阶跃型滑坡综合变形预测及监测预警方法研究

大型库岸滑坡的长期变形在汛期降雨作用下呈现明显的周期性“阶跃式”陡增特征。针对阶跃型滑坡的变形特征,本文提出了一种多源数据“融合-预测-预警”的三步式滑坡监测预警方法：(1)“融合”,即基于经验模态分解法将多点位移监测数据分别分解为趋势项和周期项,采用加权值法分别融合不同监测点的趋势项和周期项位移得到融合趋势项和融合周期项序列,并将两者叠加得到滑坡体的现状综合变形时间序列;(2)“预测”,即引入“一个预测周期”概念,采用滑动多项式拟合法和随机森林算法分别对融合趋势项和融合周期项进行预测并叠加得到滑坡体的预测综合变形时间序列;(3)“预警”,即基于斜率变点分析方法搜索综合变形曲线的“稳定点”和“跃迁点”,确定稳定变形和加速变形区间的斜率,建立阶跃型滑坡的四级递进式分级预警模型,基于该预警模型对滑坡现状进行预警。以向家坝水库某滑坡体自动化位移监测数据为研究对象,采用本文所提方法对该滑坡进行了综合变形预测和监测预警,结果表明：综合变形时间序列可以整体反映滑坡的变形演化规律,且预测结果可靠,根据分级预警模型判断此滑坡体当前处于稳定变形阶段(Ⅰ级预警)。

Research on comprehensive deformation prediction and monitoring and early warning method for step-like landslide

The long-term deformation of the bank slope of the large reservoir presents an obvious periodical "step-type" steep increase under the action of rainfall in flood season.Aiming at the deformation characteristics of step landslide, a three-step landslide monitoring and early warning method for multi-source data is proposed in this study:(1) "Fusion"-The multi-point monitoring data are decomposed into trend term and period term based on empirical mode decomposition method, the trend term and periodic term displacements of different monitoring points are fused by weighting method, and the obtained trend term and periodic term are superimposed to obtain the current status of the integrated landslide deformation.(2) "Prediction"-The concept of "one prediction period" is introduced, the sliding polynomial fitting method and the random forest algorithm are used to predict the fusion trend term and the fusion period term respectively and superimpose them to obtain the predicted comprehensive deformation of the landslide.(3) "Early warning"-Based on the slope change point theory, the "stable point" and "transition point" of the comprehensive deformation curve are searched, and the average slope of the stable deformation and accelerated deformation interval is determined.Then, a 4-stage progressive hierarchical warning model of step-like landslide is established, and the current safety of the bank slope is warned.Taking the slope automatic displacement monitoring data of a bank slope of Xiangjiaba Reservoir as an example, the comprehensive deformation prediction and safety warning of this bank slope are carried out by using the method proposed in this paper.The results show that the comprehensive deformation time series can reflect the deformation evolution law of the landslide as a whole, and the prediction results are reliable.According to the hierarchical warning model, this slope is currently in the stage of stable deformation (Grade I early-warning).