Bayesian operational modal analysis considering environmental effect

In long term structural health monitoring (SHM), Operational modal analysis (OMA) techniques are commonly used to identify dynamic properties of structures. However, conventional OMA methods do not consider the effect of environment, which may cause bias in the identified dynamic properties. Some machine learning models (such as Gaussian process model) can quantitatively describe the effect of environment on structure, but the associated parameters do not have physical meaning. In order to address these problems, this paper proposes a novel fusion model that combines structural dynamics model and data-driven model to identify dynamic properties with the effect of environment considered. The proposed method adopts a probabilistic framework that can incorporate the uncertainties and correlations in the data. The proposed method is validated using synthetic data and applied to the monitoring data of two full-scale bridges. The results show that it can effectively identify the modal parameters while describing the correlation between the environmental variations and the structural responses. The proposed method can facilitate robust and reliable identification of dynamic properties with structures under environmental effect.