滑坡监测点多信息相似性度量方法研究

滑坡监测点在一段时间内对外界影响因素的响应可以看作滑坡在当前时刻的状态。滑坡监测点在相似的状态下受到相似的外界激励时会表现出相似的响应。据此，可以对相似的状态进行类比，从而将相似状态运用于滑坡的稳定性分析、数据修正和预测预报等。提出一种基于多信息的状态单元矩阵的相似性度量方法，运用该方法同时对滑坡的外界影响因素和运动状态进行相似性匹配，既保证了外界影响因素的相似性，也保证了滑坡当前运动状态的相似性。为验证该方法的有效性，运用三峡库区4个滑坡34个监测点的1 770个状态作为状态单元集，对白水河滑坡2015年1月至2016年6月的位移数据进行预测。预测结果显示，该方法优于现有的BP神经网络、支持向量回归等经典预测方法。

A similarity measurement method for multiple information data of landslide

The response of a monitoring point on a landslide to external factors over a period of time can be regarded as a current state of the landslide. The monitoring points on landslide show similar responses when they are stimulated by similar external factors under similar states. Accordingly, it is possible to analogize similar states and apply it to stability analysis, data correction and prediction of landslide. In this study, a similarity measurement method for multiple information data based on status element matrix was proposed. It not only ensured the similarity of the external factors, but also guaranteed the similarity of the current state of landslide to measure external factors and movement status of landslides. To verify the effectiveness of the method, 1 770 states of 34 monitoring points of 4 landslides in the Three Gorges reservoir area were used as status element matrix to perform similarity matching, and good results were obtained to predict displacement data of Baishui River landslide from January 2015 to June 2016. Based on the results, the similarity measurement method proposed was superior to the existing classic prediction methods such as back propagation neural network, support vector regression and so on.