Production safety accident key cause link identification method based on complex network

Abstract

The invention relates to the technical field of safety production management and intelligent analysis, in particular to a production safety accident key cause link identification method based on a complex network, which comprises the following steps of: acquiring structured production safety accident report data from a safety production supervision platform by utilizing a web crawler technology; a TF-IDF algorithm is combined with an industrial standard term library to optimize Jieba word segmentation, and a chi-square statistical method is adopted to calculate the correlation between each risk keyword and an accident category. By fusing text mining, statistical analysis and complex network theories and combining field dictionary optimization word segmentation, multi-accident model adaptation and triple centrality index analysis, the problems that a traditional method is single in analysis dimension, depends on subjective experience and is insufficient in technology integration are solved; efficient utilization of unstructured data, scientific layering of risk factors and accurate identification of key cause links are realized, and objectivity and effectiveness of safety management of a complex production system are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of production safety management and intelligent analysis, and in particular to a method for identifying key causal links of production safety accidents based on complex networks. Background Art

[0002] In the field of production safety, the prevention and control of safety accidents have always been a key issue in enterprise management. With the development of industrial intelligence, the complexity of production systems has increased exponentially, and the correlation and dynamic nature of risk factors have placed higher demands on safety management. Existing accident analysis methods, such as Fault Tree Analysis (FTA), Event Tree Analysis (ETA), and Failure Mode and Effects Analysis (FMEA), have the following significant shortcomings:

[0003] Single analysis dimension: It can only handle single-factor or simple multi-factor coupling scenarios, and cannot cope with the interactive impact of multi-dimensional and multi-level risk factors in modern complex production systems.

[0004] Strong subjective dependence: Over-reliance on expert experience, lack of objective quantitative basis for risk factor screening and stratification, resulting in insufficient accuracy in identifying key causes.

[0005] Insufficient technology integration: The advantages of text mining, statistical analysis and network modeling have not been effectively integrated, making it difficult to automatically extract key information from massive accident reports and construct complex causal networks. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for identifying key causal links of production safety accidents based on complex networks, so as to solve the problem of being unable to meet the safety management requirements of complex production systems.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying key causal links of production safety accidents based on complex networks, comprising the following steps:

[0008] S1. Data collection step: using web crawler technology to collect structured production safety accident report data from the production safety supervision platform. The accident report includes fields such as accident time, location, type, casualties, and detailed description;

[0009] S2. Text processing step: Based on the accident report collected in step S1, the TF-IDF algorithm is combined with the industry standard terminology library to optimize the Jieba word segmentation, and the accident detailed description text is segmented and stop words are filtered. Risk keywords are extracted and the word frequency weights are calculated to obtain initial text feature items;

[0010] S3. Risk Factor Analysis: For the initial text feature items obtained in step S2, the chi-square statistical method is used to calculate the correlation between each risk keyword and the accident category. Highly correlated keywords with a significance level of p < 0.05 are retained to reduce the feature dimension. Based on the preset accident model, the screened risk keywords are classified, integrated, and hierarchically processed to form a set of risk factors with causal logical relationships.

[0011] S4. Network construction and analysis: Based on the risk factor set obtained in step S3 and their hierarchical causal relationships, a complex accident network model is constructed with the risk factors as nodes and the causal relationships between factors as edges; the complex network is analyzed using degree centrality, betweenness centrality, and clustering coefficient to identify key nodes, and the h-strength method is used to extract key links;

[0012] S5. Visualization display: Import the key causal nodes and key causal links identified in step S4 into Gephi software, and display the node centrality values and edge weights through a graphical interface to form a visual accident causal network map.

[0013] Preferably, the platform in step S1 includes the accident reporting database of the Ministry of Emergency Management of the People's Republic of China and the enterprise safety production management system, and the collected structured data is stored as an Excel table containing the core fields of the accident, and the unextracted fields are manually supplemented.

[0014] Preferably, the industry standard terminology library in step S2 includes the keyword system in the "Enterprise Employee Injury and Death Accident Classification Standard" (GB6441), and the TF-IDF algorithm retains the top 300 keywords with the word frequency weight as the initial text feature items. If there are less than 300 keywords, all of them are retained.

[0015] Preferably, the accident model in step S3 includes at least one of a 2-4 model, a Swiss cheese model, a STAMP model, a HFACS model and a Bow-Tie model.

[0016] Preferably, in step S3, the adaptation model is selected according to the accident type:

[0017] When using the 2-4 model, risk factors are stratified into direct causes, indirect causes, and root causes;

[0018] When using the Swiss cheese model, risk factors are stratified by the level of the defense system;

[0019] When using the STAMP model, risk factors are stratified according to the system control level;

[0020] When using the HFACS model, risk factors are stratified according to the human factor level;

[0021] When using the Bow-Tie model, risk factors are stratified according to cause chains and consequence chains.

