A dynamic evaluation method for rock-soil slope stability considering rainfall infiltration effect

Abstract

The application discloses a kind of dynamic evaluation method of rock-soil slope stability considering rainfall infiltration effect, comprising: obtaining the dynamic basic attribute data set of target rock-soil slope;Introduce pore water pressure response lag factor;Based on rainfall infiltration dynamic model, the pore water pressure space-time distribution characteristics of target rock-soil slope under time-varying rainfall boundary condition is calculated, and is corrected in real time by machine learning agent model, and the corrected pore water pressure field is obtained;Based on the corrected pore water pressure field, the dynamic stability safety factor of target rock-soil slope at different rainfall time is calculated using dynamic unsaturated soil shear strength model, and the dynamic stability evaluation result is obtained;Output the rainfall full-cycle dynamic stability evaluation report of target rock-soil slope.The application uses machine learning agent model to correct the calculated pore water pressure field in real time;Output the stability evaluation report covering the entire rainfall cycle, significantly improve the accuracy of evaluation.

Description

Technical Field

[0001] This invention relates to the fields of geotechnical engineering and slope stability analysis technology, specifically to a dynamic assessment method and system for the stability of geotechnical slopes that takes into account the effect of rainfall infiltration. Background Technology

[0002] The stability of soil and rock slopes under rainfall has always been a key research focus in geotechnical engineering. Rainfall infiltration significantly alters the seepage field and stress state within the slope, leading to a decrease in the shear strength of unsaturated soil and potentially triggering geological disasters such as landslides. Especially under conditions of heavy or continuous rainfall, the risk of slope instability increases dramatically, posing a serious threat to people's lives, property, and infrastructure. Therefore, developing dynamic assessment methods that accurately reflect the coupled processes of rainfall, infiltration, and stability is of significant engineering importance.

[0003] Traditional slope stability analysis often employs static or quasi-static methods, treating soil and rock parameters as constants and neglecting the dynamic characteristics of water transport during rainfall. Although some studies have introduced saturated and unsaturated seepage theories, Darcy's law is commonly used in simulating rainfall infiltration, failing to consider complex physical phenomena such as nonlinear flow and soil hydraulic hysteresis that exist in actual infiltration. Furthermore, most models simplify the entire rainfall process to a single infiltration boundary, making it difficult to realistically depict the dynamic transition from complete infiltration to surface runoff, resulting in significant errors in pore water pressure prediction.

[0004] In recent years, with the development of monitoring technology and numerical simulation, some dynamic assessment methods have begun to integrate real-time meteorological and displacement data. For example, Chinese invention patent CN120579350B, "A GNSS-based method and system for assessing the stability of loess slopes on mountain highways," proposes to monitor displacement using GNSS, correct displacement by combining it with Fredlund's theory of unsaturated soil shear strength, and correct soil strength parameters in real time based on cumulative rainfall. The corrected parameters are then coupled with the actual displacement to calculate the safety factor, thereby achieving early warning judgment. Although this method has made progress in multi-source data fusion and dynamic parameter correction, it still mainly relies on external monitoring to invert soil parameters. It does not model the rainfall infiltration process itself from a physical mechanism perspective, nor does it consider the dynamic division of the infiltration stage, non-Darcy flow, and the hysteresis effect of pore water pressure response. Its seepage field is still indirectly derived, lacking a mechanistic coupling of the entire process of rainfall, infiltration, and strength degradation.

[0005] In summary, existing technologies generally suffer from the following problems: the basic input data is static, failing to reflect the nonlinear functional relationship between soil and rock parameters and volumetric water content; the rainfall infiltration model is overly simplified, neglecting non-Darcy flow and hysteresis effects, and failing to reasonably distinguish between the stages of complete infiltration and surface runoff; the pore water pressure field relies on indirect inversion or idealized calculations, lacking a mechanism for real-time correction based on physical models and combined with machine learning; although the stability assessment introduces unsaturated strength theory, it is not directly coupled with the dynamic pore water pressure field driven by the refined infiltration model, making it difficult to accurately capture the instability critical point; at the same time, there is a lack of a closed-loop system that automatically identifies dangerous sliding surfaces and potential instability periods based on the entire rainfall process and triggers dynamic support decisions, which restricts the timeliness, mechanism consistency, and intelligence level of slope safety control. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a dynamic assessment method for the stability of soil and rock slopes that considers the effect of rainfall infiltration. This method can more realistically simulate the physical mechanism of water infiltration during rainfall, reflect the spatiotemporal evolution of pore water pressure, dynamically calculate the slope safety factor, and realize early identification of instability risks and intelligent support decisions, thereby improving the accuracy of slope stability assessment.

[0007] The objective of this invention is achieved through the following technical solution: A dynamic assessment method for the stability of soil and rock slopes that considers rainfall infiltration effects includes the following steps: Obtain a dynamic basic attribute dataset of the target soil and rock slope, wherein the dynamic basic attribute dataset includes slope geometric parameters, nonlinear soil and rock mass parameter functions that vary with volumetric water content, hydrological and meteorological parameters, and initial seepage field parameters. Based on the aforementioned dynamic basic attribute dataset, a rainfall infiltration dynamic model considering non-Darcy flow and hysteresis effects is constructed. The rainfall infiltration dynamic model dynamically divides a complete rainfall process into a full infiltration stage and a surface runoff stage, and introduces a pore water pressure response hysteresis factor. Based on the rainfall infiltration dynamic model, the spatiotemporal distribution characteristics of pore water pressure of the target soil and rock slope under time-varying rainfall boundary conditions are calculated, and real-time correction is performed through a machine learning surrogate model to obtain the corrected pore water pressure field. Based on the modified pore water pressure field, the dynamic stability safety factor of the target soil and rock slope at different rainfall times is calculated using the dynamic unsaturated soil shear strength model, and the dynamic stability assessment results are obtained. Based on the dynamic stability assessment results, the potential instability period and dangerous sliding surface location of the target soil and rock slope are identified, and based on the preset stability threshold, it is determined whether a dynamic support decision response is triggered. Output a dynamic stability assessment report of the target soil and rock slope throughout the entire rainfall cycle.

[0008] As a preferred embodiment, the construction of a rainfall infiltration dynamic model considering non-Darcy flow and hysteresis effects based on the dynamic basic attribute dataset further includes: Based on the real-time rainfall intensity data and dynamic saturated permeability coefficient in the hydrological and meteorological parameters, the infiltration state of the target soil and rock slope surface is determined, wherein the dynamic saturated permeability coefficient... Expressed as volumetric water content power function:

[0009] in, The dynamic saturated permeability coefficient is expressed in m / s. The initial saturated permeability coefficient is expressed in m / s. This refers to the volumetric water content. Residual moisture content; This represents the saturated moisture content. It is an experience index; When the real-time rainfall intensity is less than or equal to the dynamic saturated permeability coefficient, the current stage is divided into the complete infiltration stage, and the flow boundary condition is used to simulate the slope rainwater infiltration process. When the real-time rainfall intensity is greater than the dynamic saturated permeability coefficient, the current stage is classified as the surface runoff stage, and a slope runoff-subsurface seepage coupling coefficient is introduced. To correct the boundary conditions, where is a coefficient, with a value ranging from 0 to 1.

