A coal mine concealed mining face intelligent identification method based on air distribution gas coupling

Abstract

The present application relates to the technical fields of coal mine safety monitoring and intelligent ventilation, and discloses a coal mine concealed mining face intelligent identification method based on air distribution gas coupling, which establishes a mine static ventilation baseline and a data space mapping; injects a periodic wind pressure perturbation signal into a ventilation pipe network and starts a phase-locked joint sampling to obtain phase-aligned wind speed and gas concentration time series data on a physical time axis; constructs a measured perturbation response Jacobian matrix to solve impedance residual errors, and locks the target topology branch where an aerodynamic impedance mutation occurs; uses a legal blasting pulse to calibrate an actual longitudinal dispersion coefficient, performs inverse dispersion calculation on unexplained gas emission residual errors, calculates the actual migration distance of an unknown gas source from a downstream sensor, and combines a three-dimensional space trend curve and the actual migration distance to perform a space geometric intersection to generate an exact physical coordinate of the concealed mining face. The present application realizes spatial accurate quantitative tracing of illegal mining areas in a mine.

Description

Technical Field

[0001] This invention relates to the field of coal mine safety monitoring and intelligent ventilation technology, specifically to an intelligent identification method for concealed mining faces in coal mines based on the coupling of ventilation and gas distribution. Background Technology

[0002] Hidden mining activities in coal mines can disrupt the balance of the mine's original ventilation system, posing serious safety hazards. Within the existing mine ventilation and gas monitoring system, the investigation of hidden mining faces faces numerous technical bottlenecks.

[0003] In terms of ventilation status monitoring, traditional investigation methods mainly rely on manual wind measurement or simple static air volume comparison. This conventional method is difficult to separate the slight aerodynamic impedance change caused by the access of the concealed mining face from the normal background airflow fluctuations in the mine, making it difficult to accurately locate the target ventilation branch and concealed air use node where resistance changes occur.

[0004] Regarding gas anomaly early warning, existing monitoring systems typically rely solely on gas concentration exceeding the threshold for alarms. When unexplained gas outbursts occur, the system can only provide qualitative anomaly alerts, lacking a deep analytical mechanism based on the physical laws of gas transport. It cannot quantitatively deduce the actual physical transport distance of the unknown gas source from the downstream methane sensor. Because it is impossible to simultaneously obtain the topological location of the abnormal ventilation branch and the relative physical transport distance of the gas source, existing technologies cannot perform geometric trajectory tracing and spatial intersection of multidimensional data. Consequently, it is impossible to generate the precise physical spatial coordinates and on-site verification path of suspected concealed mining face openings. This results in mine safety inspectors lacking intuitive on-site physical location guidance when actually investigating illegal mining operations, often having to rely on experience for large-scale searches, leading to low overall investigation and evidence collection efficiency. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling, which solves the problems of existing technologies that rely heavily on manual investigation and lack quantitative spatial physical coordinate tracing methods for identifying concealed mining faces.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for intelligent identification of concealed mining faces in coal mines based on ventilation and gas coupling, comprising: Establish a static ventilation baseline and data space mapping for the mine to obtain a directed ventilation network topology, theoretical air demand baseline, and dynamic baseline of known gas emission from different sources; Inject wind pressure perturbation signals and initiate phase-locked loop sampling to acquire phase-aligned wind speed time series data and gas concentration time series data; The air distribution difference is calculated by summing the measured total return air volume and the measured air distribution volume at each major air-consuming location. If the air distribution difference exceeds the normal adjustment range, it is determined that there is an additional air demand. The wind speed time series data is converted into air volume time series data by combining the roadway cross-sectional attributes in the directed ventilation network topology diagram. The measured micro-perturbation response Jacobian matrix is ​​constructed based on the air volume time series data. The impedance residual is calculated by combining the theoretical Jacobian matrix. The ventilation branch that has a sudden change in aerodynamic impedance is locked as the target topology branch. Based on the time-series waveform of the gas concentration pulse generated by a legal blast, the actual longitudinal dispersion coefficient of a specific ventilation branch is calibrated. Based on the gas concentration time series data and the known gas emission dynamic baseline, unexplained gas emission residuals are extracted. The unexplained gas emission residuals are coherently filtered to obtain effective residual pulses. The actual longitudinal dispersion coefficient is substituted into the inverse solution model to calculate the actual migration distance of the unknown gas source from the downstream methane sensor. Extract the three-dimensional spatial trajectory curve of the target topological branch in the mining engineering plan, and combine it with the actual movement distance to perform spatial geometric intersection, thereby generating the exact physical spatial coordinates of the suspected hidden mining face area.

[0007] The core innovation of this invention lies in combining perturbation response analysis in fluid mechanics with a one-dimensional gas convection-diffusion model to establish a joint three-dimensional source tracing mechanism of impedance topology localization and dispersion distance inverse calculation. By injecting a specific frequency of wind pressure perturbation signal into the mine ventilation network, coherence filtering is used to separate the weak aerodynamic impedance changes brought about by the concealed mining face from the background noise, thereby constructing a perturbation Jacobian matrix to locate the abnormal branch. At the same time, the roadway gas dispersion channel is calibrated using legal blasting pulses, and the physical migration distance of the unexplained gas source is inversely calculated based on the waveform broadening effect. The direction of the topology branch and the actual migration distance calculated in reverse are spatially geometrically intersected to achieve accurate quantitative determination of the physical coordinates of the concealed mining face opening.

[0008] Preferably, the step of injecting wind pressure perturbation signals and initiating phase-locked loop sampling to acquire phase-aligned wind speed time series data and gas concentration time series data includes: The active perturbation angular frequency of the periodic wind pressure perturbation signal is used as a reference clock signal and sent to the wind speed sensor and methane sensor. The wind speed sensor and the methane sensor generate a synchronous sampling clock signal with a frequency that is an integer multiple of the active perturbation angular frequency based on the reference clock signal; Synchronous acquisition is performed using a synchronous sampling clock signal as the trigger beat to obtain phase-aligned wind speed time series data and gas concentration time series data on the physical time axis.

