Optimization method for generating BIM model through natural language based on large language model

Abstract

The invention relates to an optimization method for generating a BIM model by a natural language based on a large language model. The optimization method comprises the following steps: step 1, constructing an EXCEL database containing a plurality of predefined cue word templates; each cue word template comprises two columns, namely a type column and a cue word column, a BIM model function template is called in cue words, and default values of parameters are set; 2, receiving a natural language description input by a user; identifying and matching with the cue word template library by utilizing a natural language processing technology to generate a complete and structured cue word; step 3, constructing a BIM model function template library corresponding to an API (Application Program Interface) of the BIM software; 4, executing an instruction generated by the cue word template to create or modify the BIM model; according to the method, the problem that complex BIM model codes are difficult to directly generate by a large model at present is solved, the accuracy, efficiency, stability and usability of BIM model generation are remarkably improved, the professional threshold is lowered, and digital transformation of the construction engineering industry is promoted.

Description

Technical Field

[0001] The present invention relates to a building information modeling (BIM) technology, and in particular to a method for automatically creating and optimizing a BIM model by utilizing an artificial intelligence large model. Background Art

[0002] BIM (Building Information Modeling) technology plays a vital role in architectural design, construction, and operational management. It uses digital 3D models to represent a building's geometric, physical, and functional information. However, the construction of traditional BIM models often relies on manual input and design. This process is cumbersome, time-consuming, and susceptible to human influence, making it difficult to meet the modern construction industry's demands for rapid design and easy modification.

[0003] In recent years, the rapid development of artificial intelligence (AI), particularly deep learning, natural language processing, and image recognition, has opened up new possibilities for automated BIM model generation. However, due to the lack of BIM software programming knowledge in large models, the effectiveness of BIM models is poor. Modeling even slightly complex structures is inefficient, and waiting for the large model to generate all the code before executing it is very inefficient, or the generated code cannot even be debugged. Therefore, developing a method for fast and stable model generation has become an urgent problem. Summary of the invention

[0004] Aiming at the current problem that it is difficult to directly generate complex BIM model codes from large models, an optimization method for generating BIM models from natural language based on large language models is proposed.

[0005] The present invention aims to provide an optimization method for generating BIM models from text based on artificial intelligence. By creating a prompt word template function, the method enables large models to more accurately and efficiently convert natural language descriptions into structured parameters, and fill the parameters into the BIM model function template to quickly generate complex BIM models.

[0006] The technical solution of the present invention is:

[0007] An optimization method for generating a BIM model from natural language based on a large language model comprises the following steps:

[0008] Step 1. Create a prompt word template library

[0009] Build an EXCEL database containing multiple predefined prompt templates; each prompt template contains two columns. The first column is the type column, which includes the structural type of the BIM model to be created or specific modification tasks. The second column is the prompt column, which contains the prompts sent to the large model. The prompt calls the BIM model function template and sets the default values of the parameters to explain the required parameters and reference the code attached to the prompt. Each prompt column contains fixed structured statements and reserved replaceable parameters for receiving key information extracted from the natural language description and filling it into the BIM model function template through function calls.

[0010] Step 2. Natural Language Understanding and Parameter Extraction

[0011] Receive the natural language description input by the user; use natural language processing techniques to identify the category of the BIM model module that the user's needs belong to. After identifying the BIM model module, select the appropriate prompt template, use the understanding ability of the large model to match the natural language parameters with the parameter placeholders in the prompt template library, and extract the parameter values of the successfully matched natural language. Fill the extracted parameter values into the parameter placeholders of the prompt template to generate a complete and structured prompt.

[0012] Step 3. Construction of the BIM Model Function Template Library

[0013] Build a BIM model function template library corresponding to the BIM software API; each function template represents a specific BIM model creation or modification operation; the function template contains function calls of the BIM software API and reserves parameter entrances corresponding to the parameter placeholders of the prompt template; the function template library is called by the prompt template library.

