BIM data intelligent matching and conflict detection method

Abstract

The invention belongs to the technical field of data processing, and provides a BIM data intelligent matching and conflict detection method, and the method comprises the steps: generating a fusion feature of a component through fusing a geometric topology feature and a semantic constraint feature; through alignment and coupling of the installation state sequence and the compliance state sequence, a space-time state sequence of the component is formed, so that the construction constraint graph can fit the actual dynamic change of construction; the method comprises the following steps: injecting a space-time state sequence into entity nodes by taking components as the entity nodes and a design specification and a construction process as constraint nodes, constructing a construction constraint graph, and predicting a conflict propagation path between the components through a time sequence graph neural network; the prediction result of the conflict propagation path is subjected to actual construction verification, the verified prediction result is subjected to influence tracing and grading, and the comparison data of the actual construction result and the prediction result is combined to adjust the parameters of the time sequence diagram neural network and the edge weight of the construction constraint diagram, so that the prediction accuracy of the conflict propagation path is further improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method for intelligent matching and conflict detection of BIM data. Background Technology

[0002] As the construction industry transforms towards digitalization and intelligence, Building Information Modeling (BIM) has been widely applied throughout the entire lifecycle of building engineering, from design and construction to operation and maintenance, becoming one of the technical means to achieve refined management and control of projects and reduce construction risks. During the construction phase, the installation accuracy of components, the rationality of process connections, and compliance management directly affect project quality and construction efficiency. Therefore, component matching and construction conflict detection based on BIM data have become key research areas in the field of building digitalization.

[0003] Current BIM data matching and conflict detection mainly revolve around three aspects: BIM data parsing, component matching, and conflict identification. Although some progress has been made in intelligent BIM data matching and conflict detection, there are still many shortcomings in practical engineering applications, which cannot meet the needs of refined and precise construction management. Existing technologies mostly analyze the geometric topology of components in BIM data separately, resulting in low BIM data matching accuracy and creating potential deviations for subsequent conflict detection.

[0004] Existing technologies cannot reflect the temporal and attribute dimensions of components during the construction phase. That is, they cannot clearly determine whether the actual installation status of components meets compliance standards within the same construction period, which affects the targeting of conflict detection. At the same time, they do not fit the actual construction process and cannot accurately reflect the constraint relationship and temporal transmission law between components, resulting in insufficient conflict prediction accuracy and difficulty in dealing with multi-type and multi-path conflict scenarios in complex projects.

[0005] In view of the shortcomings of the existing technology, the technical problem to be solved by this application is how to achieve intelligent matching and conflict detection of BIM data in complex engineering projects. Summary of the Invention

[0006] The purpose of this application is to overcome the shortcomings of the prior art and provide a BIM data intelligent matching and conflict detection method. The method includes: acquiring the BIM data, construction schedule and design specification documents of the project, parsing the geometric topological features of the components in the BIM data, mining the semantic constraint features of the components based on the design specification documents, and fusing the geometric topological features and semantic constraint features of the same component to generate the fused features of the component.

[0007] Based on the construction schedule, the installation status sequence of components during the construction phase is generated. Based on the fusion features and the preset BIM domain knowledge graph, reasoning is performed to generate the compliance status sequence of components during the construction phase. The installation status sequence and the compliance status sequence are aligned and coupled to form the spatiotemporal status sequence of the components.

[0008] Using components as entity nodes and design specifications and construction techniques as constraint nodes, and injecting spatiotemporal state sequences into the corresponding entity nodes to construct a construction constraint diagram, the construction constraint diagram is learned through a time-series graph neural network to predict the conflict propagation path between components during the construction phase.

[0009] The predicted results of the conflict propagation path are verified. The verified prediction results are then used for impact tracing and classification. The parameters of the time series neural network and the edge weights of the construction constraint graph are adjusted by combining the comparison data of the actual construction results and the predicted results.

[0010] Optionally, the fusion features of the generated component include:

[0011] Analyze the geometric and topological features of components in BIM data, mine the semantic constraint features of components based on design specification documents, and assign domain weights by combining the engineering association between geometric topology and semantic constraints in the BIM domain.

[0012] Based on the construction stage and the engineering type of the component, fusion weights are assigned to the geometric topology features and semantic constraint features of the same component respectively. The geometric topology features and semantic constraint features are weighted and fused according to the domain weight and fusion weight to generate the initial fused features of the component.

[0013] Perform BIM domain compliance verification on the initial fusion features, reverse the domain weights based on the reasons for the verification failure, and re-execute weighted fusion to generate fusion features of the components.

[0014] Optionally, the sequence of installation states of the generated component during the construction phase includes:

[0015] The construction schedule is deconstructed hierarchically, and the time windows, connection logic and schedule deviation tolerance of different processes are extracted and associated with the component engineering identifiers to form the relationship between construction progress and components.

[0016] The definition includes an installation state system encompassing construction phases, process execution, project acceptance, and temporary fixing. The relationships are then adapted to the installation state system to generate an initial state sequence. The rationality of the initial state sequence is verified based on the construction technology and project attributes corresponding to the component's engineering type.

[0017] Based on the verification results, the connection logic and time window are corrected, the initial state sequence is iteratively adjusted, and engineering control tags are added to generate the installation state sequence of components during the construction phase.

[0018] Optionally, the sequence of compliance statuses of the generated component during the construction phase includes:

[0019] Entities, relationships, and attributes with associated features are selected and integrated from a pre-defined BIM domain knowledge graph;

[0020] By combining the constraint dimensions of the BIM domain knowledge graph with the fusion feature analysis, reasoning logic is formed. Based on the time window of the process, the related entities, relationships and attributes are traversed through the reasoning logic to infer the compliance status of the components in different time windows and generate the initial compliance status.

[0021] Verify the matching between the initial compliance status and the fusion features, as well as the temporal coherence of the initial compliance status. Based on the verification results, optimize the reasoning logic or supplement the constraint dimensions of the BIM domain knowledge graph. Iterate the reasoning and verification to generate a sequence of compliance statuses for components during the construction phase.

[0022] Optionally, the spatiotemporal state sequence of the forming component includes:

[0023] Using fusion features as attribute benchmarks and process time windows as time-series benchmarks, the installation status sequence and compliance status sequence are aligned to clarify the relationship between installation status and compliance status within the same time window;

[0024] Based on the relationship, the installation status, compliance status and corresponding attributes within the same time window are bidirectionally coupled to generate a state unit of a single time window, which is arranged according to the connection logic of the process to form the initial spatiotemporal state.

[0025] Verify the temporal coherence of the initial spatiotemporal state and the attribute matching degree between the state unit and the fusion feature. Based on the verification results, correct the correlation relationship, and re-execute the alignment and coupling to form the spatiotemporal state sequence of the component.

[0026] Optionally, the construction constraint diagram includes:

[0027] Using components as entity nodes and design specifications and construction techniques as constraint nodes, spatiotemporal state sequences, fusion features and engineering control labels are injected into entity nodes, and constraint dimensions are marked on constraint nodes.

[0028] Associate the fusion features of entity nodes with the constraint dimensions of corresponding constraint nodes, and associate the constraint relationships between entity nodes and constraint nodes according to the connection logic of the process.

[0029] Add time sequence identifiers to entity nodes, constraint nodes and constraint relationships according to time windows, and combine fusion features and engineering control tags to verify the rationality of the association between entity nodes and constraint nodes, as well as the temporal consistency between constraint relationships and spatiotemporal state sequences.

[0030] Based on the verification results, the constraint relationships and timing identifiers are corrected, and the entity nodes, constraint nodes, and corrected constraint relationships are integrated to form a construction constraint diagram.

[0031] Optionally, the predicted conflict propagation path between components during the construction phase includes:

[0032] The temporal propagation of constraint relationships in construction constraint graphs and the attribute association between fusion features and constraint dimensions are learned through hierarchical learning of temporal graph neural networks.

[0033] Based on the results of hierarchical learning, potential conflicts in the construction constraint diagram are identified, corresponding entity nodes and constraint nodes are associated to predict and generate initial conflict paths, and the triggering type and propagation order of potential conflicts are labeled.

[0034] By combining the constraint dimension and the temporal coherence of the spatiotemporal state sequence, the rationality of the propagation order and the attribute fit of the trigger type are verified. Based on the verification results, the learning weights of the initial conflict path and the temporal graph neural network are adjusted. After the verification is passed, the conflict propagation path between components during the construction stage is predicted and generated.

[0035] Optionally, the parameters of the adjusted time-series graph neural network include:

[0036] Based on the actual construction results, the predicted results of the conflict propagation path are verified. For the verified conflict propagation path, the starting component and timing identifier of the conflict are traced based on the trigger type, propagation order and constraint relationship. The impact of the conflict is classified according to the engineering control label and fusion characteristics.

[0037] Based on the results of impact tracing and classification, and combined with the comparison data between actual construction results and prediction results, the types of deviations are distinguished and the parameters of the corresponding time series neural network to be adjusted are clarified.

[0038] Based on the fusion features and constraint dimensions, adjustment weights are assigned to the corresponding parameters. After adjusting the parameters in a hierarchical and iterative manner, the conflict propagation path is verified. Once the verification is successful, the adjustment weights and parameter adjustment ranges are locked.

[0039] Optionally, adjusting the edge weights of the construction constraint diagram includes:

[0040] Based on the results of impact tracing and classification, as well as the type of deviation, the constraint relationships corresponding to the construction constraint diagram are associated;

[0041] Based on the constraint dimensions, engineering control labels, and fusion characteristics, determine the adjustment direction and allocate the adjustment magnitude for the edge weights of the constraint relationships to be adjusted;

[0042] Adjust edge weights hierarchically according to constraint relationships, verify the consistency between the adjusted constraint relationships and the time sequence identifiers, and verify the adaptability of edge weights to constraint dimensions. After the verification is passed, lock the adjustment direction and magnitude of the edge weights.

[0043] Compared with existing technologies, the beneficial effects of this application are as follows: by integrating geometric topological features and semantic constraint features to generate fused features of components, the accuracy and comprehensiveness of BIM data matching are improved, laying the foundation for the accuracy of subsequent conflict detection; by aligning and coupling installation state sequences and compliance state sequences, a spatiotemporal state sequence of components is formed, realizing a dual characteristic description of temporal and attribute dimensions, allowing the component state to dynamically reflect the temporal changes and compliance adaptation during the construction process, and enabling the construction constraint diagram to conform to the dynamic changes of actual construction.

