Building artificial intelligence integrated application platform

Abstract

The invention discloses a building artificial intelligence integrated application platform, and the platform comprises a cost intelligent pricing module which generates an optimal matching equipment item and a reference price; the bidding document auxiliary generation module is used for generating bidding document content; the equipment intelligent operation optimization module is used for generating an equipment adjustment strategy; the building environment monitoring and intelligent analysis module is used for generating environment regulation suggestions and risk prompts; the cross-system data integration and visualization module visually presents the operation state and the index change trend; the building knowledge sharing and semantic support module is used for providing semantic retrieval and knowledge recommendation services; the building operation evaluation module generates an evaluation report; the intelligent office assistant module is used for realizing intelligent guidance and active information pushing of an office process; the engineering intelligent supervision and scheduling module realizes construction site state monitoring and resource intelligent matching; and the building health detection module is used for carrying out continuous monitoring and service life management on building structures and equipment states. Building operation and maintenance management is improved.

Description

Technical Field

[0001] This invention relates to the field of digital transformation and cross-application of artificial intelligence technology in the construction industry, and in particular to an integrated application platform for artificial intelligence in construction. Background Technology

[0002] In recent years, with the rapid development of the construction industry and the iterative upgrading of information technology, the demand for informatization and intelligentization in the construction industry has experienced explosive growth. The integration and innovation of artificial intelligence technology with the construction industry has gradually become a research hotspot, as intelligent methods can break through the efficiency limitations of traditional management models.

[0003] In existing technologies, the application of artificial intelligence in the construction industry mainly revolves around office work, project management, and knowledge system construction. In office settings, addressing the traditional fragmented and inefficient information transmission and management model of the construction industry, technologies such as intelligent approval and automated document processing have enabled the digital upgrade of basic office processes. In project management, based on traditional project management software and data analysis tools, machine learning algorithms are being used to optimize construction plans and predict potential risks, providing auxiliary support for project decision-making. Regarding the construction of a building knowledge system, the use of knowledge graph technology is being explored to structure industry data such as design specifications and construction experience, compensating for the shortcomings of traditional knowledge management methods in intelligent retrieval and application.

[0004] However, existing technologies still have shortcomings in the following aspects: Technology fragmentation: Currently, most applications of artificial intelligence technology in the construction industry are still the introduction of individual technologies, lacking overall planning and system integration, resulting in serious information silos and making it difficult to form efficient synergy.

[0005] Limited data processing capabilities: Existing technologies still suffer from insufficient data processing capabilities when dealing with complex building data, especially in the analysis and mining of large-scale data, where efficient technical means are lacking.

[0006] Incomplete knowledge system construction: In terms of building a knowledge system for construction, existing technologies are mostly based on traditional knowledge management methods and lack intelligent support, making it difficult to meet the needs of complex construction projects.

[0007] While existing IT technologies in the construction industry have made some progress in areas such as smart office operations, digital project management, and knowledge system construction, problems remain, including low project collaboration efficiency under traditional management models, insufficient multi-source data fusion and analysis capabilities, significant information barriers between various business systems, and an incomplete construction knowledge system. These have become core bottlenecks restricting the industry's intelligent transformation. Summary of the Invention

[0008] In view of this, the purpose of this invention is to propose an integrated application platform for building artificial intelligence, which can realize intelligent perception, decision support and automatic task generation throughout the entire building life cycle, from early cost estimation to mid-term bidding preparation, and then to later operation management and performance evaluation, so as to comprehensively improve the level of intelligence, standardization and intensification of building operation and maintenance management.

[0009] To achieve the above-mentioned technical objectives, the technical solution adopted by this invention is as follows: This invention provides an integrated application platform for building artificial intelligence, comprising: The intelligent cost pricing module is used to extract semantic features and structured parameter information of equipment from equipment description text, match them with the equipment database, and generate the optimal matching equipment item and its reference price. The tender document generation module is used to extract the core points of the tender documents, generate the tender document content, and perform compliance verification and correction. The intelligent operation optimization module is used to generate equipment adjustment strategies based on the operating status and historical usage of various areas of the building, combined with the real-time load of the equipment, environmental changes and space usage. The building environment monitoring and intelligent analysis module is used to collect environmental data from various areas within the building, analyze trend changes and identify anomalies, and generate environmental adjustment suggestions and risk warnings. The cross-system data integration and visualization module is used to integrate multi-source heterogeneous data and present the operating status and indicator change trends in a visual way; The architectural knowledge sharing and semantic support module is used to construct a knowledge graph in the architectural field and provide semantic retrieval and knowledge recommendation services. The building operation evaluation module is used to comprehensively analyze equipment operation indicators, calculate operation and maintenance performance scores, and generate evaluation reports. The intelligent office assistant module is used to deploy a conversational AI interactive interface to enable intelligent guidance of office processes and proactive information push. The intelligent monitoring and scheduling module for construction projects is used to build an Internet of Things (IoT) network for construction sites. It combines image recognition and scheduling algorithms to achieve on-site status monitoring and intelligent resource matching. The building health monitoring module is used for continuous monitoring and lifespan management of building structures and equipment.

[0010] Furthermore, the intelligent cost pricing module includes: The text parsing unit is used during the project preparation phase to parse the equipment description text in the bill of quantities. This equipment description text includes: equipment name, functional requirements, specifications, reference price, and reference brand; specifically, it includes: Let the bill of quantities for construction projects consist of several project items, denoted as set. Among them, the i1th project entry Includes a device description text written in natural language. This is used to represent the equipment name, functional requirements, specifications, reference price, and reference brand; where n1 represents the total number of project items and n1 is a positive integer, and i1 represents the index and i1 is a positive integer. ; The data extraction unit is used to extract semantic features and structured parameter information of the device from the device description text; specifically, it includes: The device description text is first processed in two ways: firstly, through a large language model encoding function. Extract its semantic features and generate semantic vectors On the other hand, the structured information extraction function is called. From the device description text Key technical parameters are extracted from the data and a structured parameter vector is generated. ; The data matching unit is used to match the extracted semantic features and structured parameter information of the device with the semantic features and structured parameter information of the standard device items stored in the device database. If the match is successful, the result is output to the result output unit; if the match fails, no processing is performed. Specifically, it includes: Device database is represented as a set The j1st device item Includes its standardized text description Structured parameter vectors and price Where m1 represents the total number of device items and m1 is a positive integer, j1 represents the index and j Standardized text descriptions of equipment items Perform semantic encoding to obtain semantic vectors. ; for quantifying engineering entries With equipment items The degree of matching between them is defined by a scoring function that combines semantic similarity and parameter consistency:

[0011] in, These are the weighting coefficients for semantic and parameter matching. This represents a parametric inference function used for evaluation. and The degree of matching between them; The result output unit is used to generate the optimal matching equipment item and its corresponding reference price; specifically, it includes: The device with the highest score among all candidate devices is selected as the optimal matching device for the project entry. And return its reference price As a result of price arbitrage; Set acceptance threshold Only if the matching score satisfies Only then are the pricing results adopted, and the final output is a set of structured pricing results: .