[0014] Step 4. Execute the instructions generated by the prompt template to create or modify the BIM model;

[0015] Execute the code or instructions generated by the prompt template library. The instructions will automatically call the function template library containing the parameters required by the user and execute the code to create or modify the corresponding BIM model in the BIM software.

[0016] Furthermore, in step 1, the structural types include grid, bracket, and trestle, and the specific modification tasks include stretching and rotation.

[0017] Furthermore, in step 1, the prompt template is in EXCEL format, and the user can customize and expand the prompt template according to actual needs to adapt to different modeling scenarios and styles.

[0018] Preferably, in step 2, the natural language processing technology includes the jieba word segmentation library.

[0019] Further, in step 2, the template includes default settings and can also be generated if the user does not give any parameters.

[0020] Preferably, in step 3, the BIM software includes Revit and Tekla Structures.

[0021] The beneficial effects of the present invention are as follows:

[0022] The present invention significantly improves the accuracy, efficiency, stability, and usability of BIM model generation, reduces the professional threshold, and promotes the digital transformation of the construction engineering industry.

[0023] Improve the accuracy of BIM model generation: Through the constraint of the prompt template, the ambiguity of natural language understanding is reduced, enabling the large model to more accurately understand the user's modeling intention.

[0024] Enhance the efficiency of BIM model generation: By using predefined function templates, the problem of the large model directly generating complex BIM software code is avoided, accelerating the model generation process.

[0025] Enhance the stability of BIM model generation: The function template ensures that the generated code or operation instructions conform to the specifications of BIM software, reducing generation failures caused by code errors.

[0026] Lower the technical threshold of BIM modeling: Users can generate complex BIM models by simply describing them in natural language without the need to deeply understand the programming interfaces of BIM software.

[0027] Easy to expand and maintain: The prompt template and function template can be conveniently expanded and maintained according to new modeling requirements. Description of the Drawings

[0028] Figure 1 It is a schematic flow diagram of an optimized method for text generation of BIM models based on artificial intelligence provided by an embodiment of the present invention;

[0029] Figure 2 It is a list of large models and BIM software used in the present invention;

[0030] Figure 3 It is a prompt template library provided by an embodiment of the present invention;

[0031] Figure 4 It is a BIM model function template library corresponding to the BIM software Tekla Structures API provided by an embodiment of the present invention, taking the grid as an example;

[0032] Figure 5 It is a schematic diagram of the grid generated by Tekla Structures provided by an embodiment of the present invention;

[0033] Figure 6 Schematic diagram of the bracket generated by Tekla Structures provided by the embodiment of the present invention;

[0034] Figure 7 Schematic diagram of the trestle generated by Tekla Structures provided by the embodiment of the present invention. Detailed implementation manners

[0035] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the present invention, and gives detailed implementation manners and specific operation processes, but the protection scope of the present invention is not limited to the following embodiments.

[0036] As Figure 1 shown, an optimization method for generating a BIM model based on artificial intelligence provided by an embodiment of the present invention mainly uses tools as shown in Figure 2 . The tools include the large model Qwen 72b and Tekla Structures. The large model Qwen 72b understands natural language and converts it into a prompt word template, and Tekla Structures is used as the BM software of the generation model. This method covers the complete process from natural language to structured parameter generation, then to replacing the parameters of the BIM software function template, and finally generating the model. It significantly improves the generation efficiency and quality of the AI text-to-model, and reduces errors in the generation process, including the following steps:

[0037] Step 1, create a prompt word template library as shown in Figure 3 . Divide the model to be created into several modules. For example, in this example, the grid, bracket, trestle, etc. all have their own prompt word templates. When the user makes a request, the corresponding module will be automatically matched and the corresponding template will be read.

[0038] Step 2, create a function template library as shown in Figure 4 . Write the Tekla Structures generation codes of the grid, bracket, and trestle into module1, module2, and module3 (examples of several function templates created) in mypackage (a user-defined function template library name) respectively, and set parameters such as dimensions as formal parameters for convenient external call. This can avoid errors in the large model writing code and improve the code generation speed.

