Water conservancy project BIM cost AI intelligent agent-based method and system

Abstract

The invention discloses a construction cost AI intelligent agent method and system based on water conservancy project BIM. The method comprises the steps of building a water conservancy construction cost knowledge base, an interface layer, an Agent layer, an information acquisition layer, a model management layer and a functional layer, and performing result application and intelligent analysis. Water conservancy project BI M cost software components and workflow description standards are processed through a large language model, autonomous perception, thinking and action of an intelligent agent are achieved, and water conservancy project cost compilation tasks can be executed through understanding, learning and reasoning. The user interacts with the AI assistant in a natural language or character import mode, and the AI assistant autonomously calls a tool to operate, so that the compliance and accuracy of the result are ensured, and investment comparative analysis and the like can be carried out. According to the method, the problems of low efficiency, error proneness, lack of intelligent analysis and the like in the prior art are solved, and the accuracy and efficiency of hydraulic engineering cost compilation are improved.

Description

Technical Field

[0001] This application relates to the technical field of artificial intelligence for water conservancy project cost estimation, and in particular to a method and system based on AI intelligent agent for water conservancy project BIM cost estimation. Background Technology

[0002] Currently, water conservancy project cost estimation software is still an information technology software product, and its core processes have significant shortcomings: quantity calculations mainly rely on manual calculations combined with CAD software assistance; basic unit price calculations, quota pricing, and cost document report preparation, although supplemented by information technology tools, are still primarily done manually. This model leads to repeated calculations and report preparation during the cost application, approval, and auditing stages of projects, not only consuming a significant amount of time and energy for cost engineers but also making the preparation work cumbersome, with a low tolerance for error, and prone to mistakes. More importantly, existing technology cannot achieve intelligent analysis of results, such as investment comparison or risk assessment, thus limiting work efficiency and decision support capabilities.

[0003] To address the aforementioned issues, the applicant of this invention provides an AI-powered intelligent agent method and system for compiling cost estimates for water conservancy projects using BIM+AI technology. This method aims to solve the problems of "precision, accuracy, and speed" in the quantity calculation stage using BIM technology. It learns relevant information through a large language model to obtain a learned large language model, and generates a business expression chain for cost document report compilation based on fine-tuning training, enabling business applications. Users can communicate their needs with the AI ​​assistant through natural language or text import. After understanding the intent, the AI ​​assistant autonomously calls upon BIM cost estimation software components and tools, ensuring compliance and accuracy of the results while solving the problems of "precision, accuracy, and speed" in the stages of quota application and pricing calculation, and the integration of various cost document reports. Ultimately, this invention provides an efficient technical solution for improving the efficiency of water conservancy project cost estimation and reducing the time consumed in project investment application, approval, settlement, and auditing. Summary of the Invention

[0004] This application aims to at least partially address one of the technical problems in the related art.

[0005] To achieve the above objectives, the first aspect of this application proposes an AI-based intelligent agent method for BIM cost estimation in water conservancy projects. This method involves building a water conservancy cost knowledge base, collecting regulations and quotas for water conservancy project cost estimation based on the design characteristics of BIM cost estimation software components, and establishing workflow description standards. A large language model is used to register the unified access addresses of the BIM cost estimation software components and workflow description standards, generating software component registry entries and workflow description registry entries, and managing the components and standards. The large language model performs semantic governance on the input, output, and application functions of the registered software components and workflow description standards. Through the collaborative work of the BIM cost estimation software components and workflow description standards, the intelligent agent can perceive the environment, collect information, make decisions, and execute tasks within the authorized scope.

[0006] This application aims to address the problems of low efficiency, error-proneness, and lack of intelligent analysis in the existing process of compiling water conservancy project costs. It provides a method and system based on AI intelligent agent for water conservancy project cost compilation, which can achieve accuracy, efficiency, and intelligence in water conservancy project cost compilation, thereby improving work efficiency and decision support capabilities.

[0007] In addition, the AI ​​intelligent agent method and system for BIM cost estimation of water conservancy projects proposed above in this application may also have the following additional technical features:

[0008] In one embodiment of this application, the method further includes: building an interface layer to allow the large language model to learn the user's operation process and habits of operating BIM cost estimation software components through the interface layer, understand the relationship between software components and workflow description standards, and master the compilation rules for water conservancy project cost estimation; and storing the learning records and effect evaluations in a knowledge memory base.

[0009] In one embodiment of this application, the method further includes: establishing an Agent layer to accept user instructions on setting knowledge base attributes, roles, Agent collaboration, security, permissions, and personalization, enabling the Agent to execute tasks within the authorized scope according to user instructions; wherein, the setting instructions include matching project name, project number, project library name, construction nature, engineering category, and report category; matching project location, price level, and engineering elevation; role settings include compiling unit, reviewing unit, approving unit, consulting unit, and auditing unit; security and authorization settings; and personalization settings.

