An intelligent management method and system for building information based on BIM technology

Abstract

The present invention relates to a method and system for intelligent management of building information based on BIM technology. The method includes: intelligently parsing user requirements into a standardized parameter set and generating an initial BIM model in combination with a rule engine; automatically adjusting model parameters through parameter matching and an AI rule engine to output a compliant BIM model; performing simulation analysis on the compliant model for structure, energy consumption, and space utilization rate to generate a performance report containing normalized deviation values; using the deviation values as the input of the state of the deep Q network, iteratively optimizing the parameters to meet the standards through a predefined action space, and outputting a performance-optimized BIM model; obtaining construction and operation and maintenance data in real time, performing digital twin comparison with the optimized model, triggering secondary optimization of the deep Q network and generating a dynamically updated BIM data stream, synchronizing all stages of the full life cycle, so as to realize the closed-loop management of dynamic control of the full life cycle of the building, and improve the design efficiency, engineering compliance, and cross-stage collaboration ability.

Description

Technical Field

[0001] The present invention relates to the field of building information technology, and particularly to a building information intelligent management method and system based on BIM technology. Background Art

[0002] With the acceleration of global urbanization and infrastructure construction, the construction industry faces both opportunities and challenges. The scale of projects has become larger and the functions have become more complex. Society's expectations for building quality, sustainability, and operational efficiency have increased, and the traditional working mode has become difficult to adapt to. For example, when conducting demand docking, due to the lack of systematic and intelligent transformation tools and processes, the translation of user requirements is prone to errors, resulting in the design not meeting the owner's expectations and requiring repeated modifications, which not only prolongs the project cycle but also wastes resources, restricting the allocation of industry resources.

[0003] At the same time, industry technical standards and specifications are constantly updated, and design compliance has become a difficult problem. It is difficult to integrate the specifications into the design process, and designers need to make manual adjustments, which affects efficiency and is difficult to ensure compliance, bringing potential hazards to project implementation. In building performance evaluation, traditional simulation technologies are lagging and one-sided, and it is difficult to accurately evaluate core indicators such as structural stability, energy consumption, and space utilization rate. There is a lack of data basis for design optimization, and it is impossible to effectively improve building performance. In addition, during the construction and operation and maintenance stages, the information interaction among the participating parties is not smooth, and there is also a lack of a real-time feedback mechanism between the physical building data and the design model, making it difficult to optimize according to on-site and operation and maintenance data, destroying the coherence of information throughout the life cycle, and restricting the improvement of industry efficiency and sustainable development.

[0004] In the existing technology, natural language processing (NLP) demand conversion, dynamic specification adaptation, performance simulation, and digital twin technologies are mostly applied separately. For example, CN118012894A discloses a BIM parameter generation method based on NLP, but it does not solve the problems of real-time update of specifications and dynamic compliance of models. In this context, constructing a BIM technology system integrating demand conversion, model optimization, performance simulation, and information interaction is an effective attempt for the construction industry to break through the existing development bottlenecks and transform and upgrade to a higher level. Summary of the Invention

[0005] The main object of the present invention is to provide a building information intelligent management method and system based on BIM technology to achieve the full-process intelligent control of building information from intelligent parsing of user requirements, real-time compliance of design specifications, multi-objective performance independent optimization to dynamic closed-loop management throughout the life cycle, and to improve the design efficiency, engineering compliance, and resource utilization efficiency.

[0006] To achieve the above object, the present invention provides a building information intelligent management method based on BIM technology, including the following steps:

[0007] Convert user requirements into a standardized parameter set through a preset mapping table that supports online updates and natural language processing technology, and generate an initial BIM model in combination with a rule engine;

[0008] Based on a dynamically updated design specification database, automatically adjust the parameters of the initial BIM model through parameter matching and an AI rule engine, and output a specification-compliant BIM model;

[0009] Perform simulation calculations on the structural stability, energy consumption distribution, and space utilization rate of the specification-compliant BIM model to generate a performance analysis report;

[0010] Use the normalized deviation value in the performance analysis report as the state input of the deep Q network, output an optimization decision through a predefined action space, iteratively adjust the parameters until the threshold is met, and output a performance-optimized BIM model;

[0011] Obtain physical building data during the construction and operation and maintenance phases in real time through the Internet of Things interface, perform digital twin comparison with the performance-optimized BIM model, calculate parameter deviations, trigger secondary optimization of the deep Q network, generate a dynamically updated BIM data stream, and synchronize it to all phases of the entire life cycle.

[0012] Furthermore, the step of converting user requirements into a standardized parameter set through a preset mapping table that supports online updates and natural language processing technology, and generating an initial BIM model in combination with a rule engine includes:

[0013] Parse the natural language requirements input by the user, and extract keywords and constraints;

[0014] Map the keywords to BIM parameter types and value ranges through a preset mapping table, and the mapping table contains the triple relationship of user requirement semantics, BIM parameters, and associated components;

[0015] If there is no matching item in the mapping table, trigger an online update request and expand the mapping table entries based on user feedback;

[0016] Use a rule engine to generate the geometric structure and attribute labels of the initial BIM model according to parameter priorities and spatial topology rules.

[0017] Furthermore, the step of automatically adjusting the parameters of the initial BIM model based on a dynamically updated design specification database through parameter matching and an AI rule engine, and outputting a specification-compliant BIM model includes:

[0018] Grab the latest building specification text from an authoritative specification release platform through a data interface, extract the specification parameter requirements, and structurally store them in the design specification database;

[0019] Filter applicable clause sets from the design specification database based on the component types and spatial attributes of the initial BIM model;

[0020] Based on the selected clause sets, compare model parameters with specification thresholds item by item through a parameter matching algorithm to identify non-compliant item parameters;

[0021] Call a preset replacement strategy through the AI rule engine to automatically correct non-compliant parameters and generate a specification-compliant BIM model with an adjustment log.

