³Ô¹ÏºÚÁÏ

Machine learning and physics-informed AI are increasingly used to design and assess  structural response, but purely data-driven models can be opaque, poorly generalisable outside their training data, and disconnected from the underlying mechanics.

We work to combine data-driven methods with the mechanics-first principles at the core of our research programme — developing interpretable models that respect known physical constraints, and physics-informed architectures that use mechanics to guide learning rather than replace it.

Selected publications

• ).