Development of a Parallel Adaptive Cartesian Cell Code to Simulate Blast in Complex Geometries

Mr Joseph Tang (2008). Development of a Parallel Adaptive Cartesian Cell Code to Simulate Blast in Complex Geometries PhD Thesis, School of Engineering, The University of Queensland.

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
n33632575_PhD_abstract.pdf Final Thesis Abstract application/pdf 28.01KB 0
n33632575_PhD_totalthesis.pdf Final Thesis Lodgement application/pdf 22.21MB 0
Author Mr Joseph Tang
Thesis Title Development of a Parallel Adaptive Cartesian Cell Code to Simulate Blast in Complex Geometries
School, Centre or Institute School of Engineering
Institution The University of Queensland
Publication date 2008-06
Thesis type PhD Thesis
Supervisor Dr. Peter Jacobs
Dr. Michael Macrossan
Total pages 255
Total colour pages 31
Total black and white pages 224
Subjects 290000 Engineering and Technology
Formatted abstract
The modelling of blast propagation in urban environments generated by explosions
allows prediction of blast loading on structures, which in turn has useful applications
like damage assessment and improvement of structural design. However such an exercise
is often realizable only with Computational Fluid Dynamics simulations, which can be
dicult to perform because of the geometric complexity of the blast environment.
This thesis describes the development of the code OctVCE designed especially for
modelling shock and blast eects in complex structural geometries. This code is designed
for practical engineering use where high resolution is unnecessary. It uses a nite-volume
formulation of the unsteady Euler equations with second-order explicit Runge-Kutta
timestepping and linear interpolation with a minmod-based limiter. Flux solvers used
are the Advection Upwind Splitting Method variant (AUSMDV) and the Equilibrium
Flux Method (EFM). No
uid-structure coupling or chemical reactions are modelled,
and gas models can be perfect gas or the real-gas JWL model.
The code uses the Virtual Cell Embedding (VCE) Cartesian cell method to automatically
generate grids in complex geometries. This method is chosen because of
its simplicity, robustness and generality. Additional eciency in computational performance
and memory usage is obtained by implementing an octree-based mesh adaptation
scheme in the code. The parallel implementation of the code using the shared-memory
OpenMP paradigm is also described.
The code is veried to establish reliability of the numerical implementation via test
cases like the method of manufactured solutions, an ideal shock tube problem, supersonic

ow over wedge and cone geometries and a supersonic vortex problem. The code is then
validated to demonstrate its reliability and usefulness in simulating more realistic shock
and blast problems. Test cases presented increase in geometric complexity and include
unsteady shock interaction with wedge and cylinder geometries and blast interaction
with barriers, axisymmetric containers, simple arrangements of cuboidal structures and
complex cityscape buildings.
As part of a design exercise for the development of a static-ring test facility, OctVCE
is applied to modelling internal blast in a shipping container geometry. It is found that
very large amplication of pressures and impulse exists within the structure (by at
least a factor of ten) due to blast connement. It was not always easy to demonstrate
convergence, especially along edges and corners of the geometry, due to the coarsenessof the grids employed in the simulations. However, the impulse could still be computed
with fairly low error.
The serial and parallel performance of the code is measured for some of these cases.
The performance proles indicate that substantial savings in storage and execution time
is achieved on adaptive meshes compared to equivalent uniform meshes. Execution
time is also considerably shortened through the use of parallel processing. However,
code performance can still be signicantly enhanced, and several aspects of the code
are identied in the last chapter in which improvements can be made in future work.
These include more ecient parallel implementation, better adaptation indicators, less
conservative timestepping and importantly reduction of memory usage.
Keyword Blast, numerical simulation, Cartesian cell, virtual cell embedding, complex geometry
Additional Notes 39, 40, 41, 61, 81, 82, 94, 98, 100, 104, 108, 110, 111, 112, 113, 114, 121, 147, 148, 151, 152, 158, 159, 160, 172, 184, 185, 186, 190, 223, 235

Citation counts: Google Scholar Search Google Scholar
Created: Mon, 10 Nov 2008, 10:38:44 EST by Dr Joseph Tang on behalf of Library - Information Access Service