Acceleration Response Data Processing for Structural Health Monitoring of a High-Rise Building

Abstract

A 38-story steel moment-resisting frame building located in the city of San Francisco has been instrumented with 29 accelerometers for structural health monitoring. A data-recorder attached to the building basement is used to continuously store structural acceleration response data. Available recordings include both, ambient and earthquake acceleration response. In this study, three data processing techniques were applied to the collected data to define the baseline building vibration periods for monitoring of its structural condition. The power spectral density and the frequency domain decomposition techniques were applied to ambient vibration data whereas the eigensystem realization algorithm was applied to the earthquake data. In the application of these algorithms, input parameters including number of accelerometers, duration of recording, frequency resolution, and filtering of data were varied to study the variability of the results. This parametric analysis was applied to the estimated building periods to determine the level of confidence of the results. It was observed the estimated building periods are sensitive, especially near long period values, to input filtering and frequency resolution parameters. A high-pass filter was applied in the data processing to eliminate spurious period identification results, the cut-off frequency being defined based on stability of the filter algorithm. In regards of the frequency resolution, it was observed the estimated fundamental period results were scattered within a wide interval deviated from the mean when different resolutions were used. While a coarse resolution increases the precision or repeatability of the results among different datasets, a fine resolution is preferred as it allows for higher accuracy in the estimation of the structural periods. Based on the results, suggested parameters for building period estimation are listed and baseline periods are defined. The baseline periods serve as the basis to calibrate an FE model used for further studies.