摘要
【目的】基于传热反问题研究方法,采用BP神经网络对地采暖地板蓄热性能进行反演计算,为地采暖地板蓄热性能分析提供理论和方法支撑。【方法】基于CFD软件构建检测腔体数值模型,模拟获取结构单一样本在不同初始温度(50～130℃范围内每间隔5℃设定一个模拟工况)下散热形成的温度场分布数据,并将初始温度为50、60、70、80、90、100、110、120和130℃的温度场分布数据作为神经网络模型的训练集,初始温度为55、65、75、85、95、105、115和125℃的温度场分布数据作为神经网络模型的测试集。【结果】经反复训练比较,获取较优的神经网络模型,其测试集的平均相对误差(MRE)=0.68%,最大相对误差(MAE)=19.51%,均方误差(MSE)=1.18%,拟合度(R^2)=0.98。基于该神经网络模型,选取白桦、水曲柳、西南桦、柞木4种典型实木地采暖地板样本进行蓄热性能反演计算,结果显示4种实木地采暖地板的蓄热性能表现为柞木>西南桦>水曲柳>白桦。【结论】经训练的神经网络模型可有效预测不同地采暖地板的蓄热性能,基于BP神经网络反演地采暖地板蓄热性能的方法可行。
【Objective】 Based on the study method for the inverse heat transfer problem, BP neural network technique is adopted for the inverse calculation of the heat storage of wood flooring with geothermal system. Therefore, provides theoretical and method ological support for the analysis of heat storage performance of wood floor.【Method】 Firstly, the numerical model of the testing cavity is established by CFD software. The temperature field data of a single structure sample under different initial temperature range of 50-130 ℃ are obtained by simulation （different simulation conditions are divided by interval of 5 ℃）. The data are divided into training set and testing set of neural network model. The data of the initial temperature of 50,60,70,80,90,100,110,120,130 ℃ are used as the training set, while the data of the initial temperature of 55,65,75,85,95,105,115,125 ℃ are used as the testing set.【Result】 After repeated training, a better neural network model is obtained. The average values of the calculation error and the fitting degree of the testing set are： mean relative error （MRE）=0.68%, maximum relative error （MAE）=19.51%, mean square error （MSE）=1.18%, fitting degree （ R 2）=0.98. Based on this model,Betula platyphylla,Fraxinus mandshurica, Betula alnoides and Quercus mongolica are selected as the four typical solid wood floor samples for the inversion calculation of the heat storage. The results showed that the heat storage performances of the four kinds of solid wood floor are as follows： Quercus mongolica 〉 Betula alnoides 〉 Fraxinus mandshurica 〉 Betula platyphylla. 【Conclusion】 It can be concluded that the well trained neural network model could effectively predict out the heat storage performance of different wood floor samples, verifying the feasibility of the method based on BP neural network technology to retrieve the thermal storage performance of the wood floor.
作者
周世玉
杜光月
曹正彬
刘晓平
周玉成
Zhou Shiyu;Du Guangyue;Cao Zhengbin;Liu Xiaoping;Zhou Yucheng(School of Thermal Engineering,Shandong Jianzhu University Jinan 250101;School of Information and Electrical Engineering,Shandong Jianzhu University Jinan 250101)
出处
《林业科学》
EI
CAS
CSCD
北大核心
2018年第11期14-19,共6页
Scientia Silvae Sinicae
基金
泰山学者优势特色学科人才团队(2015162)
山东建筑大学校内博士基金(X18006Z)
关键词
地采暖地板
蓄热性能
传热反问题
神经网络
wood flooring with geothermal system
heat storage performance
heat transfer inverse problem
neural network
