This study proposes a novel graph embedding technique for efficient representation of 3D stiffened panels. The approach uses a Graph Sampling and Aggregation (GraphSAGE) to predict stress distributions in these panels with varying geometries. The study also includes a comparison with a finite-element-vertex graph representation to demonstrate the effectiveness of the proposed method. The results show the potential of graph neural networks as robust reduced-order models for 3D structures.
Publication date: 22 Sep 2023
Project Page: https://arxiv.org/abs/2309.13022v1
Paper: https://arxiv.org/pdf/2309.13022
