Although statistical models are efficient in most cases to analyze concrete dam displacements, these models are built on several hypotheses, leading to uncertainties especially for special periods. The special statistical models, improving estimations of the non‐stationary thermal and the non‐monotonic time‐dependent effects, are proposed for the displacements of high arch dams during their initial impoundment periods in this paper. The hierarchical clustering on principal component analysis is developed to divide thermometers into groups and to choose representative thermometers or identify major principal components on measured temperature data to represent the non‐stationary thermal effect on dam's displacements. The non‐monotonic formula for the time‐dependent deformation, emphasizing the creep and its restoration of dam concrete and its surrounding rock, is derived and further simplified when the reservoir water level exhibits an evident periodicity. Then, two improved statistical models accounting for these non‐stationary thermal and non‐monotonic time‐dependent effects are proposed. The proposed statistical models with clear physical meanings are applied to investigate the measured displacements of the Xiluodu arch dam. Model performance comparisons indicate that the proposed models have significant improvement in fitting precision and prediction ability over the traditional and more recent models. Model results confirm the influence of reservoir thermal stratification and concrete temperature rise on the thermal displacement, and the non‐monotonic effect on the time‐dependent displacement. The proposed models yield to a better identification of the deformation mechanism for high arch dams during their initial impoundment periods.

中文翻译：

---

蓄水初期高拱坝变形的统计模型

尽管统计模型在大多数情况下可以有效地分析混凝土大坝的位移，但这些模型是建立在几个假设的基础上的，从而导致不确定性，尤其是在特殊时期。针对高拱坝初始蓄水期的位移，提出了一种特殊的统计模型，以改进对非稳态热和非单调时变效应的估计。开发了基于主成分分析的层次聚类，以将温度计分为几类，并选择具有代表性的温度计或在测得的温度数据上识别主要的主要成分，以表示对大坝位移的非稳态热影响。与时间有关的变形的非单调公式，当水库水位表现出明显的周期性时，可以推导并进一步简化坝混凝土及其周围岩石的蠕变及其修复。然后，提出了两个改进的统计模型，这些模型考虑了这些非平稳的热和非单调的时间依赖性效应。提出的具有清晰物理意义的统计模型被用于调查溪洛渡拱坝的实测位移。模型性能比较表明，与传统模型和最新模型相比，所提出的模型在拟合精度和预测能力方面有显着提高。模型结果证实了储层热分层和混凝土温升对热位移的影响，以及非单调对随时间变化的位移的影响。