Railway safety production regulation compliance evaluation method and system based on large model

Abstract

The application discloses a railway safety production regulation compliance evaluation method and system based on a large model, relates to the technical field of railway safety production, and comprises the following steps: acquiring current version railway safety production supervision requirement file data and performing deep structured analysis processing to extract railway safety production supervision requirement features; performing supervision requirement feature comparison processing on the railway safety production supervision requirement features of the current version and a previous version based on a large model, and determining content change types and marking change contents; performing influence degree evaluation processing on enterprise business modules based on the content change types and the marking change contents; performing supervision requirement feature comparison processing on enterprise internal railway operation rules and regulations file data based on the influence degree evaluation processing results, and generating a revision suggestion scheme based on the supervision requirement feature comparison processing results; and the problems of low artificial evaluation efficiency are solved, and safety production regulation compliance adjustment is quickly completed.

Description

Technical Field

[0001] This invention relates to the field of railway safety production technology, and in particular to a method and system for assessing compliance with railway safety production regulations based on a large model. Background Technology

[0002] Railway safety is directly related to train operation safety and the safety of personnel. Relevant regulations and regulatory requirements are constantly updated based on industry development and accident experience. Railway transportation and maintenance companies need to adjust their internal operating rules and regulations in a timely manner to ensure compliance. However, current railway safety compliance assessments largely rely on manual processes. This not only requires significant manpower to analyze the differences between various versions of regulations but also to meticulously check internal regulations for non-compliance, resulting in low efficiency, omissions, and difficulty in timely compliance adjustments. Furthermore, manual assessments are susceptible to subjective bias from assessors, leading to inconsistent results and an inability to accurately identify internal regulations requiring priority revision. While large-scale modeling and natural language processing technologies have developed, there is currently no mature solution specifically designed for railway safety regulations that leverages large-scale modeling capabilities to automate compliance assessments and generate revision suggestions, failing to meet the needs of railway companies for efficient compliance management. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention provides a method and system for railway safety production regulation compliance assessment based on a large model. This system automates the identification of version changes in railway safety production regulatory requirements, quantifies the impact of these changes on different business modules, accurately matches internal company rules and regulations to identify non-compliant content, and automatically generates directly referable policy revision suggestions. This significantly reduces the manual cost of compliance assessment, improves assessment efficiency and accuracy, and solves the problems of low efficiency, easy omissions, and poor consistency in results associated with manual assessment. It also meets the needs of railway companies to quickly complete compliance adjustments for safety production regulations.

[0004] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0005] The first aspect of this application provides a method for assessing compliance with railway safety regulations based on a large model, including the following steps: S101. Obtain the current version of the railway safety production supervision requirements document data and perform deep structured parsing to extract the characteristics of the railway safety production supervision requirements. S102. Based on the large model, compare the current version of railway safety production supervision requirements with the previous version of railway safety production supervision requirements, and determine the content change type and mark the changed content. S103. Assess the impact of content change types and marked changes on enterprise business modules; S104. Based on the impact assessment results, the data of the enterprise's internal railway operation rules and regulations documents are compared and processed according to the regulatory requirements characteristics, and a revision suggestion is generated based on the comparison and processing results of the regulatory requirements characteristics.

[0006] Furthermore, the current version of the railway safety production supervision requirements document data is obtained and subjected to deep structured analysis to extract the characteristics of the railway safety production supervision requirements, including the following steps: The current version of the railway safety production supervision requirements document is split according to the text format, and non-text content is removed. Non-text content includes headers, footers, table of contents, and citations, while retaining the main text information of the clauses. The retained clause text information is structurally broken down according to chapters and clause levels to obtain independent control clause items; Feature extraction is performed on each independent control clause to identify the railway safety production supervision requirements characteristics corresponding to each clause. The railway safety production supervision requirements characteristics include production scenario, constraint object, control requirements, penalty standards, effective time, and related clauses.

[0007] Furthermore, the retained clause text information is structurally broken down according to chapters and clause levels to obtain independent control clause items, including the following steps: Based on the preset chapter level labeling rules, the title number of the clause text is pattern matched to locate the starting identifier of different levels; The control clauses are divided into independent control clause items according to the hierarchical logic of chapter-section-clause-sub-clause, ensuring that each control clause item corresponds to only a single safety production control requirement, and avoiding the mixing of multiple control rules that may affect the accuracy of subsequent feature extraction.

[0008] Furthermore, feature extraction is performed on each independent control clause to identify the railway safety production supervision requirements characteristics corresponding to each clause. These characteristics include production scenario, constraint object, control requirements, penalty standards, effective date, and related clauses. The steps include: Keywords containing terms related to location, work phase, or environmental conditions in the clauses are identified as production scenario characteristics. Keywords containing entity, position / personnel, or generic objects in the clause text are identified as constraint object characteristics. Keywords containing the categories of obligation, prohibition, authorization, and procedure in the clauses are identified as characteristics of control requirements. Keywords containing terms related to administrative penalties, monetary ranges, or criminal liability in the clauses are identified as characteristics of penalty standards. Keywords containing the terms "effective date of regulations", "transition period of specific clauses", or "repeal and replacement" in the clause text are identified as effective date characteristics. Keywords containing explicit references, scope references, or exception clauses in the clause text are identified as related clause characteristics.

