17 April 2013 Frequency response feature selection in a Bayesian framework
Author Affiliations +
Abstract
Frequency response related quantities are widely used for damage detection and structural health monitoring (SHM) because of their computational robustness and clear physical interpretations. In reality, there is uncertainty due to noise or other operational variability involved in any feature evaluation, which makes it hard to select a robust and sensitive SHM feature. Two specific spectra are considered in this paper, namely, frequency response function (FRF) and transmissibility, while FRF includes system resonances and transmissibility only has system zeros. A Bayesian model selection framework is adopted by comparing the Bayes factor of using either feature in structural health monitoring applications, and suggests which is better with regard to plausibility. This framework is implemented with data acquired from a lab-scale plate structure, and to be more realistic, external artificial noise is contaminated to the data imitating a more stringent test condition.
© (2013) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Zhu Mao, Michael Todd, "Frequency response feature selection in a Bayesian framework", Proc. SPIE 8695, Health Monitoring of Structural and Biological Systems 2013, 869535 (17 April 2013); doi: 10.1117/12.2009686; https://doi.org/10.1117/12.2009686
PROCEEDINGS
9 PAGES


SHARE
Back to Top