[0010] As a preferred embodiment, the step of calculating the spatiotemporal distribution characteristics of pore water pressure on the target soil and rock slope under time-varying rainfall boundary conditions based on the rainfall infiltration dynamic model, and performing real-time correction through a machine learning surrogate model to obtain a corrected pore water pressure field, further includes: The initial pore water pressure distribution is obtained by solving the seepage problem in unsaturated soil based on the Richards equation. A proxy model based on a deep neural network is constructed, which takes the initial pore water pressure distribution and dynamic basic attribute dataset as input and outputs the pore water pressure correction. The pore water pressure correction is superimposed on the initial pore water pressure distribution to obtain the corrected pore water pressure field.

[0011] As a preferred embodiment, the step of calculating the dynamic stability safety factor of the target soil and rock slope at different rainfall times using a dynamic unsaturated soil shear strength model based on the modified pore water pressure field to obtain the dynamic stability assessment result further includes: Based on the modified pore water pressure field, the dynamic volumetric water content distribution and matrix suction distribution inside the target soil and rock slope are calculated. The dynamic shear strength parameters were calculated using Fredlund's dual-stress variable strength theory, where the shear strength... The expression is:

[0012] in, Shear strength, expressed in kPa; Effective cohesion, expressed in kPa; The total normal stress is expressed in kPa. This represents pore gas pressure, expressed in kPa. Pore ​​water pressure, in kPa; The effective internal friction angle is expressed in degrees (°). The matrix suction angle is expressed in units of 1 / 2. Effective cohesion and effective internal friction angle All change dynamically with volumetric water content and matrix suction: ;

[0013] in, Effective cohesion, expressed in kPa; The initial effective cohesion is expressed in kPa. This is the cohesion attenuation coefficient; This refers to the dynamic volumetric moisture content. The effective internal friction angle is expressed in degrees (°). This is the initial effective internal friction angle, in degrees. The internal friction angle attenuation coefficient; matrix suction angle. With matrix suction The changes are based on the following empirical relationships:

[0014] in, The matrix suction angle is expressed in degrees (°). The initial matrix suction angle is expressed in degrees (°). The attenuation coefficient; Standard atmospheric pressure, unit: kPa; Calculation of dynamic stability safety factor based on limit equilibrium method :

[0015] in, The dynamic stability safety factor; For the first The effective cohesive force of the strip, expressed in kPa; For the first The length of the sliding surface of the strip, in meters; For the first Weight per unit width of the strip, in kN / m; For the first The inclination angle of the sliding surface of the strip, in degrees; For the first The pore water pressure of the strip, in kPa; For the first The base area per unit width of the strip ( (Unit: m) 2 ; For the first The effective internal friction angle of the strip, in degrees; For the first The pore air pressure of the strip, in kPa; For the first The pore water pressure of the strip, in kPa; For the first The matrix suction angle of the strip is expressed in degrees. Based on the dynamic stability safety factor at different rainfall times, the global minimum safety factor and its corresponding most dangerous potential sliding surface position are determined to form the dynamic stability assessment result.

[0016] As a preferred embodiment, the step of identifying the potential instability period and dangerous sliding surface location of the target soil and rock slope based on the dynamic stability assessment results, and determining whether to trigger a dynamic support decision response based on a preset stability threshold, further includes: The global minimum safety factor in the dynamic stability assessment result is compared with the first threshold. Second threshold Comparison, among which ,and The value range is 1.3 to 1.5. The value range is 1.0 to 1.2; If the global minimum security factor is greater than the first threshold The target soil and rock slope is determined to be in a stable state, and no support decision is triggered. If the global minimum security factor is within the first threshold With the second threshold If the target rock and soil slope is determined to be in an unstable state, the first-level dynamic support decision response is triggered, and an early warning message is output. If the global minimum security factor is less than the second threshold The target soil and rock slope is determined to be in an unstable state, triggering a second-level dynamic support decision response, and it is recommended to adopt emergency support measures.

[0017] As a preferred embodiment, the method further includes a correction step that takes into account the non-uniformity of spatial distribution of rainfall: Obtain the spatial distribution data of rainfall retrieved from meteorological radar in the area where the target soil and rock slope is located, and divide the slope surface into multiple sub-regions; Based on the rainfall infiltration dynamic model, the contribution weight of each sub-region to the pore water pressure field inside the slope is calculated, and the contribution weight is solved by a neural network based on physical information. The pore water pressure responses of each sub-region are weighted and superimposed to obtain the corrected pore water pressure field considering spatially non-uniform rainfall.

[0018] As a preferred method, the dynamic basic attribute dataset of the target soil and rock slope is obtained, wherein the nonlinear soil and rock mass parameter function that varies with volumetric water content is updated in real time based on field monitoring data using a Bayesian inversion method, and the Kalman filter algorithm is used to achieve optimal parameter estimation during the update process.

[0019] When acquiring dynamic basic attribute data of the target soil and rock slope, special attention is paid to the nonlinear characteristics of soil and rock parameters changing with volumetric water content, and these parameter functions are continuously updated using real-time field monitoring data. Using the Bayesian inversion method, information such as displacement, pore water pressure, or water content observed in the field is used as feedback to continuously correct the initial assumptions in the model regarding the relationship between key parameters such as soil and rock strength, permeability, and water content. To achieve efficient and stable parameter updates, the system employs a Kalman filter algorithm to optimize the inversion process. This algorithm can integrate model predictions and actual monitoring values, dynamically adjusting parameter estimates while considering the uncertainties of both, gradually approximating the true state. This mechanism, combining Bayesian inference and Kalman filtering, not only ensures the physical rationality of parameter updates but also possesses good noise resistance and computational efficiency, thus ensuring that slope stability analysis is always based on the most reliable soil and rock parameters, significantly improving the timeliness and accuracy of risk assessment.

[0020] As a preferred embodiment, the method further includes a quantitative assessment step of the rainfall infiltration lag effect: The lag time is defined as the time difference between the peak moment of slope rainfall and the peak moment of pore water pressure at depth z, in hours. A lag time prediction model was constructed, which comprehensively considers soil permeability, depth, and rainfall characteristic parameters; The lag time is used as an output item of the dynamic stability assessment result to correct the advance time of slope instability warning.

[0021] Considering the time required for rainwater to infiltrate deep into the soil from the slope surface, the response of pore water pressure within the slope typically lags behind surface rainfall. Therefore, a quantitative assessment of the rainfall infiltration lag effect was specifically introduced. Specifically, the lag time is defined as the time difference, expressed in hours, between the peak of surface rainfall and the peak of pore water pressure at a given depth. This time difference reflects the delayed characteristics of water migration and pressure transmission within the soil. To accurately predict the lag time at different depths, a predictive model was constructed that comprehensively considers soil permeability, monitoring point depth, rainfall intensity, and duration. Since highly permeable soils experience rapid water infiltration and short lag times, while less permeable soils or longer rainfall durations may lead to more significant delays, this model can dynamically estimate the response delay at each depth based on actual conditions. Ultimately, the obtained lag time is used as a key output of the dynamic stability assessment to adjust the release time of slope instability warnings. For example, if the lag time for a certain depth is 6 hours, the system will reserve a corresponding time window after the rainfall peak occurs before triggering the warning, thereby avoiding alarms that are too early or too late and significantly improving the accuracy and practicality of the warning.

