A Novel Simulative-Experimental Approach to Determine the Permeability of Technical Textiles

Abstract

Since years, fiber reinforced polymer composite (FRPC) parts made by Liquid composite molding (LCM) make up a significant share of the composite market. In LCM dry reinforcing structure gets impregnated with a resin system. Permeability, a material parameter with key influence on all LCM processes, quantifies the conductance of technical textiles for resin flow. Today, when a numerical filling simulation is applied for process design, large experimental test programs are required for characterization of permeability, as the permeability has to be measured for all textiles processed and also the dependency e.g. on the fiber volume content has to be considered. Together with Math2Market GmbH (M2M), the IVW currently develops a novel simulative-experimental approach (SEA), using experimental tests to calibrate a simulation model for replacing a significant amount of the experimental tests through “virtual” measurements. In a first step the functionality of the simulation and the most appropriate methods for textile modelling were investigated. For this, three routes were followed: At first a micro-computer tomograph (μCT)-scan of a glass fiber non-crimp fabric was fed into GeoDict, the material simulation software developed by M2M. Second, a digital model (DM) of the textile was created by computer modelling of basic structure and subsequent virtual compaction. μCT-model and DM were then used for computational fluid flow simulation which gives the direction-dependent permeability as an output. The DM calibrated by experiments represents the SEA and results in the digital twin. Third route was the experimental permeability measurement to generate reference values. Comparing the results of all three routes allows statements about the functionality of the simulation and accuracy of modelling. The rather deficient correlation between the results of experiments and μCT-model based simulations revealed that segmentation is a remarkable source of error despite the use of recognized methods. Different modeling approaches were followed to build up the digital twin. The best results were achieved with models undergoing a virtual compaction step, which takes various imperfections such as yarn deformation and varying nesting behavior into account. With this 2.25 - 6.75% deviation from the experimental results at an average standard deviation of 21.9 - 61.2% were achieved. Hence, the digital twin shows a better correlation than the μCT-model and high potential for substitution of experiments. Even better results are expected when in a next step a local anisotropic permeability will be allocated to the yarns.