A method for accurately determining the position of sliding surface based on the single sliding surface R-type deep hole inclinometer curve

Abstract

The present invention discloses a method for accurately determining the position of a sliding surface based on a single sliding surface R-type deep hole inclinometer curve, and the present invention belongs to the technical field of landslide detection. The main steps of the present invention are as follows: 1) drawing a cumulative displacement-time curve graph of the deep part of the borehole; 2) calculating the displacement rate of the soil at different depths during the monitoring period; 3) drawing a displacement rate-time curve graph at different depths of the borehole; 4) extracting key day data, and drawing a displacement rate-depth scatter plot with depth as the horizontal coordinate and displacement rate as the vertical coordinate; 5) DBSCAN data clustering to determine the sliding interval where the sliding surface is located and the size of the interval; 6) mathematical calculation of discrete points to identify the distribution type of discrete points and accurately calculate the specific position of the sliding surface. According to the relationship between the changing characteristics of the displacement rate of the soil at different depths during the slope deformation process and the position of the sliding surface, the present invention introduces the machine learning algorithm DBSCAN to classify the data and quickly and accurately obtain the position of the sliding surface.

Description

Technical Field

[0001] The invention belongs to the technical field of landslide monitoring, and in particular relates to a method for accurately determining the position of a sliding surface based on a single sliding surface R-shaped deep hole inclinometer curve. Background Art

[0002] In the study of landslide stability, landslide displacement monitoring is an important research direction. Through landslide displacement monitoring, not only can we obtain data such as slope deformation displacement rate and direction, but also by analyzing the monitoring data, we can deepen our understanding of the landslide deformation mechanism and deformation characteristics, and provide an important basis for the treatment of unstable slopes. In landslide displacement monitoring, determining the accurate position of the landslide sliding surface is a prerequisite for evaluating the stability of the slope and effectively treating it.

[0003] Sliding surface identification methods are divided into non-deterministic methods and deterministic methods. Non-deterministic methods mainly include simple mechanical identification method, geophysical method, numerical simulation method, etc. The sliding surface identified by non-deterministic methods is a speculative sliding surface, which needs to be verified by deterministic methods. In addition, non-deterministic methods are generally auxiliary means or means of conducting preliminary research. Deterministic methods are used to identify sliding surfaces during the investigation, design and construction stages of landslide control projects. Deterministic methods mainly include field geological identification method, exploration identification method and deep displacement observation method.

[0004] At present, the deep displacement observation method is the main method for identifying the sliding surface. This method intuitively determines the location of the sliding surface based on the deformation characteristics of the deep cumulative displacement curve of the soil during the monitoring period. Common landslide deep hole inclinometer curves (i.e. deep cumulative displacement-time curves) mainly include "V" type, "B" type, "r" type, "pendulum" type and "composite" type.

[0005] In the identification of the sliding surface position of the "r"-shaped deep hole inclinometer curve, the deep displacement observation method has some obvious defects: on the one hand, this method can only roughly judge the sliding surface position and cannot accurately obtain the sliding surface position data; on the other hand, the horizontal and vertical coordinate observation scales of the deep cumulative displacement-time curve will also greatly affect the identification of the sliding surface position. If the observation scale is too small, the irrelevant deformation features in the curve will be magnified, which is not conducive to capturing the sliding surface features; if the observation scale is too large, the change characteristics of the cumulative displacement curve are not obvious, and it is difficult to determine the exact position of the sliding surface. Summary of the invention

[0006] According to the deformation rate characteristics of the landslide body and the machine learning algorithm, the present invention proposes a method for accurately determining the sliding surface position based on the single sliding surface R-type deep hole inclinometer curve, which can make full use of the deep accumulated displacement data of the borehole and quickly and accurately determine the sliding surface position of the "R"-type deep hole inclinometer curve. The detailed technical scheme of the present invention is as follows:

