Automatic Generation of Product Manufacturing Information from 3D CAD Models Using Deep Learning

Product Manufacturing Information (PMI) is an essential component of Model-Based Definition (MBD), as it communicates manufacturing intent such as dimensions, tolerances, and annotations directly through 3D CAD models. In current industrial practice, PMI is mainly created manually by designers, which is time-consuming, experience-dependent, and prone to inconsistency. To address these limitations, this paper presents a deep learning–based approach for the automatic generation of PMI from 3D CAD models. The proposed method extracts boundary representation (B-Rep) information from STEP files and converts the geometric and topological data into a graph-based representation. A deep learning model is trained to learn the relationship between CAD geometry and manufacturing semantics. Based on this learned knowledge, the system automatically generates relevant PMI, which is then mapped back onto the CAD model. Experimental evaluation on mechanical components demonstrates that the proposed approach can successfully generate PMI for common manufacturing features, significantly reducing manual effort. This work supports intelligent CAD automation and advances AI-driven MBD workflows.