摘要
基于港口湾大坝多期变形观测数据,采用M atlab语言、小波消噪及BP神经网络分别建立了基于时间序列和基于环境因素的大坝变形监测BP神经网络模型,并利用模型分别对大坝某点变形值进行预测。时间序列BP模型具有结构简单、学习速率快的特点;环境因素BP模型精度高,可有效反映变形因素,便于拟合预测复杂的测点变形,相对前一种模型能更好地揭示大坝变形规律。两种建模方法先应用小波分析对原始观测数据消噪,训练过程中采用附加动量法等改进BP算法,大大提高了BP神经网络的计算效率,克服了其易陷入局部极小的缺陷,取得了良好的拟合效果和预测精度。
Based on deformation monitoring data of Gangkouwan dam,using Matlab,wavelet denoise,BP neural network,we establish deformation monitoring BP neural network models on the basis of time sequence and environment factors respectively.And the two models are both applied to predict deformation of a monitoring point of the dam.The time sequence BP neural network model has simple structure and quick learning speed,however,the BP neural network model based on environmental factors can effectively reflect the deformation factors,which is more efficient for fit and prediction of complex deformation and can better reveal the deformation law of a dam.Before applying BP network to forecasting,the original data is de-noised through wavelet analysis method and the advanced BP algorithm such as additional momentum method is adopted in the training process,which significantly improve the accuracy and speediness of BP network predictor and avoid it falling local minimum.Therefore,satisfying fitting effects and forecasting precision are obtained.
出处
《人民长江》
北大核心
2011年第9期90-93,共4页
Yangtze River
基金
国家公益性科研专项(200911014)
安徽省国土资源项目(200908024)
关键词
小波消噪
BP神经网络
大坝变形
变形预测
wavelet denoise
BP neural network
dam deformation
deformation prediction
