改进的BP神经网络在碾压混凝土坝温度场反分析中的应用  

Application of Improved BP Neural Network in Back-Analysis of Temperature Field in RCC Dams

作　　者：张晓飞[1] 李守义[1] 陈尧隆[1] 杨杰[1] 余猛[1] 

机构地区：[1]西安理工大学水利水电学院,陕西西安710048

出　　处：《西安理工大学学报》2009年第1期95-99,共5页Journal of Xi'an University of Technology

基　　金：国家自然科学基金资助项目(50779051)

摘　　要：采用改进的BP神经网络,建立了碾压混凝土坝热学参数反馈分析模型;利用三维有限元浮动网格法正分析得到的样本去训练网络,再利用温度实测值对热学参数进行反分析,根据反演后的热学参数进行温度场计算。计算结果和实测结果较为接近,可满足工程实际要求,且该方法具有较好的稳定性和收敛性,表明改进的BP神经网络算法用于反演混凝土坝热学参数是可行的。In this article, improved BP neural network method is used to create back-analysis model of thermal parameters of RCC dam. The network is trained by the samples obtained by means of three-dimensional finite element relocating mesh method. And the temperature actual determined values are used to carry out the back-analysis of the thermal parameters. The calculation of temperature field is conducted in terms of the back-analytical thermal parameters. The calculated results are found to be close to the actually-determined results, which can satisfy the requirements by the engineering practice. Also, this method is of better stability and convergency, whereby indicating that the improved BP neural network algorithm applied to thermal parameters of back-analysis of concrete dam is feasible.

关 键 词：BP神经网络 三维有限元浮动网格法 碾压混凝土坝 反分析 热学参数 

分 类 号：TV315[水利工程—水工结构工程]