Damage Diagnosis Algorithm for Civil Structures Using a Sequential Change Point Detection Method and Time-Series Analysis

Abstract

This paper introduces a structural damage diagnosis algorithm that uses time-series analysis to extract damage sensitive features (DSFs) from structural responses and a single fault sequential change point detection method to classify those features into damage states. This algorithm provides the optimum asymptotic solution for the problem of minimizing the fault detection time delay given a false alarm rate. The advantage of this algorithm comes from its sequential nature, which allows us to make a decision from a series of observations instead of one, eliminating the need for all past observations to be stored. For validation, we applied the algorithm to a set of white noise shake table test data collected from a four-story steel momentresisting frame. The results show that the algorithm can identify damage, particularly when it uses multi-dimensional DSFs and lower false alarm rates, but further study is necessary for damage localization.