[0012] Furthermore, the tender document generation module includes: The content extraction and encoding unit is used during the bidding stage to extract summaries and vectorize the content of the bidding documents; specifically, it includes: The received tender document text content is denoted as the text set. It includes the background of the bidding project, technical requirements, and scoring details, where k represents the total number of bidding document texts and k is a positive integer; through the extraction function Summarize and segment key points from each section of the tender document text to obtain several core summary units. Where u represents the total number of core abstract units and u is a positive integer, and the j2-th core abstract unit is... Then encoded into vectors by the large language model. They are uniformly stored in the bidding content vector database, where j2 represents the index and ; The content generation unit is used to drive the large model to generate bid document content chapter by chapter according to a predefined bid document structure. When generating each chapter, it references the corresponding bid content summary, industry standard fragments from the enterprise knowledge base, and summaries of previously generated content; specifically, it includes: Based on the predefined tender document structure template Where m2 represents the total number of chapters and m2 is a positive integer, the tender document content is generated segment by segment according to chapter-level granularity; let the currently generated chapter be... Where r represents the index and r Then the large language model generates the text for this chapter. When calling the method:

[0013] in, This represents the subset of tender summary units that have the highest semantic relevance to this chapter. For the retrieved set of industry knowledge, Indicates that the chapter was generated previously. The summary representation constitutes the generation context, and LLM stands for Large Language Model; The review sub-unit is used to perform a reasonableness check on the content of the generated bid documents according to the scoring rules and compliance requirements. If problems are found, it will provide feedback and trigger a chapter-level regeneration process; specifically, it includes: After each chapter is generated, the auxiliary review module is called to review the chapter text. A reasonableness review is conducted, which is based on a large language model comparing the original tender requirements and scoring rules, and executing a content compliance check function. Determine whether the chapter meets the corresponding requirements. ,in, This represents the set of rationality rules that need to be verified. If logical conflicts, omissions, or deviations in expression are found, the specific problem points will be located and marked for user reference or feedback to the large language model for correction. During the generation and review process, automatic iterative optimization is supported, that is, the feedback from the auxiliary review module is used as a constraint prompt to guide the large language model to generate new chapter versions. The review process is repeated multiple times until the generated content meets the preset standards of reasonableness and compliance, and the final output is a complete tender document organized in a structured and chapter-by-chapter format. .

[0014] Furthermore, the intelligent operation optimization module for the equipment specifically includes: Suppose that there exists a set of regions within the building. Where m3 represents the total number of regions and m3 is a positive integer, and the set of device types. Where n2 represents the total number of device types and is a positive integer, at time t, the operating state of device type d in region z is: The generated strategy recommendation value is The adjustment command is The strategy recommendation calculation formula is as follows:

[0015] in, This represents the real-time load of device type d within region z at time t. This indicates the maximum rated load of device type d in region z; This represents the historical energy consumption data of device type d in region z and at time t. Let be the long-term average energy consumption of equipment of type d in region z; The usage intensity factor of region z at time t is calculated by combining the dimensions of personnel density, spatial reservation status and environmental deviation. These are weighting coefficients, and When the strategy recommendation value When the threshold range is different, different control behaviors of the device are triggered, and adjustment commands are generated and issued. This enables equipment adjustment.

[0016] Furthermore, the building environment monitoring and intelligent analysis module includes: An environmental data acquisition unit is used to acquire environmental data during the building operation phase by using various sensor devices deployed within the building. The environmental data includes: indoor and outdoor temperature, relative humidity, PM2.5 concentration, carbon dioxide concentration, and light intensity. The environmental prediction unit is used to predict and analyze environmental change trends and identify abnormal states in real time based on environmental data by constructing a prediction algorithm and anomaly detection mechanism based on time series modeling, and to provide environmental regulation suggestions and operational risk warnings. The cross-system data integration and visualization module includes: The data fusion unit is used to acquire multi-source heterogeneous data from multiple systems within the building during its operational cycle by setting up data acquisition and access interfaces. These systems include a building energy consumption metering system, an environmental monitoring system, and a personnel behavior system. The building energy consumption metering system collects data on the electricity and water / gas consumption of air conditioning, elevators, and lighting equipment. The environmental monitoring system collects environmental data for each area. The personnel behavior system acquires dynamic behavioral information on the space occupancy status, entry / exit frequency, and personnel density of each functional area. The data foundation unit is used to construct a unified data structure model based on the data access results of multiple systems, so as to realize the standardized storage and integration of data from different sources, frequencies and formats in the platform, forming a unified data foundation. The visualization unit is used to visualize the equipment operation status, environmental level and personnel activity in different areas of the building at different times through a graphical visualization interface based on a data base. The graphical visualization interface supports interactive switching by time dimension, spatial dimension and equipment dimension, and provides a multi-layer linkage mode to realize real-time perception of the overall operation status of the building and historical trend analysis.

[0017] Furthermore, the architectural knowledge sharing and semantic support module includes: Graph construction units are used to build a knowledge graph of the construction domain that covers multiple entities and relationships related to component types, construction techniques, operational experience, and equipment maintenance; specifically including: Access to multi-source documents and structured data from the project implementation process, including construction drawings, process specifications, equipment operation manuals, operation and maintenance logs, and acceptance specifications; extract key entities from the multi-source documents and structured data, including component types, construction processes, operating experience, and equipment maintenance events; identify the logical relationships between key entities; and construct a knowledge graph in the construction field. The knowledge graph in the construction field expresses the semantic connections between building components, processes, equipment, and events in the form of a graph structure. It supports multi-level entity classification, multiple relationship types, and hierarchical logical organization, and is regularly supplemented with knowledge and semantic cleansing. The semantic service unit supports semantic retrieval, recommendation, and question-answering services, and serves as the underlying knowledge enhancement interface, providing semantic support and contextual supplementation to the platform; specifically, it includes: It provides semantic retrieval services to support user queries using natural language or keywords, returning relevant component information, process flow, common faults, and solutions; it also supports question-and-answer services and intelligent recommendation services based on knowledge graphs in the construction field to help users obtain targeted knowledge content; as an underlying knowledge enhancement interface, it provides semantic enhancement and contextual supplementation support for the platform's intelligent cost pricing module, tender document generation module, and intelligent equipment operation optimization module.

[0018] Furthermore, the building operation evaluation module includes: The indicator system establishment unit is used to take the multi-dimensional operation indicators collected by the platform as input, perform structured summarization and cleaning of the multi-dimensional operation indicators, and establish a unified evaluation indicator system; the multi-dimensional operation indicators include: equipment maintenance frequency, response timeliness, environmental regulation effectiveness, user satisfaction indicators, and energy consumption trends. The scoring calculation unit is used to set the weighting factors of the evaluation indicators according to the building operation and maintenance management objectives, and to calculate the overall operation and maintenance performance score of the building using a weighted scoring model, thereby forming a phased operation and maintenance score value. The report generation unit is used to generate a visualized phased operation and maintenance evaluation report based on the phased operation and maintenance score. The phased operation and maintenance evaluation report includes: the score distribution of various indicators, key issue prompts and suggested optimization directions, to help the management to have a global understanding of the current building operation status and adjust strategies.

[0019] Furthermore, the intelligent office assistant module includes: The office task scheduling unit is used to deploy a conversational AI interactive interface, supporting both voice and text input. Users can make requests to the intelligent office assistant module through natural language. The intelligent office assistant module uses semantic recognition and intent parsing to schedule and respond to office tasks. The office tasks include: expense approval process, seal application process, meeting room reservation, work order circulation and notification inquiry. Users can perform guided operations across processes and multiple steps through dialogue. The knowledge query unit is used to connect the intelligent office assistant module to the enterprise's internal knowledge base, including institutional documents, policy compilations and business guidelines. It supports semantic-level retrieval and intelligent extraction of document fragments. Users can send query requests to the intelligent office assistant module for management processes, approval requirements and operating procedures. The intelligent office assistant module automatically matches relevant content and outputs a summary of key points. The early warning push unit is used to build personalized demand models based on user behavior data and proactively push project risk warnings, policy and regulation updates, and schedule reminders.

