Intelligent inspection system for gas in coal mine based on laser perception

Abstract

This invention discloses an intelligent methane inspection system for underground coal mines based on laser sensing, belonging to the field of big data processing. It includes: a laser sensing acquisition module, an anomaly extraction module, a time-delay alignment module, a source determination module, a predictive diagnosis module, and an early warning and scheduling module. This invention improves the accuracy of anomaly source and diffusion range determination by extracting anomalies, estimating propagation time delays, and aligning time data from multiple monitoring points along the airflow path using laser sensors. After alignment, it determines the anomaly's starting point, diffusion direction, and diffusion range. This solves the problem in existing technologies where the lack of consideration for propagation time differences between measuring points leads to downstream high-concentration methane accumulation being misjudged as an anomaly source.

Description

Technical Field

[0001] This invention relates to the field of big data processing technology, and in particular to an intelligent gas inspection system for underground coal mines based on laser sensing. Background Technology

[0002] In underground coal mine operations, complex environmental conditions and variable gas compositions often harbor serious safety hazards. In particular, hazardous factors such as methane and coal dust, if their concentrations are abnormal, can easily trigger major accidents such as explosions and poisoning, threatening the lives of underground workers and causing enormous economic losses. Therefore, constructing an efficient, accurate, and stable coal mine methane inspection system has become a crucial measure to ensure safe coal mine production.

[0003] Existing technologies utilize laser methane sensors and substations deployed in areas such as the coal face, upper corner, and return airway to achieve continuous methane concentration detection, data upload, and centralized processing at a ground-based central station. By validating the monitoring data, removing anomalies, performing segmental aggregation analysis, and trend judgment, these technologies enable tiered early warning, intelligent generation of inspection tasks, online or offline inspection verification, alarm linkage, power outage interlocking, personnel location tracking, emergency broadcasting, and historical database updates. Furthermore, related coal mine safety monitoring platforms achieve a closed-loop system for gas risk management and intelligent inspection management through tiered early warning, alarm linkage, power outage interlocking, area control, personnel location tracking, inspection verification, and historical database updates.

[0004] For example, Chinese invention patent CN117808274B discloses an intelligent inspection system for underground gas safety in coal mines, which includes: an inspection position verification module, an inspection position adjustment module, a gas safety analysis module, a gas safety comprehensive analysis module, and an inspection operation quality evaluation module. By analyzing the positional offset index between the gas inspector and the inspection point, adjusting the deviation position, collecting data information from the inspection point, analyzing the comprehensive gas safety index of each inspection route, and finally analyzing the inspection operation quality corresponding to the target coal mine underground.

[0005] The above-mentioned technology has at least the following technical problems: In existing technologies, underground gas monitoring in coal mines typically involves deploying multiple monitoring points at locations such as the coal face, upper corner, and return airway. The source and scope of anomalies are analyzed by combining the concentration changes, duration, and exceedance ranges at each monitoring point. However, in the actual underground environment, gas does not appear simultaneously at all monitoring points. Instead, it gradually propagates from upstream to downstream along the airflow generated by the ventilation system. This propagation process is also affected by changes in airflow, the start and stop of local ventilation fans, differences in roadway structure, and changes in mining operations, resulting in a natural time difference in propagation between different monitoring points that varies with operating conditions. If only the concentration values ​​of multiple monitoring points at the same moment are directly compared and correlated, subsequent high-concentration accumulations at downstream monitoring points may be misjudged as the starting point of anomalies. For example, after gas is released from the coal wall near a coal face, the gas first appears on the face side and then propagates along the return air direction to the upper corner and return airway. If the system does not consider this propagation time and only sees that the value of the upper corner measuring point is higher than that of the working face measuring point at a certain moment, it may mistakenly believe that the upper corner is the source of the anomaly, which will lead to inaccurate judgment of the source of the anomaly, the direction of spread, and the direction of emergency response during inspection. Summary of the Invention

[0006] To address the technical problems of existing technologies that fail to consider the abnormal sources, diffusion directions, and inaccurate judgment of emergency response directions during inspections due to airflow, this invention provides a laser-sensing-based intelligent gas inspection system for underground coal mines. The technical solution is as follows: The laser sensing acquisition module is used to generate a laser methane sensing sequence, a multi-parameter synchronization sequence, and an airflow path sequence based on the sequence of monitoring points along the airflow direction.

[0007] The anomaly extraction module is used to determine methane anomaly events at each monitoring point based on the laser methane sensing sequence.

[0008] The time-delay alignment module is used to estimate the propagation time delay based on methane anomaly events and airflow path sequences, and to time-align the laser methane sensing sequence based on the propagation time delay.

[0009] The source determination module is used to determine the anomaly initiation point and the diffusion range along the airflow path based on the time-aligned laser methane sensing sequence.

[0010] The predictive diagnostic module is used to determine the methane prediction result and the laser sensor status result based on the time-aligned laser methane sensing sequence and the multi-parameter synchronization sequence.

[0011] The early warning and scheduling module is used to generate inspection tasks, early warning results, and alarm results based on the anomaly initiation point, diffusion range, methane prediction results, and laser sensor status results.

[0012] The beneficial effects of the technical solution provided by this invention include at least the following: 1. The intelligent methane detection system for underground coal mines based on laser sensing provided by this invention deploys laser methane sensing sequences at key points along the airflow path, such as the coal face, upper corner, and return airway in underground coal mines. Based on the collected data, it extracts abnormal events, estimates propagation time delays, and aligns the data with time. After time alignment, it determines the starting point, diffusion direction, and diffusion range of the anomaly. This transforms the comparison benchmark of multiple measurement points from the same clock moment to the same propagation stage, resulting in more accurate identification of the anomaly source, more reliable identification of the diffusion direction, and more reasonable delineation of inspection focus and early warning area. This effectively solves the problem in existing technologies where multiple measurement point data are not aligned with the airflow propagation time difference, leading to the misjudgment of downstream high-concentration accumulation as the starting point of the anomaly.

[0013] 2. This invention inputs a time-aligned laser methane sensing sequence along with temperature, humidity, local airflow velocity, gas pressure, and other gas parameters into a lightweight neural network. Combined with propagation consistency constraints, it performs advanced prediction and sensor self-diagnosis, thereby providing early warnings before anomalies develop into measured exceedances. It also reduces the interference of abnormal sensors on source location and inspection decisions, thus achieving the technical effects of issuing inspection tasks earlier and improving the reliability of underground gas inspection and handling. This effectively solves the problems of existing technologies that rely solely on current monitoring results for passive handling and are easily affected by interference from abnormal measuring points. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A structural diagram of an intelligent coal mine gas inspection system based on laser sensing provided in an embodiment of the present invention; Figure 2 A flowchart of a laser-sensing-based intelligent gas inspection system for underground coal mines, provided as an embodiment of the present invention; Figure 3 A block diagram of lightweight neural network construction and sensor self-diagnosis provided in this embodiment of the invention; Figure 4 This is a schematic diagram of a visual real-time monitoring interface provided in an embodiment of the present invention. Detailed Implementation

[0016] Embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present disclosure are shown in the drawings, it should be understood that embodiments of the present disclosure may be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure.