[0007] A method for accurately determining the position of a sliding surface based on a single sliding surface R-shaped deep hole inclinometer curve comprises the following steps:

[0008] 1) Import the displacement monitoring data of the deep hole inclinometer sensor, draw the deep hole cumulative displacement-time curve corresponding to each sensor based on the monitoring data, and determine whether the deformation feature of the curve is R-type. If it is R-type, continue to perform the following steps;

[0009] 2) Take the deep hole inclinometer in step 1), use formula ① to calculate the sensor data at each position of the deep hole inclinometer, and calculate the displacement rate of the soil of each sensor during the monitoring period:

[0010] ①

[0011] In formula ①:

[0012] t i is the acquisition time of deep displacement;

[0013] t i+1 For t i The next adjacent deep displacement acquisition time;

[0014] s i for t i Cumulative displacement during the term;

[0015] s i+1 for t i+1 Cumulative displacement during the term;

[0016] v i for t i to t i+1 The displacement rate within the time limit;

[0017] The displacement rate value of the sensor in each monitoring period is calculated, and the displacement rate-time curve diagram corresponding to the sensor is drawn; the displacement rate values ​​of the remaining sensors are calculated in turn, and the displacement rate-time curve diagrams corresponding to each sensor are drawn; thereby, the displacement rate values ​​of the soil at different depths in the monitoring period are obtained, and then the displacement rate-time curve diagrams at different depths are drawn;

[0018] 3) By using the displacement rate-time curve at different depths drawn in step 2), determining the date corresponding to the data point with the most significant clustering characteristics on the curve; extracting the soil displacement rate value of each sensor within the date in step 2);

[0019] 4) Using the displacement rate value obtained in step 3) as the ordinate and the sensor deployment depth as the abscissa, draw a displacement rate-depth scatter plot; according to the distribution characteristics of the scatter plot, identify the step stages between different clusters in the scatter plot, and the range corresponding to the horizontal interval of the step stage is the range of the sliding surface;

[0020] 5) In the scatter plot, the upper cluster data points and the lower cluster data points in the step stage are distributed in a horizontal chain. The DBSCAN algorithm is used to set the corresponding neighborhood radius (Eps) and the threshold of the number of data objects in the neighborhood (MinPts) to cluster the data;

[0021] According to the clustering results, take the end data points of the upper cluster data points adjacent to the step stage, take the end data points of the lower cluster data points adjacent to the step stage, and the interval range formed by the depth values ​​corresponding to the two end data points is the interval range where the sliding surface is located;

[0022] 6) According to the position of the discrete points in the step stage, identify the distribution type of the discrete points as single-point distribution or multi-point distribution; according to the specific distribution type of the discrete points, select the mathematical calculation method corresponding to the distribution type, substitute the values ​​of the two end data points and the discrete points obtained in step 5) into the mathematical calculation method, and calculate the accurate position of the sliding surface.

[0023] Basic principle: DBSCAN is a density-based clustering algorithm. This type of density clustering algorithm generally assumes that the category can be determined by the density of sample distribution. Samples of the same category are closely connected to each other, that is, there must be samples of the same category not far away from any sample of the category. By classifying closely connected samples into one category, a cluster category is obtained. By classifying all groups of closely connected samples into different categories, the final cluster category results are obtained. The basic principle of the DBSCAN algorithm and its calculation process are as follows:

[0024] Input: Sample set , neighborhood parameters The sample distance measurement method generally chooses Euclidean distance. The specific calculation method of Euclidean distance is shown in formula ② and formula ③:

[0025] ②

[0026] ③

[0027] Among them, ρ is the point (x 2 ,y 2 ) and (x 1 ,y 1 ) between the points; |X| is the Euclidean distance between the points (x 2 ,y2 ) to the origin.

[0028] Output: Clustering C

[0029] 1) Initialize the core object collection , initialize the number of clusters k=0, initialize the unvisited sample set , cluster partition .

