A coal mine accident cause analysis method based on a knowledge graph

Abstract

The present application relates to the technical field of coal mine accident cause analysis, and proposes a coal mine accident cause analysis method based on a knowledge graph.The specific steps of the present application include: collecting coal mine roof accident reports and cleaning the text to obtain a roof accident word set;secondly, by extracting the accident cause nodes, accident attributes and their relationships of the word set, a roof accident cause knowledge graph is constructed;then, the similarity of the accident cause nodes is calculated, and node merging is performed based on the similarity;then, the cause attribute code and the accident attribute code are obtained, the cause correlation index of the accident cause node is calculated, and the single-factor risk index and the multi-factor risk index are obtained in combination with the accident attribute code;finally, according to the single-factor risk and the multi-factor risk, the comprehensive risk index is determined, and the danger degree of the accident cause node is sorted.The present application can effectively analyze the risk factors of coal mine roof accidents, thereby reducing the potential risk of roof accidents.

Description

Technical Field

[0001] This invention relates to the field of coal mine accident cause analysis technology, specifically a coal mine accident cause analysis method based on knowledge graphs. Background Technology

[0002] Coal mine roof collapse accidents are a common and serious safety incident during mining operations, typically manifesting as the collapse, slippage, and subsidence of the roof in coal mine roadways or working faces. These accidents are closely related to multiple factors, including the mine's geological conditions, mining methods, support techniques, and roof management. During underground mining operations, insufficient understanding of geological conditions, outdated support technologies, and inadequate mine management and safety measures can lead to the destruction and collapse of rock strata or the roof, resulting in severe property damage, environmental pollution, and even numerous casualties. Therefore, introducing coal mine accident causal analysis techniques can analyze the causes of roof collapse accidents, help identify the root causes and key factors, and provide a basis for developing more scientific and reasonable safety measures, thereby reducing the incidence of coal mine accidents.

[0003] Currently, existing methods for analyzing the causes of coal mine accidents only focus on environmental or geological factors, neglecting equipment failure and human error factors, and failing to consider the correlation between these factors. This results in roof collapse accidents under the influence of multiple factors being unable to be effectively analyzed, thereby reducing the accuracy of coal mine accident cause analysis, further affecting the formulation of accident prevention and safety management measures, and increasing the potential risk of accidents.

[0004] To address this, a knowledge graph-based method for analyzing the causes of coal mine accidents is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a knowledge graph-based method for analyzing the causes of coal mine accidents. This method involves collecting coal mine roof collapse accident reports and cleaning the text to obtain a set of roof collapse accident terms. Next, by extracting the accident causation nodes, accident attributes, and their relationships from this term set, a knowledge graph of roof collapse accident causes is constructed. Then, the similarity of the accident causation nodes is calculated, and nodes are merged based on the similarity. Following this, causation attribute codes and accident attribute codes are obtained, and the causation association index of the accident causation nodes is calculated. Combined with the accident attribute codes, single-cause risk indices and multi-cause risk indices are obtained. Finally, based on single-cause risk and multi-cause risk, a comprehensive risk index is determined, and the accident causation nodes are ranked by their degree of danger. This invention can effectively analyze the risk factors of coal mine roof collapse accidents, thereby reducing the potential risk of roof collapse accidents.

[0006] A knowledge graph-based method for analyzing the causal relationships of coal mine accidents, comprising:

[0007] Collect accident reports of coal mine roof collapse accidents and obtain a set of roof collapse accident terms through text cleaning operations;

[0008] Based on the set of roof accident terms, a roof accident entity extraction model, an entity relationship classification model, and an entity attribute extraction model are constructed, and the models are trained using the set of roof accident terms. The trained models are then used to extract entities, entity relationships, and attributes from the set of roof accident terms to obtain accident cause types, accident cause nodes, roof accident nodes, cause attributes, accident attributes, and accident cause association edges.

[0009] By combining the accident cause nodes, the roof accident nodes, the cause attributes, the accident attributes, and the accident cause association edges, a roof accident cause knowledge graph is constructed.

[0010] Calculate the text similarity of the cause of the accident cause node and the attribute similarity of the corresponding cause attribute to obtain the accident cause similarity; combine the accident attribute of the accident cause node to perform a merging operation on the accident cause nodes of the roof accident cause knowledge graph;

[0011] Obtain the causal attribute code and accident attribute code, and calculate the causal association index of different accident causal nodes; combine the accident attribute code and the roof accident risk assessment function to calculate the single-cause risk index; obtain the multi-cause risk index of the accident causal node, combine it with the single-cause risk index to obtain the comprehensive risk index, and rank the accident causal nodes by degree of danger.

[0012] Preferably, the specific implementation process for obtaining the accident cause node, the roof accident node, the cause attribute, the accident attribute, and the accident cause association edge includes:

[0013] Labeling is performed on several sets of roof accident terms to obtain roof accident entity training set, entity relationship training set, and entity attribute training set; wherein, the labels of the roof accident entity training set include accident cause node labels and roof accident node labels; the labels of the entity relationship training set are accident cause association edge labels; the labels of the entity attribute training set include cause attribute labels and accident attribute labels.

