Open pit coal mine slope deformation monitoring method and system based on multi-source data

Abstract

The invention discloses an open pit coal mine slope deformation monitoring method and system based on multi-source data, and belongs to the technical field of mining safety analysis. According to the method, an improved slope stability control equation fusing mining disturbance factors is constructed, and a physical information neural network is established based on the improved slope stability control equation. A physical information neural network is driven by air-sky-ground-internal multi-source heterogeneous data, real-time synchronous solution of a slope global displacement field, a stress field, a dynamic stability coefficient, a potential sliding surface form and a damage probability curve is realized, and an early warning level is automatically generated according to the multi-source indexes. And outputting a disposal decision accurately matched with the spatial form of the sliding surface. According to the method, deep fusion of monitoring data and a physical mechanism is realized, slope stability evaluation is promoted from static analysis to dynamic prediction, and the accuracy, foresight and engineering applicability of early warning are enhanced.

Description

Technical Field

[0001] This invention relates to the field of mining safety analysis technology, and more specifically to a method and system for monitoring slope deformation in open-pit coal mines based on multi-source data. Background Technology

[0002] Currently, open-pit coal mine slope deformation monitoring mainly relies on a multi-faceted technological system. For example, it utilizes equipment such as the Global Navigation Satellite System, ground-based synthetic aperture radar (SMR), and surveying robots to continuously observe surface displacement. Simultaneously, it combines this with methods like inclinometers and microseismic monitoring to obtain internal slope deformation information, thus initially realizing the transition from single-point monitoring to multi-dimensional sensing from the sky and ground. Furthermore, numerical simulation techniques such as the limit equilibrium method and the finite element method are widely used in the mechanical analysis of slope stability based on geological exploration data.

[0003] However, existing technologies still have significant limitations. First, monitoring technologies and stability analysis models are usually independent, creating a disconnect between data acquisition and mechanism analysis. Monitoring data, such as displacement and stress, are often distributed in point or area patterns, making it difficult to directly and in real-time use to invert or correct the parameters of refined constitutive models describing the progressive failure process of slopes. This leads to a continuous accumulation of discrepancies between numerical model predictions and actual conditions. Second, traditional numerical analysis methods are computationally expensive and cannot meet the urgent need for minute-level or even second-level online real-time assessment and early warning of slope conditions. Furthermore, while purely data-driven machine learning early warning models have fast response times, their physical interpretability is poor. When monitoring data is scarce or when encountering unprecedented conditions, such as extreme rainfall or special mining disturbances, the reliability of their extrapolated predictions is difficult to guarantee.

[0004] Therefore, how to propose a method and system for monitoring slope deformation in open-pit coal mines based on multi-source data, realize online real-time analysis of slope stress and deformation state and intelligent prediction of failure modes, and improve the accuracy, real-time and reliability of monitoring results is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for monitoring the deformation of open-pit coal mine slopes based on multi-source data. By introducing a physical information neural network (PINN) enhanced by mining disturbance, a dynamic digital twin of the slope is established to realize real-time analysis of stability status, prediction of future evolution, and precise linkage and control with the production process.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: On the one hand, this invention proposes a method for monitoring slope deformation in open-pit coal mines based on multi-source data, comprising the following steps: S1. Constructing a geological model of open-pit coal mine slopes; S2. Real-time acquisition of multi-source heterogeneous monitoring data from air, space, ground, and indoor environments in open-pit coal mines; S3. Construct an improved slope stability control equation that integrates mining disturbance factors and the constitutive relationship of rock mass time-dependent damage, and establish a physical information neural network based on the improved slope stability control equation; S4. Input the slope geological model parameters, multi-source heterogeneous monitoring data and mining plan into the physical information neural network to solve in real time the displacement field, stress field, dynamic stability coefficient and potential sliding surface spatial morphology of the whole domain, and output the failure probability time series curve. The displacement field, stress field and multi-source heterogeneous monitoring data are spatiotemporally registered, and the slope geological model parameters are dynamically corrected using an ensemble Kalman filter algorithm. S5. Based on the dynamic stability coefficient, the time series curve of the failure probability and the potential sliding surface spatial morphology, output the early warning level and trigger the corresponding early warning response mechanism, and generate a disposal decision according to the early warning level and the potential sliding surface spatial morphology.