[0020] Furthermore, the intelligent monitoring and scheduling module for the project includes: The network monitoring unit is used to collect key elements of the construction site in real time by deploying an Internet of Things monitoring network and combining personnel positioning devices, equipment sensing terminals and material identification tags. The key elements include the trajectory of construction personnel entering and leaving the area, the operating parameters of key equipment, and the arrival time and batch information of building materials. The digital twin unit is used to construct a three-dimensional virtual mapping scene of the construction site based on the key elements and combined with digital twin technology, so as to realize the dynamic visualization and monitoring of the construction status. The quality and safety assessment unit is used to connect to the image acquisition equipment and environmental sensors deployed on site. The image acquisition equipment automatically identifies process quality defects based on image recognition algorithms, and judges the level of safety hazards on site by combining sensor data collected by environmental sensors. When an abnormality is identified, a rectification notice is automatically generated, pushed to the relevant responsible personnel, and the closed-loop processing progress is tracked, recording the rectification completion status and time nodes. The resource scheduling unit integrates labor and material scheduling mechanisms. By jointly modeling the distribution of on-site labor, the pace of construction tasks, and the progress of material consumption, it uses optimization scheduling algorithms to achieve intelligent matching and resource allocation of labor and materials. Based on the prediction results, it proactively recommends scheduling schemes, including adjusting the input of construction teams and optimizing the batch and time window of material transportation.

[0021] Furthermore, the building health monitoring module includes: The structural monitoring unit is used to collect key operating parameters through structural stress and strain sensors and visual inspection equipment, and to assess the safety status of the structure based on these parameters; specifically, it includes: By deploying structural health monitoring sensors in key structural parts, the stress response, deformation trend and attitude change information of the main building structure during operation can be collected in real time to determine whether there is any over-limit behavior or trend risk in the current structural state, and to generate structural health assessment results. By deploying stress-strain monitoring nodes on key building components, multidimensional structural response data is continuously collected from each monitoring node, mainly including stress σ, strain ε, and tilt angle θ. , will the i-th The state of each sensor's stress-strain monitoring node at any time point t' is represented as a three-dimensional vector. ,in, The vector transpose is represented, and the whole matrix forms the structure-state matrix. Where N is the number of sensor nodes. Represents the real number field; For the j-th in the building Each structural unit determines a set of neighboring sensors based on its geometric location. Interpolation weights are set based on the spatial distance from the sensor to the structural unit. Thus, a simplified stress estimation model is constructed; at time t', the i-th... The j-th sensor Estimated stress values ​​of each structural unit Represented as:

[0022] Wherein, the weights satisfy And it is calculated using the inverse distance weighting method:

[0023] in, Indicates the i-th The sensor and the j-th Geometric distance of each structural unit; k This indicates traversing the entire sensor set. The mark, Indicates the kth The sensor and the j-th Geometric distance of each structural unit; After obtaining the estimated stress values ​​for each structural unit, the corresponding allowable stress limits are obtained by referring to the material design specifications or structural service standards. Based on this, the safety margin factor of the structural unit is calculated:

[0024] like If the signal is positive, it indicates that the structural unit currently has a potential risk of exceeding structural limits, which will trigger an early warning and mark the number of the exceeding unit and its region. The lifecycle prediction unit is used to perform time-series modeling and trend inference of equipment operating status, generate health scores, and predict the remaining lifespan of the equipment. It supports time-series analysis and health trend prediction of equipment operating status, specifically including: During the building's operation, at fixed time intervals Collect operating status parameters of various key electromechanical equipment within the building to construct a state vector. Where e represents the device number, k0 is the dimension of the state variable monitored by the device, and a sliding window mechanism of length L is used to construct a time series segment vector of the device state:

[0025] Where, n Indicates from a point in time At the appointed time There are a total of L state point vectors; After time series segment vectors are processed by stacked Transformer layers, a deep representation of the entire sequence is obtained. , where d For the hidden layer dimension, This represents the feature extraction function composed of a deep network with transformer layers. The nth layer obtained by the transformer layer The feature vector of the layer; after further average pooling, we get:

[0026] in, This represents the vector result after average pooling of feature vectors from layers 1 to L, where i0 represents the index and i0 ; Finally, a regression mapping layer is used. : Output the predicted health score for the current time. :

[0027] in, This represents a regression mapping layer composed of a multilayer perceptron network, which reduces the vector dimension from d... The value drops to 1; the higher the value, the better the current health status of the device. Less than the health warning threshold When the equipment is in a state of aging and degradation, an automatic warning will be issued, and a suggested maintenance time window will be generated.

[0028] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art: This invention addresses the technical problems of fragmented intelligent applications, insufficient data integration, and inefficient knowledge reuse in the construction industry by constructing a full-chain technical system covering intelligent office, digital engineering management, intelligent building operation and maintenance, and knowledge system construction. It achieves intelligent collaboration and decision optimization throughout the entire building lifecycle.

[0029] In the field of AI-enabled office work, the platform revolutionizes the traditional office model of the construction industry by deeply integrating natural language processing and intelligent interaction technologies. The tender document generation module, based on semantic analysis of tender documents and combined with enterprise knowledge bases and industry standards, drives a large model to automatically generate logically rigorous and compliant tender documents. An intelligent review mechanism ensures tender document quality, significantly improving the efficiency and accuracy of tender document preparation. The intelligent office assistant module, using a conversational AI interface, intelligently guides office processes and automates task distribution. Simultaneously, it builds personalized demand models based on user behavior data, proactively pushing risk warnings, policy and regulatory information, and schedule reminders, comprehensively improving the level of intelligent office work and service response efficiency in the construction industry.

[0030] At the level of AI-enabled engineering, the platform integrates full-cycle data resources and intelligent algorithms to drive the transformation of engineering management models towards digitalization and intelligence. The intelligent cost pricing module analyzes equipment description text in the construction project bill of quantities, extracts semantic features and structured parameters, and generates accurate reference prices through intelligent matching and parameter comparison, providing a reliable basis for project cost control. The intelligent engineering supervision and scheduling module relies on IoT monitoring networks and digital twin technology to collect real-time data on personnel, machinery, and materials at the construction site, constructs a dynamic visualization model of the construction scenario, and achieves intelligent assessment of quality and safety risks through image recognition and sensor fusion technology. Combined with optimized scheduling algorithms, it completes the intelligent allocation of labor and material resources, effectively improving construction management efficiency and project execution quality.

[0031] In the field of intelligent building operation and maintenance, the platform integrates multi-source heterogeneous data acquisition and intelligent analysis technologies to achieve comprehensive perception and intelligent control of building operation status. The building environment monitoring and intelligent analysis module acquires environmental data by deploying multiple types of sensors and uses time-series modeling and anomaly detection algorithms to predict environmental changes and provide risk warnings; the equipment intelligent operation optimization module combines electromechanical equipment operation data and historical usage to dynamically recommend energy-efficient equipment adjustment strategies; the cross-system data integration and visualization module builds a unified data foundation to achieve multi-dimensional visualization of data such as energy consumption, environment, and personnel; the building operation evaluation module builds a weighted model based on core operation data to quantitatively evaluate building operation and maintenance performance and generate evaluation reports; the building health monitoring module collects structural stress and visual inspection data and combines time-series analysis to predict the health status of building structures and equipment. The collaboration of multiple modules provides scientific decision support for building operation and maintenance management.