[0014] The roof accident entity training set is input into the BERT pre-trained model to train the coal mine roof accident entities, thereby obtaining the roof accident entity extraction model; the entity relationship classification model is constructed based on the BERT pre-trained model and the classifier, and the coal mine roof entity relationship is trained according to the entity relationship training set; the entity attribute extraction model is obtained by training the coal mine roof accident entity attributes on the BERT pre-trained model through the entity attribute training set.

[0015] The entities in the roof accident term set are extracted using the roof accident entity extraction model to determine the accident cause node and the roof accident node; the relationship between different accident cause nodes and / or roof accident nodes is determined according to the entity relationship classification model to obtain the accident cause association edge; and the cause attribute of the accident cause node and the accident attribute of the roof accident node are obtained by combining the entity attribute extraction model.

[0016] Preferably, the accident causation types include geological factors, human error factors, equipment failure factors, and working environment factors; the accident causation nodes of geological factors include geological structure nodes, rock strata nodes, formation nodes, and surface water nodes; the accident causation nodes of human error factors include mining method nodes, ventilation measure nodes, support method nodes, and overload operation nodes; the accident causation nodes of equipment failure factors include support equipment nodes, ventilation equipment nodes, mechanical equipment nodes, and electrical equipment nodes; the accident causation nodes of working environment factors include mine type nodes, mine depth nodes, mine temperature nodes, mine gas concentration nodes, and mine humidity nodes; the accident attributes include location, time, accident scope, casualties, and economic losses.

[0017] Preferably, the similarity between the causal text and the attribute similarity is calculated, and the accident causal nodes in the knowledge graph of the roof accident causation are merged; the specific implementation process includes:

[0018] Based on two different accident causation nodes and Obtain each and Number of characters in the node name and pass Calculate the accident causation node and The causal text similarity between the two Where min() represents the minimum value function; max() represents the maximum value function; Used for calculation and The number of identical characters in the node name;

[0019] Obtain the two accident causation nodes respectively. and The corresponding set of causal attributes {A1,…,A m} and {B1,…,B n For each of the two different sets of causative attributes, a one-to-one match is performed to obtain k sets of causative attribute matching pairs; for any causative attribute matching pair... Calculate the attribute value similarity between the two causal attributes; combine the attribute value similarity of k sets of causal attribute matching pairs to obtain the attribute similarity. The specific calculation formula is as follows:

[0020]

[0021] Where m and n represent respectively and The number of the causative attributes; A i and B j For cause attribute matching pair The two causal attributes mentioned in the text, A i express The i-th causal attribute and i∈{1,…,m}, B j express The j-th causal attribute and j∈{1,…,n}; α i Indicates causal attribute matching pairs Influence factors;

[0022] Combining the aforementioned accident causal nodes and The aforementioned text similarity Similarity to the aforementioned attributes pass Calculate the comprehensive similarity of accident causes; where β represents The weights;

[0023] If the overall similarity of the accident causes exceeds a predetermined similarity threshold, the accident cause node will be... and Merge into one node.

[0024] Preferably, the specific implementation process for calculating the causation correlation index of different accident causation nodes is as follows:

[0025] The causative attributes and the accident attributes are encoded to obtain causative attribute codes and accident attribute codes; based on the causative attribute codes and accident attribute codes, the causative correlation index between the roof accident node and different accident causative nodes is calculated, using the following formula:

[0026]

[0027] in, Indicates the roof accident node and the u-th accident causation node The causal correlation index between them; N roof This indicates the total number of coal mine roof collapse accidents. express The total number of coal mine roof collapse accidents; COV() represents the covariance function; R represents... The numerical matrix encoding the accident attributes; C u,l express The numerical matrix encoding the causative attribute of the l-th causative attribute; σ R R represents the standard deviation; Indicate C u,l Standard deviation; γ l M represents the degree of influence of the l-th causative attribute on coal mine roof accidents; M represents The total number of the causal attributes.

[0028] Preferably, by combining the accident attribute code and the causal correlation index of different accident causal nodes, the single-cause risk index of the accident causal node causing the coal mine roof accident is quantified through the roof accident risk assessment function, and the specific formula is as follows:

[0029]

[0030] in, Indicates the roof accident node and the u-th accident causation node Causal correlation index between them; RA u express The single-cause risk index; R v express The vth accident attribute; R represents v The aforementioned roof accident risk assessment function; ω v R represents v The influence coefficient on the consequences of coal mine roof collapse accidents; K represents the total number of accident attributes.

[0031] Preferably, the multi-cause risk index for different accident causation nodes is calculated, and combined with the single-cause risk index, the comprehensive risk index for each accident causation node is obtained; the specific calculation formula is as follows:

[0032]

[0033] Among them, CR u Represents the accident causation node described in section u. The comprehensive risk index; ∈0 represents The single-cause risk index RA u Weighting coefficients; express The qth multifactor risk index; ∈ q express The corresponding weighting coefficient; Q represents The total number of the multifactor risk indices;

[0034] The accident causative nodes are ranked according to their degree of danger based on the comprehensive risk index of the different accident causative nodes.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0036] 1. The knowledge graph for roof accident causation proposed in this invention, by combining entity extraction, entity relationship classification, and entity attribute extraction models, can effectively extract accident causation nodes, roof accident nodes, causation attributes, accident attributes, and accident causation association edges from the roof accident term set, thereby improving the efficiency of accident data processing and analysis. Through BERT pre-training models trained for different training tasks, accurate entity recognition, relationship extraction, and attribute recognition are achieved, effectively improving data processing efficiency. Furthermore, constructing accident causal relationships using a knowledge graph approach helps to analyze the root causes of accidents holistically, promoting the improvement of accident prevention measures and decision support.

