A landslide early warning method and system
By acquiring data through tilt sensors, a tilt angle and time fitting function is constructed to calculate the tilt rate and tangent angle. This solves the problem of low sensitivity in the initial deformation stage of landslide early warning methods, and achieves timeliness and accuracy in landslide early warning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CENT SOUTHERN CHINA ELECTRIC POWER DESIGN INST CHINA POWER ENG CONSULTING GROUP CORP
- Filing Date
- 2026-03-31
- Publication Date
- 2026-05-29
AI Technical Summary
Existing landslide early warning methods have low sensitivity in the initial deformation stage of landslides, making it difficult to capture deformation acceleration. They are also poorly adaptable to sudden landslides and easily affected by monitoring noise, which limits the accuracy and robustness of early warning.
Data is acquired using tilt sensors. By constructing a tilt angle and time fitting function, the tilt rate and tangent angle are calculated. The optimal fitting function is selected in conjunction with the Akaike Information Criterion to determine the landslide warning level, suppress noise interference, and improve the quality of monitoring data.
It improves the timeliness and accuracy of landslide early warning, can sensitively reflect the early slight deformation of the landslide body, and forms a complete chain from data collection to risk assessment. The logic is clear and it is easy to implement in engineering.
Smart Images
Figure CN122116606A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of landslide early warning technology, and in particular to a landslide early warning method and system. Background Technology
[0002] my country's power transmission and transformation projects are widely distributed in mountainous and hilly areas with complex geological conditions and frequent tectonic activity. Landslides occur frequently due to multiple factors, including heavy rainfall and human-induced disturbances, seriously threatening the safe and stable operation of power grid infrastructure. Landslide monitoring and early warning, as a key link in geological disaster prevention and control, aims to accurately identify the evolutionary characteristics of a landslide body from initial deformation to accelerated instability, thereby achieving early identification and dynamic assessment of disaster risks.
[0003] Existing landslide early warning methods mostly employ the traditional criterion based on the displacement-time series tangent angle. This involves calculating the arctangent of the ratio of displacement increment to time increment to determine whether a landslide has entered an accelerated deformation stage. While this method has some applicability in some gently changing landslides, it essentially relies solely on the first-order characteristics of displacement, making it difficult to effectively capture deformation acceleration—a key precursor to instability. It exhibits significantly insufficient sensitivity in the early stages of non-uniform deformation, poor adaptability to abrupt landslides, and susceptibility to monitoring noise interference. Furthermore, it lacks consideration of the coupling of multiple inducing factors such as rainfall and groundwater, resulting in limited accuracy and robustness in early warning systems.
[0004] Current technologies still face multiple challenges in early landslide identification and dynamic risk assessment: First, displacement signals change weakly during the initial deformation stage of a landslide, making it difficult to reflect internal structural damage in a timely manner; second, traditional tangent angle models heavily rely on accurate identification during the uniform deformation stage, while actual landslide processes often exhibit nonlinear and unsteady-state characteristics, leading to decreased model applicability; third, relying solely on displacement indicators cannot fully characterize the entire evolution mechanism of a landslide body from deformation to rupture; finally, monitoring data is susceptible to environmental noise interference, and without effective filtering and reconstruction methods, the reliability of early warning criteria will be directly affected. These problems collectively restrict the real-time performance, accuracy, and engineering practicality of landslide early warning systems under complex geological and meteorological conditions. Summary of the Invention
[0005] The purpose of this invention is to provide a landslide early warning method and system to solve the problem of low sensitivity in the early stage of slope deformation in existing landslide early warning methods and systems.
[0006] To solve the above-mentioned technical problems, the present invention provides a landslide early warning method, comprising: Acquire tilt angle monitoring data collected by tilt angle sensors in the landslide body and perform preprocessing; Obtain the time series and the corresponding tilt angle series; Construct a tilt angle and time fitting function based on the time series and tilt angle series: Calculate the tilt rate data sequence by differentiating the fitting function for tilt angle and time; Calculate the slope of the tilt rate data sequence; and, The tangent angle is calculated based on the slope of the inclination rate data sequence, and the landslide warning level is determined.
[0007] Optionally, in constructing the tilt angle and time fitting function based on the time series and tilt angle series, the tilt angle and time fitting function is one of a first-order polynomial function, a second-order polynomial function, and an exponential function.
[0008] Optionally, constructing the tilt and time fitting functions based on the time series and tilt series includes: calculating the Akaike Information Criterion (AIC) values for a first-order polynomial function, a second-order polynomial function, and an exponential function, respectively; selecting the function with the smallest Akaike Information Criterion (AIC) value among the first-order polynomial function, second-order polynomial function, and exponential function as the optimal fitting function; if the AIC values of the first-order polynomial function, second-order polynomial function, and exponential function are equal, then the first-order polynomial function is selected as the optimal fitting function; if the AIC values of two of the first-order polynomial function, second-order polynomial function, and exponential function are equal and the smallest, and one of them has the AIC of a first-order polynomial function, then the first-order polynomial function is selected as the optimal fitting function; if it does not contain the AIC of a first-order polynomial function, then the exponential function is selected as the optimal fitting function.
