes of oscillation (i.e., K < 2). Keulegan-Carpenter N u m b e r : K
In this study, we are mainly interested in the experiments that Fig. 10 Justesen's experiments with a smooth cylinder (D = 0.5 m)
extend from the separation line and preferably cover a region as RE = 500,000

98 /Vol. 114, MAY 1992 Transactions of the ASME

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
1035, 1380 and 11240, respectively. The cd values measured phase angle, tp. It seems that phase angles below 25 deg do not
are represented by dashed lines. The Stokes solution is repre- have a major influence on the results.
sented by the chain-dotted line. The solid lines represent the So far, we have compared our predictions with measurements
drag coefficients calculated by the new drag-force model. At a for smooth cylinders. Our model is very sensitive to the ratio of
certain K value our model diverges from the Stokes solution. boundary-layer thickness to the roughness size. However, for
This depends very much on the roughness assumed for the smooth cylinders it is very difficult to estimate this ratio and
smooth cylinders. For the experiments shown in Fig. 8 we therefore it is difficult to predict results for smooth cylinders
estimated the degree of roughness arbitrarily to be about 0.75 although some improvements on Stokes solution are obtained.
mm (giving Nikuradse roughness 1.5 mm). The improvement The difficulty is greatest for low (3 experiments such as
that the new drag-force model offers compared to the Stokes Sarpkaya's.
solution is significant. Another uncertainty in our calculations was the limit of our
Figures 9 and 10 compare our model with the high RE model. We usually stopped our calculations at about the point
experiments of Rodenbusch [22] and Justesen [9]. Rodenbusch where the minimum cd is obtained in experiments, assuming this
performed his tests at SSPA's wave tank with a test cylinder of 1 to be the separation point. However, this is an assumption and
m. Justesen performed his tests at the Technical University of requires further investigation by flow visualization tests. In Fig.
Denmark with a test cylinder of 0.5 m. Since in both cases the I minimum cd points obtained from the present experiments are
diameter of cylinder is large, the roughness size of about 0.75 shown together with Hall's transition line and flat plate instability
mm (ks is about 1.5 mm) may be reasonable. line. If the cd minimum identifies separation, then our method
The results shown in Figs. 9 and 10 again show a significant can be used for K less than 3 and ft greater than 10 4 .
improvement over the Stokes solution. To detect the influence of
roughness, we used a different roughness size of 0.25 mm 3.1.2 Data From Present Experiments for Smooth Cylin-
(giving ks = 0.5 mm). Although roughness affected the results der. Some of the drag coefficients measured are shown if Fig.
of our model somewhat, improvement over the Stokes solution II for various /3 parameters.
still remains substantial. We also investigated the effect of the If (3 = 29000, the small-K experiments produce drag coeffi-

c)
. -1
P r*/D \)/
o
Stokes solution
. i . . i
P r*/D \|/ Stokes solution
29000. 0.002 25. 100000. 0.002 25.
New drag model New drag model

10°-
:e
OCD a>
o ^ \
OQj O
0 ° ra
0
a o
o
o 0
1
10" - o
o
in-2: ' "I ' i i i i i . , i
10° 101 10"1 10° 10 1
Keulegan-Carpenter Number: K Keulegan-Carpenter Number: K

b) d)
P r*/D V|/ — Stokes solution P r*/D \|/ Stokes solution
69000. 0.002 25. 130000. 0.002 25.
New drag model New drag model

c
0)
o
1 10°-J
o : >,^
o 0
en O O
ra 0
D
X > . o o 0
0
0
.ooo
\ o o
So 8
Iff1- \

5-10"2: \
10"1 10 u 10 1
Keulegan-Carpenter Number: K Keulegan-Carpenter Number: K
Fig. 11 Present experiments with smooth cylinders (D 0.4 m)

