A method and system for analyzing data on causes of safety accidents in a gas station

Abstract

This invention relates to the field of urban gas safety technology and discloses a method and system for analyzing the causal data of gas plant safety accidents. The method includes: constructing a system of causal factors for gas plant accidents through multi-source data acquisition and screening; obtaining scoring data on the correlation strength between each causal factor, converting the scoring data into ternary interval gray numbers, and then weighted and fused them after standardization, binary interval transformation, and gray-level clarification to generate a standardized and clarified direct influence matrix; normalizing the standardized and clarified direct influence matrix to obtain a normalized influence matrix, solving the comprehensive influence matrix, and calculating the influence degree, affected degree, centrality, and causal degree of each causal factor; calculating a threshold based on the comprehensive influence matrix and performing binarization to generate a reachability matrix; and dividing each causal factor into levels based on the reachability matrix to generate a multi-level hierarchical structure model. This invention can transform complex and discrete causal factors of gas plant accidents into intuitive causal structural relationships.

Description

Technical Field

[0001] This invention relates to the field of urban gas safety technology, and in particular to a method and system for analyzing the causes of safety accidents at gas plants. Background Technology

[0002] As the core hub of the urban gas transmission and distribution system, urban gas stations integrate process units such as pressure regulation, storage and distribution, filtration, and odorization. The media within these stations are flammable and explosive, and pressure conditions fluctuate frequently, resulting in a deep coupling of risks across multiple dimensions, including equipment, personnel, environment, and management. In recent years, urban gas leaks, deflagrations, and poisoning accidents have occurred frequently in China. Accident reviews show that single hidden dangers rarely directly induce accidents; most are caused by multi-level, cascading, and coupled factors. Traditional safety management focuses only on surface-level, visible hazards such as equipment damage and personnel violations, making it difficult to distinguish root-cause management deficiencies, intermediate transmission risks, and direct accident triggers. It fails to trace the complete risk transmission chain, and accident tracing and tiered prevention and control lack quantitative support.

[0003] To achieve systematic identification of the causes of accidents at gas plants, the industry has gradually introduced analytical tools such as fault trees, analytic hierarchy process, DEMATEL, and ISM to conduct risk modeling. However, existing analytical methods and corresponding patented technologies still have multiple defects, such as poor indicator compatibility, large subjective errors, inability to decompose causal links in layers, and inability to quantify the interaction of causes, making it difficult to adapt to the complex and multi-coupled risk scenarios of urban gas plants.

[0004] The invention patent (publication number CN116415847A) discloses a fuzzy comprehensive evaluation method for urban gas gate station pipeline risk based on AHP. This scheme only constructs a static weight system for pipeline failure risk at the plant and calculates the risk value of each indicator based on the analytic hierarchy process. However, this method assumes that all causes are independent of each other, ignores the bidirectional constraints and mutual transmission coupling relationships between personnel, equipment, environment, and management factors, and can only output a single risk level. It cannot identify the causal transmission path between causes, nor can it distinguish between root causes and superficial causes. It is only suitable for static risk assessment of a single point in the pipeline and cannot complete the causal tracing of the entire station's systemic accidents.

[0005] The invention patent (publication number CN117236477A) proposes a risk detection method for natural gas stations, which relies on hierarchical structure indicators and risk index models to classify the risk level of the station. This technology only realizes the quantitative scoring of risk values, but does not quantify the intensity of the mutual influence between various causes; at the same time, it lacks hierarchical decomposition logic and cannot construct a hierarchical causal structure model. It can only complete the risk level determination, but cannot clarify the multi-layer transmission mechanism of accidents.

[0006] The invention patent (publication number CN119885581A) discloses the STAMP-VSM analysis system for third-party construction accidents of gas pipelines. It only identifies control defects in a single accident scenario involving third-party damage. The causal system has a narrow coverage and does not cover the unique risks of gas plant process corrosion, overpressure, station layout, operation and maintenance management, personnel capabilities, etc. It lacks versatility and cannot be adapted to the full-dimensional accident analysis of complete plants.

[0007] The invention patent (publication number CN119150126A) is a risk level assessment scheme for gas stations. It directly constructs an evaluation matrix using conventional expert scoring, without incorporating grey theory to address the ambiguity of expert perception. Gas stations suffer from a large amount of management issues and human-related causes lack clear quantitative standards, leading to significant expert subjective bias. The original, clear numerical scores amplify evaluation errors, resulting in low reliability of the final risk ranking and cause correlation results, and lacking a noise reduction and optimization process.

[0008] For causal analysis techniques that couple DEMATEL and ISM, existing general safety assessment patents are mostly geared towards industries such as subways, building construction, and coal mines, and have not customized a complete causal system for the process characteristics of urban gas plants. Existing basic DEMATEL-ISM analysis methods directly use raw expert scores to construct matrices, failing to eliminate subjective scoring biases caused by experts' experience and professional differences; simultaneously, thresholds often use fixed constants to filter reachable matrices, easily missing weakly correlated but crucial root causes, resulting in distorted stratified results. Existing, limited gray-scale improved DEMATEL techniques are mostly applied to coal mines, foundation pits, and chemical laboratories, without establishing a dedicated causal factor indicator system adapted to urban gas plants. The indicator system is incomplete, and direct application may lead to risk omissions; furthermore, the lack of a differentiated expert weighting gray-scale fusion process, which equally weights evaluation data from experts with different qualifications, further amplifies evaluation bias.