[0009] Optionally, calculating the tangent angle and determining the landslide warning level based on the slope of the tilt rate data sequence includes: calculating the arctangent value of the slope of the tilt rate data sequence, using the arctangent value of the slope of the tilt rate data sequence as the tangent angle, dividing the landslide deformation stage according to the magnitude of the tangent angle and its changing trend, and outputting the corresponding warning level signal accordingly.
[0010] Optionally, the landslide warning level classification criteria are as follows: when the tangent angle is less than the first preset angle threshold, it is determined to be the initial deformation stage, and a blue warning is output; when the tangent angle is between the first preset angle threshold and the second preset angle threshold and shows a continuous upward trend, it is determined to be the uniform deformation stage, and a yellow warning is output; when the tangent angle is greater than the second preset angle threshold and the rate of increase accelerates, it is determined to be the accelerated instability stage, and a red warning is output.
[0011] The present invention also provides a landslide early warning system employing the above-described landslide early warning method, comprising: The data acquisition and preprocessing module is used to acquire and preprocess the tilt monitoring data collected by the tilt sensors in the landslide body. The acquisition module is used to obtain a time series and a tilt angle series corresponding to the time series; The fitting function construction module is used to construct a tilt angle and time fitting function based on the time series and tilt angle series: The tilt rate data sequence calculation module is used to calculate the tilt rate data sequence by differentiating the tilt angle and time fitting function; The slope calculation module is used to calculate the slope of the tilt rate data sequence; and The warning level determination module is used to calculate the tangent angle based on the slope of the tilt rate data sequence and determine the landslide warning level.
[0012] Optionally, in the inclination angle and time fitting function constructed by the fitting function construction module based on the time series and inclination angle series, the inclination angle and time fitting function is one of a first-order polynomial function, a second-order polynomial function, and an exponential function.
[0013] Optionally, the fitting function construction module constructs the tilt angle and time fitting functions based on the time series and tilt angle series, including: calculating the Akaike Information Criterion (AIC) values of the first-order polynomial function, the second-order polynomial function, and the exponential function, respectively; selecting the function with the smallest Akaike Information Criterion (AIC) value among the first-order polynomial function, the second-order polynomial function, and the exponential function as the optimal fitting function; if the AIC values of the first-order polynomial function, the second-order polynomial function, and the exponential function are equal, then the first-order polynomial function is selected as the optimal fitting function; if the AIC values of two of the first-order polynomial function, the second-order polynomial function, and the exponential function are equal and the smallest, and one of them has the AIC of the first-order polynomial function, then the first-order polynomial function is selected as the optimal fitting function; if it does not contain the AIC of the first-order polynomial function, then the exponential function is selected as the optimal fitting function.
[0014] Optionally, the warning level determination module includes: The arctangent calculation module calculates the arctangent value of the slope of the slope rate data sequence; A tangent angle calculation module is used to calculate the arctangent of the slope of the tilt rate data sequence as the tangent angle; and, The warning level signal output module is used to classify the landslide deformation stage based on the magnitude and trend of the tangent angle, and output the corresponding warning level signal accordingly.
[0015] Optionally, the standard for the warning level output module to classify warning levels is as follows: When the tangent angle is less than the first preset angle threshold, it is determined to be the initial deformation stage, and a blue warning is output; when the tangent angle is between the first preset angle threshold and the second preset angle threshold and shows a continuous upward trend, it is determined to be the uniform deformation stage, and a yellow warning is output; when the tangent angle is greater than the second preset angle threshold and the rate of increase accelerates, it is determined to be the acceleration and instability stage, and a red warning is output.
[0016] The landslide early warning method and system provided by this invention have the following beneficial effects: Using dip angle data instead of traditional displacement data can more sensitively reflect early, subtle deformations in the deep or critical areas of a landslide, improving the timeliness of early warning. Secondly, by constructing a dip angle and time fitting function and resampling, smooth and continuous reconstructed data is generated, effectively suppressing random noise and environmental interference in the original monitoring data, improving data quality, and laying a solid foundation for subsequent accurate analysis. Finally, this method connects three levels of features—dip angle, tilt rate, and tangent angle—to form a complete chain from data acquisition and processing to risk assessment, with clear logic and ease of engineering implementation. Attached Figure Description
[0017] Figure 1 This is a flowchart of the landslide early warning method in an embodiment of the present invention. Detailed Implementation
[0018] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, but these embodiments should not be construed as limiting the present invention.
[0019] refer to Figure 1 , Figure 1 This is a flowchart of a landslide early warning method in an embodiment of the present invention. This embodiment provides a landslide early warning method, including: Step S100: Obtain the tilt angle monitoring data collected by the tilt angle sensor in the landslide body and perform preprocessing; Step S200: Obtain the time series and the tilt angle series corresponding to the time series; Step S300: Construct a tilt angle and time fitting function based on the time series and tilt angle series: Step S400: Calculate the tilt rate data sequence by differentiating the tilt angle and time fitting functions; Step S500: Calculate the slope of the tilt rate data sequence; and, Step S600: Calculate the tangent angle based on the slope of the tilt rate data sequence and determine the landslide warning level.