[0037] 2. This invention achieves automatic merging of accident cause nodes by using textual and attribute similarity to effectively improve the construction efficiency of the knowledge graph for roof accident causes. This method not only considers the character similarity of node names but also deeply compares the attribute value similarity of cause attributes and uses a comprehensive evaluation based on influence factors to ensure the accuracy and rationality of the merging operation. By merging highly similar accident cause nodes, the knowledge graph structure is effectively simplified, redundant information is reduced, and the accuracy and utilization efficiency of the knowledge graph are improved, thereby enhancing the accuracy of coal mine accident cause analysis.

[0038] 3. This invention, through the quantitative calculation of causative correlation indices, single-cause risk indices, and multi-cause risk indices of accident causative nodes, can accurately assess the impact of various causative nodes on coal mine roof accidents and rank the hazard levels of accident causative nodes based on a comprehensive risk index. This method combines causative attribute coding and accident attribute coding to calculate causative correlation indices, comprehensively reflecting the relationship between each accident causative node and coal mine roof accidents, effectively improving the accuracy of coal mine accident risk prediction. Simultaneously, the comprehensive risk index can quantify the risk of accident causative nodes from multiple dimensions, ensuring that roof accidents under the influence of multiple factors can be effectively analyzed, thereby providing a scientific basis for coal mine safety production, reducing the risk of coal mine roof accidents, and improving the level of coal mine safety management. Attached Figure Description

[0039] Figure 1A flowchart of a knowledge graph-based causal analysis method for coal mine accidents provided in an embodiment of this invention application;

[0040] Figure 2 A schematic diagram of the nodes of the knowledge graph of roof accident causes provided in an embodiment of the present invention;

[0041] Figure 3 A flowchart for comprehensive risk index assessment provided in an embodiment of this invention application. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Coal mine roof falls are typically closely related to factors such as geological conditions, mining methods, support techniques, and roof management. As mining depth increases, geological conditions become more complex, leading to a gradual increase in the frequency and severity of roof falls. This not only causes miner injuries and fatalities but also results in production stoppages and significant economic losses. Previous methods for analyzing the causes of coal mine accidents focused solely on environmental or geological factors, neglecting the impact of equipment failures and human error, as well as the interrelationships between these factors. Consequently, they were unable to effectively analyze the risk factors contributing to roof falls.

[0044] This invention proposes a knowledge graph-based method for analyzing the causes of coal mine roof collapse accidents to identify high-risk factors. To illustrate the effectiveness of this method, detailed explanations will be provided in conjunction with the accompanying drawings and the following two embodiments.

[0045] Example 1

[0046] This application discloses a knowledge graph-based method for analyzing the causal factors of coal mine accidents, effectively analyzing high-risk factors that can lead to roof collapse accidents in coal mines. (See also...) Figure 1This is a flowchart of a knowledge graph-based method for analyzing the causes of coal mine accidents. The specific implementation steps include: S1, collecting accident reports of coal mine roof collapse accidents and obtaining a set of roof collapse accident terms through text cleaning; S2, constructing a knowledge graph of roof collapse accident causes based on the roof collapse accident term set; S3, calculating the similarity of causal texts and attributes to determine the accident cause similarity and merging accident cause nodes; S4, obtaining causal attribute codes and accident attribute codes, and calculating the causal association index of different accident cause nodes; S5, combining accident attribute codes and the roof collapse accident risk assessment function to calculate a single-cause risk index; S6, obtaining the multi-cause risk index of accident cause nodes, combining the single-cause risk index to obtain a comprehensive risk index, and ranking the accident cause nodes by their degree of danger.

[0047] Furthermore, accident reports of coal mine roof collapse accidents are collected, and a set of words related to roof collapse accidents is obtained through text cleaning operations, corresponding to step S1 above; wherein, the text cleaning operations include removing stop words, removing noise information, normalizing synonyms, part-of-speech tagging, and word segmentation.

[0048] Specifically, the system removes stop words, such as "of," "and," and "at," which are irrelevant to the analysis. It also removes noise information, such as tables, images, and special symbols, from the accident report. Synonym normalization converts similar expressions in the same accident report into a unified format, such as unifying "roof collapse," "roof collapse," and "roof instability" as "roof collapse." Finally, it performs part-of-speech tagging and word segmentation to ensure the extraction of correct nouns, verbs, and other useful keywords.

[0049] Furthermore, based on the set of terms related to roof accidents, a knowledge graph of the causes of roof accidents is constructed; corresponding to step S2 above, the specific implementation process includes:

[0050] Labeling is performed on several sets of roof accident terms to obtain roof accident entity training set, entity relationship training set, and entity attribute training set; among them, the labels of the roof accident entity training set include accident cause node labels and roof accident node labels; the labels of the entity relationship training set are accident cause association edge labels; the labels of the entity attribute training set include cause attribute labels and accident attribute labels.

