Building engineering design and management system based on BIM technology

Abstract

The invention relates to the technical field of building engineering management, and discloses a building engineering design and management system based on a BIM technology, and the system comprises a data collection and processing module which integrates a BIM model, real-time construction data and operation and maintenance records, and carries out the standardization processing; after the dynamic tensor and multi-agent collaborative optimization module receives the data, modeling and updating are carried out by utilizing a dynamic topology tensor, and collaborative optimization is carried out on a plurality of agents through a game theory; the cross-stage graph neural network reasoning module constructs a heterogeneous graph and reasones a component relationship, and provides decision support; and the decision support and closed-loop feedback module adjusts system parameters in real time and feeds back the system parameters to the above modules to form closed-loop optimization. According to the method, the cross-stage graph neural network reasoning module is adopted, a heterogeneous graph is constructed, a graph neural network is applied to conduct reasoning on the complex relation between building components, and the dependency and conflict problems of different stages in a building project are accurately recognized.

Description

Technical Field

[0001] The present invention relates to the technical field of construction project management, and in particular to a construction project design and management system based on BIM technology. Background Art

[0002] In modern construction projects, with the expansion of project scale and increase in complexity, traditional management methods can no longer meet actual needs. Construction projects involve multiple links such as design, construction, and resource scheduling. There are complex dependencies and conflicts between each link. How to achieve precise management and effective scheduling has become a major problem in construction projects. In order to solve these problems, traditional methods often rely on a large amount of manual operation and experience judgment, but this method is not only inefficient, but also prone to errors, leading to problems such as delayed construction progress and waste of resources.

[0003] In the existing technology, the construction project management system has developed to a certain extent. Based on BIM (Building Information Modeling), the current building management system can realize three-dimensional visualization of project design and optimize the workflow in the design phase through digital tools. Many systems can realize basic functions such as progress management, cost control and resource scheduling, helping project managers to grasp the progress of the project and ensure that various resources are effectively allocated. At the same time, the application of BIM technology in the design phase has greatly improved the accuracy of design drawings and reduced the occurrence of design changes and construction conflicts to a certain extent. Some systems can also track the progress in real time during construction and adjust the project plan according to actual conditions.

[0004] However, existing technologies still have many shortcomings when dealing with complex construction project management issues. First, existing systems have limited capabilities in multi-stage, cross-departmental collaborative management. Especially when faced with complex dependencies and conflicts, many traditional methods still rely on manual experience for reasoning and judgment, lacking efficient intelligent support. Second, although existing systems can provide certain real-time feedback, most systems still do not implement dynamic adaptation and closed-loop feedback mechanisms. Once changes occur during the construction process, project plans are often unable to be adjusted in a timely manner. More importantly, existing technologies still have bottlenecks in the collaborative optimization between intelligent agents, and cannot guarantee the optimal collaboration between various intelligent agents when handling complex tasks, resulting in insufficient efficiency and accuracy in resource scheduling. In addition, the application of existing technologies in graph neural network reasoning, decision support, etc. is relatively limited, and cannot comprehensively improve the intelligent level of construction project management. To this end, those skilled in the art have proposed a construction project design and management system based on BIM technology to solve the above problems. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a construction engineering design and management system based on BIM technology, which solves the problems of insufficient coordination, dynamic adjustment lag and lack of intelligent optimization capabilities in the existing construction engineering management system.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a construction engineering design and management system based on BIM technology, comprising:

[0007] Data acquisition and processing module, used to integrate BIM model data, real-time construction data and operation and maintenance records, and perform standardized processing;

[0008] A dynamic tensor and multi-agent collaborative optimization module receives the output data of the data acquisition and processing module, models the building components through the dynamic topology tensor and updates them in real time, and multiple agents make collaborative optimization decisions based on the game theory model;

[0009] A cross-stage graph neural network reasoning module receives the output data of the dynamic tensor and multi-agent collaborative optimization module, constructs a heterogeneous graph and applies a graph neural network to reason about the relationship between building components, and provides cross-stage decision support;

[0010] The decision support and closed-loop feedback module receives the output data of the cross-stage graph neural network reasoning module, adjusts the system parameters in real time based on the optimal decision of the intelligent agent and the graph neural network reasoning results, and feeds back to the data acquisition and processing module and the dynamic tensor and multi-agent collaborative optimization module to form a closed-loop optimization.

