Natural language and database driven parameterized three-dimensional modeling method and system

Abstract

The application provides a natural language and database driven parameterized three-dimensional modeling method and system, and belongs to the technical field of civil engineering modeling. The method constructs a function function library (including a structure creation unit, a section operation unit and a parameter management unit), receives user instructions through a natural language dialog box, parses the instructions into modeling parameters with the help of natural language processing technology, and automatically calls corresponding function functions in the function function library. Then, according to the function function calling result, the node coordinates, geometric topological relationship, section attribute and other information are generated and stored into the database, and finally the three-dimensional model is generated based on the database. By using the natural language and database driven parameterized three-dimensional modeling method and system, the learning and modification costs are significantly reduced, the manual input errors are reduced, the modeling efficiency and team collaboration ability are improved, and the method and system can be widely used in building, bridge, tunnel and steel structure engineering.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of civil engineering modeling, and in particular to a parameterized three-dimensional modeling method and system based on natural language and database driving. BACKGROUND

[0002] In the field of civil engineering, three-dimensional modeling is a key technical means to realize engineering design, construction simulation and project management. However, the traditional three-dimensional modeling method mainly relies on the graphical interface of professional software (such as Revit, 3D3S, AutoCAD, etc.), and users need to locate the command through complex multi-level menus, manually input a large number of parameters, and point-select to generate geometric models. This method has many shortcomings:

[0003] (1) High learning cost.

[0004] Software difference: Different modeling platforms have independent menus, toolbars, modeling logic, command sets, operation processes and functions. For example, Revit focuses on the parametric design of building information modeling (BIM), while AutoCAD pays more attention to the flexibility of two-dimensional drawing and three-dimensional modeling, and 3D3S focuses on the finite element analysis and design of steel structure. Users need to spend a lot of time and effort to familiarize themselves with the function entry and operation process of each software, which not only increases the individual learning burden, but also significantly increases the personnel training cost of enterprises.

[0005] Complex operation: Traditional modeling software usually requires users to master complex operation skills, such as accurate point selection, dragging, parameter input, etc., which is difficult for beginners and is prone to modeling errors due to improper operation.

[0006] (2) Large amount of data and prone to errors.

[0007] Parameter input is tedious: Three-dimensional modeling involves a large amount of parameter data, such as cross-sectional dimensions (width, height, thickness), material properties (elastic modulus, Poisson's ratio, thermal expansion coefficient, etc.). Traditional modeling methods rely on manual input of these complex parameter data, which not only consumes time and effort, but also easily leads to data input errors due to negligence or misoperation, thereby affecting the accuracy of the model and the reliability of subsequent analysis.

[0008] Modification is difficult: In traditional modeling software, many geometric features (such as cross-sectional width, cross-sectional height, etc.) are not fully parameterized. Once some parameters in the model need to be adjusted, for example, changing the height of all columns from 3 meters to 3.5 meters, the user must manually adjust the height of all related components, which is a huge amount of work and inefficient. This inefficient modification method greatly limits the flexibility of the design phase, which is not conducive to quickly responding to engineering change requirements.

[0009] The above problems to some extent restrict the wide promotion and application of three-dimensional modeling technology in the field of civil engineering. Specifically, it is shown that:

[0010] Industry efficiency is limited: high learning cost and cumbersome operation process increase the difficulty of popularizing three-dimensional modeling technology, so that many small enterprises or project teams have difficulty in fully utilizing its advantages, thereby affecting the overall efficiency improvement of the whole industry.

[0011] Innovation development is hindered: the inefficiency and error-prone nature of traditional modeling methods limit the innovative thinking of designers, making it difficult to quickly realize the design and optimization of complex structures, which is not conducive to the innovation development of the industry. SUMMARY

[0012] The purpose of the present application is to provide a natural language and database driven parameterized three-dimensional modeling method and system, which does not require professional software operation, significantly reduces the learning and modification cost, reduces the manual input error, improves the modeling efficiency and team collaboration ability, and can be widely used in building, bridge, tunnel and steel structure engineering.

[0013] To achieve the above purpose, the present application provides a natural language and database driven parameterized three-dimensional modeling method, comprising the following steps:

[0014] Step S1, constructing a function function library, including a structure creation unit, a section operation unit and a parameter management unit;

[0015] Step S2, receiving user instructions through a natural language dialog box;

[0016] Step S3, using natural language processing technology to parse the user instructions into modeling parameters, and automatically mapping and calling the corresponding function functions in the function function library;

[0017] Step S4, generating node coordinates, geometric topological relations and section attributes according to the function function calling results, and storing them into a database;

[0018] Step S5, generating a three-dimensional model based on the database.

