基于EMD分解和ABC-RBF模型的混凝土平板坝变形监测模型  

Monitoring Model of Concrete Slab Dam Deformation Based on EMD and ABC-RBF Neural Network Model

作　　者：赵鲲鹏[1] 

机构地区：[1]河海大学水利水电学院,南京210098

出　　处：《三峡大学学报（自然科学版）》2015年第4期29-33,共5页Journal of China Three Gorges University：Natural Sciences

基　　金：淮安市水利院士工作站资助;国家自然科学基金项目(51279052;51209077);江苏省杰出青年基金项目(BK20140039;BK2012036);江苏省"333高层次人才培养工程"(2017-B08037)

摘　　要：针对混凝土平板坝水平位移监测序列呈非线性以及周期性变化的特点,采用经验模态分解(EMD)的方法对混凝土平板坝水平位移监测进行分解,并采用计算信噪比最大值的方法对信号进行去噪,得到最优的去噪效果.面板坝的坝顶上下游方向水平位移主要受上下游水位、环境温度以及时效的影响,据此建立普通RBF和ABC-RBF神经网络模型,并对混凝土平板坝上下游方向水平位移进行预测,对比两种模型预测样本的均方根误差(RMSE)、平均绝对误差(MAE)以及负相关系数(R).结果表明,ABC-RBF神经网络模型预测效果较好,误差比较小,该方法为分析混凝土平板坝工作性态和安全监测及预测提供一种新的方法.In view of the nonlinear and cyclical change characteristics of the horizontal displacement observation data of concrete slab dam,the method of empirical mode decomposition（EMD）is used to decompose the horizontal displacement of concrete slab dam.And the method of calculating the maximum SNR is adopted to denoise;and then the most optimal denoising effect is got.Concrete slab dam horizontal displacements of dam crest upstream and downstream direction are mainly affected by upstream and downstream water level,environment temperature and time effect.Then the general RBF and ABC-RBF neural network model is established to forecast the horizontal displacement of concrete slab dam.The comparison between two kinds of models to predict samples of root mean square error（RMSE）,mean absolute error（MAE）and negative correlation coefficient（MRE）is made.The results show that the error of predicting by using ABC-RBF neural network model is better,so as to provide a new method for analyzing the workability of concrete slab dam and its safety monitoring and forecasting.

关 键 词：混凝土平板坝 水平位移 经验模态分解 ABC-RBF神经网络 

分 类 号：TV642[水利工程—水利水电工程] TV698.1