[0009] Preferably, the step of constructing the measured perturbation response Jacobian matrix based on the air volume time series data and performing impedance residual calculation in combination with the theoretical Jacobian matrix includes: Extract the air volume fluctuation amplitude of the air volume time series data at the active perturbation angular frequency; The amplitude of the micro-perturbation wind pressure is used as the system input excitation, and the amplitude of the air volume fluctuation extracted from each ventilation branch is used as the system output response to construct the measured micro-perturbation response Jacobian matrix. Based on the flow conservation and loop pressure balance relationship of the pipeline network nodes in fluid mechanics, the partial derivatives of the theoretical air volume of each ventilation branch with the theoretical air pressure of the perturbation node are calculated and combined into a theoretical Jacobian matrix. The measured perturbation response Jacobian matrix and the theoretical Jacobian matrix are inverted respectively. The inverted measured perturbation response Jacobian matrix is ​​then subtracted from the inverted theoretical Jacobian matrix to obtain the impedance residual matrix.

[0010] Preferably, locking the ventilation branch where abrupt aerodynamic impedance changes as the target topology branch includes: Perform eigenvalue decomposition on the impedance residual matrix to extract the matrix eigenvectors corresponding to eigenvalues ​​greater than a preset judgment threshold; The column indices of the matrix corresponding to the non-zero elements in the eigenvectors of the positioning matrix; Spatial comparison is performed between the matrix column index and the directed ventilation network topology graph, and the matrix column index is restored to the specific ventilation topology branch according to the spatial encoding mapping relationship; When it is determined that the slope of the wind pressure-air volume relationship curve of a specific ventilation topology branch undergoes a sudden change without a reported damper adjustment record or a connection record, the specific ventilation topology branch is locked as the target topology branch.

[0011] Preferably, the calibration of the actual longitudinal dispersion coefficient of a specific ventilation branch based on the time-series waveform of the gas concentration pulse generated by a legal blast includes: Acquire information on pyrotechnic equipment reporting events and map it to a specific ventilation branch in the directed ventilation network topology graph; When a methane sensor located in a specific ventilation branch detects a gas concentration pulse generated by a legally reported blasting operation, the time-series waveform of the gas concentration pulse in phase-locked sampling mode is extracted, and the waveform broadening variance of the time-series waveform of the gas concentration pulse on the time axis is calculated. Extract the physical distance from the known blasting location to the downstream methane sensor, and obtain the average wind speed synchronously collected by the wind speed sensor; The actual longitudinal dispersion coefficient of a specific ventilation branch under the current operating conditions is determined by using waveform widening variance, physical distance, and average wind speed. The specific calculation method for the actual longitudinal dispersion coefficient is as follows: multiply the waveform widening variance by the cube of the average wind speed, and then divide by twice the physical distance to obtain the actual longitudinal dispersion coefficient.

[0012] Preferably, the step of extracting unexplained gas emission residuals based on the gas concentration time series data and the known gas emission dynamic baseline, and performing coherent filtering on the unexplained gas emission residuals to obtain effective residual pulses includes: Obtain the total return air volume and methane concentration of the return air in the main area or the return air in the zone, and multiply the total return air volume or the return air in the zone by the methane concentration and the unit conversion factor to generate the actual exhaust gas emission. The unexplained gas emission residual is calculated by subtracting the known dynamic baseline of gas emission from different sources and the interpretation of legal blasting disturbance from the actual gas emission from ventilation. Extract the residual time-series waveform of the unexplained gas outburst residual, and perform frequency domain conversion between the residual time-series waveform and the periodic wind pressure perturbation signal; Calculate the cross-coherence function of the residual time-series waveform and the periodic wind pressure perturbation signal, and extract the coherence coefficient value of the cross-coherence function at the active perturbation angular frequency; When the coherence coefficient value is greater than the set coherence determination threshold, the residual timing waveform with active perturbation angular frequency modulation characteristics is confirmed as a valid residual pulse.

[0013] Preferably, the step of substituting the actual longitudinal dispersion coefficient into the inverse calculation model to calculate the actual migration distance of the unknown gas source from the downstream methane sensor includes: Extract the time skewness characteristics of the effective residual pulses and calculate the widening variance of the residual waveform of the effective residual pulses on the time axis; Substitute the actual longitudinal dispersion coefficient into the inverse solution model of the one-dimensional convection-diffusion equation; The actual migration distance of the unknown gas source from the downstream methane sensor is calculated by multiplying the variance of the widened residual waveform by the cube of the average wind speed and then dividing by twice the actual longitudinal dispersion coefficient.

[0014] Preferably, the step of combining the actual transport distance to perform spatial geometric intersection to generate the exact physical spatial coordinates of the suspected concealed mining face area includes: Obtain the reference spatial coordinates of the downstream methane sensor on the three-dimensional trajectory curve; Starting from the reference spatial coordinates, follow the three-dimensional spatial curve against the direction of the ventilation airflow to perform geometric trajectory tracking. When the length of the tracked curve is equal to the actual movement distance, the endpoint of the track is identified as the exact physical spatial coordinates. The additional air usage direction is obtained by tracing the air supply difference along the path from the air inlet to the air usage point to the return air, and the location of the air path change is extracted by finding the resistance anomaly along the damper and the airtight seal. By performing spatial geometric intersection of the exact physical spatial coordinates with the additional wind direction and the location of wind path changes, a suspected hidden mining face area is generated. Extract the duration and deviation of the air distribution difference and map them as the air distribution difference confidence score; extract the eigenvalue change of the target topological branch in the impedance residual matrix and map it as the impedance residual confidence score; extract the cross-coherence coefficient value of the effective residual pulse and map it as the gas residual dispersion confidence score. The confidence scores of the air distribution difference, impedance residual, and gas residual dispersion are multiplied by their respective preset weighting coefficients and then summed to generate a comprehensive early warning score for suspected hidden mining faces. When the comprehensive early warning score exceeds the preset score threshold, path planning is performed in the directed ventilation network topology map to generate a field verification path leading to the suspected hidden mining face area.

[0015] A second aspect of this invention provides an intelligent identification system for concealed mining faces in coal mines based on ventilation and gas coupling, comprising: The mine environment sensing network includes wind speed sensors and methane sensors deployed in each branch roadway to collect phase-aligned wind speed time series data and methane concentration time series data on the physical time axis. The perturbation actuator is used to inject periodic air pressure perturbation signals into the mine ventilation network; The central processing computing platform is connected to the mine environment sensing network and the micro-disturbance actuator, respectively. The central processing computing platform is configured to execute the intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling as described in the first aspect above.