[0032] In terms of building a private knowledge base and knowledge system for the construction industry, the platform constructs a dedicated knowledge graph for the construction field, integrating multi-dimensional knowledge resources such as component types, construction techniques, operational experience, and equipment maintenance to form a structured and semantic knowledge network. This knowledge graph supports semantic retrieval, intelligent recommendation, and question-and-answer services, not only providing users with an efficient channel for knowledge acquisition but also serving as a foundational knowledge enhancement interface. It is deeply integrated into business modules such as tender document generation and operation and maintenance analysis, enabling intelligent association and application of knowledge, and providing a solid foundation for the accumulation, sharing, and innovative application of knowledge in the construction industry. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a schematic diagram of an integrated application platform for building artificial intelligence provided in an embodiment of the present invention.

[0035] Figure 2 This is a flowchart illustrating a management method for an integrated building artificial intelligence platform provided in an embodiment of the present invention. Detailed Implementation

[0036] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be particularly noted that the following embodiments are for illustrative purposes only and do not limit the scope of the invention. Similarly, the following embodiments are only some, not all, embodiments of the present invention, and all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Please see Figure 1 The present invention provides an integrated application platform for building artificial intelligence, comprising: 1) The intelligent cost pricing module is used to extract the semantic features and structured parameter information of the equipment from the equipment description text, match them with the equipment database, and generate the optimal matching equipment item and its reference price; based on the unstructured description of the project item, it realizes accurate matching of equipment items and achieves intelligent cost pricing management.

[0038] In this embodiment, the intelligent cost pricing module includes: The text parsing unit is used during the project preparation phase to parse the equipment description text in the bill of quantities. This equipment description text includes: equipment name, functional requirements, specifications, reference price, and reference brand; specifically, it includes: Let the bill of quantities for construction projects consist of several project items, denoted as set. Among them, the i1th project entry Includes a device description text written in natural language. This is used to represent the equipment name, functional requirements, specifications, reference price, and reference brand; where n1 represents the total number of project items and n1 is a positive integer, and i1 represents the index and i1 is a positive integer. ; The data extraction unit is used to extract semantic features and structured parameter information of the device from the device description text; specifically, it includes: The device description text is first processed in two ways: firstly, through a large language model encoding function. Extract its semantic features and generate semantic vectors On the other hand, the structured information extraction function is called. From the device description text Key technical parameters, such as model number, interface type, and size rating, are extracted from the data and a structured parameter vector is generated. This dual-path modeling mechanism enables the platform to simultaneously capture device functional semantics and technical parameters, thereby supporting joint reasoning at the semantic and parameter levels.

[0039] The data matching unit is used to match the extracted semantic features and structured parameter information of the device with the semantic features and structured parameter information of the standard device items stored in the device database. If the match is successful, the result is output to the result output unit; if the match fails, no processing is performed. Specifically, it includes: Device database is represented as a set The j1st device item Includes its standardized text description Structured parameter vectors and price Where m1 represents the total number of device items and m1 is a positive integer, j1 represents the index and j Standardized text descriptions of equipment items Perform semantic encoding to obtain semantic vectors. ; for quantifying engineering entries With equipment items The degree of matching between them is defined by a scoring function that combines semantic similarity and parameter consistency:

[0040] in, These are the weighting coefficients for semantic and parameter matching. This represents a parametric inference function used for evaluation. and The degree of matching between them; The result output unit is used to generate the optimal matching equipment item and its corresponding reference price; specifically, it includes: The device with the highest score among all candidate devices is selected as the optimal matching device for the project entry. And return its reference price As a result of price arbitrage; Set acceptance threshold Only if the matching score satisfies Only then are the pricing results adopted, and the final output is a set of structured pricing results: .

[0041] 2) Tender document generation module, used to extract the core points of the tender documents, generate the tender document content, and perform compliance verification and correction; automatically process the content of the tender documents, and generate the tender content according to the tender requirements, so as to realize intelligent auxiliary generation management of tender documents.

[0042] In this embodiment, the tender document generation module includes: The content extraction and encoding unit is used during the bidding stage to extract summaries and vectorize the content of the bidding documents; specifically, it includes: The received tender document text content is denoted as the text set. It includes the background of the bidding project, technical requirements, and scoring details, where k represents the total number of bidding document texts and k is a positive integer; through the extraction function Summarize and segment key points from each section of the tender document text to obtain several core summary units. Where u represents the total number of core abstract units and u is a positive integer, and the j2-th core abstract unit is... Then encoded into vectors by the large language model. All data is uniformly stored in the bidding content vector database for subsequent retrieval and retrieval; where j2 represents the index and j2 ; The content generation unit is used to drive the large model to generate bid document content chapter by chapter according to a predefined bid document structure. When generating each chapter, it references the corresponding bid content summary, industry standard fragments from the enterprise knowledge base, and summaries of previously generated content to maintain overall logical consistency. Specifically, it includes: Based on the predefined tender document structure template Where m2 represents the total number of chapters and m2 is a positive integer, the tender document content is generated segment by segment according to chapter-level granularity; let the currently generated chapter be... Where r represents the index and r Then the large language model generates the text for this chapter. When calling the method:

[0043] in, This represents the subset of tender summary units that have the highest semantic relevance to this chapter. For the retrieved set of industry knowledge, Indicates that the chapter was generated previously. The abstract represents the generated context, and LLM stands for Large Language Model. This mechanism ensures that when the language model generates new chapters, it can integrate project requirements, industry background, and existing content, thus guaranteeing the consistency, relevance, and professionalism within the tender document.

[0044] The review sub-unit is used to perform a reasonableness check on the content of the generated bid documents according to the scoring rules and compliance requirements. If problems are found, it will provide feedback and trigger a chapter-level regeneration process; specifically, it includes: After each chapter is generated, the auxiliary review module is called to review the chapter text. A reasonableness review is conducted to control the quality of the generated chapters. This review is based on a large language model, comparing it against the original tender requirements and scoring rules, and executing a content compliance check function. Determine whether the chapter meets the corresponding requirements. ,in, This represents the set of rationality rules that need to be verified. If logical conflicts, omissions, or deviations in expression are found, the specific problem points will be located and marked for user reference or feedback to the language model for correction. During the generation and review process, automatic iterative optimization is supported, that is, the feedback from the auxiliary review module is used as a constraint prompt to guide the large language model to generate new chapter versions. The review process is repeated multiple times until the generated content meets the preset standards of reasonableness and compliance, and the final output is a complete tender document organized in a structured and chapter-by-chapter format. .

[0045] 3) The intelligent operation optimization module is used to improve the energy efficiency of equipment such as air conditioning, elevators, and lighting based on the operating status and historical usage of various areas of the building; it generates equipment adjustment strategies by combining the real-time load, environmental changes and space usage of equipment (electromechanical equipment such as air conditioning, elevators, and lighting) to improve the overall flexibility and energy efficiency of the building. In this embodiment, the intelligent operation optimization module specifically includes: Suppose that there exists a set of regions within the building. Where m3 represents the total number of regions and m3 is a positive integer, and the set of device types. Where n2 represents the total number of device types and is a positive integer, at time t, the operating state of device type d in region z is: The generated strategy recommendation value is The adjustment command is The strategy recommendation calculation formula is as follows:

[0046] in, This represents the real-time load of device type d within region z at time t. This indicates the maximum rated load of device type d in region z; This represents the historical energy consumption data of device type d in region z and at time t. Let be the long-term average energy consumption of equipment of type d in region z; The usage intensity factor of region z at time t is calculated by combining the dimensions of personnel density, spatial reservation status and environmental deviation. These are weighting coefficients, and When the strategy recommendation value When the threshold range is different, different control behaviors of the device are triggered, and adjustment commands are generated and issued. This enables equipment adjustment.