[0022] Preferably, the key link in step S4 is the shortest path connecting multiple high-betweenness centrality nodes, and the shortest path length between nodes is calculated by the h-strength method for screening.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] By integrating web crawlers, text mining, multi-accident model adaptation and complex network analysis technologies, the present invention has constructed a complete technical system for automatically extracting risk factors from unstructured accident reports, scientifically stratifying causal levels, and accurately identifying key nodes and shortest transmission paths. It effectively solves the problems of traditional methods such as single analysis dimension, strong subjective dependence, and insufficient technical integration, and realizes objective quantitative assessment and visual presentation of complex production system risks, significantly improves the accuracy of accident cause identification and the scientific nature of safety management decisions, and provides an efficient, customizable and intelligent solution for accident prevention in high-risk industries, effectively improving the efficiency of production safety accident management and control and reducing the incidence rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a system block diagram of the overall structure of the method for identifying key cause links of production safety accidents based on complex networks of the present invention. DETAILED DESCRIPTION

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0027] See also Figure 1 The present invention provides a technical solution: a method for identifying key cause links of production safety accidents based on complex networks, comprising the following steps:

[0028] S1. Data collection steps: Use Python to develop a web crawler program, configure crawling rules, and collect accident reports from the accident report database of the Ministry of Emergency Management of the People's Republic of China and the enterprise safety production management system for the past 10 years. The reports are stored in a structured Excel format with fields such as accident time, location, type, casualties, and detailed description. The crawling rules mainly include fields such as accident time, location, type, casualties, and detailed description. Fields that are not extracted are extracted manually.

[0029] S2. Text Processing Steps: Based on the accident reports collected in Step S1, first, construct a domain dictionary and load the keywords in the "Classification Standard for Industrial Injury Accidents of Enterprise Employees" (GB6441) (such as "mechanical injury", "violation of operation rules", etc.). Second, use the Jieba word segmentation tool to load the custom dictionary, segment the accident description text, filter out irrelevant words using the stop word list (such as "de", "le"), and finally, use the TF-IDF algorithm to calculate the word frequency weights, and retain the top 300 keywords sorted from largest to smallest weight as the initial feature terms. If the number of keywords does not reach 300, all will be retained. The TF-IDF algorithm formula is as follows:

[0030]

[0031] In the formula: w i is the word frequency weight, f i is the document frequency, N is the total number of documents, extract high-weight risk keywords to obtain the initial text feature terms;

[0032] S3. Risk Factor Analysis: For the initial text feature terms obtained in Step S2, calculate the chi-square value between each keyword in the initial text feature terms and the accident type through SPSS software. The larger the value, the stronger the correlation between variables; retain the high-correlation keywords with a significance level p < 0.05 to reduce the feature dimension. The chi-square value and p-value calculation formulas are as follows:

[0033]

[0034] P(x 2 ≥x) = 1 - P(x 2 <x)

[0035] In the formula: O is the frequency of the associated combination; E is the expected frequency, and x is the calculated chi-square statistic.

[0036] Then classify, integrate, and hierarchically process the obtained accident keywords according to the preset accident models to form a set of risk factors with causal logical relationships. The accident models include at least one of the 2-4 model, Swiss Cheese model, STAMP model, HFACS model, and Bow-Tie model. Comprehensively utilize the advantages of each model to make up for the limitations of a single model. The adaptation rules are shown in the following table.

[0037]

[0038]

[0039] S4. Network construction and analysis: Based on the risk factor set obtained in step S3 and their hierarchical causal relationships, a complex accident network model is constructed with risk factors as nodes and causal relationships between factors as edges. The complex network is analyzed using degree centrality, betweenness centrality, and clustering coefficient to identify key nodes. The key links are extracted using the h-strength method, specifically:

[0040] Taking risk factors as nodes, edges are constructed through the hierarchical structure of the model, and the betweenness centrality of high-frequency nodes is identified through degree centrality; the key hubs of the accident chain are identified through betweenness centrality; and the closeness between the node and the adjacent nodes is expressed through the clustering coefficient.

[0041] Critical link extraction: Use the h-strength method to calculate the shortest path length. The shorter the average path length, the greater the accident risk. Screen the shortest paths connecting more than three high-betweenness nodes (such as "institutional defects → poor supervision → operational errors → explosion").

