Mine production potential safety hazard control method and device and electronic equipment

Abstract

The invention belongs to the technical field of production safety prevention and control, and particularly relates to a mine production potential safety hazard control method and device and electronic equipment. According to the scheme, dynamic simulation and prediction of a risk spatio-temporal evolution rule are achieved by constructing a spatio-temporal-behavior coupling risk field model; a dominant risk type is accurately judged by means of mode recognition based on deep learning, optimal scheduling of hidden danger points is carried out based on a multi-resource collaborative scheduling strategy, and finally continuous adaptive optimization of model parameters is realized by disposing feedback data closed loop. A closed-loop control system from accurate early warning and intelligent diagnosis to optimized disposal and self-evolution is constructed, and the core pain point problems of isolated early warning, single mode recognition, experience-dependent decision making and the like existing in current mine safety production hidden danger control are effectively solved.

Description

A method, device and electronic equipment for controlling safety hazards in mine production. Technical Field

[0001] This invention belongs to the field of production safety prevention and control technology, and specifically relates to a method, device and electronic equipment for controlling potential safety hazards in mine production. Background Technology

[0002] While various monitoring systems and information platforms are widely used in current mine production safety management, the following problems still exist in the intelligent control of safety hazards: First, existing solutions typically employ simple statistical models based on threshold judgment or isolated risk factor weighted scoring methods for risk early warning. These methods only analyze data from a single monitoring point independently, ignoring the spatial diffusion and coupling effects of risk throughout the entire mine network, and thus failing to accurately define the scope of high-risk areas. Second, existing solutions usually use fixed time windows for statistical analysis, lacking dynamic simulation of the risk evolution process, making it difficult to predict future development trends. Third, existing solutions fail to dynamically couple behavioral factors such as personnel distribution and equipment operation with geological and environmental factors, leading to difficulties in complex operations. The accuracy of risk prediction is poor under certain circumstances; secondly, existing solutions only issue alarms based on the exceeding of a single or a few risk indicators, lacking the ability to intelligently identify multiple risk coupling patterns. Moreover, the weighting factors in the risk scoring model are usually set through expert evaluation. At the same time, the alarm information only tells "where the risk is high" without specifying "why the risk is high" or "what the main risk types are," resulting in a lack of targeted emergency response; thirdly, the emergency response decision-making of existing solutions mainly relies on manual dispatch and experience judgment, which is prone to resource conflicts and response delays; at the same time, when selecting emergency response strategies, existing solutions mostly only consider safety effects, lacking quantitative assessment of factors such as production stoppage losses, labor costs, and equipment consumption, which is prone to problems of over-prevention or under-prevention. Summary of the Invention

[0003] This invention provides a method, device, and electronic equipment for controlling safety hazards in mining production, effectively solving the problems existing in the prior art.

[0004] To achieve the above objectives, the technical solution adopted by this invention is as follows: a method for controlling safety hazards in mine production, comprising the following processing steps: Step S100: acquiring multi-source heterogeneous data affecting mine production safety and performing corresponding preprocessing, wherein the multi-source heterogeneous data includes environmental physical field data, personnel behavior field data, equipment status data, and terminal interaction data; Step S200: based on the acquired multi-source heterogeneous data, using a pre-constructed spatiotemporal-behavioral coupled risk field model, with risk diffusion as the benchmark, quantifying environmental anomaly data into risk source terms and quantifying personnel gathering and equipment vibration behavior into disturbance terms, and solving to obtain the current dynamic risk field; Step S300: based on the acquired multi-source heterogeneous data, obtaining fused features through feature extraction and feature fusion; based on the obtained fused features... The process involves several steps: Step S400: First, a risk pattern classification model based on a pre-trained neural network is used to obtain risk pattern classification results. Step S500: Based on the obtained dynamic risk field, a set of potential hazards to be addressed is determined. Then, based on the obtained set of hazards and the risk pattern classification results corresponding to each hazard, several potential risk management plans are selected from a pre-built plan library. These plans are then optimized through multi-hazard point multi-resource collaborative scheduling calculations to obtain the optimal management plan for each hazard. Step S600: Risk management is executed based on the obtained optimal management plan. During the execution of risk management, interactive data uploaded by operators via terminals is acquired in real time. Based on the comparison results between the interactive data and preset standard data, parameters are optimized for the spatiotemporal-behavioral coupled risk field model and the multi-hazard point multi-resource collaborative scheduling optimization process.

[0005] Further, step S200 specifically includes the following processing procedures: Step S201: Based on risk diffusion, environmental anomaly data is quantified into risk source terms and personnel gathering and equipment vibration behavior are quantified into disturbance terms to construct a spatiotemporal-behavior coupled risk field model; Step S202: Based on the obtained environmental physical field data, personnel behavior field data and equipment status data, the spatiotemporal-behavior coupled risk field model is solved by the finite difference method to obtain the corrected dynamic risk field and the prediction results of the future short-term risk field.

