Академический Документы
Профессиональный Документы
Культура Документы
SOO-S09, 2001
Offshore subsea pipelines must be stable against external loadings, which are mostly due to
waves and currents. To determine the stability of a subsea pipeline on the seabed, the Morrison
equation has been applied with prediction of inertia and drag forces. When the pipeline is placed
in a trench, the force acting on it is reduced considerably. Therefore, to consider the stability of
a pipeline in a trench, one must employ reduction factors. To investigate the stability of various
trenches, we numerically simulated flows over various trenches and compared them with
experimental data from PIV (Particle Image Velocimetry) measurements. The present results
were produced at Reynolds numbers ranging from 6 X lOS to 3 X lOS based on the diameter of the
cylinder. Quasi-periodic flow patterns computed by large-eddy simulation were compared with
experimental data in terms of mean flow characteristics for typical trench configurations (W/
H= I and H/D=3, 4). The stability for various trench conditions was addressed in terms of
mean amplitudes of oscillating lift and drag, and the reduction factor for each case was suggested
for pipeline design.
Key Words: Trench, Stability of Pipeline, Large-Eddy Simulation, Particle Image Velocimetry,
Drag Reduction Factor
economical method for transporting materials. To ristics of oscillating lift and drag coefficients, and
insure their continuous operation, adequate reduction coefficients for each trench slope can
criteria for design must be provided. A lack of also be applied to estimate the stability of a
necessary information for their design has pipeline with scour process occurring in trench
introduced high safety factors which in tum have sections.
increased the construction cost. Information con-
cerning hydrodynamic loads can help to reduce
2. Numerical Simulation
the uncertainty and therefore the overall cost
(Knoll and Herbich, 1980).
The correct estimation of external forces on the 1.1 Governing equations and subgrid-scale
subsea pipelines has been a topic of considerable modeling
interest and discussion. Many experiments have Development of an LES formulation applicable
been conducted both in the ocean and in the to a non-orthogonal grid system begins with the
laboratory for wave and current channels. The incompressible continuity, and Navier-Stokes
test results have frequently shown extreme scatter equation in integral form.
(Garrison, 1980).
[-U.-ndS=O (I)
The strategies for dealing with turbulence are
many: turbulence models (educated guesses about a -
at!apudQ + !spuunds --
the needed Reynolds stresses), statistical theory of
turbulence (to gain fundamental understanding), =fsT-ndS+fapbdQ (2)
and direct numerical simulation. The LES is one Direct filtering of the continuity equation
of the more promising modes of numerical yields:
simulation of turbulence. One of the main differ-
aUIe=O (3)
ences between the conventional method of RANS aSIe
(Reynolds-Averaged Navier-Stokes) solver and
Direct filtering of the momentum equation
the LES technique is the averaging process in-
yields:
volved. In this study, unsteady, viscous flows over
a two-dimensional cylinder in a trench are com- p[aUIe+buleu,)]=- ap + adlel + ariel (4)
at aXI aXle aXI aXI
puted by using the LES technique with
Smagorinsky's eddy viscosity model combined where riel =- p( UIeUI-UIeUI + UIeU'1 + UIU'1e
with a truncated deductive subgrid-scale model.
The turbulent time-mean flow around a subsea + U'leU' I) is called the subgrid-scale stress tensor.
pipeline resting on the ocean bottom is not sym- The filtered momentum equation IS now
metric; therefore, a non-zero mean lift must exist solvable if we provide a model for riel. The
together with waves (sumer, 1997). There have subgrid-scale stress tensor, riel, can be
been numerous studies to develop the criteria to decomposed into the subgrid-scale Leonard,
predict the hydrodynamic loads imposed by ex- cross, and Reynolds stresses based on Gaussian
ternal waves and currents. The purpose of this filtering. Smagorinsky (1963) was the first to
research is to investigate the effect of trench slope propose a model for the subgrid-scale stresses.
and depth on the hydrodynamic force acting on a His model assumes that they follow a gradient-
pipeline. Turbulent flow in a trench on which a diffusion process, similar to molecular motion. It
two-dimensional pipeline model rests was is still the most popular algebraic eddy viscosity
numerically computed and experimentally veri- model, with riel given by
fied using the PlY technique in a circulating
water channel. The reduction in force coefficients rlel=2I1TSIeI, IIT=pCs2Lf I S I (5)
due to sheltering effect was obtained and reported where liT is the subgrid eddy viscosity, C is the
for steady current in this study. The characte- Smagorinsky constant, Slel is the resolved scale
502 Seungbae Lee. Sung- Wook Jang, Chul H. Jo and Sung-Guen Hong
T I
r
lID
3D \
wan
f, '(I4o+Q~' 0_
The equation of continuity was transformed collocated, structured grid system. A major
into an equation for computating density in this advantage of this method is that it intrinsically
study. This approach of working directly with the preserves mass, and material fluxes both on local
density variable is termed DEFCON for Density and global scales.
