Loading...

Rapid Seismic Damage Estimation using Bayesian System Identification

Rostami, Parisa | 2023

85 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 56610 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Mahsuli, Mojtaba; Ghahari, Farid
  7. Abstract:
  8. This dissertation presents a comprehensive probabilistic framework based on system identification for rapid seismic damage assessment of buildings at regional scale. Given the large number of buildings within a region and the need for rapid damage detection, this research uses simplified models with low computational cost to model each building. For this purpose, stochastic filters are employed as system identification tools. In the first step, a continuous linear model consisting of a Timoshenko beam in combination with the extended Kalman filter is utilized. This model is subjected to joint state-parameter identification under both fixed- and flexible-base conditions. Additionally, the flexible-base model is identified as output-only, leading to the joint identification of the state, system and input excitation. The proposed method is first validated using a synthetic example and then applied to real data recorded at the Millikan Library at the California Institute of Technology. Due to uncertainties inherent in real-world data, the proposed framework is implemented in two steps. The first step focuses on identifying structural parameters, while the second step aims at identifying substructure parameters and input excitation. As the true values of parameters are unknown in real-world applications, the proposed framework provides a multi-start approach, conducting numerous identifications with random initial values within a high error range, up to 200%. The convergence of these identification analyses determines the true value of parameters. The results demonstrate that the proposed model exhibits high identifiability and can be identified with high accuracy and low computational cost even using a very small number of sensors, i.e., two sensors at the roof and base of the structure. This is achieved even under high noise levels, up to a signal-to-noise ratio of 10. The convergence of 50 identification analyses with random initial values to results whose coefficient of variations are less than 1% underscores the robustness of the proposed approach. As structures exhibit nonlinear behavior under moderate to severe earthquakes, a simplified nonlinear model is used in the next step for predicting the dynamic behavior of buildings. For this purpose, a discrete flexural-shear model is utilized, in which each story is modeled with nonlinear flexural and shear springs. This model is integrated into the unscented Kalman filter algorithm as a parameter estimation tool. Furthermore, an efficient approach to calculate initial estimates of identification parameters is provided merely based on observable building information. The proposed framework is validated through two three- and 12-story buildings using noisy measurements generated from a detailed finite element model. The results demonstrate that the proposed model accurately predicts the nonlinear behavior of structures, even with a limited number of sensors, i.e., up to three sensors at the roof, mid-height, and the base of the structure. Finally, the identified flexural-shear model is used to determine the damage level of both structural and nonstructural components and assess the seismic loss of the structure. This involves computing the probability of various damage states for these components based on the results of the proposed method and comparing them with those from the detailed finite element model. The results show that the proposed framework is capable of estimating the extent of damage and seismic loss of various structures with reasonable accuracy
  9. Keywords:
  10. Simplified Method ; Extended Kalman Filter ; Unscented Kalman Filter ; Nonlinear Behavior ; Damage Assessment ; Bayesian Identification ; Seismic Damage ; Rapid Damage Detection

 Digital Object List

 Bookmark

No TOC