[0011] Preferably, the data acquisition and processing module includes:

[0012] BIM model data integration unit, used to obtain building design data from BIM software and convert it into a standard format;

[0013] Sensor data acquisition unit, used to obtain real-time temperature, humidity, and stress monitoring data at the construction site and synchronize it with the static data in the BIM model;

[0014] The data standardization unit is used to unify the format and pre-process the data from BIM models, sensors and operation and maintenance records to form a unified data set.

[0015] Preferably, the dynamic tensor and multi-agent collaborative optimization module includes:

[0016] Dynamic tensor modeling unit, used to encode the geometric information, material properties, and construction status data of building components into multi-dimensional tensors and model them through tensor decomposition methods;

[0017] A multi-agent optimization unit, including a design optimization agent, a construction scheduling agent, and a resource management agent. Each agent optimizes corresponding subtasks based on the tensor data and coordinates decisions through a game theory model.

[0018] The intelligent agent collaborative decision-making unit is used to calculate the optimal decision of each intelligent agent through game theory strategies to ensure the optimal operation of the entire system.

[0019] Preferably, the cross-stage graph neural network reasoning module includes:

[0020] A graph construction unit, used to map building components into graph nodes and construct dependency and conflict graphs between building stages;

[0021] The graph neural network reasoning unit updates node features through multiple rounds of message passing and infers the complex relationships between building components;

[0022] The decision-making reasoning unit provides decision-making recommendations for design changes, construction conflict resolution, and operation and maintenance risk prediction based on the reasoning results.

[0023] Preferably, the decision support and closed-loop feedback module includes:

[0024] Decision support unit, used to generate visual reports based on the agent's optimal decision and the graph neural network's reasoning results;

[0025] The closed-loop feedback unit is used to update the dynamic tensor model and the BIM model in real time, and to feed back the updated parameters to the data acquisition and processing module and the dynamic tensor and multi-agent collaborative optimization module.

[0026] Preferably, the dynamic tensor modeling unit performs tensor modeling by the following steps:

[0027] Obtain the geometric information, material properties and construction status of building components and map them into different dimensions of tensors;

[0028] The tensor decomposition method is used to reduce the dimension of building component data;

[0029] The construction data at different time points are updated tensor-wise, and a dynamic tensor model is formed by combining the tensor decomposition method with real-time data.

[0030] Preferably, the update formula of the dynamic tensor is:

[0031]

[0032] in: is the tensor at the previous time point; U (N) is the factor matrix of dimension N; It is the real-time data increment collected by sensors; Dynamic update of tensors.

[0033] Preferably, the multi-agent optimization unit performs optimization decision-making through the following steps:

[0034] Each agent obtains the current dynamic tensor data according to the task it is responsible for;

[0035] Each agent calculates its own utility function and adjusts its decision variables based on resource, schedule, and cost constraints;

[0036] Each intelligent agent calculates the Nash equilibrium through the game theory model and obtains the optimal coordination decision.

[0037] Preferably, the game model of the agent is:

[0038]

[0039] Among them: U i (a i ,a -i ) is the utility function of agent i; a i is the decision variable of agent i; a -i are the decision variables of other agents; is the optimal decision of agent i;

[0040] The utility function is constructed based on task completion efficiency, resource consumption and conflict resolution rate.

[0041] Preferably, the graph neural network reasoning unit performs node update through the following steps:

[0042] Initialize the features of each node and update the adjacent node information according to the edge definition of the graph;

[0043] Calculate the new eigenvalue of each node through multiple rounds of message passing until convergence is reached;

[0044] The following update formula is used to calculate node features:

[0045]

[0046] in: is the feature of node v in round t; is the set of neighbor nodes of node v; is the information transmitted from neighbor node u to node v in round t.

