Railway line intelligent design and optimization method based on large language model

Abstract

The application discloses a railway line intelligent design and optimization method based on a large language model, which comprises the following steps: S1, setting a railway line intelligent design AI environment; S2, constructing a railway line intelligent design command for AI calling; S3, creating an AI command set configuration document of the railway line intelligent design; S4, constructing a railway line intelligent design workflow knowledge base in a natural language form; S5, constructing a railway line intelligent design MCP server; S6, configuring a railway line intelligent design platform; S7, based on the railway line intelligent design platform, a user inputs a natural language to drive an AI command to perform railway line intelligent design and scheme optimization; and S8, after the railway line intelligent design and optimization are completed, the user inputs a natural language to select a railway line scheme, and then outputs a result map and a result table of the line scheme. The method can greatly improve work efficiency, improve design experience, and has high engineering application value.

Description

Technical Field

[0001] This invention relates to the field of railway line design, and in particular to a method for intelligent design and optimization of railway lines based on a large language model. Background Technology

[0002] Railway line design is a core component of railway engineering construction, and intelligent design and optimization of railway lines is a crucial topic in this field. Traditional design methods heavily rely on engineers manually completing CAD drawings, parameter calculations, scheme comparisons, and output, resulting in cumbersome processes, high repetitive workloads, and low efficiency. With increasing engineering complexity and compressed design cycles, manually-led design methods are no longer sufficient to meet the demands for efficient and precise engineering.

[0003] Currently, although some CAD secondary development tools can achieve partial automation, engineers still need to write specific scripts or manually trigger commands, which cannot achieve natural language-driven intelligent interaction. At the same time, design knowledge and workflows are mostly based on engineers' personal experience, lacking standardized knowledge bases and unified calling platforms, resulting in low knowledge reuse rates and inconsistent design quality.

[0004] The development of large language models and professional domain server technology has provided a new path for intelligent engineering design. However, existing technologies have not yet been deeply integrated with railway line design scenarios. Some intelligent agents have achieved simple AI operations on design software, but require a large number of specific modifications and interface definitions to the software, resulting in a huge workload. It is difficult to synchronize software updates with AI, and it cannot meet the collaborative needs of AI calls and manual operations, making it difficult to achieve full-process automation and intelligence from scheme conception to output.

[0005] Therefore, there is an urgent need for a railway line intelligent design and optimization method based on a large language model to solve the pain points of the traditional design mode, improve design efficiency and quality, and improve the working experience of engineers. Summary of the Invention

[0006] In view of the problems and current status of existing intelligent design and optimization methods for railway lines, this invention provides an intelligent design and optimization method for railway lines based on a large language model.

[0007] Therefore, the present invention adopts the following technical solution:

[0008] A method for intelligent design and optimization of railway lines based on a large language model includes the following steps:

[0009] S1, set up the AI ​​environment for intelligent design of railway lines, including setting the working path, project configuration parameters, project basic data and file naming rules for intelligent design of railway lines;

[0010] S2, constructs intelligent design commands for railway lines for AI invocation, including tagging commands and AI commands;

[0011] S3, based on the AI ​​commands obtained in S2, creates a configuration document for the AI ​​command set of intelligent railway line design:

[0012] S4. Based on the AI ​​command set configuration document, construct a knowledge base for intelligent railway line design workflow in natural language form according to the workflow required for actual intelligent railway line design work.

[0013] S5: First, read the AI ​​command set configuration document obtained in S3, create a command list, create a tool function for intelligent design of railway lines based on the content of each AI command, then create an MCP server, and register the tool function to the MCP server to facilitate the calling of AI commands;

[0014] S6, configure the intelligent railway line design platform, including adding a large language model, adding the MCP server mentioned in S5, and loading the intelligent railway line design workflow knowledge base built in S4;

[0015] S7, based on the intelligent railway line design platform obtained from S6, allows users to drive AI commands through natural language input for intelligent railway line design and scheme optimization.

[0016] S8. After the intelligent design and optimization of the railway line is completed, the user selects the railway line scheme by inputting natural language, sets the current plane and the current longitudinal profile, generates and outputs the result map and result table. The file names of the result map and result table are determined according to the naming rules.

