DISCRETE MODELING AND MACHINE LEARNING ASSISTED CALIBRATION OF 3D PRINTED CARBON FIBER REINFORCED PLASTICS (CFRP) STRUCTURAL JOINTS

摘要

Over the past few decades, Additive Manufacturing (AM) has rapidly grown and has been very quickly adopted across a wide variety of industrial and academic fields. Although materials and structures manufactured using this technique are generally inferior in strength than those manufactured using traditional techniques, the many inherent advantages of AM still make it a great alternative to conventional composites. Fused Deposition Modeling (FDM) is by far the most widely used AM technique due to its capability to create complex parts by melting and extruding filaments of many different materials according to a specific and a-priori optimized pattern. With improvements in controls and robotics technologies, it is nowadays possible to deposit long continuous bundles of carbon fibers embedded in a snap-cure resin, allowing to greatly increase the strength-to-weight ratio of a complex structural joint where different layers of materials are thoughtfully and methodically placed to better react the loading conditions. The Discrete Model for Composites (DM4C), currently under development, is used to generate both simple and more complex joint structures, which will be calibrated using machine learning algorithms after computational results are gathered by a massively parallelized HPC code and stored in a Structured Query Language (SQL) database.