[0009] Furthermore, the current version of railway safety production supervision requirements features are compared with the previous version of railway safety production supervision requirements features, and the type of content change and the marked changes are determined, including the following steps: Based on the structured independent control clauses, a mapping relationship between the current version and the previous version is established according to the clause hierarchy number. Clauses without matching mappings are initially identified and marked as newly added control clauses or deleted and invalidated clauses. For clauses with matching mappings, compare the differences in the content of each regulatory requirement feature field one by one. If the production scenario, constraint object, penalty standard or control requirement is modified, it is marked as a control requirement modification; if the effective time field is changed, it is marked as an effective time adjustment; if the number or content of the related clause is changed, it is marked as a related clause change, thus completing the identification and marking output of all changes.

[0010] Furthermore, the construction of the large model includes the following steps: A structured clause feature vector library is constructed, which transforms each extracted railway safety production supervision requirement feature into a standardized and comparable feature vector, and stores all control clause items of different versions in vectors of the same dimension; Based on the cosine similarity of feature vectors, the feature difference degree of corresponding clauses at the same level in different versions is calculated. When the difference degree exceeds the set difference threshold, it is determined that the clause has content changes. Then, the specific change type is identified one by one for the changed feature fields, and the model-driven automated comparison and identification is completed.

[0011] Furthermore, the impact assessment of enterprise business modules based on content change type and marked content changes includes the following steps: For each changed content marked, a value is assigned according to a set weight coefficient; New control clauses and modifications to control requirements are set as the first-level weight coefficient; adjustments to the effective time and changes to related clauses are set as the second-level weight coefficient; deleted or invalidated clauses are set as the third-level weight coefficient. The first-level weight coefficient is greater than the second-level weight coefficient, and the second-level weight coefficient is greater than the third-level weight coefficient. Based on the changed content, the associated production scenarios are extracted, and the weight of the changed content is included in the enterprise business module that matches the corresponding production scenario. The impact score for each business module of the enterprise is calculated using the following expression: ; in, Let i be the score of the i-th enterprise business module. For the i-th enterprise business module, the number of changes in the t-th type of content. Set a weight coefficient for the content change type t. Let t be the safety coefficient of the i-th business module, i be the business module index (line construction, EMU maintenance, dangerous goods transportation, passenger service, special equipment operation and maintenance), t be the content change type index (new control clauses, deleted invalid clauses, control requirement modification, effective time adjustment, related clause changes), and n be the total number of content unchanged types. The impact level of the corresponding business module is determined based on the impact score. The impact level includes high impact level, medium impact level, and low impact level.

[0012] Furthermore, based on the impact assessment results, the data of the enterprise's internal railway operation rules and regulations documents are compared with regulatory requirements features, and a revision proposal is generated based on the results of the regulatory requirements feature comparison. This includes the following steps: Select business modules with a medium or higher level of impact, and extract all tagged regulatory requirement features from the selected business modules; Retrieve existing railway operation rules and regulations documents for the corresponding business modules within the enterprise, perform structured processing on the existing rules and regulations documents according to the same rules as in step S101, and extract the operation control requirements features from the existing rules and regulations. The extracted existing operational control requirements features are compared one by one with the current version of railway safety production supervision requirements features to identify the content in the existing rules and regulations that does not meet the latest supervision requirements, as well as the missing control content that needs to be supplemented. Based on the identified discrepancies and the original wording of regulatory requirements, the system automatically generates revision suggestions for the corresponding rules and regulations, including suggestions for deletion, modification, and addition of clauses. After integrating all revision suggestions, a complete revision proposal is generated.

[0013] Furthermore, the extracted existing operational control requirements features are compared one by one with the current version of railway safety production supervision requirements features to identify content in existing rules and regulations that does not comply with the latest supervision requirements, as well as missing control content that needs to be supplemented, including the following steps: For each existing regulation and control clause, match it one by one with the regulatory requirements characteristics of the current version in the corresponding production scenario; If the regulatory requirements corresponding to existing clauses have been marked as deleted or void, then non-compliant content is identified as needing to be deleted. If the current version has new regulatory requirements, but the existing rules and regulations do not have corresponding control clauses, then it is identified as missing control content that needs to be supplemented. If the objects of constraint and control requirements of the existing control provisions are inconsistent with the characteristics of the current version of regulatory requirements, they are identified as non-compliant content that needs to be modified. If the effective date or related terms of the existing terms are inconsistent with the current version, they are identified as content that needs to be updated synchronously, and all differences are identified and sorted out.