[0006] Furthermore, the spatiotemporal-behavioral coupled risk field model is specifically represented as follows: in, Risk diffusion coefficient ( ), This represents the overall weighting coefficient for behavioral disturbances; Represents the Laplace operator; For risk source functions, , and The contribution coefficients of various risk sources. For safe gas concentration, The critical stress, The spatial attenuation length of the microseismic effect; This indicates that at time t, it is located The stress occurring at the top plate, This indicates that at time t, it is located The energy released during the time of the i-th micro-earthquake occurring at point i, This indicates that at time t, it is located The concentration of methane gas at that location To assess the location of the target, Location of the epicenter of the microseismic event. Let be the behavior perturbation function. , This is the behavioral perturbation gain coefficient. For the population distribution density field, This is the equivalent vibration energy field of the equipment.

[0007] Further, step S300 specifically includes the following processing procedures: Step S301: Acquire environmental physical field data, personnel behavior field data, equipment status data, and the corrected dynamic risk field; Step S302: Extract spatial distribution features from the dynamic risk field to obtain the area of ​​high-risk regions, the maximum value of risk gradient, and the location of the risk centroid; extract the temporal features of key points in the corresponding region from the environmental physical field data to obtain the gas concentration change rate and stress accumulation rate, and construct fusion features; Step S303: Based on the constructed fusion features, calculate the risk pattern classification result of the current region at the current time through a pre-trained risk pattern classification model, wherein the risk pattern classification result includes a comprehensive risk index, the dominant risk pattern, and the probability distribution of each risk pattern.

[0008] Furthermore, the comprehensive risk index is obtained by weighted summation of each risk component indicator, as specifically expressed below: in, Let be the comprehensive risk index at time t. For the i-th normalized risk component indicator, Let be the dynamic weight of the i-th component at time t. The basic weight of the i-th item, When the risk model is When, the specific weight offset of the i-th sub-item, N is the number of sub-item indicators, and M is the number of risk patterns.

[0009] Furthermore, in the multi-hazard point multi-resource collaborative scheduling optimization calculation, when there are multiple hazard points, a mixed integer programming problem is constructed for collaborative scheduling, specifically represented as follows: in, For the decision variable of the explosion-proof terminal, it is used to indicate whether the explosion-proof terminal k is assigned to the hidden danger point h, 0 indicates no, 1 indicates assignment; The decision variable for available engineering equipment is used to indicate whether engineering equipment l is assigned to the hidden danger point h; For standby personnel, the decision variable is used to indicate whether standby personnel r is assigned to the hazard point h; Let h be the start time for hazard mitigation, and CPSG() be the function for calculating unit safety benefit cost. To implement the optimal contingency plan The estimated operation time, The time span penalty coefficient, This is the risk threshold. ( ) represents the shortest estimated travel time from the terminal's current location to the potential hazard point. For based on The predicted risks have reached The latest time to start the treatment, This represents the terminal assigned to the hazard point h, where H is the set of hazard points.

[0010] Furthermore, the risk pattern classification model based on neural networks employs a temporal convolutional network.

[0011] Furthermore, the risk sub-indicators include, but are not limited to: gas and gas-related risk indicators, including gas anomaly index, carbon monoxide index, oxygen deficiency index, and gas outburst dynamic index; roof or rock stratum stability risk indicators, including roof stress concentration index, roof delamination index, support effectiveness index, and roof subsidence rate; and rockburst or microseismic activity risk indicators, including microseismic energy release rate and microseismic event frequency.

[0012] A control device for mine production safety hazards includes a sensor, a back-end server, and a portable terminal, wherein the sensor and the portable terminal are respectively communicatively connected to the back-end server; the back-end server specifically executes the aforementioned control method for mine production safety hazards.

[0013] An electronic device includes a memory, a processor, and a computer program stored in the memory and running on it. When the processor executes the program, it implements the aforementioned method for controlling safety hazards in mine production.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: (1) The present invention provides a method, device and electronic equipment for controlling safety hazards in mine production. The scheme constructs a coupled model with risk diffusion as the core and integrates environmental risk sources and behavioral disturbance terms. Instead of analyzing individual sensor data in isolation, it accurately depicts the spatial propagation, superposition and evolution process of risk in the entire mine. This enables the scheme to accurately define the real spatial boundary of high-risk areas and predict the short-term risk situation in the future. It fundamentally solves the fundamental problems of the existing technology's vague early warning range and lack of trend prediction ability.

[0015] (2) The proposed scheme introduces a risk pattern adaptive identification and weight transfer mechanism based on deep learning, which realizes the transformation of risk warning from "numerical over-limit" to "pattern diagnosis". The scheme not only calculates the global comprehensive risk index, but also identifies the dominant risk coupling pattern (such as "gas outburst type" and "roof-micro-seismic composite type") in real time through a temporal convolutional network, and dynamically adjusts the weight of each risk component according to the identification results, so that the alarm information can clearly indicate "where the risk comes from and what type it is", providing an important qualitative basis for subsequent accurate handling, and completely changing the status quo of vague alarm information and insufficient decision support in traditional alarms.

