Machine learning-based prediction and optimization of surface roughness in FDM-printed PLA using ANN and box-behnken design

This study aims at the parametric investigation of Fused Deposition Modelling (FDM) and developed Polylactic acid (PLA) for the optimization of surface roughness. A hybrid Response Surface Optimization with Artificial Neural Network i.e., RSO-ANN methodology was adopted for the optimization of different input parameters namely number of layers, infill ratio, infill orientation and printing temperature which affect the surface roughness of PLA. The ANN model predicted with the regression values 0.9225 which represent the adequacy of the predicted model. The predicted results closely match the experimental data with an error percentage ranging from 0.83% to 1.04%. Thus, it resulted into the effectiveness of the ANN model. Under these parametric configuration, the surface roughness is significantly reduced from 1.2 to 0.8 µm.