神经网络法求解墙体传热系数关键输入变量的分析  

Analysis of key input arguments of wall heat transfer coefficient by neural network

作　　者：孙金金[1] 朱彤[1] 吴家正[1] 

机构地区：[1]同济大学机械工程学院,上海200092

出　　处：《新型建筑材料》2006年第12期61-64,共4页New Building Materials

摘　　要：在应用神经网络法求解墙体传热系数过程中,合理确定训练样本的输入变量是准确得到墙体传热系数的关键。分析了加热功率、内外侧环境温差、内外表面温差、内外表面热流密度、总厚度等对边界可测量量的影响,进而得出它们是确定墙体传热系数的关键输入变量。并对用它们作为输入变量的样本进行了训练和测试,得到了较满意的结果。11When using the neural network to solve heat transfer coefficient of the wall, confirming the proper input arguments for the training sample is the key step to obtain accurate heat transfer coefficient of the wall. The effect of some parameters such as heat power, environmental temperature difference of both sides, temperature difference in the surface of both sides, heat-flow density in the surface of both sides and total thickness of the material on variables which can be measured on the boundary are analyzed. And they are confirmed to be the key input parameters to determine heat transfer coefficient of the wall which are taken as input arguments for sample to be trained and tested, and satisfactory result is obtained.

关 键 词：导热反问题 神经网络 传热系数 数值模拟 

分 类 号：TU111.3[建筑科学—建筑理论]