.png){kind=icon}
.png){kind=icon}
Return
An efficient Expectation-Maximization algorithm for Bayesian operational modal analysis with physics-data fusion model
.png){kind=icon}
PRE
AI
DOI：10.1016/j.ymssp.2025.113144.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Operational Modal Analysis (OMA) is widely used in Structural Health Monitoring (SHM), to identify structural dynamic properties which are critical for evaluating structural integrity and safety. Considering the interference of environmental effects, physics-data fusion models are introduced to enhance the robustness of parameter identification. However, these models usually couple ambient and dynamic variables and require large amounts of monitoring data, resulting in high computational costs. To address these issues in fusion models, this paper proposes a novel method that allows the environmental impact to be addressed separately, reducing model dimensionality. The Taylor approximation is then applied to derive an approximate distribution during the inference process, enabling the efficient identification of modal parameters as well as the hyperparameters of the Gaussian Process model in a computationally feasible manner. The Expectation-Maximization (EM) algorithm is used as the iterative method to facilitate the computational feasibility and robustness of the proposed method. Validation using synthetic data and real-world monitoring data from a long-span cable-stayed bridge demonstrates that the proposed method outperforms a previous approach, FM-MCI, by achieving superior computational efficiency and delivering more consistent identification results. Furthermore, the method effectively captures the correlation between environmental effects and structural responses, offering a robust solution for modal parameter identification under varying environmental conditions.
Keywords:
Operational Modal Analysis
Structural Health Monitoring
Physics-data fusion
Gaussian Process
Expectation-Maximization algorithm
Journal
IF:
8.9
Papers:
1.3W
Citations:
6.6W
Organization
View more organizations