A novel method for detecting structural damage based on data-driven and similarity-based techniques under environmental and operational changes

The applications of time series modeling and statistical similarity methods to structural health monitoring (SHM) provide promising and capable approaches to structural damage detection. The main aim of this article is to propose an efficient univariate similarity method named as Kullback similarity (KS) for identifying the location of damage and estimating the level of damage severity. An improved feature extraction technique based on autoregressive (AR) model is presented to extract independent residuals of the AR model as damage-sensitive features. This technique emphasizes to choose a sufficient order such that the model residuals be independent. The proposed univariate similarity approach is a new application of the well-known KS method that attempts to measure a difference between two randomly distributed variables. The major contribution of the proposed KS method is that it only requires one measurement of undamaged and damaged conditions to compute the similarity between them. For the process of damage localization, the sensor location associated with the largest KS quantity is identified as the damaged area. In the damage level estimation, it is necessary to compare at least two different damaged conditions and find the maximum KS value in these conditions as the highest level of damage severity. The performance and capability of the improved and proposed methods is successfully verified by an experimental laboratory frame belonging to the Los Alamos National Laboratory. Results show that the methods are powerful and reliable tools for identifying the location of damage and estimating the level of damage severity.