[0010] In one embodiment of this application, the method further includes: establishing an information collection layer; developing and building a project division coding database and a project name item information collection prompt word template library; establishing a persistent, stateful, and controllable Agent information collection workflow prompt word template library; through the collaborative work of water conservancy engineering BIM cost software components and workflow description standards, performing project division coding matching, project name item selection, and engineering quantity calculation information collection, and generating an engineering quantity calculation list report; based on the information collection workflow prompt words, the user inputs the project name and project main feature context constraints in text or natural language, and infers and generates a project name item description chain and a quota adjustment description chain; based on the collected information, generating an engineering quantity calculation result report, a basic unit price calculation table, and a project cost compilation basic information table, and storing the results in the project database; specifically, the engineering quantity calculation result collection content includes project division coding, sub-item project name, project main features, measured engineering quantity, component name, component ID, engineering location, construction method, main materials, design parameter indicators, and other cost attribute information.

[0011] In one embodiment of this application, the method further includes: building a model management layer and using the QLoRA method to fine-tune the large language model, including: fine-tuning training for quota pricing calculation: based on the project name item description chain and quota adjustment description chain generated by reasoning, a neural network mind map algorithm is used to generate a quota pricing calculation application description chain; fine-tuning training for compilation instructions generation: based on the basic cost compilation information, cost compilation rules and project cost characteristics stored in the project database, compilation instructions for engineering quantity calculation, investment estimation, investment estimate, design estimate, engineering budget, bidding engineering quantity list and tender engineering quantity list are generated; fine-tuning training for cost report generation: based on the cost compilation rules and project cost characteristics, investment estimate report, investment estimate report, design estimate report, engineering budget report, bidding engineering quantity list report and tender engineering quantity list report are generated; a prompt word template is constructed using the cost compilation rules as context information for the large language model to obtain a learned large language model; wherein, the fine-tuning training utilizes quantization and low-rank adaptation techniques to reduce computing resources and memory usage.

[0012] In one embodiment of this application, the method further includes: building a functional layer to enable the large language model to recognize user intent and realize the business application of water conservancy project cost estimation based on BIM technology; specifically including: the function of quota pricing calculation: selecting the quota pricing calculation workflow description in the workflow description registry, the large language model calls the component according to the instruction to generate the quota pricing calculation application expression chain through the expression chain provided by the information layer. This expression chain processes the component connection relationship, input and output parameters, data source and data type, generates the building engineering unit price table and the installation engineering unit price table, and stores them in the project database; the function of compiling instructions generation: selecting the compiling instructions workflow description in the workflow registry, the large language model calls the component according to the instruction to generate the compiling instructions generation expression chain, generates the compiling instructions table, and stores it in the project database; the function of cost document report compilation: selecting the cost document report compilation generation workflow description in the workflow registry, the large language model calls the component according to the instruction to generate the cost document report compilation expression chain, generates business application documents or reports for each stage such as water conservancy project estimation, estimation, preliminary estimate, budget, and bill of quantities, and stores them in the project database.

[0013] In one embodiment of this application, the workflow description standard includes: setting instructions including creating a new project, opening a project, setting rates, collecting component attribute information, and calculating basic unit prices; business functions including creating project information, generating a quantity calculation report, compiling a basic unit price calculation table, applying quota pricing calculations, compiling a preliminary design budget report, compiling a feasibility study investment estimate report, compiling an engineering planning investment estimate report, compiling an engineering budget report, compiling a bidding quantity list report, compiling a tender quantity list, compiling an engineering settlement report, and compiling an engineering settlement audit report; and output export including exporting the quantity calculation report, exporting basic unit price calculation results, exporting the building engineering calculation table, exporting the installation engineering unit price calculation table, exporting the preliminary design budget report, exporting the feasibility study investment estimate report, exporting the engineering planning investment estimate report, exporting the engineering budget report, exporting the bidding quantity list report, exporting the tender quantity list, exporting the engineering settlement report, and exporting the engineering settlement audit report.

[0014] In one embodiment of this application, the application further includes: application of results and intelligent analysis of results: generating business applications including basic unit price calculation tables, engineering quantity calculation result reports, compilation instructions, engineering budget summary tables, engineering partial budget tables, budget appendices, preliminary design budget reports, feasibility study investment estimation reports, engineering planning investment estimate reports, engineering budget reports, bidding engineering quantity list reports, tendered engineering quantity list reports, engineering settlement reports, and engineering settlement audit reports; intelligent analysis of results includes investment comparison analysis, cost calculation and estimation, price query, progress management, and risk assessment; users interact with application requirements through natural language or text import methods, the AI ​​assistant understands the user's intent, autonomously calls BIM cost estimation software component tools to operate, realizes business function applications, and ensures compliance and accuracy of results.

[0015] A computer device includes a memory, a processor, and a computer program stored in the memory, wherein when the processor executes the computer program, it implements the steps of the method.

[0016] A computer-readable storage medium having a computer program stored thereon, which, when loaded by a processor, is capable of executing the steps of the method.

[0017] The advantages of this application compared to existing technologies are:

[0018] (1) This invention enables automated processing of each stage of water conservancy project cost preparation by autonomously calling BIM cost software component tools through AI intelligent agent, reducing manual intervention, solving the problem of repetitive calculation and report preparation in the traditional mode, greatly saving the time and energy of cost engineers, and improving work efficiency.

