.png){kind=icon}
.png){kind=icon}
Return
Towards robust statistical damage localization via model-based sensitivity clustering
.png){kind=icon}
OA
AI
DOI：10.1016/j.ymssp.2019.106341.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Damage diagnosis is a fundamental task for structural health monitoring (SHM). With the statistical sensitivity-based damage localization approach, a residual vector is computed from vibration measurements in the reference and the damaged state. The residual is analyzed statistically in hypothesis tests with respect to change directions defined by the sensitivities of the structural parameters associated to elements of a finite element (FE) model of the investigated structure. If the test for a parameter reacts, then the respective element of the structure is indicated as damaged. This approach offers a very generic and theoretically sound framework to analyze parametric changes in systems, and takes into account the intrinsic statistical uncertainty related to measurement data. Depending on the definition of the residual and of the parameterization, the approach offers a simple computation of the test statistics directly from the measurement data in the damaged system, without the need of system identification. Since an FE model is used, the approach is applicable on arbitrary structures, while no model updating is required and therefore the requirements on the FE model accuracy are less strict. While the theoretical framework has been developed previously, it lacked robustness so far for an application on real structures. The purpose of this paper is the development of this framework into a working damage localization method that is applicable on real data from complex structures. To achieve this goal, robust hypothesis tests are used, the sensitivity computation of the residual is revisited for more precision thanks to reduced modal truncation errors, and an adequate clustering approach is proposed for the case of a high-dimensional FE parameterization for complex structures. Furthermore, several robustness properties of the method are proven. Finally, an application of this framework is shown for the first time on experimental data for damage localization, namely in an ambient vibration test of a 3D steel frame at the University of British Columbia. (C) 2019 Elsevier Ltd. All rights reserved.
Keywords:
Damage localization
Structural vibration monitoring
Ambient excitation
Hypothesis testing
Subspace methods
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
8.9
Papers:
1.3W
Citations:
6.6W