[0044] By using components as entity nodes and design specifications and construction techniques as constraint nodes, spatiotemporal state sequences are injected into the entity nodes to construct construction constraint diagrams. Through time-series neural networks, conflict propagation paths between components are predicted, clarifying the triggering source, propagation sequence, and scope of impact of conflicts. This enables early prediction of conflicts, facilitating construction personnel to take control measures in advance, reducing the risk of construction conflicts, and improving construction efficiency.

[0045] By verifying the predicted conflict propagation paths through actual construction, the predicted results that pass verification are subjected to impact tracing and classification. By combining the comparison data between the actual construction results and the predicted results, the parameters of the time series graph neural network and the edge weights of the construction constraint graph are adjusted to achieve synergistic optimization of the time series graph neural network and the construction constraint graph, thereby further improving the accuracy of conflict propagation path prediction.

[0046] This application presents a comprehensive BIM data intelligent matching and conflict detection solution, encompassing the complete extraction of component features, dynamic generation of spatiotemporal state sequences, precise construction of construction constraint diagrams, accurate prediction of conflict paths, and closed-loop optimization. This ensures the accuracy and efficiency of overall conflict detection. The method is adaptable to complex construction scenarios involving multiple components, processes, and constraints, providing technical support for refined management during the construction phase. It effectively reduces delays and cost increases caused by construction conflicts, thereby improving the construction quality and management efficiency of building projects. Attached Figure Description

[0047] Figure 1 This is a flowchart of a BIM data intelligent matching and conflict detection method provided in an embodiment of this application;

[0048] Figure 2 A logic flowchart illustrating the spatiotemporal state sequence of the forming component provided in the embodiments of this application;

[0049] Figure 3 This is a logical flowchart illustrating the prediction of conflict propagation paths between components during the construction phase, as provided in an embodiment of this application. Detailed Implementation

[0050] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Furthermore, in the description of the embodiments of this application, "multiple" refers to two or more pairs. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0051] refer to Figure 1 This application provides a method for intelligent matching and conflict detection of BIM data, the method comprising:

[0052] S1. Obtain the project's BIM data, construction schedule, and design specifications. Analyze the geometric topological features of components in the BIM data and mine the semantic constraint features of components based on the design specifications. Then, fuse the geometric topological features and semantic constraint features of the same component to generate the component's fused features.

[0053] Specifically, the fusion features of the generated components include:

[0054] Analyze the geometric and topological features of components in BIM data, mine the semantic constraint features of components based on design specification documents, and assign domain weights by combining the engineering association between geometric topology and semantic constraints in the BIM domain.

[0055] Based on the construction stage and the engineering type of the component, fusion weights are assigned to the geometric topology features and semantic constraint features of the same component respectively. The geometric topology features and semantic constraint features are weighted and fused according to the domain weight and fusion weight to generate the initial fused features of the component.

[0056] Perform BIM domain compliance verification on the initial fusion features, reverse the domain weights based on the reasons for the verification failure, and re-execute weighted fusion to generate fusion features of the components.

[0057] Using features alone cannot fully cover the engineering attributes of components. If reasonable domain weights are not assigned based on engineering relationships, feature bias will occur during the subsequent fusion process, affecting the comprehensiveness and relevance of the fused features. The geometric topology features of components in BIM data are analyzed, including the shape, size parameters, spatial coordinates, connection methods and connection nodes between components, and redundant geometric parameters in the BIM model are filtered out, such as the geometric data corresponding to modeling auxiliary lines and temporary annotations. This ensures that the geometric topology features reflect the actual engineering form and connection relationship of the components, and avoids redundant data from interfering with the subsequent fusion effect.

[0058] Natural language processing was used to segment the design specification document, extract keywords, and analyze contextual relationships. Semantic constraint features directly related to the components were then selected, including design specifications, performance constraints, construction process limitations, material requirements, and acceptance standards. For example, semantic constraints such as fire resistance grade I, seismic grade II, and rigid connection with column components in beam components were extracted from the design specification document. General descriptions unrelated to the components were excluded to ensure the specificity of the semantic constraint features.

[0059] This study analyzes the engineering roles of two types of features in BIM data matching and conflict detection. For structural components, both geometric topology features and semantic constraint features have high engineering relevance, and similar domain weights are assigned to both types of features. For decorative components, semantic constraint features have higher engineering relevance than geometric topology features, so semantic constraint features are assigned higher domain weights, while geometric topology features are assigned lower domain weights. For electromechanical components, the engineering relevance of geometric topology features and semantic constraint features is balanced, and similar domain weights are assigned to ensure that the allocation of domain weights aligns with the actual engineering needs of the components.

[0060] Domain weights can only reflect the basic correlation between geometric topological features and semantic constraint features in the BIM field, and cannot adapt to the differentiated needs of components in different construction stages and different engineering types. In conjunction with the project construction schedule, the current construction stage of the component should be clarified. In the foundation construction stage, the focus should be on the geometric topological features of the component, and a higher fusion weight should be assigned to the geometric topological features, while a lower fusion weight should be assigned to the semantic constraint features. In the main construction stage, geometric topological features and semantic constraint features are equally important, and similar fusion weights should be assigned to the two types of features. In the decoration construction stage, the focus should be on the semantic constraint features of the component, and a higher fusion weight should be assigned to the semantic constraint features, while a lower fusion weight should be assigned to the geometric topological features.

[0061] Based on the engineering type of the components, structural components prioritize the accuracy of geometric topological features and the compliance of semantic constraint features, with both types of features having relatively high fusion weights. Decorative components emphasize the adaptability of semantic constraint features, with fusion weights higher than those for geometric topological features. For electromechanical components, the fusion weights of geometric topological features and semantic constraint features are balanced to ensure compatibility with the spatial layout and specification requirements of electromechanical installations. The geometric topological features and semantic constraint features of the same component are multiplied by their corresponding domain weights and fusion weights, respectively, and then summed to obtain the initial fusion features of the component. Normalization is then performed to ensure the stability and consistency of the initial fusion features, facilitating subsequent compliance verification.

[0062] This solves the problem of insufficient feature specificity caused by using only domain weight fusion; by using dual weighting of domain weight and fusion weight, it achieves accurate fusion of geometric topological features and semantic constraint features. The generated initial fused features retain the geometric attributes of the components and cover the key semantic constraint requirements, taking into account both comprehensiveness and specificity; normalization processing ensures the stability of the initial fused features, providing standardized feature data for subsequent compliance verification and avoiding verification deviations caused by differences in feature dimensions.

[0063] The initial fusion features are generated based on domain weights and fusion weights, and may not be compatible with BIM design specifications, construction specifications, and industry standards. The initial fusion features are then validated for compliance by incorporating the BIM domain knowledge graph and current industry design and construction specifications. This validation includes verifying whether the geometric topology parameters in the initial fusion features meet the component design specifications, whether they are compatible with the overall spatial layout of the BIM model, and whether there are any geometric dimension deviations or inconsistencies in connection relationships. It also verifies whether the semantic constraint features in the initial fusion features are consistent with the design documentation, whether they comply with industry compliance standards, and whether there are any issues such as substandard performance constraints or conflicts with construction techniques. Finally, it checks for logical contradictions in the fused features, specifically whether the geometric topology features and semantic constraint features are compatible and conflict-free.

[0064] Record the components that pass and fail the verification. For components that fail the verification, analyze the reasons for the failure one by one to determine whether it is due to unreasonable domain weight allocation or deviations in the feature extraction and fusion process. If it is due to unreasonable domain weight allocation, adjust the corresponding domain weights. After the domain weights are adjusted, perform a weighted fusion of the adjusted domain weights, the original fusion weights, and the corresponding geometric topological features and semantic constraint features to generate new fused features. Perform compliance verification on the new fused features again. If the verification still fails, continue processing until the fused features pass the compliance verification and the component's fused features are obtained. If the problem is due to deviations in the feature extraction or fusion process, rather than domain weight issues, re-execute the feature extraction or fusion operation to ensure that the fused features meet the compliance requirements.

[0065] By verifying compliance in the BIM domain, compliance issues in the initial fusion features can be identified in a timely manner, preventing unqualified fusion features from entering subsequent processes and effectively reducing the probability of deviations in subsequent processing. Based on the verification results, the domain weights are corrected in reverse, realizing dynamic optimization of domain weights and resolving any unreasonable issues with the initial domain weights. This ensures that the fusion features can comprehensively and accurately reflect the engineering attributes and compliance requirements of the components. Iterative verification and fusion improve the accuracy, reliability, and compliance of the fusion features, providing support for the stable operation of the overall BIM data intelligent matching and conflict detection methods.

[0066] S2. Generate the installation status sequence of components during the construction phase based on the construction schedule plan, and perform reasoning based on the fusion features and the preset BIM domain knowledge graph to generate the compliance status sequence of components during the construction phase. Align and couple the installation status sequence with the compliance status sequence to form the spatiotemporal status sequence of the components.

[0067] Furthermore, the sequence of installation states of the generated components during the construction phase includes:

[0068] The construction schedule is deconstructed hierarchically, and the time windows, connection logic and schedule deviation tolerance of different processes are extracted and associated with the component engineering identifiers to form the relationship between construction progress and components.

[0069] The definition includes an installation state system encompassing construction phases, process execution, project acceptance, and temporary fixing. The relationships are then adapted to the installation state system to generate an initial state sequence. The rationality of the initial state sequence is verified based on the construction technology and project attributes corresponding to the component's engineering type.

[0070] Based on the verification results, the connection logic and time window are corrected, the initial state sequence is iteratively adjusted, and engineering control tags are added to generate the installation state sequence of components during the construction phase.

[0071] Project construction schedules are typically macro-level plans that lack specific details regarding construction procedures, timelines, and connections for individual components, failing to directly reflect the construction arrangements for each component during its construction phase. To address this, the construction schedule is deconstructed hierarchically, breaking it down into "Overall Project Schedule → Sub-project Schedule → Itemized Project Schedule → Specific Construction Procedures." This approach breaks down the overall macro-level schedule, focusing on the specific construction procedures directly related to each component. This ensures that the deconstructed schedule information accurately corresponds to the construction process of each component, avoiding information bias caused by hierarchical ambiguity. During the deconstruction process, priority is given to retaining the schedule levels directly related to component installation, filtering out macro-level control information unrelated to component installation, such as overall project acceptance plans and personnel allocation plans, ensuring that the extracted schedule information is targeted and relevant.

