Rail transit shift change log intelligent generation and management method based on artificial intelligence

Abstract

The invention relates to the technical field of data processing and management, and discloses a rail transit shift change log intelligent generation and management method based on artificial intelligence, and the method comprises the steps: collecting various types of original handover data related to shift change, and carrying out the preprocessing of the original handover data to generate an original semantic Token; a preset multi-mode log content generation engine is adopted for processing, and a structured shift change log is obtained; performing logic consistency, knowledge association and format specification verification on the structured shift change log by adopting a preset rail transit operation knowledge graph and a logic rule to obtain a shift change log draft; pre-defining a multi-stage approval process for multi-stage approval to obtain a target shift change log; and key matter reminding of the target shift change log is realized through setting a time node, and log content risk intelligent early warning is carried out. According to the method, manual operation and an intelligent technology are deeply fused through natural language processing and a rule engine, and automatic analysis, structured generation, semantic verification, intelligent reminding and closed-loop management of the rail transit shift change logs are achieved.

Description

Technical Field

[0001] This invention relates to the field of data processing and management technology, and in particular to an artificial intelligence-based method for intelligent generation and management of rail transit shift logs. Background Technology

[0002] As a crucial component of the national transportation system, the safe and efficient operation of rail transit is of vital importance to the national economy and people's livelihood. In the daily scheduling and maintenance of rail transit, shift logs, serving as a vital link between preceding and following work processes and transmitting critical operational information, are a fundamental management tool for ensuring the safe operation and orderly scheduling of the rail transit system. Traditional rail transit shift logs are primarily paper-based records or simple forms, relying mainly on manual filling and circulation. This method has several shortcomings: 1. Manual data entry is inefficient and prone to errors. In practice, shift workers spend a significant amount of time collecting and manually organizing information, especially when dealing with data from multiple departments and dimensions (such as construction plans, equipment malfunctions, maintenance schedules, and potential hazards). This can easily lead to omissions or unclear descriptions, affecting the accurate understanding of incoming shift workers. 2. The unstructured and non-standardized nature of shift handover content is widespread. A large amount of log content is recorded in natural language, making effective classification, statistics, and querying difficult. It also lacks consistent standards, hindering long-term data accumulation and mining, and failing to support the demands of modern rail transit operation and maintenance for "intelligent data analysis and risk prediction." 3. The log transmission and approval process lacks real-time efficiency and transparency. Paper or tabular logs are often transmitted offline via USB drives or printouts, which not only delays processing but also poses risks of tampering and loss. Furthermore, log content review lacks hierarchical approval and record-keeping, resulting in blurred responsibility boundaries and difficulties in tracing the source of information.

[0003] In recent years, with the continuous improvement of the digitalization and intelligence level of rail transit, building efficient, unified, and intelligent shift logs has become an inevitable trend in the industry. To this end, some units have attempted to develop information systems for shift logs, mainly using web systems and mobile terminals to transform traditional paper logs into electronic forms. However, existing systems mostly remain at the stage of "automated information entry," lacking intelligent processing and analysis capabilities, and failing to truly achieve higher-level goals such as "structured generation, semantic extraction, knowledge accumulation, and risk identification" of log data.

[0004] Therefore, there is an urgent need for a new, AI-based online generation and management method for rail transit shift logs to compensate for the shortcomings of existing technologies in information fusion, intelligent recognition, and in-depth analysis, and to achieve a digital, intelligent, and safe upgrade of rail transit shift work. Summary of the Invention

[0005] This invention provides an intelligent generation and management method for rail transit shift logs based on artificial intelligence. By using AI natural language processing, rule engines, semantic analysis, knowledge graphs and other technologies, it deeply integrates manual operation with intelligent technology to achieve automatic parsing, structured generation, semantic verification, intelligent reminders and closed-loop management of rail transit shift log content. This significantly improves the efficiency, accuracy and intelligence level of the shift handover process and solves problems such as traditional manual recording, fragmented information and misunderstanding.

[0006] This invention provides an artificial intelligence-based method for intelligent generation and management of rail transit shift logs, comprising: S1. Collect various types of raw handover data related to shift changes, and preprocess the raw handover data to generate raw semantic tokens; wherein, the preprocessing includes standardization cleaning, synonym normalization, and format recognition. S2. The original semantic token is processed using a preset multimodal log content generation engine to obtain a structured handover log that meets the handover specifications; wherein, the multimodal log content generation engine includes a semantic classification matching unit, a structured filling unit, and a context completion unit. S3. The structured shift handover log is checked for logical consistency, knowledge association and format standardization using a preset rail transit operation knowledge graph and logical rules to obtain a draft shift handover log. S4. Predefine a multi-level approval process and combine it with the approval mechanism to conduct multi-level approval of the handover log draft, so as to obtain the target handover log and push it to the external system; S5. By setting time nodes, key items in the target shift handover log are reminded, and the target shift handover log is analyzed based on a machine learning model to provide intelligent risk warnings for the log content.

[0007] Furthermore, S1 specifically includes: S101. Manually input data through a visual interface or collect various types of original handover data related to shift changes by connecting to a data source interface; wherein, the data source interface includes a scheduling system, production planning system, maintenance platform, safety supervision platform or big data platform, and the original handover data includes the current shift's work plan, handover content, construction plan, maintenance window arrangement, fault work order, and temporary work instructions. S102. The natural language preprocessing function based on the pre-trained language model performs standardization cleaning, synonym normalization, and format recognition on the original handover data, and generates original semantic tokens for subsequent analysis; wherein, the pre-trained language model includes BERT and RoBERTa.

