Event image

SPEAKER

Dr Gareth Conduit, Department of Physics, University of Cambridge

SYNOPSIS:

Machine: Materials discovery with artificial intelligence

We have developed a computational tool that employs deep learning to discover new materials. The tool has proposed five alloys for use in jet engines, whose properties have been experimentally verified and are now undergoing testing by Rolls-Royce plc. The approach is now being developed further to combine experimental results with first principles calculations into a holistic materials design tool.

 

———–

Man: Developing real alloys in practice vs. thermodynamic database-based alloy optimisation

Professor David Dye, Department of Materials, Imperial College London

The Materials Genome Initiative and its proponents, most notably Olsen and Reed, propose that our modelling capabilities are such that alloy design can now be reduced to the development of alloy optimisation tools, reducing the need for expensive and time-consuming empirical alloy development programmes (and Professors!). In the strongest version of the vision, modelling across the length scales is said to enable the complete replacement of experimentation with design “in silico”.  Whilst this is a compelling vision, I will suggest that we might think about modelling tools as aids to the alloy designer, rather than replacements for her.  We will begin with the limitations of thermodynamic databases, and examine their empirical underpinnings.

Examples from titanium alloys and superalloys will be used to highlight the point.  Then we will go down to the atomistic length scale and examine how DFT models can be helpful, without requiring them to be completely predictive. We will then return to the integrated computational materials engineering (ICME) vision and unpick how recent successful alloy designers have gone about their craft. We will then highlight how mesoscale modelling tools have helped us to optimise processing, and visit some successes and disasters. We will conclude with some thoughts on how to use modelling in materials design in the real world.