[0022] Furthermore, the steps of performing simulation calculations on the structural stability, energy consumption distribution, and space utilization rate of the specification-compliant BIM model to generate a performance analysis report include:

[0023] Conduct structural stability analysis through the finite element method, calculate the component stress ratio and overall displacement angle in the specification-compliant BIM model, and identify the over-limit areas;

[0024] Conduct energy consumption simulation based on the specification-compliant BIM model, perform hourly load calculations throughout the year through the EnergyPlus engine, and output the peak heating and cooling loads and energy consumption density distribution;

[0025] Evaluate the space utilization rate of the specification-compliant BIM model, analyze the efficiency of the pedestrian flow path using space syntax, and calculate the space utilization rate index;

[0026] Normalize the over-limit areas, the peak heating and cooling loads, the energy consumption density distribution, and the space utilization rate index into deviation rates, and generate a performance analysis report including component correlation markings.

[0027] Furthermore, the steps of using the normalized deviation values in the performance analysis report as the state input of the deep Q-network, outputting optimization decisions through a predefined action space, iteratively adjusting parameters until the threshold is met, and outputting a performance-optimized BIM model include:

[0028] Classify and encode the structural stability deviation, energy consumption deviation, and space utilization rate deviation in the performance analysis report by parameter type, and use them as the state input vector of the deep reinforcement learning network;

[0029] Define the output actions of the deep reinforcement learning network as a set of BIM model parameter adjustment operations, including component geometric dimension adjustment, material property replacement, and space topology relationship optimization;

[0030] Configure the network training reward mechanism to give positive rewards when the model parameters meet the performance threshold, give partial rewards when the optimized deviation decreases, and give penalties when structural or energy consumption constraints are violated;

[0031] Optimize actions by exploring-exploiting balance strategy, iteratively adjust BIM model parameters and re-perform performance simulation until the model parameters reach the convergence condition after continuous optimization for multiple times.

[0032] Further, the step of obtaining real-time physical building data during the construction and operation and maintenance phases through the Internet of Things interface, performing digital twin comparison with the performance-optimized BIM model, calculating parameter deviations and triggering secondary optimization of the deep Q network, generating dynamically updated BIM data stream, and synchronizing it to all phases of the whole life cycle includes:

[0033] Obtain the three-dimensional point cloud data of the actual construction components through laser scanning, perform position deviation analysis with the theoretical coordinates of the corresponding components in the performance-optimized BIM model, and calculate the construction deviation ratio;

[0034] Collect real-time operation data of building equipment, including air-conditioning energy consumption, lighting power, and elevator usage frequency, and perform energy consumption and performance deviation analysis with the predicted values of the performance-optimized BIM model;

[0035] When the construction deviation ratio or the operation and maintenance energy consumption deviation ratio exceeds the preset threshold, automatically trigger the secondary optimization of the deep Q network, generate a dynamically updated BIM data stream including parameter adjustment records, and the BIM data stream is marked with version number, change time, and impact scope label;

[0036] Distribute the dynamically updated BIM data stream to the design, construction, and operation and maintenance ends, respectively drive design review, construction plan adjustment, and equipment parameter calibration, and perform whole life cycle management.

[0037] Further, based on the BIM data stream, monitor the model operation performance through a real-time feedback system. If there are deviations, trigger the deep Q network to optimize again, dynamically update the BIM model and iterate cyclically, including:

[0038] Deploy performance monitoring agents during the operation and maintenance phase, collect real-time building operation data and calculate key indicators;

[0039] If the indicators exceed 20% of the predicted values of the BIM model for 3 consecutive monitoring cycles, it is determined as a performance deviation;

[0040] Trigger the deep Q network to re-optimize the model parameters with the current operation and maintenance data as the input;

[0041] Mark the updated BIM model version and synchronize it to the whole life cycle data stream to form a closed-loop management.

[0042] The present invention also provides an intelligent building information management system based on BIM technology, including:

[0043] A requirements modeling unit, which is used to convert user requirements into a standardized parameter set through a preset mapping table that supports online updates and natural language processing technology, and generate an initial BIM model in combination with a rule engine;

[0044] A specification adaptation unit, which is used to automatically adjust the parameters of the initial BIM model based on a dynamically updated design specification database through parameter matching and an AI rule engine, and output a specification-compliant BIM model;

[0045] A performance simulation unit, which is used to perform simulation calculations on the structural stability, energy consumption distribution, and space utilization rate of the specification-compliant BIM model, and generate a performance analysis report;

[0046] A model optimization unit, which is used to use the normalized deviation value in the performance analysis report as the state input of a deep Q network, output an optimization decision through a predefined action space, iteratively adjust the parameters until the threshold is met, and output a performance-optimized BIM model;

[0047] A twin update unit, which is used to obtain physical building data in the construction and operation and maintenance phases in real time through an Internet of Things interface, perform digital twin comparison with the performance-optimized BIM model, calculate parameter deviations, trigger secondary optimization of the deep Q network, generate a dynamically updated BIM data stream, and synchronize it to all stages of the full life cycle.

[0048] The present invention also provides a computer device, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps of the above-mentioned intelligent building information management method based on BIM technology are implemented.

[0049] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned intelligent building information management method based on BIM technology are implemented.

[0050] The building information intelligent management method and system based on BIM technology provided by the present invention have the following beneficial effects: Through natural language processing and dynamic mapping tables, the present invention realizes the accurate conversion of unstructured user requirements into standardized BIM parameters, solves the problems of low efficiency and easy deviation in traditional manual interpretation, and improves the accuracy and efficiency of modeling; Based on a dynamically updated specification database and an AI rule engine, it can identify and correct model parameters in real time to ensure that the design complies with the latest standards throughout the process and avoid the risk of omission in manual review. By using a deep reinforcement learning network to replace manual trial and error and a multi-objective collaborative optimization mechanism, it balances energy consumption and space efficiency on the premise of ensuring structural safety and shortens the optimization cycle. The present invention integrates NLP requirement parsing, AI rule engine specification adaptation, multi-objective performance simulation, deep reinforcement learning (DQN) optimization, and digital twin feedback into a closed-loop system. For example, after the laser scan data in the construction stage triggers DQN secondary optimization, the updated BIM model directly drives the calibration of the operation and maintenance equipment parameters, realizing seamless data linkage of "design - construction - operation and maintenance" and solving the defect that traditional single-point optimization cannot adapt to dynamic changes. By integrating unstructured user requirements (natural language), structured specification libraries (dynamically captured), physical building data (Internet of Things), and reinforcement learning strategies, an adaptive decision-making network is formed. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a schematic flowchart of a building information intelligent management method based on BIM technology in an embodiment of the present invention;

[0052] Figure 2 is a structural block diagram of a building information intelligent management system based on BIM technology in an embodiment of the present invention;

[0053] Figure 3 is a schematic structural block diagram of a computer device in an embodiment of the present invention.