[0017] In step S1 above:

[0018] The project configuration parameters include planar design parameters, longitudinal section design parameters, cross section design parameters, engineering cost indicators, and comprehensive cost parameters;

[0019] The project's basic data includes topographic data, geological and geomorphological data, and route selection control vector data. The route selection control vector data includes railways, roads, rivers, lakes, environmental protection, adverse geological conditions, economic hubs, restricted areas, and planning areas.

[0020] The file naming rules include data file naming rules, command execution result file naming rules, and result chart naming rules. The command execution result file naming rule is command name + execution result.

[0021] In step S2 above:

[0022] The marking command is used to assign values ​​to the marking variables, which are global static variables. The values ​​of the marking variables are divided into AI calls and manual calls. The marking command is run before each AI call to the railway line design command of the CAD secondary development to set the marking variables to AI calls. After the design command ends, the marking variables are set to manual calls through the reactor.

[0023] The AI ​​commands are generated by modifying railway line design commands from secondary CAD development, and are used for AI invocation while also being compatible with manual invocation.

[0024] The modifications include adding recognition and processing for AI calls. When the flag variable is an AI call, the command uses the AI ​​environment settings; when the flag variable is a manual call, the command uses manually input data. The modified AI command includes both cases corresponding to AI and manual calls, achieving compatibility.

[0025] Each AI command saves the results parameters and data to a command execution result file. The file name is determined according to the file naming rules. AI commands pass parameters and data to each other through the command execution result file.

[0026] In step S3 above:

[0027] The AI ​​command set configuration document is in JSON file format;

[0028] The AI ​​command includes a command name, function description, and parameter list. Each parameter in the parameter list includes a parameter name, parameter type, parameter description, and return value type; wherein:

[0029] The command names include: Create a new project, Set project configuration parameters, Obtain terrain data, Build a digital model, Obtain geological and geomorphological data, Obtain route selection control vector data, Terrain exploration, Cost map construction, Channel search, Scheme fitting, Scheme optimization, Select current plane, Select current longitudinal section, Plane specification check, Longitudinal section specification check, Adjust broken chains, Adjust plane intersections, Adjust plane curves, Adjust longitudinal section slope change points, Adjust longitudinal section slope length and gradient, Adjust bridges, Adjust tunnels, Adjust grade-separated intersections, Adjust stations, Save horizontal and vertical data, Generate design data table, Generate scheme comparison table, Generate horizontal and vertical section diagrams, and Generate interface data.

[0030] Step S4 above includes:

[0031] Based on the workflow required for intelligent design of actual railway lines, each task is broken down into multiple steps. The information recorded in each step includes the functional description of the corresponding AI command, the completeness check of the required parameters before the AI ​​command is executed, the description of the automatic acquisition method when parameters are missing, the termination of the workflow and the interaction strategy when parameters are abnormal, and the AI ​​command for the next step determined based on the command return value and the completeness check result.

[0032] In step S5 above:

[0033] The utility function connects to the CAD software via a COM component, activates the CAD document, first sends the mark command constructed by S2, then constructs a command string based on the command name and parameter list of the AI ​​command, and then sends the command string to the CAD software;

[0034] The utility functions include docstrings, parameter type annotations, and custom signature attributes, where:

[0035] The document string is set according to the function description, parameter name, and parameter description of the AI ​​command;

[0036] The parameter type annotation is set according to the parameter name, parameter type, and parameter description;

[0037] The custom signature attributes include a parameter list and return value type annotations. First, a mapping between parameter types and Python variable types is established, and then the parameters are added according to the command list and the mapping.

[0038] In step S6 above:

[0039] When adding a large language model, integrate multiple large models that support utility function calls;

[0040] When loading the intelligent design workflow knowledge base for railway lines, specify the knowledge base name and the embedded model;

[0041] The embedding model is an AI model used to map text into numerical vectors, which facilitates retrieval and understanding by large models.

[0042] Step S7 above includes:

[0043] Input natural language containing a description of the railway line design;

[0044] The intelligent railway line design platform uses a large language model to parse the natural language input by users and understand their design intentions.