[0051] The roof accident entity training set is input into the BERT pre-trained model to train the coal mine roof accident entity and obtain the roof accident entity extraction model; an entity relationship classification model is built based on the BERT pre-trained model and classifier, and coal mine roof entity relationship is trained according to the entity relationship training set; the entity attribute training set is used to train the coal mine roof accident entity attributes of the BERT pre-trained model to obtain the entity attribute extraction model.

[0052] Entities in the roof accident term set are extracted using a roof accident entity extraction model to identify accident causal nodes and roof accident nodes. The relationships between different accident causal nodes and / or roof accident nodes are determined using an entity relationship classification model to obtain accident causal association edges. The causal attributes of accident causal nodes and accident attributes of roof accident nodes are obtained by combining an entity attribute extraction model. Finally, a roof accident causal knowledge graph is constructed by combining accident causal nodes, roof accident nodes, causal attributes, accident attributes, and accident causal association edges.

[0053] Optionally, during the model training process, for the roof accident entity extraction model, the model is trained and optimized using Dice loss; for the entity relationship classification model and the entity attribute extraction model, the model is optimized using the cross-entropy loss function.

[0054] By constructing a knowledge graph of roof accident causes and combining it with a BERT pre-trained model and related training sets for entity extraction, entity relationship classification and entity attribute extraction, accident cause nodes and roof accident nodes can be extracted efficiently and accurately from the roof accident word set, providing a reliable foundation for subsequent cause analysis.

[0055] Furthermore, the types of accident causes include geological factors, human error, equipment failure, and working environment factors; see [reference needed]. Figure 2 This is a schematic diagram of the nodes in the knowledge graph of roof fall accident causes; the accident cause nodes for geological factors include geological structure nodes, rock strata nodes, formation nodes, and surface water nodes; the accident cause nodes for human error factors include mining method nodes, ventilation measure nodes, support method nodes, and overload operation nodes; the accident cause nodes for equipment failure factors include support equipment nodes, ventilation equipment nodes, mechanical equipment nodes, and electrical equipment nodes; the accident cause nodes for working environment factors include mine type nodes, mine depth nodes, mine temperature nodes, mine gas concentration nodes, and mine humidity nodes; the accident attributes include location, time, accident scope, casualties, and economic losses.

[0056] Specifically, the causal attributes of geological structural nodes include geological structural type (such as faults, folds, and dikes), structural distribution, rock mass stability, and rock mass fragmentation; the causal attributes of strata nodes include strata thickness, strata hardness, strata brittleness, rock type, strata distribution, and strata friction coefficient; the causal attributes of stratigraphy nodes include stratigraphic type (such as coal seams, shale layers, and sandstone layers), strata thickness, and stratigraphic spacing; and the causal attributes of surface water nodes include groundwater level, groundwater flow velocity, permeability, and groundwater pressure on the roof. Force; causal attributes of mining method nodes include mining method (e.g., roadway mining and reverse mining), mining sequence, mining depth, and mining speed; causal attributes of ventilation measures nodes include ventilation method (e.g., mechanical ventilation and natural ventilation), air volume, air velocity, and ventilation duct layout; causal attributes of support method nodes include support type (e.g., hydraulic support, wooden support, and steel support), support density, and the rationality of support design; causal attributes of overload operation nodes include operating load, equipment load-bearing capacity, and operating intensity; support equipment The causal attributes of nodes include support equipment type (such as hydraulic supports and anchor bolts), equipment wear, equipment maintenance frequency, and equipment load; the causal attributes of ventilation equipment nodes include ventilation equipment type (such as fans and ducts), ventilation capacity, equipment failure frequency, and equipment operating time; the causal attributes of mechanical equipment nodes include mechanical equipment type (such as mining equipment and transportation equipment), equipment performance, and failure rate; the causal attributes of power equipment nodes include power equipment type (such as generators and power control equipment), equipment failure rate, and power supply stability; the causal attributes of mine type nodes include mine type (such as underground coal mine, metallic mine, and non-metallic mine) and mine structure; the causal attributes of mine depth nodes include mine depth and its impact on formation pressure; the causal attributes of mine temperature nodes include mine temperature and temperature variation range; the causal attributes of mine gas concentration nodes include mine gas concentration (such as methane, carbon dioxide, and hydrogen sulfide) and gas concentration variation; and the causal attributes of mine humidity nodes include mine humidity and humidity fluctuations.

[0057] By refining the types of accident causes and clarifying the classification of various causal nodes, the specific causes of coal mine roof collapse accidents can be analyzed more comprehensively and accurately. Classifying accident causes into geological factors, human error factors, equipment failure factors, and working environment factors, and further subdividing each into multiple specific nodes, helps to reveal the underlying causes of accidents. Furthermore, by comprehensively considering accident attributes (such as location, time, accident level, and accident scope), more operational decision-making basis can be provided, helping to accurately assess the scope of impact and risk level of accidents.