[0007] S6. Collect on-site execution feedback data of the disposal decision to continuously optimize the model parameters of the physical information neural network.

[0008] Preferably, in S3, the improved slope stability control equation is as follows: ; ; ; In the formula, Rock mass shear strength; Normal stress; This represents the initial cohesion of the rock mass. It is the cohesion degradation function of the rock mass itself as it evolves with equivalent plastic strain; This represents the initial cohesion of the rock mass. Let the mining disturbance damage function be... , Let be the weighting coefficient of the i-th mining disturbance component. Let i be the i-th mining disturbance component. The mining disturbance component includes the blasting vibration damage component, the excavation unloading damage component, and the groundwater damage component. The initial internal friction angle of the rock mass; The internal friction angle degradation function is the evolution of the rock mass itself with equivalent plastic strain. This is the damage sensitivity coefficient of the internal friction angle.

[0009] Preferably, S4 includes: By using forward propagation calculations through a physical information neural network and constraining the output results with the improved slope stability control equation, displacement field distribution data and stress field distribution data of the entire slope area are directly generated. Based on global stress field data, an optimization search algorithm is used to iteratively calculate the stability coefficients of potential sliding surfaces at different spatial locations, thereby determining the most dangerous potential sliding surface and its corresponding dynamic stability coefficient that makes the stability coefficient take the global minimum value. Input working condition parameters, simulate the time evolution process of slope mechanical state under corresponding working conditions through physical information neural network, and output the time series curve of slope failure probability.

[0010] Preferably, the dynamic stability coefficient The calculation formula is as follows: ; In the formula, For potential sliding surfaces The actual shear stress on the surface.

[0011] Preferably, S5 includes: The rate of slope stability deterioration is judged by the rate of change of dynamic stability coefficient over time, the intensity of risk growth is quantified by the slope of the failure probability time series curve, and the degree of failure development is characterized by the expansion rate of the spatial range of the potential sliding surface and the sliding rate. A multi-dimensional objective criterion system is constructed to output the early warning level and trigger the corresponding early warning response mechanism. Based on the spatial morphology of the potential sliding surface and the mechanical equilibrium state of the slope, the direction of reinforcement, unloading or drainage treatment is matched with the mechanical characteristics of the sliding surface. At the same time, the treatment measures are determined according to the stability deterioration rate corresponding to the warning level.

[0012] On the other hand, the present invention also proposes an open-pit coal mine slope deformation monitoring system based on multi-source data, comprising: The geological model building module is used to build geological models of open-pit coal mine slopes; The data acquisition module is used to collect multi-source heterogeneous monitoring data from the air, space, ground, and interior of open-pit coal mines in real time. The network construction module is used to construct an improved slope stability control equation that integrates mining disturbance factors and the constitutive relationship of rock mass time-dependent damage, and to establish a physical information neural network based on the improved slope stability control equation. The stability solution module is used to input slope geological model parameters, multi-source heterogeneous monitoring data and mining plans into the physical information neural network, and solve in real time to obtain the displacement field, stress field, dynamic stability coefficient and potential sliding surface spatial morphology of the whole domain, while outputting the failure probability time series curve. The early warning output module is used to output an early warning level and trigger a corresponding early warning response mechanism based on the dynamic stability coefficient, the time series curve of the failure probability and the spatial morphology of the potential sliding surface, and to generate a disposal decision based on the early warning level and the potential sliding surface.

[0013] The geological model correction module is used to perform spatiotemporal registration of the global displacement field, stress field and the multi-source heterogeneous monitoring data, and to dynamically correct the slope geological model parameters using an ensemble Kalman filter algorithm.

[0014] The model optimization module is used to collect on-site execution feedback data of the disposal decision to continuously optimize the model parameters of the physical information neural network.

[0015] As can be seen from the above technical solution, compared with the prior art, this invention discloses a method and system for monitoring slope deformation in open-pit coal mines based on multi-source data. It utilizes heterogeneous monitoring data from multiple sources (air, space, ground, and interior) and employs a physical information neural network that integrates mining disturbance factors to solve the global mechanical state of the slope in real time. Simultaneously, it outputs dynamic stability coefficients, potential sliding surface morphology, and failure probability curves, and automatically generates early warnings and precise response decisions. This invention accurately quantifies the time-dependent damage effect of mining activities on slope stability, achieves pre-simulation prediction through deep integration of data and physical laws, establishes an automatic closed loop for early warning and production control, and improves the initiative, accuracy, and system engineering level of mine safety management. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 A flowchart of the method provided by the present invention; Figure 2 The system architecture diagram provided for this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] On the one hand, such as Figure 1As shown in the figure, this invention proposes a method for monitoring slope deformation in open-pit coal mines based on multi-source data, comprising the following steps: S1. Constructing a geological model of open-pit coal mine slopes.