[0019] (2) Fine-tuning training is carried out using a large language model, and operations such as applying quota pricing are performed in combination with cost preparation rules and quota items. At the same time, the compliance and accuracy of the results are ensured, the error rate caused by human operation errors is reduced, and the accuracy of cost preparation results is improved.

[0020] (3) It can perform intelligent analysis of results such as investment comparison analysis, cost calculation and estimation, price inquiry, schedule management and risk assessment, which provides strong support for project decision-making and makes up for the shortcomings of existing technologies that cannot achieve intelligent analysis of results.

[0021] (4) By building modules at various levels, it has achieved flexible adaptation to different project information, role settings, etc., which can meet the needs of water conservancy project cost preparation at different stages and in different scenarios, and has strong adaptability and scalability.

[0022] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0023] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0024] Figure 1 This is a schematic diagram of the overall process of a method and system for BIM cost estimation of water conservancy projects according to an embodiment of this application;

[0025] Figure 2 This is a schematic diagram illustrating the basic architecture construction process of a water conservancy project BIM cost AI intelligent agent method and system according to an embodiment of this application;

[0026] Figure 3 This is a schematic diagram illustrating the Agent configuration and information collection process of an AI intelligent agent method and system for BIM cost estimation in water conservancy projects according to an embodiment of this application;

[0027] Figure 4 This is a schematic diagram illustrating the model training and function implementation process of a BIM cost AI intelligent agent method and system for water conservancy projects according to an embodiment of this application;

[0028] Figure 5 This is a schematic diagram illustrating the application and interaction process of a method and system for BIM cost estimation in water conservancy projects, according to an embodiment of this application. Detailed Implementation

[0029] Embodiments of this application are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. Rather, embodiments of this application include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.

[0030] The following describes a method and system for BIM cost estimation of water conservancy projects based on AI intelligent agents, in conjunction with the accompanying drawings.

[0031] like Figures 1-5As shown in this application, an embodiment of the present invention discloses a method and system for BIM-based cost estimation in water conservancy projects. The core of this invention is to construct a water conservancy project cost estimation system that integrates BIM technology and an AI intelligent agent. By building a multi-level architecture (water conservancy cost knowledge base → interface layer → agent layer → information collection layer → model management layer → functional layer → result application and intelligent analysis), the intelligent agent can autonomously perceive, make decisions, and execute within the water conservancy project cost estimation environment. Each level works collaboratively, and through semantic understanding, fine-tuning training, and component invocation of a large language model, the entire cost estimation process is ultimately automated and intelligent, ensuring the compliance and accuracy of the results.

[0032] Example: I. The specific implementation steps for each level are as follows:

[0033] (I) Establishing a knowledge base for water conservancy cost

[0034] 1. Data collection and standardization

[0035] In response to the design features of BIM cost estimation software components for water conservancy projects (such as concrete component calculation components and earthwork measurement components), the system first collects currently valid regulations for compiling water conservancy project costs (such as the "Regulations for Compiling Water Conservancy Project Design Estimates"), industry quotas (such as the "Water Conservancy Construction Project Estimate Quotas" and the "Water Conservancy Project Construction Machinery Hourly Rate Quotas") and local supplementary quotas to establish a basic data pool.

[0036] Example: Collect the "Materials Budget Price Adjustment Method" issued by the Water Resources Department of a certain province as the basis for calculating the basic unit price.

[0037] 2. Establishment of workflow description standards

[0038] Based on the entire cost estimation process, a standardized workflow description standard is established, clarifying the operational specifications, input and output requirements, and connection logic of each step. Specifically, this includes:

[0039] Basic operations: Create a new project (requires entering project number, construction type, and other mandatory fields), open a project (supports historical project retrieval), set a fee rate (associates with the fee standard of the project's location);

[0040] Data processing: collecting component attribute information (extracting component dimensions, materials, etc. from the BIM model) and calculating basic unit prices (including calculation rules for labor, material, and machinery unit prices);

[0041] Report preparation: Generating engineering quantity calculation results reports (specifying report formats), preparing preliminary design budget reports (clarifying chapter composition);

[0042] Exportable results: Export the bill of quantities report for tendered works (supports Excel and PDF formats), and export the project settlement audit report (with electronic signature interface).

[0043] 3. Component and workflow registration and management

[0044] Using a large language model (such as a customized model based on the GPT-4 architecture), unified access address registration is performed for water conservancy engineering BIM cost estimation software components (such as automatic quantity calculation components and quota matching components) and the aforementioned workflow description standards:

[0045] Assign a unique identifier to each component (e.g., “GCL-001” represents the concrete quantity calculation component), and record its interface address, input parameters (e.g., component volume, strength grade), and output format (e.g., quantity value).

[0046] The workflow description standard is structured and coded (e.g., "WF-005" represents the workflow for calculating the unit price). A workflow description registry is generated to store the preconditions (e.g., "the quantity of work needs to be calculated"), execution steps, and post-processing results (e.g., "generating a unit price table") for each workflow.

