The following research projects make use of the Silicon Graphics Power Challenge supercomputer at The University of Queensland and/or the QPSF SP/2 supercomputer at Griffith University. The initial data was collected in 1996 and has been updated to reflect activities in 1997. The project abstracts are grouped by Department or Research Centre.
The objective of this project is to improve the performance of unaerated wastewater ponds. In order to achieve this the effect of various pond designs and modifications need to be evaluated. Unfortunately the limitations of experimental investigations mean that a model of the ponds must be used to predictively determine the ponds performance. Traditional pond models are unable to predict the effect of geometric changes on hydraulics, thus computational fluid dynamics (CFD) is used to investigate the hydraulics in a pond, and thus determine the performance.
Publications:
[Go back to the list of titles]
Designing and Implementing High Performance Algorithms for the solution of Large-Scale Numerical Linear Algebra Problems and Differential Equations.
Publications:
[Go back to the list of titles]
The research seeks to develop 3D models of the downwind concentration of combustion products resulting from warehouse fires where the roof may be intact or collapsed. The work looks at plume lift off and the influence of the atmospheric flow field on the impact of the plume.
In particular we are using FIDAP as a CFD tool. The work involves 3D finite element simulations of the flow field which must resolve the turbulence down to a fine grid. The distances of interest for near field dispersions extend to several kilometres from the warehouse.
The work is using large scale wind tunnel test done in the UK as a means of validating the CFD approach. The modelling work considers the effect of fire growth, the condition of the warehouse and the effect of nearby structures in promoting downwash of the fire plumes.
The final outcome of this work is to develop effective shortcut techniques which can be used for landuse planning by local authorities as well as emrgency response by fire services and police.
The scale of the simulations is such that the work could not be carried out onconventional computing machines. The use of the Power Challenge array has made this work possible.
[Go back to the list of titles]
Molecular modelling: so far only dynamics studies of large peptides.
[Go back to the list of titles]
[Go back to the list of titles]
This project aims at studying local amplification of seismic waves (site effects) that are responsible for increasing the degree of devastation of numerous earthquakes. Numerical modelling using the pseudospectral method is used to better understanf these amplifications. Tests were conducted using the topography and subsurface models adjacent to Brisbane Airport and a volcano in the Reunion Island.
Publications:
[Go back to the list of titles]
A lattice solid model capable of simulating the physics of rocks and the non-linear dynamics of earthquakes has been refined. This enable the causes of seismic wave attenuation to be studied, as well as heat generation and the development and effect of fault gouge on earthquakes.
Publications:
[Go back to the list of titles]
Evaluation the performance in terms of local accuracy of various interpolation and simulation methods in the context of mining applications Five methods were implemented and tested. The quality of the estimates were obtained by comparing the estimated values with the reference data set. To proceed with the estimations a sub-sample extracted from the reference data set was used.
Publications:
[Go back to the list of titles]
Mine planning has relied on the existance of 3-D models of the subsurface. Traditionally this has been constructed using chemical assay data derived from drilling information
My study aims to investigate ways of integrating geophysical data into modelling process. It involves the investigation of existing methodologies and the development of new techniques. As this study is computer intensive, access to the HPCU is extremely desirable. This is because I don't have access to any other high-performance machines.
