Early damage detection in structural health monitoring by a sensitivity method and DBSCAN clustering

Abstract

Early damage detection is an important and initial step in structural health monitoring (SHM)<br><br>that aims to evaluate the global state of a structure and finds whether damage is available<br><br>throughout the structure. To do this, a new sensitivity function regarding modal strain energy<br><br>(MSE) is proposed to use as a novel damage-sensitive feature. The sensitivity function is<br><br>based on proposing a new equation of modal sensitivity and using the presented equation in<br><br>the global formulation of MSE. In order to apply this sensitivity function as the damagesensitive<br><br>feature, the matrix of sensitivity of MSE is converted into a vector by the vectorization<br><br>procedure. A new density-based clustering method named DBSCAN is also presented<br><br>here to detect early damage using the vector of the sensitivity of MSE. To demonstrate the<br><br>performance and reliability of the proposed methods, the ASCE benchmark structure (Phase<br><br>I) is employed as a numerical example. Results show that the proposed sensitivity function is<br><br>sensitive to damage and can be a reliable damage-sensitive feature in the applications of<br><br>SHM. Furthermore, numerical results demonstrate that the proposed DBSCAN approach is a<br><br>robust tool for detecting damage.