[0022] As a preferred embodiment, the method further includes a dynamic comparative evaluation step for different rainfall types: Multiple rainfall scenarios are set up, including uniform rainfall, front-peak rainfall, mid-peak rainfall, and back-peak rainfall; The corresponding stability assessment results were calculated under each rainfall condition. Based on the aforementioned stability assessment results, an ensemble learning algorithm is used to predict the most unfavorable rainfall type and its corresponding minimum stability safety factor for a future period of time, and an adaptive rainfall warning level is generated accordingly.

[0023] As a preferred embodiment, when the target soil slope is a multi-layered heterogeneous soil slope, the method further includes a step of processing inter-layer parameters for continuity. A transition layer model is introduced at the soil layer interface, and the hyperbolic tangent function is used to describe the smooth transition of parameters at the interface:

[0024] in, These are parameter values ​​near the interface, with units determined based on the specific parameter. , These are the parameter values ​​for the upper and lower soil layers, respectively, with units determined based on the specific parameters. These are depth coordinates, in meters (m). This refers to the interface location, in meters (m). The thickness coefficient of the transition layer is expressed in meters (m). Numerical calculations are performed within the transition layer using an adaptive mesh refinement technique, with the mesh size adaptively adjusted according to the parameter gradient to improve calculation accuracy.

[0025] The present invention has at least the following beneficial effects: This invention provides a dynamic assessment method for the stability of soil and rock slopes that considers rainfall infiltration effects. By collecting dynamic baseline data such as slope geometry, soil and rock parameters varying with water content, meteorological data, and initial seepage fields, a more realistic analytical foundation is constructed. The established rainfall infiltration model comprehensively considers the hysteresis effects of non-Darcy flow and pore water pressure response, and dynamically divides the rainfall process into two stages: complete infiltration and surface runoff, accurately reflecting the rainfall infiltration process. A machine learning proxy model is used to correct the calculated pore water pressure field in real time, improving its accuracy. Combined with a dynamic unsaturated soil shear strength model, the safety factor at different rainfall times is accurately calculated, realistically depicting the trend of slope stability changes with rainfall. Based on this, potential instability periods and dangerous sliding surfaces can be identified, and the initiation of support measures can be automatically determined based on preset thresholds, achieving early warning and intelligent decision-making. Finally, a stability assessment report covering the entire rainfall cycle is output, providing clear and practical basis for engineering management, significantly improving the accuracy, timeliness, and practicality of the assessment. Attached Figure Description

[0026] To reveal the technical details of the embodiments of the present invention, the accompanying drawings involved in the embodiments will be briefly described below. It should be emphasized that these drawings only present several embodiments of the present invention and should not be considered as defining the scope of the invention. For those skilled in the art, other related drawings can still be derived based on these drawings without inventive effort.

[0027] Figure 1 A flowchart of a dynamic assessment method for the stability of soil and rock slopes that takes into account the effect of rainfall infiltration; Figure 2 A flowchart for pore water pressure correction to account for the non-uniform spatial distribution of rainfall. Detailed Implementation

[0028] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0029] It should be noted that the specific details described below are intended to facilitate a full understanding of the exemplary embodiments by those skilled in the art and are not essential for implementing the present invention. Those skilled in the art should understand that the embodiments can still be effectively implemented even without some or all of the described details. For example, the system may be shown in block diagram form to avoid redundant details affecting the clear expression of the overall structure; in other cases, to highlight the core solution, well-known technical processes, conventional structures, or general methods, and other non-essential content, may be omitted.

[0030] like Figure 1 As shown, a dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effects includes the following steps: Obtain the dynamic basic attribute dataset of the target soil and rock slope. The dynamic basic attribute dataset includes slope geometric parameters (such as slope height, slope angle, slope length and layered structure), nonlinear soil and rock mass parameter functions that vary with volumetric water content (such as unsaturated permeability coefficient, shear strength parameters c and φ, soil-water characteristic curves, etc.), hydrological and meteorological parameters (such as rainfall intensity, rainfall duration, evaporation rate and previous rainfall), and initial seepage field parameters (such as initial pore water pressure distribution, groundwater level depth and initial water content profile).

[0031] Based on the aforementioned dynamic basic attribute dataset, a rainfall infiltration dynamic model considering non-Darcy flow and hysteresis effects is constructed. The rainfall infiltration dynamic model dynamically divides a complete rainfall process into a full infiltration stage and a surface runoff stage, and introduces a pore water pressure response hysteresis factor. Based on the rainfall infiltration dynamic model, the spatiotemporal distribution characteristics of pore water pressure of the target soil and rock slope under time-varying rainfall boundary conditions are calculated, and real-time correction is performed through a machine learning surrogate model to obtain the corrected pore water pressure field. Based on the modified pore water pressure field, the dynamic stability safety factor of the target soil and rock slope at different rainfall times is calculated using the dynamic unsaturated soil shear strength model, and the dynamic stability assessment results are obtained. Based on the dynamic stability assessment results, the potential instability period and dangerous sliding surface location of the target soil and rock slope are identified, and based on the preset stability threshold, it is determined whether a dynamic support decision response is triggered. Output a dynamic stability assessment report of the target soil and rock slope throughout the entire rainfall cycle.

[0032] In one embodiment, there is a 30-meter-high soil slope next to a highway in a mountainous area. The area experiences frequent rainy seasons and has previously experienced shallow landslides due to heavy rainfall. To prevent further instability, engineers used the method of this invention for dynamic assessment. First, through on-site surveys and a sensor network, data were acquired regarding the slope's gradient, layered structure, soil strength parameters varying with moisture content, recent rainfall forecasts, and initial groundwater levels. Then, the system constructs a dynamic model considering the nonlinear characteristics of rainwater infiltration: in the initial stages of rainfall, all rainwater infiltrates into the soil; when the rainfall intensity exceeds the soil's infiltration capacity, the model automatically switches to the surface runoff stage and introduces a hysteresis factor to simulate the time delay of water propagation in the soil. Based on this model, the changes in pore water pressure within the slope over the next 72 hours with time and depth are calculated. A machine learning model trained using existing historical monitoring data is then used to correct the calculation results in real time, improving accuracy. Next, based on the corrected pore water pressure distribution, the slope safety factor was calculated hourly. It was found that the safety factor dropped to 1.15 after 18 hours (below the warning threshold of 1.2), and the most dangerous sliding surface was located in the upper middle part of the slope. The system immediately determined that there was a risk of instability, automatically triggered an early warning, and recommended the installation of temporary drainage or micropile support in key areas. Finally, an assessment report was generated, including the stability curve for the entire process, risk periods, dangerous areas, and corresponding response recommendations, for management personnel to use in decision-making.

[0033] This invention provides a dynamic assessment method for the stability of soil and rock slopes that considers rainfall infiltration effects. It refines and dynamically models the entire physical process of rainfall infiltration, pore water pressure evolution, soil and rock strength degradation, and instability early warning. First, the system acquires the slope's geometric parameters, nonlinear soil and rock mass parameter functions varying with volumetric water content, real-time hydrological and meteorological data, and initial seepage field information to construct a dynamic dataset of basic attributes that accurately reflects the current state of the slope. Based on this, a dynamic rainfall infiltration model that more closely resembles the actual physical process is established. This model not only considers non-Darcy flow and the hysteresis effect of soil moisture response that may occur under high water content conditions, but also automatically divides a complete rainfall process into a full infiltration stage and a surface runoff stage, thus more accurately simulating how rainwater gradually infiltrates the slope over time. Subsequently, the model is used to calculate the temporal and spatial distribution characteristics of pore water pressure under time-varying rainfall boundary conditions, and a machine learning surrogate model is used to correct the results in real time, obtaining a high-precision corrected pore water pressure field. Next, combining a dynamic unsaturated soil shear strength model, the safety factor of the slope is calculated hourly based on the modified pore water pressure field, generating a continuous stability evolution curve. Based on this curve, the system can automatically identify potential dangerous sliding surface locations and the most likely time period for instability, comparing them with preset stability thresholds. Once the risk exceeds the safety limit, a dynamic support or early warning response mechanism is immediately triggered. Finally, the entire process outputs a dynamic stability assessment report covering the entire rainfall process, providing scientific, timely, and operable technical support for slope safety management.

