A method for constructing mine disaster maps based on spatiotemporal semantics

Abstract

The present invention is a method for constructing a mine disaster map based on spatiotemporal semantics, and provides a mine gas outburst prediction method based on spatiotemporal semantic map, which relates to the field of coal mine safety, including obtaining spatiotemporal perception data of gas outburst and traceability indicators of working face, constructing spatiotemporal semantic nodes of gas outburst, constructing spatiotemporal semantic adjacent edges of gas outburst, embedding spatiotemporal semantic graph of gas outburst, aggregating spatiotemporal semantics of gas outburst using graph neural network, and predicting the abnormal gas outburst location using link prediction algorithm. Through the above method steps, the abnormal gas outburst map of working face is constructed, combined with graph embedding, graph message aggregation, and link prediction algorithm, the catastrophic spatiotemporal information of abnormal gas outburst in mine and the disaster semantic knowledge are integrated, and the gas outburst location is predicted, which provides decision-making guidance for coal mine disaster prevention and disaster prevention, and is expected to bring important technical support for coal mine disaster safety management.

Description

Technical Field

[0001] The present invention relates to the field of coal mine safety technology, and in particular to a method for constructing a mine disaster map based on spatiotemporal semantics. Background Art

[0002] At present, coal mine safety is the top priority of mine mining, and the main safety protection in coal mine safety lies in the technology of identifying abnormal gas conditions.

[0003] Traditional gas anomaly identification technology focuses on mining historical gas outburst data, but fails to fully consider the factors of abnormal gas outburst and the temporal and spatial variation characteristics of the factors. Therefore, traditional gas anomaly identification technology cannot accurately predict the location of gas outburst and is increasingly unsuitable for current mine mining. Summary of the invention

[0004] In view of the above problems, the present invention proposes a method for constructing a mine disaster map based on spatiotemporal semantics.

[0005] The embodiment of the present invention provides a method for constructing a mine disaster map based on spatiotemporal semantics, the method for constructing a mine disaster map comprising:

[0006] Obtain the spatiotemporal perception data of gas outburst at the working face and the spatiotemporal perception data of causal indicators;

[0007] Using the working face gas emission spatiotemporal perception data and the traceability index spatiotemporal perception data, constructing a gas emission spatiotemporal semantic node;

[0008] Based on the gas emission spatiotemporal semantic nodes, constructing gas emission spatiotemporal semantic adjacency edges;

[0009] A gas outflow spatiotemporal semantic graph is formed based on the gas outflow spatiotemporal semantic nodes and the gas outflow spatiotemporal semantic adjacent edges;

[0010] Using a graph embedding algorithm to perform representation learning on the gas emission spatiotemporal semantic graph;

[0011] The message passing algorithm of graph neural network is used to aggregate and represent the learned spatiotemporal semantic graph of gas outburst;

[0012] Based on the aggregated spatiotemporal semantic graph of gas emission, the link prediction algorithm is used to predict the location of abnormal gas emission.

[0013] Optionally, the spatiotemporal perception data of gas outburst at the working face and the spatiotemporal perception data of the causal index are obtained, including:

[0014] Acquire the spatiotemporal perception data of gas outburst at the working face and the spatiotemporal perception data of the causal index from a database;

[0015] The spatiotemporal data of abnormal gas outburst at the working face include: monitoring time, spatial position, and concentration value of the gas sensor at the mining working face;

[0016] The temporal and spatial perception data of the traceability indicator include: the time and spatial position of the traceability indicator, and the value item of the traceability indicator;

[0017] The traceability indicator value items include: working face air volume monitoring value, air door short circuit, air inlet tunnel short circuit, tunnel air leakage, abnormal gas outburst in goaf, sensor adjustment, periodic pressure, existence of old goaf and old tunnels, drilling, roof collapse, coal seam thickening, roof spalling, blasting, excessive advancement speed, and blockage of the rear chute of top coal release.