[0047] 4) Building environment monitoring and intelligent analysis module, used to collect environmental data in various areas of the building, analyze trend changes and identify anomalies, and generate environmental adjustment suggestions and risk warnings; In this embodiment, the building environment monitoring and intelligent analysis module includes: An environmental data acquisition unit is used to acquire environmental data during the building operation phase by using various sensor devices deployed within the building. The environmental data includes: indoor and outdoor temperature, relative humidity, PM2.5 concentration, carbon dioxide concentration, and light intensity. The environmental prediction unit is used to predict and analyze environmental change trends and identify abnormal states in real time based on environmental data by constructing a prediction algorithm and anomaly detection mechanism based on time series modeling, and to provide environmental regulation suggestions and operational risk warnings. 5) Cross-system data integration and visualization module, used to integrate multi-source heterogeneous data and present the operating status and indicator change trends in a visual way; In this embodiment, the cross-system data integration and visualization module includes: The data fusion unit is used to acquire multi-source heterogeneous data from multiple systems within the building during its operational cycle by setting up data acquisition and access interfaces. These systems include a building energy consumption metering system, an environmental monitoring system, and a personnel behavior system. The building energy consumption metering system collects data on the electricity and water / gas consumption of air conditioning, elevators, and lighting equipment. The environmental monitoring system collects environmental data for each area. The personnel behavior system acquires dynamic behavioral information on the space occupancy status, entry / exit frequency, and personnel density of each functional area. The data foundation unit is used to construct a unified data structure model based on the data access results of multiple systems, so as to realize the standardized storage and integration of data from different sources, frequencies and formats in the platform, forming a unified data foundation for subsequent calling and processing; The visualization unit is used to visualize the equipment operation status, environmental level and personnel activity in different areas of the building at different times through a graphical visualization interface based on a data base. The graphical visualization interface supports interactive switching by time dimension, spatial dimension and equipment dimension, and provides a multi-layer linkage mode to realize real-time perception of the overall building operation status and historical trend analysis, which can help managers understand the changing trends of key indicators.

[0048] The graphical visualization interface can display, but is not limited to, the following: daily, weekly, and monthly distribution trend charts of energy consumption data; operating efficiency curves of specific equipment; space temperature and humidity heat maps; personnel flow heat maps; system status alarm markers, etc. Managers can use the graphical visualization interface to obtain the changing trends of key indicators of the building's current operation, assisting in energy efficiency optimization, fault location, and resource scheduling decisions.

[0049] 6) The architectural knowledge sharing and semantic support module is used to construct a knowledge graph in the architectural field, provide semantic retrieval and knowledge recommendation services, build a knowledge organization and sharing mechanism for the architectural field, improve the platform's knowledge reuse capability and intelligent service level, and realize architectural knowledge sharing management.

[0050] In this embodiment, the architectural knowledge sharing and semantic support module includes: Graph construction units are used to build a knowledge graph of the construction domain that covers multiple entities and relationships related to component types, construction techniques, operational experience, and equipment maintenance; specifically including: Access to multi-source documents and structured data from the project implementation process, including construction drawings, process specifications, equipment operation manuals, operation and maintenance logs, and acceptance specifications; extract key entities from the multi-source documents and structured data, including component types, construction processes, operating experience, and equipment maintenance events; identify the logical relationships between key entities; and construct a knowledge graph in the construction field. The knowledge graph in the construction field expresses the semantic connections between building components, processes, equipment, and events in the form of a graph structure. It supports multi-level entity classification, multiple relationship types, and hierarchical logical organization, and regularly supplements and cleans the knowledge to ensure the completeness and up-to-dateness of the knowledge content. The semantic service unit supports semantic retrieval, recommendation, and question-answering services, and serves as the underlying knowledge enhancement interface, providing semantic support and contextual supplementation to the platform; specifically, it includes: It provides semantic retrieval services to support user queries using natural language or keywords, returning relevant component information, process flow, common faults, and solutions. It also supports question-and-answer services and intelligent recommendation services based on knowledge graphs in the construction field, assisting users in obtaining targeted knowledge content. As an underlying knowledge enhancement interface, it provides semantic enhancement and contextual supplementation support for the platform's intelligent cost pricing module, tender document generation module, and intelligent equipment operation optimization module, improving project management and problem response efficiency.

[0051] 7) Building operation evaluation module, used to comprehensively analyze equipment operation indicators, calculate operation and maintenance performance scores and generate evaluation reports; based on the operation data collected and generated by each functional module in the platform, to evaluate and analyze the overall operation and maintenance performance of the building and generate operation and maintenance evaluation reports.

[0052] In this embodiment, the building operation evaluation module includes: The indicator system establishment unit is used to take the multi-dimensional operation indicators collected by the platform as input, perform structured summarization and cleaning of the multi-dimensional operation indicators, and establish a unified evaluation indicator system; the multi-dimensional operation indicators include: equipment maintenance frequency, response timeliness, environmental regulation effectiveness, user satisfaction indicators, and energy consumption trends. The scoring calculation unit is used to set the weighting factors of the evaluation indicators according to the building operation and maintenance management objectives, and to calculate the overall operation and maintenance performance score of the building using a weighted scoring model, thereby forming a phased operation and maintenance score value. The report generation unit generates a visualized phased operation and maintenance (O&M) evaluation report based on the phased O&M score. The report includes: the distribution of scores for various indicators, key issue alerts, and suggested optimization directions. This assists management in gaining a comprehensive understanding of the building's current operational status and adjusting strategies, enabling improvements in building energy efficiency and optimization of management strategies. The phased O&M evaluation report supports evaluation by period (e.g., daily, weekly, monthly) and can also be analyzed hierarchically by spatial area, equipment type, etc., identifying operational shortcomings or abnormal trends and generating a list of O&M optimization suggestions for use by other modules of the platform or for manual intervention by management personnel.

[0053] 8) Intelligent office assistant module, used to deploy a conversational AI interactive interface to realize intelligent guidance of office processes and proactive information push, thereby improving the intelligence of building office processes and service response capabilities; In this embodiment, the intelligent office assistant module includes: The office task scheduling unit is used to deploy a conversational AI interactive interface, supporting both voice and text input. Users can make requests to the intelligent office assistant module through natural language. The intelligent office assistant module uses semantic recognition and intent parsing to schedule and respond to office tasks. The office tasks include: expense approval process, seal application process, meeting room reservation, work order circulation and notification inquiry. Users can perform guided operations across processes and multiple steps through dialogue. The knowledge query unit is used to connect the intelligent office assistant module to the enterprise's internal knowledge base, including institutional documents, policy compilations and business guidelines. It supports semantic-level retrieval and intelligent extraction of document fragments. Users can send query requests to the intelligent office assistant module for management processes, approval requirements and operating procedures. The intelligent office assistant module automatically matches relevant content and outputs key summary, improving information acquisition efficiency. The early warning push unit is used to build personalized demand models based on user behavior data and proactively push project risk warnings, policy and regulation updates, and schedule reminders, which can improve the intelligence level and service response capabilities of building offices.