[0009] Another invention patent related to fault tree (publication number CN115829336A) relies on fault logic to list underlying events. It can only qualitatively sort out the logic of the accident occurrence, but cannot quantitatively calculate the comprehensive importance of each cause. It does not distinguish between the cause's influence, the affected, and the attribute of the cause. It can only review the branches of the accident occurrence after the fact, and cannot identify the risk transmission chain in advance, making it difficult to support the hierarchical management of source hazards.

[0010] In summary, existing technologies for risk assessment and accident cause identification at gas plants have four major shortcomings: First, expert scoring lacks a noise reduction mechanism, resulting in significant subjective errors and insufficient reliability of quantitative results; second, the causative indicator system does not align with the complete process scenarios of urban gas plants, leading to incomplete risk coverage; third, it can only statically calculate risk weights, failing to quantify the interactions between causative factors or differentiate accident triggers at different levels; and fourth, it cannot generate multi-layered hierarchical causal structure models, resulting in an ambiguous accident risk transmission chain and hindering precise, tiered, and graded management. Currently, the industry lacks a complete accident cause identification solution that is adapted to urban gas plants, incorporates gray theory for noise reduction, quantifies causal coupling relationships, and outputs multi-layered hierarchical causal models. Summary of the Invention

[0011] To address the aforementioned technical problems, this invention provides a method for analyzing the causal data of safety accidents in gas plants, comprising the following steps: S1: Construct a system of causal factors for gas plant accidents through multi-source data collection and screening; S2: Obtain the score data of the correlation strength between each causative factor in the accident causative factor system, convert the score data into a ternary interval gray number, perform standardization processing on the ternary interval gray number, perform binary interval conversion and gray level clarification processing to obtain a clear direct influence matrix, and perform weighted fusion on the clear direct influence matrix to generate a standard clarification direct influence matrix. S3: Normalize the standardized direct influence matrix to obtain a normalized influence matrix. Solve the comprehensive influence matrix based on the normalized influence matrix. Calculate the influence degree, affected degree, centrality, and causal degree of each causal factor based on the comprehensive influence matrix. S4: Calculate the threshold based on the comprehensive influence matrix, and binarize the sum of the comprehensive influence matrix and the identity matrix based on the threshold to generate a reachability matrix; S5: Divide the levels of each causal factor according to the reachability matrix to generate a multi-level hierarchical structure model.

[0012] Furthermore, in S1, the formula for expressing the accident causative factor system is as follows: In the formula, Let n represent the causative factors of gas plant accidents, and n be the number of causative factors.

[0013] Furthermore, in step S2, a matrix A of direct influencing factors is obtained based on the scoring data. k ,in ,element Let be the scoring data, representing the causal factors of the accident as determined by the k-th expert. Causes of the accident The strength of the association.

[0014] Furthermore, in S2, the standardization of the gray values ​​in the ternary intervals followed by binary interval conversion and grayscale enhancement specifically includes the following steps: using the formula , Binary range conversion and grayscale enhancement are performed, among which... , These represent the lower and upper bounds of the standardized ternary interval gray number after conversion to a binary interval. This represents the standardized grayscale value after grayscale sharpening processing. , , These represent the upper bound, centroid, and lower bound of the gray number in the ternary interval, respectively. , , These represent the upper bound, centroid, and lower bound of the gray number in the ternary interval after standardization.

[0015] Furthermore, in S2, the clear direct influence matrix is ​​C. k ,in To directly affect matrix C k elements, .

[0016] Furthermore, in S2, the standardization direct impact matrix is ​​C, where... Standardization directly affects the elements of matrix C. ; , This represents the expert weights after defuzzification.

[0017] Further, in S3, the normalized influence matrix is ​​Z, where ; x is a normalization factor; the comprehensive influence matrix is ​​E, where I is the identity matrix.

[0018] Furthermore, in step S3, the specific process of calculating the influence, affectedness, centrality, and causation degree of each causal factor based on the comprehensive influence matrix includes: calculating the influence degree using the DEMATEL algorithm. And the degree of influence , , The elements in the comprehensive influence matrix E; based on the influence degree And the degree of influence Calculate the centrality of the causative factors and causal degree , .

[0019] Further, in step S4, the specific process of calculating a threshold based on the comprehensive influence matrix, and binarizing the sum of the comprehensive influence matrix and the identity matrix based on the threshold to generate the reachability matrix includes: calculating a threshold φ based on the mean and standard deviation of the statistical distribution. In the formula To comprehensively influence the mean of all elements in the matrix, The standard deviation of all elements in the comprehensive influence matrix is ​​calculated; the comprehensive influence matrix E is added to the identity matrix I to obtain the overall influence matrix T; based on the value of the threshold φ, the overall influence matrix T is simplified by binarization to obtain the reachability matrix K, calculated as follows: T = E + I; .