[0047] The system manages the lifecycle (addition, update, and deactivation) of components and workflows through a unified management module.

[0048] 4. Semantic Governance

[0049] The large language model performs semantic processing on registered components and workflows to ensure that the inputs and outputs of each element are consistent at the semantic level. For example:

[0050] Standardize the description of "Project Name (Main Features of Project)", such as "Normal Concrete Dam (Main Dam, C30F50W6 (Secondary Mixture), Pouring Thickness ≤ 1.5m, Mechanized Pouring". The "Project Name and Main Features of Project" are provided with unified descriptive semantics by the "Project Name List Information Collection Prompt Word Template Library". The "Project Name" is semantically processed according to the cost preparation regulations, and the "Main Features of Project" is semantically processed according to the quota regulations.

[0051] Clarify the semantic boundaries of "quantity calculation" (distinguish the calculation logic between "design quantity" and "measured quantity").

[0052] 5. Basic Construction of Intelligent Agent Collaboration

[0053] By associating components with workflows (e.g., associating “quota pricing workflow” with “quantity calculation component” and “quota matching component”), the agent can perceive the current compilation environment (e.g., in the “preliminary design stage”) and trigger information collection (e.g., calling “component attribute collection component”), decision-making (e.g., choosing “water conservancy construction quota” or “water conservancy installation quota”) and task execution (e.g., automatically filling the unit price table) based on the workflow.

[0054] (II) Building the Interface Layer

[0055] 1. User operation learning mechanism

[0056] The interface layer uses an API interface to interact with the water conservancy project BIM cost estimation software, and records the user's operation trajectory in real time.

[0057] Operation process: When users compile basic unit prices, they usually first enter the "material name" → select the "transportation method" → calculate the "freight and miscellaneous charges". The system records this process as the default recommended path.

[0058] User habits: For example, if a design institute user prefers to "divide the list by engineering part", the system will automatically adjust the item sorting rules after learning the user's preferences.

[0059] 2. Rule Acquisition and Storage

[0060] The large language model parses the relationship between software components and workflows through the interface layer (e.g., "compiling a feasibility study investment estimation report" requires calling the "investment estimation index component"), and learns cost preparation rules by combining historical project data (e.g., "basic contingency fee for hub projects is calculated at 5%).

[0061] Learning outcomes (such as operation logs and rule weight values) are stored in a knowledge memory (using a MySQL database, with table structures containing fields for "operation ID, component name, rule description, and number of applications"), for subsequent model optimization.

[0062] (III) Building the Agent Layer

[0063] 1. User command reception and parsing

[0064] The Agent layer receives system configuration commands from users through a graphical interface, specifically including:

[0065] Project attribute matching: When the user enters "Project Name: XX Reservoir Reinforcement Project" and "Project Location: XX Province XX City", the system will automatically match the local price level (such as the material price index of XX City in May 2024) and the project elevation (which affects the adjustment coefficient of labor and machinery efficiency);

[0066] Role settings: After a user selects the "Compilation Unit" role, the system grants permissions for "Preliminary Budget Compilation", "Basic Budget Compilation", "Estimated Budget Compilation", and "Results Export", while restricting the "Approval Adjustment" and "Audit Adjustment" functions;

[0067] Security and Authorization: Connect to the enterprise authentication system via LDAP protocol and set three levels of permissions: "read-only users", "editing users", and "approval users". For example, construction unit users can only view the list data of their own projects.

[0068] Personalized settings: Supports users to customize report headers (such as adding a "Bidding Unit Remarks" column) and set up commonly used quota libraries (such as loading "Water Conservancy and Hydropower Engineering Budget Quotas" by default).

[0069] 2. Instruction Execution and Access Control

[0070] The Agent layer converts the parsed instructions into system-executable parameters (such as converting the "Audit Unit" role into the permission code "SQ-003"), ensuring that the agent executes tasks within the authorized scope (such as the audit unit user can only call the "Settlement Comparison Component" and cannot modify the original project quantity).

[0071] (iv) Establishing an information collection layer

[0072] 1. Database setup

[0073] The project classification and coding database is stored in a tree structure. The root node is "Building Engineering, Mechanical and Electrical Equipment and Installation Engineering, Metal Structure Equipment and Installation Engineering". For example, the sub-node of "Building Engineering" is divided into 6 levels: "Building Engineering → Key Project → Water Retaining Project → Dam Category → Category Project Name → Sub-item Project Name". Each node corresponds to a unique code (e.g., the code for "Key Project → Water Retaining Project → Concrete Dam (Sluice) Project → Earthwork Excavation → General Earthwork Excavation" is "10-10.10.10.10.01").

[0074] Project Name Item Information Collection Prompt Template Library: Stores standardized prompts to guide user input (e.g., "Please describe the project name and main characteristics," or "General earthwork excavation (dam foundation, Class III soil, 2m)"). 3 Excavator, 20t dump truck, transport distance 3.5km)”), supports natural language understanding.

