Structural Health Monitoring

This work addressed the application of a novel family of signal processing tools, known as modern spectral estimation (MSE) techniques, to multi-channel accelerometer data collected on structures. MSE techniques hold the promise of providing a fully automatable means of extracting modal parameter estimates (i.e., eigenfrequencies and mode shape vectors) from time-series accelerometer data. The method involves computing the SVD of a Hankel matrix, which is built up from accelerometer signal time-domain auto- and cross-correlation statistics, and using the decomposition to isolate the eigenfrequencies via matrix algebra methods.

The effort under this project focused exclusively on the problem of eigenfrequency estimation, i.e., determining the modal resonance frequencies of a vibrating mechanical structure. In analyses performed on experimental data acquired from a vibrating test article, the effectiveness of the approach was confirmed insofar as it demonstrated a strong correspondence between MSE-calculated response frequencies and power spectra peaks calculated using a discrete Fourier transform (DFT). All of the major peaks that appear in the DFT power spectra are identified and their widths characterized accurately. Tracking of MSE values over time can potentially provide indication of small and subtle shifts in modal parameter values, and thereby early warning of structural damage. These methods lend themselves to nonintrusive, in situ monitoring based on natural excitation and can be used in forward-fit (e.g., composite structures), as well as in retrofit, applications.

Future efforts will expand upon this work and lead to estimation of the mode shape vectors. Such modal parameter estimates will be useful for making inferences about the underlying health of a mechanical structure via diagnostic tracking techniques.