[0016] This invention provides an intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling. It has the following beneficial effects: 1. This invention injects periodic wind pressure perturbation signals into the mine ventilation network and combines phase-locked loop sampling to obtain air volume time series data. It constructs the measured and theoretical perturbation response Jacobian matrix for impedance residual calculation, which can separate the weak aerodynamic impedance mutation characteristics caused by the connection of the hidden mining face to the ventilation system from the background fluctuations. This allows for accurate location of the target topological branch where resistance anomalies occur, solving the problem that traditional methods relying solely on manual wind measurement are insufficient to detect hidden air-using nodes.

[0017] 2. This invention uses the time-series waveform of gas concentration pulse generated by legal blasting to calibrate the actual longitudinal dispersion coefficient of ventilation branches, and substitutes the waveform broadening variance of the extracted effective residual pulse into the inverse solution model of the one-dimensional convection-diffusion equation, thereby realizing the quantitative calculation of the physical transport distance of an unknown gas source from the downstream methane sensor. This overcomes the technical defect of existing technologies where gas anomaly alarms can only provide qualitative prompts and cannot quantitatively deduce the physical distance of the gas source.

[0018] 3. This invention extracts the three-dimensional spatial trajectory curve of the target topological branch in the mining engineering plan, combines it with the actual transport distance obtained by reverse calculation to perform geometric trajectory tracking and spatial intersection, and performs weighted comprehensive scoring on multi-dimensional evidence such as ventilation difference, impedance residual and gas residual to generate the exact physical spatial coordinates of suspected hidden mining face areas and on-site verification paths, providing mine safety supervisors with intuitive on-site physical location guidance and improving the efficiency of investigating illegal mining operations in mines. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system architecture diagram of the present invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see the appendix Figure 1-2 This invention provides an intelligent identification method for concealed mining faces in coal mines based on ventilation-gas coupling. The method relies on a precise tracing system for concealed mining faces. The hardware architecture of the precise tracing system for concealed mining faces mainly includes a perturbation actuator, a mine environment sensing network, and a central processing computing platform. In this embodiment, the physical implementation of the perturbation actuator can be a frequency converter installed on the main ventilation fan in the mine, or an intelligent regulating damper set at key ventilation nodes underground. The mine environment sensing network specifically includes wind speed sensors and methane sensors deployed in each branch roadway. The precise tracing system for concealed mining faces also includes a phase-locked loop (PLL) sampling control unit. The PLL sampling control unit is specifically implemented as an analog-to-digital converter control board or PLL circuit with an external clock synchronization interface configured inside each sensor, used to receive the baseband signal sent by the perturbation actuator and synchronously sample the clock cycle.

[0022] The specific process of the above method is as follows: S100: Establish the mine static ventilation baseline and data space mapping. The system converts the mine ventilation system diagram into a directed ventilation network topology diagram containing nodes, edges and attributes, and performs spatial coding mapping on the locations of wind measurement stations, gas sensors and explosives reporting locations. Based on this, the theoretical air volume baseline under the legal production state within the time window is calculated, and the baseline of known gas emission from different sources is constructed simultaneously.

[0023] S200, injecting wind pressure perturbation signal and starting phase-locked sampling, the system injects periodic wind pressure excitation signal into the mine ventilation network through perturbation actuator. The central processing computing platform sends the active perturbation angular frequency as a reference clock signal to each wind speed sensor and methane sensor underground. Each sensor uses the active perturbation angular frequency as a trigger source for synchronous sampling through the built-in phase-locked sampling control unit. The synchronous sampling operation enables the wind speed time series data and methane concentration time series data acquired by the system to achieve phase alignment at the physical level.

[0024] S300, Measured Impedance Residual Calculation and Abnormal Branch Location: The system calculates the static air distribution difference between the measured total air volume and the total air distribution at the main air-using locations. When the static air distribution difference exceeds the allowable error and cannot be explained by the reporting record, the system extracts the fluctuation amplitude of each wind speed sensor at the active perturbation angular frequency, constructs the measured perturbation response Jacobian matrix, synchronously calls the theoretical Jacobian matrix under the current topology, calculates the system's impedance residual matrix, and filters out the matrix column indexes in the non-zero residual set by performing eigenvector calculation on the impedance residual matrix. The specific ventilation topology branch with unplanned aerodynamic impedance changes is locked as the roadway where the suspected concealed mining face is located.

[0025] S400, based on the dispersion channel calibration of the legal blasting pulse of pyrotechnics, the system acquires the pyrotechnics reporting information and maps it to the corresponding ventilation branch. When a gas concentration pulse is detected from a legally reported blasting operation, the gas concentration pulse is used to calibrate the physical parameters of gas transmission in the roadway under the current ventilation conditions. The time-series waveform broadening variance of the legal gas pulse under phase-locked sampling is extracted. Combined with the physical distance from the reported known blasting location to the downstream sensor, the actual longitudinal dispersion coefficient is inferred.

[0026] S500, Gas Residual Dispersion Inverse Solution and Physical Distance Measurement: The system calculates the actual gas emission from the total return air or zoned return air, extracts unexplained gas emission residuals, verifies the cross-coherence between the unexplained gas emission residual waveform and the wind pressure perturbation signal in the frequency domain, extracts the waveform broadening variance of effective residual pulses with angular frequency modulation characteristics, uses the longitudinal dispersion coefficient obtained from calibration in S400 as a known environmental parameter, and calculates the actual migration distance of the unknown gas source from the downstream sensor by inversely solving the diffusion equation.

[0027] S600, a multi-dimensional evidence fusion and spatial coordinate tracing output, extracts the target topology branch locked in S300 and the migration distance calculated in S500. It performs spatial geometric superposition of the one-dimensional line of the branch topology and the relative distance value to determine the physical coordinates of the suspected hidden mining face opening. It integrates static air distribution difference, resistance anomaly, gas residual and explosive consistency evidence to generate tracing output results including the coordinates of the abnormal area and the on-site verification path.