[0075] 2. Information Collection Process

[0076] Collaborative data collection: Through the collaboration of the "Quantity Calculation Component" and the "Project Division and Coding Workflow," the following tasks are completed:

[0077] ① Project classification and coding matching (e.g., "General earthwork excavation of dam foundation" matches the code "10-10.10.10.10.01");

[0078] ② Select the project name (e.g., based on the feature of "arc gate", automatically select "arc steel gate fabrication and installation");

[0079] ③ Collection of engineering quantity calculation information (extract “gate size 3m×5m” from BIM model, calculate weight and associate measurement rules).

[0080] User Input and Reasoning: For example, if a user inputs "Project Name: Masonry Slope Protection, Main Features: Curved Surface, M7.5 Mortar", the system will combine the prompts to generate the following reasoning:

[0081] Project Name List Description Chain (“Masonry Slope Protection → Curved Surface → M7.5 Mortar”);

[0082] Quota adjustment statement chain ("According to Article 3.2 of XX quota, the quota labor adjustment coefficient is 1.05").

[0083] The quantities of work are calculated, collected, and statistically analyzed by the BIM cost estimation software components by selecting the component IDs of the BIM model.

[0084] 3. Results Generation and Storage

[0085] Three types of core tables are generated based on the collected information:

[0086] The engineering quantity calculation results table includes project classification code, project name, component name, component ID, measured engineering quantity, etc., such as "10-10.10.10.10.01, general earthwork excavation, dam foundation excavation, component IDXX, 10056m". 3 ”);

[0087] Basic unit price calculation sheet (such as labor budget unit price, material budget unit price, machine hourly unit price, ventilation, water and electricity budget unit price, etc.);

[0088] Basic information table for project cost estimation (including project overview, basis for estimation, applicable quotas, price level, etc.).

[0089] All deliverables are stored in the project database (using PostgreSQL, which supports associated storage of BIM model files).

[0090] (V) Building the Model Management Layer

[0091] 1. Fine-tuning training of large language models

[0092] The QLoRA (Quantized Low-Rank Adaptation) method is used to fine-tune large language models (such as Llama2-70B) to reduce computational resource consumption. Specifically, this includes:

[0093] Fine-tuning of quota pricing calculation: Input "item description chain + quota adjustment statement chain", combine neural network map algorithm (model sub-projects and quota items as nodes, feature similarity as edge weights), train model to match the optimal quota (e.g. "mortar masonry slope protection" matches "quota number 030015" and automatically applies adjustment coefficient);

[0094] Compilation instructions for fine-tuning: Using "project overview + fee standard + and project cost characteristics" from the project database as input, the model is trained to generate the compilation instructions for the specifications (such as "the unit price of labor budget for this project shall be implemented in accordance with the 2024 standard of XX Province").

[0095] Cost report generation fine-tuning: Input “Bill of Quantities + Basic Unit Price + Fee Rate”, train the model to generate a report according to the chapter structure (e.g., “Chapter 1 Project Overview → Chapter 2 Key Investment Indicators → Chapter 3 Budget Table”).

[0096] 2. Construction of the prompt word template library

[0097] A project name list prompt template library was constructed based on cost estimation rules and quota regulations.

[0098] Templates are used as contextual information input to the model to improve output accuracy.

[0099] (vi) Building the functional layer

[0100] 1. User Intent Recognition and Function Mapping

[0101] The large language model parses user requirements through natural language processing. For example, if a user inputs "generate unit price table for dam concrete engineering", the system recognizes it as "call the quota pricing function" and matches the corresponding workflow description.

[0102] 2. Implementation of core functions

[0103] Calculation of quota pricing: In the workflow description registry, call the "Quota pricing workflow" to generate the business function description chain "call concrete engineering quantity component → match quota item → input basic unit price information → calculate comprehensive unit price", and finally generate the "Construction Engineering Unit Price Table" (including details of labor cost, material cost and machinery cost);

[0104] Compilation instructions generation: The "Compilation Instructions Workflow" is invoked, and the description chain guides the system to extract information such as "compilation basis and basic unit price calculation method" from the project database to generate the "Compilation Instructions Table";

[0105] Cost document report preparation: The "Cost Report Preparation Workflow" is invoked, and the data is integrated according to the "Cover → Table of Contents → Main Text → Appendices" structure to generate documents such as the "Preliminary Design Estimate Report" and automatically stored in the project database.

[0106] (VII) Application of Results and Intelligent Analysis

[0107] 1. Business Application Generation

[0108] Based on the output of the functional layer, the system generates end-to-end business application results, including:

[0109] Basic data: Labor cost unit price list, construction machinery hourly rate unit price list;

[0110] Intermediate deliverables: Quantity calculation report, partial cost estimate;

[0111] Final report types: Feasibility study investment estimation report, preliminary design investment estimate report, and tender bill of quantities report (with electronic signature).

[0112] 2. Intelligent analysis function

[0113] Investment Comparison Analysis: Compare the differences between the "Preliminary Design Estimate" and the "Feasibility Study Estimate" to generate an "Investment Deviation Analysis Table" (marked "Dam Project Cost Overrun 15%, Reason: Rising Steel Price").

