A model for quantifying uncertainty in the estimation of noise-contaminated measurements of transmissibility

作者：Mao Zhu; Todd Michael^*

来源：Mechanical Systems and Signal Processing, 2012, 28: 470-481.

DOI：10.1016/j.ymssp.2011.10.002

• System identification
• Transmissibility
• Transfer function
• Uncertainty quantification
• Statistical modeling
• Structural health monitoring

System identification in the frequency domain is a very important process in many aspects of engineering. Among many forms of frequency domain system identification such as frequency response function analysis and modal decomposition, transmissibility (output-to-output relationship) estimation has been regarded as one of the most practical tools for its clear physical interpretation, its compatibility with output-only data, and its sensitivity to local changes of structural parameters. Due to operational and environmental variability in any real system, quantization and estimation error, and extraneous measurement noise, the computation of transmissibility may contain a significant level of uncertainty and variability, and these sources propagate to degrade system identification quality and in some cases to system mischaracterization. In this paper, the uncertainty of the magnitude of a transmissibility estimator via output auto-power density spectra is quantified, an exact probability density function for the estimates is derived analytically via a Chi-square bivariate approach, and it is validated with Monte Carlo simulation. Validation shows very consistent results between the observed histogram and predicted distribution for different estimation and noise conditions.

• 出版日期2012-4