[0008] Furthermore, S2 specifically includes: S201. The semantic classification and matching unit classifies the original semantic tokens according to a preset machine learning model and generates standard statements based on log generation templates; wherein, the categories include operating status, plan changes, equipment failure, and security risks; S202. The structured filling unit categorizes and fills the standard statements according to log fields, and outputs structured JSON format log items. S203. The context completion module intelligently completes incomplete or vaguely described content in the structured JSON format log items based on historical shift experience logs and knowledge graphs to obtain a structured shift handover log that meets the handover specifications.

[0009] Furthermore, S3 specifically includes: S301. Based on the AI ​​model, verify the logical consistency of the internal descriptions of the structured handover log content, mark potentially conflicting items, and prompt for manual confirmation; S302. Establish entity relationships between the content of the structured shift handover log and the preset rail transit operation knowledge graph to identify missing items and content deviations; wherein, the rail transit operation knowledge graph includes train plans, construction arrangements, fault types, and shift handover specifications; S303. Combine the preset logical rules to check whether there are logical conflicts in the content of the structured handover log, and at the same time standardize the language style and field format of the structured handover log.

[0010] Furthermore, S4 specifically includes: S401. Predefine multi-level approval processes based on approval nodes, and introduce AI-assisted review in the approval nodes to mark potentially high-risk items for focused review; S402. At the approval node, the current version of the handover log draft is compared with the historical version, and the differences are highlighted. A signing confirmation mechanism is set up to determine that the approval at the current approval node is completed. S403. After completing the approval process according to the multi-level approval process, the target shift handover log is generated and archived in multiple formats, and pushed to the required external system through the interface; among them, the multiple formats include PDF, JSON, and CSV.

[0011] Furthermore, S5 specifically includes: S501. Set a shift handover time node and remind the incoming shift personnel of key items in the target shift handover log; wherein, the key items include delays, temporary construction, and equipment status not restored; S502. Based on the content of the target shift handover log, analyze the existing abnormal situations using a machine learning model, output risk information and issue warnings to the visualization interface; wherein, the machine learning model continuously learns various log errors and accident precursor patterns to continuously optimize the warning capability.

[0012] This invention also provides an artificial intelligence-based intelligent generation and management device for rail transit shift logs, based on the aforementioned artificial intelligence-based intelligent generation and management method for rail transit shift logs, characterized in that the device comprises: The data acquisition module is used to collect various types of raw handover data and preprocess the raw handover data to generate raw semantic tokens; wherein, the preprocessing includes standardization cleaning, synonym normalization, and format recognition. The generation module is used to process the original semantic token using a preset multimodal log content generation engine to obtain a structured handover log that meets the handover specifications; wherein, the multimodal log content generation engine includes a semantic classification matching unit, a structured filling unit, and a context completion unit; The verification module is used to verify the logical consistency, knowledge association and format standardization of the structured shift log using a preset rail transit operation knowledge graph and logical rules, so as to obtain a draft of the shift log. The approval module is used to predefine multi-level approval processes and, in conjunction with the approval mechanism, to conduct multi-level approvals on the draft handover log to obtain the target handover log for push to external systems. The early warning module is used to remind users of key matters in the target shift handover log by setting time nodes, and to analyze the target shift handover log based on a machine learning model to provide intelligent early warning of risks in the log content.

[0013] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0015] The beneficial effects of this invention are as follows: This invention utilizes multi-source data fusion preprocessing to improve semantic recognition accuracy and input data comprehensiveness, reduce manual data entry and classification, enhance data consistency and semantic understanding depth, and solidify the foundation of operational knowledge. The automatic log generation and completion mechanism significantly reduces manual intervention, improves generation efficiency and content completeness, avoids information omissions and non-standard terminology, and ensures accurate and clear handover information. Semantic verification and knowledge matching mechanisms enhance log accuracy, rigor, and verifiability, avoid logical contradictions, and contribute to the formation of high-quality operational data assets. The closed-loop approval process improves approval efficiency and traceability, achieves digital archiving, prevents content tampering and ambiguity of responsibility, and provides a secure and controllable management basis. AI risk warning and continuous learning mechanisms strengthen proactive security protection, with early warning intervention reducing operational risks. Through continuous optimization, a self-evolving risk control system is formed, comprehensively achieving efficient, standardized, and secure shift handover work. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the intelligent generation and management method for rail transit shift logs based on artificial intelligence, as described in this invention.

[0017] Figure 2 This is a flowchart illustrating step S1 in this invention.

[0018] Figure 3 This is a flowchart illustrating step S2 in this invention.

[0019] Figure 4 This is a schematic diagram of the device structure according to an embodiment of the present invention.

[0020] Figure 5 This is a schematic diagram of the internal structure of a computer device according to an embodiment of the present invention.