[0114] Cost calculation and risk assessment: Based on historical data, predict the "risk of material price fluctuations during the construction period" and give the suggestion of "stockpiling 30% of steel in advance";

[0115] User interaction support: Users can query "basics for earthwork quota adjustment for XX project" using natural language. The AI ​​assistant will then call the quota database and description chain to return the specific terms and calculation process.

[0116] II. Computer Equipment and Storage Media:

[0117] 1. Computer equipment

[0118] Using server-grade hardware configurations (such as an Intel Xeon Gold 6330 processor, 128GB of memory, and 2TB of SSD storage), and installing a Linux operating system and a Python 3.9 environment, the processor executes the computer program stored in the memory to achieve all the steps of the above method.

[0119] 2. Computer-readable storage medium

[0120] The computer program can be stored on media such as USB flash drives, external hard drives, or optical discs. When the program is loaded by the processor, it automatically executes the functional code at each level, ensuring the system's portability across different devices.

[0121] III. Summary of Technical Results:

[0122] This embodiment achieves the following technological breakthroughs through the collaborative design of a multi-level architecture:

[0123] Automation: It replaces traditional manual calculations and repetitive compilation work, improving the efficiency of engineering quantity calculation;

[0124] Intelligentization: Through learning and reasoning, AI agents can autonomously complete complex tasks such as quota matching and report generation;

[0125] Compliance: Strictly adheres to industry norms and quota standards, and the results are ensured to be compliant through built-in verification modules (such as "fee standard compliance check");

[0126] Scalability: Supports the addition of new software components (such as the "Green Construction Cost Adjustment Component") and workflows to adapt to industry technology development.

[0127] This invention provides a fully intelligent solution for the preparation of water conservancy project costs, which significantly reduces labor costs and improves the accuracy of results and decision-making efficiency.

[0128] Example: Specific usage process

[0129] Taking the cost estimation of a "large reservoir reinforcement project" as an example, the complete workflow of this invention is as follows:

[0130] 1. System initialization and knowledge base loading

[0131] After the system is started, the water conservancy cost knowledge base automatically loads the "Regulations for the Compilation of Water Conservancy Engineering Design Estimates (Estimations)" (2024 Edition) and local supplementary quotas, and simultaneously activates the workflow description standards (including core processes such as "Preliminary Design Estimate Compilation" and "Bill of Quantities Generation"). The software component registry completes the registration of 12 core components, such as the "Concrete Quantity Calculation Component" and the "Foundation Unit Price Calculation Component". The large language model performs semantic verification on the component inputs and outputs (such as unifying the feature description format of "dam concrete" to "concrete strength, frost resistance, and impermeability grade + pouring method + component size parameters, etc.").

[0132] 2. Interface layer adaptive learning

[0133] The system reads the project's historical compilation records through the interface layer, learns that "the design institutes in this region are accustomed to dividing the engineering parts into 'dam body → spillway → water conveyance tunnel'", and automatically adjusts the item sorting rules; at the same time, it parses out the compilation rule that "the unit price of labor budget is implemented according to the provincial-level published 82 yuan / man-day" and stores it in the knowledge memory base.

[0134] 3. Agent layer parameter configuration

[0135] Users input project information through a graphical interface:

[0136] Project attributes: Name: "XX Reservoir Reinforcement and Upgrading Project", Location: "XX County, XX Province" (altitude 2300m), Project category: "Large-scale water conservancy project";

[0137] Role settings: If "Compilation Unit" is selected, the system will grant permissions for "Preliminary Budget Compilation", "Basic Budget Compilation", "Estimated Budget Compilation", and "Results Export", while restricting the "Approval Adjustment" and "Audit Adjustment" functions;

[0138] Personalized configuration: Custom reports must include a "Construction Period" column, and the default is to load "Water Conservancy Construction Engineering Budget Quota (2024 Edition)".

[0139] The Agent layer converts parameters into system commands, such as "Altitude 2300m → Construction machinery hourly rate multiplied by a coefficient of 1.02" and "Design unit role → Permission code SJ-001".

[0140] 4. Data Acquisition at the Information Acquisition Layer

[0141] The project is divided into a coding database. For example, the code for "hub project → water-retaining project → concrete dam project → earthwork excavation → general earthwork excavation" in the construction project is "10-10.10.10.10.01".

[0142] Import the BIM model of this project, and call the BIM cost estimation software component to automatically collect the attribute information of the "C20 roller-compacted concrete cofferdam" (such as length 150m, top width 8m, bottom width 36.2m, height 15.6m), and calculate the project volume as V = 51714.00m. 3 ;

[0143] The user inputs a natural language description: "The cofferdam is constructed using roller-compacted concrete, with grade III crushed stone as aggregate." The system generates a project name item description chain: "C25 roller-compacted concrete cofferdam → grade III crushed stone → length 150m × (top width 8m + bottom width 36.2m) / 2 × height 15.6m," and infers a quota adjustment statement chain: "According to quota 4-2.3, the corresponding labor and machinery adjustment coefficient for roller-compacted concrete cofferdam construction is 0.90."

