Supervised damage and deterioration detection in building structures using an enhanced autoregressive time-series approach

In this paper, a supervised learning approach is introduced for detecting both damage and deterioration in two building models under ambient and forced vibrations. The coefficients and residuals of autoregressive (AR) time-series models are utilized for extracting features through some statistical indices. Moreover, a novel algorithm called best-uncorrelated features selection (BUFS) is proposed and utilized in order to select the most sensitive and uncorrelated features, which are used as predictors. Accordingly, a common set of predictors capable of detecting both damage and deterioration is established and used in order to form a general pattern of the structural condition. Besides, the BUFS algorithm can also be utilized with other features as well as different types of structures and depicts the most sensitive predictors. The results indicate that the proposed method is capable of detecting damage and deterioration in both models precisely, even in a noisy environment, and the appropriate features are introduced.