[0020] First, using dip angle data instead of traditional displacement data can more sensitively reflect early, subtle deformations in the deeper or critical areas of a landslide, improving the timeliness of early warning. Second, by constructing a dip angle and time fitting function and resampling, smooth and continuous reconstructed data is generated, effectively suppressing random noise and environmental interference in the original monitoring data, improving data quality, and laying a solid foundation for subsequent accurate analysis. Finally, this method connects three levels of features—dip angle, tilt rate, and tangent angle—to form a complete chain from data acquisition and processing to risk assessment, with clear logic and ease of engineering implementation.
[0021] Among them, the tilt rate sequence characterizes the change in the deformation rate of the landslide body.
[0022] In step S100, the tilt monitoring data collected by the tilt sensors in the landslide body is acquired and preprocessed. The tilt sensors are deployed near the potential sliding surface or key deformation areas of the landslide body. The sampling frequency is not lower than a preset frequency threshold, and the accuracy of a single measurement is better than a preset accuracy threshold. Data transmission uses wired or low-power wireless communication to ensure data integrity and timeliness. Deploying the sensors near the potential sliding surface or key deformation areas ensures that the monitoring data can directly reflect the core deformation characteristics of the landslide body, improving the representativeness and effectiveness of the monitoring. Setting the sampling frequency and accuracy threshold ensures the sufficiency and accuracy of data acquisition, providing a data foundation for subsequent high-precision analysis. Using wired or low-power wireless communication ensures data transmission stability while meeting the long-term, low-power monitoring requirements in harsh field environments, improving the system's engineering applicability.
[0023] Preferably, tilt sensors are deployed at both the top and bottom of the landslide. Deploying tilt sensors simultaneously at both the top and bottom of the landslide allows for the acquisition of deformation differences at different locations. By comparing and analyzing the rate and direction of tilt change at the top and bottom, a more comprehensive understanding of the overall movement pattern and deformation field distribution of the landslide can be obtained, effectively identifying the difference between local sliding and overall instability, and providing multi-source data support for more refined landslide risk assessment.
[0024] In step S100, the preprocessing of tilt monitoring data collected by tilt sensors in the landslide body involves filtering and noise reduction. The filtering and noise reduction methods are selected from one of the following: weighted moving average, wavelet denoising, local weighted regression, SG digital filtering algorithm, and Empirical Mode Decomposition (EMD) filtering method, with weighted moving average being the preferred method. Several mature and reliable data filtering and noise reduction algorithms are provided as options. For example, weighted moving average is suitable for smoothing stationary trend data, wavelet denoising is suitable for processing non-stationary signals, and EMD filtering is suitable for adaptive decomposition of nonlinear and non-stationary signals. This flexible selectivity allows the method of this invention to select the most suitable preprocessing method based on the characteristics of the actual monitoring data (such as noise type and trend), thereby maximizing the preservation of the true deformation signal and improving the accuracy of data reconstruction.
[0025] The weighted moving average method assigns different weights to each observation value, calculates a moving average based on these weights, and uses this final moving average as the basis for determining the forecast value. The weighted moving average method is used because recent observations within the observation period have a greater impact on the forecast value, better reflecting recent market trends. Therefore, observations closer to the forecast period are given larger weights, while observations further away are given smaller weights. These different weights adjust the influence of each observation on the forecast value, allowing the forecast value to more closely reflect future market trends. The formula for the weighted moving average method is as follows: Among them, p t x: The smoothed value after weighted moving average at time t. n: The half-length of the moving window, indicating that n data points are taken before and after time t to participate in the averaging, and the total window length is 2n+1 (the first n points + the current point + the next n points). t-i : The original observation at time ti (the i-th time before t). t+i : The original observation at time t+i (the i-th time after t). t : The original observation value at time t itself. w t : Corresponding to x t The weight represents the degree of influence of the current observation. t-i : Corresponding to x t-i The weight represents the degree of influence of the historical observation on the smoothing result. t+i : Corresponding to x t+i The weight represents the degree of influence of the future observation on the smoothing result.
[0026] The above formula represents averaging each observation by its weight. This method is suitable when the observations themselves have a confidence level. The weighted moving average method is sensitive to recent trends, but if a set of data has a significant seasonal effect, the predicted values obtained using the weighted moving average method may be biased. Therefore, this method is not suitable for data affected by seasonality.
