A BIM-driven full-process digital delivery system for fabricated concrete structures

Abstract

The application discloses a BIM-driven full-process digital delivery system for fabricated concrete structures, which comprises a digital twin construction module, a BIM-GIS fusion module, an intelligent hoisting planning module, a 4D construction simulation module, an intelligent production scheduling module and a real-time monitoring and tracking module.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of building informatization, and particularly relates to a BIM-driven full-process digital delivery system for fabricated concrete structures. BACKGROUND

[0002] As an important form of building industrialization, fabricated concrete structures realize the standardization and industrialization of building construction through the mode of factory prefabrication and on-site assembly. However, the existing fabricated buildings have the following technical problems in digital management:

[0003] (1) The information islands of design, production, transportation and construction are serious, the BIM model is disconnected with the actual construction, and it is difficult to realize the information penetration of the whole process;

[0004] (2) The construction site lacks effective digital management means, the hoisting path planning relies on manual experience, and the safety risk of multi-tower crane collaborative operation is high;

[0005] (3) The whole process from production to installation of components lacks precise real-time monitoring, and quality tracing is difficult;

[0006] (4) The traditional construction management mode is difficult to realize the dynamic optimization of the construction process, and the resource scheduling efficiency is low.

[0007] Therefore, a system capable of realizing the full-process digital management of fabricated concrete structures is urgently needed to improve construction efficiency, ensure engineering quality and reduce safety risks. SUMMARY

[0008] The purpose of the application is to provide a BIM-driven full-process digital delivery system for fabricated concrete structures, to solve the technical problems of low digital management level, poor construction efficiency and difficult quality tracing of fabricated buildings in the prior art.

[0009] To achieve the above-mentioned purpose, the application provides the following technical scheme:

[0010] A BIM-driven full-process digital delivery system for fabricated concrete structures, the system comprising:

[0011] A digital twin construction module for establishing a real-time digital twin model of the construction site, collecting on-site sensor data through edge computing nodes, realizing dynamic mapping of the physical construction environment and the virtual model, the digital twin model comprising component state information, equipment operation parameters, environmental condition data, and realizing virtual-real synchronous updating through a time series data fusion algorithm;

[0012] A BIM-GIS fusion module is configured to realize deep fusion of a building information model and a geographic information system, to establish a unified spatial reference system through a coordinate conversion algorithm, to associate and map fine component-level information of the BIM model and site environment information of the GIS, and to support multi-scale information display and query from a macro site to a micro component.

[0013] An intelligent hoisting planning module is configured to generate an optimized tower crane operation path, based on a path search algorithm and by introducing three types of constraint conditions, i.e., wind load calculation, bearing capacity checking, and dynamic obstacle detection. The path search algorithm increases a dynamic cost function and a multi-constraint pruning strategy based on a traditional A* algorithm. The dynamic cost function dynamically adjusts a path cost value according to a real-time wind speed and a component gravity center position. The multi-constraint pruning strategy excludes path nodes that violate safety constraints in advance in a search process. A multi-tower crane anti-collision algorithm is used to realize collaborative operation control of multiple tower cranes, and a hoisting operation instruction manual including a hoisting point, a travel path, and a positioning coordinate is automatically generated.

[0014] A 4D construction simulation module is configured to realize construction process simulation based on a time dimension, to associate a construction period plan with a 3D BIM model, to evaluate a construction period risk through a Monte Carlo simulation method, to automatically identify a critical path and detect resource conflicts, and to generate construction progress deviation early warning information.

[0015] An intelligent production scheduling module is configured to optimize production and transportation plans of prefabricated components, to obtain production capacity, transportation resources, and construction demand information through multi-source data fusion, to generate a production scheduling scheme that meets just-in-time production requirements based on a constraint satisfaction problem solving algorithm, and to dynamically adjust the production scheduling scheme according to actual execution.

[0016] A real-time monitoring and tracking module is configured to realize position tracking and state monitoring of components in a whole life cycle, to obtain component identity information through an RFID tag, to track a transportation process through a GPS locator, to monitor hoisting safety states through an inclination sensor, to detect installation and positioning quality through a vibration sensor, and to form a complete quality traceability chain.

[0017] A mixed reality acceptance module is configured to provide an engineering acceptance interface based on mixed reality, to accurately superimpose a BIM model on a physical component through a mixed reality device, to perform three-dimensional reconstruction through a point cloud scanning technology, to automatically identify size deviations and surface defects, and to generate a digital acceptance report containing deviation annotations.

[0018] Further, the digital twin construction module adopts an edge-cloud collaborative architecture, a light inference model is deployed on an edge to realize real-time data processing, complex simulation algorithms are run on a cloud to optimize decision-making, and asynchronous communication between the edge and the cloud is realized through a message queue middleware.

