A BIM-based building project management system

Abstract

The application belongs to the technical field of building information model and construction project intelligent management, and discloses a building project management system based on BIM, which comprises a BIM model semantic analysis and component coding module, a field multi-source state mapping and credible calibration module, a component-task-resource heterogeneous graph construction module, a constraint perception construction period deviation propagation prediction module, a multi-target rolling construction scheme optimization module, and an execution feedback and BIM state updating module; the system establishes heterogeneous relationships among components, tasks, resources, spaces and quality events, predicts construction period deviation and delay propagation paths according to task readiness, relationship types, remaining time difference and state credibility, and selects a rolling construction scheme through BIM construction state deduction and multi-target evaluation, so that the application can improve construction period risk prediction accuracy, reduce resource conflicts, space conflicts and frequent scheme changes, and realize closed-loop management of building projects.

Description

Technical Field

[0001] This invention belongs to the field of building information modeling and intelligent management of construction projects, specifically relating to a BIM-based building project management system. Background Technology

[0002] Construction projects involve component installation, material supply, personnel allocation, machinery scheduling, work site handover, and quality acceptance. Different construction activities are subject to process dependencies, space occupancy, and resource competition. BIM can carry the geometric information, professional information, and construction attributes of building components. Therefore, it is necessary to combine the BIM model with the construction site status, construction plan, and resource allocation to realize construction status analysis, schedule risk prediction, and construction plan adjustment.

[0003] Existing building project management systems mainly associate BIM components with construction plans and display component status based on the completion rate of the plan and manually reported progress. Data on component arrival, material arrival, personnel attendance, machinery operation, quality acceptance, and space occupancy are scattered and lack a unified object mapping and reliable status calibration mechanism. Furthermore, the dynamic relationship between components, tasks, resources, construction areas, and quality events has not been established, making it difficult to identify the actual source and scope of impact of construction task deviations.

[0004] Existing schedule risk analysis typically employs fixed thresholds, critical path calculations, and single-task time series forecasting. It fails to differentiate the different impacts of pre-processing, space occupancy, resource sharing, and quality rework on task status propagation, nor does it determine whether the deviation of pre-processing has exceeded the remaining time difference. This can easily amplify minor delay errors that could be absorbed by task time differences. Furthermore, the forecast results mainly rely on manual handling by management personnel, which can easily lead to resource conflicts, spatial conflicts, and frequent changes in construction plans. Summary of the Invention

[0005] To address the shortcomings of existing building project management systems, which primarily visualize and link BIM models with construction plans but lack unified modeling of the dynamic impact relationships between components, tasks, resources, construction spaces, and quality events, and are unable to accurately assess the propagation impact of delays in preceding tasks, resource competition, spatial conflicts, and quality rework on subsequent construction tasks, this invention provides a BIM-based building project management system.

[0006] The present invention provides a BIM-based building project management system, including a BIM model semantic parsing and component coding module, a site multi-source state mapping and reliable calibration module, a component-task-resource heterogeneous graph construction module, a constraint-aware schedule deviation propagation prediction module, a multi-objective rolling construction scheme optimization module, and an execution feedback and BIM state update module.

[0007] The BIM model semantic parsing and component coding module parses the component categories, spatial locations, floors, disciplines, and construction areas in the BIM model, generating BIM component codes, construction task codes, component spatial bounding volumes, and component-task mapping relationships.

[0008] The on-site multi-source state mapping and reliable calibration module maps component arrival data, material arrival data, personnel attendance data, machinery operation data, construction progress data, quality acceptance data, space occupancy data, and on-site image data to corresponding BIM components, construction tasks, and construction areas, and generates state reliability based on data integrity, time freshness, source consistency, and continuity of change.

[0009] The component-task-resource heterogeneous diagram construction module establishes BIM component nodes, construction task nodes, construction team nodes, construction machinery nodes, material batch nodes, construction area nodes, and quality event nodes, and establishes process prerequisite relationship edges, space occupation relationship edges, resource sharing relationship edges, component ownership relationship edges, material supply relationship edges, and quality rework relationship edges to generate the component-task-resource heterogeneous diagram.

[0010] The component-task-resource heterogeneous graph construction module further generates the task readiness level based on the completion level of the preceding tasks, the material satisfaction level, the personnel satisfaction level, the machinery satisfaction level, the work surface satisfaction level, the quality acceptance level, and the task status credibility. This reduces the overall task readiness level when any key construction condition is insufficient.

[0011] The constraint-aware schedule deviation propagation prediction module uses a relationship type-aware graph attention network to process pre-process relationships, space occupancy relationships, resource sharing relationships, component ownership relationships, material supply relationships, and quality rework relationships respectively. It generates relationship attention weights based on node dynamic status, edge constraint characteristics, and relationship status credibility.

[0012] The constraint-aware schedule deviation propagation prediction module generates predicted schedule deviations based on task node characteristics, task readiness, relationship attention weights, previous task schedule deviations, remaining time differences, and task status uncertainties. It then traces back the relationship attention weights from task nodes where the predicted schedule deviation exceeds the warning conditions to generate delay propagation paths.

[0013] If the schedule deviation of a preceding task does not exceed the corresponding remaining time difference, the schedule deviation will not be propagated to subsequent construction tasks. If the space occupancy time of two construction tasks does not overlap, spatial conflicts will not be propagated. If there is remaining capacity in shared resources, resource contention will not be propagated.

[0014] The multi-objective rolling construction scheme optimization module generates construction sequence adjustment actions, construction team reallocation actions, construction machinery time adjustment actions, material supply priority adjustment actions, and construction area switching actions based on the predicted schedule deviation and delay propagation path, and combines multiple compatible adjustment actions into candidate construction schemes.

[0015] The multi-objective rolling construction scheme optimization module predicts the task status, resource status, and construction area status of candidate construction schemes through BIM construction status simulation. It verifies the pre-process relationships, resource capacity, space occupation, safety passages, and quality acceptance control points of the candidate construction schemes, and selects the rolling construction scheme based on project schedule, construction cost, resource fluctuations, spatial conflicts, scheme change range, and prediction uncertainty.