[0047] The present invention provides a construction engineering design and management system based on BIM technology.

[0048] Beneficial effects:

[0049] 1. The present invention adopts a cross-stage graph neural network reasoning module. By constructing a heterogeneous graph and applying the graph neural network to reason about the complex relationships between building components, it can accurately identify the dependencies and conflicts in different stages of the construction project. Compared with the existing solutions that simply rely on rule reasoning or traditional optimization algorithms, the present invention can better cope with the dynamically changing construction environment, solve the limitations of traditional methods in dealing with complex dependencies, and make construction project management more flexible and intelligent.

[0050] 2. The decision support and closed-loop feedback module of the present invention improves the adaptability and response speed of construction project management through real-time feedback of system parameters and dynamic adjustment of BIM models. Unlike the static feedback mechanism in the prior art, the present invention can continuously optimize model parameters during actual operation, ensuring that the system can self-adjust and maintain stability in the face of emergencies or changes, thus solving the lag problem of existing methods when dealing with complex working conditions.

[0051] 3. The multi-agent collaborative optimization module of the present invention works in conjunction with the cross-stage graph neural network reasoning module to achieve precise scheduling and resource optimization of construction projects. Through the multi-party collaboration of intelligent agents and the deep reasoning of graph neural networks, the present invention can comprehensively consider the constraints of different components, tasks and resources in the decision-making process. Compared with traditional single decision-making optimization solutions, the present invention can more accurately predict potential conflicts and provide optimization suggestions, solving the problems of uneven resource allocation and low scheduling efficiency in the existing technology.

[0052] 4. The present invention realizes the full tracking and real-time updating of construction project data through the seamless connection of dynamic tensor model and BIM model. Different from the manual updating and isolated data processing in the prior art, the present invention can automatically adjust the model according to the real-time status of the project to ensure the high consistency and real-time performance of the data, and solves the problem of frequent data updates and difficulty in synchronizing with the actual progress in the prior art, making the engineering decision-making process more accurate and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 Schematic diagram of the system architecture of the present invention;

[0054] Figure 2 This is a schematic diagram of the data acquisition and processing module architecture of the present invention;

[0055] Figure 3 This is a schematic diagram of the dynamic tensor and multi-agent collaborative optimization module architecture of the present invention;

[0056] Figure 4 This is a schematic diagram of the cross-stage graph neural network reasoning module architecture of the present invention;

[0057] Figure 5 This is a schematic diagram of the decision support and closed-loop feedback module architecture of the present invention. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0059] Please see the attached Figure 1 -Attached Figure 5 The embodiment of the present invention provides a construction engineering design and management system based on BIM technology, including:

[0060] Data acquisition and processing module, used to integrate BIM model data, real-time construction data and operation and maintenance records, and perform standardized processing;

[0061] Specifically, the data acquisition and processing module of the present invention, as the foundational module of the construction project management system, is primarily responsible for collecting, organizing, and processing data from multiple data sources, converting this data into standardized datasets, and providing reliable data support for subsequent analysis, optimization, and decision-making. By seamlessly connecting with BIM models, sensor equipment, and operation and maintenance records, this module ensures that dynamic data from the real-time construction process can be effectively integrated with static design data, providing a comprehensive and accurate information foundation for the full lifecycle management of engineering projects.

[0062] In this embodiment, the data acquisition and processing module includes three main components: a BIM model data integration unit, a sensor data acquisition unit, and a data standardization unit.

[0063] The BIM model data integration unit is used to obtain architectural design data from BIM software and convert it into a standard format. Specifically, the BIM model data integration unit can parse the design files generated on the BIM platform (such as IFC, gbXML and other formats) and extract key information such as the building model's geometric information, component attribute information, material specifications, and spatial position. Through this unit, the system can convert the three-dimensional data of building components in BIM (such as walls, floors, doors and windows, etc.) into a standard format for data processing and optimization analysis for further integration and processing.

