Concurrent multiscale micromechanical modelling of fiber networks

This PhD project aims to develop an advanced multiscale model for fibre network materials by integrating machine learning, surrogate modelling, and concurrent multiscale methods.

The primary goal is to accurately predict the mechanical and hygroscopic behaviour of fibre networks under various environmental conditions and understand the role of fibre joint compliances on their mechanical response.The project comprises four work packages (WPs): fibre network generation, fibre joint compliance surrogate modelling, development of a general concurrent multiscale scheme, and model evaluation and application.

A 3D Convolutional Neural Network (CNN) will be used for fibre recognition and segmentation in 3D tomography images, facilitating the reconstruction of realistic fibre network models.

Gaussian Process Regression (GPR) surrogate models will be employed to predict fibre joint compliances and integrated into fibre network simulations.

The concurrent multiscale approach will incorporate material non-linearities, hygroexpansion, viscoelasticity, damage and handle arbitrary geometries of fibre joints.Leveraging unique numerical and experimental facilities at KTH, the methods developed in this project will contribute to solving the outstanding problems that fibre network materials can exhibit in industrial and end-use applications.

Therefore, these methods will aid in developing resilient fibre network materials with tailored properties to meet specific requirements.