Physics-informed machine learning meets engineering seminar series

Introduction

Machine learning (ML)-based methods have found a wide range of applications in engineering. They can be applied to small and large-scale problems, from analysing large amounts of data to optimise designs and predict failures, to improving manufacturing processes.  

ML methods are a popular choice due to their proficiency at finding hidden patterns, even in scenarios where the complexity prohibits analytical descriptions. However, the necessity of using large training datasets together with the difficulty of interpreting internal workings of ML models may compromise interpretability, which is necessary for practical applications.  

Physics-informed machine learning (Φ-ML) is an emerging subfield that aims to tackle such caveats. Φ-ML integrates physical laws and principles into the design and training of machine learning algorithms.

By incorporating prior knowledge from physics into machine models, Φ-ML can lead to more accurate predictions and better generalisations.