Funded PhD project:
Design of experiments for computer model calibration with application to sensor placement for source term inversion
University of Southampton
Supported by the UK Defence Science and Technology Laboratory (Dstl)
Supervisor: Professor Dave Woods, in collaboration with Professor Veronica Bowman (DSTL) and Dr Tim Waite (University of Manchester)
Start date: October 2020
Applications will be considered as they are submitted, up to a deadline of 29 May 2020.
Apply via the below link, quoting the project title and supervisor name.
https://www.southampton.ac.uk/maths/postgraduate/research_degrees/apply.page?&utm_source=mathspgr&utm_medium=mathspgrCTAapply&utm_campaign=mathspgrCTAapply
The study of many important real-world phenomena, e.g. atmospheric dispersion, climate change etc., is facilitated by the use of complex computer simulation models. Such a simulator often depends upon unknown parameters, which must be calibrated using data from a physical experiment. To deal with the high computational expense of the simulator, when calibrating the unknown parameters it is usual to approximate the simulator using a much cheaper Gaussian process emulator, built using data from a computer experiment on the simulator.
One important example of a simulator calibration problem is source term inversion, which concerns the estimation of the location of an unknown chemical or biological release using physical data from sensors together with a simulator of atmospheric dispersion processes.
The initial goal of this project would be to extend and adapt existing calibration design methodology to address the problem of choosing optimal sensor locations in source term inversion. In particular, methods would be developed to address the highly multivariate nature of the output from the dispersion model, and the existence of large regions where the simulator is constant (corresponding to no deposition in this location). Time allowing, extensions to the case of moving sensors would be considered, and links to emerging work on optimal dynamic designs.
This PhD position is open to UK nations meeting certain residential eligibility requirements.
Subject to approval from Dstl, funding is available at standard UKRI rates (fees + stipend), along with a research training grant to support conference attendance etc.
For enquiries about the project, please contact Professor Dave Woods (This email address is being protected from spambots. You need JavaScript enabled to view it.).
