基于相空间重构及PLS法的冻土路基变形预测

为了正确预测多年冻土路基变形,将多年冻土路基变形单变量混沌时间序列进行相空间重构,依重构空间嵌入维数38在多维空间中定义自变量、因变量和样本数,利用偏最小二乘法对所构造的自变量进行主成分提取,建立路基变形预测模型,借助多种评判指标进行模型精度分析,并绘制了预测值与实测值散点图。对比分析结果表明所有样本点都集中在散点图对角线附近,相关系数达0.889 4,预测值与实测值最大相对误差为9.75%,说明建立的预测模型合理可信,利用该预测模型进行多年冻土路基变形预测可行。

Permafrost subgrade deformation forecast based on phase space reconstruction and partial least squares regression

In order to accurately forecast permafrost subgrade deformation, a single variable chaos time series of permafrost subgrade deformation was constructed in multidimensional phase space,some independent variables,function and sample numbers were defined according to embeded 38 dimensions of the space, the main components of independent variables were picked by using partial least squares regression method,a deformation forecast model was constructed,the precision of the model was analyzed by recurring multiplicate evaluation indices,and the comparative figure between forecast values and measure values was plotted.Analysis result shows that all sample points concentrate near the diagonal line of the figure,their correlation index is 0.889 4,their maximum relative error is 9.75%,which indicates that the model is reasonable and credible,can forecast subgrade deformation in permafrost regions.2 tabs,8 figs,10 refs.