Publications:
[Go back to the list of titles]
The minimum energy pathways (MEP1-MEP7 associated with the corresponding intermediates 1-7) of the bond-rupture processes involving in the NH2+NO reaction have been characterized using spin-unrestricted density functional theory with the geometry of each point along the reaction coordinate (R) optimized at the UB3LYP/6-311G(d,p) level of theory. MEP1 describes a NN bond-breaking path from the intermediate 1 directly dissociating to form the reactants NH2+NO with a spin purity breakpoint occurring around R=3D2.1 =8F. Following the NO bond-dissociation paths of MEP2-MEP7 leads to the radical product channel (HN2+OH) with the breakpoint occurring around R=3D2.0 =8F. However, it has also been shown that there is a common local minimum, whose geometry and energetics are very close to those of the dissociation products, involved in the reaction paths MEP2, MEP6 and MEP7 and that the energy barriers connecting this local minimum with the corresponding intermediates (2, 6 or 7) are insignificant. Therefore, the isomerization from intermediate 2 to 6 or 7 followed by rapid dissociation through the N2+H2O product channel would be another alternative for MEP2, in addition to the direct dissociation forming HN2+OH. Another local minimum has been found in the bond-breaking path of MEP3 but without affecting the trend of further dissociation. For MEP4 and MEP5, only part of the reaction path has been characterized due to the difficulty of convergence of the energy optimization along their individual reaction coordinates. According to the results of MEP1 and MEP2 studied with three different basis sets [6-311G(d,p), 6-311+G(d,p) and cc-pVQZ], the entire single-bond-breaking potential curve for the system predicted at the UB3LYP level appears to be relatively insensitive to the basis functions. The geometries of the separating fragments associated with the various MEP=92s have been found to relax quickly along the respective reaction coordinates, in particular reaching the asymptotic bond angles and bond lengths around R=3D2 =8F for= MEP1, MEP2, and MEP3.
Publications:
[Go back to the list of titles]
An objective of our research which runs parallel to the achievement of fundamental advances in unimolecular rate theory is the continued improvement of methods for computing thermal rate coefficients as a function of temperature and pressure, accurate estimates of which are vital for combustion and atmospheric modelling. In recent years, we have developed new theoretical methods for the implementation of microcanonical variational transition state theory (mVTST) which have enabled a dramatic enhancement in the speed of rate coefficient calculations for dissociation and recombination reactions proceeding via loose (e.g., radical-radical) transition states [1,2]. These developments have opened up new possibilities for accurate modelling of complex radical-radical reactions involving multiple channels and multiple wells [e.g., 3]. Ab initio quantum chemical calculations provide details of the potential energy surface which are crucial to the reliability and predicitve capability of the kinetics calculations [4].
[1] S.C. Smith, "Rapid Algorithms for Microcanonical Variational
Rice-Ramsperger-Kassel-Marcus Theory", J. Phys. Chem. 97, 7034 (1993).
[2] S.C. Smith, "Flux Factors in Variational Transition State
Theory", J. Phys. Chem. 98, 6496 (1994).
[3] E. W.-G. Diau and S.C. Smith, "Calculation of the Temperature
Dependence of Rate Coefficients and Branching Ratios for the NH2+NO Reaction
via Microcanonical Variational Transition State Theory", J. Phys. Chem.
100, 12349 (1996).
[4] E. W.-G. Diau and S.C. Smith, "Theoretical Investigation
of the Potential Energy Surface for the NH2+NO Reaction via
Density Functional Theory and Ab Initio Molecular Electronic Structure
Theory", J. Chem. Phys., 106(22) (1997).
[Go back to the list of titles]
Quantum mechanical (ab initio and semi-empirical) calculations on the rearrangements of highly reactive intermediates.
[Go back to the list of titles]
Molecular dynamics studies have been carried out in order to determine
fluid properties with particular interest in their chaotic properties. The
work can be divided into three projects:
(i) Large scale simulations have been carried out on liquid and model
systems in order to investigate the dependence of the maximum Lyapunov
exponent on the number of degrees of freedom (or number of particles, N) in
the system. Whereas thermodynamic and transport properties are found to
converge as 1/N, the maximum Lyapunov exponent, which is a measure of
chaos, appears to diverge logarithmically.
(ii) The behaviour of the thermal conductivity and the isothermal
compressibility of a Lennard-Jones fluid in the viscinity of the
gas/vapour critical point has been investigated usine molecular dynamics
and nonequilibrium molecular dynamics simulations. Current results are
consistent with theory and experiment.
