Generative geotechnical engineering design system and equipment based on artificial intelligence

Abstract

The present invention belongs to the field of geotechnical engineering design, and specifically relates to a generative geotechnical engineering design system and equipment based on artificial intelligence, which aims to solve the problem of uneven economy, safety and rationality of the design results of existing geotechnical engineering design methods. The present invention includes: collecting design parameters specified by users; based on the design parameters, generating an initial design scheme through a geotechnical engineering-specific artificial intelligence model; evaluating the initial design scheme with set dimensional indicators to obtain multiple dimensional indicator evaluation results; optimizing the initial design scheme selected by the customer through an optimization algorithm based on the multiple indicator evaluation results to obtain an optimized design method; accepting user interactive instructions to adjust the optimized design scheme to obtain a final design scheme; exporting the final design scheme in a set standard format. The present invention uses a generative model to quickly generate economic, safe and reasonable solutions that meet specific conditions.

Description

Background Art

[0002] Geotechnical engineering design refers to the process of analyzing and calculating the mechanical properties and engineering behaviors of soil and rock in construction projects according to geological conditions and soil characteristics, combined with engineering requirements and safety standards, to determine appropriate engineering design solutions and construction methods. Geotechnical engineering design is affected by multiple constraints such as regionality, risk, comprehensiveness, practicality, and normativeness. It is necessary to comprehensively consider geological conditions, soil parameters, foundation design, construction technology, material selection, construction monitoring, and environmental protection to ensure the rationality, safety, and sustainability of engineering design and construction processes. The particularity of the industry has caused the existing geotechnical engineering design results to be limited by the level of designers, resulting in large differences in the economy, safety, and rationality of geotechnical engineering design results. Generative artificial intelligence can quickly generate design solutions that meet specific conditions through the analysis and learning of large amounts of data, greatly improving design efficiency. Secondly, generative artificial intelligence can predict possible problems and risks of design solutions in actual construction through simulation and analysis, which helps designers to find problems in advance and optimize design solutions. In addition, generative artificial intelligence can also assist designers in more refined designs. Applying generative artificial intelligence to geotechnical engineering design can effectively make up for the user's lack of experience and significantly improve the level of geotechnical engineering design. Summary of the invention

[0003] In order to solve the above-mentioned problems in the prior art, namely, the problems that the design results of the existing geotechnical engineering design methods are insufficient in economy, safety and rationality, the present invention provides a generative geotechnical engineering design system based on artificial intelligence, the system comprising:

[0004] A design parameter input unit, used to collect design parameters specified by a user;

[0005] A scheme generating unit, used for generating an initial design scheme based on the design parameters by using a geotechnical engineering-specific artificial intelligence model;

[0006] An evaluation and analysis unit, used to evaluate the initial design scheme according to set dimensional indicators and obtain evaluation results of multiple dimensional indicators;

[0007] The optimization design unit is used to optimize the initial design solution selected by the customer through the optimization algorithm according to the evaluation results of multiple indicators to obtain the optimized design method;

[0008] The human-computer interaction unit is used to display the current optimized design scheme and accept the user's interactive instructions to adjust the optimized design scheme to obtain the final design scheme;

[0009] The export unit is used to export the final design plan in a set standard format.

[0010] Furthermore, the design parameters include: design space dimensions, site rock and soil performance indicators, constraints, deformation requirements and durability requirements.

[0011] Furthermore, the solution generating unit includes:

[0012] Pre-dividing and storing the historical geotechnical engineering design into independent component modules;

[0013] Component modules are selected and combined according to the design conditions input by the user and the functions required for geotechnical engineering, and an initial design scheme that meets the constraints and performance indicators is generated through a generative artificial intelligence algorithm.

[0014] Furthermore, the component modules include: soil nails, anchor rods, anchor heads, prestressed anchor rods, various types of I-steels, various types of T-steels, various types of steel pipes, geogrids, geomembranes, geotextiles, shotcrete, pipe wells, light well points, underground continuous walls, composite soil nail walls, steel sheet piles, cast-in-place piles, compression pile reinforcement models, and bending pile reinforcement models;

[0015] Each component module comes with corresponding functional attributes, which include the basic functions, applicable scenarios, performance indicators, reliability, maintainability and durability of the module.

[0016] Furthermore, the set dimensional indicators include performance, cost, deformation, settlement, construction feasibility, and structural stability.

[0017] Furthermore, the optimized design method is obtained by:

[0018] According to the initial design scheme selected by the user, the objective function is to minimize weight, maximize stiffness or minimize engineering quantity. Iterative optimization is performed within the scope of satisfying constraints and dimensional indicators to obtain the optimized design scheme.

[0019] Furthermore, the geotechnical engineering-specific artificial intelligence model, whose basic model includes a scheme understanding model, a terminology understanding model, a component function learning model, a component relationship function learning model, a terminology-function mapping model, and a scheme generation model, and the geotechnical engineering-specific artificial intelligence model training method includes:

[0020] Step A1, collect historical geotechnical engineering design data and clean, organize and annotate them;

[0021] Step A2, learning geotechnical engineering industry terminology and generating an initial technical solution through the geotechnical engineering-specific artificial intelligence model to be trained;

[0022] Step A3, calculating the accuracy, recall and F1 score of each basic model in the trained geotechnical engineering-specific artificial intelligence model according to the initial technical solution;

[0023] Step A4, adjust the parameters of the model, repeat steps A1 to A3 until the accuracy, recall rate and F1 score all reach the respectively set thresholds, and obtain a trained artificial intelligence model for geotechnical engineering.