[0021] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0023] Artificial intelligence technologies, particularly Natural Language Processing (NLP), semantic analysis, machine learning, and knowledge graph construction, have been successfully applied across various industries, providing a new technological path for the intelligent transformation of rail transit shift logs. Leveraging AI algorithms, it is possible to achieve semantic understanding of massive amounts of log content, extraction of key information, verification of content logic consistency, automatic structured storage, and multi-dimensional statistical analysis, thereby significantly improving the accuracy, real-time performance, and intelligence of shift handover work. Currently, there is a lack of online shift handover methods that deeply integrate artificial intelligence with rail transit shift logs, ensuring both the integrity of the data structure and the accuracy of the shift handover logic, while also enabling intelligent guidance, content correction, and risk warnings throughout the shift handover process. Therefore, this invention provides an AI-based intelligent generation and management method for rail transit shift logs, comprising the following five steps: data input and AI preprocessing, intelligent generation of shift log content, semantic verification and knowledge mapping, log approval and push management, and reminder and risk warning mechanisms.

[0024] This invention aims to improve the efficiency, accuracy, and intelligent management of shift handover logs through intelligent technologies such as natural language processing, semantic analysis, and knowledge graphs. The following is a detailed implementation of this method.

[0025] like Figure 1 As shown, this invention provides an intelligent generation and management method for rail transit shift logs based on artificial intelligence, including: S1. Collect various types of raw handover data related to shift changes, and preprocess the raw handover data to generate raw semantic tokens; wherein, the preprocessing includes standardization cleaning, synonym normalization, and format recognition. Figure 2 As shown, it specifically includes: S101. Manually input data through a visual interface (the handover personnel input handover information through the visual interface), or collect various types of original handover data related to the handover by connecting to a data source interface; wherein, the data source interface includes a scheduling system, production planning system, maintenance platform, safety supervision platform or big data platform, and the original handover data includes the current shift's work plan, handover content, construction plan, track closure arrangement, fault work order, and temporary work instructions. The system proactively adapts and integrates with the core business systems of rail transit, supporting access to multiple data sources including dispatching systems, production planning systems, maintenance platforms, safety monitoring platforms, and big data platforms, enabling real-time data synchronization. It retrieves current shift work plans and formal handover details from the dispatching and production planning systems; construction plans and maintenance window arrangements (including window times, work areas, and responsible personnel) from construction management modules; fault work orders (including faulty equipment, fault descriptions, and processing progress) and safety hazard records from the maintenance and safety monitoring platforms; and synchronizes dynamic data such as temporary work instructions (e.g., emergency maintenance tasks and operational adjustment notices) in real time from the dispatching system. The data acquisition process supports breakpoint resumption and duplicate data filtering to ensure the integrity and uniqueness of the collected data and avoid redundant information interference.

[0026] S102. The natural language preprocessing function based on the pre-trained language model performs standardization cleaning, synonym normalization, and format recognition on the original handover data, and generates original semantic tokens for use as semantic input for subsequent structured generation and understanding; wherein, the pre-trained language model includes BERT and RoBERTa.

[0027] Standardized cleaning: Identify and correct spelling errors, punctuation misuse, and other issues in text; unify data formats, standardizing time representations and equipment number formats from different sources into a unified system standard; filter meaningless and redundant information (such as duplicated instructions, invalid spaces, and symbols).

[0028] Synonym unification: Relying on the rail transit industry terminology database, terms and phrases with the same meaning but different expressions are unified into standard expressions (such as "skylight maintenance plan" and "skylight arrangement" are unified into "skylight operation plan", and "signal malfunction" and "signal equipment abnormality" are unified into "signal system malfunction") to ensure data semantic consistency.

[0029] Format recognition and entity extraction: Identify key entity information in the text, including time, location, equipment name, responsible unit, fault type, etc., and mark field attributes; at the same time, identify the data format type (such as text description, tabular data, work order number, etc.) to provide a classification basis for subsequent structured processing.

[0030] Generate raw semantic tokens: All the data processed above is transformed into a sequence of semantic tokens that can be recognized by the pre-trained language model. Each token corresponds to a complete semantic unit (e.g., the track maintenance work in section 3 of Line 3 from 10:00 to 12:00 on May 20, 2024 is broken down into associated semantic tokens containing time, line, and work type). Finally, a standardized semantic input package is output for use in the subsequent structured generation and semantic analysis in step two.

[0031] As described in steps S101-S102 above, at the initial stage of shift handover, the system collects relevant raw data from multiple channels, including manual input and automatic interface input. Shift workers can fill in information such as the operational status, construction schedule, equipment malfunctions, and safety hazards related to their shift through a visual interface. Simultaneously, the system can connect to data sources such as the dispatch system, maintenance schedule system, equipment maintenance platform, safety monitoring system, and big data platform to automatically collect relevant information and avoid omissions. All input content undergoes standardized processing in the AI ​​preprocessing module. This module integrates a natural language understanding system built on mainstream pre-trained language models such as BERT and RoBERTa, possessing capabilities such as word normalization, format recognition, and named entity extraction. The system automatically corrects spelling errors, recognizes time expressions, standardizes terminology, and converts it into a standard token format for subsequent semantic modeling and structured processing. This step ensures comprehensive and diverse data sources and accurate and unified preprocessing, laying a solid foundation for the subsequent structured and intelligent implementation of shift handover logs.