[0054] 9) The intelligent supervision and scheduling module for construction projects is used to build an Internet of Things network for construction sites. By combining image recognition and scheduling algorithms, it enables on-site status monitoring and intelligent resource matching, thereby improving the level of information management during the construction phase. In this embodiment, the intelligent engineering monitoring and scheduling module includes: The network monitoring unit is used to collect key elements of the construction site in real time by deploying an Internet of Things monitoring network and combining personnel positioning devices, equipment sensing terminals and material identification tags. The key elements include the trajectory of construction personnel entering and leaving the area, the operating parameters of key equipment, and the arrival time and batch information of building materials. The digital twin unit is used to construct a three-dimensional virtual mapping scene of the construction site based on the key elements and combined with digital twin technology, so as to realize the dynamic visualization and monitoring of the construction status. The quality and safety assessment unit is used to connect to the image acquisition equipment and environmental sensors deployed on site. The image acquisition equipment automatically identifies process quality defects based on image recognition algorithms, and judges the level of safety hazards on site by combining sensor data collected by environmental sensors. When an abnormality is identified, a rectification notice is automatically generated, pushed to the relevant responsible personnel, and the closed-loop processing progress is tracked, recording the rectification completion status and time nodes. The resource scheduling unit integrates labor and material scheduling mechanisms. By jointly modeling the distribution of on-site labor, the pace of construction tasks, and the progress of material consumption, it uses optimized scheduling algorithms to achieve intelligent matching and resource allocation of labor and materials. Based on the prediction results, it proactively recommends scheduling schemes, including adjusting the input of construction teams, optimizing material transportation batches and time windows, and improving the response efficiency and resource utilization of construction organization.

[0055] 10) Building health monitoring module, used for continuous monitoring and life management of building structure and equipment status, improving the safety of building operation and the foresight of maintenance work.

[0056] In this embodiment, the building health monitoring module includes: The structural monitoring unit is used to collect key operating parameters through structural stress and strain sensors and visual inspection equipment, and to assess the safety status of the structure based on these parameters; specifically, it includes: By deploying structural health monitoring sensors in key structural parts, the stress response, deformation trend and attitude change information of the main building structure during operation can be collected in real time to determine whether there is any over-limit behavior or trend risk in the current structural state, and to generate structural health assessment results. By deploying stress-strain monitoring nodes on key building components, multidimensional structural response data is continuously collected from each monitoring node, mainly including stress σ, strain ε, and tilt angle θ. , will the i-th The state of each sensor's stress-strain monitoring node at any time point t' is represented as a three-dimensional vector. ,in, The vector transpose is represented, and the whole matrix forms the structure-state matrix. Where N is the number of sensor nodes. Represents the real number field; For the j-th in the building Each structural unit determines a set of neighboring sensors based on its geometric location. Interpolation weights are set based on the spatial distance from the sensor to the structural unit. Thus, a simplified stress estimation model is constructed; at time t', the i-th... The j-th sensor Estimated stress values ​​of each structural unit Represented as:

[0057] Wherein, the weights satisfy And it is calculated using the inverse distance weighting method:

[0058] in, Indicates the i-th The sensor and the j-th Geometric distance of each structural unit; k This indicates traversing the entire sensor set. The mark, Indicates the kth The sensor and the j-th The geometric distance between structural elements. This method is simpler than the finite element method and is suitable for scenarios where sensor points are sparse and model information is limited in practical engineering deployments.

[0059] After obtaining the estimated stress values ​​for each structural unit, the corresponding allowable stress limits are obtained by referring to the material design specifications or structural service standards. Based on this, the safety margin factor of the structural unit is calculated:

[0060] like If the signal is positive, it indicates that the structural unit currently has a potential risk of exceeding structural limits, which will trigger an early warning and mark the number of the exceeding unit and its region. The lifecycle prediction unit is used to perform time-series modeling and trend inference of equipment operating status, generate health scores, and predict the remaining lifespan of the equipment. It supports time-series analysis and health trend prediction of equipment operating status, specifically including: During the building's operation, at fixed time intervals Collect operating status parameters of various key electromechanical equipment within the building to construct a state vector. Where e represents the device number, k0 is the dimension of the state variable monitored by the device, and a sliding window mechanism of length L is used to construct a time series segment vector of the device state:

[0061] Where, n Indicates from a point in time At the appointed time There are a total of L state point vectors; The time series segment vectors are processed through stacked Transformer layers (the Transformer layer is the basic module of the Transformer model, which achieves efficient feature interaction through self-attention mechanism) to obtain a deep representation of the entire sequence. , where d For the hidden layer dimension, This represents the feature extraction function composed of a deep network with transformer layers. The nth layer obtained by the transformer layer The feature vector of the layer; after further average pooling, we get:

[0062] in, This represents the vector result after average pooling of feature vectors from layers 1 to L, where i0 represents the index and i0 ; Finally, a regression mapping layer is used. : Output the predicted health score for the current time. :

[0063] in, This represents a regression mapping layer composed of a multilayer perceptron network, which reduces the vector dimension from d... Reduced to 1, The higher the value, the better the current health status of the device; when Less than the health warning threshold When the equipment is in a state of aging and degradation, an automatic warning will be issued, and a suggested maintenance time window will be generated.

[0064] like Figure 2 As shown, the present invention also provides a management method for an integrated building artificial intelligence platform, the method comprising: Step 1: Extract the semantic features and structured parameter information of the equipment from the equipment description text, match them with the equipment database, and generate the best matching equipment item and its reference price; Step 2: Extract the core points of the bidding documents, generate the content of the bid documents, and perform compliance verification and correction; Step 3: Based on the operating status and historical usage of each area of ​​the building, combined with the real-time load of the equipment, environmental changes and space usage, generate equipment adjustment strategies; Step 4: Collect environmental data for each area within the building, analyze trend changes and identify anomalies, and generate environmental adjustment suggestions and risk warnings; Step 5: Integrate multi-source heterogeneous data and present the operating status and indicator change trends in a visual manner; Step 6: Construct a knowledge graph for the architectural field to provide semantic retrieval and knowledge recommendation services; Step 7: Analyze the equipment operation indicators, calculate the operation and maintenance performance score, and generate an evaluation report; Step 8: Deploy a conversational AI interactive interface to enable intelligent guidance of office processes and proactive information push; Step 9: Construct a construction site Internet of Things (IoT) network, and combine image recognition and scheduling algorithms to achieve on-site status monitoring and intelligent resource matching. Step 10: Conduct continuous monitoring and lifespan management of the building structure and equipment status.

[0065] The above description is only a part of the embodiments of the present invention and does not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made based on the content of the present invention specification and drawings, or direct or indirect application in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. An integrated application platform for building artificial intelligence, characterized in that, include: The intelligent cost pricing module is used to extract semantic features and structured parameter information of equipment from equipment description text, match them with the equipment database, and generate the optimal matching equipment item and its reference price. The tender document generation module is used to extract the core points of the tender documents, generate the tender document content, and perform compliance verification and correction. The intelligent operation optimization module is used to generate equipment adjustment strategies based on the operating status and historical usage of various areas of the building, combined with the real-time load of the equipment, environmental changes and space usage. The building environment monitoring and intelligent analysis module is used to collect environmental data from various areas within the building, analyze trend changes and identify anomalies, and generate environmental adjustment suggestions and risk warnings. The cross-system data integration and visualization module is used to integrate multi-source heterogeneous data and present the operating status and indicator change trends in a visual way; The architectural knowledge sharing and semantic support module is used to construct a knowledge graph in the architectural field and provide semantic retrieval and knowledge recommendation services. The building operation evaluation module is used to comprehensively analyze equipment operation indicators, calculate operation and maintenance performance scores, and generate evaluation reports. The intelligent office assistant module is used to deploy a conversational AI interactive interface to enable intelligent guidance of office processes and proactive information push. The intelligent monitoring and scheduling module for construction projects is used to build an Internet of Things (IoT) network for construction sites. It combines image recognition and scheduling algorithms to achieve on-site status monitoring and intelligent resource matching. The building health monitoring module is used for continuous monitoring and lifespan management of building structures and equipment.