[0028] The process of establishing the mine static ventilation baseline and data space mapping as described in S100 is executed. S100 specifically includes S101 to S103.

[0029] S101, Construct a ventilation network topology diagram and perform data spatial mapping. The central processing computing platform receives the mine ventilation system diagram and the mining engineering plan, and abstracts the mine physical roadway network into a directed ventilation network topology diagram containing a set of nodes, a set of edges, and a set of attributes. In the directed ventilation network topology diagram, the set of nodes represents the physical location of shafts, roadway intersections, air doors, air windows, local ventilation fans, coal mining faces, tunneling faces, drilling sites, chambers, explosives storage facilities, return air confluence points, and sensor measuring points. The set of edges represents air intake channels, return air channels, local ventilation channels, and leakage channels. The set of attributes includes the cross-section, length, wind resistance, design air volume, approved operation status, associated sensors, associated explosives locations, and data reliability of nodes and edges. The central processing computing platform maps each air volume measuring point, methane sensor, local ventilation fan, air duct, and blasting location to the corresponding node or edge in the directed ventilation network topology diagram through unified spatial coding, forming a unified three-dimensional data analysis framework.

[0030] S102. Establish a dynamic baseline for theoretical air demand under legal production conditions. The central processing and computing platform constructs a theoretical air demand baseline for a specific time window based on approved coal mining faces, approved tunneling faces, chambers, underground explosives depots, charging chambers, drilling sites, gas extraction locations, number of underground workers, blasting operations, equipment operation, climate conditions, and air leakage margin. The theoretical air demand baseline is obtained by summing the air demand of approved coal mining faces, approved tunneling faces, chambers, drilling sites, auxiliary systems, air leakage, and margin air demand. The central processing and computing platform dynamically updates all parameters in the theoretical air demand baseline according to shift plans, mining and tunneling status, blasting plans, and equipment operating conditions.

[0031] S103. Construct a dynamic baseline for known gas emission from different sources. The central processing and computing platform performs source-specific modeling for known gas sources, integrating the gas emission from approved coal mining faces, approved tunneling faces, gas emission from drilling sites and extraction disturbances, background gas emission from goaf or confined areas, the influence of the gas extraction system, and sensor background drift. The aforementioned gas emission and influence quantities are added together to generate a total gas emission model from different sources. The known dynamic baseline for gas emission from different sources serves as a comparative reference standard for subsequent identification of unexplained gas residuals.

[0032] The process of injecting wind pressure perturbation signal and starting phase-locked sampling as described in S200 is executed. S200 specifically includes S201 to S204.

[0033] S201, Construct a mathematical model of the wind pressure perturbation signal and set safety constraint parameters. The central processing platform generates the wind pressure perturbation signal for active identification. The mathematical model of the wind pressure perturbation signal is expressed as follows:

[0034] In the formula, This represents the instantaneous wind pressure at the target perturbation node in the ventilation network. This represents the steady-state foundation working wind pressure of the target perturbation node. This indicates the amplitude of minor wind pressure disturbance. Indicates the active perturbation angular frequency. Representing the time variable, to prevent active perturbations from causing underground gas exceedances or airflow reversal accidents, the central processing computing platform sets the perturbation air pressure amplitude to be less than the safe threshold for air pressure fluctuations allowed by the mine safety regulations. At the same time, to ensure that the air pressure perturbation signal can be effectively propagated in the mine roadway network and is not attenuated by noise, the central processing computing platform sets the active perturbation angular frequency to the millihertz level.

[0035] S202 relies on a perturbation actuator to implement physical injection control of the wind pressure perturbation signal. The central processing computing platform converts the constructed wind pressure perturbation signal into electrical control commands and sends the electrical control commands to the perturbation actuator. The perturbation actuator is specifically implemented as a frequency converter deployed in the control cabinet of the main ventilation fan on the mine surface, or as an intelligent regulating damper installed at the ventilation node underground. When the perturbation actuator is a frequency converter, the frequency converter periodically adjusts the motor speed of the main ventilation fan according to the electrical control commands to change the total wind pressure of the entire mine. When the perturbation actuator is an intelligent regulating damper, the servo motor configured inside the intelligent regulating damper periodically fine-tunes the opening angle of the damper blades according to the electrical control commands to change the local ventilation resistance. Through the periodic physical action of the frequency converter or the intelligent regulating damper, the wind pressure perturbation signal is superimposed on the stable airflow of the mine ventilation network.

[0036] S203, configure sensor hardware and send a reference clock signal. The central processing computing platform sends the active perturbation angular frequency as the reference clock signal to the wind speed sensor and methane sensor deployed in each branch roadway of the mine. The central processing computing platform controls the wind speed sensor and methane sensor to synchronize time. The wind speed sensor and methane sensor receive the reference clock signal sent by the central processing computing platform. The wind speed sensor and methane sensor generate a synchronous sampling clock signal based on the reference clock signal. The frequency of the synchronous sampling clock signal is set to an integer multiple of the active perturbation angular frequency.

[0037] S204, execute phase-locked linkage sampling. The wind speed sensor and methane sensor stop the independent timing discrete sampling mode. The wind speed sensor and methane sensor synchronously acquire data with the synchronous sampling clock signal as the trigger. The wind speed sensor acquires wind speed time series data, and the methane sensor synchronously acquires gas concentration time series data. The synchronous sampling operation aligns the wind speed time series data and gas concentration time series data acquired by the wind speed sensor and methane sensor distributed throughout the mine on the physical time axis. The synchronous sampling operation eliminates the signal phase difference caused by the drift of the independent hardware clock of the wind speed sensor and methane sensor. The synchronous sampling operation provides a phase reference for subsequent extraction of perturbation response features in the frequency domain and calculation of residual coherence.

[0038] Perform the measured impedance residual calculation and abnormal branch location as described in S300. S300 specifically includes S301 to S303.