[0058] Further, the similarity of the causal text of the accident causal node and the attribute similarity of the corresponding causal attribute are calculated to obtain the accident causal similarity; combined with the accident attributes of the accident causal node, the accident causal nodes of the roof accident causal knowledge graph are merged; corresponding to the above S3 step, the specific implementation process includes:

[0059] Based on two different accident causation nodes and Obtain each and Number of characters in the node name and pass Calculate the accident causation nodes and Causal text similarity between Where min() represents the minimum value function; max() represents the maximum value function; Used for calculation and The number of identical characters in the node name;

[0060] Obtain the two accident cause nodes respectively and The corresponding set of causal attributes {A1,…,A m} and {B1,…,B n For each of the two different sets of causal attributes, a one-to-one match is performed to obtain k sets of causal attribute matching pairs; for any causal attribute matching pair Calculate the attribute value similarity between two causal attributes; combine the attribute value similarity of k sets of causal attribute matching pairs to obtain the attribute similarity score. The specific calculation formula is as follows:

[0061]

[0062] Where m and n represent respectively and The number of causal attributes; A i and B j For cause attribute matching pair The two causal attributes in A i express The i-th causal attribute and i∈{1,…,m}, B j Let α represent the j-th causal attribute of B, where j∈{1,…,n}; i Indicates causal attribute matching pairs Influence factors;

[0063] Combining the causal nodes of the accident and Causes of text similarity Similarity of attributes pass Calculate the comprehensive similarity of accident causes; where β represents The weights are such that 0 < β < 1;

[0064] If the overall similarity of the accident causes exceeds a predetermined similarity threshold, the accident cause node will be... and Merge into one node.

[0065] To illustrate the effectiveness of the accident cause node merging operation proposed in the embodiments of this application, Table 1 exemplarily lists different accident cause node merging processes; wherein, the predetermined similarity threshold is 0.85.

[0066] Table 1. Merging Process of Different Accident Causation Nodes

[0067] Node 1 Node 2 Text similarity Attribute similarity Overall similarity Preset similarity threshold Merge or not JD01 JD02 0.875 0.8 0.838 0.85 no JD02 JD03 0.857 0.9 0.878 0.85 yes JD03 JD04 1 0.7 0.86 0.85 yes JD04 JD05 0.555 0.6 0.578 0.85 no

[0068] By calculating the text similarity of accident cause nodes and the attribute similarity of cause attributes, the similarity between different accident cause nodes can be accurately assessed, thereby effectively merging similar cause nodes and reducing redundant information. This process ensures the accuracy of node merging through precise text and attribute matching algorithms combined with similarity thresholds. By comprehensively considering the similarity of text and attributes, the correlation between accident cause nodes can be reflected more comprehensively, improving the accuracy and consistency of the knowledge graph and reducing the graph complexity caused by node redundancy.

[0069] Furthermore, the causative attribute codes and accident attribute codes are obtained, and the causative correlation index of different accident causative nodes is calculated, corresponding to... Figure 1 Step S4; see Figure 3 This is a flowchart for the comprehensive risk index assessment, detailing the implementation process:

[0070] The causal attributes and accident attributes are encoded to obtain causal attribute codes and accident attribute codes. Based on the causal attribute codes and accident attribute codes, the causal correlation index between the roof accident node and different accident causal nodes is calculated. The specific calculation formula is as follows:

[0071]

[0072] in, Indicates the roof accident node and the u-th accident causation node The causal correlation index between them; N roof This indicates the total number of coal mine roof collapse accidents. express The total number of coal mine roof collapse accidents; COV() represents the covariance function; R represents... The numerical matrix encoding the accident attributes; C u,l express The numerical matrix encoding the causative attribute of the l-th causative attribute; σ R R represents the standard deviation; Indicate C u,l Standard deviation; γ l M represents the degree of influence of the l-th causative attribute on coal mine roof collapse accidents; M represents... The total number of causal attributes.

[0073] This application's embodiments combine causative attribute coding and accident attribute coding to calculate the causative correlation index of accident causative nodes, enabling a more accurate quantification of the relationship between different causative factors and coal mine roof accidents. By introducing covariance function and standard deviation calculation methods, it is possible not only to accurately identify the degree of influence of each accident causative node, but also to quantify the specific impact of causative attributes on coal mine roof accidents, thereby providing a more precise decision-making basis for coal mine safety management.

[0074] Furthermore, by combining accident attribute coding and roof accident risk assessment functions, a single-cause risk index is calculated; corresponding to... Figure 1 The S5 step, specifically, is implemented as follows:

[0075] Combining accident attribute coding and causal correlation indices of different accident causal nodes, the single-cause risk index of coal mine roof accidents caused by accident causal nodes is quantified through a roof accident risk assessment function. The specific formula is as follows:

[0076]

[0077] in, Indicates the roof accident node and the u-th accident causation node Causal correlation index between them; RA u express The single-cause risk index; R v express The vth accident attribute; R represents v The roof accident risk assessment function, For R v The exponential coefficient; ω v R represents v The influence coefficient on the consequences of coal mine roof collapse accidents; K represents the total number of accident attributes.

[0078] By combining accident attribute coding and roof accident risk assessment functions, we can accurately calculate the single-cause risk index, further quantify the risk of coal mine roof accidents caused by each accident causal node, and reasonably assess the potential impact of different causes on coal mine roof accidents. This can improve the accuracy of accident risk identification and thus reduce the accident incidence rate.