[0072] From the deconstructed construction schedule, information corresponding to different processes is extracted, including the time window, connection logic, and tolerance for schedule deviations for each process. The time window clearly defines the planned start and finish time of each process; the connection logic clarifies the sequence of adjacent processes, including the correspondence between preceding and subsequent processes, and clearly marks the dependencies between processes to ensure the sequential continuity of component installation; the tolerance for schedule deviations is determined based on the engineering type, structural importance, and construction difficulty of the components. The tolerance for schedule deviations is lower for core load-bearing components, relatively higher for decorative components, and a reasonable deviation range is determined based on the installation complexity of electromechanical components, reserving reasonable space for subsequent schedule adjustments.

[0073] The component engineering identifiers adopt the preset identifiers in the project BIM model. The time window, connection logic and schedule deviation tolerance of each process are associated with the component engineering identifiers one by one, forming a relationship between construction progress and components. During the association process, the engineering type and engineering attributes of the components are clarified in combination with the integration characteristics of the components, so as to ensure that the association information of components of different engineering types is consistent with the construction characteristics. For example, structural components are mainly associated with the time window and connection logic of load-bearing processes, and decorative components are mainly associated with the relevant information of installation and acceptance processes, so as to avoid the deviation of subsequent state sequence due to misalignment of association.

[0074] By deconstructing the construction schedule hierarchically, the macro-schedule can be precisely broken down into process information directly related to the components, effectively solving the problem of the construction schedule being disconnected from the construction of individual components. The extracted time windows, connection logic, and schedule deviation tolerance provide a basis for the sequential description of the component installation status, avoiding the defect of the status sequence lacking temporal support. The association with the component engineering identifier realizes the correspondence between the schedule information and the component, ensuring that each component has a clear construction sequence and connection requirements, providing accurate and comprehensive basic data for the subsequent initial installation status sequence, and improving the pertinence and rationality of the association relationship.

[0075] The lack of a unified standard for describing the installation status of components leads to inconsistencies in the definition of installation status among different construction personnel and different work processes, resulting in chaotic installation status and an inability to uniformly manage it. This paper defines an installation status system that includes construction stages, work process execution, project acceptance, and temporary fixing. This system ensures a comprehensive and standardized description of the installation status of components during the construction stages, which include foundation construction, main structure construction, decoration construction, and electromechanical installation. The construction content and component status requirements for each stage are clearly defined. Work process execution describes the specific execution status of the corresponding work processes for each component, including five states: not started, construction preparation, construction in progress, work process completed, and work process paused. Each state has clearly defined judgment criteria.

[0076] The project acceptance describes the acceptance status of components after the completion of the process, including four states: not accepted, under acceptance, accepted and unaccepted. The acceptance standards are aligned with the semantic constraints of the components in the integration features. The temporary fixing describes the temporary fixing status of components during construction, including four states: not fixed, temporary fixing in progress, temporary fixing completed and temporary fixing removed, to ensure construction safety and accuracy.

[0077] Based on the time windows of the association relationships, determine the construction stage of the component in different time windows; combine with the connection logic to determine the corresponding process execution status of the component in each construction stage; match the project acceptance status according to the completion status of the process; determine the temporary fixing status according to the project type and construction technology of the component; arrange the installation status of different time windows in chronological order to form the initial state sequence of the component; perform a rationality check on the initial state sequence according to the construction technology and project attributes corresponding to the project type, including checking whether the installation status of different time windows in the initial state sequence conforms to the connection logic.

[0078] Simultaneously check whether the installation status in the initial sequence conforms to the construction process corresponding to the component engineering type. For example, the temporary fixing status of structural components should be earlier than the completion status of the process, and the acceptance status of decorative components should closely follow the completion status of the process to avoid conflicts with the construction process. Check whether the initial sequence conforms to the engineering attributes of the components. For example, the acceptance status standard of core load-bearing components should be higher than that of ordinary components, and the duration of the temporary fixing status should meet the requirements of its structural characteristics. Record the verification results and clarify the reasons for the verification failure, such as disordered timing, inconsistent process, and mismatched attributes, to provide a clear basis for subsequent correction work.

[0079] This solves the problem of non-standard and inconsistent descriptions of component installation status, ensuring the integrity and standardization of installation status, facilitating subsequent unified management, verification, and adaptation; it enables the accurate conversion of construction progress information into installation status sequences, generating initial status sequences with clear timing, strong relevance, and close alignment with component construction arrangements; and the rationality verification effectively identifies timing, process, and attribute-related deviations in the initial status sequence, preventing unqualified initial status sequences from entering subsequent steps and reducing the probability of deviations in subsequent work.

[0080] The initial state sequence, after undergoing rationality verification, still has issues such as unreasonable connection logic and unscientific time window settings. Direct use of these could affect the accuracy of subsequent state sequence alignment and conflict detection. Based on the verification results, the initial state sequence was corrected. For the unreasonable connection logic issues found during verification, the preceding and subsequent procedures corresponding to the components were re-examined, and the execution order and corresponding time nodes of the procedures in the installation state sequence were adjusted to ensure smooth connection logic. For example, the acceptance qualified state was moved to after the procedure completion state, and the temporary fixed completion state was moved to before the procedure was in progress, in line with the actual construction process.

[0081] To address the issue of unscientific time window settings, the time windows for each process are adjusted based on the tolerance for schedule deviations and the construction techniques of the components. This shortens the time window deviation range for core load-bearing components and optimizes the time window settings for decorative and electromechanical components. The goal is to ensure that the time windows meet both the construction schedule requirements and the construction difficulty and schedule deviation needs of the components. To address the issues of inconsistent processes and mismatched attributes, the state dimensions in the initial state sequence are adjusted. For example, the acceptance state standards for core load-bearing components are revised, and details of the temporary fixed state for large components are added to ensure that the sequence fully matches the actual construction techniques and engineering attributes of the components.

[0082] After the correction is completed, the rationality check is performed again on the adjusted initial state sequence to check whether there are still deviations in the corrected sequence. If there are still cases where the check fails, the adjustment continues until the sequence passes the rationality check. In the iteratively optimized initial state sequence, engineering control labels are added. The engineering control labels are set based on the fusion features and component construction characteristics, highlighting the control priority and control requirements of the components. The control priority is divided into three levels: high, medium and low. Core load-bearing components and key electromechanical components are marked as high control priority, ordinary decorative components and auxiliary electromechanical components are marked as medium control priority, and temporary components and auxiliary support components are marked as low control priority.

[0083] The control requirements involve combining the semantic constraints of components to identify key control points during installation. For example, high-priority components require key control over construction accuracy and acceptance quality, while low-priority components require key control over construction progress. The system integrates and iterates the revised state sequence with supplementary engineering control tags, organizing them chronologically to form the component's installation state sequence during the construction phase. This effectively resolves discrepancies in timing, process, and attributes within the initial state sequence, ensuring the installation state sequence aligns with the actual construction process and requirements of the component, providing accurate state support for subsequent operations. Furthermore, it clarifies the control priorities and key areas of the components, facilitating targeted control and optimization in subsequent processing, and improving the overall technical solution's control efficiency and accuracy. This ensures the installation state sequence possesses both temporal completeness and targeted control capabilities.

[0084] Furthermore, the sequence of compliance statuses for generated components during the construction phase includes:

[0085] Entities, relationships, and attributes with associated features are selected and integrated from a pre-defined BIM domain knowledge graph;

[0086] By combining the constraint dimensions of the BIM domain knowledge graph with the fusion feature analysis, reasoning logic is formed. Based on the time window of the process, the related entities, relationships and attributes are traversed through the reasoning logic to infer the compliance status of the components in different time windows and generate the initial compliance status.

[0087] Verify the matching between the initial compliance status and the fusion features, as well as the temporal coherence of the initial compliance status. Based on the verification results, optimize the reasoning logic or supplement the constraint dimensions of the BIM domain knowledge graph. Iterate the reasoning and verification to generate a sequence of compliance statuses for components during the construction phase.

[0088] The BIM knowledge graph encompasses a vast amount of entity, relationship, and attribute information across the entire BIM process. It contains a large amount of redundant knowledge unrelated to the current component. Directly performing subsequent reasoning on such knowledge would lead to low reasoning efficiency, significant deviations in the reasoning results, and an inability to accurately match the compliance requirements of the current component. The BIM knowledge graph pre-integrates current design specifications, construction process standards, component attribute requirements, and acceptance specifications in the BIM field. It includes various entities such as component, specification, process, and attribute categories, as well as the relationships between entities and entity attributes. Entity attributes include specification constraints and component compliance parameters.

[0089] The system retrieves and integrates features to identify the attributes of the current component, including its engineering type, geometric topology parameters, and semantic constraints. It then filters entities, relationships, and attributes associated with the integrated features from a pre-defined BIM domain knowledge graph. This includes entities of the same type as the component's engineering type, specification entities corresponding to the component's semantic constraints, and process entities corresponding to the component's construction process. It also considers the compatibility between the component and specifications, the correspondence between the component and the process, and the fulfillment of attributes and constraints. Furthermore, it includes the constraint parameters of specification entities, the compliance attributes of component entities, and the operational standards of process entities.

[0090] By filtering related information, redundant knowledge in the BIM knowledge graph can be effectively eliminated, avoiding interference from irrelevant information in the subsequent reasoning process and significantly improving the efficiency of subsequent compliance status reasoning. The filtered entities, relationships, and attributes are highly matched with the component fusion features, ensuring the relevance of subsequent compliance status reasoning and solving the problem of the disconnect between compliance requirements and the component's own attributes in conventional reasoning. At the same time, the classification and organization of related knowledge provides clear knowledge support for subsequent processing, ensuring the orderly conduct of the reasoning process, further improving the accuracy of the reasoning results, and providing high-quality knowledge basis for the generation of the initial compliance status.

[0091] The selected related entities, relationships, and attributes are still scattered, lacking clear constraint standards and reasoning rules, and cannot be directly used for reasoning about compliance status. Combining fusion features with the selected related knowledge, the constraint dimensions of component compliance status are analyzed. These constraint dimensions include performance compliance, process compliance, and acceptance compliance. Geometric compliance clarifies the compliance standards for component geometry, spatial location, and connection methods; performance compliance clarifies the compliance requirements for component fire resistance, seismic resistance, and load-bearing capacity; process compliance clarifies the operational specifications and compliance standards for different construction procedures; and acceptance compliance clarifies the acceptance standards and compliance judgment criteria after each procedure of the component is completed.

