In the context of three-dimensional numerical control process reuse

the need for pre-defined local machining features are complex and dependent heavily on the expert knowledge or experience. This seriously affects the accuracy of similar process retrieval and subsequent reusability. To address these issues

a novel method based on the point cloud deep learning for local feature retrieval from 3D CAD models is proposed. The part's three-dimensional geometric model is transformed into a point cloud

and uniform down sampling reconstruction and normal vector calculation are applied to the part's point cloud model to achieve a three-dimensional point cloud representation of the part's geometry. Subsequently

based on the PointNet++ network architecture

the method extracts the local features of the part by reducing the number of network layers

applying max pooling operations

and removing fully connected layers. Then

a similarity retrieval method based on the correlation between the part's local features and the design requirements is used to retrieve the similar manufacturing processes. The present method is validated on a cavity machining feature

and the results demonstrate its effectiveness in extracting local part features and retrieving the similar manufacturing processes

supporting the rapid process design by subsequent process personnel.