Abstract
This paper deals with a modal identification method for large-scale structures such as boiler plants in thermal power stations. The proposed method utilizes seismic records which include multidirectional base acceleration data, and multi-I/O identification is carried out. This method is composed of two stages: (1) processing of frequency transfer functions by multi-variate autoregressive (AR) spectral analysis, and (2) curve-fitting to extract modal parameters. This method is examined using time-series data of seismic response simulation, and a comparative study is carried out between multi-input and single-input identification. It is shown that asymmetrical structures excited by earthquakes should be identified using multi-I/O identification. This method has actually been applied to in-situ seismic observation data of a boiler plant in operation. As a result of these investigations, this kind of modal identification has proven to be an effective tool for seismic modeling of large-scale structures.