[0020] Constructing a geological model of an open-pit coal mine slope requires comprehensive acquisition of basic data through systematic geological exploration. This includes: conducting on-site geological surveys of the entire slope area, combining drilling and pit exploration methods to clarify the slope's topographic features, including overall slope direction, height difference, and surface cover; systematically reviewing stratigraphic and lithological data to determine the lithological composition, particle size distribution, and cementation state of each stratum, and to identify the thickness distribution and spatial contact relationships of different strata; thoroughly investigating the development of geological structures, accurately identifying the attitude, scale, extension length, filling material, and mechanical properties of faults, joints, and fissures, and highlighting the distribution range, burial depth, and physical and mechanical parameters of weak interlayers; and collecting regional hydrogeological data to clarify the types of groundwater, water level changes, aquifer distribution, and permeability coefficients.

[0021] Based on the above-mentioned geological data, three-dimensional geological modeling technology was used for data integration and model construction. Through refined grid division and parameter assignment, the three-dimensional geological structure of the slope was fully restored. Key geological elements affecting stability, such as weak interlayers of rock mass structure, were fully incorporated to ensure that the model can truly reflect the actual geological conditions of the slope and provide an accurate geological framework for subsequent stability analysis and interpretation of monitoring data.

[0022] S2. Real-time acquisition of multi-source heterogeneous monitoring data from air, space, ground, and interior of open-pit coal mines.

[0023] Aerial monitoring relies on satellite remote sensing technology, selects satellite images of appropriate resolution, and regularly acquires large-scale topographic deformation data of slope areas to capture the overall trend of surface changes over long-term periods.

[0024] Low-altitude monitoring employs an unmanned aerial vehicle (UAV) aerial survey system equipped with high-definition imaging equipment and lidar sensors. It conducts flight monitoring at a set frequency to acquire detailed topographic data and local micro-deformation information of slope surfaces within the mid-to-low altitude range, thus compensating for the shortcomings of satellite remote sensing in detailed monitoring.

[0025] The surface monitoring system deploys equipment such as global navigation satellite system receivers, slope radar, and measurement robots. Monitoring points are set up in key areas and potentially dangerous sections of the slope to continuously collect data such as the three-dimensional coordinate displacement and vibration amplitude of the surface, enabling real-time dynamic tracking of surface deformation.

[0026] Internal slope monitoring involves drilling and installing equipment such as inclinometers, strain gauges, stress sensors, and piezometers to reach different depths of the slope and monitor the displacement, stress distribution, and groundwater level and pressure changes within the rock mass.

[0027] All monitoring equipment operates stably according to the preset sampling frequency, ensuring the synchronization and timeliness of various data acquisitions. The types of data collected cover displacement data, stress data, vibration data, groundwater data, topographic data, etc., forming a multi-dimensional and multi-level heterogeneous monitoring dataset, which provides comprehensive and reliable data support for the subsequent training of physical information neural networks and the solution of slope stability parameters.

[0028] S3. Construct an improved slope stability control equation that integrates mining disturbance factors and the constitutive relationship of rock mass time-dependent damage, and establish a physical information neural network based on the improved slope stability control equation.

[0029] The improved slope stability governing equations are as follows: ; Cohesion and internal friction angle Both are time t and equivalent plastic strain Evolutionary variables.

[0030] ; ; In the formula, Rock mass shear strength; Normal stress; It is the cohesion degradation function of the rock mass itself as it evolves with equivalent plastic strain; This represents the initial cohesion of the rock mass. The initial internal friction angle of the rock mass; The internal friction angle degradation function is the evolution of the rock mass itself with equivalent plastic strain. This is the damage sensitivity coefficient of the internal friction angle.

