Numerical simulations are commonly used to predict resin flow in fibrous reinforcements but exhibit a trade-off between accuracy and computational cost. As an alternative, machine learning (ML) based models pose as a potential tool to accelerate or replace such costly simulations. This work proposes an open-source image-based deep learning framework to estimate the permeability of unidirectional microstructures in arbitrarily sized domains. This presents a scalable step towards estimating the permeability of large meso-domains. First, we present two robust and accurate surrogate models capable of predicting microstructure velocity and pressure fields with varying physical dimensions, fiber diameter, and volume fraction. These predictions achieve 5% error on the training set and 8% error on unseen microstructures. Secondly, based on those predicted flow fields, we infer the permeability of the microstructures with respectively 4% and 6% deviation for the training and validation sets. Third, opposed to previous works limited to microstructures with a fixed aspect ratio, we propose a so-called sliding window procedure, based on physics-based principles to predict the resin velocity and pressure field in microstructures with different aspect ratios. The method is validated against high-fidelity numerical simulations, and its predictive performance and computational efficiency are confirmed with μ-CT scans of real microstructures. Finally, the presented code and surrogate model are open-sourced to promote further exploration by the scientific community.

A new and sustainable membrane manufacturing method is 3D printing, which reduces the number of fabrication steps, waste production, and the corresponding CO₂emissions. It further enables fabricating membranes with well-defined pore size, shape, and configuration. Here, we study 3D printing of microfiltration membranes using a novel dual-wavelength microstereolithography method. Via the gradient descent method, we are able to calculate and control a printable membrane with micrometer precision, enabling the possibility of printing membranes directly. Hydrophilic porous membranes with cylindrical microscale pores (≈10 μm in diameter) are printed from polyethylene glycol diacrylate (PEGDA). Membrane printing procedure and postprocessing steps are thoroughly investigated to print consistent membranes with uniform thickness. The membranes are fully characterized using SEM, FTIR, contact angle, and surface roughness measurements. The pure water permeability and separation performance of the 3D-printed membrane are further investigated and compared with those of commercial hydrophilic PTFE membranes. The 3D-printed membranes show similar permeability values to those of commercial membranes and could successfully separate oil droplets from oil-in-water emulsions. The membranes’ permeability is further predicted using a 1D tube model and numerical modeling. The effect of material’s property (e.g., swelling) and pore deformation during pressurization are studied to understand the discrepancy between the calculated and the experimental permeability values. The results provide valuable insights into the permeability prediction of 3D-printed membranes and the corresponding design optimization.

Multiple phenomena occurring at the microscopic scale affect the final mechanical performance of composite parts manufactured through processes involving impregnation of dry fibers, such as resin transfer molding. Formation of fiber-poor areas in specific locations or air entrapment within the resin are issues that commonly arise during the impregnation. Such challenges have motivated the use of numerical simulations to understand the manufacturing processes better and to optimize the process design. However, the limitation imposed by their computational cost has encouraged the use of machine learning (ML) to replace them. Thus far, the state of the art has focused on predicting the permeability of fiber-reinforced microstructures. We expand the limits by proposing an ML-based surrogate for microscale steady-state velocity prediction of a fluid flowing through a fibrous microstructure. This model, inspired by the U-net architecture, takes as input the image representation of fiber-reinforced composite microstructures. It subsequently outputs the resin velocity field around the fibers based on prescribed boundary conditions. Those results are further used to estimate the permeability of the microstructures, thus encompassing previous works. We describe in this work the computational pipeline of our approach, starting from generation of the ground truth data to the optimization of the UNet hyperparameters.