INVESTIGATION OF DRAG & ADDED MASS

PROPERTIES OF MID-WATER ARCH STRUCTURE

FOR RISER DESIGN Presenter: L iangl i (Michael) L iIndustry Supervisor: Dan Brooker ( Intecsea, Perth)Academic Superv isor: Stuart Higgins (Curt in)

2

1. Introduction

3

2. Hydrodynamic Force Analysis Morison Equation

, Reference Area

Flow Velocity

Structure Volume

Fluid Density

4

3. Analyzing Using Existing code MWA Design – buoyancy cylinder, gutters

Ca & Cd Analysis◦ Code method - DNV-RP-C205 ◦ Numerical approach–

◦ Panel Method, Ca◦ Computational Fluid Dynamics, Cd

Compare to Code Method

Verify the results

5

4. Panel Method

Directions DNV-RP-205Circular Cylinder

DNV-RP-205Square Cylinder

Panel Method

Surge, ZZ 0.919 0.72 1.6Heave, YY 0.919 0.72 1.0Sway, XX 0.137 0.125 0.08

3D Model

ANSYS APDL

HydroD

Added Mass Matrix

Symmetry

Plane

6

5. CFD Simulations

Velocity Reynolds Number, Re

Circular Cylinder

Square Cylinder

CFD

0.5 m/s 1.8381x106 0.523 1.61 1.2076

1 m/s 3.6762x106 0.523 1.61 1.1978

1.5 m/s 5.5143x106 0.523 1.61 1.1464

2 m/s 7.3524x106 0.523 1.61 1.1396

3D Model

ANSYS ICEM

Fluent

Forces over time

Flow

Flow

7

6. Result Verification & Validation

Grid and Fluid Domain◦ Grid Quality: 99.87% OK◦ Wall interference◦ Meshing size,5.4 million cells

Turbulence Models

Compare to Published Cylinder Result

Compare to Published MWA Results

Flow

Symmetry Plane

8

Turbulence ModelOptions Turbulence Model Options Applications Limitations

RANS

Spalart-Allmaras Model (1-eq)Developed for aerospace applications, suitable for external flow and flow with adverse pressure gradient

Work well only on simple geometry. Only work under low Reynolds number flow

k-ε Model (2-eqn)

Suitable for internal flow and heat transfer applications

Limited to high Re flow condition, and not work well if flow encounters bluff bodies

k-ω Model (2-eqn)

Also developed for aerospace applications, suitable for external flow and flow with adverse pressure gradient, in both low and high Re flow

Sensitive to free stream turbulence

Laminar-Turbulent Transition Models (3-4 eqn)

Provide prediction of laminar to turbulent transition of boundary layer

Rely on mesh refinement near wall and inlet initial condition.

RSM (5 or 7-eqn) Work with any complex flows, including strong swirl and rotation

Most costly to run

SRS

SAS (RANS-LES hybrid)

Provide more detailed vortex shedding simulation (using von-Karman length scale)

High dependence on grid size and time-steps

DES(RANS-LES hybrid)

Provide more detailed turbulence simulation (turbulence length scale)

High dependence on grid size and time-steps

LES(RANS-DNS hybrid)

Used for research purpose on sophisticated geometry or small geometry scale

Only work with very fine grid size and small time-steps

DNS Used for 2D problems Most costly to run.0 5 10 15 20 25 30 35 40 45 50

0.0

0.5

1.0

1.5

2.0

2.5

2m/s Case Drag Coefficient Time History

2m/s k-w-SST

2m/s DES-k-w-SST

Time (s)

Cd Transient response

9

Turbulence Model

10

Compare to Published Cylinder Result

Cases Reynolds Number

Cd after finite length reduction (as pre DNV Code)

Present simulation hexahedron, 3D k-ω-SST

3.6762x106 0.314

Present simulation tetrahedron, 3D k-ω-SST

3.6762x106 0.313

DNV-RP-C205 (Book: Fluid-Dynamic Drag, 1965)

>1.0x106 0.523

Ong et al. 2009, 2D k-ε 3.6x106 0.368

Catalano et al. 2003, 2D k-ε 4.0x106 0.370

Shih et al, 1993, Exp ~3.5x106 ~0.33

Schewe, 1983, Exp 4.0x106 ~0.41

11

Comparing with MWA Results

Cd = ~0.85Re > 107

0 0.5 1 1.5 2 2.5 3 3.5 4 4.50

20

40

60

80

100

120

140

160Drag Force vs. Velocity^2

K-w-SSTLinear (K-w-SST)DES-k-w-SST

Velocity^2

Drag

(kN)

0 0.5 1 1.5 2 2.5 3 3.5 40

50

100

150

200

250

Drag Force vs V^2

Velocity^2

Drag

(kN)

Cd = ~1.17Re > 106 Cd = ~1.18

Re > 105

Cd = ~0.86Re > 105

12

Closing RemarksAdded Mass Coefficient• Ca varies with MWA design• Code method has limitations• Circular cylinder better than square cylinder• Panel method gives better prediction of Ca

Drag Force Coefficient• Cd varies with MWA design• Code method gives a upper-lower bound• Cd independent to Re, matches well to Morison eqn• CFD gives better prediction of Cd

Possible Design Improvement

Future Work – other flow directions, roughness effect, model testing

13

Cd comparison DNV code vs. Others

14

0 10 20 30 40 50 60 70 80 90 1000.0

0.5

1.0

1.5

2.0

2.5

Drag Coefficient Over Time

0.5m/s k-w-SST

1.0m/s k-w-SST

1.5m/s k-w-SST

2.0m/s k-w-SST

Time (s)

Cd

MWA and Cylinder Plots

0 20 40 60 80 100 1200.0

0.1

0.2

0.3

0.4

0.5

0.6

Cd of Cylinder

Time (s)

Cd

15

Quality Check by ICEM

16

Allocating 12080x6232=75282560 pixel map.

57773920 pixels filled, area = 57.7731

Projected Area, Solidworks and Fluent