Building specification intelligent consultation system and method based on knowledge graph enhancement

Abstract

The invention discloses a building specification intelligent consultation system and method based on knowledge graph enhancement, and relates to the technical field of artificial intelligence. The system comprises a multi-modal data processing module which is used for carrying out intelligent analysis on a building specification document in a PDF format and realizing extraction of text, chart and formula contents; the knowledge graph construction module is used for converting the structured data into a complete knowledge network and establishing an association relationship between specifications; the graph enhancement retrieval module is used for realizing deep semantic retrieval based on a knowledge graph and supporting multi-hop path discovery; the multi-Agent collaboration module is used for completing a complex consultation task through collaboration of a plurality of professional Agents; the user interaction module provides a query input and result display interface and supports multiple interaction modes; in addition, according to the system, the invention further provides a building specification intelligent consultation method based on knowledge graph enhancement. According to the method, standard documents are processed through a multi-modal AI technology, a complete knowledge graph is constructed, deep semantic understanding is achieved through a graph enhancement retrieval technology, and professional consultation services are provided through a multi-agent cooperation mechanism.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and specifically to an artificial intelligence building specification intelligent consultation system and method based on knowledge graph enhancement, a corresponding computer program product, a computer-readable storage medium, and an electronic device. Background Art

[0002] The specification and standard system of the construction industry is huge and complex, involving technical details and professional requirements in multiple links such as design, construction, and management. Engineering technicians need to frequently query and understand these specification requirements in their daily work to ensure that the design plan meets the relevant standards. At present, there are mainly the following problems in specification query and understanding:

[0003] First, regulatory documents usually exist in unstructured formats such as PDF, and contain content in multiple forms such as text, tables, and drawings. Traditional text retrieval methods are difficult to effectively process such multimodal information, resulting in low query efficiency.

[0004] Secondly, there are a large number of cross-references and associations between specifications. These complex knowledge networks are difficult to effectively express and utilize through simple database storage and retrieval. Existing systems often separate specifications and lack a deep understanding of knowledge associations.

[0005] Thirdly, traditional retrieval augmentation generation (RAG) technology has limitations when dealing with professional domain knowledge and cannot fully utilize the associations between specifications, resulting in a lack of completeness and relevance in retrieval results.

[0006] In addition, existing question-and-answer systems usually adopt a single processing mode, which makes it difficult to cope with complex consulting needs. Especially in scenarios that require cross-standard analysis and comprehensive judgment, the system's understanding ability and answer quality are difficult to meet the needs of professional users.

[0007] In order to solve the above problems, the present invention proposes an artificial intelligence building specification intelligent consultation method based on knowledge graph enhancement. Summary of the invention

[0008] The purpose of the present invention is to propose an intelligent consulting system and method for building specifications based on knowledge graph enhancement to solve the problems raised in the background technology.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] Multimodal data processing module:

[0011] The present invention uses multimodal deep learning technology to process building specification documents to achieve effective extraction of different types of information. Specifically, the system converts PDF format documents into structured data through layout analysis and structure recognition. Optical character recognition (OCR) technology is used to extract text information, and image recognition technology is combined to analyze charts and formula content. Natural language processing technology is used to understand the extracted text semantically. All extracted information will be integrated to ensure that the associations between them can remain consistent in subsequent processing.

[0012] Knowledge graph building blocks:

[0013] The structured data obtained through multimodal processing is further converted into a knowledge graph. The construction of the graph includes identifying entities and extracting the relationships between entities, and using the graph database to establish multidimensional associations between entities. At the same time, the graph will also express the reference relationship and dependency relationship between building specifications to ensure the organic combination of different specifications. In order to ensure the real-time and high efficiency of the graph, a graph database is used for storage, and an efficient storage and indexing mechanism is designed for fast query and update.

[0014] Graph Enhanced Retrieval Module (GraphRAG):

[0015] The graph-enhanced retrieval module (GraphRAG) is a key technology of the present invention, which can realize deep semantic retrieval based on the knowledge graph. The system analyzes the user's query intent, identifies the key concepts in the query, and locates the relevant nodes in the knowledge graph. Through multi-hop path search, the system can deeply explore the complex relationships between specifications, extract path-related contextual information, and then generate comprehensive retrieval results. The GraphRAG module can handle multiple query methods and support flexible query requirements. Through the above method, multi-hop paths related to the query problem can be quickly and efficiently located in a complex knowledge graph, thereby improving the accuracy and response time of the query.