[0054] The realization, functional features, and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] In order to make the object, technical solution, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0056] Refer to Figure 1 , which is a schematic flowchart of a building information intelligent management method based on BIM technology proposed by the present invention, including the following steps:

[0057] S1. Convert the user requirements into a standardized parameter set through a preset mapping table that supports online updates and natural language processing technology, and generate an initial BIM model in combination with a rule engine;

[0058] S2. Based on a dynamically updated design specification database, automatically adjust the parameters of the initial BIM model through parameter matching and an AI rule engine, and output a specification-compliant BIM model;

[0059] S3. Perform simulation calculations on the structural stability, energy consumption distribution, and space utilization rate of the specification-compliant BIM model to generate a performance analysis report;

[0060] S4. Use the normalized deviation value in the performance analysis report as the state input of the deep Q network, output an optimization decision through a predefined action space, iteratively adjust the parameters until the threshold is met, and output a performance-optimized BIM model;

[0061] S5. Real-time obtain physical building data during the construction and operation and maintenance phases through an IoT interface, perform digital twin comparison with the performance-optimized BIM model, calculate parameter deviations and trigger secondary optimization of the deep Q network, generate a dynamically updated BIM data stream, and synchronize it to all phases of the full life cycle.

[0062] As described in the above step S1, parse the natural language requirements input by the user, and extract keywords and constraints. Use natural language processing techniques (such as fine-tuning the BERT-Base model, with a training set containing 100,000 pieces of text in the construction field, such as specifications, tender documents, etc., the loss function is cross-entropy loss, the learning rate is 2e-5, and the BiLSTM-CRF model is combined to achieve entity recognition of component types and material attributes) to parse the unstructured requirements input by the user. When using the BERT model to parse the unstructured requirements input by the user, fine-tune the BERT model with a domain-specific corpus to improve the understanding ability of complex sentence patterns and professional terms. The domain-specific corpus collects a large number of text materials related to the construction industry, including design specifications, project documents, etc., which can enable the BERT model to more accurately extract keywords and constraints. For example, for complex requirement descriptions such as "This building should meet the three-star green building standard, and adopt prefabricated building technology, and the assembly rate of prefabricated components is not less than 70%", the fine-tuned BERT model can accurately identify key information such as "three-star green building standard", "prefabricated building technology", and "prefabricated component assembly rate ≥ 70%". At the same time, introduce semantic role labeling (SRL) technology to further clarify the semantic relationships between entities in the requirements. For example, in "High-rise office buildings need to meet the first-level fire protection standard, and the area of each fire compartment does not exceed 1500 square meters", through SRL technology, it can be marked that "high-rise office buildings" are the main body, and "first-level fire protection standard" and "the area of each fire compartment does not exceed 1500 square meters" are descriptions of the attributes and restrictions of the main body, helping the system to more deeply understand the user's requirements.

[0063] Map the keyword to the BIM parameter type and value range through a preset mapping table, where the mapping table contains the triple relationship of user requirement semantics, BIM parameters, and associated components; if there is no matching item in the mapping table, trigger an online update request and expand the mapping table entries based on user feedback. Among them, the preset mapping table is stored in the form of a knowledge graph and also adopts a distributed storage technology to disperse the mapping table data on multiple nodes, improving the data reading speed and the fault tolerance of the system. When the system triggers the online update mechanism, to ensure the accuracy and standardization of the newly added entries, an audit process is introduced. After the administrator or user submits the newly added entries, the system automatically performs preliminary format and logic verification, and then it is audited by domain experts. Only the entries that pass the audit can be officially added to the mapping table. For example, for the requirements related to new green building materials, if the user submits a new material parameter and the corresponding BIM parameter mapping relationship, the auditor will check its rationality and accuracy based on industry standards and actual engineering experience to avoid incorrect data from entering the mapping table and affecting subsequent modeling work. Among them, the dynamic expansion process of the mapping table triples includes that when the user inputs an unmatched requirement (such as "carbon fiber reinforced concrete exterior wall"), the system extracts the keywords "carbon fiber" and "exterior wall" through the fine-tuned BERT model, calls the Wikidata API to verify that the material compressive strength ≥ 50 MPa and the fire resistance rating is A1, and associates the component type "Structural Wall". Subsequently, based on the Neo4j graph database, perform a consistency check. If there is no conflict with the existing entries (such as no conflict with the fire protection code), then add a new triple (requirement semantics: high-strength exterior wall, BIM parameter: Material.Strength = 55 MPa, associated component: Structural Wall), mark the version number as V2.1.3 and take effect immediately. The historical update records are traceable and support version rollback.

[0064] Use a rules engine to generate the geometric structure and attribute tags of the initial BIM model according to parameter priorities and spatial topology rules. When calling the Revit API to generate the initial BIM model based on parameter priorities and spatial topology rules, a conflict resolution algorithm is adopted to better handle conflicts between parameters. When the settings of multiple parameters conflict with each other, the algorithm processes them according to the preset priorities and logical rules. For example, when setting the floor height, it is necessary to meet the building function requirements (such as the net height requirements of office spaces) and consider the structural design specifications (such as the impact of beam height on floor net height). The conflict resolution algorithm will comprehensively weigh these factors, give priority to meeting the requirements of more important parameters, and at the same time adjust other relevant parameters to balance various needs as much as possible. In addition, the parameter priorities are dynamically adjusted. According to the specific requirements and actual situation of the project, users or system administrators are allowed to flexibly adjust the priority order of parameters. For example, in some projects with extremely high energy consumption requirements, the priorities of energy consumption efficiency-related parameters can be raised above those of other parts of the parameters to ensure that energy consumption issues are focused on during the initial modeling stage. Use the Drools rules engine to manage parameter priorities and spatial topology rules, and integrate with Revit through the Java API. For example, for the default priority of "structural safety > energy consumption efficiency > cost control", define a rule: when the design specification database is updated (such as adding a clause of "fire compartment area ≤ 1500㎡"), the system matches the thresholds in the specification text with the parameter types through regular expressions, automatically generates Drools rules and loads them into the engine; in the parameter conflict handling, fuzzy logic is introduced to evaluate multi-objective trade-offs. For example, when "floor net height" conflicts with "beam height", the algorithm calculates the satisfaction degree of each parameter based on the membership function and selects the solution with the highest comprehensive membership degree (such as giving priority to meeting the net height requirements and at the same time adjusting the cross-sectional shape of the beam to reduce the impact on the floor height).

