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.
| Authors | Betkier, I.; Oszczypała, M.; Pobożniak, J.; Sobieski, S.; Betkier, P. |
|---|---|
| Year | 2023 |
| Venue | SoftwareX 23, 101466 |
| DOI | 10.1016/j.softx.2023.101466 |
| Source Database | Backtracking |
| Bridge-to-GNN | Category A |
| GNN Architecture | GCN (heuristic — verify) |
| Graph Encoding | Not yet extracted from PDF |
| AEC Task | Not yet extracted from PDF |
| Cohort | Consolidation (2023) |
| Dataset Size | 64084 nodes; 600 models (heuristic — verify) |
|---|---|
| Implementation Framework | PyTorch Geometric (PyG), PyTorch, Keras, NetworkX (heuristic — verify) |
| Key Hyperparameters | 3 layers; 2000 epochs; lr 0.001; optimizer=ADAM optimizer (heuristic — verify) |
| Primary Metric | Accuracy (heuristic — verify) |
| Primary Value | 95% (heuristic — verify) |
| Key Finding | software tool called PocketFinderGNN for the recognition of a critical manufacturing feature named close pocket in 3D models (heuristic — verify) |
| Quality Assessment | 3/6 · rigor: Medium · Code/Data: ✓Baseline: ✓Reproducible: ✓ (heuristic — verify) |
| GNN architectures detected | Knowledge Graph Embedding (TransE/RotatE/DistMult) · GCN |
|---|---|
| Frameworks / libraries | PyTorch Geometric (PyG) · PyTorch · Keras · NetworkX |
| Code repositories found in text | https://github.com/ElsevierSoftwareX/SOFTX-D-23-00227 · https://github.com/betanddontcare/PocketFinderGNN · https://github.com/betanddontcare/PocketFinderGNN/blob/main/ |
| Hyperparameters (regex-detected) | layers=3 · epochs=2000 · learning_rate=0.001 · optimizer=ADAM optimizer |
| Dataset stats found | 600 models · 576 models · 64084 nodes · 156699 edges · 137 nodes · 338 edges |
| Reported metrics + values | Accuracy: 95% · Accuracy: 99% |
| Quality (heuristic, 6-flag) | 3/6 · rigor: Medium Code/Data: ✓Benchmark: —Baseline: ✓CV/Split: —Ablation: —Reproducible: ✓ |
Part of the GML/GNN in AEC Systematic Review (PRISMA 2020) — 112 papers, 2020–2026