[0034] Taking a typical red clay slope as an example: the slope is 12m high and initially in an unsaturated state, with a shallow pore water pressure of -30kPa. The soil-water characteristic curve (SWCC) fitted based on the FredlundXing model and its corresponding unsaturated permeability coefficient function are presented. The Richards equations were used to simulate the transient seepage process under a post-peak rainfall event (20 mm / h in the first 6 hours and 60 mm / h in the last 6 hours). To balance physical mechanisms and observational information, an ensemble Kalman filter (EnKF) was introduced as a data assimilation framework, integrating real-time monitoring data from three pore pressure sensors and two TDR moisture probes deployed in the field. This dynamically corrected the numerical simulation results, effectively avoiding overfitting issues common in purely data-driven methods under small sample conditions. Based on this, the Fredlund dual-stress variable strength model was used to calculate the transient shear strength.

[0035] in , The linear suction contribution was limited to conditions with suction <50 kPa. As the corrected shallow pore water pressure increased from -30 kPa to -5 kPa, the equivalent cohesion decreased from 14.4 kPa to 9.1 kPa. The high-precision pore pressure field at each moment was substituted into the Morgenstern-Price limit equilibrium method, and the safety factor was updated hourly to obtain the stability evolution curve: The value decreased continuously from an initial 1.42, reaching 1.03 by the 12th hour. Based on this, the system automatically identified this as the most likely period of instability and located the shallow circular arc slip surface as a high-risk area. When the threshold approaches the preset warning threshold (e.g., 1.10), the platform automatically triggers a level 3 warning and generates a dynamic assessment report that includes the spatiotemporal pore pressure distribution, safety factor time series, sliding surface evolution, and disposal suggestions, providing closed-loop support for emergency decision-making.

[0036] In a preferred embodiment, the step of constructing a dynamic rainfall infiltration model considering non-Darcy flow and hysteresis effects based on the dynamic basic attribute dataset further includes: Based on the real-time rainfall intensity data and dynamic saturated permeability coefficient in the hydrological and meteorological parameters, the infiltration state of the target soil and rock slope surface is determined, wherein the dynamic saturated permeability coefficient... Expressed as volumetric water content power function:

[0037] in, The dynamic saturated permeability coefficient is expressed in m / s. The initial saturated permeability coefficient is expressed in m / s. This refers to the volumetric water content. Residual moisture content; This represents the saturated moisture content. It is an experience index; When the real-time rainfall intensity is less than or equal to the dynamic saturated permeability coefficient, the current stage is divided into the complete infiltration stage, and the flow boundary condition is used to simulate the slope rainwater infiltration process. When the real-time rainfall intensity is greater than the dynamic saturated permeability coefficient, the current stage is classified as the surface runoff stage, and a slope runoff-subsurface seepage coupling coefficient is introduced. To correct the boundary conditions, where is a coefficient, with a value ranging from 0 to 1.

[0038] In a preferred embodiment, the step of calculating the spatiotemporal distribution characteristics of pore water pressure on the target soil and rock slope under time-varying rainfall boundary conditions based on the rainfall infiltration dynamic model, and performing real-time correction using a machine learning surrogate model to obtain a corrected pore water pressure field, further includes: The initial pore water pressure distribution is obtained by solving the seepage problem in unsaturated soil based on the Richards equation. Construct a proxy model based on a deep neural network (such as a multilayer perceptron MLP, a convolutional neural network CNN, or a physical information neural network), take the initial pore water pressure distribution and dynamic basic attribute dataset as input, and output the pore water pressure correction amount; The pore water pressure correction is superimposed on the initial pore water pressure distribution to obtain the corrected pore water pressure field.

[0039] This embodiment combines physical models with data-driven technology to dynamically improve the calculation accuracy of pore water pressure fields on slopes. First, the classical Richards equations are used to numerically simulate the rainfall infiltration process in unsaturated soil, obtaining an initial spatiotemporal distribution of pore water pressure under given time-varying rainfall boundary conditions. However, due to uncertainties in model parameters, simplified boundary conditions, or geological heterogeneity, there may be deviations from actual monitoring values. To compensate for this discrepancy, the system further introduces a machine learning surrogate model based on a deep neural network. This model takes the initial pore water pressure distribution and real-time collected slope basic attribute data (such as soil type, moisture content, and topographic features) as input. By learning from historical or current measured data from monitoring points, it automatically predicts the required "correction amount." Finally, this correction amount is superimposed on the initial pressure field, generating a corrected pore water pressure field that conforms to physical laws and closely matches the actual measured data. The entire process integrates physical mechanisms with data intelligence, not only improving computational efficiency but also significantly enhancing the reliability and real-time adaptability of the results across the entire spatial domain.

[0040] In a preferred embodiment, a rainfall infiltration dynamic model considering non-Darcy flow and hysteresis effects is constructed, further including a refined infiltration stage discrimination step based on a non-Darcy flow threshold. Specifically, the hydraulic gradient modulus in the slope or shallow soil is calculated. and the non-Darcy initial gradient of the soil. Comparison:

[0041] in, The hydraulic gradient modulus is dimensionless. The initiation gradient for non-Darcy flow is dimensionless and calibrated through indoor experiments. When this criterion holds, the non-Darcy flow constitutive model is activated, and the volumetric flow rate vector is calculated using Forchheimer-type equations. :

[0042] in, This is a volumetric flow rate vector, with units of m / s; The dynamic saturated permeability coefficient is expressed in m / s. This is the head gradient vector, which is dimensionless. The non-Darcy flow inertial drag coefficient, in units of The modulus of volumetric flow rate is expressed in m / s.

[0043] In a preferred embodiment, a surrogate model based on a deep neural network is constructed, employing a Physics-Informed Neural Network (PINN) architecture, with the unsaturated soil seepage control equations embedded as physical constraints in the loss function. The surrogate model is defined as a mapping function:

[0044] in For spatial coordinates, For time, This is the correction amount for pore water pressure. These are the network parameters. Their composite loss function... The structure is as follows:

[0045] in, The data-driven loss term is used to calculate the model output. The mean square error of the residual between the observed pore water pressure values ​​at the field monitoring points; The physical constraint loss term is used to calculate the degree of violation of the Richards equations by the surrogate model output through automatic differentiation. Considering that rainfall infiltration is mainly applied through boundary conditions, the internal source-sink terms can be ignored (i.e., ...). Then, the physical constraint loss term is specifically:

[0046] in, Total head (unit: m) The initial pore water pressure, This is the correction amount for pore water pressure. The density of water, It is the acceleration due to gravity; Specific water density function (unit: m) -1 ), The unsaturated permeability coefficient is a function of the unsaturated permeability coefficient (unit: m / s). The number of configuration points for random sampling; This is the physical loss weighting coefficient, used to balance the accuracy of data fitting with the degree of adherence to physical laws. Among them, and All were normalized according to their respective physical characteristic scales to eliminate dimensional differences. This was achieved by minimizing... Training network parameters The resulting proxy model can not only fit the measured data of a limited number of monitoring points with high accuracy, but also ensure that the corrected pore water pressure field strictly satisfies the basic physical laws of unsaturated seepage throughout the entire slope domain, effectively overcoming the non-physical or numerical oscillation problems that may occur in the monitoring blind area of ​​the pure data-driven model.