[0020] This invention also provides a data analysis system for the causes of safety accidents at gas plants, comprising: The causative factor system construction module is used to construct a causative factor system for gas plant accidents through multi-source data collection and screening. The matrix generation module is used to obtain the score data of the correlation strength between each causal factor in the accident causal factor system, convert the score data into a ternary interval gray number, perform standardization processing on the ternary interval gray number, perform binary interval conversion and gray level clarification processing to obtain a clear direct influence matrix, and perform weighted fusion on the clear direct influence matrix to generate a standard clarified direct influence matrix. The DEMATEL analysis module is used to normalize the standardized clear direct influence matrix to obtain a normalized influence matrix, solve the comprehensive influence matrix based on the normalized influence matrix, and calculate the influence degree, affected degree, centrality degree, and causal degree of each causal factor based on the comprehensive influence matrix. The ISM analysis module is used to calculate a threshold based on the comprehensive influence matrix, perform binarization processing on the sum of the comprehensive influence matrix and the identity matrix based on the threshold to generate an reachability matrix, and divide the hierarchy of each causative factor based on the reachability matrix. The model output module is used to generate a multi-level hierarchical structure model based on the divided levels and output the accident cause path analysis results.

[0021] The embodiments of the present invention have the following technical effects: (1) In the process of identifying the causes of safety accidents at urban gas stations, this invention standardizes and clarifies the expert interval scoring information based on grey theory, effectively weakens the evaluation bias and randomness problems caused by traditional subjective scoring, greatly reduces the interference of human subjective evaluation on the cause analysis results, makes the data processing of the causes of gas station accidents more rigorous and objective, ensures the scientificity and accuracy of the subsequent cause analysis results, and provides a reliable data foundation for the identification of the causes of gas station accidents.

[0022] (2) This invention integrates the DEMATEL model and the ISM model to conduct coupled analysis, deeply explore the correlation, constraint relationship and transmission mechanism among the various accident-causing factors of urban gas stations, and accurately distinguish key causes, transmission causes and result factors by quantitatively calculating the influence, affected degree, centrality and causal degree of each risk factor. It breaks through the limitation of traditional gas station risk analysis that can only qualitatively list factors and cannot quantify the correlation strength, and realizes the qualitative and quantitative integrated analysis of the causes of gas station accidents.

[0023] (3) Based on the quantitative calculation results, this invention constructs a hierarchical structure model of the causes of gas plant accidents, sorts out the complex, discrete and intertwined causes of plant safety accidents into a hierarchical causal structure with clear hierarchy and logical clarity, effectively simplifies the complex risk system with multiple factors coupled, and intuitively presents the hierarchical distribution, correlation path and transmission chain of each cause factor, solving the problems of messy factors, ambiguous causal relationship and difficulty in clarifying the risk transmission law in traditional analysis methods.

[0024] (4) This invention can accurately identify the essential causes and key transmission paths of safety accidents at urban gas plants, distinguish between core risk sources and secondary influencing factors from the source, accurately locate the weak links and key nodes in the safety management of plants, change the traditional gas safety management situation of being all-encompassing but not highlighting the key points, and provide clear theoretical basis and technical support for the prediction of gas plant accidents, the management of hidden dangers, and the prevention of risks.

[0025] (5) This invention realizes the systematic, structural and hierarchical analysis of the causes of gas plant accidents through the Grey-DEMATEL-ISM coupling method. It can stably, efficiently and accurately complete the risk identification and mechanism analysis of the plant, effectively guide the daily hidden danger investigation, risk classification and control, accident prevention and emergency management of urban gas plants, significantly improve the scientific nature and pertinence of gas plant safety management, and ensure the long-term safe, stable and reliable operation of gas plants. Attached Figure Description

[0026] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0027] Figure 1 This is a flowchart of the data analysis method for the causes of safety accidents at gas plants provided in this embodiment of the invention; Figure 2 This is a multi-level hierarchical explanation structure model diagram of the causes of safety accidents in gas plants provided in this embodiment of the invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are part of this invention. Specific Implementation Example 1 To address the aforementioned technical problems, this invention provides a method for analyzing the causal data of safety accidents in gas plants, comprising the following steps: S1: Construct a system of causal factors for gas plant accidents through multi-source data collection and screening; S2: Obtain the score data of the correlation strength between each causative factor in the accident causative factor system, convert the score data into a ternary interval gray number, perform standardization processing on the ternary interval gray number, perform binary interval conversion and gray level clarification processing to obtain a clear direct influence matrix, and perform weighted fusion on the clear direct influence matrix to generate a standard clarification direct influence matrix. S3: Normalize the standardized direct influence matrix to obtain a normalized influence matrix. Solve the comprehensive influence matrix based on the normalized influence matrix. Calculate the influence degree, affected degree, centrality, and causal degree of each causal factor based on the comprehensive influence matrix. S4: Calculate the threshold based on the comprehensive influence matrix, and binarize the sum of the comprehensive influence matrix and the identity matrix based on the threshold to generate a reachability matrix; S5: Divide the levels of each causal factor according to the reachability matrix to generate a multi-level hierarchical structure model.

[0030] Furthermore, in S1, the formula for expressing the accident causative factor system is as follows: In the formula, Let n represent the causative factors of gas plant accidents, and n be the number of causative factors.

[0031] Furthermore, in step S2, a matrix A of direct influencing factors is obtained based on the scoring data. k ,in ,element Let be the scoring data, representing the causal factors of the accident as determined by the k-th expert. Causes of the accident The strength of the association.