(iii) The Lyapunov spectrum of the hard disk model of liquids has been
calculated from the stability matrix of a single trajectory. The results
are consistent with results obtained by simulation tangent trajectories.
Publications:
[Go back to the list of titles]
Development of methods for computing chemical rate coefficients over a wide range of tempartures and pressures by variational transition state theory and the solution of multidimensional master equations which allow for the non-equilibrium distributions of internal energy and angular momentum in the reacting molecules.
Publications:
[Go back to the list of titles]
Development of high performance algorithms for quantum wave packet propagation in the simulation of dissociation and recombination reactions. Calculation of molecular rovibrational bound states and resonances.
[Go back to the list of titles]
The aim is to study recently discovered new classes of reactive intermediates and new chemical rearrangement reactions. The interplay between theory and experiment is extremely important. The very nature of the intermediates makes the collection of reliable experimental data difficult. High level ab inito calculations using the GAUSSIAN94 package help by providing structures, energies absorption spectra and reaction pathways, and provide useful guides in the search of novel chemical species.
Publications:
[Go back to the list of titles]
We are developing computer code for the prediction of bound-state eigenvalues associated with the floppy modes in molecular dimer clusters. The Hamiltonian operator formulated for a reduced dimensional model in which the two monomers of the cluster are treated as rigid rotors, which amounts to a six-dimensional computation for two non-linear species. Bound states are computed using our recently-developed MINRES Filter Diagonalization approach [1] which is based on a Lanczos subspace but suffers none of the complications of ghost eigenvalues which make interpretation of the Lanczos spectrum problematic. We utilize an extension of the pseudo-spectral algorithm used in our studies of loose transition states [2] for the action of the Hamiltonian. In this approach, the action of the potential is effected on a six-dimensional spatial grid, but in order to act with the rotational kinetic energy operator a transformation of the wavefunction to a body-fixed Wigner function representation is made.
[1] H.G. Yu and S.C. Smith, "The Calculation of Vibrational Eigenstates
by MINRES Filter Diagonalization", Ber. Bunsenges. Phys. Chem., 101,
400 (1997).
[2] S.C. Smith, "A Pseudo-Spectral Algorithm for the Computation
of Transitional Mode Eigenfunctions in Loose Transition States", Chem.
Phys. Lett. 243, 359 (1995).
[Go back to the list of titles]
We have recently developed new spectral filtering and filter diagonalization methods, based on Lanczos and related Krylov-space algorithms, which enable the efficient calculation of highly-excited bound states and resonances [1-5]. A particularly useful feature of our method is the elimination of the ghosting problem which normally complicates the application of the Lanczos method. As a result of these new developments, it is possible to determine complete spectral information including bound state energies, resonance energies and lifetimes from a single iterative calculation [5].
[1] H.G. Yu and S.C. Smith, "Restarted Krylov-Space Spectral
Filtering", Faraday Transactions, 93, 861 (1997).
[2] H.G. Yu and S.C. Smith, Ber. Bunsenges. Phys. Chem., 101,
400 (1997).
[3] H.G. Yu and S.C. Smith, "The Elimination of Lanczos Ghosting
by MINRES Filter Diagonalization", J. Comp. Phys., submitted.
[4] H.G. Yu and S.C. Smith, "The Imposition of Outgoing-Wave
Boundary Conditions via a Symmetrically-Damped Hermitian Hamiltonian Operator",
J. Chem. Phys., submitted.
[5] H.G. Yu and S.C. Smith, "Calculation of Quantum Resonance
Energies and Lifetimes via Quasi-Minimum Residual Filter Diagonalization",
Chem. Phys. Lett., submitted.
[Go back to the list of titles]
Coset enumeration is a most important procedure for investigating finitely presented groups. We present a practical parallel procedure for coset enumeration on shared memory processors. The shared memory architecture is particularly interesting because such parallel computation is both faster and cheaper. The lower cost comes when the program requires large amounts of memory, and additional CPU's allow us to lower the time that the expensive memory is being used.