[0016] Professional answer agent is responsible for handling various specific issues, including:

[0017] (1) Code interpretation: The Code Interpretation Agent is responsible for interpreting the provisions of building codes and providing detailed meanings of the provisions and their scope of application.

[0018] (2) Calculation Verification: The calculation verification agent is responsible for calculating and verifying numerical problems to ensure that all calculations meet the requirements of building regulations.

[0019] (3) Solution recommendation: The solution recommendation agent provides multiple solution suggestions based on user needs, supporting users to make choices among multiple options.

[0020] (4) Conflict Check: The conflict check agent is responsible for checking conflicts between specifications and providing solutions.

[0021] The multi-agent collaborative processing system also includes the following additional mechanisms:

[0022] (1) Message passing mechanism: Different agents exchange information through an efficient message passing mechanism to ensure task coordination and data consistency.

[0023] (2) Task coordination and resource allocation: The task coordination mechanism ensures the reasonable allocation of tasks for each agent and dynamically schedules resources according to processing capabilities.

[0024] (3) Parallel processing and exception management: The system supports multiple agents to process tasks in parallel and has a built-in exception handling mechanism to ensure stable operation of the system under high load conditions.

[0025] (4) Result caching and sharing mechanism: Intermediate results are shared with each agent through a caching mechanism to speed up information transmission and avoid repeated calculations.

[0026] (5) Execution monitoring and management: Through the execution status monitoring system, the execution status of the entire intelligent system is monitored in real time to ensure the smooth completion of the task.

[0027] Regulatory query and consultation functions support:

[0028] The system supports two query modes: general mode and advanced mode. General mode is suitable for basic standard queries, while advanced mode is suitable for complex consulting issues. The system supports multiple rounds of interaction, problem clarification, and historical query management, and can flexibly switch between the two query modes to meet the needs of different users.

[0029] Computer equipment implementation plan:

[0030] The present invention also provides a computer device, comprising a processor and a memory, wherein the memory stores a program code for implementing the method of the present invention, and the processor completes various functions of intelligent consultation on building specifications by executing the program code.

[0031] Compared with the prior art, the present invention provides an intelligent consulting system and method for building specifications based on knowledge graph enhancement, which has the following beneficial effects:

[0032] (1) Knowledge processing ability:

[0033] Multimodal processing technology enables intelligent parsing of regulatory documents; knowledge graphs build a complete regulatory knowledge system; graph-enhanced retrieval technology supports deep semantic understanding; and multi-agent collaboration provides professional consulting capabilities.

[0034] (2) System performance:

[0035] General query response time ≤ 3 seconds; complex consultation response time ≤ 10 seconds; knowledge graph supports millions of nodes; concurrent processing capacity ≥ 100 users; system availability ≥ 99.9%.

[0036] (3) Query results:

[0037] General query accuracy ≥98%; complex consultation satisfaction ≥95%; professionalism score ≥4.5 / 5; standard citation accuracy 100%; answer completeness ≥92%.

[0038] (4) Expansion capability:

[0039] Supports continuous updating of standardized knowledge; has incremental learning capabilities; can be expanded to multiple professional fields; supports custom knowledge bases; and provides API calling interfaces.

[0040] (5) Application value:

[0041] Significantly improve the efficiency of regulatory queries; reduce the workload of professionals; reduce the possibility of human errors; speed up the training of new personnel; and promote knowledge sharing and inheritance.

[0042] In summary, the present invention proposes an intelligent consulting system and method for building specifications based on knowledge graph enhancement, which processes specification documents through multimodal AI technology, constructs a complete knowledge graph, adopts graph-enhanced retrieval technology to achieve deep semantic understanding, and provides professional consulting services through a multi-agent collaborative mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the overall architecture of the building specification intelligent consulting system based on knowledge graph enhancement mentioned in an embodiment of the present invention;

[0044] Figure 2 Schematic diagram of the multimodal data processing flow mentioned in an embodiment of the present invention;

[0045] Figure 3 A schematic diagram of the knowledge graph construction process mentioned in an embodiment of the present invention;

[0046] Figure 4 It is a schematic diagram of the graph enhanced retrieval process mentioned in an embodiment of the present invention;

[0047] Figure 5 Schematic diagram of multi-agent collaborative workflow mentioned in an embodiment of the present invention;

[0048] Figure 6 Schematic diagram of the user interface mentioned in the embodiment of the present invention. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0050] Example:

[0051] This invention proposes an intelligent consulting system for building specifications based on knowledge graph enhancement. Figure 1 , where: 1-1 is a multimodal data processing module, which is used to intelligently parse building specification documents in PDF format and extract content in various forms such as text, charts, and formulas; 1-2 is a knowledge graph construction module, which is used to convert structured data into a complete knowledge network and establish associations between specifications; 1-3 is a graph-enhanced retrieval module, which implements deep semantic retrieval based on knowledge graphs and supports multi-hop path discovery; 1-4 is a multi-agent collaboration module, which completes complex consulting tasks through the collaboration of multiple professional agents; 1-5 is a user interaction module, which provides query input and result display interfaces and supports multiple interaction modes; the arrows in the figure indicate data flow and processing directions.

