A method for predicting creep failure of landslides
By constructing a landslide creep model with time-varying parameters and dynamically updating it with real-time monitoring data, the accuracy and reliability issues of landslide creep failure prediction in existing technologies have been solved, achieving accurate prediction and adaptive early warning throughout the entire life cycle.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA THREE GORGES UNIV
- Filing Date
- 2026-02-28
- Publication Date
- 2026-06-19
AI Technical Summary
Existing landslide creep failure prediction methods lack unified quantitative criteria, are susceptible to environmental noise, cannot cover the creep stages of the entire life cycle, and cannot dynamically update model parameters, resulting in increased prediction errors and making it difficult to meet the long-term safety monitoring needs of engineering projects.
A creep model with time-varying parameters is constructed. Observable physical quantities during the landslide creep process are collected by deploying monitoring equipment. The model parameters are dynamically updated, and rolling predictions are achieved by combining real-time monitoring data. A unified quantitative description and mathematical and physical basis for the creep stage are established.
It achieves a unified quantitative description of the creep behavior of landslides at all stages, enhances the accuracy and reliability of prediction, can adapt to changes in the external environment, is suitable for complex load scenarios, and supports integration with automated monitoring systems.
Smart Images
Figure CN122241060A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of geological disaster monitoring and early warning technology, specifically to a method for predicting landslide creep failure. Background Technology
[0002] Landslides, a typical and frequent geological hazard worldwide, generally exhibit a full-stage creep characteristic of "deceleration-isolation-acceleration" in their evolution. Predicting the failure time during the accelerated creep stage is crucial for mitigating casualties and reducing economic losses in engineering projects. However, existing landslide creep failure prediction methods mainly focus on three directions: empirical model fitting, physical and mechanical mechanism analysis, and single monitoring index criteria. These methods still have the following limitations in practical engineering applications: First, existing methods often rely on expert experience or single monitoring indicators to define creep stages, lacking unified quantitative criteria and easily susceptible to misjudgments due to environmental noise and local disturbances. Second, traditional prediction models are mostly designed for the accelerated stage at the end of creep, failing to cover the iso- and deceleration stages, which are crucial for early stability assessment, thus limiting their application in landslide life-cycle monitoring. Furthermore, most of the model parameters are empirical constants and lack clear physical and mechanical significance; thirdly, most existing methods make one-time predictions based on monitoring data for fixed periods, and cannot update model parameters and prediction results on a rolling basis according to newly added real-time monitoring data. This makes it difficult to cope with sudden changes in landslide state caused by factors such as rainfall infiltration, reservoir water level fluctuations, and changes in external loads, resulting in the prediction results gradually increasing in error over time, which cannot meet the needs of long-term safety monitoring of engineering projects.
[0003] Therefore, in view of the shortcomings and deficiencies of the existing technologies, there is an urgent need to develop a landslide creep failure prediction method that is universal, theoretical and dynamic. By constructing a unified cross-stage creep model, clarifying the physical meaning of the parameters, and combining real-time monitoring data to achieve dynamic update prediction, the application limitations of traditional methods can be broken through, and the accuracy and reliability of landslide failure prediction can be improved. Summary of the Invention
[0004] The purpose of this invention is to solve the technical problems mentioned above and to propose a method for predicting landslide creep failure, comprising the following steps: S1. Construct a creep model incorporating time-varying parameters, damage accumulation, and displacement potential: (1); In the formula, Let Ω(t) be the creep rate, Ω(t) be the creep displacement, A(t) be the time-varying parameter characterizing damage accumulation, and n(t) be the time-varying parameter characterizing displacement potential. S2. By deploying monitoring equipment at key parts of the landslide, time-series monitoring data of the observable physical quantity Ω(t) during the landslide creep process are collected; S3. Based on the time-series monitoring data collected in step S2, solve the time-varying parameters n(t) and A(t) in the landslide creep model, and determine the current creep stage of the landslide based on the value of n(t). S4. When step S3 determines that the landslide is currently in the accelerated creep stage, i.e., n(t) > 1, the characteristic sequence is calculated using the current value of n(t) and the corresponding Ω(t) data. ; S5, Feature Sequence Using the ordinate as the vertical axis and the corresponding time t as the horizontal axis, perform linear regression to obtain a fitted line; extrapolate this fitted line to the ordinate. The intersection of the x-coordinates at the point is the predicted landslide failure time tf; S6. As new time-series monitoring data is continuously acquired, repeat steps S3 to S5 to continuously update the time-varying parameters, feature sequences, and predicted failure time tf.
