Physics informed neural networks fro harmonic Maxell equations – MAXINET

The MAXINET project aims to develop a new generation of artificial intelligence (AI)-enhanced digital engineering tools for solving harmonic Maxwell's equations applied to the design of radio-frequency components and systems. These tools are based on the Physics-Informed Neural Network (PINN) concept, which is a modern concept in Scientific Machine Learning (SciML) research. MAXINET proposes to explore this concept by focusing on the scientific obstacles identified for its effective application to the specificities of electromagnetic wave propagation in the harmonic regime. These include non-trivial boundary conditions, complex geometries and multi-material domains (heterogeneous media). Another important research direction in MAXINET is to build ultra-fast substitution models for multi-parameter studies, based on the neural operator concept. This will pave the way towards exploiting the AI-enhanced digital engineering tools produced by the project for the reverse design of radio-frequency components and systems.