基于人工神经网络的土石坝震损病害的预测研究

The study of the prediction of seismic damages of earth and rock dams based on artificial neural network

本文介绍了基于L-M(Levenberg-Marquardt)算法的BP神经网络模型,利用模型对土石坝进行损害预测。选取243座震损土石坝水库为研究对象,确定土石坝裂缝、渗漏、滑坡为本次预测的指标。同时采集243座水库震损土石坝的裂缝、渗漏、滑坡等三大震损病害资料,通过专家意见划分其等级,作为实测值。经过神经网络模型反复的信息正向传递和误差反向传递的学习训练,使预测值与实测值的误差尽可能的满足精度要求。研究表明BP神经网络可用于土石坝震损病害预测。

This paper introduced BP neural network model which predicting seismic damage of earth and rock dams based on LM（Levenberg-Marquardt） algorithm.243 earth and rock dams which had seismic damages are selected for the study,the indicators for predicting the dam cracks,leaks and landslide are chosen.In the meantime the information of cracks,leaks and landslide about these dams was collected,and was divided and the grades were given by exports,which were used as actual measured values.Experiencing the neural network model simulation the information transmission forward and error transmission backward repeat and repeat,it made the error of the forecast values and actual measured values to meet the accuracy requirement.The research shows that BP neural network can be applied in the forecast of the earth and dam damaged by earthquake.