[0024] Furthermore, it also includes a verification module:

[0025] Collect and summarize verification, parameters and summary verification conditions, use multimodal technology to identify the geotechnical engineering investigation report of the site in the specified design parameters, and obtain the geotechnical parameters of the engineering site;

[0026] Geotechnical engineering designers set design conditions in advance according to the target area and store industry-standard verification methods in the algorithm library;

[0027] Based on the geotechnical parameters of the engineering site, the corresponding data required by the standard verification methods in each industry are called, and verification is performed according to the standard verification methods in the industry, and the verification results and verification suggestions are output.

[0028] A second aspect of the present invention provides an electronic device, comprising:

[0029] at least one processor; and

[0030] a memory communicatively connected to at least one of the processors; wherein,

[0031] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the above-mentioned artificial intelligence-based generative geotechnical engineering design system.

[0032] In a third aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the above-mentioned artificial intelligence-based generative geotechnical engineering design system.

[0033] Beneficial effects of the present invention:

[0034] (1) The present invention uses a generative model to quickly generate solutions that meet specific conditions and can predict possible problems and risks of the design solutions in actual construction, which helps designers to discover problems in advance and optimize the design solutions.

[0035] (2) The system can understand geotechnical engineering terminology, and geotechnical engineers can interact with the system through professional terminology and natural language to modify and improve the system, and finally output geotechnical engineering design construction drawings and bills of quantities. This method can greatly improve the efficiency of geotechnical engineering design.

[0036] (3) The present invention innovatively proposes a geotechnical engineering design method and system based on a geotechnical engineering large model. Users can interact with the geotechnical engineering large model through professional terms. The geotechnical engineering large model generates physical graphics according to the interactive content and assigns unique codes to the physical graphics. Users use natural language to interact with the system when generating or modifying design solutions. The system knows the physical meaning of the geometric figures, which is very different from traditional engineering design software. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0038] Figure 1 It is a basic structural diagram of the generative geotechnical engineering design system based on artificial intelligence of the present invention;

[0039] Figure 2 It is a schematic diagram of the architecture of the complete functions of the artificial intelligence-based generative geotechnical engineering design system of the present invention. DETAILED DESCRIPTION

[0040] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0041] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0042] In order to more clearly explain the generative geotechnical engineering design system based on artificial intelligence of the present invention, the following is combined with Figure 1 Each step in the embodiment of the present invention is described in detail.

[0043] The present invention proposes a generative artificial intelligence design system for geotechnical engineering design, which is used to automatically generate and optimize geotechnical engineering design solutions. It integrates an artificial intelligence large model with a graphic design tool into a software system dedicated to geotechnical engineering design.

[0044] Figure 1is a basic structural diagram of a generative geotechnical engineering design system based on artificial intelligence in a real-time example of the present invention, Figure 2 The schematic diagram of the complete functional architecture of the artificial intelligence-based generative geotechnical engineering design system of the present invention is shown in FIG. Figure 2 It includes a large model module for geotechnical engineering, an intelligent deformation calculation module, a project quantity statistics and project cost analysis module, a design result generation module, a built-in geotechnical model module, and a database module.

[0045] The artificial intelligence-based generative geotechnical engineering design system of the first embodiment of the present invention includes a design parameter input unit, a solution generation unit, an evaluation and analysis unit, an optimization design unit, a human-computer interaction unit and an export unit. The functional modules are described in detail as follows:

[0046] A design parameter input unit, used to collect design parameters specified by a user;

[0047] In this embodiment, the design parameters include: design space size, site rock and soil performance index, constraint conditions, deformation requirements and durability requirements. Design parameters are used to define the scope and limitations of the design problem.

[0048] A scheme generating unit, used for generating an initial design scheme based on the design parameters by using a geotechnical engineering-specific artificial intelligence model;

[0049] In this embodiment, the solution generating unit includes:

[0050] Pre-dividing and storing the historical geotechnical engineering design into independent component modules;

[0051] In this embodiment, the component modules include: soil nails, anchor rods, anchor heads, prestressed anchor rods, various types of I-beams, various types of T-beams, various types of steel pipes, geogrids, geomembranes, geotextiles, shotcrete, pipe wells, light well points, underground continuous walls, composite soil nail walls, steel sheet piles, cast-in-place piles, compression pile reinforcement models, and bending pile reinforcement models;

[0052] Each component module comes with corresponding functional attributes, including the basic functions, applicable scenarios, performance indicators, reliability, maintainability and durability of the module.

[0053] Modular design of geotechnical engineering has the following advantages:

[0054] Improve maintainability: Modular design reduces the coupling between modules by splitting the geotechnical engineering design scheme into independent small modules, each of which is responsible for a specific function. In this way, when modifications are needed, only the specific module can be operated without affecting the entire design scheme, thereby reducing the coupling between modules.

[0055] Enhanced reusability: Modularity allows users to create modules that can be reused in different projects. This saves design time and improves efficiency.

[0056] Facilitates scalability: The modular design makes the system easier to expand. When new processes need to be added, new modules can be simply added.

[0057] Maintaining output consistency: Modular design helps maintain consistency in output throughout the development process.

[0058] Improve design efficiency: Modularity can simplify the design process and enable users to focus on the design of specific modules, thereby improving overall design efficiency.

[0059] Supports rapid satisfaction of personalized needs: Through modularization, different modules can be quickly combined to meet the personalized needs of users.

[0060] In the design software used in existing geotechnical engineering design, geotechnical engineering designers represent corresponding physical objects by drawing corresponding graphic elements or symbols in the software, but the design software does not understand the practical significance of each graphic element or symbol. In this embodiment, the design scheme can be generated by recommending a special artificial intelligence model for geotechnical engineering, and the scheme can be modified through geotechnical engineering professional terms. In this embodiment, graphic elements with physical meaning are directly generated, and material and deformation properties are assigned to each graphic element. For example, "prestressed anchor rod" or "pile" can be input through the interactive window to generate corresponding graphic elements with physical meaning.