[0045] Based on the S6-loaded intelligent railway line design workflow knowledge base, semantic matching is performed by embedding numerical vectors of model mapping to generate a workflow containing intelligent railway line design AI commands; the workflow clearly defines the tool function call order, parameters and execution logic;

[0046] Then, the intelligent railway line design platform calls the tool functions registered in the MCP server in sequence according to the generated workflow; the tool functions drive AI commands to complete the intelligent design of railway lines, generate railway line schemes, and draw the line plane and longitudinal profile.

[0047] In step S7 above:

[0048] Users input natural language that describes the optimization of the generated railway route plan;

[0049] The intelligent railway line design platform once again uses a large language model and a knowledge base for intelligent railway line design workflows to generate a workflow containing AI commands for optimizing corresponding line schemes. The intelligent railway line design platform calls tool functions in sequence according to the generated workflow, driving the AI ​​commands to adjust the line schemes and achieve optimization of the railway line schemes.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. This invention innovatively redesigns traditional CAD secondary development commands, eliminating the need for extensive modifications to the core code and operational logic of existing railway line design commands. It rapidly constructs AI commands for intelligent railway line design that are compatible with both AI and manual operations, without requiring separate development of adaptations. This effectively solves the technical pain points of existing technologies, such as the inability of AI calls and manual operations to coordinate, the large workload of software modifications, and the lack of synchronization between AI and software updates. Furthermore, this invention abandons the complex mode of simply passing parameters and data directly between traditional tool functions. Instead, it innovatively uses command execution result files as the carrier for passing parameters and data between AI commands. This not only significantly reduces the number and complexity of parameters passed by the tool functions themselves but also achieves standardized and regulated exchange of design data, improving the efficiency of design data exchange.

[0052] 2. Based on the creation of AI command set configuration documents, this invention innovatively and uniformly constructs a professional railway line intelligent design MCP server and design workflow knowledge base. It systematically organizes and standardizes the scattered individual design experience, professional knowledge, industry standards and design processes, enabling the large language model to correctly understand professional requirements and design functions. This completely solves the core defects of existing intelligent design technologies that "do not understand the profession and cannot accurately respond to design requirements", and significantly improves the reusability of design knowledge and the stability of design quality.

[0053] 3. This invention, based on a large language model, a professional MCP server, and a workflow knowledge base, constructs an intelligent railway line design platform. It innovatively realizes natural language-driven railway line scheme design, optimization, and output of results charts. This method is novel and unique, with a significantly improved level of automation and intelligence compared to existing technologies. By simply inputting design requirements in natural language, the entire workflow from scheme conception, parameter calculation, scheme comparison to result output can be completed, significantly improving work efficiency and design experience. While meeting design requirements, it greatly reduces manual triggering operations, possessing high engineering application value and promising prospects for widespread adoption. (CAD secondary development of railway lines) Attached Figure Description

[0054] Figure 1 This is a flowchart of the intelligent design and optimization method for railway lines according to an embodiment of the present invention. Detailed Implementation

[0055] The technical solution of the invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. Obviously, the following embodiments are only some embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Example

[0057] See Figure 1 The present invention provides a method for intelligent design and optimization of railway lines based on a large language model, comprising the following steps:

[0058] S1, setting up the AI ​​environment for intelligent railway line design, including setting the work path, project configuration parameters, project basic data, and file naming rules for intelligent railway line design, among which:

[0059] The project configuration parameters include planar design parameters, longitudinal section design parameters, cross section design parameters, engineering cost indicators, and comprehensive cost parameters;

[0060] The project's basic data includes topographic data, geological and geomorphological data, and route selection control vector data. The route selection control vector data includes railways, roads, rivers, lakes, environmental protection areas, adverse geological conditions, economic hubs, restricted areas, and planning areas.

[0061] The file naming rules include data file naming rules, command execution result file naming rules, and result chart naming rules. The command execution result file naming rule is command name + execution result.

