AI & Material Talks: Ehsan Ghane on Hybrid AI Models

Welcome to the first installment of a new seminar series called: “AI & Material Talks”!

Key questions explored in the seminar include:

– How much data is actually needed to train reliable material models?
– How much faster can AI-assisted models be compared to traditional simulations?
– Can AI models capture real physical mechanisms rather than just curve fitting?

_____________

12.00 – Drop in and lunch

12.15 Welcome – IMA, Pia Lindström (7 mins)

12. 25 AI Öst – Oscar Spaak (5mins)

12.30 Ehsan Ghane, PhD – (30 mins incl Q&A)

_____________

In many industrial applications, predicting how materials behave under complex cyclic loading, plastic deformation, or manufacturing processes requires extensive testing and computationally expensive simulations. Hybrid AI models offer a promising alternative by learning from a limited amount of data while still respecting the underlying physics of the material.

In this seminar, Ehsan Ghane, postdoctoral researcher at Linköping University, will discuss how hybrid AI models can help engineers better predict material behaviour. The idea is to combine machine learning with physics-based models to capture complex behaviours that are difficult and time-consuming to simulate or measure experimentally.

The talk, titled “Hybrid AI Models for Multiscale Material Modelling in Engineering Applications” will present examples from engineering materials such as woven composites and metal forming, and show how these models can connect detailed material behaviour with practical engineering predictions used in design and production.

The seminar aims to open a discussion between researchers, engineers, and industry on how AI can complement traditional material modelling and support future engineering applications.