考虑多重共线性影响的特高拱坝时空监控模型

为了解决特高拱坝时空监控模型中因子数目众多、因子之间存在多重共线性以及各测点之间存在空间关联性的问题，基于大坝变形原型监测资料，采用核独立分量分析(KICA)方法提取独立分量，将多个测点信息转化为少数几个综合指标；将提取的独立分量代入利用灰狼优化(GWO)算法优化的支持向量机(SVM)模型，对特高拱坝空间测点进行回归预测，构建了KICA-GWO-SVM特高拱坝时空监控模型。工程实例分析结果表明，KICA-GWO-SVM特高拱坝时空监控模型与多元回归模型、BP模型及SVM模型相比，其非线性表达能力强且性能良好，能够降低多重共线性对大坝变形监测的影响，对特高拱坝变形序列的拟合与预测精度高，可以更加准确全面地表征大坝整体的时空变形性态。

Spatio-temporal monitoring model for super-high arch dams considering multicollinearity effects

In order to solve the problem of large number of factors, multicollinearity between factors and spatial correlation among measuring points in the spatio-temporal monitoring model of super-high arch dams, the kernel independent component analysis (KICA)method was used to extract independent components based on the prototype monitoring data of dam deformation, and the information of multiple measuring points was transformed into a few comprehensive indicators. On this basis, the support vector machine (SVM) model was optimized by using the advantages of grey wolf optimization (GWO) algorithm with good convergence speed and solution accuracy in parameter optimization. The extracted independent components were substituted into the SVM model optimized by GWO algorithm, the regression prediction on the spatial measuring points of the super-high arch dam was performed, and the KICA-GWO-SVM spatio-temporal monitoring model for super-high arch dams was constructed. Analysis of engineering examples shows that, compared with the multiple regression model, BP model and SVM model, the KICA-GWO-SVM spatio-temporal monitoring model has strong nonlinear expression ability and good performance, which can reduce the influence of multicollinearity on dam deformation monitoring, and the fitting and prediction accuracy of the deformation sequence of super-high arch dams is excellent. It can more accurately and comprehensively grasp the overall spatio-temporal deformation behavior of the dam.