Rather than report on a suite of test cases, we take a single, typical case, and analyze the performance factors in-depth. The parallelization is achieved through a master-slave architecture. This results in an interesting phenomenon, whereby the CPU time is divided into a sequential and a parallel portion, and the parallel part demonstrates a speedup that is linear in the number of processors. We describe an early version for which only 40% of the program was parallelized, and we describe how this was modified to achieve 90% parallelization while using 15 slave processors and a master. In the latter case, a sequential time of 158 seconds was reduced to 29 seconds using 15 slaves.
Publications:
[Go back to the list of titles]
The aim of this project was to construct a computer model to simulate microstructural development during liquid phase sintering. We had previously developed a model which could simulate some aspects of solid state sintering in a two dimensional, single phase system. In that case, a cross section through a wire winding was represented by a series of close packed circles. This is a simple system and therefore a reasonable starting point. Sintering was simulated using a Monte Carlo technique which we had adapted from one which had been developed by others to model recrystallization. A continuum structure is superimposed onto a 2-D lattice. Each lattice site is assigned a number corresponding to either an atom or a vacancy. The number of vacancies can vary with the simulation temperature, while a cluster of vacancies is a pore. The energy of each lattice site is defined in terms of the interactions with neighbouring sites. A lattice site is picked at random and re-oriented in terms of the atomistic model governing mass transport. The probability of the change being accepted is dependent on the change in the number of nearest neighbours, which is a measure of the energy of the site. Progress is monitored by measuring the density, defined by the pore area; and the pore shape, defined by the rugoscity coefficient. The latter is the perimeter of the pore divided by the circumference of a circle of equal area as the pore The aim of the new project was to extend the model to simulate liquid phase sintering. We had initially confined ourselves to the solid state because of its relative simplicity. However, liquid phase sintering has wider applicability. Most metallic systems and many ceramic systems are sintered in the presence of a liquid phase. It is a much more prevalent commercial process. This was the rationale for the change in direction.
It soon became apparent that this was a very complicated system for the relatively simple model we had developed to simulate solid state sintering. In order to reduce the complexity of the problem to manageable proportions, we limited our investigations to the final stages of sintering where shrinkage is almost complete and coarsening of the solid phase occurs by Ostwald ripening.
Interface control occurs when the coarsening rate is determined by reactions at the interface. Under diffusion control, the particle size, d, is proportional to t3. Under interface control, d is proportional to t2. In our simplified system, diffusion through the liquid is assumed to be infinitely fast and interface control is expected. Our data tends to fall on a straight line when d0.5 is plotted against time. This also serves to validate our results.
We used the Power Challenge to carry out the simulation. We usually ran batch jobs and processed the results locally. The code relied on the compiler to generate an efficient executable file. The turn round time was usually overnight.
Without the Power Challenge our progress would have been considerably slower. Given the simplicity of the model, it is evident that we shall have to look for methods less demanding of resources, however the eventual scale of the problem means that we shall almost certainly need cutting edge performance to make a realistic simulation. One of the solutions which we investigated briefly involved the use of genetic algorithms. This greatly enhances computational efficiency allowing us to tackle complex problems in a shorter time frame. This will be further explored in future projects.
[Go back to the list of titles]
Use of CPLEX to solve an integer programming problem.
[Go back to the list of titles]
Research projects are mainly part of my PhD thesis. The thesis will study parallel methods based on Iterated Multistep Runge-Kutta (IMRK) of Radau type for solving Ordinary Differential Equations (ODEs), Partial Differential Equations (PDEs), and Hamiltonian Differential Equations (HDEs). The work on these topics are motivated by results of van der Houwen and Sommeijer , the existence of MRK of higher order, 2*s+k-2, where k and s respectively denote the number of past solutions and stages required by the methods. The result of van der Houwen et al. show that it is possible to construct parallel iterated methods based on implicit Runge-Kutta methods, which are seldom used in codes due to a number of coupled nonlinear algebraic equations which have to be solved on everystep. Following their ideas, there will be a great advantages when the core methods is MRK from which one could obtain parallel high order methods. Both non-stiff and stiff problems will be considered and the methods' performance will be compared with the available parallel methods.
