A landslide time-to-failure prediction method based on adaptive time window and multi-criteria
By using an adaptive time window and a multi-criteria approach, the problems of sensitivity of tangent angle calculation to time windows and insufficient identification by a single criterion are solved, thus achieving high reliability and accuracy in landslide time prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING INST OF TECH
- Filing Date
- 2022-10-24
- Publication Date
- 2026-06-26
Smart Images
Figure CN115563802B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of alarm technology for responding to disaster events, and specifically to a method for predicting the imminent landslide time based on adaptive time windows and multiple criteria. Background Technology
[0002] Based on M. Saito's three-stage deformation theory of landslides, T.A. Fukuzono demonstrated through analysis and experiments that there is a linear relationship between the reciprocal of velocity and time during the landslide's imminent stage. As the landslide deformation rate increases, the predicted landslide time can be calculated using the reciprocal velocity method, but this prediction is only reliable when the landslide is in its imminent stage. Currently, the mainstream prediction method uses tangent angle values to determine the imminent stage of the landslide and the reciprocal velocity method for time prediction. This method has several drawbacks: 1) There is no consensus on the time window for calculating the tangent angle and velocity. Some researchers use the sampling time interval of monitoring data points, while others use fixed time windows (e.g., 3 hours, 7 hours). The tangent angle value is highly sensitive to the time window, and an inappropriate time window can easily lead to errors in the tangent angle calculation; 2) Relying solely on the tangent angle value as a criterion is insufficient for accurately identifying the imminent stage. Summary of the Invention
[0003] In view of this, the present invention provides a landslide imminent time prediction method based on adaptive time window and multiple criteria, which can effectively improve the reliability of landslide time prediction.
[0004] The landslide imminent time prediction method based on adaptive time windows and multiple criteria of the present invention includes:
[0005] S1, monitor the scene and obtain the deformation time curve; adaptively obtain the optimal time window; the specific methods for obtaining the optimal time window include:
[0006] S11: Continuously monitor the scene in real time and obtain the deformation time curve; set multiple fixed time windows, calculate the tangent angle of the deformation time curve under each fixed time window, move the fixed time window, and obtain the tangent angle curve corresponding to each fixed time window.
[0007] S12. Sort the fixed time windows in ascending order and calculate the overlap rate of the tangent angle curves under adjacent fixed time windows in turn; the fixed time window corresponding to the first time the overlap rate exceeds the set threshold B is the optimal time window.
[0008] S2, Based on the optimal time window obtained in step S1, calculate the tangent angle at each time point;
[0009] S3. Based on the optimal time window obtained in step S1, calculate the reciprocal of the velocity at each moment, and then obtain the Pearson correlation coefficient at each moment.
[0010] S4 extracts the moments when the tangent angle is greater than the threshold θ and the Pearson correlation coefficient is less than the threshold β, and outputs the landslide imminent time predicted for these moments using the inverse velocity method.
[0011] A better method for calculating the tangent angle under a fixed time window Δt is as follows:
[0012] First, calculate the cumulative deformation S. i Average rate B i :
[0013]
[0014] Among them, S1, S i These are time points t1 and t2 in the deformation time curve, respectively. i The cumulative deformation corresponding to each moment;
[0015] Then, use the cumulative deformation S i Divided by average speed B i To obtain T with dimensions consistent with time t i :
[0016]
[0017] Then t i At time t, the deformation-time curve is within a fixed time window Δt (Δt = t). i -t i-n The tangent angle α under) i for:
[0018]
[0019] Preferably, in step S12, the method for calculating the overlap rate of tangent angle curves within adjacent fixed time windows is as follows:
[0020] Calculate the difference in tangent angles on the tangent angle curves corresponding to adjacent time windows at the same time. If the difference is less than or equal to a set threshold A, the two tangent angle curves are considered to overlap at that time. The ratio of the overlap time of the two tangent angle curves to the total time is the overlap rate of the two curves.
[0021] Preferably, in S3, the Pearson correlation coefficient is calculated as follows:
[0022] Calculate the reciprocal of velocity IV under a fixed time window Δt. i :
[0023]
[0024] In equation (7), Δt is the optimal time window obtained by S1; S i S i-n The t values in the deformation time curves are respectivelyi Time and t i-n The cumulative deformation corresponding to each moment;
[0025] Then t i Pearson correlation coefficient r at time 1 i for:
[0026]
[0027] Among them, IV i-n ~IV i t i-n ~t i The speed reciprocal of time.
[0028] In S3, the reciprocal of the velocity at each moment is filtered to remove maximum and negative values.