Journal of Offshore Mechanics and Arctic Engineering MAY 1992, Vol. 1 1 4 / 99

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
cients close to Stokes solution. One may expect attached laminar literature. The most detailed small-amplitude experiments were
flow perhaps up to K about 2. Our model does not provide a performed by Sarpkaya [23, 24] (see Figs. I2(a, b)) albeit at
good prediction of the experiments at low JC, perhaps as a result smaller j3 than required. These experiments were performed in a
of the arbitrary roughness height of 0.75 mm used in the U-tube with the test diameters of 14 and 17 cm yielding j3 of
calculations. When K lies around 3, prediction by the new drag 3598 and 9354, respectively. Roughness to diameter ratio was
model is better than the Stokes solution. kept as 1/50 giving ks/D ratio of about 0.04. In Figs. 12(«, b)),
With increasing (3, the drag coefficients measured diverge we investigated the effect of the phase angle, \p, for the turbulent
from Stokes solution even at smaller K values. The improvement part of the flow over the cylinder surface. It is clear that the
our new model offers over Stokes solution is remarkable. phase-angle effect is not significant. The improvement of our
In all measurements with smooth cylinders we observed a drag-coefficient calculations over the Stokes solution is signifi-
decreasing drag coefficient with increasing K up to about 2.5. cant.
After this point an increase in the drag coefficient is observed. To be able to verify the model for high Reynolds-number and
This trend is similar to that found in measurements made by low-K experiments, we compared our results with Justesen's
others at smaller (3 numbers. experiments. These experiments were performed with a 50-cm-
diameter test cylinder at constant RE(=/3*K). Roughness-to-
3.2 Rough Circular Cylinders. It was easier to perform diameter ratio was kept as 1/50 yielding a K s /Z> ratio of
comparisons with the rough-cylinder experiments since rough- approximately 0.04. Figures 12(c, d) show comparisons for RE
ness (which is an input in our model) is known. So far, there is numbers 250,000 and 500,000. The agreement of our model
no theoretical model available to predict the transition region with the experiments is encouraging.
from laminar to turbulent flow over rough cylinders. In our
opinion, this region is very small, therefore there would hardly
be any interest in practical applications. 3.2.2 Data from the Present Experiments for Rough
Cylinders. Since even fewer data were available for rough
3.2.1 Data From Literature for Rough Cylinders. There circular cylinders at high (3 values, a rough cylinder has also
are not many experiments with rough cylinders available in the been tested at our facility. The same cylinder was used as before,

; P r*/D Stokes solution Stokes solution


0.02 RE rVD \\i
3598 250000. 0.02 25. New drag model
Experiments Experiments
10
S
X

N ®x s
N

x * c
X

/
o

o
o

N ^S&.-O0 8 io°- V ^ 3
a - *

\ ^ 1 5 ' o
Drag

"^
X "^.^
"V.
-
1 --^
10' - •^
irr1- 1 J ~i—r 1
I ' ' 5-10" 2;
r—i—r- I I I I I I I I I I I I I* 1 1—i—i i i i 11 1 1

10" 10u 101 2-10"' 10" 10u 10'


Keulegan-Carpenter Number: K Keulegan-Carpenter Number: K
a) Sarpkaya's rough-cylinder experiments in a U tube (D= 14 cm) c) Justesen's rough-cylinder experiments (D=0.5 m)

P r*/D . — Stokes solution Stokes solution


9354 0.02 — New drag model RE r*/D \|/ New drag model
. - - Experiments 500000. 0.02 25. Experiments
10'

8 10U
o
CO
Q

10
5-10' -rr,— 1 1— | 1 1—i i i I i i|
10"1 10° 101 2-10" 10"1
10° 101
Keulegan-Carpenter Number: K Keulegan-Carpenter Number: K
b) Sarpkaya's rough-cylinder experiments in a U tube (D= 17 cm) d) Justesen's rough-cylinder experiments (D=0.5 m)
Fig. 12 Experiments with rough cylinders, from literature

100 / V o l . 114, MAY 1 9 9 2 Transactions of the ASME

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
but with extra roughness attached to the surface. The base 4 Conclusions
diameter of the cylinder was 0.41 m and the average roughness
size was 1 cm. Barnacles of varying sizes were attached to the In this study, we developed a new drag-force model that can
cylinder surface in a random pattern. be used to estimate the drag forces acting on an oscillating
The drag coefficients measured are shown in Fig. 13 for cylinder at high Reynolds number (fi) and low Keulegan-
varying /? parameters. The experiments with rough cylinders do Carpenter numbers (K < 3). It is assumed-that flow in this region
not show a sharp dip in drag coefficient, unlike those with the is attached and Stokes solution is not valid. The model conve-
smooth cylinders, and therefore it is not clear where the point of niently provides drag coefficients that are applicable to the
separation lies. This flatness in cd is possibly a result of the fact present practice, i.e., Morison's equation.
that the drag forces on a rough cylinder are more nearly quadrat- The model is a semi-empirical method which uses the friction
ically related to the velocity in the attached flow regime, as is factors measured for flat plates. Our investigation showed that
predicted by our model. Minimum cd values obtained from the the phase angle between velocity and shear stress for the turbu-
present experiments (with some difficulty) are shown in Fig. 1. lent part of the boundary layer does not influence results signifi-
Although agreement is not very good at low /3, it improves cantly below 25 deg, and therefore it can be kept around 25 deg
with higher fi. This may be due to the neglect of the Honji for practical purposes. The only other input required is the
instability route to turbulence in our model. friction factor for flat plates.
The improvement of the new drag force model over Stokes The method has been tested on the data available from the
solution particularly at high /3 is quite substantial. For example, literature and from our experiments. Although it was not so
at (3 = 130,000, see Fig. 14, Stokes solution gives a cd of 0.03, simple to use the model for smooth cylinders because of the
whereas our model (in agreement with the experiments) gives a uncertain thickness of the surface roughness, the results still
cd of 0.22, for a smooth cylinder, which is about a factor of 7. showed a significant improvement over Stokes solution. The new
The corresponding ratio for the rough cylinder is about 24. The drag-force model showed good agreement with experiments for
impact of this ratio on the damping forces acting on offshore rough circular cylinders.
structures, and consequently on the response of the structure, The increase in force calculated by the new method over
may be substantial! Stokes solution is estimated to be about 24 times at high (3

