Bio

Mark Hobbs is a Postdoctoral Research Fellow at the University of Exeter and The Alan Turing Institute, where he works under Professor Tim Dodwell in the data-centric engineering group. He gained his PhD in the Department of Engineering at the University of Cambridge (Christ's College), where his research focussed on the development of a numerical model to predict the failure behaviour of quasi-brittle materials. The model was developed using the peridynamic theory of solid mechanics, an integral formulation of continuum mechanics that allows for the natural inclusion of fracture behaviour.

More generally, he is interested in the emerging field of 'learning to simulate' - the application of modern machine learning methods to the field of computational mechanics. This is motivated by the need to reduce the computational expense of costly simulations, and to infer model parameters with a quantified level of uncertainty. Mark has broad research interests in peridynamics, computational mechanics, high-performance computing, simulation, model validation, and uncertainty quantification.

Research interests

Mark is currently working on the development of PeriPy, a lightweight, open-source and high-performance package for peridynamic simulations written in Python. The development of this package is motivated by the need for fast analysis tools to achieve the large number of simulations required for 'outer-loop' applications, including sensitivity analysis, uncertainty quantification and optimisation. The developed code utilises the heterogeneous nature of OpenCL so that it can be executed on any platform with CPU or GPU cores.

The GPU implementation provides a significant performance gain (4-7 times faster) over a similar OpenMP implementation. Mark is open to academic and industrial collaborations, and he is particularly interested in the simulation of non-deterministic problems. He would be very interested to hear about experimental or numerical problems that exhibit different failure modes due to small changes (or inherent randomness) in the initial conditions.