2. The integrated application platform for building artificial intelligence as described in claim 1, characterized in that, The intelligent cost pricing module includes: The text parsing unit is used during the project preparation phase to parse the equipment description text in the bill of quantities. This equipment description text includes: equipment name, functional requirements, specifications, reference price, and reference brand; specifically, it includes: Let the bill of quantities for construction projects consist of several project items, denoted as set. Among them, the i1th project entry Includes a device description text written in natural language. This is used to represent the equipment name, functional requirements, specifications, reference price, and reference brand; where n1 represents the total number of project items and n1 is a positive integer, and i1 represents the index and i1 is a positive integer. ; The data extraction unit is used to extract semantic features and structured parameter information of the device from the device description text; specifically, it includes: The device description text is first processed in two ways: firstly, through a large language model encoding function. Extract its semantic features and generate semantic vectors On the other hand, the structured information extraction function is called. From the device description text Key technical parameters are extracted from the data and a structured parameter vector is generated. ; The data matching unit is used to match the extracted semantic features and structured parameter information of the device with the semantic features and structured parameter information of the standard device items stored in the device database. If the match is successful, the result is output to the result output unit; if the match fails, no processing is performed. Specifically, it includes: Device database is represented as a set The j1st device item Includes its standardized text description Structured parameter vectors and price Where m1 represents the total number of device items and m1 is a positive integer, j1 represents the index and j Standardized text descriptions of equipment items Perform semantic encoding to obtain semantic vectors. ; for quantifying engineering entries With equipment items The degree of matching between them is defined by a scoring function that combines semantic similarity and parameter consistency: in, These are the weighting coefficients for semantic and parameter matching. This represents a parametric inference function used for evaluation. and The degree of matching between them; The result output unit is used to generate the optimal matching equipment item and its corresponding reference price; specifically, it includes: The device with the highest score among all candidate devices is selected as the optimal matching device for the project entry. And return its reference price As a result of price arbitrage; Set acceptance threshold Only if the matching score satisfies Only then are the pricing results adopted, and the final output is a set of structured pricing results: 。 3. The integrated application platform for building artificial intelligence as described in claim 1, characterized in that, The tender document generation module includes: The content extraction and encoding unit is used during the bidding stage to extract summaries and vectorize the content of the bidding documents; specifically, it includes: The received tender document text content is denoted as the text set. It includes the background of the bidding project, technical requirements, and scoring details, where k represents the total number of bidding document texts and k is a positive integer; through the extraction function Summarize and segment key points from each section of the tender document text to obtain several core summary units. Where u represents the total number of core abstract units and u is a positive integer, and the j2-th core abstract unit is... Then encoded into vectors by the large language model. They are uniformly stored in the bidding content vector database, where j2 represents the index and j2 ; The content generation unit is used to drive the large model to generate bid document content chapter by chapter according to a predefined bid document structure. When generating each chapter, it references the corresponding bid content summary, industry standard fragments from the enterprise knowledge base, and summaries of previously generated content; specifically, it includes: Based on the predefined tender document structure template Where m2 represents the total number of chapters and m2 is a positive integer, the tender document content is generated segment by segment according to chapter-level granularity; let the currently generated chapter be... Where r represents the index and r Then the large language model generates the text for this chapter. When calling the method: in, This represents the subset of tender summary units that have the highest semantic relevance to this chapter. For the retrieved set of industry knowledge, Indicates that the chapter was generated previously. The summary representation constitutes the generation context, and LLM stands for Large Language Model; The review sub-unit is used to perform a reasonableness check on the content of the generated bid documents according to the scoring rules and compliance requirements. If problems are found, it will provide feedback and trigger a chapter-level regeneration process; specifically, it includes: After each chapter is generated, the auxiliary review module is called to review the chapter text. A reasonableness review is conducted, which is based on a large language model comparing the original tender requirements and scoring rules, and executing a content compliance check function. Determine whether the chapter meets the corresponding requirements. ,in, This represents the set of rationality rules that need to be verified. If logical conflicts, omissions, or deviations in expression are found, the specific problem points will be located and marked for user reference or feedback to the large language model for correction. During the generation and review process, automatic iterative optimization is supported, that is, the feedback from the auxiliary review module is used as a constraint prompt to guide the large language model to generate new chapter versions. The review process is repeated multiple times until the generated content meets the preset standards of reasonableness and compliance, and the final output is a complete tender document organized in a structured and chapter-by-chapter format. .

4. The integrated application platform for building artificial intelligence as described in claim 1, characterized in that, The intelligent operation optimization module for the equipment specifically includes: Suppose that there exists a set of regions within the building. Where m3 represents the total number of regions and m3 is a positive integer, and the set of device types. Where n2 represents the total number of device types and is a positive integer, at time t, the operating state of device type d in region z is: The generated strategy recommendation value is The adjustment command is The strategy recommendation calculation formula is as follows: in, This represents the real-time load of device type d within region z at time t. This indicates the maximum rated load of device type d in region z; This represents the historical energy consumption data of device type d in region z and at time t. Let be the long-term average energy consumption of equipment of type d in region z; The usage intensity factor of region z at time t is calculated by combining the dimensions of personnel density, spatial reservation status and environmental deviation. These are weighting coefficients, and When the strategy recommendation value When the threshold range is different, different control behaviors of the device are triggered, and adjustment commands are generated and issued. This enables equipment adjustment.

5. The integrated application platform for building artificial intelligence as described in claim 1, characterized in that, The building environment monitoring and intelligent analysis module includes: An environmental data acquisition unit is used to acquire environmental data during the building operation phase by using various sensor devices deployed within the building. The environmental data includes: indoor and outdoor temperature, relative humidity, PM2.5 concentration, carbon dioxide concentration, and light intensity. The environmental prediction unit is used to predict and analyze environmental change trends and identify abnormal states in real time based on environmental data by constructing a prediction algorithm and anomaly detection mechanism based on time series modeling, and to provide environmental regulation suggestions and operational risk warnings. The cross-system data integration and visualization module includes: The data fusion unit is used to acquire multi-source heterogeneous data from multiple systems within the building during its operational cycle by setting up data acquisition and access interfaces. These systems include a building energy consumption metering system, an environmental monitoring system, and a personnel behavior system. The building energy consumption metering system collects data on the electricity and water / gas consumption of air conditioning, elevators, and lighting equipment. The environmental monitoring system collects environmental data for each area. The personnel behavior system acquires dynamic behavioral information on the space occupancy status, entry / exit frequency, and personnel density of each functional area. The data foundation unit is used to construct a unified data structure model based on the data access results of multiple systems, so as to realize the standardized storage and integration of data from different sources, frequencies and formats in the platform, forming a unified data foundation. The visualization unit is used to visualize the equipment operation status, environmental level and personnel activity in different areas of the building at different times through a graphical visualization interface based on a data base. The graphical visualization interface supports interactive switching by time dimension, spatial dimension and equipment dimension, and provides a multi-layer linkage mode to realize real-time perception of the overall operation status of the building and historical trend analysis.