[0065] As described in step S2 above, obtain the latest building specification text from the authoritative specification release platform through the data interface, extract the specification parameter requirements and store them in the design specification database in a structured manner. When obtaining the latest building specification text from the authoritative specification release platform through the data interface, multiple strategies are adopted to cope with the platform's anti-crawler mechanism. On the one hand, set a reasonable request interval time to simulate the browsing behavior of human users and avoid being blocked from accessing the IP due to frequent requests; on the other hand, use a proxy server pool to dynamically switch the IP address for data scraping. When extracting key parameters, use the named entity recognition (NER) technology in natural language processing and combine it with a professional dictionary in the construction field to improve the accuracy of parameter extraction. For some ambiguous or ambiguous specification content, accurate interpretation is carried out through semantic understanding and context analysis. For example, when interpreting the clause on "number of safety exits" in the fire protection specification, combine information such as different building types and building areas to accurately extract the corresponding quantity requirements.

[0066] Based on the component types and spatial attributes of the initial BIM model, screen the applicable clause sets from the design specification database; based on the screened applicable clause sets, compare the model parameters with the specification thresholds item by item through a parameter matching algorithm to identify non-compliant item parameters. When screening the applicable clause sets from the specification library according to the types and spatial attributes of BIM components, a classification model based on deep learning is introduced. The classification model takes information such as component types and spatial attributes as inputs and is trained with a large amount of specification clause data, and can quickly and accurately screen out the most relevant specification clauses. Compared with traditional rule-based screening methods, the deep learning model can better handle complex situations and implicit relationships. In terms of the parameter matching algorithm, a matching method based on cosine similarity is adopted and combined with the edit distance algorithm to perform multi-dimensional comparison of the model parameters and the specification thresholds. This can more precisely judge the similarity and differences between parameters and improve the recognition accuracy of non-compliant item parameters. For example, for the fire resistance limit parameter of a firewall, not only the specific value is compared, but also the unit, precision of the value and its relevance to other related parameters are considered to ensure accurate identification of non-compliant situations.

[0067] Call the preset replacement strategy through the AI rule engine to automatically correct non-compliant parameters and generate a specification-compliant BIM model with an adjustment log. When the AI rule engine calls the preset policy library to correct non-compliant parameters, case-based reasoning (CBR) technology is introduced. When encountering non-compliant parameters, the system first searches for similar problem cases in the case library and refers to historical solutions for correction. If there is no exactly matching case in the case library, a new solution is generated by combining rule reasoning and expert experience, and the new case is stored in the case library for future reference. At the same time, the preset policy library is expanded to add more correction strategies, such as component replacement, connection method adjustment, etc. For example, when it is found that the steel reinforcement of a certain beam does not meet the specification requirements, in addition to adjusting the quantity and specification of the steel reinforcement, the type of the beam can also be considered for replacement (such as replacing a rectangular beam with a T-shaped beam) or the connection method between the beam and other components can be optimized to meet the specification requirements.

[0068] As described in step S3 above, structural stability analysis is carried out by the finite element method to calculate the member stress ratio and the overall displacement angle in the specification-compliant BIM model, and the over-limit areas are identified. When using finite element analysis software (such as ANSYS) to calculate the member stress ratio and the overall displacement angle, refined modeling techniques are adopted. For key members and complex joints, more detailed finite element models are established, considering factors such as material nonlinearity, geometric nonlinearity, and contact nonlinearity to improve the accuracy of the analysis results. For example, when analyzing the bottom frame columns of a high-rise building, the mechanical behavior of the columns under vertical and horizontal loads is more realistically simulated through refined modeling to accurately calculate the stress ratio and displacement angle. At the same time, a reliability analysis method is introduced to evaluate the reliability of the structural stability analysis results. By considering uncertain factors such as material properties and load values, the failure probability of the structure at different reliability levels is calculated. For example, the stress ratio and displacement angle of a certain floor beam at a 95% reliability level are calculated to evaluate whether it meets the design requirements.

[0069] Based on the specification-compliant BIM model, energy consumption simulation is carried out, and the hourly load calculation for the whole year is performed through the EnergyPlus engine to output the peak values of heating and cooling loads and the energy consumption density distribution. When performing the hourly load calculation for the whole year based on the EnergyPlus engine, more influencing factors are considered to improve the accuracy of the energy consumption simulation. In addition to the building envelope structure and equipment operation parameters, the dynamic changes in meteorological data, personnel behavior patterns, and internal heat gain factors in the building (such as lighting and heat generation of electrical equipment) are also included. By establishing a personnel behavior model, the impact of personnel activities on building energy consumption at different time periods is simulated. For example, during office hours, personnel activities are frequent, lighting and air conditioning equipment are used more, and the energy consumption increases accordingly; while at night or on holidays, personnel activities decrease and the energy consumption decreases. In addition, the energy consumption simulation results are analyzed in depth to identify high-energy consumption areas, and the key factors affecting energy consumption are determined through sensitivity analysis. For example, by changing parameters such as the heat transfer coefficient and shading coefficient of the external windows, the degree of their impact on the overall building energy consumption is analyzed.

[0070] Evaluate the space utilization rate of the compliant BIM model against the specification, analyze the efficiency of the pedestrian flow path using space syntax, and calculate the space utilization rate index; normalize the out-of-limit area, the peak heating and cooling load, the energy consumption density distribution, and the space utilization rate index into deviation rates, and generate a performance analysis report including component correlation markings. When analyzing the efficiency of the pedestrian flow path using space syntax, combine virtual reality (VR) and augmented reality (AR) technologies to conduct a more intuitive analysis and evaluation of the space. By creating a virtual building space model, users can take a virtual walk to observe and experience the accessibility and line-of-sight integration of the space in real time. At the same time, use VR / AR technologies to collect the behavior data of users in the virtual space, such as walking paths and staying times, to further optimize the calculation method of the space utilization rate index. When proposing layout optimization suggestions, use intelligent optimization algorithms such as genetic algorithms to search for and optimize multiple layout schemes. The algorithm takes the space utilization rate index as the objective function, considers constraints such as building functions and structural limitations, and automatically generates the optimal layout scheme. For example, when optimizing the office area layout of an office building, the genetic algorithm can quickly search for the most reasonable room division and passage setting to improve the space utilization efficiency.

