温度变形效应认知不确定性影响下高拱坝位移置信区间的预测方法

Prediction method of confidence interval for displacement of high arch dams under influence of cognitive uncertainty of temperature deformation effects

为解决传统预测模型未考虑拱坝变形因果关系的不确定性,导致所建立的位移静态置信区间缺乏合理的因果机理,提出了温度变形效应认知不确定性影响下高拱坝位移置信区间的预测方法。采用动态时间规整法衡量坝体温度测点之间的时间序列相似性,构建最小相似性实测温度变形因子筛选准则,基于支持向量机构建不放回采样和正交试验设计采样的机器学习模型,通过统计多模型预测值的分布规律来拟定动态变化的置信区间。以锦屏一级拱坝为例的预测结果表明该预测方法及筛选准则可有效实现高拱坝最具表征性温度测点的优选,基于最小相似性实测温度因子的正交试验设计误差小、建模效率高,所得位移预测置信区间更符合因果机理;高拱坝的最优实测温度因子组合是动态变化的。

Traditional prediction models are difficult to consider the uncertainty of causality of arch dam deformation,leading to the lack of causality mechanism in the established static displacement confidence interval.To solve the difficulty,a prediction method which considers the influence of cognitive uncertainty of temperature deformation effects,was proposed for the confidence interval of high arch dams’displacement.The dynamic time warping method was used to calculate the similarity between the measured time series of multiple temperature monitoring points on the dam body,and a criterion was established to select the measured temperature deformation factors with the minimum similarity.Multiple support vector machine models were then established,in which the orthogonal experimental sampling of selected temperature factor was used as inputs.The distribution of multi-model predicted displacements was analyzed and used to establish a dynamic changed confidence interval.Research results of the Jinping I Arch Dam indicate that the proposed methods and criteria can effectively optimize the most representative dam temperature monitoring points.The minimum similarity temperature factor-based orthogonal experiment design has small error and high modeling efficiency.The confidence interval of displacement prediction is more consistent with the causal mechanism and the optimal measured temperature factor combination of the high arch dam is dynamic.