基于深度学习的结构位移场预测方法

打开文本图片集

中图分类号：TP181；TP391.92 文献标志码：B

Structural displacement field prediction method based on deep learning

HU Yezhi' ZHANG Yaqi2 LU Changhong （1.DNE Technology Co.，Ltd.，Shanghai 200030，China；2.Allbright Lawofices（Hefei），Hefei 230001，China； 3.Hefei Hanwang Software Technology Co.，Ltd.，Hefei 23OOO1，China）

Abstract：Based on the basic equations of finite element method，the feature extension layer is embedded into the deep learning model，and a training set generator is developed using Abaqus software interface to achieve single model prediction of the full displacement component of the structure. Using the Keras API under the TensorFlow framework to train a deep learning model for spatial thin shell structures，a quantitative analysis of the prediction performance is conducted. The results show that： the computational efficiency of the deep learning model is significantly improved compared to the simulation model，and the prediction of the maximum displacement and distribution pattern is basically consistent with the simulation results，however there is an increase in error at the O displacement boundary.

Key words： deep learning；model training；displacement field；prediction； spatial thin shell structure

0 引言

将现实世界的几何形态映射到虚拟世界并建立系统进行分析管理，是目前数字孪生研究的主流方向，但这尚未跳出BIM的工作范畴。（剩余8494字）