[0018] Optionally, the gas outflow spatiotemporal semantic nodes include: a gas outflow abnormal event node, a traceable event node, a gas outflow abnormal time node, a traceable event time node, a gas outflow abnormal spatial position node, and a traceable indicator event spatial position node;

[0019] Using the working face gas emission spatiotemporal perception data and the traceability index spatiotemporal perception data, a gas emission spatiotemporal semantic node is constructed, including:

[0020] Based on the concentration value, construct a gas outburst abnormal event node;

[0021] Based on the traceability index value item, a traceability event node is constructed. The traceability event node is represented by a character string, including: air door short circuit event node, air intake lane short circuit event node, lane air leakage event node, goaf gas abnormality event node, sensor adjustment event node, periodic pressure node, old goaf and old lane event node, drilling event node, roof collapse event node, coal seam thickening event node, roof spalling event node, blasting event node, excessive advancement speed event node, and top coal rear blockage event node;

[0022] Based on the monitoring time in the spatiotemporal data of abnormal gas emission in the working face, a time node of abnormal gas emission is constructed, wherein the time node of abnormal gas emission is represented by a timestamp of the occurrence of the abnormal gas event;

[0023] Based on the time of the traceability indicator, construct a traceability event time node, wherein the traceability event time node is represented by a timestamp of the occurrence of the traceability event;

[0024] Based on the spatial position in the spatiotemporal data of abnormal gas outburst in the working face, a spatial position node of abnormal gas outburst is constructed, wherein the spatial position node of abnormal gas outburst is represented by a spatial coordinate triple (x, y, z);

[0025] Based on the spatial position of the traceability indicator, a spatial position node of the traceability indicator event is constructed, and the spatial position node of the traceability indicator event is represented by a spatial coordinate triple (x, y, z).

[0026] Optionally, the constructed gas emission spatiotemporal semantic adjacent edge is a weighted directed edge, which is connected from the head node to the tail node in the gas emission spatiotemporal semantic node, and includes:

[0027] Temporal adjacency edges, spatial connection edges, and semantic connection edges.

[0028] Optionally, the temporal adjacent edge is a weighted directed edge, connected from a head node to a tail node in the gas outburst spatiotemporal semantic node, including:

[0029] According to the order of occurrence, the abnormal gas outburst time node and the traceable event time node are connected, and the edge weight is expressed as: after that, the head node is the abnormal gas outburst time node and / or the traceable event time node, and the tail node is the abnormal gas outburst time node and / or the traceable event time node;

[0030] Connect the gas outflow abnormal event node to the gas outflow abnormal time node, and the edge weight is represented by: a character string of the occurrence time, the head node is the gas outflow abnormal event node, and the tail node is the gas outflow abnormal time node;

[0031] Connect the traceable event node to the causal event time node, and the edge weight is expressed as: a string of occurrence time, the head node is the traceable event node, and the tail node is the causal event time node.

[0032] Optionally, the spatial adjacent edge is a weighted directed edge, connected from a head node to a tail node in the gas outburst spatiotemporal semantic node, including:

[0033] According to the spatial position relationship, the spatial position node of the abnormal gas outburst and / or the spatial position node of the traceable event are connected, and the edge weight is expressed as: the spatial distance between the two nodes, the head node is the spatial position node of the abnormal gas outburst and / or the spatial position node of the traceable event, and the tail node is the spatial position node of the abnormal gas outburst and / or the spatial position node of the traceable event;

[0034] Connect the gas outflow abnormal event node to the gas outflow abnormal spatial location node, and the edge weight is represented by: a character string of the occurrence location, the head node is the gas outflow abnormal event node, and the tail node is the gas outflow abnormal spatial location node;

[0035] Connect the traceable event node to the spatial location node of the traceable event, and the edge weight is expressed as: a character string of the occurrence location, the head node is the traceable event node, and the tail node is the spatial location node of the traceable event.

[0036] Optionally, the semantic adjacent edge is a directed weighted edge, connected from the head node to the tail node in the gas outburst spatiotemporal semantic node, including:

[0037] Connect the gas outflow abnormal event node to the traceable event node, and the edge weight is represented by: the character string causing the disaster, the head node is the gas outflow abnormal event node, and the tail node is the traceable event node;

[0038] The traceable event nodes are connected, and the edge weight is expressed as: an internally associated string, the head node is the traceable event node, and the tail node is the traceable event node.

[0039] Optionally, using a graph embedding algorithm to perform representation learning on the gas emission spatiotemporal semantic graph includes:

[0040] The gas emission spatiotemporal semantic graph is embedded using the RotatE algorithm, specifically including:

[0041] The triple (h, r, t) consisting of the head entity h, edge weight r, and tail entity t in the mine disaster map is mapped to the vector space. The RotatE distance of the triple (h, r, t) is expressed as:

[0042] dr(h, t) = ∥h⊙rt∥

[0043] In the above formula, ⊙ represents the Hadamard product, and ∥ represents the Euclidean norm.

