Numerical simulation and machine learning-based prediction and optimization of mechanical properties of FDM-printed PLA

Abstract

In recent years, the interest towards the use of Polylactic Acid (PLA) structures produced via the fused deposition modeling (FDM) technique in the field of sustainable 4D printing has increased rapidly, owing to the properties of shape memory and biodegradability. In this research, the effects of strip thickness, raster angle and programming temperature on thermo-mechanical performance of PLA smart structures are explored by means of a combined method of finite element analysis (FEA), adaptive neuro fuzzy inference system (ANFIS) and genetic algorithm (GA). In order to investigate the influence of the factors on output responses of maximum induced stress and total strain energy, a full factorial DoE including 64 simulation cases is generated through ANSYS Workbench 2024. The results indicate that strip thickness has the highest effect on the responses, whereas raster angle has the lowest one. It is observed that the increases in strip thickness and temperature cause increases and decreases in stress resistance and strain energy storage, respectively, owing to thermal softening. The ANFIS models are then built for the prediction of the outputs and show very good predictive performance, with mean absolute errors of 6.23% and 4.50% for maximum induced stress and total strain energy, respectively. Finally, a multi-objective GA is incorporated with ANFIS for optimization of the process parameters by minimizing stress and maximizing strain energy simultaneously.