Time-series coefficient-based deterioration detection algorithm

Regardless how well structures were designed, all existing buildings deteriorate over time due to environmental effects, varying service loads and aging. Hence, it is importent to continuously track the deterioration condition of structures. Structural health monitoring (SHM) based assessment procedures can be employed to evaluate this. In this regard, vibration-based methods are amongst the most effective ones as they can be used in ambient vibration and operational loading conditions. Each building has a unique set of vibration characteristics which change with accumulated deterioration and damage. However, the changes due to deterioration are generally subtler than changes due to damage; consequently, more difficult to detect. Therefore, deterioration detection procedures need to be more accurate and sensitive to these changes. This paper presents a nonparametric statistical test to capture changes in dynamic characteristics of structures using SHM data. First, coefficients of an autoregressive (AR) time-series model are made to represent the vibration response data of a structure. Then, statistical hypotheses of Two-sample t-tests are performed and a function of the resulting P-values is used to detect changes in the structure due to deterioration. A novel AR model order estimation procedure was proposed in order to enhance the sensitivity of the method. The proposed vibration-based deterioration identification method was successfully verified utilizing sensor data from an experiment carried out on a bookshelf structure. Results show that the proposed methodology can clearly detect deterioration.