• A novel GNN-based surrogate model predicts material deformation with high accuracy. • Recurrent structures capture temporal dynamics in large material deformations. • Graph U-Net architecture efficiently processes large and complex geometries. • New contact representation method improves efficiency in large contact interactions. • The model facilitates manufacturability assessment in material forming. In recent years, various artificial intelligence-based surrogate models have been proposed to provide rapid manufacturability predictions of material forming processes. However, traditional AI-based surrogate models, typically built with scalar or image-based neural networks, are limited in their ability to capture complex 3D spatial relationships and to operate in a permutation-invariant manner. To overcome these issues, emerging graph-based surrogate models are developed using graph neural networks. This study developed a new graph neural network surrogate model named Recurrent U Net-based Graph Neural Network (RUGNN). The RUGNN model can achieve accurate predictions of sheet material deformation fields across multiple forming timesteps. The RUGNN model incorporates Gated Recurrent Units (GRUs) to model temporal dynamics and a U-Net inspired graph-based downsample/upsample mechanism to handle spatial long-range dependencies. A novel ’node-to-surface’ contact representation method was proposed, offering significant improvements in computational efficiency for large-scale contact interactions. The RUGNN model was validated using a cold forming case study and a more complex hot forming case study using aluminium alloys. Results demonstrate that the RUGNN model provides accurate deformation predictions closely matching ground truth FE simulations and outperforming several baseline GNN architectures. Model tuning was also performed to identify suitable hyperparameters, training strategies, and input feature representations. These results demonstrate that RUGNN is a reliable approach to support sheet material forming design by enabling accurate manufacturability predictions.

Surrogate models with structured data representations, mainly images and graphs, have been widely investigated in various domains, including automotive manufacturing. Despite advances, existing approaches still face significant challenges in terms of accuracy, efficiency, and generalisability. To address these challenges, a promising direction is to combine the advantages of both images and graphs. This study proposes a graph-based surrogate model, which has a multi-level architecture with enhanced graph convolutional operations using image-inspired spatial edge weights. To evaluate its performance, three additional graph-based surrogate models are developed for comparison, each differing in the formulation of spatial edge weights. All four models are assessed on a real-world hot-stamping B-pillar case study, which involves variations in blank shapes under multiple parameterisations and post-stamped thickness distributions exhibiting complicated local patterns. The proposed architecture significantly outperforms the three comparison models, achieving high accuracy with a relatively low computational burden during training and deployment. Furthermore, it demonstrates strong robustness in hyperparameter calibration and shows the potential for generalisation to other manufacturability-related real-time simulation problems. This study presents an effective methodology for future surrogate model development by integrating the advantages of different structured data representations.