A railway electric service engineering management method, device, electronic equipment and program product

Abstract

The application relates to the technical field of railway electric power engineering, and discloses a railway electric power engineering management method and device, an electronic device and a program product. The method is based on a situation awareness module of a train operation protection intelligent agent, regularly acquires train position information, train speed information and state information of a construction section, the state information of the construction section includes an occupation state and a safety state; based on the train position information, the train speed information and the state information of the construction section, the spatial overlap length and the time overlap section of a train operation section and the construction section are calculated; risk judgment is performed based on the spatial overlap length, the time overlap section and the train speed information, and construction safety early warning information is generated; and the construction cooperation intelligent agent can generate railway electric power management information based on the construction safety early warning information. In this way, the management efficiency of railway electric power can be effectively improved, the safety of railway electric power engineering construction operation is ensured through intelligent early warning of risk preposition.

Description

Technical Field

[0001] This application relates to the field of railway electrical engineering technology, specifically to a railway electrical engineering management method, device, electronic equipment, and program product. Background Technology

[0002] Railway electrical engineering is the core of ensuring train operation safety and control. It covers key facilities such as signaling, communication, interlocking, train control, track circuits, turnout switching equipment and indoor and outdoor cables. It is characterized by numerous points and long lines, dense work near operating lines, complex cross-disciplinary collaboration, and high safety requirements.

[0003] Currently, railway electrical engineering management still relies mainly on manual verification, manual filling, and manual scheduling. In scenarios where construction and train operation plans conflict, it is difficult to accurately determine the spatiotemporal overlap between train operation sections and construction sections through manual means, resulting in low efficiency in railway electrical engineering management. At the same time, it is impossible to accurately ensure the safety of construction workers entering the line.

[0004] It is evident that the railway electrical engineering management schemes in related technologies suffer from poor management efficiency and inadequate safety performance during construction operations. Summary of the Invention

[0005] The present invention aims to solve the problems of poor management efficiency and poor safety performance of railway electrical engineering management schemes in related technologies, and provides a railway electrical engineering management method, device, electronic equipment and program product.

[0006] To achieve the above objectives, the first aspect of this application provides a railway electrical engineering management method, applied to an engineering construction management intelligent agent system. The engineering construction management intelligent agent system includes a train operation protection intelligent agent, a construction collaboration intelligent agent, and a mobile terminal intelligent agent. The train operation protection intelligent agent, the construction collaboration intelligent agent, and the mobile terminal intelligent agent all adopt a hybrid architecture of BDI model and finite state machine, and coordinate through an event-driven bus. The method includes:

[0007] Based on the situational awareness module of the aforementioned train protection intelligent agent, train location information, train speed information, and construction section status information are periodically acquired. The train speed information is used to characterize the train's travel speed, and the construction section status information includes occupancy status and safety status. Based on the train location information, the train speed information, and the status information of the construction section, calculate the spatial overlap length and temporal overlap interval between the train operation section and the construction section; Risk assessment is performed based on the spatial overlap length, the temporal overlap interval, and the train speed information to generate construction safety early warning information, which is then sent to the construction collaborative intelligent agent and the mobile terminal intelligent agent. The construction collaborative intelligent agent can generate railway electrical management information based on the construction safety early warning information. The railway electrical management information includes at least one of the following: construction plan adjustment information to achieve the safe evacuation of construction personnel and train operation adjustment information.

[0008] A second aspect of this application provides a railway electrical engineering management device applied to an engineering construction management intelligent agent system. The engineering construction management intelligent agent system includes a train operation protection intelligent agent, a construction collaboration intelligent agent, and a mobile terminal intelligent agent. The train operation protection intelligent agent, the construction collaboration intelligent agent, and the mobile terminal intelligent agent all adopt a hybrid architecture of BDI model and finite state machine, and coordinate through an event-driven bus. The device includes: The information acquisition module is used to periodically acquire train location information, train speed information, and construction section status information based on the situational awareness module of the train protection intelligent agent. The train speed information is used to characterize the train's travel speed, and the construction section status information includes occupancy status and safety status. The calculation module is used to calculate the spatial overlap length and temporal overlap interval between the train operating section and the construction section based on the train position information, the train speed information, and the status information of the construction section. The risk assessment module is used to assess risks based on the spatial overlap length, the temporal overlap interval, and the train speed information, generate construction safety early warning information, and send the construction safety early warning information to the construction collaborative intelligent agent and the mobile terminal intelligent agent. The generation module is used by the construction collaborative intelligent agent and the mobile terminal intelligent agent to generate railway electrical management information based on the construction safety early warning information. The railway electrical management information includes at least one of the following: construction plan adjustment information to achieve the safe evacuation of construction personnel and train operation adjustment information.

[0009] A third aspect of this application provides an electronic device including a processor and a memory, wherein the memory stores a program or instructions executable on the processor, the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect.

[0010] A fourth aspect of this application provides a computer program product stored in a storage medium, which is executed by at least one processor to perform the steps of the method as described in the first aspect.

[0011] Compared with the prior art, this application has the following beneficial effects: The railway electrical engineering management method provided in this application adopts a train operation protection intelligent agent with a hybrid architecture of BDI model and finite state machine, combined with an event-driven bus collaborative mechanism. This allows for comprehensive perception of train location information, train speed information, and multi-dimensional status information regarding the occupancy and safety of construction sections. By quantitatively calculating the spatial and temporal overlap lengths between train operating sections and construction sections, and combining this with train speed information for multi-dimensional and accurate risk classification, the method achieves early intelligent warning of construction-train intersection conflicts. This effectively improves the management efficiency of railway electrical engineering, and more importantly, ensures the safety of railway electrical engineering construction operations through early intelligent risk warning. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating a railway electrical engineering management method is shown schematically. Figure 2 The diagram schematically illustrates a modular design, construction, operation, and maintenance integrated digital platform for railway electrical systems. Figure 3 A schematic diagram of a design subsystem is shown. Figure 4 The diagram illustrates a station-level statistical table. Figure 5 This diagram schematically illustrates the signal equipment and cable layout of a three-dimensional train control system implemented using a design subsystem. Figure 6 This schematic diagram illustrates one of the module diagrams of an intelligent agent system for engineering construction management. Figure 7 This schematic diagram illustrates the second module diagram of an intelligent engineering construction management system. Figure 8 This diagram schematically illustrates a network diagram of an intelligent agent system for engineering construction management. Figure 9 This diagram schematically illustrates a module diagram of an operations and maintenance subsystem. Figure 10 This diagram illustrates a modeling schematic of an intelligent agent system. Figure 11 This illustration schematically shows a formal definition and component design diagram of a BDI model; Figure 12 The diagram illustrates the interaction flow between the operation and maintenance subsystem, the design subsystem, and the railway train control system. Detailed Implementation

[0014] To facilitate understanding of this application, the following description will be more comprehensive and detailed in conjunction with the accompanying drawings and preferred embodiments, but the scope of protection of this application is not limited to the following specific embodiments.

[0015] Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by those skilled in the art. The technical terms used herein are for the purpose of describing particular embodiments only and are not intended to limit the scope of this application.

[0016] Please see Figure 1 This application provides a railway electrical engineering management method, applied to an engineering construction management intelligent agent system. The system includes a train operation protection intelligent agent, a construction collaboration intelligent agent, and a mobile terminal intelligent agent. All three agents—train operation protection intelligent agent, construction collaboration intelligent agent, and mobile terminal intelligent agent—adopt a hybrid architecture combining a BDI model and a finite state machine, and coordinate through an event-driven bus. The method can be executed by electronic devices such as mobile phones and computers, and includes the following steps: S1. Based on the situational awareness module of the train protection intelligent agent, the train position information, train speed information and construction section status information are periodically acquired. The train speed information is used to characterize the train's travel speed, and the construction section status information includes occupancy status and safety status. S2. Based on the train location information, the train speed information, and the status information of the construction section, calculate the spatial overlap length and temporal overlap interval between the train operation section and the construction section; S3. Based on the spatial overlap length, the temporal overlap interval, and the train speed information, a risk assessment is performed, construction safety early warning information is generated, and the construction safety early warning information is sent to the construction collaborative intelligent agent and the mobile terminal intelligent agent. S4. Generation module, used by the construction collaborative intelligent agent and the mobile terminal intelligent agent to generate railway electrical management information based on the construction safety early warning information. The railway electrical management information includes at least one of the following: construction plan adjustment information to achieve safe evacuation of construction personnel and train operation adjustment information.

[0017] In some embodiments, the aforementioned spatial overlap length refers to the mileage range of the train running line section and the mileage range of the railway signaling construction section that overlap in the longitudinal direction of the line, and the unit can be kilometers; in other words, it is to determine whether the construction section is on the train's travel path.

[0018] In some embodiments, the train running section and the construction occupied section can be calculated in milliseconds. By comparing the endpoints and quantifying the overlap length, the missed detection and false detection caused by the ambiguity of the section boundary and the mileage error can be completely avoided. At the same time, Beidou positioning can be introduced to calibrate the train position in real time, correct the mileage deviation, and improve the spatial detection accuracy so that the obtained spatial overlap length can accurately characterize whether the construction section is on the train's travel path.

[0019] In some embodiments, the train running time and construction occupation time are bidirectionally verified, the time overlap interval is calculated, and misjudgments caused by instantaneous time fluctuations and data synchronization delays are eliminated; at the same time, the real-time time series data collected by the situational awareness module is correlated to dynamically correct the time window and ensure the accuracy of time matching.

[0020] In some embodiments, the vehicle location information, train speed information, and construction section status information can be updated every 100 seconds to ensure the real-time nature of the train's trajectory and construction occupation status, allowing sufficient time for early warning.

[0021] In some embodiments, the mobile terminal intelligent agent can collect personnel location, trajectory, on-site images, and electronic fence intrusion status in real time through intelligent safety helmets and on-site operation terminals; the data can be transmitted back to the train protection intelligent agent in real time as the basis for personnel lingering judgment, early warning triggering, and evacuation verification, so as to achieve precise personnel protection and improve the safety of railway electrical engineering operations.

[0022] In some embodiments, the status information of the construction section can be obtained through a construction collaborative intelligent agent. That is, the construction collaborative intelligent agent can obtain the corresponding construction information such as construction scope, construction period, work permit, and section occupancy, and generate the status information of the construction section based on the obtained construction information. The generated status information of the construction section can be sent to the train protection intelligent agent so that the train protection intelligent agent can make risk assessment based on the obtained train position information, train speed information and the status information of the construction section.

[0023] In some embodiments, construction safety early warning information generated by the train safety protection intelligent agent is sent to the construction collaborative intelligent agent, so that the construction collaborative intelligent agent can generate corresponding railway electrical management information based on the received construction safety early warning information, thereby improving the safety management of railway electrical systems. The railway electrical management information includes at least one of construction plan adjustment information and train operation adjustment information.