[0032] Furthermore, in S2, the standardization of the gray values ​​in the ternary intervals followed by binary interval conversion and grayscale enhancement specifically includes the following steps: using the formula , Binary range conversion and grayscale enhancement are performed, among which... , These represent the lower and upper bounds of the standardized ternary interval gray number after conversion to a binary interval. This represents the standardized grayscale value after grayscale sharpening processing. , , These represent the upper bound, centroid, and lower bound of the gray number in the ternary interval, respectively. , , These represent the upper bound, centroid, and lower bound of the gray number in the ternary interval after standardization.

[0033] Furthermore, in S2, the clear direct influence matrix is ​​C. k ,in To directly affect matrix C k elements, .

[0034] Furthermore, in S2, the standardization direct impact matrix is ​​C, where... Standardization directly affects the elements of matrix C. ; , This represents the expert weights after defuzzification.

[0035] Further, in S3, the normalized influence matrix is ​​Z, where ; x is a normalization factor; the comprehensive influence matrix is ​​E, where I is the identity matrix.

[0036] Furthermore, in step S3, the specific process of calculating the influence, affectedness, centrality, and causation degree of each causal factor based on the comprehensive influence matrix includes: calculating the influence degree using the DEMATEL algorithm. And the degree of influence , , The elements in the comprehensive influence matrix E; based on the influence degree And the degree of influence Calculate the centrality of the causative factors and causal degree , .

[0037] Further, in step S4, the specific process of calculating a threshold based on the comprehensive influence matrix, and binarizing the sum of the comprehensive influence matrix and the identity matrix based on the threshold to generate the reachability matrix includes: calculating a threshold φ based on the mean and standard deviation of the statistical distribution. In the formula To comprehensively influence the mean of all elements in the matrix, The standard deviation of all elements in the comprehensive influence matrix is ​​calculated; the comprehensive influence matrix E is added to the identity matrix I to obtain the overall influence matrix T; based on the value of the threshold φ, the overall influence matrix T is simplified by binarization to obtain the reachability matrix K, calculated as follows: T = E + I; .

[0038] This invention also provides a data analysis system for the causes of safety accidents at gas plants, comprising: The causative factor system construction module is used to construct a causative factor system for gas plant accidents through multi-source data collection and screening. The matrix generation module is used to obtain the score data of the correlation strength between each causal factor in the accident causal factor system, convert the score data into a ternary interval gray number, perform standardization processing on the ternary interval gray number, perform binary interval conversion and gray level clarification processing to obtain a clear direct influence matrix, and perform weighted fusion on the clear direct influence matrix to generate a standard clarified direct influence matrix. The DEMATEL analysis module is used to normalize the standardized clear direct influence matrix to obtain a normalized influence matrix, solve the comprehensive influence matrix based on the normalized influence matrix, and calculate the influence degree, affected degree, centrality degree, and causal degree of each causal factor based on the comprehensive influence matrix. The ISM analysis module is used to calculate a threshold based on the comprehensive influence matrix, perform binarization processing on the sum of the comprehensive influence matrix and the identity matrix based on the threshold to generate an reachability matrix, and divide the hierarchy of each causative factor based on the reachability matrix. The model output module is used to generate a multi-level hierarchical structure model based on the divided levels and output the accident cause path analysis results. Specific Implementation Example 2 This invention, through on-site surveys and investigations of urban gas plants in multiple locations, reveals that the safety production system of urban gas plants covers a complete range of operational aspects, including equipment operation, process operation, environmental conditions, personnel management, operation and maintenance control, and emergency response. Each main aspect includes multiple subdivided work processes. For example, the equipment operation aspect includes subdivided work such as pipeline pressure monitoring, valve control, pressure regulation and stabilization, equipment corrosion prevention, and leak detection. These work processes are interconnected and coupled, resulting in a complex overall system structure and a high degree of intertwined risks. Each operational aspect contains numerous potential causal factors that could lead to safety accidents at gas plants.

[0040] In this embodiment, a preliminary list of causal factors for safety accidents at urban gas plants is first compiled through literature review, accident report compilation, on-site inspection of gas plants, and collection of historical data. Senior industry experts in the field of gas safety are invited to screen, revise, and validate the causal indicators, verifying the scientific validity and applicability of the indicator system. Based on this, a comprehensive causal indicator system for urban gas plant safety accidents is constructed. Furthermore, this invention introduces grey theory to mitigate the bias of expert subjective evaluation, integrates DEMATEL and ISM coupled analysis methods, quantitatively analyzes the correlation strength and causal relationship of each accident causal factor, and constructs a multi-layered hierarchical explanation structure model for the causes of urban gas plant accidents. This clearly outlines the risk transmission path and hierarchical structure of plant accidents. Finally, the practicality and effectiveness of this identification method are verified by combining actual gas plant operating conditions, providing a scientific basis and technical support for accurate identification, risk prevention and control, and safety management of urban gas plant accidents.

[0041] like Figure 1 This invention provides a method for identifying the causes of accidents at urban gas plants, which is described in detail in the following steps.

[0042] Step S1: Construct a system of causal factors for plant accidents.

[0043] The process of establishing the causal factor system for plant accidents in this invention is as follows: Figure 2 As shown. Identifying the causal factors of safety accidents at urban gas plants is a systematic, complex, and labor-intensive fundamental task. In carrying out the identification of accident causes and indicator screening at urban gas plants, this invention strictly adheres to the following basic principles: (1) Comprehensiveness: It fully covers all aspects of risk in urban gas plants, including equipment, environment, personnel, management, processes, and operation and maintenance. It systematically sorts out all potential causal indicators that may induce gas plant safety accidents, ensuring that the identification of causal factors is complete and without omission.