[0005] In the preferred scheme, in step S1, the creep model The universality of this method was verified in the following way: Different types of landslide examples were selected, and the results of each example were analyzed. Transform Ω(t) time series data into After obtaining the coordinate data, the least squares method was used for linear fitting. The fitting results all satisfied R² > 0.9 and p < 0.05, indicating that the rate-displacement evolution of different types of landslides all followed the power law law of this model. When a landslide is in motion, regardless of whether it is in the acceleration, constant speed or deceleration stage, the value of Ω(t) is always positive. By linearly fitting the log rate and log displacement data, a unified fit of the entire creep process can be achieved.
[0006] In the preferred embodiment, in step S2, the observable physical quantity Ω(t) is selected from surface displacement or deep displacement; Monitoring equipment includes GNSS fully automatic displacement monitoring equipment, inclinometers, or crack gauges; Monitoring points were set up in the main sliding zone of the landslide, the tension cracks at the rear edge, and the key location at the shear outlet.
[0007] In the preferred embodiment, step 3 specifically includes: landslide creep model Taking the logarithm of both sides, we get the linearized expression: (2); exist In the coordinate system, a sliding window linear fitting is performed using the time-series monitoring data collected in step 2. The window length is adaptively adjusted according to the monitoring frequency and landslide deformation characteristics. The time-varying parameter n(t) is directly solved through fitting. ; The creep stage is determined based on the time-varying parameter n(t): when n(t)≈0, it is determined to be the constant-rate creep stage; when n(t)>1, it is determined to be the accelerated creep stage; when n(t)<0, it is determined to be the decelerated creep stage. Once a landslide enters the accelerated creep stage, the damage becomes irreversible under continuous load, leading to deformation and eventual failure.
[0008] In the preferred embodiment, in step S3, during the sliding window linear fitting, a differentiated monitoring time interval is adopted for different creep stages: in the accelerated creep stage, the monitoring time interval is set to less than or equal to 24 hours; in the constant-speed and deceleration creep stages, the monitoring time interval is set to less than or equal to 48 hours.
[0009] In the preferred embodiment, step S3: n(t) represents the displacement potential energy. In the constant creep stage, n(t)≈0; in the accelerated creep stage, n(t)>1; in the decelerated creep stage, n(t)<0; after the landslide creep terminates, n(t) resets to the initial state; the decrease of n(t) indicates that the displacement potential energy is being consumed rapidly; when the external load exceeds the critical threshold, the energy state of the system changes abruptly, causing n(t) to increase sharply. A(t) characterizes the degree of damage accumulation. An increase in its value indicates the continuous accumulation of internal damage, making the landslide more prone to deformation at the same displacement level. In the initial creep stage, A(t) → 0, indicating that the landslide has high strength and stability. In the accelerated failure stage, A(t) gradually increases, indicating that the landslide has lower stability. When the landslide creep terminates, A(t) will reset to the initial state.
[0010] In the preferred scheme, the parameters n(t) and A(t) evolve according to the shear strength of the soil and rock mass structure, the periodic fluctuation of the reservoir water level, the rainfall infiltration, and the changes in external loads.
[0011] In the preferred embodiment, in step S5, based on the landslide creep model Derive the expression for the remaining failure time due to landslide creep; the derivation process specifically includes: Treat A(t) and n(t) as constants A and n, and solve the model; When n≠1, the relationship between creep rate and time is: (3); In the formula, Ω0 is the displacement at the initial time t0; When in the accelerated creep stage and n>1, assume that the creep displacement at failure time tf tends to infinity. →∞), the remaining failure time from the current time t is derived from equation (3). expression: (4); In the formula, tf represents the landslide failure time. This represents the displacement at the current time t; Linear extrapolation prediction method based on A linear negative correlation was established with time t during the accelerated creep phase. When n≠1, equation (3) is rewritten in the form of a standard linear equation, i.e., the expression for creep failure time: (5); In the formula, the slope and intercept All are instantaneous constants; When n < 1, It is positively correlated with time t; when n>1, It is negatively correlated with time t, and as t increases, Gradually decrease; Will Linear extrapolation of the temporal relationship with time t is performed to extrapolate to... When the intersection point is reached, the time corresponding to the landslide prediction failure time tf is taken. The feature sequence is performed only if n(t) > 1. The calculation; when n(t) < 1, It is positively correlated with time t. The fact that the value gradually increases but does not decrease and approaches zero indicates that the landslide will not fail within a finite time.