[0061] According to the design space size and design conditions input by the user, the geotechnical engineering-specific artificial intelligence model selects component modules for combination based on the acquired engineering geological conditions, and generates an initial design plan that meets the constraints and performance indicators through a generative artificial intelligence algorithm.

[0062] The geotechnical engineering-specific artificial intelligence model, whose basic model includes a scheme understanding model, a terminology understanding model, a component function learning model, a component relationship function learning model, a terminology-function mapping model, and a scheme generation model. The geotechnical engineering-specific artificial intelligence model training method includes:

[0063] Step A1, collect historical geotechnical engineering design data and clean, organize and annotate them; in this embodiment, a large amount of geotechnical engineering related data is collected on the basis of the general training data set, including geotechnical engineering professional terms, geotechnical engineering related theories and calculation formulas, geotechnical engineering design methods, geological survey reports, construction drawings, meteorological data, etc. These data need to be cleaned and organized to ensure their quality and accuracy. Labels or annotations are added to the collected data to convert the data into the format required for model training to help model learning.

[0064] Step A2, learning geotechnical engineering industry terminology and generating an initial technical solution through a geotechnical engineering-specific artificial intelligence model to be trained; the geotechnical engineering-specific artificial intelligence model includes a deep neural network and a generative adversarial network;

[0065] The terms include soil nails, anchor rods, anchor heads, prestressed anchor rods, anti-floating anchor rods, geogrids, geomembranes, drainage boards, geotextiles, pipe wells, light well points, underground continuous walls, composite soil nail walls, piles, steel sheet piles, steel piles, precast pipe piles, precast square piles, cast-in-place piles, pull-out piles, anti-sliding piles, cement-soil piles, lime-soil piles, CFG piles, sand and gravel piles, crushed stone piles, row piles, mixing piles, jet jet piles, underground continuous walls, cement-soil walls, retaining walls, soil nail walls, waist beams, crown beams, pile-anchor support, internal support, inclined column support and cushion layers, etc.

[0066] Step A3, calculating the accuracy, recall and F1 score of each basic model in the trained geotechnical engineering-specific artificial intelligence model according to the initial technical solution;

[0067] Step A4, adjust the parameters of the model, repeat steps A1 to A3 until the accuracy, recall and F1 scores all reach the respectively set thresholds, and obtain a trained geotechnical engineering-specific artificial intelligence model. According to the evaluation results, adjust and optimize the model to improve its performance. Cooperate with experts in the field of geotechnical engineering to ensure the accuracy and reliability of the model. Through the above steps, the generative artificial intelligence model is professionally trained to become a geotechnical engineering intelligent expert, providing more efficient and accurate services for geotechnical engineering design.

[0068] Through professional training of the model, the artificial intelligence model dedicated to geotechnical engineering understands the professional terms in the geotechnical engineering industry, such as: physical property parameters of rock and soil, compressibility, permeability, soil stability, rock stability, cohesion, internal friction angle, compression modulus, shear strength, earth pressure, lateral pressure coefficient, groundwater head, rock quality indicators, foundation bearing capacity, foundation settlement, foundation treatment, foundation deformation, soil nail wall, anchor rod, prestressed anchor rod, geosynthetics, geogrid, geomembrane, geotextile, shotcrete, grouting, drainage system, underground continuous wall, foundation pit support, composite soil nail wall, steel sheet pile, cast-in-place pile, mixing pile, precast pile, earth excavation, fill compaction, compaction coefficient, pile foundation, slope stability analysis, slope protection, geological disasters, landslide, mudslide, ground subsidence, settlement deformation, soil improvement, foundation reinforcement, engineering monitoring, etc.

[0069] The dedicated artificial intelligence model for geotechnical engineering can output corresponding geotechnical engineering graphics through prompt words, generate corresponding graphics in the user interface through prompt words, automatically encode the generated graphics and assign material properties, and export calculation books and bills of quantities through prompt words.

[0070] An evaluation and analysis unit, used to evaluate the initial design scheme according to set dimensional indicators and obtain evaluation results of multiple dimensional indicators;

[0071] In this embodiment, the set dimensional indicators include performance, cost, deformation, settlement, construction feasibility, and structural stability. The geotechnical engineering design is evaluated by simulating, analyzing, and optimizing the design scheme.

[0072] The geotechnical engineering-specific artificial intelligence model includes at least one of each of a solution understanding model, a terminology understanding model, a component function learning model, a component relationship function learning model, a terminology-function mapping model, and a solution generation model;

[0073] The scheme understanding model is implemented based on a hybrid model jointly constituted by graph neural networks (GNNs) and natural language processing (NLP). The training process is to add three-level labels of overall-combination between components-component module to the geotechnical engineering design scheme that has been completed by manual annotation, and the hybrid model is used to understand the applied component modules of the geotechnical engineering design scheme that has been completed. It is essentially a classification model. When the probability difference of a certain component or a certain combination of components belonging to multiple categories is less than the set distinction threshold, or the probability of belonging to each existing class is lower than the recognition threshold, a new category is automatically created, and the specific spatial size, structure, and material type are added to the component module library. The present invention realizes the automatic update iteration and learning from scratch of the present application by setting an automatically learned scheme understanding model, and can automatically learn the component modules needed in the subsequent geotechnical engineering design of the completed scheme, and understand what kind of component modules can be combined and what elements a complete geotechnical engineering design scheme should have. And the learned implicit function is added as a constraint condition to the solution generation model, so that the generated solutions are all under a mature framework to avoid engineering errors.

