Embedding Physics in Digital Twins: The Statistical Finite Element Method and Recent Progress

Eky Febrianto (Uni of Glasgow (James Watt School of Engineering))

Wednesday 18th February 12:00-13:00
Maths 311B

Abstract

Digital twins enable continuous monitoring and assessment of engineering assets. Modern sensor instrumentation produces rich streams of data, but interpreting these measurements typically requires strong domain expertise. In parallel, these assets are designed and analysed using numerical models based on deterministic physical laws, most commonly discretised with the finite element method (FEM). In practice, data and models are used together to support more robust decision making, for example, in risk assessment and maintenance planning. The statistical finite element method (statFEM) provides a principled link between physics-based FEM models and observational data, which enables coherent inference of the underlying true state of an asset and its parameters. In this talk, I will introduce statFEM and illustrate how it can be used in real applications, from true state estimation in infrastructure to the identification of material properties within physics-based models. I will also highlight recent progress on a key computational bottleneck in digital twins: repeated evaluation of forward model. In particular, I will discuss surrogate modelling strategies and outline emerging opportunities such as quantum computing for accelerating inference workflows.

Add to your calendar

Download event information as iCalendar file (only this event)