[0071] As described in step S4 above, classify and code the structural stability deviation, energy consumption deviation, and space utilization deviation in the performance analysis report according to parameter types, and use them as the state input vector of the deep reinforcement learning network. The deep Q network adopts a three-layer fully connected structure. The dimension of the input layer is 128 (corresponding to normalized parameters such as structural stability deviation, energy consumption deviation, and space utilization deviation), the number of hidden layer nodes is 512, 256, and 128, the activation function selects ReLU to enhance the non-linear modeling ability, the dimension of the output layer is 50 (corresponding to predefined actions such as component size adjustment and material replacement), and a linear activation function is used. During training, set the capacity of the experience replay buffer to 10,000 entries, adopt a priority sampling strategy based on TD error, the initial exploration rate ε = 0.9 and linearly decay to 0.1. The reward mechanism is specifically as follows: +10 for meeting the structural deviation standard, +5 for every 5% reduction in energy consumption, -20 for violating the structural constraint, and -5 for cost overrun. The training process includes steps such as initializing the network, ε-greedy action selection, experience storage, and batch parameter update. The convergence condition is that the total deviation change rate < 2% after 10 consecutive optimizations.

[0072] In the Deep Q-Network (DQN), the Double Q-Network and Prioritized Experience Replay techniques are adopted to improve the training efficiency and convergence speed of the network. The Double Q-Network reduces the overestimation problem of Q-value estimation by decoupling action selection and action evaluation, making the training more stable. Prioritized Experience Replay samples according to the importance of experience samples, preferentially selecting samples that contribute more to network training for learning, and accelerating the convergence speed of the network. Define the output action of the deep reinforcement learning network as a set of BIM model parameter adjustment operations, including component geometric dimension adjustment, material property replacement, and spatial topology relationship optimization. Further expand the predefined action space by adding more parameter adjustment operations. For example, in terms of material property replacement, in addition to replacing the outer window glass type and increasing the wall insulation layer thickness, operations such as replacing building facade materials and adjusting the thermal conductivity of insulation materials can also be considered; in terms of spatial topology optimization, actions such as re-planning the internal functional zoning of the building and adjusting the public space layout are added. At the same time, perform a hierarchical design on the action space. According to different optimization goals and difficulty levels, divide the actions into different levels. In the initial stage of optimization, preferentially select simple and easy-to-implement actions that have a greater impact on performance improvement; as the optimization progresses, gradually try more complex actions. For example, when optimizing the building structure performance, first try simple actions such as adjusting the cross-sectional dimensions of beams and columns. If the effect is not obvious, then consider complex actions such as adjusting the structural system.

[0073] Configure the network training reward mechanism to give positive rewards when the model parameters meet the performance threshold, partial rewards when the deviation after optimization decreases, and penalties when structural or energy consumption constraints are violated; select optimization actions through an exploration-exploitation balance strategy, iteratively adjust the BIM model parameters and re-perform performance simulations until the model parameters reach the convergence condition after continuous optimization for multiple times. Refine the design of the reward mechanism. In addition to considering whether the parameters meet the threshold and whether the deviation rate decreases, more factors are also incorporated. For example, consider the implementation cost of the optimization action, its impact on other performance indicators, and the timeliness of optimization. If an optimization action can reduce the energy consumption deviation rate but has too high an implementation cost or has an adverse impact on structural stability, then a lower reward or even a penalty is given. At the same time, set different reward weights according to different optimization stages. In the initial stage of optimization, encourage exploring more actions and give certain rewards for trying new actions; in the later stage of optimization, pay more attention to the quality of the optimization results and give higher rewards to actions that meet the performance threshold and significantly reduce the deviation rate. For example, in the first half of the optimization stage, set the reward value for exploring new actions to +0.3; in the second half, increase the reward value for meeting the performance threshold to +1.5, guiding the network to find the optimal solution faster.

[0074] As described in step S5 above, three-dimensional point cloud data of the actual construction components is obtained through laser scanning, and position deviation analysis is carried out with the theoretical coordinates of the corresponding components in the performance-optimized BIM model to calculate the construction deviation ratio. The operating data of building equipment, including air-conditioning energy consumption, lighting power, and elevator usage frequency, is collected in real time, and energy consumption and performance deviation analysis is carried out with the predicted values in the performance-optimized BIM model. When obtaining the on-site component point cloud data through laser scanning, multi-sensor fusion technology is adopted. By combining multiple sensors such as lidar and cameras, more comprehensive and accurate component information can be obtained. The position and shape of the component are measured by lidar, and the surface texture and color information of the component (defects such as cracks and corrosion on the component surface) are obtained by the camera. The combination of the two can more intuitively judge the state and quality of the component. When collecting the equipment operating data in real time, edge computing technology is adopted to perform preliminary processing and analysis of the data at the equipment end, and only the key and compressed data is transmitted to the cloud, reducing the data transmission volume and latency. For example, when collecting the operating data of air-conditioning equipment, indicators such as the energy consumption and COP value of the air conditioner are calculated in real time through edge computing, and the data is subjected to anomaly detection. When the data is abnormal or further analysis is required, the data is uploaded to the cloud to improve the efficiency and real-time performance of data processing.

[0075] When the construction deviation ratio or the operation and maintenance energy consumption deviation ratio exceeds the preset threshold, the secondary optimization of the deep Q network is automatically triggered to generate a dynamically updated BIM data stream containing parameter adjustment records. The BIM data stream is marked with a version number, a change time, and an impact range label. When performing digital twin comparison with the BIM model, image matching and feature recognition technologies based on deep learning are adopted to improve the comparison accuracy and efficiency. By training a convolutional neural network (CNN) model, feature extraction and matching are carried out on the point cloud data obtained by laser scanning and the three-dimensional model in the BIM model to calculate the position deviation and shape deviation. When triggering the DQN secondary optimization, model order reduction technology is adopted to simplify the complex BIM model, extract key parameters and features, and reduce the computational complexity and accelerate the optimization speed while ensuring the optimization accuracy.