[0042] The formula for the h-strength method is:

[0043]

[0044] Where: L is the shortest path length, d ij is the number of connecting edges between nodes i and j, and n is the number of nodes;

[0045] S5. Visualization: Import the key causal nodes and key causal links identified in step S4 into the Gephi software, and display the node centrality value (node size) and edge weight (line thickness) through a graphical interface to form a visual accident causal network map.

[0046] The present invention realizes the intelligent identification and visualization of key causal links of accidents by integrating text mining, model selection and complex network analysis. First, the TF-IDF algorithm is combined with domain-specific dictionaries to optimize word segmentation, which greatly improves the recognition accuracy of professional terms for safe production. Second, the present invention is compatible with multiple theoretical frameworks such as the 2-4 model, Swiss cheese model, STAMP model, HFACS model, Bow-Tie model, etc., which can be selected or combined as needed to achieve flexible and comprehensive risk stratification. The accident complex network is constructed through risk stratification, and topological characteristic indicators such as degree centrality and betweenness centrality are comprehensively used, combined with the h-strength method to accurately identify the key nodes and paths of accident propagation. Finally, with the help of the Gephi visualization tool, the analysis results are intuitively presented, providing strong support for safety management decision-making. It can be widely used in safety risk assessment and accident prevention in high-risk industries such as chemical, construction, and transportation, providing an effective tool for achieving safety management. Through the implementation of the present invention, the level of enterprise safety production management can be significantly improved, and the probability of accidents and the degree of losses can be reduced.

[0047] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0048] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may 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 identifying key causal links of production safety accidents based on complex networks, characterized by: The following steps are involved: S1. Data collection step: using web crawler technology to collect structured production safety accident report data from the production safety supervision platform. The accident report includes fields such as accident time, location, type, casualties, and detailed description; S2. Text processing step: Based on the accident report collected in step S1, the TF-IDF algorithm is combined with the industry standard terminology library to optimize the Jieba word segmentation, and the accident detailed description text is segmented and stop words are filtered. Risk keywords are extracted and the word frequency weights are calculated to obtain initial text feature items; S3. Risk Factor Analysis: For the initial text feature items obtained in step S2, the chi-square statistical method is used to calculate the correlation between each risk keyword and the accident category. Highly correlated keywords with a significance level of p < 0.05 are retained to reduce the feature dimension. Based on the preset accident model, the screened risk keywords are classified, integrated, and hierarchically processed to form a set of risk factors with causal logical relationships. S4. Network construction and analysis: Based on the risk factor set obtained in step S3 and their hierarchical causal relationships, a complex accident network model is constructed with the risk factors as nodes and the causal relationships between factors as edges; the complex network is analyzed using degree centrality, betweenness centrality, and clustering coefficient to identify key nodes, and the h-strength method is used to extract key links; S5. Visualization display: Import the key causal nodes and key causal links identified in step S4 into Gephi software, and display the node centrality values and edge weights through a graphical interface to form a visual accident causal network map.

2. The complex network-based method for identifying key causal links of production safety accidents according to claim 1 is characterized by: In step S1, the platform includes the accident report database of the Ministry of Emergency Management of the People's Republic of China and the enterprise safety production management system. The collected structured data is stored as an Excel table containing core fields of the accident, and the unextracted fields are manually supplemented.

3. The complex network-based method for identifying key causal links of production safety accidents according to claim 2 is characterized by: The industry standard terminology library in step S2 includes the keyword system in the "Enterprise Employee Injury and Death Accident Classification Standard" (GB6441), and the TF-IDF algorithm retains the top 300 keywords with the word frequency weight as the initial text feature items. If there are less than 300 keywords, all of them are retained.

4. The complex network-based method for identifying key causal links of production safety accidents according to claim 3 is characterized by: The accident model in step S3 includes at least one of a 2-4 model, a Swiss cheese model, a STAMP model, a HFACS model, and a Bow-Tie model.

5. The complex network-based method for identifying key causal links of production safety accidents according to claim 4 is characterized by: In step S3, the adaptation model is selected according to the accident type: When using the 2-4 model, risk factors are stratified into direct causes, indirect causes, and root causes; When using the Swiss cheese model, risk factors are stratified by the level of the defense system; When using the STAMP model, risk factors are stratified according to the system control level; When using the HFACS model, risk factors are stratified according to the human factor level; When using the Bow-Tie model, risk factors are stratified according to cause chains and consequence chains.

6. The complex network-based method for identifying key causal links of production safety accidents according to claim 5 is characterized by: In step S4, the key link is the shortest path connecting multiple high-betweenness centrality nodes, and the shortest path length between nodes is calculated by the h-strength method for screening.

Images