Slope instability probability dynamic evaluation system based on multi-source data and artificial intelligence

Abstract

The present application provides a slope instability probability dynamic evaluation system based on multi-source data and artificial intelligence, relates to the technical field of slope safety monitoring and big data processing, and comprises: a multi-dimensional state sensing module for acquiring multi-source heterogeneous data such as surface macroscopic deformation, slope body fine structure and rock-soil internal physical field from macro remote sensing, close-range observation and deep sensing array; and an intelligent evaluation module for calculating response sensitivity index, field state offset index and structure damage index from three dimensions of external disturbance response, macroscopic state evolution and internal microscopic damage respectively. The three indexes are weighted and fused to generate a comprehensive stability index, based on which the dynamic and accurate judgment of the slope stability grade and the early warning grade is realized, and the corresponding graded response measures are triggered. The present application establishes an evaluation system from the surface to the inside and from the result to the cause, can effectively capture the weak precursor information before instability, and improves the accuracy, reliability and timeliness of the early warning.

Description

Technical Field

[0001] This invention relates to the fields of slope safety monitoring and big data processing technology, specifically a dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence. Background Technology

[0002] Slope stability is crucial for the safe operation of open-pit mines, large-scale water conservancy projects, major transportation routes, and mountain infrastructure. Due to the complexity of geological structures and the continuous disturbance of dynamic changes in the hydrological environment, slopes may deform or even become unstable, posing a serious threat to life and property. Therefore, establishing a precise and efficient slope safety monitoring and early warning system is of paramount importance for disaster prevention.

[0003] Traditional slope monitoring methods mainly rely on manual on-site inspections and the deployment of contact sensors (such as crack gauges, displacement gauges, and inclinometers). These methods suffer from limited monitoring range, low data collection frequency, high labor intensity, and heavy dependence on personnel experience. More importantly, they typically only detect anomalies when deformation has become quite significant, resulting in poor early warning timeliness and difficulty in capturing early, subtle signs of instability.

[0004] With technological advancements, non-contact, large-scale monitoring technologies, such as Global Navigation Satellite System (GNSS), Synthetic Aperture Radar Interferometry (InSAR), and 3D laser scanning, have become widely used. These technologies can acquire the macroscopic surface deformation field of slopes, improving the automation level and coverage of monitoring to some extent. However, these technologies primarily focus on the "result" of the slope, namely macroscopic surface deformation. They reflect the final external manifestation of the adjustment of the internal stress field and the accumulation of structural damage, making them a "lagging" indicator. In many sudden landslide events, macroscopic surface deformation may not be obvious before instability, making it difficult for these methods to achieve effective early warning. Furthermore, their warning logic is often based on simple deformation rate or cumulative displacement thresholds. This linear threshold judgment method cannot effectively cope with nonlinear and sudden instability modes caused by complex coupling of multiple factors, easily leading to missed or false alarms. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence includes a multi-dimensional state perception module and an intelligent assessment module. The multidimensional state perception module is configured to acquire surface macro deformation data, slope fine structure data, and soil internal physical field data of the monitoring area by connecting a macro remote sensing system, a near-field monitoring system, and a deep sensing array, and classify them into macro datasets, structural datasets, and physical field datasets. The intelligent assessment module includes a response assessment unit, a state assessment unit, a damage assessment unit, and a risk management unit. The response assessment unit evaluates the sensitivity of each monitoring zone to external disturbances based on the macro dataset and generates a corresponding response sensitivity index. The state assessment unit evaluates the degree of deviation of the evolution of the physical field within each monitoring zone based on the structural dataset and generates a corresponding field state offset index. The damage assessment unit analyzes the degree of damage accumulation in the microstructure of the soil and rock mass based on the physical field dataset and generates a corresponding structural damage index. The risk management unit sets a fixed range of stability threshold intervals and warning threshold intervals, and combines the response sensitivity index, field state offset index, and structural damage index to determine the stability level, warning level, and instability probability, outputs the assessment results, and triggers corresponding response measures.

[0007] Furthermore, the macroscopic dataset includes the three-dimensional displacement field of the land surface, the phase change rate of synthetic aperture radar interferometry, the gradient of the land surface temperature field, the regional equivalent damping ratio, the far-field seismic wave amplification coefficient, the tidal load strain response coefficient, the atmospheric pressure load strain response coefficient, the regional rainfall infiltration rate, and the land surface vegetation cover index for each partition of the monitoring area.

[0008] Furthermore, the structured dataset includes a digital elevation model for each monitoring zone, a high-precision three-dimensional point cloud model, the orientation of the dominant structural surface, the fractal dimension of the fracture network, the surface roughness of the slope, the density of gully development, the distribution map of seepage outflow points, the spatial distribution map of the self-electric field, and the apparent resistivity tomography model.

[0009] Furthermore, the physical field dataset includes at least one of the following depth characteristic parameters within each monitoring zone: the self-potential gradient tensor measured by the electrode array, the complex resistivity phase angle measured by the multi-frequency induced polarization method, the temporal variability of dissolved radon concentration measured by the hydrological monitoring well, the relative change rate of the wake interference wave velocity calculated by repeated source monitoring, the Kaiser effect breakdown ratio calculated by the acoustic emission monitoring system, the inelastic energy dissipation rate calculated by the distributed fiber optic strain sensor, the microgravity anomaly gradient obtained by time-series gravity measurement, the rock mass dielectric constant heterogeneity index obtained by ground penetrating radar inversion, the T2 spectrum distribution morphology parameter measured by borehole nuclear magnetic resonance, and the surface mineral characteristic absorption depth ratio obtained by hyperspectral imaging analysis.

[0010] Furthermore, the calculation process for the response sensitivity index is as follows: S11. Extract parameters characterizing external disturbances from the macro dataset, including far-field seismic wave amplification coefficient, tidal load strain response coefficient, atmospheric pressure load strain response coefficient, and regional rainfall infiltration rate. After normalizing the above parameters, perform weighted summation to obtain the comprehensive quantitative value E of external disturbance. S12. Extract parameters characterizing the slope response from the macro dataset, including the three-dimensional displacement field of the ground surface, the phase change rate of synthetic aperture radar interferometry, and the gradient of the ground surface temperature field. After normalizing the above parameters, perform weighted summation to obtain the comprehensive quantitative value of the slope response. S13. Divide the comprehensive quantitative value of the slope response by the comprehensive quantitative value of the external disturbance to obtain the initial sensitivity value. Then, combine the regional equivalent damping ratio and the surface vegetation cover index to correct the initial sensitivity value, and finally obtain the response sensitivity index.

[0011] Furthermore, the calculation process for the field state offset index is as follows: S21. Based on the structural dataset of historical stable periods, a reference benchmark model is established for each monitoring zone. The model includes the fractal dimension of the reference fracture network, the development density of the reference gully, the spatial distribution map of the reference self-electric field, and the reference apparent resistivity tomography model. S22. Compare the currently acquired fractal dimension of the fracture network and gully development density with the corresponding reference values ​​in the reference benchmark model, calculate their relative rate of change, and obtain the structural morphology deviation. S23. Perform spatial difference operation on the currently acquired self-electric field spatial distribution map and apparent resistivity tomography model with the corresponding reference map and model in the reference benchmark model, calculate its spatial inconsistency norm, and obtain the physical field morphology deviation. S24. The structural morphological deviation and the physical field morphological deviation are weighted and fused to obtain the field state offset index.

