Agent-based arch dam digital design method and device

Abstract

The application relates to the field of intelligent design of arch dams, and discloses an arch dam digital design method and device based on an intelligent agent, which comprises the following steps: taking a large language model as a decision-making large model for calling various arch dam design tools; collecting and pre-processing arch dam design field data to form an arch dam design data set, and constructing an arch dam design knowledge base based on the data set by using a knowledge base construction tool; extracting an initial design parameter list from an initial arch dam design file by using a design parameter extraction tool, and determining an arch dam design necessary design parameter list from the arch dam design knowledge base, and performing a comparison check, and reminding a user to supplement if there is a missing part; calling an arch dam modeling tool, an agent model construction tool and an arch dam body shape optimization tool to realize automatic generation and optimization of a complete arch dam body shape with a dam foundation. The method is used to solve the problems of dependence on manual work, low efficiency and difficult experience inheritance in traditional arch dam design, and can realize efficient, high-quality and intelligent arch dam design.

Description

Technical Field

[0001] This invention relates to the field of intelligent arch dam design, and specifically to a digital intelligent design method and device for arch dams based on intelligent agents. Background Technology

[0002] Dams play a crucial role in safeguarding water resources and providing energy. Among various dam types, arch dams are widely used due to their unique structural advantages. With the rise of digitalization and intelligentization, the water conservancy industry is moving towards intelligent dam design, which not only improves project quality and efficiency but also aligns with the development needs of modern intelligent construction technologies.

[0003] However, current arch dam structural design still largely relies on traditional engineer-led drafting and modeling methods, with relatively insufficient levels of intelligent technology. The current design model heavily depends on the personal experience of senior engineers, which brings several challenges: First, numerous completed arch dam projects have accumulated a wealth of design reports, publications, and valuable experience. However, acquiring this experience is costly, and it is difficult to systematically accumulate and reuse it, leading to a bottleneck in the transmission of professional knowledge. Second, due to the inherent limitations of the manual design process, engineers' productivity is restricted, making it difficult to significantly improve design efficiency. The traditional design model, relying on manual experience, has gradually revealed its shortcomings in knowledge transmission and the reuse of results, urgently requiring an intelligent revolution. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, the present invention provides a digital design method and device for arch dams based on intelligent agents, which utilizes artificial intelligence technology to automatically extract design parameters and efficiently carry out arch dam design.

[0005] In a first aspect, the present invention provides a digital intelligent design method for arch dams based on intelligent agents, comprising: Obtain the initial design documents for the arch dam; The design parameter extraction tool is invoked to extract the list of necessary design parameters for the arch dam design from the pre-built arch dam design knowledge base, and to extract the list of initial design parameters from the initial arch dam design file. The initial design parameter list is compared with the necessary design parameter list; if any are missing, they are added until all parameters are checked and a complete design parameter list is output. The arch dam modeling tool is invoked to perform parametric geometric modeling based on the design parameter list, generating a three-dimensional geometric model of the arch dam. Based on the three-dimensional geometric model of the arch dam and the complete list of design parameters, the structural performance of the arch dam is calculated. The performance calculation results are input into a proxy model building tool to construct a proxy model. The arch dam shape optimization tool is then invoked to perform a dual-objective iterative optimization of the arch dam shape, with the optimization objectives being to minimize the total cost of each grade of concrete in different zones of the dam body and to minimize the volume of the tensile stress zone of the arch dam. During the iteration process, the candidate schemes generated by the arch dam shape optimization tool are evaluated using the proxy model. When the evaluation results meet the preset conditions, the iterative optimization stops, and the optimized Pareto front shape scheme is output.

[0006] As a further technical solution, the construction of the intelligent agent includes: A large decision-making model is constructed based on a large language model. This model is used to invoke and execute tools such as knowledge base construction tools, design parameter extraction tools, arch dam modeling tools, finite element analysis tools, surrogate model construction tools, and arch dam shape optimization tools. The knowledge base building tool is used to build the arch dam design knowledge base; The design parameter extraction tool is used to extract parameters from the arch dam design knowledge base and the initial design file; The arch dam modeling tool is used to construct a three-dimensional geometric model of the arch dam; The finite element analysis tool is used to perform precise calculations of the performance of the arch dam structure and output the calculation results of the arch dam performance. The proxy model building tool is used to build proxy models; The arch dam shape optimization tool is used to iteratively optimize the performance data of the arch dam and generate candidate design scheme data.

