Modeling the influence of fused filament fabrication processing parameters on the mechanical properties of ABS parts

Modeling the influence of the processing parameters of fused filament fabrication (FFF) on the mechanical properties of FFF fabricated parts is a challenging task due to the complex dynamics and the large number of factors that affect the quality of the fabricated parts. Therefore, optimizing the mechanical properties of parts fabricated using FFF usually requires a considerable number of test samples. Most past studies have focused on the main effects of the processing parameters and have ignored the interactions between the parameters or their non-linear effects on the mechanical properties of FFF fabricated parts. In the work presented, the effects of the layer thickness, nozzle temperature, infill percentage, and infill pattern are investigated to achieve the fabricated parts' optimum strength and stiffness. A group of Artificial Neural Networks (ANN) was used to model the influence of these processing parameters on the mechanical properties of FFF fabricated ABS parts. The Design of Experiments (DOE) approach was utilized to minimize the number of tests needed to study the investigated parameters. Response Surface Methodology (RSM) and Taguchi's orthogonal arrays were used to generate the training and testing data sets used to develop a group of ANN models and evaluate their performance. Furthermore, the fitness of ANN models was compared to the regression models of the RSM and Taguchi's DOE. It was found that some parameters exhibit strong interaction and nonlinear effects on the strength, stiffness, and ductility of FFF fabricated parts. The coefficient of determination R2 indicates that the ANN models were more accurate at predicting the mechanical properties than DOE regression models. The contour plots show that generally, increasing the layer thickness and infill percentage increase the tensile strength and the elastic modulus and that increasing the nozzle temperature is important when thick layers are used. For the ductility, the infill pattern is the most significant parameter, with linear infill yielding the highest ductility and triangular yielding the lowest. Lower nozzle temperature generally improves the ductility which is the exact opposite of what is required to maximize the elastic modulus and tensile strength. Finally, a multi-objective particle swarm optimization algorithm was used to obtain the non-dominated Pareto-optimal solutions that can lead to an optimal combination of the mechanical properties.

Automated summary

Modeling the influence of FFF processing parameters on mechanical properties is complex, with interactions and non-linear effects being important factors. ANN outperformed DOE regression models in predicting mechanical properties, showing that parameters like layer thickness, infill pattern, and nozzle temperature significantly affect strength, stiffness, and ductility.