[0024] In some embodiments, construction plan adjustment information can be understood to include at least information on adjusting the construction plan for the safe evacuation of construction personnel, and may also include other instructions, such as adjusting the construction by suspending construction and transferring construction personnel and facilities, to ensure the normal operation of the train. After the train has passed, the construction plan adjustment information can be further adjusted to continue construction, thereby ensuring a balance between safety and construction efficiency.

[0025] In some embodiments, if the train is traveling at normal speed but construction facilities and personnel cannot be moved within a preset time, train operation adjustment information can be sent to the train so that the train can adjust its operating speed based on the train operation adjustment information, such as reducing the train's speed until the train waits in a safe area, so as to allow sufficient time for the construction facilities and personnel to be moved, thereby improving the safety of railway electrical engineering operations.

[0026] The railway electrical engineering management method provided in this application, through the use of a train operation protection intelligent agent with a hybrid architecture of BDI model and finite state machine, combined with an event-driven bus collaborative mechanism, can comprehensively perceive train location information, train speed information, and multi-dimensional status information such as occupancy and safety of construction sections. Furthermore, by quantitatively calculating the spatial and temporal overlap lengths between train operation sections and construction sections, and combining this with train speed information for multi-dimensional and accurate risk classification, it achieves early intelligent warning of construction-train intersection conflict risks. This effectively improves the management efficiency of railway electrical engineering, and more importantly, ensures the safety of railway electrical engineering construction operations through early intelligent risk warning.

[0027] In this application, the vehicle protection intelligent agent adopts a hybrid architecture of BDI model and finite state machine, which can realize the parallel execution of conflict detection and early warning triggering without waiting for other tasks to complete. At the same time, through the event-driven bus, the construction safety early warning information is quickly pushed to the vehicle protection intelligent agent and mobile terminal intelligent agent of the engineering construction management intelligent agent system, avoiding information transmission delay and further ensuring the early warning time.

[0028] In some embodiments, the step of determining risk based on the spatial overlap length, the temporal overlap interval, and the train speed information, and outputting construction safety early warning information, includes: If the spatial overlap length is greater than zero and the time overlap interval is completely contained, a first-level construction safety early warning message is output. The complete inclusion of the time overlap interval is used to indicate that the construction process of the construction section is within the time period covered by the train's operation. If the spatial overlap length is greater than zero, the time overlap interval duration is greater than or equal to the preset duration threshold, and the time overlap interval is not completely contained, a secondary construction safety early warning message is output. If the spatial overlap length is greater than zero and the duration of the temporal overlap interval is less than a preset duration threshold, a level three construction safety early warning message is output.

[0029] In this embodiment, risk assessment is performed by combining multiple parameters such as spatial overlap length, temporal overlap interval, and train speed to achieve graded and precise early warning of conflicts between railway line construction and train operation risks. A Level 1 warning is specifically designated for extremely high-risk conditions with complete spatial and temporal overlap, allowing for priority management of major safety hazards that cover the entire train passage window during construction. A Level 2 warning is issued for long-term spatial and temporal overlap scenarios, and a Level 3 warning is issued for short-term spatial and temporal overlap scenarios, enabling accurate identification of transient overlap risks.

[0030] The preset speed can be set to 120km / h.

[0031] By adopting the railway electrical engineering management method of this application, the accuracy rate of collision detection between trains and construction sections can reach over 99.5%.

[0032] In some embodiments, all levels of construction safety early warning information meet the pre-triggering conditions: Based on the remaining distance between the train and the construction section, and the train's speed, the arrival time is calculated, which is used to characterize the time required for the train to arrive at the construction section. If the arrival time is greater than or equal to a preset time, construction safety warning information at all levels will be triggered accordingly, and the preset time is greater than or equal to 2 minutes.

[0033] In this embodiment, a logic for determining the pre-arrival time of the train in the construction section is added. This logic accurately calculates the time required for the train to reach the construction section based on the remaining distance and real-time speed, and strictly limits the pre-warning trigger window to no less than 2 minutes, providing sufficient safety response buffer time for all levels of construction safety warnings. This setup ensures that before the train enters the dangerous construction area, sufficient time is reserved for construction hazard avoidance, train speed reduction, on-site emergency response, and personnel evacuation, completely avoiding safety accidents caused by hasty alarms near the point of conflict and insufficient time to avoid risks on-site.

[0034] In some embodiments, the greater the travel speed, the greater the weight of the train speed information in the risk assessment of the output of the construction safety early warning information.

[0035] Understandably, under high-speed conditions, vehicle braking distance is longer, the window for hazard avoidance is narrower, and the collision damage and accident risk increase dramatically. In this embodiment, by simultaneously amplifying the judgment weight of the speed dimension, the risk sensitivity in high-speed driving scenarios can be accurately enhanced, avoiding excessive warnings at low and normal speeds, while eliminating safety loopholes such as underestimating risks and low warning levels in high-speed and high-risk conditions.

[0036] This application also provides a railway electrical engineering management device, applied to an engineering construction management intelligent agent system. The engineering construction management intelligent agent system includes a train operation protection intelligent agent, a construction collaboration intelligent agent, and a mobile terminal intelligent agent. The train operation protection intelligent agent, the construction collaboration intelligent agent, and the mobile terminal intelligent agent all adopt a hybrid architecture of BDI model and finite state machine, and coordinate through an event-driven bus. The device includes: The information acquisition module is used to periodically acquire train location information, train speed information, and construction section status information based on the situational awareness module of the train protection intelligent agent. The train speed information is used to characterize the train's travel speed, and the construction section status information includes occupancy status and safety status. The calculation module is used to calculate the spatial overlap length and temporal overlap interval between the train operating section and the construction section based on the train position information, the train speed information, and the status information of the construction section. The risk assessment module is used to assess risks based on the spatial overlap length, the temporal overlap interval, and the train speed information, generate construction safety early warning information, and send the construction safety early warning information to the construction collaborative intelligent agent and the mobile terminal intelligent agent. The generation module is used by the construction collaborative intelligent agent to generate railway electrical management information based on the construction safety early warning information. The railway electrical management information includes at least one of the following: construction plan adjustment information to achieve the safe evacuation of construction personnel and train operation adjustment information.

[0037] Optionally, the risk assessment module is specifically used for: If the spatial overlap length is greater than zero and the time overlap interval is completely contained, a first-level construction safety early warning message is output. The complete inclusion of the time overlap interval is used to indicate that the construction process of the construction section is within the time period covered by the train's operation. If the spatial overlap length is greater than zero, the time overlap interval duration is greater than or equal to the preset duration threshold, and the time overlap interval is not completely contained, a secondary construction safety early warning message is output. If the spatial overlap length is greater than zero and the duration of the temporal overlap interval is less than a preset duration threshold, a level three construction safety early warning message is output.

[0038] Optionally, before triggering construction safety early warning information at all levels, the following may also be included: Based on the remaining distance between the train and the construction section, and the train's speed, the arrival time is calculated, which is used to characterize the time required for the train to arrive at the construction section. If the arrival time is greater than or equal to a preset time, construction safety warning information at all levels will be triggered accordingly, and the preset time is greater than or equal to 2 minutes.

[0039] Optionally, the greater the travel speed, the greater the weight of the train speed information in the risk assessment of the construction safety early warning information output.

[0040] The railway electrical engineering management device provided in this application can realize the various processes of the railway electrical engineering management method embodiment of this application and achieve the same beneficial effects. To avoid repetition, it will not be described again here.

[0041] Please see Figure 2 This application also provides a digital platform integrating railway signaling design, construction, operation, and maintenance, including a design subsystem, the aforementioned engineering construction management intelligent system, and an operation and maintenance subsystem; wherein: The design subsystem is used to generate 3D diagrams and bills of quantities based on the input station layout map, line data, and signal equipment parameters, and then send the 3D diagrams and bills of quantities to the engineering construction management intelligent agent system and the operation and maintenance subsystem. The design subsystem is also used to receive contact forms from the engineering construction management intelligent agent system, generate as-built drawings based on the contact forms, and send the as-built drawings to the operation and maintenance subsystem. Furthermore, the design subsystem is used to receive alarm information from the operation and maintenance subsystem, generate fault handling operation instructions based on the alarm information, and send the fault handling operation instructions to maintenance personnel to perform maintenance operations. The engineering construction management intelligent system is used to decompose the bill of quantities based on the engineering station work teams, section work teams and work site quantities on site, carry out on-site construction based on the decomposition results and the 3D drawings, and send the contact forms during the construction process to the design subsystem after the construction is completed. The operation and maintenance subsystem is used to monitor the construction status during construction and the failures of all equipment and infrastructure during operation online. When construction and equipment operation abnormalities are detected, alarm information is generated and sent to the design subsystem. The operation and maintenance subsystem is also used to receive maintenance work results from maintenance personnel and archive the information.

[0042] In some embodiments, the engineering construction management intelligent agent system includes a vehicle safety intelligent agent, a construction collaboration intelligent agent, a progress and investment intelligent agent, a disaster prevention and control intelligent agent, an operation and maintenance diagnosis intelligent agent, and a mobile terminal intelligent agent; wherein, the vehicle safety intelligent agent, the construction collaboration intelligent agent, and the mobile terminal intelligent agent based on the engineering construction management intelligent agent system can achieve the following: Figure 1 The various processes of the railway electrical engineering management method embodiment shown, as well as the methods for achieving the same beneficial effects, will not be repeated here to avoid repetition.