[0064] In one possible implementation, the BIM model data integration unit encodes each component into a tensor-oriented component object, and the generated data will contain the following field information:

[0065] Geometric topological information: represents the spatial position and shape of the component;

[0066] Material performance parameters: including component material, durability, compressive strength, etc.;

[0067] Spatial position information: including the relative position of components in the building, such as coordinates, azimuth, etc.;

[0068] List of associated components: describes the structural associations and dependencies between components, such as load-bearing relationships, support relationships, etc.

[0069] The sensor data acquisition unit is used to obtain dynamic data from the construction site in real time, such as temperature, humidity, and stress monitoring data. Sensors are placed at key node components and high-risk areas to ensure real-time monitoring of the construction environment and component status. The data collected by the sensors includes not only environmental parameters (such as temperature, humidity, light intensity, etc.), but also structural health monitoring data, such as component stress, vibration frequency, displacement and other information. The sensor data acquisition unit synchronizes these data with the static data in the BIM model, so that all information of the construction project can reflect the latest construction status at any time.

[0070] In some embodiments, the data collected by the sensors can be transmitted to a data processing center via wired or wireless means. This data set is then matched against corresponding components in the design model. For example, if the stress value of a load-bearing wall detected at the construction site exceeds the design standard, the system will automatically compare it with the component relationships in the BIM model to identify potential construction risks in that area.

[0071] The data standardization unit unifies and preprocesses data from BIM models, sensors, and maintenance records, ensuring that all data can be stored and further analyzed in a standardized format. Specifically, the data standardization unit cleans, removes noise, and fills in missing values ​​on the raw data according to pre-set data format specifications. All numerical data is converted to standardized units, and text data has standardized field names to ensure data consistency and integrity.

[0072] In a possible implementation, the data standardization unit implements data preprocessing through the following steps:

[0073] Missing value filling: supplement missing data through interpolation, average value, maximum value and other methods to ensure the completeness of the data set.

[0074] Outlier elimination: Use statistical methods (such as Z-score) to eliminate outliers to ensure data quality.

[0075] Unit conversion: Data from different sources may have different units. The data standardization unit will convert all values ​​into specified units according to actual needs (such as temperature to Celsius and pressure to Pascal).

[0076] Timestamp alignment: Sensor data and BIM model design data usually have different collection times. The data standardization unit synchronizes data from different time periods through a timestamp alignment algorithm (such as time window-based aggregation).

[0077] Through the collaborative work of these three units, the data acquisition and processing module ensures that all types of data are quickly and accurately converted into standardized datasets. This standardized data provides the necessary information foundation for subsequent dynamic analysis, intelligent optimization, and decision support systems. Furthermore, this module's data processing capabilities ensure the system can respond in real time and provide reliable data support in complex construction environments. For anomalies or deviations during construction, the system can provide timely alerts or feedback through real-time monitoring and data processing, ensuring that the project proceeds as planned.

[0078] The dynamic tensor and multi-agent collaborative optimization module receives the output data of the data acquisition and processing module, models the building components through the dynamic topology tensor and updates them in real time. Multiple agents make collaborative optimization decisions based on the game theory model.

[0079] Specifically, the dynamic tensor and multi-agent collaborative optimization module of the present invention is one of the core modules of the construction engineering design and management system. It is responsible for real-time modeling and optimization decisions for building components through dynamic tensor modeling and multi-agent collaborative optimization algorithms. This module, closely integrated with the data acquisition and processing module, constructs a multidimensional tensor model based on real-time building data. It then uses game-theoretic multi-agent optimization strategies to intelligently optimize key tasks such as construction progress and resource scheduling, ensuring maximum overall project operational efficiency and resource utilization.

