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.