Acceleration-based damage indicators for building structures using neural network emulators

Yuyin Qian, Akira Mita

Research output: Contribution to journalArticle

19 Citations (Scopus)

Abstract

In this paper we propose the use of artificial neural network (ANN) emulators for an acceleration-based approach to evaluating building structures under earthquake excitation. The input layer of the ANN is a forced vibration, described as ground acceleration and the acceleration data of several floors. The approach is improved by using the acceleration at later time steps as the output of the neural network (NN). This time delay is optimized as a tunable band to provide the most sensitive signals. Minimally, this approach requires only one sensor, making it highly practicable and flexible. It is applicable to structures under diverse excitations including even very small impacts. Based on numerical simulation of a 5-story shear structure, we determined appropriate parameters for use of an NN and studied the generality and efficacy of the approach. The damage index, the relative root mean square error, was calculated for the case of a single structural damage as well as for cases of double damages at different damage locations, and appropriate parameters for the NN emulator were proposed according to the damage patterns. Variant ground motions were used to certify the generality of the approach. The numerical simulations of the proposed approach were verified experimentally.

Original languageEnglish
Pages (from-to)901-920
Number of pages20
JournalStructural Control and Health Monitoring
Volume15
Issue number6
DOIs
Publication statusPublished - 2008 Oct 1

Keywords

  • Building structures
  • Damage identification
  • Damage indicators
  • Neural network emulator
  • Structural health monitoring

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Building and Construction
  • Mechanics of Materials

Fingerprint Dive into the research topics of 'Acceleration-based damage indicators for building structures using neural network emulators'. Together they form a unique fingerprint.

  • Cite this