[0092] Based on constraint dimensions and associated knowledge, a reasoning logic for compliance status is formed. The reasoning logic includes association rules, temporal rules, and judgment rules. Association rules clarify the correspondence between different constraint dimensions and fusion features and knowledge graph attributes. Temporal rules clarify the priority and reasoning focus of constraint dimensions corresponding to different process time windows. For example, in the time window of the basic construction stage, geometric compliance dimension and process compliance dimension are reasoned first, and in the time window of the acceptance stage, acceptance compliance dimension and performance compliance dimension are reasoned first. Judgment rules clarify the compliance judgment conditions for each constraint dimension. If the actual state of a component in a certain dimension conforms to the constraint attributes in the knowledge graph, it is judged as compliant; otherwise, it is judged as non-compliant. At the same time, the initial correction direction is clarified when non-compliance occurs.

[0093] Based on the time windows of the construction process, the related entities, relationships, and attributes are traversed through reasoning logic to infer the compliance status of components within different time windows. That is, according to the order of the time windows, for the construction process corresponding to each time window, based on the temporal rules of the reasoning logic, the constraint dimensions with corresponding priority are inferred first, and then the compliance status of other constraint dimensions is supplemented to generate the initial compliance status of a single time window. For example, in the time window of beam components in the main construction stage, geometric compliance and process compliance are inferred first to clarify the allowable range of geometric dimension deviations and construction specifications of connection nodes during the installation process of beam components. At the same time, the preliminary requirements for performance compliance are supplemented to form the initial compliance status of this time window.

[0094] This clarifies the key points and direction of compliance status reasoning, solves the problems of scattered related knowledge and lack of unified constraint standards, and ensures the pertinence and comprehensiveness of compliance status reasoning; unified reasoning logic standardizes the reasoning process and judgment criteria, avoids standard confusion and result deviation in sequential reasoning, and ensures the consistency of compliance status reasoning logic across different time windows; it realizes the sequentialization of compliance status, synchronizing the initial compliance status with the construction progress and component installation status, solving the problem of the disconnect between compliance requirements and construction progress; the initial compliance status provides objects for subsequent verification iterations and provides compliance guidance for component construction, ensuring that the construction process is based on evidence.

[0095] Verify the compatibility between the initial compliance status and the fusion features, and confirm that the requirements of geometric compliance, performance compliance, and other dimensions are consistent with the geometric topology parameters and semantic constraints in the fusion features. For example, if the cross-sectional dimensions of a beam member in the fusion features are specific values, it is necessary to verify whether the corresponding dimensional constraints in the initial compliance status match these values. If the requirement of fire resistance level 1 in the semantic constraints is met, it is necessary to verify whether the corresponding fire resistance standard is matched in the performance compliance dimension. At the same time, it is confirmed that the compliance requirements do not exceed the load-bearing capacity of the member's own properties.

[0096] Simultaneously verify the temporal continuity of the initial compliance status. According to the order of time windows, check whether the compliance status of different time windows conforms to the process connection logic and whether there are any temporal contradictions or sudden changes in compliance requirements. For example, if the previous time window is the beam component installation process, the compliance status focuses on installation accuracy, and the next time window is the acceptance process, the compliance status focuses on accuracy acceptance. It is necessary to ensure that the compliance requirements of the two are consistent and do not contradict each other. At the same time, confirm that the compliance status of each time window matches the process type and there is no misalignment.

[0097] Record the verification results and clarify the reasons for failure, including incomplete reasoning logic, missing constraint dimensions, and deviations in the reasoning process. For example, unreasonable priority settings for constraint dimensions may lead to a disconnect between compliance requirements and construction priorities, or the special compliance requirements for a certain type of component may not be reflected in the knowledge graph, resulting in incomplete reasoning results, or the correspondence between time windows and compliance status may be misaligned. Based on the verification results, optimize the reasoning logic or supplement the constraint dimensions of the BIM domain knowledge graph. For example, adjust the priority of constraint dimensions and correct the correspondence between time windows and compliance dimensions to ensure that the reasoning logic is more in line with component needs and construction processes. For example, if the compliance requirements for pipe connections of a certain electromechanical component are missing, the corresponding specification entities, constraint attributes, and relationships need to be supplemented, and the knowledge graph updated to ensure that subsequent reasoning has complete knowledge support. If the reasoning process is flawed, re-combine the reasoning logic with the time window and correct the initial compliance status of the corresponding time window.

[0098] After optimization and improvement, the reasoning process is re-executed to generate a new initial compliance state. The aforementioned dual verification process is then executed again, iterating repeatedly until a compliance state sequence for the component during the construction phase is generated. This effectively solves the problems of disconnect between compliance state and component integration features, as well as inconsistent timing, ensuring the relevance and coherence of the compliance state sequence. Based on the iterative optimization of the verification results, the reasoning logic is improved and the knowledge graph is supplemented, addressing the shortcomings of conventional compliance reasoning, such as lack of iteration and large result deviations. The final generated compliance state sequence is clear in timing and requirements, highly adapted to integration features and construction progress, providing precise time-sequential compliance guidance for component construction and high-quality compliance basis for subsequent steps.

[0099] like Figure 2 As shown, the spatiotemporal state sequence forming the component includes:

[0100] Using fusion features as attribute benchmarks and process time windows as time-series benchmarks, the installation status sequence and compliance status sequence are aligned to clarify the relationship between installation status and compliance status within the same time window;

[0101] Based on the relationship, the installation status, compliance status and corresponding attributes within the same time window are bidirectionally coupled to generate a state unit of a single time window, which is arranged according to the connection logic of the process to form the initial spatiotemporal state.

[0102] Verify the temporal coherence of the initial spatiotemporal state and the attribute matching degree between the state unit and the fusion feature. Based on the verification results, correct the correlation relationship, and re-execute the alignment and coupling to form the spatiotemporal state sequence of the component.

[0103] The installation status sequence and the compliance status sequence are currently isolated from each other, lacking a clear spatiotemporal correlation, and cannot reflect the correspondence between the actual status and compliance requirements within the same time window. By using fusion features as the attribute benchmark and the time window of the process as the temporal benchmark, the installation status sequence and the compliance status sequence are aligned to ensure that the actual status within the same time window corresponds to the compliance standard. During the alignment process, the attribute compatibility of the two sequence nodes within each time window is verified in conjunction with the attribute benchmark. For example, if the installation status of a structural beam component in a certain time window is "process in progress," the corresponding compliance status needs to be aligned to geometric compliance and process compliance to ensure attribute matching and avoid alignment misalignment where the installation status and compliance requirements are unrelated.

[0104] After alignment, the relationship between installation status and compliance status within the same time window is clearly defined, thus avoiding alignment deviations caused by focusing only on timing and ignoring attribute matching in conventional alignment. Establishing a clear relationship between installation status and compliance status within the same time window breaks the limitation of the two sequences being isolated from each other, realizes the spatiotemporal correspondence between actual construction status and compliance standards, and makes subsequent integration more targeted. It provides a clear basis for subsequent coupling operations and verification corrections, ensuring that the alignment results are consistent with component attributes and actual construction.

[0105] Aligning the two sequences only clarifies the relationship, but fails to achieve a deep integration of the actual state and compliance standards, thus failing to reflect the spatiotemporal integration attribute of the component within a specific time window. Based on the relationship, the installation state, compliance state, and corresponding attributes within the same time window are bidirectionally coupled. This bidirectional coupling refers to the bidirectional linkage of "adaptation coupling of installation state → compliance state" and "verification coupling of compliance state → installation state." On the one hand, the actual information of the installation state is adapted to the compliance requirements of the corresponding compliance state, clarifying whether the actual state meets the compliance standards and marking the adaptation results. On the other hand, the actual information of the installation state is verified with reference to the compliance standards of the compliance state, and the verification results are marked, achieving a deep binding between the two.

[0106] Based on the bidirectional coupling results, the installation status and compliance status are integrated to generate a single time window status unit. For structural components, the status unit focuses on strengthening the coupling information of geometric compliance and performance compliance with the structural attributes of the fusion characteristics, and annotates the adaptation verification results of the core load-bearing parts. For decorative components, the status unit focuses on strengthening the coupling information of process compliance and acceptance compliance with the decorative attributes of the fusion characteristics, and annotates the adaptation verification results of appearance and materials. For electromechanical components, the status unit focuses on strengthening the coupling information of geometric layout and installation specifications with the electromechanical attributes of the fusion characteristics, and annotates the adaptation verification results of installation spacing and connection methods.

[0107] According to the connection logic of the process, the generated state units are arranged in an orderly manner to ensure that the initial spatiotemporal state after arrangement is synchronized with the time sequence and continuous with the time window. For example, the state units of the foundation construction stage are arranged first, and then the state units of the main construction stage are arranged. The arrangement of each state unit is in line with the connection logic of the process to avoid time sequence disorder and process reversal. The connection relationship between adjacent state units is marked to ensure the temporal continuity of the initial spatiotemporal state. The arranged state units are integrated to form the initial spatiotemporal state of the component.

[0108] This achieves deep integration of installation status and compliance status, solving the problem of isolated sequences that cannot reflect spatiotemporal integration attributes. It ensures that each status unit can reflect the actual construction status and compliance requirements of the component within the same time window. The status unit provides a unified carrier for the spatiotemporal attributes of the component, facilitating the injection of nodes and information retrieval in subsequent construction constraint diagrams. The arrangement according to the process connection logic ensures the temporal continuity of the initial spatiotemporal status, conforms to the actual construction process, improves the completeness and relevance of the attributes of the initial spatiotemporal status, and provides high-quality basic data for subsequent verification iterations and construction constraint diagram construction.

[0109] According to the time window sequence, verify the temporal continuity of the initial spatiotemporal state and confirm the reasonable connection relationship between adjacent state units; for example, whether the arrangement and connection logic of state units and the construction schedule are consistent, and confirm that the time window and process corresponding to each state unit are consistent with the actual construction process. For example, the state unit of beam installation process should be arranged after the foundation construction state unit and before the acceptance process state unit; verify the attribute matching degree between state unit and fusion feature, and confirm that the coupling information of state unit and fusion feature attributes are not contradictory, and that the compliance and installation status of state unit are completely matched with fusion feature; for example, the geometric compliance requirements of state unit of structural beam component must be completely matched with the beam cross-sectional dimensions and connection method in fusion feature.