[0007] As a further technical solution, the pre-construction of the arch dam design knowledge base includes: We acquired multi-source text data in the field of arch dam design, and used natural language processing technology to transform, standardize and clean the acquired data to form an arch dam design dataset, including: design specifications, manuals, design reports of existing and under-construction arch dams, monographs related to arch dam design and shape optimization, textbooks related to concrete dams, and acceptance and inspection procedures for concrete dams. Natural language processing technology is used to transform, standardize, and clean the acquired data. Specifically, OCR technology and computer vision technology are combined to divide the multi-source text into page layouts, identify elements such as headers, footers, images, formulas, tables, and body text, and convert the formulas, tables, and body text into editable .md, .json, or .txt files. Then, a data processing script is written to correct typos and line breaks that occur during the transformation process, and a stop word list is used to clean the text data content to obtain the arch dam design dataset. Based on the arch dam design dataset, a knowledge base for arch dam design is constructed using a knowledge base construction tool, which includes a knowledge graph construction sub-tool and a vector knowledge base construction sub-tool. The knowledge graph construction sub-tool first divides the dataset text into paragraphs, and then performs entity extraction and relation extraction on each paragraph one by one based on deep learning algorithms to obtain the arch dam design knowledge graph; The vector knowledge base construction sub-tool dynamically generates semantically complete paragraphs based on paragraph end symbols and character length limitations during the text segmentation stage, and assigns a number to each paragraph; subsequently, it uses a text embedding model to map these paragraphs to a semantic vector space to obtain a vector database.

[0008] As a further technical solution, the step of calling the design parameter extraction tool includes: The design parameter extraction tool is invoked to extract a list of necessary design parameters for arch dam design from the arch dam design knowledge base based on text semantic retrieval functionality; the design parameter extraction tool is constructed based on a semantic retrieval algorithm. The design parameter extraction tool is invoked to extract the initial design parameter list from the initial design document of the arch dam.

[0009] As a further technical solution, the steps of the comparative inspection include: The initial design parameter list is checked against the necessary design parameter list. If any of the uploaded initial design parameters are found to be missing, a prompt message will be generated and the missing design parameters will be fed back. After they are supplemented, the comparison check will be performed again, and the comparison check will be performed repeatedly until a complete list of design parameters is obtained. The complete list of design parameters includes shape-related design parameters and load-related design parameters.

[0010] As a further technical solution, the step of calling the arch dam modeling tool includes: Based on the complete list of design parameters, the preset arch dam modeling code is called, and the parametric design method is used to realize the digital expression of the arch crown beam section and the horizontal arch ring respectively. The simulated terrain line is extracted through the terrain plan view to construct the dam foundation slope excavation outline. The arch dam modeling tool is invoked to automatically generate a complete geometric model of the arch dam, including the dam foundation.

[0011] As a further technical solution, the iterative optimization steps include: The finite element analysis tool is invoked to complete the finite element analysis process, including mesh generation, load application, solution, and post-processing, based on the arch dam geometric model and the complete design parameter list. This enables accurate numerical simulation of the three-dimensional arch dam model and obtains finite element training samples. Once the number of finite element training samples reaches the training requirements of the surrogate model, the decision-making big model calls the surrogate model building tool to build a surrogate model based on the finite element training samples; the surrogate model is then called to perform rapid performance prediction on the candidate design points output by the arch dam shape optimization tool, replacing the finite element analysis tool for approximate calculation and re-analysis; The arch dam shape optimization tool is invoked, and combined with the evaluation capabilities of the surrogate model, the optimization objectives are to minimize the total cost of each grade of concrete in different zones of the dam and to minimize the volume of the tensile stress zone of the arch dam. Taking into account constraints such as dam geometry and finite element equivalent stress, the design parameters of the arch dam shape are iteratively optimized, outputting candidate arch dam shapes. In each iteration, the surrogate model evaluates the performance and constraint satisfaction of the candidate arch dam shapes and feeds the evaluation results back to the arch dam shape optimization tool. The arch dam shape optimization tool updates the population or candidate solution set based on the optimization algorithm and Pareto dominance relationship, selecting a better scheme to continue iterating. When the evaluation results meet the convergence accuracy requirements of the optimization objectives or the number of iterations reaches a preset maximum value, the iteration stops, resulting in a Pareto front shape scheme that balances economy and safety.

[0012] Secondly, the present invention provides a digital design device for arch dams based on intelligent agents, comprising: The first processing module is used to obtain the initial design files for the arch dam. The second processing module is used to call the design parameter extraction tool to extract the list of necessary design parameters for the arch dam design from the pre-built arch dam design knowledge base, and to extract the list of initial design parameters from the arch dam design initial file. The third processing module is used to compare and check the initial design parameter list with the necessary design parameter list; if there are any missing parameters, they are added until all parameters are checked and a complete design parameter list is output. The fourth processing module is used to call the arch dam modeling tool based on the complete design parameter list, perform parametric geometric modeling based on the design parameter list, and generate a three-dimensional geometric model of the arch dam. The fifth processing module is used to perform performance calculations on the arch dam structure based on the three-dimensional geometric model and complete design parameter list. The performance calculation results are then input into a proxy model building tool to construct a proxy model. The module then calls an arch dam shape optimization tool to perform a dual-objective iterative optimization of the arch dam shape, with the optimization objectives being the minimum total cost of concrete of each grade in different dam sections and the minimum volume of the tensile stress zone of the arch dam. During the iteration process, the proxy model is used to evaluate the candidate schemes generated by the arch dam shape optimization tool. When the evaluation results meet preset conditions, the iterative optimization stops, and the optimized Pareto front shape scheme is output.