[0016] (3) The proposed solution proposes a control decision mechanism based on cost-effectiveness optimization and multi-resource collaborative scheduling. It incorporates economic indicators such as the direct cost of disposal measures and estimated production stoppage losses into the optimization objectives. The solution aims to minimize the cost-benefit ratio (CPSG) of the unit safety and optimizes resource collaborative scheduling in scenarios with multiple hidden danger points. The solution achieves the best balance between safety and efficiency by minimizing production disturbance and resource waste while ensuring the bottom line of safety. It effectively avoids the drawbacks of over-control or under-control.

[0017] (4) The scheme forms an intelligent safety control scheme with self-evolution capability by executing a closed-loop verification and continuously and automatically optimizing the model parameters. The scheme uses the on-site verification data (images, measurements) uploaded by the portable explosion-proof terminal to compare with the preset standard to complete the closed-loop confirmation of the treatment effect. At the same time, it uses the effect deviation data to iteratively optimize the model parameters used above and continuously update and enrich the pre-built contingency plan library, so that the prediction accuracy, decision economy and pattern recognition capability of the overall scheme can be continuously improved with the accumulation of running time. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below: Figure 1 is a flowchart of a method for controlling safety hazards in mine production according to an embodiment of the present invention; Figure 2 is a flowchart of risk field solving according to an embodiment of the present invention; Figure 3 is a flowchart of risk pattern classification result acquisition according to an embodiment of the present invention; Figure 4 is a schematic diagram of a control device for safety hazards in mine production according to an embodiment of the present invention; Figure 5 is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0019] To better understand the above-mentioned objectives, features and advantages of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0020] Numerous specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways than those described herein, and therefore the invention is not limited to the specific embodiments disclosed in the following specification.

[0021] Example 1: The following describes in detail a method for controlling potential safety hazards in mine production, with reference to the accompanying drawings.

[0022] As shown in Figure 1, a method for controlling safety hazards in mine production includes the following processing steps: Step S100: Acquire multi-source heterogeneous data affecting mine production safety and perform corresponding preprocessing, wherein the multi-source heterogeneous data includes environmental physical field data, personnel behavior field data, equipment status data, and terminal interaction data; Step S200: Based on the acquired multi-source heterogeneous data, through a pre-constructed spatiotemporal-behavioral coupled risk field model, taking risk diffusion as a benchmark, quantify environmental anomaly data into risk source terms and quantify personnel gathering and equipment vibration behavior into disturbance terms, and solve to obtain the current dynamic risk field; Step S300: Based on the acquired multi-source heterogeneous data, obtain fused features through feature extraction and feature fusion; based on the obtained fused features, through pre-training... A risk pattern classification model based on a neural network is trained to obtain risk pattern classification results; Step S400: Based on the obtained dynamic risk field, a set of potential hazards to be addressed is determined. Based on the obtained set of potential hazards and the risk pattern classification results corresponding to each potential hazard, several potential risk handling plans are selected from a pre-built plan library. Through multi-potential hazard multi-resource collaborative scheduling optimization calculation, the optimal handling plan for each potential hazard is obtained; Step S500: Risk handling is executed based on the obtained optimal handling plan. During the execution of risk handling, interactive data uploaded by operators through terminals is acquired in real time; Based on the comparison results of the interactive data and the preset standard data, the parameters of the spatiotemporal-behavior coupled risk field model and the multi-potential hazard multi-resource collaborative scheduling optimization process are optimized.

[0023] In specific implementation, step S200 includes the following processing steps: Step S201: Based on risk diffusion, environmental anomaly data is quantified into risk source terms and personnel gathering and equipment vibration behavior are quantified into disturbance terms to construct a spatiotemporal-behavior coupled risk field model; Step S202: Based on the obtained environmental physical field data, personnel behavior field data and equipment status data, the spatiotemporal-behavior coupled risk field model is solved using the finite difference method to obtain the corrected dynamic risk field and the prediction results of the short-term risk field.

[0024] In practical implementation, the spatiotemporal-behavioral coupled risk field model is specifically represented as follows: in, Risk diffusion coefficient ( ), This represents the overall weighting coefficient for behavioral disturbances; Represents the Laplace operator; For risk source functions, , and The contribution coefficients of various risk sources. For safe gas concentration, The critical stress, The spatial attenuation length of the microseismic effect; This indicates that at time t, it is located The stress occurring at the top plate, This indicates that at time t, it is located The energy released during the time of the i-th micro-earthquake occurring at point i, This indicates that at time t, it is located The concentration of methane gas at that location To assess the location of the target, Location of the epicenter of the microseismic event. Let be the behavior perturbation function. , This is the behavioral perturbation gain coefficient. For the population distribution density field, This is the equivalent vibration energy field of the equipment.

[0025] In specific implementation, step S300 includes the following processing steps: Step S301: Acquire environmental physical field data, personnel behavior field data, equipment status data, and the corrected dynamic risk field; Step S302: Extract spatial distribution features from the dynamic risk field to obtain the area of ​​high-risk regions, the maximum value of risk gradient, and the location of the risk centroid; extract the temporal features of key points in the corresponding region from the environmental physical field data to obtain the gas concentration change rate and stress accumulation rate, and construct fusion features; Step S303: Based on the constructed fusion features, calculate the risk pattern classification result of the current region at the current time using a pre-trained risk pattern classification model, wherein the risk pattern classification result includes a comprehensive risk index, the dominant risk pattern, and the probability distribution of each risk pattern.

