Accurate mathematical models of composite materials in Aerospace Engineering are computationally expensive even on modern supercomputers. On the other hand, data from experimental testing is limited, and sparse, and comes from various sources and experiments at different length scales. As a consequence, modern Bayesian statistical inference, which typically requires an extensive number of model evaluations, is an extremely challenging task.
This post is key to the EPSRC Programme Grant CerTest, which seeks to introduce Bayesian statistics, numerical methods and high-performance computing tools for the manufacture of the safe and sustainable aircraft of the future. Currently, small scale testing, combined with a frequentist view of statistics, introduces conservatism at the structural level. The main objective of CerTest is to create an ensemble of statistical, numerical and engineering tools to determine accurate safety margins for structural capacity and, at the same time, to develop “digital twins” that generate virtual testing data to reduce experimental costs. This will allow to design more structurally efficient and lightweight aerostructures to meet future fuel and cost efficiency challenges.
You will constantly interact with a team of researchers including statisticians, applied mathematicians and mechanical engineers and will have the opportunity to contribute to the development of new statistical methods and mathematical theory in the context of large-scale scientific computing. Specifically, this position will develop:
High dimensional Markov chain Monte Carlo samplers for large-scale, nonlinear models of composite materials and structuresStatistical design and Bayesian inference methods to integrate real aircraft testing data from different scales of the structure with information about manufacturing defects and expert prior informationThe role will involve a close collaboration with the research group lead by Prof Tim Dodwell at the University of Exeter and at the Alan Turing Institute (within the Data Centric Engineering Programme), as well as with the research group lead by Prof. Rob Scheichl at the University of Heidelberg in Germany. It is envisaged that you will spend short periods at each of these institutions for collaboration.
Applicants should hold a PhD in computational statistics or applied mathematics and have substantial experience with Bayesian sampling methodologies in high dimensions (eg in areas such as uncertainty quantification or imaging), as well as a broad knowledge of statistical modelling.
Specific experience with modelling in structural mechanics, design of experiments or reliability (survival) analysis would be an advantage, as would experience with different computing platforms such as R, C++, Python, and general HPC platforms.
CerTest - full title 'Certification for Design - Reshaping the Testing Pyramid' – is a £6.9m research investment of the UK Engineering and Physical Sciences Research Council (EPSRC) in the form of a so-called Programme Grant. CerTest is conducted in a close partnership between the academic partners University of Bristol (lead), University of Bath, University of Exeter and the University of Southampton, with strong industrial and stakeholder support by Airbus, Rolls Royce, BAE Systems, GKN Aerospace, CFMS, the National Composites Centre (NCC), the Alan Turing Institute, and with close interaction with the European Aviation Safety Agency (EASA). CerTest addresses barriers to validation and certification of composite aerostructures posed by the so-called ‘building block approach’ (or ‘testing pyramid’) which is the backbone of current validation and certification processes.
This is a fixed-term contract initially for 2 years.
For more details about the role please contact Dr Karim Anaya-Izquierdo (kai21@bath.ac.uk) or Prof Richard Butler (R.Butler@bath.ac.uk), however, please ensure that your application is submitted via the University website.
Closing Date: 14 Jun 2020
Type: Education & Research