[0039] S301, Perform airflow closure verification and air distribution difference calculation. The central processing platform obtains the measured total airflow and the sum of the measured air distribution volumes at each major air-consuming location within a specific time window. The central processing platform performs a quality verification operation to determine whether the difference between the measured total airflow and the sum of the measured air distribution volumes at each major air-consuming location is within the range of measurement error, leakage estimation, and normal adjustment. If the difference between the measured total airflow and the sum of the measured air distribution volumes at each major air-consuming location is within the range of measurement error, leakage estimation, and normal adjustment, the central processing platform further calculates the air distribution. The air distribution difference is calculated by subtracting the theoretical required air volume baseline from the measured total air volume or total return air volume. When the air distribution difference is continuously positive and the value is greater than the allowable error and safety margin, and the air distribution difference cannot be explained by the approved coal mining face, approved tunneling face, chamber, drilling site, post-blasting smoke exhaust, maintenance ventilation or ventilation adjustment records, the central processing computing platform determines that there is an additional air demand in the mine. The central processing computing platform uses the determination result of the additional air demand as a prerequisite for triggering the system to extract the micro-disturbance response characteristics and start the subsequent impedance residual calculation and abnormal branch location.

[0040] S302. Constructing the measured perturbation response Jacobian matrix and the theoretical Jacobian matrix: Assuming the central processing platform determines there is additional ventilation demand in the mine, it retrieves wind speed time-series data from wind speed sensors distributed throughout the mine, obtained through phase-locked loop sampling. Combining this with the roadway cross-sectional attributes corresponding to the edge sets in the directed ventilation network topology, the platform converts the wind speed time-series data into air volume time-series data. It then extracts the air volume fluctuation amplitude at the active perturbation angular frequency from the air volume time-series data. The platform uses the perturbation pressure amplitude fed back by the perturbation actuator as the system input excitation and the air volume fluctuation amplitude extracted from each ventilation branch as the system output response, constructing the measured perturbation response Jacobian matrix. The mathematical expression of the measured perturbation response Jacobian matrix is ​​as follows:

[0041] In the formula, This represents the measured perturbation response Jacobian matrix. In the directed ventilation network topology graph, the first... The air volume of the ventilation branch affects the first Partial derivatives of wind pressure at each perturbation node This represents the total number of ventilation branches contained in the directed ventilation network topology graph. This represents the total number of nodes in the directed ventilation network topology. Simultaneously, the central processing platform retrieves the directed ventilation network topology under the condition of no concealed mining face access and the theoretical air volume baseline. Based on the fluid dynamics of the flow conservation at network nodes and the loop pressure balance, the central processing platform calculates the partial derivatives of the theoretical air volume of each ventilation branch with respect to the theoretical air pressure of the perturbed nodes. The central processing platform then combines the derived sets of partial derivatives of the theoretical air volume of each ventilation branch with respect to the theoretical air pressure of the perturbed nodes into a theoretical Jacobian matrix. The theoretical Jacobian matrix is ​​symbolically represented as... The measured perturbation response Jacobian matrix and the theoretical Jacobian matrix constitute the data basis for the system to identify abrupt changes in aerodynamic impedance.

[0042] In step S303, impedance residual calculation is performed and the abnormal ventilation branch is located. The central processing platform obtains the measured perturbation response Jacobian matrix and the theoretical Jacobian matrix constructed in step S302. Matrix inversion is performed on both the measured and theoretical Jacobian matrices. The inverted measured perturbation response Jacobian matrix is ​​then subtracted from the inverted theoretical Jacobian matrix to obtain the system's impedance residual matrix. The mathematical expression for the impedance residual matrix is ​​as follows:

[0043] In the formula , represents the impedance residual matrix of the system. This represents the inverted measured perturbation response Jacobian matrix. The theoretical Jacobian matrix after inversion is represented. The central processing computing platform further performs eigenvalue decomposition on the impedance residual matrix, extracts the matrix eigenvectors corresponding to the eigenvalues ​​in the impedance residual matrix that are greater than the preset judgment threshold, locates the matrix column indexes corresponding to the non-zero elements in the matrix eigenvectors, and performs spatial comparison between the matrix column indexes and the directed ventilation network topology map. According to the spatial coding mapping relationship, the matrix column indexes are restored to the specific ventilation topology branches. When the slope of the wind pressure and air volume relationship curve of the specific ventilation topology branch changes abruptly without a reported damper adjustment record or a connection record, the central processing computing platform determines that the specific ventilation topology branch has traces of newly added air passages or unplanned air supply with closed opening. The specific ventilation topology branch is locked as the target topology branch where the suspected hidden mining face is located, and the coordinate information of the target topology branch is used as the basic data for subsequent spatial inversion and multi-dimensional evidence fusion scoring.

[0044] Perform the dispersion channel calibration based on the legal blasting pulse of pyrotechnics as described in S400, which specifically includes S401 to S402.

[0045] S401, executes spatiotemporal mapping and waveform extraction of explosives registration information. The central processing computing platform obtains explosives registration event information, which includes explosives receipt records, return records, usage location, blasting time, blasting operation quantity, blasting personnel, and operation approval records. The central processing computing platform maps the explosives registration event information to a specific ventilation branch of the directed ventilation network topology. When the methane sensor located in the specific ventilation branch detects the gas concentration pulse generated by a legally registered blasting operation, the central processing computing platform extracts the time-series waveform of the gas concentration pulse in phase-locked sampling mode and calculates the waveform broadening variance of the time-series waveform of the gas concentration pulse on the time axis.

[0046] S402, perform dispersion parameter calibration based on the legal blast pulse. The transport law of methane gas in the roadway airflow follows a one-dimensional convection-diffusion equation. The central processing computing platform extracts the physical distance from the known blast location to the downstream methane sensor from the directed ventilation network topology diagram, obtains the average wind speed synchronously collected by the wind speed sensor, and uses waveform broadening variance, physical distance, and average wind speed to calibrate the actual longitudinal dispersion coefficient of a specific ventilation branch under the current operating conditions. The formula for calculating the actual longitudinal dispersion coefficient is expressed as follows:

[0047] In the formula, This represents the actual longitudinal dispersion coefficient. The waveform broadening variance of the time-series waveform representing the gas concentration pulse on the time axis. This represents the average wind speed collected by the wind speed sensor. The physical distance from the legal blasting site to the downstream methane sensor is represented by the central processing computing platform. The actual longitudinal dispersion coefficient is calculated and used as a known physical parameter for the inverse calculation of the physical transport distance of the unexplained gas emission residual.