[0110] Record the verification results and clarify the reasons for verification failure, including unreasonable correlation and coupling or layout deviations; correct the correlation based on the verification results to ensure that the installation status and compliance status within the same time window are accurately correlated and the attributes are matched; if it is a coupling or layout deviation, correct the adaptation and verification logic in the bidirectional coupling process, regenerate the state unit of the corresponding time window, and adjust the arrangement order of the state unit to ensure the temporal continuity of the initial spatiotemporal state; after the correction is completed, regenerate a new initial spatiotemporal state and perform verification, iterate repeatedly until the verification is passed, and form the spatiotemporal state sequence of the component.

[0111] This allows for a comprehensive investigation of temporal and attribute deviations in the initial spatiotemporal state, effectively resolving issues of discontinuous temporal sequences and mismatched attributes, and ensuring the accuracy and reliability of the spatiotemporal state sequence. Based on iterative optimization of the verification results, the correlation, coupling logic, and arrangement order are continuously improved. Ultimately, the spatiotemporal state sequence achieves integrated fusion of temporal and attribute dimensions, providing accurate and comprehensive support for subsequent construction constraint diagram construction and conflict detection.

[0112] S3. Using components as entity nodes and design specifications and construction techniques as constraint nodes, and injecting spatiotemporal state sequences into the corresponding entity nodes to construct a construction constraint diagram, the construction constraint diagram is learned through a time-series graph neural network to predict the conflict propagation path between components during the construction phase.

[0113] Furthermore, constructing the construction constraint diagram includes:

[0114] Using components as entity nodes and design specifications and construction techniques as constraint nodes, spatiotemporal state sequences, fusion features and engineering control labels are injected into entity nodes, and constraint dimensions are marked on constraint nodes.

[0115] Associate the fusion features of entity nodes with the constraint dimensions of corresponding constraint nodes, and associate the constraint relationships between entity nodes and constraint nodes according to the connection logic of the process.

[0116] Add time sequence identifiers to entity nodes, constraint nodes and constraint relationships according to time windows, and combine fusion features and engineering control tags to verify the rationality of the association between entity nodes and constraint nodes, as well as the temporal consistency between constraint relationships and spatiotemporal state sequences.

[0117] Based on the verification results, the constraint relationships and timing identifiers are corrected, and the entity nodes, constraint nodes, and corrected constraint relationships are integrated to form a construction constraint diagram.

[0118] The system uses components as entity nodes and design specifications and construction techniques as constraint nodes. Design specifications cover geometric compliance, performance compliance, and acceptance compliance, while construction techniques cover construction operations, temporary fixing, and process connections. For example, the cross-sectional dimensions of concrete structures and the installation techniques of beam components are both independent constraint nodes. Spatiotemporal state sequences, fusion features, and engineering control tags are injected into entity nodes to ensure that they reflect the spatiotemporal attributes of components in each time window. The system also clarifies the constraint focus of different entity nodes and optimizes the information injection order according to the engineering type of the component. For structural components, geometric topological features, performance semantic constraints, and geometric compliance coupling information of spatiotemporal state sequences are injected first. For decorative components, process semantic constraints and process compliance coupling information of spatiotemporal state sequences are injected first.

[0119] The constraint dimensions are labeled to the constraint nodes. For example, constraint nodes corresponding to the cross-sectional dimensions of concrete structures are labeled with geometric compliance dimensions; constraint nodes corresponding to the seismic construction process of beam members are labeled with performance compliance and process compliance dimensions; constraint nodes corresponding to the acceptance specifications of decorative components are labeled with acceptance compliance dimensions. This achieves a precise correspondence between the component entity and the construction constraints, ensures the continuity and completeness of the injected information, avoids interference from redundant information, and enables the entity nodes to fully reflect the spatiotemporal attributes, inherent properties, and control requirements of the component. It provides a basis for the subsequent association between entity nodes and constraint nodes, solves the problem of unfounded and chaotic association of nodes in conventional constraint diagrams, and ensures that the construction constraint diagram is built to fit the actual components and construction needs.

[0120] Entity nodes and constraint nodes are still isolated from each other and have not formed an effective constraint relationship. The integration characteristics of entity nodes and the constraint dimensions of corresponding constraint nodes should be linked to ensure that the relationship conforms to the component attributes and constraint requirements. For example, entity nodes of structural beam components should be linked to the constraint nodes corresponding to the concrete structure cross-sectional dimension specifications with geometric compliance dimension and the constraint nodes corresponding to the seismic design specifications with performance compliance dimension. Entity nodes of decorative wall components should be linked to the constraint nodes corresponding to the wall decoration construction process with process compliance dimension and the constraint nodes corresponding to the decoration engineering acceptance specifications with acceptance compliance dimension.

[0121] During the association process, the semantic similarity between the fusion features of entity nodes and the constraint dimensions of constraint nodes is compared, and unreasonable associations with low semantic similarity are eliminated. The constraint relationship between entity nodes and constraint nodes is associated in combination with the connection logic of the construction process to ensure that the constraint relationship fits the actual construction process and to clarify the constraint relationship between components of different construction processes. For entity nodes with connection relationships, such as beam components and column components, the constraint nodes of the two are associated to ensure that the constraint relationship can reflect the constraint requirements in the connection of construction processes.

[0122] This solves the problems of unfounded and chaotic node associations in conventional constraint diagrams, ensuring the accuracy of the association between entity nodes and constraint nodes; by optimizing the associations through process connection logic, the constraint relationships are made to fit the actual construction process, avoiding the defect of constraint relationships being out of sync with the construction sequence; it ensures that the constraint relationships of core components are more comprehensive, laying the foundation for targeted conflict detection in the future; and it achieves the organic combination of nodes and constraint relationships, breaking the limitation of isolated nodes and providing a basis for adding subsequent sequence identifiers.

[0123] The construction process is highly time-dependent, with differences in component status and constraint requirements across different time windows, making it impossible to reflect the temporal changes in constraint relationships and node status during construction. By adding time-series markers to nodes and constraint relationships according to time windows, the validity of each node and constraint relationship within different time windows can be clearly defined. For example, for a certain foundation component node, a corresponding time-series marker is marked within the foundation construction time window. Outside of this time window, if some constraints of this node become invalid, the time-series marker is also updated accordingly.

[0124] By combining fusion features and engineering control tags, the rationality of the association between entity nodes and constraint nodes is verified, as well as the temporal consistency between constraint relationships and spatiotemporal state sequences. This is to confirm that the fusion features of entity nodes match the constraint dimensions of constraint nodes, and that they conform to the component engineering type, engineering control priority, and actual construction. For example, for core load-bearing components with high control priority, it is checked whether they are associated with core constraint nodes, and whether decorative components are not associated with unrelated structural constraint nodes. At the same time, it is confirmed that the temporal effective range of constraint relationships is consistent with the time window of the spatiotemporal state sequence, and that the temporal changes of constraint relationships are synchronized with the state changes of components in the spatiotemporal state sequence. For example, the constraint relationships of a component within the installation time window must be synchronized with the installation status of the time window.

[0125] This injects a temporal dimension into the construction constraint diagram, enabling the graph structure to reflect changes in node states and constraint relationships within different time windows, thus adapting to the temporal learning needs of subsequent temporal graph neural networks. Dual verification comprehensively checks for defects such as unreasonable associations and temporal contradictions in the graph structure, effectively reducing the probability of deviation in subsequent conflict prediction. The verification process combines fusion features and engineering management tags to align with component attributes and engineering management requirements, ensuring the temporal consistency and rationality of the graph structure and further improving its functionality.

[0126] Based on the verification results, the constraint relationships and timing identifiers are corrected. For constraint relationships with unreasonable associations, invalid associations are deleted, such as the association between decorative components and structural code constraint nodes, and missing associations are added to ensure that the association between entity nodes and constraint nodes fits and integrates the characteristics and constraint dimensions. For constraint relationships with inconsistent timing, the effective time range and timing identifier of the constraint relationship are adjusted to ensure that it is synchronized with the time window and timing identifier of the spatiotemporal state sequence. For example, the effective time window of a constraint relationship is corrected to be consistent with the component installation time window to avoid timing contradictions. After the correction is completed, the verification is re-executed, and after repeated iterations, the entity nodes, constraint nodes and the corrected constraint relationships are integrated to form the construction constraint diagram.

[0127] This ensures the logical coherence and timing accuracy of the construction constraint diagram, avoiding the problems of conventional constraint diagrams lacking iteration and leaving residual defects. The final construction constraint diagram integrates nodes, constraints, timing, and attributes, ensuring that the graph structure can accurately support subsequent timing graph neural network learning and conflict prediction, providing graph structure support for the timing and accuracy of the overall technical solution.

[0128] like Figure 3 As shown, the predicted conflict propagation paths between components during the construction phase include:

[0129] The temporal propagation of constraint relationships in construction constraint graphs and the attribute association between fusion features and constraint dimensions are learned through hierarchical learning of temporal graph neural networks.

[0130] Based on the results of hierarchical learning, potential conflicts in the construction constraint diagram are identified, corresponding entity nodes and constraint nodes are associated to predict and generate initial conflict paths, and the triggering type and propagation order of potential conflicts are labeled.

[0131] By combining the constraint dimension and the temporal coherence of the spatiotemporal state sequence, the rationality of the propagation order and the attribute fit of the trigger type are verified. Based on the verification results, the learning weights of the initial conflict path and the temporal graph neural network are adjusted. After the verification is passed, the conflict propagation path between components during the construction stage is predicted and generated.

[0132] Conventional processing methods for temporal graph neural networks cannot accurately capture the temporal transmission patterns of constraint relationships or effectively learn the attribute associations between fusion features and constraint dimensions, thus failing to provide reliable support for subsequent conflict identification and path prediction. This paper clarifies that temporal graph neural networks consist of a temporal transmission layer and an attribute association layer. It learns layer by layer, according to time windows, the effectiveness changes of constraint relationships within different time windows, the temporal dependencies between constraint relationships, and the transmission logic of constraint relationships between different processes. It also clarifies the constraint transmission patterns between entity nodes caused by process connections and, combined with the temporal changes of spatiotemporal state sequences, captures the impact of component state changes on the temporal transmission of constraint relationships, ensuring that the temporal transmission layer can accurately grasp the dynamic temporal changes of constraint relationships.

