Physics informed neural networks for multiscale poroelastic biomaterials: Improving mechanical characterisation to inform innovation in medical technologies
- Academic lead
- Dr Greg de Boer, School of Mechanical Engineering, g.n.deboer@leeds.ac.uk
- Co-supervisor(s)
- Dr Shufan Yang, School of Mechanical Engineering, s.f.yang@leeds.ac.uk, Dr David Head, School of Computer Science, d.head@leeds.ac.uk , Dr Mark Walkley, School of Computer Science, m.a.walkley@leeds.ac.uk
- Project themes
- Data-driven methods, Health, Multiphysics & Complex Fluids
This project builds on a recently established multiscale poroelastic simulation tool developed for cartilage (Figure 1) which utilises homogenisation to couple simulations of the biofibres and material across disparate scales. Advancement of the tool will be carried out to simulate mechanical responses of a range of biomaterials as observed experimentally, with concurrent data collection to be undertaken using an indentation/sliding rig (Figure 2) and/or through a critical review of literature. Physics informed neural network modelling will be applied to validate the tool and extract the relevant multiscale material properties. The project will investigate cartilage, brain tissue, and edematous soft tissues, to demonstrate the applicability of the tool for accurately characterising the mechanical response across a varied selection of biomaterials.
This new method for obtaining multiscale biomaterial properties will subsequently be exploited to inform innovations in medical technology. Cartilage repair and synthetic tissue replacement treatments will be enhanced through a better understanding of the mechanical response under load, and the capability to rapidly pre-screen putative formulations and solutions. Safety devices such as helmets and shock-absorbers will also benefit directly from a deeper understanding of how biomaterials such as brain or other soft tissues behave in the body.