ABSTRACT
Structural health monitoring methods that utilize statistical pattern recognition method shave become increasingly popular (see Chang 1999, 2001, 2003, 2005 and 2007). In all these methods, either an autoregressive, AR, (e.g., Sohn et al. 2005, Nair and Kiremidjian 2006, 2007, Lynch et al. 2006, Noh et al. 2007) or a wavelet (Nair 2007) is fitted to measurement data and the parameters of the model are used in formulating a damage sensitive feature. In addition, Nair and Kiremidjian (2006) showed that changes in the first three AR coefficients are related to changes in the stiffness of the structure. These methods have gained popularity because of their computational efficiency making them particularly suitable for embedding on a small form microprocessor such as those used in wireless structural monitoring systems. The initial testing and validation of these methods was limited to the ASCE Benchmark structure and the results were indeed very promising. Additional testing and validation is needed, however, for data collect on variety of structures under different loading and environmental conditions.