[0133] Then, learning is categorized by constraint dimensions to clarify the correlation strength between different fusion features and corresponding constraint dimensions. For example, the correlation strength between the geometric topology features of structural components and the geometric compliance dimension is higher than that of decorative components. The matching logic between the fusion features of entity nodes and the constraint dimensions of constraint nodes is learned to identify potential correlation contradictions where fusion features do not meet the requirements of constraint dimensions. At the same time, combined with engineering control tags, the attribute correlation learning of high-control-priority components is strengthened to ensure that the attribute correlation layer can accurately capture the matching relationship and potential contradictions between component attributes and constraint requirements. This ensures that the layered learning results can truly reflect the temporal transmission rules and attribute correlation logic of construction constraint diagrams.

[0134] This effectively avoids interference from redundant information, improves the accuracy and relevance of time-series neural network learning, and solves the shortcomings of conventional learning in capturing the temporal transmission of constraint relationships and attribute associations. The learning of the temporal transmission layer accurately grasps the temporal dynamic change law of constraint relationships, providing reliable support for the temporal prediction of subsequent conflict propagation paths. The learning of the attribute association layer accurately captures the matching relationship between component attributes and constraint requirements and potential contradictions, laying the foundation for subsequent identification of potential conflicts. It ensures that the learning process is consistent with the actual construction and component attributes, providing high-quality learning basis for subsequent conflict identification and path prediction.

[0135] After the hierarchical learning is completed, only the temporal transmission and attribute association rules of the construction constraint diagram are obtained. However, potential conflicts are not clearly identified, nor are specific conflict propagation paths formed, which makes it impossible to carry out targeted conflict management and path optimization in the future. Based on the hierarchical learning results, potential conflicts in the construction constraint diagram are identified. According to the potential contradictions between the fusion features output by the attribute association layer and the constraint dimensions, potential conflicts of attribute type are identified, that is, conflicts in which the fusion features of entity nodes do not meet the constraint dimension requirements of the corresponding constraint nodes. According to the temporal transmission rules of constraint relationships output by the temporal transmission layer, potential conflicts of temporal type are identified, that is, conflicts in which the temporal transmission of constraint relationships does not conform to the process connection logic or is inconsistent with the temporal sequence of time and space states.

[0136] The system associates the entity nodes and constraint nodes corresponding to potential conflicts, predicts and generates initial conflict paths. For each identified potential conflict, starting from the conflict trigger node, it associates the entity nodes and constraint nodes subsequently affected by the conflict based on the temporal transmission rules of the constraint relationships learned by the temporal transmission layer, sorts out the transmission logic of the conflict, and forms an initial conflict path. For example, if the geometric dimensions of a beam component exceed the requirements of the geometric compliance constraint node, it associates the entity nodes of the beam component and the corresponding geometric compliance constraint node as the starting point, and associates the entity nodes of the column components affected by it and the subsequent acceptance constraint nodes to form an initial conflict path of "beam component entity node → geometric compliance constraint node → column component entity node → acceptance constraint node".

[0137] The trigger types and propagation order of potential conflicts are marked. The trigger types include five categories: geometric conflicts, performance conflicts, process conflicts, acceptance conflicts, and timing conflicts. For example, in attribute-type conflicts, geometric dimension discrepancies are marked as geometric conflicts, performance index discrepancies are marked as performance conflicts, and timing-type conflicts are marked as timing conflicts. The propagation order is marked according to the transmission logic of the initial conflict path and the timing identifier of the construction constraint diagram. The propagation steps of the conflict from the starting node to the subsequent nodes are marked according to the chronological order and the order of constraint relationship transmission. The time window, associated nodes, and constraint relationships corresponding to each propagation step are clearly defined.

[0138] This addresses the shortcomings of conventional conflict identification methods, such as their simplistic approach, numerous omissions, and false conflicts. It ensures that identified potential conflicts align with actual construction conflicts. The initial conflict path identifies the conflict point and its propagation path, clarifying the scope of impact and transmission logic. The annotation of trigger types and propagation order makes the initial conflict path information more complete, providing a clear basis for subsequent verification and adjustment. It ensures that the initial conflict path and annotation information align with actual construction, component attributes, and constraint requirements, laying the foundation for subsequent verification and adjustment.

[0139] By combining the temporal coherence of constraint dimensions and spatiotemporal state sequences, the rationality of the propagation order and the attribute compatibility of trigger types are verified to confirm that the propagation order conforms to the temporal transmission rules of constraint relationships and the logic of process connection. For example, the propagation order of geometric conflicts of beam components must conform to the process logic of "beam installation → beam acceptance → subsequent component connection". There should be no situation where the conflict of subsequent component connection propagates before the conflict of beam installation. It is ensured that the time window of conflict propagation is synchronized with the state changes of components in the spatiotemporal state sequence. At the same time, it is confirmed that the actual attributes and constraint dimensions of the trigger type and potential conflict are completely matched. For example, conflicts caused by geometric dimension discrepancies in fusion features are marked as geometric conflicts, which are consistent with the geometric compliance dimension of constraint nodes; conflicts caused by performance index discrepancies are marked as performance conflicts, which are consistent with the performance compliance dimension of constraint nodes.

[0140] Record the verification results and clarify the reasons for verification failures, including deviations in the initial conflict path, trigger type labeling, and learning weights of the time-series neural network. Adjust the deviations based on the verification results. For paths with unreasonable propagation order, re-organize the conflict propagation order and correct the arrangement of path nodes by combining process connection logic and time sequence identifiers. For cases of missing paths, supplement the entity nodes and constraint nodes affected by the conflict based on the hierarchical learning results to improve the conflict propagation path. For cases of misaligned paths, delete invalid nodes and re-associate the corresponding nodes to ensure that the path is reasonable.

[0141] For annotations with insufficient attribute fit, the trigger type is re-annotated by combining the constraint dimension and the actual conflict attribute to ensure fit with the constraint dimension and component attribute. For problems caused by learning weight deviation, the weight allocation of the two learning layers is adjusted by combining the layered learning results and the verification results. The learning weight related to the process connection logic in the time sequence transmission layer is strengthened to improve the accuracy of constraint relationship time sequence transmission learning. The learning weight related to high control priority components and constraint dimensions in the attribute association layer is strengthened to improve the accuracy of component attribute and constraint dimension association learning. The learning weight corresponding to redundant information that causes learning deviation is weakened.

[0142] After adjustment, a new initial conflict path is regenerated, and the process is iterated repeatedly to organize the verified conflict propagation paths, forming conflict propagation paths between components during the construction phase. This comprehensively identifies deviations between the initial conflict paths and the annotation information, effectively resolving issues such as unreasonable propagation order and misaligned trigger type annotations, ensuring the accuracy of the conflict propagation paths and the accuracy of the annotation information. The conflict propagation paths clearly define the triggering source, propagation order, and scope of impact of the conflict, aligning with actual construction and project management needs, and providing a precise basis for subsequent conflict management, parameter adjustment, and edge weight optimization.

[0143] S4. Verify the prediction results of the conflict propagation path, perform impact tracing and classification on the verified prediction results, and adjust the parameters of the time series neural network and the edge weights of the construction constraint graph by combining the comparison data of the actual construction results and the prediction results.

[0144] Specifically, adjusting the parameters of the time-series graph neural network includes:

[0145] Based on the actual construction results, the predicted results of the conflict propagation path are verified. For the verified conflict propagation path, the starting component and timing identifier of the conflict are traced based on the trigger type, propagation order and constraint relationship. The impact of the conflict is classified according to the engineering control label and fusion characteristics.

[0146] Based on the results of impact tracing and classification, and combined with the comparison data between actual construction results and prediction results, the types of deviations are distinguished and the parameters of the corresponding time series neural network to be adjusted are clarified.

[0147] Based on the fusion features and constraint dimensions, adjustment weights are assigned to the corresponding parameters. After adjusting the parameters in a hierarchical and iterative manner, the conflict propagation path is verified. Once the verification is successful, the adjustment weights and parameter adjustment ranges are locked.

[0148] The conflict propagation path was not verified in conjunction with actual construction results, making it impossible to determine the accuracy and fit of the prediction results. The prediction results of the conflict propagation path were verified in conjunction with actual construction results to confirm whether the prediction results are consistent with the actual construction, distinguishing between conflict propagation paths that passed verification and those that failed verification. For paths that failed verification, parameter adjustments were not included; only the deviations were recorded for re-verification after subsequent adjustments. For conflict propagation paths that passed verification, conflict impact tracing and classification were carried out. Based on trigger type, propagation order, and constraint relationships, the starting components and timing identifiers of the conflict were traced. The starting components of the source conflict were clearly identified as the entity nodes that triggered the conflict, and the timing identifiers of the source conflict clearly identified the specific time window and corresponding process stage of the conflict trigger. Combined with the process connection logic, the correlation between conflict triggering and timing transmission was confirmed, and the parameter deviation points existing in the learning process of the time sequence graph neural network within that time window were identified.

[0149] Based on engineering control labels and fusion characteristics, the impact of conflicts corresponding to verified conflict propagation paths is graded. First, the basic level is determined by control priority based on engineering control labels. Conflicts triggered by high control priority components are classified as high-level; conflicts triggered by medium control priority components are classified as medium-level; and conflicts triggered by low control priority components are classified as low-level. Then, the basic level is fine-tuned based on fusion characteristics. For example, for core load-bearing components that belong to high control priority, if their fusion characteristics include key performance constraints such as seismic resistance and load-bearing capacity, and the conflict triggering type is performance conflict, their conflict impact level is upgraded to the key level within the high-level category. If the conflict triggering type is general geometric conflict and does not affect core performance, the high-level basic level is maintained.

[0150] After grading, the impact level of each verification path through conflict propagation is clearly defined, and the grading basis is marked. This avoids directional deviations caused by adjusting parameters based on inaccurate prediction results, accurately identifies the initiating component and timing identifier of the conflict, clarifies the learning level where parameter deviations may exist, and provides a clear positioning basis for subsequent parameter adjustments. By grading according to engineering control tags and fusion characteristics, the impact degree of different conflicts is clarified, and the priority of parameter adjustments is realized, ensuring that subsequent parameter adjustments can focus on key points and improve the pertinence and efficiency of parameter adjustments. This ensures that the verification, tracing, and grading work are aligned with the actual construction and component attributes.