[0044] (2) Independence: Each accident cause indicator is independent of the others and has clear boundaries. Each factor can independently characterize the safety risk characteristics of a single dimension, avoiding the problem of high overlap and repetitive characterization between indicators.

[0045] (3) Representativeness: On the premise of ensuring the integrity and comprehensiveness of the causal system, the initial complex causal factors are clustered and integrated and the dimensions are simplified. Redundant, invalid and weakly correlated indicators are eliminated, and highly representative indicators are retained. The indicator system structure is optimized as much as possible while ensuring the accuracy of the analysis.

[0046] (4) Objectivity. The selection of indicators closely matches the operating process, work scenarios, equipment characteristics and current safety management status of urban gas plants, and fully aligns with the industry reality and scenario specificity of gas plant safety production.

[0047] (5) Consistency. All causal indicators are uniformly constructed around the core objectives of identifying the causes of safety accidents in gas plants, analyzing risk transmission, and optimizing safety management and control. The indicators have unified connotations, consistent logic, and do not conflict with each other.

[0048] This invention combines the safety production process, accident characteristics, and risk evolution patterns of urban gas plant stations to specifically identify accident-causing factors in each operational stage and module, from planning and design to construction and operation and maintenance. Based on this, a safety accident causation index system for urban gas plant stations is established. First, through extensive industry literature, safety regulations, on-site investigations, accident cases, and standard provisions, an initial list of causes of gas plant safety accidents is preliminarily identified, summarized, and compiled. Combining expert opinions and merging or summarizing semantically similar factors, a total of 28 accident causation factor indicators affecting the safe production of urban gas plant stations are finally selected, as shown in the table below: Step S2: Construct the direct influence matrix.

[0049] Based on the aforementioned system of causal factors for plant accidents, the pairwise interactions between risk factors are described. This invention invited 10 experts or senior practitioners in the urban gas sector to conduct a questionnaire survey. Experts assigned scores to the degree of influence between factors using a five-point scale. Semantic variables were defined as follows: 0 indicates no influence, 1 indicates minimal influence, 2 indicates minor influence, 3 indicates significant influence, and 4 indicates extremely high influence. This led to the initial matrix A of directly influencing factors. k Elements in matrix Ak This indicates the k-th expert's assessment of the causal factors of the accident. Causes of the accident The direct influence strength, when i=j, is assigned a value of 0, and the initial direct influence factor matrix A k The formula is as follows: (1) For the initial direct influence matrix A k Fuzzy processing is performed to reduce the interference of expert subjectivity during the indicator assignment process, thereby improving the objectivity and accuracy of expert evaluation results. This involves fuzzifying the accident causative factors. Causes of the accident degree of influence The results are converted into ternary interval gray numbers to quantify the fuzziness, uncertainty, and cognitive biases in expert evaluations, achieving complete preservation and accurate representation of uncertain evaluation information. Simultaneously, considering the differences in professional experience, on-site cognition, and expertise among experts, to avoid evaluation distortion caused by equal weighting, differentiated weights are assigned to each expert, and these expert weights are converted into corresponding binary interval gray numbers. This achieves a fuzzy quantitative expression of expert weight information, further weakening human scoring bias and improving the stability and scientific rigor of the multi-expert evaluation fusion results. The specific semantic variables and their corresponding interval gray numbers are shown in the table below: For the initial direct influence matrix A k Defuzzification processing is performed. This invention addresses the incompleteness of causal evaluation information and the ambiguity of expert judgment boundaries in gas plant accidents. It eliminates differences in evaluation dimensions and numerical range deviations by standardizing the upper, centroid, and lower bounds of the ternary interval gray number. This achieves unified normalization of fuzzy evaluation information from multiple experts and multiple indicators, significantly reducing the interference of extreme value biases in human scoring on the overall evaluation system. The standardized calculation formula is as follows: (2) (3) (4) In the formula , , These represent the upper bound, centroid, and lower bound of expert k's score on the causative factors, respectively, after being converted into interval gray numbers. , , These represent the upper bound, centroid, and lower bound of the gray number in the interval after standardization transformation, respectively, so as to uniformly map all causal evaluation data to the same standard interval and ensure that the multi-dimensional gas plant accident causal factor evaluation data are comparable and consistent.

[0050] The standardized ternary interval gray numbers are then subjected to binary interval conversion and gray-scale clarification. Since the fuzzy information regarding the causes of gas plant accidents often exhibits interval uncertainty, binary interval conversion simplifies the complex ternary gray number structure, preserving core risk uncertainty information while reducing computational redundancy. Furthermore, the gray number clarification formula transforms the interval fuzzy quantities into deterministic evaluation values, achieving precise quantification of fuzzy risk information. The specific calculation formula is as follows: (5) (6) (7) In the formula , Let represent the lower and upper bounds of the binary interval transformation of the standardized interval gray number, respectively. This represents a standardized grayscale value. This step effectively addresses the industry pain points of "incomplete information, ambiguous judgments, and large differences in experience" in the risk assessment of gas plants, transforming vague qualitative judgments by experts into precise quantitative values ​​and suppressing subjective errors from a data perspective.

