基于卷积神经网络模型数值求解双曲型偏微分方程的研究

人工神经网络近年来得到了快速发展,将此方法应用于数值求解偏微分方程是学者们关注的热点问题.相比于传统方法其具有应用范围广泛(即同一种模型可用于求解多种类型方程)、网格剖分条件要求低等优势,并且能够利用训练好的模型直接计算区域中任意点的数值.该文基于卷积神经网络模型,对传统有限体积法格式中的权重系数进行优化,以得到在粗粒...

Study on Numerical Solutions to Hyperbolic Partial Differential Equations Based on the Convolutional Neural Network Model

In recent years, artificial neural networks developed rapidly. Application of this method to partial differential equations became a new idea for exploring numerical solutions to differential equations. Compared with the traditional methods, it has some advantages, such as a wide range of applications (i.e. the same model can be used to solve multiple types of equations) and low meshing requirements. In addition, the trained model can be directly used to calculate the numerical solution at any point in the computation domain. The weight coefficients in the traditional finite volume method were optimized based on the convolutional neural network model to get a new numerical scheme with high resolution results on the coarse grid. The proposed model helps solve the Burgers and level set equations efficiently and stably with high accuracy.