A Soft Post-Earthquake Damage Identification Methodology Using Vibration Time Series
A neural-network-based post-earthquake damage identification methodology for smart structures with the direct use of vibration measurements is developed. Two neural networks are constructed to facilitate the process of post-earthquake damage identification. The rationality of the proposed methodology is explained and the theory basis for the construction of an emulator neural network (ENN) and a parametric evaluation neural network (PENN) are described according to the discrete time solution of the structural state space equation. An evaluation index called the root mean square of the prediction difference vector (RMSPDV) is presented to evaluate the condition of different associated structures. Based on the trained ENN, which is a non-parametric model of the object structure in a healthy state, and the PENN that describes the relation between structural parameters and the components of the corresponding RMSPDVs, the inter-storey stiffness of the damaged object structure is identified. The accuracy, sensibility and efficacy of the proposed strategy for different ground excitations are also examined using a multi-storey shear building structure by numerical simulations. Since the methodology does not require the extraction of structural dynamic characteristics such as frequencies and mode shapes from measurements, it has the potential of being a practical tool for health monitoring of smart engineering structures.
B. Xu et al., "A Soft Post-Earthquake Damage Identification Methodology Using Vibration Time Series," Smart Materials and Structures, Institute of Physics - IOP Publishing, Jun 2005.
The definitive version is available at http://dx.doi.org/10.1088/0964-1726/14/3/014
Civil, Architectural and Environmental Engineering
Keywords and Phrases
Emulator Neural Network (ENN); Multi-Storey Shear Building; Parametric Evaluation Neural Network (PENN); Root Mean Square; Structural State Space Equation
Library of Congress Subject Headings
Neural networks (Computer science)
Article - Journal
© 2005 Institute of Physics - IOP Publishing, All rights reserved.