[0027] Typically, in the front-end sensing layer, multiple high-precision tilt sensor arrays are deployed near the potential sliding surface of the landslide (usually located 1–3 meters below the interface between the sliding bed and the sliding mass) and in key deformation areas (such as the toe, shoulder, and crack propagation front). Each sensor array uses a triaxial MEMS (Micro-Electro-Mechanical Systems) tiltmeter with a single-axis measurement range of ±90°, static accuracy better than ±0.005°, dynamic resolution up to 0.001°, and a sampling frequency configurable from 1 to 10 times per hour, with a default setting of 1 time per hour to meet analytical standards. The tilt sensor housing is encapsulated in IP68-rated stainless steel and incorporates a built-in temperature compensation module to suppress zero-point drift caused by diurnal temperature variations. Each sensor connects to the edge computing node via an RS-485 bus or a LoRaWAN low-power wide-area network (LPWAN): For sites less than 500 meters away where cabling is feasible, shielded twisted-pair cables are used to form an RS-485 daisy-chain topology, with a power supply voltage of DC 24V and a signal level conforming to the TTL standard; for dispersed deployments or areas with terrain obstructions, LoRa wireless modules (operating frequency band 470–510MHz, transmit power ≤20dBm) are used, with data packet format following a custom binary protocol, including timestamp, sensor ID, triaxial tilt angle values (represented by 16-bit signed integers in units of 0.001°) and checksum, ensuring communication reliability and anti-interference capability in complex mountainous environments.
[0028] In step S200, the time series and the tilt angle sequence corresponding to the time series are obtained, and the time series set X = {x} i ,x i+1 ,x i+2 ,…,x i+23}, where x i This represents the i-th monitoring time (i=1,2,…,n-23, where n is the total number of monitoring samples, 26≤n). The set of tilt angle sequences corresponding to each timestamp sequence set T is Y={y i ,y i+1 ,y i+2 ,…,y i+23}, y i The tilt angle value measured at the i-th monitoring time.
[0029] In step S300, when constructing the tilt angle and time fitting function based on the time series and tilt angle series, the tilt angle and time fitting function is one of a first-order polynomial function, a second-order polynomial function, and an exponential function.
[0030] In step S300, constructing the tilt and time fitting functions based on the time series and tilt series includes: calculating the Akaike Information Criterion (AIC) values for a first-order polynomial function, a second-order polynomial function, and an exponential function, respectively; selecting the function with the smallest Akaike Information Criterion (AIC) value among the first-order polynomial function, second-order polynomial function, and exponential function as the optimal fitting function; if the AIC values of the first-order polynomial function, second-order polynomial function, and exponential function are equal, then the first-order polynomial function is selected as the optimal fitting function; if the AIC values of two of the first-order polynomial function, second-order polynomial function, and exponential function are equal and the smallest, and one of them has the AIC value of a first-order polynomial function, then the first-order polynomial function is selected as the optimal fitting function; if it does not contain the AIC value of a first-order polynomial function, then the exponential function is selected as the optimal fitting function. For example, during a period of continuous rainfall, the dip angle data showed a clear accelerating upward trend. At this time, the AIC value of the exponential model was significantly lower than that of the first-order polynomial function and the second-order polynomial function. The system then selected the exponential model as the fitting equation for the current period, ensuring that the model neither overfits nor underfits, accurately characterizing the nonlinear deformation features. The adaptive selection of the fitting function was achieved through the Akaike Information Criterion (AIC). This dynamic function selection strategy greatly enhances the adaptability of the method of this invention to different landslide deformation characteristics (linear, nonlinear, data anomalies), ensuring the reliability and effectiveness of the data substitution model.
[0031] Among them, the AIC of linear polynomial functions, quadratic polynomial functions, and exponential functions is AIC = nln(RSS / n) + 2p. Where RSS is the sum of squared residuals of the corresponding function, n is the total number of data points, and p is the number of parameters to be estimated in the function (including the intercept term). For a linear polynomial function Where S(k,d) is equal to RSS, x i For time, y i Let be the inclination angle, K be the rate of change of the line, and d be the intercept (the fitted value when x=0).
[0032] For quadratic polynomial functions , where x i For time, y i Let A be the inclination angle, B be the coefficient of the quadratic term, and C be the coefficient of the linear term. Let be the predicted value of the fitting function at the i-th sample, and n be the total number of data points.
[0033] For exponential functions, , where xi is time, yi is tilt angle, n is the total number of data points, a is the initial value coefficient, b is the exponential growth rate, and is the predicted value of the fitting function at the i-th sample.
[0034] The linear polynomial function is described below: y=Kx+d Where x is time, y is the inclination angle, K is the rate of change of the line, and d is the intercept (the fitted value when x=0).
[0035] The least squares method is used to solve for the parameters K and d, i.e., the fitting objective of the first-order polynomial function is to minimize the sum of squared residuals S(K,d). S(K,d) is the sum of squared residuals, x i For time, y i Let be the inclination angle, K be the rate of change of the line, and d be the intercept (the fitted value when x=0).
[0036] right and Taking the partial derivative and setting it to zero, we get: The formula for calculating the K value is: in, The average value over time. This represents the average angle of inclination.
[0037] The formula for calculating the d value is: in, The average value over time. This represents the average angle of inclination.
[0038] The quadratic polynomial function is as follows: y = Ax 2 +Bx+C Where x is time, y is the tilt angle, A is the coefficient of the quadratic term, B is the coefficient of the linear term, and C is the constant term.
