Oolitic karst collapse column activation instability risk evaluation and dynamic early warning method and system

Abstract

The present application relates to the technical field of coal mine water disaster prevention and mine safety, and specifically discloses a method and system for evaluating and dynamically warning the activation and instability risk of Ordovician limestone collapse column. The method comprises: establishing a hydrological prior model of the collapse column structure; constructing a risk driving function that integrates the coupling effect of mining and water pressure, and calculating the activation and instability risk index; connecting and fusing the multi-source monitoring data stream of the whole mining process; dynamically calculating the self-adaptive warning threshold based on the real-time risk index and the monitoring sequence fluctuation; triggering the graded warning and corresponding treatment measures according to the risk level, and forming a closed loop of evaluation-warning-treatment. The system correspondingly comprises the modules of exploration data integration and modeling, risk calculation engine, multi-source monitoring data fusion processing, self-adaptive warning decision and instruction execution, and closed loop feedback. The present application realizes the quantitative and dynamic evaluation and warning of the activation risk of the collapse column, and improves the accuracy and timeliness of the coal mine floor water disaster prevention and control.

Description

Technical Field

[0001] This application relates to the field of mine safety and automation control technology, and in particular to a method and system for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns. Background Technology

[0002] In pressurized mining of areas with developed karst collapse columns, risk assessment and early warning of collapse column activation and instability (delayed water inrush) are the core of water prevention and control work.

[0003] Current technologies primarily focus on the detailed detection of collapse columns (such as using CSAMT, downhole electrical resistivity tomography, and drilling to determine their extent, structure, and water conductivity) and the classification of water inrush hazard based on static parameters. These methods have significant limitations: First, risk assessments often rely on the results of a single exploration before mining and empirical criteria (such as the water inrush coefficient). The assessment conclusions are static and fail to incorporate the core disaster-causing factors such as dynamic changes in mining stress, floor damage, and water pressure fluctuations during the mining process into a unified and quantitative risk evolution model. This results in the assessment results being out of sync with the actual mining process and makes it difficult to effectively warn of "delayed activation".

[0004] Secondly, even if water level and resistivity are monitored during the mining process, fixed thresholds are often used for single-parameter alarms. This fails to deeply integrate multi-source and heterogeneous time-series monitoring information (such as microseismic activity, spontaneous potential, deformation, etc.) within a unified probabilistic framework to dynamically correct risk perception. Rigid warning thresholds can easily lead to false alarms or missed alarms.

[0005] Finally, existing early warning methods often lack mandatory and traceable closed-loop linkage with specific engineering measures (such as adjusting coal pillars, controlling mining speed, and increasing exploration density). After the early warning is issued, there may still be a dilemma of "conservatively leaving large coal pillars" or "blindly and recklessly advancing", which makes it impossible to achieve safe and efficient recovery of resources under controllable risks. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this application aims to provide a method and system for assessing and dynamically warning of the activation and instability risk of Ordovician limestone karst collapse columns. This system uses the coupling effect of mining stress field, fracture field, and Ordovician limestone water pressure field as the core risk driver. It integrates prior structural exploration information of the collapse column (column segmentation, distribution / integrity of mudstone slurry sections, development of column wall fractures, and possibility of water-conducting channels) with multi-source time-series monitoring information throughout the mining process (floor damage depth / deformation, evolution of spontaneous potential / resistivity, Ordovician limestone water level / pressure, microseismic / acoustic emission, etc.) to construct a dynamically updatable activation and instability risk index and graded early warning thresholds. The early warning results are then linked in a closed loop to the verification of waterproof coal pillars, monitoring intensification strategies, and allowable advance rate control, thereby balancing the safety of pressurized mining and the rationality of resource recovery under controllable risk.

[0007] To achieve the above objectives, this application adopts the following technical solution: This application provides a method for risk assessment and dynamic early warning of activation and instability of karst collapse columns in Ordovician limestone, including the following steps: S1. Establish a prior hydrological model of the collapse column structure. For target collapse columns within the influence range of the same working face, based on the results of controlled source magnetotelluric exploration (CSAMT), downhole high-density electrical resistivity tomography (HOT), borehole television, and verification drilling, characterize the spatial distribution and internal structural segmentation of the column. At least distinguish the rockfill section, mudstone slurry section, and possible column wall fracture zone. Provide a prior probabilistic description of the water-bearing layer or water-conducting channel between the column and the top surface of the Ordovician limestone. The output is a set of structural parameters and prior weights. S2 constructs a risk driving function that couples mining and water pressure. The depth of the base plate failure and the expansion of the disturbance zone caused by mining, the reduction of the effective water-proof thickness of the collapsed column, especially the mudstone section, and the confined water head of the Ordovician limestone are used as common driving terms. A continuously calculable activation instability risk index R is defined, which makes R monotonically increase as the depth of the base plate failure approaches the effective water-proof thickness, as the conductivity of the column wall cracks increases, as the effective water-proof thickness decreases, and as the water pressure of the Ordovician limestone increases. The structural parameters obtained in S1 are embedded as constraints into this risk driving function so that different column types can obtain different risk baselines. S3 connects to the multi-source time-series monitoring stream of the entire mining process. During the advance of the working face, it continuously acquires monitoring subsets corresponding to the time sequence, including at least parameters of bottom plate damage or disturbance response, water chemistry or potential parameters, and Ordovician limestone water level or water pressure parameters, and aligns the monitoring subsets into feature vectors within the same time window according to the spatiotemporal reference. S4 dynamically calculates the current risk index and adaptively calibrates the early warning threshold. Substitutes the feature vector of S3 into the risk driving function of S2 to obtain the current risk index Rt. At the same time, based on the statistical fluctuation of the recent monitoring sequence and the prior confidence of S1, adaptive thresholds T1 and T2 are calculated for Rt, so that the adaptive thresholds dynamically fluctuate with data quality, sampling rate stage and exploration credibility rather than being rigidly fixed as a single constant. S5 outputs graded early warnings and triggers corresponding response strategies. When Rt is less than T1, it outputs a normal advance state but still refreshes the model at a preset frequency. When Rt is greater than or equal to T1 and less than T2, it outputs a warning state and triggers local measures including encrypted natural potential or resistivity monitoring, drilling verification holes in the column's influence zone to verify the water-conducting channel, and reducing the daily advance rate. When Rt is greater than or equal to T2, it outputs a dangerous state and triggers strong response measures including suspending advance towards the column side, verifying whether the boundary of the waterproof coal pillar meets the criterion that the depth of bottom plate damage is less than the effective water-proof thickness, and widening temporary seepage prevention or drainage and pressure reduction measures when necessary. All early warning and response actions are recorded in a versioned manner, forming a closed loop from risk assessment to engineering response.

[0008] As a preferred technical solution, the structural hydrological prior model described in step S1 is a digital product that integrates and interprets the results of multiple exploration methods and performs three-dimensional modeling. These multiple exploration methods include at least magnetotelluric methods for exploring deep electrical structures, downhole electrical resistivity methods for finely characterizing shallow resistivity distribution, and borehole optical imaging technology that can directly observe the lithology and structure within the borehole. By fusing the above data, the geometric boundary of the collapse column is reconstructed in three-dimensional space, and structural units with different physical, mechanical, and hydrogeological characteristics are divided within it. These structural units include at least rockfill sections and mudstone sections with different material compositions and degrees of cementation. The development characteristics of fractures in the possible contact zone with the surrounding rock are also identified. At the same time, the relative water-impermeable performance of the column itself and the possibility of hydraulic connection with the underlying strong aquifer are evaluated based on the results of core drilling, water pressure tests, or hydrological observations. Finally, a comprehensive prior geological model is formed that includes spatial morphology, structural attributes, hydrological characteristics, and corresponding confidence levels.

[0009] As a preferred technical solution, the risk-driving function described in step S2 is a mathematical model that comprehensively quantifies multiple key risk influencing factors. This mathematical model takes the degree of damage to the coal seam floor caused by mining, the integrity and thickness of the effective water-blocking structure within the collapse column, the water pressure of the Ordovician limestone confined water, and the development degree of possible water-conducting channels as core input variables, and establishes a quantitative functional relationship between these variables and an index value characterizing the comprehensive activation and instability risk. This functional relationship is constructed such that the risk index value shows a monotonically increasing trend as the degree of floor damage intensifies, the effective water-blocking thickness decreases, the confined water head increases, or the water-conducting capacity increases. Furthermore, the weight coefficients of different variables in the function can be adaptively adjusted based on the judgment of the collapse column structure type and characteristics obtained in the prior model in step S1.

[0010] As a preferred technical solution, the multi-source time-series monitoring stream mentioned in step S3 originates from a three-dimensional monitoring network specifically deployed for the mining-hydrological coupling process. The time-series data collected by this three-dimensional monitoring network includes: monitoring data reflecting the stress, strain, and fracture state of the rock mass, acquired through mechanical sensors and microseismic monitoring systems deployed in the coal seam floor or adjacent roadways; monitoring data reflecting groundwater seepage and geochemical field changes, acquired through geophysical and hydrogeochemical monitoring equipment deployed in boreholes or roadways; and borehole group water level or water pressure monitoring data reflecting the dynamic state of the Ordovician limestone aquifer. After being connected, all monitoring data undergoes a series of standardized processing steps, including time synchronization, spatial positioning, noise filtering, and missing value interpolation, to form a standardized spatiotemporal data sequence that is strictly aligned with the mining progress and can be used for real-time model updates.

[0011] As a preferred technical solution, the adaptive thresholds T1 and T2 mentioned in step S4 are comprehensively set by dynamically analyzing the statistical characteristics of the recent risk index sequence and combining the degree of certainty of the prior model. The determination of T1 is based on the historical baseline level of the risk index and its normal fluctuation range in the recent monitoring period, and the tolerance of the fluctuation range is corrected by considering the credibility of the prior geological exploration results. T2 adds a dynamic adjustment amount that is positively correlated with the intensity of the current mining activity on the basis of T1.

[0012] As a preferred technical solution, the local measures triggered under the alert state in step S5 include: firstly, automatically increasing the monitoring frequency and data analysis depth of key hydrogeological parameters such as spontaneous potential and resistivity; secondly, intelligently generating design schemes and target point suggestions for construction verification boreholes based on the spatial orientation of risk anomalies to verify or correct the judgment on water conductivity in the prior model; and simultaneously outputting control suggestions to the production system to limit the mining intensity, including operable instructions such as reducing the daily advance speed. The system will continuously track and evaluate the change in risk index response after the implementation of these measures to form a preliminary closed loop of response feedback.

[0013] As a preferred technical solution, the emergency response measures triggered under dangerous conditions in step S5 include: automatically issuing instructions to stop further progress towards high-risk areas, and immediately calling up the latest data on the depth of bottom plate damage and the thickness of the effective water-resistant layer to perform a safety review and calculation of the existing waterproof coal pillar design. At the same time, the system will check the emergency capability and status of the mine drainage system based on real-time water pressure data, and provide decision support and a contingency plan call interface for further engineering measures such as grouting and sealing or pressure reduction and drainage that may be required.

[0014] As a preferred technical solution, the method is based on a case-based learning model and knowledge optimization mechanism: the entire process experienced by each completed mining face, including the prior model, all monitoring data, risk index curve, early warning trigger records to the final engineering outcome, will be archived by the system as a complete learning case. These learning cases are used to inversely correct and optimize the key parameter relationships in the risk driving function and the logical rules in the calculation of early warning thresholds.

[0015] As a preferred technical solution, the implementation of the method relies on a software platform that integrates data management, model calculation, decision analysis, and control interaction functions. This software platform has standard interfaces for data exchange with mine geological databases, safety monitoring systems, and production management systems. It can automatically acquire the exploration data required for modeling, access monitoring data streams in real time, perform risk calculations and early warning judgments, and convert disposal strategies into executable instructions or suggestions to push to relevant systems and personnel, thereby realizing full-process digital and intelligent management and control from geological data to risk assessment, and then to early warning decisions and engineering responses.

[0016] This application also provides a system for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns, used to implement the above-mentioned method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns. The system includes: The exploration data integration and modeling module is used to import and process various geophysical exploration and drilling data to construct a priori three-dimensional structural and hydrological model of the target collapse column. The risk calculation engine module, which embeds a risk driving function that couples mining and water pressure, can dynamically calculate the activation instability risk index based on prior models and real-time monitoring data streams. The multi-source monitoring data fusion and processing module is used to access, align, and preprocess real-time data streams from foundation damage monitoring, groundwater chemical and potential monitoring, and Ordos limestone water pressure monitoring. The adaptive early warning decision module is used to calculate the dynamic early warning threshold in real time based on the risk index sequence, and generate corresponding early warning signals and handling strategy suggestions according to the risk level. The instruction execution and closed-loop feedback module is used to convert early warning strategies into specific control instructions or operational suggestions and send them to relevant systems, and to track and record the execution and effects of the handling process; The above modules exchange data and coordinate commands through the data center bus.

[0017] Compared with the prior art, the beneficial effects of this application are as follows: This application addresses the shortcomings of existing technologies by constructing a closed-loop method that encompasses the entire process of "geological prior knowledge - mining dynamic monitoring - risk quantification calculation - threshold adaptive floating - hierarchical linkage of measures".

[0018] First, this method is based on a structural hydrological prior model established by integrating multi-source exploration data. It constructs a risk-driven function with the core variables of mining-induced floor failure depth, effective water-blocking thickness, Ordovician limestone water pressure, and the development degree of water-conducting channels. This enables risk assessment to be coupled with mining activities in real time, and the risk index can dynamically reflect the evolution of disaster-causing conditions under the coupling effect of "mining-water pressure", thus realizing continuous and quantitative tracking of risk status.

[0019] Secondly, by real-time access and fusion of multiple types of time-series monitoring data throughout the mining process, and using them as dynamic inputs to the risk-driven function, while adaptively setting early warning thresholds based on the statistical characteristics and prior confidence of the monitoring sequence, risk assessment not only makes full use of real-time information, but also avoids rigid misjudgments caused by normal data fluctuations or exploration uncertainties, significantly improving the intelligence and reliability of early warning.

[0020] Finally, this method establishes a clear hierarchical early warning and response strategy linkage mechanism. It automatically triggers engineering response measures of increasing intensity for different levels of risk, from monitoring intensification and mining rate limits to mining stoppage review and emergency preparedness. All early warning and response actions are versioned and recorded, forming a complete management closed loop from risk perception, assessment, early warning to engineering execution and feedback. This ensures that precise and timely intervention measures can be taken under the premise of real-time risk identification, and ultimately maximize the protection of mining safety and resource recovery under controllable risk levels. Attached Figure Description

[0021] Figure 1 A schematic diagram of the multi-source monitoring network layout; Figure 2 This is a flowchart of the three-level early warning and response process; Figure 3 This is a module diagram of the system architecture for risk assessment and dynamic early warning of activation and instability of karst collapse columns in Ordovician limestone. Detailed Implementation

[0022] To enable those skilled in the art to better understand the present application, the technical solutions in specific embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0023] This application provides a method for risk assessment and dynamic early warning of activation and instability of karst collapse columns in Ordovician limestone, including the following steps: S1. Establish a prior hydrological model of the collapse column structure. For target collapse columns within the same working face influence range, based on the results of controlled source magnetotelluric exploration (CSAMT), downhole high-density electrical resistivity tomography (HOT), borehole television, and verification drilling, characterize the spatial distribution and internal structural segmentation of the column. At least distinguish the rockfill section, mudstone slurry section, and possible column wall fracture zone. Provide a prior probabilistic description of the water-bearing layer or water-conducting channel between the column and the top surface of the Ordovician limestone. The output is a set of structural parameters and prior weights.

[0024] First, surface CSAMT exploration data were collected and processed to obtain a two-dimensional inversion section reflecting the electrical structure from deep within the Ordovician limestone top surface.

[0025] Secondly, by integrating high-resolution resistivity data obtained from downhole high-density electrical resistivity methods in roadways, the boundary morphology of the collapse column near the coal seam and the development of column wall fractures are precisely characterized.

[0026] Then, all borehole data that revealed the collapse column were combined, including borehole wall images from borehole television viewing, lithology and recovery rates from core logging, and results from simple borehole hydrological observations.

[0027] Using this data, a three-dimensional solid model of the collapse column and its surrounding rock was constructed in professional 3D geological modeling software, employing a combination of deterministic modeling with multiple constraints and stochastic simulation. The 3D solid model needs to clearly identify the rockfill section composed of disordered breccia, the relatively water-impermeable mudstone section composed of silty cement, and the fractured zone where the column contacts the surrounding rock.

[0028] For mudstone sections, their thickness, spatial continuity, and degree of cementation need to be assessed, and a quantitative index characterizing their water-tightness should be assigned based on water pressure tests or water inflow data.

[0029] The final output prior weights are used to characterize the credibility of different parts of the model. For example, the weight of mudstone sections verified by drilling is high, while the weight of boundaries inferred solely by geophysical exploration is low.

[0030] S1 integrates geophysical, drilling, and geological information to construct a digital geological body that closely approximates reality, providing a static but uncertain spatial framework for subsequent dynamic risk analysis. This upgrades the geological understanding of collapse columns from scattered reports, drawings, and descriptions to a unified, quantifiable, and computationally quantifiable three-dimensional digital model, laying a precise spatial data foundation for risk assessment.

[0031] S2 constructs a risk driving function that couples mining and water pressure. The depth of the base plate failure and the expansion of the disturbance zone caused by mining, the reduction of the effective water-proof thickness of the collapsed column, especially the mudstone section, and the confined water head of the Ordovician limestone are used as common driving terms. A continuously calculable activation instability risk index R is defined, which increases monotonically as the depth of the base plate failure approaches the effective water-proof thickness, as the conductivity of the column wall cracks increases, as the effective water-proof thickness decreases, and as the water pressure of the Ordovician limestone increases. The structural parameters obtained in S1 are embedded into this risk driving function as constraints, so that different column types can obtain different risk baselines.

[0032] Risk-driven function The mathematical expression can be: ; in, It is the depth of the mining-induced floor failure as assessed or predicted in real time. It is the effective aquitard thickness within the collapse column below the current mining location, extracted from a priori models. It is a non-linear function, such as an exponential function, such that when near hour, A sharp increase in the value indicates a risk of a breakout. It is Ordovician limestone water pressure. It is a certain critical reference water pressure. yes An increasing function. It is the column wall crack conduction coefficient, which can be quantified based on the degree of microseismic event concentration near the column wall or the dynamic decrease in resistivity. yes An increasing function. , , These are weighting coefficients determined based on the S1 prior model. For example, for a collapse column dominated by thick, intact mudstone sections, the following weighting coefficients are assigned: A higher weight is given, emphasizing the crucial role of effective waterproof thickness. For collapse columns with highly developed wall fissures, a higher weight is assigned... Higher weight. ,initial Parameters such as these are obtained directly from the prior model.

[0033] S2 abstracts the complex interactions of mining, structure, and water pressure into a computable mathematical function, mapping geological conditions, mining activities, and hydrological dynamics onto a risk index between 0 and 1, thus achieving quantitative and dynamic characterization of risk. It creates a core algorithm that comprehensively reflects the coupled evolution of multiple disaster-causing factors, making risk no longer a vague qualitative concept, but a quantitative indicator that can be calculated, tracked, and compared in real time, providing precise judgment criteria for intelligent early warning.

[0034] S3 connects to the multi-source time-series monitoring stream of the entire mining process. During the advance of the working face, it continuously acquires monitoring subsets corresponding to the time sequence, including at least parameters of bottom plate damage or disturbance response, water chemistry or potential parameters, and Ordovician limestone water level or water pressure parameters. The monitoring subsets are then aligned into feature vectors within the same time window according to the spatiotemporal reference.

[0035] S3 constructs a three-dimensional spatiotemporal monitoring network and achieves data fusion.

[0036] like Figure 1 As shown, the failure response parameters of the floor strata are obtained through borehole stress gauges, multi-point displacement gauges, and microseismic monitoring systems arranged in the floor roadway, and are used to invert the stress changes, deformation, and fracture depth of the floor strata.

[0037] Water chemistry or potential parameters are obtained by using natural potential monitoring electrode arrays or high-density resistivity dynamic monitoring systems arranged in boreholes in the floor or in tunnels to capture geoelectric field and resistivity anomalies caused by water seepage.

[0038] The water level and pressure parameters of the Ordos limestone formation are acquired in real time through a specially constructed long-term hydrological observation well group equipped with submersible pressure sensors.

[0039] All sensor data is transmitted to a unified data platform via an industrial ring network. The platform assigns a precise timestamp and three-dimensional spatial coordinates to each data point. The data fusion engine collects readings from all sensors within a preset time window, such as every hour or every 5 meters of advance. After preprocessing including filtering, noise reduction, gross error removal, and spatial interpolation, a standardized, multi-dimensional feature vector is generated. ; This represents the estimated depth of damage to the base plate. It is stress. It is the number of microseismic events. It is resistivity. It is the natural potential. It's water pressure.

[0040] S3 uses IoT technology to capture multi-dimensional response signals of the rock mechanics field, groundwater seepage field, and confined hydrodynamic field in real time during the mining process, and synchronizes them to provide vivid feature inputs that reflect the current system state for the risk model. The beneficial effect is that it enables multi-dimensional, continuous, and three-dimensional perception of the disaster-causing process, upgrading traditional discrete and static monitoring to continuous and integrated dynamic monitoring, and providing a reliable data stream for real-time risk calculation.

[0041] S4 dynamically calculates the current risk index and adaptively calibrates the early warning threshold. Substitutes the feature vector of S3 into the risk driving function of S2 to obtain the current risk index Rt. At the same time, based on the statistical fluctuation of the recent monitoring sequence and the prior confidence of S1, adaptive thresholds T1 and T2 are calculated for Rt, so that the adaptive thresholds dynamically fluctuate with data quality, sampling rate stage and exploration credibility rather than being rigidly fixed as a single constant.

[0042] S4 involves the real-time calculation of risk indices and dynamic management of thresholds.

[0043] The risk calculation engine receives the latest feature vectors every second or every minute. Substitute it into the risk-driven function defined in S2 In this process, the risk index at the current moment is calculated. .

[0044] Warning threshold and It is not a fixed value. The warning threshold can be calculated using the following formula: ; in It is a static risk baseline value estimated based on prior models and mining design. It's the past The standard deviation of the risk index series over a given time period represents the normal level of risk fluctuation. This is an adjustment coefficient, inversely proportional to the average confidence level of the prior model in S1. The lower the model confidence level, the higher the confidence level. A larger value results in a wider threshold range, reducing false alarms caused by model uncertainty.

[0045] The danger threshold can be calculated using the following formula: ; in This is the average daily advance rate of the current working face. It is a proportionality coefficient. This means that as mining intensity increases, the danger threshold automatically rises, making it more sensitive to changes in risk.

[0046] S4 operates a real-time risk model, combining historical data volatility and the reliability of prior knowledge to dynamically determine reasonable early warning thresholds, enabling the early warning system to adapt to different environments. This significantly improves the intelligence and accuracy of early warnings, avoiding frequent false alarms when data fluctuates normally or geological understanding is unclear, and automatically increasing alertness when mining activities intensify, thus achieving precise early warnings.

[0047] S5 outputs tiered early warnings and coordinates response strategies, such as... Figure 2 As shown, when Rt is less than T1, the normal advance state is output, but the model is refreshed at the preset frequency. When Rt is greater than or equal to T1 and less than T2, the warning state is output and local measures are triggered, including encrypted natural potential or resistivity monitoring, additional verification holes are drilled in the influence zone of the column to verify the water channel, and the daily advance rate is reduced. When Rt is greater than or equal to T2, the danger state is output and strong measures are triggered, including suspending the advance towards the column side, verifying whether the boundary of the waterproof coal pillar meets the criterion that the depth of the bottom plate damage is less than the effective water-proof thickness, and widening the temporary seepage prevention or drainage pressure reduction measures when necessary. All warning and response actions are recorded in a versioned manner, forming a closed loop from risk assessment to engineering response.

[0048] S5 enables closed-loop management of early warning and response.

[0049] The system's human-machine interface displays risk status using different colors.

[0050] When in alert mode, the system automatically executes a preset response script, such as increasing the natural potential monitoring frequency from once per hour to once every 10 minutes, and automatically generating a verification drilling design suggestion sheet containing hole location, angle, and depth, which is then pushed to the geological survey department. Simultaneously, through the production management system interface, it sends a suggested instruction to the workface to limit the daily progress speed to within 80% of the original plan, requiring confirmation from the shift leader.

[0051] When in a dangerous state, the system automatically sends a lockout command to the coal mining machine control system, prohibiting cutting towards the collapse pillar side, and pops up a waterproof coal pillar verification interface, automatically filling in the latest information. and The value is verified.

[0052] All early warning triggers, the issuance of handling instructions, personnel confirmations, and execution feedback are recorded in a blockchain-style log, forming an immutable and complete chain of evidence.

[0053] S5 forcibly links risk assessment results with specific engineering technologies and management measures through a rule engine and software interface, ensuring that once a risk is identified, corresponding control measures are implemented and their effectiveness is tracked. From risk perception to risk control, an executable, traceable, and auditable proactive prevention and control closed loop is established, achieving intelligent early warning and greatly improving the speed of mine emergency response and the scientific nature of decision-making.

[0054] As a preferred technical solution, the structural hydrological prior model in step S1 is a digital product that integrates and interprets the results of multiple exploration methods and performs three-dimensional modeling. These multiple exploration methods include at least magnetotelluric methods for exploring deep electrical structures, downhole electrical methods for finely characterizing shallow resistivity distributions, and borehole optical imaging technology that can directly observe the lithology and structure within the borehole. By fusing the above data, the geometric boundary of the collapse column is reconstructed in three-dimensional space, and structural units with different physical, mechanical, and hydrogeological properties are divided within it. The structural units include at least rockfill sections and mudstone sections with different material compositions and degrees of cementation. The development characteristics of fractures in the possible contact zone with the surrounding rock are also identified. At the same time, the relative water-impermeable performance of the column itself and the possibility of hydraulic connection with the underlying strong aquifer are evaluated based on the results of core drilling, water pressure tests, or hydrological observations. Finally, a comprehensive prior geological model is formed that includes spatial morphology, structural properties, hydrological characteristics, and corresponding confidence levels.

[0055] First, by using ground-based CSAMT resistivity inversion sections and combining them with known geological information, we can preliminarily interpret the top burial depth, center location, and vertical extension range of the collapse column.

[0056] Secondly, the high-density electrical resistivity data collected in the underground roadway is used to generate resistivity inversion data volume. This high-resolution data volume is used to correct the shallow boundary interpreted by CSAMT and to finely depict the morphology of the column around the roadway and the possible low-resistivity fracture zone.

[0057] Then, the trajectories, lithological stratifications, borehole television images, and RQD values ​​of all boreholes were imported. Using borehole data as hard data and geophysical inversion results as soft constraints, geostatistical methods such as sequential indicator simulation were employed to establish a three-dimensional lithofacies model of the rockfill section, slurry section, fracture zone, and intact surrounding rock within the collapse column. For each lithofacies unit, preliminary mechanical and hydrological parameters were assigned according to its definition.

[0058] Finally, based on the permeability of the pressure test or the borehole water inflow, the water-tightness of the mudstone slurry section is evaluated by zone, and areas suspected of having high water conductivity are marked.

[0059] The entire process was completed in 3D geological modeling software, and the output was a 3D mesh model containing multiple attribute fields.

[0060] By employing a technical approach of geophysical exploration morphology, drilling to determine attributes, and statistical simulation to fill in the gaps, a reliable 3D geological model is constructed by maximizing the use of limited exploration information. This achieves transparency and quantification of the geological structure of the collapse column, providing a unique and accurate 3D data source for all subsequent numerical analysis, risk calculation, and engineering decision-making.

[0061] As a preferred technical solution, the risk driving function in step S2 is a mathematical model that comprehensively quantifies multiple key risk influencing factors. This mathematical model takes the degree of damage to the coal seam floor caused by mining, the integrity and thickness of the effective water-blocking structure within the collapse column, the water pressure of the Ordovician limestone confined water, and the development degree of possible water-conducting channels as core input variables, and establishes a quantitative functional relationship between these variables and an index value characterizing the comprehensive activation and instability risk. This functional relationship is constructed such that the risk index value shows a monotonically increasing trend as the degree of floor damage intensifies, the effective water-blocking thickness decreases, the confined water head increases, or the water-conducting capacity increases. Furthermore, the weight coefficients of different variables in the function can be adaptively adjusted based on the judgment of the collapse column structure type and characteristics obtained in the prior model in step S1.

[0062] The general form of a function is: ; in Representing different risk factors, It is its normalized value. It's a mapping function. Weights Determining the key is crucial.

[0063] The system has a built-in knowledge base that classifies the features of the prior model of the collapse column. For example, class A is defined as the dominant type with thick mudstone slurry, class B as the fractured type, and class C as the loose rockfill type. For class A, the function related to the effective impermeable thickness... weight The setting is the highest because its risk primarily depends on whether the mudstone section is breached. For Class B, regarding the water-conducting channel coefficient... weight These classification rules and weight presets are set to the highest level. These rules and weight presets are developed by domain experts and stored as rule tables.

[0064] In practical applications, the system analyzes the prior model of the target collapse column, automatically matches its category, and loads the corresponding set of weight parameters. This allows the same mathematical model to be flexibly applied to collapse columns with different geological characteristics, making the evaluation more targeted.

[0065] This step introduces a rule-based expert system, enabling the risk model to have a certain ability to determine geological classifications and achieve semi-automatic adaptation of model parameters. The beneficial effect is that it improves the universality and professionalism of the risk assessment model, allowing it to more accurately reflect the differences in dominant risk factors under different geological conditions, thus providing risk assessment results that are more in line with reality.

[0066] As a preferred technical solution, the multi-source time-series monitoring stream in step S3 originates from a three-dimensional monitoring network specifically deployed for the mining-hydrological coupling process. The time-series data collected by this network includes: monitoring data reflecting the stress, strain, and fracture state of the rock mass, acquired through mechanical sensors and microseismic monitoring systems deployed in the coal seam floor or adjacent roadways; monitoring data reflecting groundwater seepage and geochemical field changes, acquired through geophysical and hydrogeochemical monitoring equipment deployed in boreholes or roadways; and borehole water level or water pressure monitoring data reflecting the dynamic state of the Ordovician limestone aquifer. All monitoring data undergoes a series of standardized processing steps after access, including time synchronization, spatial positioning, noise filtering, and missing value interpolation, forming a standardized spatiotemporal data sequence strictly aligned with the mining progress and suitable for real-time model updates.

[0067] Before mining, a monitoring scheme was designed around the area affected by the collapse column. Specialized monitoring boreholes were drilled in the coal seam floor, and fiber optic strain sensors and piezometers were installed. In adjacent roadways, roof and floor separation meters and roadway surface convergence monitoring stations were arranged at regular intervals. Microseismic geophones were densified between the two roadways of the working face and between the collapse column. Multiple sets of spontaneous potential and resistivity monitoring boreholes were drilled in the floor to form a fixed profile.

[0068] All observation wells in the Ordos limestone formation are connected to an automated water level telemetry system. A data platform establishes a stream processing pipeline to perform real-time source location and energy calculation on microseismic data, inverse imaging on resistivity data, and calculation of the rate of change on deformation data. Then, using the mining progress as the baseline time axis, the data from each sensor are interpolated to a unified time point and associated with their three-dimensional coordinates. For missing data, nearest-neighbor interpolation or prediction models based on historical sequences are used to fill in the gaps. Finally, a snapshot at each moment is a standard vector containing dozens of features.

[0069] We have constructed a fully automated data processing chain, from physical signal acquisition to feature information extraction and spatiotemporal standardization integration. This provides high-quality input data that is clean, well-organized, and spatiotemporally correlated for the upper-level risk model, which is the fundamental guarantee for ensuring the accuracy and reliability of risk calculation results.

[0070] As a preferred technical solution, the adaptive thresholds T1 and T2 in step S4 are set by dynamically analyzing the statistical characteristics of the recent risk index sequence and combining the degree of certainty of the prior model. T1 is determined based on the historical baseline level of the risk index and its normal fluctuation range in the recent monitoring period, and the tolerance of the fluctuation range is corrected by considering the credibility of the prior geological exploration results. T2 adds a dynamic adjustment amount that is positively correlated with the intensity of the current mining activity on the basis of T1.

[0071] ; in It is a basic coefficient. The smaller the value, the larger the tolerance range for fluctuations.

[0072] ; Mining activity intensity The step distance can be characterized by the average advance speed or cycle over the most recent 24 hours.

[0073] The algorithm automatically recalculates T1 and T2 at regular intervals or when V changes significantly.

[0074] By employing statistical process control and adaptive filtering, the early warning threshold can automatically adapt to the system's background noise and current operating status. This gives the early warning system good robustness and environmental adaptability, preventing erroneous judgments due to uncertainties in the geological model itself or normal process fluctuations, and ensuring it does not react sluggishly when mining intensifies or risks accumulate rapidly.

[0075] As a preferred technical solution, the local measures triggered under the alert state in step S5 include: firstly, automatically increasing the monitoring frequency and data analysis depth of key hydrogeological parameters such as spontaneous potential and resistivity; secondly, intelligently generating design schemes and target point suggestions for construction verification boreholes based on the spatial orientation of risk anomalies to verify or correct the judgment on water conductivity in the prior model; and simultaneously outputting control suggestions to the production system to limit the mining intensity, including operable instructions such as reducing the daily advance speed. The system will continuously track and evaluate the change in risk index response after the implementation of these measures to form a preliminary closed loop of response feedback.

[0076] When the system enters a state of alert, it automatically executes enhanced monitoring plans, such as accelerating resistivity inversion calculations from once every 6 hours to once per hour. Simultaneously, the risk calculation engine outputs the main contributing factors that cause an increase in Rt and their spatial locations. For example, if... If the contribution of the factor increases suddenly and the microseismic events are concentrated on one side of the collapse column, the system will automatically extrapolate a certain distance on that side based on the three-dimensional model, and intelligently recommend the opening position, orientation, dip angle and design depth of one or more verification boreholes to form a preliminary borehole design document.

[0077] Production control instructions are sent to the production execution system via message middleware, and these instructions include specific speed limits. The system will mark the time point that triggers the alert state. and continuously monitor In the period of time that followed Curve. If If a downward trend is observed, the feedback measures are effective. If If the number continues to climb, it suggests that the measures may be insufficient, providing a basis for escalating the response.

[0078] It deeply integrates early warning with investigation and production management workflows, and provides data-driven post-response evaluation of the handling effects. It achieves automation and intelligence from risk alarm to initial response, shortens the time for manual analysis and decision-making, and forms a preliminary "measure-effect" feedback learning cycle.

[0079] As a preferred technical solution, the strong response measures triggered in the dangerous state in step S5 include automatically issuing an instruction to stop the continued advancement towards the high-risk area, and immediately calling the latest data on the depth of the floor damage and the thickness of the effective water-proof layer to perform a safety review and calculation of the existing waterproof coal pillar design. At the same time, the system will check the emergency capability and status of the mine drainage system based on real-time water pressure data, and provide decision support and a contingency plan call interface for further engineering measures such as grouting and sealing or pressure reduction and drainage that may be required.

[0080] Upon triggering of a hazardous condition, the system sends an emergency shutdown interlock signal to the electro-hydraulic control system at the working face via hardwired connection or high-priority network messages. Simultaneously, it automatically retrieves the latest waterproof coal pillar design drawings for the area from the database, as well as the real-time floor failure depth obtained from monitoring data. And the effective waterproof thickness at that location extracted from the prior model The system automatically performs the verification; if... The interface will then display "Coal Pillar Failure Risk" in flashing red, and automatically pop up the key points of the "Base Plate Reinforcement" and "Ordovician Ash Pressure Reduction" plans from the emergency plan library.

[0081] The system will also read the water level in the central pump room's reservoir, the operating status of the pumps, and the drainage capacity in real time, and calculate the current Ordos limestone water pressure. If a sudden flood occurs, how long can the existing drainage system keep the water level below the safety line, and will the assessment results be displayed in real time?

[0082] Under extreme risks, it enables rapid linkage and information aggregation across multiple systems, including production control, engineering calculations, and emergency support, providing commanders with one-stop comprehensive situational awareness and decision support. In critical situations, it can maximize the use of automation to gain response time, quickly provide critical safety status information and feasible emergency response plan references, and support scientific disaster relief decision-making.

[0083] As a preferred technical solution, the method is based on a case-based learning model and knowledge optimization mechanism. The entire process experienced by each completed mining face, including the prior model, all monitoring data, risk index curve, early warning trigger records, and the final engineering outcome, will be archived by the system as a complete learning case. These learning cases are used to inversely correct and optimize the key parameter relationships in the risk driving function and the logical rules in the calculation of early warning thresholds.

[0084] After each working face is completed and the hydrogeological conditions are stable, the technicians will confirm the final outcome label of the water control case for that working face, such as "safe mining", "micro-seepage occurred", or "water inrush occurred".

[0085] The system packages all the data from the working face, including the initial 3D geological model file, the complete time-series monitoring database, the calculated full-cycle risk index sequence, all early warning logs, disposal records, and final outcome labels, forming a structured case.

[0086] During the model optimization phase, a batch learning approach is adopted, using all historical case data to optimize the risk-driving function. Weight parameters in Mapping function Shape parameters and coefficients in the threshold calculation formula , Then, retraining is performed. The training goal is to ensure that the risk curve calculated by the model throughout the entire mining cycle has the best statistical correlation with the final engineering outcome. For example, the risk curve for the "safe mining" case should remain at a low value for most of the time, while the risk curve for the "water inrush" case should show a peak value that significantly exceeds the threshold before the water inrush.

[0087] The optimization algorithm can be either gradient descent or evolutionary algorithm.

[0088] By leveraging historical data for supervised learning, the model parameters evolve through data-driven processes, leading to increasingly accurate predictions. This endows the system with the ability to continuously improve. As mining cases accumulate, the system's risk assessment and early warning accuracy will increase, creating a virtuous cycle of becoming smarter with use and possessing long-term value.

[0089] As a preferred technical solution, the implementation of the method relies on a software platform that integrates data management, model calculation, decision analysis and control interaction functions. This software platform has a standard interface for data exchange with mine geological databases, safety monitoring systems and production management systems. It can automatically acquire the exploration data required for modeling, access monitoring data streams in real time, perform risk calculation and early warning judgment, and convert disposal strategies into executable instructions or suggestions and push them to relevant systems and personnel, thereby realizing the full-process digital and intelligent management and control from geological data to risk assessment, and then to early warning decision and engineering response.

