Process optimization and stochastic modeling of void contents and mechanical properties in additively manufactured composites

Abstract

The effects of process parameters (print temperature, bed temperature, and print speed) on dimensional accuracy, void contents, and mechanical properties are experimentally investigated for the additively manufactured short carbon fiber reinforced composites. Such studies are carried out at bead, lamina, and laminate levels to identify the process-microstructure-property relationship. Different dimensional parameters such as height, width, and deposition-alignment are monitored for different process parameters at the bead level and its effects on lamina, and laminate are studied in a multi-level manner. The various sources of uncertainties in fused filament fabrication (FFF) based additive manufacturing (AM) are identified, and their adverse effects on microstructure and performance are analyzed. Utilizing the experimentally extracted data for dimensional variability and mechanical property at each level, physical models were adopted to accurately quantify the uncertainty. A non-intrusive polynomial chaos (NIPC) based uncertainty analysis was introduced to improve the computational efficiency and reliability of the physical models. The classical lamination theory (CLT) is used with a slight modification to account for different kinds of voids of short fiber reinforced composites manufactured by FFF. The adjusted process parameters for 5% carbon fiber reinforced polylactic acid (CF/PLA) composite showed minimum dimensional variability and maximum structural performance at a print temperature of 220°C, bed temperature of 80°C, and print speed of 20 mm/s. The maximum dimensional accuracy, minimum void contents, and improved mechanical properties support these optimized processed parameters. These optimized parameters may be related to the viscosity of the material to identify the same parameters for other material systems. The NIPC accurately predicted intra-bead, inter-bead and interfacial-bead voids contents with the overall void contents in the range of 20–26%. All these predicted void contents were incorporated in the CLT to predict the stochastic load distribution of the laminate. The stochastic model closely predicted the laminate properties in terms of axial load distributions which were validated by experimentation of [00]s, [900]s, [450]s, [00/900]s, and [±450]s laminate testing.