[0062] S2, constructs intelligent design commands for railway lines for AI invocation, including tagging commands and AI commands, wherein:

[0063] The tagging command is used to assign values ​​to tagging variables (global static variables, the values ​​of which are divided into AI calls and manual calls).

[0064] Before each AI calls the railway line design command developed by CAD, a marking command is run to set the marking variable to AI call. After the design command ends, the marking variable is set to manual call via the reactor.

[0065] The AI ​​commands are generated by modifying (reconstructing) the railway line design commands developed through secondary CAD software. They are used for AI invocation while also being compatible with manual invocation.

[0066] The modifications include adding recognition and processing for AI calls. When the marker variable is an AI call, the command uses the AI ​​environment settings; when the marker variable is a manual call, the command uses manually input data. The modified AI command includes both cases corresponding to AI and manual calls, achieving compatibility.

[0067] Each AI command saves the results parameters and data to a command execution result file. The file name is determined according to the file naming rules. AI commands pass parameters and data to each other through the command execution result file.

[0068] S3, configuration document for creating AI command set for intelligent railway line design:

[0069] The AI ​​command set configuration document is created based on the AI ​​commands obtained in S2. Specifically, the command content of the AI ​​commands obtained in S2 is manually organized to form an AI command set configuration document in JSON file format.

[0070] Each AI command includes a command name, a function description, and a parameter list. Each parameter in the parameter list includes a parameter name, parameter type, parameter description, and return value type.

[0071] The command names include: Create a new project, Set project configuration parameters, Obtain terrain data, Build a digital model, Obtain geological and geomorphological data, Obtain route selection control vector data, Terrain exploration, Cost map construction, Channel search, Scheme fitting, Scheme optimization, Select current plane, Select current longitudinal section, Plane specification check, Longitudinal section specification check, Adjust broken chains, Adjust plane intersections, Adjust plane curves, Adjust longitudinal section slope change points, Adjust longitudinal section slope length and gradient, Adjust bridges, Adjust tunnels, Adjust grade-separated intersections, Adjust stations, Save horizontal and vertical data, Generate design data table, Generate scheme comparison table, Generate horizontal and vertical section diagrams, and Generate interface data.

[0072] In this embodiment, the information in the AI ​​command set configuration document is arranged in the order of the command names mentioned above.

[0073] S4, building a knowledge base for intelligent railway line design workflow:

[0074] Based on the AI ​​command set configuration document for intelligent railway line design, a natural language-based knowledge base for intelligent railway line design workflow is manually constructed according to the workflow required for actual intelligent railway line design work. The specific steps are as follows:

[0075] Each task is broken down into multiple steps. The information recorded in each step includes a functional description of the corresponding AI command, a completeness check of the required parameters before the AI ​​command is executed, an explanation of the automatic parameter acquisition method when parameters are missing, workflow termination and interaction strategies when parameters are abnormal, and the AI ​​command for the next step determined based on the command return value and the completeness check results. The railway line intelligent design workflow knowledge base is written in text files.

[0076] S5, building an intelligent design MCP server for railway lines:

[0077] First, read the AI ​​command set configuration document obtained from S3, create a command list, create a tool function for intelligent design of railway lines based on the content of each AI command, then create an MCP server, and register the tool function to the MCP server for easy calling of AI commands.

[0078] The utility function connects to the CAD software via a COM component, activates the CAD document, first sends the mark command constructed by S2, then constructs a command string based on the command name and parameter list of the AI ​​command, and then sends the command string to the CAD software.

[0079] The utility function includes a docstring, parameter type annotations, and custom signature attributes, which are used by the MCP server to read and understand the utility function, wherein:

[0080] The utility functions include docstrings, parameter type annotations, and custom signature attributes, where:

[0081] The document string is set according to the function description, parameter name, and parameter description of the AI ​​command;

[0082] The parameter type annotation is set according to the parameter name, parameter type, and parameter description;

[0083] The custom signature attributes include a parameter list and return value type annotations. First, a mapping between parameter types and Python variable types is established, and then the parameters are added according to the command list and the mapping.

[0084] S6, equipped with a railway line intelligent design platform:

[0085] The configuration process includes adding a large language model, adding the MCP server described in S5, and loading the railway line intelligent design workflow knowledge base described in S4.

