Pile foundation design method and device based on large language model

Abstract

The invention discloses a pile foundation design method and device based on a large language model, and belongs to the technical field of pile foundation engineering and computer artificial intelligence application crossing. By means of the pile foundation design method based on the large language model, a single person can complete pile foundation design work, results can be checked through the large language model, and the pile foundation design efficiency is improved. Errors are avoided; according to the method, expert experience is used as a local knowledge base, so that young designers can smoothly complete design tasks. Through the mode, the method can be realized by combining software with a universal hardware platform, can also be realized through hardware, can exist in a software product form, and can be stored in different types of computer readable storage media, such as ROM / RAM (Read Over Memory / Random Access Memory), disks, optical disks and the like. The software products comprise a series of instructions, so that the computer equipment can execute implementation steps of related technical schemes.

Description

Technical Field

[0001] The present invention relates to the cross-technical field of pile foundation engineering and computer artificial intelligence application, and in particular to a pile foundation design method and device based on a large language model. Background Art

[0002] Pile foundation plays a vital role in civil engineering and water conservancy projects, especially when the bearing capacity of the foundation is insufficient or the geological conditions are complex. Pile foundation design is a highly complex and special link in the field of geotechnical engineering. Different types of projects have significantly different design requirements for pile foundations. For example, building construction projects focus on the bearing capacity and settlement control of pile foundations, while water conservancy projects pay more attention to the stability control of pile foundations under complex geological conditions. In addition, the diversity of geographical environments also puts forward different requirements for pile foundation design. The geological conditions, climate environment and hydrological conditions in different regions will affect the design of the scheme.

[0003] Nowadays, the big language model in artificial intelligence technology has developed rapidly in all walks of life, covering many fields such as medical care, finance, manufacturing, transportation, agriculture, etc., showing strong text understanding and generation capabilities, greatly improving the efficiency and level of traditional working methods. The big language model also has great application potential in pile foundation design. However, since pile foundation engineering belongs to a relatively professional and niche field, and the training data of the big language model usually comes from a wide range of large-domain knowledge, covering a large amount of general information and general language applications. Therefore, there may be certain limitations in directly using the big language model to solve pile foundation problems. Although the big language model has strong language generation and understanding capabilities, it lacks targeted professional knowledge and cannot deeply understand the specific technical requirements, geological conditions, construction specifications, etc. in pile foundation design.

[0004] However, it is difficult to train a dedicated model for pile foundation design. Existing pile foundation design solutions are generally completed by professional engineering designers who need to have many years of professional experience. In addition, the entire design process is time-consuming, labor-intensive, inefficient, and prone to errors. How to use large language model technology to assist pile foundation design to improve work efficiency and level is an urgent problem to be solved in the cross-technical field of pile foundation engineering and computer artificial intelligence applications.

[0005] Based on this, the present invention designs a pile foundation design method and device based on a large language model to solve the above problems. Summary of the invention

[0006] In view of the above-mentioned shortcomings of the prior art, the present invention provides a pile foundation design method and device based on a large language model.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0008] A pile foundation design method based on a large language model comprises the following steps:

[0009] Step 1. Collect information on engineering geological survey reports and preliminary engineering design reports;

[0010] Step 2, select the large language model and input the data of the engineering geological survey report and the engineering preliminary design report. The large language model extracts the engineering information of the engineering geological survey report and the engineering preliminary design report;

[0011] Step 3: The large language model selects a design solution based on the extracted engineering information and the pile foundation applicability condition document customized by industry experts;

[0012] Step 4: The large language model calculates the engineering information and the applicable condition files of the pile foundation, obtains the calculation results and verifies them, thereby generating a standardized calculation report and finally generating a pile foundation design calculation file;

[0013] Step 5: Generate the pile foundation layout required for the project based on the engineering information, design plan and calculation results in the above steps.

[0014] Furthermore, in Step 1, the engineering geological survey report includes the engineering overview, topography, hydrogeological conditions and geological problems;

[0015] The preliminary design report of the project includes project overview, design plan, construction organization, investment estimate, environmental impact and soil and water conservation, etc.