[0051] Perform inverse calculation of the scoring to generate the expert k-clear direct influence matrix C. k : (8) In the formula To directly affect matrix C k The elements represent the factors used in calculating the expert k-score. Factors The standard for the degree of impact is clearly defined, accurately restoring the objective evaluation results of experts on the causal relationships of gas plant accidents.

[0052] By introducing differentiated expert weights to achieve multi-source evaluation fusion, a final standard-clarifying direct impact matrix C is constructed. Experts in different gas fields exhibit significant differences in years of experience, field experience, and research directions; equal weighting can easily lead to biased and distorted evaluation results. This invention simultaneously performs fuzzification and defuzzification optimization on the gray numbers of expert weights, achieving scientific fusion of multi-expert evaluation information through weighted summation. This weakens individual subjective biases, preserves the consensus patterns of the group, and significantly improves the stability and credibility of causal evaluation results for gas plants. The calculation formula is as follows: (9) (10) In the formula The elements that directly affect the standardization and clarification of matrix C represent the factors in the calculation of all expert scores. Factors Clarification of the standard for the degree of impact; This represents the expert weights after defuzzification.

[0053] Step S3: Construct the comprehensive influence matrix E.

[0054] First, the clear direct impact matrix C, after incorporating the weights of multiple experts, is standardized. The causative factors of accidents at urban gas plants are complex and vary greatly in strength. The original direct impact matrix exhibits uneven distribution of numerical magnitudes across rows and columns, leading to extreme value biases. Directly performing matrix iterations can easily mask weak causative factors and overemphasize strong ones, resulting in distorted subsequent correlation analyses. Therefore, this invention divides the standardized clear direct impact matrix C by the maximum sum of all rows in the matrix. Global normalization compresses the numerical range, eliminating computational interference caused by differences in the evaluation magnitudes of different causative factors. This yields a standardized matrix suitable for causal analysis at gas plants, effectively ensuring the fairness and accuracy of multi-dimensional risk factor calculations. The normalization coefficients and the formulas for calculating the normalized impact matrix are as follows: (11) (12) In the formula, x is the normalization factor; n is the number of influencing factors, i.e., n is 28; Z is the normalized influence matrix.

[0055] The comprehensive risk impact matrix E is calculated based on the normalized impact matrix Z, using the following formula: (13) In the formula, I is the identity matrix.

[0056] Step S4: Calculate the influence, affectedness, centrality, and causal degree of the accident-causing factors for each gas plant.

[0057] The influence degree is calculated using the DEMATEL algorithm. And the degree of influence Influence degree represents the overall impact of a factor on all other factors, reflecting its ability to actively influence the system. Influence level represents the overall influence of all other factors on the factor, reflecting its passive response characteristics within the system. In the comprehensive influence matrix T, summing the elements in the i-th row yields the influence degree of the corresponding element, and summing the elements in the j-th column yields the influence level of the corresponding element in that column. The calculation formula is as follows: (14) In the formula For the elements in the comprehensive influence matrix E, This represents the influence degree of the corresponding element in E. This represents the degree to which the corresponding element in E is affected.

[0058] Based on the influence calculated above And the degree of influence Calculate the centrality of the causative factors and causal degree Centrality represents the overall importance of a factor in a system; a higher value indicates a stronger correlation. Causality is used to distinguish factor attributes; when... A centrality value greater than 0 indicates that factor Mi is a causal factor, meaning it has a significant impact on other factors in the system; conversely, a value greater than 0 indicates that it is an outcome factor, meaning it is susceptible to the influence of other factors. (Centrality) and causal degree The calculation formula is as follows: (15) This invention calculates the influence of the index system of various accident causative factors. Degree of influence Centrality and causal degree As shown in the table below: Step S5: Construct the reachability matrix to complete the data transformation from the DEMATEL model to the ISM model.

[0059] The comprehensive influence matrix E obtained by solving the Grey-DEMATEL model is a continuous numerical matrix, which can only quantify the correlation strength between the causal factors of each accident at the urban gas plant. If the ISM model is used to generate a clear hierarchical causal structure of the factor system, it is necessary to complete the conversion process from the quantitative results of the DEMATEL model to the ISM modeling basis.

[0060] First, a threshold φ is constructed. To remove redundant information from the comprehensive influence matrix E and reduce the subjectivity of expert judgment, the threshold φ is calculated based on the mean and standard deviation of the statistical distribution. The calculation formula is as follows: (16) In the formula To comprehensively influence the mean of all elements in the matrix, The standard deviation of all elements in the comprehensive influence matrix is ​​used. This adaptive threshold adapts to the risk distribution pattern of gas plant stations, avoids the loss of core causal links due to fixed thresholds, and provides an objective boundary standard for subsequent binarization conversion of the ISM model.

[0061] Adding the comprehensive influence matrix E to the identity matrix I yields the overall influence matrix T. Based on the threshold φ, the overall influence matrix T is simplified through binarization to obtain the reachability matrix K, calculated using the following formula: T = E + I (17) (18) The results of calculating the reachability matrix K in this invention are shown below: Step S6: Divide the accident causative factors into levels and construct a multi-level hierarchical directed structure model of the causes of safety accidents at urban gas plants.

