A visual-cognition-inspired model for machining feature recognition

Automatic feature recognition technology is the key support of intelligent design and manufacturing. Using artificial neural networks (ANNs) to identify machining features is a significant interdisciplinary research direction. Although the ANNs have the ability to learn and generalize, and process faster, they can only process numerical input and perform arithmetic operations, not logical operations, thus restricting the application of ANNs in the field of CAD machining feature recognition. This paper establishes a new Visual-cognition-inspired Model (VCIM) for machining feature recognition by imitating the visual cognition process of the human brain and related neural mechanisms. The VCIM has a structure closer to the cerebral cortex than ANNs, uses a 3D CAD model as a direct input, and has three new activation functions that can perform logical operations. The VCIM have been tested and verified to identify four different types of machining features, and new features can be recognized by changing the activation function definition.