Scientific and Physics-Informed Machine Learning for Industrial Applications

Nowadays, Scientific Machine Learning (SciML) is revolutionizing the academic and industrial world like a storm. It combines traditional scientific mechanistic modelling (differential equations) with the machine and deep learning methodologies. As it is well known, traditional Deep Learning suffers some issues like interpretability and enforcing physical constraints; combining such methodologies with numerical analysis and differential equations can bring to a new field of research through new methods, architectures and algorithms. SciML techniques aim to overcome the classical barriers of the data-driven approaches like (i) the significant amount of data required from data-driven models to identify and interpret events/signals, (ii) the generation and collection of data often not fitting the purpose. It can be stated that incorporating physical models will bring us many benefits to the machine learning approaches. By deeply looking at the industrial scenario, if we consider the manufacturing task, SciML is about applying physical principles and laws to process materials into useful products