Nicholas McClendon Final Presentation

Report
EFFECTS OF FLEXIBLE
MOTION ON TSUNAMI
WALL EFFICACY
HARP REU 2011
Nicholas McClendon, Rice University
Mentors: H.R. Riggs, Sungsu Lee, Krystian Paczkowski
Contents
• (1) Flexible wall study
• Background
• Models and Methods
• Adaptive Mesh Refinement
• Results
• (2) Adaptive mesh study
• Decomposed domain
• Whole domain
• (3) Domain decomposition study
• Setup
• Results
FLEXIBLE WALL STUDY
Background: Tsunami Walls
• Tsunami walls are designed
to reduce the damage done
by tsunami waves on coastal
structures
• Typical design parameters:
• Material
• Typically reinforced concrete, can
be steel
• Physical dimensions
• 1 – 12 meters high
• 100s of meters long
• Location
• Distance from sea affects forces
sustained
• Distance from the coastal
structures it’s protecting affects
damage done
Flexible Wall Study
Motivation
Goal
• Rigid walls are subject to
• To develop an
quick and unpredictable
failure
• Flexible walls absorb the
force of incoming waves,
increasing durability,
reducing impact force,
and allowing failure to be
predicted more readily
understanding of the
effects of flexible motion
on tsunami wall efficacy
for possible real-world
application
• How? Compare forces
sustained by a flexible
wall to the forces
sustained by a rigid one
Models and Methods
• Mathematical models:
• K-Epsilon turbulence
model
• Newtonian transport model
• Linear viscous fluid model
• Numerical solvers
• interFoam solver
• Multiphase solver for two
incompressible fluids
• Numerical technique is the
Volume of Fluid Method
• interDyMFoam solver
• interFoam + mesh
modification capabilities
• Flexible wall is modeled
using a rigid block with a
torsional spring applied
at the base
Adaptive Mesh Refinement
• AMR refers to
refinement of mesh
during computation
• Static AMR
• Superimposes finer subgrids on areas of interest
• Dynamic AMR
• Alters size, shape,
orientation, and/or number
of cells
• Layering, remeshing, and
smoothing
Domain Setup
• Setup:
• 20m x 1.8m x 0.1m domain
• 2-dimensional simulation
• 2 phases
• Liquid (dambreak scenario)
• Gas
• Boundary conditions:
• Rigid wall on sides/bottom
• Atmosphere on top (permits
inflow/outflow)
• 90,000-cell mesh
• Simulation’s dimensions
are reasonable to test
experimentally
• Test the simulation many
times, varying the spring
constant each time to
determine how the spring
constant and angle of
deflection relate to the
forces sustained by the
wall
Domain Setup
Flexible Wall Simulations
Results (spring constant variance)
Results (spring constant variance)
Results (spring constant variance)
Results (spring constant variance)
• Conclusion: Allowing for deflection of the wall to absorb
impact force of the wave does reduce the forces
sustained by approximately 1 percent per degree of
deflection.
• Flexible walls could provide effective impact force
reduction of tsunami bores
• Further study:
• Walls which use the impact and uplift forces of the tsunami bore to
raise into place
Results (mesh convergence)
Results (mesh convergence)
ADAPTIVE MESH STUDY
Adaptive Mesh Study
• Goal: Observe the
effects of adaptive
meshing on the
consistency of results of
a control case
• interFoam
• Wall is simply part of
boundary
• interDyMFoam
• Wall is a separate (fixed)
object
• Adaptive meshing used
Adaptive Mesh Study
interFoam, 8 processors
interDyMFoam, 8 processors
Adaptive Mesh Study
interFoam, 8 processors
interDyMFoam, 8 processors
Velocity magnitude (m/s)
snapshot at 1.05 seconds
Velocity magnitude (m/s)
snapshot at 1.05 seconds
Adaptive Mesh Study
interFoam vs. interDyMFoam, 1 processor
Adaptive Mesh Study
• Conclusions:
• Subdomain coupling for dynamic meshing
(dynamicMotionSolverFvMesh) needs to be fixed if it is to be used
in the future (otherwise cannot trust results from runs done in
parallel)
• On a single processor, results obtained from the control case using
static and dynamic meshing align very closely, so the use of
dynamic meshing tools is validated for the 1-processor case
DOMAIN DECOMPOSITION
STUDY
Domain Decomposition Study
• Setup a dambreak scenario
• Domain is 40m x 3.2m x 1m
• Dam is 1m x 20m
• Decomposed the domain
into 1, 2, 4, 8, 16, 32, 64,
and 128 subdomains
• Solved on HOSC using
interFoam for multiphase
• Measured several key
values (eg. splash height,
computation time) from
each trial
Credit also goes to Adam Koenig
and Trent Thurston for gathering
and analyzing data for this study.
Results
Number of
processors
1
2
4
8
16
32
64
128
Time to reach wall
4.40 s
4.40 s
4.40 s
4.35 s
-
4.40 s
4.40 s
4.35 s
Peak force on wall
22397 N 16696 N 14451 N 22259 N (5.2s)
(5.1s)
(5.1s)
(5.15s)
13748 N (5.15s)
-
Max splash height
2.04 m
(5.2s)
1.74 m
(5.1s)
1.60 m
(5.1s)
2.09 m
(5.15s)
-
1.64 m
(5.15s)
1.89 m
(5.15s)
1.87 m
(5.15s)
Computation time
(ClockTime)
182 h
131 h
93 h
31 h
-
8h
7.6 h
5.6 h
Processor time
182 h
262 h
373 h
245 h
-
237 h
786 h
718 h
* Red text denotes measurements extrapolated from partially-completed simulations.
Results
# processors vs. clock time
# processors vs. processor time
Results
• Conclusions:
• Alternating domain decomposition can result in unpredictable
variations in results
• Computation clocktime reduces with increasing numbers of
processors (up to 32 processors), then levels out
• Processor time remains low until # of processors exceeds 32
• Too few processors doesn’t experience the benefits of parallel
computing
• Too many processors loses time in communication between nodes
• It seems like 16 or 32 processors would be ideal, at least
for this simulation, as computations can be executed most
quickly without wasting resources
References
• http://www.tfd.chalmers.se/~hani/kurser/OS_CFD_2007/PiroozMorad
•
•
•
•
•
nia/OpenFOAM-rapport.pdf
http://perso.crans.org/kassiotis/openfoam/movingmesh.pdf
http://openfoamwiki.net/index.php/Main_FAQ
http://www.openfoam.com/docs/user/
http://www.tfd.chalmers.se/~hani/kurser/OS_CFD_2008/ErikEkedahl/
6dofbeamer.pdf
http://web.student.chalmers.se/groups/ofw5/Advanced_Training/Dyna
micMesh.pdf
Acknowledgment and Disclaimer
I’d like to thank the following people and organizations for
their help and support: Dr. Brown, Dr. H.R. Riggs, Prof.
Sungsu Lee, Krystian Paczkowski, The University of Hawaii
at Manoa, UHM College of Engineering, National Science
Foundation, the OpenFOAM community.
This material is based upon work supported by the National
Science Foundation under Grant No. 0852082. Any
opinions, findings, and conclusions or recommendations
expressed in this material are those of the author(s) and do
not necessarily reflect the views of the National Science
Foundation.

similar documents