Feature Extraction within the Fei-Tsui Arch Dam under Environmental Variations

The objective of this research is to develop methods for analyzing the seismic response data and the long-term static data of the Fei-tsui arch dam, and based on the result of analysis to set an early warning threshold level for dam safety early warning evaluation. First, the input/output subspace identification technique is used to analysis the recorded seismic data from 84 earthquake events in order to identify the modal properties of the dam under different water level. Considering the spatial variability of input excitation, two kinds of system model are applied to subspace identification technique: the single-input and the multiple-input system. The regression curves between the identified system natural frequencies and water level are developed from the statistical analysis of identification results. Second, two different approaches are applied to extract features of the long-term data of the dam. The methods include the singular spectrum analysis with AR model (SSA-AR) and the nonlinear principal component analysis (NPCA) using auto-associate neural network method (AANN). By using these methods, the residual deformation between the estimated and the recorded data was generated, through statistical analysis, the threshold level of the dam static deformation can be determined. Discussion on (1) the difference between two kinds of input model for subspace identification and (2) proposed methods to extract static data are also made in this research.