[0019] Preferably, in step S1, the user instructions include input modeling requirements and parameters.

[0020] Preferably, in step S1, the function function library contains a chain calling mechanism: when the target function function is executed, the dependent pre-function functions are automatically triggered and sequentially executed.

[0021] Preferably, in step S4, the database includes:

[0022] A node table for storing node numbers and three-dimensional coordinates of rod end points;

[0023] section attribute table, storing section number, section shape code, size number;

[0024] section size table, storing section shape code, size number, section size;

[0025] bar table, storing bar number, bar end node number, section number, azimuth angle;

[0026] member table, storing member number, bar number list;

[0027] After modifying a single data table record, all associated data tables are automatically updated.

[0028] Preferably, in step S5, the modeling process specifically includes:

[0029] drawing point objects according to node table coordinates;

[0030] generating section objects according to section attribute table and section size table parameters;

[0031] forming bars based on bar table information along end point straight lines;

[0032] integrating all bar information belonging to the same member into the member table.

[0033] Preferably, in step S5, when the included angle between adjacent bars is less than a preset angle, the bars are automatically identified as the same member object during member generation.

[0034] Preferably, the automatic identification of the same member object is performed according to the following steps:

[0035] Step S51, iterate through all bars to generate an adjacent bar dictionary, with the key being the bar number and the value being a list of all adjacent bar numbers, denoted as list 1;

[0036] Step S52, select one bar from the bars not assigned to any member as the initial bar, create an initial bar number list, denoted as list 2, and add the number of the initial bar to list 2;

[0037] Step S53, find the list of adjacent bar numbers corresponding to the initial bar in the adjacent bar dictionary, and iterate through each bar number in the list of adjacent bar numbers;

[0038] Step S54, for the currently iterated bar, determine whether the number of the currently iterated bar already exists in list 2, and if it exists, skip the currently iterated bar;

[0039] Step S55, if the number of the current traversed bar member is not in list 2, the angle value between the current traversed bar member and the initial bar member is calculated according to the direction vectors of the two bar members, if the angle is less than the preset angle, the number of the current traversed bar member is added to list 2;

[0040] Step S56, the newly added bar member is taken as a new initial bar member, the corresponding adjacent bar member number list is found in the adjacent bar member dictionary, and steps S54 to S55 are repeated;

[0041] Step S57, when there is no new bar member added in the bar member number list, the component generation is completed, list 2 is saved to the component table, and all bar members in the component are marked as allocated state;

[0042] Step S58, a new initial bar member is selected from the remaining unallocated bar members, and steps S52 to S57 are repeated until all bar members are completed.

[0043] The application also provides a natural language and database driven parameterized three-dimensional modeling system, comprising:

[0044] A function function library module is used for constructing and storing a function function library, the library contains a structure creation unit, a section operation unit and a parameter management unit, and has a chain calling mechanism, which can automatically trigger and sequentially execute the dependent pre-function functions;

[0045] A natural language interaction module: provides a natural language dialogue box, receives user input modeling requirements and parameters, and uses natural language processing technology to parse user instructions into modeling parameters, automatically maps and calls corresponding function functions in the function function library;

[0046] A database module: used for storing and managing node coordinates, geometric topological relations, section attributes, bar member information and component information generated in the modeling process, including node table, section attribute table, section size table, bar member table and component table, and has a data linkage updating function, which automatically updates all associated data tables after modifying any table record;

[0047] A three-dimensional model generation module: generates a three-dimensional model based on the data in the database in real time, including drawing a point object according to the node table coordinates, generating a section object according to the section attribute table and section size table parameters, laying out a bar member along the endpoint straight line based on the bar member table information, and integrating all bar member information belonging to the same component into the component table, while having intelligent merging logic, which automatically identifies the same component object when the angle between adjacent bar members is less than the preset angle.

[0048] Therefore, the application adopts the above-mentioned natural language and database driven parameterized three-dimensional modeling method and system, and has the following beneficial technical effects:

[0049] (1) Significantly reduce the learning cost.

[0050] Users do not need to spend a lot of time and effort to familiarize themselves with the complex interface layout, modeling logic and operation process of different modeling software. They can complete the generation of three-dimensional models by inputting modeling requirements and parameters in any order and format through a simple natural language dialog box. This reduces the learning threshold, enabling beginners and non-professionals to quickly get started and reducing personnel training costs.

[0051] (2) Improve modeling efficiency and accuracy.

[0052] The present application stores modeling information such as nodes, cross sections, rods and components in specialized data tables through a database-driven parameterized modeling method, and realizes the association between data tables, avoiding the tedious operation of manual adjustment one by one in traditional modeling, greatly improving the modeling efficiency.

