Archives by date

You are browsing the site archives by date.

Graph neural network-based stress field prediction in stiffened panels

Our GNN-based approach for stress prediction effectively captures entire stress field in stiffened panels under various loadings and with wide geometrical variations. For more, please see: Y. Cai, J. Jelovica (2024) Efficient graph representation in graph neural networks for stress predictions in stiffened panels. Thin-Walled Structures, Vol. 203, 112157.

KATO: Neural-reparameterized topology optimization framework

Our novel neural-reparameterization for topology optimization generates stress-optimal structures by training a convolutional KAN network. The framework is completely data-free, outperforming MMA and CNN models. For more, see: S. Yan, J. Jelovica (2025) KATO: Neural-reparameterized topology optimization using convolutional Kolmogorov-Arnold network for stress minimization, International Journal for Numerical Methods in Engineering, Vol. 126:e70034,