Mesoscale Results from the Virtual Permeability Benchmark on Engineering Textiles
Topic(s) : Special Sessions
Co-authors :
Elena SYERKO (FRANCE), Tim SCHMIDT (GERMANY), David MAY (GERMANY), Christophe BINETRUY (FRANCE), Stephan LOMOV (BELGIUM), Suresh ADVANI (UNITED STATES), Luisa SILVA (FRANCE)Abstract :
In order to optimize liquid composite molding processes, filling simulations can be carried out to ensure a fast and void-free impregnation of the fibrous reinforcement. A key input parameter required for the filling simulation is the permeability, which quantifies the resistance to flow of the fibrous structure. Permeability is usually determined experimentally. The procedure has been recently standardized in ISO 4410. However, experimental measurements are time-consuming, require specialized equipment and show a high variation of at least 20% [1],[2]. Alternative determination of permeability numerically using a Representative Volume Element (RVE) does not have an established methodology yet. A large variety of methods was observed in the first stage of the Virtual Permeability Benchmark (VPB) [3], where participants had to calculate the permeability on a REV at the scale of fibers. Hence, a second VPB stage was planned.
In the second stage a 3D scan of a textile stack with 14 layers of a glass twill-weave fabric HexForce 01102 was considered at the scale of the textile. The RVE fiber volume fraction of 54 %, representative of processing conditions, allowed for flow at the inter-tow and intra-tow scale. The objectives of the VPB are to identify possible sources of error in different approaches and ultimately develop the guidelines for the accurate permeability prediction. Participants of the second stage were provided with segmented images of the RVE with dimensions 700x680x345 voxels (8 µm/voxel) and orientations map indicating the orientations of warp/weft yarns in order to assign corresponding local tow permeability tensor from the first VPB stage. Mandatory tasks were the calculation of the permeability tensor: i) neglecting the intra-tow flow (single-scale flow) and ii) taking the intra-tow flow into account (dual-scale flow), in order to assess the caused error in case of simplification of a computationally expensive problem.
In the second stage of the VPB 15 participants submitted in total 64 results for the single-scale and dual-scale flow cases. After preliminary elimination of evident outliers, the results of dual-scale simulations showed the following coefficients of variation for the principal components of the permeability tensor: 55% and 81% for Kxx and Kyy in-plane components respectively, and 83% for out-of-plane component, Kzz. Since both stages of the VPB use the same fabrics from the experimental permeability benchmark [1],[2], the numerical predictions were compared to the measurements: the predicted values either fell within the main cluster of experimental results, or were lower. Some participants calculated permeability not only for the provided RVE of twill fabrics but also on a bigger volume that included more information on the material variability. However, these calculations did not result in a closer match with the experimental measurements. Further detailed analysis of the results will be presented in the talk.
In the second stage a 3D scan of a textile stack with 14 layers of a glass twill-weave fabric HexForce 01102 was considered at the scale of the textile. The RVE fiber volume fraction of 54 %, representative of processing conditions, allowed for flow at the inter-tow and intra-tow scale. The objectives of the VPB are to identify possible sources of error in different approaches and ultimately develop the guidelines for the accurate permeability prediction. Participants of the second stage were provided with segmented images of the RVE with dimensions 700x680x345 voxels (8 µm/voxel) and orientations map indicating the orientations of warp/weft yarns in order to assign corresponding local tow permeability tensor from the first VPB stage. Mandatory tasks were the calculation of the permeability tensor: i) neglecting the intra-tow flow (single-scale flow) and ii) taking the intra-tow flow into account (dual-scale flow), in order to assess the caused error in case of simplification of a computationally expensive problem.
In the second stage of the VPB 15 participants submitted in total 64 results for the single-scale and dual-scale flow cases. After preliminary elimination of evident outliers, the results of dual-scale simulations showed the following coefficients of variation for the principal components of the permeability tensor: 55% and 81% for Kxx and Kyy in-plane components respectively, and 83% for out-of-plane component, Kzz. Since both stages of the VPB use the same fabrics from the experimental permeability benchmark [1],[2], the numerical predictions were compared to the measurements: the predicted values either fell within the main cluster of experimental results, or were lower. Some participants calculated permeability not only for the provided RVE of twill fabrics but also on a bigger volume that included more information on the material variability. However, these calculations did not result in a closer match with the experimental measurements. Further detailed analysis of the results will be presented in the talk.