[0019] Further, the BIM-GIS fusion module establishes a multi-level spatial index structure, including R-tree index for spatial range query, quadtree index for geographic feature management, and octree index for three-dimensional component retrieval, to realize efficient spatial query across scales.

[0020] Further, the intelligent hoisting planning module introduces a dynamic obstacle avoidance mechanism, real-time perceives dynamic obstacles in the construction site, and adjusts the hoisting path online through a fast re-planning algorithm to ensure operation safety.

[0021] Further, the 4D construction simulation module integrates a resource balancing optimization algorithm, automatically generates a resource allocation scheme after identifying resource conflicts, solves a multi-objective optimization problem through a genetic algorithm, and balances three optimization objectives of duration, cost, and resource utilization.

[0022] Further, the intelligent production scheduling module establishes a component priority scoring mechanism, calculates priority scores according to three dimensions of construction critical path, storage cost, and transportation window, and preferentially arranges production and distribution of key components.

[0023] Further, the system further comprises a data standardization interface, supports model import and export of IFC standard, provides RESTful API for third-party system integration, and realizes data exchange with ERP and MES enterprise information systems.

[0024] Further, the workflow of the system comprises the following steps:

[0025] S1, digital twin environment initialization and model construction: deploy edge computing nodes, configure temperature and humidity sensors, anemometers, inclination sensors, vibration sensors, and video cameras, import an initial three-dimensional model of the construction site from the BIM-GIS fusion module, determine coordinate conversion parameters through calibration control points, and establish a mapping relationship between the physical space and the digital space;

[0026] S2, BIM-GIS data fusion and spatial index construction: receive BIM models and GIS data, perform coordinate system unification, data format conversion, and attribute information integration, realize coordinate system conversion through a seven-parameter conversion model, construct R-tree index and octree index, and manage two-dimensional geographic features and three-dimensional building components, respectively;

[0027] S3, intelligent production scheduling and resource allocation: based on project progress requirements and construction drawings, comprehensively considering factory production capacity, mold resources, maintenance site, and transportation vehicle constraints, calculating priority based on component position in the critical path, storage cost, and installation time window, generating a production scheduling scheme including production batch, start time, completion time, and transportation arrangement;

[0028] S4, Construction Process 4D Simulation and Risk Assessment: Receive the production plan and BIM-GIS fusion data, simulate the construction process through discrete event simulation, use Monte Carlo method for multiple iterations, assess the schedule risk and resource conflict, identify the critical path, and generate a risk assessment report feedback to the intelligent production scheduling module;

[0029] S5, Hoisting Path Intelligent Planning and Anti-collision Control: Obtain wind speed, wind direction, and equipment state data from the digital twin construction module, query spatial constraints from the BIM-GIS fusion module, use an improved path search algorithm to calculate the optimal trajectory by integrating wind load, obstacles, and load rate, and construct a space-time occupancy graph in a multi-crane scenario to ensure job safety;

[0030] S6, Real-time Monitoring and Tracking and State Synchronization Update: Identify component identity through RFID, track transportation trajectory through GPS, monitor hoisting posture through inclination sensor, detect installation quality through vibration sensor, update digital twin model after filtering data, and send warning to related modules when deviation is detected;

[0031] S7, Dynamic Scheduling Optimization and Plan Adjustment: Receive real-time monitoring deviation information, hoisting adjustment request and simulation optimization suggestion, use rolling time domain strategy to re-optimize component production order, transportation batch and hoisting time sequence in future fixed time window, and notify the execution unit through message push;

[0032] S8, Mixed Reality Quality Inspection and Deviation Analysis: Superimpose BIM model on physical components through MR equipment, identify ArUco markers for coarse registration, use point cloud registration algorithm for accurate alignment, calculate deviation distribution, and generate digital inspection report containing deviation statistics, quality evaluation and rectification suggestions;

[0033] S9, Data Archiving and Knowledge Extraction: Structured storage of sensor monitoring data, execution deviation records, optimization adjustment history and quality inspection results, extraction of construction experience and optimization mode through data mining, and updating of system knowledge base;

[0034] S10, Full-process Digital Delivery and File Generation: Integrate design models, construction records, quality archives and inspection reports, use IFC standard format to form structured digital archives containing component lifecycle information, support post-operation management and traceability query.

[0035] Further, in step S3, a priority-based heuristic algorithm is used to solve the constraint satisfaction problem, and the priority is calculated based on three dimensions of critical path, storage cost and time window.

[0036] Further, in the step S5, the dynamic penalty term of the path search algorithm includes a wind load penalty, an obstacle distance penalty and a load rate penalty, which are comprehensively calculated by weighting coefficients.

