Experimental Tests of Substructure Approaches for Health Monitoring System

Abstract

Two substructure approaches combined with different statistical techniques are presented in the paper to reduce the challenges in the health monitoring system, such as the computation complexity, the number of unknown parameters and the number of sensors. In the approaches, a complete structure is divided into smaller substructures which are modelled as a series of single-degree-of-freedom (SDOF) systems by manipulating the equations of motion of the substructures. Newmark’s method is utilized to construct the discrete systems only containing accelerations from the continuous systems. The structural parameters are then separately identified in each system using statistical techniques. A new substructure approach incorporated with the constrained least square algorithm is proposed by the authors. This new approach along with the previous substructure approach based on the ARMAX models are both evaluated and compared in two laboratory experiments including system identification of a three-story structure and damage detection of a two-story structure. All the possible ways to segment the shear structure into substructures are also investigated in the study, where the substructures with lower model errors are selected. The proposed substructure approaches are able to estimate structural parameters every two seconds, and the structural performance is evaluated through the rates of changes of structural parameters. A real-time structural health monitoring (SHM) system can be realized based on the proposed substructure approaches with processing several accelerations each time.