[0086] When adding a large language model, multiple large models that support utility function calls are integrated (API key and API address are set), and a custom name for the large language model is set.

[0087] When loading the knowledge base for intelligent design workflow of railway lines, specify the knowledge base name and the embedded model. The embedded model is an AI model used to map text into numerical vectors, which facilitates retrieval and understanding by large models.

[0088] S7, Intelligent Design and Optimization of Railway Lines:

[0089] Based on the S6-based intelligent railway line design platform, users can use natural language input to drive AI commands for intelligent railway line design and scheme optimization, as detailed below:

[0090] Input natural language containing a description of the railway line design;

[0091] The large language model of the railway line intelligent design platform parses the natural language and understands the user's design intent. Based on the loaded railway line intelligent design workflow knowledge base, it performs semantic matching by embedding numerical vectors of the model mapping to generate a workflow containing railway line intelligent design AI commands. The workflow clarifies the tool function call order, parameters and execution logic. Then, the platform calls the tool functions registered in the added MCP server in sequence according to the generated workflow.

[0092] The tool functions drive AI commands to complete the intelligent design of railway lines, generate railway line schemes, and draw the line plane and longitudinal profile.

[0093] Users input natural language that describes the optimization of the generated railway route plan;

[0094] The intelligent railway line design platform once again uses the large language model and intelligent railway line design workflow knowledge base configured in S6 to generate a workflow containing AI commands for optimizing the corresponding line scheme. The intelligent railway line design platform calls tool functions in sequence according to the generated workflow, driving the AI ​​commands to adjust the line scheme and optimize the railway line scheme.

[0095] S8, intelligent output of results charts:

[0096] After the intelligent design and optimization of the railway line is completed, the user selects the railway line scheme by inputting natural language, sets the current horizontal and vertical alignment, and generates and outputs the result maps and tables. The file names of the result maps and tables are determined according to the naming rules.

Claims

1. A method for intelligent design and optimization of railway lines based on a large language model, characterized in that, Includes the following steps: S1, set up the AI ​​environment for intelligent design of railway lines, including setting the working path, project configuration parameters, project basic data and file naming rules for intelligent design of railway lines; S2, constructs intelligent design commands for railway lines for AI invocation, including tagging commands and AI commands, wherein: The marking command is used to assign values ​​to the marking variables, which are global static variables. The values ​​of the marking variables are divided into AI calls and manual calls. The marking command is run before each AI call to the railway line design command of the CAD secondary development to set the marking variables to AI calls. After the design command ends, the marking variables are set to manual calls through the reactor. The AI ​​commands are generated by modifying the railway line design commands in the secondary development of CAD, and are used for AI invocation while also being compatible with manual invocation. The modifications include adding recognition and processing for AI calls. When the flag variable is an AI call, the command uses the AI ​​environment settings; when the flag variable is a manual call, the command uses manually input data. The modified AI command includes both cases corresponding to AI and manual calls, achieving compatibility. Each AI command saves the result parameters and data to the command execution result file. The file name is determined according to the file naming rules. AI commands pass parameters and data to each other through the command execution result file. S3, based on the AI ​​commands obtained in S2, creates an AI command set configuration document for intelligent railway line design, including: The AI ​​command set configuration document is in JSON file format; the content of the AI ​​command includes command name, function description and parameter list, and each parameter in the parameter list includes parameter name, parameter type, parameter description and return value type; S4. Based on the AI ​​command set configuration document, construct a knowledge base for intelligent railway line design workflow in natural language form according to the workflow required for actual intelligent railway line design work. S5: First, read the AI ​​command set configuration document obtained in S3, create a command list, create a tool function for intelligent railway line design based on the content of each AI command, then create an MCP server, and register the tool function to the MCP server for easy invocation of AI commands; wherein: The utility function connects to the CAD software via a COM component, activates the CAD document, first sends the mark command constructed by S2, then constructs a command string based on the command name and parameter list of the AI ​​command, and then sends the command string to the CAD software; The utility functions include docstrings, parameter type annotations, and custom signature attributes, where: The document string is set according to the function description, parameter name, and parameter description of the AI ​​command; The parameter type annotation is set according to the parameter name, parameter type, and parameter description; S6, configure the intelligent railway line design platform, including adding a large language model, adding the MCP server mentioned in S5, and loading the intelligent railway line design workflow knowledge base built in S4; S7, based on the intelligent railway line design platform obtained from S6, allows users to perform intelligent railway line design and scheme optimization by inputting natural language-driven AI commands. The intelligent railway line design includes: Input natural language containing a description of the railway line design; The intelligent railway line design platform uses a large language model to parse the natural language input by users and understand their design intentions. Based on the S6-loaded intelligent railway line design workflow knowledge base, semantic matching is performed by embedding numerical vectors of model mapping to generate a workflow containing intelligent railway line design AI commands; the workflow clearly defines the tool function call order, parameters and execution logic; Then, the intelligent railway line design platform calls the tool functions registered in the MCP server in sequence according to the generated workflow; the tool functions drive AI commands to complete the intelligent design of railway lines, generate railway line schemes, and draw the line plane and longitudinal profile. S8. After the intelligent design and optimization of the railway line is completed, the user selects the railway line scheme by inputting natural language, sets the current plane and the current longitudinal profile, generates and outputs the result map and result table. The file names of the result map and result table are determined according to the naming rules.