[0037] Further, the steps S6 and S7 form a closed-loop feedback mechanism, and when the actual execution deviates from the plan by more than a threshold value, a dynamic adjustment is automatically triggered, and the adjusted scheme returns to step S5 for re-execution. Advantages

[0038] 1. Information island elimination: through digital twin technology and BIM-GIS fusion, the information of design, production, transportation and construction is connected, and the problem of disconnection between BIM model and actual construction is solved.

[0039] 2. Intelligent construction management: through intelligent hoisting planning and 4D simulation technology, automatic optimization of hoisting path and collaborative control of multi-tower crane are realized, which changes the traditional mode of relying on manual experience and reduces the safety risk.

[0040] 3. Full-process precise tracing: through multi-sensor collaborative monitoring, a complete tracing chain from production to installation is established, solving the problems of difficult component whole-process monitoring and quality tracing.

[0041] 4. Dynamic optimization scheduling: through real-time feedback and rolling optimization mechanism, dynamic adjustment of the construction process is realized, the resource scheduling efficiency is improved, and the static limitation of the traditional management mode is overcome. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 The figure shows the architecture schematic diagram of the system of the application;

[0043] Figure 2 The figure shows the working step flow chart of the system of the application. DETAILED DESCRIPTION

[0044] The exemplary embodiments of the application will be described in detail below with reference to the accompanying drawings.

[0045] In combination Figure 1 , the application provides a BIM-driven prefabricated concrete structure full-process digital delivery system, which comprises a digital twin construction module, a BIM-GIS fusion module, an intelligent hoisting planning module, a 4D construction simulation module, an intelligent production scheduling module, a real-time monitoring and tracking module and a hybrid physical experiment module. The modules are deeply cooperated through standardized data interface and message passing mechanism, forming an organic whole system.

[0046] The digital twin construction module serves as the data perception and state mapping foundation of the entire system, responsible for establishing a real-time virtual mirror of the construction site. This module adopts a layered architecture with edge-cloud collaboration, deploying multiple edge computing nodes at the construction site, each responsible for data collection and preprocessing in a specific area. The edge nodes are equipped with ARM architecture embedded processors running lightweight Linux operating systems, connected to temperature and humidity sensors, anemometers, tilt sensors, vibration sensors, and video cameras through GPIO interfaces. Temperature and humidity sensors sample once per minute, wind speed sensors sample 10 times per second, and video streams are transmitted at a rate of 25 frames per second. The sampling frequencies of different sensors are determined based on monitoring requirements and data characteristics.

[0047] The edge nodes perform preliminary processing on the collected raw data, including data cleaning, format conversion, outlier detection, and data compression. Data cleaning uses a sliding window method, calculating the deviation of each data point from other data points within the window. When the deviation exceeds three standard deviations, it is marked as an outlier. For video data, the edge nodes run a lightweight target detection model based on MobileNet, identifying personnel, equipment, and components in the construction site, and transmitting the detection results in structured data format, significantly reducing network bandwidth usage.

[0048] The processed data is sent to the cloud server through the MQTT protocol, with a JSON-encoded message format containing four necessary fields: timestamp, device identification, data type, and data value. The cloud server deploys a message queue service and uses a publish-subscribe model to manage data flow. The digital twin construction module subscribes to all sensor topics, receives data, stores it in a time-series database, and triggers the state update process.

[0049] In constructing the digital twin model, the invention uses a state estimation method based on extended Kalman filtering to fuse discrete observation data into continuous state descriptions. The system state vector contains the position, velocity, and attitude information of all monitored objects, with dimensions dynamically adjusted according to the size of the construction site. The state transition equation is expressed as:

[0050]

[0051] where: is the system state vector at time t; is the system state vector at time t-1; is the control input vector at time t-1, including tower crane operation instructions and transport vehicle scheduling instructions; is a nonlinear state transition function; is the process noise vector at time t-1, with its covariance matrix derived from historical data statistics.

[0052] The observation equation relates the actual sensor measurements to the system state:

[0053]

[0054] where: is the observation vector at time t, containing the measurements of all sensors; is the nonlinear observation function; is the observation noise vector at time t, whose covariance matrix is determined through sensor calibration experiments.

[0055] The prediction step of the extended Kalman filter computes the state prediction and the error covariance:

[0056]

[0057]

[0058] where: is the prediction of the state at time t based on the information at time t-1; is the optimal state estimate at time t-1; is the prediction error covariance matrix at time t; is the estimation error covariance matrix at time t-1; is the Jacobian matrix of the state transition function at time t, obtained by taking the partial derivative of with respect to the state variable; is the transpose of ; and is the process noise covariance matrix.

