PocketFinderGNN: A manufacturing feature recognition software based on Graph Neural Networks (GNNs) using PyTorch Geometric and NetworkX

In this paper, we present a software tool called PocketFinderGNN for the recognition of a critical manufacturing feature named close pocket in 3D models. The close pocket is a pocket which is surrounded by material everywhere along its circumference. This feature makes a crucial role in the manufacturing industry, and its recognition is essential for the automation and optimization of machining processes. PocketFinderGNN converts the .stp file generated by CAD/CAM systems to a graph representation of the 3D model and utilizes a Graph Convolutional Network (GCN) to predict which nodes consist of the close pocket feature. The proposed tool was implemented using PyTorch Geometric and NetworkX frameworks. We trained our model on a dataset of 576 3D models obtained from the electromechanical industry and achieved an accuracy of 95% for the correct recognition of the faces forming the closed pocket feature. Our method outperforms state-of-the-art techniques and demonstrates its robustness to noise and perturbations in the input data. The results show that our approach is an effective tool for the recognition of complex manufacturing features in 3D models, which can significantly improve the efficiency and accuracy of manufacturing processes.