[0047] In a preferred embodiment, the step of calculating the dynamic stability safety factor of the target soil and rock slope at different rainfall times using a dynamic unsaturated soil shear strength model based on the modified pore water pressure field to obtain the dynamic stability assessment result further includes: Based on the modified pore water pressure field, the dynamic volumetric water content distribution and matrix suction distribution inside the target soil and rock slope are calculated. The dynamic shear strength parameters were calculated using Fredlund's dual-stress variable strength theory, where the shear strength... The expression is:

[0048] in, Shear strength, expressed in kPa; Effective cohesion, expressed in kPa; The total normal stress is expressed in kPa. This represents pore gas pressure, expressed in kPa. Pore ​​water pressure, in kPa; The effective internal friction angle is expressed in degrees (°). The matrix suction angle is expressed in units of 1 / 2. Effective cohesion and effective internal friction angle All change dynamically with volumetric water content and matrix suction: ;

[0049] in, Effective cohesion, expressed in kPa; The initial effective cohesion is expressed in kPa. This is the cohesion attenuation coefficient; This refers to the dynamic volumetric moisture content. The effective internal friction angle is expressed in degrees (°). This is the initial effective internal friction angle, in degrees. The internal friction angle attenuation coefficient; matrix suction angle. With matrix suction The changes are based on the following empirical relationships:

[0050] in, The matrix suction angle is expressed in degrees (°). The initial matrix suction angle is expressed in degrees (°). The attenuation coefficient; Standard atmospheric pressure, unit: kPa; Calculation of dynamic stability safety factor based on limit equilibrium method (Calculated per unit width):

[0051] in, The dynamic stability safety factor; For the first The effective cohesive force of the strip, expressed in kPa; For the first The length of the sliding surface of the strip, in meters; For the first Weight per unit width of the strip, in kN / m; For the first The inclination angle of the sliding surface of the strip, in degrees; For the first The pore water pressure of the strip, in kPa; For the first The base area per unit width of the strip ( (Unit: m) 2 ; For the first The effective internal friction angle of the strip, in degrees; For the first The pore air pressure of the strip, in kPa; For the first The pore water pressure of the strip, in kPa; For the first The matrix suction angle of the strip is expressed in degrees; all terms represent forces per unit width, expressed in kN / m, ensuring unit consistency. Based on the dynamic stability safety factors described at different rainfall times, the global minimum safety factor and its corresponding most dangerous potential sliding surface location are determined, forming the dynamic stability assessment results.

[0052] In a preferred embodiment, the step of identifying the potential instability period and dangerous sliding surface location of the target soil and rock slope based on the dynamic stability assessment results, and determining whether to trigger a dynamic support decision response based on a preset stability threshold, further includes: The global minimum safety factor in the dynamic stability assessment result is compared with the first threshold. Second threshold Comparison, among which ,and The value range is 1.3 to 1.5. The value range is 1.0 to 1.2; If the global minimum security factor is greater than the first threshold The target soil and rock slope is determined to be in a stable state, and no support decision is triggered. If the global minimum security factor is within the first threshold With the second threshold If the target rock and soil slope is determined to be in an unstable state, the first-level dynamic support decision response is triggered, and an early warning message is output. If the global minimum security factor is less than the second threshold The target soil and rock slope is determined to be in an unstable state, triggering a second-level dynamic support decision response, and it is recommended to adopt emergency support measures.

[0053] This embodiment achieves dynamic classification and response to the safety status of soil and rock slopes by setting two levels of stability thresholds. The calculated global minimum safety factor is used as the key indicator of slope stability and compared with two preset thresholds: the higher first threshold ( Typically, a value of 1.3 to 1.5 represents a sufficient safety margin boundary, with a lower second threshold ( A safety factor (usually between 1.0 and 1.2) corresponds to the warning line for critical instability. When the safety factor is higher than... When the safety factor is within a certain range, it indicates that the slope is generally stable under the current rainfall and geological conditions, and no intervention is required; when the safety factor decreases to a certain level... and If the slope stability is weakened during this period, although it has not yet collapsed, there is a potential risk. At this point, the system automatically triggers the first-level response, such as issuing an early warning, increasing the monitoring frequency, or activating contingency plans. However, once the safety factor falls below a certain threshold... This indicates that the slope is highly likely to slide and fail in the short term. The system then activates a Level II emergency response, recommending immediate reinforcement, unloading, or personnel evacuation as support measures. This tiered judgment mechanism not only reflects the gradual evolution of a slope from stability to instability but also enables intelligent decision-making from early warning to emergency response, improving the timeliness and relevance of slope risk management.

[0054] In a preferred embodiment, the first threshold Second threshold The adaptive adjustment rule is based on the current rainfall intensity. With cumulative rainfall Quantization is performed. Specifically, a dynamic threshold adjustment factor is defined. for:

[0055] in, and These are weighting coefficients, with values ​​ranging from [0.05, 0.15]. For reference rainfall intensity, a value of 50 mm / h is used; The cumulative rainfall is taken as a reference, with a value of 200 mm. To avoid triggering overly conservative warnings under extreme rainfall conditions, a setting is used. The upper limit is 1.3. Therefore, the adaptive threshold calculation formula is:

[0056] in, and These are the baseline thresholds, with values ​​of 1.4 and 1.1 respectively. This rule dynamically tightens the criteria for judging slope stability under high-intensity or long-duration rainfall conditions, thereby triggering early warnings or support responses before the safety factor drops to the normal warning level, providing a greater risk buffer margin for slope safety, and achieving true adaptive management of rainfall risks.

[0057] In a preferred embodiment, the first threshold Second threshold The threshold is dynamic and adaptively adjusts according to rainfall intensity and duration. Specifically, the adjustment rule is as follows: when rainfall intensity increases or rainfall duration increases, the first threshold... Second threshold The corresponding increase.

[0058] Considering that rainfall is the main external factor inducing slope instability, the first threshold Second threshold These are not fixed but rather adaptive parameters that dynamically adjust based on actual rainfall conditions. As rainfall intensity increases or duration lengthens, soil moisture content rises and pore water pressure increases, gradually weakening the overall anti-sliding capacity of the slope. Even if the safety factor value does not change drastically, the actual risk level has significantly increased. Therefore, the system will adjust accordingly. and The value of this value is equivalent to applying stricter safety criteria under more severe weather conditions. For example, during light rain or short-duration rainfall, Set to 1.3; however, in cases of heavy rainfall or continuous rainfall over several days, It may automatically increase to 1.45 or even higher. Simultaneously, it is also improved. This dynamic adjustment mechanism enables stability assessments to more sensitively reflect the evolution of risks under real working conditions, avoiding underestimation of danger due to the use of fixed thresholds, thereby ensuring that early warning and support decisions are always matched with current environmental conditions, and improving the adaptability and reliability of the intelligent slope protection system.

