基于VMD的多尺度变量提取法在混凝土坝变形预测中的应用

VMD-based Multi-scale Variables Extraction Model for Deformation Prediction of Concrete Dam

大坝变形通常呈非线性、非稳定特征,利用可靠的数据驱动模型分析大坝变形变化趋势具有重要意义。为此,提出了一种基于变分模态分解(VMD)的多尺度变量提取大坝变形预测模型,首先利用VMD将变形原始序列分解为若干具有不同频率的子序列,以降低原始序列的非平稳性;然后引入初始变量集,并通过平均影响值-极限学习机模型(MIV-ELM)分别提取出与变形子序列相关度较高的变量;最后构建ELM模型对大坝变形进行多尺度预测。针对某混凝土坝水平位移的分析结果表明,提出的模型能有效减小非平稳性对预测结果的影响,实现大坝变形的准确预测。

Dam deformation usually presents nonlinear and nonstationary characteristics, and it is important to analyze the variation trend of dam deformation with reliable data-driven models. In this paper, a multi-scale variable extraction dam deformation prediction model based on variational modal decomposition(VMD) is proposed. Firstly, the original deformation series is decomposed into several subseries with different frequencies by the VMD to reduce the nonstationarity of the original series. Then the initial set of variables is introduced and the variables that have higher correlation with the deformation subseries are extracted by the mean impact value-extreme learning machine model(MIV-ELM). Finally, the ELM models are constructed for multi-scale prediction of dam deformation. The analysis results for the horizontal displacement of a concrete dam show that the proposed model can effectively reduce the impact of nonstationarity on the prediction results, and achieve accurate prediction of dam deformation.