9319829 Petropulu It has been shown, through theory and observations, that conditions at the site of a constructed facility can amplify the seismic ground motions significantly, as compared to those recorded on bedrock. This amplification increases the potential damage and collapse of engineered structures. The objective of this research is to examine the feasibility of applying a blind deconvolution algorithm to site response analysis; and to investigate whether the approach can lead to a better understanding of the complex phenomena underlying this problem. The approach utilizes the blind deconvolution algorithm and the recordings of ground motions at two stations on a sediment site in order to reconstruct the seismic input at the bedrock-layer interface and to recover the characteristics of the site through impulse responses (reflectivities) between the input location and the recording stations. The site characteristics are estimated by combining higher-order statistics of the seismograms and the theory of signal reconstruction from higher-order spectral phase only. Potential advantages of this approach over existing techniques are that (i) it does not require recordings at nearby rock sites or downhole recordings for the site- specific analysis, but reconstructs the seismic input motion at the base of the sediments from two surface records; (ii) it identifies the reflectivities between the bedrock-layer interface and each recording station rather than one one- dimensional site response function from the downhole (or nearby rock) station and one station on the surface of the sediment site; and (iii) it remains robust even in the presence of additive white Gaussian noise in the observations and in the input seismic motion at the base of the sediments. This exploratory study concentrates on well documented earthquakes recorded by the SMART-1 accelerograph array in Taiwan. ***