[0053] The present application can accurately map the user's natural language instructions to the pre-designed function library and automatically identify the specific functions of each function and the required modeling parameters. In scenarios involving chain operations, the chain invocation mechanism in the function library ensures that each modification operation is executed in the correct modeling process, avoiding modeling failures caused by incorrect operation order and improving the accuracy and reliability of modeling.

[0054] (3) Enhance team collaboration efficiency.

[0055] In traditional three-dimensional modeling, if you need to select a component, you usually have to point to all the rods included in the component one by one, which is inefficient. The present application proposes that when the angle between adjacent rods is less than a preset angle (to accommodate curved components), they can be automatically identified and merged into the same component object. This automatic identification and merging mechanism not only simplifies the modeling process, but also facilitates collaboration and communication among team members, avoiding repetitive work and errors caused by information asymmetry, and enhancing team collaboration efficiency.

[0056] (4) Reduce data input errors.

[0057] The present application notes the specific function of each function, the required modeling parameters, the specific meaning of each modeling parameter and the default parameter values commonly used in the preset project in the function library. If the user does not provide specific input, the system will automatically use the default values to execute. This design not only simplifies user operations, but also reduces the risk of model construction failure caused by missing or input errors. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 Flowchart of the natural language and database-driven parameterized three-dimensional modeling method of the first embodiment of the present application.

[0059] Figure 2 Database table relationship architecture diagram for structure member data association. DETAILED DESCRIPTION

[0060] The technical solutions of the present application are further described below by means of the accompanying drawings and examples.

[0061] Unless otherwise defined, technical or scientific terms used in the present application shall have the usual meaning understood by one of ordinary skill in the art to which the present application pertains.

[0062] Example 1

[0063] As shown in the following figure, the natural language and database-driven parameterized three-dimensional modeling method comprises the following steps: Figure 1

[0064] Step S1, a function library is constructed to connect DeepSeek and the database-driven parameterized three-dimensional modeling function library, which includes different functional units such as structure creation unit, cross-section operation unit and parameter management unit.

[0065] The structure creation unit is used to convert natural language instructions into geometric data of plate shell structures, space truss structures, etc.

[0066] The cross-section operation unit is used to perform cross-section lookup and size modification operations.

[0067] The parameter management unit is used to associate parameter descriptions and preset default values for each operation.

[0068] Each functional unit is provided with a corresponding function function, and the function function is annotated with specific function functions, required modeling parameters, specific meanings of each modeling parameter and preset default parameter values commonly used in engineering.

[0069] If the user does not provide specific input, the default value will be automatically executed. This can simplify user operation and reduce model construction failure caused by parameter omission or input error.

[0070] The function function library contains a chain calling mechanism: when the target function function is executed, the dependent pre-function functions are automatically triggered and executed in sequence. For example, when the user gives the instruction "change the cross-section height to 450mm" in the natural language dialogue box, the normal modification process should be "get cross-section shape - modify cross-section size". Since DeepSeek itself cannot directly get the shape of the cross-section in the current model, the function library will automatically execute the following process:

[0071] (1) Execute the function function of looking up the cross-section shape.

[0072] ​In the function function of modifying the section size, add a note "execute the function function of searching the section shape before executing this function", when DeepSeek receives the user instruction and executes the function function of modifying the section size, it will identify the note and execute the function function of searching the section shape first.

[0073] (2) Deliver the query result.

[0074] DeepSeek obtains the query result, and then determines to execute the corresponding function function of modifying the section size according to the section shape.

[0075] Step S2, receive the modeling requirements and parameters input by the user through the natural language dialogue box.

[0076] The constructed system provides a simple natural language dialogue box, and the user inputs the modeling requirements and modeling parameters (such as geometric size, material attribute, etc.) in any order and format through natural language. The user does not need to remember complex operation instructions or be familiar with the interfaces of various platforms to complete the generation of three-dimensional models, which greatly reduces the learning threshold and improves the modeling efficiency.

[0077] Step S3, based on the tool_calls technology of DeepSeek, the natural language instruction of the user can be mapped to the pre-designed function library, and the specific function of each function function in the library and the required modeling parameters can be automatically recognized. DeepSeek understands the modeling requirements of the user through the natural language input by the user, analyzes the specific modeling parameters, and autonomously decides to call the specific function function in the function library.

[0078] Step S4, according to the function function call result, generate node coordinates, geometric topological relationship, and section attribute, and store them in the database.

