Application of time series analysis in structural damage evaluation

The time series analysis approach for damage detection employed by Los Alamos National Laboratory (LANL) is modified and applied to the benchmark building designed by ASCE task group for damage detection and localization. A two-tier prediction model, using the auto-regressive (AR) and the autoregressive model with exogenous inputs (ARX) is used for obtaining the damage sensitive feature. The damage detection approach is modified to consider the excitation and ARX prediction model order variability on the original damage-detection feature as defined by LANL. In this paper, the reference database consists of the normalized relative interstory acceleration. An empirical distribution of the residual errors is developed from the undamaged signals, which can be used to set rigorous threshold limits to check if the new signal that is generated is obtained from a damaged structure or not. This algorithm is validated for different damage patterns in the ASCE benchmark structure including minor damage patterns. Source: Proceedings of the International Conference on Structural Health Monitoring, Tokyo, Japan, November 14-16, 2003. tions. Based on the empirical distribution of the residual errors from the undamaged signals, rigorous threshold limits are set to check whether the new signal that is generated is obtained from a damaged structure or not. The applicability of this modified algorithm has been used for different combinations of structural finite element models, excitation conditions and damage patterns in the benchmark study. The results indicate that the damage feature used is also effective in identifying minor damage patterns, which was not the case with the earlier model. 2 DAMAGE DETECTION ALGORITHM OF VIBRATION SIGNALS USING TIME SERIES