A slope safety monitoring and early warning method based on intelligent particles

Abstract

This invention proposes a slope safety monitoring and early warning method based on intelligent particles. The method includes the following steps: S1. intelligent particle deployment and monitoring system construction; S2. intelligent particle cumulative rotation angle outlier identification based on the Z-score; S3. intelligent particle cumulative rotation angle time registration based on Kalman filtering and cubic interpolation; S4. intelligent particle cumulative rotation angle fusion based on an adaptive fusion algorithm; S5. slope deformation stage classification based on the cumulative rotation angle time series and the tangent angle method; S6. slope failure moment estimation based on the cumulative rotation angle time series and the cumulative rotation angle velocity inverse method. This invention applies intelligent particle sensors to slope safety monitoring and early warning, using the cumulative rotation angle as an indicator for classifying slope deformation stages and estimating slope failure moments. This method can sensitively reflect changes in slope stability and issue timely early warnings.

Description

Technical Field

[0001] The present invention belongs to the field of civil engineering, and specifically designs a slope safety monitoring and early warning method based on intelligent particles. Background Art

[0002] Slope collapse is a major geological hazard that can cause massive loss of property and life, and the potential risk of landslides exists in all regions of the world. Currently, advanced technologies for slope monitoring and early warning include global navigation satellite systems, interferometric radar, lidar, fiber-optic sensing, and optical remote sensing. However, these technologies primarily monitor surface displacement changes. When a slope's stability deteriorates due to environmental influences, deformation and stress begin to change internally. Only when the internal deformation of the sliding mass is sufficient will significant macroscopic deformation appear on the surface of the landslide. Surface monitoring equipment can only detect these changes when macroscopic deformation exceeds a certain level. Therefore, surface deformation is not a sufficient condition for landslide occurrence, making it difficult to provide accurate early warning of landslides. Although technologies such as inclinometers and array displacement meters can be used to monitor deep slope displacement, their limited measurement range and high cost have prevented widespread and long-term application. Inclinometers installed on attached structures can also reflect slope stability through changes in their inclination angle, but these only utilize a portion of the sensor's attitude data and fail to fully utilize the sensor's attitude information. Therefore, for attitude sensing technology represented by smart particles, developing a set of monitoring and early warning methods that can be directly embedded into the slope is of great significance for slope health assessment and engineering safety assurance. Summary of the Invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a slope safety monitoring and early warning method based on intelligent particles, which uses the quaternion data obtained by intelligent particle monitoring to calculate the cumulative rotation angle, achieve more sensitive slope deformation stage division and slope failure time estimation, and timely release early warning information.

[0004] Technical solution, in order to solve the above technical problems and achieve the purpose of the present invention, the present invention proposes a slope safety monitoring and early warning method based on smart particles, comprising the following steps:

[0005] S1. Drill holes in the slope and bury smart particle sensors, configure data transmission hardware equipment, and build an intelligent particle monitoring system;

[0006] S2, based on Z-score, identifies and eliminates outliers in the accumulated rotation angle of smart particles. This method calculates the angular distance and accumulated rotation angle time series of each smart particle sensor's quaternion time series, and detects and eliminates outliers.

[0007] S3. Time-align the cumulative rotation angle time series of each smart particle sensor after removing outliers based on Kalman filtering and cubic interpolation, so that each smart particle sensor has a cumulative rotation angle value at the monitoring time;

[0008] S4, processing the cumulative rotation angle time series of each intelligent particle sensor after time registration obtained in S3 using an adaptive fusion algorithm to obtain a fused cumulative rotation angle time series;

[0009] S5. The fused cumulative rotation angle time series obtained in S4 is processed using the tangent angle method to obtain the division of slope deformation stages and conduct a qualitative analysis of the slope safety and stability.

[0010] S6. The fused cumulative rotation angle time series obtained in S4 is processed using the inverse cumulative rotation angle velocity method to obtain an estimated value of the slope failure moment and perform a quantitative analysis of the slope safety and stability.