Solving ODES resulting from discretizing PDEs in space will be the second part of the thesis. Some areas of this topic have been done by some researchers. These covers parallelism across the space which are handled by operator splitting and waveform relaxation. But the problem will become more complex when parallel across the method is also considered. Parallel LSODE code, which are under construction of its author, and parallel Iterated Runge-Kutta code will be good partners to compare with parallel codes based on IMRK.
The final part of thesis will cover the possibilities of constructing parallel methods based on IMRK for solving Hamiltonian problems. The problems, which have gained much attention recently, are solved by using so-called symplectic or canonical methods. The results of people working on this area show that the problem is not suitably solved by non-symplectic methods.
[Go back to the list of titles]
Develop theories and computational models for investigating the vibratory characteristics of general laminated plates.
Publications:
[Go back to the list of titles]
A finite element (non-linear) program is used to describe how the plasticity spreads in a box girder bridge.
[Go back to the list of titles]
This project has involved the use of FIDAP to simulate the three-dimensional turbulent flow in an irregular open channel.
[Go back to the list of titles]
The project involved exploring techniques for adapting speech recognition systems trained for one speaker or set of conditions to another. This is of interest because optimum recognition performance for a system is achieved by training that system with samples from the same speakers which will use it, recorded in the same acoustic environment (which is also preferably noise free) in which it will be used. These requirements frequently are impractical. Hence, methods which instead adapt an existing system to better match a new speaker may be useful.
The techniques used are based on gradient descent maximisation of the likelihood functions for the hidden Markov models which have become a de facto standard for acoustic modelling in speech recognition.
Publications:
[Go back to the list of titles]
The project is an experimental one where I am taking turbulent measurements in a highly heated air jet to determine the dissipation time scale ratio. Measurements of velocity and temperature are taken with a triple hot wire set up. An A/D card samples three channels simultaneously (2 velocity and 1 temperature) at a maximum of 80khz (total) for about 40 seconds, resulting in a 5 MB data file. This data file is for one point and on average I take 40 points per profile and about 5 profiles per run, so I need to quicky reduce my data files and obviously the disc space it uses. I use the C-compiler on Moreton and Fraser to run my data reduction program.
[Go back to the list of titles]
Numerical evaluation of top Lyapunov exponents are used to characterise behaviour of several practical mechanical systems.
[Go back to the list of titles]
Investigation of the control requirements of a hypersonic air-breathing vehicle. These type of aircraft are inherently unstable with a high degree of intergration between airframe, engine, and dynamics. Simulation of the non linear dynamics coupled with an aerodynamic model is required and to date Matlab has been used for this.
[Go back to the list of titles]
This project will use PHOENICS to simulate the flow around the UQ Solar Car.
[Go back to the list of titles]
Scramjet engines exhibit high enthalpy supersonic flows in the combustor region that can be computationally very expensive to model. To faithfully model the flow, finite rate chemical kinetics and hypersonic viscous flow interactions need to be simulated but at the same time a need for fast solution turn around is also important. The design of such combustion chambers has historically been based on experimental analysis and simplified computational modeling. This method of "design and try" cycling has been adequate for the most part however it inherently takes a lot of time and money before the optimum design is achieved.
Publications:
[Go back to the list of titles]
The design of a square section nozzles for the small hypersonic shock tunnel facility is currently underway in the Department of Mechanical engineering. A square section nozzle produces a cleaner test flow in comparison to conventional axisymmetric nozzles. Axisymmetric nozzles focus disturbances at the center and these disturbances propagate to the test core. Square section nozzles do not focus disturbances to a point but rather to the two symmetry lines which is more dissipative.