[0014] Furthermore, based on the identified discrepancies and the original wording of regulatory requirements, revision suggestions for corresponding rules and regulations are automatically generated. These suggestions include deletion, modification, and addition of new clauses. Integrating all revision suggestions to generate a complete revision proposal involves the following steps: For the identified regulatory clauses that have been marked for deletion or invalidation, a deletion suggestion for the corresponding clause is directly generated, clearly indicating the clause number to be deleted and the original content, explaining that the basis for deletion is that the corresponding regulatory requirements have been officially invalidated; For the identified missing new regulatory requirements, the original structural features of the regulatory requirements are extracted, and corresponding suggestions for new clauses are generated by combining them with the expression style of the company's existing systems. The corresponding chapter position where the new clauses should be inserted is clearly marked, and the complete content of the control requirements is listed. For the identified content that needs to be modified, compare the differences between the existing clauses and regulatory requirements, generate modification suggestions with the differences marked, retain the existing clauses that comply with regulatory requirements, and only provide the complete modified statements for the differences. For effective dates and related clauses that need to be updated synchronously, update suggestions for the corresponding fields are generated directly, and the accurate updated content is marked.

[0015] The second aspect of this application provides a railway safety production regulation compliance assessment system based on a large model, including: The first data processing unit is used to perform deep structured parsing on the current version of the railway safety production supervision requirements document data in order to extract the characteristics of the railway safety production supervision requirements. The second data processing unit is used to compare the current version of railway safety production supervision requirements with the previous version of railway safety production supervision requirements based on the large model, and to determine the type of content change and mark the changed content. The third data processing unit is used to assess the impact on enterprise business modules based on the type of content change and the marked changes in content. The fourth data processing unit is used to perform regulatory requirement feature comparison processing on the data of internal railway operation rules and regulations documents of the enterprise based on the impact degree assessment processing results, and generate revision suggestions based on the regulatory requirement feature comparison processing results.

[0016] The beneficial effects of this application are as follows: It automates the entire process of compliance assessment for railway safety production regulations. Relying on the natural language processing capabilities of large-scale models, it completes the comparison of version differences of regulatory requirements, the assessment of the degree of impact, and the generation of revision suggestions. Compared with traditional manual assessment methods, it significantly shortens the compliance assessment cycle, reduces the workload of manual sorting and investigation, and effectively avoids the omissions in manual assessment. Through a quantitative impact assessment model, it eliminates the assessment bias caused by the subjective experience of assessors, ensures the consistency of compliance assessment results in different scenarios, and can accurately identify internal system clauses that need to be revised first. This helps railway enterprises quickly complete the internal system compliance adjustments after the updates of regulatory requirements, meets the actual needs of railway enterprises for efficient compliance management, and fills the application gap of large-scale model technology in the field of railway safety production compliance assessment. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram illustrating the steps of the railway safety production regulation compliance assessment method based on a large model, as proposed in this invention. Detailed Implementation

[0019] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0020] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] Example 1 The railway safety production regulation compliance assessment method based on a large model includes the following steps: S101. Obtain the current version of the railway safety production supervision requirements document data and perform deep structured parsing to extract the characteristics of the railway safety production supervision requirements. This process involves acquiring data on the current version of railway safety production supervision requirements documents. Previous versions include regulatory requirements documents of legal and regulatory types and industry standard types. Through deep structural analysis of the unstructured data, the characteristics of railway safety production supervision requirements are extracted. These characteristics include production scenarios, constrained objects, control requirements, penalty standards, effective dates, and related clauses. Production scenarios include railway line construction operations, EMU maintenance, dangerous goods transportation, passenger boarding and alighting organization, and special equipment operation and maintenance. Constrained objects include railway production and operation units, front-line workers, responsible personnel, and safety management personnel. Control requirements include prohibitive requirements, mandatory requirements, and standard requirements. Penalty standards include the level of punishment for units and the severity of punishment for responsible personnel. The effective date is the date the corresponding regulations or standards officially come into effect. Related clauses are other clauses in the same regulations that are related to the current control requirements. This process organizes and categorizes fragmented, unstructured regulatory texts, providing a clear structured data foundation for subsequent compliance assessments.

[0022] The process of obtaining the current version of the railway safety production supervision requirements document data and performing deep structured analysis to extract the characteristics of the railway safety production supervision requirements includes the following steps: The current version of the railway safety production supervision requirements document is split according to the text format, and non-text content is removed. Non-text content includes headers, footers, table of contents, and citations, while retaining the main text information of the clauses. The retained clause text information is structurally broken down according to chapters and clause levels to obtain independent control clause items; Feature extraction is performed on each independent control clause to identify the railway safety production supervision requirements characteristics corresponding to each clause. The railway safety production supervision requirements characteristics include production scenario, constraint object, control requirements, penalty standards, effective time, and related clauses.

[0023] The retained clause text information is structurally broken down according to chapters and clause levels to obtain independent control clause items, including the following steps: Based on the preset chapter level labeling rules, the title number of the clause text is pattern matched to locate the starting identifier of different levels; The control clauses are divided into independent control clause items according to the hierarchical logic of chapter-section-clause-sub-clause, ensuring that each control clause item corresponds to only a single safety production control requirement, and avoiding the mixing of multiple control rules that may affect the accuracy of subsequent feature extraction.