[0052] The processing flow of the above multimodal data processing module is as follows: Figure 2 As shown, among them: 2-1 is the PDF document input unit, which is responsible for receiving and preprocessing building specification documents to ensure document integrity and format consistency; 2-2 is the layout analysis unit, which uses a deep learning model to identify the document structure, divide the content area, and mark special elements; 2-3 is the text recognition unit, which uses OCR technology to extract text content, process professional terms, and maintain format features; 2-4 is the chart processing unit, which identifies and analyzes non-text content such as graphics and tables to extract key information; 2-5 is the structured conversion unit, which converts the extracted information into a standard format to support subsequent processing; the arrows in the figure indicate the sequence of the processing flow and the data transfer relationship.

[0053] The processing flow of the above knowledge graph construction module is as follows Figure 3As shown, 3-1 is an entity recognition unit, which uses a deep learning model to identify professional concepts, terms and key elements in standard documents with an accuracy rate of no less than 95%; 3-2 is a relationship extraction unit, which is responsible for identifying various relationship types such as references, dependencies, and constraints between entities and building a relationship network; 3-3 is an attribute annotation unit, which supplements the attributes and features of entities and supports multi-dimensional information description; 3-4 is a graph construction unit, which builds a knowledge graph based on the identified entities and relationships to ensure the integrity of the knowledge structure; 3-5 is a graph optimization unit, which can improve the query performance and storage efficiency of the graph database or knowledge graph, and can also enhance the scalability and intelligent reasoning ability of the system. 3-5-1 is a storage management unit, including: 3-5-1 is a graph database storage unit, which uses a distributed graph database to store knowledge graphs and supports efficient queries; 3-5-2 is an index management unit, which establishes a multi-dimensional indexing mechanism and optimizes retrieval performance; 3-5-3 is a version control unit, which manages the version evolution of the knowledge graph and supports incremental updates; the arrows in the figure represent the processing of knowledge graph construction.

[0054] The processing flow of the above graph enhanced retrieval module is as follows: Figure 4 As shown, among them: 4-1 is the query analysis unit, which is responsible for semantic understanding of user queries, identifying query intent and key concepts; 4-2 is the node positioning unit, which locates relevant entity nodes in the knowledge graph and establishes query mapping; 4-3 is the path search unit, which uses heuristic algorithms to explore multi-hop paths and discover knowledge associations; 4-4 is the context generation unit, which extracts path-related context information and constructs a complete knowledge background; 4-5 is the result optimization unit, which sorts and deduplicates the retrieval results according to their relevance to ensure output quality; the arrows in the figure indicate the execution order and data flow of the retrieval process.

[0055] The processing flow of the above multi-agent collaboration module is as follows: Figure 5As shown, 5-1 is the problem reconstruction agent, which is responsible for receiving user input, analyzing query intent, clarifying problem requirements, and standardizing query expressions; 5-2 is the task planning agent, which decomposes tasks according to the complexity of the problem, formulates processing strategies, and allocates execution resources; 5-3 is the professional answer agent group, including: 5-3-1 is the specification interpretation agent, which is responsible for interpreting the clauses and analyzing the applicability; 5-3-2 is the calculation verification agent, which is responsible for numerical calculation and indicator verification; 5-3-3 is the solution recommendation agent, which is responsible for providing design suggestions and optimization solutions; 5-4 is the result integration agent, which is responsible for collecting and merging the processing results of each agent to ensure the integrity and consistency of the output; 5-5 is the output optimization agent, which is responsible for format standardization, professional optimization and readability improvement of the final result; 5-6 is the intelligent feedback loop mechanism, which is responsible for improving the system's adaptive ability and intelligent decision-making ability, and optimizing system behavior and performance through real-time feedback and self-adjustment. 5-7 is the data sharing and synchronization mechanism, which is responsible for the timely and effective exchange and update of data between multiple systems, nodes or components to ensure the consistency and integrity of information. 5-7 is an anomaly detection and fault tolerance mechanism, which is responsible for the system to detect and take measures in time when errors or failures occur, to ensure the stability and reliability of the system. The solid arrows in the figure represent task allocation and data flow.