[0151] By analyzing the comparison data between actual construction results and predicted results, and combining the impact tracing and classification results, deviation manifestations corresponding to the conflict propagation path are extracted and verified. Based on the deviation manifestations, tracing results, and conflict triggering types, deviation types are distinguished to ensure that the deviation types accurately correspond to the learning structure and parameter types of the time-series graph neural network. These include time-series deviations, attribute-series deviations, and comprehensive deviations. Time-series deviations correspond to learning deviations in the time-series propagation layer. The main deviation manifestations are the deviations in the conflict propagation order and the prediction of time-series identifiers, and the inaccurate learning of the time-series propagation rules of constraint relationships. The tracing results show that the deviations are located in the time-series propagation layer. The core issue is that the parameters of the time-series propagation layer in the time-series graph neural network do not accurately capture the time-series dependency rules of constraint relationships.

[0152] The attribute-type bias corresponds to the learning bias in the attribute association layer. The bias manifests mainly as bias in trigger type prediction and bias in identifying conflict initiation components. The source results show that the bias is located in the attribute association layer, and the core issue is that the parameters of the attribute association layer do not accurately capture the association logic between the fusion features and the constraint dimensions. The comprehensive bias corresponds to the collaborative bias of the two learning structures. The bias manifests as the overall bias in the conflict propagation path. The source results show that the bias involves the temporal transmission layer and the attribute association layer, and is mostly caused by the joint conflict of temporal and attribute types. The core issue is the insufficient synergy of the learning parameters of the two layers.

[0153] Based on the identified deviation types, the parameters that need to be adjusted in the time-series graph neural network are determined accordingly. For time-series deviations, the core parameters of the time-series propagation layer are adjusted, including the time-series dependency weights of constraint relationships, time window recognition parameters, and process connection logic learning parameters. These parameters directly affect the learning accuracy of the time-series propagation rules of constraint relationships. For attribute-related deviations, the core parameters of the attribute association layer are adjusted, including the association strength parameters between fused features and constraint dimensions, and the matching parameters between component attribute classification parameters and constraint dimensions. These parameters directly affect the learning accuracy of the association between fused features and constraint dimensions. For comprehensive deviations, the coordination parameters of the two-layer learning structure are adjusted, including the weight allocation parameters of the two layers and the learning result fusion parameters. These parameters directly affect the coordination of the two layers of learning, avoiding the overall deviation caused by the disconnect of layered learning.

[0154] This approach clarifies the core sources of parameter deviations, providing a clear directional basis for parameter adjustments. The identification of corresponding adjustment parameters ensures a correspondence between deviation types, learning levels, and adjustment parameters, avoiding inefficiency or new deviations caused by blind parameter adjustments. By combining the differentiation of deviation types based on source tracing and grading results, it ensures that parameter adjustments focus on key deviations and parameters, aligning with the level of conflict impact, thus improving the efficiency and accuracy of parameter adjustments. Simultaneously, it links with the hierarchical learning structure, ensuring consistency between parameter adjustments and the learning structure, providing a clear parameter range for subsequent hierarchical iterative adjustments.

[0155] After clarifying the adjustment parameters and deviation types, applying a uniform adjustment range and weight to all parameters will result in insufficient adjustment of key parameters and excessive adjustment of non-key parameters, failing to achieve precise optimization. Based on the fusion features and constraint dimensions, adjustment weights are assigned to each parameter, with higher adjustment weights assigned to parameters corresponding to core components and key attributes. For example, in the attribute association layer, the association strength parameter between the fusion features and performance constraint dimensions of core load-bearing components is assigned a higher adjustment weight than the corresponding parameter for ordinary components. Similarly, higher adjustment weights are assigned to adjustment parameters corresponding to key constraint dimensions. For example, in the time-series transmission layer, the time-series transmission parameter of geometric compliance constraints is assigned a higher adjustment weight than the corresponding parameter for ordinary process constraints.

[0156] The parameters of the time-series graph neural network are adjusted iteratively in a hierarchical manner based on weight adjustment. First, the parameters of the attribute association layer are adjusted, then the parameters of the time-series propagation layer are adjusted, and finally the parameters of the two collaborative layers are adjusted. Each layer of parameter adjustment is carried out according to the weight priority, with high-weight parameters adjusted first and with a larger adjustment magnitude, and low-weight parameters adjusted secondarily and with a moderate adjustment magnitude. After each parameter adjustment, the conflict propagation path is re-predicted and the fit between the new prediction result and the actual construction is verified. If the verification passes, the current adjustment weight and parameter adjustment magnitude are locked. If the verification fails, the reasons for the unimproved deviation are analyzed based on the verification results, the adjustment weight allocation and parameter adjustment magnitude are finely adjusted, and the adjustment and verification are carried out again until the prediction results of all conflict propagation paths can be verified by the actual construction and the prediction accuracy meets the construction requirements.

[0157] This addresses the shortcomings of conventional parameter adjustments, which lack weighting and prioritization, thereby improving the targeting and efficiency of parameter adjustments. Layered iterative adjustments and closed-loop verification avoid deviations caused by single parameter adjustments, ensuring the effectiveness of parameter adjustments. The adjustment range and weight allocation are gradually optimized to achieve the optimal configuration of parameters in the time-series neural network. After parameter locking, the accuracy of subsequent conflict propagation path prediction is ensured to be stable, providing model support for the accuracy of overall BIM data intelligent matching and conflict detection methods. This achieves a deep fit between parameter adjustments and actual construction, component attributes, and constraint requirements.

[0158] Specifically, adjusting the edge weights of the construction constraint diagram includes:

[0159] Based on the results of impact tracing and classification, as well as the type of deviation, the constraint relationships corresponding to the construction constraint diagram are associated;

[0160] Based on the constraint dimensions, engineering control labels, and fusion characteristics, determine the adjustment direction and allocate the adjustment magnitude for the edge weights of the constraint relationships to be adjusted;

[0161] Adjust edge weights hierarchically according to constraint relationships, verify the consistency between the adjusted constraint relationships and the time sequence identifiers, and the adaptability of edge weights to constraint dimensions. After the verification is passed, lock the adjustment direction and adjustment range of the edge weights.

[0162] Based on the impact tracing and classification results, as well as the deviation types, the constraint relationships in the construction constraint diagram are correlated. First, the role of each constraint relationship in the conflict propagation process is clarified, such as the initial constraint relationship that triggers the conflict and the intermediate constraint relationships during propagation. Combining the starting component and timing identifier of the tracing, the entity node and time window corresponding to the constraint relationship are identified, ensuring that the correlated constraint relationships are consistent with the conflict source and the timing transmission logic. Second, based on the conflict impact classification results, the constraint relationships involved in conflicts of corresponding levels are correlated. Constraint relationships associated with high-impact level conflicts are prioritized, while constraints associated with medium- and low-impact level conflicts are correlated sequentially according to priority, clarifying the priority of constraint relationships corresponding to different impact levels.

[0163] Finally, based on the deviation type, the corresponding constraint relationships are associated. For time-series deviations, the focus is on associating them with time-series constraints; for attribute-series deviations, the focus is on associating them with compliance constraints; and for comprehensive deviations, both time-series and compliance constraints are associated to ensure that the associated constraint relationships accurately correspond to the parameter deviation types. At the same time, constraint relationships in the construction constraint diagram that do not participate in any conflict propagation, have no deviation associations, have low impact levels, and have no corresponding parameter deviations are removed. These constraint relationships do not need to be adjusted and their initial edge weights are maintained. Only the constraint relationships that need to be adjusted are included in the subsequent adjustment scope.

[0164] This clarifies the scope and priority of the constraints to be adjusted, ensuring that edge weight adjustments focus on key areas; it achieves deep linkage between impact tracing, hierarchical results, deviation types, and construction constraint diagram structure, ensuring that the associated constraints align with the actual conflict propagation, temporal logic, and component attributes; it provides a clear object basis for subsequent adjustment work, avoids ineffective adjustments, improves the efficiency and targeting of edge weight adjustments, and lays a solid foundation for determining the direction and magnitude of subsequent adjustments.

[0165] Once the constraints to be adjusted are clarified, conventional edge weight adjustments simply change the numerical value without specifying the direction or allocating a reasonable adjustment range. This results in the adjusted edge weights failing to reflect the priority and actual needs of the constraints and failing to adapt to the adjusted time-series graph neural network parameters. By combining constraint dimensions, engineering control labels, and fusion features, the adjustment direction of edge weights for the constraints to be adjusted is determined. For time-series constraints with time-series deviations, if the deviation manifests as inaccurate learning of the time-series propagation rules of the constraints, the adjustment direction is determined to be to increase the edge weights, strengthening the time-series propagation weight of the constraint in the graph structure, ensuring its coordination with the adjusted time-series propagation layer parameters, and improving the learning and prediction accuracy of the time-series logic. If the constraint is over-strengthened, leading to a propagation order deviation, the adjustment direction is determined to be to decrease the edge weights.

[0166] For compliance constraint relationships associated with attribute-type deviations, if the deviation manifests as inaccurate learning of the association between fused features and constraint dimensions, the adjustment direction is determined to be to increase edge weights, strengthen the attribute matching weight of the compliance constraint relationship in the graph structure, and adapt to the adjusted attribute association layer parameters. If the compliance constraint relationship is over-strengthened, leading to deviation in trigger type labeling, the adjustment direction is determined to be to decrease edge weights. For the two types of constraint relationships associated with comprehensive deviations, the adjustment direction is determined based on the dominant type of deviation manifestation. If the dominant time-series deviation is the first to be adjusted, the edge weights of the time-series constraint relationship will be adjusted first; if the dominant attribute-type deviation is the first to be adjusted, the edge weights of the compliance constraint relationship will be adjusted first, ensuring that the adjustment direction is consistent with the deviation type and model parameter adjustment logic.

[0167] Combining constraint dimensions, engineering control tags, and fusion features, adjustment ranges are assigned to the constraint relationships to be adjusted. Constraint relationships corresponding to core constraint dimensions are assigned larger adjustment ranges, while those corresponding to ordinary constraint dimensions are assigned moderate adjustment ranges. For example, the adjustment range for performance compliance constraints is greater than that for process compliance constraints, ensuring more thorough weight optimization of core constraints. Constraint relationships associated with high-control-priority components receive further increased adjustment ranges, while those associated with medium- and low-control-priority components decrease sequentially. For instance, the performance compliance constraints for core load-bearing components receive a higher adjustment range than those for ordinary decorative components. Constraint relationships corresponding to key attributes within the fusion features of core components are assigned larger adjustment ranges, while those corresponding to ordinary attributes are assigned moderate adjustment ranges, ensuring that the adjustment ranges align with the core requirements of the components.

