Substructural Identification For Damage Detection During Seismic Events

Abstract

The substructural identification method was applied to help reduce the required sensors and the unknown values that need to be monitored at the same time when monitoring the health of a portion of the whole structure. Two types of data were tested separately, including acceleration data and displacement data, to see which one provided better substructural identification using the Craig-Bampton equation of motion representation, Extended Kalman Filter, and the Bayesian method. The Bathe method was used to simulate the necessary data for the first example, but the acceleration and displacement data were measured on a physical demo for the second example. The equations of motion of the substructure were rewritten into the state space form. Then the Extended Kalman Filter was applied to estimate the unknown displacement, velocity, stiffness, and damping coefficients. Furthermore, the Bayesian method estimated in real-time the noise parameters and quantified the associated estimation uncertainty. Results were examined for a variety of damage cases and load scenarios, including random external forces, random ground motion, and four kinds of earthquakes such as the Northridge, Mendocino, Kobe, and El-Centro earthquakes. The substructure identification method was shown to detect changes in stiffness, noise parameters, and damping parameters. However, when using displacement data, the value of the actual stiffness could not be correctly found, and the spring stiffness values at the interface and the noise parameters were not updated correctly. The amount of stiffness change also affects the ability to detect damage, and a more realistic equation of motion tends to give better results when identifying a physical structure.