CFD Study of the Flow in the
Vicinity of a Subsea Pipeline
Khalid M. Saqr, Mohamed Saber, Amr A. Hassan,
Mohamed A. Kotb
College of Engineering and Technology
Arab Academy for Science, Technology and Maritime
Transport
1029 Abu Qir, Alexandria – EGYPT
k.saqr@aast.edu
1. Problem outlines
• Subsea pipelines are subjected to hydrodynamic
stresses due to marine currents
• These stresses may rupture the pipeline and
cause financial losses and environmental
hazards.
• There is a demand to improve the methods used
to protect subsea pipelines from hydrodynamic
stresses
• This paper presents a comparison between two
protection methods.
1. Problem Outlines
• Current protection methods
– Trenching/Burying the pipeline into seabed.
– Concrete weight coating.
– Concrete mattress adding.
– Rock dumping (covering).
1. Problem Outlines
• The proposed double barrier method
Pipeline
Barrier
Seabed
2. Methodology: Physical Model
• Computational Fluid Dynamics (CFD) model
L
Trench in seabed
b
U
H
a
Y
X
Trenching method
Pipeline
b

a
α
Barrier
b
Seabed
ranges from 0.1 to 0.75
a
Double barrier method
2. Methodology: CFD Approach
A survey of relevant
literature showed that
the current approaches
involve:
1. Two and three
dimensional models
2. Finite volume
framework
3. RANS turbulence
models
2.
Methodology: Governing Equations
• Continuity:
U i
0
xi
• Momentum: 

P

2Sij 
U jU i  u j ui  

x j
xi x j
(1)


12
k ij
• Reynolds stress closure:   u j ui  2T S ij 
3
(2)
(3)
• Turbulence models:
– k – ε model
Turbulence kinetic energy
T

 

U i k  




xi
x j 
k
 

 x j

k   T S 2   

(4)
2.
Methodology: Governing Equations
turbulence dissipation rate
T   

 

2
2
U i           C 1 T S  C 2 

xi
x j 
   x j 
k
k
– Eddy viscosity
T  C 
k2
(5)
Cμ = 0.09

– Realizable k-ε model
T

 

U i   

  
xi
x j 


 

C1  max 0.43,


5



 

 x j
C 


2
   C1 S   C 2 
k  

1
A0  AsU 
1

As  6 cos arccos 6W  W 
3

k

U *  S ij S ij   ij  ij
8Sij S jk S ki
S3
(6)
1  ui u j 
 ij 

2  x j xi 
2.
Methodology: Governing Equations
– k-ω turbulence model


Ui
k
xi
x j


*
  T
x j




Ui

xi
x j
T 
k




k    ij
U i   * k
xi


 


2









U





T
ij
i

x
k

x


j
j
5

9
3

40
9
1
*
 
  
100
2
*
– SST k-ω turbulence model
A hybrid model which applies the standard k-ε model in
the near wall region and k-ω in the main stream
region
2. Methodology:
CFD Model Reliability Check
Elementary computational model
VERIFICATION
Different grid resolutions
Compare flow field obtained by
different grids
Refine grid resolution
NO
Predictions agree ?
YES
Select the optimum grid
VALIDATION
Test turbulence model
Change model
NO
Best agreement
with measurements ?
Select best turbulence model
Optimize numerical scheme
Final Computational Model
2. Methodology: Validation
•CFD Model Validation
Comparison
between CFD
predicted pressure
coefficient using four
turbulence models
and experimental
measurements of [9]
on the pipe wall.
Cp 
P  P
1
U 2
2
3. Results: Flow structure
α=0
0.0
0.3
0.6
0.9
1.2
1.5
Flow
direction
180o
90o
α=0
270o
0o
Figure 5. Contours of normalized velocity magnitude and vectors over a bare pipe
Flow structure of the bare pipe
3. Results: Flow structure
α = 0.1
α = 0.1
α = 0.25
α = 0.25
α = 0.5
α = 0.5
α = 0.75
α = 0.75
0.0
0.3
0.6
0.9
1.2
1.5
3. Results: α = 0.1
3. Results: α = 0.25
3. Results: α = 0.5
3. Results: α = 0.5
4. Conclusions
1. It can be concluded that the double barrier method is a
prospective alternative to trenching at small aspect
ratios.
2. With the difficulties faced during the trenching process,
especially when the pipeline route passes a rocky
terrain, the double barrier method appears as an
efficient and reliable alternative.
3. The present work also reveals that the low-Reynolds
number turbulence models (k-ω) performs better than
the high-Reynolds number models in the present
problem.
4. With proper construction of the non-uniform grid, a
number of cells as small as 3×104 can be sufficient to
produce accurate results.