[0016] Furthermore, the implementation process of Step 2 includes the following steps:

[0017] ①. Use Python to parse the engineering geological survey report and engineering preliminary design report files into plain text format;

[0018] ② Use Python to process the error format in the parsed text;

[0019] ③. Choose a large language model: GPT model or BERT model;

[0020] ④. Construct specific prompts in the large language model to extract engineering information from engineering geological survey reports and engineering preliminary design reports;

[0021] ⑤. Input the processed plain text engineering geological survey report, engineering preliminary design report file and the constructed specific prompt into the selected large language model;

[0022] ⑥ Input the engineering information extracted by the command into the large language model, and save the extracted engineering information in an easy-to-store CSV format or JSON format to generate an engineering information file.

[0023] Furthermore, Prompt includes Role, Definition, Example, and Demand;

[0024] The Role part sets the system as a professional geotechnical engineering and structural analysis expert, which can help the big language model to extract information from a professional perspective and determine the professionalism and accuracy of the information when performing tasks;

[0025] The Definition part clarifies the information to be extracted, helps the large language model understand the scope and objectives of the task, and ensures that the large language model does not miss key content when extracting information; in addition, the Definition part can also help the large language model complete the reasoning process;

[0026] The Example section provides sample output after the large language model extracts information, which helps the large language model understand the expected output format and standards, and provides consistent structured information to reduce ambiguity;

[0027] The Demand section describes the expectations and standards of the task in detail, and performs text extraction based on the provided report. Information not mentioned in the report is extracted as None to prevent large models from randomly generating useless interference information.

[0028] Furthermore, the implementation process of Step 3 includes the following steps:

[0029] ①. The large language model prepares and analyzes engineering information files and pile foundation applicable condition files;

[0030] ② The large language model generates engineering condition Embedding vectors and pile foundation applicability condition Embedding vectors from engineering information files and pile foundation applicability condition files respectively. The engineering condition Embedding vectors and pile foundation applicability condition Embedding vectors place similar data in similar positions in the vector space, thereby capturing their semantic or feature similarities.

[0031] ③. The large language model performs semantic search and matching based on semantic similarity. The large language model performs semantic similarity search on the engineering condition embedding vector and the pile foundation applicable condition embedding vector, and uses the vector database to perform K-nearest neighbor search to find the pile foundation condition most similar to the input. The matching results are then sorted according to the similarity score to ensure that the most relevant matching results are displayed;

[0032] ④. Set a similarity threshold, and the large language model will further process the matching results obtained through semantic search to screen out the optimal pile foundation type.

[0033] Furthermore, the similarity score is calculated using a cosine similarity calculation method, a Euclidean distance calculation method, a Manhattan distance calculation method, and the like.

[0034] Furthermore, the implementation process of Step 4 includes the following steps:

[0035] ①. The large language model preliminarily organizes and defines the basic parameters for pile foundation design calculation and automatically derives auxiliary parameters;

[0036] ② The large language model constructs a specific prompt to perform automatic bearing capacity calculation of pile foundation and output the calculation results. The constructed prompt includes the standard formulas and engineering experience parameters involved in pile foundation design;

[0037] ③. Based on the output calculation results, the large language model automatically generates a standardized calculation report that complies with engineering specifications; the calculation report includes detailed design basis, calculation formula and data source;

[0038] ④. After the standardized calculation report is generated, the large language model automatically verifies the calculation results;

[0039] ⑤. The large language model generates pile foundation design calculation files based on the calculation results, and saves the calculation results in JSON format, CSV format or PDF format that is easy to store and retrieve.

[0040] Furthermore, the implementation process of Step 5 includes the following steps:

[0041] ① The large language model performs data analysis on the key parameters of the layout drawing, including the geometric dimensions of the building base, pile spacing, pile diameter, pile length and pile foundation layout mode;

[0042] ②, Use prompt to issue instructions to the large language model to formulate the logic of generating layout diagram;

[0043] ③. Generate pile position data using a large language model, i.e., a set of two-dimensional coordinate points for pile foundation arrangement;

[0044] ④. Through the combination of the communication interface and the graphics generation tool, the large language model converts the generated pile position data into a graphical visualization of the pile foundation layout according to the drawing parameters. The drawing parameters include: pile position number and coordinates, pile foundation diameter and building floor contour information;

[0045] ⑤. The large language model can be used to automatically check the generated layout diagram and make corresponding adjustments.

