Statistical model-based damage detection and localization: subspace-based residuals and damage-to-noise sensitivity ratios - Archive ouverte HAL Access content directly
Reports (Research Report) Year : 2002

Statistical model-based damage detection and localization: subspace-based residuals and damage-to-noise sensitivity ratios

Michèle Basseville
Laurent Mevel

Abstract

The vibration-based structural health monitoring problem is addressed as the double task of detecting damages modeled as changes in the eigenstructure of a linear dynamic system, and localizing the detected damages within (a FEM of) the monitored structure. The proposed damage detection algorithm is based on a residual generated from a stochastic subspace-based covariance driven identification method and on the statistical local approach to the design of detection algorithms. This algorithm basically computes a global test, which performs a sensitivity analysis of the residuals to the damages, relative to uncertainties and noises. How this residual relates to some residuals for damage localization and model updating is discussed. Damage localization is stated as a detection problem. This problem is addressed by plugging aggregated sensitivities of the modes and mode-shapes w.r.t. FEM structural parameters in the above setting. This results in directional tests, which perform the same type of damage-to-noise sensitivity analysis of the residual as for damage detection. How the sensitivity aggregation mechanism relates to sub-structuring is outlined. Numerical results obtained on one example are reported.
Fichier principal
Vignette du fichier
RR-4645.pdf (435.97 Ko) Télécharger le fichier

Dates and versions

inria-00071940 , version 1 (23-05-2006)

Identifiers

  • HAL Id : inria-00071940 , version 1

Cite

Michèle Basseville, Laurent Mevel, Maurice Goursat. Statistical model-based damage detection and localization: subspace-based residuals and damage-to-noise sensitivity ratios. [Research Report] RR-4645, INRIA. 2002. ⟨inria-00071940⟩
226 View
560 Download

Share

Gmail Mastodon Facebook X LinkedIn More