[0074] The component function learning model and the component relationship function learning model are both learned in the form of knowledge graphs, with component modules as graph nodes, which contain their attributes (such as function descriptions and function dependencies); the relationships between nodes (combination relationships, dependency relationships) are used as edges to build knowledge graphs. Function learning and reasoning are performed using a knowledge graph reasoning engine or a machine learning model based on a knowledge graph (such as KG-BERT). For example, by traversing graph paths, calculating path scores, or running reasoning rules, the effects of functional combinations between components are learned.

[0075] The term-function mapping model, in order to adapt to different term-to-function mappings and accurately and comprehensively ensure that the generated technical solutions meet the requirements and implicit requirements of designers, this system pre-stores a variety of term-function mapping models in the library, including word embedding models (Word Embeddings), convolutional neural networks (Convolutional Neural Networks, CNNs), recurrent neural networks (Recurrent Neural Networks, RNNs), long short-term memory networks (Long Short-Term Memory, LSTM), Transformer models, dual encoder architectures (such as Siamese Network, BERT for Sentence Pair Classification), and models enhanced by attention mechanisms (Attention Mechanisms) such as self-attention (Self-Attention) or cross-attention (Cross-Attention), and is selected according to task requirements and the length of the term description, the degree of structure, whether it contains time series information, whether it is a simple classification task or a task that requires understanding of deep semantic relationships, and the computing power and memory required for model training and reasoning.

[0076] For example, use pre-trained word embedding models (such as Word2Vec, FastText, GloVe, etc.) to convert terms into fixed-dimensional vector representations, which contain the semantic information of the vocabulary. You can build a simple multi-layer perceptron (MLP) or logistic regression model, use the embedding vector of the term as input, and the component function as a label for supervised learning. The model captures the association between the term vector and the component function category by learning the weight matrix.

[0077] If the term description contains multiple words or phrases, CNN can be used to process the text sequence. CNN extracts local features through convolutional layers and aggregates these features through pooling layers to identify patterns related to component module functions. The input is the word embedding sequence of the term, and the output is the classification probability of the component function. The filter weights learned by the model can reveal the connection between different word combinations and component module functions.

[0078] For terms that contain time series information or long text descriptions, RNN or LSTM can capture contextual dependencies within the text. They process the sequence recursively, remembering past information to influence the current output. The input is the word embedding sequence of the term, and the output is the classification result of the component function. Through training, the model learns how to dynamically determine the component function associated with it based on the word sequence.

[0079] If there are description texts of terms and component functions at the same time, a dual encoder architecture (such as SiameseNetwork, BERT for Sentence Pair Classification, etc.) can be used. The two encoders encode the terms and component function descriptions respectively, and then calculate the vector similarity between the two or fuse the two vectors through the attention mechanism to predict the association between the two.

[0080] The terminology understanding model in this embodiment can be implemented through the existing semantic recognition model. Since the meanings referred to by different people or different expressions are essentially the same, and specific terms are not necessarily completely directed to the functions of component modules, such as environment, humidity, seasonal changes, temperature differences and other information, it is necessary to use a terminology understanding model as a supplement to the terminology-function mapping model. The two models are used together to guide the work of the solution generation model.

[0081] It is particularly noted that the term-function model and term understanding model in this embodiment need to be trained after the component function learning model and the component relationship function learning model are trained, so as to fully understand and cover all component modules and their functions. The component function learning model and the component relationship function learning model need to be trained after the solution understanding model is trained, and after the new component modules are added to the preset component module library, to avoid the wrong mapping of function-component modules.

[0082] The scheme generation model can be generated by two models in two stages or a single model in one stage. In the two models in two stages, the first model needs to include a model for selecting component modules and component module combinations based on the understood terms, which can be constructed by a deep learning recommendation model, that is, the understood terms are regarded as a recommendation problem, and a suitable component combination is recommended given a design task or scenario. Deep learning recommendation systems (such as collaborative filtering based on deep neural networks, matrix decomposition, deep factor decomposition machines, etc.) can be used to learn the implicit associations and functional complementarity between components. The input can be a feature vector of a component (such as functional description, application scenario, etc.), user behavior data (such as the use of components in historical design solutions), characteristics of design tasks, etc. The output is a list of recommended component combinations for new design tasks, and the recommended combination list is further generated into a complete geotechnical engineering design solution. The single model in the first stage can be implemented by a reinforcement learning model (RL) or a Transformer-based sequence generation model. The reinforcement learning model selects, combines and adjusts component modules to form a design solution, and the environment gives a reward signal based on the design quality. You can use a policy gradient-based approach (such as REINFORCE), Q-learning, or an Actor-Critic framework. The state can represent the current state of the design, and the action corresponds to the operation of selecting or modifying a component. By training the model to understand and generate sequences that describe the design, you can guide the model to combine functional components in a certain logical order and rules to generate a new design. Alternatively, you can generate the model through a generative model GAN ​​network trained against adversarial training.

[0083] When the geotechnical engineering-specific artificial intelligence model provided in this embodiment needs to be updated and iterated, it is only necessary to input the completed design plan into the plan understanding model to achieve automatic updating of the entire model.

[0084] The optimization design unit is used to optimize the initial design solution selected by the customer through the optimization algorithm according to the evaluation results of multiple indicators to obtain the optimized design method;

[0085] In this embodiment, the optimized design method is obtained by:

[0086] According to the initial design scheme selected by the user, the objective function is to minimize weight, maximize stiffness or minimize engineering quantity. Iterative optimization is performed within the scope of satisfying constraints and dimensional indicators to obtain the optimized design scheme.

