A machining feature recognition approach based on hierarchical neural network for multi-feature point cloud models

Most mechanical part models in current industrial manufacturing are composed of multiple different machining features. However, the traditional rule-based feature recognition methods are only suitable for analyzing simple and specific features. Although the existing methods based on deep learning are no longer limited to recognizing particular features, they cannot recognize complex overlapping features. To solve the above issues, this paper proposed a machining feature recognition approach based on the hierarchical neural network to recognize the multiple features on point cloud models. Firstly, the 3D models were converted into point cloud samples to construct the dataset, so that the approach could be applied to different 3D model formats. Then a hierarchical neural network called PointNet++ for single feature recognition was constructed. For the multi-feature point cloud models, a feature segmentation method was proposed to divide a complex multi-feature model into single feature models for recognition. Finally, the approach was evaluated on the created test data sets. The test results show that the overlapping machining feature on point cloud models can be accurately recognized with low computational cost.