基于ELM模型的混凝土坝坝基渗压预测  

Seepage Pressure Prediction of Dam Foundation Based on ELM

作　　者：张蒙 喻和平[1] 陈玉江 ZHANG Meng;YU Heping;CHEN Yujiang(School of Hydraulic Engineering,Changsha University of Scienceand Technology,Changsha,Hunan 410004,China)

机构地区：[1]长沙理工大学水利工程学院,湖南长沙410004

出　　处：《水利与建筑工程学报》2019年第1期159-162,220,共5页Journal of Water Resources and Architectural Engineering

基　　金：湖南省大坝安全与病害防治工程技术研究中心开放基金资助项目(hndbgczx001)

摘　　要：为更精确的预测某混凝土坝坝基渗压变化趋势以保证大坝安全,利用混凝土重力坝渗压监测数据建立基于极限学习机(Extreme Learning Machine, ELM)的大坝基础渗压预测模型,并与传统逐步回归和传统BP神经网络方法进行对比。结果表明:ELM模型能够准确反映大坝坝基渗透系统的不确定性非线性关系,相比于逐步回归模型,ELM模型则可使h_(rmse)减幅至少有34.1%,误差区间降低有36.5%。ELM模型在精度和稳定性上均优于其余2种模型,其仿真曲线与测点渗压实测动态基本一致。该模型可作为渗透压力预测的推荐模型。In order to more accurately predict the seepage pressure variation trend of a concrete dam foundation, the seepage monitoring data of concrete gravity dams are used to establish the extreme learning machine(ELM) model. ELM model is compared with traditional stepwise regression and BP neural network. The results showed that the ELM model could accurately reflect the uncertainty non-linear relationship of dam foundation seepage system, compared with stepwise regression model, ELM model can reduce the root mean square error(hrmse) by at least 34.1% and error interval by 36.5%. Obviously, the ELM model is better than the other 2 models in accuracy and stability, and its simulation curve is basically consistent with the measured seepage pressure. The ELM could be used as a recommended model to predict seepage pressure.

关 键 词：混凝土坝 渗透压力 神经网络 极限学习机 预测 

分 类 号：TV642[水利工程—水利水电工程]