混凝土坝变形的空间聚类及预测模型优选  

Spatial Clustering and Prediction Model Selection of Concrete Dam Deformation

作　　者：詹明强 陈波[1,2] 刘庭赫 陈伟楠 ZHAN Ming-qiang;CHEN Bo;LIU Ting-he;CHEN Wei-nan(College of Water Conservancy and Hydropower Engincering,Hohai University,Nanjing 210098,China;State Key Laboratory of Hydrology Water Resources and Hydraulic Enigineering,Hohai University,Nanjing 210098,China;China Water Northeastern Investigaion,Design&Research Co.,Ltd.,Changchun 130021,China)

机构地区：[1]河海大学水利水电学院,江苏南京210098 [2]河海大学水文水资源与水利工程科学国家重点实验室,江苏南京210098 [3]中水东北勘测设计研究有限责任公司,吉林长春130021

出　　处：《水电能源科学》2022年第8期96-100,共5页Water Resources and Power

基　　金：国家重点研发计划(2018YFC0407104);国家自然科学基金项目(52079049,51609074,51739003);中央高校基本科研业务费(B200202160)。

摘　　要：目前,混凝土坝变形预测方法主要针对单测点,未考虑各测点在空间维度上的关联,且在多测点情况下获取最优模型效率较低。对此,提出一种混凝土坝变形的区域化预测模型。首先构建了变形综合相似性指标作为FCM聚类的目标函数,实现对混凝土坝的变形分区;接着对不同区域关键性测点采用几种热门的机器学习预测模型进行训练,在优先考虑精度的前提下结合模型运行速度等指标优选出一种代表性模型,并外延应用至整个分区内的其他测点。工程实例表明,该方法提高了模型的外延能力,在保证预测效率的同时实现了高精度预测。The prediction method of concrete dam deformation is mainly aimed at single measuring point without considering the correlation of each measuring point in spatial dimension,and it is inefficient to obtain the optimal model in the case of multiple measuring points.A zoning prediction model of concrete dam was proposed.Firstly,the comprehensive deformation similarity index was constructed as the objective function of FCM clustering to achieve the deformation zoning of concrete dam.Then,several popular machine learning prediction models were used for training of the key measurement points in different regions.Under the premise of prioritizing accuracy,a representative model was selected and applied to other measurement points in the whole region by denotation.An engineering example shows that this method can improve the epitaxial capacity of the model,ensure the prediction efficiency and achieve high precision prediction.

关 键 词：混凝土坝 变形 空间聚类 机器学习 预测模型优选 

分 类 号：TV698.1[水利工程—水利水电工程]