[0013] Thirdly, the present invention provides an electronic device, comprising: At least one memory for storing programs; At least one processor is configured to execute a program stored in memory, wherein, when the program stored in memory is executed, the processor is configured to perform the method described in the first aspect or any possible implementation thereof.

[0014] Fourthly, the present invention provides a non-transitory computer-readable storage medium, comprising: A computer program instruction, when executed on a digital design apparatus, causes the digital design apparatus to perform the method described in the first aspect or any possible implementation thereof.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a digital intelligent design method and device for arch dams based on intelligent agents. For the design process of arch dams, including design parameter extraction, parameter completeness checking, arch ring design, arch centerline design, thickness design, and upstream surface curve design, it collects existing knowledge such as relevant design specifications, manuals, and design reports of existing arch dams. Using natural language processing methods, this knowledge is aggregated, processed, and formatted into a standardized knowledge base. Leveraging the reading and understanding capabilities of a large-scale decision model, it enables the learning, inheritance, and engineering reuse of relevant knowledge. A design parameter extraction tool automatically extracts existing parameters from the initial design documents. Through iterative self-checking and user interaction, it assists engineers in completing the necessary design parameters. Based on the shape design parameters, an arch dam modeling tool is used to construct a 3D model of points, lines, surfaces, and volumes. Based on the generated 3D model and load-related parameters, further structural performance evaluation and analysis are conducted using finite element analysis software and a proxy model. This invention adopts an agent-led overall framework, using a large language model to simulate the design behavior of engineers, flexibly calling knowledge base construction tools, design parameter extraction tools, and arch dam modeling tools, thereby realizing high-quality arch dam design and accelerating the transformation of arch dam design from experience-based manual drawing design to intelligent and efficient design. Attached Figure Description

[0016] To more clearly illustrate the implementation process of the technical solution described in this invention, the following will provide an explanatory description of the method execution process through embodiments and in conjunction with the accompanying drawings. Obviously, the following drawings and embodiments are only one or more possible situations involving this invention. For those skilled in the art, other implementation methods can be derived by referring to the following embodiments and drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the system architecture of the intelligent design method and device for arch dams based on intelligent agents in an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the intelligent agent architecture in an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram illustrating the implementation process of the intelligent design method and device for arch dams based on intelligent agents in an embodiment of the present invention.

[0020] Figure 4 This is a schematic diagram of a knowledge graph construction model provided in an embodiment of the present invention.

[0021] Figure 5 This is a schematic diagram of a design parameter extraction process provided in an embodiment of the present invention.

[0022] Figure 6 This is a schematic diagram of the shape design process in an arch dam modeling tool provided in an embodiment of the present invention.

[0023] Figure 7 This is a schematic diagram of an operating device for an intelligent digital design method for arch dams, provided in an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0025] The intelligent design method and apparatus for arch dams based on intelligent agents provided in this invention have the following system architecture: Figure 1 As shown, it runs in a collaborative computing environment consisting of a terminal (101) and a server (103). The terminal (101) establishes a communication link with the server (103) through the network. The server (103) can access an independently set up or cloud-based data repository (105) for persistent storage of knowledge, model parameters and design process data in the field of arch dam design.

[0026] The terminal (101) serves as the user interaction entry point, used to upload multi-source text (such as language descriptions, requirement reports, Excel documents, etc.) or initial design files containing initial design intent. The server (103), as the core of the intelligent agent, receives data and, through the decision-making model, sequentially executes the following autonomous, process-oriented task chain: By calling NLP processing tools and integrating text extraction and computer vision technologies, the uploaded multi-source heterogeneous text is intelligently parsed, and elements such as tables, formulas, and body text are identified and deeply cleaned, transforming it into a high-quality, structured arch dam design dataset, laying the foundation for knowledge base construction. The knowledge base construction tool performs deep semantic analysis on the structured dataset, extracts entities and relations in parallel through deep learning algorithms, constructs a knowledge graph and a vector database, forms a dam design knowledge base that can be queried and reasoned by intelligent agents, and stores the results in the data storage repository (105) so that the construction step is eliminated when using it, thus improving system efficiency; The design parameter extraction tool is activated. This tool has two capabilities: first, based on a semantic retrieval algorithm, it intelligently locates the list of necessary design parameters for the arch dam design from the knowledge base; second, relying on the contextual understanding capabilities of a large language model, it accurately extracts design parameters from the initial design files uploaded by the user. Subsequently, the tool initiates a cyclical self-checking mechanism, comparing the extracted parameters with the list of necessary design parameters. If any are found to be missing, a prompt is immediately generated and fed back to the terminal (101), requesting the user to supplement the parameters, until a complete and compliant set of design parameters is obtained.