[0029] Preferably, the threshold θ depends on the landslide type, θ∈[60°,85°].
[0030] Preferably, the threshold β depends on the landslide type, β∈[-0.90,-0.70].
[0031] Preferably, in step S1, the deformation time curve is obtained using ground-based GNSS equipment or ground-based interferometric radar.
[0032] Preferably, in step S1, ground-based interferometric radar is used to acquire the surface cumulative deformation of the monitoring scene; the tangent angle is determined based on the surface cumulative deformation, specifically including:
[0033] S101, Design multiple deformation thresholds based on the maximum value in the cumulative surface deformation;
[0034] S102, Obtain the area-time curves corresponding to each deformation threshold:
[0035] For each deformation threshold, the total area of pixels whose cumulative deformation exceeds the deformation threshold is calculated in real time. Each time point corresponds to a total area value of pixels that exceeds the deformation threshold, which forms an area-time curve as time progresses.
[0036] S103. Based on the adaptive time window method, for each fixed time window, calculate the tangent angle of the area-time curve corresponding to each deformation threshold at the current time. The largest of the tangent angles of the area-time curves corresponding to each deformation threshold is the tangent angle at the current time corresponding to the fixed time window.
[0037] Preferably, in step S101, N deformation thresholds are designed in the following manner:
[0038] The maximum cumulative deformation value S in the monitoring scenario during the current measurement period maxWhen ≤100mm, Deformation threshold S k =10·k, k=1~N, Indicates rounding down;
[0039] The maximum cumulative deformation value S in the monitoring scenario during the current measurement period max When >100mm, Deformation threshold
[0040] Beneficial effects:
[0041] (1) This invention first sets multiple time windows and selects the optimal time window based on the tangent angle time curve, avoiding tangent angle calculation errors and improving the intelligence and accuracy of landslide time prediction. Then, it uses the Pearson correlation coefficient as the second criterion for identifying the imminent landslide stage. That is, when the tangent angle value is large enough, the inverse velocity method is used to predict the landslide time. When the Pearson correlation coefficient reaches a certain range, the landslide time prediction result is input. This invention adaptively determines the optimal time window and simultaneously uses dual criteria to ensure the accuracy of imminent landslide stage identification, thereby improving the reliability of the prediction.
[0042] (2) Different tangent angle curves are judged by the distance between the tangent angles corresponding to the same time to determine whether the tangent angles coincide. This method is simple and highly operable.
[0043] (3) When using ground-based interferometric radar to obtain the surface cumulative deformation of the monitoring scene, the tangent angle can also be determined based on the surface cumulative deformation. Compared with the traditional early warning method based on single-point deformation information, the method of using area time series curves can identify the overall trend of landslide surface deformation. Its early warning information is more comprehensive and accurate, which can effectively improve the accuracy of landslide early warning stage judgment. Attached Figure Description
[0044] Figure 1 This is a flowchart of the present invention.
[0045] Figure 2 This is the tangent angle α curve under multiple time windows.
[0046] Figure 3 The overlap rate of the tangent angle α curve under multiple time windows.
[0047] Figure 4 This is the tangent angle α curve under the adaptive time window.
[0048] Figure 5 This represents the reciprocal of the velocity at each moment.
[0049] Figure 6 The velocity at each moment is the reciprocal (after filtering).
[0050] Figure 7 Here is the Pearson correlation coefficient at each time point.
[0051] Figure 8 For t i =Time forecast at 51.33h.
[0052] Figure 9 This is for landslide timing prediction based on adaptive time windows and multiple criteria.
[0053] Figure 10 Obtain a flowchart for adaptive time windows.
[0054] Figure 11 This is a method for obtaining the tangent angle based on the surface time curve. Detailed Implementation
[0055] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0056] This invention provides a method for predicting the imminent landslide time based on adaptive time windows and multiple criteria, comprising the following steps:
[0057] 1. Adaptive time window acquisition
[0058] When calculating the tangent angle based on the deformation time curve, the calculated value of the tangent angle is highly sensitive to the time window. For example, a time window that is too small can easily cause the tangent angle to jump, while a time window that is too large will lead to a lag in early warning information. Moreover, for deformation measurement equipment with different monitoring frequencies and deformation characteristics of different types of disasters, there is no fixed time window suitable for all conditions. Therefore, this step first adaptively selects the optimal time window from multiple time windows.
[0059] First, calculate the cumulative deformation S. i Average rate B i .
[0060]
[0061] Among them, S i S1 and S1 are respectively t i The cumulative deformation at time t1 and time t2.
[0062] Then, use the cumulative deformation S i Divided by average speed B i To obtain T with dimensions consistent with time t i .