[0011] Furthermore, in step S1, the smart particle is a sensor, equipped with an accelerometer, a gyroscope, a magnetometer, a thermometer, and a pressure gauge. The outer shape of the smart particle is a cube, the edge length is 27-40 mm, and the apparent density is 2.18-2.30 g / cm 3 , the intelligent particle placement and monitoring system is built as follows:

[0012] (2.1) Setting up monitoring points: drilling holes on the slope surface and burying smart particle sensors. The drilling depth and the burial depth of the smart particles should be shallower than the potential sliding surface, with the lowest burial depth being 1m below the slope surface. The monitoring points should be arranged on cross-section lines that are parallel to or perpendicular to the slope direction and the sliding direction of the landslide, and should be arranged 10m outside the sliding influence range. For a single slope, cross-section lines should be arranged in the middle and on both sides of the slope or landslide, with no less than three lines. For a strip slope group, a number of cross-section lines greater than a preset number should be selected. The spacing between monitoring points should be 5m to 40m, with a maximum horizontal spacing of no more than 40m and a maximum vertical spacing of no more than 20m. The preset number is obtained by dividing the length of the strip slope by the horizontal spacing.

[0013] (2.2) Backfilling of monitoring points: using backfill materials to fill holes in layers, and filling each layer before the next layer. The backfill material is plain soil or prepared fly ash soil around the monitoring slope;

[0014] (2.3) Monitoring system hardware configuration: remote roadside signal collectors are set up at preset locations on the monitoring slope. Data is transmitted to the remote roadside signal collectors based on the ZigBee protocol and uploaded to the cloud storage in real time.

[0015] Furthermore, in step (2.1), after the smart particles are started, they enter a periodic sleep and wake-up mode. The smart particles should monitor once every 7 to 14 days, and each wake-up operation should last for 1 to 2 minutes. During the flood season, rainy season, and construction period of prevention and control projects, monitoring should be intensified, once a day or once every few hours.

[0016] The indicator used for slope safety monitoring and early warning is the cumulative rotation angle of the smart particle sensor, which is calculated by the quaternion attitude data jointly calculated by the accelerometer, gyroscope and magnetometer data of the smart particle. The smart particle is at the monitoring time t i After waking up, it enters sleep mode and waits for the next wake-up. The smart particle records several quaternions according to the working frequency, a total of n. These quaternions are first averaged and then normalized as the quaternion data of the monitoring time t:

[0017]

[0018] In formulas (1)-(3), is the monitoring time t i At the beginning of the awakening process, the intelligent particle records the mth original quaternion data, a total of n, i, j, k are imaginary numbers, and their coefficients are is the real part of the mth original quaternion data; the monitoring time starts at t0 and the monitoring ends at t N , that is, there are N data points in the time series;

[0019] represents the monitoring time t i The average quaternion data of the smart particles, the coefficients of the imaginary parts i, j, k are The real part is represents the monitoring time t i The normalized average quaternion data of the smart particles has a modulus of 1;

[0020] The angular distance is calculated based on the quaternion of adjacent monitoring moments. The cumulative sum of the angular distances is the cumulative rotation angle of the smart particle over time:

[0021]

[0022] In formulas (4) and (5), is the previous monitoring time t i-1 To monitoring time t i angular distance; real is the name of the function that takes the real part of the quaternion; represents the previous monitoring time t i-1 The conjugate quaternion of the normalized mean quaternion; From the initial monitoring time t0 to the current monitoring time t i The total angular distance traveled by the smart particles is the cumulative turning angle.

[0023] Furthermore, in step S2, for the cumulative rotation angle time series from each smart particle, outlier values ​​of the smart particle cumulative rotation angle are identified and eliminated based on the Z-score standard score, as follows:

[0024] ①Determine the sliding window size as N window ;

[0025] ②For a certain monitoring time t i , calculate the local mean using the data points in the window and local standard deviation The angle distance time series of a certain intelligent particle sensor is That is, if the time series has N data points, then:

[0026]

[0027] ③For each angle distance value, calculate its Z-score:

[0028]

[0029] ④Set threshold Z threshold , if |Z t Exceeds the preset threshold Z threshold , then it is believed that If it is an outlier, the data determined as an outlier will be eliminated, and the angle distance after elimination will be accumulated according to time to obtain the cumulative angle time series after outlier elimination. The outlier is identified and eliminated for the angle distance of each smart particle, and the cumulative angle time series is calculated.

