go to UNSW home page

The 15th Biennial Computational Techniques and Applications Conference CTAC2010
28 Nov - 1 Dec 2010, UNSW, Sydney
Home Invited speakers Proceedings Register Login Accommodation Program Papers Committees Sponsors More Info

Invited speakers

Tim Cornwell (CSIRO) Large computational and algorithmic challenges of the square kilometre array
Paul Durbin (Iowa State University) Turbulence flow through a serpentine channel
Bjorn Engquist (Uni of Texas at Austin and Royal Institute of Technology at Stockholm) Multiscale method for wave propagation
Mark Knackstedt (Australian National University) The digital materials laboratory
Frances Kuo (University of New South Wales) Lifting the Curse of Dimensionality - Quasi Monte Carlo Methods for High Dimensional Integration
Robert Scheichl (University of Bath) Multilevel simulation under uncertainty
Jamie Sethian (University of California at Berkeley) In Pursuit of Interfaces - Inkjet Plotters, Coating Rollers, Semiconductors, Retinopathy, and Chemical Pathways
Large computational and algorithmic challenges of the square kilometre array

Tim Cornwell
Over much of its 80 year history, radio astronomy has been enabled by and dependent on three technologies: digital signal processing, computing, and networking technologies. By the end of this decade, radio astronomers working in an international consortium will have built the world's largest, most sensitive, fastest radio telescope, the Square Kilometre Array. The SKA will be one of the foremost scientific instruments in the world, addressing some of the most important questions in astrophysics and cosmology. The SKA will naturally stress and stretch the state of the art in the three technology areas mentioned above. One of the most notable technical aspects of the SKA is the very large data rate - exceeding 10 Pbit/s over long-distance networks into the digital signal processing chain, requiring about 1 Exaflop/s for processing into images, and producing 1 Exabyte of science data per week. The data rate raises many issues throughout the entire system design. I will concentrate on the interaction between the computing architecture and the physics and algorithmics of the measurement process.

Turbulent Flow Through a Serpentine Channel

Paul Durbin
Iowa State University
The serpentine geometry is meant to isolate the influence of curvature on turbulence. Periodicity is imposed between inlet and exit and a fully developed flow field develops in our direct numerical simulation. The duct is rotated to examine the combined effects of curvature and rotation. The theoretical framework will be sketched. Stabilizing and destabilizing influences of curvature and rotation are seen in the DNS data. Large, but quite weak, streamwise vortices form due to convex curvature; but they break up under rotation, possibly because rotation destroys the symmetry between the two bends of the serpentine. Within the bends turbulent intensities are enhanced or suppressed roughly in accord to theory. Particles are tracked in the flow and some rather interesting distributions of particle accumulation are seen. Light particles acculumlate somewhat along the walls. It is unclear whether turbophoresis plays a role, or whether this is a balance of centrifugal acceleration and turbulent dispersion. Heavy particles form into jets, leaving the inner wall, and bands of concentration are caused by multiple reflections from the outer wall. The concentration bands spread by turbulent mixing. Standard drag formulas for one-way coupling are used for the particle trajectories. The domain is partitioned geometrically, so the high concentrations cause the particle tracking to be poorly load balanced.

Multiscale Methods for Wave Propagation

Bjorn Engquist
The University of Texas at Austin and Royal Institute of Technology at Stockholm
We will give a brief overview of multiscale modeling for wave equation problems and then focus on two techniques. One is the heterogeneous multiscale method applied to problems with oscillatory velocity fields. In this technique calculations in time domain on a refined grid is performed on small subsets of the computational domain to achieve efficiency. The other technique is a new class of sweeping preconditioners for frequency domain simulation with variable coefficients resulting in computational procedures that essentially scale linearly even in the high frequency regime.

The Digital Materials Laboratory

Mark Knackstedt
Australian National University
One of the main obstacles to real progress in the science of complex real world materials has been the need to accurately characterise material structure and thereby to predict properties of materials in three dimensions. We describe the development of a new quantitative numerical laboratory approach to the study of complex real world materials in 3D. A first part incorporates the development and integration of experimental 3D imaging facilities for characterizing materials at multiple scales. A second part is the development of computational infrastructure for image reconstruction, phase identification, multiscale mapping, 3D visualisation, structural characterisation and prediction of physical properties of material properties from digitised 3D images. We describe examples where the ability to quantitatively measure and characterize the structure and properties of complex materials in 3D has impacted on applied sciences including materials design and bone health diagnosis. We also describe proven commercial applications of the technology in the resources sector.

Lifting the Curse of Dimensionality - Quasi Monte Carlo Methods for High Dimensional Integration

Frances Kuo
University of New South Wales
High dimensional problems, that is, problems with a very large number of variables, are coming to play an ever more important role in applications. These include, for example, option pricing problems in mathematical finance, maximum likelihood problems in statistics, and porous flow problems in computational physics. High dimensional problems pose immense challenges for practical computation, because of a nearly inevitable tendency for the costs of computation to increase exponentially with dimension: this is the celebrated "curse of dimensionality". In this talk I will give an introduction to "quasi-Monte Carlo methods" for tackling high dimensional integrals, with a focus on "lattice rules", and discuss the challenges that we face while attempting to lift the curse of dimensionality.

Multilevel Simulation under Uncertainty

Robert Scheichl
University of Bath
The quantification of uncertainty in groundwater flow plays a central role in the safety assessment of radioactive waste disposal and of CO2 capture and storage underground. Stochastic modelling of data uncertainties in the rock permeabilities lead to elliptic PDEs with random coefficients. A typical computational goal is the estimation of the expected value or higher order moments of some relevant quantities of interest, such as the effective permeability or the breakthrough time of a plume of radionuclides. Because of the typically large variances and short correlation lengths in groundwater flow applications, methods based on truncated Karhunen-Loeve expansions are only of limited use and Monte Carlo type methods are still most commonly used in practice. To overcome the notoriously slow convergence of conventional Monte Carlo, we formulate and implement novel methods based on (i) deterministic rules to cover probability space (Quasi-Monte Carlo) and (ii) hierarchies of spatial grids (multilevel Monte Carlo). It has been proven theoretically for certain classes of problems that both of these approaches have the potential to significantly outperform conventional Monte Carlo. A full theoretical justification that the groundwater flow applications discussed here belong to those problem classes are under current investigation. However, experimentally our numerical results show that both methods do indeed always clearly outperform conventional Monte Carlo even within this more complicated setting, to the extent that asymptotically the computational cost is proportional to the cost of solving one deterministic PDE to the same accuracy.

In Pursuit of Interfaces - Inkjet Plotters, Coating Rollers, Semiconductors, Retinopathy, and Chemical Pathways

Jamie Sethian
University of California, Berkeley
Propagating interfaces occur in a wide variety of settings, and include ocean waves, burning flames, and material boundaries. Less obvious boundaries are equally important, and include iso-intensity contours in images, handwritten characters, and shapes against boundaries. In addition, some static problems can be recast as advancing fronts, including robotic navigation and finding shortest paths on contorted surfaces. One way to frame moving interfaces is to recast them as solutions to fixed domain Eulerian partial differential equations, and this has led to a collection of PDE-based techniques, including level set methods, fast marching methods, and ordered upwind methods. These techniques easily accommodate merging boundaries and the delicate 3D physics of interface motion. In many settings, they been proven valuable. In this talk, we will focus on scientific and engineering applications of these techniques. This will include "How do home inkjet plotters work?". "What happens when my faucet drips?" "How can we guide chemical probes through complex materials?". "How are semiconductors built?". And, "How can we automate the early detection of eye disease?".