[0063]
[0064] Calculate T i The tangent angle (slope) α on the time curve within the time window Δt. i αi It is a dimensionless natural number.
[0065]
[0066] In equation (3), Δt is the tangent angle time window, and α i For t i The tangent angle at time S i t i S represents the cumulative deformation and time corresponding to the i-th monitoring data, respectively. i-1 t i-1 These represent the cumulative deformation and time corresponding to the (i-1)th monitoring data point, respectively.
[0067] Now, Δt is set to multiple fixed values, and the method for calculating the tangent angle is changed from formula (3) to formula (4).
[0068]
[0069] In equation (4), Δt is a fixed value, and the value of n depends on Δt.
[0070] Then select m fixed time windows:
[0071] Δt j =j,j=1,2,3,…,m (5)
[0072] In equation (5), Δt j The unit is h, and m is set according to the deformation frequency of the monitored object.
[0073] Calculate the time window Δt j The corresponding tangent angle α i,j This yields m tangent angle-time curves.
[0074]
[0075] Sort the fixed time windows in ascending order and calculate t at each time point. i Below, the difference in tangent angle C between adjacent time windows i,j :
[0076] C i,j =|α i,j -α i,j+1 | (7)
[0077] C i,j When ≤3, t is considered to be i Time α i,j Curve and α i,j+1 The ratio of the overlap time to the total time of curve information is α. i,j Curve and α i,j+1 The overlap rate C of the curves j(Formula (8)). Curves with an overlap rate of 80% have high similarity, therefore when C j When ≥80% of the conditions are met for the first time, the model adaptively identifies j hours as the optimal time window.
[0078]
[0079] 2. Preliminary determination of tangent angle
[0080] Substitute the optimal time window into equation (3) to calculate the tangent angle α at each time point. If the tangent angle α is greater than or equal to the threshold θ, then time forecast calculation is required; otherwise, no time forecast is required. The value of the threshold θ depends on the landslide type, and is usually θ∈[60°, 85°].
[0081] 3. Secondary determination of Pearson correlation coefficient
[0082] The reciprocal of velocity at the point of landslide imminent has a linear relationship with time. By calculating the Pearson correlation coefficient of the reciprocal of velocity, the linear correlation between the reciprocal of velocity and time can be measured, which can determine whether a landslide is in an imminent state.
[0083] Speed countdown IV i The calculation uses the same time window as the tangent angle.
[0084]
[0085] In the above formula, Δt is the optimal time window calculated earlier; S i S i-n The t values in the deformation time curves are respectively i Time and t i-n The cumulative deformation corresponding to each moment.
[0086] Then t i Pearson correlation coefficient r at time 1 i for:
[0087]
[0088] Among them, IV i-n ~IV i t i-n ~t i The speed reciprocal of time.
[0089] If the Pearson correlation coefficient is less than or equal to the threshold β, then time forecast calculation is required; otherwise, no time forecast is required. The value of the threshold β depends on the landslide type, and is typically β∈[-0.90,-0.70].
[0090] 4. Output time forecast
[0091] When both the tangent angle and Pearson correlation coefficient meet the above conditions, a linear fit is performed on the scatter plot of the reciprocal velocity within the time window (the horizontal axis represents time, and the vertical axis represents the reciprocal velocity value). The intersection of the fitted line and the horizontal axis is the predicted time of disaster occurrence. Finally, the landslide forecast time is output.
[0092] The deformation time curve of the present invention can be obtained using ground GNSS equipment or ground-based interferometric radar (GB-InSAR).
[0093] Specifically, for monitoring using GB-InSAR, considering its technical advantage in monitoring surface deformation, and based on the characteristic of regular expansion of deformation areas during landslide development, a method for calculating the tangent angle based on area-time curves is proposed. By setting multiple deformation thresholds, the tangent angle calculation results from multiple deformation-time curves are fused to obtain the optimal tangent angle for landslide early warning stage identification. Compared to traditional early warning methods based on single-point deformation information, the method using area-time curves can identify the overall trend of landslide surface deformation, providing more comprehensive and accurate early warning information, and effectively improving the accuracy of landslide early warning stage identification. The specific method flow is as follows: Figure 6 As shown, the specific steps include the following:
[0094] S101, Design Deformation Threshold
[0095] GB-InSAR is used to monitor the scene and obtain the cumulative deformation of the scene within the current time window Δt. The maximum value S of the cumulative deformation of the scene within the current time window Δt is then used. max Design N deformation thresholds: The number and value of deformation thresholds can be determined empirically or through numerical simulation. This embodiment is designed as follows:
[0096] When S max When ≤100mm, Deformation threshold S k =10·k, k=1~N, This indicates rounding down to the nearest integer.
