摘要
针对目前常规组合模型在残差序列信息挖掘上的不足,借助小波分析、逐步回归分析和粒子群算法对运行期沉降进行分层次预测,充分提取沉降观测中的非线性和不确定性信息,即利用小波去噪对观测数据进行处理,结合逐步回归模型(SR)对沉降变形的整体趋势层进行预测,并通过粒子群-极限学习机(PSO-ELM)对观测数据中的残差层进行第二层建模预测,采用分层预测后的叠加值来反映测点的沉降预测值,据此构建小波SR-PSO-ELM模型.结合某已建工程实例对本模型效果进行验证,对比其它模型的预测结果表明,提出的组合模型有着更高的精度和更优的泛化能力,能对面板堆石坝运行期沉降变形进行有效预测.
Aiming at the current shortcomings of conventional combined models in the residual sequence information mining,this paper uses wavelet analysis,stepwise regression analysis and particle swarm algorithm to predict the settlement at different levels during operation,and fully extract the nonlinear and uncertain information in the settlement observation.That is,wavelet denoising is used to process the observed data with combination of the stepwise regression model(SR)to predict the overall trend layer of settlement deformation and particle swarm-extreme learning machine algorithm to model the residual layer in the observation data.The summation after hierarchical prediction is used to reflect the settlement prediction value of the measuring point.Based on previous processes,the wavelet SR-PSO-ELM prediction model is constructed and applied to predict a rockfill dam.The prediction result is compared with other methods.The results show that the combined prediction model proposed in this paper is more accurate and robust.It can effectively predict the settlement of CFRD during operation.
作者
燕乔
高名杨
梁明浩
王硕
YAN Qiao;GAO Mingyang;LIANG Minghao;WANG Shuo(College of Hydraulic&Environmental Engineering,China Three Gorges Univ.,Yichang 443002,China)
出处
《三峡大学学报（自然科学版）》
CAS
2021年第5期1-5,共5页
Journal of China Three Gorges University：Natural Sciences
基金
国家大坝安全工程技术研究中心开放基金项目(NDSKFJJ1202)。
关键词
面板堆石坝
沉降预测
小波变换
极限学习机
粒子群优化算法
concrete faced rockfill dam(CFRD)
settlement prediction
wavelet transform
extreme learning machine
particle swarm optimization algorithm