[0030] 2) For j=1,2,...m , follow these steps to find all core objects:

[0031] a) Find sample x by distance metric j of Number of domain subsamples ;

[0032] b） If the number of samples in the subsample set satisfies , the sample x j Added core object sample collection: .

[0033] 3) If the core object collection , the algorithm ends, otherwise it continues to the next step.

[0034] 4) Randomly select a core object σ from the core object set Ω and initialize the core object list of the current cluster , initialize the category number k=k+1, initialize the current cluster sample set , update the unvisited sample set .

[0035] 5) If the current cluster core object queue , then the current cluster C k Completed, update cluster division , update the core object collection , go to step 3. Otherwise update the core object set .

[0036] 6) In the current cluster core object queue Take out a core object σ' and pass the neighborhood distance threshold Find all Neighborhood subsample set ,make , update the current cluster sample set , update the unvisited sample set renew Go to step 5).

[0037] The output result is: cluster division .

[0038] According to the movement and deformation characteristics of the landslide body, the present invention constructs a mathematical calculation method corresponding to the distribution of different discrete points in the step stage. The specific derivation process is as follows:

[0039] The “r” type inclinometer curve has a complete and continuous sliding surface, and the sliding body ( Figure 1 The upper region in a) and the immobile body ( Figure 1 The motion deformation between the lower and middle areas of a can be simplified as Figure 1 The mechanical model is shown in a. Its deformation process is as follows Figure 1 As shown in Figure 2, this motion change process can be simplified into a mechanical model in which two rods are connected by a sliding hinge support. Figure 1 c, deformation curve of the rod Figure 1 d is similar to the "r" type inclinometer curve.

[0040] Among them, the sliding surface is located in the sliding interval ( Figure 1 c), the soil in the sliding interval is generally homogeneous sliding zone soil. It can be assumed that the displacement rate of the soil in the sliding interval changes linearly from top to bottom along the depth. The displacement rate at the top of the sliding interval is equal to the displacement rate of the scattered points on the left end of the step stage, and the displacement rate at the bottom of the sliding interval is equal to the displacement rate of the scattered points on the right end of the step stage. The midpoint value of the displacement rate in the sliding interval (V 中 ) is the depth corresponding to the sliding surface, and this depth should be the midpoint of the sliding interval. However, actual data acquisition is affected by the sampling depth interval, and the velocity value corresponding to the midpoint of the sliding interval is not the midpoint value of the interval displacement velocity (V 中 ):

[0041] 1) The actual speed corresponding to the midpoint position (V 测 ) is greater than V 中 When the sliding surface is at V 测 Corresponding depth (h 测 ) to the right (i.e. deeper);

[0042] 2) V 测 Less than V 中 When the sliding surface is at V 测 Corresponding depth (h 测 ) to the left (i.e. shallower);

[0043] According to the above assumptions and analysis criteria, mathematical calculation methods corresponding to single-point distribution and multi-point distribution are constructed respectively.

[0044] I Single point distribution:

[0045] The distribution type of discrete points within a step stage can be further divided into "within the stage" and "outside the stage".

[0046] ①In stage

[0047] During the step phase, according to V 中 With V 测 The size is divided into V 中 <V 测 and V 中 >V 测 Two situations.

[0048] Among them, V 中 =(V 上 +V 下 ) / 2; (V 上 and V 下 are the displacement rates at the upper and lower ends of the sliding range respectively; V 测 is the displacement rate of discrete points in the sliding interval).

[0049] a. If V 中 <V 测 , the relationship is as follows Figure 2 As shown, the calculation process is as follows:

[0050] V 中 =(V 上 +V 下 ) / 2

[0051] tanθ=(V 上 -V 下 ) / (h 下 -h 上 )

[0052] L 1 = V 测 -V 中

[0053] L 1 / L 2 = tanθ

[0054] h 滑 =h 测 +L 2

[0055] Then, the depth of the sliding surface is h 滑 .

