Loading...

Structural Health Monitoring using Bayesian Optimization of the finite element model of structures and Kalman filter

Sadegh, Alireza | 2023

96 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56719 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Bakhshi, Ali
  7. Abstract:
  8. With confidence in the recorded observations, the RLS method no longer estimates the recorded measurements by sensors, i.e. the displacement and speed of the floors, and only estimates the parameters. In contrast, in the EKF method, in addition to estimating the structure's parameters, a more precise estimation of the observations recorded by the sensors has been done by accepting the noise in the recorded observations. These methods, which are based on the Bayesian updating, investigate the two primary sources of uncertainty in a problem: a) measurement noise or observation noise, and b) process noise, which includes modeling errors. In these methodologies, the unknown system parameters, such as the stiffness of the floors, are added to the system's state vector. During each time step, multiple procedures are carried out. Initially, the initial state of the system is estimated utilizing the equation of motion, the posterior values of the unknown parameters, and the previous input excitation. The input excitation is then applied in the current time step, and the system state vector and parameter values are then updated based on the measured response of the current time step. Throughout the stimulation, the prediction and update operations are repeated for each time step. In order to evaluate the accuracy of the two methods of the Extended Kalman filter and RLS, the state vector was estimated using Bayesian updating, and the ability of these three methods to estimate similar state vectors was compared. In this thesis, the estimated values of the parameters of the state vector have been updated in three distinct probabilistic ways, based on the observation that it is actually the time history of the structure's response to an arbitrary vibration plus some noise. In fact, in these three methodologies, the parameters are determined so that they have the smallest possible variance. The calculated response of a four-story shear building contaminated with noise, and the structure's parameters, as well as the response without noise, have been identified by these three methods in order to validate them. In addition, the results of this study demonstrate the cost, accuracy, and measurement accuracy differences between each of these three methods. Finally, the ability of constrained Extended Kalman filter to estimate the non-linear parameters of the Bouc-Wen model of a bridge pier was investigated using an experimental example, and the estimated and actual hysteresis curves were compared
  9. Keywords:
  10. Structural Health Monitoring ; Kalman Filters ; Least Squares Method ; Extended Kalman Filter ; Bouc-Wen Dynamic Model ; Bayesian Filtering ; Structural Estimation ; Structural Parameters ; Recursive Least Square Estimation

 Digital Object List

 Bookmark

No TOC