[0080] In this embodiment, the dynamic tensor and multi-agent collaborative optimization module includes a dynamic tensor modeling unit, a multi-agent optimization unit and an agent collaborative decision-making unit.

[0081] The dynamic tensor modeling unit's primary task is to encode the geometric information, material properties, and construction status data of building components into a multidimensional tensor and perform dimensionality reduction modeling using tensor decomposition. Specifically, the dynamic tensor modeling unit first obtains the geometric information, material properties, and construction status of the building components and maps this information into different dimensions of the tensor. For example, geometric information can be mapped to the first dimension of the tensor, material properties to the second, and construction status data to the third. In this way, the tensor model can fully reflect the multidimensional information of the building project.

[0082] In some embodiments, the dynamic tensor modeling unit uses tensor decomposition to reduce the dimensionality of building component data. Tensor decomposition can effectively reduce computational complexity while preserving key information to a certain extent. In this way, the constructed tensor can be more effectively integrated with subsequent dynamic data updates, thereby improving the system's real-time responsiveness.

[0083] Specifically, the dynamic update of tensors is based on the following formula:

[0084]

[0085] in: is the tensor at the previous time point; U (k) is the factor matrix of the kth dimension; It is the real-time data increment collected by sensors; Dynamic update of tensors.

[0086] In this way, the dynamic tensor modeling unit is able to model building components in real time and continuously update the model over time to adapt to real-time changes during the construction process.

[0087] The multi-agent optimization unit is primarily responsible for optimizing various tasks in the construction process based on dynamic tensor data, including design optimization, construction scheduling, and resource management. Each agent calculates the optimal decision variables based on its own tasks and goals by analyzing the current dynamic tensor data. Specifically, each agent evaluates the effectiveness of its current decision by setting a utility function and adjusts it according to the constraints. The utility function of an agent typically takes into account factors such as task completion efficiency, resource consumption, and conflict resolution rate. In some embodiments, the utility function can be expressed as:

[0088] U i (a i ,a -i )=f(a i ,a -i );

[0089] Among them: U i (a i ,a -i ) is the utility function of agent i; a i is the decision variable of agent i; a -i are the decision variables of other agents; f(a i ,a -i ) represents the impact of the agent’s decision on the overall system.

[0090] The agent collaborative decision-making unit is responsible for calculating the optimal decision for each agent based on a game theory model and coordinating their actions to ensure optimal operation of the entire system. Through game theory, the decision-making between agents will lead to a Nash equilibrium solution through interaction and collaboration. This Nash equilibrium solution ensures that no single agent in the multi-agent system can improve its own utility by changing its strategy, thus achieving optimal operation of the entire system.

[0091] Specifically, the game model of the agent can be expressed as:

[0092]

[0093] Among them: U i (a i ,a -i ) is the utility function of agent i; a i is the decision variable of agent i; a -i are the decision variables of other agents; is the optimal decision of agent i.

[0094] In this way, the agent can continuously adjust its strategy during the game and ultimately calculate the global optimal decision through iteration.

[0095] The cross-stage graph neural network reasoning module receives the output data of the dynamic tensor and multi-agent collaborative optimization module, constructs a heterogeneous graph and applies the graph neural network to reason about the relationship between building components, providing cross-stage decision support;

[0096] Specifically, to achieve in-depth identification and intelligent reasoning of complex dependencies and conflict factors throughout the construction process, the proposed cross-stage graph neural network reasoning module is integrated after the dynamic tensor and multi-agent collaborative optimization module to further explore the structural semantic relationships between components and generate cross-stage decision-making basis. This module receives dynamic tensor data output by upstream modules and constructs a heterogeneous graph structure based on construction progress, component attributes, and task dependency information. It then uses graph neural networks for relational reasoning, providing refined auxiliary support for design change optimization, conflict identification, and operation and maintenance risk assessment.

[0097] In this embodiment, the cross-stage graph neural network reasoning module includes a graph construction unit, a graph neural network reasoning unit and a decision reasoning unit.

