This is an Innovation Fellowship funded by the EPSRC Future Composites Manufacturing Research Hub.
Accurate permeability data is key to high-fidelity resin flow simulations for Liquid Composites Moulding (LCM) yet remains an unsolved challenge due to variabilities at different length scales. This project will bring powerful numerical methods for statistically accurate permeability calculations and filling fundamental knowledge gaps.
Three work packages (WPs) are
• WP1 Development of efficient numerical solvers for predicting the dual-scale flow in textile fabrics. To alleviate the issue related to computational cost, this WP will establish a large-scale simulation framework based on the Fast Fourier Transform (FFT) methods. This framework will enable efficient parallel computing with high performance computing (HPC) systems.
• WP2 Permeability uncertainty correlated to microstructural variability for NCFs. This WP will involve a large number of experimental tests, mainly using X-ray Computed Tomography (XCT). Advanced and dedicated image processing techniques will be applied and developed for the 3D images of various NCFs. The extracted realistic microstructures will be input into the FFT solvers to calculate the effective permeability using full-field homogenisation techniques.
• WP3 Development of stochastic surrogate models. Although the FFT solvers developed in the previous WP can be efficient, they will still be computationally demanding. To reduce the requirement of HPC capacity, this WP will develop surrogate models based on machine learning techniques, such as convolutional neural networks and physics-informed neural networks. The big datasets produced from the XCT and the FFT simulations will be essential for training these surrogate models.