[0012] Furthermore, the calculation process for the structural damage index is as follows: S31. The depth characteristic parameters in the physical field dataset are divided into micro-fracture damage parameters and fluid-structure interaction degradation parameters according to their physical meaning. Among them, micro-fracture damage parameters include Kaiser effect breakdown ratio, relative change rate of wake interference wave velocity and inelastic energy dissipation rate; fluid-structure interaction degradation parameters include complex resistivity phase angle and time-series variation rate of dissolved radon gas concentration. S32. Normalize each type of parameter and calculate its first derivative over time to characterize its dynamic rate of change. S33. Perform weighted summation on the normalized values of the micro - fracture damage - type parameters and their change rates to obtain the micro - fracture damage component; perform weighted summation on the normalized values of the fluid - solid coupling deterioration - type parameters and their change rates to obtain the fluid - solid coupling deterioration component. S34. According to the type of rock and soil mass in the monitoring area, assign different weight coefficients to the micro - fracture damage component and the fluid - solid coupling deterioration component and then sum them to obtain the structural damage degree index.

[0013] Further, the specific process for the risk management unit to judge the stability level is as follows: Denote the upper limit of the stability threshold interval as maxSY and the lower limit as minSY; obtain the comprehensive stability index CSI by performing weighted summation on the response sensitivity index RSI, the field state deviation index FSDI, and the structural damage degree index SDI. If CSI < minSY, the stability level is "stable", and the response measure is to maintain the conventional monitoring frequency; if minSY ≤ CSI ≤ maxSY, the stability level is "basically stable", and the response measure is to automatically increase the data acquisition frequency of the deep - sensing array by 25% - 50% and conduct data correlation review on the most significantly changing parameters; if CSI > maxSY, the stability level is "sub - stable", and the response measure is to start the encrypted observation mode and notify the management personnel to pay attention.

[0014] Further, the specific process for the risk management unit to judge the warning level is as follows: Denote the upper limit of the warning threshold interval as maxWY and the lower limit as minWY; if the comprehensive stability index CSI < minWY, the warning level is "no warning", and no additional response strategy is triggered; if minWY ≤ CSI ≤ maxWY, the warning level is "first - level warning", and the first - level response strategy is triggered; if CSI > maxWY, the warning level is "second - level warning", and the second - level response strategy is triggered.

[0015] Further, the first - level response strategy includes: adopting systematic bolt support with a spacing of 4.4 m - 4.8 m, and supplementing with pressure grouting at a standard of 15 kg - 16 kg per square meter for repair; and adding drainage facilities at a density of 0.9 - 1.0 per 100 square meters to improve the internal environment. The second - level response strategy includes: adopting systematic bolt support with a spacing of 3.1 m - 3.5 m, and supplementing with pressure grouting at a standard of 22 kg - 23 kg per square meter for repair; and adding drainage facilities at a density of 1.3 - 1.5 per 100 square meters to improve the internal environment.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention, through the Structural Damage Index (SDI), can directly capture early, weak precursory information such as micro-fractures in rock masses before instability, significantly advancing the early warning threshold, gaining valuable time for disaster prevention and mitigation, and effectively avoiding major losses caused by sudden landslides.

[0017] This invention constructs a more robust and comprehensive Comprehensive Stability Index (CSI) by integrating three physically distinct and independent indices: RSI, FSDI, and SDI. This effectively eliminates interference from single data sources, improving the accuracy and reliability of the assessment results. This invention not only determines whether a slope is "dangerous," but also reveals the dominant factors leading to increased risk by analyzing the weights and trends of the three sub-indices. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the execution of the core logic flow nodes of the overall system of the present invention; Figure 2 This is a schematic diagram of the system process execution of the present invention. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Please see Figures 1 to 2 This invention provides a technical solution: a dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence, comprising a multi-dimensional state perception module and an intelligent assessment module. The multidimensional state perception module is configured to acquire surface macro deformation data, slope fine structure data, and soil internal physical field data of the monitoring area by connecting macro remote sensing system, near-field monitoring system, and deep sensing array, and classify them into macro dataset, structural dataset, and physical field dataset. The intelligent assessment module includes a response assessment unit, a state assessment unit, a damage assessment unit, and a risk management unit. The response assessment unit evaluates the sensitivity of each monitoring zone to external disturbances based on the macro dataset and generates the corresponding response sensitivity index (RSI). The state assessment unit evaluates the degree of deviation of the evolution of the physical field within each monitoring zone based on the structural dataset and generates the corresponding field state deviation index (FSDI). The damage assessment unit analyzes the degree of damage accumulation in the microstructure of the soil and rock mass based on the physical field dataset and generates the corresponding structural damage index (SDI). The risk management unit sets a fixed range of stability threshold interval (SY) and warning threshold interval (WY). Combining the response sensitivity index (RSI), field state deviation index (FSDI), and structural damage index (SDI), it determines the stability level, warning level, and instability probability, outputs the assessment results, and triggers corresponding response measures.

[0022] The macro dataset includes the three-dimensional displacement field of the land surface, the phase change rate of synthetic aperture radar interferometry, the gradient of the land surface temperature field, the regional equivalent damping ratio, the far-field seismic wave amplification factor, the tidal load strain response coefficient, the atmospheric pressure load strain response coefficient, the regional rainfall infiltration rate, and the land surface vegetation cover index for each partition of the monitoring area.

[0023] The structural dataset includes a digital elevation model (DEM) for each monitoring zone, a high-precision 3D point cloud model, the orientation of dominant structural surfaces, the fractal dimension of the fracture network, the surface roughness of the slope, the density of gully development, the distribution map of seepage outflow points, the spatial distribution map of the self-electric field (SP), and the apparent resistivity tomography (ERT) model.

[0024] The physical field dataset includes at least one of the following depth characteristic parameters within each monitoring zone: the self-potential gradient tensor measured by the electrode array, the complex resistivity phase angle measured by the multi-frequency induced polarization method, the temporal variability of dissolved radon concentration measured by the hydrological monitoring well, the relative change rate of wake interference wave velocity calculated by repeated source monitoring, the Kaiser effect breakdown ratio calculated by the acoustic emission monitoring system, the inelastic energy dissipation rate calculated by distributed fiber optic strain sensing, the microgravity anomaly gradient obtained by time-series gravity measurement, the rock mass dielectric constant heterogeneity index obtained by ground penetrating radar inversion, the T2 spectrum distribution morphology parameter measured by borehole nuclear magnetic resonance, and the surface mineral characteristic absorption depth ratio obtained by hyperspectral imaging analysis.