Equation Formulation of Continuity equation. The numerical integration starts with the as-
The governing Navier-Stokes equations are sumption of an integration profile for the state
integrated by Finite Volume Method (FVM) for a variable. The CONDIF (Runchal, 1987) scheme,
which is stable and second order accurate, is
employed in this research (see Runchal(1987. In
this work, the Alternating Direction Implicit
(ADI) method was used, which solves the set of
algebraic equations in three sweeping directions.
This temporal discretization of the ADI method is
proven to be unconditionally stable in a linear
sense. But it may not be accurate in time if the
time step is large due to neglecting the third-order
term which is essential to factorization. The de-
tails of the formulation have been described by
Lee et al. (1999) along with some validation
Fig.:1 Schematic diagram of pipe in trench studies.
To numerically simulate turbulent flows in the
trenches, H'-type grid systems of 141 x67, 150X
67, and 160X67 (in x and y directions) were used
with the inflow and outflow boundaries placed at
8D and 240, 20D and 40D, 32D and 640 for the
cases of W/H = 1.0, 3.0, and 5.0, respectively (see
Fig. 1). As an exit boundary condition, the sim-
plified convective one among non-reflecting
boundary conditions was also employed (Hayder
and Turkel, 1995).
-.<,
2
6tH
. AII' \
0
0
4070
~
V JI
..,m
Fig.6 Effects of boundary layer thicknesses on drag reduction coefficients, and CD-CL characteristics for HI
0=3, WIH= I. and Reo=6 X 1<J3
Flow Around a Pipeline and Its Stability in Subsea Trench 505
8 0 C85
(;
u,
'"
e
-
0 0 "'"
D.2I '0
'E
O~
.!I!
0.2S
0
0
."
= O~
---.
.2
13 0000
"a:.
:>
0.2.
-401 -407 -4OS <lOS no' r1d 3"1 0'
c. Re.
02'
10
Tin e (sec)
t5 2. 00 05 '0
Tome (sec)
2.0
Fig. 7 Effectsof Reynolds numbers on drag reduction coefficients, and CD-CL characteristics for 8/H =0.1, HI
0=3, W/H=I
. o.
~----k=-.
1-- - -- - - - - - - -
CJ~ 01
Q
CJ 0.0
4.
42
4.L..- -'- .J
o 10
Time(see)
(a) W/ H = I.O
o.
I- - - -- -----h.-.- 03
f - - -- - - - - ....- - - - 02
f--- - -- - - - - - _ .-
1- -- - - - - _ .- - Q
<J 00
f-- - - - - - - _ ._. _
--- 4 .
42
20 30 010 ~ ClO 70
T1me(oee)
(b) W/H=3.0
L. 03
02
o~ at c,
~ - a
- - ---- -
_ . _
Q
<J 0 0
:-<~~~~~~.
.0 1
42
I ~ ~ 30 ~ ~ m ~
I Time(see)
(c) W/ H = 5.0
Fig.8 Mean-streamline patterns and unsteady CD-CL characteristics for each W/H with H/D=3, 8/H=O.I,
and ReD=6 X lOS fixed
020 r---.,....-----....------~~ same density with water were seeded in the tunnel
~
o
u.. 0 1. e _ - - - - - e after static electricity was suppressed.
ell
a
'0 018
The visual image was captured by a Charge
Coupled Device (CCD) camera (SONY-
;:., XC77RR) and stored on a tape via a video
1:S 0'4
E
8 recorder. The captured images controlled by
U 0_'2
c:
.2
AOM were sent to odd and even fields and
g 0.10 processed by aq image board (DT3155 . 640 X480
-g e
It: 001 L..-......... -'-- --''--...J pixels) on a host computer (250MHz CPU). After
2
Openln9 Length (WIH)
the removal of background noise and the com-
Fig. 9 Drag reduction coefficients vs.opening length pensation of rotation. the searching area was set
(H/D=3.0, 8/H=O.I, and Reo=6X lOS) to have a radius of 25 pixels with correlation area
Flow Around a Pipeline and Its Stability in Subsea Trench 507
0.'
/H11>2
0.0 ~""lr'\""'\r'\,r- I/"I/,'\/,,'\ "'\ J""l "" "'\"'\,,",, J~\l
..
e 0.30
A 20 simulation of a circular cylinder in a free-
If stream is known to under-predict the formation
CIl
.3 0.25 length of the primary vortex while over-
-
j
o
020
predicting the strength and the shedding fre-
quency. However, vortex shedding may be
!
c
o
01'
suppressed for a cylinder near a wall. Span wise
structures are also assumed not to modify the
g 0,'0
characteristic flow patterns in the trench
e--- e
considerably.
! 006L-......... '- .......- l
The mean flow around a near-wall cylinder is
2 3
Cover Depth (HID) not symmetric; therefore, a non-zero mean lift
must exist. When the cylindrical pipeline is placed
Fig. 11 Drag reduction coefficients vs, cover depth
ratios(W/H=l.O, 8/H=O.I, and Reo=6X in a trench hole, the forces are reduced
10') considerably. Both the drag and the lift forces are
reduced by a factor 5 to 10, depending on the
radius of 30 pixels. Optimized searching of cross position of the pipe in the trench hole. This is due
-correlation coefficients was conducted to deter- to the sheltering effect by the trench. It was
mine the velocity vectors by using Eq. (7) (Raffel, reported that in the case of a same sized trench
et.al, 1997). hole but with a steeper slope, the reductions in the
forces are much larger.