a) c)
10'
: P rVD V r*/D
1 33000 0.02 25
Stokes solution p
102000 0.02
V
25
Stokes solution
- .
-

c •

(D .5>
o o cyPo?,
'% 10°-
o : 0° O Io 10°. 0 0 OO
o
oO) j
\\ o
o s V
o
0 °
a ra \ o o o
a \\
%\
o° o
0 \
\
\
10-1- 10"'- \
\ \
5-10-2' 1 1 1 1—J— 1 M | 1 -1—1 1 1—,—, 5-10' 1 1—I- • ' ' 1 1 ' '
10" 10° 10 1 10" 1 10° 10 1
Keulegan-Carpenter Number: K Keulegan-Carpenter Number: K

b) d)

Stokes solution
New drag model

Keulegan-Carpenter Number: K Keulegan-Carpenter Number: K


Fig. 13 Present experiments with a rough cylinder (D = 0.41 m)

Journal of Offshore Mechanics and Arctic Engineering MAY 1992, Vol. 1 1 4 / 101

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
a) c)
0.4 0.4

P r*/D V Stokes solution


P r*/D 4* Stokes solution 102000 0.02 25 New drag model
0.3 33000 0.02 25 New drag model 0.3

0.2 0.2

|
i
2 0.1 I 0.1

0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Keulegan-Carpenter Number: K Keulegan-Carpenter Number: K

b) d)
1
0.4-1 P rVD Stokes solution
68000 0.02 25 • New drag model Stokes solution
New drag model

0.3-

| "

0.1

0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Keulegan-Carpenter Number: K Keulegan-Carpenter Number: K
Fig. 14 Present experiment with a rough cylinder compared with (Eq. (34))