[0030] Furthermore, in step S3, the intelligent particle cumulative rotation time registration based on Kalman filtering and cubic interpolation is performed as follows:

[0031] ① Perform Kalman filtering on the cumulative rotation angle time series of each smart particle sensor to remove noise and smooth the data. The data points on the rotation angle time series of a smart particle patent after Kalman filtering are:

[0032]

[0033] ② Unified time axis, which is composed of the wake-up working time of all smart particle sensors;

[0034] ③ Perform cubic interpolation. For the cumulative rotation angle time series of a certain smart particle, the time t to be interpolated is iAs a benchmark, determine the two most recent cumulative rotation angle data forward and backward respectively, and construct a cubic polynomial F(x)=A i (t i -x) 3 +B i (t i -x) 2 +C i (t i -x)+D i , the independent variable of the function is x, which represents the monitoring time, and the dependent variable of the function is the cumulative rotation angle;

[0035] ④ Solve the coefficient A of the cubic polynomial i ,B i ,C i ,D i , interpolate to obtain F(x), so that the cubic polynomial satisfies: the function values ​​at the interpolation points are equal, The first-order derivative is continuous at the interpolation point, The second-order derivative is continuous at the interpolation point, The cumulative rotation angle time series of all smart particles are time-aligned so that the cumulative rotation angle value of each smart particle exists at all monitoring moments.

[0036] Furthermore, in step S4, the accumulated rotation angles of all intelligent particles are fused based on the adaptive fusion algorithm to merge the accumulated rotation angle time series of all intelligent particles into one accumulated rotation angle time series for early warning. At the same monitoring moment, the accumulated rotation angles from different intelligent particles are fused according to the weight coefficient. The size of the weight coefficient depends on the difference from the average value of the accumulated rotation angle:

[0037]

[0038] In formulas (10) and (11), SAD ti,l is the monitoring time t i The original cumulative rotation angle from the lth smart particle, there are L smart particles in total, is the monitoring time t i The variance of the average cumulative turning angle of the lth smart particle; is the monitoring time t i The weight coefficient of the cumulative turning angle of the lth smart particle; is the monitoring time t i Cumulative rotation angle of fused intelligent particles after adaptive weighted fusion.

[0039] Furthermore, in step S5, based on the cumulative rotation angle time series and the slope deformation stage division using the tangent angle method, a qualitative analysis of the slope safety and stability is performed as follows:

[0040] The improved fused cumulative rotation angle time series is obtained by dividing the fused intelligent particle cumulative rotation angle rate by the fused intelligent particle cumulative rotation angle rate during the uniform deformation phase of the slope. The value of the fused intelligent particle cumulative rotation angle rate from the start of monitoring to the present monitoring moment is taken and continuously updated. When the slope enters the initial acceleration phase according to the tangent angle method of the cumulative rotation angle time series, the updating of the fused intelligent particle cumulative rotation angle rate during the uniform deformation phase of the slope is stopped.

[0041] According to the improved tangent angle of the fused cumulative rotation angle time series, the slope deformation stages are divided as follows:

[0042] Select adjacent monitoring data of the improved fusion cumulative rotation angle time series, calculate the slope between the two points, and convert it into an angle, i.e., the tangent angle;

[0043] When the tangent angle of the improved fused cumulative rotation angle time series is 45°, the slope is in the uniform deformation stage, and the warning level is blue, which is a caution warning;

[0044] When the tangent angle of the improved fused cumulative rotation angle time series is in the range of 45° to 80°, the slope is in the initial accelerated deformation stage, and the warning level is yellow or warning level warning;

[0045] When the tangent angle of the improved fused cumulative rotation angle time series is in the range of 80° to 85°, the slope is in the medium-accelerated deformation stage, and the warning level is orange, a warning level warning;

[0046] When the tangent angle of the improved fused cumulative rotation angle time series is in the range of 85° to 90°, the slope is in the accelerated deformation and imminent sliding stage, and the warning level is red, an alarm-level warning.

[0047] Furthermore, in step S6, based on the cumulative rotation angle time series and the slope failure moment estimation using the cumulative rotation angle inverse method, a quantitative analysis of the slope safety and stability is performed, as follows:

[0048] ① Based on the fused cumulative rotation angle time series, the cumulative rotation angle velocity time series is obtained by difference;

[0049] ② Take the inverse of the cumulative angular velocity time series;

[0050] ③ Perform linear fitting on the inverse time series of the cumulative angular velocity. The intersection of the fitting line and the horizontal axis, i.e., the time axis, represents the moment of slope failure.