[0056] b. If V 中 >V 测 , the relationship is as follows Figure 3 As shown, the calculation process is as follows:

[0057] V 中 =(V 上 +V下 ) / 2

[0058] tanθ=(V 上 -V 下 ) / (h 下 -h 上 )

[0059] L 1 = V 中 -V 测

[0060] tanθ=L 1 / L 2

[0061] h 滑 =h 测 -L 2

[0062] Then, the depth of the sliding surface is h 滑 .

[0063] ②Outside the stage

[0064] V 中 <V 测 There is a special case where the discrete point falls outside the step stage. This is because the displacement rate of the sliding interval is greater than the displacement rate of the upper sliding body. Considering the symmetry, V 上 The discrete points are converted into discrete points within the step phase by the horizontal symmetry axis, then V 测 becomes V* 测 .like Figure 4 As shown:

[0065] V 中 =(V 上 +V 下 ) / 2

[0066] tanθ=(V 上 -V 下 ) / (h 下 -h 上 )

[0067] L 0 =V 测 -V 上

[0068] L 1 = (V 上 -V 中 )-L 0

[0069] L 1 / L 2 = tanθ

[0070] h 滑 =h 测 +L 2

[0071] Then, the depth of the sliding surface is h 滑 .

[0072] II Multi-point distribution:

[0073] The distribution types of multiple discrete points in the step stage can be divided into "two-point distribution" and "three-point distribution". This is because the sliding interval is large and the sensor interval is small, so that multiple sensors fall within the sliding interval, which leads to multiple discrete points in the step stage; while the range of the soil sliding interval of the R-type inclinometer curve is generally small, ranging from a few centimeters to several meters. In actual survey and design, the sensor interval is 0.5m, so the number of discrete points in the step stage generally does not exceed three.

[0074] ① Two-point distribution

[0075] According to the equivalent displacement rate (V e ) and the size in V can be divided into two calculation cases: V e >V neutralizes V e <V 中 Among them, V 1 and V 2 are two discrete points in the step phase, V e are the equivalent discrete points of these two discrete points, that is, converting the multi-point distribution into a single-point distribution.

[0076] a. If V e >V 中 , the calculation process is as follows, Figure 5 As shown:

[0077] V e =(V 1 +V 2 ) / 2

[0078] h e =(h 1 +h 2 ) / 2

[0079] V 中 =(V 上 +V 下 ) / 2

[0080] tanθ=(V 上 -V 下 ) / (h 下 -h 上 )

[0081] L 1 = Ve -V 中

[0082] L 1 / L 2 = tanθ

[0083] Because V e >V 中 , the sliding surface position should be at h e On the right side, h 滑 =h e +L 2；

[0084] Then, the depth of the sliding surface is h 滑 .

[0085] b. If V e <V 中 , the calculation process is as follows, Figure 6 As shown:

[0086] V e =(V 1 +V 2 ) / 2

[0087] h e =(h 1 +h 2 ) / 2

[0088] V 中 =(V 上 +V 下 ) / 2

[0089] tanθ=(V 上 -V 下 ) / (h 下 -h 上 )

[0090] L 1 = V 中 -V e

[0091] L 1 / L 2 = tanθ

[0092] Because V e <V 中 , the sliding surface position should be at h e On the left side, h 滑 =h e -L 2；

[0093] Then, the depth of the sliding surface is h 滑 .

[0094] ②Three-point distribution

[0095] According to the equivalent displacement rate (V e ) and V 中 The size of can also be divided into two calculation cases:

[0096] V e <V 中 and V e >V 中 Among them, V 1 、V 2 and V 3 are three discrete points in the step phase, V e are equivalent discrete points of these three discrete points.

