基于相空间重构的事件性时间序列片段的提取

Eventuality segments extraction from a time series based on phase space reconstruction

时间序列中的相似性片段包含着一些潜在有用的信息。在状态监测系统中,从信号采集系统得到的时间序列中找出相似性片段,是进一步特征提取及状态分析等数据处理的基础。现有的相似性片段寻找算法通常是用滑动长度固定的窗提取等长的子序列进行相似性匹配计算,但在实际应用中,具有相似特征的子序列长度未必一定相等,致使找到的相似性片段可能不具有事件性。针对这一问题研究了一种基于相空间重构的相似性片段寻找方法。首先,将一维时间序列以时延方式嵌入到二维相空间中,计算得到与时间序列相对应的峰值特征序列;其次,根据已有的经验知识通过峰值特征序列确定子序列分割的位置,划分出不等长的事件性片段;最后,对得到的事件性片段进行聚类,从而找出其中的相似性片段。心电信号和故障轴承振动信号的实验结果表明,使用本方法找到的相似性片段长度不等,并具有事件性特征。

In a condition monitoring system,extracting similarity segments containing some potentially useful imformation from a time series obtained with a signal acquisition system is the basis for further data processing,such as,feature extraction and state analysis.The existing similarity subsequences searching algorithm usually uses a fixed sliding length window to extract equal length subsequences for similarity matching calculation.However,in practical applications,subsequences with similar characteristics are not inevitably equal in length to make the extracted similarity segments not have eventuality.Here,aiming at this problem,a similarity subsequence extraction method was studied based on phase space reconstruction.Firstly,a 1-D time series was embedded in a 2-D phase space according to the time delay method,and a peak feature sequence corresponding to the time series was calculated.Then,the existing empirical knowledge was used to determine the position of subsequence segmentation with the peak feature sequence,and divide eventuality subsequences with unequal lengths.Finally,the acquired eventuality subsequences were clustered to extract similarity subsequences.Test results of ECG signals and faulty bearing vibration signals indicated that the similarity segments extracted using the proposed method are unequal in length and have eventuality characteristics.