[0026] In practice, the comprehensive risk index is obtained by weighted summation of each risk component indicator, as shown below: in, Let be the comprehensive risk index at time t. For the i-th normalized risk component indicator, Let be the dynamic weight of the i-th component at time t. The basic weight of the i-th item, When the risk model is When, the specific weight offset of the i-th sub-item, N is the number of sub-item indicators, and M is the number of risk patterns.

[0027] In practical implementation, during the multi-hazard point multi-resource collaborative scheduling optimization calculation, when multiple hazard points exist, a mixed integer programming problem is constructed for collaborative scheduling, specifically as follows: in, For the decision variable of the explosion-proof terminal, it is used to indicate whether the explosion-proof terminal k is assigned to the hidden danger point h, 0 indicates no, 1 indicates assignment; The decision variable for available engineering equipment is used to indicate whether engineering equipment l is assigned to the hidden danger point h; For standby personnel, the decision variable is used to indicate whether standby personnel r is assigned to the hazard point h; Let h be the start time for hazard mitigation, and CPSG() be the function for calculating unit safety benefit cost. To implement the optimal contingency plan The estimated operation time, The time span penalty coefficient, This is the risk threshold. ( ) represents the shortest estimated travel time from the terminal's current location to the potential hazard point. For based on The predicted risks have reached The latest time to start the treatment, This represents the terminal assigned to the hazard point h, where H is the set of hazard points.

[0028] In specific implementation, step S100: acquire multi-source heterogeneous data affecting mine production safety and perform corresponding preprocessing, wherein the multi-source heterogeneous data includes environmental physical field data, personnel behavior field data, equipment status data and terminal interaction data, specifically including the following processing steps: step S101: acquisition of multi-source heterogeneous data (1) environmental physical field data : Through a pre-defined fixed sensor network, at time t, from spatial coordinates Obtained from: gas concentration vector ;in, This refers to the methane concentration. This refers to the carbon monoxide concentration. This refers to the concentration of carbon dioxide. Oxygen concentration; environmental vector ,in, For temperature, For humidity, Wind speed; geotropic component ,in, For the stress of the top plate, This represents the energy of a microseismic event.

[0029] (2) Personnel behavior field data Acquired through positioning and operation systems, including: personnel location trajectory matrix: ;in, For personnel Position at time t; Personnel state vector: ,in, For personnel The working status at time t, such as: drilling, support or transportation; (3) Equipment status data Data is acquired through the equipment's PLC and a pre-built status monitoring system, including: Equipment operating status: Key timing parameters: Such as current, oil pressure, and vibration frequency.

[0030] (4) Terminal interaction data Data is acquired via an explosion-proof smart terminal (portable terminal), including inspection records and hazard reports (including text, images, and location data). (and instruction confirmation receipt).

[0031] Step S102: Data Preprocessing (1) Spatiotemporal Alignment: Synchronize the timestamps t of all data using the NTP (Network Time Protocol) protocol, and the spatial coordinates Unify to the mine's geographical coordinate system.

[0032] (2) Outlier cleaning: 1) For environmental physical field data : in, This represents the average of the data within the previous window (e.g., 5 minutes) before the current time point. This represents the standard deviation of the data within the corresponding window.

[0033] 2) For personnel behavior field data: use logical consistency verification and filter out impossible position jumps by speed threshold; for example: the speed threshold cannot exceed 5m / s; and ensure that the same person's state is not empty.

[0034] 3) For equipment status data: use status-parameter consistency verification, for example: when the status is running, the current should be greater than the no-load value; and for timing parameters, set reasonable upper and lower limits for filtering according to the equipment model.

[0035] 4) For terminal interaction data: use basic validity checks, such as location. Is it within the underground geographical coordinate range? And is the timestamp reasonable?

[0036] (3) Missing value imputation: 1) For missing environmental physical field data: For missing data of a single point sensor, the average of adjacent time is used for imputation; for continuous missing data caused by single point sensor failure, the average of data of adjacent spatial sensors is used for imputation.

[0037] 2) Missing data on personnel behavior: personnel location For the missing values, kinematic interpolation is used: in, The time point where the missing value is located. For personnel The average velocity vector before and after the missing period. This is the last valid time point before the missing time point.

[0038] When personnel status is missing, a method of status maintenance and work logic inference is adopted. Specifically: Basic rule: It is assumed that the personnel will maintain the previous valid status for a short period of time (such as 5 minutes); Logical verification: If the interpolated status contradicts the current position or equipment status (for example: the status is "drilling" but the position is not at any tunneling face), then it is adjusted to "moving" or "standby" status; Shift change point handling: If a status is missing near the shift change time, the pre-built team work plan table is used as the primary reference for assignment.

[0039] 3) For missing equipment status data: If the missing equipment operating status is less than the typical start-stop cycle of the equipment, the previous valid status will be maintained. If it exceeds the typical cycle or there are related production system log records during this period, it will be set to "offline" or "unknown".

