基于面板数据分析方法的混凝土拱坝变形数据时空聚类模型

变形数据可以直接表征混凝土拱坝安全状况,但传统变形分析一般仅针对单个监测点,不同监测点之间变形的相似和关联性质仍有待挖掘。基于时空数据挖掘领域的聚类方法,分析混凝土拱坝变形序列的变化过程,提取变形序列的相似性特征;提出混凝土拱坝变形数据不同时间截面、不同测点变形序列的绝对距离、增量距离、增速距离3种相似性指标及相应的综合距离指标,由此定量分析变形时间截面和变形空间测点的相似程度;利用Ward联结聚类方法,对混凝土坝变形时段及相应变形区域进行划分;在此基础上,建立基于面板数据分析方法的高混凝土坝变形测点聚类分析模型,并结合工程实例,验证时空聚类模型的聚类效果。结果表明所提出的聚类相似性指标合理,时空聚类模型便捷、有效,可用于研究大坝相时段变形状态。

Spatial and Temporal Clustering Model of Concrete Arch Dam Deformation Data Based on Panel Data Analysis Method

Deformation data directly characterizes the safety condition of concrete arch dam. However, traditional deformation analysis targets only a single monitoring point. The similarity and association of deformation between different monitoring points still need to be excavated. In this research, a clustering analysis model for the deformation of high concrete arch dam is established based on panel data analysis method. The deformation sequences of concrete arch dam are analyzed at first, and the similarities of temporal and spatial deformation sequences are extracted based on the clustering method in space-time data mining. Three similarity indicators, namely, absolute distance, incremental distance, and growth distance of the deformation sequences at different time sections and different positions as well as their corresponding comprehensive distance indicators are proposed to quantify the similarity between temporal and spatial sequences. Moreover, Ward's junction clustering method is adopted to divide the time periods and the corresponding deformation areas. Practical engineering case study verifies the rationality of the selected similarity indicators and the effectiveness of the model.