[0044] Optionally, a message passing algorithm of a graph neural network is used to aggregate and represent the learned spatiotemporal semantic graph of gas emission, including:

[0045] Aggregating the information of the node itself in the gas outburst spatiotemporal semantic node by a determined aggregation function, and then merging and updating the aggregation value with the information of the node itself;

[0046] The aggregation function is expressed as:

[0047] b v (k) =AGG (k) ({T u (k-1) :u∈N(v)}),

[0048] T v (k) =COM (k) (T v (k-1), b v (k) )

[0049] In the above formula, b is the aggregation result, u is the element in the node set directly connected to node v in the embedding representation of the spatiotemporal semantic graph of gas outburst, N(v) is the node set directly adjacent to node v, k and k-1 are the number of aggregations, T is the combination of the information from the upper layer and the aggregation result of this layer, and AGG and COM are mean aggregators.

[0050] Optionally, based on the aggregated spatiotemporal semantic graph of gas emission, a link prediction algorithm is used to predict the abnormal gas emission location, including:

[0051] The link prediction algorithm is used to predict the gas outburst location, and node u is used to represent the gas abnormal event node to be predicted. V is a set of all historical gas outburst event space nodes v, V = {v: historical gas outburst event space node}. By taking the dot product of u and each element v in V, the probability that u corresponds to different outburst abnormal locations can be obtained, and the gas outburst event space node with the largest probability is taken to represent the predicted gas outburst location.

[0052] The method for constructing a mine disaster map based on spatiotemporal semantics provided by the present invention first obtains the spatiotemporal perception data of gas outbursts at the working face and the spatiotemporal perception data of causal indicators; then uses the spatiotemporal perception data of gas outbursts at the working face and the spatiotemporal perception data of causal indicators to construct the spatiotemporal semantic nodes of gas outbursts.

[0053] Based on the spatiotemporal semantic nodes of gas emission, the spatiotemporal semantic adjacent edges of gas emission are constructed. Based on the spatiotemporal semantic nodes of gas emission and the spatiotemporal semantic adjacent edges of gas emission, the spatiotemporal semantic graph of gas emission is constructed. The graph embedding algorithm is used to perform representation learning on the spatiotemporal semantic graph of gas emission.

[0054] The message passing algorithm of graph neural network is used to aggregate the learned spatiotemporal semantic graph of gas emission. Finally, based on the aggregated spatiotemporal semantic graph of gas emission, the link prediction algorithm is used to predict the location of abnormal gas emission.

[0055] In view of the fact that traditional gas anomaly identification technology focuses on mining historical gas emission data but fails to fully consider the factors of abnormal gas emission and the spatiotemporal variation characteristics of the factors, the present invention creatively proposes to use knowledge graph technology to integrate and analyze the spatiotemporal semantic associations between abnormal gas emission and working face air volume, air gate short circuit, air inlet lane short circuit, lane leakage, goaf gas emission abnormal values, sensor adjustment, periodic pressure, existence of old goaf and old lanes, drilling, roof collapse, coal seam thickening, roof spalling, blasting, excessive advancement speed, and blockage of the rear part of top coal caving, and uses graph embedding learning and link prediction algorithm to identify the factors of abnormal gas emission.

[0056] Not only the semantic knowledge of the causal relationship between abnormal gas outburst and disaster-causing factors is considered, but also the knowledge of the temporal and spatial variation characteristics of disaster-causing factors is considered, realizing the prediction and causal analysis of abnormal gas outburst. Through the above method steps, the abnormal gas outburst map of the working face is constructed, combined with graph embedding, graph message aggregation, and link prediction algorithm, the disaster spatiotemporal information of abnormal gas outburst in the mine and the semantic knowledge of disaster-causing factors are integrated to predict the location of gas outburst, which provides decision-making guidance for coal mine disaster prevention and disaster prevention, and is expected to bring important technical support to coal mine disaster safety management, with high practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0058] Figure 1 is a flow chart of a method for constructing a mine disaster map based on spatiotemporal semantics according to an embodiment of the present invention;

[0059] Figure 2 It is a structural diagram of predicting the location of abnormal gas outburst using the mine disaster map construction method based on spatiotemporal semantics proposed in the embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, are only part of the embodiments of the present invention, not all of the embodiments, and are not used to limit the present invention.