[0043] Specifically, in some feasible ways, the six types of intelligent agents and their responsibility boundaries are shown in Table 1 below: Table 1. Six types of intelligent agents and their boundaries of responsibility

[0044] In some embodiments, the mobile terminal intelligent agent serves as the data entry point, collecting real-time personnel and machinery locations and on-site images to provide a basic data source for all modules. The vehicle protection intelligent agent acts as the safety red line: it issues vehicle information to the construction collaboration intelligent agent to guide the construction operation area (electronic fence) and strictly prevent encroachment conflicts; it receives equipment fault warnings from the maintenance diagnosis intelligent agent, dynamically adjusts protection strategies, and simultaneously adjusts relevant schedules and funding nodes for the progress investment intelligent agent. The construction collaboration intelligent agent acts as the planning hub: combining early warning information from vehicle protection and disaster prevention linkage, it simulates and adjusts construction plans in real time; it provides feedback on actual construction progress to the progress investment intelligent agent, supporting precise control of progress and funding. The progress investment intelligent agent acts as the resource core: based on the progress and status of construction, vehicle operation, and maintenance, it coordinates WBS nodes and contract payments; it links with the maintenance diagnosis intelligent agent to predict maintenance investment and ensure the orderly progress of the project and equipment. The disaster prevention linkage intelligent agent acts as an emergency backup: once a disaster warning such as weather or displacement is triggered, it immediately synchronizes with the construction, vehicle, and mobile terminals to initiate emergency evacuation and protection adjustments for all personnel and equipment. Operation and maintenance diagnostic intelligent agents serve as the cornerstone of equipment: By analyzing the time-series data of equipment collected by mobile terminals, fault trends are predicted and root causes are located. Equipment health status is synchronized with vehicle safety (adjusting driving strategies) and schedule investment (adjusting maintenance schedules and costs) to ensure vehicle safety and controllable total project costs. As can be seen, the six intelligent agents in the above solution are interconnected: the mobile terminal intelligent agent serves as the on-site data entry point; the vehicle safety intelligent agent acts as the safety baseline; the construction collaboration intelligent agent connects the implementation of plans; the schedule investment intelligent agent manages resource input; the disaster prevention linkage intelligent agent mitigates external risks; and the operation and maintenance diagnosis intelligent agent ensures equipment infrastructure. These six intelligent agents are interlinked and data flows bidirectionally, achieving a closed-loop collaborative process encompassing on-site data collection, vehicle safety, construction coordination, schedule investment, disaster prevention and emergency response, and equipment operation and maintenance. This maximizes the overall efficiency of ensuring operational safety, vehicle safety, and project management.

[0045] In one embodiment, such as Figure 3 As shown, the design subsystem includes: The drawing recognition and extraction module is used to extract work site information, material information, and building element information from station micromaps, line data, and signal equipment parameters using the YOLOv11 model, and generate graphic elements and text attribute information. The design data processing module is used to generate signal design drawings, wiring tables, data tables, and bills of quantities based on the graphic elements and text attribute information. The station three-dimensional generation module includes a workshop module for generating workshop statistical tables and a station module for generating station statistical tables, wherein the workshop module and the station module are communicatively connected. The design document generation module is used to generate 3D diagrams based on the signal design drawings, wiring tables, data tables, bill of quantities, workshop statistics tables, and station statistics tables.

[0046] In this embodiment, the workshop module compiles information within its jurisdiction, including the section plan, section cable wiring diagram, section layout data table, section cable route diagram, section rack cabinet layout diagram, logic check circuit, and train empty data table, to generate a workshop statistical table. The station module compiles information within the station area, including the interlocking indoor diagram, interlocking outdoor diagram, coded indoor diagram, coded outdoor diagram, interlocking table, train control data table, transponder data table, and route data table, to form a station-level station statistical table. Details are as follows: Figure 4 As shown.

[0047] It is worth noting that the design subsystem adopts a Master Data Management (MDM) system based on Service-Oriented Architecture (SOA) to implement the encoding, publishing, cleaning, integration, sharing, and governance functions of master data, including turnout control circuits, signal and lighting circuits, and side wiring tables. It is also used to create, edit, import, and export basic files, including plan views, standard diagrams, and side wiring tables. Using a standard diagram segment association method, signaling equipment in the plan view is abstracted into directional nodes. Combining the basic data dictionary and model database, and based on the topology of the plan view, device relationships are searched and matched, outputting an association diagram including circuit diagrams and wiring tables.

[0048] The system uses an SQL Server database and a WebService platform as a unified database access interface. It utilizes virtualization technology to pool hardware resources. The server-side employs a Windows cluster architecture with primary / standby redundancy failover. The client-side uses remote desktop and terminal cluster load balancing technology to enable access to a unified system client interface via a dedicated system network. By leveraging virtualization technology, the system integrates server, network, and other hardware resources into a dynamically managed "resource pool," improving the utilization of CPU, memory, and disk I / O resources. The server-side deployment in a Windows cluster architecture with automatic primary / standby node redundancy failover ensures high system availability.

[0049] The design subsystem receives external station layout maps, track data, and signal equipment parameters. Through internal processing, it generates 3D models and detailed bills of quantities (including equipment quantities and cable lengths), which are then sent to the engineering construction management intelligent agent system (for guiding construction and material procurement) and the operation and maintenance subsystem (as foundational data for operation and maintenance). Conversely, when changes occur during construction, the design subsystem receives a contact form from the engineering construction management intelligent agent system, automatically updates the design model based on the changes, generates accurate as-built drawings, and transfers them to the operation and maintenance subsystem. Furthermore, during the operation phase, the design subsystem receives alarm information (such as a red light on a track circuit) reported by the operation and maintenance subsystem in real time. By linking design data with a fault knowledge base, it automatically generates fault handling instructions containing fault location, handling steps, and required tools and materials, and pushes these instructions to the smart terminals of designated maintenance personnel. In this way, by using the YOLOv11 model to achieve intelligent and automatic extraction of drawing information, and combining the collaborative modules of design data processing, station standardization ("generalized engineering design", "factory-based equipment production", and "plug-in on-site construction") generation, and design document generation, the entire process of railway electrical design, from drawing analysis, data processing, statistical reports to 3D model generation, is automated and intelligent. The standardization of train control system modules, the standardized combination of the "three-standardization" modules, and the classification and connection of various cables between the train control cabinet, frequency shift cabinet, and integrated cabinet, significantly improve design efficiency and accuracy, reduce labor costs and error rates, and provide a precise and standardized digital twin data foundation for the entire construction and operation process, further strengthening the integrated management and control capabilities and technological advancement of the platform of this invention.

[0050] The design subsystem automatically generates general-type combinations corresponding to the controlled equipment for splicing and configuration diagrams, and uses charts to show equipment names, arrangement positions, and interface connection directions, improving work efficiency and ensuring design quality. The design subsystem automatically exports production assembly drawings for all equipment, including standardized combinations, rack wiring, bundled cables, and connectors. All wiring work is tested and verified in the factory. The design subsystem's implementation of the "three-in-one" train control system signal equipment and cable layout diagram is shown below. Figure 5 As shown, the design subsystem generates the configuration rules and manufacturing standards for the three-standardized cables. The design subsystem automatically exports Table 2 below, realizing the configuration and production of the three-standardized standardization modules. This significantly improves the efficiency of engineering construction and shortens the data center implementation time by approximately 50%.

[0051] Table 2. Rules for the Configuration and Manufacturing Standards of Three-Cable Generation by the Design Subsystem

[0052] In one embodiment, such as Figure 6 As shown, the engineering construction management intelligent agent system includes: 1. Vehicle monitoring module This module achieves real-time data synchronization through an interface with the existing CTC / TDCS (Train Dispatch and Command System) query system.

[0053] Data synchronization includes: real-time acquisition of train timetables, phase plans, actual timetables, and train operation information from the CTC query system. The system establishes virtual electronic fences at stations and in sections, and automatically pushes early warnings to associated construction protection terminals before a train approaches a construction site.

[0054] The plan integration includes: performing collision detection on driving plans and construction plans, automatically identifying plan conflicts, and highlighting the conflict risk on the dispatch screen.

[0055] 2. Construction Plan Module Achieve full-process electronic closed-loop management of construction application and cancellation points.

[0056] Electronic application and approval: Construction units submit construction plans (including scope of impact, work content, personnel and machinery configuration) through mobile terminals. The system has a built-in conflict detection algorithm that automatically compares the plans with the driving plans to assist managers in the approval process.

[0057] Visual simulation, combining BIM model and GIS map, performs 3D time simulation of large or complex construction plans to predict the impact on traffic clearance during construction.

[0058] For point-of-sale management, after construction is completed, the on-site supervisor confirms via mobile terminal that personnel and equipment have been evacuated to a safe area. The system automatically compares the remaining items within the electronic fence, and once confirmed to be correct, the point of sale is cleared and the traffic restrictions are lifted.

[0059] In addition, the intelligent agent system for engineering construction management also includes: 3. Project Schedule Module Establish a work breakdown structure tree based on lines, sections, unit projects, and sub-items of projects.

[0060] It supports directly importing Gantt charts generated by planning software such as Project (P6) into GIS maps, enabling a visual presentation of the plan. For example, on a single map, lines of different colors represent the planned duration of different work sites.

[0061] The system automatically collects and reports on-site progress via mobile devices. On-site technicians use mobile terminals to report the completion status of key processes daily or weekly using a "point-and-click + photo" method (e.g., 5km of fiber optic cable laid today). The system automatically links to hidden works image data to ensure the traceability of progress data.

[0062] The system automatically plots a "progress leader line" on the planning Gantt chart based on real-time data submitted from the site. Managers can then easily identify which work sites are behind schedule and by how many days on the system or the command center's large screen.

[0063] Automatic early warning: If the actual progress of a certain process is later than the planned progress and the deviation exceeds the set threshold (such as 3 days), the system will automatically trigger an early warning and push suggestions to the project manager and chief engineer to expedite the work.

[0064] The 3D visual progress display combines the BIM model with a timeline. By dragging the time slider on the large visual scheduling screen, you can see the construction status of bridges, tunnels, roadbeds, laying of electrical and electronic cables, installation of electrical cabinets, commissioning of electrical systems, and implementation of interfaces between electrical systems and the station front at different times.

[0065] 4. Investment Management Module Investment breakdown and contract linkage. Preliminary budget breakdown involves decomposing the approved construction drawing budget according to a Work Breakdown Structure (WBS) to form the project's "cost budget red line."

[0066] Contract linking associates each construction contract and procurement contract with its corresponding WBS node. The system automatically summarizes the total contract price, measured amount, and paid amount, forming a dynamic investment ledger.

[0067] The investment monitoring section, presented in a single chart, allows for real-time comparison of three curves: 1) Contract budget line, which is theoretically how much each WBS node should cost.

[0068] 2) Completed production line (payable): The cumulative amount payable is automatically calculated based on the "actual completed work volume" filled in by the project schedule module and the contract unit price.

[0069] 3) Actual payment line (actual payment): The amount actually paid by the financial system.

[0070] When the "completed output value" exceeds the "contract budget" or the "actual payment" exceeds the "completed output value", the system will automatically issue a red warning.

[0071] Dynamic cash flow forecasting: Based on the monthly funding plan and the project schedule for the next 1-3 months, the system automatically predicts the amount of money to be paid to construction companies and material suppliers in the following month. The finance department can then use this information to raise funds in advance and achieve "expenditure based on revenue".

[0072] Cash flow analysis predicts the project's cash flow trends throughout its entire lifecycle, helping the command center plan and allocate funds in advance. The system issues a warning when cumulative payments approach 90% of the contract price; at 95%, it mandates final settlement procedures.

[0073] 5. Disaster Prevention Module Integrate multi-source data to build a situational awareness capability for construction safety risks.

[0074] External early warning access is provided through an interface with the disaster prevention center, enabling real-time access to early warning information for disasters such as typhoons, rainstorms, and landslides.

