.png){kind=icon}
返回
FluidMLP: A general method for learning Lagrangian fluid simulation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.simpat.2022.102623.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Learning simulators from data using deep learning methods is a popular research area in simulation modeling. We proposed a general data-driven Lagrangian fluid prediction model to solve the problem of low accuracy in long series prediction and improve the network's generalization. Although the model based on multi-layer shared perceptron has a simple structure and good generalization, it obtains a more effective prediction. The network directly takes the fluid motion formulations of physics as prior knowledge rather than traditional discretization methods for fluid simulations. Experimental results show that compared with the existing approaches, the proposed model can significantly improve the accuracy of fluid prediction without increasing the time consumption, which makes the high accuracy of prediction in continuous long time series, and can deal with fluid-solid interaction more reasonably. Moreover, the proposed model has better generalization and can deal with various scenes after training.
Keyword:
Simulation analysis
Point cloud
Deep learning
Fluid simulation
期刊
IF:
4.6
论文数:
2.6K
被引数:
4.8K
机构
暂无机构信息
.png){kind=icon}
没有更多内容