Model-free identification of multiple periodic excitations and detection of structural anomaly using noisy response measurements

Z. G. Ying, Y. W. Wang, Y. Q. Ni, C. Xu

Research output: Journal article publicationJournal articleAcademic researchpeer-review

Abstract

Anomaly and damage detection is an important research topic in the field of structural health monitoring (SHM). It is in general difficult to establish a precise computational model and measure multiple dynamic loads for complex structures. Model-free identification methods using only response measurements are therefore highly desired. Based on second-order statistics blind separation (SOSBS), this study explores response-only blind excitation separation and structural feature extraction when the structure is subject to multiple periodic excitations. The proposed method proceeds with two steps: (i) a transformation to convert the measurement space to eigenspace with identity covariance matrix and compact the measurement dimension to independent source dimension; and (ii) joint diagonalization of covariances with various time shifts to determine the mixture features. Neither structural model nor measurement of excitations is required in this method, and the extracted mixture matrix representative of structural dynamic characteristics can be used for structural anomaly detection and damage diagnosis. Both numerical simulation of a 3-degree-of-freedom vibration system and experimental study of a 5-story physical structure are conducted to verify the proposed method.

Original languageEnglish
Pages (from-to)407-423
Number of pages17
JournalSmart Structures and Systems
Volume28
Issue number3
DOIs
Publication statusPublished - Sep 2021

Keywords

  • Blind excitation separation
  • Model-free and response-only approach
  • Multiple periodic excitations
  • Second-order statistics
  • Structural anomaly detection

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Computer Science Applications
  • Electrical and Electronic Engineering

Cite this