[0027] Once the parameters are complete, the decision-making big model, acting as the "brain" of the intelligent agent, is activated. Based on all input parameters, it autonomously invokes the arch dam modeling tool using the model context protocol. This tool, based on the design principles of arch dams, automatically executes a series of digital modeling steps, including arch ring layering, centerline fitting, arch crown beam thickness fitting, and upstream surface fitting, ultimately driving the 3D modeling engine to generate an accurate, modifiable arch dam shape model.

[0028] Use finite element analysis tools to obtain a sufficient number of finite element training samples through high-precision numerical simulation until the training requirements of the surrogate model are met. The proxy model building tool is called to establish a nonlinear mapping relationship between the shape parameters of the arch dam and the output results of the finite element method, and a proxy model is generated. Test set data is imported into the proxy model, the performance of the test set samples is predicted, and the accuracy of the proxy model is verified by the test set samples. The arch dam shape optimization tool is invoked, a surrogate model is imported, and candidate arch dam design schemes are generated based on the optimization algorithm. The surrogate model quickly evaluates the performance of the schemes, outputs the performance evaluation results, and feeds them back to the shape optimization tool. The shape optimization tool updates the population or candidate solution set based on the optimization algorithm and Pareto dominance relationship, selects the better scheme, and continues to iterate until a Pareto frontier shape scheme that balances economy and safety is found.

[0029] The server (103) feeds back the final generated digital model of the arch dam and the design report to the terminal (101). This method emphasizes the autonomous decision-making and tool calling capabilities of the agent, and its entire process can also be completed in a single server (103) or a powerful terminal (101).

[0030] The terminal (101) includes, but is not limited to, various workstations (such as high-performance desktop computers, graphics workstations), mobile devices (smartphones, tablets, smartwatches), and field survey equipment. The server (103) can be implemented using a high-performance standalone server, a distributed server cluster, or a cloud server. In this embodiment, the terminal (101) is mainly operated by the designer, who is responsible for initiating tasks and confirming parameters; the server (103) carries the core reasoning, decision-making, and execution functions of the intelligent agent.

[0031] The following is combined with Figure 2 , Figure 3 , Figure 4 , Figure 5 as well as Figure 6 The content shown is further explained through specific embodiments.

[0032] In one illustrative embodiment, the intelligent agent architecture of the intelligent agent-based arch dam digital design method and apparatus is as follows: Figure 2 As shown. The decision big model is used to receive design requirements from the terminal (101), analyze them and distribute the tasks to suitable tools. After the tools execute the tasks, they return the results to the decision big model. The decision big model analyzes them again and distributes the tasks to other tools. The above steps are repeated until the design requirements are met, and the design shape of the arch dam can be obtained. The design results are then transmitted back to the terminal (101).

[0033] In one illustrative embodiment, the implementation process of the intelligent agent-based digital design method and device for arch dams is given, including... Figure 3 The steps shown are as follows: Step 1: Obtain multi-source text data in the field of arch dam design; Step 2: Use natural language processing technology to transform, standardize and clean the acquired data to form an arch dam design dataset; Step 3: Based on the arch dam design dataset, construct an arch dam design knowledge base using a knowledge base building tool; Step 4: Use the design parameter extraction tool to automatically extract design parameters from the initial design file; Step 5: Automatically check if any design parameters are missing, remind the user to supplement any missing necessary design parameters, and repeat the self-check until complete design parameters are obtained; complete design parameters should include shape-related design parameters and load-related design parameters. Step 6: Based on the complete shape-related design parameters, call the arch dam modeling tool to automatically generate the complete arch dam shape, including the dam foundation; based on the generated arch dam geometric model and the complete design parameter list, call the finite element analysis tool to automatically calculate and analyze the arch dam structural performance; based on the generated initial shape and finite element analysis model, call the proxy model building tool and the arch dam shape optimization tool, and continuously iterate with the optimization objectives of minimizing the total cost of each grade of concrete in different dam sections and minimizing the volume of the tensile stress zone of the arch dam, until the optimal Pareto front shape with dual objectives is obtained, thus completing the arch dam design.

[0034] Step 1 involves acquiring multi-source text data in the field of arch dam design, including but not limited to: National and industry-issued arch dam design specifications, design manuals, engineering design reports, geological survey reports, special research reports, academic monographs on arch dam shape optimization and structural analysis, textbooks on concrete dams, and regulations for dam safety monitoring and acceptance inspection, etc.

[0035] Step 2 involves using natural language processing techniques to transform, standardize, and clean the acquired data to form an arch dam design dataset, including but not limited to: Combining OCR and computer vision technologies, this method divides multi-source text into page layouts and identifies elements such as headers, footers, images, formulas, tables, and body text. Convert formulas, tables, and text into editable .md, .json, or .txt files; A data processing script was written to correct typos and incorrect line breaks during the conversion process. Based on the publicly available Sichuan University stop word list, and combined with arch dam design documents, common stop words in the field of arch dam design were added to form a general stop word list. Then, the text data content was cleaned using the general stop word list to obtain the arch dam design dataset.