[0039] The least squares method is used to solve for parameters A, B, and C, i.e., the objective of fitting the quadratic polynomial function is to minimize the sum of squared residuals S(A,B,C). Where xi is time, yi is tilt angle, A is quadratic coefficient, B is linear coefficient, C is constant term, is the predicted value of the fitting function at the i-th sample, and n is the total number of data points.
[0040] right , and Taking the partial derivatives and setting them to zero, we obtain the following system of equations: right Find the partial derivative: right Find the partial derivative: right Find the partial derivative: After simplification, we obtain the standard system of equations: The above system of equations can be expressed in matrix form. : Where M is the sum of x i The constructed 3×3 coefficient matrix, where N is the coefficient of x i y i Construct a constant term vector, X=[A, B, C] T Let be the vector of coefficients to be determined.
[0041] Solving for: Calculate the coefficients: in: , , , , , , , , Among them, SS xx Let x be the sum of squared deviations from the mean. SS yy Let y be the sum of squared deviations from the mean. SS xy Let x and y be the cross sum of their deviations from the mean. SS x2x For x 2 The sum of the crosses with x, SS x2y For x 2 The sum of the intersections with y, SS x2x2 For x 2 The sum of squared deviations from the mean.
[0042] The exponential function is as follows: Where a is the initial value coefficient, b is the exponential growth rate, x is time, and y is the tilt angle.
[0043] The least squares method is used to solve for the parameters a and b, that is, the goal of the exponential function fitting is to minimize the sum of squared residuals S(a,b).
[0044] in, Taking the natural logarithm of the exponential function: lny = lna + bx, and letting Y = lny and A = lna, we transform it into a linear model Y = A + bx.
[0045] For the transformed data (x) i lny i Perform linear least squares fitting to solve for A and b, then restore a=e. A In step S400, the tilt rate data sequence is calculated by differentiating the fitting functions for tilt angle and time. The analytical differentiation process uses corresponding mathematical expressions for different fitting function forms: the derivative of a first-order polynomial is a constant, the derivative of a second-order polynomial is a linear function, and the derivative of an exponential function is itself multiplied by the exponential coefficient. The resulting tilt rate is expressed in degrees per unit time. This scheme clarifies the specific implementation of analytical differentiation in engineering practice. Precise derivative calculation formulas are provided for common fitting models such as first-order polynomials, second-order polynomials, and exponential functions. This allows those skilled in the art to directly apply the corresponding differentiation rules based on the type of fitting function constructed, efficiently and accurately obtaining the tilt rate function and quantifying the result into a data sequence with clear physical meaning (degrees per unit time), enhancing the operability and practicality of the method.
[0046] In calculating the tilt rate data sequence by differentiating the fitting function (optimal fitting function) for tilt angle and time, if the optimal fitting function is a first-order polynomial y = kx + d, then the tilt rate ω is dy / dx = k. As mentioned above, the tilt rate sequence ω = {k} can be obtained using the least squares method. i ,k i+1 , …,k n-23}; If the optimal fitting function is a quadratic polynomial y = Ax 2 +Bx+C, then the tilt rate ω is dy / dx =2Ax+B (a linear function). Based on A, B, and x obtained above, the tilt rate sequence ω={2A... i x i +B i 2A i+1 x i+1 +B i+1, …,2A n-23 x n-23 +B n-23}; If the optimal fitting function is the exponential function Y = a·e bx Then the tilt rate ω is dy / dx = ab·e bx Based on a, b, and x obtained above, the tilt rate sequence ω = {a} is obtained. i b i e bixi , a i+1 b i+1 e bi+1xi+1 , …,a n-23 b n-23 e bn -23xn-23}
[0047] In step S500, the slope of the tilt rate data sequence is calculated by taking three adjacent data points as a group and calculating the slope m of the tilt rate sequence in sequence, which is the surrogate index of acceleration.
[0048] Among them, t i Let ω be the slope of the skew rate sequence formed by grouping three adjacent data points at the i-th monitoring time (i=1,2,…,n-23, where n is the total number of monitoring samples, 26≤n). i θ represents the tilt rate.
[0049] Substituting the obtained slope rate sequence ω and the corresponding time series T into the above formula, the set of slopes M={m} of the slope rate data sequence is calculated. i ,m i+1 ,…, m n-25}
[0050] In step S600, calculating the tangent angle and determining the landslide warning level based on the slope of the tilt rate data sequence includes: Step S610: Calculate the arctangent of the slope of the tilt rate data sequence; Step S620: Use the arctangent of the slope of the tilt rate data sequence as the tangent angle; Step S630: Divide the landslide deformation stage according to the magnitude and trend of the tangent angle, and output the corresponding early warning level signal accordingly.
[0051] First, by using the arctangent of the slope of the tilt rate data sequence as the tangent angle, the changing trend of the tilt rate can be captured in real time, avoiding misjudgments caused by non-equidistant data or local anomalies. This scheme makes the division of landslide deformation stages more objective and stable, thereby improving the accuracy and reliability of early warning level determination.
[0052] In step S620, the tangent angle , where θ is the tangent angle and M is the set of slopes of the tilt rate data sequence.
