人工神经网络和可靠度理论用于拱坝坝踵开裂风险的研究  

Study of risk of arch dam heel cracking based on artificial neural network and reliability theory

作　　者：徐福卫[1,2] 田斌[1] 陈海玉[2] 

机构地区：[1]三峡大学水利与环境学院,湖北宜昌443002 [2]湖北文理学院建筑工程学院,湖北襄阳441053

出　　处：《武汉大学学报（工学版）》2016年第5期787-790,800,共5页Engineering Journal of Wuhan University

基　　金：湖北省教育厅科学研究计划项目(编号:D20162603);国家自然科学基金项目(编号:50679036)

摘　　要：借用部分同级别坝高、高宽比接近拱坝的相关数据,利用人工神经网络建立的拱坝坝踵应力和上游水位之间的函数关系σ=f(H),计算出坝踵应力的均值、标准差.根据可靠度理论分析选取上游水位和坝踵应力为随机变量和根据最大拉应力破坏准则建立坝踵混凝土开裂的功能函数.最后,利用蒙特卡罗法进行坝踵破坏(失效)概率计算,得出拱坝坝踵的破坏(失效)概率.在拱坝可行性设计论证阶段,借用部分同级别坝高、高宽比接近拱坝的相关数据,根据拟建拱坝的上游设计水位可以初步估算坝踵区的拉应力值,采用此方法论证拱坝坝踵开裂失效概率是可行的.High arch dam is statically indeterminate structure. The failure probability of arch dam heel cracking has not been well solved and been widely focused on . In this paper, referencing the parameters and related random variables of some similar arch dams, using the BP network method to establish the nonlinear function between the tensile stress of dam heel and the water-head and to calculate the mean value and standard deviation of the tensile stress, the Monte Carlo method of structure reliability theory has been employed to study the risk probability of dam heel cracking. In the dam design stage,to reference part of the same level of high dams,according to the proposed design water level upstream of arch dam, the ten- sile stress of dam heel can be calculated preliminalily. The results show that the method employed to study the risk probability of arch dam heel cracking is feasible.

关 键 词：人工神经网络 可靠度理论 拱坝 坝踵开裂 风险 

分 类 号：TV321[水利工程—水工结构工程]