Gas detection alarm method and system for mining safety equipment

Abstract

The invention discloses a gas detection alarm method and system for mining safety equipment, and relates to the technical field related to gas detection, and the method comprises the steps: obtaining basic gas detection related equipment information of a target mine, and constructing a mine topological graph; performing incremental coding identification on the node gas correlation detection data sequence of the M equipment topology nodes in the preset detection window; performing multi-dimensional structured interaction on the M signal token sequences to obtain a local association sub-graph set; and in combination with the M signal token sequences, carrying out local gas detection anomaly identification on the local association sub-graph set, and obtaining gas detection alarm information according to an identification result set. The technical problems that in the prior art, the semantic relation between signals is difficult to accurately understand in the complex and heterogeneous mine environment, local gas abnormity is difficult to recognize in time, and consequently the accuracy and timeliness of gas detection alarming are poor are solved, and the technical effects that abnormity accurate positioning is achieved, and the accuracy and timeliness of gas detection alarming are improved are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gas detection, and particularly relates to a gas detection alarm method and system of a mine safety device. BACKGROUND

[0002] In the underground mine environment such as coal mines and metal mines, there are various harmful gases such as methane, carbon monoxide, and hydrogen sulfide. These gases may cause serious accidents such as explosion and poisoning under a specific concentration, threatening the safety of miners and mine facilities. However, the mine environment is complex and changeable, and the gas distribution is affected by various factors such as ventilation system, mining activities, and geological structure, showing high spatial heterogeneity and dynamics. The traditional gas detection mainly relies on single sensor equipment for fixed-point monitoring, and triggers an alarm by setting a threshold, which is difficult to fully reflect the gas state of the entire mine and is prone to false alarms, missed alarms, or failure to identify local gas abnormal accumulation in time. By deploying various gas sensors, anemometers, temperature sensors, and other equipment, a mine gas detection network can be constructed to collect multi-source and real-time environmental data, but how to extract effective information from massive and heterogeneous data and accurately identify gas abnormalities becomes a key issue.

[0003] Therefore, in the related art, there is a technical problem that it is difficult to accurately understand the semantic relationship between signals in a complex and heterogeneous mine environment and timely identify local gas abnormalities, resulting in poor accuracy and timeliness of gas detection alarm. SUMMARY

[0004] The present application provides a gas detection alarm method and system of a mine safety device, which solves the technical problem that it is difficult to accurately understand the semantic relationship between signals in a complex and heterogeneous mine environment and timely identify local gas abnormalities in the prior art, resulting in poor accuracy and timeliness of gas detection alarm, and achieves the technical effects of accurate positioning of abnormalities and improved accuracy and timeliness of gas detection alarm.

[0005] The present application provides a gas detection alarm method of a mine safety device, which comprises: acquiring basic gas detection associated equipment information of a target mine, and constructing a mine topology graph, wherein the mine topology graph comprises M device topology nodes and Q heterogeneous edges, and M and Q are positive integers; incrementally encoding and identifying M node gas associated detection data sequences of the M device topology nodes within a preset detection window to obtain M signal token sequences; performing multi-dimensional structured interaction on the M signal token sequences in combination with the mine topology graph to obtain a local associated subgraph set; performing local gas detection abnormality identification on the local associated subgraph set in combination with the M signal token sequences, and obtaining gas detection alarm information according to a recognition result set.

[0006] In a possible implementation, the gas detection alarm method of the mine safety equipment further performs the following processing: extracting the device type and device layout position in the basic gas detection associated device information, obtaining M device topology nodes with position identification; constructing an associated edge between two associated device topology nodes in the M device topology nodes based on the ventilation connection relationship, the electrical feeder connection relationship, the physical adjacency relationship and the operation association relationship in the basic gas detection associated device information, and type-identifying the associated edge according to the relationship type, to obtain Q heterogeneous edges; and constructing the mine topology graph based on the Q heterogeneous edges and the M device topology nodes with position identification.

[0007] In a possible implementation, the gas detection alarm method of the mine safety equipment further performs the following processing: performing normalization processing on the M node gas associated detection data sequences respectively, and performing adjacency difference operation on the normalized sequences to determine data increments, to obtain M node gas associated monitoring data increment sequences; and performing discrete symbol coding on the M node gas associated detection data sequences according to a preset coding rule based on the direction and amplitude of the data increments in the M node gas associated monitoring data increment sequences, to obtain the M signal token sequences.

[0008] In a possible implementation, the gas detection alarm method of the mine safety equipment further performs the following processing: the preset coding rule is that when the direction of the data increment is downward and the amplitude is less than a preset first amplitude, first type discrete code coding is performed; when the direction of the data increment is downward and the amplitude is greater than or equal to the preset first amplitude and less than a preset second amplitude, second type discrete code coding is performed; when the amplitude of the data increment is 0, third type discrete code coding is performed; when the direction of the data increment is upward and the amplitude is greater than a preset third amplitude and less than or equal to a preset fourth amplitude, fourth type discrete code coding is performed; and when the direction of the data increment is upward and the amplitude is greater than the preset fourth amplitude, fifth type discrete code coding is performed.

