National University Transportation Consortium: A Speaker Recognition Based Damage Detection

In this study, an adaptation of Mel-Frequency Cepstral Coefficients as damage sensitive features for structural health monitoring of civil structures was addressed. Typically used in speaker recognition methodologies, these indices offer an extremely easy extracting process with a few user-defined parameters and a low computational burden and they have been shown to be an effective alternative to other features for damage detection problems. To remove environmental effects from the coefficient estimation, a technique called "Cointegration", quite popular in econometrics, has been applied. Two study cases were presented: (1) a numerical simulation of a cantilever beam subject to environmental variations both in undamaged as well as damaged conditions, and (2) the benchmark case of the Z24 bridge, a structure in Switzerland that was recently demolished and that was fully instrumented, during operational conditions as well as during demolition. From the results of this study, it appears that the following conclusions can be drawn: (1) the cepstral coefficients have the potential to become quite useful damage sensitive features that can be used on bridge structures: they are compact features, easy to obtain, and require little input from the user, and (2) the cointegration technique appears to be a very effective technique to remove non-stationary effects such as those induced by the environment temperature. As shown in this report, the analyses conducted on data from the tests run on the Z24 bridge show great potential for both techniques and warrants further investigation.


  • English

Media Info

  • Media Type: Digital/other
  • Edition: Final Report
  • Features: Figures; Photos; References; Tables;
  • Pagination: 38p

Subject/Index Terms

Filing Info

  • Accession Number: 01715251
  • Record Type: Publication
  • Report/Paper Numbers: CAIT-UTC-NC47
  • Contract Numbers: DTRT13-G-UTC28
  • Created Date: Aug 29 2019 5:28PM