[0036] The knowledge base construction tools and arch dam design knowledge base mentioned in step 3 include, but are not limited to: The knowledge base construction tool includes a knowledge graph construction sub-tool based on deep learning algorithms. Scripts can be written to divide paragraphs into segments based on fixed length or line breaks. Text knowledge extraction algorithms are used to extract entities and relationships from each paragraph. Entities and relationships are then grouped into five-tuple knowledge chains in the form of (head entity - head entity type - entity relationship - tail entity - tail entity type). Subsequently, tail entities with the same entity are connected to the head entity of the next five-tuple, and this process is continuously extended to obtain the arch dam design knowledge graph. This knowledge base building tool includes a sub-tool for building a vector knowledge base based on text embedding models. It automatically divides text into paragraphs, numbers the paragraphs, and then uses text embedding models to map the text paragraphs to a semantic space. These text embedding models include, but are not limited to, BGE series models, Text2Vec models, Word2vec models, and Multilingual series models. The mapping results are then saved to the local disk for later use. The arch dam design knowledge base contains a knowledge graph that includes arch dam design steps, selection principles, methods for determining horizontal arch rings, methods for determining arch crown beams, design specifications, and existing design results. The arch dam design knowledge base contains textual materials covering a wider range of fields, including hydraulic structure design, construction, operation and maintenance, as well as vector knowledge bases including arch dam design steps, selection principles, methods for determining horizontal arch rings, methods for determining arch crown beams, design specifications, and existing design results.

[0037] The design parameter extraction tool mentioned in step 4 includes, but is not limited to: The text semantic retrieval function relies on semantic retrieval algorithms to determine the list of necessary design parameters for arch dam design from the arch dam design knowledge base; The design parameter extraction function relies on a large language model to automatically extract the initial design parameter list from the initial design file of the arch dam. At the same time, it performs a cyclic self-check on the automatically extracted design parameters based on the necessary design parameters until a complete design parameter list is obtained.

[0038] Step 5 involves automatically checking for missing design parameters and prompting the user to supplement any missing necessary design parameters until complete design parameters are obtained, including but not limited to: The comparison check involves comparing the extracted list of initial design parameters with the list of necessary design parameters required for the arch dam design. The system generates prompts; if any missing parameters are found, the self-checking system generates prompts to remind the user to supplement them.

[0039] Step 6 describes the use of a large decision model, based on complete design parameters, to call arch dam modeling tools, finite element analysis tools, surrogate model building tools, and arch dam shape optimization tools to achieve automated generation of arch dam shape, structural performance evaluation, and iterative optimization, including but not limited to: Calling the arch dam modeling tool refers to using the model context protocol to call the written arch dam modeling code. Based on the necessary design parameters, the arch dam is digitally represented through steps such as arch ring layering, arch ring centerline fitting, arch crown beam thickness fitting, upstream surface fitting at the arch crown, valley slope topographic line fitting, and dam body connection with the terrain. Then, the complete arch dam body shape including the dam foundation is automatically generated based on 3D modeling tools such as Rhino. Calling the finite element analysis tool for arch dams refers to calling the written finite element calculation code for arch dams based on the generated three-dimensional model of the arch dam and the necessary load design parameters. This includes a complete finite element analysis process, including mesh generation, load application, solution, and post-processing, thereby realizing the automated calculation and analysis of the performance of the arch dam structure. The arch dam shape optimization tool and the proxy model construction tool are used to carry out optimization with the dual objectives of minimizing the total cost of concrete of different grades in different zones of the dam and minimizing the volume of the tensile stress zone of the arch dam. When the evaluation results meet the convergence accuracy requirements of the optimization objectives or the number of iterations reaches the preset maximum value, such as 100 iterations, it is considered that the optimization has basically converged, and the iteration is stopped. The optimal Pareto front shape for the dual objectives is obtained, and the most suitable Pareto front is selected as the final design shape.

[0040] In another illustrative embodiment, see Figure 4 The knowledge graph construction model includes a text representation module, a feature partitioning module, a feature fusion module, and a feature classification module, wherein: The text representation module receives input text from the field of arch dam design and maps each word or character therein to a distributed vector representation rich in semantic information, thereby transforming the original text into a numerical feature sequence that can be processed by a machine. The feature partitioning module, connected to the text representation module, is used for deep analysis of the numerical feature sequence. This module dynamically divides the feature sequence into feature components focused on entity recognition and feature components focused on relation extraction. During the partitioning process, this module combines the features of the current word with the historical feature information of the preceding words, enabling the feature representation of each word to perceive its local context, thereby providing a more discriminative feature foundation for subsequent tasks. The feature fusion module, connected to the feature partitioning module, is used to perform deep interaction and fusion of entity-related features and relationship-related features obtained after partitioning at a global level. This module leverages the powerful nonlinear mapping capabilities of neural networks to capture complex, latent dependencies between features, enhancing the model's understanding of deep text semantics and obtaining more global and robust fused feature representations of entities and relationships. The feature classification module, connected to the feature fusion module, is used to predict the final knowledge units based on the fused features. This module includes an entity classification layer and a relation classification layer. The entity classification layer is responsible for identifying the boundaries (start and end positions) of entities in the text and determining their types (e.g., "dam material", "dam height"); the relation classification layer is responsible for identifying the semantic relationships between different entities and determining their types (e.g., "height", "number of layers", "has attributes"). Through the collaborative work of this module, the model can synchronously extract structured knowledge units from the input text. Finally, by organizing the extracted entities and relationships into predefined structures (such as triples and quintuples), the knowledge graph for arch dam design can be automatically constructed or expanded.

