Automatic Generation of Design History-Based CAD Model Dataset for Machine Learning

Commercial CAD (computer-aided design) programs generally support design history-based CAD modeling methods, which facilitate shape modification. However, CAD models often lack design history, making design changes difficult and resulting in significant industrial costs. To address this issue, several studies have focused on reconstructing the design history of CAD models using machine learning. However, this approach requires a dataset of CAD models that include design history, and currently, no suitable dataset exists for this purpose. Therefore, this study proposes an automatic generation method for a design history-based CAD model dataset. In the proposed method, features and their parameter values are randomly selected, with constraints applied to the parameters to generate realistic shapes. Finally, to validate the proposed method, it was implemented, and experiments demonstrated the successful generation of a CAD model dataset that includes design history.