基于径向基网络模型的面板堆石坝坝顶沉降量预测

Predicting crest settlement of concrete face rockfill dams using RBF network

随着国家在水利基础建设方面的大力投入以及面板堆石坝自身的优越性,近几年该坝型逐步向超高坝型发展,同时坝体变形的预测也面临着诸多困难。本文提出了一种基于径向基(RBF)网络的面板堆石坝的变形预测模型。该模型充分利用了径向基网络的非线性映射能力,利用收集的历史样本信息,即可预测出面板堆石坝沉降。以水布垭面板堆石坝为例,预测得到的竣工期和满蓄5年后的沉降位移分别为2.156m和2.491m,与实测位移基本一致,相对误差分别为0.748%和0.400%。结果表明预测位移在设计允许范围之内,RBF网络模型具有建模速度快、预测精度高的特点。

With the vigorous investment getting involved in the field of water conservancy infrastructure construction in China, and the superiority of concrete face rockfill dams (CFRD),the ultra-high CFRD dams are constructed gradually in recent years.At the same time,the prediction of the crest settlement of CFRD is confronted with many difficulties.This paper pro-posed a model for deformation prediction of CFRD based on RBF (RBF)networks.The model makes full use of the nonlinear mapping ability of the RBF networks,using the collected information about the history samples,to predict the settlement of CFRDs.Take Shuibuya CFRD as an example,the predicted settlement values by RBF during completion period and storage af-ter 5 years of reservoir fully filled are 2.1 56m and 2.491m,respectively.The predicted results are in accordance with the meas-ured ones,and the relative error is 0.748% and 0.400%,respectively.The results demonstrate that the dam deformation is within a reasonable range,and RBF network model has the characteristics of high speed and accuracy of modeling.