[0031] The strain softening function of the rock mass itself and It needs to be determined through indoor triaxial compression tests, especially cyclic loading and unloading tests. This embodiment uses a negative exponential function or similar fitting method, for example: ; ; In the formula, It is the ratio of residual strength to peak strength; The cohesion softening coefficient; The ratio of the residual values ​​of the internal friction angle, i.e., the residual internal friction angle. With peak internal friction angle The ratio; is the softening coefficient of the internal friction angle.

[0032] Let the mining disturbance damage function be... , Let be the weighting coefficient of the i-th mining disturbance component. Let i be the i-th mining disturbance component. The mining disturbance component includes: Explosion vibration damage component This characterizes the instantaneous damage and cumulative fatigue effect of rock mass structures caused by single or multiple blasting vibrations. Its calculation is based on the classical blasting vibration attenuation law and damage mechanics model, and the specific expression is as follows: ; In the formula, For the first The maximum amount of explosives per stage in a single blast; For the first The distance from the blast center of the second blast to the slope calculation point; For the first The moment of the blast; The current moment; , The blasting vibration attenuation parameters, which are related to the site's geological conditions, were obtained through regression analysis of on-site blasting vibration monitoring data. The damage recovery coefficient represents the partial recovery capacity of rock mass damage over time, and is calibrated by the microseismic activity attenuation law; It is a unit step function, ensuring that damage contribution only begins after the blast occurs.

[0033] Excavation and unloading damage components The deterioration of rock mass strength caused by stress relief and the formation of a free face due to slope toe excavation is directly related to the excavation process and stress state changes. The specific expression is as follows: ; In the formula, For a moment The excavation rate is derived from the mining production plan; The excavation unloading damage sensitivity coefficient was obtained by comparing data from historical excavation stages with those from the period of accelerated slope deformation. The change in the maximum principal stress at the calculation point due to excavation (unloading results in a negative value) can be estimated using theoretical formulas or simplified numerical analysis. It represents the uniaxial compressive strength of the rock mass.

[0034] Groundwater damage component This describes the effect of groundwater level fluctuations or rainfall infiltration causing changes in pore water pressure, which in turn reduces the effective stress and shear strength of rock masses (especially weak interlayers). The specific expression is as follows: ; In the formula, The cumulative effective rainfall at time t (or the pore water pressure near the slip surface obtained through the seepage model). The rainfall / pore water pressure threshold for triggering damage; For reference rainfall / pore water pressure, used for normalization; This refers to the groundwater level elevation monitored within the slope. The rate of change in groundwater level contributes more to damage when it rises rapidly. , These are the damage weighting coefficients for hydrostatic pressure and hydrodynamic pressure, respectively, determined through correlation analysis of groundwater level, rainfall records, and slope deformation response.

[0035] Weighting coefficients of the above damage components , , satisfy The value of the weighting coefficient reflects the relative importance of different disturbance mechanisms to the stability of a specific slope. The calibration of the weighting coefficients is an inversion optimization process based on historical data: collecting complete disturbance sequences over a sufficiently long historical period (e.g., 1-3 years). And the corresponding slope deformation response (such as key point displacement time series data), adjusted through optimization algorithms. , so that the damage function The deformation trend predicted by the driven slope mechanics model achieves the best fit with the actual monitoring data.

[0036] A physical information neural network is constructed based on the improved slope stability control equation. This network adopts a multi-layer feedforward neural network as its basic architecture: the input layer includes the spatial coordinates of any point within the slope computation domain. The time variable is t, and the initial rock mass parameters corresponding to this point are ( , and the current value of the perturbation function. The output layer is designed with multiple outputs, including the displacement components at that point. Stress tensor components and equivalent plastic strain .

[0037] The depth and width of the hidden layers in the network are configured according to the complexity of the slope model. This embodiment uses an architecture with eight hidden layers, each with 256 neurons, and the activation function is the hyperbolic tangent function, which has smoothing properties. The total loss of this network is... Data loss Physical loss and boundary condition loss The weighted summation consists of: ; Data loss Calculate the mean square error of the network output at the sensor deployment location compared to the measured displacement and stress data. Physical loss. It is the core of the physical constraints of the entire network. The network output field is calculated using automatic differentiation technology to satisfy the aforementioned improved slope stability control equations and static equilibrium equations. The residuals. Boundary condition loss. Ensure that the network solution conforms to the known displacement or stress boundary conditions. Minimize this composite loss function through an optimization algorithm, forcing the neural network to strictly obey defined physical laws while fitting the measured data.

