Structural Health Monitoring of Historical Structures and New Techniques for Data Analysis

Structural health monitoring of historical structures typically involves collecting continuous vibration data with very low signal-to-noise ratios (SNR). Fourier-based standard spectral analysis techniques are not appropriate to analyze such data, because low SNR can cause large errors in spectral analysis, particularly when they involve spectral ratios. This is particularly true for historical structures, because most historical structures are fairly massif and rigid. In this paper, we demonstrate the types and the magnitudes of errors that the standard spectral analysis can produce, and introduces several new techniques to analyze ambient vibration data. They include leastsquares approximations of Fourier transforms, correlation-based methods, and eigen-based methods. The superiority of the new methods over the spectral analysis methods is demonstrated by using a records from a historical mosque in Istanbul.