[0075] Threshold warning: Based on displacement monitoring data obtained from Beidou reference stations, it monitors the settlement and displacement of high embankments, deep foundation pits, and adjacent existing railway line subgrades at the millimeter level, and immediately alarms when the threshold is exceeded.

[0076] In a coordinated response, the system automatically matches emergency plans based on the warning level and issues stop-work orders or evacuation notices to construction managers and supervisors in the affected areas with a single click through the visual dispatch module.

[0077] 6. Video surveillance module Construct a comprehensive video perception system that combines fixed and mobile sensors.

[0078] Fixed monitoring is implemented by deploying high-definition PTZ cameras at key construction sites such as mixing plants, steel processing plants, tunnel entrances, and overpasses, and connecting them to the platform to enable remote supervision of construction processes and safe and civilized construction practices.

[0079] Mobile monitoring equips on-site safety officers and supervisors with law enforcement recorders, installs vehicle-mounted video on large machinery, and transmits the data back via 5G network to achieve on-site control from a mobile perspective.

[0080] AI recognition, with AI algorithms configured on the backend, automatically identifies violations or abnormal behaviors such as not wearing a safety helmet, climbing over guardrails, smoke, and flames, and captures images to trigger an alarm.

[0081] 7. Operational communication and protection module This addresses the pain point of "being able to hear but not see" in traditional protection, achieving dual protection through a combination of human and technological defenses.

[0082] BeiDou precise positioning, relying on the BeiDou ground-based augmentation reference stations built along the route, provides real-time centimeter-level and post-event millimeter-level positioning services for on-site safety personnel, construction supervisors, and large machinery. Mobile terminals can display the wearer's specific location on a particular track and at a specific mileage in real time.

[0083] The electronic protection system utilizes BeiDou positioning and GIS electronic fences to construct a "virtual protective wall." Once construction personnel or machinery intrude into the vehicle clearance or operate beyond the designated area, the terminal immediately triggers an audible and visual alarm.

[0084] The system triggers a proximity alarm, automatically identifying individuals in danger zones based on train approach information and real-time personnel locations. A voice announcement is then broadcast via the terminal: "Train approaching, proceed with caution."

[0085] 8. Visualized mobile terminal module A "command center in your hand" for on-site personnel (project manager, chief engineer, safety officer).

[0086] The system provides a comprehensive view of all elements, allowing users to view real-time GIS maps, driving plans, construction task orders, and surrounding video surveillance footage overlaid with their location on handheld devices or tablets.

[0087] The trajectory playback function records the historical trajectories of all personnel and machinery entering the site, supporting post-event queries for analyzing work efficiency or tracing the causes of accidents. The engineering construction management intelligent agent system modules are as follows: Figure 7 As shown, the network topology is as follows: Figure 8 As shown.

[0088] On-site data collection supports on-site photography and short video uploads, enabling image recording of concealed works and key processes, which are then linked to construction logs to achieve quality traceability.

[0089] In one embodiment, such as Figure 9 As shown, the operation and maintenance subsystem includes: The data acquisition and preprocessing module is used to aggregate the operating status and self-test messages of turnouts, track circuits, and interlocking systems to obtain time-series data, and to clean the noise of the time-series data using sliding window filtering and missing value interpolation algorithms. The equipment and cable full-element management module is used to associate components in 3D drawings with 3D physical models, and display cable models, laying paths, connection relationships and operation and maintenance records based on the association results; The concealed works visualization module is used to restore the actual layout of cables in concealed spaces.

[0090] In practical implementation, the system first integrates multi-source heterogeneous data through a data acquisition and preprocessing module, aggregating the operating status, self-test messages, and sensor data from equipment such as turnouts, track circuits, and interlocking systems. For this time-series data, the system uses sliding window filtering and missing value interpolation algorithms to remove noise and ensure the quality of the input data. In the core alarm judgment stage, the system mainly relies on the following algorithms: The threshold and logic interlocking algorithm, based on a finite state machine model, compares equipment parameters (such as current and voltage) with preset safety thresholds and combines Boolean logic operations such as signal light combinations and route conflicts to quickly identify hardware faults or operational violations.

[0091] Trend analysis and prediction algorithms, employing time series analysis (such as exponential smoothing) or autoregressive models, monitor the deterioration trends of key equipment indicators. When parameters deviate continuously from the baseline within limits, an early warning is triggered, shifting from "post-event alerts" to "pre-event predictions."

[0092] The system uses event correlation analysis, based on decision trees or a pre-defined expert rule base, to pinpoint the root cause of concurrent alarms, filter out accompanying secondary alarms, and accurately output the source of the fault and handling suggestions. The entire process adopts an event-driven architecture, ensuring that the end-to-end latency from data changes to interface pop-ups and audio-visual prompts is controlled within milliseconds.

[0093] In one example, the design subsystem, the engineering construction management intelligent agent system, and the operation and maintenance subsystem all satisfy at least one of the following constraints during the engineering construction process: a) Time of completion of the floor plan drawing Input completion time greater than or equal to that of a perforation plot ; b) The output time of the i-th type circuit in the design subsystem is greater than or equal to the time when the as-built drawing is completed. ,Right now The i-th type of circuit includes at least one of the following: signal lighting circuit, turnout control circuit, track circuit, direction circuit, scattered circuit, outdoor cable route diagram, outdoor cable network diagram, train control data table, interlocking table, and side wiring. c) Output time of the bill of quantities ; d) Administrator approval start time Approval completion time ,in This is the parameter for the approval duration; e) Construction start time ; f) As-built drawing output time Greater than or equal to the completion time of the drawing liaison form ; g) The wiring correctness of the i-th type of circuit satisfies The accuracy of operation and maintenance meets the requirements. ,in This is the minimum wiring accuracy parameter. This is the minimum operational accuracy parameter; The optimization objective of the mathematical model is to minimize the as-built drawing output time. .

[0094] Specifically, the steps for minimizing the as-built drawing output time are as follows: Define the output time range and segments of the bill of quantities, and discretize the continuous time axis into 25 segments; The Sigmoid function is used to capture nonlinear patterns, and then piecewise linear interpolation is used to transform the discrete 25 time segments into a linear form. Use binary variables Perform interval selection using continuous variables Perform a linear approximation within the interval; Ultimately, with the goal of minimizing the completion time, the optimal solution for the as-built drawing output time is found through a linear programming solver.

[0095] The above steps satisfy the following relationship: By employing the approach of "discretizing continuous problems → linear approximating nonlinear functions → mixed-integer linear programming modeling," the optimization problem of outputting the bill of quantities is transformed into a solvable linear programming problem. 1. First, define the time range and segments, and discretize the continuous time axis into 25 segments; 2. Use the Sigmoid function to capture nonlinear patterns, and then transform them into a linear form through piecewise linear interpolation; 3. Using binary variables To achieve "range selection", use continuous variables Achieve "linear approximation within the interval"; 4. Ultimately, with the goal of minimizing the completion time, the optimal solution is found using a linear programming solver.

[0096] Specifically, 1. The input range is defined as follows: ; Function: Specify the output time of the bill of quantities The feasible region, namely: .

[0097] In this way, Based on this, limit the upper and lower bounds of the output time (such as a reasonable range under engineering experience or resource constraints) to avoid the solution from exceeding practical significance.

[0098] 2. Piecewise discretization (number of segments and interval) is as follows: ; The function is to convert consecutive time intervals Divide the data into 25 segments and calculate the time interval for each segment. .

[0099] Its function is to transform continuous variables into a discrete structure of "piecewise approximation", preparing for subsequent linearization of nonlinear functions (such as Sigmoid) and construction of linear constraints.

[0100] 3. The segmentation points are generated as follows:

[0101] The function is: generate Each segment node ( (From 0 to 25, a total of 26 points), covering the entire time interval.

[0102] The purpose is to serve as anchor points for subsequent calculations of the Sigmoid function value and the construction of linear constraints, essentially "dividing" the continuous time axis into discrete reference points.

[0103] 4. The Sigmoid function is calculated as follows: The function is: for each segmentation point Calculate the Sigmoid function value ,in It is the slope parameter (which controls how steep the function is). It is the center point (the center of symmetry of the function), specifically the inflection point of the sigmoid function, i.e., the function value. The corresponding moment. It determines the moment when the nonlinear change is most drastic. For the first The Sigmoid function value at each discrete point represents the value at time point (i.e., at time 1). The normalized output after nonlinear transformation has a value range of (0,1). It is typically used to characterize project progress, risk probability, or a certain degree of "activation". For time, in this application, it is a continuous time variable, with units of days, hours, or minutes, representing the time coordinates of the construction process, train approach process, or equipment degradation process. B is the slope parameter (kurtosis coefficient), used to control the Sigmoid curve's... The steepness of the surrounding area. The larger the value, the more drastic the curve jumps from 0 to 1, and the more obvious the simulated "interval effect" becomes; The smaller the value, the smoother the transition. It is a natural constant.

[0104] Its function is as follows: The Sigmoid function is a typical nonlinear function, often used to simulate the "saturation effect" (such as the saturation growth of project progress over time). Here, the Sigmoid values ​​at discrete points provide a data basis for the subsequent "piecewise linear approximation" (approximating the nonlinear curve with linear segments).

[0105] 5. The sum constraints of binary-assisted variable stars are as follows:

[0106] Function: Constrain binary variables The sum of is 1.

[0107] Its function is to ensure that only one exists. The rest are 0. Combined with subsequent constraints, this can be understood as "selecting a certain interval". "Approximate" The interval in which it is located implements the logic of "segmented selection".

[0108] 6. The Big M method constraints (interval selection) are as follows:

[0109] The function is: to utilize the Big M method ( (for a sufficiently large constant), Falling into a certain segment interval Transform it into a linear constraint.

[0110] Function: For when When, the constraints are simplified to Forced Falling in The interval is given by the formula. For the first A selectable time period; when hour, Xiang Yin Extremely large and "loose" constraints (without affecting feasibility) ensure that only the selected intervals are affected. )right There are practical limitations.

[0111] 7. Continuous variable stars The constraints are as follows: Function: To constrain continuous variables The value of, combined with binary nature, implementation It is only valid within the selected range.

[0112] Its function is: when hour, Must meet and ,Right now and Strong correlation; when hour, and (because Therefore ),at last That is, the unselected interval No contribution.

[0113] 8. The objective function related (piecewise linear approximation) is as follows:

[0114] Its function is to construct the core terms of the objective function. The integral or heterogeneous effect of the Sigmoid function is approximated by piecewise linear interpolation.

[0115] Its function is: Is the Sigmoid in the interval The slope (a linear approximation of "slope"); in, It is the "weight" or "length" within that interval; It is the Sigmoid value at the starting point of the interval (since there is only one). (Only the starting value of the corresponding interval is included). Overall, Regarding the Sigmoid function The linear approximation of the interval transforms the nonlinear relationship into a form that can be handled by linear programming.