[0059] In a preferred embodiment, see Figure 2 The method further includes a correction step that takes into account the spatial non-uniformity of rainfall distribution: Obtain the spatial distribution data of rainfall retrieved from meteorological radar in the area where the target soil and rock slope is located, and divide the slope surface into multiple sub-regions; Based on the rainfall infiltration dynamic model, the contribution weight of each sub-region to the pore water pressure field inside the slope is calculated, and the contribution weight is solved by a neural network based on physical information. The pore water pressure responses of each sub-region are weighted and superimposed to obtain the corrected pore water pressure field considering spatially non-uniform rainfall.

[0060] To more accurately reflect the impact of actual rainfall on slope stability, this invention further introduces a correction for the non-uniformity of rainfall spatial distribution. Since natural rainfall is often not uniformly distributed across a slope, some areas may experience concentrated rainfall while others receive relatively less. Treating all areas as average rainfall can easily lead to inaccuracies in pore water pressure estimation. Therefore, the system first uses meteorological radar to obtain high-resolution spatial distribution data of rainfall and divides the entire slope surface into several sub-regions accordingly. Then, combining the physical processes of rainfall infiltration, a neural network model based on physical information is used to calculate the degree of influence of rainfall infiltration on the pore water pressure field within each sub-region, assigning a contribution weight to each region. This weight comprehensively reflects the effects of local rainfall intensity, topographic slope, and soil permeability on water migration and pressure accumulation. Finally, the pore water pressure responses caused by each sub-region are weighted and superimposed according to their respective weights to generate a more realistic corrected pore water pressure field that considers the effects of spatially non-uniform rainfall. This correction step significantly improves the accuracy of subsequent stability analysis, enabling slope risk assessment to capture potential landslide hazards caused by localized heavy rainfall.

[0061] In a preferred embodiment, the dynamic basic attribute dataset of the target soil and rock slope is obtained, wherein the nonlinear soil and rock mass parameter function that varies with volumetric water content is updated in real time based on field monitoring data using a Bayesian inversion method, and the Kalman filter algorithm is used to achieve optimal parameter estimation during the update process.

[0062] In a preferred embodiment, the real-time updating process of the nonlinear soil and rock mass parameter function that varies with volumetric water content employs a joint inversion framework that integrates hysteresis effects and dynamic parameter updates. The parameter state vector to be inverted is defined. for:

[0063] in, These are the parameters for the van Genuchten soil-water characteristic curve model. The unit is 1 / m. Dimensionless; The initial saturated permeability coefficient is expressed in m / s. This is an empirical index of the permeability coefficient; As a lag factor. Based on field monitoring data. (Including pore water pressure and volumetric water content), establish a joint observation equation:

[0064] in, For the observation operator, For observing the noise vector. Effective matrix suction. Due to lag factor Dynamic interpolation yields:

[0065] in, Effective matrix suction is defined as pore pressure. With pore water pressure The difference (i.e.) (), the unit is kPa; The dehumidification master curve function is expressed in kPa. This is the master moisture absorption curve function, with units in kPa; is the lag factor, with a value range of [0, 1], used to characterize the current hydraulic path state. The state vector is processed using a Kalman filter algorithm. Optimal estimation is performed to achieve synchronous dynamic updating of geotechnical parameters and hysteresis characteristics.

[0066] In a preferred embodiment, the method further includes a quantitative assessment step of the rainfall infiltration lag effect: The lag time is defined as the time difference between the peak moment of slope rainfall and the peak moment of pore water pressure at depth z, in hours. A lag time prediction model was constructed, which comprehensively considers soil permeability, depth, and rainfall characteristic parameters; The lag time is used as an output item of the dynamic stability assessment result to correct the advance time of slope instability warning.

[0067] In a preferred embodiment, the step of outputting a dynamic stability assessment report for the target soil and rock slope throughout the rainfall cycle further includes: Calculate and output the cumulative infiltration and leakage of the target soil and rock slope, in cubic meters; When the ratio of the cumulative infiltration to the total rainfall is less than a preset threshold, a runoff warning is triggered. The preset threshold is set based on the slope surface characteristics and vegetation cover.

[0068] When generating a dynamic stability assessment report for the target soil and rock slope throughout its entire rainfall cycle, the system not only focuses on the internal mechanical response of the slope but also simultaneously calculates the cumulative infiltration and cumulative leakage, both closely related to hydrological processes, in cubic meters. The cumulative infiltration reflects the total amount of rainwater actually infiltrating into the slope soil, while the cumulative leakage represents the amount of water that has flowed out through the bottom or sides of the slope. These two indicators together characterize the slope's ability to absorb and discharge rainfall. Based on this, the system further analyzes the ratio of cumulative infiltration to total rainfall to determine whether a large amount of runoff that cannot infiltrate has been generated on the surface. When this ratio is below a preset threshold, it indicates that most of the rainfall has failed to infiltrate effectively and has instead been converted into surface runoff, potentially leading to risks such as scouring, erosion, or even shallow landslides. In this case, the system will automatically trigger a runoff warning. This threshold is not a fixed value, but is reasonably set according to the surface characteristics of the slope (such as slope and degree of soil exposure) and vegetation cover (such as root soil stabilization capacity and canopy interception effect) to ensure that the early warning mechanism is both in line with physical reality and regionally adaptable, thereby providing a more comprehensive hydrological basis for disaster prevention decisions.

[0069] In a preferred embodiment, the method further includes a dynamic comparative evaluation step of different rainfall types: Multiple rainfall scenarios are set up, including uniform rainfall, front-peak rainfall, mid-peak rainfall, and back-peak rainfall; The corresponding stability assessment results were calculated under each rainfall condition. Based on the aforementioned stability assessment results, an ensemble learning algorithm is used to predict the most unfavorable rainfall type and its corresponding minimum stability safety factor for a future period of time, and an adaptive rainfall warning level is generated accordingly.

[0070] In a preferred embodiment, when the target soil slope is a multi-layered heterogeneous soil slope, the method further includes a step of inter-layer parameter continuity processing: A transition layer model is introduced at the soil layer interface, and the hyperbolic tangent function is used to describe the smooth transition of parameters at the interface:

[0071] in, These are parameter values ​​near the interface, with units determined based on the specific parameter. , These are the parameter values ​​for the upper and lower soil layers, respectively, with units determined based on the specific parameters. These are depth coordinates, in meters (m). This refers to the interface location, in meters (m). The thickness coefficient of the transition layer is expressed in meters (m). Numerical calculations are performed within the transition layer using an adaptive mesh refinement technique, with the mesh size adaptively adjusted according to the parameter gradient to improve calculation accuracy.

[0072] A dynamic assessment system for the stability of soil and rock slopes that considers rainfall infiltration effects includes: The data acquisition unit is used to acquire the dynamic basic attribute dataset of the target soil and rock slope, including slope geometric parameters, soil and rock mass parameters, hydrological and meteorological parameters, and initial seepage field parameters. The model building unit is used to build a dynamic model of rainfall infiltration that considers non-Darcy flow and hysteresis effects based on the dynamic basic attribute dataset. The pore water pressure calculation unit is used to calculate the spatiotemporal distribution characteristics of pore water pressure based on the rainfall infiltration dynamic model, and to make real-time corrections through a machine learning proxy model. The stability assessment unit is used to calculate the dynamic stability safety factor based on the modified pore water pressure field and obtain the dynamic stability assessment results. The support decision unit is used to compare the dynamic stability assessment results with a preset threshold to determine whether to trigger a dynamic support decision response. The report output unit is used to output a dynamic stability assessment report for the entire rainfall cycle.

