Regression-based Statistical Change Point Analysis for Damage Localization.

摘要

Structural health monitoring (SHM) research has become a vital tool in maintaining the integrity of structures that has been refined over the years. There are numerous methods for damage detection and localization; yet some are not efficient. For example, researchers have used dynamic properties as damage features to monitor a structure because they change in the presence of damage; however, these methods are global in nature. Research in improving them (i.e. having automated, statistical monitoring techniques) is critical to the advancement of the civil engineering field. This thesis presents the implementation of damage detection methods using an experimental structure. Damage features are created from linear regression models and are utilized in control charts to localize damage because they represent the changing properties of a structure in the event of damage. Therefore, this thesis evaluates the performance of different damage features and change point analysis methods in detecting and localizing damage.