[0009] In a possible implementation, the gas detection alarm method of the mine safety device further performs the following processing: single-point anomaly identification is performed on the M signal token sequences to obtain a set of abnormal signal token sequences; a first abnormal signal token sequence is extracted from the set of abnormal signal token sequences, and a first device topology node corresponding to the first abnormal signal token sequence is combined with a position identifier to perform one-level multi-dimensional structured interaction in the mine topology graph to obtain a set of one-level associated device topology nodes; two-level multi-dimensional structured interaction is performed on the set of one-level associated device topology nodes in the mine topology graph to obtain a set of two-level associated device topology nodes; the first device topology node, the set of one-level associated device topology nodes, and the set of two-level associated device topology nodes are combined to construct a first local associated subgraph, and the first local associated subgraph is added to a set of local associated subgraphs.

[0010] In a possible implementation, the gas detection alarm method of the mine safety device further performs the following processing: a single-point anomaly identifier is pre-constructed, wherein the single-point anomaly identifier includes a single-point anomaly threshold identification branch and a single-point anomaly trend identification branch; threshold anomaly identification is performed on the M signal token sequences based on the single-point anomaly threshold identification branch to obtain a set of threshold abnormal signal token sequences; trend anomaly identification is performed on the M signal token sequences based on the single-point anomaly trend identification branch to obtain a set of trend abnormal signal token sequences; a union set of the set of threshold abnormal signal token sequences and the set of trend abnormal signal token sequences is obtained to obtain the set of abnormal signal token sequences.

[0011] In a possible implementation, the gas detection alarm method of the mine safety device further performs the following processing: a first local associated subgraph is extracted from the set of local associated subgraphs; a set of related first local signal token sequences in the M signal token sequences is extracted based on the first local associated subgraph; scene-based local gas detection anomaly identification is performed based on the set of first local signal token sequences to obtain a first abnormal identification result, and the first abnormal identification result is added to a set of identification results.

[0012] In a possible implementation, the gas detection alarm method of the mine safety device further performs the following processing: a set of historical gas detection anomaly interaction logs is pre-constructed; scene-based semantic clustering analysis is performed on the set of historical gas detection anomaly interaction logs to construct a scene-based semantic template library; the set of first local signal token sequences is matched with the scene-based semantic template library to obtain the first abnormal identification result.

[0013] In a possible implementation, the gas detection alarm method of the mine safety device further performs the following processing: performing semantic clustering on the set of historical gas detection abnormal interaction logs according to a preset clustering granularity, to obtain a plurality of sets of clustered historical gas detection abnormal interaction logs; and traversing the plurality of sets of clustered historical gas detection abnormal interaction logs to perform scene-based template extraction, to obtain the scene-based semantic template library.

[0014] The application further provides a gas detection alarm system of a mine safety device, which comprises: a topology graph construction module, configured to acquire basic gas detection associated device information of a target mine, and construct a mine topology graph, wherein the mine topology graph comprises M device topology nodes and Q heterogeneous edges, and M and Q are positive integers; an incremental encoding identification module, configured to respectively perform incremental encoding identification on M node gas associated detection data sequences of the M device topology nodes within a preset detection window, to obtain M signal token sequences; a structured interaction module, configured to perform multi-dimensional structured interaction on the M signal token sequences in combination with the mine topology graph, to obtain a set of local associated subgraphs; and a detection abnormality identification module, configured to respectively perform local gas detection abnormality identification on the set of local associated subgraphs in combination with the M signal token sequences, and acquire gas detection alarm information according to a set of identification results.

[0015] The gas detection alarm method and system of the mine safety device provided in the application can acquire basic gas detection associated device information of a target mine to construct a mine topology graph, perform incremental encoding identification on node gas associated detection data sequences of M device topology nodes within a preset detection window, perform multi-dimensional structured interaction on M signal token sequences to obtain a set of local associated subgraphs, respectively perform local gas detection abnormality identification on the set of local associated subgraphs in combination with the M signal token sequences, and acquire gas detection alarm information according to a set of identification results. The technical problems that it is difficult to accurately understand the semantic relationship between signals in a complex heterogeneous mine environment and timely identify local gas abnormalities in the prior art, resulting in poor accuracy and timeliness of gas detection alarm, are solved, and the technical effects of realizing accurate abnormal positioning and improving the accuracy and timeliness of gas detection alarm are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments of the present disclosure will be briefly introduced below. In the present application, a flowchart is used to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the foregoing or the following operations are not necessarily performed in sequence. On the contrary, according to the needs, various steps can be processed in reverse order or simultaneously. Meanwhile, other operations can be added to these processes, or a step or several steps can be removed from these processes.

[0017] Figure 1A gas detection alarm method of a mine safety device provided by the embodiment of the application is shown in a flowchart.

[0018] Figure 2 A structure diagram of a gas detection alarm system of a mine safety device provided by the embodiment of the application is shown.