[0097] a. If V e <V 中 , the calculation process is as follows, Figure 7 As shown:

[0098] V e =(V 1 +V 3 ) / 2

[0099] tanθ 1 =(V 1 -V 3 ) / (h 3 -h 1 )

[0100] L 1 =V 2 -V e

[0101] L 1 / L 2 = tanθ 1

[0102] If V 2 >V e , then h e =h 2 +L 2 ;

[0103] If V 2 <V e , then h e =h 2 -L 2 ;

[0104] Therefore, the coordinates of the equivalent discrete points are (h e , V e );

[0105] V 中 =(V上 +V 下 ) / 2

[0106] tanθ 2 =(V 上 -V e ) / (h e -h 上 )

[0107] L 3 =V 中 -V e

[0108] L 3 / L 4 = tanθ 2

[0109] Because V 中 >V e , then h 滑 =h e -L 4 ;

[0110] Therefore, the depth of the sliding surface is h 滑 .

[0111] b. If V e >V, such as Figure 8 As shown, the calculation process is as follows:

[0112] V e =(V 1 +V 3 ) / 2

[0113] tanθ 1 =(V 1 -V 3 ) / (h 3 -h 1 )

[0114] L 1 =V 2 -V e

[0115] L 1 / L 2 = tanθ 1

[0116] If V 2 >V e , then h e =h 2 +L 2 ;

[0117] If V 2 <V e , then he =h 2 -L 2 ;

[0118] Therefore, the coordinates of the equivalent discrete points are (h e , V e );

[0119] V 中 =(V 上 +V 下 ) / 2

[0120] tanθ 2 =(V 上 -V e ) / (h e -h 上 )

[0121] L 3 =V 中 -V e

[0122] L 3 / L 4 = tanθ 2

[0123] Because V 中 <V e , then h 滑 =h e +L 4 ;

[0124] Therefore, the depth of the sliding surface is h 滑 .

[0125] The beneficial effects of the present invention are:

[0126] 1. According to the deformation characteristics of landslide, the present invention takes the "r"-type deep hole inclinometer curve as the research object, explores the relationship between the changing characteristics of soil displacement rate at different depths during slope deformation and the position of the sliding surface, introduces DBSCAN to classify the data, quickly and accurately obtains the position information of the slope sliding interval, and obtains the mathematical calculation method corresponding to the discrete points in the step stage under different distribution types by constructing a landslide motion mechanics model, and then substitutes the actual data to complete the accurate determination of the sliding surface position.

[0127] 2. Through in-depth exploration and analysis of the displacement rate of the slope, the present invention proposes to use the displacement rate change characteristics as the basis for identifying the position of the sliding surface. It is found that there is an obvious step stage in the displacement rate between the landslide body and the immovable body, and a corresponding mathematical calculation method is constructed based on the distribution characteristics of discrete points in the step stage. A complete method that can accurately determine the position of the sliding surface is proposed.

[0128] 3. The present invention conducts in-depth mining of borehole deep displacement monitoring data and improves the utilization of data; on the other hand, the algorithm DBSCAN is introduced to identify data features and accurately extract the position information of the sliding interval, which can provide a basis and reference for the accurate identification of the sliding surface position.

[0129] 4. Compared with the traditional method, the present invention adopts a more efficient data processing method based on actual monitoring data, which can quickly and accurately determine the position of the sliding surface, greatly improve work efficiency, and effectively avoid the situation where the error is large due to human subjective judgment of the sliding surface position. BRIEF DESCRIPTION OF THE DRAWINGS

[0130] Figure 1 It is a simplified schematic diagram of the sliding body model;

[0131] Figure 2 It is the calculation relationship in the single point distribution stage Figure 1 ;

[0132] Figure 3 It is the calculation relationship in the single point distribution stage Figure 2 ;

[0133] Figure 4 It is a graph of external calculation relationships in the single-point distribution stage;

[0134] Figure 5 It is the two-point distribution calculation relationship Figure 1 ;