[0062] Influencing factors can be determined based on the reachability matrix K. reachable set and antecedent set and intersection The reachable set represents the set of all causative factors in urban gas plants that can directly or indirectly influence cause i. The extraction rule is to select the set of influencing factors whose elements in the i-th row of the reachable matrix K are all equal to 1. The antecedent set represents the set of all preceding causative factors that can directly or indirectly induce cause i. The extraction rule is to select the set of influencing factors whose elements in the i-th column of the reachable matrix K are all equal to 1. and antecedent set The intersection is The causative factors that have a bidirectional coupling feedback relationship with causative factor i are represented by the following formula: (19) (20) (twenty one) If reachable set and antecedent set satisfy Then the collection can be reached. The corresponding factors can all be found in the preceding set. The process involves identifying the antecedent, indicating that all influencing factors of this cause have prior triggering sources and no underlying root cause effect that propagates downwards. This is considered a surface-level accident triggering factor that only exposes risks and does not generate new hidden dangers. This factor is listed as a top-level factor, and then the row and column corresponding to this factor are removed from the reachability matrix K. This step is repeated until all rows and columns corresponding to all causal factors in the reachability matrix are removed, and all causes are hierarchically categorized.

[0063] Finally, following the order in which the causative factors are iteratively eliminated—that is, the first eliminated factors are the top-level factors, and the last eliminated are the bottom-level root causes—the causative factors at each level are arranged from top to bottom. Based on the binary association logic within the reachability matrix, directed transmission lines between levels are drawn, ultimately yielding a multi-level hierarchical directed structural model of the causes of safety accidents at urban gas plants, as shown below. Figure 2 .

[0064] Through analysis Figure 2The multi-layered hierarchical explanation model of the causes of safety accidents at urban gas plants reveals that the causes affecting the safe production of urban gas plants can be divided into deep-seated root causes, shallow-level transmission factors, intermediate-level transitional factors, surface-level inducing factors, and direct accident causes. These factors are interconnected, forming a complex multi-layered hierarchical structure that reflects the impact mechanism of various risk factors on the safe production of urban gas plants and the transmission relationship between the causes.

[0065] Based on the aforementioned multi-layered hierarchical structure, the deep-seated root causes include inadequate implementation of responsibilities (A22), ineffective safety supervision (A23), and low investment in safety assurance (A28). These are the most basic and critical factors for ensuring the safe operation of urban gas plants. Above these deep-seated root causes, the superficial transmission factors are all risk factors derived from the plant's internal management system, including shortcomings at the management level such as imperfect operating procedures (A24), chaotic technical management (A25), inadequate emergency management (A26), and imperfect safety inspection system (A27). These factors stem from shortcomings in lower-level supervision and investment. Defects directly contribute to gas plant safety risks, having a wider impact and longer duration. Therefore, in actual urban gas plant hazard investigation and accident tracing, these factors are often the core targets for long-term management. Addressing these factors can effectively mitigate the evolution of various risks at higher levels, thereby achieving the goal of preventing gas plant accidents at their source. Management defects, after being transmitted upwards, form intermediate transitional factors, encompassing deficiencies in professional training (A2), lack of professional background (A3), lack of multi-skilled personnel (A4), lack of public auxiliary facilities (A12), harsh climate (A18), and geological conditions. Intermediate risks related to personnel, environment, and process layout, such as complexity (A19), insufficient distance to sensitive areas (A20), unreasonable process design (A15), and unreasonable station layout (A21), are at this level and serve as a bridge between higher and lower levels, continuously amplifying management loopholes at the lower levels. These are key areas for routine plant maintenance and inspection. Mid-level transitional factors accumulate and further form surface-level triggering factors, including operator physical and mental health (A5), poor initial equipment quality (A7), equipment aging (A8), inadequate monitoring (A9), incomplete protective measures (A10), vent pipe location (A13), and sewage pipes. Location (A14) and non-standard equipment maintenance (A16) are mostly visible hidden dangers in equipment and personnel on site. They are the easiest risks to be found in the current safety inspection of gas plants, but they are not fundamental solutions. The top layer is the direct accident inducing factors, including human operation errors (A1), corrosion and thinning of process facilities (A6), overpressure operation (A11), and poor reliability of protective equipment (A17). These direct accident inducing factors are transmitted and superimposed from the lower layers of risks, and there is a fixed risk transmission path. Once they occur, they will directly induce gas safety accidents such as leaks, explosions, and poisoning.

