Φ-ML Meets Engineering Seminar Series

1 minute read

Φ-ML Meets Engineering” is a newly launched bi-monthly seminar series discussing applications of Physics-enhanced Machine Learning (Φ-ML) methods in Engineering practice.

Machine learning (ML) applications in engineering have been applied to both small and large-scale problems. However, the necessity of using large training datasets together with the ‘black box’ nature of ML learning methods may compromise their interpretability, which is necessary for practical applications. In this context, Physics-informed machine learning (Φ-ML) is an emerging subfield to tackle such difficulties. Φ-ML methods aim to integrate known physical understanding of a phenomenon, e.g. expressed as ODE/PDE/SDEs, into the ML-learning framework.

Our very first seminar was held online last Thursday, 25th March 2020 and was a great success with 100+ subscribers and 60+ attendees at our first session. Our first speaker was Dr Gabriel D Weymouth, who is an Associate Professor at the University of Southampton and a Research affiliate at the Turing. Dr Weymouth shared his Φ-ML approach with applications in Fluid Mechanics. In general, Fluid Mechanics simulation are incredibly costly because of the vast range of relevant interacting time and length scales. As such, machine learning approaches to speed up fluid mechanics predictions are appealing - but hard to realize in practice. Dr Weymouth’s group has been working on a range of different approaches to solving this problem, including physics-based feature spaces and using CNNs to map from 2D to 3D turbulence simulations. Their recent attempts to use Optimal Transport for simulation interpolation was much less successful but highlighted key issues such as boundary condition and solution representation.

A recording of the first seminar will be uploaded to the seminar website shortly. Please visit the seminar website to subscribe to the seminar series and to learn about upcoming seminars.

The seminar series is co-organised by Andrea Pizzoferrato, Themis Botsas and Zack Xuereb Conti.

Updated: