Intelligent early warning method and system for stope roof risk based on uncertainty fusion model

Abstract

The invention discloses a stope roof risk intelligent early warning method and system based on an uncertainty fusion model, and the method comprises the steps: determining static and dynamic indexes needed for judgment, and building a risk grade judgment standard; dividing the corresponding relation between the data range of each index and the grade division and the top plate safety risk grade; adopting a dynamic data acquisition scheme to obtain static and dynamic indexes, and performing defuzzification and standardization processing on the data; data features are extracted by integrating various subjective and objective weight determination methods, and a comprehensive index weight and a redistribution weight under the condition of incomplete indexes are obtained; judging the safety risk level of the roof through a maximum membership degree principle to obtain a risk tendency judgment result and visualize the risk tendency judgment result; performing linkage early warning based on the judgment result; the system comprises an acquisition module, a weight management module, an evaluation standard module, an arithmetic logic module, a risk level evaluation module, an early warning module, a visualization module and a dynamic optimization module. According to the invention, accurate identification and effective early warning of the roof risk can be realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent early warning technology for mine roof safety, and specifically provides a method and system for intelligent early warning of mine roof risks based on an uncertainty fusion model. Background Art

[0002] China boasts a wide variety of mineral resources and abundant reserves. However, as mining progresses, the availability of mineral resources suitable for shallow open-pit mining has become increasingly scarce, leading to a significant increase in the proportion of underground mines. During production activities in underground mines, machinery, equipment, and personnel are exposed beneath the roof, which also provides communication with the strata above. Therefore, roof safety risks not only endanger personnel and equipment below but can also further induce surface collapse and connect water-conducting fissures above, potentially leading to mine water inrush. Roof safety risks are volatile and sudden, making early warning of mine roof safety risks throughout their lifecycle crucial for ensuring safe mine operations. However, current roof safety risk management in underground mines relies primarily on extensive redundant design by designers during the design phase, coupled with enhanced support and empirical inspections by technical personnel during production. This results in an overly reactive roof management approach that fails to effectively identify and warn of roof safety risks in a timely and effective manner. Summary of the Invention

[0003] To address the problems of the above-mentioned prior art, the present invention provides a method and system for intelligent warning of mine roof risk based on an uncertainty fusion model. This method is simple to implement and highly intelligent. It can efficiently and accurately assess the roof risk level and perform coordinated warning actions according to different roof risk levels. This is extremely valuable for ensuring safe production operations in mines. The system has a simple structure and a high degree of intelligence. It can intelligently and accurately identify the risk level of mine roofs and provide dynamic warnings, effectively improving the safety management level of mine roofs.

[0004] In order to achieve the above object, the present invention provides an intelligent early warning method for stope roof risk based on uncertainty fusion model, comprising the following steps: Step 1: Based on the requirements of underground mine engineering, determine the static and dynamic indicator data required for roof safety risk assessment and establish the roof safety risk level assessment criteria; Step 2: Divide the data range and scoring range corresponding to the indicators at different safety levels based on numerical simulation and engineering standards; Step 3: Based on data requirements, a dynamic data collection solution is used to obtain static and dynamic indicator data, and the obtained data is defuzzified and standardized; Step 4: Based on data-driven, a variety of subjective and objective weight determination methods are integrated to extract data features, obtain comprehensive indicator weights, and redistribute weights under incomplete indicator conditions; Step 5: Based on the current optimal weight, the roof safety risk level is evaluated using the maximum membership principle, and the risk tendency judgment result is obtained and visualized through the risk assessment basic cloud model; Step 6: Based on the risk threshold set by the evaluation results, the early warning equipment is automatically activated to provide synchronous sound, light, and electricity warnings and automatically report the warning information; Step 7: By collecting roof risk warning information and forming a comparison database with actual on-site information, dynamically optimize the indicator weights and modify the basic cloud model for risk assessment.

[0005] As a preferred embodiment, in step one, the static indicator data include roof rock grade, stress field data, mining technology, and historical accident conditions; the dynamic indicator data include roof displacement data, stress data, and crack conditions; the roof safety risk levels include Level I roof risk, Level II roof risk, Level III roof risk, Level IV roof risk, and Level V roof risk; wherein, the scoring range of Level I roof risk is that the roof safety risk is greater than or equal to 81 and less than or equal to 100, the scoring range of Level II roof risk is that the roof safety risk is greater than or equal to 61 and less than or equal to 80, the scoring range of Level III roof risk is that the roof safety risk is greater than or equal to 41 and less than or equal to 60, the scoring range of Level IV roof risk is that the roof safety risk is greater than or equal to 21 and less than or equal to 40, and the scoring range of Level V roof risk is that the roof safety risk is greater than or equal to 0 and less than or equal to 20.

[0006] As a preferred embodiment, the corresponding relationship in step 2 is as follows: For the roof rock mass grade, it is divided into five levels according to the rock mass stability: extremely stable, stable, moderate, unstable and extremely unstable, and corresponds to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, extremely stable corresponds to 81≤rock mass stability≤100, stable corresponds to 61≤rock mass stability≤80, moderate corresponds to 41≤rock mass stability≤60, unstable corresponds to 21≤rock mass stability≤40, and extremely unstable corresponds to 0≤rock mass stability≤20; For stress field data, according to the magnitude of ground stress, it is divided into five levels: no risk, low risk, potential risk, high risk and extremely high risk, which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, no risk corresponds to ground stress <5Mpa, low risk corresponds to 5Mpa≤ground stress <10Mpa, potential risk corresponds to 10Mpa≤ground stress <15Mpa, high risk corresponds to 15Mpa≤ground stress <20Mpa, and extremely high risk corresponds to ground stress >20Mpa; For mining technology, it is divided into five levels according to the disturbance force of the mining method, namely extremely low disturbance, low disturbance, medium disturbance, high disturbance and relatively high disturbance, which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, the mining methods corresponding to extremely low disturbance are bench mining method, residual ore recovery method and shallow stratification method, the mining methods corresponding to low disturbance are stratification mining method, shallow room-and-pillar method and strip method, the mining methods corresponding to medium disturbance are room-and-pillar method, longwall mining method, sublevel caving method and room-and-pillar method with ore retention zone, the mining methods corresponding to high disturbance are sublevel filling mining method, stage open-pit method, stage filling method and stage caving method, and the mining methods corresponding to relatively high disturbance are deep full stoping method, large stope unsupported caving method, block caving method, continuous stoping method and full-section mining; For historical accident situations, they are divided into five levels: safe, low risk, medium risk, high risk and extremely high risk according to the average annual number of roof accidents, and correspond to Level I roof risk, Level II roof risk, Level III roof risk, Level IV roof risk and Level V roof risk respectively. Among them, safe corresponds to no mine roof accidents, low risk corresponds to 0 times < mine annual average roof accidents ≤ 1 time, medium risk corresponds to 1 time < mine annual average roof accidents ≤ 2 times, high risk corresponds to 2 times < mine annual average roof accidents ≤ 3 times, and extremely high risk corresponds to mine annual average roof accidents > 3 times; For roof displacement data, it is divided into five levels according to the roof displacement situation: low risk, relatively low risk, medium risk, high risk and extremely high risk, which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, low risk corresponds to 0≤ daily average displacement rate≤1mm / d, and 0≤ cumulative displacement≤10mm, relatively low risk corresponds to 1 mm / d< daily average displacement rate≤3mm / d, and 10 mm< cumulative displacement≤20mm, medium risk corresponds to 3mm / d< daily average displacement rate≤5mm / d, and 20 mm< cumulative displacement≤40mm, high risk corresponds to 5mm / d< daily average displacement rate≤8mm / d, and 40 mm< cumulative displacement≤60mm, and extremely high risk corresponds to daily average displacement rate>8mm / d, and cumulative displacement>60mm; For stress data, it is divided into five levels according to the stress conditions of the support / column, namely stable pressure, appropriate pressure, leakage pressure, heavy pressure and overpressure, and corresponds to Level I roof risk, Level II roof risk, Level III roof risk, Level IV roof risk and Level V roof risk respectively. Among them, stable pressure corresponds to 30% design bearing capacity ≤ support / column force < 50% design bearing capacity, appropriate pressure corresponds to 50% design bearing capacity ≤ support / column force < 70% design bearing capacity, leakage pressure corresponds to support / column force < 30% design bearing capacity, heavy pressure corresponds to 70% design bearing capacity ≤ support / column force < 90% design bearing capacity, and overpressure corresponds to 70% design bearing capacity ≥ 90% design bearing capacity; For the crack situation, they are divided into five levels according to the crack size, namely no cracks, fine cracks, stable cracks, expanding cracks and through cracks, which correspond to Level I roof risk, Level II roof risk, Level III roof risk, Level IV roof risk and Level V roof risk respectively. Among them, no cracks correspond to no obvious cracks and no cracks, fine cracks correspond to the existence of small cracks, and the crack width is less than 1mm, stable cracks correspond to the existence of non-penetrating cracks, and the crack width is 1mm≤<3mm, expanding cracks correspond to the existence of non-penetrating cracks, and the crack width is 3mm≤<5mm, and through cracks correspond to the existence of through cracks, and the crack width is ≥5mm.

[0007] As a preferred embodiment, in step three, for static indicator data, technical and design personnel obtain the data through simulation calculation based on the mine geological report and input it into the data processing terminal through manual input. For dynamic indicator data, monitoring equipment arranged at the mine site is used to collect the data in real time and automatically, thereby obtaining static indicator data and dynamic indicator data. Among them, the monitoring equipment adopts differentiated layout methods in coal mines and non-coal mines.

[0008] To ensure high-quality monitoring data, in step three, the specific process of defuzzification and standardization of the obtained data is as follows: S31; Introduce the rule mapping defuzzification processing mechanism in fuzzy set theory to convert qualitative indicators into quantitative data, and obtain the first i The corresponding level score of each indicator ; (1); Where, Indicates the corresponding indicator level; S32: For indicators with actual data, linear normalization is performed; for indicators with data interval , the score interval is , data value X exist When , the corresponding score interval is , according to formula (2), the score of the jth indicator after nonlinear normalization mapping is obtained ; (2). As a preferred embodiment, the specific process of step 4 is as follows: S41: At the subjective level, an expert evaluation matrix is ​​constructed based on the hierarchical analysis method to extract expert knowledge and experience; at the objective level, the entropy method and the coefficient of variation method are combined to quantify the importance of indicators from the perspectives of information distribution and volatility of indicators respectively; finally, a linear weighted fusion strategy is used to integrate the three methods to form a unified weight system, providing accurate input feature guarantee for the subsequent risk level identification model; specifically, the comprehensive weight is obtained according to formula (3): ; (3); Where, is the AHP weight; is the entropy method weight; is the weight of the coefficient of variation method; , represents the reliability coefficient of the AHP method; , is the reliability coefficient of the entropy method; is the reliability coefficient of the coefficient of variation method; ; S42: For a weight vector containing n indicators , , When m indicators are missing, the first indicator after redistribution is obtained according to formula (4). i The weight of the indicator , according to formula (5), we can obtain the weight vector containing k indicators ; (4) ; (5) ; Where, Before redistribution i The weight of each indicator; is the weight sum of the residual indicators before redistribution; k is the number of participating indicators, k=mn; .

[0009] As a preferred embodiment, the specific process of step five is as follows: S51: Real-time collection of static and dynamic indicator data for the assessed area, calculation of the final evaluation score based on the established roof safety risk level evaluation criteria, the corresponding relationship between the data range and level division of various indicators and the roof safety risk level, the comprehensive indicator weight, and the redistribution of weights under the condition of incomplete indicators, and calculation of the membership of the final evaluation score under each risk level, as shown in formula (6); calculation of the membership of each level for the data group at the corresponding moment, and determination of the maximum membership as the roof safety risk level of the corresponding area at the current moment; (6); Where, Indicates the final evaluation score i The degree of membership under risk level, express i The central value of membership under risk level, , x Indicates the final evaluation score; S52: Based on the established roof safety risk level evaluation criteria, a risk assessment basic cloud model corresponding to each risk level is constructed. Each risk assessment basic cloud model is defined by three parameters: expected value, entropy, and super entropy, forming a standard risk distribution cloud corresponding to different risk levels; S53: Based on the static indicator and dynamic indicator data collected in real time, the evaluation score is calculated and converted into the risk assessment basic cloud model parameters to form the current risk evaluation cloud; S54: Compare the risk assessment basic cloud model with the current risk judgment cloud to obtain the roof safety risk tendency judgment result to provide a basis for auxiliary decision-making.

[0010] In order to perform different levels of warning actions according to different risk levels, the specific process of step six is ​​as follows: When the roof safety risk reaches the set threshold score, the early warning equipment arranged in the working face and tunnel is controlled to execute the sound and light warning action and send it to the control detection end. Among them, when the level I roof risk occurs, the early warning equipment is controlled to keep the green light on and maintain normal power supply. At the same time, it generates early warning information for daily inspections and maintaining the current monitoring and support measures unchanged; when the level II roof risk occurs, the early warning equipment is controlled to keep the blue light on and activate the early warning system. At the same time, it generates early warning information for strengthening the inspection frequency and continuously observing the changes in monitoring data; when the level III roof risk occurs, the early warning equipment is controlled to keep the yellow light on, and a prompt voice is broadcast, the power supply of non-critical equipment is disconnected, and only important systems are retained. Power supply, at the same time, generate early warning information for adding monitoring points, conducting special inspections, and optimizing support measures; when there is a level IV roof risk, control the early warning equipment to flash with an orange light, and make continuous beeps and voice broadcasts, cut off power to the set area, and link the safety lighting and broadcast systems. At the same time, generate early warning information for starting the early warning response process, arranging personnel to evacuate relevant areas, and strengthening support structures; when there is a level V roof risk, control the early warning equipment to flash with a red light at a high frequency, continuously perform long beeps and voice broadcasts, cut off power completely, and stop power supply to all work areas. At the same time, generate early warning information for immediate implementation of personnel evacuation, suspension of operations, and implementation of emergency reinforcement measures.

[0011] In order to achieve adaptive update of the risk assessment model, the specific process of step seven is as follows: S71: Continuously collect warning data generated by roof risk warning, and simultaneously collect actual on-site monitoring data and disposal feedback information; S72: By building a comparison database of warning information and actual on-site data, the matching degree and deviation between the warning results and actual risk events are analyzed; S73: Based on the comparative analysis results, repeat steps 4 and 5 using a dynamic weight adjustment algorithm to correct the weight distribution of risk assessment indicators in real time, eliminate the impact of misleading indicators, and enhance the weight sensitivity of key indicators, thereby optimizing the performance of the risk assessment model.

[0012] In this invention, the multi-source data of static and dynamic indicators required for roof safety risk assessment is first determined based on underground mine engineering requirements, and a roof safety risk level assessment standard is established. Then, a corresponding relationship between the data range and risk score for each indicator corresponding to the risk level is established. This facilitates the coordinated analysis of static indicators (such as geological structure) and dynamic indicators (such as real-time stress changes) to form a more comprehensive multi-dimensional safety risk assessment framework, effectively ensuring the accurate assessment of roof risk levels. A dynamic data acquisition scheme is used to obtain static and dynamic indicator data, and qualitative indicators are quantified and analyzed through defuzzification. Standardization methods are used to improve the comparability of quantitative indicator data, effectively ensuring that the obtained multi-source indicator data accurately reflects roof status information. A subjective and objective weight fusion mechanism is used to extract data features, providing accurate input features for the subsequent risk level identification model. Simultaneously obtaining comprehensive indicator weights and redistributing weights under incomplete indicator conditions enables the method to support both comprehensive indicators and weight redistribution under incomplete indicator conditions, ensuring accurate and effective assessment of roof safety risks even when some indicator data is missing, effectively enhancing the adaptability of the indicator system. The risk level can be dynamically determined through the principle of maximum membership, and the risk tendency can be judged to accurately display the degree of risk deviation and development trend. Then, the visual mapping of the risk level can be achieved through visual display, which can intuitively present the risk level and dynamically present the risk situation in real time. Based on the evaluation results, the early warning equipment is automatically activated to perform synchronous sound, light, and electricity early warnings, which can realize the early warning linkage of sound, light, and electricity, ensuring that effective emergency response measures can be taken in a timely and effective manner when dangerous situations occur. By using the dynamic optimization indicator weights to correct the evaluation module, the risk assessment model can be adaptively updated through a closed-loop method of monitoring-early warning-feedback-optimization, ensuring that it has high prediction accuracy and stability under different working conditions and at different time nodes.

[0013] The implementation process of this method is simple and highly intelligent. It can effectively and accurately evaluate the roof risk level and perform corresponding linkage early warning actions according to different roof risk levels. It is of great value in ensuring safe production operations in mines.

[0014] The present invention also provides an intelligent early warning method for stope roof risk based on an uncertainty fusion model, which is used to implement an intelligent early warning method for stope roof risk based on an uncertainty fusion model, including an acquisition module, a weight management module, an evaluation standard module, an operation logic module, a risk level evaluation module, an early warning module, a visualization module and a dynamic optimization module; The acquisition module is used to collect multi-source monitoring data in the area to be evaluated, and record the spatial layout information and data acquisition frequency of the monitoring equipment to form an original data set, and send the original data set to the weight management module and the risk level assessment module; The weight management module is used to extract data set features based on the volatility of the acquired data, the correlation between indicators, and historical evaluation performance, initialize the weight vector of each evaluation indicator, support the allocation of incomplete indicator weights and dynamic weight adjustment based on subsequent model feedback, and output the revised indicator weight set to the risk level assessment module; The evaluation criteria module is used to construct a multi-level risk rating system. Based on expert experience, policies and regulations, historical cases and numerical analysis results, the boundary thresholds and judgment logic of each level are defined to form an underground mine roof risk rating evaluation standard set and send it to the risk rating evaluation module; The operation logic module is used to perform normalization and defuzzification processing on the collected data information, and calculate the relevant parameters and results of the processed data through fuzzy comprehensive evaluation, cloud model or multi-factor fusion algorithm, and send them to the risk level assessment module; The risk level evaluation module is used to receive the indicator data collected by the acquisition module, the indicator weights output by the weight management module, the quantitative analysis results output by the operation logic module, and the underground mine roof risk level evaluation standard set output by the evaluation standard module, and classify the risk levels according to the maximum membership principle and the evaluation standard module, and send the risk level classification results to the early warning module and the visualization module; The early warning module is used to execute the corresponding level of sound, light and electricity early warning measures according to the risk level output by the risk level assessment module; The visualization module is used to graphically display real-time assessment results, historical trend evolution, risk area distribution and warning status, and is used to provide a human-computer interaction interface, supporting two-dimensional and three-dimensional risk maps, time series diagrams, risk cloud maps and console operation interfaces; The dynamic optimization module is used to collect deviation information between the risk level assessment results and actual on-site feedback, build an assessment-measurement comparison database, and dynamically correct the indicator weights based on the comparison database. At the same time, the corrected indicator weights are sent to the weight management module to achieve adaptive optimization of the assessment model.

[0015] The system has a simple structure and a high degree of intelligence. It can intelligently and accurately identify the risk level of the mine roof and provide dynamic early warning, effectively improving the safety management level of the mine roof. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a workflow diagram of the intelligent early warning method of the present invention; Figure 2 Schematic diagram of the arrangement of monitoring equipment in a coal mine working face according to the present invention; Figure 3 Schematic diagram of the arrangement of monitoring equipment in a non-coal working face of the present invention; Figure 4 Schematic diagram of the arrangement of monitoring equipment in the main lanes of the present invention; Figure 5 This is a visual standard risk distribution cloud map in the present invention; Figure 6 This is a cloud diagram for visually judging risk distribution in the present invention; Figure 7 This is the risk tendency comparison cloud map in the present invention; Figure 8 It is a structural diagram of the intelligent early warning system in the present invention.

[0017] In the figure: 201, working face roof 1; 202, support stress sensor; 203, hydraulic support; 204, working face; 205, scraper conveyor; 206, working face tunnel communication and amplification device 1; 207, hydraulic support telescopic column; 208, goaf; 301. Ore blast pile; 302. Working face roof (II); 303. Working face support stress sensor; 304. Working face hydraulic support; 305. Working face tunnel communication and amplification device (II); 401, tunnel roof; 402, support anchor; 403, pillar stress sensor; 404, anchor stress sensor; 405, anchor tray; 406, displacement sensor; 407, roof monitoring equipment; 408, tunnel communication and amplification device; 409, tunnel hydraulic support; 410, tunnel side; 411, tunnel floor; 501, V-level evaluation cloud; 502, IV-level evaluation cloud; 503, III-level risk evaluation cloud; 502, II-level evaluation cloud; 503, I-level risk evaluation cloud; 701. Level V evaluation cloud; 702. Level IV evaluation cloud; 703. Level III risk evaluation cloud; 704. Risk evaluation cloud; 706. Level II evaluation cloud; 707. Level I risk evaluation cloud. DETAILED DESCRIPTION

[0018] The present invention will be further described below with reference to the accompanying drawings.

[0019] An embodiment of the present invention provides a method and system for intelligent early warning of stope roof risk based on an uncertainty fusion model. The method and system can be implemented by an electronic device, which can be a control terminal or a server. Figure 1 A workflow diagram of an intelligent early warning method for stope roof risk based on an uncertainty fusion model is given. The processing flow of this method may include the following steps: Step 1: Based on the requirements of underground mine engineering, determine the static and dynamic indicator data required for roof safety risk assessment and establish the roof safety risk level assessment criteria; Through the massive research process of current mine roof accidents, the recommended data indicators are determined. According to the data changes, the data are determined as static indicator data and dynamic indicator data. Among them, the static indicator data include roof rock grade, stress field data, mining technology, and historical accident conditions. The dynamic indicator data include roof displacement data, stress data, and crack conditions. According to the degree of danger to personnel and equipment caused by the corresponding indicators, the roof safety risk level is divided into five levels. The meaning and score corresponding to the evaluation level are shown in Table 1; specifically, the roof safety risk level includes level I roof risk, level II roof risk, level III roof risk, Level IV roof risk and Level V roof risk, among which, the scoring range of Level I roof risk is that the roof safety risk is greater than or equal to 81 and less than or equal to 100, the scoring range of Level II roof risk is that the roof safety risk is greater than or equal to 61 and less than or equal to 80, the scoring range of Level III roof risk is that the roof safety risk is greater than or equal to 41 and less than or equal to 60, the scoring range of Level IV roof risk is that the roof safety risk is greater than or equal to 21 and less than or equal to 40, and the scoring range of Level V roof risk is that the roof safety risk is greater than or equal to 0 and less than or equal to 20.

[0020] The explanations and recommended measures corresponding to the five levels are as follows: Level I roof risk indicates extremely low risk (safety level), and daily inspections are recommended to maintain current measures; Level II roof risk indicates a lower risk (controllable level), and it is recommended to strengthen inspections and pay attention to dynamic changes in monitoring data; Level III roof risk indicates a moderate risk (level of concern), and it is recommended to add monitoring points, initiate special inspections, and optimize support measures when necessary; Level IV roof risk indicates a high risk (dangerous level). It is recommended to initiate risk warning procedures, arrange for personnel to evacuate the risk area, and strengthen support. Level V roof risk indicates extremely high risk (danger level), and it is recommended to take emergency measures, evacuate personnel immediately, suspend operations, and implement emergency reinforcement measures.

[0021] Step 2: Based on numerical simulation and engineering standards, the corresponding data range and scoring range of indicators at different safety levels are divided; the data range of evaluation indicators such as roof rock mass grade, stress field data, mining technology, historical accident conditions, roof displacement, stress, and crack conditions are matched with the grade division and scoring standards; The roof rock mass grade index, rock mass quality rating index score and roof rock mass safety risk score can be directly used in a one-to-one correspondence. The corresponding safety risk assessment score of the underground mine roof rock mass based on the rock mass quality rating score is shown in Table 2: For the roof rock mass grade, it is divided into five grades according to the rock mass stability: extremely stable (grade I), stable (grade II), moderate (grade III), unstable (grade IV) and extremely unstable (grade V), which correspond to grade I roof risk, grade II roof risk, grade III roof risk, grade IV roof risk and grade V roof risk respectively. Among them, extremely stable corresponds to 81≤rock mass stability≤100, stable corresponds to 61≤rock mass stability≤80, moderate corresponds to 41≤rock mass stability≤60, unstable corresponds to 21≤rock mass stability≤40, and extremely unstable corresponds to 0≤rock mass stability≤20; The stress field data indicators are graded and scored based on the magnitude of mine ground stress and numerical simulation methods. The classification results are as follows: The stress field data are divided into five levels according to the magnitude of the ground stress, namely no risk (level I), relatively low risk (level II), potential risk (level III), relatively high risk (level IV) and extremely high risk (level V), which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. No risk corresponds to ground stress < 5 MPa, indicating that there is no significant safety risk to the roof; relatively low risk corresponds to ground stress 5 MPa ≤ < 10 MPa, indicating that there is a relatively low safety risk to the roof; potential risk corresponds to ground stress 10 MPa ≤ < 15 MPa, indicating that there is a potential safety risk to the roof; relatively high risk corresponds to ground stress 15 MPa ≤ < 20 MPa, indicating that there is a relatively high safety risk to the roof; extremely high risk corresponds to ground stress > 20 MPa, indicating that there is an extremely high safety risk to the roof; quantitative values ​​are assigned to the five ground stress ranges, as shown in Table 3: Mining process indicators, based on the relevant specifications of the "Safety Regulations for Metal and Non-metal Underground Mines" and the "Safety Regulations for Coal Mines", divide the mining methods currently used in underground mines into five levels. The classification results are as follows: For mining technology, it is divided into five levels according to the disturbance force of the mining method, namely extremely low disturbance (level I), low disturbance (level II), medium disturbance (level III), high disturbance (level IV) and relatively high disturbance (level V), which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, the extremely low disturbance ground pressure activity is weak and the disturbance to the roof is small. The corresponding mining methods are step mining method, residual ore recovery method and shallow stratification method. The low disturbance has slight stress concentration and the roof is relatively stable. The corresponding mining methods are layered mining method. , shallow room-and-pillar method and strip method, medium disturbance ground stress is obvious, moderate disturbance to the roof, its corresponding mining methods are room-and-pillar method, longwall mining method, sublevel caving method, room-and-pillar method with ore retention, high disturbance rock mass is broken, large goaf has strong disturbance to the roof, its corresponding mining methods are sublevel filling mining method, stage empty field method, stage filling method, stage caving method, high disturbance has extremely strong disturbance to the roof, its corresponding mining methods are deep full stoping method, large stope unsupported caving method, block caving method, continuous stoping method, full-section mining; quantitative points are given to the five levels, as shown in Table 4: The historical accident situation index is based on the "Mine Safety Accident Classification Standard", which divides the historical roof safety accident situations of mines into 5 levels: For historical accidents, they are divided into five levels according to the average annual number of roof accidents: safe (level I), low risk (level II), medium risk (level III), high risk (level IV) and extremely high risk (level V), which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, safe corresponds to no mine roof accidents, and the mine has no roof accident records; low risk corresponds to 0 times < mine annual average roof accident ≤ 1 time, and the mine has had roof accidents due to Roof accidents are small-scale sporadic accidents with no casualties. Medium risk corresponds to 1 roof accident < the average annual number of roof accidents in the mine ≤ 2 times, and the mine has had small-scale injuries or dangerous situations due to roof accidents. High risk corresponds to 2 roof accidents < the average annual number of roof accidents in the mine ≤ 3 times, and the mine has had personnel injuries or equipment damage due to roof accidents. Extremely high risk corresponds to 3 roof accidents > the average annual number of roof accidents in the mine, and the mine has had multiple casualties or major equipment damage due to roof accidents. Explanations and quantitative scores are given for the five levels, as shown in Table 5: The roof displacement index is based on the deformation response law of the roof rock layer under mining disturbance. The impact of the roof displacement on the roof safety risk is divided into 5 levels, as shown in Table 6: For roof displacement data, according to the roof displacement situation, it is divided into five levels: low risk (level I), relatively low risk (level II), medium risk (level III), high risk (level IV) and extremely high risk (level V), which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, low risk corresponds to 0≤ daily average displacement rate≤1mm / d, and 0≤ cumulative displacement≤10mm, relatively low risk corresponds to 1 mm / d< daily average displacement rate≤3mm / d, and 10 mm< cumulative displacement≤20mm, medium risk corresponds to 3mm / d< daily average displacement rate≤5mm / d, and 20mm< cumulative displacement≤40mm, high risk corresponds to 5mm / d< daily average displacement rate≤8mm / d, and 40 mm< cumulative displacement≤60mm, and extremely high risk corresponds to daily average displacement rate>8mm / d, and cumulative displacement>60mm; Stress index, based on the monitoring stress and the design bearing capacity of the support, the safety risk level of the roof is divided into: For stress data, it is divided into five levels according to the stress conditions of the support / column: stable pressure (level I), appropriate pressure (level II), leakage pressure (level III), heavy pressure (level IV) and overpressure (level V), which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, stable pressure refers to the state where the support / column stress is within the initial support force range, which corresponds to 30% of the design bearing capacity ≤ support / column stress (stress monitored by monitoring equipment) < 50% of the design bearing capacity, and appropriate pressure refers to the state where the support / column is subjected to The influence of certain stress concentration corresponds to 50% design bearing capacity ≤ support / column force < 70% design bearing capacity. Leakage pressure is an abnormal situation where the support / column force is far lower than the design bearing capacity, corresponding to support / column force < 30% design bearing capacity. Heavy pressure is a state where the support / column force is close to the limit, corresponding to 70% design bearing capacity ≤ support / column force < 90% design bearing capacity. Overpressure is a state where the support force reaches the critical value, corresponding to 70% design bearing capacity ≥ 90% design bearing capacity. The explanation and scoring of the corresponding levels are shown in Table 7: The crack condition indicators are based on image recognition (using high-definition image acquisition equipment for image acquisition and recognition) and crack measurement, and are divided into 5 levels according to their impact on roof safety risks: For crack conditions, they are divided into five levels according to the size of the cracks: no cracks (level I), fine cracks (level II), stable cracks (level III), expanding cracks (level IV) and through cracks (level V), which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, no cracks correspond to no obvious cracks or no cracks, fine cracks correspond to the presence of small cracks, stable cracks, slow expansion, and crack width <1mm, stable cracks correspond to the presence of cracks that do not penetrate, that is, there are some cracks, the cracks expand but do not penetrate, and the crack width is 1mm≤<3mm, expanding cracks correspond to the presence of cracks that do not penetrate, that is, there are obvious cracks, the crack expansion is accelerated, and the crack width is 3mm≤<5mm, through cracks correspond to the presence of through cracks, cracks are obviously penetrated, and the crack width is ≥5mm. The explanation and scoring of the cracks corresponding to the roof risk level are shown in Table 8: Step 3: Based on data requirements, a dynamic data collection solution is used to obtain static and dynamic indicator data, and the obtained data is defuzzified and standardized; As a preferred method, for static indicator data (roof rock mass grade, stress field data, mining technology, historical accident conditions), technical and design personnel obtain them through simulation calculations based on mine geological reports and input them into the data processing terminal through manual input. For dynamic indicator data (roof displacement, stress, crack conditions), monitoring equipment arranged at the mine site is used to collect them in real time and automatically, thereby obtaining static indicator data and dynamic indicator data. Among them, the monitoring equipment adopts differentiated layout methods in coal mines and non-coal mines.

[0022] Based on the supplementary transformation of existing underground mine equipment, the equipment layout is divided into two categories: underground mine working face layout and main tunnel layout.

[0023] In the coal mine working face, the support pressure sensor 202 is set above the hydraulic support 203 to monitor the stress of the working face roof in real time. The working face tunnel communication and amplification device 206 is set on the support telescopic column 207 for roof early warning. Figure 2 The layout of underground coal mine working face equipment is shown, including but not limited to: A support pressure sensor 202 located above a hydraulic support 203 between the working face 204 and the goaf 208 monitors stress in the working face roof 201 above. A working face tunnel communication and amplification device 206, located on the hydraulic support telescopic column 207, provides early warning of roof safety risks and protects key working face equipment such as the scraper conveyor 205.

[0024] In non-coal mines, the working face support stress sensor 303 is set between the hydraulic support 304 and the roof to monitor the impact of ore explosion on the roof. The tunnel communication and amplification device 305 is also used for early warning. Figure 3 The layout of some equipment in an underground non-coal mining working face is shown, including but not limited to: Working face support stress sensors 303, also located between the working face roof 302 and the working face hydraulic support 304, are used to monitor the impact of the ore burst 301 formation process on the roof. Working face tunnel communication and amplification devices 305, fixed to the working face hydraulic support, are used to warn of roof safety risks.

[0025] In the main tunnel, the pillar stress sensor 403 and the anchor stress sensor 404 jointly monitor the roof stress, the displacement sensor 406 is used to monitor the roof displacement, the crack monitoring device (which can use a high-definition camera) is used to track the development of roof cracks, and the sound amplification device sends out sound and light warning signals; Figure 4 The layout of the main tunnel monitoring equipment in an underground mine is shown, including but not limited to: tunnel hydraulic struts 409 installed between the tunnel floor 411 and the tunnel roof 401. A strut stress sensor 403 located above the hydraulic struts 409 and an anchor stress sensor 404 placed between the anchor support tray 405 of the support anchor 402 and the tunnel roof 401 jointly monitor roof stress. Displacement sensors 406 located on the tunnel side 410 monitor roof displacement; roof monitoring equipment 407 monitors roof cracks; and tunnel communication and loudspeaker systems 408 provide audible and visual alarms to indicate roof safety risks.

[0026] In order to ensure high-quality monitoring data, the specific process of defuzzification and standardization of the obtained data is as follows: S31; In order to realize the quantitative expression of qualitative indicators, the rule mapping defuzzification processing mechanism in fuzzy set theory is introduced to convert qualitative indicators (mining technology, historical accident situation) into quantitative data. According to formula (1), the first i The corresponding level score of each indicator ; (1); Where, Indicates the corresponding indicator level, corresponding to level I, level II, level III, level IV, and level V respectively; Among them, the quantitative relationship corresponding to the five risk levels of the roof is shown in Table 9: S32: Standardization is based on the data range of its historical data, corresponding to four levels and 0-100 scores. For indicators with actual data (roof rock mass grade, stress field data, roof displacement data, stress data, crack conditions), linear standardization is performed; for indicators with data range of , the score interval is , data value X exist When , the corresponding score interval is , according to formula (2), the score of the jth indicator after nonlinear normalization mapping is obtained ; (2).

[0027] Step 4: Based on data-driven, a variety of subjective and objective weight determination methods are integrated to extract data features, obtain comprehensive indicator weights, and redistribute weights under incomplete indicator conditions; As a preferred embodiment, the specific process is as follows: S41: Based on the established evaluation criteria and data defuzzification and standardization criteria. On the subjective level, an expert evaluation matrix is ​​constructed based on the analytic hierarchy process (AHP) to extract expert knowledge and experience; on the objective level, the entropy value method (EWM) and the coefficient of variation method (VCM) are combined to quantify the importance of indicators from the perspectives of information distribution and volatility of indicators respectively; finally, through the linear weighted fusion strategy, the three methods are integrated to form a unified weight system to provide accurate input feature guarantee for the subsequent risk level identification model; specifically, the comprehensive weight is obtained according to formula (3) ; (3); Where, is the AHP weight; is the entropy method weight; is the weight of the coefficient of variation method; , represents the reliability coefficient of the AHP method; , is the reliability coefficient of the entropy method; is the reliability coefficient of the coefficient of variation method; ; As a preferred embodiment, the reliability coefficient values ​​are shown in Table 10: S42: Considering that some evaluation indicators may be affected by monitoring failure, data loss or unavailability in actual applications, the present invention proposes an adaptive weight redistribution mechanism to ensure the versatility of the model. , , When m indicators are missing, the first indicator after redistribution is obtained according to formula (4).i The weight of the indicator , according to formula (5), we can obtain the weight vector containing k indicators ; (4) ; (5) ; Where, Before redistribution i The weight of each indicator; is the weight sum of the residual indicators before redistribution; k is the number of participating indicators, k=mn; .

[0028] Step 5: Based on the current optimal weight, the roof safety risk level is evaluated using the maximum membership principle, and the risk tendency judgment result is obtained and visualized through the risk assessment basic cloud model; As a preferred embodiment, the specific process is as follows: S51: Use the maximum membership principle to judge the roof risk level: The static and dynamic indicator data of the assessed area are collected in real time. The final evaluation score is calculated based on the established roof safety risk level evaluation criteria, the corresponding relationship between the data range and level division of various indicators and the roof safety risk level, the comprehensive indicator weight, and the redistribution of weights under the condition of incomplete indicators. The membership of the final evaluation score under each risk level is calculated, as shown in formula (6). The membership of the data group at the corresponding moment to each level is calculated, and the maximum membership is determined as the roof safety risk level of the corresponding area at the current moment. (6); Where, Indicates the final evaluation score i The degree of membership under risk level, express i The central value of membership under risk level, , x Indicates the final evaluation score; S52: Visualize roof risk based on cloud model; Based on the established roof safety risk level evaluation standard, construct the risk assessment basic cloud model corresponding to each risk level. Each risk assessment basic cloud model is defined by three parameters: expected value (Ex), entropy (En) and super entropy (He), forming a standard risk distribution cloud corresponding to different risk levels (level I to level V), such as Figure 5 As shown; S53: Risk evaluation cloud; based on the static and dynamic indicator data collected in real time, the evaluation score is calculated and converted into the risk assessment basic cloud model parameters (Ex, En, He) to form the current risk evaluation cloud, such as Figure 6As shown; S54: Comparison and risk tendency judgment: Compare the risk assessment basic cloud model with the current risk assessment cloud to obtain the roof safety risk tendency judgment result to provide a basis for auxiliary decision-making, such as Figure 7 shown.

[0029] Step 6: Based on the risk threshold set by the evaluation results, the early warning equipment is automatically activated to provide synchronous sound, light, and electricity warnings and automatically report the warning information; the specific process is as follows: When the roof safety risk reaches the set threshold score, the early warning equipment arranged in the working face and tunnel is controlled to execute the sound and light warning action and send it to the control detection end. The sound and light warning measures of the corresponding levels are shown in Table 11; among them, when the level I roof risk occurs, the early warning equipment is controlled to keep the green light on and maintain normal power supply. At the same time, it generates daily inspections and maintains the current monitoring and support measures. Early warning information; when the level II roof risk occurs, the early warning equipment is controlled to keep the blue light on and activate the early warning system. At the same time, it generates early warning information to strengthen the inspection frequency and continuously observe the changes in monitoring data; when the level III roof risk occurs, the early warning equipment is controlled to keep the yellow light on, and a prompt voice is broadcast, disconnecting non-critical equipment Power supply, only retain the power supply of important systems. At the same time, generate early warning information for adding monitoring points, conducting special inspections, and optimizing support measures; when there is a level IV roof risk, control the early warning equipment to flash with orange lights, and make continuous beeps and voice prompts, cut off power to the set area, and link the safety lighting and broadcast systems. At the same time, generate early warning information to start the early warning response process, arrange for personnel to evacuate related areas, and strengthen the support structure; when there is a level V roof risk, control the early warning equipment to flash with red lights at high frequencies, continue to perform long beeps and voice broadcasts, cut off power completely, and stop the power supply to all work areas. At the same time, generate early warning information for immediate implementation of personnel evacuation, suspension of operations, and implementation of emergency reinforcement measures. Step 7: By collecting roof risk warning information and comparing it with actual on-site information to form a database, dynamically optimize the indicator weights and modify the basic cloud model for risk assessment. The specific process is as follows: S71: Continuously collect warning data generated by roof risk warning, and simultaneously collect actual on-site monitoring data and disposal feedback information; S72: By building a comparison database of warning information and actual on-site data, the matching degree and deviation between the warning results and actual risk events are analyzed; S73: Based on the comparative analysis results, repeat steps 4 and 5 using a dynamic weight adjustment algorithm to correct the weight distribution of risk assessment indicators in real time, eliminate the impact of misleading indicators, and enhance the weight sensitivity of key indicators, thereby optimizing the performance of the risk assessment model.

[0030] This process iterates and continuously refines and calibrates model parameters, enabling adaptive updates of the risk assessment model. This ensures high predictive accuracy and stability across different operating conditions and timeframes. Furthermore, the dynamic optimization process balances the model's generalization capabilities with its local adaptability, enhancing its responsiveness to changing risks in complex mine rooftop environments.

[0031] In this invention, the multi-source data of static and dynamic indicators required for roof safety risk assessment is first determined based on underground mine engineering requirements, and a roof safety risk level assessment standard is established. Then, a corresponding relationship between the data range and risk score for each indicator corresponding to the risk level is established. This facilitates the coordinated analysis of static indicators (such as geological structure) and dynamic indicators (such as real-time stress changes) to form a more comprehensive multi-dimensional safety risk assessment framework, effectively ensuring the accurate assessment of roof risk levels. A dynamic data acquisition scheme is used to obtain static and dynamic indicator data, and qualitative indicators are quantified and analyzed through defuzzification. Standardization methods are used to improve the comparability of quantitative indicator data, effectively ensuring that the obtained multi-source indicator data accurately reflects roof status information. A subjective and objective weight fusion mechanism is used to extract data features, providing accurate input features for the subsequent risk level identification model. Simultaneously obtaining comprehensive indicator weights and redistributing weights under incomplete indicator conditions enables the method to support both comprehensive indicators and weight redistribution under incomplete indicator conditions, ensuring accurate and effective assessment of roof safety risks even when some indicator data is missing, effectively enhancing the adaptability of the indicator system. The risk level can be dynamically determined through the principle of maximum membership, and the risk tendency can be judged to accurately display the degree of risk deviation and development trend. Then, the visual mapping of the risk level can be achieved through visual display, which can intuitively present the risk level and dynamically present the risk situation in real time. Based on the evaluation results, the early warning equipment is automatically activated to perform synchronous sound, light, and electricity early warnings, which can realize the early warning linkage of sound, light, and electricity, ensuring that effective emergency response measures can be taken in a timely and effective manner when dangerous situations occur. By using the dynamic optimization indicator weights to correct the evaluation module, the risk assessment model can be adaptively updated through a closed-loop method of monitoring-early warning-feedback-optimization, ensuring that it has high prediction accuracy and stability under different working conditions and at different time nodes.

[0032] The implementation process of this method is simple and highly intelligent. It can effectively and accurately evaluate the roof risk level and perform corresponding linkage early warning actions according to different roof risk levels. It is of great value in ensuring safe production operations in mines.

[0033] like Figure 8As shown, the present invention also provides an intelligent early warning method for stope roof risk based on an uncertainty fusion model, which is used to implement an intelligent early warning method for stope roof risk based on an uncertainty fusion model, including an acquisition module, a weight management module, an evaluation standard module, an operation logic module, a risk level evaluation module, an early warning module, a visualization module and a dynamic optimization module; The acquisition module is used to collect multi-source monitoring data in the area to be evaluated, and record the spatial layout information and data acquisition frequency of the monitoring equipment to form an original data set, and send the original data set to the weight management module and the risk level assessment module; The weight management module is used to extract data set features based on the volatility of the acquired data, the correlation between indicators, and historical evaluation performance, initialize the weight vector of each evaluation indicator, support the allocation of incomplete indicator weights and dynamic weight adjustment based on subsequent model feedback, and output the revised indicator weight set to the risk level assessment module; The evaluation criteria module is used to construct a multi-level risk rating system. Based on expert experience, policies and regulations, historical cases and numerical analysis results, the boundary thresholds and judgment logic of each level are defined to form an underground mine roof risk rating evaluation standard set and send it to the risk rating evaluation module; The operation logic module is used to perform normalization and defuzzification processing on the collected data information, and calculate the relevant parameters and results of the processed data through fuzzy comprehensive evaluation, cloud model or multi-factor fusion algorithm, and send them to the risk level assessment module to further support the efficient operation of the assessment module; The risk level evaluation module is used to receive the indicator data collected by the acquisition module, the indicator weights output by the weight management module, the quantitative analysis results output by the operation logic module, and the underground mine roof risk level evaluation standard set output by the evaluation standard module, and classify the risk levels according to the maximum membership principle and the evaluation standard module, and send the risk level classification results to the early warning module and the visualization module; The early warning module is used to execute the corresponding level of sound, light and electricity early warning measures according to the risk level output by the risk level assessment module; The visualization module is used to graphically display real-time assessment results, historical trend evolution, risk area distribution and warning status, and to provide a human-computer interaction interface, supporting two-dimensional and three-dimensional risk maps, time series diagrams, risk cloud maps and console operation interfaces, thereby improving information comprehension and interaction efficiency; The dynamic optimization module is used to collect deviation information between the risk level assessment results and actual on-site feedback, build an assessment-measurement comparison database, and dynamically correct the indicator weights based on the comparison database. At the same time, the corrected indicator weights are sent to the weight management module to achieve adaptive optimization of the assessment model.

[0034] The system has a simple structure and a high degree of intelligence. It can intelligently and accurately identify the risk level of the mine roof and provide dynamic early warning, effectively improving the safety management level of the mine roof.

Claims

1. An intelligent early warning method for stope roof risk based on uncertainty fusion model, characterized by: The following steps are included: Step 1: Based on the requirements of underground mine engineering, determine the static and dynamic indicator data required for roof safety risk assessment and establish the roof safety risk level assessment criteria; Step 2: Divide the data range and scoring range corresponding to the indicators at different safety levels based on numerical simulation and engineering standards; Step 3: Based on data requirements, a dynamic data collection solution is used to obtain static and dynamic indicator data, and the obtained data is defuzzified and standardized; Step 4: Based on data-driven, a variety of subjective and objective weight determination methods are integrated to extract data features, obtain comprehensive indicator weights, and redistribute weights under incomplete indicator conditions; Step 5: Based on the current optimal weight, the roof safety risk level is evaluated using the maximum membership principle, and the risk tendency judgment result is obtained and visualized through the risk assessment basic cloud model; Step 6: Based on the risk threshold set by the evaluation results, the early warning equipment is automatically activated to provide synchronous sound, light, and electricity warnings and automatically report the warning information; Step 7: By collecting roof risk warning information and forming a comparison database with actual on-site information, dynamically optimize the indicator weights and modify the basic cloud model for risk assessment.

2. The intelligent early warning method for stope roof risk based on uncertainty fusion model according to claim 1 is characterized in that: In step one, the static indicator data include roof rock mass grade, stress field data, mining technology, and historical accident conditions; the dynamic indicator data include roof displacement data, stress data, and crack conditions; the roof safety risk levels include Level I roof risk, Level II roof risk, Level III roof risk, Level IV roof risk, and Level V roof risk; among them, the scoring range of Level I roof risk is that the roof safety risk is greater than or equal to 81 and less than or equal to 100, the scoring range of Level II roof risk is that the roof safety risk is greater than or equal to 61 and less than or equal to 80, the scoring range of Level III roof risk is that the roof safety risk is greater than or equal to 41 and less than or equal to 60, the scoring range of Level IV roof risk is that the roof safety risk is greater than or equal to 21 and less than or equal to 40, and the scoring range of Level V roof risk is that the roof safety risk is greater than or equal to 0 and less than or equal to 20.

3. The intelligent early warning method for stope roof risk based on uncertainty fusion model according to claim 2 is characterized in that: The corresponding relationship in step 2 is as follows: For the roof rock mass grade, it is divided into five levels according to the rock mass stability: extremely stable, stable, moderate, unstable and extremely unstable, and corresponds to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, extremely stable corresponds to 81≤rock mass stability≤100, stable corresponds to 61≤rock mass stability≤80, moderate corresponds to 41≤rock mass stability≤60, unstable corresponds to 21≤rock mass stability≤40, and extremely unstable corresponds to 0≤rock mass stability≤20; For stress field data, according to the magnitude of ground stress, it is divided into five levels: no risk, low risk, potential risk, high risk and extremely high risk, which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, no risk corresponds to ground stress <5Mpa, low risk corresponds to 5Mpa≤ground stress <10Mpa, potential risk corresponds to 10Mpa≤ground stress <15Mpa, high risk corresponds to 15Mpa≤ground stress <20Mpa, and extremely high risk corresponds to ground stress >20Mpa; For mining technology, it is divided into five levels according to the disturbance force of the mining method, namely extremely low disturbance, low disturbance, medium disturbance, high disturbance and relatively high disturbance, which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, the mining methods corresponding to extremely low disturbance are bench mining method, residual ore recovery method and shallow stratification method, the mining methods corresponding to low disturbance are stratification mining method, shallow room-and-pillar method and strip method, the mining methods corresponding to medium disturbance are room-and-pillar method, longwall mining method, sublevel caving method and room-and-pillar method with ore retention zone, the mining methods corresponding to high disturbance are sublevel filling mining method, stage open-pit method, stage filling method and stage caving method, and the mining methods corresponding to relatively high disturbance are deep full stoping method, large stope unsupported caving method, block caving method, continuous stoping method and full-section mining; For historical accident situations, they are divided into five levels: safe, low risk, medium risk, high risk and extremely high risk according to the average annual number of roof accidents, and correspond to Level I roof risk, Level II roof risk, Level III roof risk, Level IV roof risk and Level V roof risk respectively. Among them, safe corresponds to no mine roof accidents, low risk corresponds to 0 times < mine annual average roof accidents ≤ 1 time, medium risk corresponds to 1 time < mine annual average roof accidents ≤ 2 times, high risk corresponds to 2 times < mine annual average roof accidents ≤ 3 times, and extremely high risk corresponds to mine annual average roof accidents > 3 times; For roof displacement data, it is divided into five levels according to the roof displacement situation: low risk, relatively low risk, medium risk, high risk and extremely high risk, which correspond to level I roof risk, level II roof risk, level III roof risk, level IV roof risk and level V roof risk respectively. Among them, low risk corresponds to 0≤ daily average displacement rate≤1mm / d, and 0≤ cumulative displacement≤10mm, relatively low risk corresponds to 1 mm / d< daily average displacement rate≤3mm / d, and 10 mm< cumulative displacement≤20mm, medium risk corresponds to 3mm / d< daily average displacement rate≤5mm / d, and 20 mm< cumulative displacement≤40mm, high risk corresponds to 5mm / d< daily average displacement rate≤8mm / d, and 40 mm< cumulative displacement≤60mm, and extremely high risk corresponds to daily average displacement rate>8mm / d, and cumulative displacement>60mm; For stress data, it is divided into five levels according to the stress conditions of the support / column, namely stable pressure, appropriate pressure, leakage pressure, heavy pressure and overpressure, and corresponds to Level I roof risk, Level II roof risk, Level III roof risk, Level IV roof risk and Level V roof risk respectively. Among them, stable pressure corresponds to 30% design bearing capacity ≤ support / column force < 50% design bearing capacity, appropriate pressure corresponds to 50% design bearing capacity ≤ support / column force < 70% design bearing capacity, leakage pressure corresponds to support / column force < 30% design bearing capacity, heavy pressure corresponds to 70% design bearing capacity ≤ support / column force < 90% design bearing capacity, and overpressure corresponds to 70% design bearing capacity ≥ 90% design bearing capacity; For the crack situation, they are divided into five levels according to the crack size, namely no cracks, fine cracks, stable cracks, expanding cracks and through cracks, which correspond to Level I roof risk, Level II roof risk, Level III roof risk, Level IV roof risk and Level V roof risk respectively. Among them, no cracks correspond to no obvious cracks and no cracks, fine cracks correspond to the existence of small cracks, and the crack width is less than 1mm, stable cracks correspond to the existence of non-penetrating cracks, and the crack width is 1mm≤<3mm, expanding cracks correspond to the existence of non-penetrating cracks, and the crack width is 3mm≤<5mm, and through cracks correspond to the existence of through cracks, and the crack width is ≥5mm.

4. The method for intelligent early warning of stope roof risk based on uncertainty fusion model according to claim 3 is characterized in that: In step three, for static indicator data, technical and design personnel obtain the data through simulation calculation based on the mine geological report and input it into the data processing terminal through manual input. For dynamic indicator data, monitoring equipment arranged at the mine site is used to collect the data in real time and automatically, thereby obtaining static indicator data and dynamic indicator data. Among them, the monitoring equipment adopts differentiated layout methods in coal mines and non-coal mines.

5. The method for intelligent early warning of stope roof risk based on uncertainty fusion model according to claim 4 is characterized in that: In step 3, the specific process of defuzzifying and standardizing the obtained data is as follows: S31; Introduce the rule mapping defuzzification processing mechanism in fuzzy set theory to convert qualitative indicators into quantitative data, and obtain the first i The corresponding level score of each indicator ; （1）； Where, Indicates the corresponding indicator level; S32: For indicators with actual data, linear normalization is performed; for indicators with data interval , the score interval is , data value X exist When , the corresponding score interval is , according to formula (2), the score of the jth indicator after nonlinear normalization mapping is obtained ; （2）。 6. The method for intelligent early warning of stope roof risk based on uncertainty fusion model according to claim 5 is characterized in that: The specific process of step 4 is as follows: S41: On a subjective level, an expert evaluation matrix is ​​constructed based on the analytic hierarchy process to extract expert knowledge and experience. On an objective level, the entropy method and the coefficient of variation method are combined to quantify the importance of indicators from the perspectives of information distribution and volatility, respectively. Finally, a linear weighted fusion strategy is used to integrate the three methods to form a unified weighting system, providing accurate input features for the subsequent risk level identification model. Specifically, the comprehensive weight is obtained according to formula (3): ; （3）； Where, is the AHP weight; is the entropy method weight; is the weight of the coefficient of variation method; , represents the reliability coefficient of the AHP method; , is the reliability coefficient of the entropy method; is the reliability coefficient of the coefficient of variation method; ; S42: For a weight vector containing n indicators , , When m indicators are missing, the first indicator after redistribution is obtained according to formula (4). i The weight of the indicator , according to formula (5), we can obtain the weight vector containing k indicators ; （4） ； （5） ； Where, Before redistribution i The weight of each indicator; is the weight sum of the residual indicators before redistribution; k is the number of participating indicators, k=mn; .

7. The method for intelligent early warning of stope roof risk based on uncertainty fusion model according to claim 6, characterized in that: The specific process of step five is as follows: S51: Real-time collection of static and dynamic indicator data of the assessed area, calculation of the final evaluation score based on the established roof safety risk level evaluation criteria, the corresponding relationship between the data range and level classification of various indicators and the roof safety risk level, the comprehensive indicator weights, and the redistribution of weights under the condition of incomplete indicators, and calculation of the membership degree of the final evaluation score under each risk level, as shown in formula (6); Calculate the membership degree of each level of the data group at the corresponding moment, and determine the maximum membership degree as the roof safety risk level of the corresponding area at the current moment; （6）； Where, Indicates the final evaluation score i The degree of membership under risk level, express i The central value of membership under risk level, , x Indicates the final evaluation score; S52: Based on the established roof safety risk level evaluation criteria, a risk assessment basic cloud model corresponding to each risk level is constructed. Each risk assessment basic cloud model is defined by three parameters: expected value, entropy, and super entropy, forming a standard risk distribution cloud corresponding to different risk levels; S53: Based on the static indicator and dynamic indicator data collected in real time, the evaluation score is calculated and converted into the risk assessment basic cloud model parameters to form the current risk evaluation cloud; S54: Compare the risk assessment basic cloud model with the current risk judgment cloud to obtain the roof safety risk tendency judgment result to provide a basis for auxiliary decision-making.

8. The method for intelligent early warning of stope roof risk based on uncertainty fusion model according to claim 7 is characterized in that: The specific process of step six is ​​as follows: When the roof safety risk reaches the set threshold score, the early warning equipment arranged in the working face and tunnel is controlled to execute the sound and light warning action and send it to the control detection end. Among them, when the level I roof risk occurs, the early warning equipment is controlled to keep the green light on and maintain normal power supply. At the same time, it generates early warning information for daily inspections and maintaining the current monitoring and support measures unchanged; when the level II roof risk occurs, the early warning equipment is controlled to keep the blue light on and activate the early warning system. At the same time, it generates early warning information for strengthening the inspection frequency and continuously observing the changes in monitoring data; when the level III roof risk occurs, the early warning equipment is controlled to keep the yellow light on, and a prompt voice is broadcast, the power supply of non-critical equipment is disconnected, and only important systems are retained. Power supply, at the same time, generate early warning information for adding monitoring points, conducting special inspections, and optimizing support measures; when there is a level IV roof risk, control the early warning equipment to flash with an orange light, and make continuous beeps and voice broadcasts, cut off power to the set area, and link the safety lighting and broadcast systems. At the same time, generate early warning information for starting the early warning response process, arranging personnel to evacuate relevant areas, and strengthening support structures; when there is a level V roof risk, control the early warning equipment to flash with a red light at a high frequency, continuously perform long beeps and voice broadcasts, cut off power completely, and stop power supply to all work areas. At the same time, generate early warning information for immediate implementation of personnel evacuation, suspension of operations, and implementation of emergency reinforcement measures.

9. The method for intelligent early warning of stope roof risk based on uncertainty fusion model according to claim 8, characterized in that: The specific process of step seven is as follows: S71: Continuously collect warning data generated by roof risk warning, and simultaneously collect actual on-site monitoring data and disposal feedback information; S72: By building a comparison database of warning information and actual on-site data, the matching degree and deviation between the warning results and actual risk events are analyzed; S73: Based on the comparative analysis results, repeat steps 4 and 5 using a dynamic weight adjustment algorithm to correct the weight distribution of risk assessment indicators in real time, eliminate the impact of misleading indicators, and enhance the weight sensitivity of key indicators, thereby optimizing the performance of the risk assessment model.

10. An intelligent early warning method for stope roof risk based on an uncertainty fusion model, used to implement the intelligent early warning method for stope roof risk based on an uncertainty fusion model as claimed in any one of claims 1 to 9, characterized in that: It includes acquisition module, weight management module, evaluation standard module, operation logic module, risk level evaluation module, early warning module, visualization module and dynamic optimization module; The acquisition module is used to collect multi-source monitoring data in the area to be evaluated, and record the spatial layout information and data acquisition frequency of the monitoring equipment to form an original data set, and send the original data set to the weight management module and the risk level assessment module; The weight management module is used to extract data set features based on the volatility of the acquired data, the correlation between indicators, and historical evaluation performance, initialize the weight vector of each evaluation indicator, support the allocation of incomplete indicator weights and dynamic weight adjustment based on subsequent model feedback, and output the revised indicator weight set to the risk level assessment module; The evaluation criteria module is used to construct a multi-level risk rating system. Based on expert experience, policies and regulations, historical cases and numerical analysis results, the boundary thresholds and judgment logic of each level are defined to form an underground mine roof risk rating evaluation standard set and send it to the risk rating evaluation module; The operation logic module is used to perform normalization and defuzzification processing on the collected data information, and calculate the relevant parameters and results of the processed data through fuzzy comprehensive evaluation, cloud model or multi-factor fusion algorithm, and send them to the risk level assessment module; The risk level evaluation module is used to receive the indicator data collected by the acquisition module, the indicator weights output by the weight management module, the quantitative analysis results output by the operation logic module, and the underground mine roof risk level evaluation standard set output by the evaluation standard module, and classify the risk levels according to the maximum membership principle and the evaluation standard module, and send the risk level classification results to the early warning module and the visualization module; The early warning module is used to execute the corresponding level of sound, light and electricity early warning measures according to the risk level output by the risk level assessment module; The visualization module is used to graphically display real-time assessment results, historical trend evolution, risk area distribution and warning status, and is used to provide a human-computer interaction interface, supporting two-dimensional and three-dimensional risk maps, time series diagrams, risk cloud maps and console operation interfaces; The dynamic optimization module is used to collect deviation information between the risk level assessment results and actual on-site feedback, build an assessment-measurement comparison database, and dynamically correct the indicator weights based on the comparison database. At the same time, the corrected indicator weights are sent to the weight management module to achieve adaptive optimization of the assessment model.

Images