The solution of partial differential equation of heat conduction is an important calculation in industrial applications.To solve the problem that the traditional forward calculation of heat conduction equation takes too long,this paper proposes a solver of heat conduction equation inverse problem based on the new network structure FluxNet.The data set of temperature field distribution images and corresponding heat flux images are obtained by forward operation of heat conduction equation.The data set is trained based on deep learning.It establishes a heat flux network structure FluxNet model with convolution layer and deconvolution layer.The prediction heat flux images of the solver and the reality data are tested and verified.Experimental re-sults show that the structure similarity of heat flux predicted by the solver reached more than 90％,which meets the require-ments of industrial applications.
关键词
深度学习/求解器预测模型/热传导偏微分方程/热流密度/温度场分布
Key words
deep learning/solver prediction model/partial differential equation of heat conduction/heat flux/temperature field distribution
维普
万方数据