Applying machine learning methods to engineering simulation
29 April - 9 June 2024
Biography
Dr Evins’ research group has pioneered the use of machine learning algorithms as ‘surrogate models’ for detailed engineering simulations of building energy use, achieving huge improvements in speed and usability. Models that provide fast approximations using machine learning can be used in online dashboards to explore complex engineering design problems in an interactive manner, revolutionising the role of such tools in the building design process. He also works on the application of machine learning algorithms directly to data about buildings, from IoT sensors, smart meters and thermostats up to city-wide catalogues of the building stock. Prior to starting his research group in Canada, he worked at Empa / ETH Zurich in Switzerland and Buro Happold Ltd in London, where he competed and Engineering Doctorate with the Systems Centre at the University of Bristol.
He has published 16 journal papers and 31 conference papers in the past 4 years (in engineering and computer science, conference papers are significant contributions). His work has been cited 4,199 times with an h-index of 26. In the last 7 years he has supervised 15 graduate students and 8 post-doctoral researchers, funded by $3.8M CAD in competitive grants. He is Director of the ReBuild Initiative (2022-2026, $2M CAD), which brings together 15 industry and government partners to apply machine learning methods to building data. He has led the development of two open-source software platforms ($1M CAD government funding) to make these methods accessible for use in academia and industry. He has recently begun to commercialise this research directly through his consulting company, Edifican.
Project Summary
While visiting the University of Bristol, Dr Evins will build connections between the application area of low-energy building design and modelling (and the wider field of engineering simulation) and emerging developments in machine learning (ML) and artificial intelligence (AI). This research area has been a very fruitful field of study for Dr Evins, where he has successfully introduced the use of machine learning algorithms as ‘surrogate models’ for detailed simulations of building energy use, achieving huge improvements in speed and usability. Models that provide fast approximations using machine learning can be used in online dashboards to explore complex engineering design problems in an interactive manner, revolutionising the role of such tools in the design process. This fledgling field has grown dramatically since his first published work in 2019, with great interest from practitioners and policy-makers as well as academics.
This project will be realised through Dr Evins collaborating with Professor Peter Flach and many other researchers and students across the CDT in Interactive Artificial Intelligence at the University of Bristol. In bringing his expertise in surrogate modelling to Bristol, Dr Evins will raise awareness of these emerging methods with students and researchers across campus. The underlying methods are very widely applicable, as they can be used in any area of engineering simulation. The benefits of increased speed, flexibility and ease of deployment can bring benefits in any engineering discipline. Dr Evins will collaborate with all interested researchers to help them apply these methods to their specific challenges.
Dates, times and venues of Dr Evins' lectures and seminars will be posted on the Events page in due course, in the meantime please contact Dr Evins' host, Professor Peter Flach, for further information.