[0016] The execution feedback and BIM status update module sends the rolling construction plan to the project management terminal and the on-site execution terminal, collects on-site confirmation information, task execution status and quality acceptance results, updates the construction status, actual start time, actual completion time and quality status of BIM components, and reconstructs the affected nodes and relation edges based on the update results.

[0017] Furthermore, this invention provides a method for operating a BIM-based building project management system, including step S1, parsing the BIM model and establishing component-task mapping relationships; step S2, collecting multi-source data from the site and generating status credibility; step S3, establishing a component-task-resource heterogeneity diagram; step S4, generating task start-up readiness; step S5, generating predicted schedule deviations and delay propagation paths through a relationship type-aware graph attention network; step S6, generating and validating candidate construction schemes; step S7, selecting a rolling construction scheme based on the multi-objective scheme evaluation results; and step S8, collecting execution feedback and updating the BIM status and the component-task-resource heterogeneity diagram.

[0018] The beneficial effects of this invention are as follows: (1) By encoding BIM objects and modeling heterogeneous graph relationships, BIM components, construction tasks, construction teams, construction machinery, material batches, construction areas and quality events are set as different types of graph nodes. The associations are established through process prerequisite relationships, space occupation relationships, resource sharing relationships and quality rework relationships, so as to realize the unified expression of building models, construction plans, site conditions and construction resources, and solve the problem that the existing BIM project management system has scattered data and cannot describe the dynamic influence relationship of construction elements.

[0019] (2) By using task start-up readiness, relationship type perception graph attention and remaining time difference gating, the task construction conditions are determined based on the preceding tasks, materials, personnel, machinery, work surface and quality acceptance status. The propagation of schedule deviation is controlled based on relationship type and status credibility, and slight delays that do not exceed the remaining time difference are prevented from spreading to subsequent tasks. This achieves accurate generation of predicted schedule deviation and delay propagation paths, solving the problems of high false alarm rate and inability to locate the source of delay in fixed threshold and single task prediction methods.

[0020] (3) Through BIM construction status simulation and multi-objective rolling optimization, the candidate construction schemes formed by adjusting the construction sequence, redistributing the work teams, adjusting the machinery time period, adjusting the material supply priority and switching the construction area are verified for construction constraints. The construction period, cost, resource fluctuations, spatial conflicts, scheme change range and prediction uncertainty are comprehensively evaluated to achieve rolling scheme selection that meets the on-site construction conditions and the change range is controllable, and solves the problem that manual adjustment of the construction plan can easily cause resource conflicts, spatial conflicts and repeated changes in the scheme. Attached Figure Description

[0021] The accompanying drawings are provided to further understand the present invention and form part of the specification. They are used together with the embodiments of the present invention to explain the invention and do not constitute a limitation thereof.

[0022] Figure 1 This is the overall architecture diagram of the construction project management system of the present invention; Figure 2 This is a diagram generated for predicting the propagation of construction period deviations and generating a rolling construction scheme according to the present invention. Figure 3 This is a diagram showing the relationship between the prediction test results of the project schedule deviation of this invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the scope of protection of the present invention.

[0024] In the embodiments provided in this specification, data acquisition, model training, and construction management operations without specific technical specifications are implemented in accordance with conventional technical specifications in the fields of building information modeling, construction planning management, and graph neural networks. Sensors, computing devices, and field terminals without specified manufacturers are all conventional products obtained through legitimate channels.

[0025] The following embodiments are limited to the construction phase of prefabricated public building main structure and electromechanical installation. The construction site includes activities such as hoisting of prefabricated components, binding of steel bars, pouring of concrete, installation of electromechanical pipelines, quality acceptance and handover of work surface. Different construction tasks have simultaneous relationships of process prerequisite, space occupation, resource sharing and quality rework.

[0026] The BIM model in the following embodiments includes a building model, a structural model, and a mechanical and electrical model. BIM components have component codes, component categories, spatial locations, design dimensions, construction floors, construction areas, and professional attributes. Construction plan data includes construction tasks, planned start time, planned completion time, prerequisite tasks, planned duration, and planned resources.

[0027] The multi-source data in the following embodiments includes component arrival data, material arrival data, personnel attendance data, machinery operation data, construction progress data, quality acceptance data, space occupancy data, and on-site image data. All multi-source data are associated with each other according to BIM component codes, construction task codes, construction area codes, and collection time.

[0028] Example 1, see Figures 1-3 The present invention provides a BIM-based building project management system, including a BIM model semantic parsing and component coding module, a site multi-source state mapping and reliable calibration module, a component-task-resource heterogeneous graph construction module, a constraint-aware schedule deviation propagation prediction module, a multi-objective rolling construction scheme optimization module, and an execution feedback and BIM state update module.

[0029] The BIM model semantic parsing and component coding module receives basic building project data in addition to BIM model, construction plan data, and PCB design data. It performs semantic parsing on building components, structural components, MEP components, and construction areas in the BIM model, generates BIM component nodes, construction area nodes, and component spatial bounding volumes, and establishes associations between BIM components and construction tasks.

[0030] The on-site multi-source state mapping and reliable calibration module receives component arrival data, material arrival data, personnel attendance data, machinery operation data, construction progress data, quality acceptance data, space occupancy data, and on-site image data. It maps the on-site multi-source data to the corresponding BIM components, construction tasks, and construction areas, and generates state reliability based on data integrity, time freshness, source consistency, and continuity of change.

[0031] The component-task-resource heterogeneous diagram construction module sets BIM components, construction tasks, construction teams, construction machinery, material batches, construction areas, and quality events as different types of diagram nodes. Based on the construction plan, BIM spatial relationships, resource allocation records, and quality acceptance records, it establishes edges for process prerequisite relationships, space occupancy relationships, resource sharing relationships, component ownership relationships, material supply relationships, and quality rework relationships, thereby generating the component-task-resource heterogeneous diagram.

[0032] The constraint-aware schedule deviation propagation prediction module generates task readiness based on the completion level of prerequisite tasks, material satisfaction level, personnel satisfaction level, machinery satisfaction level, work surface satisfaction level, and quality acceptance level. It processes the component-task-resource heterogeneous graph through the relation type awareness graph and the network, and predicts the task schedule deviation and project delay propagation path based on task time difference, relation type, node status reliability, and construction constraint status.