[0019] The reference signs are explained as follows: a topology graph construction module 10, an incremental coding identification module 20, a structured interaction module 30, and an abnormality detection identification module 40. DETAILED DESCRIPTION

[0020] The above description is only a summary of the technical scheme of the application. In order to make the technical means of the application more clearly understood, the application can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the application more obvious and easy to understand, the following specific embodiments of the application are described.

[0021] In order to make the purposes, technical schemes and advantages of the application more clear, the following will further describe the application in combination with the drawings, and the described embodiments should not be regarded as limiting the application. All other embodiments obtained by those skilled in the art without making creative efforts belong to the protection scope of the application.

[0022] In the following description, "some embodiments" are related to a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The term "first\second" is only to distinguish similar objects, and does not represent a specific order of the objects. The terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art in the technical field of the application. The terms used herein are only for the purpose of describing the embodiments of the application.

[0023] The embodiment of the application provides a gas detection alarm method of a mine safety device, as shown in the figure, the method comprises the steps of: Figure 1 S100, acquiring basic gas detection associated device information of a target mine, and constructing a mine topology graph, wherein the mine topology graph comprises M device topology nodes and Q heterogeneous edges, and M and Q are positive integers.

[0024] ​The step S100 further comprises a step S110 of extracting device types and device layout positions in the basic gas detection associated device information to obtain M device topology nodes with position identifiers; a step S120 of constructing associated edges between two associated device topology nodes in the M device topology nodes based on ventilation connection relationships, electrical feeder connection relationships, physical adjacency relationships and operation associated relationships in the basic gas detection associated device information, and type-identifying the associated edges according to relationship types to obtain Q heterogeneous edges; and a step S130 of constructing the mine topology graph based on the Q heterogeneous edges and the M device topology nodes with position identifiers.

[0025] Preferably, all gas detection related device lists and attribute data thereof are collected from a monitoring system, a device management system or a database of a target mine to obtain basic gas detection associated device information, which can include unique IDs, device names and other device identifiers, device types such as methane sensors, carbon monoxide sensors, wind speed sensors, air door opening and closing sensors, device start and stop sensors and device layout positions such as specific three-dimensional coordinates or installation position descriptions in a mine roadway map; each physical device is abstracted as a data node, and each node contains key attributes of device types and accurate device layout positions, thereby obtaining M device topology nodes with position identifiers, wherein M is a positive integer representing the number of devices with position identifiers.

[0026] Preferably, associated edges between two associated device topology nodes in the M device topology nodes are constructed based on ventilation connection relationships, electrical feeder connection relationships, physical adjacency relationships and operation associated relationships in the basic gas detection associated device information, wherein the ventilation connection relationship refers to whether the roadways where the two nodes are located are connected by main air flow or branch air flow, the electrical feeder connection relationship refers to whether the devices represented by the two nodes are powered by the same power line or are in the same power supply loop, the physical adjacency relationship indicates whether the two nodes are directly adjacent or very close in space and geography, and the operation associated relationship refers to whether the devices represented by the two nodes belong to the same mining operation or process flow; the associated edges are then type-identified according to relationship types, i.e., relationship type labels are identified for the associated edges, to obtain Q heterogeneous edges, wherein Q is a positive integer representing the number of associated relationships. Finally, the Q heterogeneous edges and the M device topology nodes with position identifiers are combined to determine the mine topology graph, which contains device spatial distribution and multiple different types of functional logical relationships between devices.

[0027] The step S200 incrementally encodes and identifies M node gas associated detection data sequences of the M device topology nodes in a preset detection window respectively to obtain M signal token sequences.

[0028] The step S200 further comprises a step S210 of normalizing the M node gas correlation detection data sequences respectively, and performing adjacent difference operation on the normalized sequences to determine data increments and obtain M node gas correlation monitoring data increment sequences; and a step S220 of performing discrete symbol coding on the M node gas correlation detection data sequences according to a preset coding rule based on the direction and amplitude of the data increments in the M node gas correlation monitoring data increment sequences, and obtaining the M signal token sequences.

[0029] Preferably, for any one of the M device topology nodes, a plurality of readings are collected in time sequence in a preset detection window to form a node gas correlation detection data sequence, and then M node gas correlation detection data sequences are obtained, and the M node gas correlation detection data sequences are incrementally coded and recognized. Specifically, the M node gas correlation detection data sequences are normalized, i.e., each concentration value in the sequence is mathematically transformed to eliminate the numerical range difference caused by different sensors and different dimensions, so that it is converted to a unified scale, such as the interval [0, 1], to ensure that the data of all devices are processed on the same order of magnitude, for example, 1% concentration of methane and 24 ppm of carbon monoxide are both mapped to a normalized value of approximately 0.5, to obtain M normalized sequences; then adjacent difference operation is performed on the normalized sequences, i.e., the difference between the data of two adjacent time points is calculated as a data increment, to obtain M node gas correlation monitoring data increment sequences representing the change amount and change direction of the gas concentration in each sampling interval, a positive increment indicating an increase in concentration, a negative increment indicating a decrease in concentration, and a zero increment indicating no change in concentration. The direction and amplitude of the data increments in the M node gas correlation monitoring data increment sequences are recognized, and the M node gas correlation detection data sequences are discretely coded according to a preset coding rule, each increment value is converted into a symbol representing a certain change pattern, a signal token is obtained, and finally M signal token sequences are obtained to describe the dynamic change pattern of the gas concentration in the detection window of the device, thereby reducing the complexity of the data and highlighting the key change features.