[0061] Reference Figure 1 , shows a flow chart of a method for constructing a mine disaster map based on spatiotemporal semantics according to an embodiment of the present invention, the method for constructing a mine disaster map comprising:

[0062] Step 101: Obtain the spatiotemporal perception data of gas outburst at the working face and the spatiotemporal perception data of causal indicators.

[0063] The method for constructing a mine disaster map based on spatiotemporal semantics proposed in the present invention first needs to obtain the spatiotemporal perception data of gas outburst at the working face and the spatiotemporal perception data of causal indicators.

[0064] A better choice is to obtain the spatiotemporal perception data of gas outburst at the working face and the spatiotemporal perception data of the traceability index from the database. Of course, the spatiotemporal perception data of gas outburst at the working face and the spatiotemporal perception data of the traceability index can also be obtained from other channels or other methods, which will not be elaborated one by one.

[0065] The spatiotemporal data of abnormal gas outburst at the working face include but are not limited to: the monitoring time, spatial position, and concentration value of the gas sensor at the mining working face; the spatiotemporal perception data of the causal indicators include but are not limited to: the time, spatial position, and value items of the causal indicators.

[0066] Among them, the tracing indicator values ​​include: working face air volume monitoring value, air door short circuit, air inlet tunnel short circuit, tunnel air leakage, abnormal gas outburst in goaf, sensor adjustment, periodic pressure, existence of old goaf and old tunnels, drilling, roof collapse, coal seam thickening, roof spalling, blasting, excessive advancement speed, and blockage of the rear channel of top coal release.

[0067] Step 102: Use the spatiotemporal perception data of gas outburst in the working face and the spatiotemporal perception data of the causal indicators to construct a spatiotemporal semantic node of gas outburst.

[0068] After obtaining the spatiotemporal perception data of gas outburst in the working face and the spatiotemporal perception data of causal indicators, the spatiotemporal semantic nodes of gas outburst can be constructed by using the spatiotemporal perception data of gas outburst in the working face and the spatiotemporal perception data of causal indicators.

[0069] Preferably, the spatiotemporal semantic nodes may include: abnormal gas outburst event node, tracing event node, abnormal gas outburst time node, tracing event time node, abnormal gas outburst spatial position node, and tracing indicator event spatial position node.

[0070] A better method to construct spatiotemporal semantic nodes of gas outburst by using spatiotemporal perception data of gas outburst in working face and spatiotemporal perception data of traceability indicators includes:

[0071] Step S1: constructing a gas outburst abnormal event node based on the concentration value;

[0072] Step S2: Based on the traceability index value item, construct a traceability event node, which is represented by a character string, including: air door short circuit event node, air inlet lane short circuit event node, lane air leakage event node, goaf gas abnormality event node, sensor adjustment event node, periodic pressure node, old goaf and old lane event node, drilling event node, roof collapse event node, coal seam thickening event node, roof spalling event node, blasting event node, excessive advancement speed event node, and top coal rear blockage event node;

[0073] Step S3: constructing a gas abnormal outburst time node based on the monitoring time in the spatiotemporal data of abnormal gas outburst in the working face, wherein the gas abnormal outburst time node is represented by a timestamp of the occurrence of the gas abnormal event;

[0074] Step S4: constructing a traceability event time node based on the time of the traceability indicator, and the traceability event time node is represented by the timestamp of the traceability event;

[0075] Step S5: constructing a spatial position node of abnormal gas outburst based on the spatial position in the spatiotemporal data of abnormal gas outburst in the working face, wherein the spatial position node of abnormal gas outburst is represented by a spatial coordinate triple (x, y, z);

[0076] Step S6: Based on the spatial position of the traceability indicator, a spatial position node of the traceability indicator event is constructed, and the spatial position node of the traceability indicator event is represented by a spatial coordinate triple (x, y, z).

[0077] Step 103: Based on the gas emission spatiotemporal semantic nodes, construct gas emission spatiotemporal semantic adjacent edges.

[0078] After the gas outflow spatiotemporal semantic node is constructed, the gas outflow spatiotemporal semantic adjacent edge can be constructed based on the gas outflow spatiotemporal semantic node. The constructed gas outflow spatiotemporal semantic adjacent edge is a weighted directed edge, which is connected from the head node to the tail node in the gas outflow spatiotemporal semantic node, and includes:

[0079] Temporal adjacency edges, spatial connection edges, and semantic connection edges.