[0087] The human-computer interaction unit is used to display the current optimized design scheme and accept the user's interactive instructions to adjust the optimized design scheme to obtain the final design scheme;

[0088] Through the visual interface, users can view the generated design scheme, result analysis and engineering quantities. Users can interact with the generated design scheme through natural language to modify and improve the geotechnical engineering design scheme.

[0089] This embodiment provides a complete and fully functional architecture diagram, such as Figure 2 As shown, it includes intelligent deformation calculation module, engineering quantity statistics and engineering cost analysis module, design result generation module, built-in geotechnical model module, and database module. The design result generation module, built-in geotechnical model module, and database module jointly realize the function of the scheme generation unit, and the intelligent deformation calculation module, engineering quantity statistics and engineering cost analysis module jointly complete the content of the evaluation and analysis unit. And it is called and linked through the visual interface in the human-computer interaction unit.

[0090] The visualization interface includes four partitions: artificial intelligence information interaction area, basic conditions display and design results display area, built-in module and engineering quantity display area, and drawing tool area.

[0091] The artificial intelligence model information interaction area is located on the left side of the user interface. It is a geotechnical engineering-specific artificial intelligence model data interaction interface. There is a content input window at the bottom of the interface. Users can input information through the window to achieve information interaction with the geotechnical engineering-specific artificial intelligence model. The text part of the geotechnical engineering survey report, the geotechnical engineering survey plan, and the geotechnical engineering survey profile can be imported through the information interaction window. All user-artificial intelligence model interaction data are displayed sequentially at the top of the interactive area dialogue window.

[0092] The basic conditions display and design results display area is located in the middle of the user interface. Engineering geological conditions, environmental conditions, and geotechnical engineering design results can be displayed through this area, which includes the following functions: Graphics window: This is the main area for creating, displaying, and editing graphics. It provides a virtual three-dimensional space. The geotechnical engineering design scheme recommended by artificial intelligence can be displayed in this area. Users can draw plane graphics, create three-dimensional models, and modify the geotechnical engineering design graphics recommended by artificial intelligence in this space. Mobile interaction dialog box: When the user selects a graphic unit, the mobile interaction dialog box is displayed at the cursor position. Users can enter information or suggestions in the mobile interaction dialog box to interact with the artificial intelligence model data.

[0093] The built-in modules and engineering quantity display area are located on the right side of the user interface. The upper part dynamically displays the engineering quantities and material usage in the current design scheme. The lower part recommends the design modules that can be used in the design. The characteristic of geotechnical engineering design is that the design results are highly targeted. Each project needs to be designed based on geological conditions, soil layer parameters, building foundation conditions, construction technology, material selection, construction monitoring and environmental protection and other factors. However, the selection of geotechnical engineering design schemes is repetitive. The system of the present invention fully utilizes the design characteristics of the industry and decomposes the geotechnical engineering design scheme into modules. Users can select standard modules for design or design modification, which greatly improves the efficiency of geotechnical engineering design.

[0094] The drawing tool area includes the Quick Access Toolbar, the Ribbon, the File Tab Bar, the Command Line, and the Status Bar. Quick Access Toolbar: Located at the top of the interface, it provides quick access to new projects, open projects, save projects, and more. Ribbon: Located at the top of the interface, it is an area for placing various tool icons. Users can use these tool icons to call drawing, editing, and management functions. The Ribbon usually contains multiple panels, each of which has related tools and controls. Users can click the tabs at the top of the Ribbon to switch between different panels. File Tab Bar: Displays the name of the currently opened file. Users can switch between different drawings by clicking the file tab bar. In addition, right-clicking the tab can also save or close the drawing. Command Line: Located at the bottom of the interface, it is used to enter commands and display information during the execution of commands. The command line is an important interactive tool that can be used to complete some complex operations. Status Bar: Located at the bottom of the interface, it displays the current working status and some auxiliary tools, such as snap, grid, orthogonal mode, etc.

[0095] In this embodiment, a verification module is also included:

[0096] The optimized design scheme is verified based on site conditions and usage requirements, specifically:

[0097] Step B1, collect summary verification, parameters and summary verification conditions, use multimodal technology such as mm Transformer or GANs model to identify the geotechnical engineering investigation report of the site in the specified design parameters, and obtain the geotechnical parameters of the engineering site;

[0098] Step B2: The geotechnical engineering designer sets the design conditions in advance according to the target area; and stores the industry standard verification method in the algorithm library;

[0099] Step B3, based on the geotechnical parameters of the engineering site, call the corresponding data required by the standard verification methods in various industries, perform verification according to the standard verification methods in the industry, and output the verification results and verification suggestions.

[0100] Multimodal technology identifies the geotechnical parameters of the project site in the geotechnical engineering investigation report, obtains the surrounding conditions and deformation requirements of the foundation pit through interaction with the user, and verifies the stability of the foundation pit support system according to site conditions and environmental constraints to ensure safety and stability during construction and use. Verification includes stability and deformation calculations: the resistance verification of the support structure includes stability and deformation problems that may occur during construction and use. Groundwater control design verification includes foundation pit anti-seepage stability calculation, pit bottom sudden surge stability verification, groundwater level control calculation, foundation treatment bearing capacity verification, foundation deformation verification, foundation stability verification and comprehensive analysis verification;

[0101] The anti-seepage stability calculation of the foundation pit is as follows: in areas with high groundwater levels, when the bottom of the foundation pit is gravel and sand and there is seepage water pressure, an anti-seepage stability verification is performed to avoid soil flow; this involves the insertion depth or embedding length of the retaining structure, which must meet the anti-seepage stability conditions.