[0079] Among them, the generation method of node coordinates is as follows:

[0080] Integrate node coordinate information based on a series of interface programs of commonly used finite element software such as SAP2000, MIDAS, and 3D3S;

[0081] Generate node coordinate information based on a grid structure parameterization program of mathematical equations (such as spherical equation x 2 +y 2 +z 2 =R 2 , x, y, z represent coordinate variables, and R represents the radius of the sphere);

[0082] The grid structure parameterization program based on an arbitrary curved surface generates node coordinate information, for example, the curved surface is projected onto an xy plane, x and y coordinates of each grid point are obtained based on a preset equal division number on the xy plane, and then the grid points on the xy plane are projected along a vertical direction to the curved surface, so that node coordinates (x, y, z) on the arbitrary curved surface are obtained.

[0083] Different data tables are divided in the database, and different data tables store different types of information respectively.

[0084] The node table stores node numbers and three-dimensional coordinates of the end points of the rod members.

[0085] The cross-section attribute table stores a cross-section number, a cross-section shape code (such as a box cross-section and a rectangular cross-section), and a size number.

[0086] The cross-section size table stores a cross-section shape code, a size number, and a cross-section size.

[0087] The rod member table stores a rod member number, end point node numbers of the rod member, a cross-section number, and an azimuth angle.

[0088] The component table stores a component number and a rod member number list (converted into a text format and stored in the database).

[0089] As shown in FIG. 1, the association modes between the data tables are as follows. Figure 2

[0090] The node table and the rod member table: the node numbers of the node table correspond to the end point node numbers of the rod member table, and the node information of the rod member can be obtained by querying the node numbers of the node table through the end point node numbers of the rod member table.

[0091] The rod member table and the cross-section attribute table: the cross-section numbers of the cross-section attribute table correspond to the cross-section numbers of the rod member table, and the cross-section attribute of the rod member can be obtained by querying the cross-section numbers of the cross-section attribute table through the cross-section numbers of the rod member table.

[0092] The cross-section attribute table and the cross-section size table: the size information can be obtained by querying the size numbers of the corresponding cross-section size table through the size numbers of the cross-section attribute table according to the shape code of the cross-section attribute table.

[0093] The rod member table and the component table: the rod member number list of the component table contains all the rod member numbers of the component, and the rod member numbers correspond to the rod member numbers of the rod member table.

[0094] Step S5: generating a three-dimensional model based on the database.

[0095] A point object is drawn according to the coordinates of the node table.

[0096] ​Generating a section object according to the section attribute table and section size table parameters;

[0097] Based on the bar table information, a bar is lofted along the straight line of the end points.

[0098] All bar information belonging to the same component is integrated into the component table.

[0099] The component generation includes intelligent merging logic: when the included angle between adjacent bars is less than a preset angle, it is automatically identified as the same component object.

[0100] The automatic identification algorithm includes the following steps (taking 5° as the preset angle):

[0101] Step S51, all bars are traversed to generate an adjacent bar dictionary, the key of the dictionary is the bar number, and the value is a list composed of all bar numbers adjacent to the bar, denoted as list 1.

[0102] Step S52, select one bar from the bars not assigned to any component as the initial bar, create an initial bar number list denoted as list 2, and add the number of the initial bar to list 2.

[0103] Step S53, find the adjacent bar number list corresponding to the initial bar in the adjacent bar dictionary, and traverse each bar number in the adjacent bar number list in turn.

[0104] Step S54, for the bar currently traversed, determine whether the number of the bar currently traversed exists in list 2, if it exists, skip the bar currently traversed.

[0105] Step S55, if the number of the bar currently traversed is not in list 2, calculate the included angle value between the bar currently traversed and the initial bar according to the direction vectors of the two bars.

[0106] If the included angle is less than 5°, add the bar number to list 2; if the included angle is greater than or equal to 5°, determine that the bar does not belong to the current component, do not add it to the component list and skip it.

[0107] Step S56, take the newly added bar as a new initial bar, find the corresponding adjacent bar number list in the adjacent bar dictionary, and repeat steps S54 to S55.

[0108] Step S57, when there is no new bar added to the bar number list, the component generation is completed, list 2 is saved to the component table, and all bars in the component are marked as assigned.

[0109] Step S58, select a new initial bar from the remaining unassigned bars, repeat steps S52 to S57 until all bars are assigned.

[0110] It is worth noting that the content not elaborated in the present application is all the prior art, which is well known to those skilled in the art.

[0111] Therefore, the present application adopts the above-mentioned natural language and database-driven parameterized three-dimensional modeling method and system, which does not require professional software operation, significantly reduces the learning and modification cost, reduces the manual input error, improves the modeling efficiency and team collaboration ability, and can be widely used in building, bridge, tunnel and steel structure engineering.