[0038] The training of this network employs a two-stage strategy. The first stage is offline pre-training, which utilizes finite element numerical simulation software to generate a large amount of synthetic sample data based on a three-dimensional geological model, covering different combinations of rock mass parameters, different mining steps, and disturbance scenarios. This initial training allows the network to learn the basic laws of slope mechanical response. The second stage is online adaptive training, which uses real-time slope monitoring data streams as supervisory data to continuously and rapidly fine-tune the pre-trained network, enabling it to dynamically track and approximate the mechanical state of the real slope, thus forming a digital twin of the slope.

[0039] S4. Input the slope geological model parameters, multi-source heterogeneous monitoring data, and mining plan into the physical information neural network to solve in real time the displacement field, stress field, dynamic stability coefficient, and potential sliding surface spatial morphology of the entire domain. Simultaneously, output the failure probability time series curve, including: S41. By using forward propagation calculations through a physical information neural network and constraining the output results with the improved slope stability control equation, displacement field distribution data and stress field distribution data of the entire slope area are directly generated.

[0040] After training the physical information neural network, it was put into real-time operation as the core solver. During the solution process, the slope geological model parameters at the current moment, preprocessed multi-source heterogeneous monitoring data, and short-term mining plans were used as inputs to the network. The network directly outputs the mechanical state variables of all discrete points within the slope computational domain through a single efficient forward propagation calculation. Specifically, based on internally stored mapping relationships trained under physical law constraints, the network processes the input data to directly generate two key distribution fields: one is the displacement field distribution data across the entire slope domain, which describes the movement vector of the slope rock mass at every point in space. Secondly, the stress field distribution data of the entire slope area describes the stress tensor state at every point within the rock mass. This process enables real-time and continuous insight into the mechanical state of the slope.

[0041] S42. Based on global stress field data, an optimization search algorithm is adopted to iteratively calculate the stability coefficients of potential sliding surfaces at different spatial locations, thereby determining the most dangerous potential sliding surface and its corresponding dynamic stability coefficient that minimizes the global stability coefficient.

[0042] This embodiment uses an integrated optimization search algorithm to intelligently generate a large number of potential sliding surfaces S with different locations and shapes in areas where slope failure may occur. For each assumed sliding surface S, based on the strength criterion provided by the improved slope stability control equation and the stress data on the sliding surface extracted from the global stress field, its corresponding stability coefficient is calculated. Specifically, the ratio of the anti-sliding force to the sliding force on the sliding surface S is the dynamic stability coefficient. It can be calculated using the following integral formula: ; In the formula, Potential sliding surfaces obtained by interpolation from the global stress field The actual shear stress on the surface.

[0043] The optimized search algorithm automatically finds the optimal solution through iterative optimization. The sliding surface whose value reaches the global minimum is identified as the most dangerous potential sliding surface at the current moment. The minimum value is the dynamic stability coefficient. At the same time, the three-dimensional spatial coordinates, morphology and other parameters of the sliding surface are accurately extracted and output as the potential sliding surface spatial morphology.

[0044] S43. Input working condition parameters, simulate the time evolution process of slope mechanical state under the corresponding working condition through physical information neural network, and output the time series curve of slope failure probability.

[0045] First, based on short-term weather forecasts, mining plans, and historical statistical information, N sets of random disturbance condition sequences for the next few days, conforming to a probability distribution, are automatically generated. Each set of conditions includes different combinations of rainfall events, blasting event sequences, and excavation progress. Then, the parameters of each set of future conditions are used as time-varying boundary conditions and sequentially input into a trained physical information neural network. The network acts as a high-speed simulator, rapidly simulating the continuous evolution of the slope's mechanical state from its current state to various future time points under these specific disturbance conditions. For each future time point... The system records the number of working conditions in all N sets of simulation results where the slope stability was determined to be in failure. The criterion for failure can be set as simultaneously satisfying the dynamic stability coefficient. <1.0 and the potential sliding surface plastic strain zone is completely penetrated. Then, at this future time point... Slope failure probability That is: ; By calculating and connecting multiple consecutive time points from the present to the future By obtaining the probability of failure, we can obtain the time series curve P(t) of the slope failure probability changing over time.