[0033] The multi-objective rolling construction scheme optimization module generates multiple candidate construction schemes based on task schedule deviations, project delay propagation paths, available construction resources, construction area occupancy status, construction quality control points, and the current construction plan. It then selects the rolling construction scheme based on project schedule, construction cost, resource fluctuations, spatial conflicts, scheme change range, and prediction uncertainty.

[0034] The execution feedback and BIM status update module sends the rolling construction plan to the project management terminal and the on-site execution terminal, collects on-site confirmation information, task execution status and quality acceptance results, updates the construction status, actual start time, actual completion time and quality status of BIM components according to the task execution status, and resends the updated BIM status to the on-site multi-source status mapping and reliable calibration module.

[0035] The building project in this embodiment is a prefabricated public building with nine floors above ground and two floors below ground, with a total construction area of ​​48,600 m². 2 The main structure includes precast columns, precast beams, composite floor slabs and cast-in-place core tubes. The electromechanical engineering includes ventilation ducts, fire protection ducts, water supply and drainage ducts and cable trays. The project construction period is 420 days.

[0036] The project's BIM model adopts a unified coordinate system. The BIM component code consists of project code, professional code, floor code, construction area code, component category code, and sequence code. The construction task code consists of professional code, floor code, construction flow section code, work type code, and sequence code.

[0037] The BIM model semantic parsing and component coding module associates precast components, cast-in-place components, and electromechanical components with their corresponding installation tasks. It also groups tasks with process dependencies within the same construction area into a construction task chain, enabling changes in component status to be transmitted to construction tasks and allowing adjustments to construction tasks to update the planned status of components.

[0038] This embodiment does not merely use the BIM model as a 3D visualization interface, but converts BIM components, construction tasks, resource entities, construction areas, and quality events into heterogeneous graph nodes with relational types and dynamic states. It uses construction constraints and graph relationships to jointly predict the propagation of schedule deviations and feeds the prediction results back to the construction scheme optimization and BIM status update process.

[0039] Example 2: Based on all the above examples, the BIM model semantic parsing and component coding module specifically includes a model lightweighting unit, a component semantic parsing unit, a component coding unit, a task association unit, and a spatial partitioning unit.

[0040] The lightweight model unit removes material rendering data and non-construction management attributes from the BIM model, while retaining the component geometry, component center coordinates, component spatial bounding volume, component category, floor information, professional information, and construction attributes, and then creates a lightweight BIM model.

[0041] The component semantic parsing unit reads the entity type, family type, spatial level and attribute fields of each component in the BIM model, divides the components into structural components, architectural components and mechanical and electrical components, and determines the floor and construction area to which the component belongs based on the component center coordinates and the component spatial enclosing volume.

[0042] The component coding unit generates a unique BIM component code according to the project code, professional code, floor code, construction area code, component category code and sequence code, and writes the unique BIM component code into the lightweight BIM model and project management database.

[0043] The task association unit reads the work breakdown structure from the construction plan data and establishes a many-to-many association between construction tasks and BIM component codes. The same construction task can be associated with multiple BIM components, and the same BIM component can be associated with component arrival tasks, component hoisting tasks, connection construction tasks, and quality acceptance tasks in sequence.

[0044] The spatial zoning unit divides the construction area according to the construction flow section, floor, axis grid and fire compartment, and assigns BIM components to the corresponding construction area according to the overlap relationship between the component space enclosure and the construction area. Each construction area records the number of people allowed to work at the same time, the number of machines allowed, the material stacking capacity and the safety passage occupancy limit.

[0045] For precast component hoisting tasks, the task association unit establishes a connection between precast components, transport vehicles, tower cranes, hoisting teams, component storage yards, and installation work areas. For electromechanical pipeline installation tasks, the task association unit establishes a connection between electromechanical components, installation teams, work platforms, material batches, and corresponding construction areas.

[0046] When the construction plan is adjusted, the task-related unit retains the original construction task code and the adjusted version number, and records the changes in the plan start time, plan completion time, resource allocation and prerequisite tasks, so as to avoid the inability to track the source of the schedule deviation after directly overwriting the original construction plan.

[0047] Example 3: Based on all the above examples, the on-site multi-source state mapping and trusted calibration module specifically includes a data access unit, a timing synchronization unit, a BIM object mapping unit, a state conflict calibration unit, and a state trustworthiness generation unit.

[0048] The data access unit receives component QR code scanning data, material warehousing data, personnel attendance data, machinery operation data, mobile terminal progress reporting data, quality acceptance data, construction area positioning data, and on-site image recognition data, and adds data source, collection time, construction area code, and corresponding object code to each data entry.

[0049] The time synchronization unit aggregates mechanical operation data and construction area positioning data according to a 1-minute time window, aggregates personnel attendance data and material warehousing data according to a 30-minute time window, records construction progress data and quality acceptance data according to the event occurrence time, and maps the aggregated multi-source on-site data to a unified project timeline.

[0050] The BIM object mapping unit first performs direct mapping based on BIM component codes, construction task codes, and construction area codes. When the site data lacks object codes, it performs auxiliary mapping based on the data acquisition location, acquisition time, site work team, and the current work area in the construction plan.

[0051] The status conflict calibration unit compares the construction status of the same task generated by different data sources. When the mobile terminal reports that the task has been completed but the corresponding BIM component has not passed the quality acceptance, the task status is set to construction completed and awaiting acceptance. When the mechanical operation data shows that the tower crane has not entered the corresponding hoisting period and the component QR code has not been installed and scanned, the individual progress report result is not determined as the component installation completed.

[0052] The state credibility generation unit generates state credibility based on data integrity, collection time freshness, consistency of multiple source states, and continuity of state changes. State credibility is reduced when data sources are missing, collection time exceeds the corresponding data update cycle, or multiple source states conflict with each other.

[0053] When the task status credibility is below 0.60, the constraint-aware schedule deviation propagation prediction module retains the credibility status of the task in the previous time window and marks the task as a task to be reviewed. When the status credibility reaches 0.60 but is below 0.80, the weight of the task propagating status information to adjacent tasks is reduced. When the status credibility reaches 0.80, the component-task-resource heterogeneous graph is updated using the current task status.