[0116] 9. The final objective function is: min

[0117] The function is to minimize the completion time. To optimize the objective.

[0118] Its function is to combine all the above constraints (range, piecewise division, linearization, variable association, etc.) and use a linear programming solver to find the optimal solution that satisfies all the conditions. ,make Minimize (e.g., in project schedule optimization, minimize the total completion time).

[0119] Specifically, when solving for the minimum as-built drawing output time under the constraints of the aforementioned digital platform, the output time of the bill of quantities is... Time to output as-built drawings The nonlinear temporal relationship between them is addressed by using the approach of "continuous-time discretization - nonlinear function linear approximation - mixed-integer linear programming modeling," which transforms the complex engineering schedule optimization problem into an efficient linear programming problem.

[0120] The specific implementation steps are as follows: 1. Determine the feasible time region: Input the completion time using a millimeter chart. Based on this benchmark and combined with engineering experience, the time for outputting the bill of quantities is set. The reasonable upper and lower bounds are defined as follows: ,in 1000 (Time unit is set according to the actual project).

[0121] 2. Discretization of continuous time axis: Discretizing the continuous time interval Evenly divided into Section (This invention takes) ), calculate time step and generate Discrete segmented nodes .

[0122] 3. Capturing and linearizing nonlinear patterns: using the Sigmoid function Calculate each segment node function value To capture the nonlinear saturation growth pattern of project progress over time, where k is the growth rate factor and t is time, a piecewise linear interpolation method is then used to interpolate the Sigmoid function across each segment interval. The inner approximation is a linear function.

[0123] 4. Construct a mixed-integer linear programming model: Introducing binary auxiliary variables Used for range selection, constraints Ensure that exactly one interval is selected.

[0124] Constructing linear constraints using the Big M method to force variables Falling within the selected interval Inside.

[0125] Introducing continuous auxiliary variables Characterization The specific position and weight within the selected interval.

[0126] Constructing the relevant terms of the objective function Based on piecewise linear slope and Calculate the linear approximate cumulative effect of the Sigmoid function, thereby establishing and A linear mapping relationship between them.

[0127] 5. Solve the optimization problem: minimize the as-built drawing output time. With the goal of solving the mixed-integer linear programming model, a linear programming solver is used to obtain the optimal output time of the bill of quantities and the corresponding minimum output time of the as-built drawings.

[0128] In some embodiments, pseudocode for calculating the completion time is provided to verify this application as follows: I. Input Parameters and Constant Definitions: Input parameters: text Parameter T_thousand: The completion time for the thousandthogram input (unit: time unit, such as day or hour); Parameter T_approval_duration: Duration of the approval process; Parameter T_contact: Completion time of the drawing contact form; Parameter k: Slope parameter of the Sigmoid function (default value 0.01); Parameter t0: Offset of the center point of the Sigmoid function (default value 500); Parameter alpha: Progress conversion factor (default value 0.5); Parameter beta: Fixed time offset (default value 10); System default constants: text Constant M: Large M constant, taking the value of a sufficiently large positive number (e.g., 1 × 10⁻⁶). 6 ); Constant n: The number of interval segments for piecewise linear approximation (e.g., 25). II. Variable Definition and Feasible Region Partition Step 1: Determine the search range for the output time T_engineering of the bill of quantities; text Lower time limit T_min ← T_thousand; The maximum time limit T_max ← T_thousand + 1000; The interval step size delta_t ← (T_max - T_min) / n; Step 2: Generate piecewise linear approximation nodes and calculate the Sigmoid function value; text Initialize array t[0..n] as a floating-point array; Initialize the array y[0..n] as a floating-point array; For j, loop from 0 to n: t[j] ← T_min + j × delta_t; y[j] ← 1 / (1 + exp(-k × (t[j]- t0))); End the loop III. Construction of Mixed Integer Linear Programming Model Step 3: Initialize the optimization problem; text Create a linear programming problem instance named "Railway_Completion_Minimization" with the objective of minimizing; Step 4: Define decision variables text Define a continuous variable T_engineering, with a value range of [T_min, T_max]; Define a continuous variable T_completion, with a value range of [0, +∞); Define a binary variable array delta[0..n-1], where each element takes the value 0 or 1; Define a continuous auxiliary variable array omega[0..n-1], where each element takes values ​​in the range [0, T_max]. Define a continuous variable T_plane, with a value range of [0, +∞) / / Plane completion time, for illustration only. IV. Linearization of Constraints; Step 5: Basic constraints of sequential logic (list the core dependencies as an example); text Add constraint: T_plane ≥ T_thousand; Add constraint: T_engineering ≥ T_plane + 50 / / This constant 50 is an example value, and the actual value is determined by the maximum completion time of the ten types of circuits; Add constraint: T_completion ≥ T_contact; Note: In the complete model, this section should include all the timing dependency constraints present in the actual project, for example: T_engineering ≥ max(completion time of various circuits T_circuit_i); T_completion ≥ T_engineering + approval duration + summary time for each specialty, etc.

[0129] To simplify the description, only the constraint structure directly related to the core optimization variables is shown here.

[0130] Step 6: Select Uniqueness Constraints for Intervals text Add the constraint: ∑_{j=0}^{n-1} delta[j] = 1; This constraint ensures that the value of the variable T_engineering falls exactly into one of the n segmented intervals.

[0131] Step 7: Large M method interval positioning constraint; text For j, loop from 0 to n-1: Add constraint: t[j] - M × (1 - delta[j]) ≤ T_engineering; Add constraint: T_engineering ≤ t[j+1] + M × (1 - delta[j]); End the loop The purpose of this constraint group is to force T_engineering to satisfy t[j]≤ T_engineering ≤ t[j+1] if and only if delta[j] = 1.

[0132] Step 8: Linearization constraints on the continuous auxiliary variable omega; text For j, loop from 0 to n-1: Add constraint: omega[j] ≤ T_max × delta[j]; Add constraint: omega[j] ≤ T_engineering; Add constraint: omega[j] ≥ T_engineering - T_max × (1 - delta[j]); End the loop: This constraint group performs the following function: when delta[j] = 1, omega[j] = T_engineering; when delta[j] = 0, omega[j] = 0.

[0133] Step 9: Constructing a linear approximation of the Sigmoid cumulative term A; text Create a linear expression A, initialized to 0; For j, loop from 0 to n-1: Calculate the piecewise slope as slope ← (y[j+1] - y[j]) / delta_t; Update expression A: A ← A + slope × omega[j] + y[j] × delta[j]; End the loop Here, a piecewise linear function is used to approximate the Sigmoid shape, transforming the nonlinear relationship into a linear constraint that can be handled by MILP.

[0134] Step 10: Constraints on the completion time correlation equation; text Add the constraint: T_completion = T_engineering + alpha × A + beta; This equation links the output time of the bill of quantities with the output time of the as-built drawings through a sigmoid cumulative term, reflecting the nonlinear impact of the progress of the electrical engineering project on subsequent completion work.

[0135] V. Objective Function Setting and Solution Step 11: Set optimization goals text The objective function for the problem is set as: minimizing T_completion; Step 12: Call the solver to perform calculations text The above problem can be solved by calling a mixed-integer linear programming solver (such as the open-source solver CBC); Set the solver output message mode to off (msg=False); Step 13: Result Analysis and Output; text Get the solution status; If the status is "Optimal" (the optimal solution has been found): Optimal as-built drawing output time T_opt ← Get the optimal value of variable T_completion; Optimal Bill of Quantities Output Time T_eng_opt ← Get the optimal value of variable T_engineering; Returns (T_opt, T_eng_opt); otherwise: Throws an exception or returns an error flag, indicating that the problem is unsolvable; VI. Complete Main Process Pseudocode text Algorithm: Minimizing the completion time of railway electrical engineering; Input: T_thousand, T_approval_duration, T_contact, k, t0, alpha, beta; Output: Optimal completion time T_completion and the corresponding bill of quantities output time T_engineering ; Process optimization completion time: / / 1. Initialize segmentation parameters; T_min ← T_thousand; T_max ← T_thousand + 1000; n ← 25; delta_t ← (T_max - T_min) / n; M ← 1,000,000; Define floating-point arrays t[0..n] and y[0..n]; For j ← 0 to n, execute: t[j] ← T_min + j × delta_t; y[j] ← 1 / (1 + exp(-k × (t[j]- t0))); End the loop; / / 2. Establish a MILP model; Create a minimum problem prob named "Railway_Completion_Minimization"; Define a continuous variable T_engineering ∈ [T_min, T_max]; Define a continuous variable T_completion ≥ 0; Define a binary variable delta[j] ∈ {0,1} for j = 0..n-1; Define a continuous auxiliary variable omega[j] ∈ [0, T_max] for j = 0..n-1; Define a continuous variable T_plane ≥ 0; / / 3. Add timing constraints (example); Add the constraint T_plane ≥ T_thousand; Add the constraint T_engineering ≥ T_plane + 50; Add the constraint T_completion ≥ T_contact; / / 4. Add piecewise linearization constraints; Add the constraint ∑_{j=0}^{n-1} delta[j] = 1; For j ← 0 to n-1, execute: Add the constraint t[j] - M × (1 - delta[j]) ≤ T_engineering; Add the constraint T_engineering ≤ t[j+1] + M × (1 - delta[j]); Add the constraint omega[j] ≤ T_max × delta[j]; Add the constraint omega[j] ≤ T_engineering; Add the constraint omega[j] ≥ T_engineering - T_max × (1 - delta[j]); End the loop; / / 5. Construct a linear expression for the cumulative term A; Initialize the linear expression A ← 0; For j ← 0 to n-1, execute: slope ← (y[j+1] - y[j]) / delta_t; A ← A + slope × omega[j] + y[j]× delta[j]; End the loop; / / 6. Add a completion time correlation equation; Add the constraint T_completion = T_engineering + alpha × A + beta; / / 7. Set the objective and solve it; Set the target of prob to minimize T_completion; Call the solver to solve for prob; If the solution state is optimal: Returns (the optimal value of T_completion, the optimal value of T_engineering); otherwise: The system returned the error message "No feasible solution". The process is over.

[0136] In some implementations, the engineering construction management intelligent agent system is modeled as six types of intelligent agents: vehicle protection intelligent agent, construction collaboration intelligent agent, progress investment intelligent agent, disaster prevention linkage intelligent agent, operation and maintenance diagnosis intelligent agent, and mobile terminal intelligent agent. Each type of intelligent agent adopts a hybrid architecture of BDI model and finite state machine, and coordinates through an event-driven bus.