[0012] In the preferred embodiment, in step S5, the feature sequence is... When performing linear regression, the goodness of fit R2 is required to be greater than 0.9, and the extrapolation prediction of failure time tf is made with a confidence interval of 95% or higher. When the goodness of fit R2 < 0.9, the old data needs to be removed so that R2 > 0.9.
[0013] In the preferred embodiment, in step S6, the dynamic update process can automatically identify the deformation trend change caused by the change of external load; through rolling calculation, multiple continuous time-varying parameter n(t) sequences and corresponding predicted failure times tf are obtained, so that the prediction results can flexibly adapt to the dynamic changes of the landslide state, thereby realizing the continuous correction of the early warning information.
[0014] Compared with the prior art, the beneficial effects of the present invention are: by constructing a creep model containing time-varying parameters... This achieves a unified quantitative description and objective diagnosis of creep behavior across all stages of landslide deceleration, constant velocity, and acceleration. Based on mathematical derivation, a creep stage system was established. The linear theoretical relationship with time t elevates failure time prediction from empirical fitting to analytical extrapolation with clear mathematical and physical basis. Simultaneously, the model parameters possess clear physical meaning (damage accumulation and displacement potential), and their dynamic evolution is directly related to the internal damage mechanism of the landslide, enhancing the model's interpretability. This method supports rolling updates and dynamic corrections based on real-time monitoring data, enabling it to adapt to changes in the external environment. It is suitable for complex, non-constant load scenarios: in actual geological disasters, stress fields are complex and difficult to measure directly and accurately, while displacement can be directly and continuously monitored using GNSS, InSAR, and other methods. Furthermore, this method has a clear process, high computational efficiency, and is easily integrated with automated monitoring systems, demonstrating significant engineering practical value. Attached Figure Description
[0015] Figure 1 This is a schematic diagram of the landslide creep failure prediction method in the embodiments of this specification.
[0016] Figure 2 This is a graph showing the rate-displacement power law relationship of the landslide in the deceleration, constant velocity, and accelerated creep stages in the embodiments of this specification.
[0017] Figure 3 This refers to n(t) and n(t) at various creep stages of a landslide in the embodiments of this specification. Change curve graph.
[0018] Figure 4 yes Figure 3 Predicted results of creep failure of a landslide in China during days 1 to 13. Detailed Implementation
[0019] Combination Figures 1-4 The specific embodiments of the present invention will be further described in detail below. A method for predicting landslide creep failure, such as... Figure 1 As shown, the method specifically includes the following steps: Step 1: Establish a landslide creep model. The landslide creep model is used to describe the rate of landslide creep. The power-law relationship between creep and displacement Ω(t) applies to all types of landslides in all stages of creep (deceleration, constant velocity, acceleration), and is expressed as follows: (1) In the formula, Ω(t) represents the creep rate, Ω(t) represents the creep displacement (surface displacement or deep displacement), A(t) represents the time-varying parameter characterizing damage accumulation, and n(t) represents the time-varying parameter characterizing displacement potential. creep model The universality of this method was verified in the following way: Different types of landslide examples were selected, and the results of each example were analyzed. Transform Ω(t) time series data into After obtaining the coordinate data, a linear fit is performed using the least squares method. For example... Figure 2 As shown, the universality of the model was verified through multiple landslide examples (including soil landslides and rock landslides). Data from multiple landslides at deceleration, isochronous, and accelerated creep stages were presented. The figures generally show a significant linear relationship (R² > 0.9, p < 0.05), indicating that although landslide creep modes vary, their rate-displacement evolution follows a power-law law according to this model. When the landslide is in motion ( When the creep traverse is in the acceleration, constant speed, or deceleration phase, the value of Ω(t) is always positive. Therefore, it is safe to take the logarithm to transform it into a linear relationship, thereby achieving a uniform fit for the entire creep process.