[0098] The primary function of the graph construction unit is to map data from dynamic tensors into nodes and edges in a graph structure, forming a graph topology suitable for graph neural network operations. Typically, the graph construction unit maps building components (such as beams, columns, walls, equipment pipelines, etc.) as nodes in the graph, and the dependencies between components (such as construction sequence, spatial interference, functional interdependence, etc.) as edges in the graph.

[0099] Specifically, in some embodiments, the constructed graph is a heterogeneous graph (HeterogeneousGraph), whose node types include structural component nodes, electromechanical component nodes, construction task nodes, etc., and edge types include construction sequence relationship edges, spatial overlapping relationship edges, functional coupling relationship edges, etc.

[0100] As an option, the node features in the graph construction unit may include the following information:

[0101] Geometric attribute characteristics (such as volume, height, spatial coordinates, etc.);

[0102] Material and mechanical parameters (such as elastic modulus, durability grade, reinforcement arrangement);

[0103] Time attributes (such as planned start time, actual progress, lag time, etc.);

[0104] Resource information (such as required manpower, equipment, material supply cycle, etc.);

[0105] Historical conflict records and security risk level labels.

[0106] Edge features can include conflict weight, dependency priority, risk propagation probability, etc.

[0107] The graph neural network inference unit is used to perform deep reasoning on the relationships between building components based on the graph structure described above. Typically, this unit updates node features multiple times through a graph convolution mechanism to capture high-level semantic relationships between components.

[0108] In one possible implementation, node update follows these steps:

[0109] First, initialize the feature vector of each node And obtain the initial state through component attribute tensor mapping;

[0110] Then, for each round t, the message is generated by aggregating the information of neighbor nodes u∈N(v) The message takes into account edge weights and structural relevance;

[0111] The node feature update formula is as follows:

[0112]

[0113] in: is the feature of node v in round t; is the set of neighbor nodes of node v; is the information transmitted from neighbor node u to node v in round t.

[0114] Usually defined as:

[0115]

[0116] Among them: φ is a nonlinear transformation function; e uv The eigenvector representing the edge connecting u and v.

[0117] In some embodiments, the message aggregation function ψ can adopt the form of weighted average, maximum pooling, attention mechanism, etc. to improve the modeling ability under complex dependency structures.

[0118] After multiple rounds of message propagation, when the node features converge or reach the preset number of propagation rounds T, the final node state will serve as the basis for relational reasoning.

[0119] The decision reasoning unit generates recommended results for engineering management decisions based on the node state vectors obtained by graph neural network reasoning and the predicted conflict results and dependency risks between components.

[0120] Specifically, in one possible implementation, the unit supports the following three types of output tasks:

[0121] Design change suggestions: When a component's state vector shows high-risk attribute markers (e.g., frequent conflicts, inconsistent functions), the system will propose a design optimization path.

[0122] Construction conflict resolution: Use the reasoning results between nodes and edges to mark component combinations with temporal or spatial conflicts, and output suggestions for adjusting priorities or construction sequences;

[0123] Operation and maintenance risk prediction: Based on the historical evolution path of the node and its neighbor relationship, the potential risk level of the component in the future cycle is inferred and a risk warning report is generated.

[0124] As an extension mechanism, the decision reasoning unit can access historical decision feedback data for reverse reinforcement learning updates to improve the dynamic adaptability of the system.

[0125] The decision support and closed-loop feedback module receives the output data of the cross-stage graph neural network reasoning module, adjusts the system parameters in real time based on the optimal decision of the intelligent agent and the graph neural network reasoning results, and feeds back to the data acquisition and processing module and the dynamic tensor and multi-agent collaborative optimization module to form a closed-loop optimization.