[0046] S5. Based on the dynamic stability coefficient, the time series curve of the failure probability, and the spatial morphology of the potential sliding surface, output the early warning level and trigger the corresponding early warning response mechanism. Generate a disposal decision based on the early warning level and the spatial morphology of the potential sliding surface, including: The rate of slope stability deterioration is determined by the rate of change of the dynamic stability coefficient over time. The intensity of risk growth is quantified by the slope of the failure probability time series curve. The degree of failure development is characterized by the expansion rate of the potential sliding surface and the slip rate. A multi-dimensional objective criterion system is constructed to output early warning levels and trigger corresponding early warning response mechanisms.

[0047] Based on the rate of change of dynamic stability coefficient over time Assess the rate of slope stability deterioration and set a threshold. and Distinguish between slow, medium, and rapid degradation, specifically: Slow degradation:

[0048] Moderate degradation:

[0049] Rapid degradation:

[0050] Using the instantaneous slope of the probability of destruction curve Quantifying the intensity of risk growth through thresholds and The growth is categorized into low, medium, and high risk levels, specifically: Low-risk growth:

[0051] Medium risk of growth:

[0052] High-risk growth:

[0053] Further extract the spatial extent expansion rate of the potential sliding surface With average slip rate (A is the projected area of ​​the sliding surface on the horizontal plane,) This represents the total number of nodes selected on the sliding surface. The rate of change of the displacement of the i-th node over time is used as a direct measure of the degree of activity of the sliding surface, and a comprehensive judgment is made on whether the slope is in the stage of accelerated deformation or slippage.

[0054] During the slope stabilization or slow deformation phase, the system continuously calculates... and The moving average is obtained. and The sliding surface is considered to have entered the acceleration phase when both of the following conditions are met: 1. and ;in and The acceleration factor is set based on engineering experience, and its typical value range is [value range missing]. This is used to identify a significant increase in rate relative to the historical average. 2. and The instantaneous value exceeds the absolute threshold set according to the soil and rock type and slope grade. and .

[0055] Based on the above criteria, the warning level is divided into four levels, and the corresponding warning response mechanism is triggered: Blue Alert (Attention Level): Triggered when Fs>1.2, P(t)<0.1, and all criteria are within the low-risk range. Response mechanisms include: increasing monitoring frequency to once per hour and generating observation reports.

[0056] Yellow alert (warning level): When 1.1 < The event is triggered when P(t) ≤ 1.2 or 0.1 ≤ P(t) < 0.3, and either criterion falls within the medium-risk range. The response mechanism includes: initiating on-site inspections and preparing emergency response plans.

[0057] Orange alert (Level 1): When 1.0 < The event is triggered when P(t) ≤ 1.1 or 0.3 ≤ P(t) < 0.6, and at least two of the criteria fall within the medium-to-high risk range. The response mechanism includes: restricting work near the slope and activating emergency response teams.

[0058] Red Alert (Critical Level): When The event is triggered when P(t) is ≤1.0 or P(t)≥0.6, and the sliding surface is in a state of accelerated expansion or slippage. The response mechanism includes: immediately evacuating the danger zone and initiating emergency rescue.

[0059] Based on the spatial morphology of the potential sliding surface and the slope's mechanical equilibrium state, the direction of reinforcement, unloading, or drainage treatment is matched to the mechanical characteristics of the sliding surface. Simultaneously, based on the stability deterioration rate corresponding to the warning level, treatment measures are determined, including: By extracting the location, depth H, dip angle α, area A of the sliding surface, and its relationship with weak interlayers, the mechanical failure mode is determined. Based on this, appropriate engineering treatment directions are recommended: for the middle and upper steeply dipping sliding surface, reinforcement measures such as anti-slide piles and prestressed anchor cables are recommended; for shallow sliding at the toe of the slope, unloading methods such as slope cutting and load reduction are recommended; if the sliding surface is closely related to groundwater activity, drainage measures such as intercepting ditches and horizontal drainage holes should be prioritized. The priority and combination of treatment measures are dynamically determined according to the stability deterioration rate corresponding to the warning level, and a decision report including engineering quantity estimates, construction location, construction period recommendations, and key points for monitoring and adjustment is output, forming a complete closed loop from warning to treatment.