[0024] Feature extraction is performed on each independent control clause to identify the railway safety production supervision requirements characteristics corresponding to each clause. The railway safety production supervision requirements characteristics include production scenario, constraint object, control requirements, penalty standards, effective time, and related clauses. The steps include: Keywords in the clauses containing categories such as location, work phase, and environmental conditions are identified as production scenario features. Among them, keywords related to location include sections of railway lines, stations, tunnels, and bridges; keywords related to work phase include construction and maintenance, loading and unloading operations, and train operation; and keywords related to environmental conditions include night work, severe weather, and high temperatures. Keywords containing entity-type, position / personnel-type, and generic-type objects in the clause text are identified as constraint object characteristics; among them, entity-type keywords include railway transportation enterprises, equipment manufacturing enterprises, and local governments; position / personnel-type keywords include duty officers, dispatchers, and safety management personnel; generic-type keywords include any unit, any individual, and employees. Keywords in the clauses that are obligatory, prohibitory, authorizing, or procedural are identified as characteristics of control requirements. Among them, obligatory keywords include "shall," "should," and "must"; prohibitory keywords include "strictly prohibited," "prohibited," and "must not"; authorizing keywords include "may," "have the right," and "enjoy"; and procedural keywords include "must be," "report," and "approval." Keywords containing terms related to administrative penalties, monetary ranges, and criminal liability in the clauses are identified as characteristics of penalty standards. Among them, keywords related to administrative penalties include warnings, fines, and orders to rectify; keywords related to monetary ranges include fines with a + numerical range; and keywords related to criminal liability include those that constitute a crime and will be prosecuted according to law. Keywords containing phrases related to the effective date of regulations, transitional periods for specific provisions, or repeal / replacement clauses in the clause text are identified as having an effective date characteristic. Specifically, keywords related to the effective date of regulations include "effective from [year / month / day]"; keywords related to the transitional period for specific provisions include "existing facilities shall be rectified within [number] years / month / days" from the effective date; and keywords related to repeal / replacement clauses include "original [article / section / title]". The terms are hereby repealed. Keywords containing explicit references, scope references, or exception clauses in the clause text are identified as related clause characteristics; among them, explicit references include keywords such as "by clause..." Article's provisions; the scope of cited keywords includes the first... Where the provisions of this article are otherwise provided, those provisions shall prevail; the key terms for exception clauses include those not provided in the preceding paragraph.

[0025] For example, regarding the clause in a certain "Regulations on the Safety Management of Railway Dangerous Goods Transportation" that states, "Railway stations transporting explosives shall be equipped with explosion-proof testing equipment that meets national standards and arrange personnel with corresponding qualifications to conduct inspections," it can be identified that the production scenario corresponding to this clause is the transportation of dangerous goods, the subject of the constraint is railway transportation production and operation units, and the control requirements are standard requirements, namely, equipping with corresponding equipment and arranging qualified personnel. If the clause does not directly indicate the penalty, the penalty information is linked to the corresponding penalty range in the superior law "Regulations on Railway Safety Management" of the same regulation, the effective date is the official implementation date stipulated in the regulations, and the associated clause is the corresponding clause in the regulations regarding the qualification management of dangerous goods transportation, thus completing the structured feature extraction of a single clause.

[0026] For example, the clause in the "Regulations on Railway Safety Management" that "railway safety protection zones shall be established on both sides of railway lines" can be identified as having a production scenario of railway line operation and maintenance, a binding object of railway transportation enterprises, a mandatory control requirement, a clear standard for the delineation of safety protection zones, an effective date of the corresponding regulations, and related clauses that are subsequent clauses on prohibited behaviors and corresponding penalties within the safety protection zones, thus completing the structured feature extraction of a single clause.

[0027] It should be noted that the starting identifier specifically refers to the starting character of the numbering of different levels of chapters and clauses. The numbering is split from top to bottom, and each independent control requirement at the lowest level is treated as an independent control clause item. After splitting, the complete level number corresponding to each independent control clause item is retained, which facilitates quick location of the corresponding content when comparing versions in a later time and avoids confusion and matching issues between different chapters and clauses.

[0028] S102. Based on the large model, compare the current version of railway safety production supervision requirements with the previous version of railway safety production supervision requirements, and determine the content change type and mark the changed content. By performing deep, structured analysis on the current version of railway safety production regulatory requirements documents, the characteristics of the current version's requirements are extracted and compared with those of the previous version. This process identifies and marks the changes in the current version's requirements. Based on these changes, the type of content change is determined, enabling precise identification of changes in railway safety production regulatory requirements and avoiding the problems of missed or incorrect assessments caused by traditional manual analysis. The types of content changes include newly added control clauses, deleted or invalidated clauses, modifications to control requirements, adjustments to effective dates, and changes to related clauses. A corresponding change tag is generated for each type of change to facilitate targeted compliance adjustments later.