[0059] The update step corrects the prediction with the new observation value, first computing the Kalman gain:

[0060]

[0061] where: is the Kalman gain matrix at time t; is the Jacobian matrix of the observation function at time t, obtained by taking the partial derivative of with respect to the state variable; is the transpose of ; and is the observation noise covariance matrix; denotes the inverse of the matrix .

[0062] The Kalman gain is used to update the state estimate and the error covariance:

[0063]

[0064]

[0065] wherein: is the optimal state estimation after observation at fusion time t; is the observation residual, representing the difference between actual observation and predicted observation; is the estimation error covariance matrix at time t; is an identity matrix with the same dimension as the state vector.

[0066] The digital twin construction module synchronizes the updated state information to other modules in real time. The intelligent hoisting planning module obtains the real-time positions of the tower crane and the components by subscribing to state update events; the 4D construction simulation module verifies the accuracy of the simulation model based on the state information; and the real-time monitoring and tracking module generates a visual monitoring interface based on the state data. This event-driven information sharing mechanism ensures that each module always makes decisions based on the latest on-site state.

[0067] The BIM-GIS fusion module is responsible for integrating the data of the building information model and the geographic information system, and providing unified spatial information services for other modules. This module needs to handle the coordinate differences and data model differences between the two systems. The BIM system uses the right-hand Cartesian coordinate system, with the origin at the reference point of the building; the GIS system uses the geographic coordinate system, using latitude, longitude and elevation to represent the position.

[0068] The coordinate conversion process is divided into two stages. The first stage converts the BIM local coordinates to the engineering coordinate system. This conversion is achieved by selecting at least four control points in the BIM model and the construction site, and using the least squares method to solve the conversion parameters. Let n be the number of control points, the coordinates of the i-th control point in the BIM coordinate system be , and the coordinates in the engineering coordinate system be , then the conversion relationship is represented as:

[0069]

[0070] wherein: are the x, y, z coordinate values of the i-th control point in the engineering coordinate system; are the x, y, z coordinate values of the i-th control point in the BIM coordinate system; is the scale factor; is the rotation matrix determined by Euler angles, is the rotation angle around the x-axis, is the rotation angle around the y-axis, is the rotation angle around the z-axis; are the translation parameters in the x, y, z directions, respectively.

[0071] Solve the parameters by minimizing the transformation error of all control points:

[0072]

[0073] where: denotes the Euclidean norm of a vector; is the total number of control points; the summation symbol indicates the accumulation of errors for all control points.

[0074] After completing the coordinate transformation, the BIM-GIS fusion module constructs a multi-level spatial index structure to support efficient querying. For two-dimensional geographic features, R-tree indexing is used, and for three-dimensional BIM components, octree indexing is used. The construction process of R-tree uses the STR algorithm, which first sorts the spatial objects by x-coordinate, divides them into groups, and then sorts them by y-coordinate within each group to form leaf nodes, where is the total number of spatial objects. The construction of octree starts from the smallest bounding box containing all components, recursively divides the space into eight subspaces until the number of components contained in each leaf node is less than the preset threshold.

[0075] The BIM-GIS fusion module provides a unified spatial query interface to other modules. When the intelligent hoisting planning module needs to obtain obstacle information around a certain position, it calls the range query interface and inputs the query center point coordinates and the query radius . The BIM-GIS fusion module searches in R-tree and octree in parallel, merges the query results of the two indexes, and returns all spatial objects that meet the conditions. This unified spatial information service avoids repeated storage and processing of spatial data by each module, improving the overall efficiency of the system.

[0076] The intelligent hoisting planning module is responsible for generating an optimized tower crane operation path. This module closely cooperates with the digital twin construction module and the BIM-GIS fusion module to obtain real-time construction site state and spatial constraint information. Before path planning begins, the intelligent hoisting planning module obtains the current wind speed, wind direction, and tower crane attitude parameters from the digital twin construction module, and queries the location of the components to be hoisted, the distribution of surrounding obstacles, and the operation range of other tower cranes from the BIM-GIS fusion module.

[0077] The path planning uses an improved A* algorithm, which introduces a dynamic cost function and a multi-constraint pruning strategy based on traditional heuristic search. The search space is discretized into a three-dimensional grid, and each grid node represents a possible position of the tower crane hook. The evaluation function of a node is defined as:

[0078]

[0079] where: The total cost of the node ; The actual path cost from the starting node to the node , calculated by accumulating the lengths of each segment on the path; The heuristic estimated cost from the node to the target node, using the Euclidean distance as the estimated value; The penalty term for handling dynamic constraints.

[0080] The penalty term considers three factors: wind load, obstacle distance, and tower load rate:

[0081]

[0082] Where: is the wind load penalty term for the node ; is the obstacle distance penalty term for the node ; is the load rate penalty term for the node ; is the wind load weight coefficient; is the obstacle weight coefficient; is the load rate weight coefficient; the three weight coefficients satisfy .