[0056] The user interface designed based on the above user interaction module is as follows Figure 6 As shown, among them: 6-1 is the query input area, which provides a natural language input interface and supports multiple input methods; 6-2 is a mode switching control, which is used to switch between general query and advanced consulting modes; 6-3 is the question-and-answer interaction area, which displays multi-round dialogue content and supports dynamic interaction; 6-4 is the result display area, which presents query results and suggestions in a structured manner; 6-5 is the feedback operation area, which provides rating, collection and other functions to collect user feedback; each area works together to provide users with a smooth interactive experience.

[0057] Based on the above system, the present invention further proposes a building specification intelligent consultation method based on knowledge graph enhancement, including:

[0058] Building specification documents are processed based on multimodal deep learning technology to extract various forms of content such as text, charts, formulas, etc. in the documents; the processing process includes layout analysis, structure recognition, text extraction, image processing and semantic understanding of PDF format specification documents.

[0059] The extracted content is structured and a knowledge graph is constructed, wherein the knowledge graph expresses the association relationship between specifications; the process includes the steps of entity recognition, relationship extraction, multi-dimensional association establishment, reference dependency analysis, etc.

[0060] A graph-enhanced retrieval module (GraphRAG) is constructed to implement deep semantic retrieval based on the multi-hop path of the knowledge graph; this module realizes intelligent retrieval and association analysis of normative knowledge through heuristic path exploration, relevance scoring, path optimization and other technologies.

[0061] Construct a multi-agent collaborative processing system, including problem reconstruction agent, task planning agent, professional answer agent, result integration agent and output optimization agent; through the collaborative work of multiple professional agents, realize the intelligent processing of complex consulting problems.

[0062] Based on the graph enhanced retrieval module and the multi-agent collaborative processing system, standard query and consultation functions are realized, supporting two modes of general query and advanced consultation to meet the needs of users at different levels.

[0063] The processing of multimodal deep learning technology includes:

[0064] 1. Document preprocessing:

[0065] Layout analysis: using deep learning models to identify document structure

[0066] Content classification: distinguish different types of content such as text, charts, formulas, etc.

[0067] Format standardization: unified processing format for easy subsequent analysis

[0068] 2. Content Extraction:

[0069] Text recognition: Use OCR technology to extract text content, with an accuracy rate of ≥ 98%

[0070] Image processing: Identify engineering drawings and schematics and extract key information

[0071] Table analysis: identify table structure and convert it into structured data

[0072] Formula processing: recognize mathematical formulas and maintain their logical structure

[0073] The process of building a knowledge graph includes:

[0074] 1. Entity Recognition and Attribute Extraction:

[0075] Use deep learning models to identify specialized concepts and terminology

[0076] Extract entity attributes and feature information

[0077] Establishing an entity classification system

[0078] Perform entity disambiguation and synonym processing

[0079] 2. Relationship building:

[0080] Explicit relationship extraction: such as normative references, definition explanations, etc.

[0081] Implicit relationship discovery: based on semantic analysis and rule reasoning

[0082] Relationship weight calculation: evaluate the strength of association

[0083] Relationship verification: ensuring the accuracy of relationships

[0084] 3. Knowledge organization:

[0085] Use graph database to store knowledge structure

[0086] Establish a multi-dimensional indexing mechanism

[0087] Support real-time update and version management

[0088] Achieve incremental knowledge expansion

[0089] The implementation of the graph enhanced retrieval module includes:

[0090] 1. Query analysis:

[0091] Use natural language processing techniques to understand query intent

[0092] Identify key concepts and constraints in queries

[0093] Map queries to knowledge graph entities

[0094] Generate search strategy

[0095] 2. Path search:

[0096] Implement multi-hop path exploration algorithm

[0097] Design path scoring mechanism:

[0098] οRelationship weight: 0.4

[0099] οNode relevance: 0.3

[0100] ο Path length penalty: 0.2

[0101] οTimeliness weight: 0.1

[0102] Execution path optimization and pruning

[0103] Generate optimal search path

[0104] 3. Context Generation:

[0105] Extract path node information

[0106] Organization-related content

[0107] Generate structured context

[0108] Optimize search results

[0109] The workflow of the multi-agent collaborative processing system includes:

[0110] 1. Problem reconstruction agent:

[0111] Receive user input

[0112] Analyze query intent

[0113] Extract key information

[0114] Standardized query expressions

[0115] Clarify needs through interaction

[0116] 2.Task Planning Agent:

[0117] Assess the complexity of the problem

[0118] Develop a treatment strategy

[0119] ·Decompose the task objectives

[0120] Allocate processing resources

[0121] Coordinate execution process

[0122] The workflow of the multi-agent collaborative processing system includes:

[0123] 1. Problem reconstruction agent:

[0124] Receive user input

[0125] Analyze query intent

[0126] Extract key information

[0127] Standardized query expressions

[0128] Clarify needs through interaction

[0129] 2.Task Planning Agent:

[0130] Assess the complexity of the problem

[0131] Develop a treatment strategy

[0132] ·Decompose the task objectives

[0133] Allocate processing resources

[0134] Coordinate execution process

[0135] Monitor processing progress

[0136] ·Choose the appropriate professional answer agent according to the task type

[0137] Based on the assignment of the task planning agent, the system starts the corresponding professional answer agent to process:

[0138] 1. Professional answer agent includes:

[0139] · Standard interpretation agent:

[0140] ο Responsible for understanding and interpreting the regulations

[0141] οHandling professional terminology interpretation

[0142] ο Provide applicable provisions

[0143] οSolution rate ≥ 95%

[0144] ·Computation verification agent:

[0145] οPerform parameter calculations and verification

[0146] οComparison of processing numerical indicators

[0147] οDerive the formula

[0148] οCalculation accuracy ≥ 99%

[0149] ·Solution Recommendation Agent:

[0150] οGenerate suggestions based on context

[0151] οProvide design reference solutions

[0152] οProvide optimization suggestions

[0153] οRecommendation rationality ≥ 90%

[0154] Conflict Check Agent:

[0155] οIdentify normative conflicts

[0156] οAnalysis applicable conditions

[0157] οProvide treatment suggestions

[0158] οConflict recognition rate ≥ 92%

[0159] 2. Result integration agent:

[0160] Collect the processing results of each agent

[0161] Resolving conflicting results

[0162] Supplement missing information

[0163] Ensure logical consistency

[0164] Generate a unified answer

[0165] 3. Output optimization Agent:

[0166] Standardize professional terminology

[0167] Adjust the way you express yourself

[0168] Optimize content structure

[0169] Add citation instructions

[0170] Improve readability

[0171] The operation process of the multi-agent collaborative processing system is as follows:

[0172] 1. General query mode:

[0173] ·User input simple standard query

[0174] The system directly retrieves relevant content through GraphRAG

[0175] A single agent processes query results

[0176] Generate standardized answers quickly

[0177] Response time ≤ 1 second

[0178] 2. Advanced Consulting Model:

[0179] Receive complex consultation questions

[0180] Start multi-agent collaborative processing

[0181] Support multiple rounds of interaction to clarify requirements

[0182] Generate professional solutions

[0183] Response time ≤ 3 seconds

[0184] Based on the above content, the present invention further implements and explains the proposed building specification intelligent consulting system and method based on knowledge graph enhancement through specific examples, and its implementation environment requirements are as follows:

[0185] 1. Hardware environment:

[0186] Server configuration:

[0187] οCPU: High-performance multi-core processor

[0188] οMemory: no less than 64GB

[0189] ο Storage: High-speed SSD, capacity not less than 2TB

[0190] ο Network: Gigabit Ethernet interface

[0191] GPU resources:

[0192] οAI Inference Accelerator Card

[0193] ο Video memory no less than 16GB

[0194] οSupport parallel computing

[0195] οCUDA compatible

[0196] 2. Software environment:

[0197] Basic environment:

[0198] οOperating system: Linux

[0199] οContainerized deployment: Docker

[0200] οService Orchestration: Kubernetes

[0201] οLoad balancing: Nginx

[0202] Development framework:

[0203] ο Backend: Python

[0204] οGraph Database: Neo4j

[0205] ο Cache system: Redis

[0206] οAPI interface: FastAPI

[0207] 3. Deployment architecture:

[0208] Microservices organization:

[0209] οData processing services

[0210] οKnowledge Graph Service

[0211] οMulti-Agent Coordination Service

[0212] οAPI Gateway Service

[0213] Data Flow:

[0214] οAsynchronous message queue

[0215] οReal-time data synchronization

[0216] οState management mechanism

[0217] ο Resource Scheduling Strategy

[0218] Example 1:

[0219] Simple specification query

[0220] User input: "What is the minimum floor space requirement for bedrooms in residential buildings?"

