A DeepONet Surrogate Model for Stress Fields in Parametric Bracket Families — dataset and trained model
收藏资源简介：
Dataset and trained model accompanying the MBA monograph "A DeepONet Surrogate Model for Stress Fields in Parametric Bracket Families: The Cost of Geometric Generalization" (ICMC/USP, 2026). 55,803 finite element solutions (quadratic tetrahedral elements, verified against cubic elements) of six parametric bracket families, split into one archive per family: L 9,999, Z 10,000, U 9,999, T 8,751, O 8,486 and G 8,568 cases. A compact version without mesh connectivity, sufficient for training, is also provided. MANIFEST_sha256.txt lists the SHA-256 of every case. The trained model, its training log and the validation split (8,370 cases) are included. Field descriptions and code: see the linked GitHub repository. File names: tcc (trabalho de conclusão de curso) refers to the monograph, and treino means training. Field names inside the .npz files are also in Portuguese; the GitHub repository's README has a glossary.