[0083] The wind load penalty term is calculated according to the real-time wind speed and component characteristics:

[0084]

[0085] Where: is the current wind speed, provided in real time by the digital twin construction module; is the safe operation wind speed threshold, generally taken as 6 m / s; is the stop operation wind speed threshold, generally taken as 12 m / s; is the wind load influence coefficient, determined according to the component shape; is the effective windward area of the component, calculated by the component cross-sectional area and wind direction angle.

[0086] The obstacle distance penalty term ensures that the path maintains a safe distance from obstacles:

[0087]

[0088] Where: is the number of obstacles within the influence range of the node ; is the shortest distance from the node to the th obstacle. is the distance decay parameter, controlling the decay speed of the penalty value, typical value is 2 meters; the summation sign means the penalty values of all obstacles are accumulated.

[0089] The load rate penalty term considers the influence of the current load of the tower crane on path selection:

[0090]

[0091] wherein: is the load influence coefficient; is the weight of the current hoisting component; is the maximum lifting weight of the tower crane at the node position; is the horizontal distance from the node to the center of the tower crane.

[0092] In the multi-tower crane collaborative operation scene, the intelligent hoisting planning module needs to perform anti-collision detection and coordination. Suppose there are tower cranes in the construction site, the hook position of the th tower crane at time is , and the speed is . Collision detection between two tower cranes is achieved by calculating the minimum distance:

[0093]

[0094] wherein: is the Euclidean distance between the hooks of the th and th tower cranes at time ; is the three-dimensional coordinate of the hook of the th tower crane at time ; is the three-dimensional coordinate of the hook of the th tower crane at time .

[0095] The safety constraint condition is:

[0096]

[0097] wherein: is the minimum safety distance, generally taken as 3 meters; is the enclosing sphere radius of the hook of the th tower crane and the hoisting component; is the enclosing sphere radius of the hook of the th tower crane and the hoisting component.

[0098] When a collision is predicted to occur at a future time, the intelligent lifting planning module adopts a spatio-temporal coordination strategy to avoid it. This strategy builds a four-dimensional spatio-temporal occupancy graph where each voxel represents the occupancy state at time is represented as:

[0099]

[0100] By searching for a collision-free path in the spatio-temporal occupancy graph, each tower crane is assigned a non-interfering operation period and space. When the operation plan needs to be adjusted, the intelligent lifting planning module sends an adjustment request to the intelligent production scheduling module, including the component number that needs to be delayed, the suggested new time window, and the adjustment reason. After receiving the request, the intelligent production scheduling module re-evaluates the priority of the related components, generates a new production scheduling scheme, and feeds it back to the intelligent lifting planning module.

[0101] The 4D construction simulation module is responsible for introducing the time dimension into the three-dimensional BIM model, realizing the dynamic simulation and optimization of the construction process. This module obtains the initial construction plan from the intelligent production scheduling module, imports the construction site model from the BIM-GIS fusion module, and obtains the lifting scheme from the intelligent lifting planning module, and integrates these information to build a four-dimensional construction model.

[0102] The construction process is modeled using the discrete event simulation method, and each construction activity is represented as an event object, containing the following attributes: activity identifier, activity type, start time, duration, required resources, pre-activity set, and post-activity set. The logical relationship between activities is represented by a directed acyclic graph (DAG), where nodes represent activities and edges represent dependency relationships.

[0103] The uncertainty of activity duration is modeled using a three-point estimation method:

[0104]

[0105]

[0106] where: is the expected duration of the activity; is the optimistic time estimate, representing the completion time under the most ideal circumstances; is the most likely time estimate, based on the mode of historical data; is the pessimistic time estimate, considering the longest time under various unfavorable factors; is the standard deviation of the duration, used to quantify the uncertainty.

[0107] Based on the time model above, the 4D construction simulation module evaluates the project duration risk using Monte Carlo method. In each simulation iteration, the duration of each activity is randomly sampled from the beta distribution:

[0108]

[0109] where the shape parameters of the beta distribution are calculated by three-point estimation:

[0110]

[0111]

[0112] where: is the first shape parameter of the beta distribution; is the second shape parameter of the beta distribution.

[0113] The total project duration is calculated using the critical path method (CPM) in each simulation iteration. For each activity in the activity network, its earliest start time , earliest finish time , latest start time , latest finish time and total float are calculated as follows:

[0114]

[0115]

[0116]

[0117]

[0118] where: is the set of predecessor activities of activity ; is the set of successor activities of activity ; is the duration of activity ; denotes the max operation; denotes the min operation.