[0017] It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure. In the description of the embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "this embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., can refer to different or the same objects. Furthermore, the specific number of terms such as "continuous sampling period" and "continuous effective monitoring points" described in the embodiments are freely set by those skilled in the art based on actual circumstances, provided that the disclosed content is met. The content disclosed in this invention is only illustrative and does not limit the specific number.

[0018] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0019] like Figure 1 The diagram shown illustrates the structure of a laser-sensing-based intelligent gas inspection system for underground coal mines, as provided in this embodiment of the invention. It includes an edge server comprised of master-slave data servers, a web server, operator stations, and engineer stations; a network switch; and a 10-gigabit industrial Ethernet ring network platform composed of 10-gigabit switches for connecting the surface and underground. Additionally, it includes a data acquisition gateway, a multi-parameter host, a backup power supply, and a display screen, facilitating timely monitoring of the overall safety situation underground by back-end operators.

[0020] Before executing specific operations, the system first establishes the airflow path relationship between monitoring points along the actual airflow direction in the coal mine, and sequentially numbers each monitoring point located on the same path. For example, in a typical working face scenario, the monitoring point of the coal mining face can be designated as the upstream monitoring point, the upper corner monitoring point as the midstream monitoring point, and the return airway monitoring point as the downstream monitoring point. Intrinsically safe laser multi-parameter monitoring hosts are densely deployed at key locations such as underground coal mining faces, tunneling faces, and return airways to achieve real-time perception of multi-dimensional parameters of the underground environment, including laser gas sensors, temperature and humidity sensors, carbon monoxide sensors, probe oxygen sensors, and sensors for carbon dioxide, hydrogen sulfide, and ethylene. Simultaneously, local airflow velocity and air pressure information near the monitoring points are collected. The local airflow velocity mainly refers to the actual airflow velocity formed by the underground ventilation system in the upstream and downstream adjacent ranges near the current monitoring point, which is used to reflect the speed of gas transport and accumulation tendency near the monitoring point.

[0021] Furthermore, to enable real-time and rapid data transmission from underground coal mines, the system employs both wired and wireless transmission methods. Wired transmission utilizes a mine-grade flame-retardant optical cable to construct an industrial Ethernet network, with explosion-proof and intrinsically safe fiber optic Ethernet switches deployed in main roadways and nodes to ensure rapid data aggregation and transmission. Wireless transmission addresses the needs of manual inspections and temporary monitoring by employing mine-grade Wi-Fi, 4G, or 5G communication technologies to facilitate data interaction between handheld terminals and the monitoring center, suitable for areas where cabling is difficult. After configuring the hardware, the system begins collecting and analyzing data.

[0022] like Figure 2 The diagram shown is a flowchart of a laser-sensing-based intelligent gas inspection system for underground coal mines provided in an embodiment of the present invention, including the following: The laser sensing and acquisition module is used to perform laser methane sensing and multi-parameter synchronous acquisition at each monitoring point. The laser gas sensor emits detection light into the characteristic absorption band of methane and receives the received light signal after passing through the gas to be measured. This received light signal is compared with a reference light signal pre-calibrated under the same optical path conditions. Then, based on the pre-established correspondence between the reference light signal and the methane concentration calibration, the original methane concentration for the current sampling period is determined. The characteristic absorption band of methane is determined as follows: During the system deployment or calibration phase, a certain number of candidate absorption bands are pre-selected based on the methane absorption spectrum; under fixed optical path conditions, zero gas and standard methane gas of different concentrations are introduced into each candidate absorption band, and the received light signal is continuously collected for no less than 5 sampling periods at each standard methane gas concentration. The maximum and minimum values ​​are removed, and the average value is calculated; the average received light signal is compared with the corresponding reference light signal. For the same candidate absorption band, if the comparison results obtained from the received optical signal and the reference optical signal always change in the same direction as the standard methane gas concentration increases stepwise according to the concentration nodes, and the difference between the comparison results of adjacent concentration nodes is not less than the minimum distinguishable change corresponding to the sensor resolution, then the candidate absorption band is considered to be usable for distinguishing different methane concentrations. Subsequently, under the same standard methane gas concentration, the temperature and humidity values ​​were changed respectively, and coexisting gases such as carbon monoxide and carbon dioxide of preset concentrations were introduced. The received optical signals were repeatedly collected, and the original methane concentration was calculated in the aforementioned manner. The candidate absorption band with the smaller absolute value of the difference and the smaller dispersion of the results of multiple repeated acquisitions is the characteristic absorption band of methane.

[0023] The method for establishing the correspondence between the reference optical signal and the methane concentration calibration is as follows: During the system deployment or calibration phase, under fixed optical path conditions, zero gas is first introduced into the laser gas sensor, and the received optical signal is continuously collected for no less than 5 sampling periods. After removing the maximum and minimum values, the average value is calculated and recorded as the reference optical signal under this optical path condition. Next, multiple standard methane gases of known concentrations are introduced into the laser gas sensor, and the received optical signal is continuously collected for no less than 5 sampling periods at each standard methane gas concentration. After removing the maximum and minimum values, the average value is calculated and compared with the reference optical signal. The comparison result is then recorded one-to-one with the corresponding standard methane gas concentration. After completing the above operations for all standard methane gas concentrations, a set of calibration data consisting of the comparison results and their corresponding methane concentrations is obtained. The system writes the calibration data into the calibration table in order of comparison results. When the result obtained after comparing the received optical signal with the reference optical signal in a certain sampling period during actual operation falls exactly on an existing record in the calibration table, the corresponding methane concentration is directly read as the original methane concentration. When the result falls between two adjacent calibration records, the methane concentrations corresponding to the two adjacent calibration records are linearly interpolated to obtain the original methane concentration of the current sampling period. When the result exceeds the coverage of the calibration table, the methane concentration corresponding to the record closest to the boundary is used, and a calibration boundary mark is added to the sampling period.