[0020] Step 2: Collect landslide creep time-series monitoring data. Using monitoring equipment deployed at key locations of the landslide, collect time-series monitoring data of the observable physical quantity Ω(t) during the landslide creep process; The observable physical quantity Ω(t) is selected from surface displacement or deep displacement; the monitoring equipment includes GNSS fully automatic displacement monitoring equipment, inclinometers or crack gauges; monitoring points are deployed in the main sliding zone of the landslide, the rear edge tension cracks, and key locations at the shear exit. The raw monitoring data can be time-series smoothed to remove noise interference and obtain a more robust parameter estimation and extrapolation basis.
[0021] Step 3: Solve for time-varying parameters and determine the creep stage. Based on the time-series monitoring data collected in Step 2, solve for the time-varying parameters n(t) and A(t) in the landslide creep model, and determine the current creep stage of the landslide based on the value of n(t). For the landslide creep model Taking the logarithm of both sides, we get the linearized expression: (2) exist In the coordinate system, a sliding window linear fitting is performed using the time-series monitoring data collected in step 2. The window length is adaptively adjusted according to the monitoring frequency and landslide deformation characteristics. The time-varying parameter n(t) is directly solved through fitting. ; In the sliding window linear fitting, different monitoring time intervals are used for different creep stages: in the accelerated creep stage, the monitoring time interval is set to less than or equal to 24 hours; in the constant and decelerated creep stages, the monitoring time interval is set to less than or equal to 48 hours.
[0022] like Figure 3As shown, the creep stage is determined based on the time-varying parameter n(t) of each creep stage of a landslide: when n(t)≈0, it is determined to be a constant-rate creep stage; when n(t)>1, it is determined to be an accelerated creep stage; and when n(t)<0, it is determined to be a decelerated creep stage. After the landslide enters the accelerated creep stage, the damage is irreversible under continuous load, and it will deform and evolve into failure.
[0023] The time-varying parameters n(t) and A(t) have clear physical meanings and evolutionary laws: n(t) represents the displacement potential. During the constant-rate creep stage, n(t) ≈ 0; during the accelerated creep stage, n(t) > 1; during the decelerated creep stage, n(t) < 0. After the landslide creep terminates, n(t) resets to its initial state. A decrease in n(t) indicates that the displacement potential is being rapidly depleted. A larger n(t) value means that the landslide has a large reserve of displacement potential, requiring more energy to drive deformation. When the external load exceeds the critical threshold, the system's energy state undergoes a sudden change, leading to a sharp increase in n(t). This phenomenon signifies a shift in the deformation mechanism from quasi-static dislocation slip to an accelerated failure stage dominated by dynamic crack propagation, providing a crucial signal for failure early warning.
[0024] A(t) characterizes the degree of damage accumulation. An increase in its value indicates the continuous accumulation of internal damage, making the landslide more prone to deformation at the same displacement level. In the initial creep stage, A(t) → 0, indicating that the landslide has high strength and stability. In the accelerated failure stage, A(t) gradually increases, indicating that the landslide has lower stability. When the landslide creep terminates, A(t) will reset to the initial state, which can be physically interpreted as damage healing or partial recovery of bearing capacity.
[0025] The evolution of the parameters n(t) and A(t) is affected by factors such as the shear strength of the soil and rock mass structure, the periodic fluctuation of the reservoir water level, the amount of rainfall infiltration, and changes in external loads.
[0026] Step 4: Identify and calculate the accelerated creep stage. Timing. When step 3 determines that the landslide is currently in the accelerated creep stage, i.e., n(t) > 1, the characteristic sequence is calculated using the current value of n(t) and the corresponding Ω(t) data. ; Step 5: Predict failure time based on linear extrapolation. The feature sequence calculated in Step 4... Using the ordinate as the vertical axis and the corresponding time t as the horizontal axis, perform linear regression to obtain a fitted line; extrapolate this fitted line to the ordinate. The intersection of the x-coordinates at the point is the predicted landslide failure time tf; Based on landslide creep model Derive the expression for the remaining failure time due to landslide creep; the derivation process specifically includes: Treat A(t) and n(t) as constants A and n, and solve the model; When n≠1, the relationship between creep rate and time is: (3) In the formula, Ω0 is the displacement at the initial time t0; When in the accelerated creep stage and n>1, assume that the creep displacement at failure time tf tends to infinity. →∞), the remaining failure time from the current time t can be derived from equation (3). expression: (4) In the formula, tf represents the landslide failure time. Let be the displacement at the current time t. Equation (4) shows the remaining failure time. Power of (1-n) of the current displacement Proportional; the linear extrapolation prediction method in step five is based on A linear negative correlation was established between time t and the time during the accelerated creep phase.
