| Numerical Modeling of Fiber Bundle Architecture in the Robotic Coreless Filament Winding Process |
| The coreless filament winding technique is a novel robot assisted manufacturing process, where pre-impregnated tows are wound around anchors in a predefined pattern without a mould resulting in shell-like truss or lattice structures. The flexibility in shape design and lightweight construction potential make the process suitable not only for the architectural sector, but also as topology-optimized fiber-reinforced connecting elements or satellite structures. The final fiber architecture results from fiber-fiber and fiber-anchor-interaction during the process, as well as anchor point positioning and orientation. Due to compaction of the fibers during the process and in the typically large number of tow crossing points in coreless wound structures, analytical calculation of fiber architecture and bundle cross-sections of complex structures is challenging. A precise fiber-bundle cross-section is required to reduce over-dimensioning and material waste. To obtain detailed data prior to fabrication and virtually replicate manufacturing as basis for structural and process optimization, a process simulation workflow has been developed using an explicit finite element approach. This showed that the modelling of fiber deformation must be included in the simulation for a realistic prediction of fiber architecture and cross-sections.
The finite element approach is able to capture the elastic deformation and friction, but requires a long computation time, limiting the applicability for large structures. In this work, an approach to model the fiber architecture and cross-section using multiple filaments in a post-processing step is presented. This is based on the work of Jones using the solver SimTex. The explicit finite element solver was originally developed for digital element simulation with an implementation of fast contact algorithms to reduce computation time for textile forming processes. In the modelling approach, a single virtual filament fiber path is transferred in an overlapping multi-filament cross-section that is subsequently expanding due to penalty stiffness in the contact among the individual virtual filaments. Such multifilament cross-section is derived by a division of the single virtual filament into multiple virtual filaments with a smaller diameter, preserving the fiber volume ratio of the resulting cross-section. The ends of the fibers are fixed in axial direction, but free to move perpendicular to it, allowing the filaments to move freely in crossing areas or at contact areas with anchors. As previous research of the authors showed that the diameter of the virtual filaments requires additional scaling to more closely match experimental results, this work extends the method described above with an optimization loop. This loop uses LS-Opt to evaluate selected cross-sections of a simulation result and modify the parameters filament diameter, static coefficient of friction, youngs modulus, pretension and penalty-stiffness until a termination criterion based on the deviation of the simulative and experimentally measured cross-section is reached. To quantify this deviation, the area of the cross-section and the moment of inertia to obtain the shape of the cross-section are used.
In order to validate simulation results, a small specimen with two anchors as basis for the optimisation and a larger specimen with four anchors for evaluation of the approach are prepared. To impregnate the fibers, an inline impregnation system mounted on a winding head of a robotic system is used with a controlled pretension during winding and an initial fiber volume fraction of 40 %. To scan the geometry after curing, an industrial laser scanner ATOS 3 and a scanning spray is used. Three tilting angles of the sample holder and multiple repositions around the rotation axis enable a high coverage ratio during scanning. The resulting point clouds of all specimens for each type are merged and filtered for outliers to average the data for a comparison with simulation results. Additionally, microsections of selected zones are evaluated with the microscope to validate scanning results.
The approach presented allows modelling of complex deformations of fiber bundles occurring during the winding process, improving prediction of the resulting fiber architecture. The additional optimization loop presented in this work showed that the rescaled simulation model can predict fiber architectures in good agreement with measured cross-sections. Nevertheless, a general statement on the applicability of this method for other structural topologies requires further investigation.
|
|
|
