Time domain modal identification of Tsing Ma suspension bridge

This paper presents results of a research on modal identification of the Tsing Ma bridge by using an improved eigensystem realization algorithm on the basis of ambient testing data. The testing was carried out before the bridge was open to traffic and after the completion of surfacing. The data were acquired group by group because of limited number of accelerometers and channels of data loggers available. Without traffic load, the responses of bridge to only less intensive ambient excitations were significantly influenced by the irregular movement of heavy construction vehicles on the deck. To cut off noises in the testing data and make the ambient signals more stationary, the Chebyshev digital filter was used in stead of window digital filters. A numbers of groups of random decrement (RD) functions to convert the ambient responses to free vibrations were built with different triggering levels for different measurement groups. An improved Eigensystem Realization Algorithm (FERA) method was employed to improve accuracy and efficiency of modal identification. FERA uses cross-correlation functions of RD functions to form the Hankel matrix instead of RD functions and used eigenvalue decomposition instead of singular value decomposition. The modes were identified group by group and then assembled by using messages acquired at reference points to form modes of the complete bridge. 79 modes of the Tsing Ma bridge had been identified, including 5 complex modes formed in accordance with unevenly distributed damping in the bridge. The identified modes in time domain were then compared with those identified in frequency domain and finite element analytical results.