[0080] Specifically: The temporal adjacent edge is a weighted directed edge, which connects the head node to the tail node in the gas outburst spatiotemporal semantic node, and includes:

[0081] According to the order of occurrence, the abnormal gas outburst time node and the traceable event time node are connected, and the edge weight is expressed as: "afterwards", the head node is the abnormal gas outburst time node and / or the traceable event time node, and the tail node is the abnormal gas outburst time node and / or the traceable event time node. It should be noted that the "and / or" here and below means both or one of them is selected, for example: the head node is the abnormal gas outburst time node and / or the traceable event time node, which means: the head node is the abnormal gas outburst time node and the traceable event time node, or the head node is one of the abnormal gas outburst time node and the traceable event time node. The rest of the cases are similar and will not be described one by one.

[0082] Connect the gas outflow abnormal event node to the gas outflow abnormal time node. The edge weight is represented by a string of "occurrence time". The head node is the gas outflow abnormal event node, and the tail node is the gas outflow abnormal time node.

[0083] Connect the traceable event node to the causal event time node. The edge weight is represented by a string of "occurrence time". The head node is the traceable event node, and the tail node is the causal event time node.

[0084] The spatial adjacent edge is also a weighted directed edge, which is connected from the head node to the tail node in the gas outburst spatiotemporal semantic node, including:

[0085] According to the spatial position relationship, the spatial position nodes of abnormal gas outburst and / or the spatial position nodes of the causal events are connected, and the edge weight is expressed as: the spatial distance between the two nodes, the head node is the spatial position node of abnormal gas outburst and / or the spatial position node of the causal event, and the tail node is the spatial position node of abnormal gas outburst and / or the spatial position node of the causal event.

[0086] Connect the gas outburst abnormal event node to the gas outburst abnormal spatial location node. The edge weight is represented by a string of "occurrence location". The head node is the gas outburst abnormal event node, and the tail node is the gas outburst abnormal spatial location node.

[0087] Connect the traceable event node to the spatial location node of the traceable event, and the edge weight is expressed as a string of "occurrence location". The head node is the traceable event node, and the tail node is the spatial location node of the traceable event.

[0088] The semantic adjacent edge is also a directed weighted edge, which is connected from the head node to the tail node in the gas outburst spatiotemporal semantic node, including:

[0089] Connect the gas outflow abnormal event node to the tracing event node. The edge weight is represented by the string of "causing disaster". The head node is the gas outflow abnormal event node, and the tail node is the tracing event node.

[0090] Connect the traceable event nodes, the edge weight is expressed as a string of "internal association", the head node is the traceable event node, and the tail node is the traceable event node.

[0091] The above method can be used to obtain the spatiotemporal semantic adjacent edges of gas outburst with weighted directed edges.

[0092] Step 104: construct a gas emission spatiotemporal semantic graph based on the gas emission spatiotemporal semantic nodes and the gas emission spatiotemporal semantic adjacent edges.

[0093] After step 103, a gas outflow spatiotemporal semantic graph is constructed based on the gas outflow spatiotemporal semantic nodes and the gas outflow spatiotemporal semantic adjacent edges. The specific method of constructing the semantic graph can refer to the currently known method of constructing the semantic graph, which will not be described in detail.

[0094] Step 105: Use a graph embedding algorithm to perform representation learning on the spatiotemporal semantic graph of gas outburst.

[0095] A variety of graph embedding algorithms can be used to represent and learn the spatiotemporal semantic graph of gas outburst. A better way is to embed the spatiotemporal semantic graph of gas outburst using the RotatE algorithm, which specifically includes:

[0096] The triple (h, r, t) consisting of the head entity h, edge weight r, and tail entity t in the mine disaster map is mapped to the vector space. The RotatE distance of the triple (h, r, t) is expressed as:

[0097] dr(h, t) = ∥h⊙rt∥

[0098] In the above formula, ⊙ represents the Hadamard product, and ∥ represents the Euclidean norm.

[0099] Step 106: Use the message passing algorithm of the graph neural network to aggregate and represent the learned spatiotemporal semantic graph of gas emission.