[0041] In another illustrative embodiment, see Figure 5 The design parameter extraction process includes: (1) A semantic retrieval algorithm based on an embedding model is used to query and determine the list of parameters necessary for arch dam design from a pre-built arch dam design knowledge base, and this list is used as a criterion for verifying the completeness of parameters.

[0042] (2) Use a large language model to parse the initial design documents of the arch dam provided by the user, or directly receive the structured data input by the user and automatically extract the preliminary design parameter set from it.

[0043] (3) Compare the design parameters extracted in step (2) with the list of required parameters determined in step (1) to check for any missing items.

[0044] (4) Based on the comparison results of step (3), determine whether the currently extracted design parameters are complete. If complete, the process proceeds to step (5). If the parameters are incomplete, the process proceeds to the loop supplementation step, that is, generating specific prompt information to clearly inform the user of the missing necessary parameters and requesting the user to supplement them. The system receives the information supplemented by the user according to the prompt (which may be supplemented files or directly input data descriptions), and automatically extracts the design parameters from them again using the large language model. The newly extracted parameters will be fed back and merged into the parameter set of step (2). Return to step (3) and re-perform parameter completeness verification, thereby forming a closed-loop self-checking mechanism of "verification-prompt-supplementation-re-verification" until all necessary parameters are completely extracted, then the process proceeds to step (5).

[0045] (5) Output a complete set of design parameters for use in the subsequent arch dam shape generation and mechanical calculation module.

[0046] In one illustrative embodiment, see Figure 6 The arch dam modeling tool used for arch dam shape design includes four core steps. Each step achieves accurate modeling through parametric operations and geometric logic. The specific process is as follows: Step 1: Generate arch points on the characteristic elevation of the arch ring through parametric description equations, connect them and stack them in space, and further transform them into smooth curved surface entities through curve lofting technology to form the basic arc outline of the arch dam (smooth dam entity); establish the flood discharge orifice cutting auxiliary block by positioning the "orifice cutting center", and combine rotational cutting and depth parameter control to complete the geometric modeling of functional openings such as surface holes and deep holes in sequence (opening model); according to the design parameters, divide the dam body into independent dam sections through transverse joint rotational cutting operation to form a joint model to ensure that the structural segmentation meets the engineering specifications; Step 2: Using the topographic map as input, extract the topographic lines as the boundary constraints for slope modeling, simplify the topographic lines, retain key undulation features, and generate a slope surface that matches the terrain through surface fitting technology, so that the slope shape conforms to the actual geological conditions, laying the foundation for the subsequent connection between the dam body and the terrain. Step 3: Based on the spatial relationship between the bottom foundation surface outline of the dam body and the terrain, determine the excavation area. Use the excavation line connection operation to geometrically integrate the dam body with the terrain to ensure the rationality of the dam body's embedding into the terrain. To further ensure that the model boundary effect meets the requirements of finite element analysis, extend the left and right banks and upstream and downstream directions to form a complete connection structure including the slope body and the riverbed dam foundation, thereby enhancing the spatial continuity of the model. Step 4: Integrate elements such as the dam body, left and right bank topography, upstream and downstream reservoir water areas and dam foundation to form an overall model containing all engineering components, clarify the spatial relationship of each part, and further use APDL script to perform structured meshing of the overall model, transforming the continuous geometric model into discrete units that can be used for numerical calculations, and use them to obtain data samples for subsequent finite element stress static and dynamic calculations under different working conditions. Step 5: Call the finite element analysis tool for arch dams to perform batch finite element calculations on the three-dimensional models of arch dams under different shape parameters. Obtain the calculation results of concrete volume, dam stress, displacement, tensile stress zone volume and stability safety index of different dam sections, and form a sample dataset composed of shape design parameters, material zoning parameters and finite element response results.

[0047] Once the sample size meets the training requirements for the surrogate model, the decision-making large model is used to call the surrogate model building tool, import the training set data, and conduct surrogate model training based on the mapping relationship between input parameters and finite element output response. This establishes a nonlinear mapping relationship between the arch dam shape parameters and the finite element output results, generating the surrogate model. The surrogate model is then imported into the test set data, and performance prediction is performed on the test set samples. The prediction results are compared with the finite element true values ​​to verify the accuracy of the surrogate model. The surrogate model that meets the accuracy requirements is used as a fast evaluator for subsequent shape optimization.