[0137] The intelligent agent for train protection integrates a spatiotemporal collision detection algorithm. This algorithm calculates the spatial and temporal overlap between the train's operating section and the construction section, and introduces a train speed weighting factor to output a three-level conflict warning. Based on the planned duration and actual completion percentage of the WBS nodes, it calculates the lag days and automatically generates resource-duration compression suggestions when the threshold is exceeded. It compares the contract budget line, completed output value line, and actual payment line in real time, and uses exponential smoothing to handle data jitter, achieving automatic warnings for over-budget and over-payment. It uses the Holt-Winters three-parameter model to predict the degradation trend of equipment parameters and uses a decision tree model to perform root cause reasoning on concurrent alarms, filtering accompanying alarms. The intelligent agent system adopts a hierarchical hybrid architecture, containing 6 types of autonomous agents and 1 cooperative bus. Each type of agent uses a combination of the BDI (Belief-Desire-Intention) model and finite state machine for behavior modeling.

[0138] The intelligent agent system is modeled as follows: Figure 10 As shown: The intelligent agent system adopts a hierarchical hybrid architecture, comprising six types of autonomous agents and one cooperative bus. Within each type of agent, behavior modeling is performed using a combination of the BDI (Belief-Desire-Intention) model and a finite state machine.

[0139] In one example, such as Figure 11 As shown, the formal definition and component design of the BDI model are as follows: Beliefs are the agent's perception of the current state of the environment and its own state. They are stored using timestamped RDF triples and maintained in a blackboard structure.

[0140] Data structure: Each belief is in the form of (subject, predicate, object, timestamp, confidence).

[0141] For example: (TrainG102, hasPosition, KM125+300, 2025-03-15T14:23:10Z, 0.99).

[0142] For example: (ConstructionSiteA, status, ACTIVE, 2025-03-15T14:23:10Z, 1.0) Sources of belief include: Direct perception: Real-time acquisition of CTC / TDCS data, BeiDou positioning, sensor readings, and video AI results via interfaces.

[0143] Communication beliefs: Extracted from messages received by other agents via the event bus.

[0144] Reasoning beliefs: Derived from existing beliefs by a built-in rule engine (such as RDFox), for example: (Train G102, willReachSiteA, within 2 min) is deduced from the train's position, speed, and the location of the construction area.

[0145] Belief update: An incremental update + version number mechanism is adopted. Each belief has a life cycle (if it is not updated for 10 seconds, it is marked as expired and the confidence level decays).

[0146] Furthermore, a wish is a target state that an agent needs to achieve, organized in the form of a goal tree, which supports goal decomposition and priority ranking.

[0147] The target types are as follows: Hard targets: Safety objectives that must be achieved (such as "avoiding conflicts between traffic and construction") and cannot be violated.

[0148] Soft objectives: Optimization objectives (such as "minimize schedule lag days") that allow for trade-offs.

[0149] Target tree structure: The root target is the core mission of the agent (e.g., the root target of the driving protection agent is "ensure driving safety"), which is then decomposed into sub-targets ("detect conflict" → "issue early warning" → "track the failure point").

[0150] Target tree structure: The root target is the core mission of the agent (e.g., the root target of the driving protection agent is "ensure driving safety"), which is then decomposed into sub-targets ("detect conflict" → "issue early warning" → "track the failure point").

[0151] Desire activation conditions: Each desire is associated with an activation rule (e.g., "Activate the 'Approach Warning' desire when there is active construction and the train distance is <5km"). After activation, the desire is placed in the Desire Pool and assigned a dynamic priority.

[0152] Furthermore, an intent is a currently committed instance of a plan. Each intent contains: Plan body: a series of action sequences (e.g., queryTrainPosition → computeDistance → ifdistance<2km then sendAlert).

[0153] Contextual conditions: Prerequisites for the application of this plan (e.g., "electronic fences must have been established").

[0154] Progress status: Train approach trigger → PENDING (command preparation) → RUNNING (command being issued) → SUCCEEDED (protection in place). When an anomaly occurs, the path is: PENDING → RUNNING → (no feedback after timeout) → FAILED (fault alarm); PENDING → (manual operation) → SUSPENDED (temporary suspension).

[0155] Intent queue: Multiple concurrent intents are managed using a priority queue, and preemption is supported (a high-priority alert intent can interrupt a low-priority data recording intent).

[0156] The BDI interpreter loops as follows: Each agent internally runs a continuous interpreter loop at a frequency of 10 Hz (which can be dynamically adjusted according to the task), performing the following steps: The design and integration of the Finite State Machine (FSM) with the BDI are as follows: The BDI model provides decision-making capabilities, but railway safety scenarios require certain responses to be completed within a fixed time budget (e.g., <50ms), and the behavior must be deterministic and verifiable. Therefore, a hierarchical finite state machine (HFSM) is embedded within each agent as the response layer, and the BDI model is responsible for calculating the conditions for state transitions and planning actions across states.

[0157] The hierarchical finite state machine structure is as follows: HFSM consists of a top-level state machine (macro-level behavioral pattern) and several sub-state machines (micro-level processing flow).

[0158] Taking a vehicle safety intelligent agent as an example: Top-level status: IDLE: No construction activity, only data monitoring.

[0159] ARMED: Construction is underway, an electronic fence has been established, and the system is on alert.

[0160] PRE_ALERT: Train distance from construction area Prepare for early warning.

[0161] ALERTING: Train distance from construction area An alert is being sent out.

[0162] EMERGENCY: An intrusion or timeout failure is detected, triggering emergency procedures.

[0163] RESOLVED: The train has passed safely or construction has been completed.

[0164] Sub-state machine (taking the alerting state as an example): SEND_ALERT → WAIT_ACK → ESCALATE (If no confirmation is received from the safety officer within 30 seconds, the alarm will be escalated).

[0165] The bridging mechanism between BDI and FSM is as follows: The BDI model interacts with the FSM through three interfaces: 1. State Transition Decision Interface: The state transition conditions of the FSM are not hard-coded Boolean expressions, but rather call the evaluation function `evaluateTransition(currentState, event)` in the BDI layer. This function queries the current belief set and uses the rule engine to determine whether the transition conditions are met. The transition conditions can be dynamically learned or adjusted without recompiling the FSM.

[0166] 2. Mapping of BDI intentions to FSM actions: The plans generated by the BDI layer are eventually compiled into an FSM action sequence. Each action corresponds to an atomic operation (such as sendMessage, updateDatabase, triggerAlarm).

[0167] When the BDI selects an intent (e.g., "Execute Level 3 Proximity Warning"), the planner for that intent is broken down into a series of actions and written to the FSM's action queue. The FSM's entry actions in the current state or These atomic operations are executed sequentially during the action.

[0168] 3. FSM Events Triggering BDI Wishes: During state transitions, the FSM emits system events (such as onEnterState(ALERTING)). These events are captured by the BDI interpreter and used to activate new wishes (such as "Request video linkage") or adjust the priority of existing wishes.

[0169] Example of a hybrid architecture workflow (train protection agent handling "train approach") 1. Perception Phase (BDI): The agent obtains the train position update belief from the CTC interface: (TrainG102, distanceToSite, 1800m).

[0170] 2. BDI Interpreter: The rule engine detects distance. constructionStatus == ACTIVE, activating the Wish_IssueApproachAlert.

[0171] 3. Select Intent: Select the highest priority intent Intent_IssueAlert_Level2 from the wish pool (the plan includes: calculating the warning level → querying the protector terminal ID → pushing the warning message → recording the log).

[0172] 4. Execution Intent: The first step of the intent, computeAlertLevel, is executed, resulting in Level 2 (distance 1800m, train speed). ).

[0173] 5. FSM Intervention: The BDI layer calls the FSM's migration evaluation function to determine whether the current state ARMED should migrate to PRE_ALERT. The evaluation result is true.

[0174] 6. FSM State Transition: Execute the exit action of ARMED (clean up temporary variables), then execute the transition action transitionTo(PRE_ALERT), and finally execute the entry action of PRE_ALERT (start the timer and prepare for alerts).

[0175] 7. Action Execution: In the do action of PRE_ALERT, FSM sequentially executes the atomic operations that BDI intents are placed into the action queue: sendAlertToTerminal(terminalID, "Train approaching, take precautions") and publishEvent("TrainApproaching").

[0176] 8. Anti-housing and closed-loop: If the protector confirms within 30 seconds, the FSM migrates to ALERTING; otherwise, the BDI layer will activate the higher priority Wish_EscalateAlert and replace the current intent.

[0177] 3. Design details are as follows: 3.1 Belief-State Consistency Maintenance: The publish-subscribe pattern is used: any update to a belief automatically triggers a re-evaluation of the state transition conditions that depend on that belief. If the condition changes from false to true, the FSM performs the transition asynchronously (without interrupting the currently executing action, but checking between actions).

[0178] A double-buffering mechanism is used: belief updates are written to a background buffer and atomically swapped at the end of each interpreter cycle to avoid state inconsistencies.

[0179] 3.2 Preemption and Recovery of Intent: High-priority intents (such as emergency braking suggestions) can preempt the currently executing intent. Interrupted intents are saved to the pause stack and resumed after the high-priority intent completes.

[0180] Resume points are marked by checkpoints in the plan body (such as those automatically set after each action). Upon resumption, execution continues from the last checkpoint.

[0181] 3.3 Verifiability and Determinism: While the BDI model introduces nondeterminism (due to noise in the source of beliefs), the FSM part guarantees the determinism of the critical safety path: for inputs that trigger safety alerts, the FSM's migration path is unique and formally verifiable. This invention performs reachability analysis and deadlock detection on the FSM through model testing (e.g., using the UPPAAL tool).

[0182] The rule engine in the BDI layer uses deterministic rule ordering (Rete algorithm) to avoid decision jitter.

[0183] 3.4 Cross-Agent BDI Collaboration When an agent's BDI layer cannot fulfill a certain request independently (e.g., a construction coordination agent needs the real-time location of a train), it publishes a RequestBelief event to the coordination bus. Other agents holding the belief (such as the train protection agent) will automatically respond. This mechanism achieves distributed BDI, where multiple agents work together to complete a complex goal, but each agent still independently runs its own BDI interpreter and FSM.

[0184] 4. The advantages compared with traditional methods are shown in Table 3 below: Table 3 Comparison between intelligent agents and traditional methods

[0185] The agent is formalized as follows: Each agent can be represented as a quintuple as follows: ; Perception: The perception interface receives event streams from sensors, databases, or other intelligent agents.

[0186] Belief: The current environmental state (such as train location, construction progress, equipment parameters) stored in an RDF graph or time series database.

[0187] Desire: A set of target states (such as "ensuring no conflict between traffic and construction", "schedule deviation") sky").

[0188] Intention: The currently active plan (such as "Triggering a nearing warning" or "Generifying a rush task suggestion").