[0027] When n≠1, equation (3) is rewritten in the form of a standard linear equation, i.e., the expression for creep failure time: (5) In the formula, the slope and intercept These are all instantaneous constants. The theoretical basis for a linear relationship with time t. When n < 1, It is positively correlated with time t; when n>1, It is negatively correlated with time t, and as t increases, Gradually decreasing. Based on this, a landslide creep failure prediction method is proposed: [The method is described in the original text, but the context is unclear.] Linear extrapolation of the temporal relationship with time t is performed to extrapolate to... ( When the displacement is the amount of displacement at the moment of failure, the time corresponding to the intersection point is the predicted failure time tf of the landslide. The feature sequence is performed only if n(t) > 1. The calculation; if n(t) < 1, then It is positively correlated with time t. The fact that the value gradually increases but does not decrease and approaches zero indicates that the landslide will not fail within a finite time.
[0028] For feature sequences When performing linear regression, a goodness-of-fit R² > 0.9 is required, and the extrapolation prediction of the failure time tf should be performed with a confidence interval of 95% or higher. When the goodness-of-fit R² < 0.9, older data from earlier periods need to be removed to ensure that R² > 0.9.
[0029] Step 6: Dynamically update early warning information. As new time-series monitoring data is continuously acquired, repeat steps 3 to 5 to continuously update time-varying parameters, feature sequences, and predicted failure times (tf) to achieve dynamic correction of early warning information.
[0030] The dynamic update process can automatically identify the deformation trend change caused by external load changes; through rolling calculation, multiple continuous time-varying parameter n(t) sequences and corresponding predicted failure times tf are obtained, so that the prediction results can flexibly adapt to the dynamic changes of landslide status, thereby realizing the continuous correction of early warning information.
[0031] To verify the effectiveness and engineering applicability of the method of the present invention, this embodiment selects a typical accretionary landslide for case verification, as follows: The verification object is an accretionary landslide located on the right bank of the Zhaxi River, a tributary of the Yangtze River, 10.8 km from the Yangtze River estuary, and administratively belonging to Shangba Village, Shuitianba Township, Zigui County. The landslide body has a typical chair-shaped depression topography, is a dip slope, and extends eastward towards the Zhaxi River; the rear edge of the landslide body is located at the foot of the bedrock steep slope (elevation 370m), and the front edge reaches the bank of the Zhaxi River (elevation 160m), with gullies as natural boundaries on both the north and south sides. The landslide has an overall slope of 27°, an east-west length of 450m, a north-south width of 350m, an average thickness of about 20m, a total volume of 315×104 m³, and a main sliding direction of 64°. The landslide material is mainly a mixture of boulders and gravelly soil. The sliding surface is the contact surface between the soil and the bedrock. The sliding bed consists of Jurassic sandstone and mudstone layers with a rock stratum dip of 30°∠15°. It exhibits typical creep deformation and failure geological conditions, making it suitable for verifying the method of this invention.
[0032] This verification uses GNSS surface displacement time-series monitoring data of the landslide in 2020 (monitoring interval 24 hours). During this monitoring period, the landslide precisely completed the transition from isochronous creep to accelerated creep, which perfectly matches the core application scenario of the invention's "accelerated creep stage identification - failure time prediction". The verification process strictly follows the core procedures of steps 3 to 5 of the invention: First, based on A sliding window linear fitting is performed using a coordinate system to solve for the time-varying parameter n(t) in real time; such as Figure 3 As shown, the evolution trend of n(t) can be accurately tracked by fitting the results of two adjacent time points. When the monitoring data shows that n(t) exceeds the 1.0 threshold, it is determined that the landslide has entered the accelerated creep stage, and the characteristic sequence is then activated. The calculation.
[0033] like Figure 4 As shown, according to the requirements of step 5, the calculated feature sequence is... A linear regression was performed using the vertical axis and the corresponding time t as the horizontal axis. After obtaining the fitted line, it was extrapolated to the vertical axis. The intersection of the x-coordinates at the given location is the predicted failure time tf. The results show that n(t) was detected on the 9th and 10th days (corresponding to n values of 2.2 and 1.7, respectively), which meets the limiting condition of "calculating the feature sequence when n(t)>1" in this invention. Therefore, subsequent prediction calculations were only carried out for these two time points.