[0032] S2. The original semantic token is processed using a preset multimodal log content generation engine to transform the input natural language or structured fragments into a structured handover log that conforms to the handover specifications. The engine then organizes the item order and generates suggested content based on contextual semantics and knowledge rules. The multimodal log content generation engine includes a semantic classification and matching unit, a structured completion unit, and a contextual completion unit. Figure 3 As shown, it specifically includes: S201. The semantic classification and matching unit classifies the original semantic tokens based on a preset machine learning model and generates standard statements based on a log generation template, achieving automatic conversion from "natural language" to "standard statements". The categories include operational status, plan changes, equipment failures, and security risks. The semantic classification and matching unit relies on a pre-trained machine learning model (based on massive historical rail transit shift log samples, this model integrates semantic classification and element extraction tasks through a multi-task learning mechanism, possessing extremely strong industry semantic recognition capabilities). The model receives the original semantic token sequence and, by analyzing the semantic relationships, keyword features, and contextual logic between tokens, automatically classifies the log content into preset categories, including four core categories: operational status, plan changes, equipment failures, and safety hazards. A confidence level judgment mechanism is introduced during the classification process. If the classification confidence level of a certain semantic token is lower than a preset threshold (e.g., 85%), it is marked as a category to be confirmed and subsequently pushed to the manual review stage to ensure classification accuracy.

[0033] A log generation template library aligning with rail transit shift handover standards is established. Templates are designed according to the aforementioned classification results, with each template clearly defining standard sentence structure, terminology, and core information dimensions (e.g., an equipment fault template includes core elements such as faulty equipment, fault location, fault description, processing progress, and pending tasks). After semantic classification, log templates corresponding to the specified categories are matched, and variable fields in the templates are populated based on key information from the original semantic tokens, achieving automatic conversion from "natural language" to "standard statements."

[0034] S202. The structured filling unit categorizes and fills the standard statements according to log fields, and outputs structured JSON format log items. The structured data entry unit reads the generated standard statements and, in conjunction with a pre-defined handover log field system (fields covering basic information, event category, core content, responsible party, time node, processing status, and pending items, with each dimension containing sub-fields, such as basic information including shift, handover person, and handover time), accurately extracts and categorizes key information from the standard statements. Following the field definitions, the extracted information is then filled into the corresponding fields, forming key-value pairs of structured data, ensuring that each field is complete and unambiguous. Finally, the structured data is encapsulated into JSON format log items for easy storage, parsing, and cross-system transmission.

[0035] S203. The context completion module intelligently completes incomplete or vaguely described content in the structured JSON format log items based on historical shift experience logs and knowledge graphs to obtain a structured shift handover log that meets the handover specifications.

[0036] The context completion module performs a full field traversal inspection on structured JSON format log items, focusing on identifying two types of problems: missing fields and ambiguous descriptions. It then calls upon the historical shift experience log database (which stores anonymized past valid shift handover logs, indexed by category and scenario) and the rail transit operation knowledge graph (which includes entities and relationships such as train lines, equipment parameters, construction specifications, fault handling procedures, and shift handover standards).

[0037] To address the issue of missing fields, based on the core information of log entries, complete logs of similar scenarios are retrieved from the historical log database to extract common field information for similar faults. Simultaneously, by combining the relationships between turnouts, their associated lines, and the responsible work teams in the knowledge graph, field completion suggestions are generated. To address the issue of ambiguous descriptions, semantic relationships are analyzed using the knowledge graph, and combined with standard expressions for similar historical adjustments, the ambiguous descriptions are refined into specific content.

[0038] Step S2 is completed by an AI-driven log content generation engine. Its core objective is to transform the input natural language content or data fragments into structured log records that conform to specifications. An integrated multimodal semantic recognition and template matching engine first performs semantic classification on the input content, automatically dividing the text into categories such as equipment failure, maintenance window arrangements, construction plans, and risk warnings, and matching the corresponding structured templates. The structured completion module maps the identified key information to preset fields, such as time, event description, responsible unit, and processing status, outputting structured log items in JSON format. For cases with ambiguous descriptions or missing fields, the system will invoke the context completion module, combining historical log database and knowledge graph content to infer and recommend supplementary items or rewrite suggestions. This step employs a multi-task learning approach, integrating a joint training mechanism for tasks such as semantic classification, information extraction, and content generation. This enables it to understand industry language and output highly consistent and readable handover entries, significantly reducing manual editing time.

[0039] S3. Using a pre-defined rail transit operation knowledge graph and logical rules, the structured shift handover log is checked for logical consistency, knowledge association, and format standardization to obtain a draft shift handover log. Specifically, this includes: S301. Based on the AI ​​model, verify the logical consistency of the internal descriptions of the structured handover log content, mark potentially conflicting items, and prompt for manual confirmation; A specialized verification model based on BERT and RoBERTa pre-trained models is adopted. This model is fine-tuned and trained using massive samples of rail transit shift logs, and has the ability to identify semantic inconsistencies and logical conflicts within the logs. It can accurately capture consistency issues in dimensions such as time, equipment, process, and data. Logical consistency verification includes time dimension, equipment-event dimension, and process dimension verification. Specifically, Time-based verification: Automatically compares the logical rationality of various time information in the logs, including whether the start and end times of the construction plan overlap with the scheduled maintenance window, whether the fault occurrence time is later than the handling start time, and whether the temporary work instruction time is within the current shift. Equipment-event-based verification: Verifies the consistency between equipment status and event descriptions. For example, if a piece of equipment has been marked as completed, subsequent entries should not contain descriptions indicating that the equipment is still malfunctioning. Process-based verification: Verifies the continuity of event handling steps in the logs according to the standard rail transit operation and maintenance process.

[0040] After verification, potential conflicting items will be presented in a format of "highlighted annotation + explanation of the reason for the conflict". Manual confirmation is supported, and the shift handover personnel can view the conflict details.