[0046] Furthermore, the verification includes the pile spacing, the distance from the edge piles to the building boundary and the bearing capacity.

[0047] A pile foundation design device includes a processor and a memory; the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method. The pile foundation design device is an electronic device, and the processor and the memory are connected to a communication interface through a communication bus, and the communication interface is connected to an external device.

[0048] Compared with the prior art, the present invention has the following beneficial effects: 1. It proposes a pile foundation design method and device based on a large language model, which fills the gap in this field and has guiding significance for related projects;

[0049] 2. The traditional pile foundation design method is time-consuming and laborious, requires the cooperation of multiple people, and is prone to errors. By using the pile foundation design method based on the large language model proposed by the present invention, a single person can complete the pile foundation design work, and the large language model can also be used to check the results to avoid errors;

[0050] 3. Traditional pile foundation design is highly dependent on engineering experience, and requires designers with many years of engineering experience to complete the pile foundation design work. The pile foundation design method based on the large language model proposed in the present invention can adopt the method of using expert experience as a local knowledge base, so that young designers can also successfully complete the design task. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0052] Figure 1 A flowchart of a pile foundation design method based on a large language model of the present invention;

[0053] Figure 2 It is a schematic diagram of the engineering information extraction process of the present invention;

[0054] Figure 3 This is a Prompt example diagram of the present invention;

[0055] Figure 4 It is a schematic diagram of the design scheme selection process of the present invention;

[0056] Figure 5 It is a schematic diagram of the pile foundation design calculation process of the present invention;

[0057] Figure 6A schematic diagram of the process of drawing a pile foundation layout diagram of the present invention;

[0058] Figure 7 It is a schematic diagram of the basic structure of the electronic device of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] Embodiment 1: In some embodiments, please refer to the drawings of the specification Figure 1-Figure 6 , a pile foundation design method based on a large language model, comprising the following steps:

[0061] Step 1. Collect information on engineering geological survey reports and preliminary engineering design reports;

[0062] The engineering geological survey report includes the project overview, topography, hydrogeological conditions, and geological problem evaluation.

[0063] The preliminary design report of the project includes project overview, design plan, construction organization, investment estimate, environmental impact, soil and water conservation, etc.

[0064] The above two reports are very important reports in general water conservancy projects, and pile foundation design work needs to be carried out based on these two reports.

[0065] However, the two reports contain a lot of content, and each report is generally hundreds of pages or even thousands of pages. Therefore, when designing pile foundations, you need to read the reports thoroughly and find the key information you need.

[0066] The preliminary design report of a general project should include a pile foundation design plan, so the preliminary design report mentioned here refers to a report on the pile foundation design plan that has not yet been completed, and the design plan for the main building has been basically determined.

[0067] Step 2, select the large language model and input the data of the engineering geological survey report and the engineering preliminary design report. The large language model extracts the engineering information of the engineering geological survey report and the engineering preliminary design report;

[0068] The specific steps include:

[0069] ①. Use Python or other tools to parse the engineering geological survey report and engineering preliminary design report files into plain text format; the plain text format can be .txt format or other formats.

[0070] ② Use Python or other tools to process the incorrect formats in the parsed text. The text directly taken from the engineering geological survey report and the engineering preliminary design report file will have incorrect formats. The incorrect formats include extra line breaks, blank characters or disordered text order, so these problems need to be processed.

[0071] ③. Choose a large language model: GPT model (Generative Pre-Trained Transformer, a deep learning model for text generation based on Internet text data and trained with available data) or other models such as BERT model.

[0072] ④. Construct specific prompts in the large language model to extract engineering information from engineering geological survey reports and engineering preliminary design reports;

[0073] Prompt mainly extracts soil layer information (name of soil layer, thickness of soil layer, physical and mechanical properties of soil layer), seismic parameters of the project area, and pile foundation design considerations from the engineering geological survey report;

[0074] Prompt mainly extracts information such as the main building information (type, level, base plate size, base reaction value), the surrounding conditions of the project (whether it is in the urban area, whether there are other important buildings), and whether the construction period is tight from the preliminary design report of the project.