[0030] Further, the step S220 further comprises that the preset coding rule is to perform first type discrete code coding when the direction of the data increment is downward and the amplitude is less than a preset first amplitude; perform second type discrete code coding when the direction of the data increment is downward and the amplitude is greater than or equal to the preset first amplitude and less than a preset second amplitude; perform third type discrete code coding when the amplitude of the data increment is 0; perform fourth type discrete code coding when the direction of the data increment is upward and the amplitude is greater than a preset third amplitude and less than or equal to a preset fourth amplitude; and perform fifth type discrete code coding when the direction of the data increment is upward and the amplitude is greater than the preset fourth amplitude.

[0031] Preferably, the preset encoding rule refers to quantizing the continuous change sequence into discrete symbol encoding according to the direction and amplitude of the increment value. Specifically, when the direction of the data increment is downward and the amplitude is less than a preset first amplitude, a first type of discrete code encoding is performed, and it is encoded as "↓", indicating rapid downward; when the direction of the data increment is downward and the amplitude is greater than or equal to the preset first amplitude and less than a preset second amplitude, a second type of discrete code encoding is performed, and it is encoded as "-", indicating slow downward; when the amplitude of the data increment is 0, a third type of discrete code encoding is performed, and it is encoded as "0", indicating stability; when the direction of the data increment is upward and the amplitude is greater than a preset third amplitude and less than or equal to a preset fourth amplitude, a fourth type of discrete code encoding is performed, and it is encoded as "+", indicating slow upward; when the direction of the data increment is upward and the amplitude is greater than the preset fourth amplitude, a fifth type of discrete code encoding is performed, and it is encoded as "↑", indicating rapid upward; wherein the preset first amplitude, the preset second amplitude, the preset third amplitude and the preset fourth amplitude are dynamic thresholds adaptively determined according to sensor noise characteristics and safety standards.

[0032] Step S300, combining the mine topological graph, the M signal token sequences are subjected to multi-dimensional structured interaction to obtain a local correlation subgraph set.

[0033] Step S300 further comprises the following steps: step S310, single-point anomaly identification is performed on the M signal token sequences to obtain an abnormal signal token sequence set; step S320, a first abnormal signal token sequence is extracted from the abnormal signal token sequence set, and a position identifier of a first device topological node corresponding to the first abnormal signal token sequence is combined to perform one-level multi-dimensional structured interaction in the mine topological graph to obtain a one-level associated device topological node set; step S330, two-level multi-dimensional structured interaction is performed on the one-level associated device topological node set in the mine topological graph to obtain a two-level associated device topological node set; step S340, the first device topological node, the one-level associated device topological node set and the two-level associated device topological node set are combined to construct a first local correlation subgraph, and the first local correlation subgraph is added to the local correlation subgraph set.

[0034] Preferably, the M signal token sequences are subjected to multi-dimensional structured interaction in combination with the mine topology map, specifically, single-point anomaly identification analysis is performed on each signal token sequence to determine whether a token representing a dangerous pattern such as rapid rise appears in the sequence, and a set of abnormal signal token sequences, i.e., a list of all devices preliminarily determined to be abnormal, is obtained; then a first abnormal signal token sequence is randomly extracted from the set of abnormal signal token sequences, the corresponding first device topology node and its location identifier are obtained, and then the device node is taken as the center to perform one-level multi-dimensional structured interaction in the mine topology map, i.e., all device nodes directly connected are identified and determined along all types of heterogeneous edges, for example, nodes directly connected are determined through ventilation connection relationship, electrical connection relationship, physical adjacency relationship and work association relationship, and then combined to form a one-level associated device topology node set.

[0035] Preferably, the one-level associated device topology node set is subjected to two-level multi-dimensional structured interaction in the mine topology map, i.e., the same identification process is performed on each node in the one-level associated device topology node set, nodes directly connected are determined through ventilation connection relationship, electrical connection relationship, physical adjacency relationship and work association relationship, and the identified device nodes are avoided from being repeatedly included, and then a two-level associated device topology node set is obtained. Then the first device topology node, the one-level associated device topology node set and the two-level associated device topology node set are merged, i.e., all these nodes and all edges connecting these nodes are extracted from the mine topology map to generate a first local associated subgraph, which is used to describe the potential abnormal influence propagation range covering the directly and indirectly associated devices centered on the initial abnormal device. Finally, all first local associated subgraphs are combined to form a local associated subgraph set.