112 Even though the incoming flow is steady, the
~(fi-li)(gi- gil cylinder in the trench may experience an
;=1
n2 nZ
(7)
~(fi- Ji)2~(gi- gi)2
oscillatory flow, and an additional parameter
i=1 i=1 called Keulegan-Carpenter (KC) number should
be considered. It is defined by KC = U Tw/D.
4. Results and Discussion When KC is small, separation behind the cylinder
in the free stream may not occur even at high
The wake flow behind a circular cylinder has Reynolds numbers (Sumer, 1997).
been considered as a fundamental study about the The trenches considered in this study may be
wake of a bluff body of arbitrar shape. This
1
problem, which is called Strouhal shedding or
classified as open cavities because a shear layer
separated from the corner of the backward facing
Karman vortex, has many engineering slope hits on the other side of the forward facing
applications in flows around rows of tubes in heat slope (Tracy and Plentovich, 1997). Therefore,
exchangers or marine structures. oscillatory behavior of lift and drag forces is
508 Seungbae Lee, Sung- Wook Jang, Chu/ H. Jo and Sung-Guen Hong
observed even in a steady current. A cylinder The phase characteristics between Co and CL
placed on the sea bottom would feel the are depicted, along with drag reduction factors, in
approaching flow turbulence that is generated Fig. 6 for different incoming boundary layer
within the bottom boundary layer. In the case of thicknesses. As the boundary-layer thickness
a cylinder in a trench, external turbulence acts on increases, both the mean lift and drag coefficients
the cylinder indirectly via oscillatory motion. The tend to decrease continuously in the absolute
boundary layer approaching the trench was as- sense for the steepest trench among the types
sumed to be laminar in this study and the flows considered (H/D=3, Reo=6 X loa).
may become turbulent in the trench. In Fig. 7, the dependence of drag reduction
The parameters of geometrical importance may coefficient and CD-CL characteristics on
be listed as the boundary layer thickness (8), the Reynolds number can be found. As the Reynolds
opening length (W), the depth of trench (H), and number increases, the mean lift coefficients
the cylinder diameter (D). Therefore, three non- increase, while the mean drag coefficient de-
dimensional parameters 8/H, H/D, and W/H are creases in the absolute sense for the case of 8/H=
formed. The non-dimensional numbers of 0.1, H/D=3, and W/H=1. The K-C numbers
dynamic importance in this study are the are found to be approximatly 12.0 for the range of
Reynolds (Re) and Keulegan-Carpenter (K-C) Reynolds numbers considered.
numbers. The mean flow patterns and time histories of Co
The effect of incoming boundary-layer thick- and CL are compared for different opening
ness on oscillatory motions in the trench was lengths with the trench depth ratio, incoming
numerically studied for the cases of 8/H=0.1, O. boundary-layer thickness ratio, and Reynolds
15, and 0.2 with H/D=2.0, W/H= 1.0, and Reo= number fixed as 3.0, 0.1, and 6 X loa, respectively
6 X loa. To figure out the Reynolds number effect (see Fig. 8). The oscillating amplitudes of lift
on the lift and drag, the case of Reo=6 X 103, coefficient for the case of W/H=3.0 are found to
which is the same condition as experimental be greater than those for the widest opening case
measurement, and the cases of 6 X 10 and 3 X I rf' of W/H =5.0, due to the unsteady nature of
were implemented in the numerical tests. To esti- separated shear-layer when W/H is equal to 3.0,
mate the stability of a pipe for each configuration, which means more susceptibility to external
numerical tests were conducted for the cases of H/ disturbances. The drag reduction coefficients be-
0=2, 3, 4, and W/H = I, 3, 5 at the Reo=6 X loa come greater as the trench slope decreases, as
and 8/0=0.1. shown in Fig. 9. Figures 10 and II also indicate
that the reduction factor decreases as the cover
4.1 Mean flow patterns in open trench depth increases for the case of W/H = 1.0, 8/H =
The computed and measured velocity vectors 0.1, and Reo=6X loa.
are compared for H/D=3 and 4 with Reo=6X
loa and 8/H=0.1 in Fig. 5. The separated shear 5. Conclusions
-layer in the fore part of the trench produces
several vortex motions, and the primary vortex The reduction factors for various trench slopes
induces reverse flow over the cylinder. This jet- and cover depths at three Reynolds numbers were
like flow interacting with the separated shear- investigated numerically and experimentally.
layer creates oscillatory motions in the trench. From the numerical/experimental results, the fol-
lowing are observed:
4.2 Force coefficients for pipelines in open (I) As the trench slope and the cover depth
trench ratio (H/D) increase, the reduction factors de-
The drag reduction coefficient is defined as the crease. Less scouring is expected.
drag force on a cylinder in an open trench divided (2) The external forces acting on the cylinder
by the drag force on a cylinder on sea bottom. get reduced as the laminar incoming boundary-
Flow Around a Pipeline and Its Stability in Subsea Trench 509