边坡非线性位移的分段最小二乘-时间序列分析

Segment least squares-time series analysis of slope nonlinear displacement

利用最小二乘法对位移时间序列进行分段拟合以提取时间序列的趋势项,从而将非稳定时间序列稳定化,对去除趋势项的稳定时间序列即残差进行常规AR(p)时序分析。结合滚动预测方法,建立边坡失稳预测的叠合模型,以长江三峡某边坡为例,对该模型进行的检验表明:新模型的预测精度较高、实时可靠。

Using least-square procedure to extract trend term of time series for piecewise fitting time series of displacement for transforming unstable time series into stabilization, the authors made a conventional analysis on stable time series of eliminating trend term by using a model named AR （p）. The congruence model was established with rolling prediction method, which can forecast slope instability and has been verified by a side slope in the Three Gorges. The result shows that the new model is higher in prediction accuracy and reliable in real time.