[0135] Figure 6 It is the two-point distribution calculation relationship Figure 2 ;

[0136] Figure 7 It is the three-point distribution calculation relationship Figure 1 ;

[0137] Figure 8 It is the three-point distribution calculation relationship Figure 2 ;

[0138] Fig. 9 is a cumulative displacement-time curve diagram of an embodiment of the present invention;

[0139] Fig.10 is a displacement rate-time curve diagram of different sensors in the embodiments of the present invention;

[0140] Fig.11 is a displacement rate-depth curve diagram of each sensor on the 461.5th day of an embodiment of the present invention;

[0141] Fig.12 It is the displacement rate-depth clustering diagram of each sensor on the 461.5th day of the embodiment of the present invention;

[0142] Fig.13It is a scatter point coordinate diagram of the present invention;

[0143] Fig.14 is a calculation relationship diagram of an embodiment of the present invention;

[0144] Fig.15 It is a display diagram of the calculation results of the present invention;

[0145] Fig.16 It is a system principle diagram of the present invention. DETAILED DESCRIPTION

[0146] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, which are deep hole displacement monitoring data of a landslide location in Yunnan:

[0147] 1) Import data and draw the deep cumulative displacement-time curve, such as Fig. 9 In the figure, the ordinate is the monitoring time, and the abscissa is the displacement value of each sensor; in this embodiment, the monitoring time is from 2018 / 5 / 15 to 2020 / 1 / 10, and a total of 29 data are collected.

[0148] from Fig. 9 It can be found that the curve is "r" shaped, and the sliding surface of the slope has been formed, and the position is approximately 24m. In order to further determine the exact position of the sliding surface, continue to perform the following steps.

[0149] 2) Formula ① can be used to calculate the displacement rate of the soil at different depths during the monitoring period, and the displacement rate-time curves of different sensors can be drawn, such as Fig.10 As shown:

[0150] From the displacement rate-time curve Fig.10 It can be found that the velocity change characteristics of the soil are significant, the initial velocity fluctuation is significant, and it gradually becomes gentle over time. Since the sliding surface has been formed, the displacement velocity will show obvious clustering characteristics one day. Fig.10 It can be found that the clustering characteristics are significant on the 461.5th day, and the displacement rate data at different depths on that day can be extracted for subsequent analysis.

[0151] 3) Extract the displacement rate data of each sensor on the 461.5th day, and draw a displacement rate-depth curve with depth as the horizontal axis and displacement rate as the vertical axis. Fig.11 shown.

[0152] Depend on Fig.11 It can be found that the displacement rate-depth scatter plot of the day is clearly divided into two clusters, there is an obvious step stage between the two clusters of scatter data, and there is a significant discrete point in the step stage. The range corresponding to the horizontal interval of the step stage is the range where the sliding surface is located, and the sensor depth corresponding to the discrete point is the location of the sliding surface.

[0153] 4) Use the DBSCAN algorithm to set Eps and MinPts according to the distance between points in each cluster and between clusters. In this case, Eps=0.500675 and MinPts=1 are set. The algorithm clustering results are as follows: Fig.12 As shown:

[0154] Depend on Fig.12 It can be found that DBSCAN divides the scattered data into three clusters, which are respectively recorded as "upper cluster", "middle cluster" and "lower cluster". The depth corresponding to the right endpoint of the upper cluster is 23m; the middle cluster has only one data point, and its corresponding depth is 23.5m; the depth corresponding to the left endpoint of the lower cluster is 24m.

[0155] Therefore, it can be known that the sliding section is located in the range of 23m to 24m underground.