[0051] When the slope is in the uniform deformation stage, no estimation of the slope failure moment is performed;

[0052] When the slope is in the initial accelerated deformation stage, the failure time estimated by the cumulative angular velocity and the inverse of the cumulative angular velocity method in this stage is the yellow and warning level failure time tcaution ;

[0053] When the slope is in the medium accelerated deformation stage, the estimated failure time using the angular velocity and the inverse method of the cumulative angular velocity in this stage is the orange and warning level failure time t warning ;

[0054] When the slope is in the stage of accelerated deformation and imminent sliding, the estimated failure time using the cumulative angular velocity and the inverse method of the cumulative angular velocity in this stage is the red and alarm level failure time t danger .

[0055] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0056] (1) A slope monitoring and early warning system was built based on intelligent particle sensors, roadside signal collectors, and cloud servers. It can sensitively analyze slope stability and issue early warnings, providing support for construction projects and operation and maintenance.

[0057] (2) Based on Z-score, Kalman filtering, cubic interpolation and adaptive fusion algorithms, the intelligent particle monitoring data was processed by outlier identification and elimination, smoothing and completion, and fusion, which improved the reliability of monitoring and early warning.

[0058] (3) In view of the uncertain relationship between sensor monitoring data and slope status, the accumulated rotation angle of smart particles is associated with the slope deformation stage based on the improved tangent angle method, which is used as a reference for relevant departments to take corresponding treatment measures.

[0059] (4) Estimate the time of slope failure in stages to provide time reference for relevant departments to evacuate the people. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments will be briefly introduced below. Obviously, the drawings in the following description only relate to some embodiments of the present invention, but are not intended to limit the present invention.

[0061] Figure 1 A technical roadmap for a slope safety monitoring and early warning method based on smart particles provided in this application;

[0062] Figure 2 This is the intelligent particle layout diagram of a single slope in this application example;

[0063] Figure 3 This is a schematic diagram of the slope deformation stage division based on the fusion of cumulative rotation angle time series and the improved tangent angle method in the example of this application;

[0064] Figure 4This is a schematic diagram of slope failure moment estimation based on the fusion of cumulative rotation angle time series and cumulative rotation angle velocity inverse method in this application example. DETAILED DESCRIPTION

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Unless otherwise defined, the technical or scientific terms used herein shall have the common meanings understood by persons of ordinary skill in the art to which the present invention belongs.

[0066] Example 1

[0067] like Figure 1-4 As shown, the present invention proposes a slope safety monitoring and early warning method based on smart particles. The method takes the smart particle sensor as the core and monitors and warns the slope through smart particle deployment, system construction, monitoring data preprocessing and calculation. The method specifically includes the following steps:

[0068] Step 1: Intelligent particle placement and monitoring system construction.

[0069] like Figure 2 As shown in the figure, a single slope with a height of 20m and a slope angle of 60° is constructed. Five section lines parallel to the slope direction are set, with a horizontal spacing of 20m between adjacent section lines. Section lines 1 and 5 are located outside the potential landslide range. An intelligent particle sensor is embedded at the top and foot of the slope, at depths of 5m and 2m, respectively. Section lines 2, 3, and 4 are located within the potential landslide range. An intelligent particle sensor is embedded at the top, mid-slope, and foot of the slope, at depths of 5m, 10m, and 2m, respectively. The drill holes are spaced 5m horizontally and 8m vertically on the slope surface.

[0070] Use a slope drilling machine to drill holes to the specified depth at designated monitoring points, embed smart particle sensors, and backfill the area in layers using a mixture of clay and fly ash mixed with half soil. The minimum thickness of each layer is 2m, and the maximum is 4m. After each layer is filled, it should be compacted before the next layer is added.

[0071] A remote roadside signal collector is installed in a geologically stable area near the monitored slope. Data is transmitted to the remote roadside signal collector using the ZigBee protocol and uploaded to the cloud in real time. The GLOBAL01 roadside signal collector, equipped with a battery and solar panel, is a terminal capable of outdoor reception, computational processing, remote monitoring, and cloud-based big data interfacing.