The design of a square section nozzle is a highly three- dimensional problem which cannot be handled with traditional design techniques. The approach being taken for this study is computational. A three-dimensional fluid dynamic flow solver has been written (which was developed on the SGI) that can simulate the flow through arbitrary ducts with high fidelity. The solver is being used to simulate the flow through a base line nozzle design of a square section nozzle and then an optimisation routine iteratively makes slight modifications to the design to achieve the desired flow conditions at the exit. Each nozzle simulation takes approximately 24hrs (based on current SGI load) to complete and the optimiser takes anywhere from 20 to 100 iterations to reach a converged solution.
Currently, the SGI is being used to find an efficient base line design. It is crucial to start with a promising design before the optimisation routine is started since many hours of CPU time could be wasted. Clearly, looking at the solution time scale, a fast machine is required to solve this design problem.
[Go back to the list of titles]
To design an efficient heat exchangers inside the thermoacoustic engine, simulation of heat exchanger inside the thermoacoutic engine is carried out using Phoenics. This is the study of heat transfer within the sound wave with large temperature gradient and without the temperature gradient.
[Go back to the list of titles]
When high Mach number nozzles are used on reflected-shock tunnels, there is a significant (and undesirable) delay before the test flow settles to a useable steady state. Previous studies, in which the transients flow through a typical nozzles have been simulated, have not been able to identify the mechanisms causing the delay and so we shall attempt to simulate the flow in an entire shock-tube/nozzle combination. The simulation tools to be used are MB_CNS, which is a finite-volume code which now runs in parallel on the Power Challenge, and CEVCATSN, which is a multiblock, multigrid code developed at DLR, Gottingen.
Publications:
[Go back to the list of titles]
The SGI was used in the development of a new Computational Fluid Dynamics computer program for the calculation of hypersonic flows over blunt bodies. The CFD program has been used to verify experimental results obtained from the Mechanical Engineering Department's T4 hypervelocity shock tunnel, and expansion tube. The code has also been used to investigate the feasibility of a numerical Air Data System calibration.
When designing the shock tunnel experiments, results obtained on the SGI were used to determine the optimum size and location of models to be tested in the shock tunnel. Also, after the experiments were run, the code aided in the interpretation of results from experiments that had not been conducted under "ideal" test conditions.
The availability of support from the SGI administrators, as well as the development environment supplied on the machine, were both important factors which helped in the coding, testing, and optimizing of the program.
Publications:
[Go back to the list of titles]
This project will use PHOENICS to model the mass transfer processes in erosion/corrosion.
[Go back to the list of titles]
An application has been made to ARC to develop a high enthalpy test facilty for rarefied gas flow. Simulation calculations are being performed to aid in the design of a nozzle to be fitted to the expansion tube.
Future work will involve comparison of calculations of rarefied gas flow about simple shapes and experimental data from the new facility. This will allow development and verification of collision models and surface-gas interaction models in the rarefied gas regime.
[Go back to the list of titles]
The flow of a low density, high speed gas cannot be adequately described by the continuum Navier-Stokes equations; the particultae nature of the gas must be considered. The standard computational method for calculating the flow of a sufficiently rarefied gas is the Direct Simulation Monte Carlo (DSMC) method. There is a regime of intermediate rarefaction between where the Navier-Stokes equations become invalid and where it is computationally feasible to use direct simulation of molecular motions in the flow using DSMC.
This project is concerned with making the DSMC method more computaionally efficient (using modified procedures at higher densities) so that the "computational gap" between continuum and raefied gas flow can be closed.
[Go back to the list of titles]
Using parallel computation to successfully simulate the starting flow characteristics of an Expansion tube as the primary and secondary diaphragms rupture. The main goals are to increase resolution of the simulation and improve turn around time for achieving a solution as compared to previous work carried out in this area.
