神经网络模型在面板坝堆石体施工期沉降变形预测中的应用

在分析面板坝堆石体施工期坝体沉降影响因素的基础上,将影响沉降的主要因素作为网络输入参数,以测点沉降量作为网络的输出,建立了神经网络模型.以水布垭面板坝堆石体为例,将模型预测值与实测结果进行了对比,结果表明,预测值与实测结果比较接近,该神经网络能很好地反映面板坝堆石体施工期沉降变形与其影响因素之间的非线性映射关系,预测结果可作为后期填筑反馈设计的依据,同时可通过网络输入参数的调整检验某一因素对坝体沉降的影响程度.

Application of neural network model to prediction of settlement deformation of rockfill body of CFRD during construction period

Based on analysis of the influencing factors on the settlement of concrete face rockfill dams(CFRD) during the construction period,a BP neural network model for settlement prediction was developed.In the model,the main influencing factors on settlement were taken as the input parameters of the network,and the settlement at measurement points was taken as the output of the model.Case study on Shuibuya CFRD shows that the predicted result is much close to the measured data,and that the neural network can well reflect the nonlinear relationship between the settlement deformation of CFRD and its influencing factors.The predicted result can provide a basis for feedback design of concrete filling at the later stage,and the influence analysis of each factor on the settlement could be made by adjustment of input parameters of the network.