[0048] The reverse calculation of gas residual dispersion and physical distance calculation described in S500 are executed. S500 specifically includes S501 to S502, which are the execution processes for extracting gas residuals in the zone and coherent filtering and denoising, specifically corresponding to S501.

[0049] S501, performs gas residual extraction and coherence filtering for noise reduction. The central processing platform obtains the total return air volume and methane concentration of the total return air or zoned return air. It multiplies the total return air volume or zoned return air volume by the methane concentration and a unit conversion factor to generate the actual exhaust gas emission. The unexplained gas emission residual is calculated by subtracting the known dynamic baseline of gas emission from different sources and the legal blasting disturbance interpretation amount from the actual exhaust gas emission. When the unexplained gas emission residual is positive and continuously or periodically increases, and cannot be explained by legal mining progress, drilling site construction, extraction adjustments, and explosives reporting, the central processing platform extracts the unexplained gas emission residual. The residual time-series waveform of the explained gas outburst residual in phase-locked sampling mode is used to perform frequency domain conversion on the residual time-series waveform and the wind pressure perturbation signal. The cross-coherence function of the residual time-series waveform and the wind pressure perturbation signal is calculated, and the coherence coefficient value of the cross-coherence function at the active perturbation angular frequency is extracted. When the coherence coefficient value is greater than the set coherence judgment threshold, the central processing computing platform determines that the unexplained gas outburst residual originates from the physical space connected to the mine ventilation network. The residual time-series waveform with active perturbation angular frequency modulation characteristics is confirmed as a valid residual pulse, and the valid residual pulse is output as the target data to be solved to the physical distance reverse solution process.

[0050] The process of performing the inverse solution of gas residual dispersion and physical distance measurement described in S500 is as follows: S500 specifically includes S501 to S502, which describes the execution process of inverse solution of dispersion equation and quantitative measurement of physical distance. Specifically, corresponding to S502, the central processing computing platform receives the effective residual pulse output by S501, extracts the time skewness characteristics of the effective residual pulse, calculates the variance of the residual waveform broadening on the time axis, obtains the average wind speed synchronously collected by the wind speed sensor of the corresponding ventilation branch in the directed ventilation network topology diagram, calls the actual longitudinal dispersion coefficient obtained from the calibration in S400, substitutes the actual longitudinal dispersion coefficient into the inverse solution model of the one-dimensional convection-diffusion equation, and back-calculates the gas source area along the return air branch and sensor concentration gradient to calculate the actual migration distance of the unknown gas source from the downstream methane sensor. The formula for calculating the actual migration distance is expressed as follows:

[0051] In the formula, This indicates the actual migration distance of the unknown gas source from the downstream methane sensor. This represents the variance of the widened residual waveform on the time axis of the effective residual pulse. This represents the average wind speed collected by the wind speed sensor. This indicates the actual longitudinal dispersion coefficient obtained from the calibration. The central processing computing platform completes the calculation of the actual transport distance and outputs the actual transport distance to the multidimensional evidence fusion and spatial coordinate tracing output process.

[0052] The execution of the multi-dimensional evidence fusion and spatial coordinate tracing output described in S600, specifically including S601 to S602, focuses on the execution process of three-dimensional spatial geometric superposition. Specifically, corresponding to S601, the central processing computing platform extracts the target topology branch where the suspected concealed mining face is located, which was locked in S300. Simultaneously, it receives the actual migration distance of the unknown gas source from the downstream methane sensor output by S500, retrieves the mining engineering plan containing the absolute elevation and planar position information of the mine, extracts the three-dimensional spatial orientation curve of the target topology branch in the mining engineering plan, and obtains the reference space of the downstream methane sensor on the three-dimensional spatial orientation curve. Using the reference spatial coordinates as the starting point, the geometric trajectory is tracked along the three-dimensional spatial curve against the direction of ventilation airflow. When the length of the tracked curve is equal to the actual movement distance, the endpoint of the track is confirmed as the exact physical spatial coordinates of the suspected hidden mining face opening. The additional air supply direction obtained by tracing the air supply difference along the path from the air intake to the air supply point to the return air is retrieved. The position of the airway change obtained by finding the resistance anomaly along the air door and the sealed area is extracted. The exact physical spatial coordinates, the additional air supply direction, and the position of the airway change are spatially geometrically intersected within the mining area or at the end of the connecting roadway. The suspected hidden mining face area is generated based on the result of the spatial geometric intersection.

[0053] The multi-dimensional evidence fusion and spatial coordinate tracing output described in S600 are executed. S600 specifically includes S601 to S602, which are the execution processes for the comprehensive scoring of the multi-dimensional evidence chain and the generation of on-site verification paths. Specifically, corresponding to S602, the central processing computing platform obtains the suspected concealed mining face area and its exact physical spatial coordinates generated in S601. It extracts the duration and deviation of the air distribution difference and maps them to the air distribution difference confidence score. It extracts the eigenvalue changes of the target topological branch in the impedance residual matrix and maps them to the impedance residual confidence score. It extracts the cross-coherence coefficient values ​​of the effective residual pulses and maps them to the gas residual dispersion confidence score. The air distribution difference confidence score, impedance residual confidence score, and gas residual dispersion confidence score are multiplied by their respective preset weighting coefficients and then weighted and summed. A comprehensive early warning score is generated for suspected concealed mining faces. If the comprehensive early warning score is greater than a preset score threshold, it is determined that illegal mining operations exist in the suspected concealed mining face area. An alarm command containing precise physical spatial coordinates is generated. The edge set length attribute and roadway passage status attribute in the directed ventilation network topology graph are invoked. Starting from the nearest legal operation location to the suspected concealed mining face area and ending at the precise physical spatial coordinates, a path planning is performed in the directed ventilation network topology graph to generate an on-site verification path to the suspected concealed mining face area. The alarm command, comprehensive early warning score, and on-site verification path are simultaneously sent to the display terminal of the mine surface dispatch control center and the mobile terminal of the underground safety supervisors, guiding the safety supervisors to reach the suspected concealed mining face area for on-site verification along the on-site verification path.