[0034] For the prediction calculation on day 9 (n=2.2): Initially, linear regression was performed using full-time series data from days 1 to 9. The goodness of fit R² < 0.9, which did not meet the accuracy requirement of R² > 0.9 in this invention. Following the rule in step 5 of "removing older data when R² < 0.9", after removing older data from days 1 to 3, the data from days 4 to 9 was used for refitting. The goodness of fit improved to R² = 0.904 > 0.9, meeting the accuracy requirement. Based on this fitted line, the prediction failure time tf = 10 days was extrapolated, and the prediction interval was days 10 to 13. For the prediction calculation on day 10 (n=1.7): Initially, after fitting with full-time series data from days 1 to 10, R² did not reach 0.9. After removing older data from day 1, the data from days 2 to 10 was used for refitting. R² = 0.924 > 0.9, meeting the accuracy requirement. The extrapolated prediction interval covered days 11 to 13. The actual failure time (tf) of the landslide was day 13. The predicted interval completely covered the actual failure time, and the predicted failure time was earlier than the actual failure time.
[0035] Verification results show that the prediction results obtained through the complete process of "data filtering - feature sequence calculation - linear fitting - extrapolation prediction" based on the monitoring data corresponding to n=2.2 on day 9 and n=1.7 on day 10 meet the core requirements of engineering early warning in terms of both accuracy and reliability. It can also be observed that during the accelerated creep stage, the larger the value of n(t) (e.g., n=2.2 on day 9 compared to n=1.7 on day 10), the higher the feature sequence... The faster the decay rate, the longer the extrapolated remaining failure time tf. The shorter t is, the more consistent this evolutionary law is with the theoretical derivation of formula (4) of this invention, further confirming the scientific nature and theoretical rigor of the method of this invention.
[0036] In summary, the landslide creep failure prediction method constructed in this invention, verified through typical landslide examples, can accurately identify the accelerated creep stage, ensure fitting accuracy through a data filtering mechanism, and stably output reliable failure time prediction results, demonstrating significant novelty, inventiveness, and engineering applicability.
[0037] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting landslide creep failure, characterized in that, Includes the following steps: S1. Construct a creep model incorporating time-varying parameters, damage accumulation, and displacement potential: (1)； In the formula, Let Ω(t) be the creep rate, Ω(t) be the creep displacement, A(t) be the time-varying parameter characterizing damage accumulation, and n(t) be the time-varying parameter characterizing displacement potential. S2. By deploying monitoring equipment at key parts of the landslide, time-series monitoring data of the observable physical quantity Ω(t) during the landslide creep process are collected; S3. Based on the time-series monitoring data collected in step S2, solve the time-varying parameters n(t) and A(t) in the landslide creep model, and determine the current creep stage of the landslide based on the value of n(t). S4. When step S3 determines that the landslide is currently in the accelerated creep stage, i.e., n(t) > 1, the characteristic sequence is calculated using the current value of n(t) and the corresponding Ω(t) data. ; S5, Feature Sequence Using the ordinate as the vertical axis and the corresponding time t as the horizontal axis, perform linear regression to obtain a fitted line; extrapolate this fitted line to the ordinate. The intersection of the x-coordinates at the point is the predicted landslide failure time tf; S6. As new time-series monitoring data is continuously acquired, repeat steps S3 to S5 to continuously update the time-varying parameters, feature sequences, and predicted failure time tf.
2. The landslide creep failure prediction method according to claim 1, characterized in that, In step S1, the creep model The universality of this method was verified in the following way: Different types of landslide examples were selected, and the results of each example were analyzed. Transform Ω(t) time series data into After obtaining the coordinate data, the least squares method was used for linear fitting. The fitting results all satisfied R² > 0.9 and p < 0.05, indicating that the rate-displacement evolution of different types of landslides all followed the power law law of this model. When a landslide is in motion, regardless of whether it is in the acceleration, constant speed or deceleration stage, the value of Ω(t) is always positive. By linearly fitting the log rate and log displacement data, a unified fit of the entire creep process can be achieved.
3. The landslide creep failure prediction method according to claim 1, characterized in that, In step S2, the observable physical quantity Ω(t) is selected from surface displacement or deep displacement; Monitoring equipment includes GNSS fully automatic displacement monitoring equipment, inclinometers, or crack gauges; Monitoring points were set up in the main sliding zone of the landslide, the tension cracks at the rear edge, and the key location at the shear outlet.
