Gaussian process regression for optimizing L-DED processing parameters in AlSi10Mg additive manufacturing

Abstract

Abstract This work presents an experimental workflow coupled with statistical tools to optimize L-DED processing parameters to build defect-free AlSi10Mg multilayer parts. Experimental steps included the quantitative evaluation of single beads, single layers, and multilayers to define L-DED processing windows. Regression and desirability functions, together with multi-objective Bayesian optimization, were used to optimize the main L-DED process parameters to meet target single-layer features. The systematic collection of reliable experimental data was critical to the success of the optimization models, and a detailed methodology is provided. The proposed workflow enabled the production of crack- and macroporosity-free multilayer parts with high geometrical accuracy and metallurgical bond integrity, achieving densities from 99.55% to 99.80% and deposition efficiencies between 87 and 97%, significantly exceeding previously reported values. This work provides practical guidelines to evaluate the outcomes of the AM L-DED process and obtain multilayer AlSi10Mg parts free from deleterious defects.