[0072] The intelligent particles are set to monitor once every 7 days, with each wake-up operation lasting 1 minute; during the flood season, rainy season and construction period of prevention and control projects, monitoring is intensified, once a day.

[0073] Step 2: Perform Z-score-based intelligent particle cumulative angle outlier identification.

[0074] Calculate the cumulative turning angle. For a certain monitoring time t i , calculate the local mean using the data points in the window and local standard deviation The angle distance time series of a certain intelligent particle sensor is but:

[0075]

[0076] For each angular distance value, calculate its Z-score:

[0077]

[0078] Set the threshold Z threshold =3, and identify outliers. If Exceeds the preset threshold Z threshold , then it is believed that If the outlier is detected, the data identified as outliers will be removed. The angle distance after removal is accumulated over time to obtain the cumulative angle time series after outlier removal.

[0079] Step 3: Intelligent particle cumulative rotation angle time registration based on Kalman filtering and cubic interpolation.

[0080] The accumulated rotation angle time series of each smart particle sensor is processed by Kalman filtering to remove noise and smooth the data. The rotation angle time series of a smart particle patent after Kalman filtering is: The unified time axis is composed of the awakening working time of all smart particle sensors. In general, the interval of monitoring data points is one week. During the flood season, rainy season and construction period of prevention and control projects, the monitoring is intensified, once a day, and the data interval is one day. At a certain time t i If a certain smart particle monitoring value is excluded due to abnormality, while the monitoring values ​​of other smart particles are normal and recorded and uploaded, the cumulative rotation angle of the smart particle at time t i Completion. The time t to be interpolated i As a benchmark, determine the two most recent cumulative rotation angle data forward and backward respectively, and construct a cubic polynomial F(x)=a i (t i -x) 3 +B i (ti -x) 2 +C i (t i -x)+D i ,, the independent variable of the function is x, which represents the monitoring time, and the dependent variable of the function is the accumulated rotation angle. Solve the coefficient A of the cubic polynomial i ,B i ,C i ,D i , so that the cubic polynomial satisfies: the function values ​​at the interpolation points are equal, The first-order derivative is continuous at the interpolation point, The second-order derivative is continuous at the interpolation point, Substituting the coefficient and monitoring time, we can get the complete cumulative angle data

[0081] Step 4: Intelligent particle cumulative corner fusion based on adaptive fusion algorithm.

[0082] The cumulative rotation angle time series of all smart particles are fused into one cumulative rotation angle time series for early warning. At the same monitoring moment, the cumulative rotation angles from different smart particles are fused according to the weight coefficient. The size of the weight coefficient depends on the difference from the average value of the cumulative rotation angle:

[0083]

[0084] In formulas (4)-(6), is the monitoring time t i The original cumulative rotation angle from the lth smart particle, there are L smart particles in total, is the monitoring time t i The variance of the average cumulative turning angle of the lth smart particle; is the monitoring time t i The weight coefficient of the cumulative turning angle of the lth smart particle; is the monitoring time t i Cumulative rotation angle of fused intelligent particles after adaptive weighted fusion.

[0085] Step 5: Slope deformation stage division based on the cumulative rotation angle time series and the improved tangent angle method.

[0086] like Figure 3As shown in Figure 2, the fused intelligent particle cumulative rotation angle time series is divided by the fused intelligent particle cumulative rotation angle rate during the uniform deformation phase of the slope to obtain the improved fused cumulative rotation angle time series. The cumulative rotation angle rate of the fused intelligent particles from the start of monitoring to the present is continuously updated. When the improved tangent angle method of the cumulative rotation angle time series indicates that the slope has entered the initial acceleration phase, the update of the cumulative rotation angle rate of the fused intelligent particles during the uniform deformation phase of the slope is stopped.