[0024] It should be noted that, since changes in temperature and humidity can disturb the gas state and detection results, the system pre-establishes a compensation table mapping temperature, humidity, and methane concentration corrections. Based on the temperature, humidity, and original methane concentration of the current sampling period, the corresponding signed methane concentration correction is looked up in the compensation table. The original methane concentration is then directly added to the methane concentration correction to obtain the compensated laser methane sensing value. When the sum is less than 0, it is recorded as 0; when the sum exceeds the upper limit of the device's range, it is recorded as the upper limit. Specifically, during the system deployment or calibration phase, the applicable temperature operating range, humidity operating range, and methane concentration measurement range of the device are first determined, and temperature calibration nodes, humidity calibration nodes, and methane concentration calibration nodes are set respectively. Specifically, temperature calibration nodes should cover the temperature range that the equipment may encounter when used downhole, with an interval of no more than 5°C between adjacent temperature calibration nodes. Humidity calibration nodes should cover the relative humidity range that the equipment may encounter when used downhole, with an interval of no more than 10%RH between adjacent humidity calibration nodes. Methane concentration calibration nodes should at least cover the zero point, the vicinity of the warning threshold, the vicinity of the alarm threshold, and several concentration intervals within the measurement range. It should be noted that the intervals between adjacent temperature and humidity calibration nodes can be preset according to the applicable environment of the equipment, the compensation accuracy requirements, and the calibration workload. This embodiment is only an example for illustration, and technicians can adjust it according to the actual situation.

[0025] Subsequently, under each combination of temperature, humidity, and methane concentration calibration nodes, standard methane gas of the corresponding concentration is introduced into the laser gas sensor. After the detected value stabilizes, the raw methane concentration is continuously collected for no less than 5 sampling periods. The maximum and minimum values ​​are removed, and the average value is taken as the calibration measurement value under that combination condition. The difference between the standard methane gas concentration and the calibration measurement value is determined as the methane concentration correction amount under that combination condition, and the temperature, humidity, and methane concentration calibration nodes, along with their corresponding methane concentration correction amounts, are written into a compensation table. When the current temperature, humidity, or raw methane concentration falls between two adjacent calibration nodes, the system uses the linear interpolation result of the correction amount corresponding to the adjacent nodes as the current correction amount; when the current temperature, humidity, or raw methane concentration exceeds the coverage range of the compensation table, the system uses the correction amount closest to the boundary node and adds a compensation boundary marker to the data of that sampling period.

[0026] Simultaneously, temperature, humidity, carbon monoxide, oxygen, carbon dioxide, hydrogen sulfide, ethylene, as well as local airflow velocity and air pressure, are uniformly mapped onto the same time axis and organized according to a pre-established airflow path sequence. Specifically, a unified sampling period is first set, and a unified sampling time sequence is formed according to this unified sampling period. For parameters with multiple sampled values ​​within the current sampling period, the average value within that period is used; for parameters with only one sampled value within the current sampling period, that value is directly used; for parameters with no valid sampled value at the current sampling time, and whose consecutive missing duration from the current sampling time forward does not exceed the allowable missing duration, it is recorded as short-term missing, and the previous valid value is retained; for parameters with no valid sampled value at the current sampling time, and whose consecutive missing duration from the current sampling time forward exceeds the allowable missing duration, it is recorded as continuous missing and marked as invalid data. The allowable missing duration is pre-calculated and set by technicians before system deployment. Its calculation basis includes the unified sampling period and the maximum number of sampling periods that the parameter is allowed to have consecutive missing. The allowable missing duration is equal to the product of the unified sampling period and the maximum number of sampling periods. For core parameters directly involved in anomaly extraction, source location, risk confirmation, or alarm judgment, such as laser-induced methane, oxygen, and carbon monoxide, the maximum number of sampling periods is preferably 1 to 2. For auxiliary parameters used only for environmental compensation, auxiliary interpretation, or prediction enhancement, such as temperature, humidity, air pressure, carbon dioxide, and local airflow velocity, the maximum number of sampling periods is preferably 2 to 3. Taking a uniform sampling period of 10 seconds as an example, if the maximum sampling period for the laser-induced methane parameter is 1, its allowable missing time is 10 seconds; if the maximum sampling period for the oxygen parameter is 2, its allowable missing time is 20 seconds; and if the maximum sampling period for the temperature parameter is 3, its allowable missing time is 30 seconds.

[0027] For common data quality issues such as missing values, occasional spikes, or short-term communication jitter, only necessary engineering processing is performed. For example, short-term previous value preservation or neighborhood interpolation is used to ensure the continuity of the time axis. Those skilled in the art should be able to understand and implement this, and it will not be elaborated further in this embodiment. After the above processing, three types of effective sequences under a unified time axis can be obtained. The first is the laser methane sensing sequence, arranged chronologically, with each record including the monitoring point identifier, sampling time, and compensated laser methane sensing value. The second is the multi-parameter synchronization sequence, arranged chronologically, with each record including the monitoring point identifier, sampling time, and the corresponding temperature, humidity, carbon monoxide, oxygen, carbon dioxide, hydrogen sulfide, ethylene, local airflow velocity, and air pressure values. The third is the airflow path sequence, arranged according to the airflow direction, with each record including the path identifier, monitoring point identifier, and the sequential position of the monitoring point in the path, providing a consistent data foundation for subsequent anomaly extraction and propagation relationship analysis.

[0028] The anomaly extraction module is used to extract abnormal events based on the laser methane sensing sequence. To avoid misjudging minor daily fluctuations at monitoring points as anomalies, a historical baseline window is set for each monitoring point. The historical baseline window refers to the interval of compensated laser methane sensing values ​​recorded over the most recent consecutive sampling periods before the current sampling time. When constructing this window, records corresponding to confirmed abnormal events and data records marked as invalid are removed. Then, the median or the mean after removing extreme high and low values ​​is used to obtain the current baseline methane value for that monitoring point.

[0029] Next, the compensated laser methane sensing value for the current sampling period is compared with the sum of the current baseline methane value and the preset methane rise threshold for that monitoring point. Both the methane rise threshold and the current baseline methane value are methane concentration values. Their sum forms the abnormal rise judgment line for the current sampling period at that monitoring point. Historical normal fluctuation ranges are statistically obtained from the system's historical normal operation data for that monitoring point. Specifically, the system selects historical data that did not trigger methane alarms, did not have sensor failure flags, was not under equipment maintenance, and was not during ventilation switching periods as normal samples. At each normal sample sampling time, the deviation of the compensated laser methane sensing value relative to the corresponding baseline methane value is calculated. The deviations of the same monitoring point within the preset statistical period are sorted by value, and the statistical value closest to the upper bound is taken as the historical normal fluctuation range for that monitoring point. For example, a percentile statistical value or the maximum deviation after removing abnormally high values ​​is used. The methane rise threshold is a preset base threshold, which can be corrected by combining the historical normal fluctuation range of the monitoring point and the sensor resolution. Specifically, during system deployment, a base methane rise threshold is first set for each monitoring point. After obtaining the historical normal fluctuation range, the base methane rise threshold, a preset multiple of the historical normal fluctuation range, and a preset multiple of the sensor resolution are compared, and the larger of these is taken as the current preset methane rise threshold used for that monitoring point. If a monitoring point has not yet accumulated enough historical normal operation data, the larger of the base methane rise threshold and the preset multiple of the sensor resolution is temporarily used until the number of normal samples reaches the preset number, at which point a correction is performed. When the laser methane sensing value of a monitoring point is higher than the abnormal rise judgment line of that monitoring point for N consecutive sampling cycles, it is determined that an abnormal methane rise event has occurred at that point. The system immediately records the start time, peak time, and event window of the event. The start time is the beginning of the consecutive occurrence of the anomaly, the peak time is the highest value of the laser methane sensing value in the event window of the anomaly compared to other sampling times, and the event window is the entire time range from the start time to the end time of the anomaly. It should be noted that the event window is different from the baseline window. The former is used to describe the current abnormal process, while the latter is used to characterize the recent normal level before the abnormality occurred.