[0036] Further, step S310 further includes step S311 of pre-constructing a single-point anomaly identifier, wherein the single-point anomaly identifier includes a single-point anomaly threshold identification branch and a single-point anomaly trend identification branch; step S312 of performing threshold anomaly discrimination on the M signal token sequences based on the single-point anomaly threshold identification branch to obtain a threshold abnormal signal token sequence set; step S313 of performing trend anomaly discrimination on the M signal token sequences based on the single-point anomaly trend identification branch to obtain a trend abnormal signal token sequence set; and step S314 of performing set union on the threshold abnormal signal token sequence set and the trend abnormal signal token sequence set to obtain the abnormal signal token sequence set.

[0037] Preferably, a single-point anomaly threshold identification branch and a single-point anomaly trend identification branch are constructed to generate a single-point anomaly identifier in combination for screening single-point abnormal devices. Specifically, the single-point anomaly threshold identification branch scans and identifies whether specific discrete symbols representing rapid changes occur in each device signal token sequence, including traversing M signal token sequences, identifying all "↓" and "↑" tokens, and determining the corresponding signal token sequence as abnormal, and then outputting all device sequences containing rapid change tokens as a threshold abnormal signal token sequence set. The single-point anomaly trend identification branch is used to analyze the persistent change pattern exhibited by the entire sequence, i.e., to identify and determine patterns that can indicate a worsening situation, such as the continuous occurrence of multiple tokens representing an increase, and to determine whether the state meets the preset abnormal trend pattern, and then output all sequence patterns exhibiting persistent adverse trends as a trend abnormal signal token sequence set. Finally, the threshold abnormal signal token sequence set and the trend abnormal signal token sequence set are combined to obtain the union set. As long as a device sequence is determined to be abnormal by either branch, it is marked as abnormal, and the final abnormal signal token sequence set is obtained.

[0038] Step S400, combining the M signal token sequences, respectively performing local correlation subgraph set local gas detection anomaly identification, and obtaining gas detection alarm information according to the identification result set.

[0039] Step S400 further includes step S410, extracting a first local correlation subgraph from the local correlation subgraph set; step S420, extracting a related first local signal token sequence set from the M signal token sequences based on the first local correlation subgraph; step S430, performing scenario-based local gas detection anomaly identification based on the first local signal token sequence set to obtain a first anomaly identification result, and adding the first anomaly identification result to the identification result set.

[0040] Preferably, the local correlation subgraph set is subjected to local gas detection anomaly identification in combination with M signal token sequences to determine whether the overall data mode of all related devices in the local correlation subgraph region constitutes a real abnormal scene. Specifically, a first local correlation subgraph is randomly extracted from the local correlation subgraph set, and analyzed to identify the initial abnormal device node and its associated device nodes constituting the subgraph. Then, all device node corresponding signal token sequences are identified from the M signal token sequences to form a first local signal token sequence set. The first local signal token sequence set is then subjected to scene-based local gas detection anomaly identification. According to the pre-constructed scene-based semantic template library, it is determined whether it matches the known abnormal scene mode. The pre-constructed scene-based semantic template library is used to store multiple device collaborative change modes in various historical abnormal events, thereby obtaining a first abnormal identification result and accompanying abnormal type, matching degree, and impact range. Similarly, the M signal token sequences are subjected to scene-based local gas detection anomaly identification to obtain a set of identification results. The identification result set is then decided, i.e., the safety state of the target mine is comprehensively evaluated, including sorting all identification results according to the abnormal type and severity level, and merging multiple adjacent local correlation subgraphs of the identified type abnormality into a more serious alarm region. Finally, gas detection alarm information is output, which may include alarm levels such as first alarm-danger, second alarm-warning, and normal; alarm types such as 305 working face area gas accumulation alarm; alarm core area, i.e., the most critical local correlation subgraph position triggering the alarm; key abnormal device list of the alarm, and preset disposal suggestions based on the abnormal type and position. Thus, accurate positioning of the abnormality is achieved, and the accuracy and timeliness of the gas detection alarm are ensured.

[0041] Further, step S430 further includes step S431 of pre-constructing a set of historical gas detection anomaly interaction logs; step S432 of traversing the set of historical gas detection anomaly interaction logs for scene-based semantic clustering analysis to construct a scene-based semantic template library; and step S433 of matching the first local signal token sequence set with the scene-based semantic template library to obtain a first abnormal identification result.

[0042] Step S320 further includes semantic clustering of the set of historical gas detection anomaly interaction logs according to a preset clustering granularity to obtain a plurality of clustered historical gas detection anomaly interaction log sets; and traversing the plurality of clustered historical gas detection anomaly interaction log sets for scene-based template extraction to obtain the scene-based semantic template library.

