24.12.2014 Views

Earthquake Engineering Research - HKU Libraries - The University ...

Earthquake Engineering Research - HKU Libraries - The University ...

Earthquake Engineering Research - HKU Libraries - The University ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Proceedings of the International Conference on<br />

Advances and New Challenges in <strong>Earthquake</strong><br />

<strong>Engineering</strong> <strong>Research</strong>, Hong Kong Volume<br />

555<br />

UNSCENTED PARTICLE FILTER FOR TIME DOMAIN<br />

IDENTIFICATION OF NONLINEAR STRUCTURAL DYNAMIC<br />

SYSTEMS<br />

K.Y. Koo and C.B. Yun<br />

Department of Civil and Environmental <strong>Engineering</strong>,<br />

Korea Advanced Institute of Science and Technology, Daejoen, Korea<br />

ABSTRACTS<br />

In this study, a recently developed unscented particle filter (UPF) technique is studied for<br />

identification of nonlinear structural dynamic systems. It is well-known that there is no general<br />

solution for parameter identification of nonlinear structural system. As an alternative in the sense<br />

of linear approximation solution, the Extended Kalman filter (EKF) has been frequently employed<br />

for identification of time-varying structural parameters. However, the EKF has several drawbacks<br />

such as biased estimations and erroneous estimations especially for highly nonlinear dynamic<br />

systems due to its crude linearization scheme. To overcome the weakness of the EKF, the UPF was<br />

recently developed. <strong>The</strong> UPF is a novel method for nonlinear, non-Gaussian, and on-line<br />

estimation. <strong>The</strong> algorithm is a particle filter that uses an unscented Kalman filter (UKF) to generate<br />

the importance proposal distribution.<br />

Numerical simulation studies have been carried out on SDOF and MDOF systems. <strong>The</strong> results on<br />

the linear and nonlinear SDOF systems show that the UPF gives more accurate and robust estimates<br />

under the existence of the system and measurement errors with rough initial guesses for the states<br />

and the error covariance matrices. <strong>The</strong> results on a five-story building structure subjected to<br />

nonlinear behavior at the bottoms of the columns show that the UPF has good estimation capability.<br />

<strong>The</strong> results from a series of numerical simulations indicate that the UPF is superior to the EKF for<br />

the system identification of nonlinear dynamic systems especially for highly nonlinear systems.<br />

1. INTRODUCTION<br />

Detection of structural change or damage is one of the most important and challenging issues in<br />

the structural health monitoring system. Statistical inference approaches are more appropriate in<br />

the detection of the structural change due to the complex nature of civil infrastructures and<br />

noise-polluted measurements. Especially Bayesian approach in statistical inference has useful<br />

structure in on-line estimation and has been applied to various fields, i.e., physics, control system<br />

and signal processing.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!