[0168] This ensures consistency between the adjustment direction and deviation type, as well as the logic of model parameter adjustment, achieving coordinated adaptation of the construction constraint diagram and the neural network parameters of the time sequence diagram, avoiding the defect of their disconnect. The differentiated allocation of adjustment magnitude, combined with constraint dimensions, engineering control labels, and fusion features, reflects the priority of constraint relationships and actual needs, ensuring more thorough weight optimization of core constraints and high-impact constraints, and improving the targeting and rationality of edge weight adjustments. It provides a clear basis for subsequent layered adjustment work, ensuring that the adjustment process is orderly, accurate, and in line with actual construction and component attributes, laying the foundation for subsequent verification and locking.

[0169] After clarifying the direction and magnitude of the adjustment, arbitrarily adjusting all constraints may lead to confusion in the temporal logic of the construction constraint diagram and make it impossible to accurately control the adjustment effect. Adjust the edge weights layer by layer according to the constraint relationship to ensure that the adjusted edge weights match the importance of the constraint dimension and the component attributes. For example, increase the edge weights of performance compliance constraints by a certain magnitude, and adjust the edge weights of process compliance constraints by a certain magnitude. Also, ensure that the adjusted edge weights match the temporal transmission rules and time window requirements. For example, increase the edge weights of temporal constraints that trigger temporal deviations by a certain magnitude to strengthen their temporal transmission effect.

[0170] Verify the consistency between the adjusted constraint relationships and the time sequence identifiers, as well as the adaptability of the edge weights and constraint dimensions. Specifically, confirm that after the edge weights are adjusted, the effective time range of the constraint relationship remains consistent with the time sequence identifier. For example, after adjusting the edge weights of a certain time sequence constraint relationship, its corresponding time sequence identifier still matches the beam installation time window. Also, confirm that the magnitude of the edge weights is adapted to the importance of the constraint dimensions. The edge weights corresponding to high-importance constraint dimensions are higher than those corresponding to low-importance constraint dimensions. The adjustment range meets the requirements for matching the constraint dimensions, engineering control labels, and fusion features. For example, the adjustment range of the edge weights corresponding to the performance compliance constraint dimension meets the attribute requirements of high-priority control components, with no adaptation deviation.

[0171] After successful verification, the adjustment direction and magnitude of the edge weights are locked, and the adjusted edge weights are updated in the constructed construction constraint diagram to form an optimized construction constraint diagram. This ensures that the graph structure and the parameters of the time sequence diagram neural network are compatible and coordinated. This solves the problems of time sequence chaos and poor adjustment effect caused by disordered adjustment, ensuring that the adjustment process is orderly and accurate, and improving adjustment efficiency. It effectively solves the defects of edge weight adjustment deviation, time sequence inconsistency and insufficient adaptability, ensuring that the adjusted edge weights are accurate and reliable. The locking of edge weights and the generation of the optimized construction constraint diagram realize the collaborative adaptation of the construction constraint diagram and the parameters of the time sequence diagram neural network, enabling the construction constraint diagram to more accurately reflect the importance of constraint relationships and the level of conflict impact, further improving the accuracy of the overall BIM data intelligent matching and conflict detection methods.

Claims

1. A method for intelligent matching and conflict detection of BIM data, characterized in that, include: Acquire the project's BIM data, construction schedule, and design specifications; parse the geometric topological features of components in the BIM data; mine the semantic constraint features of components based on the design specifications; and fuse the geometric topological features and semantic constraint features of the same component to generate the component's fused features. Based on the construction schedule, the installation status sequence of components during the construction phase is generated. Based on the fusion features and the preset BIM domain knowledge graph, reasoning is performed to generate the compliance status sequence of components during the construction phase. The installation status sequence and the compliance status sequence are aligned and coupled to form the spatiotemporal status sequence of the components. Using components as entity nodes and design specifications and construction techniques as constraint nodes, and injecting spatiotemporal state sequences into the corresponding entity nodes to construct a construction constraint diagram, the construction constraint diagram is learned through a time-series graph neural network to predict the conflict propagation path between components during the construction phase. The predicted results of the conflict propagation path are verified. The verified prediction results are then used for impact tracing and classification. The parameters of the time series neural network and the edge weights of the construction constraint graph are adjusted by combining the comparison data of the actual construction results and the predicted results.

2. The BIM data intelligent matching and conflict detection method as described in claim 1, characterized in that, The fusion features of the generated components include: Analyze the geometric topological features of components in BIM data, mine the semantic constraint features of components based on design specification documents, and combine the engineering association between geometric topology and semantic constraints in the BIM domain to assign domain weights. Based on the construction stage and the engineering type of the component, fusion weights are assigned to the geometric topology features and semantic constraint features of the same component respectively. The geometric topology features and semantic constraint features are weighted and fused according to the domain weight and fusion weight to generate the initial fused features of the component. Perform BIM domain compliance verification on the initial fusion features, reverse the domain weights based on the reasons for the verification failure, and re-execute weighted fusion to generate fusion features of the components.

3. The BIM data intelligent matching and conflict detection method as described in claim 2, characterized in that, The sequence of installation states of the generated components during the construction phase includes: The construction schedule is deconstructed hierarchically, and the time windows, connection logic and schedule deviation tolerance of different processes are extracted and associated with the component engineering identifiers to form the relationship between construction progress and components. The definition includes an installation state system encompassing construction phases, process execution, project acceptance, and temporary fixing. The relationships are then adapted to the installation state system to generate an initial state sequence. The rationality of the initial state sequence is verified based on the construction technology and project attributes corresponding to the component's engineering type. Based on the verification results, the connection logic and time window are corrected, the initial state sequence is iteratively adjusted, and engineering control tags are added to generate the installation state sequence of components during the construction phase.

4. The BIM data intelligent matching and conflict detection method as described in claim 3, characterized in that, The compliance status sequence of the generated components during the construction phase includes: Entities, relationships, and attributes with associated features are selected and integrated from a pre-defined BIM domain knowledge graph; By combining the constraint dimensions of the BIM domain knowledge graph with the fusion feature analysis, reasoning logic is formed. Based on the time window of the process, the related entities, relationships and attributes are traversed through the reasoning logic to infer the compliance status of the components in different time windows and generate the initial compliance status. Verify the matching between the initial compliance status and the fusion features, as well as the temporal coherence of the initial compliance status. Based on the verification results, optimize the reasoning logic or supplement the constraint dimensions of the BIM domain knowledge graph. Iterate the reasoning and verification to generate a sequence of compliance statuses for components during the construction phase.

5. The BIM data intelligent matching and conflict detection method as described in claim 4, characterized in that, The spatiotemporal state sequence of the forming component includes: Using fusion features as attribute benchmarks and process time windows as time-series benchmarks, the installation status sequence and compliance status sequence are aligned to clarify the relationship between installation status and compliance status within the same time window; Based on the relationship, the installation status, compliance status and corresponding attributes within the same time window are bidirectionally coupled to generate a state unit of a single time window, which is arranged according to the connection logic of the process to form the initial spatiotemporal state. Verify the temporal coherence of the initial spatiotemporal state and the attribute matching degree between the state unit and the fusion feature. Based on the verification results, correct the correlation relationship, and re-execute the alignment and coupling to form the spatiotemporal state sequence of the component.

6. The BIM data intelligent matching and conflict detection method as described in claim 5, characterized in that, The construction constraint diagram includes: Using components as entity nodes and design specifications and construction techniques as constraint nodes, spatiotemporal state sequences, fusion features and engineering control labels are injected into entity nodes, and constraint dimensions are marked on constraint nodes. Associate the fusion features of entity nodes with the constraint dimensions of corresponding constraint nodes, and associate the constraint relationships between entity nodes and constraint nodes according to the connection logic of the process. Add time sequence identifiers to entity nodes, constraint nodes and constraint relationships according to time windows, and combine fusion features and engineering control tags to verify the rationality of the association between entity nodes and constraint nodes, as well as the temporal consistency between constraint relationships and spatiotemporal state sequences. Based on the verification results, the constraint relationships and timing identifiers are corrected, and the entity nodes, constraint nodes, and corrected constraint relationships are integrated to form a construction constraint diagram.

7. The BIM data intelligent matching and conflict detection method as described in claim 6, characterized in that, The predicted conflict propagation paths between components during the construction phase include: The temporal propagation of constraint relationships in construction constraint graphs and the attribute association between fusion features and constraint dimensions are learned through hierarchical learning of temporal graph neural networks. Based on the results of hierarchical learning, potential conflicts in the construction constraint diagram are identified, corresponding entity nodes and constraint nodes are associated to predict and generate initial conflict paths, and the triggering type and propagation order of potential conflicts are labeled. By combining the constraint dimension and the temporal coherence of the spatiotemporal state sequence, the rationality of the propagation order and the attribute fit of the trigger type are verified. Based on the verification results, the learning weights of the initial conflict path and the temporal graph neural network are adjusted. After the verification is passed, the conflict propagation path between components during the construction stage is predicted and generated.

8. The BIM data intelligent matching and conflict detection method as described in claim 7, characterized in that, The parameters of the adjusted time-series neural network include: Based on the actual construction results, the predicted results of the conflict propagation path are verified. For the verified conflict propagation path, the starting component and timing identifier of the conflict are traced based on the trigger type, propagation order and constraint relationship. The impact of the conflict is classified according to the engineering control label and fusion characteristics. Based on the results of impact tracing and classification, and combined with the comparison data between actual construction results and prediction results, the types of deviations are distinguished and the parameters of the corresponding time series neural network to be adjusted are clarified. Based on the fusion features and constraint dimensions, adjustment weights are assigned to the corresponding parameters. After adjusting the parameters in a hierarchical and iterative manner, the conflict propagation path is verified. Once the verification is successful, the adjustment weights and parameter adjustment ranges are locked.

9. The BIM data intelligent matching and conflict detection method as described in claim 8, characterized in that, Adjusting the edge weights of the construction constraint diagram includes: Based on the results of impact tracing and classification, as well as the type of deviation, the constraint relationships corresponding to the construction constraint diagram are associated; Based on the constraint dimensions, engineering control labels, and fusion characteristics, determine the adjustment direction and allocate the adjustment magnitude for the edge weights of the constraint relationships to be adjusted; Adjust edge weights hierarchically according to constraint relationships, verify the consistency between the adjusted constraint relationships and the time sequence identifiers, and the adaptability of edge weights to constraint dimensions. After the verification is passed, lock the adjustment direction and adjustment range of the edge weights.

Images