[0075] Prompt includes Role, Definition, Example, and Demand.

[0076] The Role part can set the system as a professional geotechnical engineering and structural analysis expert, which can help the large language model to extract information from a professional perspective and determine the professionalism and accuracy of the information when performing tasks.

[0077] The Definition part specifies the information to be extracted. For example, the soil layer information is defined in detail to refer to the name of the soil layer under the base of the main building, the thickness of the soil layer, the allowable bearing capacity and mechanical strength of the soil layer, which helps the large language model understand the scope and objectives of the task and ensures that the large language model does not miss key content when extracting information. In addition, the definition can also help the large language model complete the reasoning process. For example, we need to extract information on whether the project schedule is tight, but the preliminary design report of the project does not clearly state whether the project schedule is tight. We can define whether the schedule is tight based on engineering experience. We define that if the estimated price of the project exceeds 150 million yuan and the planned construction period is less than 12 months, the project schedule can be defined as tight. The estimated price and planned construction period of the project are generally clearly given in the preliminary design report. We use the definition to help the large language model complete the reasoning of whether the construction period is tight.

[0078] The Example section provides sample output after the large language model extracts information, for example: The construction period of this project is tight because the estimated cost of this project is 150 million yuan, but the planned construction period is only 10 months. This helps the large language model understand the expected output format and standards, and provide consistent structured information to reduce ambiguity.

[0079] The Demand section describes the expectations and standards of the task in detail, such as explicitly referring to the definition above and performing text extraction based on the provided report. Information not mentioned in the report will be extracted as None.

[0080] ⑤. Input the processed plain text engineering geological survey report, engineering preliminary design report file and the constructed specific prompt into the selected large language model;

[0081] ⑥ Input the engineering information extracted by the command into the large language model, and save the extracted engineering information in an easy-to-store CSV format or JSON format to generate an engineering information file.

[0082] Step 3: The large language model selects a design solution based on the extracted engineering information and the pile foundation applicability condition document customized by industry experts;

[0083] The following steps are included: ①, the large language model prepares and parses the engineering information file and the pile foundation applicable condition file;

[0084] The content of the pile foundation applicable conditions document is the applicable conditions for different pile foundations and the pile foundation design experience. This document needs to be produced by industry experts and used within the company as an expert knowledge base. The applicable conditions of different pile foundations should first meet the requirements of the corresponding specifications. Different project types, different regions, and the situation of each company may affect the determination of the optimal pile foundation solution. Therefore, this document should be customized by experts in the industry.

[0085] Data analysis is to use Python and other tools to extract key information from engineering information files and pile foundation applicability condition files, and then store them in a structured manner.

[0086] Extract engineering geological conditions, building information, etc. from the project information file extracted in Step 2. You can use dictionary form to store data, for example: project_info = {"soil_type":"softclay","soil_thickness":10,#meters"building_height":30,#meters"building_type":"office","construction_time":180#days};

[0087] Extract the conditions applicable to different pile foundations from the pile foundation applicable condition file, and also store these conditions as dictionary type data, for example: pile_options = [{"type":"bored_pile","soil_type":["clay","softsoil"],"max_height":50,"priority":"quickconstruction"},{"type":"prefab_pile","soil_type":["softclay"],"max_height":30,"priority":"lowcost"},#more pile foundation types and conditions].

[0088] ② The large language model generates engineering condition Embedding vectors and pile foundation applicability condition Embedding vectors from engineering information files and pile foundation applicability condition files respectively. The engineering condition Embedding vectors and pile foundation applicability condition Embedding vectors place similar data in similar positions in the vector space, thereby capturing their semantic or feature similarities.

[0089] Embedding vector is a method of converting text, words, phrases or other types of data (such as images, speech, etc.) into a low-dimensional continuous vector. This representation method can capture the semantics or characteristics of the data, allowing machine learning models to efficiently process complex data;

[0090] The core idea of ​​Embedding is to place similar data in close locations in the vector space to capture their semantic or feature similarities.