[0156] 5) By Fig.12 It can be seen that there is only one scattered point in the step stage, and this point is located between the upper and lower cluster scattered points in the step stage, which belongs to the stage type in the single point distribution. The coordinates of the rightmost scattered point of the upper cluster scattered points, the coordinates of the discrete point, and the coordinates of the leftmost scattered point of the lower cluster scattered points are extracted respectively (such as Fig.13 ), substitute into the mathematical calculation method in the single point distribution stage:

[0157] V 测 =0.15190,

[0158] V 中 =(V 上 +V 下 ) / 2=(0.17810+0.01780) / 2=0.09795

[0159] V 测 > V 中 , so V in the single point distribution stage is selected 中 <V 测 Mathematical models such as Fig.14 As shown:

[0160] Substitute the data into the calculation formula:

[0161] V 中 =(V 上 +V 下 ) / 2 = 0.09795

[0162] tanθ=(V 上 -V 下 ) / (h 下 -h 上 ) = (0.17810 − 0.01780) / (24 − 23) = 0.1603

[0163] L1 = V 测 -V 中 = 0.15190−0.09795 = 0.05395

[0164] L 2 =L 1 / tanθ=0.05395 / 0.1603 = 0.3365

[0165] h 滑 =h 测 +L 2 = 23.5+0.3365 = 23.8365

[0166] Therefore, the depth of the sliding surface is 23.8365m.

[0167] The calculation results are shown as Fig.15 As shown, the black horizontal line (-23.8365m) is the calculated sliding surface position.

Claims

1. A method for accurately determining the position of a sliding surface based on a single sliding surface R-type deep hole inclinometer curve, characterized in that: The following steps are involved: 1) Import the displacement monitoring data of the deep hole inclinometer sensor, draw the deep hole cumulative displacement-time curve corresponding to each sensor based on the monitoring data, and determine whether the deformation feature of the curve is R-type. If it is R-type, continue to perform the following steps; 2) Take the deep hole inclinometer in step 1), use formula ① to calculate the sensor data at each position of the deep hole inclinometer, and calculate the displacement rate of the soil of each sensor during the monitoring period: ① In formula ①: t i is the acquisition time of deep displacement; t i+1 For t i The next adjacent deep displacement acquisition time; s i for t i Cumulative displacement during the term; s i+1 for t i+1 Cumulative displacement during the term; v i for t i to t i+1 The displacement rate within the time limit; The displacement rate value of the sensor in each monitoring period is calculated, and the displacement rate-time curve diagram corresponding to the sensor is drawn; the displacement rate values ​​of the remaining sensors are calculated in turn, and the displacement rate-time curve diagrams corresponding to each sensor are drawn; thereby, the displacement rate values ​​of the soil at different depths in the monitoring period are obtained, and then the displacement rate-time curve diagrams at different depths are drawn; 3) By using the displacement rate-time curve at different depths drawn in step 2), determining the date corresponding to the data point with the most significant clustering characteristics on the curve; extracting the soil displacement rate value of each sensor within the date in step 2); 4) Using the displacement rate value obtained in step 3) as the ordinate and the sensor deployment depth as the abscissa, draw a displacement rate-depth scatter plot; according to the distribution characteristics of the scatter plot, identify the step stages between different clusters in the scatter plot, and the range corresponding to the horizontal interval of the step stage is the range of the sliding surface; 5) In the scatter plot, the upper cluster data points and the lower cluster data points in the step stage are distributed in a horizontal chain. The DBSCAN algorithm is used to set the corresponding neighborhood radius and the threshold of the number of data objects in the neighborhood to cluster the data; According to the clustering results, take the end data points of the upper cluster data points adjacent to the step stage, take the end data points of the lower cluster data points adjacent to the step stage, and the interval range formed by the depth values ​​corresponding to the two end data points is the interval range where the sliding surface is located; 6) According to the position of the discrete points in the step stage, identify the distribution type of the discrete points as single-point distribution or multi-point distribution; according to the specific distribution type of the discrete points, select the mathematical calculation method corresponding to the distribution type, substitute the values ​​of the two end data points and the discrete points obtained in step 5) into the mathematical calculation method, and calculate the accurate position of the sliding surface.

Images