[0060] S6. Collect on-site execution feedback data for disposal decisions to continuously optimize the model parameters of the physical information neural network.

[0061] This embodiment also includes: spatiotemporal registration of displacement field, stress field and multi-source heterogeneous monitoring data, and dynamic correction of slope geological model parameters using ensemble Kalman filtering algorithm.

[0062] Spatiotemporal registration first unifies all monitoring data to the model's time reference and three-dimensional spatial coordinate system. The ensemble Kalman filter algorithm then applies the model's parameters to be corrected (initial cohesion of each soil element). and internal friction angle Defined as a state vector The registered monitoring data is defined as the observation vector. The algorithm maintains a set of parameters (e.g., 100 sets) to characterize the uncertainty of the parameters. Each assimilation cycle consists of two steps: a prediction step, which drives the physical information neural network with the current set of parameters to obtain the predicted observations. ; Update step, to incorporate actual observations With prediction Comparisons were made by calculating the Kalman gain matrix. To optimally adjust the parameter set:

[0063] In the formula, and These are the parameter sets before and after the update, and the Kalman gain. The parameters are determined by both the prediction covariance and the observation error. The mean of the updated parameter set serves as the corrected optimal parameter, which is fed back into the model in real time. This process forms a closed loop of "monitoring-assimilation-correction," enabling the digital twin to dynamically absorb the latest observation data, automatically calibrate the model, and significantly improve the reliability of its long-term evolution and early warning.

[0064] On the other hand, reference Figure 2The present invention also proposes an open-pit coal mine slope deformation monitoring system based on multi-source data, comprising: The geological model building module is used to build geological models of open-pit coal mine slopes; The data acquisition module is used to collect multi-source heterogeneous monitoring data from the air, space, ground, and interior of open-pit coal mines in real time. The network construction module is used to construct an improved slope stability control equation that integrates mining disturbance factors and the constitutive relationship of rock mass time-dependent damage, and to establish a physical information neural network based on the improved slope stability control equation. The stability solution module is used to input slope geological model parameters, multi-source heterogeneous monitoring data and mining plans into the physical information neural network, and solve in real time to obtain the displacement field, stress field, dynamic stability coefficient and potential sliding surface spatial morphology of the whole domain, while outputting the failure probability time series curve. The early warning output module is used to output early warning levels and trigger corresponding early warning response mechanisms based on dynamic stability coefficients, time-series curves of failure probability and potential sliding surface spatial morphology, and to generate disposal decisions based on the early warning level and potential sliding surface.

[0065] The geological model correction module is used to perform spatiotemporal registration of the global displacement field, stress field and multi-source heterogeneous monitoring data, and to dynamically correct the slope geological model parameters using an ensemble Kalman filter algorithm.

[0066] The model optimization module is used to collect on-site execution feedback data for handling decisions and to continuously optimize the model parameters of the physical information neural network.

[0067] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0068] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring slope deformation in open-pit coal mines based on multi-source data, characterized in that, Includes the following steps: S1. Constructing a geological model of open-pit coal mine slopes; S2. Real-time acquisition of multi-source heterogeneous monitoring data from air, space, ground, and indoor environments in open-pit coal mines; S3. Construct an improved slope stability control equation that integrates mining disturbance factors and the constitutive relationship of rock mass time-dependent damage, and establish a physical information neural network based on the improved slope stability control equation; S4. Input the slope geological model parameters, multi-source heterogeneous monitoring data and mining plan into the physical information neural network to solve in real time the displacement field, stress field, dynamic stability coefficient and potential sliding surface spatial morphology of the whole domain, and output the failure probability time series curve. S5. Based on the dynamic stability coefficient, the time series curve of the failure probability and the potential sliding surface spatial morphology, output the early warning level and trigger the corresponding early warning response mechanism, and generate a disposal decision according to the early warning level and the potential sliding surface spatial morphology.

2. The method for monitoring slope deformation in open-pit coal mines based on multi-source data according to claim 1, characterized in that, In S3, the improved slope stability control equation is as follows: ； ； ； In the formula, Rock mass shear strength; Normal stress; This represents the initial cohesion of the rock mass. It is the cohesion degradation function of the rock mass itself as it evolves with equivalent plastic strain; This represents the initial cohesion of the rock mass. Let the mining disturbance damage function be... , Let be the weighting coefficient of the i-th mining disturbance component. Let i be the i-th mining disturbance component. The mining disturbance component includes the blasting vibration damage component, the excavation unloading damage component, and the groundwater damage component. The initial internal friction angle of the rock mass; The internal friction angle degradation function is the evolution of the rock mass itself with equivalent plastic strain. This is the damage sensitivity coefficient of the internal friction angle.