As well as using large parallel computing facilities such as the SGI Power Challenge and the IBM SP2, we are building a dedicated parallel computer called NOVA. NOVA is a Beowulf class computational facility, consisting of several IBM-PC clone computers connected together to produce a distributed, parallel computer capable of solving large problems. NOVA relies upon using free operating systems and standards to help reduce the system cost, yielding significant computational power for the cost of the hardware only, and can easily be scaled by adding more computers. Typically, as of 1997, NOVA's is estimated to cost $50 per Mflop (Million floating point operations), whereas UQ's SGI PowerChallenge Array is estimated to cost $400 per Mflop.
Publications:
[Go back to the list of titles]
In essence the project has been primarily a numerical investigation of chaotic instabilities that can occur in 3 multibody dynamical configurations pertaining to spacecraft. The aim has then been to control these instabilities using various feedback control methods.
Publications:
[Go back to the list of titles]
The steady state and fully pulsed turbulent round air jet into quiescent air is considered. The main objective is to investigate the ability of the Reynolds stress model to predict the fully pulsed jet. It was previously shown that the k-e turbulencs model fails to predict the change in slope of the velocity decay where the jet changes from pulsed to steady jet behaviour. Both turbulencs models will be applied to simulate the steady state and the pulsed jet.
The PHOENICS flow simulation code to simulate the turbulent flow within the pulsed jet.
[Go back to the list of titles]
Development of 2-D and 3-D models of diaphragm rupture in shock tunnels and expansion tubes.
Publications:
[Go back to the list of titles]
The design of a new balance for the measurement of multiple force and moment components on a scaled model of the HYFLEX vehicle is being investigated. The technique used is an extension of the stress wave force measurement technique. This technique relies on the interpretation of stress wave propagating within a model and its support, as a result of impulsive aerodynamic loads generated through shock tunnel testing. Dynamic finite element modelling is used to evaluate proposed force balance designs. The distribution of aerodynamic loads was obtained from a Newtonian code in which the model scale and test flow conditions were determined using binary scaling. The results from preliminary modelling were used to produce a final balance design. Simulations show this design to be capable of measuring aerodynamic loads on a model of the HYFLEX vehicle.
The project makes use of PATRAN for mesh preparation and NASTRAN for the dynamic analysis of the force balances
[Go back to the list of titles]
This project will use the GASP simulation code to model the flow of carbon-dioxide at hypersonic speeds.
[Go back to the list of titles]
The numerical simulation of stochastical instabilities in a two-degree-of-freedom Hooke's joint system as well as a multi-degree-of-freedom space craft system.
[Go back to the list of titles]
Development of a Navier-Stokes incompressible flow solver, and a Particle tracking code using Random-Walk techniques. These codes are used in the prediction of droplet dispersion. The aim of the project was to acheive some method of lower off-target spray drift.
[Go back to the list of titles]
This project will use the GASP simulation code to model the flow around blunt-body hypersonic flight vehicles.
[Go back to the list of titles]
In Large-Eddy Simulation (LES), large-scale velocity fields are directly computed while small-scale structures are modelled. The rationale behind LES is based on two observations. Firstly, most of turbulence energy is carried by large-scale structures. Secondly, small-scale eddies are isotropic, and easy to be modelled. In this project, LES and Direct Navier-Stokes simulations will be used to study turbulent flow over an obstacle.
[Go back to the list of titles]
For exploration and mining purposes EM amplitude tomography provides an image of appararent attenuations between two boreholes from which data has been collected. The tomograms are used within the AIM (Approximate Inverse Mapping) method to obtain a physically valid model representative of an actual measurable property of the geology between the two boreholes (i.e. conductivity).
The approximate inverse mapping method requires a considerable amount of forward modelling which can be very time consuming. Using FRASER can reduce the required amount of time for these forward modellings sometimes by as much as a factor of 10. For this reason it has been a definite advantage in the project allowing more trials to be made in less time.
[Go back to the list of titles]