Identification of unusual events in multi-channel bridge monitoring data

Piotr Omenzetter*, James Mark William Brownjohn, Pilate Moyo

*Corresponding author for this work

Research output: Contribution to journalArticle

61 Citations (Scopus)
6 Downloads (Pure)

Abstract

Continuously operating instrumented structural health monitoring (SHM) systems are becoming a practical alternative to replace visual inspection for assessment of condition and soundness of civil infrastructure such as bridges. However, converting large amounts of data from an SHM system into usable information is a great challenge to which special signal processing techniques must be applied. This study is devoted to identification of abrupt, anomalous and potentially onerous events in the time histories of static, hourly sampled strains recorded by a multi-sensor SHM system installed in a major bridge structure and operating continuously for a long time. Such events may result, among other causes, from sudden settlement of foundation, ground movement, excessive traffic load or failure of post-tensioning cables. A method of outlier detection in multivariate data has been applied to the problem of finding and localising sudden events in the strain data. For sharp discrimination of abrupt strain changes from slowly varying ones wavelet transform has been used. The proposed method has been successfully tested using known events recorded during construction of the bridge, and later effectively used for detection of anomalous post-construction events.

Original languageEnglish
Pages (from-to)409-430
Number of pages22
JournalMechanical Systems and Signal Processing
Volume18
Issue number2
Early online date23 May 2003
DOIs
Publication statusPublished - 1 Mar 2004

Keywords

  • bridges
  • multivariate statistics
  • novelty detection
  • outlier analysis
  • wavelet transform

Fingerprint Dive into the research topics of 'Identification of unusual events in multi-channel bridge monitoring data'. Together they form a unique fingerprint.

Cite this