[0221] System processing flow:

[0222] 1. Problem reconstruction Agent processing:

[0223] Identify keywords: "residential building", "bedroom", "minimum usable area"

[0224] Confirm query completeness

[0225] Select general query mode

[0226] 2.GraphRAG search:

[0227] Locate relevant specification nodes

[0228] Extraction of specific requirements

[0229] Organizing contextual information

[0230] System output:

[0231] "According to Article 5.2.1 of the Residential Design Code GB 50096-2011, the usable area of ​​a bedroom should not be less than 9 square meters, and the usable area of ​​a master bedroom should not be less than 12 square meters."

[0232] Example 2:

[0233] Complex regulatory consulting

[0234] User input: "What are the key requirements to consider when designing a hospital operating room?"

[0235] System processing flow:

[0236] 1. Problem reconstruction Agent processing:

[0237] Identify areas that require clarification

[0238] Generate clarifying questions: "What type of operating room is this? Are there any special equipment requirements?"

[0239] User response: "It is necessary to design a Class II clean operating room for a Class A tertiary hospital, mainly for general surgical operations"

[0240] Form clear query requirements

[0241] 2.Task Planning Agent Analysis:

[0242] Break the problem down into subtasks:

[0243] οSpace size requirements

[0244] ο Requirements for purification air conditioner

[0245] οMedical equipment requirements

[0246] οBuilding fire protection requirements

[0247] οElectrical installation requirements

[0248] Develop parallel processing strategies

[0249] ·Assign corresponding professional Agents

[0250] 3. Multiple professional answer agents process in parallel:

[0251] · Standard interpretation agent:

[0252] οSearch for relevant clauses of the "Hospital Building Design Code"

[0253] οInterpretation of the basic requirements of Class II clean operating rooms

[0254] οExtract key design parameters

[0255] ·Computation verification agent:

[0256] οCalculate the minimum usable area

[0257] οVerify clear height requirements

[0258] ο Calculate ventilation times

[0259] οConfirm technical indicators

[0260] ·Solution Recommendation Agent:

[0261] οProvide standard layout suggestions

[0262] οProvide equipment configuration plan

[0263] οRecommended material selection scheme

[0264] Conflict Check Agent:

[0265] ο Check the consistency of various professional specifications

[0266] οIdentify possible design conflicts

[0267] οProvide coordination advice

[0268] 4. Result integration agent processing:

[0269] Summarize the processing results of each agent

[0270] Categorize information by importance

[0271] Ensure coordination of requirements

[0272] ·Form a complete solution

[0273] 5. Output optimization Agent processing:

[0274] Standardized professional terminology

[0275] Organize content by design stage

[0276] Add specification basis description

[0277] Generate the final answer

[0278] System output:

[0279] “After analysis, the key design requirements for Class II clean operating rooms are as follows:

[0280] 1. Space requirements:

[0281] The usable area of ​​a single operating room should not be less than 40 square meters

[0282] ·Clear height should not be less than 3.0 meters

[0283] ·A hand scrub room and an equipment preparation room are required (according to: Article 4.3.2 of the "Hospital Building Design Code" GB 51039-2014)

[0284] 2. Cleanliness requirements:

[0285] Air cleanliness level: Class II (10,000 levels)

[0286] Fresh air volume: ≥2000m 3 / h

[0287] Ventilation frequency: ≥20 times / h (Based on: "Technical Specifications for Clean Operating Room Construction in Hospitals" GB 50333-2013)

[0288] 3. Facility requirements:

[0289] The operating room needs to be equipped with:

[0290] ο Panoramic shadowless light

[0291] οSurgical Instrument Table

[0292] οAnesthesia Machine

[0293] οPatient monitor

[0294] Gas interface configuration:

[0295] οOxygen: 2

[0296] οNegative pressure suction: 2

[0297] ο Compressed air: 1 (Based on: "Electrical Design Code for Medical Buildings" JGJ 312-2013)

[0298] 4. Building fire protection:

[0299] The operating room should be set up independently

[0300] Fire resistance level should not be lower than level 2

[0301] Fire partitions should be set up

[0302] There should be no less than 2 evacuation exits (Based on: Code for Fire Protection Design of Buildings GB 50016-2014)

[0303] 5. Special Notes:

[0304] Strictly control airflow organization

[0305] Pay attention to the layout of medical equipment and pipeline reservation

[0306] Consider the possibility of future equipment upgrades

[0307] Focus on the selection of seamless clean materials

[0308] 6. Suggested design process:

[0309] First determine the floor plan

[0310] ·Design airflow organization

[0311] ·Coordinate various professional pipelines

[0312] ·Detailed decoration methods

[0313] Example 3:

[0314] Regulatory conflict resolution

[0315] User input: "A project needs to meet both the green building three-star and ultra-low energy consumption building standards. How should the exterior window design be handled?"

