Optimising process parameters for enhanced performance in PLA-based fused filament fabrication: A multi-objective approach

Abstract

Manufacturing technologies, particularly additive manufacturing (AM), have ushered in innovative practices for complex part production, prominently exemplified by fused filament fabrication (FFF) for the production of complex parts. Significant efforts have been ongoing with a focus on improving performance measures of biodegradable bioplastics, notably polylactic acid (PLA), within the eco-conscious packaging industry. Findings in this field have given rise to FFF technologies for enhancing critical performance measures such as surface roughness (Ra) of food packaging by investigating optimal process parameters such as layer thickness, raster width, raster angle, printing speed, air gap and bed temperature, to name but a few. Additional performance measures affecting geometric part quality, such as thickness variations along horizontal and vertical geometric planes, have been analysed in previous studies to derive solutions for enhancing the correlation between process parameter input and dimensional accuracy. Despite these advancements, research on 3D-printed specimens has typically focused on manufacturing standards with thickness greater than 1 mm, as opposed to the thinner requirements of some plastic packaging products. This practice has left a gap of uncertainty regarding the potential mechanical performance of thin-film 3D-printed products for current and future applications. The necessity to induce fewer infill layers during the FFF process has sparked interest in evaluating the efficacy of process parameters in achieving reliable mechanical performance in surface quality and Z-axial dimensional accuracy compared to conventionally manufactured plastic products. This article, therefore, sought to address this uncertainty using a design of experiments (DoE) approach followed by multi-objective optimisation through the application of genetic algorithm (GA) techniques. From these efforts, optimal process parameters of layer thickness (0.151 mm), raster width (0.514 mm) and raster angle (82.75^o) for the minimisation of Ra and Delta-Z (∇Z) variation in PLA-based FFF specimens. Further data analysis noted variability in the accuracy of Ra predicted R² values depending on the raster angle and void defects at the surface layer, while additional observations indicated an increase in Z-axial variation when a layer thickness of 50% of the overall specimen thickness was utilised. This study concluded by underscoring the necessity for future research in integrating multi-objective optimisation further, with the aim of not only improving the predictability of process parameter optimisation but also the detection, analysis and minimisation of defects using machine learning tools and techniques.