[0102] The pit bottom sudden surge stability verification is to perform the pit bottom sudden surge verification when the pit bottom is an impermeable layer and there is pressurized water under the impermeable layer.

[0103] The groundwater level control calculation is as follows: the design and construction of groundwater control should meet the requirements of environmental protection around the foundation pit, and should be comprehensively analyzed and verified based on the site and surrounding engineering geological conditions, hydrogeological conditions and environmental conditions, combined with the foundation pit support and foundation construction plan.

[0104] The bearing capacity verification of the foundation treatment is as follows: the bearing capacity verification of the bearing layer foundation should meet the conditions of pk≤fa, pkmax≤1.2fa; the bearing capacity verification of the soft underlying layer foundation should meet the condition of pz+pcz≤faz.

[0105] The foundation deformation verification is to verify the maximum foundation deformation characteristics, which include settlement, settlement difference, inclination and local inclination, etc.

[0106] The foundation stability verification is to verify the foundation stability using the circular sliding surface stability analysis method.

[0107] After the verification module completes the verification, the system will give the verification results and verification suggestions. Geotechnical engineers can adjust the generated design plan through the interactive interface and finally determine the best geotechnical engineering design plan.

[0108] The corresponding verification process and parameter values ​​can be exported in Word format for verification and analysis by geotechnical engineering designers.

[0109] Project monitoring plan layout, specifically for foundation pit support design, slope and landslide control design, deformation monitoring plan needs to be formulated: This system can recommend and generate deformation monitoring plan according to the nature of the project and industry specifications, including monitoring purpose and basis: clarify the monitoring objectives, including ensuring project safety, preventing and controlling potential risks, etc., as well as relevant regulations and technical standards for formulating monitoring plans. Monitoring content and items: list all items that need to be monitored, such as displacement and settlement of support structures, displacement of underground facilities, surface settlement and cracks, water seepage, etc. Monitoring methods and accuracy requirements: select appropriate monitoring methods, such as measurement technology, instrument type, etc., and specify corresponding accuracy requirements to ensure the accuracy of monitoring data. Measurement point layout: determine the location and number of monitoring points to facilitate comprehensive and accurate data collection. Determination of alarm values ​​and emergency measures: set early warning values ​​and alarm values ​​for each monitoring indicator, and formulate emergency measures to be taken once these values ​​are reached. Monitoring time limit and frequency: plan the monitoring schedule and frequency to adapt to different stages of project progress.

[0110] Calculation and statistics of bill of quantities: The bill of quantities and material list corresponding to each entity part can be output, including detailed lists of specific materials such as steel, concrete, cement, etc.

[0111] Vector graphics drawing: The dedicated artificial intelligence model for geotechnical engineering can draw vector graphics in the basic conditions display and design results display area of ​​this system by compiling codes, and assign material properties and unique codes to each graphic.

[0112] The export unit is used to export the final design plan in a set standard format. Once the final design plan is confirmed, the system can export two-dimensional and three-dimensional drawings and engineering cost lists.

[0113] In this embodiment, machine learning technology is also used to learn from user feedback and improve subsequent generated designs, which helps the tool better adapt to user needs and provide more accurate design solutions.

[0114] In this embodiment, a design habit preference learning module is also included, and conditional generative adversarial networks (cGANs) can be used for learning.

[0115] The conditional adversarial generative network adds additional conditional information such as user preference labels to the traditional GAN ​​structure, so that the generator can generate a design scheme of a specific style or in accordance with certain customary preferences based on these preference labels. During training, only the difference between the geotechnical engineering design scheme generated by the present invention and the geotechnical engineering design scheme modified by the user is learned to avoid the situation where the originally set technical standards are not met in order to adjust the design representation.

[0116] The artificial intelligence-based generative geotechnical engineering design system provided by this embodiment has a wide range of applications and can be used for pile foundation engineering design: This includes pile design, such as selecting the type and arrangement of piles, calculating the bearing capacity and settlement of single piles or pile groups, performing reinforcement design, and pile construction and inspection and acceptance. Foundation engineering design: Foundation engineering involves the use of different foundation treatment technologies to design foundation schemes, such as replacement cushion method, preloading method, vibration method, sandstone pile method, strong tamping method, deep mixing method, high-pressure jet grouting method, etc. Slope engineering design: Including slope stability design and protection measures, as well as the design of related structures such as road slopes, water conservancy embankments and earth-rock dams. Foundation pit engineering design: This includes the design of foundation pits and underground engineering, groundwater control, and the construction of retaining and support systems. Geological disaster prevention and control design: Identify the basic work of geological disasters, and design geological disaster control measures such as breakwaters, reinforced slopes, drainage projects, and protective nets. Other fields: It may also include design work in tunnels and underground engineering, earthquake engineering, and blasting engineering.

[0117] Although the various steps in the above embodiment are described in the above-mentioned order, those skilled in the art can understand that in order to achieve the effect of this embodiment, different steps do not have to be executed in such an order. They can be executed simultaneously (in parallel) or in a reverse order. These simple changes are within the scope of protection of the present invention.

[0118] The second embodiment of the present invention is described by taking a foundation pit support design project as an example:

[0119] Open the human-computer interaction unit, enter the menu bar, click Start, select New Project, and enter the project name and design parameters:

[0120] Import the engineering plan from the AI ​​model information interaction window. The platform graphics can be in dwg format, pdf format or image format. The system will identify the plan according to the annotated information and convert it into vector graphics. The imported engineering plan is displayed in the basic conditions display and design results display area of ​​the user interface. The geotechnical engineering special large model intelligently identifies the building red line range, the relationship between the buildings and pipelines around the foundation pit, the outline of the proposed building, the elevation or depth of the building base and other design information.