[0054] The on-site multi-source state mapping and trusted calibration module generates the full project baseline state at 02:00 every day and updates the on-site dynamic state every 15 minutes. When events such as component hoisting completion, material not arriving, quality acceptance failure, and construction area closure occur, incremental state updates are immediately triggered.

[0055] Example 4: Based on all the above examples, the component-task-resource heterogeneous graph construction module specifically includes a heterogeneous node generation unit, a relation edge generation unit, a task readiness calculation unit, a relation type encoding unit, and a dynamic graph update unit.

[0056] The heterogeneous node generation unit establishes BIM component nodes, construction task nodes, construction team nodes, construction machinery nodes, material batch nodes, construction area nodes, and quality event nodes respectively. Each node includes static attributes, dynamic status, update time, and status reliability.

[0057] The static attributes of construction task nodes include task category, planned duration, planned start time, planned completion time, relevant specialty, and relevant construction area. The dynamic status of construction task nodes includes actual completion rate, actual number of personnel involved, material availability, machinery availability, work surface availability, quality acceptance rate, and time already consumed.

[0058] The relationship edge generation unit establishes process prerequisite relationship edges based on the logical relationships in the construction plan, establishes space occupancy relationship edges based on the spatial enclosure of BIM components and the occupancy time of the construction area, establishes resource sharing relationship edges based on the configuration of work teams, machinery and materials, establishes component ownership relationship edges based on the mapping between BIM components and construction tasks, establishes material supply relationship edges based on the supply records between material batches and construction tasks, and establishes quality rework relationship edges based on quality events and rework tasks.

[0059] The edge attributes of the process prerequisite relationship edge include the prerequisite relationship type, planned time difference, remaining time difference, and completion delivery conditions. The edge attributes of the space occupation relationship edge include the space overlap ratio, the degree of time overlap, and the safety interval requirement. The edge attributes of the resource sharing relationship edge include the shared resource type, resource demand, and resource competition level. The edge attributes of the quality rework relationship edge include the quality problem level, the estimated rework period, and the number of affected components.

[0060] The task readiness calculation unit calculates the task start-up readiness based on the completion level of prerequisite tasks, material availability, personnel availability, machinery availability, work surface availability, quality acceptance level, and task status reliability, according to Formula 1:

[0061] In Formula 1, Let represent the task readiness level for the i-th construction task at time t. For task status credibility, This indicates the degree of completion of the prerequisite tasks. To meet the material requirements, To the degree of staff satisfaction, For the degree of mechanical satisfaction, To meet the requirements of the working surface, For the degree of quality acceptance, These are the weights for prerequisite tasks, materials, personnel, machinery, working surface, and quality.

[0062] The completion status of prerequisite tasks, material availability, personnel availability, machinery availability, work surface availability, and quality acceptance status are all normalized to the range of 0-1. The sum of the weights of each item is set to 1. A weighted product method is used to reduce the overall task readiness when any key construction condition is insufficient.

[0063] In this embodiment, the weights of the pre-construction component hoisting task, materials, personnel, machinery, working surface, and quality are set to 0.20, 0.20, 0.10, 0.25, 0.15, and 0.10, respectively. The six weights corresponding to the electromechanical pipeline installation task are set to 0.15, 0.20, 0.20, 0.10, 0.25, and 0.10, respectively.

[0064] The relation type encoding unit sets independent relation codes and feature transformation parameters for different types of relation edges. The relation type perception graph attention network distinguishes state propagation caused by process dependence, spatial conflict, resource competition and quality rework based on the relation codes.

[0065] To ensure that graph attention weights simultaneously reflect node state, relation type, edge constraints, and data reliability, the attention weight from node j to node i in a relation edge of type r is calculated according to Formula 2:

[0066] In Formula 2, For the r-th type of relation edge, the relation attention weight is used to propagate the state information of node j to node i at time t. Let i and j be the dynamic state characteristics of nodes i and j, respectively. Let r be the feature transformation matrix corresponding to the r-th type of relation. Let be the edge constraint features between node i and node j. This is the attention vector corresponding to the r-th type of relation. Indicates feature channel splicing. For the reliability of the relation state, This is a credibility penalty coefficient. Let i be the set of adjacent nodes connected to node i through edge of type r.

[0067] The edge constraint characteristics of the process prerequisite relationship edge include the remaining time difference, the degree of completion of the prerequisite task, and the completion transfer conditions; the edge constraint characteristics of the space occupation relationship edge include the spatial overlap ratio and the degree of overlap of occupation time; the edge constraint characteristics of the resource sharing relationship edge include the resource gap and the degree of resource competition; and the edge constraint characteristics of the quality rework relationship edge include the expected rework period and the number of affected components.

[0068] When the site state changes, the dynamic graph update unit only updates the corresponding nodes, relationship edges and their two-hop neighborhoods, without rebuilding the entire component-task-resource heterogeneous graph, so that the construction site state can enter the relationship type-aware graph attention network in an incremental manner.

[0069] Regarding the adjustment of task readiness parameters, the quality weight is increased when there are unskippable quality acceptance control points in the task, the machinery weight is increased when the task relies on large machinery, and the work surface weight is increased when the task is located in a multi-disciplinary cross-construction area. When any key condition is zero, the task readiness is simultaneously reduced to zero.

[0070] Regarding the adjustment of relationship attention parameters, a higher residual time difference sensitivity is used for process-preceding relationships, a higher spatial overlap sensitivity is used for space occupancy relationships, a higher resource gap sensitivity is used for resource sharing relationships, and a higher rework estimated time sensitivity is used for quality rework relationships. When the reliability of the relationship status decreases, the attention weight of the corresponding relationship is reduced through a reliability penalty item.

[0071] Example 5: Based on all the above examples, the constraint-aware schedule deviation propagation prediction module specifically includes a time window generation unit, a heterogeneous graph feature propagation unit, a construction constraint gating unit, a schedule deviation prediction unit, and a delay propagation path extraction unit.

[0072] The time window generation unit takes the current management time as the end point of the window, extracts the node status and relation edge status of the component-task-resource heterogeneous graph in the previous seven natural days, and sets the next three natural days as the project schedule deviation prediction interval.

[0073] The heterogeneous graph feature propagation unit includes three layers of relation type perception graph attention layers. Each layer of relation type perception graph attention layer processes the pre-process relationship, space occupation relationship, resource sharing relationship, component ownership relationship, material supply relationship and quality rework relationship respectively, and retains the original dynamic state of the construction task node through residual connection.