[0048] Step 6: Using the constructed complete model with dam foundation and arch dam as the initial shape, call the arch dam shape optimization tool, import the proxy model, and based on the optimization algorithm, take the minimum total cost of each grade of concrete in different dam sections as the economic objective. The calculation method is to multiply the volume of each grade of concrete section by the corresponding cost unit price; at the same time, take the volume of the tensile stress zone of the dam as the structural safety objective, and generate arch dam design candidate schemes.

[0049] The proxy model is responsible for quickly evaluating the performance of the schemes, while verifying whether each scheme meets the specification constraints (such as stress and stability constraints), and outputting the scheme performance evaluation results and constraint satisfaction status; The arch dam shape optimization tool uses Pareto dominance relations to screen non-dominated solutions and update the population, progressively pushing candidate solutions toward the dual optimization objectives, and the Pareto front solution set gradually stabilizes. The surrogate model is continuously evaluated during the iteration process. When the evaluation results meet the convergence accuracy requirements of the optimization objective or the number of iterations reaches a preset maximum value, such as 100 iterations, the optimization is considered to have basically converged, and the iteration stops. The optimization tool outputs the Pareto front non-dominated solution set of the arch dam shape. Each non-dominated solution corresponds to a candidate shape scheme, all of which meet the specification constraints and achieve the optimal trade-off between the total cost of concrete of different grades in different dam sections and the volume of the tensile stress zone of the arch dam. The best arch dam shape corresponding to the Pareto front is obtained as the final design result.

[0050] In one illustrative embodiment, see Figure 7 Furthermore, a computing device is provided for implementing the above-described agent-based intelligent design method for arch dams. The device includes one or more processors and a memory coupled to the processors; the memory stores instructions configured to be executed by the processors, which, when executed, cause the computing device to perform the steps described in any of the preceding method embodiments.

[0051] This invention also provides a computer-readable storage medium. This medium is substantially non-transitory and contains computer-executable instructions encoded thereon. When these instructions are loaded and executed on one or more computing devices, they can drive the computing devices to complete the described agent-based intelligent arch dam digital design method.

[0052] Regarding storage and processing resources: In the context of the various embodiments of this invention, the term "memory" is intended to encompass at least one type of volatile memory and non-volatile memory. Exemplary, and not exhaustive, volatile memory may include dynamic random access memory (DRAM) and static random access memory (SRAM), etc.; non-volatile memory may include flash memory, read-only memory (ROM), phase-change memory (PCM), and storage devices based on emerging storage technologies (such as resistive random access memory, RRAM), etc. The processors involved may include, but are not limited to, hardware cores with computing capabilities such as central processing units (CPUs), graphics processing units (GPUs), tensor processing units (TPUs), application-specific integrated circuits (ASICs), or field-programmable gate arrays (FPGAs). The database systems involved may include relational databases (such as SQL databases) or non-relational databases (such as Neo4j databases, graph databases), and may be deployed in a distributed or centralized architecture.

[0053] Those skilled in the art will understand that the entirety or part of the method flow described in the embodiments of the present invention can be implemented by relevant hardware components of a computing device through program instructions. Such programs can be stored in the aforementioned computer-readable storage medium. Specific forms of storage media include, but are not limited to: USB flash drives, portable hard drives, optical discs, server storage space, or distributed network storage nodes, etc.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do 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 digital design method for arch dams based on intelligent agents, characterized in that, include: Obtain the initial design documents for the arch dam; The design parameter extraction tool is invoked to extract the list of necessary design parameters for the arch dam design from the pre-built arch dam design knowledge base, and to extract the list of initial design parameters from the initial arch dam design file. The initial design parameter list is compared with the necessary design parameter list; if any are missing, they are added until all parameters are checked and a complete design parameter list is output. The arch dam modeling tool is invoked to perform parametric geometric modeling based on the complete list of design parameters, generating a three-dimensional geometric model of the arch dam. Based on the three-dimensional geometric model of the arch dam and the complete list of design parameters, the performance of the arch dam structure is calculated. The performance calculation results are then input into a proxy model building tool to build a proxy model. The arch dam shape optimization tool was invoked to perform dual-objective iterative optimization of the arch dam shape, with the optimization objectives being to minimize the total cost of each grade of concrete in different zones of the dam and to minimize the volume of the tensile stress zone of the arch dam. During the iteration process, the candidate schemes generated by the arch dam shape optimization tool are evaluated using the surrogate model; when the evaluation results meet the preset conditions, the iteration optimization stops and the optimized Pareto front shape scheme is output.

2. The intelligent design method for arch dams based on intelligent agents according to claim 1, characterized in that, The construction of the intelligent agent includes: A large decision-making model is constructed based on a large language model. This model is used to invoke and execute tools such as knowledge base construction tools, design parameter extraction tools, arch dam modeling tools, finite element analysis tools, surrogate model construction tools, and arch dam shape optimization tools. The knowledge base building tool is used to build the arch dam design knowledge base; The design parameter extraction tool is used to extract parameters from the arch dam design knowledge base and the initial design file; The arch dam modeling tool is used to construct a three-dimensional geometric model of the arch dam; The finite element analysis tool is used to perform precise calculations of the performance of the arch dam structure and output the calculation results of the arch dam performance. The proxy model building tool is used for proxy model training and accuracy verification; The arch dam shape optimization tool is used to iteratively optimize the performance data of the arch dam and generate candidate design scheme data.