[0144] Automatically generate a quantity calculation result table (including project classification code, project name, component name, component ID, quantity, etc.) and a basic unit price calculation table (e.g., labor 82 yuan / man-day, cement 420 yuan / ton, crushed stone 85 yuan / m³). 3 (etc.), stored in the project database.

[0145] 5. Model management level fine-tuning and inference

[0146] QLoRA fine-tuning is initiated based on the collected data:

[0147] Training on quota pricing: Input the project name, item description, and quota adjustment description chain. The neural network mind map model matches "Quota No. 040005" (C25 roller-compacted concrete cofferdam). The quota labor and machinery consumption are multiplied by an adjustment coefficient of 0.90.

[0148] Training on compiling instructions: Based on the "Project Overview + Fee Standard" in the project database, generate a standard text stating that "If the unit price of labor for this project is implemented according to the standard for the second category of hardship and remote areas, other direct costs will be calculated at 8%, indirect costs will be calculated according to the key project / concrete pouring project (rate at 26%), and taxes will be calculated at 9%."

[0149] After fine-tuning, the model output accuracy is high, meeting engineering precision requirements.

[0150] 6. Generation of functional layer results

[0151] The user inputs the command "Generate Preliminary Design Budget Report", and the system executes the following:

[0152] Call the "Preliminary Estimate Preparation Workflow" to generate a description chain: "Call Project Division Component → Engineering Quantity Calculation Component → BIM Information Collection Component → Quota Pricing Calculation Component → Summary and Generation of Master Table";

[0153] The system automatically generates a "Building Engineering Unit Price List," such as a C25 roller-compacted concrete cofferdam with a comprehensive unit price of 580 yuan / m. 3 And add cost status information for “feasibility study stage, preliminary design stage, and construction stage” in a user-defined format and store it in the BIM model attribute set;

[0154] Results are automatically linked to the project database, supporting online preview and version rollback.

[0155] 7. Application of Results and Intelligent Analysis

[0156] The system outputs complete deliverables, including a basic unit price calculation sheet, quantity calculation results, and a preliminary design budget report (with electronic signature).

[0157] The intelligent analysis module automatically compares the "feasibility study estimate (12 million yuan)" with the "preliminary design estimate (12.6 million yuan)" and generates a deviation analysis: "The cofferdam project over budget by 600,000 yuan, mainly due to an 8% increase in the price of granite aggregate."

[0158] When a user queries whether the overspending complies with regulations, the AI ​​assistant returns the following: "Article 4.2 of the 'Measures for Adjusting Investment Price Differences in Water Conservancy Projects' states that material price fluctuations within ±10% are considered reasonable deviations," and suggests reserving 5% of the material price difference as a contingency fund in the bidding control price.

[0159] 8. Results Export and Archiving

[0160] When a user selects the "Export PDF + Excel" format, the system generates a preliminary estimate report with an electronic signature and an editable bill of quantities according to the workflow description standards. This is then automatically synchronized to the company's document system, completing the entire cost preparation process.

[0161] Through the above process, the preliminary design cost estimate preparation, which originally required multiple cost engineers to complete over several working days, can now be completed by only one engineer in a few hours using this system. Furthermore, the results are verified by the built-in compliance verification module and meet industry standards, thus achieving the goal of "accurate, efficient, and intelligent" cost estimate preparation for water conservancy projects.

[0162] In summary, the method and system based on AI intelligent agent for BIM cost estimation of water conservancy projects in this application aims to solve the problems of low efficiency, error-proneness, and lack of intelligent analysis in the existing process of compiling cost estimates for water conservancy projects. It provides a method and system based on AI intelligent agent for BIM cost estimation of water conservancy projects to achieve accurate, efficient, and intelligent cost estimation of water conservancy projects, thereby improving work efficiency and decision support capabilities.

[0163] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A hydraulic engineering BIM cost AI agent method, characterized in that, Comprise: Build a water conservancy cost knowledge base, collect water conservancy cost preparation regulations and quota regulations according to the design characteristics of water conservancy BIM cost software components, and establish workflow description standards; Use large language model to register unified access address of water conservancy BIM cost software components and workflow description standards, generate software component registry and workflow description registry, and manage components and standards; The large language model semantically governs the input and output and application functions of the registered software components and workflow description standards; Through the collaborative work of water conservancy BIM cost software components and workflow description standards, intelligent agents can perceive the environment, collect information, make decisions and perform tasks within the authorized scope.

2. The water conservancy BIM cost AI agent method according to claim 1, characterized in that, Also include: Build an interface layer, let the large language model learn the operation process and habits of users operating BIM cost software components through the interface layer, understand the relationship between software components and workflow description standards, and master the preparation rules of water conservancy cost preparation; Store learning records and effect evaluation in knowledge memory bank.

3. The water conservancy BIM cost AI agent method according to claim 1, characterized in that, Also include: Build Agent layer, accept user's setting instructions for knowledge base attributes, roles, Agent collaboration, security, permissions and personalization, and make Agent execute tasks within authorized scope according to user's instructions; Among them, the setting instructions include matching of project name, project number, project library name, construction nature, engineering category, report category; Matching of project location, price level, engineering elevation; Role setting includes preparation unit, audit unit, approval unit, consulting unit, audit unit; Security and authorization settings; Personalized settings.