[0189] Action: The executor that outputs instructions, pushes messages, or updates the database.

[0190] In one example, we will use Algorithm 1 as an example of the spatiotemporal collision detection algorithm between driving plans and construction plans, and explain it in detail: enter: Train operation plan Each line, start_km, end_km, start_time, end_time, speed

[0191] Construction Plan Each (zone_km, occupation_start_time, occupation_end_time, type) Output: List of Conflicts conflict_level

[0192] step: 1. Spatial Filtering: Calculate the overlap length between the train operating section and the construction section. end_km, zone_end start_km, zone_start .like ,jump over.

[0193] 2. Time Filtering: Calculate the time overlap. start_time,occupation_start end_time, occupation_end .like If not empty, proceed to the next step.

[0194] 3. Conflict Level Calculation:

[0195] 4. Output: Level 1 conflicts are highlighted in red, Level 2 in orange, and Level 3 in yellow on the scheduling screen.

[0196] In this way, a speed weighting factor is introduced, and high-speed trains are assigned a high risk level even if they overlap briefly.

[0197] In one example, Algorithm 2 is an algorithm for automatically drawing and dynamically correcting the progress front line: enter: WBS Node Plan start time Planned end time Planned construction period

[0198] Current inspection date Cumulative actual completion percentage

[0199] Output: Forward positions, number of days behind schedule, suggestions for expediting the process. step: 1. Theoretical completion percentage (by time):

[0200] 2. Forward Positions: On the Gantt chart, for each... Draw a vertical line on the date Mark the actual completion point ( . ).

[0201] 3. Lag days:

[0202] like (Threshold, such as 3 days), triggers an alert.

[0203] 4. Expedited Work Request Generation: Based on the resource-time compression model, recommended expedited work strategies are generated. If the delay is due to insufficient resources, the amount of resources will be increased. .

[0204] If the delay is due to process logic, it is recommended to perform parallel operations (such as changing the serial operation to partial overlap).

[0205] In one example, Algorithm 3 is: Investment Control Three-Curve Linkage Early Warning Model.

[0206] Define three curves: Contract budget line (allocated according to the planned completion ratio) Completed production value (based on actual completed work volume × contract unit price) Actual payment line (synchronized from the financial system) Warning rules: 1. Over-probability warning: If If a red alert is triggered; This triggered an orange alert.

[0207] 2. Overpayment warning: If And the difference is greater than the contract price. This triggered an "overpayment risk" warning.

[0208] 3. Delayed payment early warning: If The total contract amount is displayed with the message "The amount due but not yet paid is too high".

[0209] The algorithm is implemented as follows: Based on WBS nodes, a rough estimate is automatically calculated daily. Exponential smoothing is used to handle data fluctuations.

[0210] in, For the current moment The original data collected from the field. The current time after algorithm correction The best estimate, This is the smoothed value calculated at the previous time (usually yesterday). As a smoothing coefficient, in one example, .

[0211] In one example, Algorithm 4 is: Beidou electronic fence and train approach warning algorithm; enter: Construction area polygon (Defined by latitude and longitude sequence).

[0212] Real-time location of personnel or machinery (lat, lon).

[0213] Real-time train location (train_km, speed), remaining distance to the construction area ; Output: Alarm commands (audio-visual, voice).

[0214] step: 1. Intrusion Detection: Calculate the shortest distance from a point to a polygon. .like This indicates an intrusion and immediately triggers a "boundary intrusion" alarm.

[0215] 2. Approach Warning: When the train is approaching the construction area At that time, calculate the time it takes for the train to arrive at the construction area. speed. If Send an "emergency exit" command; if Send a "Caution: Avoid" alert.

[0216] 3. Multi-target optimized broadcast: The system indexes all online terminals using a K-D tree and only pushes warnings to personnel within 500 m of the construction area to avoid information overload.

[0217] In one example, Algorithm 5 is: Operation and Maintenance Time Series Data Trend Prediction and Root Cause Localization Algorithm 5.1 Data Preprocessing: Use sliding window mid-range filtering to remove transient spikes:

[0218] Missing value imputation: linear interpolation or prediction imputation based on the ARIMA model.

[0219] 5.2 Trend Forecasting (Exponential Smoothing Method): For device parameter sequence (For example, turnout switching current), the Holt-Winters three-parameter model is used:

[0220] in For level terms, For trend items, For periodic terms (periodic) (Minutes, i.e., daily cycle). When the predicted value for the next hour exceeds a threshold (such as the rated current). This triggers an alert.

[0221] 5.3 Root Cause Analysis (Based on Decision Tree) Construct feature vectors: Extract the following from alarm events: {Device type, alarm code, occurrence time, associated route, and status of adjacent devices}.

[0222] Train the decision tree model (C5.0 algorithm) and output the root cause of the failure.

[0223] text IF (turnout voltage < 20V) AND (red light band on adjacent track circuit) AND (turnout switching time > 5s).

[0224] THEN Root Cause "Dao Gong's secret post is not good."

[0225] For new alarms, the source of the fault is determined through model reasoning, and accompanying alarms (such as subsequent track circuit alarms caused by turnout faults) are filtered out.

[0226] In one example, Algorithm 6 is: A1 Video Violation Detection Algorithm A lightweight convolutional neural network (improved YOLOv8) is used for deployment at the edge. The network structure includes: Backbone: CSPDarknet, for extracting feature maps.

[0227] Neck: PANet, which fuses multi-scale features.

[0228] Head: Decoupled detection head, outputs bounding box and category (safety helmet, guardrail climbing, smoke, flame, etc.).

[0229] Loss function: CloU Loss + Focal Loss to solve the imbalance between positive and negative samples.

[0230] In this way, introducing a temporal attention module to vote on the detection results of multiple consecutive frames can reduce false alarms.

[0231] In one example, Algorithm 7 is: Dynamic Cash Flow Prediction Algorithm Input: future Monthly progress ( (month) (per WBS node) Contract Unit Price

[0232] Payment delay distribution (historical data statistics) Output: Monthly payable amount (Payable) Monthly actual payment forecast step: 1. Accounts payable calculations:

[0233] 2. Actual Payment Prediction: Introducing a Payment Delay Convolution Kernel (like Payment in the current month Payment will be made the following month. (Third month payment):

[0234] 3. Cash flow gap warning: If Available funds We suggest adjusting your payment plan or applying for a loan.

[0235] (III) Agent Modeling and Algorithm Integration Methods This application adopts a model-driven architecture (MDA), where the aforementioned algorithm is encapsulated and can be understood as a skill component of the agent. Internally, the agent binds perceived events with algorithm calls through an extensible rule engine (such as Drools).

[0236] Regarding the accuracy of collision detection: simulation tests show that the accuracy of collision detection between driving and construction plans is ≥99.5%, and the warning lead time is ≥2 minutes.

[0237] In terms of progress control efficiency: the automatic drawing of the forward line takes less than 1 second, and the delay warning covers 100% of the WBS nodes.

[0238] Regarding investment control: the three-curve model provides an early warning of over-probability risks 30 days in advance, with an accuracy rate of 92%.

[0239] Operation and maintenance prediction: The exponential smoothing method has an accuracy of 85% in predicting turnout failures in advance, with an average prediction time of 24 hours.

[0240] Edge AI: Video recognition inference speed of 30 FPS (on Jetson Xavier NX), with mAP@0.5 reaching 87%.

[0241] For further explanation and verification, this application provides the following pseudocode for the vehicle protection intelligent agent control method: 1. Belief Data Recording: text Record Belief: train_distance (floating-point number, unit: meters) is the distance between the train and the construction site. The train's current speed is train_speed (a floating-point number in kilometers per hour). The site is active (boolean value). worker_present (boolean value) indicates the presence status of the workers. 2. Desire Data Recording; text Record of Desire: Wish name (string); Priority (integer, the larger the value, the higher the priority); 3. Intention data recording; text Recording Intention: The corresponding wish is desire (Desire type); The action plan (a list of actions) is a sequence of actions to be executed. The current execution step index is current_step (an integer, initially set to 0). 4. Executable action records (Action); text Record Action: Action name (string); executor (a reference to the specific execution process); 5. Definition of states in a hierarchical state machine; text State (enumeration type): IDLE / / Idle state; ARMED / / Ready / Ready status; WARNING / / Warning status; EMERGENCY / / Emergency braking status; 6. Definition of state transition flags; text Enumeration type Transition: IDLE_TO_ARMED / / Idle → Ready; ARMED_TO_WARNING / / Ready → Warning; WARNING_TO_EMERGENCY / / Warning → Emergency; EMERGENCY_TO_IDLE / / Emergency → Idle; ANY_TO_IDLE / / Any state → Idle (used for reset); II. Description of Core Functional Components; Component 1: BeliefBase; Function: Maintain the agent's cognitive information about the current environment.

[0242] operate: Update sensing data (distance, speed, construction site status, personnel status): Receive input from external sensors and update the corresponding fields in the Belief record.

[0243] Get and clear change event list(): Returns a list of new internal events generated since the last query and clears the event cache.

[0244] Component 2: DesirePool; Function: Stores and manages multiple currently active wishes.

[0245] operate: Activate a wish: Add a wish to the wish pool.

[0246] Undo Wish (name): Remove the wish by name.

[0247] Get Highest Priority Wish(): Returns the wish with the highest priority value in the current pool; returns null if the pool is empty.

[0248] Component 3: IntentionQueue; Function: Manages the intentions that the current agent has committed to fulfilling.

[0249] operate: Add an intention: Add the intention to the end of the queue.

[0250] View Current Intent(): Returns the intent at the head of the queue but does not remove it.

[0251] Remove Current Intent(): Removes and returns the intent at the head of the queue.

[0252] Component 4: Rule Engine; Function: Based on the current state in the belief library, trigger the corresponding wish activation or deactivation rules.

[0253] Rule set (example): Rule 1: If the train distance is less than 2000 meters and the construction point is activated, then activate the wish "Wish_IssueApproachAlert" with a priority of 10.

[0254] Rule 2: If the train distance is less than 500 meters, the train speed is greater than 40 km / h, and the personnel are present, then activate the wish "Wish_EmergencyBrake" with a priority of 100.

[0255] Rule 3: If a construction site is not activated, all wishes in the wish pool will be cleared.

[0256] Component 5: PlanLibrary; Function: Instantiate specific execution intentions based on the selected wishes (generate action sequences).

[0257] Plan mapping relationship: Wish "Wish_IssueApproachAlert" → Action sequence: ["Play proximity warning sound", "Send proximity alarm to construction site"].

[0258] Wish "Wish_EmergencyBrake" → Action sequence: ["Send emergency braking command", "Activate on-site audible and visual alarm"].

