摘要
大坝安全预警模型可以理解为根据特定的映射关系由影响因素域到大坝性态效应量域的计算求解问题.对于多因素综合影响下的大坝系统,这种映射关系一般为非线性的.从机器学习的角度,本文应用粗集理论和SVM理论,研究了对上述关系的拟合.首先,利用粗集理论智能数据分析方法,对大坝安全监测信息进行预处理,抽取关键成分作为映射关系的输入,从而确定映射关系的初始拓扑结构.在此基础上,应用最小二乘支持向量机算法,以训练误差作为优化问题的约束条件,以置信范围值最小化作为优化目标,从大坝安全原型观测数据中学习归纳出大坝系统运行规律,从而实现对大坝安全预警模型的构建.实例分析表明,该模型能够有效的模拟和预测大坝工作性态与主要影响因素的关系.
The early-warning models of dam safety was a expression as the non-linear mapping relation between two field of dam behavior and their influence factors. Two tools of machine learn, rough sets theory and support vector machine （SVM） method, were used to build above relation. Data pretreatment on dam safety monitoring was implemented by the intelligent method analyzing data in rough sets theory. Main factors influencing dam safety were mined. The initial topological structure of SVM model was determined. An early-warning model was built with least squares support vector machines. Training error was the restriction condition of above optimization problem. The principles of structural risk minimization was used. Based on prototype observations of dam safety, the relation between dam behavior and their main influence factors was approximated and predicted. Experimental result indicated that the proposed model was an effective method for early-warning of dam safety.
出处
《应用基础与工程科学学报》
EI
CSCD
2009年第1期40-48,共9页
Journal of Basic Science and Engineering
基金
国家自然科学基金重点项目(50809025
50539110
50539010)
国家科技支撑计划项目(2008BAB29B03
20006BAC14B03)
关键词
大坝安全
预警模型
机器学习
SVM理论
dam safety
early-warning model
machine learning
support vector machine
