Displacement observation data-based structural health monitoring of concrete dams: A state-of-art review

Abstract

Identifying and evaluating the structural state through massive monitoring data is one of the key issues in the structural health monitoring of concrete dams. Data-driven models play an important role in interpreting and predicting the deformation behaviour of concrete dams, and there are a large number of statistical models, hybrid models and machine learning models, but the used modelling factors and methods in each case are different. In contrast to existing review papers focused on dam health monitoring, this paper provides a detailed review of the research status of monitoring models only for the displacement of concrete dams, and contains three aspects: optimization of modelling factors, improvement of modelling methods, and monitoring model-based structural health diagnosis. In the first part, the paper summarizes the purpose, ideas, implementation methods and effects of adding new modelling factors and optimizing temperature deformation modelling factors. Then, some issues related to the performance of machine learning models, including parameter optimization, kernel function selection, methods to alleviate overfitting, causal interpretation ability exploring and combination modelling strategy, are discussed in detail. The measured displacement-based monitoring index and real-time risk rate of concrete dams are analyzed. Furthermore, models and methods for diagnosing the spatial deformation behaviour of super-high concrete dams are outlined. In the future, in addition to using advanced mathematical methods to establish displacement monitoring models, it is recommended to strengthen the integration of mathematical models with the deformation mechanism of concrete dams, and improve the rationality and universal applicability of the models, rather than just comparing the prediction performance on a specific case.