.png){kind=icon}
返回
A two-step failure identification approach using a stochastic optimization-based ensemble learning model for beams without pristine data
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engstruct.2025.1202533.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Most vibration-based structural health monitoring techniques evaluate structural conditions by relying on data from both intact and damaged states. Data from a healthy structure is highly valuable for detecting damage. While it is often abundant in newly monitored structures, it may be limited or unavailable in existing bridges, where monitoring usually starts after years of service-related deterioration. Additionally, noise is inherently present in measured data. Therefore, combining data from both healthy and deteriorated states introduces a higher risk of inaccuracies. To address this, the present study proposes a two-step approach for damage identification that relies solely on the damaged state. First, a modified mode shape-based damage index is employed to localize failures using only the available damaged-state data. Then an ensemble learning model improved by a stochastic optimization process is implemented to estimate damage extent. In this study, damage scenarios involving single and multiple cuts are examined instead of assuming stiffness loss through Young's modulus reduction. A flat beam, developed based on experimental data, and a concrete T-girder bridge are utilized to validate the applicability of the proposed approach. Numerical results indicate that damage locations can be identified using the first four modes, even under noisy conditions. Furthermore, optimization-based ensemble learning models demonstrate greater efficiency and stability compared to traditional models in extent estimation through cut length.
Keyword:
Dynamic properties
Gapped smoothing method
Optimization
Ensemble learning model
期刊
IF:
6.4
论文数:
2.1W
被引数:
8.7W
机构
引用论文
暂无论文信息