[0041] S302. Establish entity relationships between the content of the structured shift handover log and the preset rail transit operation knowledge graph to identify missing items and content deviations; wherein, the rail transit operation knowledge graph includes train plans, construction arrangements, fault types, and shift handover specifications; The pre-built knowledge graph for rail transit operations has completed the construction of all entities and their relationships. The core includes four main categories of entities and their associated rules: Train planning entities: lines, train schedules, operating sections, stops, planned duration, and line-train-operating section relationships; Construction arrangement entities: construction type, work area, required maintenance window time, safety protection requirements, and construction type-maintenance window time matching rules; Fault type entities: equipment category (signals, switches, power supply equipment, etc.), fault phenomenon, fault level, standard handling procedure, and equipment category-fault phenomenon-handling procedure relationships; Shift handover specifications entities: required fields, terminology standards, item classification rules, and event category-required field correspondence.

[0042] Extract core entities from structured shift handover logs and accurately match them with entities in the knowledge graph using semantic similarity algorithms to establish one-to-one or one-to-many relationships.

[0043] Based on knowledge graph association rules, the system verifies whether the logs are missing any required information, marks missing items, and generates supplementary suggestions. It also verifies the consistency between log entities and knowledge graph rules. If a turnout jamming fault in the log is categorized as a power supply equipment fault, it is considered a category deviation. If the construction type is track grinding but no maintenance window is specified, it conflicts with the knowledge graph rule that track grinding requires a maintenance window. The system marks the deviation and suggests directions for correction.

[0044] S303. Combine the preset logical rules to check whether there are logical conflicts in the content of the structured handover log, and at the same time standardize the language style and field format of the structured handover log.

[0045] The rule base integrates the business logic and management requirements of rail transit operation and maintenance, including hard logic rules and flexible specification rules. Hard logic rules stipulate that construction plans must not conflict with train operation plans, fault work orders must indicate the fault level, and temporary work instructions must include the execution deadline, etc. Flexible specification rules stipulate the standard of terminology, field format requirements, etc.

[0046] The structured logs are validated one by one according to the rule base order, with a focus on verifying the compliance of hard logical rules. For logical conflicts found during validation, a distinction is made between serious conflicts (which require mandatory correction) and general conflicts (where the circumstances can be explained). Serious conflicts (such as plan conflicts that affect driving safety) must be corrected before proceeding to the next stage.

[0047] Based on a terminology database for the rail transit industry, non-standard expressions and colloquialisms in the logs are automatically replaced to ensure consistent terminology. Fields such as time, equipment number, and fault level are formatted uniformly. Finally, the results of logical consistency checks, knowledge graph mapping, rule validation, and standardization are summarized, and combined with corrections made after manual confirmation, the structured shift handover logs are optimized to produce a high-quality draft, which is then simultaneously pushed to the approval process in step S4.

[0048] As described in steps S301-S303 above, to ensure the logical accuracy and industry standardization of the shift handover log, this invention incorporates semantic verification and knowledge mapping. The semantic consistency verification module automatically identifies conflicts between descriptions within the log content and highlights conflicting content for manual confirmation. The log content is mapped to a rail transit operation knowledge graph, and entity relationship matching identifies missing content and contextual discrepancies, providing improvement suggestions. Combined with manually set or imported business rule bases, the system automatically verifies whether the log content conforms to logical rules, ensuring its compliance and auditability. This step generates a structured draft log, providing a high-quality content source for subsequent approval and dissemination.

[0049] S4. Predefine a multi-level approval process and, in conjunction with the approval mechanism, conduct multi-level approvals on the draft handover log to obtain the target handover log for push to the external system. This step provides an intelligent process management mechanism after the log content is generated, ensuring the streamlined and closed-loop control of the handover content review, signing, and archiving process. Specifically, it includes: S401. Predefine multi-level approval processes based on approval nodes (e.g., writer → on-site foreman → operation shift worker → dispatcher, supporting customizable paths), and introduce AI-assisted review in approval nodes to mark potentially high-risk items for key review, with process traceability and support for review and approval traceability. It offers predefined multi-level approval processes and supports personalized process customization, allowing users to flexibly adjust approval nodes, roles, and workflow order based on their organizational structure and business permission divisions. Each approval node is bound to the corresponding role's operating permissions, clearly defining review responsibilities and preventing unauthorized operations or ambiguous review responsibilities.

[0050] AI-assisted review is based on a knowledge graph of rail transit operations, a historical accident case library, and a business rule library. It automatically identifies high-risk content in logs and highlights key problem scenarios, such as unresolved faults related to train operation safety, potential conflicts between construction plans and train operation plans, and vague or inconsistent descriptions of the status of key equipment. In the review interface, high-risk items are highlighted in red with risk level labels and explanations, helping reviewers quickly focus on key points and improve review efficiency.

[0051] During the approval process, key information at each node is dynamically recorded, including the reviewer, review time, review comments (approval / rejection / modification suggestions), and operation IP address, forming an immutable approval log to ensure that every modification and transfer is traceable.