[0025] To make the objectives, technical solutions, and advantages of this technical solution clearer, a detailed description of this technical solution will be provided below with reference to the accompanying drawings and specific embodiments. Please refer to... Figure 1This invention provides a technical solution: a dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence. This embodiment uses the steep northern slope of a large open-pit mine as the monitoring area. This area has a complex geological structure, containing various lithologies (granite, schist) and weak structural surfaces, and is significantly affected by seasonal heavy rainfall disturbances.

[0026] This system includes a multi-dimensional state perception module and an intelligent evaluation module.

[0027] I. Multidimensional State Perception Module: The multidimensional state perception module is configured to acquire surface macroscopic deformation data, slope fine structure data, and soil internal physical field data of the monitoring area by connecting to the macroscopic remote sensing system, near-field monitoring system, and deep sensing array, and classify them into macroscopic datasets, structural datasets, and physical field datasets.

[0028] 1.1 Data Acquisition and Sensor Network Layout: To construct a comprehensive, high-fidelity dataset, this embodiment deploys a three-layer, three-dimensional monitoring network: A macroscopic remote sensing system primarily utilizes C-band synthetic aperture radar (SAR) data from satellite constellations, acquiring interferometric wide-swath (IW) imagery covering the entire mining area with a 12-day revisit cycle. Simultaneously, high-resolution optical satellite imagery (such as Gaofen-2) is combined for calculating the land cover index (NDVI). Furthermore, a regional meteorological station network provides rainfall and air pressure data, and the national seismic network provides far-field ground motion records.

[0029] Near-field monitoring system: A long-distance ground-based 3D laser scanner is deployed on the stable bedrock opposite the slope to scan the slope daily, generating a high-density 3D point cloud model. Simultaneously, high-precision GPS / GNSS receivers are deployed at key locations on the slope to monitor 3D displacement in real time.

[0030] Deep sensor array: Deep sensors are deployed through boreholes within each monitoring zone. Includes: Fiber Optic Strain Sensor (BOTDA): Deployed along the entire length of the borehole to sense the strain distribution inside the rock and soil.

[0031] Multi-electrode array: used to perform self-electric field (SP) and apparent resistivity tomography (ERT) measurements.

[0032] Acoustic emission (AE) sensor: captures elastic wave signals generated by microfractures in rock mass.

[0033] Hydrological monitoring wells: Sensors are installed to monitor groundwater levels and the concentration of dissolved radon in water samples.

[0034] 1.2 Dataset Construction and Structuring; All sensor data is connected to a central data processing server. The server is configured with a unified data acquisition and processing cycle; for example, macroscopic data is updated daily, near-field data is updated every 6 hours, and deep data is updated every hour. In each processing cycle t, the multi-dimensional state perception module generates structured data records for each monitoring zone and stores them in three databases respectively.

[0035] 1.2.1 The macro-dataset focuses on describing the overall response of the slope to external environmental disturbances. Examples of data record fields are shown in Table 1 below: Table 1: Examples of data records for the macro dataset

[0036] The far-field seismic wave amplification factor, denoted as Pext,1, is used to quantify the amplification effect of seismic waves on the slope relative to stable bedrock. This phenomenon is known as the "site effect." When seismic waves propagate from deep within the Earth to the surface, energy redistribution and focusing occur due to local topography (such as ridges and slopes) and geological conditions (such as loose overburden layers), resulting in an actual vibration intensity on the slope being much greater than that in the surrounding stable areas. A higher amplification factor means that the slope will experience stronger shaking during an earthquake, and the risk of instability will increase accordingly. This embodiment uses the classic standard spectral ratio method in geotechnical engineering seismology to calculate the amplification factor. A seismometer is installed on (or near) the monitored slope to record seismic wave data. A seismometer of the same model is installed at an exposed, hard, and intact bedrock (such as granite) a short distance from the slope as a reference point; the recorded seismic waves can be approximated as unamplified "input" ground motion. Multiple records of moderate-to-strong earthquakes occurring at distant locations (typically greater than 100 kilometers) were selected to ensure that the seismic waves arriving at the two stations had essentially consistent incident characteristics. Baseline correction and instrument response removal were performed on the data of the same earthquake event recorded simultaneously at both stations. The S-wave (shear wave) portion with a high signal-to-noise ratio was extracted and converted from a time-domain signal to a frequency-domain signal using Fourier transform (FFT) to obtain the Fourier amplitude spectrum A(f) for each station, where f represents the frequency. The amplitude spectrum of the target station was divided by the amplitude spectrum of the reference station to obtain the spectral ratio SR(f) of the earthquake event: SR(f) = Atarget_station(f) / Areference_station(f). This ratio reflects the vibration amplification factor of the slope site relative to the bedrock at various frequencies. To eliminate the randomness of a single earthquake event, the spectral ratios calculated from multiple earthquake events were geometrically averaged to obtain a stable and reliable average spectral ratio curve. The amplification factor of far-field seismic waves is usually taken as the peak or average value of the frequency band of interest in the project (e.g., 1-10 Hz) on this average spectral ratio curve, and is used as the final value of the amplification factor of far-field seismic waves.

[0037] The tidal load strain response coefficient, denoted as Pext,2, is used to quantify the sensitivity of slope soil and rock masses to periodic stress changes induced by solid tides, a phenomenon reflecting the Earth's elastic response. Due to the gravitational pull of the Moon and the Sun, the Earth as a whole undergoes minute, regular deformations, which act as loads on slopes. A rock mass with an intact internal structure and excellent mechanical properties responds weakly to this load, while a rock mass with a network of fractures or weak interlayers will exhibit a more significant strain response. A higher response coefficient indicates poorer structural integrity of the slope, greater sensitivity to long-term, low-frequency, minute stress disturbances, and a correspondingly increased risk of potential creep failure.

[0038] This embodiment employs the classic load-response linear regression method from the fields of geophysics and statistics to calculate the data. A high-precision borehole strain gauge is installed deep within the slope to continuously record the micro-strain time series data ε(t) of the rock mass over a long period. Simultaneously, based on the precise geographical coordinates and astronomical calendar of the monitoring point, a mature geophysical model is used to calculate the theoretical solid tidal stress time series Ttheo(t) at that point. This theoretical value can be considered as a precisely known "input" load. To establish a robust statistical relationship, it is necessary to obtain long-term (usually several weeks to several months) synchronous continuous observation data.

[0039] The acquired strain data underwent de-drift and de-trend processing to eliminate interference from the instrument itself and long-term tectonic movements. Using the preprocessed strain time series ε(t) as the dependent variable and the theoretical solid tidal series Ttheo(t) as the independent variable, a linear regression model was constructed: ε(t) = Ctidal × Ttheo(t) + εnoise. The least squares method was used to fit this model, and the slope that best describes the correlation between the two was obtained. This slope is the tidal load strain response coefficient Ctidal, whose physical meaning is the actual rock mass strain induced by a unit of theoretical tidal stress.