[0259] Component 6: Hierarchical State Machine (HierarchicalFSM); Functions: Manage the behavioral state of the agent, execute state entry / exit actions, handle state transition conditions, and execute specific actions from the intent queue.

[0260] operate: State transition (target state): Execute the exit callback of the current state, update the current state to the target state, and execute the entry callback of the target state.

[0261] Single-step advancement(): Execute the loop callback (do action) for the current state.

[0262] Execute all pending actions in the action queue.

[0263] Check the transition conditions associated with the current state. If any condition is met, execute the state transition (only one transition is triggered per single step).

[0264] III. Main Control Flow of the Intelligent Agent; Initialization process: Create and initialize instances of the Beliefs Library, Wish Pool, Intent Queue, Rule Engine, and Plan Library.

[0265] Construct a hierarchical state machine and set the initial state to IDLE.

[0266] Define entry callback, do callback, and exit callback for each state.

[0267] Configure a state transition condition mapping table to associate the Transition identifier with a specific condition judgment function.

[0268] Main loop (called once for each simulation clock step, denoted as the tick process): text Process agent clock advancement(): / / Step 1: Belief Updating and Rule-Based Reasoning; Call the rule engine to trigger all rules(); / / This step internally reads the current value of the belief library and modifies the wish pool accordingly; / / Step 2: Intent selection (if there are no intents to be executed and the wish pool is not empty); If the intention queue is empty () and the non-wish pool is empty (): Highest Wish ← Wish Pool, obtain the highest priority wish(); If the highest wish is not empty: New Intention ← Instantiation of the Project Library (Highest Wish); Add an intent (new intent) to the intent queue; Output debugging information: "Selected new intent: [Wish name]"; / / Step 3: Intent Execution and Action Dispatch; Current Intent ← View the current intent in the intent queue (); If the current intent is not empty: Action to be executed ← Current intent Get next action(); If the action to be executed is not empty: Enqueue (actions to be executed) into the state machine action queue. If the current intent has been executed: Completed intent ← Remove the current intent from the intent queue (); Output debugging information: "Intent completed: [Intent name completed]"; / / Step 4: Execute the hierarchical state machine; Call the state machine to advance step by step (); / / The state machine will execute the do callback of the current state, process the enqueued action, and evaluate the state transition conditions in sequence; / / Step 5: Event Collection and Publication; Change Event List ← Belief Library retrieves and clears the Change Event List (); If the change event list is not empty: Output or distribute a list of change events; The process is over; IV. Definition of State Transition Conditions The following transition condition judgment rules are predefined in the state machine: IDLE → ARMED transition condition: The activation status of the construction point in the belief library is true.

[0269] ARMED → WARNING Conversion condition: The distance between the train in the belief database and the construction point is less than 2000 meters.

[0270] WARNING → EMERGENCY Conversion Conditions: The following conditions must be met simultaneously: (train distance < 500 meters) (train speed > 40 km / h) (operators present).

[0271] EMERGENCY → IDLE conversion condition: The distance between the train in the belief database and the construction point is >2000 meters.

[0272] The condition for transitioning from any state to IDLE is: the activation state of the construction point in the belief library is false. In some embodiments, such as Figure 12 As shown, a large model is introduced into the design subsystem, and the interactions between the operation and maintenance subsystem, the design subsystem, and the railway train control system are as follows: The operations and maintenance subsystem will send the collected abnormal event JSON data to the design subsystem's large model; The design subsystem large model is based on the received abnormal event information and initiates a train control system circuit type identification request to the railway train control system circuit schematic knowledge base; The railway train control system circuit schematic knowledge base matches corresponding fault scenarios and feeds back the train control system circuit diagram and electrical equation data to the design subsystem; Based on the circuit diagram and electrical equation data, the design subsystem infers and generates fault node assumptions for the train control system schematic diagram. The design subsystem is based on the fault node assumption and retrieves the field wiring location information corresponding to the fault from the railway train control system circuit schematic knowledge base. The railway train control system circuit schematic knowledge base returns the corresponding circuit terminal number and physical location information to the design subsystem; Simultaneously, the design subsystem calls down to the underlying fault analysis engine: it simulates the circuit operation characteristics through the SPICE simulator, performs timing logic deduction based on the graph reasoning engine, and matches the wiring information in the drawings with the help of the vector retrieval engine. It also completes the comparison between the theoretical and actual values ​​of the node electrical quantities and the backtracking verification of the timing logic chain status. Combined with the feedback information from the aforementioned knowledge base, it generates a structured analysis report of train control faults in an integrated manner. The design subsystem will send the generated JSON / PDF format structured fault analysis report back to the operation and maintenance subsystem; The operation and maintenance subsystem will push the final fault analysis results to the digital twin integrated platform to complete the intelligent fault interaction throughout the entire process.

[0273] The above-described interaction process, based on a multimodal slicing strategy and a retrieval enhancement generation strategy, employs the following technical points, described in Table 4, and data reserves, as shown in Table 5: Table 4 Technical Point Description

[0274] Table 5 Data Storage Table

[0275] This application also provides an electronic device including a processor and a memory, wherein the memory stores a program or instructions executable on the processor, and the program or instructions, when executed by the processor, implement the steps of the method described above. This electronic device can implement various embodiments of the above-described intelligent engineering construction management system and achieve the same beneficial effects; further details are omitted here.

[0276] This application also provides a computer program product stored in a storage medium, which is executed by at least one processor to implement the steps of the method described above. This computer program product can implement various embodiments of the above-described engineering construction management intelligent agent system and achieve the same beneficial effects, which will not be elaborated upon here.

[0277] The above are merely preferred embodiments of this application. It should be noted that this application is not limited to the above embodiments. For those skilled in the art, several improvements and modifications can be made without departing from the principles of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should also be considered within the scope of protection of this application.

Claims

1. A railway electrical engineering management method, characterized in that, An intelligent agent system for engineering construction management is applied, comprising a vehicle safety agent, a construction collaboration agent, and a mobile terminal agent. The vehicle safety agent, the construction collaboration agent, and the mobile terminal agent all employ a hybrid architecture combining a BDI model and a finite state machine, and coordinate through an event-driven bus. The method includes: Based on the situational awareness module of the aforementioned train protection intelligent agent, train location information, train speed information, and construction section status information are periodically acquired. The train speed information is used to characterize the train's travel speed, and the construction section status information includes occupancy status and safety status. Based on the train location information, the train speed information, and the status information of the construction section, calculate the spatial overlap length and temporal overlap interval between the train operation section and the construction section; Risk assessment is performed based on the spatial overlap length, the temporal overlap interval, and the train speed information to generate construction safety early warning information, which is then sent to the construction collaborative intelligent agent and the mobile terminal intelligent agent. The construction collaborative intelligent agent can generate railway electrical management information based on the construction safety early warning information. The railway electrical management information includes at least one of the following: construction plan adjustment information to achieve the safe evacuation of construction personnel and train operation adjustment information.

2. The method according to claim 1, characterized in that, The risk assessment based on the spatial overlap length, the temporal overlap interval, and the train speed information, and the output of construction safety early warning information, includes: If the spatial overlap length is greater than zero and the time overlap interval is completely contained, a first-level construction safety early warning message is output. The complete inclusion of the time overlap interval is used to indicate that the construction process of the construction section is within the time period covered by the train's operation. If the spatial overlap length is greater than zero, the time overlap interval duration is greater than or equal to the preset duration threshold, and the time overlap interval is not completely contained, a secondary construction safety early warning message is output. If the spatial overlap length is greater than zero and the duration of the temporal overlap interval is less than a preset duration threshold, a level three construction safety early warning message is output.

3. The method according to claim 2, characterized in that, All levels of construction safety early warning information meet the pre-triggering conditions: Based on the remaining distance between the train and the construction section, and the train's speed, the arrival time is calculated, which is used to characterize the time required for the train to arrive at the construction section. If the arrival time is greater than or equal to a preset time, construction safety warning information at all levels will be triggered accordingly, and the preset time is greater than or equal to 2 minutes.

4. The method according to any one of claims 1 to 3, characterized in that, The greater the travel speed, the greater the weight of the train speed information in the risk assessment of the output construction safety early warning information.

5. A railway electrical engineering management device, characterized in that, An intelligent agent system for engineering construction management is applied, comprising a vehicle safety agent, a construction collaboration agent, and a mobile terminal agent. The vehicle safety agent, the construction collaboration agent, and the mobile terminal agent all adopt a hybrid architecture combining the BDI model and a finite state machine, and coordinate through an event-driven bus. The device includes: The information acquisition module is used to periodically acquire train location information, train speed information, and construction section status information based on the situational awareness module of the train protection intelligent agent. The train speed information is used to characterize the train's travel speed, and the construction section status information includes occupancy status and safety status. The calculation module is used to calculate the spatial overlap length and temporal overlap interval between the train operating section and the construction section based on the train position information, the train speed information, and the status information of the construction section. The risk assessment module is used to assess risks based on the spatial overlap length, the temporal overlap interval, and the train speed information, generate construction safety early warning information, and send the construction safety early warning information to the construction collaborative intelligent agent and the mobile terminal intelligent agent. The generation module is used by the construction collaborative intelligent agent and the mobile terminal intelligent agent to generate railway electrical management information based on the construction safety early warning information. The railway electrical management information includes at least one of the following: construction plan adjustment information to achieve the safe evacuation of construction personnel and train operation adjustment information.

6. The apparatus according to claim 5, characterized in that, The risk assessment module is specifically used for: If the spatial overlap length is greater than zero and the time overlap interval is completely contained, a first-level construction safety early warning message is output. The complete inclusion of the time overlap interval is used to indicate that the construction process of the construction section is within the time period covered by the train's operation. If the spatial overlap length is greater than zero, the time overlap interval duration is greater than or equal to the preset duration threshold, and the time overlap interval is not completely contained, a secondary construction safety early warning message is output. If the spatial overlap length is greater than zero and the duration of the temporal overlap interval is less than a preset duration threshold, a level three construction safety early warning message is output.

7. The apparatus according to claim 6, characterized in that, Before triggering construction safety early warning information at all levels, it also includes: Based on the remaining distance between the train and the construction section, and the train's speed, the arrival time is calculated, which is used to characterize the time required for the train to arrive at the construction section. If the arrival time is greater than or equal to a preset time, construction safety warning information at all levels will be triggered accordingly, and the preset time is greater than or equal to 2 minutes.

8. The apparatus according to any one of claims 5 to 7, characterized in that, The greater the travel speed, the greater the weight of the train speed information in the risk assessment of the output construction safety early warning information.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the method as described in any one of claims 1 to 4.

10. A computer program product, characterized in that, The program product is stored in a storage medium and is executed by at least one processor to implement the steps of the method as described in any one of claims 1 to 4.

Images