4. The water conservancy BIM cost AI agent method according to claim 1, wherein, Also include: Build information collection layer, develop project division coding database and project name column item information collection prompt word template library, and establish persistent, stateful and controllable Agent information collection workflow prompt word template library; Through the collaborative work of water conservancy BIM cost software components and workflow description standards, carry out project division coding matching, project name column item selection, and engineering quantity calculation information collection, and generate engineering quantity calculation list report; Users input project name and project main feature context constraints in text or natural language according to information collection workflow prompts, and infer to generate project name column item description chain and quota adjustment chain; Generate engineering quantity calculation result report, basic unit price calculation table and project cost preparation basic information table according to collected information, and store the results in project database; The specific engineering quantity calculation result collection content includes project division coding, sub-item project name, project main feature, measured engineering quantity, component name, component ID, engineering part, construction method, main material, design parameter index and other cost attribute information.

5. The water conservancy BIM cost AI agent method according to claim 1, wherein, Also include: Build model management layer, use QLoRA method for fine-tuning training by large language model, including: Quota group price calculation fine-tuning training: according to the project name column item description chain and quota adjustment chain generated by inference, use neural network guide chart algorithm to generate quota group price calculation application chain; The preparation of the explanatory generation fine-tuning training: according to the project database stored cost preparation basic information, cost preparation rules and project cost characteristics, the quantity calculation, investment estimation, investment estimation, design budget, engineering budget, tender quantity list, tender quantity list of the preparation of the explanatory generation; The cost report generation fine-tuning training: according to the cost preparation rules and the project cost characteristics, the investment estimation report, the investment estimation report, the design budget report, the engineering budget report, the tender quantity list report, the tender quantity list report is generated; The cost preparation rules are used to build the prompt word template as the context information of the large language model, and a learned large language model is obtained; Among them, the fine-tuning training uses quantization and low rank adaptation technology to reduce the calculation resources and memory occupation.

6. The water conservancy BIM cost AI agent method according to claim 1, wherein, Also includes: Building a functional layer, let the large language model identify user intent, realize the business application of water conservancy project cost based on BIM technology preparation; Specifically includes: The function of the set of quota group price calculation is realized: select the set of quota group price calculation workflow description in the workflow description registry, and the large language model generates a quota group price calculation application expression chain through the information layer provided by the expression chain according to the instruction calling component. The expression chain handles the component connection relationship, input and output parameters, data source and data type, generates building engineering unit price table and installation engineering unit price table, and stores them into project database; The function of the preparation of the explanatory generation is realized: select the preparation of the explanatory generation workflow description in the workflow registry, and the large language model generates a preparation of the explanatory generation expression chain after calling the component according to the instruction, generates a preparation of the explanatory generation table, and stores it into project database; The function of the cost file report preparation is realized: select the cost file report preparation generation workflow description in the workflow registry, and the large language model generates a cost file report preparation expression chain after calling the component according to the instruction, generates water conservancy project estimation, estimation, budget, budget, quantity list and other business application files or reports at each stage, and stores them into project database.

7. The water conservancy BIM cost AI agent method according to claim 1, wherein, The workflow description standard includes: The setting instruction includes new project, open project, set rate, collect component attribute information, basic unit price calculation; Business functions include creating project information, generating quantity calculation result report, preparing basic unit price calculation table, setting quota group price calculation, preparing preliminary design budget report, preparing feasibility study investment estimation report, preparing engineering planning investment estimation report, preparing engineering budget report, preparing tender quantity list report, preparing tender quantity list, preparing engineering settlement report, preparing engineering settlement audit report; The result export includes export of quantity calculation result report, export of basic unit price calculation result, export of building engineering calculation table, export of installation engineering unit price calculation table, export of preliminary design budget report, export of feasibility study investment estimation report, export of engineering planning investment estimation report, export of engineering budget report, export of tender quantity list report, export of tender quantity list, export of engineering settlement report, export of engineering settlement audit report.

8. The water conservancy BIM cost AI agent method according to claim 6, wherein, Also includes: Result application and result intelligent analysis: The generated business application includes a basic unit price calculation table, an engineering quantity calculation result report, a preparation statement, an engineering budget summary table, an engineering budget table, an engineering budget table, a preliminary design budget report, a feasibility study investment estimation report, an engineering planning investment estimation report, an engineering budget report, a bidding engineering quantity list report, a bidding engineering quantity list, an engineering settlement report, and an engineering settlement audit report. The intelligent analysis of the results includes investment comparison analysis, cost estimation and estimation, price inquiry, progress management and risk assessment. Users interact with application requirements in natural language or text import mode, AI assistants understand user intent, and independently call BIM cost software component tools to operate, realize business function application, and ensure that the results are compliant and accurate.

9. A computer device comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program comprises instructions that, when executed by the processor, cause the processor to perform the method of any one of claims 1-8. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is loaded by the processor to execute the steps of the method of any one of claims 1 to 8.

Images