3. The method for monitoring slope deformation in open-pit coal mines based on multi-source data according to claim 2, characterized in that, S4 includes: By using forward propagation calculations through a physical information neural network and constraining the output results with the improved slope stability control equation, displacement field distribution data and stress field distribution data of the entire slope area are directly generated. Based on global stress field data, an optimization search algorithm is used to iteratively calculate the stability coefficients of potential sliding surfaces at different spatial locations, thereby determining the most dangerous potential sliding surface and its corresponding dynamic stability coefficient that makes the stability coefficient take the global minimum value. Input working condition parameters, simulate the time evolution process of slope mechanical state under corresponding working conditions through physical information neural network, and output the time series curve of slope failure probability.

4. The method for monitoring slope deformation in open-pit coal mines based on multi-source data according to claim 3, characterized in that, Dynamic stability coefficient The calculation formula is as follows: ； In the formula, For potential sliding surfaces The actual shear stress on the surface.

5. The method for monitoring slope deformation in open-pit coal mines based on multi-source data according to claim 1, characterized in that, S5 include: The rate of slope stability deterioration is judged by the rate of change of dynamic stability coefficient over time, the intensity of risk growth is quantified by the slope of the failure probability time series curve, and the degree of failure development is characterized by the expansion rate of the spatial range of the potential sliding surface and the sliding rate. A multi-dimensional objective criterion system is constructed to output the early warning level and trigger the corresponding early warning response mechanism. Based on the spatial morphology of the potential sliding surface and the mechanical equilibrium state of the slope, the direction of reinforcement, unloading or drainage treatment is matched with the mechanical characteristics of the sliding surface. At the same time, the treatment measures are determined according to the stability deterioration rate corresponding to the warning level.

6. The method for monitoring slope deformation in open-pit coal mines based on multi-source data according to claim 1, characterized in that, Also includes: The displacement field, stress field and multi-source heterogeneous monitoring data are spatiotemporally registered, and the slope geological model parameters are dynamically corrected using an ensemble Kalman filter algorithm.

7. The method for monitoring slope deformation in open-pit coal mines based on multi-source data according to claim 1, characterized in that, Also includes: The on-site execution feedback data of the disposal decision is collected to continuously optimize the model parameters of the physical information neural network.

8. A slope deformation monitoring system for open-pit coal mines based on multi-source data, characterized in that, include: The geological model building module is used to build geological models of open-pit coal mine slopes; The data acquisition module is used to collect multi-source heterogeneous monitoring data from the air, space, ground, and interior of open-pit coal mines in real time. The network construction module is used to construct an improved slope stability control equation that integrates mining disturbance factors and the constitutive relationship of rock mass time-dependent damage, and to establish a physical information neural network based on the improved slope stability control equation. The stability solution module is used to input slope geological model parameters, multi-source heterogeneous monitoring data and mining plans into the physical information neural network, and solve in real time to obtain the displacement field, stress field, dynamic stability coefficient and potential sliding surface spatial morphology of the whole domain, while outputting the failure probability time series curve. The early warning output module is used to output an early warning level and trigger a corresponding early warning response mechanism based on the dynamic stability coefficient, the time series curve of the failure probability and the spatial morphology of the potential sliding surface, and to generate a disposal decision based on the early warning level and the potential sliding surface.

9. A method for monitoring slope deformation in open-pit coal mines based on multi-source data as described in claim 8, characterized in that, It also includes a geological model correction module, which is used to perform spatiotemporal registration of the global displacement field, stress field and the multi-source heterogeneous monitoring data, and to dynamically correct the slope geological model parameters using an ensemble Kalman filter algorithm.

10. A method for monitoring slope deformation in open-pit coal mines based on multi-source data as described in claim 8, characterized in that, It also includes a model optimization module, which is used to collect on-site execution feedback data of the disposal decision to continuously optimize the model parameters of the physical information neural network.

Images