[0040] For missing time-series parameters (such as current, pressure, and vibration parameters), if the equipment status is "running", the historical average value under the same operating condition is used for interpolation, for example, the historical average value of the current of the coal mining machine under the "normal coal cutting" operating condition; if the equipment status is "stopped", the parameter interpolation is zero or no-load value.

[0041] 4) Missing terminal interaction data: Missing terminal interaction data does not exist by default.

[0042] Ultimately, standardized data is obtained: Among them, environmental physical field data and some human behavior field data are used to drive the evolution of risk field, and all data are used for risk pattern recognition after feature extraction.

[0043] In specific implementation, step S200: Based on the obtained multi-source heterogeneous data, using a pre-constructed spatiotemporal-behavioral coupled risk field model, with risk diffusion as the benchmark, environmental anomaly data is quantified into risk source terms and personnel gathering and equipment vibration behavior are quantified into disturbance terms, and the current dynamic risk field is obtained by solving; as shown in Figure 2, the specific processing steps include the following: Step S201: With risk diffusion as the benchmark, environmental anomaly data is quantified into risk source terms and personnel gathering and equipment vibration behavior are quantified into disturbance terms, and a spatiotemporal-behavioral coupled risk field model is constructed; the construction of the spatiotemporal-behavioral coupled risk field model specifically includes the following processing steps: First, environmental physical field data, personnel behavior field data, and equipment status data are obtained, and based on the personnel behavior field data and equipment status data, personnel distribution density fields are constructed respectively. Equivalent vibration energy field of equipment Specifically, it is expressed as follows: in, For time t at The number of workers per unit volume or unit area around the location. The kernel function is used (the Gaussian kernel function is used in this embodiment). For the target assessment location, N is the total number of personnel. Let be the coordinates of the j-th person at time t. The smoothing radius is used to transform discrete personnel positions into a continuous density field. For time t at The equivalent energy intensity of equipment vibration at the location, The weighting coefficient for device k is... Let the vibration signal of device k at time t be represented in the frequency domain. ( ) represents the Dirac function. Let k be the position of the device itself.

[0044] The environmental physical field data includes, but is not limited to: top plate stress. Micro-vibration energy and its location, gas concentration etc.; secondly, model construction, including: (1) defining the global risk field It indicates position Risk level at time t.

[0045] (2) Establish a coupled partial differential equation (PDE) model (i.e., a spatiotemporal-behavioral coupled risk field model): in, Risk diffusion coefficient ( This is positively correlated with rock mass properties and tunnel connectivity, and can be specifically determined through historical data inversion. This represents the overall weighting coefficient for behavioral disturbances; This represents the Laplace operator, used to characterize the isotropic diffusion of risk in space; This is the risk source function, used to transform environmental anomalies into risk generation. Specifically: in, , and The contribution coefficients of various risk sources. For safe gas concentration, The critical stress, The spatial attenuation length of the microseismic effect; This indicates that at time t, it is located The stress occurring at the top plate, This indicates that at time t, it is located The energy released during the time of the i-th micro-earthquake occurring at point i, This indicates that at time t, it is located The concentration of methane gas at that location To assess the location of the target, This indicates the location of the hypocenter of the microseismic event.

[0046] This is the behavioral perturbation function, used to quantify the impact of human activities on the risk field. Specifically: in, , This is the behavioral perturbation gain coefficient. This indicates the nonlinear amplification effect of people gathering.

[0047] Step S202: Based on the obtained environmental physical field data, personnel behavior field data, and equipment status data, the spatiotemporal-behavior coupled risk field model is solved using the finite difference method to obtain the corrected dynamic risk field and the prediction results of the short-term risk field. Specifically, the PDE is solved on the computational grid using the finite difference method to obtain the predicted risk field value. At the same time, the measured risk values ​​of the sensor locations are used. (Depend on The model is assimilated and corrected by mapping (the scheme described in this embodiment uses ensemble Kalman filtering) to continuously correct the model state and parameters in order to reduce prediction error.

[0048] Correction and update formula: in, Let t be the model's prior state vector (predicted value). Let be the observation vector at time t. For observation operators (mapping model states to observation space). The Kalman gain matrix is ​​dynamically calculated to control the model's confidence level in the observations. Let t be the posterior state vector of the model at time t (i.e., the correction value).

[0049] Finally, the obtained output includes: the corrected dynamic risk field. (Right now (This is used to describe the actual spatial distribution of risk in the entire mine at the current moment.)

[0050] Future short-term (e.g., 2-hour) risk field forecast Spatial boundaries and evolution trends of high / low risk areas.

