A two fluid model of the Limpet oscillating water column

advertisement

A TWO-FLUID NUMERICAL MODEL

OF THE LIMPET OWC

CG Mingham, L Qian, DM Causon and DM Ingram

Centre for Mathematical Modelling and Flow Analysis

Manchester Metropolitan University, Chester Street,

Manchester M1 5GD, U.K.

M Folley and TJT Whittaker

School of Civil Engineering,

Queen’s University, Belfast

Acknowledgement

• EPSRC (UK) for funding the project

(grant number GR/S12333)

Background

• LIMPET: a wave energy converter based on the Oscillating Water

Column (OWC) principle.

• LIMPET installation on Islay,

Scotland (75kw).

• Small scale experimental trials at

Queen’s University, Belfast.

Background

• The problem involves both water and air flows, wave breaking, non-sinusoidal waves, vortex formation and air entrainment.

• Linear wave theory is not suitable for modelling such flow problems.

• A two-fluid (water/air) non-linear model

(Qian,Causon,Ingram and Mingham,

Journal of Hydraulic Engineering,

Vol.129, no.9, 2003) has been applied in the present study.

AMAZON-SC: Numerical

Wave Flume

• Two fluid (air/water), boundary conforming, time accurate, conservation law based, flow code utilising the surface capturing approach.

• Cartesian cut cell techniques are used to represent solid static or moving boundaries.

Governing equations

• 2D incompressible, Euler equations with variable density.

 t



Q

   

S

F .

n

 s

 

B

  where

Q

 f

I 

 u

 u

 v

 u

2  p p

T

, F

 uv

 f

I n x u

T

 g

I n y

, B and g

I 

 v

0

 uv

0

  g

 v

2  p

0

T v

T

 is the coefficient of artificial compressibility

Discretisation

• The equations are discretised using a finite volume formulation

Q i

V i

 t

  

B V ij

 

R

  i j

 k ( i )

F ij

D l j

Where Q i is the average value of Q in cell i (stored at the cell centre), V volume of the cell, F ij i is the is the numerical flux across the interface between cells i and j and and D l j is the length of side j.

Convective fluxes

• The convective flux (F

Riemann solver. ij

) is evaluated using Roe’s approximate

F ij

I 

1

2

F

I

    ij ij

R

L ( Q

 ij

Q

 ij

)

• To ensure second order accuracy,

MUSCL reconstruction is used

Q ( x , y )

Q ij

 D

Q ij

 r where (x,y) is a point inside the cell ij, r is the coordinate vector of

(x,y) relative to ij and D Q ij slope limited gradient.

is the

Time discretisation

The implicit backward Euler scheme is used together with an artificial time variable t (to ensure a divergence free velocity field) and a linearised RHS.

I m

V

I ta where

R ( Q n

1 , m

Q n

1 , m

Q

)

Q n , m

I m

D t diag

D

1 t

Q n

1 , m

1

V

1

D t

R

D

1 t

Q n

1 , m

Q n

1 , m

1

D t

1

D t

1

D t

D

1 t

The resulting system is solved using an approximate LU factorisation.

Computer Implementation

• A Jameson-type dual time iteration is used to eliminate t at each real (outer) iteration.

• The code vectorises efficiently with simulations typically taking about three hours to run on an NEC SX6i deskside supercomputer.

Boundary Conditions

Seaward boundary – a solid moving paddle (boundary) is used to generate waves (wave-maker).

Atmospheric boundary – a constant atmospheric pressure gradient is applied.

Spray and water passing out of this boundary are lost from the computation.

Landward boundary – a solid wall boundary condition is used for the landward end of the domain.

Bed and wave power device – modelled using Cartesian cut cell techniques.

Cartesian Cut Cell Method

Automatic mesh generation

• Boundary fitted

• Extends to moving boundaries

Cartesian Cut Cells

• Input vertices of solid boundary (and domain)

Cartesian Cut Cells

• Input vertices of solid boundary (and domain)

• Overlay Cartesian grid

Cartesian Cut Cells

• Input vertices of solid boundary (and domain)

• Overlay Cartesian grid

• Identify Cut Cells and compute intersection points.

Wave Generation

• Waves are generated using a moving paddle with prescribed velocity:

U=-0.2sin(2  t)

• 0.3m Still Water Level; 6.0m long wave tank

• Using 120x40 grid cells

• 10 waves simulated, starting from still water

Wave Generation

• Comparison with experimental results for free surface elevation at two locations

LIMPET OWC Simulation

• Wave Conditions: Regular waves with wave length L  1.5m , period T=1.0s and still water level H = 0.15m.

• Device located at about 2 wave lengths from the moving paddle

• 5 seconds simulated, starting from still water

Small Scale Test at QUB

LIMPET OWC Simulation

Free surface position and velocity vectors at T=4.0s

LIMPET OWC Simulation

Free surface position and velocity vectors at T=4.2s

LIMPET OWC Simulation

Free surface position and velocity vectors at T=4.4s

LIMPET OWC Simulation

Free surface position and velocity vectors at T=4.6s

LIMPET OWC Simulation

Free surface position and velocity vectors at T = 4.8s

LIMPET OWC Simulation

Free surface position and velocity vectors at T = 5.0s.

Conclusions

• Some initial results have been presented for simulation of LIMPET OWC device using a surface capturing method in a

Cartesian cut cell framework.

– The method is computationally efficient,

– Capable of modelling both water and air, as well as their interface

– Can handle both static and moving boundary easily.

• Detailed comparisons with the small scale test from QUB using the same wave conditions are in progress.

• The numerical model is generic and can be used to model a wide range of wave energy devices.

Download