[0126] Specifically, to achieve full-process closed-loop control of the construction engineering management system, the present invention further provides a decision support and closed-loop feedback module as the final output link in the system logic chain. This module is used to integrate the structural relationship reasoning results of the cross-stage graph neural network reasoning module and the optimal decision results of the multi-agent collaborative optimization module, thereby forming an intelligent closed-loop system with reasoning-optimization-feedback capabilities. Generally, after receiving the reasoning results, this module adjusts and feeds back the tensor model and BIM parameters, enabling the system to have dynamic self-adaptation and parameter closed-loop correction capabilities, thereby improving the stability and intelligence level of the overall system operation.

[0127] In this embodiment, the decision support and closed-loop feedback module mainly includes a decision support unit and a closed-loop feedback unit.

[0128] The decision support unit (DSU) generates decision-support information output for engineering managers based on the optimal decision results from the multi-agent optimization unit and the node status and risk assessment information output by the cross-stage graph neural network inference module. Generally, the DSU can perform structured analysis on component-level status changes, the spatial location of conflicting nodes, and the effects of resource scheduling strategy adjustments, providing corresponding graphical displays and textual interpretations.

[0129] In one possible implementation, the decision support unit maps data from multiple heterogeneous sources into a unified visualization matrix, which is defined as follows:

[0130] D visual (k, j) = η·S k +λ·R j +μ·C kj ;

[0131] Where: D visual (k, j) represents the comprehensive visual value of the k-th component in the j-th task dimension; S k R represents the agent's decision score for component k; j represents the task risk coefficient identified by graph neural network reasoning; C kj represents the dependency strength between component k and task j; η, λ, μ are decision weight coefficients, which can be dynamically set according to engineering requirements.

[0132] As an option, the output content of the decision support unit includes but is not limited to: construction path diagram, conflict component list, tensor comparison diagram before and after adjustment, optimization scheduling strategy report, etc.

[0133] The closed-loop feedback unit is used to feed system outputs back to upstream modules, enabling dynamic parameter-level adjustments and forming a feedback loop for data updates. Generally, this unit includes three processing steps: a parameter correction module, a model reconstruction module, and a feedback distribution module.

[0134] Specifically, in some embodiments, the closed-loop feedback unit updates the tensor model parameters based on the node risk level, and the revised tensor is recorded as:

[0135]

[0136] in; Represents the updated tensor model; is the tensor of the previous iteration; ΔT risk The risk corresponding to the tensor correction term identified by graph neural network reasoning; ΔT decision It is the structural optimization tensor correction term corresponding to the agent's decision feedback; γ and δ are tensor adjustment coefficients used to control the feedback intensity.

[0137] As an implementation method, the closed-loop feedback unit maps the above correction results to the BIM model parameter space, such as adjusting the thickness, displacement properties, material grade or construction sequence of a wall component, and synchronizes them to the data acquisition and processing module to ensure the continuity of subsequent data collection and model consistency.

[0138] Furthermore, the feedback distribution module can simultaneously transmit partial parameter update results to the dynamic tensor modeling unit and the BIM model interface based on feedback path selection, forming a bidirectional feedback mechanism. This approach ensures that upper-level decisions can quickly penetrate the underlying data end, forming a complete closed loop where reasoning drives optimization and optimization drives re-perception.

[0139] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A construction engineering design and management system based on BIM technology, characterized in that: include: Data acquisition and processing module, used to integrate BIM model data, real-time construction data and operation and maintenance records, and perform standardized processing; A dynamic tensor and multi-agent collaborative optimization module receives the output data of the data acquisition and processing module, models the building components through the dynamic topology tensor and updates them in real time, and multiple agents make collaborative optimization decisions based on the game theory model; A cross-stage graph neural network reasoning module receives the output data of the dynamic tensor and multi-agent collaborative optimization module, constructs a heterogeneous graph and applies a graph neural network to reason about the relationship between building components, and provides cross-stage decision support; The decision support and closed-loop feedback module receives the output data of the cross-stage graph neural network reasoning module, adjusts the system parameters in real time based on the optimal decision of the intelligent agent and the graph neural network reasoning results, and feeds back to the data acquisition and processing module and the dynamic tensor and multi-agent collaborative optimization module to form a closed-loop optimization.