[0073] This dynamic assessment system for soil and rock slope stability focuses on the rainfall infiltration effect, achieving real-time and accurate assessment of slope safety status through multi-module collaborative operation. The system first acquires a dynamic dataset of basic attributes of the target slope using a data acquisition unit. This includes the slope's geometry (e.g., slope height, slope angle), soil and rock physical and mechanical parameters (e.g., permeability coefficient, shear strength), hydrological and meteorological information (e.g., rainfall intensity, duration), and the initial seepage field state, providing comprehensive input for subsequent analysis. Based on this, the model building unit integrates non-Darcy flow theory and the rainfall infiltration hysteresis effect to establish a more realistic dynamic infiltration model. Non-Darcy flow describes the nonlinear flow characteristics of water in low-permeability soil under heavy rainfall, while the hysteresis effect reflects the time delay in pore water pressure response to rainfall. Subsequently, the pore water pressure calculation unit uses this model to simulate the temporal and spatial evolution of pore water pressure during rainfall and introduces a machine learning surrogate model to correct the calculation results in real time, compensating for biases caused by model simplification or parameter uncertainties, and improving prediction accuracy. The stability assessment unit dynamically calculates the slope's safety factor based on the corrected pore water pressure field, generating a stability assessment result that updates over time. When this safety factor falls below a preset threshold, the support decision-making unit immediately activates the response mechanism to determine whether intervention measures such as reinforcement, drainage, or personnel evacuation are necessary. Finally, the report output unit integrates all the data and analysis results to generate a dynamic stability assessment report covering the entire rainfall cycle, providing engineering managers with scientific, intuitive, and actionable decision support. The entire system achieves a closed-loop process from data perception, model simulation, intelligent correction to risk warning and decision response, significantly improving the timeliness and reliability of slope disaster prevention and control.

[0074] In a preferred embodiment, the system further includes: The rainfall spatial distribution correction unit is used to acquire rainfall spatial distribution data retrieved from meteorological radar, divide the slope into multiple sub-regions, and calculate the contribution weight of each sub-region to the pore water pressure field. The real-time parameter update unit is used to update the parameters of the soil and rock mass in real time using the Bayesian inversion method and the Kalman filter algorithm. The lag effect assessment unit is used to quantitatively assess the lag effect of rainfall infiltration, construct a lag time prediction model, and correct the advance time of slope instability warning.

[0075] To further improve assessment accuracy, the system integrates three key enhancement modules. First, the rainfall spatial distribution correction unit utilizes high-resolution rainfall data retrieved from meteorological radar to divide the entire slope into multiple sub-regions, identifying the differences in actual rainfall intensity experienced at different locations. Based on this, it calculates the contribution weight of each sub-region to the overall pore water pressure field by combining its area, slope, and infiltration capacity, thus avoiding errors caused by the assumption of uniform rainfall across the entire slope in traditional methods. Second, the real-time parameter update unit continuously optimizes key parameters of the soil and rock mass, such as permeability coefficient and water-holding capacity, using field monitoring data such as water content, displacement, or pore pressure changes by fusing Bayesian inversion and Kalman filtering algorithms. Bayesian inversion continuously corrects the probability distribution of parameters based on observation information, while Kalman filtering achieves efficient and stable optimal estimation over time, ensuring the model always closely reflects the true physical state of the slope. Finally, the hysteresis effect assessment unit specifically quantifies the time delay of rainwater infiltration from the slope into the deep soil and the resulting peak pore water pressure. By constructing a hysteresis time prediction model that comprehensively considers soil permeability, burial depth, and rainfall characteristics, it accurately determines the response lag at different depths and adjusts the release time of instability warnings accordingly. For example, even after heavy rainfall, the hysteresis effect may still cause a sharp drop in stability several hours later; the system will extend the warning window accordingly to avoid missed warnings. These three units together enhance the system's ability to characterize complex rainfall infiltration instability processes, making the dynamic assessment results more refined, reliable, and forward-looking.

[0076] In a preferred embodiment, the system is deployed on a cloud computing platform, supports concurrent access by multiple users and real-time data processing, and is equipped with a visual interface to display dynamic stability assessment results and early warning information.

[0077] Deployed on a cloud computing platform, the system fully leverages the elastic computing, high availability, and distributed processing capabilities of cloud architecture. It can simultaneously respond to multiple user access requests and efficiently process real-time data streams from various sensors and meteorological systems. When multiple slope monitoring points simultaneously upload information such as rainfall, pore water pressure, or displacement, the cloud platform dynamically allocates computing resources to ensure that model calculations and evaluation result generation are unaffected by concurrent loads, guaranteeing response speed and system stability. Furthermore, the system features an intuitive visualization interface that presents complex dynamic stability assessment results graphically. For example, it displays changes in safety factors using time curves, maps risk levels for different areas of the slope using color gradients, or overlays warning areas and lag effect prediction durations onto a 3D topographic map. Whether in the office or on-site, users can view the current slope status, historical evolution trends, and system warnings in real time via a browser, significantly improving information acquisition efficiency and the timeliness of emergency decision-making. This cloud computing plus front-end visualization working mode not only reduces reliance on local hardware but also enables remote, collaborative, and intelligent slope safety monitoring.

[0078] The above embodiments are merely illustrative of the technical solutions of the present invention and are not intended to limit its scope of protection. Those skilled in the art can make various modifications, substitutions, or improvements to the above embodiments without departing from the core concept of the present invention. All equivalent changes, structural substitutions, parameter adjustments, or functional extensions made based on the technical essence of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effects, characterized in that, Includes the following steps: Obtain a dynamic basic attribute dataset of the target soil and rock slope, wherein the dynamic basic attribute dataset includes slope geometric parameters, nonlinear soil and rock mass parameter functions that vary with volumetric water content, hydrological and meteorological parameters, and initial seepage field parameters. Based on the aforementioned dynamic basic attribute dataset, a rainfall infiltration dynamic model considering non-Darcy flow and hysteresis effects is constructed. The rainfall infiltration dynamic model dynamically divides a complete rainfall process into a full infiltration stage and a surface runoff stage, and introduces a pore water pressure response hysteresis factor. Based on the rainfall infiltration dynamic model, the spatiotemporal distribution characteristics of pore water pressure of the target soil and rock slope under time-varying rainfall boundary conditions are calculated, and real-time correction is performed through a machine learning surrogate model to obtain the corrected pore water pressure field. Based on the modified pore water pressure field, the dynamic stability safety factor of the target soil and rock slope at different rainfall times is calculated using the dynamic unsaturated soil shear strength model, and the dynamic stability assessment results are obtained. Based on the dynamic stability assessment results, the potential instability period and dangerous sliding surface location of the target soil and rock slope are identified, and based on the preset stability threshold, it is determined whether a dynamic support decision response is triggered. Output a dynamic stability assessment report of the target soil and rock slope throughout the entire rainfall cycle.

