Nonintrusive Local/Global Coupling With Local Deep Learning-Based Models for the Effective Simulation of Spotwelded Structures Under Impact

The article tackles the challenge of effective modeling and simulation of large mechanical structures exhibiting numerous local complex behaviors, as encountered with spot welds in automotive crash numerical analysis. To address this challenge, we propose a nonintrusive local/global coupling strategy, where the local model is a neural network-based reduced model, specifically a physics-guided neural network (PGANN). This multiscale strategy enables accurate modeling of complex localized behaviors while maintaining computational efficiency, without modifying the global solver. The proposed approach is validated through a series of structural examples, including full 3D industrial structures with multiple spot welds.