[0079] Furthermore, the multi-cause risk index of the accident causative nodes is obtained, and combined with the single-cause risk index to obtain a comprehensive risk index. The accident causative nodes are then ranked by their degree of danger. Figure 1 Step S6 involves calculating the multi-cause risk index for different accident causal nodes, and combining it with the single-cause risk index to obtain the comprehensive risk index for each accident causal node; the specific calculation formula is as follows:

[0080]

[0081] Among them, CR u Represents the u-th accident causation node The comprehensive risk index; ∈0 indicates The single-cause risk index RA u Weighting coefficients; express The qth multifactor risk index; ∈ q express The corresponding weighting coefficient; Q represents The total number of multifactor risk indices;

[0082]

[0083] in, Indicates inclusion The total number of coal mine roof collapse accidents caused by the qth multi-accident causative node; δ q Indicating the q-th multi-accident causative node Weighting of the impact on coal mine roof collapse accidents;

[0084] Based on the comprehensive risk index of different accident causative nodes, the accident causative nodes are ranked by their degree of danger, as shown in Table 2. Table 2 lists the comprehensive risk index of some accident causative nodes as an example.

[0085] Table 2. Comprehensive Risk Index of Some Accident Causation Nodes

[0086]

[0087] By combining single-cause and multi-cause risk indices, a comprehensive risk index is calculated for each accident causation node, enabling precise quantification of the potential hazards of coal mine roof collapse accidents. Ranking the hazard levels of accident causation nodes helps clarify the priority of various risks. This method not only improves the accuracy and comprehensiveness of accident causation analysis but also effectively identifies the interactions between multiple causative factors, promoting the targetedness and effectiveness of accident prevention measures, thereby reducing safety risks in coal mine production.

[0088] This application's embodiments optimize the construction of accident causal relationships by combining entity extraction, entity relationship classification, and entity attribute extraction models, significantly improving the processing and analysis efficiency of coal mine roof accident data. By automatically merging similar accident causal nodes, redundant information is effectively reduced, the knowledge graph structure is simplified, and the accuracy and utilization efficiency of the graph are improved, thereby enhancing the precision of coal mine accident causal analysis. Simultaneously, by quantitatively calculating the causal association index, single-cause risk index, and multi-cause risk index of accident causal nodes, a precise assessment of coal mine roof accident risk is achieved. This comprehensively reflects the influence degree of various causal nodes and ranks accident causal nodes by risk level based on the comprehensive risk index, providing a scientific basis for coal mine safety management and reducing the risk of roof accidents.

[0089] Example 2

[0090] This application applies a knowledge graph-based coal mine accident causation analysis method to the unmined area of ​​Coal Mine A to predict risk factors in coal mine mining management in the area, thereby preventing coal mine roof collapse accidents in advance.

[0091] Specifically, accident reports of coal mine roof collapse accidents are collected, and a set of words related to roof collapse accidents is obtained through text cleaning operations. The text cleaning operations include removing stop words, removing noise information, normalizing synonyms, tagging parts of speech, and word segmentation.

[0092] Furthermore, based on the set of terms related to roof accidents, a knowledge graph of the causes of roof accidents is constructed; the specific implementation process includes:

[0093] Labeling is performed on several sets of roof accident terms to obtain roof accident entity training set, entity relationship training set, and entity attribute training set; among them, the labels of the roof accident entity training set include accident cause node labels and roof accident node labels; the labels of the entity relationship training set are accident cause association edge labels; the labels of the entity attribute training set include cause attribute labels and accident attribute labels.

[0094] The roof accident entity training set is input into the BERT pre-trained model to train the coal mine roof accident entity and obtain the roof accident entity extraction model; an entity relationship classification model is built based on the BERT pre-trained model and classifier, and coal mine roof entity relationship is trained according to the entity relationship training set; the entity attribute training set is used to train the coal mine roof accident entity attributes of the BERT pre-trained model to obtain the entity attribute extraction model.

[0095] Entities in the roof accident term set are extracted using a roof accident entity extraction model to identify accident causal nodes and roof accident nodes. The relationships between different accident causal nodes and / or roof accident nodes are determined using an entity relationship classification model to obtain accident causal association edges. The causal attributes of accident causal nodes and accident attributes of roof accident nodes are obtained by combining an entity attribute extraction model. Finally, a roof accident causal knowledge graph is constructed by combining accident causal nodes, roof accident nodes, causal attributes, accident attributes, and accident causal association edges.

[0096] Furthermore, the accident causation types include geological factors, human error factors, equipment failure factors, and working environment factors; the accident causation nodes for geological factors include geological structure nodes, rock strata nodes, formation nodes, and surface water nodes; the accident causation nodes for human error factors include mining method nodes, ventilation measure nodes, support method nodes, and overload operation nodes; the accident causation nodes for equipment failure factors include support equipment nodes, ventilation equipment nodes, mechanical equipment nodes, and electrical equipment nodes; the accident causation nodes for working environment factors include mine type nodes, mine depth nodes, mine temperature nodes, mine gas concentration nodes, and mine humidity nodes; the accident attributes include location, time, accident scope, casualties, and economic losses.

[0097] Furthermore, the similarity of the cause text of the accident cause node and the similarity of the corresponding cause attribute are calculated to obtain the accident cause similarity; combined with the accident attributes of the accident cause node, the accident cause nodes of the roof accident cause knowledge graph are merged.