[0091] Similarity can be calculated using these vectors, and the similarity score can be calculated using a cosine similarity calculation method, a Euclidean distance calculation method, a Manhattan distance calculation method, and the like.

[0092] ③. The large language model performs semantic search and matching based on semantic similarity, performs semantic similarity search on the pile foundation applicability condition Embedding vector and the engineering condition Embedding vector, and uses the vector database to perform K-nearest neighbor search to find the pile foundation condition most similar to the input. The matching results are then sorted according to the similarity score (such as distance or cosine similarity) to ensure that the most relevant matching results are displayed.

[0093] Semantic search refers to the process of finding information by understanding the meaning of text, rather than just by matching keywords. It relies on the generation of embedding vectors, allowing the system to efficiently match text based on semantic similarity. This approach is more accurate than traditional keyword-based search because it understands context and potential semantic relationships.

[0094] ④. Set a similarity threshold, and the large language model will further process the matching results obtained through semantic search to screen out the optimal pile foundation type.

[0095] By setting a similarity threshold, the pile foundation solutions with scores below the threshold are filtered out to ensure the relevance of the matching results. The results are then adjusted according to priority. For example, precast piles can be recommended first for emergency projects.

[0096] Step 4: The large language model calculates the engineering information and the applicable condition files of the pile foundation, obtains the calculation results and verifies them, thereby generating a standardized calculation report and finally generating a pile foundation design calculation file;

[0097] Generating pile foundation design calculation files is the core step of pile foundation engineering. By utilizing the big language model, the complex calculation process can be broken down into automated and standardized steps, thereby improving efficiency and accuracy.

[0098] The specific steps of generating pile foundation design calculation files are as follows: ①, preliminarily organize and define the basic parameters of pile foundation design calculation and automatically derive auxiliary parameters using the large language model;

[0099] Before starting the design calculation, the basic parameters required for the design and the auxiliary parameters are automatically derived. These parameters come from the information extracted in Step 2 and Step 3, including soil layer information, building information, and the selected pile type. The large language model can perform preliminary parameter sorting and derivation based on this information.

[0100] Extraction of foundation parameters: Pile foundation type (such as cast-in-place piles, precast piles), soil layer information (such as soil type, thickness, strength) and building information (such as building load, base plate size, etc.) extracted from Step 2 and Step 3.

[0101] Automatic derivation of auxiliary parameters: The large language model derives other auxiliary design parameters, such as the slenderness ratio of the pile foundation, allowable bearing capacity, and lateral friction coefficient, based on the applicable condition file or specification of the pile foundation in Step 3.

[0102] ② The large language model builds a specific prompt to perform automatic bearing capacity calculation of pile foundation and output the calculation results;

[0103] The constructed prompt includes the standard formulas and engineering experience parameters involved in pile foundation design.

[0104] Example Prompt:

[0105] You are a professional geotechnical engineer. Please calculate the bearing capacity of the pile foundation based on the following data: pile diameter: 1.2 meters, pile length: 30 meters, soil type: soft clay; building load: 5000kN, side friction coefficient: 0.35, pile end bearing coefficient: 0.7; please output the total bearing capacity calculation result of the pile foundation based on the ultimate bearing capacity formula and explain the calculation process step by step.

[0106] ③. Based on the output calculation results, the large language model automatically generates a standardized calculation report that complies with engineering specifications; the calculation report includes detailed design basis, calculation formula and data source; the large language model can automatically format and generate report content.

[0107] ④. After the standardized calculation report is generated, the large language model automatically verifies the calculation results;

[0108] The verification contents include: checking whether the bearing capacity and length of the pile foundation meet the requirements of the specifications; checking whether the pile foundation safety factor meets the design requirements according to the design specifications; if the bearing capacity is insufficient or the design does not meet the requirements, the model should be able to prompt the design defects and provide improvement suggestions, etc.

[0109] Example Prompt:

[0110] Based on the above pile foundation bearing capacity calculation results, please check whether they meet the requirements of "Code for Design of Building Foundations" GB50007-2011. If not, please give suggestions.

