An optimized zonal deformation prediction model for super-high arch dams

Abstract

The previous zonal deformation prediction models for super-high arch dams focus on the measured deformation law, which cannot capture the trend, periodic and fluctuating characteristics. This greatly limits their physical interpretation and practical applications. To address this issue, this paper proposes an optimized zonal deformation prediction model. First, the variational mode decomposition algorithm is applied to split displacement into trend, periodic and fluctuating components. Representative environmental and load factors are determined by the hierarchical clustering on principal components, and then also decomposed into the trend, low- and high-frequency components according to their physical meanings. Second, an optimized dynamic time warping (DTW) based on shape-based distance is employed to divide the displacement components of measured points into different zones. The centroid sequences are calculated to capture the shared characteristics of the corresponding deformation zones. The zonal data sets of the centroid sequences of displacement components and their strongly related influencing factor components are established. Third, the optimized zonal deformation prediction models using three machine learning algorithms (random forest, least squares support vector machines, and boosted regression tree) and the improved hydrostatic-thermal-time model are constructed. The effectiveness of the optimized model is verified using the measured data collected from the Xiluodu dam. The case study shows that the optimized model can well explain the spatio-temporal characteristics of the trend, periodic and fluctuating displacements. Affected by the geological condition, the radial displacement distribution is not completely symmetrical under high reservoir water level.