基于模式形态距离的桥梁监测测点相似性聚类研究

为挖掘桥梁检测各测点之间的相似关系,提出基于模式形态距离的时间序列相似性度量方法。该方法首先根据监测时间序列的形态特征将序列划分成若干模式,然后以各模式形态的动态变化趋势差异为依据进行相似性的判别,并定义了各类判别结果的距离函数,最后得出各测点间的模式形态距离。在此基础上,对玉峰大桥监测点的相似性进行层次聚类分析,结果反映出的信息与桥梁的真实结构状况一致。监测点的相似性分析为桥梁结构提供了更深层次信息挖掘的可能,为传感器的坏点排查以及结构的异常数据判别提供了科学的依据。

Bridge monitoring points similarity clustering analysis based on pattern-shape distance

The bridge monitoring system consists of a large amount of sensors,and there are likely to be some similarities among the data collected by different sensors.To figure out the similarity relations,a pattern-shape distance based similarity measurement approach was proposed.Firstly,the monitoring time series was divided into several patterns according to the morphological characteristics.Secondly,similarity discrimination based on the difference of dynamic change trend of each pattern shape was discussed,and the distance function of each discrimination result was defined.Thus,the pattern shape distance of different monitoring points could be calculated.On this basis,similarity analysis of monitoring points on Yufeng Bridge was discussed by means of hierarchy clustering.The clustering result shows good consistence with bridge real states.The similarity analysis of monitoring points provides the potential of deeper information mining of bridge structures and scientific basis in sensor troubleshooting and discrimination of abnormal data.