摘要
大坝变形是同一时刻内外多种荷载综合作用的结果,挖掘位移监测数据潜在规律和发展趋势是大坝变形预测诊断的关键技术,但常规GNSS+棱镜、自动化监测系统等进行观测存在较大非线性误差。为实现非线性、非平稳序列大坝变形数据的平稳化拟合处理,基于大坝变形位移关联性函数,构建了大坝变形预测BP神经网络模型。BP预测模型主要根据水压、温度和时效因子的特点,经实测数据的自适应学习训练获得能真实反映坝体变形规律及趋势的竖向位移预测数据,可为大坝变形安全预测与分析提供详实准确的数据支撑和技术保障。
Dam deformation is a result of the combined action of multiple loads at the same time.It is a key technology for prediction and diagnosis of deformation by exploring the potential laws and development trends of displacement monitoring data.However,the existing normal analysis methods such as GNSS+prism,and automatic monitoring system have large non-linear error.In order to realize the smooth fitting process of non-linear and non-stationary dam deformation data,the BP neural network model for dam deformation prediction is constructed based on the correlation function of dam deformation and displacement.Mainly based on the characteristics of water pressure,temperature,and time-depended factors,the BP prediction model can obtain vertical displacement prediction data which can truly show the deformation law and trend of dam body through self-adaptive learning and training of measured data.The model proposed would provide detailed and accurate data and technical support for safety prediction and analysis of dam deformation.
作者
林智艳
LIN Zhiyan(Water Authority of Huanren Manchu Autonomous County,Liaoning Province,Huanren 117200,China)
出处
《人民珠江》
2020年第6期74-78,共5页
Pearl River
关键词
BP神经网络
大坝变形预测
竖向位移
时效因子
BP neural network
dam deformation prediction
vertical displacement
time-depended factors