[0074] The construction constraint gating unit controls the propagation of delays based on the remaining time difference of the preceding process relationship edge, the occupation time of the space occupation relationship edge, the resource gap of the resource sharing relationship edge, and the quality event status of the quality rework relationship edge. When the schedule deviation of the preceding task has not exhausted the remaining time difference, the delay will not be propagated to the subsequent task. When the space occupation time of two tasks does not overlap, the space conflict will not be propagated. When there is still remaining capacity in the shared resources, the resource competition will not be propagated.

[0075] The schedule deviation prediction unit predicts the task schedule deviation based on the characteristics of task nodes updated through the attention layer of the relation type perception graph, task start readiness, schedule deviation of preceding tasks exceeding the remaining time difference, and state uncertainty, and calculates it according to Formula 3:

[0076] In Formula 3, For the i-th construction task within the prediction interval Deviation in predicted project duration within the timeframe, Let b be the node feature after the task node passes through the Lth layer of the relationship type perception graph attention layer, and let b be the schedule deviation mapping vector. For task readiness, This is the readiness gap coefficient. Let be the set of prerequisite tasks for the i-th construction task. Pay attention to the weights of the relationships corresponding to the prerequisite relationships of the processes. The current schedule deviation of the preceding task j. Let $\frac{j}{i}$ be the remaining time difference between the preceding task $j$ and task $i$. Due to the uncertainty of the task status, is the uncertainty coefficient.

[0077] Formula 3 only propagates the portion of the schedule deviation of the preceding tasks that exceeds the remaining float to subsequent tasks, so that minor delays in preceding tasks with sufficient float will not be directly amplified into project delays. It also reflects the combined impact of materials, personnel, machinery, work surface, and quality status on the schedule through task start-up readiness.

[0078] The delay propagation path extraction unit traces the relationship attention weights in reverse from construction task nodes where the predicted schedule deviation exceeds 0.5 days. It extracts the main upstream nodes and relationship edges according to the relationship attention weights from high to low, and generates material supply delay paths, resource competition delay paths, spatial conflict delay paths, process pre-delay delay paths, and quality rework delay paths.

[0079] In this embodiment, tasks with a predicted schedule deviation of 1 day and a relational attention weight of 0.35 are set as key warning tasks, and tasks with a predicted schedule deviation of 2 days are set as high-risk tasks. In the BIM model, the associated components of key warning tasks and high-risk tasks are marked with different line types.

[0080] Regarding the adjustment of prediction window parameters, a seven-day historical window and a three-day prediction interval are used during the main structure construction stage, and a five-day historical window and a two-day prediction interval are used during the electromechanical centralized installation stage. If the historical window is too short, it cannot identify continuous changes in resource supply and construction efficiency, and if the historical window is too long, the early construction status will weaken the impact of recent emergencies.

[0081] Regarding the adjustment of uncertainty parameters, the uncertainty coefficient is reduced when the task status confidence reaches 0.80, and the uncertainty coefficient is increased and manual verification marks are added when the task status confidence is below 0.60, so that data missing and multi-source status conflicts will not be directly interpreted as deterministic schedule deviations.

[0082] Example 6: Based on all the above examples, the multi-objective rolling construction scheme optimization module specifically includes a candidate action generation unit, a BIM construction status deduction unit, a multi-objective scheme evaluation unit, a construction constraint verification unit, and a rolling scheme selection unit.

[0083] The candidate action generation unit generates construction sequence adjustment actions, construction team reallocation actions, construction machinery time period adjustment actions, material supply priority adjustment actions, and construction area switching actions based on the delay propagation path, and combines multiple compatible adjustment actions into candidate construction schemes.

[0084] The BIM construction status simulation unit replicates the current BIM component status, task status, resource status, and construction area status, loads each candidate construction scheme, and simulates the construction status for the next seven calendar days based on the pre-process relationships, space occupancy relationships, resource sharing relationships, and quality acceptance conditions among construction tasks.

[0085] The construction constraint verification unit checks whether the candidate construction schemes meet the requirements of non-skippable pre-process relationships, maximum number of workers in the construction team, maximum operating time of construction machinery, available quantity of materials, maximum occupancy of the construction area, safety passage maintenance conditions, and quality acceptance control points. Candidate construction schemes that do not meet the construction constraints are deleted.

[0086] The multi-objective scheme evaluation unit calculates the comprehensive management cost of candidate construction schemes based on predicted schedule deviations, construction costs, resource fluctuations, spatial conflicts, scheme change ranges, and prediction uncertainties, and selects the candidate construction scheme with the minimum comprehensive management cost according to Formula 4:

[0087] In Formula 4, For the selected rolling construction scheme, Let N be the set of candidate construction schemes that have passed the construction constraint verification, and let N be the number of construction tasks within the prediction interval. Let x be the predicted schedule deviation for the i-th construction task under candidate construction scheme x. To increase construction costs, For resource allocation fluctuations, This refers to the amount of space occupancy conflict. This represents the extent of change of the candidate construction plan compared to the previous rolling construction plan. Risks arising from uncertainty in the state of the proposed solution. These are the weights of the corresponding evaluation items.

[0088] In this embodiment, the weights for schedule deviation, construction cost, resource fluctuation, spatial conflict, scheme change, and uncertainty are set to 0.30, 0.15, 0.15, 0.20, 0.10, and 0.10, respectively.

[0089] The rolling scheme selection unit performs a full-project construction scheme optimization once a day at 06:00. It also performs event-triggered optimization when high-risk tasks occur, key machinery stops, key materials are delayed, or quality acceptance fails. Event-triggered optimization only adjusts the affected tasks and tasks within their three-hop neighborhood.

[0090] When the difference in the overall management cost of multiple candidate construction plans is less than 3%, the candidate construction plan with the smaller change compared to the previous rolling construction plan should be given priority to avoid frequent adjustments to construction teams, machinery hours and construction areas, which could cause chaos in on-site execution.

[0091] Regarding the adjustment of evaluation weights, the weight of schedule deviation is increased when key project milestones are approaching, the weight of spatial conflict is increased when multiple disciplines are working together, the weight of resource fluctuation is increased when the number of construction personnel and machinery changes frequently, and the weight of uncertainty is increased when the credibility of multi-source data on site decreases.