[0111] ⑤. The large language model generates a pile foundation design calculation file based on the calculation results, and saves the calculation results in a format that is easy to store and retrieve, such as JSON, CSV or PDF format, for subsequent use.

[0112] Step 5: Generate the pile foundation layout required for the project based on the engineering information, design plan and calculation results in the above steps.

[0113] With the support of a large language model, this step can be performed in an automated manner and seamlessly connected with the previous four steps.

[0114] The specific implementation steps are as follows: ①. Use the large language model to analyze the key parameters of the layout drawing; the data for drawing the pile foundation layout drawing comes from Step 2, Step 3 and Step 4, which include geological conditions, building information, pile type and size, etc. This step uses the large language model to extract and organize the key parameters of the layout drawing, such as the geometric dimensions of the building base, pile spacing, pile diameter, pile length, and pile foundation layout mode.

[0115] ②. Use Prompt to issue instructions to the large language model to formulate the logic for generating layout drawings; for example: evenly arrange pile foundations within the building base area; arrange pile foundations in different areas according to the load and geological conditions.

[0116] ③. Generate pile position data, i.e., the two-dimensional coordinate point set of pile foundation layout, using the large language model. When generating pile position data, factors such as edge effects and pile spacing requirements in the design specifications need to be considered. The large language model can directly generate the coordinate information of the pile foundation by understanding the layout rules, forming a specific pile position table.

[0117] ④. Through the communication interface and the combination of graphics generation tools (such as AutoCAD, Python's matplotlib library, etc.), the large language model converts the generated pile position data into a graphical visualization of the pile foundation layout according to the drawing parameters. The drawing parameters include: pile position number and coordinates; pile foundation diameter; building base plate outline and other information.

[0118] ⑤. The large language model can be used to automatically check the generated layout and make corresponding adjustments. The check content includes the pile spacing, the distance from the side pile to the building boundary, the bearing capacity, etc.

[0119] The above implementation steps can be implemented by software combined with a general hardware platform, or by hardware. They can exist in the form of software products and be stored in different types of computer-readable storage media, such as ROM / RAM, magnetic disks, optical disks, etc. These software products contain a series of instructions that enable a computer device (which can be a personal computer, server or network device, etc.) to execute the implementation steps of the relevant technical solution.

[0120] Embodiment 2: In some embodiments, Figure 7As shown, as a preferred embodiment of the present invention, a pile foundation design device includes a processor and a memory; the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0121] The pile foundation design device is an electronic device, the processor and the memory are connected to the communication interface through a communication bus, and the communication interface is connected to external equipment.

[0122] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A pile foundation design method based on a large language model, characterized by: The following steps are involved: Step 1. Collect information on engineering geological survey reports and preliminary engineering design reports; Step 2, select the large language model and input the data of the engineering geological survey report and the engineering preliminary design report. The large language model extracts the engineering information of the engineering geological survey report and the engineering preliminary design report; Step 3: The large language model selects a design solution based on the extracted engineering information and the pile foundation applicability condition document customized by industry experts; Step 4: The large language model calculates the engineering information and the applicable condition files of the pile foundation, obtains the calculation results and verifies them, thereby generating a standardized calculation report and finally generating a pile foundation design calculation file; Step 5: Generate the pile foundation layout required for the project based on the engineering information, design plan and calculation results in the above steps.

2. The pile foundation design method based on a large language model according to claim 1, characterized in that: In Step 1, the engineering geological survey report includes the engineering overview, topography, hydrogeological conditions and geological problems; The preliminary design report of the project includes project overview, design plan, construction organization, investment estimate, environmental impact and soil and water conservation.

3. The pile foundation design method based on a large language model according to claim 2, characterized in that: The implementation process of Step 2 includes the following steps: ①. Use Python to parse the engineering geological survey report and engineering preliminary design report files into plain text format; ② Use Python to process the error format in the parsed text; ③. Choose a large language model: GPT model or BERT model; ④. Construct specific prompts in the large language model to extract engineering information from engineering geological survey reports and engineering preliminary design reports; ⑤. Input the processed plain text engineering geological survey report, engineering preliminary design report file and the constructed specific prompt into the selected large language model; ⑥ Input the engineering information extracted by the command into the large language model, and save the extracted engineering information in an easy-to-store CSV format or JSON format to generate an engineering information file.

