D-ECOmposer: Sustainable part decomposition for additive manufacturing using machine learning based life cycle assessment

Abstract

Additive Manufacturing (AM) has garnered significant attention due to its potential for sustainable production. To further enhance this potential, Design for Additive Manufacturing (DfAM) methodologies are frequently employed. However, traditional design approaches often fall short in addressing the inherent limitations of AM, such as build size constraints, extended lead time, and the necessity for support structure. However, due to these limitations, part decomposition (PD) has recently gained prominence as a viable solution. While the benefits of PD might be less pronounced if a model can be produced in its entirety on a single AM device, this study assumes scenarios where the model is too large for the build space of the AM device, making decomposition necessary. This study proposes a grid-based PD method that utilizes machine learning-based Life Cycle Assessment (LCA) to minimize environmental impact. The experimental data in this study were collected and analyzed based on the FDM(Fused deposition modeling) process. Initially, a predictive model is developed to quickly and accurately estimate the carbon footprint of a design candidate based on the geometric characteristics of a 3D model. This predictive model is subsequently employed as the objective function in the optimization of PD using a genetic algorithm (GA). To validate the efficacy of the proposed method, experiments were conducted on four test models using FDM. While this study focuses on FDM, the proposed methodology has potential applicability to other AM processes. The experimental results clearly demonstrate that the proposed method outperforms traditional empirical approaches in reducing the carbon footprint.