[0092] Regarding the adjustment of the rolling range parameter, the rolling range for regular daily plan adjustments is seven days in the future, while the rolling range for critical nodes is fourteen days in the future. If the rolling range is too short, it is difficult to identify the delay and transfer of the adjustment action to subsequent tasks. If the rolling range is too long, the uncertainty of the construction status increases and the reliability of the candidate scheme evaluation results is reduced.

[0093] Example 7: Based on all the above examples, this example provides a method for operating a BIM-based building project management system, which includes the following steps.

[0094] Step S1: Receive BIM model and construction plan data, parse the component category, spatial location, floor, discipline, and construction area of ​​BIM components, and generate BIM component codes, construction task codes, and component-task mapping relationships.

[0095] Step S2: Collect component arrival data, material arrival data, personnel attendance data, machinery operation data, construction progress data, quality acceptance data, space occupancy data, and on-site image data, and form on-site multi-source status data according to a unified project timeline.

[0096] Step S3: Map the multi-source status data on site to the corresponding BIM components, construction tasks and construction areas, and generate status credibility based on data integrity, time freshness, source consistency and change continuity.

[0097] Step S4: Establish BIM component nodes, construction task nodes, construction team nodes, construction machinery nodes, material batch nodes, construction area nodes, and quality event nodes respectively. Also, establish process prerequisite relationship edges, space occupation relationship edges, resource sharing relationship edges, component ownership relationship edges, material supply relationship edges, and quality rework relationship edges to generate a component-task-resource heterogeneous diagram.

[0098] Step S5: Generate task readiness based on the completion level of prerequisite tasks, material availability, personnel availability, machinery availability, work surface availability, quality acceptance level, and task status reliability.

[0099] Step S6: Update the status of construction task nodes through the relation type perception graph attention network, and generate predicted schedule deviation and delay propagation path based on task start readiness, relation attention weight, previous task schedule deviation, remaining time difference and task status uncertainty.

[0100] Step S7: Generate multiple candidate construction schemes based on the predicted schedule deviation and delay propagation path, perform BIM construction status simulation and construction constraint verification on the candidate construction schemes, and select a rolling construction scheme based on the project schedule, construction cost, resource fluctuations, spatial conflicts, scheme change range and prediction uncertainty.

[0101] Step S8: Send the rolling construction plan to the project management terminal and the on-site execution terminal, collect on-site confirmation information, task execution status and quality acceptance results, and update the BIM component status, construction task status and resource status.

[0102] Step S9: Determine whether new schedule deviations, resource conflicts, spatial conflicts, and quality rework events have occurred in the updated construction task. If a new abnormal state occurs, return to step S4 to update the component-task-resource heterogeneous diagram. If no new abnormal state occurs, return to step S2 to continue collecting multi-source status data on site.

[0103] Comparative Example 1: Comparative Example 1 adopts the traditional BIM four-dimensional schedule management method, which associates BIM components with construction tasks and planned time. It displays the planned status of components based on the current date and the planned completion percentage, but does not create a component-task-resource heterogeneous diagram, does not calculate task start-up readiness, and does not predict delay propagation paths.

[0104] The first comparison model generates progress warnings based on the difference between the planned completion rate and the actual reported completion rate. However, it cannot distinguish between schedule deviations caused by insufficient materials, resource competition, spatial conflicts, and quality rework, and it cannot identify whether the delay of the preceding tasks has consumed the task time difference.

[0105] Comparative Example 2: Comparative Example 2 uses a Long Short-Term Memory (LSTM) network to predict the schedule deviation of each construction task. The input data includes the historical completion rate of the task, the actual number of personnel involved, and the time already consumed. No graph relationship is established between the construction task and BIM components, construction areas, resource entities, and quality events.

[0106] Comparative Example 2 can predict schedule deviations based on historical changes of individual construction tasks, but it cannot identify the propagation of delays caused by shared tower cranes, occupation of the same construction area, and quality rework events among multiple construction tasks.

[0107] Comparative Example 3: Comparative Example 3 uses a homogeneous graph neural network to process construction task nodes, setting all relationship edges to the same type, without distinguishing between process prerequisite relationships, space occupation relationships, resource sharing relationships, and quality rework relationships, and without adjusting relationship weights based on state credibility.

[0108] Comparative Example 3 can propagate the status of adjacent tasks, but it will treat the effects of different construction relationships in the same way, easily identifying spatially adjacent tasks but with non-overlapping construction times as conflicting tasks, and amplifying field data with low status credibility to adjacent tasks.

[0109] Comparative Example 4: Comparative Example 4 adopts the schedule deviation propagation prediction method in Example 1, but cancels the multi-objective rolling construction scheme optimization. Project managers manually adjust the construction sequence and resource allocation based on the predicted schedule deviation.

[0110] Comparative Example 4 can identify the path of delay propagation, but the manual adjustment process cannot simultaneously verify the pre-process relationships, construction area occupancy, resource capacity and quality acceptance control points, and is prone to frequent plan changes in multiple consecutive management cycles.

[0111] Experimental Example 1: Prediction Test of Schedule Deviation This experimental case collected project operation data for the aforementioned prefabricated public building over a continuous twelve-week period, including 1,856 BIM components, 642 construction tasks, 12 construction teams, 8 construction machines, 326 material batches, 54 construction areas, and 137 quality incidents.

[0112] Using the project status over seven consecutive days as input and the actual schedule deviation generated by the construction tasks within the next three days as verification results, schedule deviation prediction was performed using Example 1, Comparative Example 1, Comparative Example 2, and Comparative Example 3, with each group running five times and the average value being taken.

[0113] The mean absolute error of the project duration deviation is determined based on the absolute difference between the predicted project duration deviation and the actual project duration deviation. The recall rate of high-risk tasks is determined based on the proportion of tasks whose actual project duration deviation reaches 2 days and are identified in advance. The false alarm rate is determined based on the proportion of tasks whose actual project duration deviation is less than 0.5 days but are identified as high-risk tasks.

[0114] The test results are shown in Table 1.

