基于数据融合的边坡临滑状态确定方法

安全监测及数值模拟是评估边坡稳定性状态的两种重要手段，但是，如何根据监测信息及数值模拟成果确定边坡的临界失稳状态一直是边坡工程领域关注的重点问题。首先基于层次聚类方法计算边坡不同类型状态变量监测点的欧式距离，根据距离大小筛选出边坡的有效监测点；然后，计算同种类型状态变量有效监测点的熵权值，采用熵权融合方法对有效监测点进行数据层融合，得到不同类型状态变量对应的融合监测指标曲线；其次，采用主成分分析法对多种数据层融合监测指标曲线进行特征层融合，得到可反映全部状态变量信息特征的综合监测信息曲线，进而构建了边坡渐进失稳过程中不同监测变量的信息挖掘融合框架；最后，提出了一种缓变曲线的变点搜索方法，采用该方法对综合监测信息曲线开展变点分析，搜索边坡状态渐进演化的突变点（即边坡临滑状态点）。将所提出的方法应用到某公路边坡临界失稳状态的确定中，结果表明，单个监测点或单个融合指标的累积值、变化速率作为边坡失稳判据存在不唯一性，融合多种监测数据建立的综合监测信息序列可以较好地反映边坡状态演化特征，可避免单一状态变量的单一监测点数据对边坡状态的误判，验证了所提方法的可行性和适用性。

Determination method of slope critical failure state based on monitoring data fusion

Safety monitoring and numerical simulation are two significant tools for assessing the slope stability. However, how to determine the slope critical failure state according to the monitoring data and numerical simulation results has always been the focus of the slope engineering. In this study, the Euclidean distances among the slope monitoring points of different types of state variables are calculated based on the hierarchical clustering method, and the effective monitoring points of slope are selected according to the distance. Then, the entropy weight of time series of effective monitoring points of the same type of state variables is calculated, and the entropy weight fusion method is used to perform data layer fusion for effective monitoring points of the same type of state variables, and the fusion monitoring index curves corresponding to different types of state variables are obtained. After that, principal component analysis method is adopted to perform feature level fusion for various fusion monitoring index curves, and a comprehensive monitoring information curve that can reflect the information characteristics of all state variables is obtained, and then an information mining and fusion framework for different monitoring variables in the process of slope progressive instability is constructed by using a variety of mathematical and statistical methods. Finally, a change point search method of graded curve is proposed to search the abrupt change point of slope state gradual evolution (i.e. slope critical failure state). The proposed method is applied to determining the critical instability of a highway slope. The results show that the cumulative value and change rate of a single monitoring point or a single fusion index are not unique as the criterion of slope instability. The comprehensive monitoring information sequence established by integrating multiple monitoring data can better reflect the evolution characteristics of slope state and avoid misjudgment of slope state by data of a single monitoring point with a single state variable, which verifies the feasibility and applicability of the proposed method.