[0043] Preferably, historical data records of various types of abnormal events that have occurred in the mine in the past and have been confirmed are collected, a pre-constructed historical gas detection abnormal interaction log set is constructed, containing a device set, signal token sequences of the devices before and after the event, event labels such as abnormal types and results, for example, gas accumulation, early coal spontaneous combustion, ventilation obstruction or misreporting of blasting operation interference; then scene semantic clustering analysis is performed on the historical gas detection abnormal interaction log set, that is, using clustering algorithms such as K-means and DBSCAN, based on the overall pattern similarity of the multi-device signal token sequences in the log, the historical gas detection abnormal interaction log set is automatically grouped according to the preset clustering granularity, and a plurality of clustered historical gas detection abnormal interaction log sets are obtained, wherein the preset clustering granularity refers to the requirement for pattern similarity, for example, all logs presenting the pattern of "device sharp rise, device slow rise, device wind speed drop" are grouped into a group, and then scene template extraction is performed on the plurality of clustered historical gas detection abnormal interaction log sets, that is, the common features of all logs in each group are extracted to form representative scene semantic templates, representing typical abnormal interaction patterns verified by historical data, and a scene semantic template library is constructed. When judging the first local signal token sequence set, a pattern matching or similarity measurement algorithm is used to calculate the matching degree of each historical template pattern in the scene semantic template library, and the historical template abnormal pattern with high matching degree is taken as the first abnormal recognition result, for example, the best matching template is gas accumulation, and the first abnormal recognition result is to confirm the gas accumulation abnormality; if the matching degree with all templates is low, the first abnormal recognition result may be an unknown type of abnormality, which needs to be checked manually.

[0044] In the foregoing, with reference to Figure 1 The gas detection alarm method of the mine safety device according to the embodiment of the present application is described in detail. Next, with reference to Figure 2 The gas detection alarm system of the mine safety device according to the embodiment of the present application is described.

[0045] The gas detection alarm system of the mine safety device according to the embodiment of the present application solves the technical problem that it is difficult to accurately understand the semantic relationship between signals and timely recognize local gas abnormalities in a complex heterogeneous mine environment in the prior art, and improves the accuracy and timeliness of gas detection alarm. As shown in Figure 2 The gas detection alarm system of the mine safety device includes a topology graph construction module 10, an incremental encoding identification module 20, a structured interaction module 30 and a detection abnormality identification module 40.

[0046] The topology graph construction module 10 is configured to obtain basic gas detection associated equipment information of a target mine, and construct a mine topology graph, wherein the mine topology graph comprises M device topology nodes and Q heterogeneous edges, and M and Q are positive integers; the incremental encoding identification module 20 is configured to respectively perform incremental encoding identification on M node gas associated detection data sequences of the M device topology nodes in a preset detection window, and obtain M signal token sequences; the structured interaction module 30 is configured to perform multi-dimensional structured interaction on the M signal token sequences in combination with the mine topology graph, and obtain a local associated subgraph set; and the detection anomaly identification module 40 is configured to respectively perform local gas detection anomaly identification on the local associated subgraph set in combination with the M signal token sequences, and obtain gas detection alarm information according to a result set.

[0047] Next, the specific configuration of the topology graph construction module 10 will be described in detail. The topology graph construction module 10 further comprises: extracting device types and device layout positions in the basic gas detection associated equipment information, and obtaining M device topology nodes with position identifiers; constructing associated edges between two associated device topology nodes in the M device topology nodes based on ventilation connection relationships, electrical feeder connection relationships, physical adjacency relationships and operation associated relationships in the basic gas detection associated equipment information, and identifying the associated edges according to relationship types, and obtaining Q heterogeneous edges; and constructing the mine topology graph based on the Q heterogeneous edges and the M device topology nodes with position identifiers.

[0048] Next, the specific configuration of the incremental encoding identification module 20 will be described in detail. The incremental encoding identification module 20 further comprises: respectively performing normalization processing on the M node gas associated detection data sequences, and performing adjacency difference operation on the normalized sequences to determine data increments, and obtaining M node gas associated monitoring data increment sequences; and performing discrete symbol encoding on the M node gas associated detection data sequences according to a preset encoding rule based on directions and amplitudes of the data increments in the M node gas associated monitoring data increment sequences, and obtaining the M signal token sequences.

[0049] The specific configuration of the incremental encoding recognition module 20 will be described in detail below. The preset encoding rule of the incremental encoding recognition module 20 further comprises: when the direction of the data increment is downward and the amplitude is less than a preset first amplitude, performing first type discrete code encoding; when the direction of the data increment is downward and the amplitude is greater than or equal to the preset first amplitude and less than a preset second amplitude, performing second type discrete code encoding; when the amplitude of the data increment is 0, performing third type discrete code encoding; when the direction of the data increment is upward and the amplitude is greater than a preset third amplitude and less than or equal to a preset fourth amplitude, performing fourth type discrete code encoding; and when the direction of the data increment is upward and the amplitude is greater than the preset fourth amplitude, performing fifth type discrete code encoding.