[0076] Distribute the dynamically updated BIM data stream to the design, construction, and operation and maintenance ends, respectively driving design review, construction plan adjustment, and equipment parameter calibration for full life cycle management. When synchronizing the data stream throughout the life cycle, a combination of blockchain technology and distributed ledger is adopted. Each data update operation is recorded in the distributed ledger and encrypted through the encryption algorithm of the blockchain to prevent data tampering and forgery. And establish a data sharing and collaboration platform to achieve data interaction and collaborative work in each stage of design, construction, and operation and maintenance. For example, designers can view the data feedback from the construction and operation and maintenance stages in real time and adjust the design plan; construction personnel can adjust the construction plan and methods according to design changes and operation and maintenance requirements; operation and maintenance personnel can feedback the equipment operation data and fault information to the design and construction parties in real time to jointly solve problems. To improve the compatibility and openness of the system, standardized data interfaces and formats are adopted to support data interaction and integration between different software and systems.

[0077] In one embodiment, taking the comprehensive office building project in a small cultural and creative park as an example, natural language processing technology combining a word vector model (such as Word2Vec) and a BERT model is used to analyze user requirements. For requirements such as "the office building needs to have sufficient natural lighting, and the creative studio should have a flexible space layout", the Word2Vec first maps the words in the text to the vector space to capture the semantic similarity between words, and the BERT model further conducts in-depth semantic understanding of the sentence, accurately extracting keywords ("natural lighting", "flexible space layout") and constraint conditions (such as the proportion of daylighting area, requirements for space transformability). The grammatical relationships of each part in the requirements are clarified. For example, the "office building" is the main body, and "sufficient natural lighting" is the conditional limitation. An efficient in-memory database is used to store the preset mapping table. For the new requirement "using environmentally friendly soundproof materials for studio partitions", an online update mechanism is triggered. With the help of a knowledge graph expansion tool, the BIM parameters (such as the sound insulation coefficient, environmental protection level, thickness, etc.) and value ranges corresponding to the "environmentally friendly soundproof materials" are found. After review and confirmation, the new entry is added to the mapping table, and the relevant semantic associations are updated. The parametric-driven geometric modeling technology is used to call the Revit API to generate an initial BIM model. According to the building functions and spatial topology rules, multiple initial layout schemes that meet the basic requirements are automatically generated. For example, for the daylighting requirement, different combinations of window orientations, sizes, and positions are considered; for the flexible space layout, different studio partition layout schemes are generated. These schemes are screened by the NSGA-II multi-objective optimization algorithm, with the daylighting effect, spatial flexibility, and structural rationality as the optimization objectives to determine the optimal initial layout. Combining the parameter priorities (such as natural lighting > spatial flexibility > structural cost), the geometric structure (such as wall thickness, window size, floor height) and attribute labels (such as material sound insulation performance, fire protection grade) of the model are determined. The finally output initial BIM model in IFC format comes with a detailed design parameter description.

[0078] The anti-crawler mechanism is bypassed by using random request intervals and disguised request headers to obtain the latest specifications. The rule engine is used to identify common specification parameters (such as evacuation passage width, fire compartment area), and then the text classification model trained by an LSTM-based model is used to process clauses with complex semantics, such as the usage specifications of new environmentally friendly materials. According to the types, spatial attributes, and functional areas of BIM components, the decision tree algorithm is used to screen applicable specification clauses. In the parameter matching process, the grey relational analysis algorithm is introduced to handle the complex relationship between the model parameters and the specification thresholds to judge compliance. The adjusted specification-compliant BIM model uses blockchain technology to record the modification history, and each modification operation is recorded as a blockchain transaction, including information such as the modification time, modification content, operator, and basis clause. At the same time, a detailed adjustment report is generated to show the changes before and after the model adjustment in the form of a table and a 3D model comparison.

[0079] Structural stability analysis was conducted using ANSYS, with finite element substructuring methods employed for detailed simulation of key areas (such as stairwells and cantilever structures). Based on the overall model analysis, detailed submodels were constructed for key locations, taking into account material nonlinearities and complex boundary conditions. This included accurately calculating the stress distribution and displacement of stair steps and landings when simulating the stairwell structure. Energy consumption simulation was conducted using the EnergyPlus engine. Geographic Information System (GIS) information was used to obtain information about the surrounding terrain, buildings, and vegetation. A radiosity algorithm was used to simulate the distribution of solar radiation on the building surface and calculate heat gain to the building envelope. To simulate the impact of occupant behavior on energy consumption, a cluster analysis-based occupant pattern recognition method was employed. Based on historical occupant activity data, occupant density and equipment usage time were predicted for each mode to simulate the building's dynamic energy consumption. For example, during weekday office hours, office lighting and equipment energy consumption are higher. Space syntax is used to analyze pedestrian path efficiency. Combined with virtual reality (VR) technology, users can take a virtual walk in a virtual building model from a first-person perspective. The system records data such as users' walking paths, stop locations, and duration in real time to optimize the calculation of space utilization indicators.

[0080] During the deep Q network optimization process, the Rainbow algorithm is employed, integrating multiple techniques to improve network performance. The Rainbow algorithm combines dual Q networks, prioritized experience replay, multi-step learning, and distributed reinforcement learning techniques to reduce Q-value estimation bias and improve training efficiency and stability. In state space construction, in addition to structural stability bias, energy consumption bias, and space utilization bias, factors such as cost bias and construction cycle bias are also incorporated to comprehensively inform network decision-making. Action adjustments for intelligent control of building equipment and energy management strategies have been added, including intelligent dimming strategies for lighting systems (automatically adjusting brightness based on time periods and light levels) and optimized zoned control strategies for air conditioning systems (controlling different areas based on occupancy density and temperature requirements). A reward mechanism has been designed that comprehensively considers multiple factors. In addition to basic rewards for parameter compliance, deviation reduction, and non-compliance penalties, rewards are also added for the sustainability and innovation of optimization actions. Additional rewards are given to optimization actions that utilize energy-saving technologies and improve building sustainability without compromising other performance characteristics. Incentives are also given for actions that utilize solar photovoltaic power generation and reduce the building's carbon emissions.