[0051] In the specific implementation, as shown in Figure 3, step S300: Based on the obtained multi-source heterogeneous data, a fused feature is obtained through feature extraction and feature fusion; based on the obtained fused feature, a risk pattern classification result is obtained through a pre-trained neural network-based risk pattern classification model, specifically including the following processing steps: Step S301: Based on the obtained environmental physical field data, personnel behavior field data, equipment status data, and the corrected dynamic risk field; Step S302: Spatial distribution features are extracted from the dynamic risk field to obtain the area of ​​high-risk regions, the maximum value of risk gradient, and the location of the risk centroid; Temporal features of key points in the corresponding region are extracted from the environmental physical field data to obtain the gas concentration change rate and stress accumulation rate, and fused features are constructed; Step S303: Based on the constructed fused feature, the risk pattern classification result of the current region at the current time is calculated through a pre-trained risk pattern classification model (in this embodiment, a temporal convolutional network is used), wherein the risk pattern classification result includes a comprehensive risk index, a dominant risk pattern, and the probability distribution of each risk pattern.

[0052] In specific implementation, the feature extraction specifically involves: extracting features from a dynamic risk field. Spatial distribution features are extracted, including but not limited to: the area of ​​high-risk regions. Maximum risk gradient And the location of the center of mass of risk; from environmental physical field data Extract key point temporal features, including but not limited to the rate of change of gas concentration. and stress accumulation rate .

[0053] Constructing fused feature vectors .

[0054] In specific implementation, the identification of the risk pattern involves: Input a pre-trained risk pattern classification model to obtain the risk pattern belonging to a pre-defined M-class. probability distribution Among them, the M-type risk model Including but not limited to: mining-induced roof delamination-abnormal gas outburst composite type and fault activation accompanied by water inrush and rockburst type.

[0055] In one or more embodiments, the risk pattern classification result includes a comprehensive risk index, a dominant risk pattern, and the probability distribution of each risk pattern. Specifically, the comprehensive risk index... From each risk sub-indicator The weighted sum is obtained, but the weights... Probability of the current risk pattern The function, specifically: in, For the i-th normalized risk component indicator, The basic weight of the i-th item, When the risk model is At that time, the specific weight offset of the i-th component constitutes a weight offset matrix. This is learned through historical case studies, such as: for the "gas outburst type" model, the gas-related sub-items... It is positive and takes a larger value.

[0056] In one or more embodiments, the risk sub-indicators include, but are not limited to: gas and gas-related risk indicators, such as: gas anomaly index, carbon monoxide index, oxygen deficiency index, and gas outburst dynamic index; roof or rock strata stability risk indicators, such as: roof stress concentration index, roof delamination index, support effectiveness index, and roof subsidence rate; and rockburst or microseismic activity risk indicators, such as: microseismic energy release rate and microseismic event frequency.

[0057] Ultimately, the dominant risk pattern was determined: ; and, the pattern probability distribution vector ; and, a comprehensive risk index with pattern interpretation. It also includes an analysis report on the contribution of each sub-item.

[0058] In specific implementation, step S400: Based on the obtained dynamic risk field, determine the set of hidden danger points to be treated. Based on the obtained set of hidden danger points and the risk mode classification results corresponding to each hidden danger point, select several risk disposal plans to be selected from the pre-constructed plan library, and obtain the optimal disposal plan for each hidden danger point through multi-hidden danger point multi-resource collaborative scheduling optimization calculation. Specifically, it includes the following processing steps: (1) Data input, including: corrected dynamic risk field and dominant risk model Set of potential hazards to be addressed ,Depend on Trigger recognition, among which, D is the preset threshold, and D is the number of potential hazards.

[0059] Available resource status: Set of idle explosion-proof terminals {term k}、Standing personnel assemble {p j} and the set of available engineering equipment {eq l}

[0060] Cost control data: various control measures Estimated direct costs Losses per unit of downtime (Yuan / minute)

[0061] (2) Data processing: 1) Optimization of cost-effectiveness of single-point emergency plans: For each potential hazard point According to its dominant risk model Retrieve K applicable contingency plans from the contingency plan database. Calculate the cost-benefit-per-unit (CPSG) for each contingency plan and select the optimal one, as shown below: in, To implement the contingency plan The total cost, in detail: Contingency Plan The expected production downtime Contingency Plan After implementation, potential hazards The predicted risk reduction value is calculated by simulation using a risk field model or obtained through regression based on historical data. This is the minimum threshold that risk must reach to ensure that risk is reduced below the safety line.

[0062] It should be noted that the contingency plan library needs to be built in advance. The contingency plan library stores control schemes adapted to different risk modes, and each control scheme is set with the expected associated risk mode.

[0063] 2) Multi-hazard point and multi-resource collaborative scheduling optimization: When there are multiple hazard points (i.e. When a mixed-integer programming problem is constructed for coordinated scheduling, it is specifically represented as follows: in, For the decision variable of the explosion-proof terminal, it is used to indicate whether the explosion-proof terminal k is assigned to the hidden danger point h, 0 indicates no, 1 indicates assignment; The decision variable for available engineering equipment is used to indicate whether engineering equipment l is assigned to the hidden danger point h; For standby personnel, the decision variable is used to indicate whether standby personnel r is assigned to the hazard point h; The start time for hazard mitigation at hazard point h. To implement the optimal contingency plan The estimated operation time, The time span penalty coefficient, This is the risk threshold. ( ) represents the shortest estimated travel time from the terminal's current location to the potential hazard point. For based on The predicted risks have reached The latest time to start the treatment, This indicates the terminal assigned to the hazard point h.

