基于Kohonen聚类的特高拱坝变形分区

Deformation Partitioning of Super-high Arch Dam Based on Kohonen Clustering

目前,特高拱坝监测系统日益完善,变形监测点数量众多,但传统变形分析模型主要研究单测点监测序列,较难体现不同测点间的时空关联性.本文研究了Kohonen聚类算法的基本原理,结合大坝变形实测数据训练自组织神经网络,对上下游方向变形测点进行分类,依据测点实际位置实现特高拱坝变形分区,为整体变形监测提供辅助信息.某特高拱坝研究结果表明:Kohonen聚类方法能够有效探测变形数据空间聚集状态,描述特高拱坝实际变形规律,为特高拱坝变形分析提供新方法.

Nowadays,the monitoring system of super-high arch dam is becoming increasingly perfect with large numbers of deformation monitoring points.The traditional deformation analysis model of dams is mainly based on the time series of a single point,which is difficult to reflect the spatiotemporal correlation between different measuring points.This paper studies the principle of Kohonen clustering algorithm and carries out classification of points with deformation in the direction of upstream and downstream through selforganizing neural network trained by measured data.Finally,according to spatial location of these points,the deformation partitioning of the dam is realized to provide auxiliary information for overall deformation monitoring.The practical application results of a super-high arch dam show that Kohonen clustering is capable of effectively detecting spatial clustering of deformation points.It can describe the spatial distribution characteristics of dam deformation and provide new methods for the super-high arch dam deformation monitoring analysis.