[0081] During the construction and operation and maintenance phases, low-power and high-precision sensors are used in combination with narrowband Internet of Things (NB-IoT) technology for data collection and transmission. Temperature and humidity sensors, energy consumption sensors, displacement sensors, etc. are arranged in the building to collect physical building data in real time. On the edge computing device, lightweight machine learning models (such as decision trees for equipment status monitoring and K-means clustering for energy consumption anomaly detection) are used to perform real-time analysis and anomaly judgment on the data. After obtaining the three-dimensional point cloud data of the actual construction components through laser scanning, a fast registration algorithm based on feature point matching is used to compare with the BIM model to calculate the position deviation and shape deviation, trigger DQN secondary optimization, and combine the real-time feedback mechanism of digital twin technology to feed back the real-time status information of the physical building (such as actual displacement and temperature change) into the BIM model for dynamic synchronization between the model and the actual building. In specific implementation, it is found that there is a deviation between the actual position and the designed position of a certain wall. Through the real-time feedback mechanism, this information is transmitted to the BIM model, and DQN re-optimizes the layout and structure of the corresponding area according to the new information. A full-life-cycle data stream management platform is built using blockchain technology. Each data update operation is recorded as a block on the blockchain, including information such as the time, content, source, and target of the data update. Automatic synchronization and collaboration of data in different stages are achieved through smart contracts.

[0082] In this embodiment, the dynamic specification adaptation module captures and parses more than 97% of the newly released specification clauses in real time, and the design compliance review time is shortened from 32 hours of traditional manual verification to 9 hours; the DQN optimization module balances structural safety and energy consumption efficiency through the reward mechanism, compressing the optimization cycle from 2 weeks of manual trial and error to 48 hours, and verifying the non-obviousness of the technology combination through the synergistic effect of each step.

[0083] Refer to Figure 2 , which is the structural block diagram of the building information intelligent management system based on BIM technology in an embodiment of the present invention, including:

[0084] A requirements modeling unit for converting user requirements into a standardized parameter set through a preset mapping table that supports online updates and natural language processing technology, and generating an initial BIM model in combination with a rule engine;

[0085] A specification adaptation unit for automatically adjusting the parameters of the initial BIM model based on a dynamically updated design specification database through parameter matching and an AI rule engine, and outputting a specification-compliant BIM model;

[0086] A performance simulation unit for performing simulation calculations on the structural stability, energy consumption distribution, and space utilization rate of the specification-compliant BIM model, and generating a performance analysis report;

[0087] A model optimization unit, configured to use the normalized deviation value in the performance analysis report as the state input of the deep Q-network, output an optimization decision through a predefined action space, iteratively adjust the parameters until the threshold is met, and output a performance-optimized BIM model;

[0088] A twin update unit, configured to obtain real-time physical building data during the construction and operation and maintenance phases through the Internet of Things interface, perform digital twin comparison with the performance-optimized BIM model, calculate parameter deviations, trigger secondary optimization of the deep Q-network, generate a dynamically updated BIM data stream, and synchronize it to all phases of the entire life cycle.

[0089] For the specific implementation of each unit in the above device instance, please refer to that described in the above method embodiment, and details will not be elaborated here.

[0090] Refer to Figure 3 , in an embodiment of the present invention, a computer device is further provided. This computer device can be a server, and its internal structure can be as Figure 3 shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0091] Those skilled in the art can understand that Figure 3 the structure shown in

[0092] is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0093] In summary, through a preset mapping table that supports online updates and natural language processing technology, the present invention converts user requirements into a standardized parameter set, combines a rule engine to generate an initial BIM model; based on a dynamically updated design specification database, through parameter matching and an AI rule engine, automatically adjusts the parameters of the initial BIM model to output a specification-compliant BIM model; performs simulation calculations on the structural stability, energy consumption distribution, and space utilization rate of the specification-compliant BIM model to generate a performance analysis report; uses the normalized deviation value in the performance analysis report as the state input of the deep Q network, outputs an optimization decision through a predefined action space, iteratively adjusts the parameters until the threshold is met, and outputs a performance-optimized BIM model; obtains real-time physical building data during the construction and operation and maintenance phases through an Internet of Things interface, performs digital twin comparison with the performance-optimized BIM model, calculates parameter deviations, and triggers secondary optimization of the deep Q network to generate a dynamically updated BIM data stream, and synchronizes it to all stages of the entire life cycle, so as to achieve the purpose of intelligent control of the entire process from intelligent parsing of user requirements, real-time compliance with design specifications, autonomous optimization of multi-objective performance to dynamic closed-loop management of the entire life cycle, and improve design efficiency, engineering compliance, and resource utilization efficiency.

[0094] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or an external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0095] It should be noted that, in this document, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, apparatus, article or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, apparatus, article or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, apparatus, article or method including such element.

[0096] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall equally be included in the patent protection scope of the present invention.

Claims

1. An intelligent management method for building information based on BIM technology, characterized in that Including the following steps: Converting user requirements into a standardized parameter set through a preset mapping table supporting online update and natural language processing technology, and generating an initial BIM model in combination with a rule engine; Based on a dynamically updated design specification database, automatically adjusting the parameters of the initial BIM model through parameter matching and an AI rule engine, and outputting a specification-compliant BIM model; Performing simulation calculations on the structural stability, energy consumption distribution, and space utilization rate of the specification-compliant BIM model to generate a performance analysis report; Taking the normalized deviation value in the performance analysis report as the state input of the deep Q network, outputting an optimization decision through a predefined action space, iteratively adjusting the parameters until the threshold is met, and outputting a performance-optimized BIM model; Real-time obtaining physical building data in the construction and operation and maintenance stages through an Internet of Things interface, performing digital twin comparison with the performance-optimized BIM model, calculating parameter deviations, triggering secondary optimization of the deep Q network, generating a dynamically updated BIM data stream, and synchronizing it to all stages of the full life cycle; The step of real-time obtaining physical building data in the construction and operation and maintenance stages through an Internet of Things interface, performing digital twin comparison with the performance-optimized BIM model, calculating parameter deviations, triggering secondary optimization of the deep Q network, generating a dynamically updated BIM data stream, and synchronizing it to all stages of the full life cycle includes: Obtaining three-dimensional point cloud data of actual construction components through laser scanning, performing position deviation analysis on the theoretical coordinates of the corresponding components in the performance-optimized BIM model, and calculating the construction deviation ratio; Real-time collecting building equipment operation data, including air conditioning energy consumption, lighting power, and elevator usage frequency, and performing energy consumption and performance deviation analysis on the predicted values of the performance-optimized BIM model; When the construction deviation ratio or the operation and maintenance energy consumption deviation ratio exceeds a preset threshold, automatically triggering secondary optimization of the deep Q network, generating a dynamically updated BIM data stream including parameter adjustment records, and the BIM data stream is marked with a version number, a change time, and an impact scope label; Distributing the dynamically updated BIM data stream to the design, construction, and operation and maintenance ends, respectively driving design review, construction plan adjustment, and equipment parameter calibration, and performing full life cycle management.

