基于聚类分区和MO-LSSVR的高拱坝变形预测模型

为解决传统单测点监控模型未考虑多测点间的内在关联，难以反映高拱坝变形区域分布特征的问题，提出了基于聚类分区和多输出最小二乘支持向量回归机(MO-LSSVR)的高拱坝变形预测模型。模型基于测点之间的复合相似性指标，借助层次凝聚聚类(HAC)算法实现空间测点的聚类分区，再利用融合测点关联特性的MO-LSSVR对分区内多测点进行建模。工程实例验证表明，模型聚类分区结果与坝体变形空间分布特征较吻合，具有较高的准确性和稳健性，为从多测点关联维度预测坝体变形和监控大坝整体安全性态提供了一种新方法。

Deformation prediction model of a high arch dam based on clustering and MO-LSSVR

To solve the problem that the internal correlation between multiple measuring points cannot be considered by traditional single measuring point monitoring models, which is difficult to reflect the regional characteristics of high arch dam deformation in space. A high arch dam deformation prediction model based on clustering partition and a multi-output least square support vector regression machine (MO-LSSVR) algorithm is proposed. Based on the composite similarity index between the measuring points, the clustering partition of spatial correlation measuring points is realized by hierarchical agglomerative clustering (HAC) algorithm. The MO-LSSVR algorithm integrating the correlation characteristics of measuring points is then used to model the points in the partition. The engineering example results show that the clustering partition results are consistent with the spatial distribution characteristics of dam deformation. The MO-LSSVR model based on the reasonable partition results has high accuracy and robustness, which provides a new method to accurately predict the dam deformation and monitoring the overall safety state of the dam from the multi-measuring points correlation dimension.