[0030] While identifying laser-induced methane anomalies, values ​​for carbon monoxide, oxygen, carbon dioxide, hydrogen sulfide, ethylene, local airflow velocity, and air pressure are extracted from the event window corresponding to the detection point. These values ​​serve as auxiliary inputs for subsequent risk confirmation and short-term prediction. After the above operations, a set of laser-induced methane anomalies for each monitoring point is obtained. Each record in the set includes the monitoring point identifier, the corresponding airflow path identifier, the event start time, peak time, event end time, event window, event peak value, and the multi-parameter event window corresponding to that event.

[0031] The time delay alignment module is used to perform propagation time delay estimation and time alignment along the airflow path. Since the same gas stream requires actual propagation time to travel from upstream to downstream, the laser methane sensing values ​​at different measuring points at the same moment cannot be directly compared. Instead, the propagation time delay between the upstream and downstream monitoring points should be estimated first, and then the delayed response from the downstream should be adjusted back to a unified reference time before comparison.

[0032] In practice, for each pair of adjacent monitoring points along the same path, methane anomaly events at both upstream and downstream monitoring points are first merged into common propagation events. Specifically, using each methane anomaly event at an upstream monitoring point as a baseline, a candidate event search window is established in the set of methane anomaly events at downstream monitoring points according to the event's start time. The window's start time is one sampling period after the start time of the upstream anomaly event, and its end time is the start time of the upstream anomaly event plus the maximum propagation time delay. At downstream monitoring points, if the event's start time falls within the candidate event search window, it is considered a candidate downstream anomaly event. When multiple candidate downstream anomalies exist, the time difference between the start time of each candidate downstream anomaly event and the start time of the upstream anomaly event is calculated, and the event window of the candidate downstream anomaly event is pushed back according to this time difference. Then, on a unified sampling time sequence, the number of sampling times jointly covered by the event window of the upstream anomaly event and the event window of the pushed-back candidate downstream anomaly event is counted. The more sampling times jointly covered, the higher the degree of correspondence between the candidate downstream anomaly event and the upstream anomaly event after the pushback. The system prioritizes the candidate downstream abnormal event with the largest number of shared coverage sampling times as the corresponding downstream abnormal event; when the number of shared coverage sampling times is the same, the event with the smaller peak time difference and the smaller starting time difference are selected as the corresponding downstream abnormal event in sequence.

[0033] Each downstream anomaly is only allowed to establish a correspondence with one upstream anomaly in the same path's adjacent point pair. Downstream anomalies that have already established a correspondence will not participate in pairing with other upstream anomalies within the same point pair. If no candidate downstream anomaly is found, the upstream anomaly is marked as not having a corresponding downstream event in the adjacent point pair. After the above merging, propagation event pairs between adjacent monitoring points are established, including the upstream anomaly, the corresponding downstream anomaly, the event start time difference, the event peak time difference, and the event window correspondence. Subsequent propagation delay estimation, supplementary propagation delay calculation, peak time deviation calculation, and source localization are only performed on anomalies for which propagation event pairs have already been established.

[0034] Next, the incremental sequence is calculated from the laser methane sensing sequence. The incremental sequence is formed by arranging the differences between the laser methane sensing values ​​of two adjacent sampling times in chronological order; that is, subtracting the laser methane sensing value of the previous sampling time from the laser methane sensing value of the next sampling time. The resulting differences are recorded sequentially according to the sampling times to characterize the magnitude of methane change trends. Subsequently, within the preset maximum propagation delay range, candidate delays are generated one by one according to a uniform sampling period. Candidate delays refer to the various time delay values ​​to be tested that are above 0 and do not exceed the maximum propagation delay range. For example, when the uniform sampling period is 10 seconds and the maximum propagation delay range is 60 seconds, the candidate delays are 10 seconds, 20 seconds, 30 seconds, 40 seconds, 50 seconds, and 60 seconds respectively. For each candidate delay, the incremental sequence of the downstream monitoring point is shifted backward according to the candidate delay, and then propagation matching is performed with the incremental sequence of the upstream monitoring point within the same window. The system uses fixed rules for judgment, namely, calculating the following quantitative indicators respectively: First, the consistency rate of change direction. For each valid comparison moment within the comparison window, if the increment of the upstream monitoring point and the increment of the downstream monitoring point after the pullback are both positive or both negative, it is recorded as consistent direction; if one is positive and the other is negative, it is recorded as inconsistent direction; when the absolute value of either increment is less than the preset increment noise threshold, that moment is not included in the comparison. The consistency rate of change direction is the proportion of moments with consistent direction to the total number of valid comparison moments. Second, the rise time deviation. Within the comparison window, the moment when a monitoring point first experiences two consecutive sampling time increments not less than the preset rise judgment threshold is recorded as the rise time of that monitoring point within the window. The absolute value of the difference between the rise time of the upstream monitoring point and the rise time of the downstream monitoring point after the pullback is the rise time deviation. Third, the peak time deviation. Within the comparison window, the moment when the laser methane sensing value of a monitoring point reaches its maximum value is recorded as the peak time of that monitoring point within the window. The absolute value of the difference between the peak time of the upstream monitoring point and the peak time of the downstream monitoring point after the pullback is the peak time deviation. Fourth, the overlap rate of effective change segments. From the moment of rise until the moment when the absolute value of the increment of two consecutive sampling moments is less than the preset stability judgment threshold for the first time after the peak, the effective change segment of the monitoring point within the current comparison window is defined as the effective change segment of the monitoring point. The overlap time of the effective change segment of the upstream monitoring point and the effective change segment of the downstream monitoring point after the pullback, divided by the union time of the two, is the effective change segment overlap rate.

