Study on the parameters optimization of 3D printing continuous carbon fiber-reinforced composites based on CNN and NSGA-II

In 3D printing of critical structural components made from continuous carbon fiber-reinforced composites (CCFRCs), mechanical performance and manufacturing efficiency are mutually constrained. This paper introduces a novel closed-loop iterative optimization method that swiftly identifies the optimal balance between performance and efficiency for the best overall results. It combines the forecasting capability of Convolutional Neural Networks (CNN) with the optimization strength of Non-dominated Sorting Genetic Algorithm II (NSGA-II). The study found that the optimal parameters as a layup angle of 0°, nozzle temperature of 260 °C, fiber filling density of 80 %, layer thickness of 0.6 mm, and fiber printing speed of 10 mm/s. The results of the optimized process parameters show a 53 % increase in mechanical performance and a 27 % improvement in manufacturing efficiency compared to the sampling experiment results. Therefore, the proposed parameter optimization strategy can quickly determine the optimal process parameters for the given conditions without requiring additional guidance.