3. The method according to claim 1, characterized in that, The pre-construction of the arch dam design knowledge base includes: Acquire multi-source text data in the field of arch dam design and process it to obtain an arch dam design dataset; Based on the arch dam design dataset, a knowledge base for arch dam design is constructed using a knowledge base construction tool, which includes a knowledge graph construction sub-tool and a vector knowledge base construction sub-tool. The knowledge graph construction sub-tool first divides the dataset text into paragraphs, and then performs entity extraction and relation extraction on each paragraph one by one based on deep learning algorithms to obtain the arch dam design knowledge graph; The vector knowledge base construction sub-tool dynamically generates semantically complete paragraphs based on paragraph end symbols and character length limitations during the text segmentation stage, and assigns a number to each paragraph; subsequently, it uses a text embedding model to map these paragraphs to a semantic vector space to obtain a vector database.

4. The intelligent design method for arch dams based on intelligent agents according to claim 1, characterized in that, The steps of calling the design parameter extraction tool include: The design parameter extraction tool is invoked to extract a list of necessary design parameters for arch dam design from the arch dam design knowledge base based on text semantic retrieval functionality; the design parameter extraction tool is constructed based on a semantic retrieval algorithm. The design parameter extraction tool is invoked to extract the initial design parameter list from the initial design document of the arch dam.

5. The intelligent design method for arch dams based on intelligent agents according to claim 1, characterized in that, The steps of the control test include: The initial design parameter list is checked against the necessary design parameter list. If any of the uploaded initial design parameters are missing, a prompt message is generated and the missing design parameters are fed back. After they are supplemented, the check is performed again. The check steps are repeated until a complete design parameter list is obtained. The complete design parameter list includes shape-related design parameters and load-related design parameters.

6. The intelligent design method for arch dams based on intelligent agents according to claim 1, characterized in that, The steps for calling the arch dam modeling tool include: Based on the complete list of design parameters, the preset arch dam modeling code is called, and the parametric design method is used to realize the digital expression of the arch crown beam section and the horizontal arch ring respectively. The simulated terrain line is extracted through the terrain plan view to construct the dam foundation slope excavation outline. The arch dam modeling tool is invoked to automatically generate a complete geometric model of the arch dam, including the dam foundation.

7. The intelligent design method for arch dams based on intelligent agents according to claim 1, characterized in that, The iterative optimization steps include: The arch dam shape optimization tool is invoked, and combined with the evaluation capability of the proxy model, the arch dam shape design parameters are iteratively optimized with the optimization objectives of minimizing the total cost of each grade of concrete in different dam sections and minimizing the volume of the tensile stress zone of the arch dam, and candidate arch dam shape schemes are output. In each iteration, the surrogate model evaluates the performance of the candidate arch dam shapes and feeds the evaluation results back to the arch dam shape optimization tool. The arch dam shape optimization tool updates the population or candidate solution set based on the optimization algorithm and Pareto dominance relationship, selects a better solution to continue iterating until the convergence accuracy requirement of the optimization objective is met or the number of iterations reaches the preset maximum value, at which point the iterative optimization stops and the Pareto front shape solution is obtained.

8. A digital design device for arch dams based on intelligent agents, characterized in that, include: The first processing module is used to obtain the initial design files for the arch dam. The second processing module is used to call the design parameter extraction tool to extract the list of necessary design parameters for the arch dam design from the pre-built arch dam design knowledge base, and to extract the list of initial design parameters from the arch dam design initial file. The third processing module is used to compare and check the initial design parameter list with the necessary design parameter list; if there are any missing parameters, they are added until all parameters are checked and a complete design parameter list is output. The fourth processing module calls the arch dam modeling tool to perform parametric geometric modeling based on the complete design parameter list, generating a three-dimensional geometric model of the arch dam. The fifth processing module is used to perform performance calculations of the arch dam structure based on the three-dimensional geometric model and complete design parameter list of the arch dam, and input the performance calculation results into the proxy model building tool to build a proxy model; The arch dam shape optimization tool was invoked to perform dual-objective iterative optimization of the arch dam shape, with the optimization objectives being to minimize the total cost of each grade of concrete in different zones of the dam and to minimize the volume of the tensile stress zone of the arch dam. During the iteration process, the candidate schemes generated by the arch dam shape optimization tool are evaluated using the surrogate model; when the evaluation results meet the preset conditions, the iteration optimization stops and the optimized Pareto front shape scheme is output.

9. An electronic device, characterized in that, The device includes a memory and a processor, wherein the memory stores program instructions that are executed by the processor, and the processor invokes the program instructions to execute the agent-based digital design method and apparatus for arch dams as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, include: A computer program instruction, when executed on a digital design apparatus, causes the digital design apparatus to perform the method as described in any one of claims 1 to 7.

Images