[0040] The atmospheric pressure load strain response coefficient, denoted as Pext,3, is used to quantify the degree to which slope soil and rock masses respond to changes in surface atmospheric pressure. This phenomenon is known as the "breathing effect" of soil and rock masses. Fluctuations in surface atmospheric pressure exert a uniformly varying load on the Earth's crust, causing the soil and rock masses to compress or rebound. A dense, poorly porous rock mass is less affected, while a slope with well-developed pores and fractures, especially containing confined aquifers, will exhibit a more significant deformation response. A higher response coefficient indicates that the slope is more sensitive to medium- to high-frequency load changes, potentially indicating the presence of dominant seepage channels or higher pore pressure, and consequently, increased susceptibility to instability.

[0041] This embodiment also employs the load-response linear regression method for calculation; a high-precision borehole strain gauge is installed deep within the slope to acquire micro-strain data ε(t) of the rock mass. Simultaneously, high-precision barometers are deployed within the monitoring area to synchronously record the continuous time series P(t) of atmospheric pressure, which is considered the "input" load acting on the slope. This ensures that the strain and pressure data have strict time synchronization and consistent sampling frequency over long time scales (weeks to months).

[0042] The collected strain data were preprocessed to remove interference signals unrelated to atmospheric pressure. A linear regression model was constructed using the preprocessed strain time series ε(t) as the dependent variable and the atmospheric pressure time series P(t) as the independent variable: ε(t) = Cpressure × P(t) + εnoise. The optimal regression coefficient was obtained by fitting the data using the least squares method. This regression coefficient is the atmospheric pressure load strain response coefficient Cpressure, which physically represents the actual rock mass strain induced by a unit change in atmospheric pressure.

[0043] Regional rainfall infiltration rate, denoted as Pext,4, is used to quantify the proportion of effective water infiltrating into the slope during a rainfall event, representing the core driving mechanism of rainfall-induced landslides. After rainfall reaches the slope surface, some flows away as runoff, while the rest infiltrates into the soil and rock mass. The infiltrated water increases the slope's weight, softens the soil and rock, and raises the groundwater level, generating pore water pressure and significantly weakening the slope's shear strength. A higher infiltration rate means a weaker "seepage prevention" capacity for the slope; under the same rainfall conditions, the internal hazardous water pressure will reach the critical threshold more quickly, correspondingly increasing the risk of instability. This embodiment uses the classic SCS-CN (Soil and Water Conservation Bureau Curve Value) model in the field of hydrology for calculation. First, through remote sensing image interpretation, geological surveys, and soil testing, the land use type (such as vegetation cover) and hydrological soil grouping (i.e., soil permeability) of the monitoring zone are comprehensively evaluated. Then, a comprehensive empirical parameter, the curve value (CN), is determined for the zone from a standardized SCS-CN lookup table. Finally, the total rainfall P for a specific rainfall event is accurately obtained using rain gauges deployed in the monitoring area.

[0044] Substituting the obtained CN value into the formula S=(25400 / CN)-254, the potential maximum water storage capacity S of the region is calculated. Then, substituting the total rainfall P and the calculated S value into the runoff calculation formula Q=(P-0.2×S)² / (P+0.8×S), the total surface runoff Q generated by this rainfall event is estimated. According to the water balance principle, the total infiltration F equals the total rainfall P minus the surface runoff Q. The regional rainfall infiltration rate is the ratio of total infiltration to total rainfall, calculated using the formula infiltration rate=F / P=(PQ) / P. This ratio serves as the final value of the regional rainfall infiltration rate.

[0045] 1.2.2 The structure dataset is used to focus on characterizing the current geometric shape and physical field structure features of the slope. Examples of the data record fields are shown in Table 2 below: Table 2: Examples of data records in a structured dataset

[0046] The physics dataset contains refined parameters obtained from deep sensing arrays that directly reflect the internal state of the soil and rock mass.

[0047] The intelligent assessment module receives three datasets processed by the multi-dimensional state perception module and performs multi-level, multi-dimensional assessments through its internal response assessment unit, state assessment unit, damage assessment unit, and risk management unit. The response assessment unit calculates the response sensitivity index (RSI); this unit aims to assess the sensitivity of the monitoring zone to external disturbances. Its calculation process strictly follows steps S11 to S13, which are described in detail in this embodiment.

[0048] The calculation process of the response sensitivity index RSI is as follows: S11. Extract parameters characterizing external disturbances from the macro dataset, including far-field seismic wave amplification factor, tidal load strain response coefficient, atmospheric pressure load strain response coefficient and regional rainfall infiltration rate. After normalizing the above parameters, the weighted sum is obtained to obtain the comprehensive quantitative value E of external disturbance. S12. Extract parameters characterizing the slope response from the macro dataset, including the three-dimensional displacement field of the ground surface, the phase change rate of synthetic aperture radar interferometry, and the gradient of the ground surface temperature field. After normalizing the above parameters, perform weighted summation to obtain the comprehensive quantitative value M of the slope response. S13. Divide the comprehensive quantitative value of slope response M by the comprehensive quantitative value of external disturbance E to obtain the initial sensitivity value. Then, combine the regional equivalent damping ratio and the surface vegetation cover index to correct the initial sensitivity value, and finally obtain the response sensitivity index RSI.

[0049] S11. Calculation of the comprehensive quantitative value E of external disturbances: Parameter Extraction: Four external disturbance parameters were extracted from the macroscopic dataset: far-field seismic wave amplification factor Pext,1, tidal load strain response factor Pext,2, atmospheric pressure load strain response factor Pext,3, and regional rainfall infiltration rate Pext,4. For each parameter, based on its historical data range [min i ,max iMax-min normalization is performed: The preset external disturbance weight vector Wext={0.2,0.1,0.1,0.6} is applied to calculate E. This weight vector is calibrated using the following joint optimization method, where rainfall infiltration rate has the highest weight (0.6), because for the slope type in this region, rainfall is the most significant inducing factor for instability. E=0.2×N(Pext,1)+0.1×N(Pext,2)+0.1×N(Pext,3)+0.6×N(Pext,4); N(Pext,1), N(Pext,2), N(Pext,3), and N(Pext,4) are the normalized far-field seismic wave amplification factor, tidal load strain response factor, atmospheric pressure load strain response factor, and regional rainfall infiltration rate.

[0050] S12. Calculation of the comprehensive quantitative value M of slope response: Parameter Extraction: Three slope response parameters are extracted from the macro dataset: the magnitude of the three-dimensional surface displacement field (Presp,1), the absolute value of the phase change rate of the synthetic aperture radar interferometry (Presp,2), and the surface temperature gradient (Presp,3). For each parameter, max-min normalization is performed based on its historical data range [mini,maxi]. A preset slope response weight vector Wresp={0.5,0.4,0.1} is applied to calculate M. Displacement and deformation rate have the highest weights because they are the most direct manifestations of the slope response. M=0.5×N(Presp,1)+0.4×N(Presp,2)+0.1×N(Presp,3); Where N(Presp,1), N(Presp,2) and N(Presp,3) are the normalized modulus of the three-dimensional displacement field of the Earth's surface, the absolute value of the phase change rate of the synthetic aperture radar interferometry, and the gradient of the Earth's surface temperature field, respectively. S13. Correction of initial sensitivity value and generation of RSI, calculation of initial sensitivity: RSI initial =M / (E+1×10 -6 ), of which 1×10 -6 This is to avoid the smallest positive constant with a denominator of zero. The regional equivalent damping ratio (Deq) and the vegetation cover index (VCI) are extracted from the macro dataset. The final RSI is obtained by applying a correction formula. The corrected weighting coefficients are wd=0.3 and wv=0.2. RSI = RSI initial ×(1-0.3×Deq)×(1-0.2×VCI); The physical meaning of this correction is that a higher damping ratio (Deq) means that the slope has a stronger ability to dissipate energy and will respond more slowly; higher vegetation cover (VCI) can enhance slope stability through root stabilization and rainwater interception. Both will reduce the slope's sensitivity to the same external disturbance.