[0035] A candidate time delay is considered valid when it simultaneously meets the following propagation matching conditions: the consistency rate of change direction is not less than 70%, the deviation of the rise time does not exceed 2 sampling periods, the deviation of the peak time does not exceed 2 sampling periods, and the overlap rate of the effective change segment is not less than 60%. Among all valid candidate time delays, the one with the highest consistency rate of change direction is preferred as the propagation time delay for the pair of adjacent monitoring points; when there are ties, the one with the smaller rise time deviation, the one with the smaller peak time deviation, and the one with the smaller candidate time delay are preferred in that order. If no valid candidate time delay that meets the propagation matching conditions is found within the maximum propagation time delay range, the start time of the corresponding abnormal event at the downstream monitoring point is subtracted from the start time of the corresponding abnormal event at the upstream monitoring point as the supplementary propagation time delay for the pair of adjacent monitoring points; where the corresponding abnormal event refers to the upstream abnormal event and the corresponding downstream abnormal event that have established a propagation event pair, and the start time difference refers to the time difference between the start time of the abnormal event at the downstream monitoring point and the start time of the abnormal event at the upstream monitoring point in the same propagation event on the same path. If the start time difference is negative or exceeds the maximum propagation time delay range, the point will not participate in source localization in this period. If no propagation event pair is established between adjacent monitoring points, then the adjacent point pair will not participate in the supplementary propagation delay calculation and source location within this cycle.

[0036] For adjacent point pairs with confirmed effective propagation delays that are not marked as not participating in source localization, their propagation delays are accumulated segment by segment according to the airflow direction to obtain the cumulative propagation delay of each monitoring point relative to the path reference point. Subsequently, the monitoring point with the earliest sequential position on the same airflow path is taken as the upstream reference point of the path, and the start time of the abnormal event corresponding to the upstream reference point of the path is taken as the unified reference time. The abnormal event corresponding to the upstream reference point of the path should be the starting abnormal event in the current propagation event chain. When there are multiple propagation event chains within the same path, the cumulative propagation delay calculation and time alignment are performed on each propagation event chain separately to avoid mixing different propagation event chains. The laser methane sensing sequences of each monitoring point are reversed according to their respective cumulative propagation delays to obtain the time-aligned laser methane sensing sequences. When the cumulative propagation delay is exactly an integer multiple of the unified sampling period, it is directly reversed according to the corresponding sampling steps; when the cumulative propagation delay is not an integer multiple of the unified sampling period, the corresponding value after reversal is determined by linear interpolation between adjacent sampling records. After this step, the multiple abnormal responses that originally appeared at different times were restored to the same propagation reference time, providing a correct temporal basis for subsequent source identification.

[0037] The source determination module is used to determine the source of an anomaly and the diffusion range along the airflow path based on the time-aligned path sequence. First, it determines continuous effective monitoring points based on the availability of adjacent point pairs in the time-delay alignment module. A continuous effective monitoring point refers to a segment of continuous monitoring points along the same airflow path where adjacent point pairs all have reliable propagation time delays and there are no breaks in the chain. The system pre-sets a preset number of continuous effective monitoring points to represent the minimum number of continuous effective monitoring points required for source location. The preset number can be set according to the monitoring point density along the airflow path, the spacing between adjacent monitoring points, and the minimum number of comparison objects required for source location. In this embodiment, it is preferred to set the preset number to 2. For example, when both the point pairs from the working face to the upper corner and from the upper corner to the return airway are available for location, the working face, upper corner, and return airway constitute a set of continuous effective monitoring points; if any of these point pairs is unavailable, the continuous chain is broken at that point. The system can only perform source localization when the number of consecutive valid monitoring points on the same path is not less than the preset number. Otherwise, the system only outputs the single-point anomaly pending verification result and does not directly provide a source determination. For paths where the number of consecutive valid monitoring points on the same airflow path is not less than the preset number, and the adjacent point pairs between these consecutive valid monitoring points have obtained propagation delays that can be used for localization and completed time alignment in the time-delay alignment module, within the same event window after time alignment, the system sequentially checks the methane status of each valid monitoring point from upstream to downstream along the airflow direction. When a monitoring point first reaches the methane warning threshold or methane alarm threshold, and its upstream adjacent valid monitoring points have not yet reached the same level threshold at the same reference time, that monitoring point is determined as the starting point of this abnormal event. It should be noted that when the abnormal starting point reaches the alarm threshold, the alarm threshold should be used as the same level threshold for this abnormal event, and only the downstream consecutive valid monitoring points that have reached the alarm threshold are included in the alarm-level diffusion range; for monitoring points that have only reached the warning threshold but not the alarm threshold, they can be recorded as warning impact points or peripheral impact areas, and are not included in the alarm-level diffusion range.

[0038] After identifying the initial point of the anomaly, downstream continuous effective monitoring points are checked along the airflow direction. All downstream monitoring points reaching the same threshold level within the same event window are sequentially included in the anomaly's spread range until the first point fails to reach the same threshold level or the point-to-point connection is interrupted. A unified threshold system is used for both the methane warning threshold and the methane alarm threshold. The alarm threshold is set according to applicable mine safety regulations and on-site procedures, while the warning threshold is set below the alarm threshold. The methane sensor alarm concentration at the return air corner of the coal face, in low-gas and high-gas mines, and in the return air roadway of the coal face is ≥1.0%. Therefore, this embodiment sets the corresponding methane warning threshold to 0.8% for example, to organize inspections and increase the priority of handling before reaching the official alarm line.

[0039] After determining the source and diffusion range, the system further confirms the risk of the anomaly. For the risk of hypoxia, the system primarily judges whether the absolute oxygen concentration is below a preset hypoxia threshold, and may add the condition that it is met for several consecutive sampling periods to avoid spurious misjudgments. In this embodiment, an oxygen volume fraction below 20% is used as the hypoxia warning reference line, and corresponding alarm lines are set in conjunction with on-site management requirements. For the risk of combustion, based on the establishment of a methane anomaly, it is required that within the same event window, the carbon monoxide concentration for several consecutive sampling periods be higher than the sum of the current baseline carbon monoxide value and the carbon monoxide rise threshold at that monitoring point, i.e., the carbon monoxide anomaly rise judgment line for the current sampling period, and that the oxygen concentration for several consecutive sampling periods be lower than the baseline minus the oxygen decrease threshold or lower than the hypoxia warning line. When all the above conditions are met, the risk nature can be confirmed, i.e., the methane anomaly is also accompanied by combustion risk or incomplete combustion risk. In this way, the system outputs the anomaly initiation point, diffusion direction, diffusion range, and risk nature.