[0053] In step S630, the landslide early warning level classification standard is as follows: when the tangent angle is less than a first preset angle threshold, it is determined to be in the initial deformation stage, and a blue warning is output; when the tangent angle is between the first and second preset angle thresholds and shows a continuous upward trend, it is determined to be in the uniform deformation stage, and a yellow warning is output; when the tangent angle is greater than the second preset angle threshold and the rate of increase accelerates, it is determined to be in the accelerated instability stage, and a red warning is output. A clear and quantifiable mapping relationship is established between landslide deformation stages (initial deformation, uniform deformation, accelerated instability) and warning levels (blue, yellow, red). This standard, based on the numerical magnitude and changing trend of the tangent angle, is logically rigorous and highly consistent with the physical process of landslide evolution (from slow initiation to uniform development and then to accelerated destruction). This clear criterion makes the warning results intuitive and easy to understand, providing direct and effective support for disaster prevention decision-making.
[0054] Specifically, if all θ < 5° and there is no continuous upward trend (defined as 3 consecutive θ increments > 0.5°), it is determined to be the initial deformation stage, and a blue warning is output. If there exists θ∈[5°,15°], and θ shows a monotonically increasing trend over the past 6 hours (then it is determined to be a uniform deformation stage, and a yellow warning is issued); If any θ > 15°, and the θ of the previous moment is less than the current moment (i.e., the growth rate is accelerating), then it is determined to be in the acceleration and instability stage, and a red warning is issued.
[0055] The system of this invention is deployed in high-risk landslide areas along power transmission and transformation projects. It uploads the early warning results to the power grid geological disaster monitoring platform in real time through a remote communication module. The response delay does not exceed a preset time threshold and supports multi-point concurrent monitoring and centralized management.
[0056] This embodiment also provides a landslide early warning system employing the landslide early warning method described in the above embodiments, including: The data acquisition and preprocessing module is used to acquire and preprocess the tilt monitoring data collected by the tilt sensors in the landslide body. The acquisition module is used to obtain a time series and a tilt angle series corresponding to the time series; The fitting function construction module is used to construct a tilt angle and time fitting function based on the time series and tilt angle series: The tilt rate data sequence calculation module is used to calculate the tilt rate data sequence by differentiating the tilt angle and time fitting function; The slope calculation module is used to calculate the slope of the tilt rate data sequence; and The warning level determination module is used to calculate the tangent angle based on the slope of the tilt rate data sequence and determine the landslide warning level.
[0057] At the start of each analysis period (e.g., 00:00 UTC+8 daily), the data acquisition and preprocessing module is activated. This module first reads the accumulated raw tilt data points (1 time / hour) from each tilt sensor over the past n hours. Subsequently, the module performs filtering and noise reduction processing using a weighted moving average method.
[0058] For each set of timestamp sequences X and tilt sequence Y (there are a total of n-23 sets of X and Y sequences), this module sequentially tries three candidate functions: a first-order polynomial (y = kx + d), a second-order polynomial (y = Ax... 2 +Bx+C), exponential function (y= a·e) bt ), x i Substitute into the formula, and at the same time, x i Corresponding y i Substitute into the formula.
[0059] For each function, the parameters (k,d,A,B,C,a,b) are solved using the least squares method, and the corresponding K={k i ,k i+1 , …,K n-23 Let d = {di, d i+1, …, dn-23}, A = {Ai, A i+1, …, An-23}, B = {Bi, Bi+1, …, Bn-23}, C = {Ci, Ci+1, …, Cn-23}, a = {ai, a i+1, …, an-23}, b = {bi, bi+1, …, bn-23}, and calculate the corresponding determination coefficients Akaike Information Criterion (AIC).
[0060] The optimal fitting function is automatically selected according to preset rules: the function with the smallest Akaike Information Criterion (AIC) value among the first-order polynomial, second-order polynomial, and exponential functions is selected as the optimal fitting function; if the AIC values of the first-order polynomial, second-order polynomial, and exponential functions are equal, the first-order polynomial function is selected as the optimal fitting function; if the AIC values of two of the first-order polynomial, second-order polynomial, and exponential functions are equal and the smallest, and one of them is a first-order polynomial function, the first-order polynomial function is selected as the optimal fitting function; if the AIC value of the first-order polynomial function is not included, the exponential function is selected as the optimal fitting function.
[0061] After selecting the optimal fitting function through the above calculations, the tilt rate data sequence is calculated by taking the derivative of the optimal fitting function through the tilt rate data sequence calculation module.