[0098] Furthermore, the causative attribute code and accident attribute code are obtained, and the causative correlation index of different accident causative nodes is calculated; specific implementation process:

[0099] The causal attributes and accident attributes are encoded to obtain causal attribute codes and accident attribute codes. Based on the causal attribute codes and accident attribute codes, the causal correlation index between the roof accident node and different accident causal nodes is calculated. The specific calculation formula is as follows:

[0100]

[0101] in, Indicates the roof accident node and the u-th accident causation node The causal correlation index between them; N roof This indicates the total number of coal mine roof collapse accidents. express The total number of coal mine roof collapse accidents; COV() represents the covariance function; R represents... The numerical matrix encoding the accident attributes; C u,l express The numerical matrix encoding the causative attribute of the l-th causative attribute; σ R R represents the standard deviation; Indicate C u,l Standard deviation; γ l M represents the degree of influence of the l-th causative attribute on coal mine roof collapse accidents; M represents... The total number of causal attributes.

[0102] Furthermore, by combining accident attribute coding and roof accident risk assessment functions, a single-cause risk index is calculated; the specific implementation process is as follows:

[0103] Combining accident attribute coding and causal correlation indices of different accident causal nodes, the single-cause risk index of coal mine roof accidents caused by accident causal nodes is quantified through a roof accident risk assessment function. The specific formula is as follows:

[0104]

[0105] in, Indicates the roof accident node and the u-th accident causation node Causal correlation index between them; RA u express The single-cause risk index; R v express The vth accident attribute; R represents v The roof accident risk assessment function; ω v R represents v The influence coefficient on the consequences of coal mine roof collapse accidents; K represents the total number of accident attributes.

[0106] Furthermore, the multi-factor risk index of the accident causative nodes is obtained, and combined with the single-factor risk index to obtain the comprehensive risk index. The accident causative nodes are then ranked by their degree of danger. Specifically, the multi-factor risk index of different accident causative nodes is calculated, and combined with the single-factor risk index, the comprehensive risk index of each accident causative node is obtained. The specific calculation formula is as follows:

[0107]

[0108] Among them, CRu Represents the u-th accident causation node The comprehensive risk index; ∈0 indicates The single-cause risk index RA u Weighting coefficients; express The qth multifactor risk index; ∈ q express The corresponding weighting coefficient; Q represents The total number of multifactor risk indices;

[0109] Based on the comprehensive risk index of different accident causal nodes, the accident causal nodes are ranked according to their degree of danger.

[0110] Optionally, based on the comprehensive risk index of different accident causal nodes, high-risk factors are identified, such as roof rock structure, unstable mining faces, improper mining methods, and insufficient support measures. These risk factors are then prioritized. Specific preventive measures are developed for high-risk factors, such as strengthening roof stability monitoring, conducting geological exploration, optimizing working face design, using hydraulic supports, support nets, and other support systems, and ensuring the effectiveness of the support equipment. Furthermore, real-time data on roof deformation, stress distribution, and geological changes in the unmined area of ​​Coal Mine A are acquired. Combined with the causal analysis results, a dynamic early warning system is established to promptly identify potential risks and take corresponding measures. A continuous optimization and improvement mechanism is established, collecting accident and monitoring data through a feedback mechanism, analyzing the effectiveness of preventive measures, adjusting plans based on new data, and continuously updating the knowledge graph to ensure the accuracy of causal analysis and optimize the implementation of preventive measures.

[0111] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing the causal relationships of coal mine accidents based on knowledge graphs, characterized in that, include: Collect accident reports of coal mine roof collapse accidents and obtain a set of roof collapse accident terms through text cleaning operations; Based on the set of roof accident terms, a roof accident entity extraction model, an entity relationship classification model, and an entity attribute extraction model are constructed, and the models are trained using the set of roof accident terms. The trained models are then used to extract entities, entity relationships, and attributes from the set of roof accident terms to obtain accident cause types, accident cause nodes, roof accident nodes, cause attributes, accident attributes, and accident cause association edges. By combining the accident cause nodes, the roof accident nodes, the cause attributes, the accident attributes, and the accident cause association edges, a roof accident cause knowledge graph is constructed. Calculate the text similarity of the cause nodes and the attribute similarity of the corresponding cause attributes to obtain the accident cause similarity. Based on the accident attributes of the accident cause nodes, a merging operation is performed on the accident cause nodes of the roof accident cause knowledge graph; Obtain the causal attribute code and accident attribute code, and calculate the causal correlation index of different accident causal nodes; combine the accident attribute code and the roof accident risk assessment function to calculate the single-cause risk index; Obtain the multi-cause risk index of the accident causative node, combine it with the single-cause risk index to obtain a comprehensive risk index, and rank the accident causative nodes according to their degree of danger. The specific implementation process for calculating the causal correlation index of different accident causal nodes is as follows: The causative attributes and the accident attributes are encoded to obtain causative attribute codes and accident attribute codes; based on the causative attribute codes and accident attribute codes, the causative correlation index between the roof accident node and different accident causative nodes is calculated, using the following formula: ； in, Indicates the roof accident node and the The accident-causing nodes mentioned above The causal correlation index between them; This indicates the total number of coal mine roof collapse accidents. express The total number of coal mine roof collapse accidents; Represents the covariance function; express The numerical matrix encoding the accident attributes; express No. A numerical matrix encoding the causative attributes of each of the causative attributes; express Standard deviation; express Standard deviation; Indicates the first The degree of influence of the aforementioned causative attributes on coal mine roof accidents; express The total number of the causative attributes; Combining the accident attribute code and the causal correlation index of different accident causal nodes, the single-cause risk index of the accident causal node causing the coal mine roof accident is quantified through the roof accident risk assessment function. The specific formula is as follows: ； in, Indicates the roof accident node and the The accident-causing nodes mentioned above The causal correlation index between them; express The single-cause risk index; express The The aforementioned accident attributes; express The aforementioned roof accident risk assessment function; express The impact coefficient on the consequences of coal mine roof collapse accidents; This indicates the total number of the aforementioned accident attributes; Calculate the multi-cause risk index for different accident causal nodes, and combine it with the single-cause risk index to obtain the comprehensive risk index for each accident causal node; the specific calculation formula is as follows: ； in, Indicates the first The accident causation node The aforementioned comprehensive risk index; express The single-cause risk index Weighting coefficients; express The The aforementioned multifactor risk index; express The corresponding weighting coefficients; express The total number of the multifactor risk indices; The calculation method is as follows: ； in, Indicates inclusion The The total number of coal mine roof collapse accidents caused by multiple accident-causing nodes; Indicates the first Among the multiple accident-causing nodes Weighting of the impact on coal mine roof collapse accidents; The accident causative nodes are ranked according to their degree of danger based on the comprehensive risk index of the different accident causative nodes.