[0040] The predictive diagnostic module is used to perform short-term prediction and sensor self-diagnosis based on the time-aligned laser methane sensing sequence, such as... Figure 3The diagram shown illustrates the lightweight neural network construction and sensor self-diagnosis block diagram provided in this embodiment of the invention. Considering downhole deployment conditions, this embodiment preferably uses a lightweight neural network, such as a two-layer GRU network, as a short-term prediction model, which facilitates model training and deployment at the edge. The two-layer GRU network includes a single-point temporal coding layer, a path temporal fusion layer, a prediction output layer, and a reconstruction output layer. The single-point temporal coding layer encodes the historical sequence of a single key monitoring point within the current prediction window. The path temporal fusion layer fuses the coding results of the current key monitoring point and its upstream and downstream adjacent monitoring points according to the airflow path sequence. The prediction output layer outputs the methane prediction value for the future prediction period, and the reconstruction output layer outputs the methane reconstruction value for the current sampling period. The input consists of two parts: first, a time-aligned laser methane sensing sequence, reflecting the actual propagation changes along the airflow path at each key monitoring point; second, a multi-parameter auxiliary input sequence extracted from the multi-parameter synchronization sequence, corresponding to the current prediction window, including temperature, humidity, carbon monoxide, oxygen, carbon dioxide, hydrogen sulfide, ethylene, local airflow velocity, and air pressure. During model training, the system extracts the time-aligned laser methane sensing sequence, multi-parameter synchronization sequence, and airflow path sequence from the historical database, and removes samples with sensor failure markers, maintenance status, continuous missing durations exceeding the allowable missing time, and signal disconnections. For each key monitoring point, the input window is L consecutive sampling periods from before the current sampling time to the current sampling time. The data in the input window includes the time-aligned laser methane sensing value of the key monitoring point, as well as the temperature, humidity, carbon monoxide, oxygen, carbon dioxide, hydrogen sulfide, ethylene, local airflow velocity, and air pressure values ​​corresponding to the same sampling time. For samples involving path fusion, the corresponding data of the upstream and downstream adjacent monitoring points of the key monitoring point within the same input window are simultaneously extracted. Training labels include prediction labels and reconstruction labels. The prediction label is the actual laser-induced methane sensing sequence of the key monitoring point after the current sampling time and within the future prediction duration; the reconstruction label is the laser-induced methane sensing value marked as valid for the key monitoring point within the current sampling period. During training, the system constructs the training error using the deviation between the predicted output and the predicted label, and the deviation between the reconstructed output and the reconstructed label, and iteratively updates the parameters of the two-layer GRU network using historical samples. After training, the model is deployed to the edge or ground center station for online prediction and sensor self-diagnosis.

[0041] The model first encodes the methane time series of a single key monitoring point within the most recent historical window. Specifically, for each key monitoring point, the laser methane sensing values ​​recorded at the most recent L sampling times are extracted, and the multi-parameter auxiliary inputs corresponding to the same sampling time are concatenated into the input vector for that sampling time. This vector is then input into the first-layer GRU time series encoding unit in chronological order, and the hidden state corresponding to the last sampling time is read as the single-point time series encoding result for that monitoring point, i.e., the numerical representation of the single-point time change characteristics of that monitoring point within the current prediction window. Subsequently, the time series encoding results of the current monitoring point and its upstream and downstream adjacent monitoring points are combined in the order of the airflow path, so that the prediction model simultaneously receives the single-point time series encoding result of the current monitoring point and the path time series encoding results of adjacent monitoring points, thereby outputting the methane prediction values ​​of each key monitoring point within the future prediction period. The future prediction period is a preset short-term prediction interval, preferably set to 30 seconds to 3 minutes; when the sampling period is 10 seconds, it can be preferably set to 60 seconds. The warning threshold used in this module is consistent with the methane warning threshold in the source determination module. When the model predicts that a key monitoring point will reach the methane warning threshold within the predicted time period, but the current measured value of the point has not yet reached the corresponding threshold, the system generates an advance warning result, writes it into the warning queue in the scheduling background, and increases the priority of the corresponding inspection task. The scheduling duty personnel notify the inspection personnel in advance to focus on checking the abnormal starting point and its downstream affected sections based on the warning result.

[0042] In addition to short-term prediction, the system also utilizes the same lightweight neural network to reconstruct the laser methane sensing value for the current sampling period. Reconstruction refers to the model providing a reference value (reconstructed value) that the point should present according to normal propagation logic in the current sampling period, based on recent historical sequences, time-aligned upstream and downstream neighboring points, and multi-parameter coupling characteristics. Then, the absolute value of the difference between the measured laser methane sensing value and the reconstructed value for the same period is recorded as the reconstruction residual. Considering that a certain error is allowed between model estimation and measured values, the system presets a residual threshold; only when the reconstruction residual exceeds this threshold is the point considered to have a significant deviation. The residual threshold is preferably calibrated based on the statistical results of reconstruction residuals under normal operating conditions. Specifically, the system selects historical samples that have not experienced methane alarms, sensor failure markers, or are not under maintenance, statistically analyzes the distribution of reconstruction residuals, and takes the statistical value closest to the upper bound of this distribution as the residual threshold. When historical samples are insufficient, the basic residual threshold set during system deployment is used first, and the residual threshold is updated after the number of normal samples reaches the preset number.

[0043] For monitoring points whose residuals do not exceed the residual threshold, the system considers them as normal usable points and continues to participate in the propagation delay estimation and source localization of the next cycle. For monitoring points whose reconstructed residuals exceed the residual threshold but are within the maximum propagation delay range, and whose upstream or downstream monitoring points have corresponding abnormal events, and whose corresponding abnormal events and current monitoring point anomalies at least simultaneously meet at least two of the following upstream and downstream propagation responses: consistent direction of change, rise time deviation not exceeding 2 sampling periods, and peak time deviation not exceeding 2 sampling periods, the system considers them as real anomalies and continues to retain them for anomaly analysis. For monitoring points whose reconstructed residuals exceed the residual threshold and do not meet any of the above conditions within the maximum propagation delay range, but this situation only occurs for no more than 2 consecutive sampling periods, the system marks them as suspected anomalies and continues to observe them. Only when the reconstructed residuals of a monitoring point exceed the residual threshold for 3 consecutive sampling periods, and no propagation response consistent with the upstream or downstream point occurs within the maximum propagation delay range, is the laser sensor corresponding to that monitoring point determined as a failed point to be reviewed, and temporarily removed from the propagation delay estimation and source localization of the next cycle. After the above processing, the system obtains the set of available monitoring points for the next cycle and uses it as one of the data bases for subsequent cyclic calculations.