[0062] If the optimal fitting function is a first-order polynomial y = kx + d, then the tilt rate ω is dy / dx = k. As mentioned above, the tilt rate sequence ω = {k} has already been obtained using the least squares method. i ,k i+1 , …,k n-23}; If the optimal fitting function is a quadratic polynomial y = Ax 2 +Bx+C, then the tilt rate ω is dy / dx =2Ax+B (a linear function). Based on A, B, and X obtained above, the tilt rate sequence ω={2A... i x i +B i 2A i+1 x i+1 +B i+1 , …,2A n-23 x n-23 +B n-23}; If the optimal fitting function is the exponential function Y = a·e bx Then the tilt rate ω is dy / dx = ab·e bx Based on a, b, and X obtained above, the tilt rate sequence ω = {a i b i e bixi , a i+1 b i+1 e bi+1xi+1 , …,a n-23 b n-23 e bn -23xn-23}
[0063] The slope of the above tilt rate data sequence is calculated. Specifically, the slope m of the tilt rate sequence is calculated sequentially for every three adjacent data points, which is a proxy index of acceleration.
[0064] Among them, t i Let ω be the slope of the skew rate sequence formed by grouping three adjacent data points at the i-th monitoring time (i=1,2,…,n-23, where n is the total number of monitoring samples, 26≤n). i θ represents the tilt rate.
[0065] Substituting the obtained slope rate sequence ω and the corresponding time series T into the above formula, the set of slopes M={m} of the slope rate data sequence is calculated. i ,m i+1 ,…, m n-25} Then, the tangent angle is calculated, and the result is stored internally in radians. Finally, it is converted to degrees for engineering interpretation.
[0066] Among them, the tangent angle .
[0067] The risk assessment logic is based on the current value of the tangent angle and its changing trend: If all θ < 5° and there is no continuous upward trend (defined as 3 consecutive θ increments > 0.5°), it is determined to be the initial deformation stage, and a blue warning is output. If there exists θ∈[5°,15°], and θ shows a monotonically increasing trend over the past 6 hours (then it is determined to be a uniform deformation stage, and a yellow warning is issued); If any θ > 15°, and the θ of the previous moment is less than the current moment (i.e., the growth rate is accelerating), then it is determined to be in the acceleration and instability stage, and a red warning is issued.
[0068] The warning level determination module includes: The arctangent calculation module calculates the arctangent value of the slope of the slope rate data sequence; A tangent angle calculation module is used to calculate the arctangent of the slope of the tilt rate data sequence as the tangent angle; and, The warning level signal output module is used to classify the landslide deformation stage based on the magnitude and trend of the tangent angle, and output the corresponding warning level signal accordingly.
[0069] The data acquisition and preprocessing module acquires and preprocesses the tilt monitoring data collected by the tilt sensors in the landslide body. The tilt sensors are deployed near the potential sliding surface or key deformation area of the landslide body, with a sampling frequency not lower than the preset frequency threshold and a single measurement accuracy better than the preset accuracy threshold. Data transmission adopts wired or low-power wireless communication to ensure data integrity and timeliness.
[0070] In this embodiment, the calculation of the landslide early warning method is implemented through a landslide early warning system. This system is deployed in an explosion-proof cabinet on-site and includes a main control processor, storage unit, power management module, and local human-machine interface. The main control processor is an industrial-grade ARM Cortex-A53 quad-core SoC (such as the NXP i.MX8M Plus), with a clock speed of 1.8GHz, integrating a hardware floating-point unit (FPU) and a neural network accelerator (although AI training is not enabled in this solution, redundant computing power is retained). The processor runs an embedded Linux system (customized by the Yocto Project) and is equipped with real-time data processing middleware. The storage unit consists of a 32GB eMMC flash memory and a 128GB microSD card; the former stores the operating system and applications, while the latter cyclically caches raw monitoring data (retaining the most recent 30 days) and supports power outage protection. The power management module is connected to a solar-battery hybrid power supply system (100W photovoltaic panel, 24V / 20Ah lithium battery pack), with low-voltage shutdown and overcharge protection functions, ensuring continuous operation for ≥7 days without mains power. The local human-machine interface includes a 4.3-inch TFT touchscreen and status indicator LEDs for on-site commissioning and fault diagnosis.
[0071] The remote communication layer is responsible for uploading early warning results to the power grid geological disaster monitoring platform. The landslide early warning system has a built-in 4GCat.1 communication module (supporting China Mobile / China Unicom / China Telecom networks), establishing a TLS encrypted connection via the MQTT protocol to push structured JSON data packets containing early warning levels, tangent angle timing, and environmental parameters to the cloud server. The communication strategy adopts an "event-triggered + timed heartbeat" mechanism: it immediately reports when the early warning level changes or data quality alarms occur; under normal conditions, a heartbeat packet is sent every 10 minutes to ensure link activity. The system is designed with a response delay of no more than 10 minutes to meet the timeliness requirements of power grid dispatch for geological disaster emergency response.
[0072] The above description is merely a description of preferred embodiments of the present invention and is not intended to limit the scope of the present invention in any way. Any changes or modifications made by those skilled in the art based on the above disclosure shall fall within the protection scope of the claims.
Claims
1. A landslide early warning method, characterized in that, include: Acquire tilt angle monitoring data collected by tilt angle sensors in the landslide body and perform preprocessing; Obtain the time series and the corresponding tilt angle series; Construct a tilt angle and time fitting function based on the time series and tilt angle series: Calculate the tilt rate data sequence by differentiating the fitting function for tilt angle and time; Calculate the slope of the tilt rate data sequence; and, The tangent angle is calculated based on the slope of the inclination rate data sequence, and the landslide warning level is determined.