[0121] Modification of intelligent recognition errors in design information. Since it is difficult to standardize and unify the imported engineering plan, key information recognition errors may occur. This can be modified through the mobile interactive dialog box. The mobile interactive dialog box is displayed at the cursor position when the user selects a graphic unit. For example, the following method can be used to modify the recognition error of the building outline. Use the mouse to operate the cursor to select the correct building outline. The mobile interactive dialog box pops up at the cursor. Enter "building outline" in the dialog box. The system memorizes the new building outline and becomes the design condition information. The building outline is modified. Errors in the recognition of information such as base depth and surrounding environment can be modified in the mobile interactive dialog box after selecting the graphic.

[0122] Determine the inner edge line of the foundation pit. You can instruct the geotechnical engineering special large model to automatically generate the inner edge line of the foundation pit from the artificial intelligence model information interaction window; you can also draw the inner edge line of the foundation pit in the drawing area and interactively feed back the geotechnical engineering special large model.

[0123] After importing the text part of the geotechnical investigation report, geotechnical investigation plan, and geotechnical investigation profile from the artificial intelligence model information interaction window, the artificial intelligence model identifies the geotechnical investigation information. The three-dimensional geological model generated based on the geotechnical investigation report information is displayed in the basic conditions display and design results display area. The three-dimensional geological model marks the location of the geotechnical investigation borehole. You can move the cursor and click on the three-dimensional geological model with the mouse to display the name of the stratum, stratum layer number, bearing capacity, cohesion, internal friction angle and other stratum information.

[0124] Enter "Generate foundation pit support plan" in the artificial intelligence model information interaction window. The large model dedicated to geotechnical engineering will analyze the known foundation pit design information and the geological conditions of the site and recommend a foundation pit support plan. In this example, the artificial intelligence model recommends the "concrete piles + prestressed anchor rods" plan. The user thinks the plan is feasible and enters "Generate plan diagram" in the artificial intelligence model information interaction window. The corresponding foundation pit support body is automatically generated along the outer edge line of the foundation pit. The automatically generated entities in this example include concrete piles, prestressed anchor rods, anchor heads, channel steel waist beams, steel mesh sprayed concrete between piles, and crown beams.

[0125] Generate concrete piles in the support structure: pile length 20 meters, pile diameter 1 meter, pile center spacing 4 meters, C25 concrete pile body, pile body equipped with steel bars, the system assigns a unique number to each concrete pile.

[0126] Prestressed anchor rods in the generated support structure: the total length of the anchor rods is 20 meters, the inclination angle is 15 degrees, the prestressed anchor rods use 1×7Φs15.2 steel strands, the standard value of tensile strength is 1860MPa, the free section length is 5 meters, the anchoring section length is 15 meters, and the slurry design strength is M30.

[0127] The channel steel waist beam is located at the anchor head of the prestressed anchor rod and surrounds the foundation pit:

[0128] Crown beam: Located at the top of the concrete pile, the crown beam is 1 meter wide and 1.5 meters high. It is poured with C30 concrete and steel bars are arranged inside the crown beam.

[0129] Geotechnical engineers can analyze the generated support scheme. If modification is required, they can select the target entity and input modification suggestions through the artificial intelligence model information interaction window. The corresponding target entity will be adjusted according to the input suggestions.

[0130] If you find that the length of the concrete pile can be reduced to 18 meters and the pile diameter can be reduced to 0.8 meters, you can modify it by the following operations. Select the pile to be modified in the drawing area, and enter "change the pile length to 18 meters and the pile diameter to 0.8 meters" in the pop-up dialog box. The large model dedicated to geotechnical engineering will understand the instructions of the geotechnical engineer and modify the selected entities. The system's default way to modify the pile length is to shorten the concrete piles in the selected area from the bottom of the pile. The system's default way to modify the pile diameter is that after the pile diameter is modified, the edge of the pile is still arranged externally along the edge line of the foundation pit. The anchor head position of the corresponding prestressed anchor rod is also modified accordingly. The target entity to be modified can be selected by box selection. Other support entities may be selected during box selection. Box selection of other support entities does not affect the execution of the modification intention.

[0131] Select the prestressed anchor to be modified in the drawing area, and enter "the total length of the anchor is changed to 25 meters, the inclination is 14 degrees, the prestressed anchor uses 1×9Φs15.2 steel strands, and the free section length is 6 meters" in the pop-up dialog box. The large model dedicated to geotechnical engineering will understand the instructions of the geotechnical engineer and modify the selected entity. The system default method for modifying the length of the prestressed anchor is to shorten the non-anchored end of the prestressed anchor in the selected area. The system default method for modifying the inclination is to keep the position of the anchor end unchanged and adjust the position of the non-anchored section to meet the inclination requirements. The target entity to be modified can be selected by box selection. Other support entities may be selected during box selection, but box selection of other support entities does not affect the execution of the modification intention.

[0132] The modification method of the support structure generated by the system is similar. Select the entity to be modified in the drawing area and enter the modification suggestions in the pop-up dialog box. The large model dedicated to geotechnical engineering will understand the instructions of the geotechnical engineer and modify the support entity.

[0133] The method of the present invention has a built-in deformation analysis module. After the support scheme is modified, "deformation analysis" can be entered in the artificial intelligence model information interaction window. The deformation analysis module calculates the deformation of the support body and generates a deformation diagram of the support structure. The magnitude of the deformation is displayed in different colors. After the deformation analysis, the geotechnical engineer can decide whether to further modify the model. The foundation pit deformation analysis calculation is not a necessary step, and the geotechnical engineer can decide whether to perform deformation calculation.