[0044] The early warning and scheduling module is used to execute tiered early warnings, issue inspection reports, and write back results. The system receives the anomaly initiation point, diffusion range, and risk confirmation results from the source determination module, as well as the methane prediction results, early warning results, and laser sensor failure markers from the prediction and diagnosis module. It then processes these results in conjunction with pre-set tiered thresholds for each parameter. These thresholds include at least a methane early warning threshold and a methane alarm threshold, and may also include carbon monoxide early warning thresholds, carbon monoxide alarm thresholds, low oxygen early warning thresholds, and low oxygen alarm thresholds. To ensure the system's response intensity matches the risk intensity, a tiered triggering mechanism is used. When the methane prediction result indicates that the methane early warning threshold will be reached within the predicted timeframe, but the current measured value has not yet exceeded the limit, the system triggers an early warning level response, generates an early warning level inspection task, and generates an inspection route according to the order of the anomaly initiation point, monitoring points within the diffusion range, and monitoring points at the end of the diffusion range. This allows inspection personnel to prioritize checking the most likely anomaly source location and its downstream impact range. When the measured levels of methane, carbon monoxide, or oxygen at a monitoring point reach the corresponding alarm threshold for two consecutive sampling cycles, the system triggers an alarm response and limits the alarm area to the determined diffusion range. If the methane anomaly is accompanied by a continuous increase in carbon monoxide and a continuous decrease in oxygen, or if it reaches a higher hazard level threshold, the alarm can be further escalated to a high-risk level to initiate higher-priority inspection and handling procedures. For laser sensors determined to be faulty and requiring verification, the system simultaneously issues a sensor verification task to prevent the fault point from continuously interfering with subsequent location results.

[0045] Finally, the system writes the laser methane sensing sequence, abnormal event set, propagation delay estimation results, cumulative propagation delay, time-aligned laser methane sensing sequence, abnormal starting point, diffusion range, risk confirmation results, methane prediction results, sensor verification results, alarm records, and inspection results back to the database. The monitoring interface provides current value display, aligned event playback, and historical query functions for subsequent traceability and continuous optimization. Figure 4 The diagram shown illustrates a real-time visualization monitoring interface provided in this embodiment of the invention. The visualization results include the number of inspection points, the number of inspection points in progress, the number of completed inspection points, the completion rate, and the number of potential hazards. Additionally, it includes detailed on-duty inspection tables, latest inspection data tables, gas inspection tracking tables, ventilation inspection data tables, monitoring point change curves, and risk / hazard area classifications. The on-duty inspection table displays the current inspection area, inspection type, inspection shift, and inspector. The latest inspection data table displays the inspection area, inspection point, methane concentration, temperature, measurement time, and corresponding inspector, with abnormal data highlighted. The gas inspection tracking table displays the inspection area and hazard status, including whether a hazard exists or is normal, the inspection time, and the inspector. The ventilation inspection data table displays the inspection area, cross-sectional area, wind speed, air volume, and detection time. The monitoring point change curves allow selection of the inspection area and curve type, displaying corresponding visualization curves based on the area and type. The risk / hazard area classifications are presented as pie charts, showing the percentage of areas where hazards occur, providing decision support for back-end control personnel.

[0046] Through the implementation of the above modules, this system uses propagation delay estimation and time alignment under airflow path constraints as its core processing principle. This allows the sequential response relationship of multiple measuring points to the same gas anomaly event to be restored to a unified reference time. Based on this unified reference time, the system then completes the identification of the anomaly initiation point, determination of the diffusion range, and hierarchical scheduling. Furthermore, the system utilizes a lightweight neural network to perform spatiotemporal sequence analysis on the correlation records of methane, local airflow velocity, gas pressure, and auxiliary parameters such as carbon monoxide and oxygen in the time and path dimensions. This enables short-term advance prediction of gas anomalies and self-diagnosis of sensor failures. Thus, the system expands from a passive sensing method that relies solely on current measured values ​​for alarms to an active predictive processing method with early warning, anomaly screening, and self-checking capabilities. This reduces the probability of downstream high-value accumulation being misjudged as an anomaly source, improves the accuracy of anomaly source identification, diffusion range delineation, and inspection route generation, and reduces the interference of sensor anomalies on the overall judgment results.

[0047] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the above functions can be divided into different functional modules to complete all or part of the functions described above.

[0048] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0049] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units, located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0050] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0051] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes, or all or part of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0052] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A coal mine underground gas intelligent inspection system based on laser sensing, characterized in that, include: The laser sensing and acquisition module is used to generate a laser methane sensing sequence, a multi-parameter synchronization sequence, and an airflow path sequence based on the sequence of monitoring points along the airflow direction. Anomaly extraction module is used to determine methane anomaly events at each monitoring point based on the laser methane sensing sequence; The time-delay alignment module is used to estimate the propagation time delay based on the methane anomaly event and the airflow path sequence, and to perform time alignment on the laser methane sensing sequence based on the propagation time delay. The source determination module is used to determine the anomaly initiation point and the diffusion range along the airflow path based on the time-aligned laser methane sensing sequence. The prediction and diagnosis module is used to determine the methane prediction result and the laser sensor status result based on the time-aligned laser methane sensing sequence and the multi-parameter synchronization sequence. The early warning and scheduling module is used to generate inspection tasks, early warning results, and alarm results based on the abnormal starting point, the diffusion range, the methane prediction result, and the laser sensor status result.

2. The intelligent coal mine gas inspection system based on laser sensing according to claim 1, characterized in that: The generation of the laser methane sensing sequence, multi-parameter synchronization sequence, and airflow path sequence includes: A laser gas sensor emits detection light into the characteristic absorption band of methane and receives the light signal. The received light signal is compared with a reference light signal, and the original methane concentration is determined based on the pre-established correspondence between the reference light signal and the methane concentration calibration. The pre-established compensation table is invoked, and the corresponding correction amount is determined based on the temperature value, humidity value and original methane concentration of the current sampling period. The original methane concentration is then corrected according to the corresponding correction amount to obtain the laser methane sensing value. Temperature, humidity, carbon monoxide, oxygen, carbon dioxide, hydrogen sulfide, ethylene, local airflow velocity, and air pressure are mapped to the same time axis and organized according to a pre-established airflow path sequence.

3. The intelligent coal mine gas inspection system based on laser sensing according to claim 2, characterized in that: The generation of the laser methane sensing sequence, multi-parameter synchronization sequence, and airflow path sequence also includes: Set a uniform sampling period, and form a uniform sampling time sequence according to the uniform sampling period; For parameters that have multiple sampled values ​​within the current sampling period, the average value within the current sampling period is used; For parameters that have only one sampled value in the current sampling period, use that sampled value; For parameters that are short-term missing and do not exceed the allowed missing duration, the previous valid value shall be used; Parameters that are continuously missing for more than the allowed duration are marked as invalid data; The allowed missing duration is determined based on the unified sampling period and the maximum number of consecutive missing sampling periods allowed by the corresponding parameters.