[0052] S402. At the approval node, the current version of the handover log draft is compared with the historical version, and the differences are highlighted. A signing and confirmation mechanism is set up, and a manual confirmation window is provided. After the incoming staff confirms that the handover is correct, the approval of the current approval node is determined to be completed. The storage log tracks all historical versions from draft generation to approval and modification at each level, culminating in the final draft. Each version is uniquely tagged with a version number, modification node, and the person who made the modification, preventing version confusion or loss. In the approval node, a one-click comparison is set up between the current version and any historical version, using different colors to highlight differences, helping subsequent approval nodes quickly understand the modification logic and avoiding duplicate reviews or misunderstandings.

[0053] After each approval node is completed, the reviewer signs off, automatically locking modification permissions for that node and allowing only the next level reviewer to modify or reject the existing version. Once all multi-level approvals are complete, the system pushes the logs to the incoming staff's visual interface, providing a dedicated manual confirmation window. The incoming staff must check the log content item by item, and only after confirming that everything is correct can the entire approval process be closed.

[0054] S403. After completing the approval process according to the multi-level approval process, the target shift handover log is generated and archived in multiple formats, and pushed to the required external system through the interface; among them, the multiple formats include PDF, JSON, and CSV.

[0055] Once the entire approval process is complete, the final version of the target handover log will be converted into three standard formats: PDF, JSON, and CSV, to meet the needs of different scenarios. The generated archived files are automatically stored in the system database or cloud storage platform according to a preset hierarchical directory, supporting quick retrieval by log number, handover person, event type, and other dimensions for convenient subsequent review and retrieval.

[0056] Complete the interface configuration with the core business system of rail transit in advance, and support the push of archived log data to the required external systems, including but not limited to operation and maintenance management platforms, big data analysis platforms, safety supervision platforms, and enterprise OA systems, through mainstream interface methods such as standard RESTful API and MQ message queue.

[0057] Following steps S401-S403 above, the log content is structured and then enters the approval process management stage. It provides flexible multi-level approval process configuration, supporting users to configure hierarchical approval paths according to job positions. Paths are customizable, and nodes can be controlled in parallel or sequentially to meet diverse organizational needs. After entering the approval interface, each level of reviewer can view AI-annotated high-risk or conflict information, and approvers can annotate, reject, or modify the logs. All approval records are logged and support full-text version comparison, facilitating backtracking and verification, and preventing users from overlooking important changes. After final approval, the handover log is automatically archived in PDF, CSV, JSON, and other formats, and can be pushed to other systems via standard RESTful APIs or MQ messaging mechanisms to achieve data synchronization and cross-platform sharing.

[0058] S5. Key reminders for the target shift handover log are implemented by setting time nodes, and the target shift handover log is analyzed based on a machine learning model to provide intelligent risk warnings for the log content. Specifically, this includes: S501. Set a shift handover time node and remind the incoming shift personnel of key items in the target shift handover log; wherein, the key items include delays, temporary construction, and equipment status not restored; Set up standard time node templates that align with the rail transit shift handover process, including reminders for handover preparation, pre-operation reminders, fault follow-up reminders, and receipt reminders. The reminder intervals and trigger conditions can be flexibly adjusted according to the line maintenance rhythm and operational complexity; shift personnel can also manually add custom reminder nodes for special matters. Extract preset core priority items (such as delays, temporary construction, and equipment status not yet restored) from the target shift handover log for reminders, ensuring targeted reminders.

[0059] S502. Based on the content of the target shift handover log, analyze the existing anomalies using a machine learning model, output risk information and issue warnings to the visualization interface; wherein, the machine learning model continuously learns various log errors and accident precursor patterns to continuously optimize the warning capability, sets up a user feedback mechanism, and improves the accuracy of AI judgment.

[0060] A risk identification model is built upon machine learning, trained on massive datasets, with core data sources including historical shift logs, accident case libraries, and business rule libraries. The model utilizes deep learning algorithms to identify three types of risks: direct identification of explicit risks, correlation mining of implicit risks, and prediction of trend risks.

[0061] Risks are categorized into high-risk, medium-risk, and low-risk based on their impact and probability of occurrence. High-risk risks directly threaten driving safety and may cause accidents; medium-risk risks affect operational efficiency and may cause work delays; low-risk risks involve minor non-standard details and do not affect core processes. The visual interface includes a risk warning zone, with different colors indicating the risk level (red = high risk, yellow = medium risk, blue = low risk). Each warning item includes a "risk description + scope of impact + handling suggestions".

[0062] Regularly extract newly added shift handover logs and risk handling results data to incrementally train the risk identification model, continuously learning new log error patterns and accident precursor characteristics. Set up a user feedback entry point where incoming shift personnel and reviewers can label warning results as accurate, misjudged, or omitted, and fill in feedback explanations. Based on user feedback data, adjust model parameter weights, generate model optimization reports, and record changes in indicators such as identification accuracy and warning response speed to ensure continuous improvement in model performance.

[0063] As described in steps S501-S502 above, to ensure timely delivery of shift handover information and proactive response to potential risks, this invention incorporates intelligent reminders and risk warnings. Based on configured parameters such as shift handover deadlines and planned milestones, reminders are sent to users before key time points. A trained machine learning model comprehensively analyzes log content to identify potential risks such as abnormal keywords, recurring faults, and high-frequency construction areas. Data sources such as a shift handover accident case library and a hazard record library are incorporated into the model training process to achieve accident precursor identification and push notifications. Anomaly detection learning is supported, continuously learning and optimizing based on constantly fed-back log data. A user feedback mechanism allows users to rate the accuracy of warnings, continuously improving the model's reliability and practical value.

