Experimental investigation and neural network development for modeling tensile properties of polymethyl methacrylate (PMMA) filament material

The present study focuses on an experimental investigation aiming at simultaneously optimizing tensile strength and elastic modulus of 3D-printed polymethyl methacrylate (PMMA) filament material while considering raster angle, printing speed, and layer thickness as the independent process-related control parameters. The Box-Behnken design of experiments (BBD) was applied to design the necessary number of experiments and establish the experimental design for fabricating experimental dog-bone samples of standard geometry. Further on, second-order regression models for tensile strength and elastic modulus were generated and employed to predict the three independent structural parameters and maximize tensile strength and elastic modulus. A neural network model was examined for its efficiency in terms of predicting the two responses. Results show a high correlation between inputs and outputs, enabling reliable modeling towards the objective of optimizing both the strength and elasticity of PMMA-fabricated parts. Regression models exhibited high correlation (R²) equal to 97.81% and 97.26% for tensile strength and elastic modulus, respectively. Simulation results referring to the neural network suggest a high correlation between outputs and targets during training (R = 0.9762), validation (R = 0.9974), and testing (R = 0.9808).