4. The intelligent coal mine gas inspection system based on laser sensing according to claim 1, characterized in that: The determination of methane anomaly events at each monitoring point includes: A historical baseline window is constructed for each monitoring point, and records corresponding to confirmed methane anomalies and records marked as invalid data are removed when constructing the historical baseline window; The current baseline methane value is determined based on the laser methane sensing value within the historical baseline window; When the laser methane sensing value at a certain monitoring point is higher than the sum of the current benchmark methane value and the preset methane rise threshold for several consecutive sampling cycles, it is determined that there is a methane anomaly event at the monitoring point, and the event start time, peak time and event window of the methane anomaly event are recorded.

5. The intelligent coal mine gas inspection system based on laser sensing according to claim 4, characterized in that: The determination of methane anomalies at each monitoring point also includes: After the methane anomaly event is determined, carbon monoxide, oxygen, carbon dioxide, hydrogen sulfide, ethylene, local airflow velocity and air pressure values ​​are extracted from the event window corresponding to the methane anomaly event to form a multi-parameter event window; The monitoring point identifier, the corresponding airflow path identifier, the event start time, the peak time, the event end time, the event window, the event peak value, and the multi-parameter event window are combined to form a set of methane abnormal events.

6. The intelligent coal mine gas inspection system based on laser sensing according to claim 1, characterized in that: The estimation of propagation time delay based on the methane anomaly event and the airflow path sequence includes: For each pair of adjacent monitoring points on the same path, an incremental sequence is generated based on its laser methane sensing sequence; Within the preset maximum propagation time delay range, candidate time delays are generated according to a uniform sampling period. For each candidate time delay, the incremental sequence of the downstream monitoring point is pushed back according to the candidate time delay, and the propagation matching judgment is performed with the incremental sequence of the upstream monitoring point in the same comparison window. The propagation matching determination includes at least the change direction consistency rate, the rise time deviation, the peak time deviation, and the effective change segment overlap rate. When a candidate time delay meets the preset propagation matching conditions, the candidate time delay is determined as a valid candidate time delay, and the propagation time delay between the corresponding adjacent monitoring points is determined according to the preset rules among all valid candidate time delays.

7. The intelligent coal mine gas inspection system based on laser sensing according to claim 6, characterized in that: The step of time-aligning the laser methane sensing sequence based on propagation delay includes: When no valid candidate time delay is found within the maximum propagation time delay range, the difference between the event start time of the methane anomaly event corresponding to the downstream monitoring point and the event start time of the methane anomaly event corresponding to the upstream monitoring point is taken as the supplementary propagation time delay. When the supplementary propagation delay is negative or exceeds the maximum propagation delay range, the corresponding adjacent monitoring points will be marked as point pairs that do not participate in source localization. For point pairs that can participate in source localization, the propagation time delay is accumulated segment by segment according to the airflow direction to obtain the cumulative propagation time delay of each monitoring point relative to the path reference point. The laser methane sensing sequence of each monitoring point is then reversed according to its own cumulative propagation time delay to obtain the time-aligned laser methane sensing sequence. When the cumulative propagation delay is not an integer multiple of the unified sampling period, the corresponding value after the callback is determined by linear interpolation between adjacent sampling records.

8. The intelligent coal mine gas inspection system based on laser sensing according to claim 1, characterized in that: The determination of the anomaly's starting point and its diffusion range along the airflow path includes: Based on the availability of point pairs output by the time-delay alignment module, continuous effective monitoring points on the same airflow path are determined; When the number of consecutive valid monitoring points on the same path is less than the preset number, output the single-point anomaly pending verification result. When the number of consecutive effective monitoring points on the same path is not less than the preset number, within the same event window after time alignment, the methane status of each effective monitoring point is checked sequentially from upstream to downstream along the airflow direction. The monitoring point that first reaches the methane warning threshold or methane alarm threshold and whose upstream adjacent effective monitoring point does not reach the same level threshold at the same reference time is determined as the abnormal starting point. The continuous effective monitoring points downstream of the abnormal starting point that reach the same level threshold within the same event window are sequentially included in the diffusion range until the first position that does not reach the same level threshold or the point pair is interrupted. When there are no continuous effective monitoring points downstream of the abnormal starting point in the diffusion range, the abnormal event is marked as a single-point abnormal event that has not formed an effective diffusion direction. The source determination module also determines the nature of the risk based on the carbon monoxide and oxygen states within the same event window.

9. The intelligent coal mine gas inspection system based on laser sensing according to claim 1, characterized in that: The determination of methane prediction results includes: A lightweight neural network is used to output methane prediction values, wherein the input of the lightweight neural network includes a time-aligned laser methane sensing sequence and a multi-parameter auxiliary input sequence corresponding to the current prediction window; The prediction and diagnosis module first performs time-series encoding on the laser methane sensing value of a single key monitoring point within the historical window, and then combines the time-series encoding results of the current key monitoring point and its upstream and downstream adjacent monitoring points according to the airflow path order, and outputs the methane prediction value of each key monitoring point within the future prediction time. After determining the methane prediction results, we proceed to determine the status results of the laser sensor.

10. The intelligent coal mine gas inspection system based on laser sensing according to claim 9, characterized in that: Determining the state result of the laser sensor includes: The lightweight neural network is used to reconstruct the laser methane sensing value of the current sampling period, and the absolute value of the difference between the measured laser methane sensing value of the current sampling period and the reconstructed value of the same period is determined as the reconstruction residual. When the reconstructed residual does not exceed the residual threshold, the corresponding monitoring point is determined as a normal and usable point; When the reconstructed residual exceeds the residual threshold and there is an upstream and downstream propagation response corresponding to the current monitoring point anomaly within the maximum propagation time delay range, the corresponding monitoring point is determined as a real anomaly point. When the reconstructed residual exceeds the residual threshold and there is no upstream or downstream propagation response corresponding to the current monitoring point anomaly within the maximum propagation time delay range, but the continuous occurrence of this state does not exceed the preset number of sampling cycles, the corresponding monitoring point will be identified as a suspected anomaly point. When the number of consecutive preset sampling periods of the reconstructed residual exceeds the residual threshold and no propagation response consistent with the upstream or downstream monitoring point appears within the maximum propagation time delay range, the corresponding monitoring point is determined as a failed point to be reviewed and removed from the propagation time delay estimation and source location of the next period. The system receives the anomaly initiation point, diffusion range, methane prediction results, and failure points to be verified. It then combines these with preset grading thresholds to generate early warning-level inspection tasks, alarm results, and sensor verification tasks. Finally, it writes back the laser methane sensing sequence, methane anomaly event set, propagation delay estimation results, time-aligned laser methane sensing sequence, anomaly initiation point, diffusion range, methane prediction results, alarm records, and inspection results to the database.

Images