[0029] The process of comparing the current version of railway safety production supervision requirements with the previous version, and determining the type of content change and marking the changed content includes the following steps: Based on the structured independent control clauses, a mapping relationship between the current version and the previous version is established according to the clause hierarchy number. Clauses without matching mappings are initially identified and marked as newly added control clauses or deleted and invalidated clauses. For clauses with matching mappings, compare the differences in the content of each regulatory requirement feature field one by one. If the production scenario, constraint object, penalty standard or control requirement is modified, it is marked as a control requirement modification; if the effective time field is changed, it is marked as an effective time adjustment; if the number or content of the related clause is changed, it is marked as a related clause change, thus completing the identification and marking output of all changes.

[0030] Large model construction includes the following steps: A structured clause feature vector library is constructed, which transforms each extracted railway safety production supervision requirement feature into a standardized and comparable feature vector, and stores all control clause items of different versions in vectors of the same dimension; Based on the cosine similarity of feature vectors, the feature difference degree of corresponding clauses at the same level in different versions is calculated. When the difference degree exceeds the set difference threshold, it is determined that the clause has content changes. Then, the specific change type is identified one by one for the changed feature fields, and the model-driven automated comparison and identification is completed.

[0031] For example, if the current version adds a new control requirement, "Regular Inspection of Emergency Equipment for Tunnel Construction," and there is no matching clause with a corresponding hierarchical number in the previous version, then this item will be marked as a new control requirement, and a corresponding change label will be generated. If the clause "Frequency Requirements for Manual Foot Inspection of Existing Lines" that existed in the previous version is completely removed in the current version and there is no corresponding matching mapping, then it will be marked as a deleted and invalid clause. If there is a matching mapping for the clause "Qualification Requirements for Safety Guards at Level Crossings," and the current version changes the qualification requirement from "Junior Worker and Above" to "Intermediate Worker and Above," which constitutes a modification of the control requirement content, then it will be marked as a modified control requirement.

[0032] For example, in determining whether a clause has changed content, a difference threshold of 0.15 can be set. If the feature difference calculated by the cosine similarity of the feature vectors of the corresponding clauses in two versions is 0.23, and the feature difference exceeds the set difference threshold, then the clause is determined to have changed content, and the corresponding change type is marked after further comparison of the specific difference fields. If the calculated difference is 0.08, which does not exceed the set difference threshold, then the clause is determined to have no substantial change in content, and there is no need to mark the change. Relying on the vector calculation capability of the large model, the automatic identification of changes in regulatory requirements in different versions can be achieved, avoiding the tediousness and omissions of manual comparison.

[0033] S103. Assess the impact of content change types and marked changes on enterprise business modules; The impact of changes on enterprise business modules is assessed by classifying the types of changes and marking them. These business modules include line construction, high-speed train maintenance, dangerous goods transportation, passenger services, and special equipment operation and maintenance. The types of changes to railway safety production supervision requirements include adding control clauses, deleting or invalidating clauses, modifying control requirements, adjusting effective dates, and changing related clauses. Optionally, different weighting coefficients can be assigned based on the type of change: adding control clauses and modifying control requirements are assigned a first-level weighting coefficient; adjusting effective dates and changing related clauses are assigned a second-level weighting coefficient; and deleting or invalidating clauses are assigned a third-level weighting coefficient. The first-level weighting coefficient is greater than the second-level weighting coefficient, and the second-level weighting coefficient is greater than the third-level weighting coefficient. By combining the production scenarios and constraints associated with the marked changes, the corresponding enterprise business modules are matched, the number of change features under each business module is counted, and the results are calculated to obtain an impact score for that business module. A higher score indicates a greater impact of the railway safety production supervision requirements on that business module.

[0034] The process of assessing the impact of content change types and the marking of changed content on enterprise business modules includes the following steps: For each changed content marked, a value is assigned according to a set weight coefficient; New control clauses and modifications to control requirements are set as the first-level weight coefficient; adjustments to the effective time and changes to related clauses are set as the second-level weight coefficient; deleted or invalidated clauses are set as the third-level weight coefficient. The first-level weight coefficient is greater than the second-level weight coefficient, and the second-level weight coefficient is greater than the third-level weight coefficient. Based on the changed content, the associated production scenarios are extracted, and the weight of the changed content is included in the enterprise business module that matches the corresponding production scenario. The impact score for each business module of the enterprise is calculated using the following expression: ; in, Let i be the score of the i-th enterprise business module. For the i-th enterprise business module, the number of changes in the t-th type of content. Set a weight coefficient for the content change type t. Let t be the safety coefficient of the i-th business module, i be the business module index (line construction, EMU maintenance, dangerous goods transportation, passenger service, special equipment operation and maintenance), t be the content change type index (new control clauses, deleted invalid clauses, control requirement modification, effective time adjustment, related clause changes), and n be the total number of content unchanged types. The impact level of the corresponding business module is determined based on the impact score. The impact level includes high impact level, medium impact level, and low impact level.