[0100] After learning the representation of the gas emission spatiotemporal semantic graph, the message passing algorithm of the graph neural network is used to aggregate the representation of the learned gas emission spatiotemporal semantic graph. A better way is:

[0101] The information of the nodes in the gas outburst spatiotemporal semantic nodes is aggregated by a certain aggregation function, and then the aggregated value is merged and updated with the node's own information. The aggregation function can be expressed as:

[0102] b v (k) =AGG (k) ({T u (k-1) :u∈Ν(v)}),

[0103] T v (k) =COM (k) (T v (k-1) ,b v (k) )

[0104] In the above formula, b is the aggregation result, u is the element in the node set directly connected to node v in the embedded representation of the spatiotemporal semantic graph of gas outburst, Ν(v) is the node set directly adjacent to node v, k and k-1 are the number of aggregations, T is the combination of the information from the upper layer and the aggregation result of this layer, and AGG and COM are mean aggregators.

[0105] Step 107: Based on the aggregated spatiotemporal semantic graph of gas emission, a link prediction algorithm is used to predict the location of abnormal gas emission.

[0106] The aggregated spatiotemporal semantic graph of gas outburst is finally used to predict the abnormal gas outburst location using the link prediction algorithm.

[0107] A better method includes: using a link prediction algorithm to predict the gas outburst location, using node u to represent the gas abnormal event node to be predicted, V is a set of all historical gas outburst event space nodes v, V = {v: historical gas outburst event space node}, through the dot product of u and each element v in V, the probability of u corresponding to different outburst abnormal locations can be obtained, and the gas outburst event space node with the largest probability is taken to represent the predicted gas outburst location.

[0108] Reference Figure 2 , which exemplarily shows the architecture diagram of predicting the location of abnormal gas outburst using the mine disaster map construction method based on spatiotemporal semantics proposed in the present invention. Figure 2 The relationship between the abnormal gas emission concentration and the abnormal gas emission space node, the abnormal gas emission time node and the abduction index value item is shown as an example in FIG.

[0109] To summarize, the method for constructing a mine disaster map based on spatiotemporal semantics of the present invention first obtains the spatiotemporal perception data of gas outbursts at the working face and the spatiotemporal perception data of causal indicators; and then uses the spatiotemporal perception data of gas outbursts at the working face and the spatiotemporal perception data of causal indicators to construct the spatiotemporal semantic nodes of gas outbursts.

[0110] Based on the spatiotemporal semantic nodes of gas emission, the spatiotemporal semantic adjacent edges of gas emission are constructed. Based on the spatiotemporal semantic nodes of gas emission and the spatiotemporal semantic adjacent edges of gas emission, the spatiotemporal semantic graph of gas emission is constructed. The graph embedding algorithm is used to perform representation learning on the spatiotemporal semantic graph of gas emission.

[0111] The message passing algorithm of graph neural network is used to aggregate the learned spatiotemporal semantic graph of gas emission. Finally, based on the aggregated spatiotemporal semantic graph of gas emission, the link prediction algorithm is used to predict the location of abnormal gas emission.

[0112] In view of the fact that traditional gas anomaly identification technology focuses on mining historical gas emission data but fails to fully consider the factors of abnormal gas emission and the spatiotemporal variation characteristics of the factors, the present invention creatively proposes to use knowledge graph technology to integrate and analyze the spatiotemporal semantic associations between abnormal gas emission and working face air volume, air gate short circuit, air inlet lane short circuit, lane leakage, goaf gas emission abnormal values, sensor adjustment, periodic pressure, existence of old goaf and old lanes, drilling, roof collapse, coal seam thickening, roof spalling, blasting, excessive advancement speed, and blockage of the rear part of top coal caving, and uses graph embedding learning and link prediction algorithm to identify the factors of abnormal gas emission.

[0113] Not only the semantic knowledge of the causal relationship between abnormal gas outburst and disaster-causing factors is considered, but also the knowledge of the temporal and spatial variation characteristics of disaster-causing factors is considered, realizing the prediction and causal analysis of abnormal gas outburst. Through the above method steps, the abnormal gas outburst map of the working face is constructed, combined with graph embedding, graph message aggregation, and link prediction algorithm, the disaster spatiotemporal information of abnormal gas outburst in the mine and the semantic knowledge of disaster-causing factors are integrated to predict the location of gas outburst, which provides decision-making guidance for coal mine disaster prevention and disaster prevention, and is expected to bring important technical support to coal mine disaster safety management, with high practicality.