[0316] System processing flow:

[0317] 1. Problem reconstruction Agent processing:

[0318] Identify key standards: Green Building Evaluation Standards and Ultra-low Energy Consumption Building Technical Standards

[0319] ·Clear focus: exterior window design requirements

[0320] Confirm that specification conflicts need to be addressed

[0321] 2.Task Planning Agent Planning:

[0322] Break down the tasks:

[0323] οExtract specific requirements of various specifications

[0324] οComparative analysis of differences

[0325] οFinding solutions

[0326] οGenerate design suggestions

[0327] Arrange processing priorities

[0328] ·Assign professional Agent

[0329] 3. Professional Agent processing:

[0330] · Standard interpretation agent:

[0331] οExtract Green Building Three Star External Windows Requirements

[0332] οExtract ultra-low energy consumption exterior window requirements

[0333] οGenerate a comparison list

[0334] ·Computation verification agent:

[0335] οVerify heat transfer coefficient index

[0336] οCalculate the shading coefficient

[0337] οAssessment of airtightness requirements

[0338] Conflict Check Agent:

[0339] οIdentify conflicting indicators

[0340] οAnalyze technical feasibility

[0341] οAssess economic rationale

[0342] 4. Result Integration Agent Integration:

[0343] Summarize all differences

[0344] Integrated solutions

[0345] Ensure the feasibility of the solution

[0346] 5. Output optimization Agent processing:

[0347] Standardize professional terminology

[0348] Clearly list the requirements

[0349] Provide specific recommendations

[0350] System output:

[0351] “After analysis, the two specifications have the following main differences in exterior window design:

[0352] 1. Heat transfer coefficient requirements:

[0353] Green Building Three Star: North District ≤1.5W / (m 2 ·K)

[0354] Ultra-low energy consumption: North area ≤0.8W / (m 2 ·K)(Required difference of 46.7%)

[0355] 2. Airtightness requirements:

[0356] Green Building Three Star: Meets ≤1.5m under 50Pa pressure 3 / (m 2 h)

[0357] Ultra-low energy consumption: ≤0.5m under 50Pa pressure 3 / (m 2 h) (required difference of 66.7%)

[0358] Suggested solution:

[0359] 1. Adopt stricter ultra-low energy consumption standards:

[0360] ·Select triple-glass double-cavity low-emissivity glass

[0361] ·Adopt thermal insulation aluminum alloy window frame

[0362] Increase sealing design

[0363] Standard requirements: K value ≤ 0.8W / (m 2 ·K)

[0364] 2. Specific technical measures:

[0365] Glass selection: Low-E coating + argon filling

[0366] Window frame structure: Nano thermal insulation bridge thermal insulation profile

[0367] Sealing design: multiple sealing strips

[0368] Installation requirements: External installation

[0369] 3. Key points of quality control:

[0370] · Carry out special design for air tightness

[0371] · Focus on controlling installation and construction quality

[0372] Increase on-site testing and verification

[0373] ·Establish model guidance mechanism

[0374] 4. Cost control suggestions:

[0375] ·Give priority to domestically produced high-quality products

[0376] Standardized design reduces costs

[0377] Bulk purchasing improves cost performance

[0378] Comprehensive evaluation of investment returns

[0379] Note:

[0380] 1. All technical indicators should meet ultra-low energy consumption standards

[0381] 2. Pay attention to the validity period of the test report

[0382] 3. Special supervision should be carried out throughout the process

[0383] 4. Pay attention to construction disclosure and process control"

[0384] Combining the above examples and comparing them with the prior art, the implementation effects are as follows:

[0385] 1. Improved efficiency:

[0386] Specification query time is reduced by 80%; solution verification time is reduced by 60%; design cycle is shortened by 30%; and drawing review efficiency is improved by 50%.

[0387] 2. Quality Improvement:

[0388] The error rate was reduced by 70%; compliance with regulations was improved; the degree of design standardization was increased; and the quality of technical briefing was improved.

[0389] 3. Cost Savings:

[0390] Labor costs are reduced by 40%; rework costs are reduced by 60%; training costs are reduced by 50%; and management costs are reduced by 35%.