2. A construction engineering design and management system based on BIM technology according to claim 1, characterized in that: The data acquisition and processing module includes: BIM model data integration unit, used to obtain building design data from BIM software and convert it into a standard format; Sensor data acquisition unit, used to obtain real-time temperature, humidity, and stress monitoring data at the construction site and synchronize it with the static data in the BIM model; The data standardization unit is used to unify the format and pre-process the data from BIM models, sensors and operation and maintenance records to form a unified data set.

3. A construction engineering design and management system based on BIM technology according to claim 1, characterized in that: The dynamic tensor and multi-agent collaborative optimization module includes: Dynamic tensor modeling unit, used to encode the geometric information, material properties, and construction status data of building components into multi-dimensional tensors and model them through tensor decomposition methods; A multi-agent optimization unit, including a design optimization agent, a construction scheduling agent, and a resource management agent. Each agent optimizes corresponding subtasks based on the tensor data and coordinates decisions through a game theory model. The intelligent agent collaborative decision-making unit is used to calculate the optimal decision of each intelligent agent through game theory strategies to ensure the optimal operation of the entire system.

4. The construction engineering design and management system based on BIM technology according to claim 1 is characterized in that: The cross-stage graph neural network reasoning module includes: A graph construction unit, used to map building components into graph nodes and construct dependency and conflict graphs between building stages; The graph neural network reasoning unit updates node features through multiple rounds of message passing and infers the complex relationships between building components; The decision-making reasoning unit provides decision-making recommendations for design changes, construction conflict resolution, and operation and maintenance risk prediction based on the reasoning results.

5. The construction engineering design and management system based on BIM technology according to claim 1 is characterized in that: The decision support and closed-loop feedback module includes: Decision support unit, used to generate visual reports based on the agent's optimal decision and the graph neural network's reasoning results; The closed-loop feedback unit is used to update the dynamic tensor model and the BIM model in real time, and to feed back the updated parameters to the data acquisition and processing module and the dynamic tensor and multi-agent collaborative optimization module.

6. The construction engineering design and management system based on BIM technology according to claim 3 is characterized in that: The dynamic tensor modeling unit performs tensor modeling by the following steps: Obtain the geometric information, material properties and construction status of building components and map them into different dimensions of tensors; The tensor decomposition method is used to reduce the dimension of building component data; The construction data at different time points are updated tensor-wise, and a dynamic tensor model is formed by combining the tensor decomposition method with real-time data.

7. The construction engineering design and management system based on BIM technology according to claim 6 is characterized in that: The update formula of the dynamic tensor is: in: is the tensor at the previous time point; U (N) is the factor matrix of dimension N; It is the real-time data increment collected by sensors; Dynamic update of tensors.

8. The construction engineering design and management system based on BIM technology according to claim 3 is characterized in that: The multi-agent optimization unit performs optimization decisions through the following steps: Each agent obtains the current dynamic tensor data according to the task it is responsible for; Each agent calculates its own utility function and adjusts its decision variables based on resource, schedule, and cost constraints; Each intelligent agent calculates the Nash equilibrium through the game theory model and obtains the optimal coordination decision.

9. The construction engineering design and management system based on BIM technology according to claim 8, characterized in that: The game model of the agent is: Among them: U i (a i ,a -i ) is the utility function of agent i; a i is the decision variable of agent i; a -i are the decision variables of other agents; is the optimal decision of agent i; The utility function is constructed based on task completion efficiency, resource consumption and conflict resolution rate.

10. The construction engineering design and management system based on BIM technology according to claim 4, characterized in that: The graph neural network reasoning unit performs node updates through the following steps: Initialize the features of each node and update the adjacent node information according to the edge definition of the graph; Calculate the new eigenvalue of each node through multiple rounds of message passing until convergence is reached; The following update formula is used to calculate node features: in: is the feature of node v in round t; is the set of neighbor nodes of node v; is the information transmitted from neighbor node u to node v in round t.

Images