A damage localization and quantification algorithm for indirect structural health monitoring of bridges using multi-task learning

Jingxiao Liu, Mario Bergés, Jacobo Bielak, James H. Garrett, Jelena Kovacevic, Hae Young Noh

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We propose a multi-task learning approach for estimating both the location and magnitude of damage occurring on an experimental bridge using acceleration signals collected from a passing vehicle. This is a low-cost and low-maintenance indirect structural health monitoring approach in which sensors on the vehicle are used to detect bridge damage. Recently, signal processing and machine learning approaches have been shown to perform well in achieving higher-level structural health monitoring objectives, such as damage localization and quantification. However, these methods not only lack robustness to measurement and model noises, but also require more physical insights. Guided by a theoretical formulation of a simple vehicle-bridge interaction system, our approach preserves the non-linearity of the trend of the acceleration signals as severity changes, and simultaneously localizes and quantifies the damage for minimizing uncertainties propagating from the location estimation. We evaluate our model on an experimental dataset. In the experiments, the damage is represented by a mass with gradually changing magnitude attached at different positions on the bridge. The results show that it can estimate locations of the damage with an accuracy of 0.08 m (3.30% of the total length of the bridge) and changes in severity level with an accuracy of 17.81 grams (8.9% of the maximum severity mass).

Original languageEnglish (US)
Title of host publication45th Annual Review of Progress in Quantitative Nondestructive Evaluation, Volume 38
EditorsSimon Laflamme, Stephen Holland, Leonard J. Bond
PublisherAmerican Institute of Physics Inc.
ISBN (Electronic)9780735418325
DOIs
StatePublished - May 8 2019
Event45th Annual Review of Progress in Quantitative Nondestructive Evaluation, QNDE 2018 - Burlington, United States
Duration: Jul 15 2018Jul 19 2018

Publication series

NameAIP Conference Proceedings
Volume2102
ISSN (Print)0094-243X
ISSN (Electronic)1551-7616

Conference

Conference45th Annual Review of Progress in Quantitative Nondestructive Evaluation, QNDE 2018
CountryUnited States
CityBurlington
Period7/15/187/19/18

Fingerprint

health monitoring
structural health monitoring
learning
damage
monitoring
artificial intelligence
preserves
vehicles
uncertainty
machine learning
signal processing
nonlinearity
maintenance
estimating
methodology
sensor
trends
formulations
sensors
estimates

ASJC Scopus subject areas

  • Ecology, Evolution, Behavior and Systematics
  • Ecology
  • Plant Science
  • Physics and Astronomy(all)
  • Nature and Landscape Conservation

Cite this

Liu, J., Bergés, M., Bielak, J., Garrett, J. H., Kovacevic, J., & Noh, H. Y. (2019). A damage localization and quantification algorithm for indirect structural health monitoring of bridges using multi-task learning. In S. Laflamme, S. Holland, & L. J. Bond (Eds.), 45th Annual Review of Progress in Quantitative Nondestructive Evaluation, Volume 38 [090003] (AIP Conference Proceedings; Vol. 2102). American Institute of Physics Inc.. https://doi.org/10.1063/1.5099821

A damage localization and quantification algorithm for indirect structural health monitoring of bridges using multi-task learning. / Liu, Jingxiao; Bergés, Mario; Bielak, Jacobo; Garrett, James H.; Kovacevic, Jelena; Noh, Hae Young.

