FDMプロセスのCRITIC-CoCoSoおよび機械学習アプローチを用いたNylonAF80フィラメントの多基準最適化

This study aimed to optimize the fused deposition modeling (FDM) process parameters for an 8% aramid fiber-reinforced polyamide filament (NylonAF80) to enhance the quality of printed parts. Six parameters were varied: the slice height, nozzle temperature, bed temperature, deposition speed, raster orientation, and build orientation. Cuboid samples with a central circular hole were printed using Taguchi's L18 orthogonal array, and evaluated for volumetric error, average roughness, and Shore D hardness. The CRITIC-CoCoSo technique identified optimal settings of 0.1 mm slice height, 255 °C nozzle temperature, 100 °C bed temperature, 40 mm/s deposition speed, 90° raster orientation, and on-edge build orientation to maximize the multi-response performance index. Sensitivity analysis using different weighting schemes and MCDM methods showed strong correlations (> 0.93) between rankings. The decision tree algorithm exhibited superior performance to k-nearest neighbor, stochastic gradient descent, and logistic regression in classifying output responses, achieving 88.9%, 94.4%, and 100% accuracies for hardness, volumetric error, and roughness, respectively, at an 80:20 split ratio. Slice height was the most influential factor for all responses. FESEM microstructural analysis revealed pores and voids in the samples printed under extreme conditions and at high slice heights. These findings provide valuable insights for optimizing FDM parameters to improve the quality of NylonAF80 parts for industrial applications.