[0035] For example, if a railway transportation company's dangerous goods transportation business module has 3 newly added control clauses, 1 control requirement modification, and 2 related clause changes, and the calculated impact score for this business module is higher than the preset high impact threshold, then the business module is classified as high impact, and the company is advised to focus on making compliance adjustments for this business module. If a line construction business module has only 1 related clause change, and the calculated impact score is between the medium and low impact thresholds, then it is classified as medium impact, and the company is advised to complete the corresponding adjustments as planned. If a passenger service business module has only 1 deleted or invalidated clause, and the score is lower than the preset low impact threshold, then it is classified as low impact, and only the filing update needs to be completed; no additional compliance rectification work is required.

[0036] S104. Based on the impact assessment results, the data of the enterprise's internal railway operation rules and regulations documents are compared with the regulatory requirements features, and a revision suggestion is generated based on the comparison results. By comparing the company's existing and valid internal railway operation regulations with the characteristics of the changed regulatory requirements, the system identifies inconsistencies between internal regulations and current regulatory requirements. Based on the severity of the impact, targeted revision suggestions are generated to achieve dynamic alignment between internal regulations and external regulatory requirements. Specifically, this involves matching the changed regulatory requirements with the corresponding clauses in the company's internal regulations to identify issues such as: internal regulations not covering new control requirements; retaining obsolete control requirements; control content inconsistent with current requirements; and outdated effective dates. These issues are then categorized and prioritized according to their severity: high-impact issues are marked as priority revisions, medium-impact issues as routine revisions, and low-impact issues as update / filing items. Finally, a revision suggestion plan is generated, including an issue list, revision directions, and completion deadlines, assisting companies in accurately updating and adjusting their internal regulations.

[0037] Based on the impact assessment results, the data of the enterprise's internal railway operation rules and regulations documents are compared with the regulatory requirements features. Based on the results of the regulatory requirements feature comparison, a revision proposal is generated, including the following steps: Select business modules with a medium or higher level of impact, and extract all tagged regulatory requirement features from the selected business modules; Retrieve existing railway operation rules and regulations documents for the corresponding business modules within the enterprise, perform structured processing on the existing rules and regulations documents according to the same rules as in step S101, and extract the operation control requirements features from the existing rules and regulations. The extracted existing operational control requirements features are compared one by one with the current version of railway safety production supervision requirements features to identify the content in the existing rules and regulations that does not meet the latest supervision requirements, as well as the missing control content that needs to be supplemented. Based on the identified discrepancies and the original wording of regulatory requirements, the system automatically generates revision suggestions for the corresponding rules and regulations, including suggestions for deletion, modification, and addition of clauses. After integrating all revision suggestions, a complete revision proposal is generated.

[0038] The extracted existing operational control requirements features are compared one by one with the current version of railway safety production supervision requirements features to identify the content in existing rules and regulations that does not meet the latest supervision requirements, as well as the missing control content that needs to be supplemented, including the following steps: For each existing regulation and control clause, match it one by one with the regulatory requirements characteristics of the current version in the corresponding production scenario; If the regulatory requirements corresponding to existing clauses have been marked as deleted or void, then non-compliant content is identified as needing to be deleted. If the current version has new regulatory requirements, but the existing rules and regulations do not have corresponding control clauses, then it is identified as missing control content that needs to be supplemented. If the objects of constraint and control requirements of the existing control provisions are inconsistent with the characteristics of the current version of regulatory requirements, they are identified as non-compliant content that needs to be modified. If the effective date or related terms of the existing terms are inconsistent with the current version, they are identified as content that needs to be updated synchronously, and all differences are identified and sorted out.

[0039] Based on the identified discrepancies and the original wording of regulatory requirements, the system automatically generates revision suggestions for the corresponding rules and regulations, including suggestions for deletion, modification, and addition of clauses. Integrating all revision suggestions to generate a complete revision proposal includes the following steps: For the identified regulatory clauses that have been marked for deletion or invalidation, a deletion suggestion for the corresponding clause is directly generated, clearly indicating the clause number to be deleted and the original content, explaining that the basis for deletion is that the corresponding regulatory requirements have been officially invalidated; For the identified missing new regulatory requirements, the original structural features of the regulatory requirements are extracted, and corresponding suggestions for new clauses are generated by combining them with the expression style of the company's existing systems. The corresponding chapter position where the new clauses should be inserted is clearly marked, and the complete content of the control requirements is listed. For the identified content that needs to be modified, compare the differences between the existing clauses and regulatory requirements, generate modification suggestions with the differences marked, retain the existing clauses that comply with regulatory requirements, and only provide the complete modified statements for the differences. For effective dates and related clauses that need to be updated synchronously, update suggestions for the corresponding fields are generated directly, and the accurate updated content is marked.

