基于人工神经网络的重力坝安全可靠度分析

Gravity dam safety reliability analysis based on artificial neural network

为了提高重力坝安全分析的可靠性,首先将影响重力坝变形的主要因素水位、气温视为随机变量,求出其概率分布特征,然后应用具有强大非线性映射能力的人工神经网络模拟了坝体变形,结合重力坝变形统计模型建立了坝体变形隐患和异常功能函数,应用可靠性理论建立了基于人工神经网络的重力坝变形可靠度及敏感性计算公式.工程实例分析结果表明:由于气温变异性较大,坝顶变形隐患概率和异常概率受气温影响较水位的影响大;坝顶变形可靠度随着库水位的增加而降低,气温对其敏感性影响较大,特别对10,5 d气温测值均值较为敏感.

In order to improve the reliability of safety analysis for gravity dams,the main factors influencing gravity dam deformation,such as water level and temperature,were considered as random variables,and their probability distributions were calculated.Then the artificial neural network（ANN）,with strong non-linear mapping ability,was employed to simulate dam deformation and to establish performance function of the hidden trouble deformation and abnormal deformation in the dam body combining with the statistical model of dam deformation.Finally,the reliability theory was used to analyze dam safety reliability and sensitivity.The analysis results of a case study show that temperature has a greater effect on the probability of dam hidden trouble deformation and abnormal deformation than reservoir water level,because of great variability of temperature.The reliability index of the dam decreases with the increase of reservoir water level,and is greatly influenced by temperature,especially the average temperatures in the first five days and ten days.