4. The pile foundation design method based on a large language model according to claim 3 is characterized in that: Prompt includes Role, Definition, Example and Demand; The Role part sets the system as a professional geotechnical engineering and structural analysis expert, which can help the big language model to extract information from a professional perspective and determine the professionalism and accuracy of the information when performing tasks; The Definition part clarifies the information to be extracted, helps the large language model understand the scope and objectives of the task, and ensures that the large language model does not miss key content when extracting information; in addition, the Definition part can also help the large language model complete the reasoning process; The Example section provides sample output after the large language model extracts information, which helps the large language model understand the expected output format and standards, and provides consistent structured information to reduce ambiguity; The Demand section describes the expectations and standards of the task in detail and performs text extraction based on the provided report. Information not mentioned in the report will be extracted as None.

5. The pile foundation design method based on a large language model according to claim 4, characterized in that: The implementation process of Step 3 includes the following steps: ①. The large language model prepares and analyzes engineering information files and pile foundation applicable condition files; ② The large language model generates engineering condition Embedding vectors and pile foundation applicability condition Embedding vectors from engineering information files and pile foundation applicability condition files respectively. The engineering condition Embedding vectors and pile foundation applicability condition Embedding vectors place similar data in similar positions in the vector space, thereby capturing their semantic or feature similarities. ③. The large language model performs semantic search and matching based on semantic similarity. The large language model performs semantic similarity search on the engineering condition embedding vector and the pile foundation applicable condition embedding vector, and uses the vector database to perform K-nearest neighbor search to find the pile foundation condition most similar to the input. The matching results are then sorted according to the similarity score to ensure that the most relevant matching results are displayed; ④. Set a similarity threshold, and the large language model will further process the matching results obtained through semantic search to screen out the optimal pile foundation type.

6. The pile foundation design method based on a large language model according to claim 5, characterized in that: The similarity score is calculated using the cosine similarity calculation method, the Euclidean distance calculation method, and the Manhattan distance calculation method.

7. The pile foundation design method based on a large language model according to claim 6, characterized in that: The implementation process of Step 4 includes the following steps: ①. The large language model preliminarily organizes and defines the basic parameters for pile foundation design calculation and automatically derives auxiliary parameters; ② The large language model constructs a specific prompt to perform automatic bearing capacity calculation of pile foundation and output the calculation results. The constructed prompt includes the standard formulas and engineering experience parameters involved in pile foundation design; ③. Based on the output calculation results, the large language model automatically generates a standardized calculation report that complies with engineering specifications; the calculation report includes detailed design basis, calculation formula and data source; ④. After the standardized calculation report is generated, the large language model automatically verifies the calculation results; ⑤. The large language model generates pile foundation design calculation files based on the calculation results, and saves the calculation results in JSON format, CSV format or PDF format that is easy to store and retrieve.

8. The pile foundation design method based on a large language model according to claim 7, characterized in that: The implementation process of Step 5 includes the following steps: ① The large language model performs data analysis on the key parameters of the layout drawing, including the geometric dimensions of the building base, pile spacing, pile diameter, pile length and pile foundation layout mode; ②, Use prompt to issue instructions to the large language model to formulate the logic of generating layout diagram; ③. Generate pile position data using a large language model, i.e., a set of two-dimensional coordinate points for pile foundation arrangement; ④. Through the combination of the communication interface and the graphics generation tool, the large language model converts the generated pile position data into a graphical visualization of the pile foundation layout according to the drawing parameters. The drawing parameters include: pile position number and coordinates, pile foundation diameter and building floor contour information; ⑤. The large language model can be used to automatically check the generated layout diagram and make corresponding adjustments.

9. The pile foundation design method based on a large language model according to claim 8, characterized in that: The verification contents include the distance between piles, the distance from the edge piles to the building boundary and the bearing capacity.

10. A pile foundation design device, characterized in that: The method comprises a processor and a memory; the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method as claimed in claim 9.

Images