[0051] The calculation process for the Field State Shift Index (FSDI) is as follows: S21. Based on the structural dataset of historical stable periods, a reference benchmark model is established for each monitoring zone. The model includes the fractal dimension of the reference fracture network, the development density of the reference gully, the spatial distribution map of the reference self-electric field, and the reference apparent resistivity tomography model. S22. Compare the currently obtained fractal dimension of the fracture network and gully development density with the corresponding reference values ​​in the reference benchmark model, calculate their relative rate of change, and obtain the deviation of the structural morphology. S23. Perform spatial difference operation on the currently acquired self-electric field spatial distribution map and apparent resistivity tomography model with the corresponding reference map and model in the reference benchmark model, calculate its spatial inconsistency norm, and obtain the deviation of the physical field morphology. The calculation process for the Field State Shift Index (FSDI) is as follows: S21. Based on the structural dataset of historical stable periods, a reference benchmark model is established for each monitoring zone. The model includes the fractal dimension of the reference fracture network, the development density of the reference gully, the spatial distribution map of the reference self-electric field, and the reference apparent resistivity tomography model. S22. Compare the currently obtained fractal dimension of the fracture network and gully development density with the corresponding reference values ​​in the reference benchmark model, calculate their relative rate of change, and obtain the deviation of the structural morphology. S23. Perform spatial difference operation on the currently acquired self-electric field spatial distribution map and apparent resistivity tomography model with the corresponding reference map and model in the reference benchmark model, calculate its spatial inconsistency norm, and obtain the deviation of the physical field morphology. S24. The structural morphology deviation and the physical field morphology deviation are weighted and fused to obtain the field state deviation index FSDI.

[0052] Based on the structure dataset, the state assessment unit evaluates the degree of deviation of the evolution of the physical field within each monitoring zone and generates the corresponding field state offset index FSDI. The state assessment unit is used to calculate the field state offset index, denoted as FSDI, to quantify the degree of deviation of the current slope state from its "healthy" baseline.

[0053] S21. The system selected monitoring data from the initial five years of mine construction, during which there were no significant deformations or geological disaster records. Based on the historical stable period structural dataset, a high-precision reference benchmark model BM was established for the i-th monitoring zone of the highway network. ref This model is the "digital fingerprint" of the partition in a healthy state, and specifically comprises four core components: the reference fracture network fractal dimension Df. ref Reference gully development density ρg ref Reference to the spatial distribution diagram of the electric field SP ref(x,y) and reference apparent resistivity tomography model ρ ref (x,z).

[0054] S22. Extract the fractal dimension Df of the fracture network for the i-th monitoring partition from the current timestamp structure dataset. cur and gully development density ρg cur Subsequently, the relative rates of change of these two macroscopic geometric parameters relative to the reference baseline were calculated, and the rate of change of the fractal dimension of the fracture network in the i-th monitoring zone was obtained using the following formula. and the rate of change in gully development density ;

[0055]

[0056] Rate of change of the fractal dimension of the fracture network for the i-th monitoring zone and the rate of change in gully development density Weighted summation yields the structural deviation Dstruc of the i-th monitoring partition. i : ; In the formula, The weights represent the relative rate of change of the fractal dimension of the fractured network. The weights representing the relative rate of change in gully development density, and For slopes with soil, colluvial deposits, or thick weathered layers, the preferred value should be set. , For steep slopes made of hard rock, the preferred value is set. , ; S23. Extract the spatial distribution map of the electric field of the i-th monitoring zone from the physical field dataset of the current timestamp. cur (x,y) and apparent resistivity tomography model ρ cur (x,z). Spatial difference operations are performed between these two internal physical field models and the corresponding reference models in the reference baseline model BMref to calculate the spatial inconsistency norm of the self-electric field in the i-th monitoring partition. and the spatial inconsistency norm of apparent resistivity ;

[0057]

[0058] The spatial inconsistency norm of the self-electric field for the i-th monitoring zone and the spatial inconsistency norm of apparent resistivity By performing a weighted summation, the physical field morphology deviation Dphys of the i-th monitoring zone is obtained. i : ; In the formula, The L2 norm operation is used to calculate the overall difference between two spatial models (matrices); β1 and β2 are the weights of the spatial inconsistency norms of the self-electric field and apparent resistivity, respectively, both constants, and β1 + β2 = 1. The physical field morphology deviation (Dphysi) aims to quantify the degree of degradation of invisible physical properties within the slope. The spatial inconsistency norm, calculated by point-by-point (or pixel-by-pixel) differencing of the two models, can accurately capture changes in internal electrical structures (such as water-bearing channels and fractured areas). The larger the deviation, the more drastic the material transport or structural changes within the slope, and the lower the internal stability. In rock slopes or gravelly soil slopes with well-developed fissures and high permeability, capturing the dynamic changes in the seepage path (i.e., where the water is flowing) is more crucial than understanding the overall water content distribution. β1=0.6~0.7, β2=0.4~0.3; for cohesive soil slopes or homogeneous soil bodies with poor permeability, water seepage is relatively slow, and the overall water content change is more important. Resistivity can better reflect the range and development of the saturation zone. β1=0.4~0.5, β2=0.6~0.5.

[0059] S24. Based on S22-S23, the structural morphology deviation Dstruc i Deviation from physical field shape By performing weighted fusion, the field state offset index (FSDI) of this monitoring zone of the highway network is calculated: ; In the formula, The weight representing the deviation of the structural form The weights represent the deviation of the physical field shape, and the sum of the weights is 1.

[0060] The calculation process for the Structural Damage Index (SDI) is as follows: S31. The depth characteristic parameters in the physical field dataset are divided into micro-fracture damage parameters and fluid-structure interaction degradation parameters according to their physical meaning. Among them, micro-fracture damage parameters include Kaiser effect breakdown ratio, relative change rate of wake interference wave velocity and inelastic energy dissipation rate; fluid-structure interaction degradation parameters include complex resistivity phase angle and time-series variation rate of dissolved radon gas concentration. S32. Normalize each type of parameter and calculate its first derivative over time to characterize its dynamic rate of change. S33. The normalized values ​​and their rates of change of the micro-fracture damage parameters are weighted and summed to obtain the micro-fracture damage component; the normalized values ​​and their rates of change of the fluid-structure interaction degradation parameters are weighted and summed to obtain the fluid-structure interaction degradation component. S34. Based on the soil and rock type of the monitoring zone, assign different weight coefficients to the micro-fracture damage component and the fluid-structure interaction deterioration component, and then sum them to obtain the structural damage index SDI.