2. The landslide early warning method as described in claim 1, characterized in that, In constructing the tilt angle and time fitting function based on the time series and tilt angle series, the tilt angle and time fitting function is one of a first-order polynomial function, a second-order polynomial function, and an exponential function.
3. The landslide early warning method as described in claim 2, characterized in that, The process of constructing the tilt and time fitting functions based on the time series and tilt series includes: calculating the Akaike Information Criterion (AIC) values for a first-order polynomial function, a second-order polynomial function, and an exponential function, respectively; selecting the function with the smallest Akaike Information Criterion (AIC) value among the first-order polynomial function, second-order polynomial function, and exponential function as the optimal fitting function; if the AIC values of the first-order polynomial function, second-order polynomial function, and exponential function are equal, then the first-order polynomial function is selected as the optimal fitting function; if the AIC values of two of the first-order polynomial function, second-order polynomial function, and exponential function are equal and the smallest, and one of them has the AIC value of a first-order polynomial function, then the first-order polynomial function is selected as the optimal fitting function; if it does not contain the AIC value of a first-order polynomial function, then the exponential function is selected as the optimal fitting function.
4. The landslide early warning method as described in claim 1, characterized in that, Calculating the tangent angle and determining the landslide warning level based on the slope of the tilt rate data sequence includes: calculating the arctangent value of the slope of the tilt rate data sequence, using the arctangent value of the slope of the tilt rate data sequence as the tangent angle, classifying the landslide deformation stage according to the magnitude of the tangent angle and its changing trend, and outputting the corresponding warning level signal accordingly.
5. The landslide early warning method as described in claim 4, characterized in that, The landslide warning level classification criteria are as follows: when the tangent angle is less than the first preset angle threshold, it is determined to be the initial deformation stage, and a blue warning is output; when the tangent angle is between the first preset angle threshold and the second preset angle threshold and shows a continuous upward trend, it is determined to be the uniform deformation stage, and a yellow warning is output; when the tangent angle is greater than the second preset angle threshold and the rate of increase accelerates, it is determined to be the accelerated instability stage, and a red warning is output.
6. A landslide early warning system employing the landslide early warning method as described in any one of claims 1-5, characterized in that, include: The data acquisition and preprocessing module is used to acquire and preprocess the tilt monitoring data collected by the tilt sensors in the landslide body. The acquisition module is used to obtain a time series and a tilt angle series corresponding to the time series; The fitting function construction module is used to construct a tilt angle and time fitting function based on the time series and tilt angle series: The tilt rate data sequence calculation module is used to calculate the tilt rate data sequence by differentiating the tilt angle and time fitting function; The slope calculation module is used to calculate the slope of the tilt rate data sequence; as well as The warning level determination module is used to calculate the tangent angle based on the slope of the tilt rate data sequence and determine the landslide warning level.
7. The landslide early warning system as described in claim 6, characterized in that, The fitting function construction module constructs an inclination and time fitting function based on the time series and inclination sequence. The inclination and time fitting function is one of a first-order polynomial function, a second-order polynomial function, and an exponential function.
8. The landslide early warning system as described in claim 7, characterized in that, The fitting function construction module constructs tilt angle and time fitting functions based on the time series and tilt angle series, including: calculating the Akaike Information Criterion (AIC) values for a first-order polynomial function, a second-order polynomial function, and an exponential function, respectively; selecting the function with the smallest Akaike Information Criterion (AIC) value among the first-order polynomial function, second-order polynomial function, and exponential function as the optimal fitting function; if the AIC values of the first-order polynomial function, second-order polynomial function, and exponential function are equal, then the first-order polynomial function is selected as the optimal fitting function; if the AIC values of two of the first-order polynomial function, second-order polynomial function, and exponential function are equal and the smallest, and one of them has the AIC value of a first-order polynomial function, then the first-order polynomial function is selected as the optimal fitting function; if it does not contain the AIC value of a first-order polynomial function, then the exponential function is selected as the optimal fitting function.
9. The landslide early warning system as described in claim 6, characterized in that, The warning level determination module includes: The arctangent calculation module calculates the arctangent value of the slope of the slope rate data sequence; A tangent angle calculation module is used to calculate the arctangent of the slope of the tilt rate data sequence as the tangent angle; and, The warning level signal output module is used to classify the landslide deformation stage based on the magnitude and trend of the tangent angle, and output the corresponding warning level signal accordingly.
10. The landslide early warning system as described in claim 9, characterized in that, The standard for classifying early warning levels by the early warning level signal output module is as follows: When the tangent angle is less than the first preset angle threshold, it is determined to be the initial deformation stage, and a blue warning is output; when the tangent angle is between the first preset angle threshold and the second preset angle threshold and shows a continuous upward trend, it is determined to be the uniform deformation stage, and a yellow warning is output; when the tangent angle is greater than the second preset angle threshold and the rate of increase accelerates, it is determined to be the acceleration and instability stage, and a red warning is output.