Development and evaluation of acceleration-based earthquake damage detection and classification algorithms
Abstract/Contents
- Abstract
- Structural Health Monitoring has a significant impact, both economic and in terms of life safety, in a wide variety of industries such as civil, manufacturing and aerospace. The timely detection of defects can prevent economic losses due to malfunctioning equipment or infrastructure and avoid potential life-threatening failures. This is especially true in the aftermath of an extreme loading event such a major earthquake where the identification of the presence, and potentially extent, of damage in a portfolio of structures can prevent further losses and allow for more informed decision making with regards to the recovery of the affected region. This dissertation presents three damage detection and classification algorithms that aim to provide reliable information on the damage state of a civil structure within a matter of minutes following an earthquake. While each algorithm corresponds to a different use case, a common objective is to keep the algorithms as simple and data driven as possible, which allows the application of the algorithms to different structure and loading types. The first algorithm presents a methodology for estimating the residual displacement of structures. It builds on previous work and utilizes stationary and multi-dimensional acceleration measurements to calculate the rotation of the sensor with respect to the direction of gravity and estimate the residual displacement based on the calculated rotation. The proposed algorithm utilizes the measurements from several sensors and estimates the residual displacement along the height of the structure using only the sensor measurements and locations as input. The optimal configuration of the algorithm with respect to parameters such as the number and location of sensors is determined, and the accuracy of the algorithm is evaluated, both using Monte Carlo simulation. In order to provide further validation of the algorithm, the effect of sensor noise and measurement error on the accuracy of the algorithm is evaluated, and recommendations on the minimum number of samples required to obtain a reliable measurement are provided. A series of Damage Sensitive Features based on the Continuous Wavelet Transform of acceleration measurements are developed. The proposed features take into account both the input excitation and the output structural response. A mathematical formulation for the combination of the input and output signals is presented, and methodologies for the extraction of the features are outlined. The correlation of the features with the extent of damage is established via frequently used damage metrics such as hysteretic energy. A damage detection scheme based on the proposed features is presented that utilizes ambient vibration measurements for establishing an undamaged baseline. The damage detection scheme is also validated using Monte Carlo simulation. Finally, a damage classification scheme is proposed where established damage indices are utilized to classify the damage sustained in different categories depending on the extent. The damage classification scheme is also validated through Monte Carlo simulation. A statistical model for the wavelet coefficients of the acceleration structural response is presented. The fundamental assumption behind the proposed model is that the wavelet coefficients at each time sample are transformed realizations of a Gaussian Process that depends only on the damage state of the structure. The model parameters are estimated using Maximum Likelihood Estimation and a systematic methodology for the implementation is proposed and validated. The statistical model is applied to experimental data as a proof of concept and a damage detection scheme based on statistical hypothesis testing is proposed. The capabilities of the rotation algorithm and the Gaussian Process statistical model are illustrated in an actual sensor deployment. These algorithms are applied to the data that were acquired from a series of shake table tests conducted on two steel frames at the National Taiwan University. The results from the algorithm applications are shown and compared to the actual damage state of the specimens.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2015 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Balafas, Konstantinos |
|---|---|
| Associated with | Stanford University, Department of Civil and Environmental Engineering. |
| Primary advisor | Kiremidjian, Anne S. (Anne Setian) |
| Thesis advisor | Kiremidjian, Anne S. (Anne Setian) |
| Thesis advisor | Baker, Jack W |
| Thesis advisor | Rajagopal, Ram |
| Advisor | Baker, Jack W |
| Advisor | Rajagopal, Ram |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Konstantinos Balafas. |
|---|---|
| Note | Submitted to the Department of Civil and Environmental Engineering. |
| Thesis | Thesis (Ph.D.)--Stanford University, 2015. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2015 by Konstantinos Balafas
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...