[0039] Step 3: Receive the natural language questions raised by the user and convert them into structured parameters according to the prompt template. For example, when generating a grid, what the user gives may be "Please help me generate a grid" or "Create a grid, x 18000,30000,24000, y 4000". According to the user's prompt, use natural language processing technology (such as the jieba word segmentation library) to identify that the user's requirement is a grid, find the grid category in the prompt template, and use the understanding ability of the large model to fill in the x and y direction data into the template in the prompt template to form a new prompt. The template contains default settings, and it can also be generated if the user does not give any parameters. The new prompt is as follows:

[0040] Please refer to the following code

[0041] from mypackage import module1

[0042] x = "18000 30000 24000"

[0043] y = "4000"

[0044] z = "0 6000 8000 9000"

[0045] module1.generate_grid(x = x, y = y, z = z)

[0046] Step 4: Execute the code generated in Step 3. As Figure 5 shown, generate a grid in Tekla Structures.

[0047] Step 5: The user continues to raise the need to generate supports and trestles, and re-execute the above steps to generate supports (such as Figure 6 ) and trestles (such as Figure 7 ) in Tekla Structures; subsequent modification functions can be executed, such as stretching, changing the cross-section, etc.

[0048] Specific practices have shown that by using the text generation model method of the embodiments of the present invention, a BIM model can be quickly generated, greatly improving the modeling efficiency; avoiding BIM software code errors generated by the large model and the problem of long waiting time caused by waiting for the code to be generated before execution.

[0049] The above embodiments only represent one implementation manner of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the appended claims.

Claims

1. An optimization method for generating a BIM model from natural language based on a large language model, characterized in that, It includes the following steps: Step 1. Create a prompt template library Build an EXCEL database containing multiple predefined prompt templates; each prompt template contains two columns. The first column is the type column, which includes the structural type of the BIM model to be created or specific modification tasks; the second column is the prompt column, which contains the prompts sent to the large model. The prompt calls the BIM model function template and sets the default values of the parameters to explain the required parameters and reference the code attached to the prompt; each prompt column contains fixed structured statements and reserved replaceable parameters for receiving key information extracted from the natural language description and filling it into the BIM model function template through function calls; Step 2. Natural language understanding and parameter extraction Receive the natural language description input by the user; use natural language processing technology to identify the category of the BIM model module required by the user; After identifying the BIM model module, select an appropriate prompt template, use the understanding ability of the large model to match the natural language parameters with the parameter placeholders in the prompt template library, and extract the parameter values of the successfully matched natural language; fill the extracted parameter values into the parameter placeholders of the prompt template to generate a complete and structured prompt; Step 3. Build a BIM model function template library Build a BIM model function template library corresponding to the BIM software API; each function template represents a specific BIM model creation or modification operation; the function template contains function calls of the BIM software API and reserves parameter entrances corresponding to the parameter placeholders of the prompt template; the function template library is called by the prompt template library; Step 4. Execute the instructions generated by the prompt template to create or modify the BIM model; Execute the code or instructions generated by the prompt template library. The instructions will automatically call the function template library containing the parameters required by the user and execute the code to create or modify the corresponding BIM model in the BIM software.

2. The optimization method for natural language generation of BIM models based on large language models according to claim 1, characterized in that In Step 1, the structural types include grid, support, and trestle, and the specific modification tasks include stretching and rotation.

3. The optimization method for generating a BIM model from natural language based on a large language model according to claim 1, characterized in that, In Step 1, the prompt template is in EXCEL format, and users can customize and expand the prompt template according to actual needs to adapt to different modeling scenarios and styles.

4. The optimization method for generating a BIM model from natural language based on a large language model according to claim 1, characterized in that, In Step 2, the natural language processing technology includes the jieba word segmentation library.

5. The optimization method for natural language generation of BIM models based on large language models according to claim 1, wherein, In Step 2, the template contains default settings, and it can also be generated if the user does not give any parameters.

6. The optimization method for generating a BIM model from natural language based on a large language model according to claim 1, wherein In Step 3, the BIM software includes Revit and Tekla Structures.

Images