Optimizing 3d Printing Process Parameters to Minimize Surface Roughness Using Bayesian Optimization

Abstract Additive manufacturing techniques, particularly Fused Deposition Modeling have been extensively used in various industries due to their advantages in part design and flexibility over the conventional subtractive manufacturing processes. The process parameters play a crucial role in FDM 3D printing and need to be carefully tuned and optimized to achieve the desired quality, mechanical strength, and accuracy of the printed parts. This paper aims to optimize the process parameters in FDM 3D printing using a Bayesian Optimization (BO) method to minimize the surface roughness of the printed part. Nozzle temperature, layer height, bed temperature, print speed, and infill density are the five continuous input process parameters considered while the surface roughness is the output response. 34 Latin Hypercube Samples (LHS) are printed on a Prusa i3MK3S+ 3D FDM printer to generate the initial samples used in the optimization. Sequential BO with expected improvement as the acquisition function is used to find the optimal process parameters by balancing exploration and exploitation. The overall mean surface roughness (Ra) of the 57 experiments conducted was 6.09 μm and the optimizer produced an optimum value of 2.78 μm when the nozzle temperature was set to 216 °C, bed temperature to 66 °C, print speed to 73 mm/s, infill density to 20% and layer height to 0.123 mm respectively.