한국전산구조공학회논문집 (Journal of the Computational Structural Engineering Institute of Korea)

  • 제37권4호
  • /
  • Pages.283-293
  • /
  • 2024
  • /
  • 1229-3059(pISSN)
  • /
  • 2287-2302(eISSN)
한국전산구조공학회 (Computational Structural Engineering Institute of Korea)
권호 표지
DOI QR코드

DOI QR Code

급격한 구조손상탐지를 위한 베이지안 필터 적용가능성 비교 검토 연구

Comparison of the Applicability of Bayesian Filters for System Identification of Sudden Structural Damage

  • 이세혁 (한국건설기술연구원 구조연구본부) ;
  • 김민규 (한국원자력연구원 구조.지진안전연구부 ) ;
  • 이상리
  • Se-Hyeok Lee (Department of Structural Engineering Research, Korea Institute of Civil Engineering and Building Technology) ;
  • Minkyu Kim (Structural and Seismic Safety Research Division, Korea Atomic Energy Research Institute) ;
  • Sang-ri Yi (Department of Civil and Environmental Engineering, University of California)
  • 투고 : 2024.07.22
  • 심사 : 2024.08.06
  • 발행 : 2024.08.31
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

본 논문에서는 지진 하중으로 인한 급격한 구조손상탐지를 수행하기 위해 분산점 칼만필터(Unscented Kalman Filter, UKF)와 파티클 필터(Particle Filter)를 소개하고 지진 손상 시나리오에 적용 및 비교·검토하였다. 이때, 비선형 전단 빌딩을 모사하기 위해 Bouc-Wen 모델을 사용하였고, 급격한 변화를 추정하기 위해 추가적으로 적응형 기법(Adaptive rule)인 Adaptive Jumping Method를 두 필터 모두에 적용하였다. 적용 결과 두 오리지날 필터 모두 급격한 손상 시점과 정도를 파악하지 못하였고, 적응형 기법을 반영하였을 경우에만 시점 파악이 가능하였다. 하지만, 여전히 손상 정도를 정확히 파악하지 못하였고, 두 방법 모두 제안된 적응형 기법을 새로이 조정하였을 경우에 정확한 추정이 가능함을 확인하였다. 최종적으로 계산시간을 고려하였을 때, 새로운 형태의 적응형 기법을 적용한 UKF 사용을 제안하는 것으로 비교 검토를 수행하였다.

In this study, advanced unscented Kalman filter (UKF) and particle filter (PF) implementations are introduced and applied to perform system identification (SI) for sudden structural damage induced by seismic loading. These two methods are then compared to validate their applicability to SI tasks. For this validation, the Bouc- Wen model is used to simulate the nonlinear shear-building response, and an adaptive rule (i.e., an adaptive tracking method) is applied to the two filter methods to improve their tracking performance during sudden changes in system properties. When the original UKF and PF are applied to an earthquake scenario, both methods fail to estimate the damage initiation time and post-damage parameter values. After applying the adaptive tracking method, it is found for both methods that although the occurrence time is identified, the estimation of the damage state is still not accurate. To improve the accuracy, an adjusted adaptive tracking method is applied, and the two methods then derive accurate estimates. Finally, when considering the computation time, UKF is promoted as a better choice for practical applications, provided that a proper adaptive tracking method is implemented.

키워드

과제정보

본 연구는 국토교통부/국토교통과학기술진흥원의 지원으로 수행되었습니다(과제번호: RS-2021-KA163162).