[0064] This invention features an intelligent semantic preprocessing mechanism driven by multi-source data fusion. It constructs an information collection mechanism integrating multiple data sources (such as scheduling systems, maintenance platforms, safety monitoring platforms, and maintenance window plans), and integrates pre-trained language models like BERT to perform semantic preprocessing on the raw input data. This achieves standardized, structured, semantically normalized, and tokenized input, providing a semantic foundation for the subsequent automatic generation of shift handover logs. This improves the accuracy of semantic recognition and the comprehensiveness of input data during shift handover log generation, significantly reduces the workload of manual data entry and classification, enhances data consistency and semantic understanding depth, and contributes to building a more complete and reliable knowledge base for rail transit operation and maintenance.

[0065] This invention establishes an AI-based automatic generation and context completion mechanism for shift handover logs. It employs a multi-task learning model that integrates semantic classification, template matching, and context completion technologies to construct a multimodal log content generation engine. This engine can automatically match structured templates, identify fields, and generate standardized log statements based on the input semantic content. Simultaneously, it intelligently completes missing content by incorporating historical experience and knowledge graphs. This significantly reduces manual intervention in the shift handover log writing process, improves generation efficiency and content completeness, effectively avoids issues such as information omissions and non-standard terminology, and enhances the consistency and readability of the handover content, providing accurate, clear, and standardized handover information for the incoming shift personnel.

[0066] This invention establishes a semantic verification mechanism oriented towards logical consistency and knowledge matching. It introduces semantic consistency checks and knowledge graph mapping mechanisms to systematically verify the logical structure of statements, knowledge entity relationships, and field specifications in shift handover logs. Combined with a rule engine, it identifies semantic conflicts, omissions, and format deviations, and provides intelligent suggestions. Through automatic semantic verification and knowledge mapping, the accuracy and rigor of shift handover log content are effectively improved, avoiding inconsistencies and illogicalities. This enhances the verifiability and system standardization of log content, contributing to the formation of high-quality rail transit operation and maintenance data assets.

[0067] This invention supports an intelligent log review and push process with closed-loop approval. It designs a multi-level approval process and review and tracking mechanism, allowing different management roles to review handover logs at different levels. It provides functions such as log version comparison and difference highlighting, risk alerts, and manual confirmation, and can automatically archive and push logs to third-party systems via API. This closed-loop log handover process improves log review efficiency and traceability, ensuring that each version of handover information undergoes multi-level confirmation and is digitally archived. This effectively avoids issues such as content tampering and ambiguous responsibility, providing a safe, controllable, and transparent management basis for rail transit operation and maintenance.

[0068] This invention features an AI-driven risk warning and continuous learning mechanism. It integrates an AI risk identification module, which can automatically identify potential risk factors based on historical log data and current shift handover content, issuing warnings and prompting manual confirmation. It also possesses continuous learning capabilities, supporting the optimization of the risk judgment model based on user feedback. This enhances the proactive safety protection capabilities of the shift handover system, enabling early warning and intervention for problems, reducing security risks caused by omissions and misjudgments during operation and maintenance. Through continuous learning and optimization, the system's risk identification accuracy is constantly improved, forming a "self-evolving" intelligent risk control system.

[0069] like Figure 4 As shown, the present invention also provides an artificial intelligence-based intelligent generation and management device for rail transit shift logs. Based on the artificial intelligence-based intelligent generation and management method for rail transit shift logs described above, the device includes: The data acquisition module 1 is used to collect various types of raw handover data and preprocess the raw handover data to generate raw semantic tokens; wherein, the preprocessing includes standardization cleaning, synonym normalization, and format recognition. The generation module 2 is used to process the original semantic token using a preset multimodal log content generation engine to obtain a structured handover log that meets the handover specifications; wherein, the multimodal log content generation engine includes a semantic classification matching unit, a structured filling unit, and a context completion unit; Verification module 3 is used to verify the logical consistency, knowledge association and format standardization of the structured shift log using a preset rail transit operation knowledge graph and logical rules, so as to obtain a draft of the shift log; Approval module 4 is used to predefine multi-level approval processes and, in conjunction with the approval mechanism, to conduct multi-level approvals on the draft handover log to obtain the target handover log for push to external systems; The early warning module 5 is used to remind key items in the target shift handover log by setting time nodes, and to analyze the target shift handover log based on a machine learning model to provide intelligent early warning of risks in the log content.

[0070] Each of the above modules is used to perform the corresponding steps in the above-mentioned AI-based intelligent generation and management method for rail transit shift logs. The specific implementation methods are as described in the above-mentioned method embodiments, and will not be repeated here.

[0071] like Figure 5 As shown, the present invention also provides a computer device, which may be a server, and its internal structure may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores all the data required for the process of intelligent generation and management of rail transit shift logs based on artificial intelligence. The network interface is used for communication with external terminals via a network connection. The computer program is executed by the processor to implement the intelligent generation and management method of rail transit shift logs based on artificial intelligence.

[0072] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment on which the present application is applied.

[0073] An embodiment of this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-described artificial intelligence-based methods for intelligent generation and management of rail transit shift logs.

[0074] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by hardware related to computer program instructions. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the above method embodiments. Any references to memory, storage, databases, or other media provided in this application and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM), such as dynamic RAM (used as main storage) or static RAM (commonly used as cache memory). By way of illustration and not limitation, RAM has various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), and Rambus DRAM (RDRAM).

