Multipoint Deformation Safety Monitoring Model for Concrete Arch Dams Based on Bayesian Model Selection and Averaging

Abstract

The deformation properties of concrete arch dams are affected by the synergistic effects of multiple factors, featuring strong, multidimensional spatialtemporal evolution and distribution characteristics. This paper proposes a zoned safety monitoring model for arch dam deformation based on spatialtemporal similarity and model optimization to evaluate the deformation safety state of arch dam structures. First, zoned clustering of the deformation monitoring points at different locations of an arch dam was performed using a panel data multi-index clustering method to determine the deformation laws at different positions of the dam. Next, multipoint comprehensive displacements of the deformation properties of each zone were extracted using principal component analysis to extract the uniform deformation law of the monitoring point in each zone. Finally, we adopted Bayesian model selection (BMS) and Bayesian model averaging (BMA) for the regression model set, considering the uncertainty of the model. The engineering case study showed that BMA yielded robust and effective prediction results for the deformation of the arch dam. The analysis of the zoned deformation mechanism indicated that the deformation of the arch dam followed the general rule. The temperature component of the arch dam was mainly reflected in the middle with a hysteresis effect, and the time-dependent component was evident in both sides of the dam shoulder. The arch dam deformation safety monitoring model proposed in this study has strong robustness and interpretability, which can provide valuable technical support for analyzing the evolution of arch dam deformation properties.