[0061] The damage assessment unit analyzes the degree of damage accumulation in the microstructure of soil and rock mass based on the physical field dataset, and generates the corresponding Structural Damage Index (SDI). The calculation process for the SDI is as follows: S31. Based on the physical damage mechanism they reflect, the depth feature parameters in the physical field dataset are divided into two categories: the micro-fracture damage parameter set Pmf and the fluid-structure interaction degradation parameter set Pfsc. The microfracture damage parameter set Pmf specifically reflects the initiation, propagation, and connection of micro-fractures within the soil and rock mass under stress. Pmf={P1,P2,P3}={Kaiser effect breakdown ratio KB, relative change rate of wake interference wave velocity Δv / v, inelastic energy dissipation rate Ene}; The fluid-structure interaction degradation parameter set Pfsc specifically reflects the softening, dissolution, and other degradation effects of groundwater activity on the structural strength of soil and rock. Pfsc = {Q1, Q2} = {complex resistivity phase angle CR, time-series variability of dissolved radon concentration (CRn)}; S32, for each parameter defined in S31 (whether it is P) j Still Q k At the current timestamp t, the following two preprocessing steps are performed: The first step is normalization: to eliminate the differences in the dimensions and numerical ranges of each parameter, the maximum-minimum method is used to map them to the [0,1] interval; the second step is dynamic change rate calculation: to characterize the dynamic evolution trend of the parameter, the first-order difference is used to calculate its change rate on the time series. S33. Weighted summation of the normalized values ​​and their rates of change of the two types of parameters is performed to obtain the micro-fracture damage component of the i-th monitoring zone at time t. and fluid-structure interaction degradation components .

[0062]

[0063]

[0064] in, and These are the normalized value and rate of change calculated from the j-th parameter (i.e., KB, Δv / v, Ene) in the Pmf set. and These are the weighting coefficients corresponding to their normalized values ​​and rates of change, respectively, and are both constants. and These are the normalized value and rate of change of the k-th parameter (i.e., CR, CRn) in the Pfsc set, respectively. and These are the weighting coefficients corresponding to their normalized values ​​and rates of change, respectively, and are both constants.

[0065] S34. Based on the specific soil and rock type of the monitoring zone, the micro-fracture damage component of the i-th monitoring zone at time t. and fluid-structure interaction degradation components Different dominant weight coefficients are assigned, and a final weighted sum is performed to obtain the structural damage index (SDI) for this partition:

[0066] In the formula, and These are the final weighting coefficients for the microfracture damage component and the fluid-structure interaction degradation component, respectively. These two coefficients are dynamically set according to the engineering geological conditions and satisfy b1+b2=1.

[0067] Specifically, the allocation of weighting coefficients b1 and b2 is crucial to this model, reflecting a professional assessment of the dominant instability factors under different geological environments. For hard rock slopes, the instability mode is often dominated by structural shear fracturing, with micro-fracture damage being the core precursor; therefore, a higher b1 can be set (b1=0.7, b2=0.3). For soil or soft rock slopes, softening and seepage pressure from water are the main causes of instability, and the fluid-structure interaction degradation effect is more critical; therefore, a higher b2 can be set (b1=0.4, b2=0.4). b 2=0.6). The final SDI is a dimensionless comprehensive index. The higher the value, the more severe the cumulative damage inside the slope, the worse the structural integrity, and the closer it is to the critical state of instability and failure.

[0068] The risk management unit has a fixed range of stability threshold interval SY and warning threshold interval WY. Combined with the response sensitivity index RSI, field state deviation index FSDI and structural damage index SDI, it determines the stability level, warning level and instability probability, outputs the assessment results and triggers corresponding response measures.

[0069] The specific process for the risk management unit to judge the stability level is as follows: Denote the upper limit of the stability threshold interval as maxSY and the lower limit as minSY; By performing a weighted sum of the response sensitivity index RSI, the field state deviation index FSDI, and the structural damage index SDI, the comprehensive stability index CSI is obtained;

[0070] Among them, s1, s2, and s3 are the weights of the response sensitivity index RSI, the field state deviation index FSDI, and the structural damage index SDI respectively, and the sum of the weights is 1. The preferred values are set as: s1 = 0.2, s2 = 0.3, s3 = 0.5; SDI is the "precursor" and "fundamental" index of instability. It directly measures irreversible damages such as micro - fractures, stress concentration, and fluid - solid coupling deterioration inside the rock and soil mass through deep - seated sensors. The accumulation of these micro - level damages is the direct physical precursor of macroscopic instability. Regardless of how the external environment changes (RSI), regardless of how much the macroscopic shape of the slope deviates from FSDI, ultimately, what determines instability is whether the internal structure can continue to bear the load. Therefore, SDI is given the highest weight because it most directly and profoundly reflects the proximity of the slope to the critical failure state. A high SDI value means that the inside of the slope is "terminally ill", and even if it seems calm on the surface, the risk of instability is high.

[0071] If CSI < minSY, the stability level is "stable", and the response measure is to maintain the conventional monitoring frequency; If minSY ≤ CSI ≤ maxSY, the stability level is "basically stable", and the response measure is to automatically increase the data acquisition frequency of the deep - seated sensor array by 25% - 50%, and conduct a data correlation review on the most significantly changing parameters; If CSI > maxSY, the stability level is "sub - stable", and the response measure is to initiate an encrypted observation mode, increase the data acquisition frequency by 100%, and notify the management to pay attention.

[0072] The specific process for the risk management unit to judge the warning level is as follows: Denote the upper limit of the warning threshold interval as maxWY and the lower limit as minWY; If the comprehensive stability index CSI < minWY, the warning level is "no warning", and no additional response strategy is triggered; If minWY ≤ CSI ≤ maxWY, the warning level is "first - level warning", and the first - level response strategy is triggered; The first - level response strategy includes: Adopting systematic bolt support with a spacing of 4.4 m - 4.8 m, and supplementing it with pressure grouting at a standard of 15 kg - 16 kg per square meter for repair; And adding drainage facilities at a density of 0.9 - 1.0 per 100 square meters to improve the internal environment; If CSI>maxWY, the warning level is "Level 2 Warning", triggering the Level 2 response strategy. The Level 2 response strategy includes: using systematic anchor bolts with a spacing of 3.1 meters to 3.5 meters for support, and supplementing it with pressure grouting at a standard of 22 kg to 23 kg / m² for repair; and adding drainage facilities at a density of 1.3 to 1.5 per 100 square meters to improve the internal environment.