[0054] In another typical operational implementation, the central processing platform acquires in real-time the measured total return air volume of the return airway in the mining area and the sum of the measured air distribution volumes at each major ventilation location within the mining area. If the difference between the measured total return air volume and the sum of the measured air distribution volumes at each major ventilation location falls outside the range of measurement error, leakage estimation, and normal adjustment, the central processing platform determines that there is an air distribution discrepancy in the mining area that cannot be explained by normal ventilation adjustment records. The central processing platform uses this air distribution discrepancy as an early warning trigger condition, sending a millihertz-level wind pressure perturbation signal to the frequency converter driver in the main ventilation fan control cabinet on the ground. The frequency converter driver periodically adjusts the motor speed of the main ventilation fan to superimpose the wind pressure perturbation signal onto the mine ventilation network. The central processing platform then controls... The wind speed sensors and methane sensors distributed within the mining area are time-synchronized and phase-locked loop sampling is initiated. The central processing platform extracts the airflow fluctuation amplitude of each ventilation branch and constructs the measured perturbation response Jacobian matrix and the theoretical Jacobian matrix by combining them with the theoretical required airflow baseline. The impedance residual matrix is ​​obtained by inverting and subtracting the measured perturbation response Jacobian matrix and the theoretical Jacobian matrix. The central processing platform performs eigenvalue decomposition on the impedance residual matrix and extracts the column index, mapping the column index to the target topological branch in the mining area where aerodynamic impedance abrupt changes occur. At the same time, the central processing platform obtains the physical distance and average wind speed from the blasting location to the downstream methane sensor from the legal blasting report, and performs waveform broadening based on the gas concentration pulse generated by the legal blasting. The variance calibration is used to determine the actual longitudinal dispersion coefficient of the return airway in the mining area. The central processing platform extracts the time-series waveform of the unexplained gas emission residual in the total return air of the mining area. After confirming the coherence of the time-series waveform with the wind pressure perturbation signal, the expanded variance of the actual longitudinal dispersion coefficient and the time-series waveform of the unexplained gas emission residual is substituted into the inverse solution model of the one-dimensional convection-diffusion equation to calculate the actual migration distance of the unknown gas source from the downstream methane sensor. The central processing platform extracts the three-dimensional spatial direction curve of the target topology branch in the mining engineering plan. Starting from the downstream methane sensor, the platform traces along the three-dimensional spatial direction curve against the airflow direction to extract the coordinate point equal to the actual migration distance. The coordinate point is then compared with the additional air supply direction pointed to by the air distribution difference. The precise physical spatial coordinates are obtained by performing spatial geometric intersection. The central processing computing platform generates a comprehensive early warning score by combining the confidence scores of the air distribution difference, impedance residuals, and gas residual dispersion. When the comprehensive early warning score is greater than the preset score threshold, an alarm command containing the precise physical spatial coordinates and an on-site verification path are generated. For the specific software implementation logic of matrix inversion, eigenvalue decomposition, graph theory pathfinding algorithm, and one-dimensional convection-diffusion equation solution, those skilled in the art can refer to existing numerical calculation, graph theory, and computational fluid dynamics foundations for programming implementation. Matrix inversion, eigenvalue decomposition, graph theory pathfinding algorithm, and one-dimensional convection-diffusion equation solution are well-known technologies in this field and will not be elaborated here.

[0055] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent identification of concealed mining faces in coal mines based on ventilation and gas coupling, characterized in that, include: S100. Establish the mine static ventilation baseline and data space mapping, and obtain the directed ventilation network topology, theoretical air demand baseline and known gas emission dynamic baseline from different sources. S200, inject wind pressure perturbation signal and start phase-locked linkage sampling to obtain phase-aligned wind speed time series data and gas concentration time series data; S300. Calculate the air distribution difference between the measured total return air volume and the sum of the measured air distribution volumes at each major air-using location. If the air distribution difference exceeds the normal adjustment range, it is determined that there is an additional air demand. Combine the roadway cross-sectional attributes in the directed ventilation network topology diagram to convert the wind speed time series data into air volume time series data. Based on the air volume time series data, construct the measured micro-perturbation response Jacobian matrix. Combine the theoretical Jacobian matrix to perform impedance residual calculation and lock the ventilation branch that has a sudden change in aerodynamic impedance as the target topology branch. S400, based on the time-series waveform of the gas concentration pulse generated by legal blasting, calibrate the actual longitudinal dispersion coefficient of a specific ventilation branch; S500. Based on the gas concentration time series data and the known gas emission dynamic baseline, extract the unexplained gas emission residual, perform coherent filtering on the unexplained gas emission residual to obtain the effective residual pulse, and substitute the actual longitudinal dispersion coefficient into the inverse solution model to calculate the actual migration distance of the unknown gas source from the downstream methane sensor. S600. Extract the three-dimensional spatial direction curve of the target topological branch in the mining engineering plan, and combine it with the actual movement distance to perform spatial geometric intersection to generate the exact physical spatial coordinates of the suspected hidden mining face area.

2. The intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling according to claim 1, characterized in that, The S200 process injects a wind pressure perturbation signal and initiates phase-locked loop sampling to acquire phase-aligned wind speed time series data and gas concentration time series data, including: The active perturbation angular frequency of the periodic wind pressure perturbation signal is used as a reference clock signal and sent to the wind speed sensor and methane sensor. The wind speed sensor and the methane sensor generate a synchronous sampling clock signal with a frequency that is an integer multiple of the active perturbation angular frequency based on the reference clock signal; Synchronous acquisition is performed using the synchronous sampling clock signal as the trigger beat to obtain the wind speed time series data and the gas concentration time series data that are phase-aligned on the physical time axis.

3. The intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling according to claim 1, characterized in that, The process of constructing the measured perturbation response Jacobian matrix based on the air volume time series data, and then calculating the impedance residual using the theoretical Jacobian matrix, includes: Extract the air volume fluctuation amplitude of the air volume time series data at the active perturbation angular frequency; The measured perturbation response Jacobian matrix is ​​constructed by using the amplitude of the micro-perturbation wind pressure as the system input excitation and the amplitude of the air volume fluctuation extracted from each ventilation branch as the system output response. Based on the fluid dynamics relationship between the flow conservation at the network node and the loop pressure balance, the partial derivatives of the theoretical air volume of each ventilation branch with respect to the theoretical pressure at the perturbation node are calculated and combined to form the theoretical Jacobian matrix. The measured perturbation response Jacobian matrix and the theoretical Jacobian matrix are inverted respectively, and the inverted measured perturbation response Jacobian matrix is ​​subtracted from the inverted theoretical Jacobian matrix to obtain the impedance residual matrix.

