cmudrc/SFEM

SFEM数据集是一个用于训练神经网络代理模型的数据集，包含大约16,000个CAD几何体及其在不同加载条件下的有限元分析结果。这些结果使得神经网络能够以前所未有的速度预测应力分布，比传统的有限元方法快几个数量级。数据集遵循PyTorch Geometric的规范，包含预处理和转换后的数据，以及原始的训练和验证数据文件。数据集的目的是用于工程设计的神经网络训练、有限元分析的代理模型开发、计算力学和机器学习研究，以及工程仿真的教育用途。

The SFEM Dataset is a collection designed for training neural network surrogate models, containing approximately 16,000 CAD geometries with finite element analysis results under various loading conditions. These results enable neural networks to predict stress distributions at orders of magnitude faster than traditional FEM simulations. The dataset conforms to PyTorch Geometric standards, including preprocessed and transformed data, as well as archived raw training and validation files. The dataset is intended for use in neural network training for engineering design applications, developing surrogate models for finite element analysis, research in computational mechanics and machine learning, and educational purposes in engineering simulation.