[0073] Figure 1 The slope profile on the left and its three-layer three-dimensional monitoring network correspond to the "multi-dimensional state perception module". Specifically, the orbital satellite image represents the "macro-level remote sensing system", the ground-based three-dimensional laser scanner image represents the "near-field monitoring system", and the borehole sensor image embedded in the slope represents the "deep sensing array"; they work together to acquire macro-level surface deformation data, fine slope structure data, and physical field data of the soil and rock mass. Figure 1 The technical roadmap on the right corresponds to the internal processing logic of the "intelligent assessment module." The "multi-dimensional state perception" diagram marks the starting point for data input; the "multi-dimensional index calculation" diagram represents the process by which the response assessment unit, state assessment unit, and damage assessment unit generate the response sensitivity index (RSI), field state offset index (FSDI), and structural damage index (SDI), respectively; the "comprehensive risk assessment" diagram represents the process by which the risk management unit weights and fuses the above three indices and compares them with a threshold to determine the stability level; finally, the "tiered early warning and response" diagram represents the final stage where the system outputs the assessment results and triggers corresponding response measures, fully embodying the technical solution of this invention.

[0074] In this embodiment, the present invention provides a dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence. Its core technical principle is that it abandons the passive early warning approach of relying on a single, lagging macroscopic deformation index in traditional slope monitoring, and constructs a "three-dimensional" dynamic assessment framework that goes from the surface to the inside and from the result to the cause, from "external disturbance response", "macroscopic structural evolution" to "internal microscopic damage".

[0075] This principle first utilizes a multi-dimensional state sensing module to achieve comprehensive and in-depth acquisition of slope conditions. It not only covers traditional surface deformation but also innovatively introduces macroscopic response parameters that reflect the slope's sensitivity to external loads (such as earthquakes, tides, and rainfall) as well as deep sensing data that characterizes changes in the internal physical fields (such as electric fields and resistivity) of the slope.

[0076] Building upon this foundation, the intelligent assessment module decouples and deeply analyzes the data using three core indices (RSI, FSDI, SDI): The RSI (Responsive Sensitivity Index) establishes a quantitative relationship between "disturbance and response." It no longer views deformation in isolation but compares it with the intensity of external disturbances in a normalized manner, thereby assessing the slope's "vulnerability" or "sensitive nature," and accurately identifying high-risk slopes highly sensitive to specific disturbances (such as rainfall).

[0077] The Field Misalignment Index (FSDI) is based on deviation quantification using a "benchmark comparison." It establishes a multi-parameter digital benchmark model for slopes in a "healthy" state, quantifying the long-term, macroscopic deterioration trend of the slope by comparing the current structure and physical field model with the benchmark, and capturing slow structural changes.

[0078] The Structural Damage Index (SDI) works by delving into the interior of the rock and soil mass to quantify and accumulate "microscopic damage." It analyzes physical parameters that directly reflect the micro-fracture and fluid-structure interaction deterioration process of the rock mass, such as acoustic emission, wave velocity changes, and electrochemical properties. This allows for the direct measurement of pre-instability internal damage, solving the "early warning blind zone" problem before macroscopic deformation occurs. The system intelligently weights and fuses these three indices, which characterize slope conditions from different dimensions and depths, to form the Comprehensive Stability Index (CSI). Based on this, it accurately judges stability and early warning levels, driving a graded and quantified closed-loop response strategy, thus forming a complete technical logic of "multi-dimensional perception - in-depth assessment - graded early warning - closed-loop response."

[0079] Compared with existing technologies, this invention, through the aforementioned technical principles, achieves a transformation in early warning modes from "lagging" to "precursor-like." Traditional methods rely on macroscopic deformation, which is the final manifestation of accumulated internal damage, resulting in poor timeliness of early warning. This invention, through the Structural Damage Index (SDI), can directly capture early, weak precursory information such as micro-fractures in the rock mass before instability, significantly advancing the early warning threshold, gaining valuable time for disaster prevention and mitigation, and effectively avoiding major losses caused by sudden landslides.

[0080] Traditional early warning logic often relies on simple displacement or rate thresholds, making it susceptible to false alarms or missed alarms due to anomalies in single sensors. This invention, by integrating three physically distinct and independent indices—RSI, FSDI, and SDI—constructs a more robust and comprehensive Comprehensive Stability Index (CSI), effectively eliminating interference from single data sources and improving the accuracy and reliability of assessment results. This invention not only determines whether a slope is "dangerous," but also reveals the dominant factors leading to increased risk by analyzing the weights and trends of the three sub-indices. For example, a sharp increase in RSI may indicate that rainfall is the primary threat, while a continuous increase in SDI suggests that internal structural damage has reached a critical stage. This diagnostic capability provides a more targeted scientific basis for subsequent engineering remediation. This system directly links accurate risk assessment results with clear and quantifiable graded response strategies, forming an automated closed-loop management system of "monitoring-assessment-early warning-response." It not only increases the frequency of data collection but also automatically triggers specific engineering intervention recommendations, reducing the delay and uncertainty of human intervention and significantly improving the intelligence level of slope safety management and emergency response efficiency.

[0081] It should be noted that all calculation formulas in this application employ regression analysis, including but not limited to machine learning algorithms, to deeply analyze the collected parameters and identify their natural trends and interrelationships. Specialized software, such as Python's Scikit-learn library or the R language, is used to automatically generate mathematical models that match the data. Then, cross-validation and other methods are used to objectively evaluate the model performance, and continuous feedback and optimization are combined to ensure that the created formulas truly reflect the inherent laws of the data, thereby guaranteeing their effectiveness and accuracy. In all calculation formulas in this application, the parameters in each formula undergo dimensionless processing within a consistent range to ensure that different physical quantities are compared on the same scale; dimensionless processing techniques include, but are not limited to, min-max-normalization and Z-score standardization. The algorithm of this invention is implemented as a Python script. Before executing the core logic, the program first executes a data loading module (e.g., using the widely used pandas library in Python) configured to read the aforementioned spreadsheet file and load its contents into the program's working memory (e.g., a DataFrame data structure). Subsequent algorithm steps will directly query and retrieve the required configuration parameters from this in-memory data structure.

[0082] It should be noted that 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence, characterized in that, Includes a multi-dimensional state perception module and an intelligent evaluation module: The multidimensional state perception module is configured to acquire surface macro deformation data, slope fine structure data, and soil internal physical field data of the monitoring area by connecting a macro remote sensing system, a near-field monitoring system, and a deep sensing array, and classify them into macro datasets, structural datasets, and physical field datasets. The intelligent assessment module includes a response assessment unit, a state assessment unit, a damage assessment unit, and a risk management unit. The response assessment unit evaluates the sensitivity of each monitoring zone to external disturbances based on the macro dataset and generates a corresponding response sensitivity index. The state assessment unit evaluates the degree of deviation of the evolution of the physical field within each monitoring zone based on the structural dataset and generates a corresponding field state offset index. The damage assessment unit analyzes the degree of damage accumulation in the microstructure of the soil and rock mass based on the physical field dataset and generates a corresponding structural damage index. The risk management unit sets a fixed range of stability threshold intervals and warning threshold intervals, and combines the response sensitivity index, field state offset index, and structural damage index to determine the stability level, warning level, and instability probability, outputs the assessment results, and triggers corresponding response measures.