[0064] 3) By solving the above optimization model, the optimal collaborative task scheduling scheme is generated and decomposed into structured task instructions, which are then sent to the designated explosion-proof smart terminal via wireless network.

[0065] In practical implementation, the optimal collaborative task scheduling scheme includes: a set of globally optimal control plans { } and its corresponding collaborative execution schedule; secondly, task instructions pushed to each terminal, including but not limited to: target location Optimal contingency plan Standard operating procedures, list of required resources and processing time limits .

[0066] In specific implementation, step S500: Based on the obtained optimal disposal plan, risk disposal is carried out, and interactive data uploaded by operators through terminals is acquired in real time during the risk disposal process; based on the comparison results of the interactive data and the preset standard data, the parameters of the spatiotemporal-behavior coupled risk field model and the multi-hazard point multi-resource collaborative scheduling optimization process are optimized, and the following processing process is specifically executed: Step S501: Data acquisition: (1) Task execution process data Uploads from explosion-proof smart terminals include, but are not limited to: personnel arrival confirmation (GPS / UWB location matching), and sign-in photos. Photographs of key processes and key points For example: photos of completed support installation, photos of equipment repositioning; confirmatory measurement data. For example, the post-treatment gas concentration measured by a portable instrument connected to the terminal. (2) Environmental feedback data after the task is completed After the treatment is completed, environmental data is continuously acquired from the fixed sensor network to calculate the actual risk reduction effect. Step S502: Data processing: (1) Intelligent verification: Image verification: The system automatically compares With contingency plan Corresponding standard operation completed image The structural similarity index (SSIM) and other indicators are used for quantification. in, , These are the mean values ​​of the image to be verified and the standard image, respectively. , These are the standard deviations of the image to be verified and the standard image, respectively. Let be the covariance between the image to be verified and the standard image; , As a stability constant, when When the preset threshold is reached, the image is deemed to have passed the acceptance test.

[0067] Data validation: Check Do the key indicators meet the contingency plan? The expected outcome standard.

[0068] (2) Closed-loop confirmation and effect evaluation: If both the image and data verification are successful, the system determines that the hazard handling is successfully closed-loop, generates a closed-loop report, updates the knowledge base, and uses this successful case as a contingency plan. A positive feedback loop on effectiveness.

[0069] (3) Continuous optimization of model parameters and updating of the contingency plan library: Periodically calculate the deviation of the treatment effect and continuously optimize the model parameters, specifically including: parameter optimization of the spatiotemporal-behavioral coupled risk field model: compare the risk reduction value predicted by the contingency plan. (This can be calculated based on the spatiotemporal-behavioral coupled risk field model in step two) and the actual risk reduction value. Using bias data, key parameters in the PDE model are iteratively optimized through gradient descent or Bayesian update methods. , as well as wait.

[0070] Parameter optimization for collaborative scheduling of multiple hidden danger points and multiple resources: reducing the actual total cost and actual risk reduction value Compared with the estimates used in decision-making ( )and By comparing and updating the unit safety benefit cost model, future CPSG calculations will be more accurate.

[0071] Enrich the contingency plan library: Successfully handled and verified cases (including their initial characteristics, patterns, and contingency plans) will be stored as new examples in the contingency plan library for training and identification of future risk patterns and contingency plan recommendations.

[0072] Example 2: In one or more embodiments, as shown in Figure 4, this embodiment provides a control device for mine production safety hazards, including sensors, a backend server, and a portable terminal. The sensors and the portable terminal are respectively communicatively connected to the backend server. The backend server specifically executes the above-described method for controlling mine production safety hazards.

[0073] In one or more embodiments, as shown in FIG5, this embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on it. When the processor executes the program, it implements the above-described method for controlling safety hazards in mine production.

[0074] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for controlling potential safety hazards in mine production, characterized in that, The process includes the following steps: Step S100: Acquire multi-source heterogeneous data affecting mine production safety and perform corresponding preprocessing, wherein the multi-source heterogeneous data includes environmental physical field data, personnel behavior field data, equipment status data, and terminal interaction data; Step S200: Based on the acquired multi-source heterogeneous data, through a pre-constructed spatiotemporal-behavioral coupled risk field model, with risk diffusion as the benchmark, quantify environmental anomaly data into risk source terms and personnel gathering and equipment vibration behavior into disturbance terms, and solve to obtain the current dynamic risk field; Step S300: Based on the acquired multi-source heterogeneous data, obtain fused features through feature extraction and feature fusion; based on the obtained fused features, through a pre-trained neural network-based risk... The model classifies risks to obtain risk pattern classification results; Step S400: Based on the obtained dynamic risk field, determine the set of hidden danger points to be treated. Based on the obtained set of hidden danger points and the risk pattern classification results corresponding to each hidden danger point, select several risk disposal plans from the pre-constructed plan library, and obtain the optimal disposal plan for each hidden danger point through multi-hidden danger point multi-resource collaborative scheduling optimization calculation; Step S500: Execute risk disposal based on the obtained optimal disposal plan. During the execution of risk disposal, obtain the interactive data uploaded by the operator through the terminal in real time; Based on the comparison results of the interactive data and the preset standard data, optimize the parameters of the spatiotemporal-behavior coupled risk field model and the multi-hidden danger point multi-resource collaborative scheduling optimization process.