[0087] According to the improved tangent angle of the fused cumulative rotation angle time series, the slope deformation stages are divided. The division criteria are:

[0088] Select adjacent monitoring data of the improved fusion cumulative rotation angle time series, calculate the slope between the two points, and convert it into an angle, i.e., the tangent angle;

[0089] When the tangent angle of the improved fused cumulative rotation angle time series is ≈45°, the slope is in the uniform deformation stage, and the warning level is blue, a caution-level warning;

[0090] When the tangent angle of the improved fused cumulative rotation angle time series is in the range of 45° to 80°, the slope is in the initial accelerated deformation stage, and the warning level is yellow or warning level warning;

[0091] When the tangent angle of the improved fused cumulative rotation angle time series is in the range of 80° to 85°, the slope is in the medium-accelerated deformation stage, and the warning level is orange, a warning level warning;

[0092] When the tangent angle of the improved fused cumulative rotation angle time series is in the range of 85° to 90°, the slope is in the accelerated deformation and imminent sliding stage, and the warning level is red, an alarm-level warning.

[0093] Step 6: Prediction of slope failure time based on the cumulative rotation angle time series and the inverse method of cumulative rotation angle velocity.

[0094] like Figure 4 As shown in the figure, based on the fused cumulative rotation angle time series, the cumulative rotation angle velocity time series is obtained by differential calculation; the inverse of the cumulative rotation angle velocity time series is taken; a linear fit is performed on the inverse of the cumulative rotation angle velocity time series, and the intersection of the fitting line and the horizontal axis, that is, the time axis, represents the moment of slope failure.

[0095] When the slope is in the uniform deformation stage, no estimation of the slope failure moment is performed;

[0096] When the slope is in the initial accelerated deformation stage, the estimated failure time using the cumulative angular velocity and the inverse of the cumulative angular velocity method in this stage is the yellow or warning level failure time t caution ;

[0097] When the slope is in the medium accelerated deformation stage, the estimated failure time using the angular velocity and the inverse method of the cumulative angular velocity in this stage is the orange and warning level failure time t warning ;

[0098] When the slope is in the stage of accelerated deformation and imminent sliding, the estimated failure time using the cumulative angular velocity and the inverse method of the cumulative angular velocity in this stage is the red and alarm level failure time t danger .

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are included in the scope of protection of the present invention.

Claims

1. A slope safety monitoring and early warning method based on intelligent particles, characterized in that: The method comprises the following steps: S1. Drill holes in the slope and bury smart particle sensors, configure data transmission hardware equipment, and build an intelligent particle monitoring system; After starting up, the smart particles enter a periodic sleep and wake-up mode. The smart particles should be monitored once every 7 to 14 days, and each wake-up should last for 1 to 2 minutes. The indicator used for slope safety monitoring and early warning is the cumulative rotation angle of the smart particle sensor, which is calculated by the quaternion attitude data jointly calculated by the accelerometer, gyroscope and magnetometer data of the smart particle. The smart particle is at the monitoring time t i After waking up, it enters sleep mode and waits for the next wake-up. The smart particle records multiple quaternions according to the working frequency, a total of n. These quaternions are first averaged and then normalized as the monitoring time t i Quaternion data: In formulas (1)-(3), is the monitoring time t i At the beginning of the awakening process, the intelligent particle records the mth original quaternion data, a total of n, i, j, k are imaginary numbers, and their coefficients are is the real part of the mth original quaternion data; the monitoring time starts at t0 and the monitoring ends at t N , that is, there are N data points in the time series; represents the monitoring time t u The average quaternion data of the smart particles, the coefficients of the imaginary parts i, j, k are The real part is represents the monitoring time t i The normalized average quaternion data of the smart particles has a modulus of 1; The angular distance is calculated based on the quaternion of adjacent monitoring moments. The cumulative sum of the angular distances is the cumulative rotation angle of the smart particle over time: In formulas (4) and (5), is the previous monitoring time t i-1 To monitoring time t i angular distance; real is the name of the function that takes the real part of the quaternion; represents the previous monitoring time t i-1 The conjugate quaternion of the normalized average quaternion; SAD ti From the initial monitoring time t0 to the current monitoring time t i The total angular distance traveled by the smart particles is the cumulative turning angle. S2, based on the Z-score, identifies and eliminates outliers in the cumulative rotation angle of smart particles. This method detects and eliminates outliers in the angular distance and cumulative rotation angle time series calculated from the quaternion time series of each smart particle sensor. S3. Time-align the cumulative rotation angle time series of each smart particle sensor after removing outliers based on Kalman filtering and cubic interpolation, so that each smart particle sensor has a cumulative rotation angle value at the monitoring time; S4, processing the cumulative rotation angle time series of each intelligent particle sensor after time registration obtained in S3 using an adaptive fusion algorithm to obtain a fused cumulative rotation angle time series; S5. The fused cumulative rotation angle time series obtained in S4 is processed using the tangent angle method to obtain the division of slope deformation stages and conduct a qualitative analysis of the slope safety and stability. S6. The fused cumulative rotation angle time series obtained in S4 is processed using the inverse cumulative rotation angle velocity method to obtain an estimated value of the slope failure moment and perform a quantitative analysis of the slope safety and stability.