[0040] For example, a railway transportation company's dangerous goods transportation business module was assessed as high-impact. The newly added control clause, "requirements for the frequency of opening and inspecting dangerous goods containers," was extracted. After comparing it with the company's existing internal rules and regulations, it was found that this content was completely missing. In this case, the system automatically generated a suggested addition clause, directly quoting the original regulatory requirements to prompt the company to supplement it into the corresponding business module's internal regulations, thus incorporating it into the revision proposal and facilitating the company's rapid completion of the supplementation. Similarly, a railway construction company's track construction business module was assessed as high-impact. The modified control requirement feature, "qualification requirements for personnel providing protection during construction near operating lines," was extracted. After structuring the company's existing internal rules and regulations, it was found that the existing regulations still retained the old qualification requirements before the modification, which did not match the latest regulatory requirements. In this case, the system automatically generated a suggested modification clause, directly quoting the standard wording of the latest regulatory requirements, prompting the company to replace and revise the clause, and marking this issue as a priority revision item, incorporating it into the final revision proposal.

[0041] Example 2 The above is the railway safety production regulation compliance assessment method based on a large model provided in the embodiments of this application. The following is the railway safety production regulation compliance assessment system based on a large model provided in the embodiments of this application.

[0042] A railway safety production regulation compliance assessment system based on a large model includes: The first data processing unit is used to perform deep structured parsing on the current version of the railway safety production supervision requirements document data in order to extract the characteristics of the railway safety production supervision requirements. The second data processing unit is used to compare the current version of railway safety production supervision requirements with the previous version of railway safety production supervision requirements based on the large model, and to determine the type of content change and mark the changed content. The third data processing unit is used to assess the impact on enterprise business modules based on the type of content change and the marked changes in content. The fourth data processing unit is used to perform regulatory requirement feature comparison processing on the data of internal railway operation rules and regulations documents of the enterprise based on the impact degree assessment processing results, and generate revision suggestions based on the regulatory requirement feature comparison processing results.

[0043] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0044] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for assessing compliance with railway safety regulations based on a large model, characterized in that, Includes the following steps: S101. Obtain the current version of the railway safety production supervision requirements document data and perform deep structured parsing to extract the characteristics of the railway safety production supervision requirements. S102. Based on the large model, compare the current version of railway safety production supervision requirements with the previous version of railway safety production supervision requirements, and determine the content change type and mark the changed content. S103. Assess the impact of content change types and marked changes on enterprise business modules; S104. Based on the impact assessment results, the data of the enterprise's internal railway operation rules and regulations documents are compared and processed according to the regulatory requirements characteristics, and a revision suggestion is generated based on the comparison and processing results of the regulatory requirements characteristics.

2. The railway safety production regulation compliance assessment method based on a large model according to claim 1, characterized in that, Step S101 includes the following steps: The current version of the railway safety production supervision requirements document is split according to the text format, retaining the main text information of the clauses; The retained clause text information is structurally broken down according to chapters and clause levels to obtain independent control clause items; Feature extraction is performed on each independent control clause to identify the railway safety production supervision requirements characteristics corresponding to each clause. The railway safety production supervision requirements characteristics include production scenario, constraint object, control requirements, penalty standards, effective time, and related clauses.

3. The railway safety production regulation compliance assessment method based on a large model according to claim 2, characterized in that, The process of extracting features from each independent control clause to identify the railway safety production supervision requirements corresponding to each clause, including production scenario, constraint object, control requirements, penalty standards, effective time, and related clauses, includes the following steps: Keywords containing terms related to location, work phase, or environmental conditions in the clauses are identified as production scenario characteristics. Keywords containing entity, position / personnel, or generic objects in the clause text are identified as constraint object characteristics. Keywords containing the categories of obligation, prohibition, authorization, and procedure in the clauses are identified as characteristics of control requirements. Keywords containing terms related to administrative penalties, monetary ranges, or criminal liability in the clauses are identified as characteristics of penalty standards. Keywords containing the terms "effective date of regulations", "transition period of specific clauses", or "repeal and replacement" in the clause text are identified as effective date characteristics. Keywords containing explicit references, scope references, or exception clauses in the clause text are identified as related clause characteristics.

4. The railway safety production regulation compliance assessment method based on a large model according to claim 1, characterized in that, Step S102 includes the following steps: Based on the structured independent control clauses, a mapping relationship between the current version and the previous version is established according to the clause hierarchy number. Clauses without matching mappings are initially identified and marked as newly added control clauses or deleted and invalidated clauses. For clauses with matching mappings, compare the differences in the content of each regulatory requirement feature field one by one. If the content of the production scenario, constraint object, penalty standard or control requirement is modified, it is marked as a control requirement modification. If the effective time field changes, it will be marked as an effective time adjustment; If the number or content of the related clause changes, it will be marked as a change to the related clause, thus completing the identification and marking of all changes.

5. The railway safety production regulation compliance assessment method based on a large model according to claim 4, characterized in that, The construction of the large model includes the following steps: A structured clause feature vector library is constructed, which transforms each extracted railway safety production supervision requirement feature into a standardized and comparable feature vector, and stores all control clause items of different versions in vectors of the same dimension; Based on the cosine similarity of feature vectors, the feature difference degree of corresponding clauses at the same level in different versions is calculated. When the difference degree exceeds the set difference threshold, it is determined that the clause has content changes. Then, the specific change type is identified one by one for the changed feature fields, and the model-driven automated comparison and identification is completed.