[0050] The specific configuration of the structured interaction module 30 will be described in detail below. The structured interaction module 30 further comprises: performing single-point anomaly recognition on M signal token sequences to obtain a set of abnormal signal token sequences; extracting a first abnormal signal token sequence from the set of abnormal signal token sequences, and performing one-level multi-dimensional structured interaction in the mine topology graph by combining the position identifier of the first device topology node corresponding to the first abnormal signal token sequence to obtain a set of one-level associated device topology nodes; performing two-level multi-dimensional structured interaction in the mine topology graph on the set of one-level associated device topology nodes to obtain a set of two-level associated device topology nodes; and constructing a first local associated subgraph by combining the first device topology node, the set of one-level associated device topology nodes, and the set of two-level associated device topology nodes, and adding the first local associated subgraph to a set of local associated subgraphs.

[0051] The specific configuration of the structured interaction module 30 will be described in detail below. The structured interaction module 30 further comprises: a pre-constructed single-point anomaly recognizer, wherein the single-point anomaly recognizer comprises a single-point anomaly threshold identification branch and a single-point anomaly trend identification branch; performing threshold anomaly discrimination on M signal token sequences based on the single-point anomaly threshold identification branch to obtain a set of threshold abnormal signal token sequences; performing trend anomaly discrimination on M signal token sequences based on the single-point anomaly trend identification branch to obtain a set of trend abnormal signal token sequences; and performing set union on the set of threshold abnormal signal token sequences and the set of trend abnormal signal token sequences to obtain the set of abnormal signal token sequences.

[0052] Next, the specific configuration of the detection anomaly identification module 40 will be described in detail. The detection anomaly identification module 40 further comprises: extracting a first local correlation subgraph from the set of local correlation subgraphs; extracting a set of relevant first local signal token sequences from the M signal token sequences based on the first local correlation subgraph; performing scenario-based local gas detection anomaly identification based on the set of first local signal token sequences to obtain a first anomaly identification result, and adding the first anomaly identification result into the set of identification results.

[0053] Next, the specific configuration of the detection anomaly identification module 40 will be described in detail. The detection anomaly identification module 40 further comprises: pre-constructing a set of historical gas detection anomaly interaction logs; performing scenario-based semantic clustering analysis by traversing the set of historical gas detection anomaly interaction logs to construct a scenario-based semantic template library; and matching the set of first local signal token sequences with the scenario-based semantic template library to obtain a first anomaly identification result.

[0054] Next, the specific configuration of the detection anomaly identification module 40 will be described in detail. The detection anomaly identification module 40 further comprises: performing semantic clustering on the set of historical gas detection anomaly interaction logs according to a preset clustering granularity to obtain a plurality of clustered sets of historical gas detection anomaly interaction logs; and performing scenario-based template extraction by traversing the plurality of clustered sets of historical gas detection anomaly interaction logs to obtain the scenario-based semantic template library.

[0055] The gas detection alarm system of the mine safety device provided in the embodiments of the present application can perform the gas detection alarm method of the mine safety device provided in any of the embodiments of the present application, and has the corresponding function modules and beneficial effects of the execution method.

[0056] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server, and the various units and modules are only divided according to the functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of mutual differentiation, and do not limit the protection scope of the present application.

[0057] The above specific embodiments do not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A gas detection and alarm method for mining safety equipment, characterized in that, The method includes: Obtain basic gas detection associated equipment information of the target mine and construct a mine topology map, wherein the mine topology map includes M equipment topology nodes and Q heterogeneous edges, where M and Q are positive integers; Incremental encoding and identification are performed on the gas correlation detection data sequences of the M device topology nodes within a preset detection window to obtain M signal token sequences; By combining the mine topology map with the M signal token sequences, a multi-dimensional structured interaction is performed to obtain a set of locally related subgraphs; By combining the M signal token sequences, local gas detection anomaly identification is performed on the local associated subgraph set, and gas detection alarm information is obtained based on the identification result set.

2. The gas detection and alarm method for mining safety equipment as described in claim 1, characterized in that, Obtain basic gas detection and related equipment information for the target mine, and construct a mine topology map, wherein the mine topology map includes M equipment topology nodes and Q heterogeneous edges, including: Extract the equipment type and equipment deployment location from the basic gas detection associated equipment information to obtain M equipment topology nodes with location identifiers; Based on the ventilation connection relationship, electrical feeder connection relationship, physical adjacency relationship and operation relationship in the basic gas detection associated equipment information, an associated edge is constructed for two related equipment topology nodes among the M equipment topology nodes, and the associated edge is type-identified according to the relationship type to obtain Q heterogeneous edges; The mine topology map is constructed based on the Q heterogeneous edges and M device topology nodes with location identifiers.

3. The gas detection and alarm method for mining safety equipment as described in claim 1, characterized in that, Incremental encoding and identification are performed on the gas correlation detection data sequences of the M device topology nodes within a preset detection window to obtain M signal token sequences, including: The gas correlation detection data sequences of the M nodes are normalized respectively, and the adjacency difference operation is performed on the normalized sequences to determine the data increment and obtain the incremental sequence of gas correlation monitoring data of the M nodes. Based on the direction and magnitude of the data increments in the incremental sequence of gas-related monitoring data from the M nodes, the incremental sequence of gas-related detection data from the M nodes is discretely symbolically encoded according to a preset encoding rule to obtain the M signal token sequences.

