Adaptive Recognition Of Machining Features In Sheet Metal Parts Based On A Graph Class-incremental Learning Strategy

This article discusses the development of an efficient and memory-saving Boundary Representation (B-Rep) graph model and Sheet-metalNet, a Graph Neural Network for machining feature identification. It also introduces a specialized dataset for sheet metal parts, showcasing the potential of deep learning in Automatic Feature Recognition (AFR) in manufacturing. The use of graph-based methods in rule-based AFR approaches is also explored, highlighting its limitations in dealing with intersecting features.