[0391] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. The intelligent consulting system for building specifications based on knowledge graph enhancement is characterized by: include: (1) Multimodal data processing module, which is used to intelligently parse building specification documents in PDF format and extract text, charts, and formula content; (2) Knowledge graph construction module, which is used to transform structured data into a complete knowledge network and establish the association relationship between specifications; (3) Graph-enhanced retrieval module, which implements deep semantic retrieval based on knowledge graph and supports multi-hop path discovery; (4) Multi-agent collaboration module, which completes complex consulting tasks through the collaboration of multiple professional agents; (5) User interaction module, which provides query input and result display interface and supports multiple interaction modes.

2. The intelligent consulting method for building specifications based on knowledge graph enhancement implemented by the system according to claim 1 is characterized in that: It includes the following: S1. Processing building specification documents based on multimodal deep learning technology to extract text, charts, and formula contents in the documents; S2. Structuring the extracted content and constructing a knowledge graph, wherein the knowledge graph is used to express the association relationship between specifications, which specifically includes: entity attribute definition, relationship type design, rule constraint expression, version management mechanism and regular update strategy; S3, build a graph-enhanced retrieval module to achieve deep semantic retrieval based on the multi-hop path of the constructed knowledge graph; S4. Construct a multi-agent collaborative processing system, the system including: a problem reconstruction agent, a task planning agent, a professional answer agent, a result integration agent and an output optimization agent; S5. Based on the constructed graph-enhanced retrieval module and multi-agent collaborative processing system, standard query and consultation functions are realized.

3. The method according to claim 2, characterized in that The processing of building specification documents based on multimodal deep learning technology described in S1 specifically includes the following contents: S1.

1. Perform layout analysis and structure recognition on PDF format standard documents; S1.2, extract text content using optical character recognition technology; S1.3, use image recognition technology to process graphics and table content; S1.4 Use natural language processing techniques to understand text semantics; S1.

5. Integrate and process various types of information and maintain their associations.

4. The method according to claim 2, characterized in that: The construction of the knowledge graph described in S2 specifically includes: S2.1, obtain structured data after multimodal processing; S2.2, perform entity recognition and relationship extraction; S2.3, establish multi-dimensional associations between entities; S2.

4. Establish references and dependencies between specifications; S2.

5. Build a storage and indexing mechanism based on graph database.

5. The method according to claim 2, characterized in that: The construction of the graph enhancement retrieval module described in S3 specifically includes the following contents: S3.

1. Analyze user query intent and key concepts; S3.2, locate relevant nodes in the knowledge graph; S3.3, perform multi-hop path search and evaluation; S3.4, extracting context information associated with the path; S3.5, generate complete search results; The multi-hop path search and evaluation described in S3.3 specifically includes: S3.3.1, heuristic path exploration algorithm; S3.3.2, path relevance score; S3.3.3, ranking of path importance; S3.3.4, dynamic weight adjustment; S3.3.5, Path pruning optimization.

6. The method according to claim 2, characterized in that The workflow of the multi-agent collaborative processing system described in S4 includes: S4.1, Question Reconstruction Agent receives and analyzes user queries; S4.2, Task Planning Agent performs problem decomposition and task allocation; S4.3, multiple professional answer agents process subtasks in parallel; S4.4, the result integration agent merges the processing results of each agent; S4.

5. Output optimization agent normalizes the final answer.

7. The method according to claim 6, characterized in that The professional answer agent described in S4.3 includes: S4.3.1, Standard interpretation Agent, responsible for interpreting the provisions; S4.3.2, calculation verification agent, responsible for numerical calculation; S4.3.3, Solution Recommendation Agent, responsible for providing suggestions; S4.3.4, conflict checking agent, responsible for standard coordination.

8. The method according to claim 6, characterized in that The multi-agent collaborative processing system in S4 further includes: S4.6, message passing mechanism between agents; S4.7, Task coordination and resource allocation strategies; S4.8, parallel processing and exception handling mechanism; S4.9, caching and sharing mechanism of intermediate results; S4.

10. Monitoring and management of execution status.

9. The method according to claim 1, characterized in that: The regulatory query and consultation functions described in S5 include: S5.1, general mode, for direct specification query; S5.2, advanced mode, for consultation on complex issues; S5.3, dynamic switching between two modes; S5.4, multiple rounds of interaction and problem clarification; S5.

5. Query history management.

10. A computer device, characterized in that: The computer device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and the instruction, program, code set or instruction set is loaded and executed by the processor to implement the building specification intelligent consultation method based on knowledge graph enhancement as described in any one of claims 1-9.

Images