[0115] Table 1. Prediction results of schedule deviation:

[0116] As shown in Table 1, the average absolute error of the schedule deviation in Example 1 is 0.63 days, the recall rate of high-risk tasks is 92.6%, and the false alarm rate is 5.8%. This indicates that processing the pre-process relationships, space occupancy relationships, resource sharing relationships, and quality rework relationships according to the relationship type, and incorporating task time difference, start-up readiness, and status credibility into the schedule deviation propagation process can improve the accuracy of schedule deviation prediction.

[0117] Comparative Example 1 only compares the planned completion rate with the actual reported completion rate, and cannot identify the causes of schedule deviations and the propagation of delays, resulting in a high-risk task recall rate of only 61.3%. Comparative Example 2 only processes the time series of a single task and cannot identify resource competition and spatial conflicts between multiple tasks. Comparative Example 3 does not distinguish relationship types and does not calibrate state credibility, which easily leads to the propagation of erroneous states.

[0118] Experiment Example 2: Optimization Experiment of Rolling Construction Scheme This test case selected a continuous eight-week construction phase with a high degree of overlap between the main structure construction and electromechanical installation. Each week included tasks such as precast component hoisting, rebar tying, concrete pouring, ventilation duct installation, fire protection pipe installation, and cable tray installation.

[0119] Construction management plans were generated using Example 1, Comparative Example 1, Comparative Example 3, and Comparative Example 4, respectively. The average task delay time, construction machinery conflict time, resource allocation volatility, quality rework waiting time, and number of construction plan changes were statistically analyzed.

[0120] The resource allocation volatility is determined based on the degree of change in the number of construction team members and the configuration of machinery within two adjacent management cycles. The construction machinery conflict time is determined based on the cumulative time that the same machinery is assigned to multiple tasks during overlapping periods. The results are shown in Table 2.

[0121] Table 2 Optimization results of the rolling construction scheme:

[0122] As shown in Table 2, Example 1 uses BIM construction status simulation, multi-objective scheme evaluation and construction constraint verification to select a rolling construction scheme, which reduces the average task delay time to 0.74 days, reduces the construction machinery conflict time to 8.2 hours, and controls the number of construction scheme changes to 14 times.

[0123] Comparative Example 1 uses a fixed plan and manual early warning, resulting in fewer changes to the construction plan but failing to address resource competition and spatial conflicts in a timely manner, with an average task delay of 1.63 days. Comparative Example 3 cannot distinguish between different relationship types, and the generated resource adjustment plan exhibits high fluctuations in resource allocation. Comparative Example 4 relies on manual processing of prediction results, resulting in 31 plan changes within multiple consecutive management cycles, indicating that providing only the project duration prediction results cannot guarantee the stability of the construction adjustment plan.

[0124] This invention discloses a BIM-based building project management system that organizes BIM components, construction tasks, construction teams, construction machinery, material batches, construction areas, and quality events into a component-task-resource heterogeneous diagram. The system uses task readiness to reflect the fulfillment of preconditions for construction tasks. A relationship type awareness diagram distinguishes between processes, spatial conflicts, resource competition, and the propagation of schedule deviations caused by quality rework. Residual time difference gating prevents minor delays not exceeding the task time difference from being erroneously amplified. Furthermore, BIM construction status simulation and multi-objective rolling optimization generate rolling construction plans that meet construction constraints and have controllable change ranges. This achieves closed-loop management of building project construction status perception, schedule deviation prediction, delay cause tracking, construction plan optimization, and BIM status updates.

[0125] Finally, it should be noted that the above embodiments, comparative examples, and test examples are only used to illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Modifications and equivalent substitutions made by those skilled in the art to the foregoing technical solutions without departing from the technical concept of the present invention should all fall within the scope of protection of the present invention.

Claims

1. A BIM-based building project management system, characterized in that, It includes modules for BIM model semantic parsing and component coding, on-site multi-source state mapping and reliable calibration, component-task-resource heterogeneous graph construction, constraint-aware schedule deviation propagation prediction, multi-objective rolling construction scheme optimization, and execution feedback and BIM state update. The BIM model semantic parsing and component coding module parses the BIM model and construction plan data to generate BIM component codes, construction task codes, component spatial attributes, and component-task mapping relationships. The on-site multi-source state mapping and reliable calibration module maps on-site multi-source data to corresponding BIM components, construction tasks and construction areas, generating node dynamic states and state reliability. The component-task-resource heterogeneous graph construction module generates different types of graph nodes based on BIM components, construction tasks, construction resources, construction areas, and quality events, and generates different types of relationship edges based on construction plan relationships, BIM spatial relationships, resource allocation relationships, and quality rework relationships, thus obtaining the component-task-resource heterogeneous graph. The constraint-aware schedule deviation propagation prediction module processes the component-task-resource heterogeneous graph based on the construction constraint status and relationship type, and generates task start-up readiness, relationship attention weight, predicted schedule deviation, and delay propagation path. The multi-objective rolling construction scheme optimization module generates candidate construction schemes based on the predicted schedule deviation and delay propagation path, performs BIM construction status simulation and construction constraint verification on the candidate construction schemes, and selects the rolling construction scheme based on the project schedule, construction cost, resource fluctuations, spatial conflicts, scheme change range and prediction uncertainty. The execution feedback and BIM status update module updates the BIM component status, construction task status, and resource status based on the execution results of the rolling construction plan, and sends the update results to the component-task-resource heterogeneous diagram construction module.

2. The BIM-based building project management system according to claim 1, characterized in that, The BIM model semantic parsing and component coding module includes a model lightweighting unit, a component semantic parsing unit, a component coding unit, a task association unit, and a spatial partitioning unit; The component semantic parsing unit extracts the component category, spatial location, floor, discipline, and construction area based on the entity type, spatial level, and attribute fields of the BIM component. The component coding unit generates BIM component codes based on project code, professional code, floor code, construction area code, component category code, and sequence code; The task association unit establishes a component-task mapping relationship between BIM component codes and construction task codes based on the work breakdown structure in the construction plan data. The on-site multi-source state mapping and reliable calibration module maps on-site multi-source data to corresponding BIM components, construction tasks and construction areas based on object codes, collection locations, collection times, on-site work teams and current work areas, and generates state reliability based on data integrity, time freshness, source consistency and change continuity.

