A Decision Tree approach for an early evaluation of 3D models in Design for Additive Manufacturing

Abstract

Metal Additive Manufacturing is an emergent production process that can realize geometries that are difficult to realize with traditional manufacturing techniques. The design rules and guidelines for Additive Manufacturing are different from the traditional approaches. One of the issues of Additive Manufacturing is the evaluation of the printability of the CAD model to be realized and the results in terms of residual stress and deformation. In the literature, there is a lack of tools and methods to rapidly evaluate the printability of the CAD models and predict the results in terms of residual stress and deformation. This paper proposes a Machine Learning-based method to confirm or not the printability of a 3D CAD model in the early design phase. This evaluation could reduce the errors during the printing phase. A Decision Tree classifier has been trained with virtual analysis. The dataset has been produced with CAD models, generated by a parametric approach, and numerical simulations used to evaluate the 3D printing output. A Knowledge-Based tool defines the list of parameters to be extracted from each CAD model. During the use of the proposed decision tool, the parameters are extracted from the CAD model and analyzed within the Decision Tree model.