Machining feature recognition method for mechanical parts based on improved MeshCNN

In order to solve the problem of low efficiency and low intelligence of feature recognition in traditional computer-aided process planning (CAPP) system, this paper proposed a novel machining feature recognition method for mechanical parts based on Mesh-Faster Region-based CNN (RCNN) by combining original MeshCNN with Faster RCNN. The method obtained the optimal neural network model by taking the customized processing dataset as the input of the neural network. Then MBD technology was used to label the machining model, and the characteristics to be processed were obtained by PMI information annotation, which were transformed into triangular mesh data. On this basis, combined with the algorithm of triangular mesh data processing, the processed machining feature data is imported into the optimal neural network model to complete the feature recognition process. Finally, the feasibility and effectiveness of the proposed method are verified by taking the key parts of the diesel marine engine as an example.