Optimizing Recycled Bulk Molding Compound (rBMC) with Machine Learning

Abstract

Carbon fiber composites (CFCs) are important materials extensively utilized in aviation, transportation, and sporting goods industries. Carbon fiber prepreg, a precursor to CFCs, is in high demand for fabricating CFC components. A significant portion of the waste stream from producing CFCs is uncured prepreg, comprising 56-70% of the total waste generated [1]. Prepreg is a high cost, high value material, repurposing scrap and expired waste can be economically viable, in addition to being environmentally friendly. Previously, uncured prepreg has been recycled by chopping into discontinuous fiber composite (DFC), but intrinsic flaws, primarily co-location of flake ends, leads to poor properties, especially with composites thinner than 3 mm in thickness. In this work, a recycled bulk molding compound (rBMC) has been developed that mixed flakes with extra resin to improve uniformity and was processed via hot pressing to produce a defect-free composite. 0.25”-1” and 0.25-0.5" flakes showed poor dispersion. 0.25”-0.125” flakes showed exhibit good dispersion, however, the interfacial contact between fibers and resin was poor, diminishing the fibers' properties in composites. Conversely, 0.25"-0.25” flakes showed superior dispersion and mechanical properties. Gaussian Process Regression (GPR) model optimized rBMC materials based on resin content.