4. The gas detection and alarm method for mining safety equipment as described in claim 3, characterized in that, The preset encoding rule is that when the direction of the data increment is downward and the magnitude is less than the preset first magnitude, the first type of discrete code encoding is performed; When the direction of the data increment is downward and the magnitude is greater than or equal to the preset first magnitude and less than the preset second magnitude, the second type of discrete code encoding is performed; When the magnitude of the data increment is 0, third-type discrete code encoding is performed; When the direction of the data increment is upward and the magnitude is greater than the preset third magnitude and less than or equal to the preset fourth magnitude, the fourth type of discrete code encoding is performed. When the direction of the data increment is upward and the magnitude is greater than the preset fourth magnitude, the fifth type of discrete code encoding is performed.

5. The gas detection and alarm method for mining safety equipment as described in claim 1, characterized in that, By combining the mine topology map with the M signal token sequences, a multi-dimensional structured interaction is performed to obtain a set of locally related subgraphs, including: Perform single-point anomaly identification on M signal token sequences to obtain a set of abnormal signal token sequences; Extract the first abnormal signal token sequence from the set of abnormal signal token sequences, and combine it with the location identifier of the first equipment topology node corresponding to the first abnormal signal token sequence. Perform a first-level multi-dimensional structured interaction on the mine topology map to obtain a set of first-level associated equipment topology nodes. In the mine topology map, a second-level multidimensional structured interaction is performed on the set of topology nodes of the first-level associated equipment to obtain a set of topology nodes of the second-level associated equipment. By combining the first device topology node, the set of first-level associated device topology nodes, and the set of second-level associated device topology nodes, a first local associated subgraph is constructed, and the first local associated subgraph is added to the set of local associated subgraphs.

6. The gas detection and alarm method for mining safety equipment as described in claim 5, characterized in that, Single-point anomaly identification is performed on M signal token sequences to obtain a set of anomalous signal token sequences, including: A pre-built single-point anomaly detector is provided, wherein the single-point anomaly detector includes a single-point anomaly threshold identification branch and a single-point anomaly trend identification branch; Based on the single-point anomaly threshold identification branch, threshold anomaly discrimination is performed on M signal token sequences to obtain a set of threshold anomaly signal token sequences; Based on the single-point anomaly trend identification branch, trend anomaly discrimination is performed on M signal token sequences to obtain a set of trend anomaly signal token sequences; The abnormal signal token sequence set is obtained by performing a union of the threshold abnormal signal token sequence set and the trend abnormal signal token sequence set.

7. The gas detection and alarm method for mining safety equipment as described in claim 1, characterized in that, By combining the M signal token sequences, local gas detection anomaly identification is performed on the local associated subgraph set, and gas detection alarm information is obtained based on the identification result set, including: Extract the first local association subgraph from the set of local association subgraphs; Based on the first local association subgraph, extract the set of related first local signal token sequences from the M signal token sequences; Based on the first set of local signal token sequences, a scenario-based local gas detection anomaly identification is performed to obtain a first anomaly identification result, which is then added to the identification result set.

8. The gas detection and alarm method for mining safety equipment as described in claim 7, characterized in that, Based on the first set of local signal token sequences, a scenario-based local gas detection anomaly identification is performed to obtain a first anomaly identification result, including: Pre-build a collection of historical gas detection anomaly interaction logs; The historical gas detection anomaly interaction log set is traversed to perform scenario-based semantic cluster analysis, and a scenario-based semantic template library is constructed. The first set of local signal token sequences is matched with the contextualized semantic template library to obtain the first anomaly identification result.

9. The gas detection and alarm method for mining safety equipment as described in claim 8, characterized in that, The historical gas detection anomaly interaction log set is traversed to perform scenario-based semantic clustering analysis, and a scenario-based semantic template library is constructed, including: Semantic clustering is performed on the historical gas detection anomaly interaction log set according to a preset clustering granularity to obtain multiple clustered historical gas detection anomaly interaction log sets. The scenario-based template library is obtained by traversing the multiple cluster historical gas detection anomaly interaction log sets.

10. A gas detection and alarm system for mining safety equipment, characterized in that, The system is used to implement the gas detection and alarm method for mining safety equipment according to any one of claims 1 to 9, the system comprising: The topology graph construction module is used to obtain basic gas detection associated equipment information of the target mine and construct a mine topology graph, wherein the mine topology graph includes M equipment topology nodes and Q heterogeneous edges, where M and Q are positive integers; The incremental coding and recognition module is used to perform incremental coding and recognition on the gas correlation detection data sequences of the M device topology nodes within a preset detection window to obtain M signal token sequences. The structured interaction module is used to perform multi-dimensional structured interaction on the M signal token sequences in conjunction with the mine topology map to obtain a set of local associated subgraphs; The anomaly detection module is used to perform local gas detection anomaly identification on the local associated subgraph set by combining the M signal token sequences, and obtain gas detection alarm information based on the identification result set.

Images