Physics-based computational models were created to build an innovative AI-driven multiscale materials modeling software, assisted by characterization facilities, for development of functional TW composites:
• Several modelling approaches were developed to address the multi-scale complexity of the TW materials. At KTH, a specialized approach using distortion maps was developed to simulate three-dimensional microstructures of orthotropic wood-based composites, including both hardwoods like birch and softwoods like spruce.
• At TU Wien, the physics-driven modelling efforts organised were dedicated to constructing robust computational frameworks that predict the mechanical, thermal, and optical properties of transparent wood and related biocomposites. Various advanced techniques were applied, namely: continuum micromechanics-based homogenization, numerical simulations such as Finite Element Analysis with phase field methods to capture complex behaviours, and computational ray tracing to simulate how light passes through the material and interacts with it.
• At the atomistic level, KTH advanced a molecular dynamics (MD) approach focused on homogeneous and defect-free material surfaces. These simulations quantified the compatibility between the material and liquid monomers, producing values not dependent on structural variances.
• Additionally, at VTT, a molecular dynamics-based workflow was constructed to determine the viscosity of infiltrating substances at the atomic scale, which was then automated using the AiiDA workflow manager.
• At the macroscale, VTT developed a finite element (FE) model of the infiltration process, leveraging the monomer-surface compatibility data from KTH’s MD simulations and viscosity values modelled at VTT. Practical experiments, including the delignification and infiltration of birch, balsa, and spruce wood samples, were executed at UNISS. In addition, the complex viscosity values of the infiltrated monomers were measured by POLITO at different temperatures.
• Further, at VTT a simulation workflow using the OpenLB framework was implemented to model the permeability of wood structures. This system was designed to incorporate distortion-map-based, realistic wood geometries and was evaluated under different strategies for handling capillarity.
• At Aalto, two main surrogate models that simplify complex computational simulations were built: one replicated KTH’s birch microstructure generator and another mimicked TU Wien’s biocomposite stiffness model. Extensive datasets, created using generators, enabled broad exploration of input parameters. The surrogate models were implemented in PyTorch and TensorFlow.
Predictive computational LCA models and software, based on the Safe and Sustainable by Design concept were created:
• The initial efforts to integrate AI-driven multiscale material models with life cycle assessment (LCA) tools for the safe and sustainable development (SSbD) of TW composites have focused on building a structured workflow and screening models thank to the joint effort of VTT, UNISS, AIMPLAS, POLITO, KTH, and industrial partners.
• A first screening LCA model for the TW manufacturing was developed using the SULCA software. This early model will serve as a baseline for future integration in the Modeling Factory. The model assimilates data such as energy consumption and process performance indicators, linking them to LCA outcomes. In addition, material properties generated from multiscale modelling (e.g. transparency, mechanical strength) are mapped to the use phase, enabling estimation of the material’s life span in different applications.
• The screening LCA model, constructed from lab-scale processing data provided by project partners, enabled a comparative assessment of different delignification and lignin modification methods throughout the manufacturing phase.