[0075] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0076] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for intelligent generation and management of rail transit shift logs based on artificial intelligence, characterized in that, include: S1. Collect various types of raw handover data related to shift changes, and preprocess the raw handover data to generate raw semantic tokens; wherein, the preprocessing includes standardization cleaning, synonym normalization, and format recognition. S2. The original semantic token is processed using a preset multimodal log content generation engine to obtain a structured handover log that meets the handover specifications; wherein, the multimodal log content generation engine includes a semantic classification matching unit, a structured filling unit, and a context completion unit. S3. The structured shift handover log is checked for logical consistency, knowledge association and format standardization using a preset rail transit operation knowledge graph and logical rules to obtain a draft shift handover log. S4. Predefine a multi-level approval process and combine it with the approval mechanism to conduct multi-level approval of the handover log draft, so as to obtain the target handover log and push it to the external system; S5. By setting time nodes, key items in the target shift handover log are reminded, and the target shift handover log is analyzed based on a machine learning model to provide intelligent risk warnings for the log content.

2. The method for intelligent generation and management of rail transit shift logs based on artificial intelligence according to claim 1, characterized in that, S1 specifically includes: S101. Manually input data through a visual interface or collect various types of original handover data related to shift changes by connecting to a data source interface; wherein, the data source interface includes a scheduling system, production planning system, maintenance platform, safety supervision platform or big data platform, and the original handover data includes the current shift's work plan, handover content, construction plan, maintenance window arrangement, fault work order, and temporary work instructions. S102. The natural language preprocessing function based on the pre-trained language model performs standardization cleaning, synonym normalization, and format recognition on the original handover data, and generates original semantic tokens for subsequent analysis; wherein, the pre-trained language model includes BERT and RoBERTa.

3. The method for intelligent generation and management of rail transit shift logs based on artificial intelligence according to claim 2, characterized in that, S2 specifically includes: S201. The semantic classification and matching unit classifies the original semantic tokens according to a preset machine learning model and generates standard statements based on log generation templates; wherein, the categories include operating status, plan changes, equipment failure, and security risks; S202. The structured filling unit categorizes and fills the standard statements according to log fields, and outputs structured JSON format log items. S203. The context completion module intelligently completes incomplete or vaguely described content in the structured JSON format log items based on historical shift experience logs and knowledge graphs to obtain a structured shift handover log that meets the handover specifications.

4. The method for intelligent generation and management of rail transit shift logs based on artificial intelligence according to claim 3, characterized in that, S3 specifically includes: S301. Based on the AI ​​model, verify the logical consistency of the internal descriptions of the structured handover log content, mark potentially conflicting items, and prompt for manual confirmation; S302. Establish entity relationships between the content of the structured shift handover log and the preset rail transit operation knowledge graph to identify missing items and content deviations; wherein, the rail transit operation knowledge graph includes train plans, construction arrangements, fault types, and shift handover specifications; S303. Combine the preset logical rules to check whether there are logical conflicts in the content of the structured handover log, and at the same time standardize the language style and field format of the structured handover log.

5. The method for intelligent generation and management of rail transit shift logs based on artificial intelligence according to claim 4, characterized in that, S4 specifically includes: S401. Predefine multi-level approval processes based on approval nodes, and introduce AI-assisted review in the approval nodes to mark potentially high-risk items for focused review; S402. At the approval node, the current version of the handover log draft is compared with the historical version, and the differences are highlighted. A signing confirmation mechanism is set up to determine that the approval at the current approval node is completed. S403. After completing the approval process according to the multi-level approval process, the target handover log is generated and archived in multiple formats, and pushed to the required external system through the interface; among them, the multiple formats include PDF, JSON and CSV.

6. The method for intelligent generation and management of rail transit shift logs based on artificial intelligence according to claim 5, characterized in that, S5 specifically includes: S501. Set a shift handover time node and remind the incoming shift personnel of key items in the target shift handover log; wherein, the key items include delays, temporary construction, and equipment status not restored; S502. Based on the content of the target shift handover log, analyze the existing abnormal situations using a machine learning model, output risk information and issue warnings to the visualization interface; wherein, the machine learning model continuously learns various log errors and accident precursor patterns to continuously optimize the warning capability.

7. An intelligent generation and management device for rail transit shift logs based on artificial intelligence, based on the intelligent generation and management method for rail transit shift logs based on artificial intelligence as described in any one of claims 1-6, characterized in that, The device includes: The data acquisition module is used to collect various types of raw handover data and preprocess the raw handover data to generate raw semantic tokens; wherein, the preprocessing includes standardization cleaning, synonym normalization, and format recognition. The generation module is used to process the original semantic token using a preset multimodal log content generation engine to obtain a structured handover log that meets the handover specifications; wherein, the multimodal log content generation engine includes a semantic classification matching unit, a structured filling unit, and a context completion unit; The verification module is used to verify the logical consistency, knowledge association and format standardization of the structured shift log using a preset rail transit operation knowledge graph and logical rules, so as to obtain a draft of the shift log. The approval module is used to predefine multi-level approval processes and, in conjunction with the approval mechanism, to conduct multi-level approvals on the draft handover log to obtain the target handover log for push to external systems. The early warning module is used to remind users of key matters in the target shift handover log by setting time nodes, and to analyze the target shift handover log based on a machine learning model to provide intelligent early warning of risks in the log content.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 16.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images