[0112] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, but not to limit them, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that: the technical solutions of the present application can still be modified or replaced by the equivalent, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for natural language based and database driven parameterized 3D modeling, characterized in that, The method comprises the following steps: Step S1, constructing a function function library, including a structure creation unit, a section operation unit and a parameter management unit; The structure creation unit is used for converting natural language instructions into geometric data of plate shell structures and three-dimensional truss structures; The section operation unit is used for performing section lookup and size modification operations; The parameter management unit is used for associating parameter descriptions and preset default values for each operation; Step S2, receiving user instructions through a natural language dialog box; Step S3, using natural language processing technology to parse the user instructions into modeling parameters, and automatically mapping and calling corresponding function functions in the function function library; Step S4, generating node coordinates, geometric topological relationships and section attributes according to the function function call results, and storing them in a database; Step S5, generating a three-dimensional model based on the database; In step S1, the function function library contains a chain calling mechanism: when the target function function is executed, the dependent pre-function functions are automatically triggered and sequentially executed; In step S4, the database includes: A node table storing node numbers and three-dimensional coordinates of rod end points; A section attribute table storing section numbers, section shape codes and size numbers; A section size table storing section shape codes, size numbers and section sizes; A rod table storing rod numbers, rod end point node numbers, section numbers and azimuth angles; A component table storing component numbers and rod number lists; After modifying a single data table record, all associated data tables are automatically updated; In step S5, the modeling process specifically includes: Drawing point objects according to the node table coordinates; Generating section objects according to the parameters in the section attribute table and the section size table; Based on the rod table information, laying out a rod along the end point straight line; Integrating all rod information belonging to the same component into the component table; In step S5, when the included angle between adjacent rods is less than a preset angle, the same component object is automatically identified.

2. The natural language based and database driven parameterized three-dimensional modeling method according to claim 1, wherein, In step S2, the user instructions include input modeling requirements and parameters.

3. The natural language based and database driven parameterized three-dimensional modeling method according to claim 1, wherein, The steps of automatically identifying the same component object are as follows: Step S51, traversing all rods to generate an adjacent rod dictionary, the key of the dictionary being a rod number and the value being a list of all adjacent rod numbers, denoted as list 1; Step S52, selecting a rod from the rods not assigned to any component as an initial rod, creating an initial rod number list denoted as list 2, and adding the number of the initial rod to list 2; Step S53, finding the adjacent rod number list corresponding to the initial rod in the adjacent rod dictionary, and sequentially traversing each rod number in the adjacent rod number list; Step S54, for the currently traversed rod, determining whether the number of the currently traversed rod exists in list 2, and if it exists, skipping the currently traversed rod; Step S55, if the number of the currently traversed rod is not in list 2, calculating the included angle between the currently traversed rod and the initial rod according to the direction vectors of the two rods, and if the included angle is less than a preset angle, adding the number of the currently traversed rod to list 2. Step S56, the newly added bar is taken as a new initial bar, the corresponding adjacent bar number list is found in the adjacent bar dictionary, and steps S54 to S55 are repeated; Step S57, when there is no new bar added in the bar number list, the component generation is completed, the list 2 is saved to the component table, and all bars in the component are marked as allocated state; Step S58, a new initial bar is selected from the remaining unallocated bars, and steps S52 to S57 are repeated until all bars are allocated.

4. A natural language and database driven parametric 3D modeling system, characterized in that, Comprise: The function function library module is used for building and storing the function function library, the library contains the structure creation unit, the section operation unit and the parameter management unit, and has chain calling mechanism, can automatically trigger and sequentially execute the dependent pre-function function; The structure creation unit is used for converting natural language instructions into geometric data of plate shell structure and three-dimensional truss structure; The section operation unit is used for executing section search and size modification operation; The parameter management unit is used for associating parameter description and preset default value for each operation; Natural language interaction module: provide natural language dialogue box, receive user input modeling demand and parameter, and use natural language processing technology to parse user instruction into modeling parameter, automatically map and call corresponding function function in function function library; Database module: for storing and managing node coordinates, geometric topological relationship, section attribute, bar information and component information generated in the modeling process, including node table, section attribute table, section size table, bar table and component table, and has data linkage update function, modifies any table record automatically updates all associated data tables; Three-dimensional model generation module: generate three-dimensional model based on data in database in real time, including drawing point object according to node table coordinates, generating section object according to section attribute table and section size table parameters, forming bar based on bar table information along endpoint straight line lofting, and integrating all bar information belonging to the same component into component table, at the same time, has intelligent merging logic, when the included angle between adjacent bars is less than the preset angle, the same component object is automatically identified.

Images