2. The method for analyzing the causal relationships of coal mine accidents based on knowledge graphs according to claim 1, characterized in that, The specific implementation process for obtaining the accident cause node, the roof accident node, the cause attribute, the accident attribute, and the accident cause association edge includes: Labeling is performed on several sets of roof accident terms to obtain roof accident entity training set, entity relationship training set, and entity attribute training set; wherein, the labels of the roof accident entity training set include accident cause node labels and roof accident node labels; the labels of the entity relationship training set are accident cause association edge labels; the labels of the entity attribute training set include cause attribute labels and accident attribute labels. The roof accident entity training set is input into the BERT pre-trained model to train the coal mine roof accident entities, thereby obtaining the roof accident entity extraction model; the entity relationship classification model is constructed based on the BERT pre-trained model and the classifier, and the coal mine roof entity relationship is trained according to the entity relationship training set; the entity attribute extraction model is obtained by training the coal mine roof accident entity attributes on the BERT pre-trained model through the entity attribute training set. The entities in the roof accident term set are extracted using the roof accident entity extraction model to determine the accident cause node and the roof accident node; the relationship between different accident cause nodes and / or roof accident nodes is determined according to the entity relationship classification model to obtain the accident cause association edge; and the cause attribute of the accident cause node and the accident attribute of the roof accident node are obtained by combining the entity attribute extraction model.

3. The method for analyzing the causal relationships of coal mine accidents based on knowledge graphs according to claim 1, characterized in that, The accident causation types include geological factors, human error factors, equipment failure factors, and working environment factors; the accident causation nodes for geological factors include geological structure nodes, rock strata nodes, formation nodes, and surface water nodes; the accident causation nodes for human error factors include mining method nodes, ventilation measure nodes, support method nodes, and overload operation nodes; the accident causation nodes for equipment failure factors include support equipment nodes, ventilation equipment nodes, mechanical equipment nodes, and electrical equipment nodes; the accident causation nodes for working environment factors include mine type nodes, mine depth nodes, mine temperature nodes, mine gas concentration nodes, and mine humidity nodes; the accident attributes include location, time, accident scope, casualties, and economic losses.

4. The method for analyzing the causal relationships of coal mine accidents based on knowledge graphs according to claim 1, characterized in that, Calculate the text similarity and attribute similarity of the causal factors, and merge the accident causal nodes in the knowledge graph of the roof accident causation; the specific implementation process includes: Based on two different accident causation nodes and Obtain each and Number of characters in the node name and ;pass Calculate the accident causation nodes and The causal text similarity between the two ;in, Describes the minimum value function; Represents the maximum value function; Used for calculation and The number of identical characters in the node name; Obtain the two accident causation nodes respectively. and The corresponding set of causal attributes and A one-to-one matching is performed on the causative attributes within two different sets of causative attributes to obtain... Group causal attribute matching pairs; for any causal attribute matching pair Calculate the similarity of attribute values ​​between the two causative attributes; and synthesize... The attribute similarity is obtained by comparing the attribute values ​​of the attribute matching pairs. The specific calculation formula is as follows: ； in, and They represent and The number of the causative attributes; and For cause attribute matching pair The two causal attributes mentioned above, express The The aforementioned causative attributes and , express The The aforementioned causative attributes and ; Indicates causal attribute matching pairs Influence factors; Combining the aforementioned accident causal nodes and The aforementioned cause text similarity Similarity to the aforementioned attributes ,pass Calculate the comprehensive similarity of accident causes; among which, express The weights; If the overall similarity of the accident causes exceeds a predetermined similarity threshold, the accident cause node will be... and Merge into one node.

Images