CFD Online Logo CFD Online URL
Home > Jobs > Job Record #15717

CFD Jobs Database - Job Record #15717

Job Record #15717
TitleParallel computation of adaptive wavelet-based 2D flood models
CategoryPhD Studentship
EmployerUniversity Of Sheffield - SHEFFIELD
LocationUnited Kingdom, South Yorkshire, Sheffield
InternationalNo, only national applications will be considered
Closure DateMonday, July 01, 2019
Computer Science graduates are invited to apply for this fully-funded 3.5 
year PhD studentship in the Department of Civil and Structural Engineering. 

This is an exciting opportunity for you to apply and develop your skills 
within an interdisciplinary project, which aims to design and execute a 
parallel implementation of an existing flood simulator. The project also 
includes regular engagement with flood model developers and consultants, and 
annual attendance of academic conferences. 
Project description: 
Many flood models rely on speed-up of parallelism to accommodate fast 
simulations over large domain [1]. Although parallelism has become 
established in flood modelling, it is often achieved a fixed mesh in order to 
alleviates many issues in both parallel computing science and practical 
modelling, e.g. domain decomposition, synchronized data distribution across 
many processors, wetting and drying treatment [2,3,4].
You will join the SEAMLESS-WAVE project team to build on their developing 
adaptive wavelet-based flood model. Wavelet adaptivity provides 35-50 speedup 
over the fixed mesh version without parallelisation. However, it entails 
constant transfer of modelled-data at different scales and changes in mesh. 
This form a non-standard challenge [5] to achieved parallelisation,
which will be tackled in this PhD, together with the standard challenges 
[2,3,4] to ultimately produce a parallelised adaptive wavelet based flood 
Phase 1- Understanding and analysing (multi)wavelet data (de)compression and 
the adaptive numerical flow solver. You will work closely with the model 
developers to provide code efficiency analysis report and documented plan(s) 
of how the code operations can be reordered in favour of parallel computing. 
Output of this phase will be an optimised code implementation alongside a 
documented algorithmic work-plan.
Phase 2- Coding of a parallelised wavelet-based flood model. Informed by 
phase 1, a choice of where to implement the parallel version will be made. 
One possible route would be to use PGI Fortran compiler with a view to 
achieve the parallelisation on GPUs. This choice is compatible with the 
current Fortran 2003 implementation of the flood model and allows flexibility 
for parallelization using any other route (i.e. migrating computational heavy 
aspects of the code to dedicated CUDA routines). The uniform and adaptive 
mesh versions will be parallelised within same code structure to allow 
quantifying the efficiency speed-up gains against overheads of both 
parallelisation and mesh adaptivity.  

Phase 3- Application to large-scale real-study sites. By this phase, the 
parallelised adaptive flood simulator will be validated in reproducing 
realistic flood scenarios over very large domains. A critical review of 
existing many case studies will be carried out to select up to four tests 
featured by large spatial coverage, fine resolution terrain data, and 
availability of reference data. Set up and runs for these case studies will 
be performed to identify the extent to which speed-up scaling is maximised.
The Candidate: Suitable for candidates holding or anticipating award of a 
Distinction/Merit MSc, or 1st/2.1 undergraduate degree in computer sciences 
discipline. The suitable candidate is also expected to have a good 
mathematical background, to be able to work in a team of mathematicians and 
Engineers. The successful candidate must start by February 2020.

Contact Information:
Please mention the CFD Jobs Database, record #15717 when responding to this ad.
NameGeorges Kesserwani
Email ApplicationYes
Record Data:
Last Modified11:28:32, Monday, March 11, 2019

[Tell a Friend About this Job Advertisement]

Go to top Go to top