4. The landslide creep failure prediction method according to claim 1, characterized in that, Step 3 specifically includes: landslide creep model Taking the logarithm of both sides, we get the linearized expression: (2)； exist In the coordinate system, a sliding window linear fitting is performed using the time-series monitoring data collected in step 2. The window length is adaptively adjusted according to the monitoring frequency and landslide deformation characteristics. The time-varying parameter n(t) is directly solved through fitting. ; The creep stage is determined based on the time-varying parameter n(t): when n(t)≈0, it is determined to be the constant-rate creep stage; when n(t)>1, it is determined to be the accelerated creep stage; when n(t)<0, it is determined to be the decelerated creep stage. Once a landslide enters the accelerated creep stage, the damage becomes irreversible under continuous load, leading to deformation and eventual failure.
5. The landslide creep failure prediction method according to claim 1, characterized in that, In step S3, during the sliding window linear fitting, different monitoring time intervals are used for different creep stages: in the accelerated creep stage, the monitoring time interval is set to less than or equal to 24 hours; in the constant and decelerated creep stages, the monitoring time interval is set to less than or equal to 48 hours.
6. The landslide creep failure prediction method according to claim 1, characterized in that, In step S3: n(t) represents the displacement potential energy. In the constant creep stage, n(t)≈0; in the accelerated creep stage, n(t)>1; in the decelerated creep stage, n(t)<0; after the landslide creep terminates, n(t) resets to the initial state; the decrease of n(t) indicates that the displacement potential energy is being consumed rapidly; when the external load exceeds the critical threshold, the energy state of the system changes abruptly, causing n(t) to increase sharply. A(t) characterizes the degree of damage accumulation. An increase in its value indicates the continuous accumulation of internal damage, making the landslide more prone to deformation at the same displacement level. In the initial creep stage, A(t) → 0, indicating that the landslide has high strength and stability. In the accelerated failure stage, A(t) gradually increases, indicating that the landslide has lower stability. When the landslide creep terminates, A(t) will reset to the initial state.
7. The landslide creep failure prediction method according to claim 6, characterized in that, The parameters n(t) and A(t) evolve based on the shear strength of the soil and rock mass structure, the periodic fluctuations of the reservoir water level, the amount of rainfall infiltration, and the changes in external loads.
8. The landslide creep failure prediction method according to claim 1, characterized in that, In step S5, based on the landslide creep model Derive the expression for the remaining failure time due to landslide creep; the derivation process specifically includes: Treat A(t) and n(t) as constants A and n, and solve the model; When n≠1, the relationship between creep rate and time is: (3)； In the formula, Ω0 is the displacement at the initial time t0; When in the accelerated creep stage and n>1, assume that the creep displacement at failure time tf tends to infinity. →∞), the remaining failure time from the current time t is derived from equation (3). expression: (4)； In the formula, tf represents the landslide failure time. This represents the displacement at the current time t; Linear extrapolation prediction method based on A linear negative correlation was established with time t during the accelerated creep phase. When n≠1, equation (3) is rewritten in the form of a standard linear equation, i.e., the expression for creep failure time: (5)； In the formula, the slope and intercept All are instantaneous constants; When n < 1, It is positively correlated with time t; when n>1, It is negatively correlated with time t, and as t increases, Gradually decrease; Will Linear extrapolation of the temporal relationship with time t is performed to extrapolate to... When the intersection point is reached, the time corresponding to the landslide prediction failure time tf is taken. The feature sequence is performed only if n(t) > 1. The calculation; when n(t) < 1, It is positively correlated with time t. The fact that the value gradually increases but does not decrease and approaches zero indicates that the landslide will not fail within a finite time.
9. The landslide creep failure prediction method according to claim 1, characterized in that, In step S5, the feature sequence is... When performing linear regression, the goodness of fit R2 is required to be greater than 0.9, and the extrapolation prediction of failure time tf is made with a confidence interval of 95% or higher. When the goodness of fit R2 < 0.9, the old data needs to be removed so that R2 > 0.
9.
10. The landslide creep failure prediction method according to claim 1, characterized in that, In step S6, the dynamic update process can automatically identify the deformation trend change caused by the change of external load; through rolling calculation, multiple continuous time-varying parameter n(t) sequences and corresponding predicted failure times tf are obtained, so that the prediction results can flexibly adapt to the dynamic changes of landslide status, thereby realizing the continuous correction of early warning information.