Information Theoretic Approach for Earthquake-Induced Building Structural Damage Diagnosis

Abstract eng:
This paper introduces a damage diagnosis algorithm using an information-theoretic approach to localize building structural damage due to earthquakes. Accurate and timely post-earthquake diagnosis of building structural damage is important for facilitating emergency responses and rehabilitations. Many prior works have focused on structural health monitoring methods to automate damage diagnosis. Among the approaches, structural vibration-based methods have been widely adopted; however most of them 1) require detailed prior information about the structure (model-based), which may not be available in reality; 2) consider each location measurement separately (signal-based), which requires dense deployment of sensors (costly to install); or 3) require post-earthquake sensing data, which is often impossible due to broken sensing/communication systems and limited availability personnel These constraints limit the applicability of existing approaches in real scenarios. This paper presents a method that localizes seismic damage in building structures by representing the wave propagation process as information exchanges between vibration signals. When a seismic wave is propagated inside building, the form of the wave changes depending on the dynamic characteristics of the structure (i.e., carries information about structural properties). Thus, the relationship between such information (i.e., information exchange) measured from different locations in the building represents the state of structure at corresponding areas. This method consists of three steps: 1) Structural vibration at each floor is recorded during an earthquake from an instrumented building. 2) A feature based on the ‘directed information’ metric from information theory is extracted to quantify the information exchange between two vibration sensor signals. 3) By detecting the changes of the information exchange feature, the evolution of seismic structural damage state is estimated. This information theoretic approach eliminates simplifying assumptions about the vibration process, such as linearity or specific distribution, which contributes to improving the damage diagnosis accuracy. The advantages of the proposed method are that it1) is accurate across different structures 2) is robust to noisy measurements 3) requires only sparse deployment of sensors and 4) does not need post-earthquake data. The algorithm is validated using experimental data from a 4-story steel moment resisting frame subjected to a series of earthquake excitations. As a result, the damage is localized with up to mean square error 4X improvement compared with the baseline method.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket