融合PSO-SVM的混凝土拱坝多测点变形监控混合模型

Multi-point Hybrid Model Based on PSO-SVM for Concrete Arch Dam Deformation Monitoring

针对混凝土拱坝单测点变形监控模型难以合理表征拱坝空间变形场协同响应特性以及传统回归方法诠释环境量与大坝变形间的复杂函数关系具有明显局限性问题,提出了融合粒子群算法优化与支持向量机(particle swarm optimization-support vector machine,PSO-SVM)的混凝土拱坝多测点变形监控混合模型。基于单测点变形监控模型构建方法,引入空间坐标并利用有限元方法计算水压分量,进而借助PSO-SVM良好的非线性处理能力对环境量与大坝变形序列进行建模和预测,从而构建了融合PSO-SVM的混凝土拱坝多测点混合模型。工程实例分析表明,所建模型具有较好的多测点变形性能分析能力,较单测点统计模型具有良好的拟合及预报精度,可有效反映大坝服役的整体安全性态。此外,所提理论和方法经一定的改进和拓展,亦可推广应用于其他水工建筑物性态安全监控模型的预报分析。

Objectives: It is difficult for the single-point deformation monitoring model for concrete arch dam to reasonably characterize the cooperative response characteristics of the arch dam spatial deformation field, and the traditional regression method has obvious limitations to interpret the complex function relationship between environmental quantity and dam deformation. Methods: On the basis of the analysis of traditional single-point deformation monitoring model construction method,a multi-point hybrid model based on support vector machine-particle swarm optimization(PSO-SVM) for concrete arch dam deformation monitoring is proposed by introducing spatial coordinates, and using the finite element method to calculate the water pressure component. SVM optimized by PSO is used for analysis and prediction of dam deformation sequence with the aid of PSO-SVM’s good nonlinear processing capabilities. Thus, a multi-point hybrid model based on PSO-SVM for concrete arch dam deformation monitoring is established.Results: To verify the effectiveness of the proposed PSO-SVM-based multi-point hybrid model, a single-point statistical model, and an SVM-based multi-point hybrid model are established at the same time. The mean absolute error and root mean square error values of the established model are lower than those of other models,and the determination coefficient R2 is closer to 1.Conclusions: The engineering example analysis shows that the proposed model has better ability to analyze the deformation performance of multi-point, and has better fitting and prediction accuracy than the statistical model of single-point, and can effectively reflect the overall behavior of the dam. The proposed theory and method can be generalized and applied to other dam behavior safety monitoring models.