Adaptable multi-objective optimization framework: application to metal additive manufacturing

Abstract

This work presents a novel adaptable framework for multi-objective optimization (MOO) in metal additive manufacturing (AM). The framework offers significant advantages by departing from the traditional design of experiments (DoE) and embracing surrogate-based optimization techniques for enhanced efficiency. It accommodates a wide range of process variables such as laser power, scan speed, hatch distance, and optimization objectives like porosity and surface roughness (SR), leveraging Bayesian optimization for continuous improvement. High-fidelity surrogate models are ensured through the implementation of space-filling design and Gaussian process regression. Sensitivity analysis (SA) is employed to quantify the influence of input parameters, while an evolutionary algorithm drives the MOO process. The efficacy of the framework is demonstrated by applying it to optimize SR and porosity in a case study, achieving a significant reduction in SR and porosity levels using data from existing literature. The Gaussian process model achieves a commendable cross-validation R2 score of 0.79, indicating a strong correlation between the predicted and actual values with minimal relative mean errors. Furthermore, the SA highlights the dominant role of hatch spacing in SR prediction and the balanced contribution of laser speed and power on porosity control. This adaptable framework offers significant potential to surpass existing optimization approaches by enabling a more comprehensive optimization, contributing to notable advancements in AM technology.