Multi-Objective Optimization of FDM-Printed Gears: A Bayesian Learning Approach to Quality and Weight Reduction

Abstract

This study presents a data-driven framework for optimizing Fused Deposition Modeling (FDM) process parameters using Bayesian Optimization (BO) to enhance the mechanical and surface properties of 3D-printed gears. Five critical parameters—layer height, nozzle temperature, print speed, infill density, and cooling rate—are systematically explored through a Box-Behnken Design (BBD) of experiments, generating 45 parameter combinations. ANSYS simulations are employed to evaluate responses, including surface roughness, weight, and tensile strength, while a Gaussian Process (GP) surrogate model captures the complex relationships between inputs and outputs. A multi-objective cost function prioritizes minimizing roughness and weight while maximizing tensile strength. Bayesian Optimization iteratively refines parameter selection, balancing exploration and exploitation to converge on optimal settings. This work demonstrates the efficacy of integrating computational modeling and machine learning for efficient process optimization in additive manufacturing, offering practical insights for industrial applications.