2. The method for slope safety monitoring and early warning based on smart particles according to claim 1 is characterized in that: In step S1, the smart particle is a sensor, which is equipped with an accelerometer, a gyroscope, a magnetometer, a thermometer and a pressure gauge. The external shape of the smart particle is a cube with an edge length of 27-40 mm and an apparent density of 2.18-2.30 g / cm 3 , the intelligent particle placement and monitoring system is built as follows: (2.1) Set up monitoring points. Drill holes on the slope surface and bury smart particle sensors. The drilling depth and the burial depth of the smart particles should be shallower than the potential sliding surface, with the lowest burial depth being 1m below the slope surface. The monitoring points should be arranged on cross-section lines that are parallel to or perpendicular to the slope direction and the landslide sliding direction, and should be arranged 10m outside the sliding influence range. For single slopes, cross-section lines should be arranged in the middle and on both sides of the slope or landslide, with no less than three cross-section lines. For strip slope groups, cross-section lines greater than the preset number should be selected. The spacing between monitoring points should be 5m to 40m, with the maximum horizontal spacing not exceeding 40m and the maximum vertical spacing not exceeding 20m. The preset number is obtained by dividing the length of the strip slope by the horizontal spacing; (2.2) Backfilling of monitoring points: using backfill materials to fill holes in layers, with each layer backfilled before the next layer. The backfill material is plain soil or prepared fly ash soil around the monitoring slope; (2.3) Monitoring system hardware configuration: remote roadside signal collectors are set up at preset locations on the monitoring slope. Data is transmitted to the remote roadside signal collectors based on the ZigBee protocol and uploaded to the cloud storage in real time.

3. The method for slope safety monitoring and early warning based on smart particles according to claim 2 is characterized in that: In step S2, for the cumulative rotation angle time series from each smart particle, the smart particle cumulative rotation angle outliers are identified and eliminated based on the Z-score standard score. The steps are as follows: ①Determine the sliding window size as N window ; ②For a certain monitoring time t i , calculate the local mean using the data points in the window and local standard deviation The angle distance time series of a certain intelligent particle sensor is That is, if the time series has N data points, then: ③For each angle distance value, calculate its Z-score: ④Set threshold Z threshold , if |Z t Exceeds the preset threshold Z threshold , then it is believed that If it is an outlier, the data determined as an outlier will be eliminated, and the angle distance after elimination will be accumulated according to time to obtain the cumulative angle time series after outlier elimination. The outlier is identified and eliminated for the angle distance of each smart particle, and the cumulative angle time series is calculated.

4. The method for slope safety monitoring and early warning based on smart particles according to claim 3 is characterized in that: In step S3, the intelligent particle cumulative rotation time registration based on Kalman filtering and cubic interpolation is performed as follows: ① Perform Kalman filtering on the cumulative rotation angle time series of each smart particle sensor to remove noise and smooth the data. The data point on the cumulative rotation angle time series of a smart particle after Kalman filtering is ② Unified time axis, which is composed of the wake-up working time of all smart particle sensors; ③ Perform cubic interpolation. For the cumulative rotation angle time series of a certain smart particle, the time t to be interpolated is i As a benchmark, determine the two most recent cumulative rotation angle data in time, and construct a cubic polynomial function F(x)=A i (t i -x) 3 +B i (t i -x) 2 +C i (t i -x)+D i , the independent variable of the function is x, which represents the monitoring time, and the dependent variable of the function is the cumulative rotation angle; ④ Solve the coefficient A of the cubic polynomial i ,B i ,C i ,D i , interpolate to obtain F(x), so that the cubic polynomial satisfies: the function values ​​at the interpolation points are equal, The first-order derivative is continuous at the interpolation point, The second-order derivative is continuous at the interpolation point, The cumulative rotation angle time series of all smart particles are time-aligned so that the cumulative rotation angle value of each smart particle exists at all monitoring moments.