2. The intelligent design and optimization method for railway lines according to claim 1, characterized in that, In S1: The project configuration parameters include planar design parameters, longitudinal section design parameters, cross section design parameters, engineering cost indicators, and comprehensive cost parameters; The project's basic data includes topographic data, geological and geomorphological data, and route selection control vector data. The route selection control vector data includes railways, roads, rivers, lakes, environmental protection, adverse geological conditions, economic hubs, restricted areas, and planning areas. The file naming rules include data file naming rules, command execution result file naming rules, and result chart naming rules. The command execution result file naming rule is command name + execution result.

3. The intelligent design and optimization method for railway lines according to claim 2, characterized in that, In S3: The command names include: Create a new project, Set project configuration parameters, Obtain terrain data, Build a digital model, Obtain geological and geomorphological data, Obtain route selection control vector data, Terrain exploration, Cost map construction, Channel search, Scheme fitting, Scheme optimization, Select current plane, Select current longitudinal section, Plane specification check, Longitudinal section specification check, Adjust broken chains, Adjust plane intersections, Adjust plane curves, Adjust longitudinal section slope change points, Adjust longitudinal section slope length and gradient, Adjust bridges, Adjust tunnels, Adjust grade-separated intersections, Adjust stations, Save horizontal and vertical data, Generate design data table, Generate scheme comparison table, Generate horizontal and vertical section diagrams, and Generate interface data.

4. The intelligent design and optimization method for railway lines according to claim 3, characterized in that, S4 includes: Based on the workflow required for intelligent design of actual railway lines, each task is broken down into multiple steps. The information recorded in each step includes the functional description of the corresponding AI command, the completeness check of the required parameters before the AI ​​command is executed, the description of the automatic acquisition method when parameters are missing, the termination of the workflow and the interaction strategy when parameters are abnormal, and the AI ​​command for determining the next step based on the command return value and the completeness check result.

5. The intelligent design and optimization method for railway lines according to claim 4, characterized in that, In S6: When adding a large language model, integrate multiple large models that support utility function calls; When loading the intelligent design workflow knowledge base for railway lines, specify the knowledge base name and the embedded model; The embedding model is an AI model used to map text into numerical vectors, facilitating retrieval and understanding by large models.

6. The intelligent design and optimization method for railway lines according to claim 5, characterized in that, In S7: Users input natural language containing a description of the optimized railway route plan in response to the generated railway route plan. The intelligent railway line design platform again uses a large language model and a knowledge base for intelligent railway line design workflows to generate a workflow containing AI commands for optimizing corresponding line schemes. The intelligent railway line design platform calls tool functions in sequence according to the generated workflow, driving the AI ​​commands to adjust the railway line schemes and optimize them.

Images