numbers. The impact of such a difference on the damping forces Journal of Waterway Harbours Coastal Engineering Division, ASCE 101
acting on offshore structures and therefore to their response may (WW2), 1975, pp. 135-144.
7 Tromans, P. S., "Stability and Transition of Periodic Pipe Flows,"
be substantial. Ph.D. thesis, University of Cambridge, 1978.
The results obtained from the present work can be used in 8 Sarpkaya, T., "Force on a Circular Cylinder in Viscous Oscillatory
dynamic analyses of jack-ups. Fluid damping is a critical vari- Flow at Low Keulegan-Carpenter Numbers," Journal of Fluid Mechanics,
able for estimating the fatigue life of such structures. Other Vol. 165, 1986, pp. 6 1 - 7 1 .
applications of the present work can be summarized as dynamic 9 Justesen, P . , "Hydrodynamic Forces on Large Cylinders in High
Reynolds Number Oscillatory Flow," BOSS Conference, 1988.
behavior of jackets during installation, SALMs, TLPs, and
10 Honji, H., "Streaked Flow Around an Oscillating Circular Cylinder,"
risers. Journal of Fluid Mechanics, Vol. 107, 1981, pp. 509-520.
11 Jensen, B. L., Sumer, B. M., and Fredsoe, J., "Transition to Turbu-
References lence at High Re Numbers in Oscillating Boundary Layers," Progress Report
66, Institute of Hydrodynamics and Hydraulic Engineering, Technical Univer-
1 Bearman, P. W., Downie, M. J., Graham, J. M. R., and Obasaju, sity of Denmark, Feb. 1988, pp. 3 - 1 5 .
E. D., "Forces on Cylinders in Viscous Oscillatory Flow at Low Keulegan- 12 Sleath, J. F. A., Sea Bed Mechanics, John Wiley and Sons, New
Carpenter Numbers," Journal of Fluid Mechanics, Vol. 154, 1985, pp. York, 1984.
337-356. 13 Sleath, J. F. A., Turbulent Oscillatory Flow Over Rough Beds,"
2 Stokes, G. G., "On the Effect of the Internal Friction of Fluids on the Journal of Fluid Mechanics, Vol. 182, 1987, pp. 369-409.
Motion of Fluids," Trans. Camb. Phil. Society, Vol. 9, 1851, pp. 8-106.
3 Wang, C. Y., "On High-Frequency Oscillating Flow," Journal of 14 Jonnson, I. G., "Wave Boundary Layers and Friction Factors,"
Fluid Mechanics, Vol. 32, 1968, pp. 55-68. Proceedings of the 10th Conference on Coastal Engineering, Tokyo,
4 Brouwers, J. J. H., and Meijssen, T.E.M., "Viscous Damping Forces 1976, pp. 127-148.
on Oscillating Cylinders," Applied Ocean Research, Vol. 7, No. 3, 1985. 15 Swart, D. H., "Coastal Sediment Transport," Computation of Long-
5 Hall, P., "On the Unsteady Boundary Layer on a Cylinder Oscillating shore Transport, Delft Hydraulics Lab., Report R968, Part 1.
Transversely in a Viscous Fluid," Journal of Fluid Mechanics, Vol. 146, 16 Batchelor, G. K., An Introduction to Fluid Mechanics, Cambridge
1984, pp. 347-367. University Press, 1967.
6 Kamphuis, J. W., "Friction Factors Under Oscillatory Waves," 17 Sleath, J. F. A., "Phase Difference Between Shear Stress and Veloc-

102 / V o l . 114, MAY 1992 Transactions of the ASME

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
icy," Note for KSEPL, University of Cambridge, 1988. 21 Anaturk, A. R., "Hydrodynamic Forces at Small Amplitude and High
18 Kajiura, K., " A Model of the Bottom Boundary Layer in Water Frequency," Applied Ocean Research, Aug. 1991.
Waves," Bulletin of Earthquake Research Institute, Vol. 46, 1968, pp. 22 Rodenbusch, G., and Kallstrom, C. "Forces on a Large Cylinder in
75-123. Random Two-Dimensional Flow," OTC, Houston, Tex., May 1986.
19 Jensen, B. L., "Experimental Investigation of Turbulent Oscillatory 23 Sarpkaya, T., "Oscillating Flow About Smooth and Rough Cylinders,"
Boundary Layers," Series Paper No. 43, Institute of Hydrodynamic and ASME JOURNAL OF OFFSHORE MECHANICS AND ARCTIC ENGINEERING, Vol. 109,
Hydraulic Engineering, Technical University of Denmark, 1988. 1987, pp. 307-313.
20 Fredsoe, J., and Justesen, P., "Turbulent Separation Around Cylin- 24 Sarpkaya, T., "In-Line and Transverse Forces on Smooth and Rough
ders in Waves," Journal of Waterway, Port, Coastal and Ocean Engi- Cylinders in Oscillatory Flow at High Reynolds Numbers," Naval Postgradu-
neering, Vol. 112, No. 2, Mar. 1986. ate School, Monterey, Calif., July 1986.

SAVE
With an ASME Standing Order Plan you can save as much as
50% on Technical Papers and Symposia/Proceedings and enjoy
the convenience of having your publications mailed to you
automatically and fast!

UP TO
You can start building a complete library in your area(s) of
interest in mechanical engineering by choosing one of the
following plans:

® All ASME Technical Papers

50%
ON
• All ASME Symposia/Proceedings
® Both Technical Papers and Symposia/Proceedings
• Technical Papers and/or Symposia/Proceedings
subject areas of mechanical engineering
in specific

PUBLICATIONS
AND
TECHNICAL
PAPERS
To receive a free brochure (with details of Plan savings) or to start a Standing Order Plan now, write or call Mr.
John Yelavich, ASME Customer Service, 22 Law Drive, Box 2350, Fairfield, NJ 07007-2350; 1-800-TH E-ASME(843-
2763). In New Jersey, call 1-201-882-1167.

Journal of Offshore Mechanics and Arctic Engineering MAY 1992, Vol. 1 1 4 / 103

Downloaded 23 Mar 2011 to 92.66.66.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

You might also like