2. The building information intelligent management method based on BIM technology according to claim 1, characterized in that, The step of converting user requirements into a standardized parameter set through a preset mapping table supporting online update and natural language processing technology, and generating an initial BIM model in combination with a rule engine includes: Parsing the natural language requirements input by the user, and extracting keywords and constraint conditions; Mapping the keywords to BIM parameter types and value ranges through a preset mapping table, and the mapping table includes a triple relationship of user requirement semantics, BIM parameters, and associated components; If there is no matching item in the mapping table, triggering an online update request and expanding the mapping table entries based on user feedback; Using a rule engine to generate the geometric structure and attribute labels of the initial BIM model according to parameter priorities and spatial topology rules.

3. The building information intelligent management method based on BIM technology according to claim 1, characterized in that The step of based on a dynamically updated design specification database, automatically adjusting the parameters of the initial BIM model through parameter matching and an AI rule engine, and outputting a specification-compliant BIM model includes: Obtain the latest building code text from the authoritative code release platform through the data interface, extract the code parameter requirements, and store them structurally in the design code database; Based on the component types and spatial attributes of the initial BIM model, screen the applicable clause sets from the design code database; Based on the selected usage clause sets, compare the model parameters with the code thresholds item by item through the parameter matching algorithm to identify non-compliant item parameters; Call the preset replacement strategy through the AI rule engine to automatically correct the non-compliant parameters and generate a code-compliant BIM model with an adjustment log; 4. The building information intelligent management method based on BIM technology according to claim 1, characterized in that The step of performing simulation calculations on the structural stability, energy consumption distribution, and space utilization rate of the code-compliant BIM model to generate a performance analysis report includes: Perform structural stability analysis by the finite element method, calculate the component stress ratio and overall displacement angle in the code-compliant BIM model, and identify the over-limit areas; Based on the code-compliant BIM model, perform energy consumption simulation, calculate the hourly load throughout the year through the EnergyPlus engine, and output the peak heating and cooling loads and the energy consumption density distribution; Evaluate the space utilization rate of the code-compliant BIM model, analyze the efficiency of the pedestrian flow path using space syntax, and calculate the space utilization rate index; Normalize the over-limit areas, the peak heating and cooling loads, the energy consumption density distribution, and the space utilization rate index into deviation rates, and generate a performance analysis report including component correlation markings; 5. The intelligent management method of building information based on BIM technology according to claim 1, characterized in that The step of using the normalized deviation values in the performance analysis report as the state input of the deep Q network, outputting optimization decisions through the predefined action space, iteratively adjusting the parameters until the threshold is met, and outputting the performance-optimized BIM model includes: Classify and code the structural stability deviation, energy consumption deviation, and space utilization deviation in the performance analysis report according to parameter types, and use them as the state input vector of the deep reinforcement learning network; Define the output action of the deep reinforcement learning network as a set of BIM model parameter adjustment operations, including component geometric dimension adjustment, material property replacement, and space topology relationship optimization; Configure the network training reward mechanism to give positive rewards when the model parameters meet the performance threshold, give partial rewards when the deviation after optimization decreases, and give penalties when structural or energy consumption constraints are violated; Select optimization actions through the exploration-exploitation balance strategy, iteratively adjust the BIM model parameters and re-perform performance simulation until the model parameters converge after continuous optimization for multiple times; 6. The building information intelligent management method based on BIM technology according to claim 1, characterized in that, Based on the BIM data stream, monitor the model operation performance through the real-time feedback system. If a deviation occurs, trigger the deep Q network to optimize again, dynamically update the BIM model and loop iteratively, including: Deploy performance monitoring agents during the operation and maintenance phase to collect building operation data in real time and calculate key indicators; If the indicators in three consecutive monitoring cycles exceed 20% of the BIM model prediction value, it is determined as a performance deviation; Trigger the deep Q network to use the current operation and maintenance data as the input to re-optimize the model parameters; Mark the updated BIM model version and synchronize it to the full life cycle data stream to form a closed-loop management; 7. An intelligent building information management system based on BIM technology, which is applied to the intelligent building information management method based on BIM technology as described in claim 1, is characterized in that, Include: A requirements modeling unit, which is used to convert user requirements into a standardized parameter set through a preset mapping table supporting online update and natural language processing technology, and generate an initial BIM model in combination with a rule engine; A specification adaptation unit, which is used to automatically adjust the parameters of the initial BIM model based on a dynamically updated design specification database through parameter matching and an AI rule engine, and output a specification-compliant BIM model; A performance simulation unit, which is used to perform simulation calculations on the structural stability, energy consumption distribution, and space utilization rate of the specification-compliant BIM model, and generate a performance analysis report; A model optimization unit, which is used to use the normalized deviation value in the performance analysis report as the state input of a deep Q network, output an optimization decision through a predefined action space, iteratively adjust the parameters until the threshold is met, and output a performance-optimized BIM model; A twin update unit, which is used to obtain physical building data in the construction and operation and maintenance stages in real time through an Internet of Things interface, perform digital twin comparison with the performance-optimized BIM model, calculate parameter deviations and trigger secondary optimization of the deep Q network, generate a dynamically updated BIM data stream, and synchronize it to all stages of the full life cycle.

8. A computer device, comprising a memory and a processor, wherein a computer program is stored in the memory, characterized in that, When the processor executes the computer program, it implements the steps of the building information intelligent management method based on BIM technology described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the building information intelligent management method based on BIM technology described in any one of claims 1 to 6.

Images