[0114] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0115] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0116] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A method for constructing a mine disaster map based on spatiotemporal semantics, characterized in that: The mine disaster map construction method comprises: Acquiring the spatiotemporal perception data of gas outburst at the working face and the spatiotemporal perception data of the traceability index, including: acquiring the spatiotemporal perception data of gas outburst at the working face and the spatiotemporal perception data of the traceability index from a database; wherein the spatiotemporal perception data of gas outburst at the working face includes: the monitoring time, spatial position, and concentration value of the gas sensor at the mining working face; the spatiotemporal perception data of the traceability index includes: the time, spatial position, and traceability index value item of the traceability index; the traceability index value item includes: the monitoring value of the working face air volume, the short circuit of the air door, the short circuit of the air inlet lane, the air leakage of the lane, the abnormal value of gas outburst in the goaf, the sensor adjustment, the periodic pressure, the existence of old goaf and old lane, drilling, roof collapse, coal seam thickening, roof leakage, blasting, excessive advancement speed, and blockage of the rear part of the top coal release; Using the working face gas outburst spatiotemporal perception data and the traceability index spatiotemporal perception data, a gas outburst spatiotemporal semantic node is constructed; the gas outburst spatiotemporal semantic node includes: a gas outburst abnormal event node, a traceability event node, a gas outburst abnormal time node, a traceability event time node, a gas outburst abnormal space location node, and a traceability index event space location node; Using the working face gas emission spatiotemporal perception data and the traceability index spatiotemporal perception data, a gas emission spatiotemporal semantic node is constructed, including: Based on the concentration value, construct a gas outburst abnormal event node; Based on the traceability index value item, a traceability event node is constructed. The traceability event node is represented by a character string, including: air door short circuit event node, air intake lane short circuit event node, lane air leakage event node, goaf gas abnormality event node, sensor adjustment event node, periodic pressure node, old goaf and old lane event node, drilling event node, roof collapse event node, coal seam thickening event node, roof spalling event node, blasting event node, excessive advancement speed event node, and top coal rear blockage event node; Based on the monitoring time in the spatiotemporal perception data of gas outburst at the working face, a gas outburst abnormal time node is constructed, and the gas outburst abnormal time node is represented by a timestamp of the occurrence of the gas abnormal event; Based on the time of the traceability indicator, construct a traceability event time node, wherein the traceability event time node is represented by a timestamp of the occurrence of the traceability event; Based on the spatial position in the spatiotemporal perception data of gas outburst at the working face, a gas abnormal outburst spatial position node is constructed, wherein the gas abnormal outburst spatial position node is represented by a spatial coordinate triple (x, y, z); Based on the spatial position in the spatiotemporal perception data of the abductive indicator, a spatial position node of the abductive indicator event is constructed, wherein the spatial position node of the abductive indicator event is represented by a spatial coordinate triple (x, y, z); Based on the gas emission spatiotemporal semantic nodes, constructing gas emission spatiotemporal semantic adjacency edges; A gas outflow spatiotemporal semantic graph is formed based on the gas outflow spatiotemporal semantic nodes and the gas outflow spatiotemporal semantic adjacent edges; Using a graph embedding algorithm to perform representation learning on the gas emission spatiotemporal semantic graph; The message passing algorithm of graph neural network is used to aggregate and represent the learned spatiotemporal semantic graph of gas outburst; Based on the aggregated spatiotemporal semantic graph of gas emission, the link prediction algorithm is used to predict the location of abnormal gas emission.

2. The method for constructing a mine disaster map based on spatiotemporal semantics according to claim 1, characterized in that: The constructed gas emission spatiotemporal semantic adjacent edge is a weighted directed edge, which is connected from the head node to the tail node in the gas emission spatiotemporal semantic node, and includes: Temporal adjacency edges, spatial adjacency edges, and semantic adjacency edges.

3. The method for constructing a mine disaster map based on spatiotemporal semantics according to claim 2, characterized in that: The temporal adjacent edge is a weighted directed edge, which is connected from the head node to the tail node in the gas outburst spatiotemporal semantic node, and includes: According to the order of occurrence, the time node of the traceable event is connected with the time node of the abnormal gas outburst, and the edge weight is expressed as: after that, the head node is the time node of the traceable event, and the tail node is the time node of the abnormal gas outburst; Connect the gas outflow abnormal event node to the gas outflow abnormal time node, and the edge weight is represented by: a character string of the occurrence time, the head node is the gas outflow abnormal event node, and the tail node is the gas outflow abnormal time node; Connect the traceable event node to the traceable event time node, and the edge weight is expressed as: a string of occurrence time, the head node is the traceable event node, and the tail node is the traceable event time node.