2. The dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effect according to claim 1, characterized in that, The construction of a rainfall infiltration dynamic model considering non-Darcy flow and hysteresis effects based on the aforementioned dynamic basic attribute dataset further includes: Based on the real-time rainfall intensity data and dynamic saturated permeability coefficient in the hydrological and meteorological parameters, the infiltration state of the target soil and rock slope surface is determined, wherein the dynamic saturated permeability coefficient... Expressed as volumetric water content power function: ； in, The dynamic saturated permeability coefficient is expressed in m / s. The initial saturated permeability coefficient is expressed in m / s. This refers to the volumetric water content. Residual moisture content; This represents the saturated moisture content. It is an experience index; When the real-time rainfall intensity is less than or equal to the dynamic saturated permeability coefficient, the current stage is divided into the complete infiltration stage, and the flow boundary condition is used to simulate the slope rainwater infiltration process. When the real-time rainfall intensity is greater than the dynamic saturated permeability coefficient, the current stage is classified as the surface runoff stage, and a slope runoff-subsurface seepage coupling coefficient is introduced. To correct the boundary conditions, where is a coefficient, with a value ranging from 0 to 1.

3. The dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effect according to claim 2, characterized in that, The process of calculating the spatiotemporal distribution characteristics of pore water pressure on the target soil and rock slope under time-varying rainfall boundary conditions based on the rainfall infiltration dynamic model, and then performing real-time correction using a machine learning surrogate model to obtain the corrected pore water pressure field, further includes: The initial pore water pressure distribution is obtained by solving the seepage problem in unsaturated soil based on the Richards equation. A proxy model based on a deep neural network is constructed, which takes the initial pore water pressure distribution and dynamic basic attribute dataset as input and outputs the pore water pressure correction. The pore water pressure correction is superimposed on the initial pore water pressure distribution to obtain the corrected pore water pressure field.

4. The dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effect according to claim 1, characterized in that, The dynamic stability safety factor of the target soil and rock slope at different rainfall times is calculated using a dynamic unsaturated soil shear strength model based on the modified pore water pressure field, and the dynamic stability assessment results are obtained. This further includes: Based on the modified pore water pressure field, the dynamic volumetric water content distribution and matrix suction distribution inside the target soil and rock slope are calculated. The dynamic shear strength parameters were calculated using Fredlund's dual-stress variable strength theory, where the shear strength... The expression is: ； in, Shear strength, expressed in kPa; Effective cohesion, expressed in kPa; The total normal stress is expressed in kPa. This represents pore gas pressure, expressed in kPa. Pore ​​water pressure, in kPa; The effective internal friction angle is expressed in degrees (°). The matrix suction angle is expressed in units of 1 / 2. Effective cohesion and effective internal friction angle All change dynamically with volumetric water content and matrix suction: ； ； in, Effective cohesion, expressed in kPa; The initial effective cohesion is expressed in kPa. This is the cohesion attenuation coefficient; This refers to the dynamic volumetric moisture content. The effective internal friction angle is expressed in degrees (°). This is the initial effective internal friction angle, in degrees. The internal friction angle attenuation coefficient; matrix suction angle. With matrix suction The changes are based on the following empirical relationships: ； in, The matrix suction angle is expressed in degrees (°). The initial matrix suction angle is expressed in degrees (°). The attenuation coefficient; Standard atmospheric pressure, unit: kPa; Calculation of dynamic stability safety factor based on limit equilibrium method : ； in, The dynamic stability safety factor; For the first The effective cohesive force of the strip, expressed in kPa; For the first The length of the sliding surface of the strip, in meters; For the first Weight per unit width of the strip, in kN / m; For the first The inclination angle of the sliding surface of the strip, in degrees; For the first The pore water pressure of the strip, in kPa; For the first The base area per unit width of the strip ( (Unit: m) 2 ; For the first The effective internal friction angle of the strip, in degrees; For the first The pore air pressure of the strip, in kPa; For the first The pore water pressure of the strip, in kPa; For the first The matrix suction angle of the strip is expressed in degrees. Based on the dynamic stability safety factor at different rainfall times, the global minimum safety factor and its corresponding most dangerous potential sliding surface position are determined to form the dynamic stability assessment result.

5. The dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effect according to claim 1, characterized in that, The step of identifying potential instability periods and dangerous sliding surface locations of the target soil and rock slope based on the dynamic stability assessment results, and determining whether to trigger a dynamic support decision response based on a preset stability threshold, further includes: The global minimum safety factor in the dynamic stability assessment result is compared with the first threshold. Second threshold Comparison, among which ,and The value range is 1.3 to 1.

5. The value range is 1.0 to 1.2; If the global minimum security factor is greater than the first threshold The target soil and rock slope is determined to be in a stable state, and no support decision is triggered. If the global minimum security factor is within the first threshold With the second threshold If the target rock and soil slope is determined to be in an unstable state, the first-level dynamic support decision response is triggered, and an early warning message is output. If the global minimum security factor is less than the second threshold The target soil and rock slope is determined to be in an unstable state, triggering a second-level dynamic support decision response, and it is recommended to adopt emergency support measures.

6. The dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effect according to claim 1, characterized in that, The method also includes a correction step that takes into account the non-uniformity of spatial distribution of rainfall: Obtain the spatial distribution data of rainfall retrieved from meteorological radar in the area where the target soil and rock slope is located, and divide the slope surface into multiple sub-regions; Based on the rainfall infiltration dynamic model, the contribution weight of each sub-region to the pore water pressure field inside the slope is calculated, and the contribution weight is solved by a neural network based on physical information. The pore water pressure responses of each sub-region are weighted and superimposed to obtain the corrected pore water pressure field considering spatially non-uniform rainfall.

7. The dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effect according to claim 1, characterized in that, The process of acquiring the dynamic basic attribute dataset of the target soil and rock slope involves the nonlinear soil and rock mass parameter function that varies with volumetric water content being updated in real time based on field monitoring data using a Bayesian inversion method. The update process employs a Kalman filter algorithm to achieve optimal parameter estimation.

8. The dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effect according to claim 1, characterized in that, The method also includes a quantitative assessment step for the rainfall infiltration lag effect: The lag time is defined as the time difference between the peak moment of slope rainfall and the peak moment of pore water pressure at depth z, in hours. A lag time prediction model was constructed, which comprehensively considers soil permeability, depth, and rainfall characteristic parameters; The lag time is used as an output item of the dynamic stability assessment result to correct the advance time of slope instability warning.

9. The dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effect according to claim 1, characterized in that, The method also includes a dynamic comparative evaluation step for different rainfall types: Multiple rainfall scenarios are set up, including uniform rainfall, front-peak rainfall, mid-peak rainfall, and back-peak rainfall; The corresponding stability assessment results were calculated under each rainfall condition. Based on the aforementioned stability assessment results, an ensemble learning algorithm is used to predict the most unfavorable rainfall type and its corresponding minimum stability safety factor for a future period of time, and an adaptive rainfall warning level is generated accordingly.

10. A dynamic assessment method for the stability of soil and rock slopes considering rainfall infiltration effect according to claim 1, characterized in that, When the target soil slope is a multi-layered heterogeneous soil slope, the method further includes a step of inter-layer parameter continuity processing: A transition layer model is introduced at the soil layer interface, and the hyperbolic tangent function is used to describe the smooth transition of parameters at the interface: ； in, These are parameter values ​​near the interface, with units determined based on the specific parameter. , These are the parameter values ​​for the upper and lower soil layers, respectively, with units determined based on the specific parameters. These are depth coordinates, in meters (m). This refers to the interface location, in meters (m). The thickness coefficient of the transition layer is expressed in meters (m). Numerical calculations are performed within the transition layer using an adaptive mesh refinement technique, with the mesh size adaptively adjusted according to the parameter gradient to improve calculation accuracy.

Images