ISSN 1000-3665 CN 11-2202/P

    基于递归图理论的沉降时间序列可预测分析

    Predictability analysis of the settlement time series based onthe theory of recurrence plots

    • 摘要: 在实际沉降监测中,由于不确定性因素的影响,导致基于实测沉降数据的工后沉降预测误差较大。采用递归图和递归定量分析方法对沉降时间序列的可预测性进行分析,选择出预测效果好的沉降时间序列,进行准确的工后沉降预测。通过对无砟轨道整体道床的沉降时间序列进行可预测性分析得出:(1)沉降时间序列具有混沌特性;(2)沉降时间序列的递归图颜色分布越规律、色彩深浅变化越均匀,沿着主对角线出现的白色区域越小,沉降时间序列的可预测性越强;(3)通过提取递归定量指标随时间变化的曲线,可以直接确定预测时间起点;(4)利用递归图求取二阶任意熵作为沉降时间序列的可预测步长因子,可推算出预测效果最佳的沉降时间序列。

       

      Abstract: Owing to the influence of uncertainties, the prediction of settlement after construction is inaccurate in the actual settlement monitoring. The predictability of the settlement time series is analyzed by using the recurrence plots and recurrence quantification analysis method. A settlement time series with high predictability is chosen to make an accurate prediction of settlement after construction. The predictability analysis of the non-ballasted-track integrated bed settlement time series show that (1) the settlement time series is chaotic; (2) the more regular of the color distribution and the more uniform in the shade, the smaller white areas appear along the main diagonal, and the stronger the predictability of settlement time series; (3) quantitative analyses of the recurrence plots on this basis may extract the recursive quantitative index curves over time and the immediate determining starting point of prediction time; (4) the recurrence plots are used to get the arbitrary entropy of the second order as the predictable time length of the settlement time series to comprehensively extrapolates the best settlement time of the forecast effect.

       

    /

    返回文章
    返回