BP网络及多元线性回归在长江堤防监测资料分析中的应用  

Application of BP Nerual Networks and Multiple Regression in Processing Data of Safety Monitoring in Yangtze River Embankment

作　　者：李刚[1] 廖勇龙[1] 徐卫军[1] 

机构地区：[1]长江科学院工程安全与病害防治研究所,湖北武汉430010

出　　处：《长江科学院院报》2004年第5期56-58,62,共4页Journal of Changjiang River Scientific Research Institute

摘　　要：为比较BP网络及多元线性回归方法在工程中的应用效果,采用这2种方法对长江重要堤防隐蔽工程典型监测断面的资料进行了分析。结果表明:2层的BP网络以及使用线性函数的3层BP网络与多元回归方法的预测效果基本一致,能够满足实际工程需要;使用"S"型活跃函数的3层BP网络预测精度较差。In order to compare their application efficiency two kinds of BP neural networks and multiple regression methods are used in processing data of safety monitoring at a typical cross、section in the Yangtze River embankment. The results show that the predicted values by two、layers BP neural networks and by three、layer one with linear element as hidden layer are basically in agreement with those found by multiple regression method. All methods can meet the practical engineering requirements, but less precision obtained from 3、layers BP neural network adopting S、type activity function.

关 键 词：BP网络 多元回归 长江堤防 渗流 安全监测 

分 类 号：TV871[水利工程—水利水电工程]