[0097] When S max When >100mm, Deformation threshold
[0098] S102, Obtain the area-time curve
[0099] Deformation-time curves based on single-point monitoring can effectively identify the landslide hazard state at a point during slope deformation. However, after a significant slope failure, the deformation rate at a single point may slow down, but the deformation area may still continue to expand. Using area-time curves can more accurately identify the overall deformation trend of the landslide surface.
[0100] For the cumulative deformation within the time window Δt, for each deformation threshold S k Calculate the time t at each time point within the time window Δt. i The cumulative deformation exceeds S k The total area A of the pixels k,i All A within the time window Δt k,i The set of these is the area-time curve A. k (t). During the landslide development process, A k (t) is an increasing curve, A k The growth rate of (t) is the expansion rate of the deformed region.
[0101] S103, Calculate the tangent angle of the time curve for area calculation
[0102] For each deformation threshold S k The corresponding area-time curve A k (t), calculate the area-time curve A k (t) from the initial time t1 to each time t i average speed B k,i (Formula (11)).
[0103]
[0104] Plot the vertical axis A of the area-time curve by dividing the area by the average velocity. k,i Transformed to T with dimensions consistent with the horizontal axis time t k,i (Formula (12)).
[0105]
[0106] Calculate t i The tangent angle (slope) α at time t is... k,i (Formula (13)), α k,i It is a dimensionless natural number.
[0107]
[0108] S104, Obtain the warning judgment tangent angle
[0109] t i At time t, each deformation threshold S k There is a corresponding tangent angle αk,i N deformation thresholds can yield N tangent angles, and the maximum tangent angle α is... i As the tangent angle of the time window (Formula (14)).
[0110]
[0111] Case
[0112] Taking the surface deformation data of the slope of the Jianshan Iron Mine in Shanxi Province monitored by ground-based interferometric radar as an example, the calculation process of landslide imminent time prediction under adaptive time window and multiple criteria is as follows:
[0113] 1) Adaptive Time Window Calculation: Set multiple time windows, Δt = 1h, 2h, 3h, 4h, 5h, 6h, ..., 10h. The tangent angle α curve under multiple time windows is shown in [reference needed]. Figure 2 The overlap rate of the tangent angle α curve under multiple time windows is shown in [reference needed]. Figure 3 The tangent angle α curves under the 4h and 5h time windows overlap first, reaching 80%, therefore the adaptive time window is 4h.
[0114] 2) Criterion 1 Tangent Angle: Substitute the 4-hour time window into formula (3) to calculate the tangent angle value at each time point, see Figure 4 The threshold θ in criterion 1 is set to 65°. Figure 4 The red portion corresponds to moments when the tangent angle is greater than 65°.
[0115] 3) Calculate the reciprocal of velocity: Substitute the 4-hour time window into formula (9) to calculate the reciprocal of velocity at each moment, see... Figure 5 .right Figure 5 The reciprocal of the velocity is filtered to remove maximum and negative values. In this embodiment, values greater than 2000 h / m and negative values are removed to obtain the effective reciprocal of the velocity ( Figure 6 ).
[0116] 4) Criterion 2: Pearson correlation coefficient: Substitute the 4-hour time window and the reciprocal of the filtered velocity into formula (10) to obtain the Pearson correlation coefficient r at each time point. i ,See Figure 7 The threshold β in criterion 2 is set to -0.75. Figure 7 The red portion corresponds to the time when the correlation coefficient is less than -0.75.
[0117] 5) Time forecast calculation: in t i Taking the time 51.33h as an example, the corresponding tangent angle and correlation coefficient are 67.79° and -0.98, respectively, satisfying the dual criteria for the skid-like stage. Therefore, time prediction calculation is required. A linear fit is performed on the reciprocal of the velocity within the 4-hour time window; the intersection of the fitted line and the horizontal axis is t. i =53.41h( Figure 8That is: t i At time 51.33h, t is calculated. i A major deformation is expected at 53.41 hours, and a forecast was issued 2.98 hours in advance.
[0118] 6) Time prediction under dual criteria:
[0119] Time forecast calculations are performed and forecast results are output when the tangent angle is ≥65° and the correlation coefficient is ≤-0.75; otherwise, no forecast is given. Figure 9 To monitor the forecast situation at various times, the horizontal axis is 51.33h and the vertical axis is 2.98h.