4. The intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling according to claim 3, characterized in that, The step of locking the ventilation branch where abrupt aerodynamic impedance changes is identified as the target topology branch includes: The impedance residual matrix is ​​subjected to eigenvalue decomposition to extract the matrix feature vectors corresponding to feature values ​​greater than a preset judgment threshold. Locate the column index of the matrix corresponding to the non-zero element in the matrix eigenvector; The matrix column index is spatially compared with the directed ventilation network topology graph, and the matrix column index is restored to a specific ventilation topology branch according to the spatial coding mapping relationship; When it is determined that the slope of the wind pressure-air volume relationship curve of the specific ventilation topology branch undergoes a sudden change without a reported damper adjustment record or a connection record, the specific ventilation topology branch is locked as the target topology branch.

5. The intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling according to claim 1, characterized in that, The time-series waveform of the gas concentration pulse generated by the legal blast is used to calibrate the actual longitudinal dispersion coefficient of a specific ventilation branch, including: Acquire information on pyrotechnic device reporting events and map it to a specific ventilation branch of the directed ventilation network topology graph; When a methane sensor located in the specific ventilation branch detects a gas concentration pulse generated by a legally reported blasting operation, the time-series waveform of the gas concentration pulse in phase-locked sampling mode is extracted, and the waveform broadening variance of the time-series waveform of the gas concentration pulse on the time axis is calculated. Extract the physical distance from the known blasting location to the downstream methane sensor, and obtain the average wind speed synchronously collected by the wind speed sensor; The actual longitudinal dispersion coefficient of the specific ventilation branch under the current operating conditions is calibrated using the waveform broadening variance, the physical distance, and the average wind speed.

6. The intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling according to claim 5, characterized in that, The formula for calibrating the actual longitudinal dispersion coefficient of a specific ventilation branch under current operating conditions using the waveform broadening variance, the physical distance, and the average wind speed is expressed as follows: The actual longitudinal dispersion coefficient is calculated by multiplying the waveform broadening variance by the cube of the average wind speed and then dividing by twice the physical distance.

7. The intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling according to claim 1, characterized in that, The step of extracting unexplained gas emission residuals based on the gas concentration time series data and the known gas emission dynamic baseline, and performing coherent filtering on the unexplained gas emission residuals to obtain effective residual pulses includes: Obtain the total return air volume and methane concentration of the return air in the main area or the return air in the zone, and multiply the total return air volume or the return air in the zone by the methane concentration and the unit conversion factor to generate the actual exhaust gas emission. The unexplained gas emission residual is calculated by subtracting the known dynamic baseline of gas emission from different sources and the explanation amount of legal blasting disturbance from the actual gas emission from ventilation. Extract the residual time-series waveform of the unexplained gas outburst residual, and perform frequency domain conversion on the residual time-series waveform and the periodic wind pressure perturbation signal; Calculate the cross-coherence function between the residual time-series waveform and the periodic wind pressure perturbation signal, and extract the coherence coefficient value of the cross-coherence function at the active perturbation angular frequency; When the coherence coefficient value is greater than the set coherence determination threshold, the residual timing waveform with active perturbation angular frequency modulation characteristics is confirmed as the effective residual pulse.

8. The intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling according to claim 1, characterized in that, The step of substituting the actual longitudinal dispersion coefficient into the inverse calculation model to calculate the actual migration distance of the unknown gas source from the downstream methane sensor includes: Extract the time skewness characteristics of the effective residual pulse, and calculate the residual waveform broadening variance of the effective residual pulse on the time axis; The actual longitudinal dispersion coefficient is substituted into the inverse solution model of the one-dimensional convection-diffusion equation; The actual migration distance of the unknown gas source from the downstream methane sensor is calculated by multiplying the variance of the stretched residual waveform by the cube of the average wind speed and then dividing by twice the actual longitudinal dispersion coefficient.

9. The intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling according to claim 1, characterized in that, The step of combining the actual transport distance with spatial geometric intersection to generate the exact physical spatial coordinates of the suspected concealed mining face area includes: Obtain the reference spatial coordinates of the downstream methane sensor on the three-dimensional spatial trajectory curve; Starting from the reference spatial coordinates, a geometric trajectory is tracked along the three-dimensional spatial curve against the direction of the ventilation airflow. When the length of the tracked curve is equal to the actual movement distance, the endpoint of the track is confirmed as the exact physical spatial coordinates. The additional air usage direction is obtained by tracing the air supply difference along the path from the air inlet to the air usage point to the return air, and the location of the air path change is obtained by extracting the resistance anomaly along the damper and the sealed area. By performing a spatial geometric intersection of the exact physical space coordinates, the additional wind direction, and the location of the wind path change, a suspected concealed mining face area is generated. Extract the duration and deviation of the air distribution difference and map them as the air distribution difference confidence score; extract the eigenvalue change of the target topological branch in the impedance residual matrix and map it as the impedance residual confidence score; extract the cross-coherence coefficient value of the effective residual pulse and map it as the gas residual dispersion confidence score. The confidence scores of the air distribution difference, the impedance residual, and the gas residual dispersion are multiplied by their respective preset weighting coefficients and then summed to generate a comprehensive early warning score for suspected hidden mining faces. When the comprehensive early warning score is greater than the preset score threshold, path planning is performed in the directed ventilation network topology map to generate a field verification path leading to the suspected hidden mining face area.

10. A smart identification system for concealed mining faces in coal mines based on ventilation and gas coupling, characterized in that, include: The mine environment sensing network includes wind speed sensors and methane sensors deployed in each branch roadway to collect phase-aligned wind speed time series data and methane concentration time series data on the physical time axis. The perturbation actuator is used to inject periodic air pressure perturbation signals into the mine ventilation network; The central processing computing platform is communicatively connected to the mine environment sensing network and the micro-perturbation actuator, respectively. The central processing computing platform is configured to execute the intelligent identification method for concealed mining faces in coal mines based on ventilation and gas coupling as described in any one of claims 1 to 9.

Images