2. The slope instability probability dynamic assessment system based on multi-source data and artificial intelligence according to claim 1, characterized in that: The macroscopic dataset includes the three-dimensional displacement field of the land surface, the phase change rate of synthetic aperture radar interferometry, the gradient of the land surface temperature field, the regional equivalent damping ratio, the far-field seismic wave amplification coefficient, the tidal load strain response coefficient, the atmospheric pressure load strain response coefficient, the regional rainfall infiltration rate, and the land surface vegetation cover index for each partition of the monitoring area.

3. The slope instability probability dynamic assessment system based on multi-source data and artificial intelligence according to claim 1, characterized in that: The structured dataset includes a digital elevation model, a high-precision three-dimensional point cloud model, the orientation of the dominant structural surface, the fractal dimension of the fracture network, the surface roughness of the slope, the density of gully development, the distribution map of seepage outflow points, the spatial distribution map of the self-electric field, and the apparent resistivity tomography model for each monitoring zone.

4. The slope instability probability dynamic assessment system based on multi-source data and artificial intelligence according to claim 1, characterized in that: The physical field dataset includes at least one of the following depth characteristic parameters within each monitoring zone: the self-potential gradient tensor measured by the electrode array, the complex resistivity phase angle measured by the multi-frequency induced polarization method, the temporal variability of dissolved radon concentration measured by the hydrological monitoring well, the relative change rate of the wake interference wave velocity calculated by repeated source monitoring, the Kaiser effect breakdown ratio calculated by the acoustic emission monitoring system, the inelastic energy dissipation rate calculated by the distributed fiber optic strain sensor, the microgravity anomaly gradient obtained by time-series gravity measurement, the rock mass dielectric constant heterogeneity index obtained by ground penetrating radar inversion, the T2 spectrum distribution morphology parameter measured by borehole nuclear magnetic resonance, and the surface mineral characteristic absorption depth ratio obtained by hyperspectral imaging analysis.

5. The dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence according to claim 1, characterized in that: The calculation process for the response sensitivity index is as follows: S11. Extract parameters characterizing external disturbances from the macro dataset, including far-field seismic wave amplification coefficient, tidal load strain response coefficient, atmospheric pressure load strain response coefficient, and regional rainfall infiltration rate. After normalizing the above parameters, perform weighted summation to obtain the comprehensive quantitative value E of external disturbance. S12. Extract parameters characterizing the slope response from the macro dataset, including the three-dimensional displacement field of the ground surface, the phase change rate of synthetic aperture radar interferometry, and the gradient of the ground surface temperature field. After normalizing the above parameters, perform weighted summation to obtain the comprehensive quantitative value of the slope response. S13. Divide the comprehensive quantitative value of the slope response by the comprehensive quantitative value of the external disturbance to obtain the initial sensitivity value. Then, combine the regional equivalent damping ratio and the surface vegetation cover index to correct the initial sensitivity value, and finally obtain the response sensitivity index.

6. The dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence according to claim 1, characterized in that: The calculation process for the field state offset index is as follows: S21. Based on the structural dataset of historical stable periods, a reference benchmark model is established for each monitoring zone. The model includes the fractal dimension of the reference fracture network, the development density of the reference gully, the spatial distribution map of the reference self-electric field, and the reference apparent resistivity tomography model. S22. Compare the currently acquired fractal dimension of the fracture network and gully development density with the corresponding reference values ​​in the reference benchmark model, calculate their relative rate of change, and obtain the structural morphology deviation. S23. Perform spatial difference operation on the currently acquired self-electric field spatial distribution map and apparent resistivity tomography model with the corresponding reference map and model in the reference benchmark model, calculate its spatial inconsistency norm, and obtain the physical field morphology deviation. S24. The structural morphological deviation and the physical field morphological deviation are weighted and fused to obtain the field state offset index.

7. The slope instability probability dynamic assessment system based on multi-source data and artificial intelligence according to claim 1, characterized in that: The calculation process for the structural damage index is as follows: S31. The depth characteristic parameters in the physical field dataset are divided into micro-fracture damage parameters and fluid-structure interaction degradation parameters according to their physical meaning. Among them, micro-fracture damage parameters include Kaiser effect breakdown ratio, relative change rate of wake interference wave velocity and inelastic energy dissipation rate; fluid-structure interaction degradation parameters include complex resistivity phase angle and time-series variation rate of dissolved radon gas concentration. S32. Normalize each type of parameter and calculate its first derivative over time to characterize its dynamic rate of change. S33. The normalized values ​​and their rates of change of the micro-fracture damage parameters are weighted and summed to obtain the micro-fracture damage component; the normalized values ​​and their rates of change of the fluid-structure interaction degradation parameters are weighted and summed to obtain the fluid-structure interaction degradation component. S34. Based on the soil and rock type of the monitoring zone, assign different weight coefficients to the micro-fracture damage component and the fluid-structure interaction deterioration component, and then sum them to obtain the structural damage index.

8. The slope instability probability dynamic assessment system based on multi-source data and artificial intelligence according to claim 1, characterized in that: The specific process for the risk management unit to determine the stability level is as follows: the upper limit of the stability threshold range is denoted as maxSY, and the lower limit is denoted as minSY; the comprehensive stability index CSI is obtained by weighted summation of the response sensitivity index RSI, the field state offset index FSDI, and the structural damage index SDI. If CSI < minSY, the stability level is "stable", and the response measure is to maintain the regular monitoring frequency; if minSY ≤ CSI ≤ maxSY, the stability level is "basically stable", and the response measure is to automatically increase the data acquisition frequency of the deep sensing array by 25% - 50%, and conduct data correlation review on the most significant parameters; if CSI > maxSY, the stability level is "less stable", and the response measure is to initiate the encrypted observation mode and notify the management to pay attention.

9. The dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence according to claim 1, characterized in that: The specific process for the risk management unit to judge the warning level is as follows: Denote the upper limit of the warning threshold interval as maxWY and the lower limit as minWY; if the comprehensive stability index CSI < minWY, the warning level is "no warning", and no additional response strategy is triggered; if minWY ≤ CSI ≤ maxWY, the warning level is "first-level warning", and the first-level response strategy is triggered; if CSI > maxWY, the warning level is "second-level warning", and the second-level response strategy is triggered.

10. The dynamic assessment system for slope instability probability based on multi-source data and artificial intelligence according to claim 9, characterized in that: The first-level response strategy includes: adopting systematic bolt support with a spacing of 4.4 m - 4.8 m, and supplementing with pressure grouting at a standard of 15 kg - 16 kg per square meter for repair; and adding drainage facilities at a density of 0.9 - 1.0 per 100 square meters to improve the internal environment; The second-level response strategy includes: adopting systematic bolt support with a spacing of 3.1 m - 3.5 m, and supplementing with pressure grouting at a standard of 22 kg - 23 kg per square meter for repair; and adding drainage facilities at a density of 1.3 - 1.5 per 100 square meters to improve the internal environment.

Images