[0120] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for predicting the imminent landslide time based on adaptive time windows and multiple criteria, characterized in that, include: S1, monitor the scene and obtain the deformation time curve; Adaptively obtain the optimal time window; specifically, the methods for obtaining the optimal time window include: S11: Continuously monitor the scene in real time and obtain the deformation time curve; set multiple fixed time windows, calculate the tangent angle of the deformation time curve under each fixed time window, move the fixed time window, and obtain the tangent angle curve corresponding to each fixed time window. S12. Sort the fixed time windows in ascending order and calculate the overlap rate of the tangent angle curves under adjacent fixed time windows in turn; the fixed time window corresponding to the first time the overlap rate exceeds the set threshold B is the optimal time window. S2, Based on the optimal time window obtained in step S1, calculate the tangent angle at each time point; S3. Based on the optimal time window obtained in step S1, calculate the reciprocal of the velocity at each moment, and then obtain the Pearson correlation coefficient at each moment. S4, extract tangent angles greater than the threshold. θ And the Pearson correlation coefficient is less than the threshold. β At each moment, output the landslide imminent time predicted using the reciprocal velocity method.
2. The landslide imminent time prediction method based on adaptive time window and multiple criteria as described in claim 1, characterized in that, Fixed time window The method for calculating the tangent angle is as follows: First, calculate the cumulative deformation. average rate : in, S 1. S i The deformation time curves are respectively t 1 moment and t i The cumulative deformation corresponding to each moment; Then, use cumulative deformation Divide by average speed Gain and Time t Dimensionally consistent : but t i At a given time, the deformation-time curve within a fixed time window ( tangent angle below for: 。 3. The landslide imminent time prediction method based on adaptive time window and multiple criteria as described in claim 1, characterized in that, In step S12, the method for calculating the overlap rate of tangent angle curves within adjacent fixed time windows is as follows: Calculate the difference in tangent angles on the tangent angle curves corresponding to adjacent time windows at the same time. If the difference is less than or equal to a set threshold A, it is considered that the two tangent angle curves overlap at that time. The ratio of the overlap time of the two tangent angle curves to the total time is the overlap rate of the two curves.
4. The landslide imminent prediction method based on adaptive time windows and multiple criteria as described in claim 1, 2, or 3, characterized in that, In S3, the Pearson correlation coefficient is calculated as follows: Calculate a fixed time window The reciprocal of the speed of the downward movement : （7） In equation (7), The optimal time window obtained for S1; S i , S i-n The deformation time curves are respectively t i Time and t i-n The cumulative deformation corresponding to each moment; but Pearson correlation coefficient at time 1 for: in, They are respectively ~ The speed reciprocal of time.
5. The landslide imminent prediction method based on adaptive time window and multiple criteria as described in claim 1, characterized in that, In S3, the reciprocal of the velocity at each moment is filtered to remove maximum and negative values.
6. The landslide imminent prediction method based on adaptive time window and multiple criteria as described in claim 1, characterized in that, The threshold θ Depending on the type of landslide, .
7. The landslide imminent prediction method based on adaptive time window and multiple criteria as described in claim 1, characterized in that, The threshold β Depending on the type of landslide, .
8. The landslide imminent prediction method based on adaptive time window and multiple criteria as described in claim 1, characterized in that, In S1, the deformation time curve is obtained using ground-based GNSS equipment or ground-based interferometric radar.
9. The landslide imminent prediction method based on adaptive time window and multiple criteria as described in claim 8, characterized in that, In step S1, ground-based interferometric radar is used to acquire the surface cumulative deformation of the monitoring scene; Determining the tangent angle based on the cumulative deformation of the surface includes: S101, Design multiple deformation thresholds based on the maximum value in the cumulative surface deformation; S102, Obtain the area-time curves corresponding to each deformation threshold: For each deformation threshold, the total area of pixels whose cumulative deformation exceeds the deformation threshold is calculated in real time. Each time point corresponds to a total area value of pixels that exceeds the deformation threshold, which forms an area-time curve as time progresses. S103. Based on the adaptive time window method, for each fixed time window, calculate the tangent angle of the area-time curve corresponding to each deformation threshold at the current time. The largest of the tangent angles of the area-time curves corresponding to each deformation threshold is the tangent angle at the current time corresponding to the fixed time window.
10. The landslide imminent prediction method based on adaptive time window and multiple criteria as described in claim 9, characterized in that, In S101, the design is as follows: N Deformation threshold: Maximum cumulative deformation value in the monitoring scenario during the current measurement period S max When ≤100 mm, Deformation threshold , Indicates rounding down; Maximum cumulative deformation value in the monitoring scenario during the current measurement period S max When >100 mm, Deformation threshold .