[0090] The platform adopts a microservice architecture, which mainly includes geological data services, monitoring data access services, model calculation services, early warning decision services, command distribution services, and front-end visualization services.

[0091] The geological data service automatically extracts exploration data from the geospatial management system via standard ODBC or API interfaces. The monitoring data access service collects data in real time from various safety monitoring subsystems using industry protocols such as OPCUA and ModbusTCP. The model calculation service deploys encapsulated risk-driven functions and data inversion algorithms. The early warning decision service runs threshold calculations and a rule engine. The instruction distribution service is responsible for converting instructions into a format recognizable by the target system and distributing them. All services communicate via the enterprise service bus.

[0092] The front end uses Web technology to provide interfaces such as 3D model display, risk dashboard, early warning information board, and disposal instructions.

[0093] Through a loosely coupled software architecture and standardized interfaces, disparate mine information systems and capabilities are integrated into a unified business intelligence platform, enabling seamless data and business process integration. This provides an engineered, implementable, integrated solution that allows advanced evaluation and early warning methods to be deployed and applied in coal mines as software products, seamlessly integrating with existing production management systems and transforming intelligent concepts into daily safety production control capabilities.

[0094] like Figure 3 As shown, this application also provides a system for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns, used to implement the above-mentioned method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns. The system includes: The exploration data integration and modeling module is used to import and process various geophysical exploration and drilling data to construct a priori three-dimensional structural hydrological model of the target collapse column.

[0095] The risk calculation engine module, which embeds a risk driving function that couples mining and water pressure, can dynamically calculate the activation instability risk index based on prior models and real-time monitoring data streams.

[0096] The multi-source monitoring data fusion processing module is used to access, align, and preprocess real-time data streams from foundation damage monitoring, groundwater chemical and potential monitoring, and Ordos limestone water pressure monitoring.

[0097] The adaptive early warning decision module is used to calculate the dynamic early warning threshold in real time based on the risk index sequence, and generate corresponding early warning signals and handling strategy suggestions according to the risk level.

[0098] The instruction execution and closed-loop feedback module is used to convert early warning strategies into specific control instructions or operational suggestions and send them to relevant systems, and to track and record the execution and effects of the handling process.

[0099] The above modules exchange data and coordinate commands through the data center bus.

[0100] The data integration and modeling module can be a workstation equipped with professional 3D modeling software. The risk calculation engine module is a core algorithm microservice deployed on a high-performance server. The multi-source monitoring data fusion processing module includes a data acquisition gateway and a stream processing framework. The adaptive early warning decision-making module is an alarm service containing a rule engine. The instruction execution and closed-loop feedback module is a work order and message push system. The data center bus can be implemented using a combination of message queues and relational databases.

[0101] Each module is deployed in the mine's data center and connected to external systems such as geological surveying, monitoring, and production via a network.

[0102] The workflow is as follows: the modeling module generates a priori models and stores them in the data center. The monitoring module continuously writes real-time data. The computing engine periodically reads the latest data, calls the model to perform calculations, and writes the results to the database. The early warning and decision-making module reads the risk index, judges the status, and generates early warning events and response strategies. The instruction execution module translates the strategies into specific tasks and distributes them.

[0103] It should be noted that the terms "first," "second," and similar terms used in this application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, "a" or "one," and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. "A plurality" or "several" indicates at least two. Unless otherwise stated, terms such as "front," "back," "left," "right," "lower," and / or "upper" are for illustrative purposes only and are not limited to a location or spatial orientation. Terms such as "comprising" or "including" indicate that the elements or objects preceding "comprising" encompass the elements or objects listed following "comprising" or "including" and their equivalents, and do not exclude other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.

[0104] The singular forms “a,” “the,” and “the” used in this application specification and appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0105] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns, characterized in that, Includes the following steps: S1. Establish a prior hydrological model of the collapse column structure. For target collapse columns within the influence range of the same working face, based on the results of controlled source magnetotelluric exploration (CSAMT), downhole high-density electrical resistivity tomography (HOT), borehole television, and verification drilling, characterize the spatial distribution and internal structural segmentation of the column. At least distinguish the rockfill section, mudstone slurry section, and possible column wall fracture zone. Provide a prior probabilistic description of the water-bearing layer or water-conducting channel between the column and the top surface of the Ordovician limestone. The output is a set of structural parameters and prior weights. S2 constructs a risk driving function that couples mining and water pressure. The depth of the base plate failure and the expansion of the disturbance zone caused by mining, the reduction of the effective water-proof thickness of the collapsed column, especially the mudstone section, and the confined water head of the Ordovician limestone are used as common driving terms. A continuously calculable activation instability risk index R is defined, which makes R monotonically increase as the depth of the base plate failure approaches the effective water-proof thickness, as the conductivity of the column wall cracks increases, as the effective water-proof thickness decreases, and as the water pressure of the Ordovician limestone increases. The structural parameters obtained in S1 are embedded as constraints into this risk driving function so that different column types can obtain different risk baselines. S3 connects to the multi-source time-series monitoring stream of the entire mining process. During the advance of the working face, it continuously acquires monitoring subsets corresponding to the time sequence, including at least parameters of bottom plate damage or disturbance response, water chemistry or potential parameters, and Ordovician limestone water level or water pressure parameters, and aligns the monitoring subsets into feature vectors within the same time window according to the spatiotemporal reference. S4 dynamically calculates the current risk index and adaptively calibrates the early warning threshold. Substitutes the feature vector of S3 into the risk driving function of S2 to obtain the current risk index Rt. At the same time, based on the statistical fluctuation of the recent monitoring sequence and the prior confidence of S1, adaptive thresholds T1 and T2 are calculated for Rt, so that the adaptive thresholds dynamically fluctuate with data quality, sampling rate stage and exploration credibility rather than being rigidly fixed as a single constant. S5 outputs graded early warnings and triggers corresponding response strategies. When Rt is less than T1, it outputs a normal advance state but still refreshes the model at a preset frequency. When Rt is greater than or equal to T1 and less than T2, it outputs a warning state and triggers local measures including encrypted natural potential or resistivity monitoring, drilling verification holes in the column's influence zone to verify the water-conducting channel, and reducing the daily advance rate. When Rt is greater than or equal to T2, it outputs a dangerous state and triggers strong response measures including suspending advance towards the column side, verifying whether the boundary of the waterproof coal pillar meets the criterion that the depth of bottom plate damage is less than the effective water-proof thickness, and widening temporary seepage prevention or drainage and pressure reduction measures when necessary. All early warning and response actions are recorded in a versioned manner, forming a closed loop from risk assessment to engineering response.