[0119] Activities with zero total float constitute the critical path, and their delay will directly affect the total project duration. After Monte Carlo simulations, the probability distribution and statistical characteristics of the project duration are obtained:

[0120]

[0121]

[0122]

[0123] in: For the first The total project duration for this simulation; This represents the expected total construction period. The variance of the total project duration; The total construction period shall not exceed The probability of; This is an indicator function that takes the value 1 when the condition is met, and 0 otherwise. The number of simulations is typically set to 5000 to ensure the statistical reliability of the results.

[0124] The 4D construction simulation module also integrates resource balancing optimization, automatically adjusting when resource conflicts are detected. Resource conflict detection is achieved by comparing resource requirements with available resources at each time point.

[0125]

[0126] in: for Total resource requirements at any given time; for The set of activities currently in progress; For the event The amount of resources required.

[0127] when Resource conflicts occur at times, among which This represents the total available resources. Resource balancing is addressed by adjusting the start times of non-critical activities, with the optimization objective being to minimize fluctuations in resource usage.

[0128]

[0129] in: For the mean squared error of resource utilization; This represents the average resource usage. This refers to the total project duration.

[0130] The optimization process employs a genetic algorithm, where chromosomes encode the start time offsets of non-critical activities, and the fitness function is the reciprocal of the resource mean square error. Iterative optimization is performed through selection, crossover, and mutation operations until a satisfactory resource allocation scheme is found. The optimized scheme is then fed back to the intelligent scheduling module to update the construction plan.

[0131] The intelligent scheduling module is responsible for coordinating the whole process of precast component production, transportation and installation. It forms a closed-loop feedback mechanism with other modules. It receives construction progress requirements from the 4D construction simulation module, actual execution conditions from the real-time monitoring and tracking module, and job adjustment requests from the intelligent hoisting planning module, and dynamically optimizes the scheduling plan based on these information.

[0132] The scheduling problem is modeled as a constraint satisfaction problem (CSP), defined as a triple where is the set of decision variables, each variable represents the start time of component production; is the set of variable domains, denotes the feasible value range of variable ; is the set of constraints, including production capacity constraints, transportation capacity constraints, timing constraints and storage space constraints.

[0133] The production capacity constraint ensures that the production load at any time does not exceed the factory capacity:

[0134]

[0135]

[0136] where: is the set of components being produced at time ; is the number of molds needed for component ; is the total number of available molds; is the set of components being maintained at time ; is the floor area of component ; is the total area of the maintenance site.

[0137] The transportation capacity constraint limits the number and weight of components being transported simultaneously:

[0138]

[0139] where: is the set of components being transported at time ; is the weight of component ; is the load capacity of a single transport vehicle; is the number of available transport vehicles; denotes the ceiling operation.

[0140] The timing constraint ensures the sequence of each stage:

[0141]

[0142]

[0143] wherein: is the production start time of component ; is the transportation start time; is the installation start time; is the production duration; is the maintenance duration; is the transportation duration.

[0144] Storage space constraints limit the component stacking at the construction site:

[0145]

[0146] wherein: is the component set stored at the construction site at time ; is the volume of component ; is the storage space capacity at the site.

[0147] The intelligent scheduling module solves the CSP problem using a priority-based heuristic algorithm. The component priority takes into account three factors: critical path, storage cost, and time window:

[0148]

[0149] wherein: is the comprehensive priority of component ; is the critical path priority, taking values 0 or 1; is the normalized storage cost priority, is the unit storage cost, is the estimated storage duration, and are the corresponding maximum values; is the time window urgency, and are the earliest and latest installation times, respectively; is the weight coefficient, satisfying .

[0150] When receiving the execution deviation feedback from the real-time monitoring and tracking module, the intelligent scheduling module triggers the rescheduling mechanism. Let the actual completion time of component be , the planned completion time be , and the deviation be .when hour( (A deviation threshold is used), and subsequent components affected by the rescheduling are rescheduled. The rescheduling adopts a rolling time-domain strategy, adjusting only the plans within a fixed future time window to maintain system stability.

[0151] The real-time monitoring and tracking module achieves precise tracking of components throughout their entire lifecycle through various sensors and identification technologies. Each prefabricated component is embedded with an RFID tag during production; the tag stores the component's unique identifier, specifications, production information, and quality inspection records. RFID readers are deployed at key locations at factory entrances and exits, transport vehicles, and construction sites to automatically identify and record the component's movement information.

[0152] GPS locators are installed on transport vehicles and report location information at a fixed frequency. Raw GPS data contains noise and drift, requiring filtering. This invention employs a Kalman filter algorithm to smooth the trajectory; the vehicle motion state vector is defined as:

[0153]

[0154] in: For the first Vehicle position coordinates at each sampling time; For the corresponding velocity components; superscript This represents the transpose of a vector.