45th Annual Review of Progress in Quantitative Nondestructive Evaluation, Volume 38. ed. / Simon Laflamme; Stephen Holland; Leonard J. Bond. American Institute of Physics Inc., 2019. 090003 (AIP Conference Proceedings; Vol. 2102).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Liu, J, Bergés, M, Bielak, J, Garrett, JH, Kovacevic, J & Noh, HY 2019, A damage localization and quantification algorithm for indirect structural health monitoring of bridges using multi-task learning. in S Laflamme, S Holland & LJ Bond (eds), 45th Annual Review of Progress in Quantitative Nondestructive Evaluation, Volume 38., 090003, AIP Conference Proceedings, vol. 2102, American Institute of Physics Inc., 45th Annual Review of Progress in Quantitative Nondestructive Evaluation, QNDE 2018, Burlington, United States, 7/15/18. https://doi.org/10.1063/1.5099821
Liu J, Bergés M, Bielak J, Garrett JH, Kovacevic J, Noh HY. A damage localization and quantification algorithm for indirect structural health monitoring of bridges using multi-task learning. In Laflamme S, Holland S, Bond LJ, editors, 45th Annual Review of Progress in Quantitative Nondestructive Evaluation, Volume 38. American Institute of Physics Inc. 2019. 090003. (AIP Conference Proceedings). https://doi.org/10.1063/1.5099821
Liu, Jingxiao ; Bergés, Mario ; Bielak, Jacobo ; Garrett, James H. ; Kovacevic, Jelena ; Noh, Hae Young. / A damage localization and quantification algorithm for indirect structural health monitoring of bridges using multi-task learning. 45th Annual Review of Progress in Quantitative Nondestructive Evaluation, Volume 38. editor / Simon Laflamme ; Stephen Holland ; Leonard J. Bond. American Institute of Physics Inc., 2019. (AIP Conference Proceedings).
@inproceedings{e152c3b977f249f38f479f537769eb61,
title = "A damage localization and quantification algorithm for indirect structural health monitoring of bridges using multi-task learning",
abstract = "We propose a multi-task learning approach for estimating both the location and magnitude of damage occurring on an experimental bridge using acceleration signals collected from a passing vehicle. This is a low-cost and low-maintenance indirect structural health monitoring approach in which sensors on the vehicle are used to detect bridge damage. Recently, signal processing and machine learning approaches have been shown to perform well in achieving higher-level structural health monitoring objectives, such as damage localization and quantification. However, these methods not only lack robustness to measurement and model noises, but also require more physical insights. Guided by a theoretical formulation of a simple vehicle-bridge interaction system, our approach preserves the non-linearity of the trend of the acceleration signals as severity changes, and simultaneously localizes and quantifies the damage for minimizing uncertainties propagating from the location estimation. We evaluate our model on an experimental dataset. In the experiments, the damage is represented by a mass with gradually changing magnitude attached at different positions on the bridge. The results show that it can estimate locations of the damage with an accuracy of 0.08 m (3.30{\%} of the total length of the bridge) and changes in severity level with an accuracy of 17.81 grams (8.9{\%} of the maximum severity mass).",
author = "Jingxiao Liu and Mario Berg{\'e}s and Jacobo Bielak and Garrett, {James H.} and Jelena Kovacevic and Noh, {Hae Young}",
year = "2019",
month = "5",
day = "8",
doi = "10.1063/1.5099821",
language = "English (US)",
series = "AIP Conference Proceedings",
publisher = "American Institute of Physics Inc.",
editor = "Simon Laflamme and Stephen Holland and Bond, {Leonard J.}",
booktitle = "45th Annual Review of Progress in Quantitative Nondestructive Evaluation, Volume 38",

}

TY - GEN

T1 - A damage localization and quantification algorithm for indirect structural health monitoring of bridges using multi-task learning

AU - Liu, Jingxiao

AU - Bergés, Mario

AU - Bielak, Jacobo

AU - Garrett, James H.

AU - Kovacevic, Jelena

AU - Noh, Hae Young

PY - 2019/5/8

Y1 - 2019/5/8

N2 - We propose a multi-task learning approach for estimating both the location and magnitude of damage occurring on an experimental bridge using acceleration signals collected from a passing vehicle. This is a low-cost and low-maintenance indirect structural health monitoring approach in which sensors on the vehicle are used to detect bridge damage. Recently, signal processing and machine learning approaches have been shown to perform well in achieving higher-level structural health monitoring objectives, such as damage localization and quantification. However, these methods not only lack robustness to measurement and model noises, but also require more physical insights. Guided by a theoretical formulation of a simple vehicle-bridge interaction system, our approach preserves the non-linearity of the trend of the acceleration signals as severity changes, and simultaneously localizes and quantifies the damage for minimizing uncertainties propagating from the location estimation. We evaluate our model on an experimental dataset. In the experiments, the damage is represented by a mass with gradually changing magnitude attached at different positions on the bridge. The results show that it can estimate locations of the damage with an accuracy of 0.08 m (3.30% of the total length of the bridge) and changes in severity level with an accuracy of 17.81 grams (8.9% of the maximum severity mass).

AB - We propose a multi-task learning approach for estimating both the location and magnitude of damage occurring on an experimental bridge using acceleration signals collected from a passing vehicle. This is a low-cost and low-maintenance indirect structural health monitoring approach in which sensors on the vehicle are used to detect bridge damage. Recently, signal processing and machine learning approaches have been shown to perform well in achieving higher-level structural health monitoring objectives, such as damage localization and quantification. However, these methods not only lack robustness to measurement and model noises, but also require more physical insights. Guided by a theoretical formulation of a simple vehicle-bridge interaction system, our approach preserves the non-linearity of the trend of the acceleration signals as severity changes, and simultaneously localizes and quantifies the damage for minimizing uncertainties propagating from the location estimation. We evaluate our model on an experimental dataset. In the experiments, the damage is represented by a mass with gradually changing magnitude attached at different positions on the bridge. The results show that it can estimate locations of the damage with an accuracy of 0.08 m (3.30% of the total length of the bridge) and changes in severity level with an accuracy of 17.81 grams (8.9% of the maximum severity mass).

UR - http://www.scopus.com/inward/record.url?scp=85066073849&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85066073849&partnerID=8YFLogxK

U2 - 10.1063/1.5099821

DO - 10.1063/1.5099821

M3 - Conference contribution

AN - SCOPUS:85066073849

T3 - AIP Conference Proceedings

BT - 45th Annual Review of Progress in Quantitative Nondestructive Evaluation, Volume 38

A2 - Laflamme, Simon

A2 - Holland, Stephen

A2 - Bond, Leonard J.

PB - American Institute of Physics Inc.

ER -