2. The method for controlling potential safety hazards in mine production as described in claim 1, characterized in that, The specific steps of step S200 include the following processing procedures: Step S201: Based on risk diffusion, the abnormal environmental data is quantified into risk source items and the personnel gathering and equipment vibration behavior are quantified into disturbance items, and a spatiotemporal-behavioral coupled risk field model is constructed. Step S202: Based on the obtained environmental physical field data, personnel behavior field data and equipment status data, the spatiotemporal-behavior coupled risk field model is solved by the finite difference method to obtain the corrected dynamic risk field and the prediction results of the short-term risk field.

3. The method for controlling potential safety hazards in mine production as described in claim 2, characterized in that, The spatiotemporal-behavioral coupled risk field model is specifically represented as follows: in, The risk diffusion coefficient, This represents the overall weighting coefficient for behavioral disturbances; Represents the Laplace operator; For risk source functions, 、 and Contribution coefficients for various risk sources For safe gas concentration, The critical stress, The spatial attenuation length of the microseismic effect; This indicates that at time t, it is located The stress occurring at the top plate, This indicates that at time t, it is located The energy released during the time interval of the i-th micro-earthquake occurring at point i, This indicates that at time t, it is located The concentration of methane gas at that location To assess the location of the target, Location of the epicenter of the microseismic event. Let be the behavior perturbation function. 、 This is the behavioral perturbation gain coefficient. For the population distribution density field, This is the equivalent vibration energy field of the equipment.

4. The method for controlling potential safety hazards in mine production as described in claim 1, characterized in that, Step S300 specifically includes the following processing procedures: Step S301: Acquire environmental physical field data, personnel behavior field data, equipment status data, and corrected dynamic risk field; Step S302: Extract spatial distribution features from the dynamic risk field to obtain the area of ​​high-risk regions, the maximum value of risk gradient, and the location of risk centroid. Extract the temporal features of key points in the corresponding area from the environmental physical field data, obtain the gas concentration change rate and stress accumulation rate, and construct fusion features; Step S303: Based on the constructed fusion features, calculate the risk pattern classification result of the current area at the current time through a pre-trained risk pattern classification model, wherein the risk pattern classification result includes a comprehensive risk index, the dominant risk pattern, and the probability distribution of each risk pattern.

5. The method for controlling potential safety hazards in mine production as described in claim 4, characterized in that, The comprehensive risk index is obtained by weighted summation of each risk component indicator, as shown below: in, Let be the comprehensive risk index at time t. For the i-th normalized risk component, Let be the dynamic weight of the i-th component at time t. The basic weight of the i-th item, When the risk model is When, the specific weight offset of the i-th sub-item, N is the number of sub-item indicators, and M is the number of risk patterns.

6. The method for controlling potential safety hazards in mine production as described in claim 1, characterized in that, In the multi-hazard point, multi-resource collaborative scheduling optimization calculation, when there are multiple hazard points, a mixed integer programming problem is constructed for collaborative scheduling, as specifically represented below: in, is the decision variable for the explosion-proof terminal, used to indicate whether the explosion-proof terminal k is assigned to the hidden danger point h, where 0 indicates no and 1 indicates assignment; The decision variable for available engineering equipment is used to indicate whether engineering equipment l is assigned to the hidden danger point h; For standby personnel, the decision variable is used to indicate whether standby personnel r are assigned to the hazard point h; Let h be the start time for hazard mitigation, and CPSG() be the function for calculating unit safety benefit cost. To implement the optimal contingency plan The estimated operation time, The time span penalty coefficient, This is the risk threshold. ( ) represents the shortest estimated travel time from the terminal's current location to the potential hazard point. For based on The predicted risks have reached The latest time to start the treatment, This represents the terminal assigned to the hazard point h, where H is the set of hazard points.

7. The method for controlling potential safety hazards in mine production as described in claim 1, characterized in that, The risk pattern classification model based on neural networks employs a temporal convolutional network.

8. The method for controlling potential safety hazards in mine production as described in claim 5, characterized in that, The risk sub-indicators include, but are not limited to: gas and gas-related risk indicators, including gas anomaly index, carbon monoxide index, oxygen deficiency index, and gas outburst dynamic index; roof or rock stratum stability risk indicators, including roof stress concentration index, roof delamination index, support effectiveness index, and roof subsidence rate; and rockburst or microseismic activity risk indicators, including microseismic energy release rate and microseismic event frequency.

9. A control device for safety hazards in mine production, comprising sensors, a backend server, and a portable terminal, wherein, The sensor and the portable terminal are respectively connected to the back-end server; the back-end server specifically executes the method for controlling safety hazards in mine production as described in any one of claims 1-8.

10. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and running thereon, characterized in that, When the processor executes the program, it implements a method for controlling potential safety hazards in mine production as described in any one of claims 1-8.

Images