3. The BIM-based building project management system according to claim 1, characterized in that, The component-task-resource heterogeneous graph construction module includes a heterogeneous node generation unit, a relation edge generation unit, a task readiness calculation unit, a relation type encoding unit, and a dynamic graph update unit; The heterogeneous node generation unit respectively establishes BIM component nodes, construction task nodes, construction team nodes, construction machinery nodes, material batch nodes, construction area nodes, and quality event nodes; The relationship edge generation unit establishes process pre-relationship edges based on the logical relationships in the construction plan, establishes space occupancy relationship edges based on the spatial enclosure of BIM components and the occupancy time of the construction area, establishes resource sharing relationship edges based on the configuration of work teams, machinery and materials, establishes component ownership relationship edges based on the mapping between BIM components and construction tasks, establishes material supply relationship edges based on the supply records between material batches and construction tasks, and establishes quality rework relationship edges based on quality events and rework tasks. The dynamic graph update unit updates the corresponding graph nodes, relation edges, and their preset neighborhoods according to changes in the field status.

4. The BIM-based building project management system according to claim 3, characterized in that, The task readiness calculation unit generates the task start-up readiness based on the completion level of the preceding tasks, the material satisfaction level, the personnel satisfaction level, the machinery satisfaction level, the work surface satisfaction level, the quality acceptance level, and the task status reliability. The constraint-aware schedule deviation propagation prediction module includes a time window generation unit, a heterogeneous graph feature propagation unit, a construction constraint gating unit, a schedule deviation prediction unit, and a delay propagation path extraction unit. The heterogeneous graph feature propagation unit includes multiple relation type-aware graph attention layers. Each relation type-aware graph attention layer generates relation attention weights based on the dynamic state of the node corresponding to the relation type, edge constraint features, feature transformation parameters, and relation state credibility, and updates the state of the construction task node according to the relation attention weights.

5. The BIM-based building project management system according to claim 4, characterized in that, The edge constraint features of the process preceding relationship edge include the remaining time difference, the degree of completion of the preceding task and the completion and transfer conditions; the edge constraint features of the space occupation relationship edge include the space overlap ratio, the degree of overlap of occupation time and the safety interval requirements; the edge constraint features of the resource sharing relationship edge include the type of shared resource, the amount of resource demand and the degree of resource competition; and the edge constraint features of the quality rework relationship edge include the quality problem level, the expected rework period and the number of affected components. The construction constraint gating unit closes the delay propagation of the corresponding process prerequisite relationship edge when the deviation of the previous task's schedule does not exceed the remaining time difference, closes the conflict propagation of the corresponding space occupation relationship edge when the space occupation time of two construction tasks does not overlap, and closes the competition propagation of the corresponding resource sharing relationship edge when there is remaining capacity in the shared resources.

6. The BIM-based building project management system according to claim 5, characterized in that, The schedule deviation prediction unit generates a predicted schedule deviation based on the construction task node status, task start readiness, the portion of the schedule deviation of the preceding task exceeding the remaining time difference, and the uncertainty of the task status updated by the attention layer of the relationship type perception graph. The delay propagation path extraction unit traces the relationship attention weights in reverse from the construction task nodes where the predicted schedule deviation reaches the early warning condition, and generates delay paths for pre-process delays, spatial conflict delays, resource competition delays, material supply delays, and quality rework delays based on the relationship type and relationship attention weights.

7. The BIM-based building project management system according to claim 6, characterized in that, The multi-objective rolling construction scheme optimization module includes a candidate action generation unit, a BIM construction status simulation unit, a multi-objective scheme evaluation unit, a construction constraint verification unit, and a rolling scheme selection unit. The candidate action generation unit generates construction sequence adjustment actions, construction team reallocation actions, construction machinery time period adjustment actions, material supply priority adjustment actions, and construction area switching actions based on the delay propagation path, and combines mutually compatible adjustment actions into candidate construction schemes. The BIM construction status simulation unit generates the task status, resource status, and construction area status within the prediction interval based on the candidate construction scheme, the component-task-resource heterogeneity diagram, and the current BIM status. The construction constraint verification unit screens candidate construction schemes based on the pre-process relationship, construction team capacity, construction machinery capacity, available material quantity, construction area occupancy, safety passage status, and quality acceptance control points. The multi-objective scheme evaluation unit generates a comprehensive management cost of candidate construction schemes based on the predicted schedule deviation, construction cost increment, resource allocation fluctuation, space occupation conflict, scheme change range, and state uncertainty. The rolling scheme selection unit selects the candidate construction scheme with the minimum comprehensive management cost as the rolling construction scheme.

8. A method for operating a BIM-based building project management system according to any one of claims 1-7, characterized in that, Includes the following steps: Step S1: Parse the BIM model and construction plan data to generate BIM component codes, construction task codes, component spatial attributes, and component-task mapping relationships; Step S2: Collect multi-source data from the site, map the multi-source data to the corresponding BIM components, construction tasks and construction areas, and generate node dynamic status and status credibility. Step S3: Create different types of graph nodes based on BIM components, construction tasks, construction resources, construction areas, and quality events; create different types of relationship edges based on construction plan relationships, BIM spatial relationships, resource allocation relationships, and quality rework relationships; and generate a component-task-resource heterogeneous graph. Step S4: Generate task readiness based on the completion level of prerequisite tasks, material availability, personnel availability, machinery availability, work surface availability, quality acceptance level, and task status reliability. Step S5: Process the heterogeneous graph of components-tasks-resources through the relation type-aware graph attention network, and generate predicted schedule deviations and delay propagation paths based on task start-up readiness, relation attention weights, previous task schedule deviations, remaining time differences, and task status uncertainties. Step S6: Generate candidate construction schemes based on predicted schedule deviations and delay propagation paths; perform BIM construction status simulation and construction constraint verification on the candidate construction schemes; and select a rolling construction scheme based on project schedule, construction cost, resource fluctuations, spatial conflicts, scheme change range, and prediction uncertainty. Step S7: Send the rolling construction plan to the project management terminal and the on-site execution terminal, and collect the task execution status and quality acceptance results corresponding to the rolling construction plan; Step S8: Update the BIM component status, construction task status, resource status, and component-task-resource heterogeneity diagram based on the task execution status and quality acceptance results.

Images