[0134] Input “monitoring point arrangement” in the artificial intelligence model information interaction window, and the system of the present invention will perform intelligent monitoring point arrangement according to the specification requirements, and the geotechnical engineer can modify the setting of the monitoring points in an interactive manner.

[0135] Enter "Generate plan view and section view" in the artificial intelligence model information interaction window, and the system will automatically generate the plan view and section view of the foundation pit support plan. If you need to adjust the position and number of the section view, you can modify it by entering the modification suggestions in the interaction window.

[0136] Enter "Generate Bill of Quantities" in the artificial intelligence model information interaction window, and the system will export the bill of quantities for this design plan.

[0137] An electronic device according to a third embodiment of the present invention includes:

[0138] at least one processor; and

[0139] a memory communicatively connected to at least one of the processors; wherein,

[0140] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the above-mentioned artificial intelligence-based generative geotechnical engineering design system.

[0141] A fourth embodiment of the present invention is a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the above-mentioned artificial intelligence-based generative geotechnical engineering design system.

[0142] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process and related instructions of the storage device and processing device described above can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0143] Those skilled in the art should be able to appreciate that the modules and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two, and the programs corresponding to the software modules and method steps can be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the technical field. In order to clearly illustrate the interchangeability of electronic hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are performed in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0144] The terms "first", "second", etc. are used to distinguish similar objects rather than to describe or indicate a particular order or sequence.

[0145] The term "comprise" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that includes a list of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article, or apparatus / device.

[0146] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A generative geotechnical engineering design system based on artificial intelligence, characterized in that: The system comprises: A design parameter input unit, used to collect design parameters specified by a user; A scheme generating unit, used for generating an initial design scheme based on the design parameters by using a geotechnical engineering-specific artificial intelligence model; The geotechnical engineering-specific artificial intelligence model, whose basic model includes a scheme understanding model, a terminology understanding model, a component function learning model, a component relationship function learning model, a terminology-function mapping model, and a scheme generation model. The geotechnical engineering-specific artificial intelligence model training method includes: Step A1, collect historical geotechnical engineering design data and clean, organize and annotate them; Step A2, learning geotechnical engineering industry terminology and generating an initial technical solution through the geotechnical engineering-specific artificial intelligence model to be trained; Step A3, calculating the accuracy, recall and F1 score of each basic model in the trained geotechnical engineering-specific artificial intelligence model according to the initial technical solution; Step A4, adjusting the parameters of the model, repeating steps A1 to A3 until the accuracy, recall rate and F1 score all reach the respectively set thresholds, and obtaining a trained geotechnical engineering-specific artificial intelligence model; An evaluation and analysis unit, used to evaluate the initial design scheme according to set dimensional indicators and obtain evaluation results of multiple dimensional indicators; The optimization design unit is used to optimize the initial design solution selected by the customer through the optimization algorithm according to the evaluation results of multiple indicators to obtain the optimized design method; The optimized design method is obtained by: According to the initial design scheme selected by the user, the objective function is to minimize weight, maximize stiffness or minimize engineering quantity, and iterative optimization is performed within the scope of satisfying constraint conditions and dimensional indicators to obtain the optimized design scheme; The human-computer interaction unit is used to display the current optimized design scheme and accept the user's interactive instructions to adjust the optimized design scheme to obtain the final design scheme; The export unit is used to export the final design plan in a set standard format.

2. The artificial intelligence-based generative geotechnical engineering design system according to claim 1, characterized in that: The design parameters include: design space dimensions, site rock and soil characteristic indicators, constraint conditions, deformation requirements and durability requirements.

3. The artificial intelligence-based generative geotechnical engineering design system according to claim 2, characterized in that: The scheme generating unit comprises: Pre-divide and store historical geotechnical engineering design plans into independent component modules; Component modules are selected and combined according to the design conditions input by the user and the functions required for geotechnical engineering, and an initial design scheme that meets the constraints and performance indicators is generated through a generative artificial intelligence algorithm.

4. The artificial intelligence-based generative geotechnical engineering design system according to claim 3, characterized in that: The component modules include: soil nails, anchor rods, anchor heads, prestressed anchor rods, various types of I-steels, various types of T-steels, various types of steel pipes, geogrids, geomembranes, geotextiles, shotcrete, pipe wells, light well points, underground continuous walls, composite soil nail walls, steel sheet piles, cast-in-place piles, compression pile reinforcement models and bending pile reinforcement models; Each component module comes with corresponding functional attributes, which include the module's basic target functions, applicable scenarios, performance indicators, reliability, maintainability and durability.

5. The artificial intelligence-based generative geotechnical engineering design system according to claim 1, characterized in that: The set dimensional indicators include performance, cost, deformation, settlement, construction feasibility, and structural stability.

6. The artificial intelligence-based generative geotechnical engineering design system according to claim 1, characterized in that: Also includes verification modules: Collect and summarize verification, parameters and summary verification conditions, use multimodal technology to identify the geotechnical engineering investigation report of the site in the specified design parameters, and obtain the geotechnical parameters of the engineering site; The design conditions set by the geotechnical designer in advance according to the target area; And pre-store the industry standard verification methods in the algorithm library; Based on the geotechnical parameters of the engineering site, the corresponding data required by the standard verification methods in each industry are called, and verification is performed according to the standard verification methods in the industry, and the verification results and verification suggestions are output.

7. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the artificial intelligence-based generative geotechnical engineering design system described in any one of claims 1-6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the artificial intelligence-based generative geotechnical engineering design system described in any one of claims 1-6.

Images