Project Overview
In this project, we aim to revolutionise the current artificial intelligence modelling approaches of textile reinforcements by integrating machine learning with the extensive wealth of physics knowledge. This facilitates the development of robust material models of textile reinforcements with unparalleled reliability (in terms of extrapolatability), thereby enabling the prediction of unseen scenarios with confidence. This new capability will enable the virtual anticipation of forthcoming threats in areas such as armored vehicles, aircraft, and infrastructure, thereby enhancing preparedness and proactive defense strategies.
To this end, large mechanical datasets of textile reinforcements will be collected by means of full-field measurements and in-silico modelling. The function space of the learners will be constructed through feature transformation based on analytically homogenised strain measures. The data-driven statistical learning framework will be developed on the basis of inverse finite element computation, which selectively identifies the optimal model from the hypothesis space. By incorporating analytical inelasticity and thermodynamic regularisation into the function space of the approximators, the generalisation error can be quantified and adapted to changes in the data, thereby ensuring extrapolability of the selected model.
Project Aims
- Collect large datasets for diverse types of textile reinforcements
- Develop a data-driven statistical learning framework that is technically robust, reliable and explainable to learn (i.e. to model) generalised mechanics of textile reinforcements from the collected data
- Integrate domain knowledge of continuum and computational mechanics into the aforementioned machine learning framework
Project Team
- Prof. Antonio J. Gil (a.j.gil@swansea.ac.uk)
- Dr Philip Harrison (philip.harrison@glasgow.ac.uk)
- Dr Anil Bastola (a.k.bastola@swansea.ac.uk)
- Charlie Patterson (charlie.patterson@glasgow.ac.uk)
- Dr Khiêm Vu Ngoc(n.k.vu@swansea.ac.uk)
Collaborators
Swansea University; University of Glasgow