참고문헌

  1. Baber, T.T., Noori, M.N. (1984) Random Vibration of Pinching Hysteretic Systems, J. Eng. Mech., 110(7), pp.1036~1049. https://doi.org/10.1061/(ASCE)0733-9399(1984)110:7(1036)
  2. Baber, T.T., Wen, Y.K. (1981) Random Vibration Hysteretic, Degrading Systems, J. Eng. Mech. Div., Am. Soc. Civil Eng., 107(6), pp.1069~1087. https://doi.org/10.1061/JMCEA3.0002768
  3. Candy, J.V. (2006) Bayesian Signal Process: Classical, Modern, and Particle Filtering Methods, John Wiley & Sons, Inc.: Hoboken, NJ, USA.
  4. Chatzi, E.N., Smyth, A.W. (2009) The Unscented Kalman Filter and Particle Filter Methods for Nonlinear Structural System Identification with Non-Collocated Heterogeneous Sensing, Struct. Control & Health Monit., 16, pp.99~123. https://doi.org/10.1002/stc.290
  5. Foliente, G.C., Singh, M.P., Noori, M.N. (1996) Equivalent Linearization of Generally Pinching Hysteretic, Degrading Systems, Earthq. Eng. & Struct. Dyn., 25, pp.611~629. https://doi.org/10.1002/(SICI)1096-9845(199606)25:6<611::AID-EQE572>3.0.CO;2-S
  6. Fujimura, K., Der Kiureghian, A. (2007) Tail-Equivalent Linearization Method for Nonlinear Random Vibration, Probab. Eng. Mech., 22, pp.63~76. https://doi.org/10.1016/j.probengmech.2006.08.001
  7. Kalman, R.E. (1960) A New approach to Linear Filtering and Prediction Problems, J. Basic Eng.-T. ASME, 82(1), pp.35~45. https://doi.org/10.1115/1.3662552
  8. Kim, M. (2019) Parameter Identification of Structures under Earthquake Excitations Using Adaptive Particle Filter and Ensemble Learning Method, Seoul National University, Master Thesis.
  9. Lee, S.-H., Kim, M., Yi, S.-R. (2024) Identification of Sudden Structural Damage using Unscented Kalman Filter with Enhanced Algorithmic Adaptive Jumping Method, submitted.
  10. Lee, S.-H., Song, J. (2017) System Identification of Spatial Distribution of Structural Parameters Using Modified Transitional Markov Chain Monte Carlo Method, J. Eng. Mech., 143(9), p.04017099.
  11. Lee, S.-H., Song, J. (2020) Regularization-Based Dual Adaptive Kalman Filter for Identification of Sudden Structural Damage Using Sparse Measurements, Appl. Sci., 10(850). DOI: 10.3390/app10030850.
  12. Lee, S.-H., Yi, S.-R., Lee, J.H. (2023) Unscented Kalman Filter with Multiple Sigma Points for Robust System Identification of Sudden Structural Damage, J. Comput. Struct. Eng. Inst. Korea, 36(4), pp.233~242. https://doi.org/10.7734/COSEIK.2023.36.4.233
  13. Liu, M., Zang, S., Zhou, D. (2005) Fast Leak Detection and Locations of Gas Pipelines Based on an Adaptive Particle Filter, Int. J. Appl. Math. Comput. Sci., 15, pp.541~550.
  14. Nguyen, V.H., Lee, S.-H., Lee, J.H. (2022) Application of the Unscented Kalman Filter to Estimate the Material Properties of a Layered Half-Space, Int. J. Appl. Mech., 15(2), p.2350005.
  15. Noori, M.N., Choi, J., Davoodi, H. (1986) Zero and Nonzero Mean Random Vibration Analysis of a New General Hysteresis Model, Probab. Eng. Mech., 1(4), pp.192~201. https://doi.org/10.1016/0266-8920(86)90012-3
  16. Papakonstantinou, K.G., Amir, M., Warn, G.P. (2020) A Scaled Spherical Simplex Filter (S3F) with a decreased n + 2 sigma points set size and equivalent 2n + 1 Unscented Kalman (UKF) accuracy, Mech. Syst. & Signal Process., 163(2022), p.107433.
  17. Song, J., Der Kiureghin, A. (2006) Generalized Bouc-Wen Model for Highly Asymmetric Hysteresis, J. Eng. Mech., 123(6), pp.610~618. https://doi.org/10.1061/(ASCE)0733-9399(2006)132:6(610)
  18. Vaiana, N., Sessa, S., Marmo, F., Rosati, L. (2018) A Class of Uniaxial Phenomenological Models for Simulating Hysteretic Phenomena in Rate-Independent Mechanical Systems and Materials, Nonlinear Dyn., 93(3), pp.1647~1669. https://doi.org/10.1007/s11071-018-4282-2
  19. Van der Merwe, R., Wan, E. (2001) The Square-Root Unscented Kalman Filter for State and Parameter-Estimation, 2001 IEEE International Conference on Acoustics, Speech, and Signal Processing, Proceedings (Cat. No.01CH37221). Salt Lake City, UT, USA, 7-11 May 2001, 6, pp.3461~3464.
  20. Wang, C.-H., Wen, Y.K. (1998) Reliability and Redundancy of Pre-Northridge Low-Rise Steel Building Under Seismic Excitation, Rep. No. UILU-ENG-99-2002, Univ. Illinois at Urbana-Champaign, Champaign, Ill.
  21. Yang, J.N., Lin, S., Huang, H., Zhou, L. (2006) An Adaptive Extended Kalman Filter for Structural Damage Identification, Struct. Control. Health Monit., 13, pp.849~867. https://doi.org/10.1002/stc.84
  22. Yi, S.-R., Song, J. (2018) Particle Filter Based Monitoring and Prediction of Spatiotemporal Corrosion Using Successive Measurements of Structural Responses, Sens., 18(3909). DOI: 10.3390/s18113909.
  23. Yi, S.-R., Wang, Z., Song, J. (2017) Bivariate Gaussian Mixture-based Equivalent Linearization Method for Stochastic Seismic Analysis of Nonlinear Structures, Earthq. Eng. Struct. Dyn., 2018(47), pp.678~696.