计算机工程与应用 ›› 2024, Vol. 60 ›› Issue (18): 306-315.DOI: 10.3778/j.issn.1002-8331.2305-0467

• 工程与应用 • 上一篇    下一篇

BSPST:形变监测仪器故障分类算法

吴晓赢,邓红霞,胡玉良,李颖,穆慧敏   

  1. 1.太原理工大学 计算机科学与技术学院,太原 030024
    2.山西省地震局 山西地震台,太原 030021
  • 出版日期:2024-09-15 发布日期:2024-09-13

BSPST: Fault Classification Algorithm for Deformation Monitoring Instruments

WU Xiaoying, DENG Hongxia, HU Yuliang, LI Ying, MU Huimin   

  1. 1.College of Computer Science and Technology, Taiyuan University of Technology, Taiyuan 030024, China
    2.Seismological Bureau of Shanxi Province, Shanxi Earthquake Agency, Taiyuan 030021, China
  • Online:2024-09-15 Published:2024-09-13

摘要: 针对现有形变监测仪器发生故障时故障类别难以准确分类的问题,提出了一种基于最大区分子序列(Shapelet)转换的时间序列分类算法(best qualify Shapelet Transform,BSPST)。为了提升Shapelet质量,利用布隆过滤器和相似度匹配保留一组高质量的候选Shapelet来构建分类模型,BSPST利用布隆过滤器筛选出同类别中重复的符号聚合近似(symbolic aggregation approximation,SAX)单词。随后通过位图中标记的单词来评价SAX单词的重复度,以此去除类别中相似的SAX单词。最终将处理后的符号聚合近似单词转化为高质量的Shapelet。通过Shapelet转换技术,对数据进行转换。最后采取集成分类器进行分类。根据地震形变仪器故障数据建立了7个地震设备故障数据集,并结合东安格利亚大学和加州大学河滨分校时间序列分类仓库中选取的44个数据集和具代表性的最先进的方法进行了充分的实验验证。结果表明,BQST算法在分类精度、分类速度上稳居前列,有效解决了形变监测仪器的故障分类问题。

关键词: 故障诊断, 时间序列分类, 最大区分子序列, 形变仪器

Abstract: A time series classification algorithm (best qualify Shapelet Transform (BSPST)) based on the maximum zone molecular sequence (Shapelet) transformation is proposed to address the problem that it is difficult to accurately classify fault categories when faults occur in existing deformation monitoring instruments. In order to improve the quality of Shapelet, a set of high-quality candidates Shapelet is retained by using Bloom filter and similarity matching to construct the classification model. BSPST uses Bloom filter to filter out the duplicate SAX words in the same category, and then evaluates the duplication of SAX words by the words marked in the bitmap, so as to remove the similar SAX words in the category. Finally, the processed symbolic aggregation approximation (SAX) words are transformed into high quality Shapelet. The data are transformed by Shapelet transformation technique. Finally, an integrated classifier is adopted for classification. Seven seismic equipment failure datasets are created based on the seismic deformation instrumentation failure data, and 44 datasets selected from the University of East Anglia and University of California Riverside (UEA&UCR) time series classification repository are combined with representative state-of-the-art classification tools. The results show that the BSPST algorithm has been fully experimentally validated with a representative state-of-the-art method. The results show that the BSPST algorithm is in the top position in terms of classification accuracy and classification speed. The problem of fault classification in deformation monitoring instruments is effectively solved.

Key words: fault diagnosis, time series classification, Shapelet, deformation instrument