Sei sulla pagina 1di 32

JHU-SPH Modeling of Water Waves

B.D. ROGERS1 & R.A. DALRYMPLE2,


1 School of Mechanical, Aeronautical & Civil Engineering, University of Manchester
2Department of Civil Engineering, Johns Hopkins University

Funded by Office of Naval Research (ONR) & now EPSRC

K. STANSBY1 & D.E. LAURENCE1


Introduction
Compressible Sub-Particle-Scale (SPS) Model
(JHU)
2-D & 3-D simulations (JHU)
Wave Breaking & Landslide-generated
Tsunamis
Non-Reflecting Open Boundaries (JHU)
Further Work
MOTIVATION : Coherent Turbulent
Structures (CTSs) in wave breaking
Obliquely Descending Eddies Downbursting
(Nadaoka et al. 1989) (Kubo & Sunamura 2001)

(after Son & Kihm 2002)

Fingers (Narayanaswamy & Dalrymple 2002)

Fingers
(Plan view)
Sub-Particle Scale (SPS) Turbulence Model
Spatial-filter over the governing equations (compressible):

(Favre-averaging)
D ~
.u
Dt
~
Du 1 1
2~
~ P g o u .
f f Dt

= SPS stress tensor with elements:~


Sij = strain tensor ij 2 t S ij 3 S kk ij 23 k ij
2 ~

Eddy viscosity: t Cs l S S 2 Sij Sij


2 1/ 2

Smagorinsky constant: Cs 0.12 (not dynamic!)


See Lo & Shao (2002), Gotoh et al. (2002) for Incompressible SPS

Shepard Filter keeps free-surface clean (Panizzo 2004)


Captured downbursting-like
phenomenon in 2-D

Splash-up Reverse breaking Kubo &


Sunamura (2001)
Three-dimensionality?? hypothesis
3-D SPH-SPS
Weakly Plunging Breaking Wave

163,000 Particles (Extension of Gomez-Gesteira & Dalrymple 2004)


Vorticity in Horizontal Planes

Resolution x = 0.008m
Velocity Vectors
in horizontal planes

(interpolated)
Velocity Vectors
in horizontal planes
Velocity Vectors
in horizontal planes
Velocity Vectors
in horizontal planes
Velocity Vectors
in horizontal planes
Velocity Vectors
in horizontal planes
Velocity Vectors
in horizontal planes
Slices of an ODE

Horizontal Slice Vertical Slice


Velocity Vectors
in vertical planes
(interpolated)
CPU Run-times
2-D
97,915 Particles - 10 days
- 3 days for segment showed here
- 2.8GHz, 1GB RAM
3-D
163,000 Particles - 23 days

- 3.2GHz, 2GB RAM


Most pressing issues at the moment are:
parallelisation!!! (we now have a 2-D code at JHU)
coupling with Boussinesq model
(most of the particles & effort is offshore!)
1-D Simple Wave Case with
Open Boundaries ...
1020

1015

Density (kg/m)
1010

1005

1000

995

Very difficult to let the wave -0.5 0 0.5 1


Distance (m)
1.5 2 2.5

crests leave domain 1020


unaltered!
1015
Density (kg/m)

1010

1005

Underlying velocity flow 1000


of 0.8m/s
995
-0.5 0 0.5 1 1.5 2 2.5
Distance (m)
The problem of using SPH near boundaries
Interior of domain Beyond open
boundary

2h
Flow
Kernel Correction Methods
Open
Boundary

We chose the Reproducing Kernel Particle Method of


W.K.Liu et al. (Northwestern University) with Essential
Boundary Conditions
1-D Case with Open Boundaries ...
RKPM can also be used to specify
solid boundaries
RKPM can be used to enforce
Non-Slip BCs

Poiseuille Flow, Re = 10 Non-periodic Open BCs


Vortex leaving the domain

Still ironing out problems with this one


Non-periodic Open BCs
Basics are done, now our next
problem to solve ...

Specifying boundary values for arbitrary free surfaces

Deactivate?
Landslide-generated Tsunami
Test case for evaluation of numerical models
at the 3rd International Workshop on
Longwave Runup Models, 2004

Sliding Wedge Experiments


by Raichlen & Synolakis,
provide wave gauge
and run-up data

(Used by Philip Liu and


Tso-Ren Wu for model
validation)
3-D Landslide-Generated Tsunami
Landslide-induced Tsunami
Surface Waves
3-D, run 30, wave gage 1 data
0.2

0.1
free-surface (m)
0
0 0.5 1 1.5 2 2.5 3
-0.1

-0.2
experimental data
-0.3
SPH 2-D
-0.4 SPH 3-D
-0.5

time (s)
Good Experimental data available
Huge drawdown resolution is main issue
resolution is less than 2-D (x = 0.1m)
with periodic boundaries at the moment
Incorrect SPS effects of non-dynamic Cs?
Conclusions and Further Work
Described approach to capturing CTSs in
SPH
Compressible SPH-SPS scheme for better
description of viscous flows
Our model predicts reverse breaking &
downbursting in 2-D
3-D JHU SPH-SPS model captures ODEs
model is too coarse and needs much finer
resolution
Sediment transport has also been attempted
at JHU by graduate student Shan Zou
Immediate Future Work (2)
Clearly, parallel codes are paramount

Coupling with a Nonlinear SWE model is essential to save time and effort
for modelling offshore (non-breaking) waves

Extension of Open Boundary conditions to free-surface flows

Better LES-type models? Dynamic SGS/SPS models are certainly


necessary near boundaries (e.g. Wedge/Landslide)
more physically correct energy transfer
between subgrid-scales and resolved scales
Issues to be addressed

Characterisation of SPS Turbulence


(To be done at Man.U. with Laurence & Stansby) Sub-
Particle scale motion and modelling of turbulence

Waves2005 conference VALIDATION TESTS


e.g. Obliquely descending eddies
BCs
Ackowledgements

JHU: Shan Zou, Muthukumar Narayanswamy


Andrea Panizzo
Moncho Gomez-Gesteira

Potrebbero piacerti anche