[0155] The state transition equation is:

[0156] Where the state transition matrix is:

[0157]

[0158] in: The sampling time interval; This is the process noise vector.

[0159] The observation equation is:

[0160] The observation matrix is:

[0161] in: These are GPS observations; This is the observed noise vector.

[0162] The filtered trajectory data is matched with the electronic fence to determine whether the vehicle is traveling along the planned route. When a vehicle deviates from the predetermined route or the estimated arrival time changes significantly, the real-time monitoring and tracking module sends an early warning to the intelligent scheduling module, triggering corresponding adjustment measures.

[0163] During lifting, the inclination sensor monitors the attitude of the component in real time to prevent safety accidents caused by excessive inclination. The inclination sensor outputs the pitch angle and roll angle , and the safety determination condition is:

[0164]

[0165] where: is the maximum allowed inclination, which is determined according to the type of component and lifting height, generally taking 5-10 degrees.

[0166] When the inclination exceeds the safety threshold, the system immediately issues an audible and visual alarm, and sends an emergency stop command to the intelligent lifting planning module. Inclination data is also used to evaluate the smoothness of lifting operations, and dangerous operations such as sudden acceleration and sudden deceleration are identified by calculating the rate of change of inclination:

[0167]

[0168] where: is the angular velocity; and are the rates of change of pitch angle and roll angle, respectively.

[0169] The vibration sensor is used to detect the quality of the component installation. When the component comes into contact with the installation position, the impact vibration signal generated contains a wealth of information. The vibration signal is subjected to fast Fourier transform (FFT) to obtain the frequency spectrum:

[0170]

[0171] where: is the frequency domain signal; is the frequency; is the imaginary unit; is the base of the natural constant.

[0172] The installation quality is judged by analyzing the spectral characteristics. During normal installation, the main frequency is concentrated in the low frequency band (10-50Hz), and the energy distribution is uniform; during abnormal installation (such as not fully in place, poor contact), there will be an increase in high frequency components, and the energy distribution will be uneven. The system extracts the following characteristic parameters:

[0173]

[0174]

[0175]

[0176] where: is the peak frequency; Total energy; High-low frequency energy ratio; Intermediate frequency, take 100Hz; Maximum analysis frequency, take 500Hz.

[0177] Based on these characteristic parameters, support vector machine (SVM) classifier is used to determine the installation quality, and the classification decision function is:

[0178]

[0179] Wherein: The feature vector to be classified; Support vector; Class label (+1 indicates qualified, -1 indicates unqualified); Lagrange multiplier; Kernel function, radial basis function is used; Bias term; The number of support vectors; Sign function.

[0180] The mixed experiment module uses MR devices such as HoloLens to superimpose BIM models on physical components for quality inspection. The key technology of this module is to realize accurate registration of virtual models and real scenes. The registration process is divided into two stages of coarse registration and fine registration.

[0181] Coarse registration is achieved by recognizing ArUco markers arranged on site. ArUco markers are two-dimensional barcodes that contain unique ID and direction information. By capturing the marker image with the camera, the four corner points of the marker are extracted using computer vision algorithms, and then the camera pose is solved by PnP (Perspective-n-Point) algorithm:

[0182]

[0183] Wherein: Image coordinates; Depth in the camera coordinate system; Camera intrinsic matrix, containing focal length and principal point coordinates; Extrinsic matrix, 3x3 rotation matrix, 3x1 translation vector; Three-dimensional point coordinates in the world coordinate system.

[0184] The camera intrinsic matrix is obtained by pre-calibration:

[0185]

[0186] Wherein: is the focal length for x and y directions; is the principal point coordinate.

[0187] Fine registration improves accuracy using point cloud registration techniques. The depth camera of the mixed reality device acquires a scene point cloud , and the point cloud generated from the BIM model is matched. The Iterative Closest Point (ICP) algorithm is used to solve the optimal transformation:

[0188]

[0189] where: is the optimal transformation matrix; is the number of corresponding point pairs; is the point in the point cloud with the nearest neighbor index; is the rigid body transformation including rotation and translation.

[0190] The ICP algorithm iteratively performs the following steps until convergence:

[0191] For each point in the point cloud , find the nearest point in the point cloud ;

[0192] Calculate the transformation matrix that minimizes the distance of the corresponding point pairs;

[0193] Apply the transformation to update the point cloud: ;

[0194] Calculate the error: ;

[0195] If , stop the iteration, otherwise go back to step 1.

[0196] where: is the convergence threshold, which is set to 0.001 meters.

