Materials discovery using deep learning

We propose a novel deep learning tool for materials discovery. The approach merges together both experimental data and computer simulations to exploit all possible sources of information. The tool has been used to design several materials that have been experimentally verified and patented. We present just one case study where we discover and characterize the new nickel-base alloy for direct laser deposition most likely to simultaneously satisfy targets of processibility, cost, density, phase stability, creep resistance, oxidation, and resistance to thermal stresses. Experimental testing confirms that the physical properties of the proposed alloy exceed those of other commercially available Ni-base alloys for combustor liner applications.

Probabilistic design of a molybdenum-base alloy using a neural network
B.D. Conduit, N.G. Jones, H.J. Stone & G.J. Conduit
Scripta Materialia 146, 82 (2018)

Materials data validation and imputation with an artificial neural network
P.C. Verpoort, P. MacDonald & G.J. Conduit
Computational Materials Science 147, 176 (2018)

Design of a nickel-base superalloy using a neural network
B.D. Conduit, N.G. Jones, H.J. Stone & G.J. Conduit
Materials & Design 131, 358 (2017)

Bio: Gareth Conduit has a track record of applying artificial intelligence to solve real-world problems. The technique, originally developed for materials design, is now being commercialized by startup Intellegens in not only materials design, but also infrastructure, drug discovery, and healthcare. Previously, Gareth had research interests in strongly correlated phenomena, in particular proposing spin spiral state in the itinerant ferromagnet that was later observed in CeFePO. Gareth's group is based at the University of Cambridge.