6. The integrated application platform for building artificial intelligence as described in claim 1, characterized in that, The architectural knowledge sharing and semantic support module includes: Graph construction units are used to build a knowledge graph of the construction domain that covers multiple entities and relationships related to component types, construction techniques, operational experience, and equipment maintenance; specifically including: Access to multi-source documents and structured data from the project implementation process, including construction drawings, process specifications, equipment operation manuals, operation and maintenance logs, and acceptance specifications; extract key entities from the multi-source documents and structured data, including component types, construction processes, operating experience, and equipment maintenance events; identify the logical relationships between key entities; and construct a knowledge graph in the construction field. The knowledge graph in the construction field expresses the semantic connections between building components, processes, equipment, and events in the form of a graph structure. It supports multi-level entity classification, multiple relationship types, and hierarchical logical organization, and is regularly supplemented with knowledge and semantic cleansing. The semantic service unit supports semantic retrieval, recommendation, and question-answering services, and serves as the underlying knowledge enhancement interface, providing semantic support and contextual supplementation to the platform; specifically, it includes: It provides semantic retrieval services to support user queries using natural language or keywords, returning relevant component information, process flow, common faults, and solutions; it also supports question-and-answer services and intelligent recommendation services based on knowledge graphs in the construction field to help users obtain targeted knowledge content; as an underlying knowledge enhancement interface, it provides semantic enhancement and contextual supplementation support for the platform's intelligent cost pricing module, tender document generation module, and intelligent equipment operation optimization module.

7. The integrated application platform for building artificial intelligence as described in claim 1, characterized in that, The building operation evaluation module includes: The indicator system establishment unit is used to take the multi-dimensional operation indicators collected by the platform as input, perform structured summarization and cleaning of the multi-dimensional operation indicators, and establish a unified evaluation indicator system; the multi-dimensional operation indicators include: equipment maintenance frequency, response timeliness, environmental regulation effectiveness, user satisfaction indicators, and energy consumption trends. The scoring calculation unit is used to set the weighting factors of the evaluation indicators according to the building operation and maintenance management objectives, and to calculate the overall operation and maintenance performance score of the building using a weighted scoring model, thereby forming a phased operation and maintenance score value. The report generation unit is used to generate a visualized phased operation and maintenance evaluation report based on the phased operation and maintenance score. The phased operation and maintenance evaluation report includes: the score distribution of various indicators, key issue prompts and suggested optimization directions, to help the management to have a global understanding of the current building operation status and adjust strategies.

8. The integrated application platform for building artificial intelligence as described in claim 1, characterized in that, The intelligent office assistant module includes: The office task scheduling unit is used to deploy a conversational AI interactive interface, supporting both voice and text input. Users can make requests to the intelligent office assistant module through natural language. The intelligent office assistant module uses semantic recognition and intent parsing to schedule and respond to office tasks. The office tasks include: expense approval process, seal application process, meeting room reservation, work order circulation and notification inquiry. Users can perform guided operations across processes and multiple steps through dialogue. The knowledge query unit is used to connect the intelligent office assistant module to the enterprise's internal knowledge base, including institutional documents, policy compilations and business guidelines. It supports semantic-level retrieval and intelligent extraction of document fragments. Users can send query requests to the intelligent office assistant module for management processes, approval requirements and operating procedures. The intelligent office assistant module automatically matches relevant content and outputs a summary of key points. The early warning push unit is used to build personalized demand models based on user behavior data and proactively push project risk warnings, policy and regulation updates, and schedule reminders.

9. The integrated application platform for building artificial intelligence as described in claim 1, characterized in that, The intelligent monitoring and scheduling module for the project includes: The network monitoring unit is used to collect key elements of the construction site in real time by deploying an Internet of Things monitoring network and combining personnel positioning devices, equipment sensing terminals and material identification tags. The key elements include the trajectory of construction personnel entering and leaving the area, the operating parameters of key equipment, and the arrival time and batch information of building materials. The digital twin unit is used to construct a three-dimensional virtual mapping scene of the construction site based on the key elements and combined with digital twin technology, so as to realize the dynamic visualization and monitoring of the construction status. The quality and safety assessment unit is used to connect to the image acquisition equipment and environmental sensors deployed on site. The image acquisition equipment automatically identifies process quality defects based on image recognition algorithms, and judges the level of safety hazards on site by combining sensor data collected by environmental sensors. When an abnormality is identified, a rectification notice is automatically generated, pushed to the relevant responsible personnel, and the closed-loop processing progress is tracked, recording the rectification completion status and time nodes. The resource scheduling unit integrates labor and material scheduling mechanisms. By jointly modeling the distribution of on-site labor, the pace of construction tasks, and the progress of material consumption, it uses optimization scheduling algorithms to achieve intelligent matching and resource allocation of labor and materials. Based on the prediction results, it proactively recommends scheduling schemes, including adjusting the input of construction teams and optimizing the batch and time window of material transportation.

10. The integrated application platform for building artificial intelligence as described in claim 1, characterized in that, The building health monitoring module includes: The structural monitoring unit is used to collect key operating parameters through structural stress and strain sensors and visual inspection equipment, and to assess the safety status of the structure based on these parameters; specifically, it includes: By deploying structural health monitoring sensors in key structural parts, the stress response, deformation trend and attitude change information of the main building structure during operation can be collected in real time to determine whether there is any over-limit behavior or trend risk in the current structural state, and to generate structural health assessment results. By deploying stress-strain monitoring nodes on key building components, multidimensional structural response data is continuously collected from each monitoring node, mainly including stress σ, strain ε, and tilt angle θ. , will the i-th The state of each sensor's stress-strain monitoring node at any time point t' is represented as a three-dimensional vector. ,in, The vector transpose is represented, and the whole matrix forms the structure-state matrix. Where N is the number of sensor nodes. Represents the real number field; For the j-th in the building Each structural unit determines a set of neighboring sensors based on its geometric location. Interpolation weights are set based on the spatial distance from the sensor to the structural unit. Thus, a simplified stress estimation model is constructed; at time t', the i-th... The j-th sensor Estimated stress values ​​of each structural unit Represented as: Wherein, the weights satisfy And it is calculated using the inverse distance weighting method: in, Indicates the i-th The sensor and the j-th Geometric distance of each structural unit; k This indicates traversing the entire sensor set. The mark, Indicates the kth The sensor and the j-th Geometric distance of each structural unit; After obtaining the estimated stress values ​​for each structural unit, the corresponding allowable stress limits are obtained by referring to the material design specifications or structural service standards. Based on this, the safety margin factor of the structural unit is calculated: like If the signal is positive, it indicates that the structural unit currently has a potential risk of exceeding structural limits, which will trigger an early warning and mark the number of the exceeding unit and its region. The lifecycle prediction unit is used to perform time-series modeling and trend inference of equipment operating status, generate health scores, and predict the remaining lifespan of the equipment. It supports time-series analysis and health trend prediction of equipment operating status, specifically including: During the building's operation, at fixed time intervals Collect operating status parameters of various key electromechanical equipment within the building to construct a state vector. Where e represents the device number, k0 is the dimension of the state variable monitored by the device, and a sliding window mechanism of length L is used to construct a time series segment vector of the device state: Where, n Indicates from a point in time At the appointed time There are a total of L state point vectors; After time series segment vectors are processed by stacked Transformer layers, a deep representation of the entire sequence is obtained. , where d For the hidden layer dimension, This represents the feature extraction function composed of a deep network with transformer layers. The nth layer obtained by the transformer layer The feature vector of the layer; after further average pooling, we get: in, This represents the vector result after average pooling of feature vectors from layers 1 to L, where i0 represents the index and i0 ; Finally, a regression mapping layer is used. : Output the predicted health score for the current time. : in, This represents a regression mapping layer composed of a multilayer perceptron network, which reduces the vector dimension from d... Reduced to 1, The higher the value, the better the current health status of the device; when Less than the health warning threshold When the equipment is in a state of aging and degradation, an automatic warning will be issued, and a suggested maintenance time window will be generated.

Images