4. The method for constructing a mine disaster map based on spatiotemporal semantics according to claim 2, characterized in that: The spatial adjacent edge is a weighted directed edge, which is connected from the head node to the tail node in the gas outburst spatiotemporal semantic node, and includes: According to the spatial position relationship, the spatial position node of the abnormal gas outburst and / or the spatial position node of the traceable indicator event are connected, and the edge weight is expressed as: the spatial distance between the two nodes, the head node is the spatial position node of the abnormal gas outburst and / or the spatial position node of the traceable indicator event, and the tail node is the spatial position node of the abnormal gas outburst and / or the spatial position node of the traceable indicator event; Connect the gas outflow abnormal event node to the gas outflow abnormal spatial location node, and the edge weight is represented by: a character string of the occurrence location, the head node is the gas outflow abnormal event node, and the tail node is the gas outflow abnormal spatial location node; Connect the traceable event node to the spatial position node of the traceable indicator event, and the edge weight is expressed as: a string of the occurrence location, the head node is the traceable event node, and the tail node is the spatial position node of the traceable indicator event.

5. The method for constructing a mine disaster map based on spatiotemporal semantics according to claim 2, characterized in that: The semantic adjacent edge is a directed weighted edge, which is connected from the head node to the tail node in the gas outburst spatiotemporal semantic node, and includes: Connect the gas outflow abnormal event node to the traceable event node, and the edge weight is represented by: the character string causing the disaster, the head node is the gas outflow abnormal event node, and the tail node is the traceable event node; The traceable event nodes are connected, and the edge weight is expressed as: an internally associated string, the head node is the traceable event node, and the tail node is the traceable event node.

6. The method for constructing a mine disaster map based on spatiotemporal semantics according to claim 1, characterized in that: The gas emission spatiotemporal semantic graph is represented and learned by using a graph embedding algorithm, including: The gas emission spatiotemporal semantic graph is embedded using the RotatE algorithm, specifically including: The triple (h, r, t) consisting of the head entity h, edge weight r, and tail entity t in the mine disaster map is mapped to the vector space. The RotatE distance of the triple (h, r, t) is expressed as: dr(h, t)= / / h☉rt / / In the above formula, ⊙ represents the Hadamard product, and / / represents the Euclidean norm.

7. The method for constructing a mine disaster map based on spatiotemporal semantics according to claim 1, characterized in that: The message passing algorithm of the graph neural network is used to aggregate the learned spatiotemporal semantic graph of gas outburst, including: Aggregating the information of the nodes themselves in the gas outflow space-time semantic node by a determined aggregation function, and then merging and updating the aggregation value with the information of the nodes themselves in the gas outflow space-time semantic node; The aggregation function is expressed as: b v (k) =AGG (k) ({T u (k-1) :u∈Ν(v)}), T v (k) =COM (k) (T v (k-1) ,b v (k) ) In the above formula, b is the aggregation result, u is the element in the node set directly connected to the gas outflow spatiotemporal semantic node v in the embedded representation of the gas outflow spatiotemporal semantic graph, Ν(v) is the node set directly adjacent to the gas outflow spatiotemporal semantic node v, k and k-1 are the aggregation times, T is the combination of the information from the upper layer and the aggregation result of this layer, and AGG and COM are mean aggregators.

8. The method for constructing a mine disaster map based on spatiotemporal semantics according to claim 1, characterized in that: Based on the aggregated spatiotemporal semantic graph of gas outburst, the link prediction algorithm is used to predict the abnormal gas outburst location, including: The link prediction algorithm is used to predict the gas outburst location, and the node u1 is used to represent the node of the gas outburst abnormal event to be predicted. V is a set of all historical gas outburst abnormal spatial position nodes v1, V = {v1: historical gas outburst abnormal spatial position node}. By taking the dot product of u1 and each element v1 in V, the probability of u1 corresponding to different outburst abnormal positions can be obtained, and the gas outburst abnormal spatial position node with the largest probability is taken to represent the predicted gas outburst location.

Images