[ Post-Grad Front Page | Maths Home Page | Maths Staff Pages | Getting to HW]

PhD Projects ( Dugald B Duncan )

My main interests are the numerical analysis of time dependent partial differential and integral equations. Some applications areas are outlined below.

Wave Propagation and Scattering - Time Domain PDEs DEMO

Wave propagation and scattering problems occur in areas as diverse as radar, underwater sonar, seismology, and ultrasonic probes in medicine. One interesting area is in how to deal with a few obstacles in an otherwise uniform medium (like a few objects flying in the air). I think that using the overlapping grid idea may give good results here, and I'm keen to investigate the numerical analysis and practical applications of that idea.

Wave Propagation and Scattering - Time Domain Boundary Integral Equations (with P J Davies, Strathclyde)

As an alternative to solving PDEs, one can instead solve time dependent integral equations for quantities defined only on the 2D surface of the scatterer. This removes the need to solve PDEs in a 3D spatial region which can be an expensive process. However, the resulting retarded potential integral equations (RPIEs) present a big challenge in understanding their numerical analysis and in constructing efficient methods for their approximation.

Reservoir Simulation DEMO

One of the problems in simulating the behaviour of oil and gas fields and groundwater flow in the presence of boreholes (wells) and other defects is the huge difference in scale between the borehole and the reservoir. This is particularly important in welltest analysis, where the pressure at a wellbore is measured and deductions are made about the properties of the reservoir based on these measurements. One approach is to use overlapping grids to make simulation more convenient, and there is still much work to be done in this area.

Convolution Integral Reaction Diffusion Equations (with G Lord)

Standard models of phase transitions and various biological processes are described by reaction-diffusion equations. An alternative model based on a convolution integral operator in place of the standard diffusion term arises in some cases (like nerve axon firing) and presents some interesting challenges in numerical and applied analysis. There is still much work to do in producing accurate and efficient numerical methods for such problems, and in proving that they work.