2. The method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns according to claim 1, characterized in that, The structural hydrological prior model described in step S1 is a digital product that integrates and interprets the results of multiple exploration methods and performs three-dimensional modeling. These multiple exploration methods include at least magnetotelluric methods for exploring deep electrical structures, downhole electrical resistivity methods for finely characterizing shallow resistivity distributions, and borehole optical imaging technology that can directly observe lithology and structure within the borehole. By fusing the above data, the geometric boundary of the collapse column is reconstructed in three-dimensional space, and structural units with different physical, mechanical, and hydrogeological properties are divided within it. These structural units include at least rockfill sections and mudstone sections with different material compositions and degrees of cementation. The development characteristics of fractures in possible contact zones with the surrounding rock are also identified. At the same time, the relative water-impermeable performance of the column itself and the possibility of hydraulic connection with the underlying strong aquifer are evaluated based on the results of core drilling, water pressure tests, or hydrological observations. Finally, a comprehensive prior geological model is formed that includes spatial morphology, structural properties, hydrological characteristics, and corresponding confidence levels.

3. The method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns according to claim 1 or 2, characterized in that, The risk-driving function described in step S2 is a mathematical model that comprehensively quantifies multiple key risk influencing factors. This mathematical model takes the degree of damage to the coal seam floor caused by mining, the integrity and thickness of the effective water-blocking structure within the collapse column, the water pressure of the Ordovician limestone confined water, and the development degree of possible water-conducting channels as core input variables, and establishes a quantitative functional relationship between these variables and an index value characterizing the comprehensive activation and instability risk. This functional relationship is constructed such that the risk index value shows a monotonically increasing trend as the degree of floor damage intensifies, the effective water-blocking thickness decreases, the confined water head increases, or the water-conducting capacity increases. Furthermore, the weight coefficients of different variables in the function can be adaptively adjusted based on the judgment of the collapse column structure type and characteristics obtained in the prior model in step S1.

4. The method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns according to claim 1, characterized in that, The multi-source time-series monitoring stream mentioned in step S3 originates from a three-dimensional monitoring network specifically deployed for the mining-hydrological coupling process. The time-series data collected by this three-dimensional monitoring network includes: monitoring data reflecting the stress, strain, and fracture state of the rock mass, acquired through mechanical sensors and microseismic monitoring systems deployed in the coal seam floor or adjacent roadways; monitoring data reflecting groundwater seepage and geochemical field changes, acquired through geophysical and hydrogeochemical monitoring equipment deployed in boreholes or roadways; and borehole group water level or water pressure monitoring data reflecting the dynamic state of the Ordovician limestone aquifer. After being connected, all monitoring data undergoes a series of standardized processing steps, including time synchronization, spatial positioning, noise filtering, and missing value interpolation, to form a standardized spatiotemporal data sequence that is strictly aligned with the mining progress and can be used for real-time model updates.

5. The method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns according to claim 1, characterized in that, The adaptive thresholds T1 and T2 mentioned in step S4 are set comprehensively by dynamically analyzing the statistical characteristics of the recent risk index sequence and combining the degree of certainty of the prior model. T1 is determined based on the historical baseline level of the risk index and its normal fluctuation range in the recent monitoring period, and the tolerance of the fluctuation range is corrected by considering the credibility of the prior geological exploration results. T2 is based on T1 and adds a dynamic adjustment amount that is positively correlated with the intensity of the current mining activity.

6. The method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns according to claim 1, characterized in that, The local measures triggered under the alert state in step S5 include: firstly, automatically increasing the monitoring frequency and data analysis depth of key hydrogeological parameters such as spontaneous potential and resistivity; secondly, intelligently generating design schemes and target point suggestions for construction verification boreholes based on the spatial orientation of risk anomalies to verify or correct the judgment on water conductivity in the prior model; and simultaneously outputting control suggestions to the production system to limit the mining intensity, including operable instructions such as reducing the daily advance speed. The system will continuously track and evaluate the change in risk index response after the implementation of these measures to form a preliminary closed loop of response feedback.

7. The method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns according to claim 1, characterized in that, The emergency response measures triggered under dangerous conditions in step S5 include: automatically issuing instructions to stop further progress towards high-risk areas, and immediately calling up the latest data on the depth of bottom plate damage and the thickness of the effective water-resistant layer to perform a safety review and calculation of the existing waterproof coal pillar design. At the same time, the system will check the emergency capability and status of the mine drainage system based on real-time water pressure data, and provide decision support and a contingency plan call interface for further engineering measures such as grouting or depressurization drainage that may be required.

8. The method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns according to claim 1, characterized in that, The method is based on a case-based learning model and knowledge optimization mechanism: the entire process experienced by each completed mining face, including the prior model, all monitoring data, risk index curve, early warning trigger records to the final engineering outcome, will be archived by the system as a complete learning case. These learning cases are used to inversely correct and optimize the key parameter relationships in the risk driving function and the logical rules in the calculation of early warning thresholds.

9. The method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns according to claim 1, characterized in that, The implementation of the method relies on a software platform that integrates data management, model calculation, decision analysis, and control interaction functions. This software platform has a standard interface for data exchange with mine geological databases, safety monitoring systems, and production management systems. It can automatically acquire the exploration data required for modeling, access monitoring data streams in real time, perform risk calculations and early warning judgments, and convert disposal strategies into executable instructions or suggestions to push to relevant systems and personnel. This enables full-process digital and intelligent management and control from geological data to risk assessment, and then to early warning decisions and engineering responses.

10. A system for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns, used to implement the method for risk assessment and dynamic early warning of activation and instability of Ordovician limestone karst collapse columns as described in any one of claims 1-9, characterized in that, The system includes: The exploration data integration and modeling module is used to import and process various geophysical exploration and drilling data to construct a priori three-dimensional structural and hydrological model of the target collapse column. The risk calculation engine module, which embeds a risk driving function that couples mining and water pressure, can dynamically calculate the activation instability risk index based on prior models and real-time monitoring data streams. The multi-source monitoring data fusion and processing module is used to access, align, and preprocess real-time data streams from foundation damage monitoring, groundwater chemical and potential monitoring, and Ordos limestone water pressure monitoring. The adaptive early warning decision module is used to calculate the dynamic early warning threshold in real time based on the risk index sequence, and generate corresponding early warning signals and handling strategy suggestions according to the risk level. The instruction execution and closed-loop feedback module is used to convert early warning strategies into specific control instructions or operational suggestions and send them to relevant systems, and to track and record the execution and effects of the handling process; The above modules exchange data and coordinate commands through the data center bus.

Images