6. The railway safety production regulation compliance assessment method based on a large model according to claim 1, characterized in that, Step S103 includes the following steps: For each changed content marked, a value is assigned according to a set weight coefficient; New control clauses and modifications to control requirements are set as the first-level weight coefficient; adjustments to the effective time and changes to related clauses are set as the second-level weight coefficient; deleted or invalidated clauses are set as the third-level weight coefficient. The first-level weight coefficient is greater than the second-level weight coefficient, and the second-level weight coefficient is greater than the third-level weight coefficient. Based on the changed content, the associated production scenarios are extracted, and the weight of the changed content is included in the enterprise business module that matches the corresponding production scenario. The impact score for each business module of the enterprise is calculated using the following expression: ； in, Let i be the score of the i-th enterprise business module. For the i-th enterprise business module, the number of changes in the t-th type of content. Set a weight coefficient for the content change type t. Let t be the security coefficient of the i-th business module, i be the index of the business module, t be the index of the content change type, and n be the total number of content unchanged types. The impact level of the corresponding business module is determined based on the impact score. The impact level includes high impact level, medium impact level, and low impact level.

7. The railway safety production regulation compliance assessment method based on a large model according to claim 1, characterized in that, Step S104 includes the following steps: Select business modules with a medium or higher level of impact, and extract all tagged regulatory requirement features from the selected business modules; Retrieve existing railway operation rules and regulations documents for the corresponding business modules within the enterprise, perform structured processing on the existing rules and regulations documents according to the same rules as in step S101, and extract the operation control requirements features from the existing rules and regulations. The extracted existing operational control requirements features are compared one by one with the current version of railway safety production supervision requirements features to identify the content in the existing rules and regulations that does not meet the latest supervision requirements, as well as the missing control content that needs to be supplemented. Based on the identified discrepancies and the original wording of regulatory requirements, the system automatically generates revision suggestions for the corresponding rules and regulations. After integrating all revision suggestions, a complete revision proposal is generated.

8. The railway safety production regulation compliance assessment method based on a large model according to claim 7, characterized in that, The process of comparing the extracted existing operational control requirements features with the current version of railway safety production supervision requirements features one by one to identify content in existing rules and regulations that does not comply with the latest supervision requirements, as well as missing control content that needs to be supplemented, includes the following steps: For each existing regulation and control clause, match it one by one with the regulatory requirements characteristics of the current version in the corresponding production scenario; If the regulatory requirements corresponding to existing clauses have been marked as deleted or void, then non-compliant content is identified as needing to be deleted. If the current version has new regulatory requirements, but the existing rules and regulations do not have corresponding control clauses, then it is identified as missing control content that needs to be supplemented. If the objects of constraint and control requirements of the existing control provisions are inconsistent with the characteristics of the current version of regulatory requirements, they are identified as non-compliant content that needs to be modified. If the effective date or related terms of the existing terms are inconsistent with the current version, they are identified as content that needs to be updated synchronously, and all differences are identified and sorted out.

9. The railway safety production regulation compliance assessment method based on a large model according to claim 7, characterized in that, Based on the identified discrepancies and the original wording of regulatory requirements, the system automatically generates revision suggestions for the corresponding rules and regulations. Integrating all revision suggestions to generate a complete revision proposal includes the following steps: For the identified regulatory clauses that have been marked for deletion or invalidation, a deletion suggestion for the corresponding clause is directly generated, clearly indicating the clause number to be deleted and the original content, explaining that the basis for deletion is that the corresponding regulatory requirements have been officially invalidated; For the identified missing new regulatory requirements, the original structural features of the regulatory requirements are extracted, and corresponding suggestions for new clauses are generated by combining them with the expression style of the company's existing systems. The corresponding chapter position where the new clauses should be inserted is clearly marked, and the complete content of the control requirements is listed. For the identified content that needs to be modified, compare the differences between the existing clauses and regulatory requirements, generate modification suggestions with the differences marked, retain the existing clauses that comply with regulatory requirements, and only provide the complete modified statements for the differences. For effective dates and related clauses that need to be updated synchronously, update suggestions for the corresponding fields are generated directly, and the accurate updated content is marked.

10. A railway safety production regulation compliance assessment system based on a large model, used to implement the railway safety production regulation compliance assessment method based on a large model as described in any one of claims 1-9, characterized in that, include: The first data processing unit is used to perform deep structured parsing on the current version of the railway safety production supervision requirements document data in order to extract the characteristics of the railway safety production supervision requirements. The second data processing unit is used to compare the current version of railway safety production supervision requirements with the previous version of railway safety production supervision requirements based on the large model, and to determine the type of content change and mark the changed content. The third data processing unit is used to assess the impact on enterprise business modules based on the type of content change and the marked changes in content. The fourth data processing unit is used to perform regulatory requirement feature comparison processing on the data of internal railway operation rules and regulations documents of the enterprise based on the impact degree assessment processing results, and generate revision suggestions based on the regulatory requirement feature comparison processing results.

Images