[0066] Therefore, when conducting causal identification and risk management for safety accidents at urban gas plants, more attention should be paid to optimizing the lower-level factors in the structure, namely, deep-seated root causes and shallow transmission factors. On the one hand, these factors have a more comprehensive and integrated impact on the overall safety risk system of urban gas plants. Controlling these risk factors will inhibit the entire risk chain of the plant, preventing the escalation of hidden dangers from the source. On the other hand, compared with the direct accident inducing factors at the upper level, the lower-level root causes and management factors can directly serve as the core anchor objects for tracing the source of accidents at urban gas plants. This helps to simplify the process of hazard investigation and accident tracing, save manpower, material resources, and financial resources, and improve the effectiveness and efficiency of safety management. In addition, by dividing risk levels and sorting out transmission paths according to the constructed multi-level hierarchical model of urban gas plant accident causes, targeted hierarchical hazard management solutions can be matched, which can reduce the probability of gas safety accidents from the root. This is of great significance for building a comprehensive and long-term urban gas plant safety risk identification and prevention system.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing the causal data of safety accidents in gas plants, characterized in that, Includes the following steps: S1: Construct a system of causal factors for gas plant accidents through multi-source data collection and screening; S2: Obtain the score data of the correlation strength between each causative factor in the accident causative factor system, convert the score data into a ternary interval gray number, perform standardization processing on the ternary interval gray number, perform binary interval conversion and gray level clarification processing to obtain a clear direct influence matrix, and perform weighted fusion on the clear direct influence matrix to generate a standard clarification direct influence matrix. S3: Normalize the standardized direct influence matrix to obtain a normalized influence matrix, obtain a comprehensive influence matrix based on the normalized influence matrix, and calculate the influence degree, affected degree, centrality degree, and causal degree of each causal factor based on the comprehensive influence matrix; S4: Calculate the threshold based on the comprehensive influence matrix, and perform binarization on the sum of the comprehensive influence matrix and the identity matrix based on the threshold to generate a reachability matrix; S5: Divide the levels of each causal factor according to the reachability matrix to generate a multi-level hierarchical structure model.

2. The method for analyzing the causes of safety accidents at gas plants according to claim 1, characterized in that, In S1, the formula for expressing the accident causative factor system is as follows: In the formula, Let n represent the causative factors of gas plant accidents, and n be the number of causative factors.

3. The method for analyzing the causal data of safety accidents in gas plants according to claim 2, characterized in that, In step S2, a matrix A of direct influencing factors is obtained based on the scoring data. k ,in , Let be the scoring data, representing the causal factors of the accident as determined by the k-th expert. Causes of the accident The strength of the association.

4. The method for analyzing the causal data of safety accidents in gas plants according to claim 3, characterized in that, In step S2, the standardization of the ternary interval gray number followed by binary interval conversion and grayscale enhancement specifically includes the following steps: using the formula , Binary range conversion and grayscale enhancement are performed, among which... , These represent the lower and upper bounds of the standardized ternary interval gray number after conversion to a binary interval. This represents the standardized grayscale value after grayscale sharpening processing. , , Let these represent the upper bound, centroid, and lower bound of the gray number in the ternary interval, respectively. , , These represent the upper bound, centroid, and lower bound of the gray number in the ternary interval after standardization.

5. The method for analyzing the causal data of safety accidents in gas plants according to claim 4, characterized in that, In S2, the clear direct influence matrix is ​​C. k ,in To directly affect matrix C k elements, .

6. The method for analyzing the causal data of safety accidents in gas plants according to claim 5, characterized in that, In S2, the standardization direct impact matrix is ​​C, where Standardization directly affects the elements of matrix C. ; , This represents the expert weights after defuzzification.

7. The method for analyzing the causal data of safety accidents in gas plants according to claim 6, characterized in that, In S3, the normalized influence matrix is ​​Z, where ; x is a normalization factor; the comprehensive influence matrix is ​​E, where I is the identity matrix.

8. The method for analyzing the causal data of safety accidents in gas plants according to claim 7, characterized in that, In step S3, the specific process of calculating the influence, affectedness, centrality, and causation of each causal factor based on the comprehensive influence matrix includes: calculating the influence using the DEMATEL algorithm. And the degree of influence , , The elements in the comprehensive influence matrix E; based on the influence degree And the degree of influence Calculate the centrality of the causal factors and causal degree , .

9. The method for analyzing the causal data of safety accidents in gas plants according to claim 8, characterized in that, In step S4, the threshold is calculated based on the comprehensive influence matrix, and the sum of the comprehensive influence matrix and the identity matrix is ​​binarized based on the threshold to generate the reachability matrix. The specific process includes: calculating the threshold φ based on the mean and standard deviation of the statistical distribution. In the formula To comprehensively influence the mean of all elements in the matrix, The standard deviation of all elements in the comprehensive influence matrix is ​​calculated; the comprehensive influence matrix E is added to the identity matrix I to obtain the overall influence matrix T; based on the value of the threshold φ, the overall influence matrix T is simplified by binarization to obtain the reachability matrix K, calculated as follows: T = E + I; .

10. A data analysis system for the causes of safety accidents at gas plants, characterized in that, include: The causative factor system construction module is used to construct a causative factor system for gas plant accidents through multi-source data collection and screening. The matrix generation module is used to obtain the score data of the correlation strength between each causal factor in the accident causal factor system, convert the score data into a ternary interval gray number, perform standardization processing on the ternary interval gray number, perform binary interval conversion and gray level clarification processing to obtain a clear direct influence matrix, and perform weighted fusion on the clear direct influence matrix to generate a standard clarified direct influence matrix. The DEMATEL analysis module is used to normalize the standardized clear direct influence matrix to obtain a normalized influence matrix, solve the comprehensive influence matrix based on the normalized influence matrix, and calculate the influence degree, affected degree, centrality degree, and causal degree of each causal factor based on the comprehensive influence matrix. The ISM analysis module is used to calculate a threshold based on the comprehensive influence matrix, perform binarization processing on the sum of the comprehensive influence matrix and the identity matrix based on the threshold to generate an reachability matrix, and divide the hierarchy of each causative factor based on the reachability matrix. The model output module is used to generate a multi-level hierarchical structure model based on the divided levels and output the accident cause path analysis results.

Images