5. The method for slope safety monitoring and early warning based on smart particles according to claim 4 is characterized in that: In step S4, the accumulated rotation angles of all intelligent particles are fused based on the adaptive fusion algorithm to merge the accumulated rotation angle time series of all intelligent particles into one accumulated rotation angle time series for early warning. At the same monitoring moment, the accumulated rotation angles from different intelligent particles are fused according to the weight coefficient. The size of the weight coefficient depends on the difference from the average value of the accumulated rotation angle: In formulas (9)-(11), is the monitoring time t i The original cumulative rotation angle from the lth smart particle, there are L smart particles in total, is the monitoring time t i The variance of the average cumulative turning angle of the lth smart particle; is the monitoring time t i The weight coefficient of the cumulative turning angle of the lth smart particle; is the monitoring time t i Cumulative rotation angle of fused intelligent particles after adaptive weighted fusion.

6. The method for slope safety monitoring and early warning based on smart particles according to claim 5 is characterized in that: In step S5, based on the cumulative rotation angle time series and the slope deformation stage division using the tangent angle method, a qualitative analysis of the slope safety and stability is performed as follows: The improved fused cumulative rotation angle time series is obtained by dividing the fused intelligent particle cumulative rotation angle rate by the fused intelligent particle cumulative rotation angle rate during the uniform deformation phase of the slope. The value of the fused intelligent particle cumulative rotation angle rate from the start of monitoring to the present monitoring moment is taken and continuously updated. When the slope enters the initial acceleration phase according to the tangent angle method of the cumulative rotation angle time series, the updating of the fused intelligent particle cumulative rotation angle rate during the uniform deformation phase of the slope is stopped. According to the improved tangent angle of the fused cumulative rotation angle time series, the slope deformation stages are divided as follows: Select adjacent monitoring data of the improved fusion cumulative rotation angle time series, calculate the slope between the two points, and convert it into an angle, i.e., the tangent angle; When the tangent angle of the improved fused cumulative rotation angle time series is 45°, the slope is in the uniform deformation stage, and the warning level is blue, which is a caution warning; When the tangent angle of the improved fused cumulative rotation angle time series is in the range of 45° to 80°, the slope is in the initial accelerated deformation stage, and the warning level is yellow, a warning level warning; When the tangent angle of the improved fused cumulative rotation angle time series is in the range of 80° to 85°, the slope is in the medium-accelerated deformation stage, and the warning level is orange, a warning level warning; When the tangent angle of the improved fused cumulative rotation angle time series is in the range of 85° to 90°, the slope is in the accelerated deformation and imminent sliding stage, and the warning level is red, an alarm-level warning.

7. The method for slope safety monitoring and early warning based on smart particles according to claim 6 is characterized in that: In step S6, based on the cumulative rotation angle time series and the slope failure moment estimation using the cumulative rotation angle inverse method, a quantitative analysis of the slope safety and stability is performed. The steps are as follows: ① Based on the fused cumulative rotation angle time series, the cumulative rotation angle velocity time series is obtained by difference; ② Take the inverse of the cumulative angular velocity time series; ③ Perform linear fitting on the inverse time series of the cumulative angular velocity. The intersection of the fitting line and the horizontal axis, i.e., the time axis, represents the moment of slope failure. When the slope is in the uniform deformation stage, no estimation of the slope failure moment is performed; When the slope is in the initial accelerated deformation stage, the failure time estimated by the cumulative angular velocity and the inverse of the cumulative angular velocity method in this stage is the yellow and warning level failure time t caution ; When the slope is in the medium accelerated deformation stage, the estimated failure time using the cumulative angular velocity and the inverse of the cumulative angular velocity method in this stage is the orange and warning level failure time t warning ; When the slope is in the stage of accelerated deformation and imminent sliding, the estimated failure time using the cumulative angular velocity and the inverse method of the cumulative angular velocity in this stage is the red and alarm level failure time t danger .

Images