[0197] After registration is complete, the system calculates the deviation of the physical components from the BIM model. For each sampling point, calculate the shortest distance from it to the model surface:

[0198]

[0199] where: is the deviation of the th sampling point; is the BIM model surface; is the Euclidean distance.

[0200] The deviation statistical indicators include:

[0201]

[0202]

[0203]

[0204] wherein: is the average deviation; is the standard deviation of the deviation; is the maximum deviation; is the total number of sampling points.

[0205] The system generates a heat map according to the deviation distribution, which intuitively displays the size deviation of each part of the component. The area where the deviation exceeds the allowed value is highlighted in red, and the normal area is displayed in green. The acceptance personnel can view detailed information through gesture interaction, including specific deviation values, cause analysis and treatment suggestions for over-standard deviation. All acceptance data are automatically recorded and a digital acceptance report is generated, including a three-dimensional deviation cloud map, statistical analysis results, acceptance conclusions and electronic signatures.

[0206] Through the cooperative work of the above-mentioned modules, the present application realizes the whole-process digital management of the fabricated concrete structure from design to construction, from production to installation. The data sharing and linkage optimization between the modules ensure the overall efficiency of the system, and provide a complete technical solution for the intelligent construction of fabricated buildings.

[0207] In combination with Figure 2 , the workflow of the system of the present application is as follows:

[0208] Step S1: initialization of digital twin environment and model construction

[0209] When the system is started, the digital twin construction module performs an initialization process. This step includes deploying edge computing nodes, configuring sensor networks, and establishing data acquisition channels. The edge nodes are communicatively connected with temperature and humidity sensors, anemometers, inclination sensors, vibration sensors, and video cameras, and the integrity of the data transmission link is verified. At the same time, the initial three-dimensional model of the construction site is imported from the BIM-GIS fusion module, and the mapping relationship between the physical space and the digital space is established. By calibrating the control points, the coordinate conversion parameters are determined, and the basic construction of the digital twin environment is completed.

[0210] Step S2: BIM-GIS data fusion and spatial index construction

[0211] The BIM-GIS fusion module receives the BIM model from the design unit and the GIS data from the surveying department. This step performs three operations: coordinate system unification, data format conversion, and attribute information integration. The precise conversion between different coordinate systems is achieved through a seven-parameter conversion model, ensuring the consistency of spatial data. Subsequently, R-tree index and octree index are constructed to manage two-dimensional geographic features and three-dimensional building components, respectively. The fused spatial data model provides a unified spatial information service interface for other modules.

[0212] Step S3: Intelligent production scheduling and resource allocation

[0213] The intelligent production scheduling module formulates the production plan of prefabricated components based on the overall project schedule requirements and construction drawings. This step considers four types of constraints: factory production capacity, mold resources, maintenance site, and transportation vehicles, and generates an initial production scheduling scheme using a priority-based heuristic algorithm. The component priority is determined comprehensively based on its position on the critical path, storage cost, and installation time window. The production scheduling plan includes the production batch, start time, completion time, and transportation arrangement of each component.

[0214] Step S4: 4D construction simulation and risk assessment

[0215] The 4D construction simulation module receives the production scheduling plan and BIM-GIS fusion data, and constructs a four-dimensional construction model. This step simulates the entire construction process through discrete event simulation, associating the time dimension with the three-dimensional spatial model. The Monte Carlo method is used for multiple simulation iterations to assess the schedule risk and resource conflicts. The simulation process identifies the critical path, detects resource bottlenecks, and analyzes potential delay factors. Based on the simulation results, a risk assessment report is generated and fed back to the intelligent production scheduling module for plan adjustment.

[0216] Step S5: Intelligent hoisting path planning and collision avoidance control

[0217] The intelligent hoisting planning module generates an optimized hoisting scheme for each component to be installed. This step obtains wind speed, wind direction, and equipment state data from the digital twin construction module, and queries spatial constraint information from the BIM-GIS fusion module. An improved path search algorithm is used to consider wind load, obstacles, and load rate, and to calculate the optimal hoisting trajectory. In the multi-crane collaborative operation scenario, a space-time occupancy graph is constructed to ensure the safety of each crane operation through a collision avoidance algorithm. The generated hoisting scheme includes the lifting point coordinates, path node sequence, speed control parameters, and operation timing arrangement.

[0218] Step S6: Real-time monitoring and tracking and state synchronization update

[0219] Real-time monitoring and tracking module continuously collects on-site data during construction execution. This step identifies component identity through RFID reader, tracks transportation trajectory with GPS locator, monitors lifting posture with inclination sensor, and detects installation quality with vibration sensor. After filtering and feature extraction, the collected data is updated to the digital twin model. When execution deviation or abnormal situation is detected, warning information is immediately sent to the relevant module. The digital twin construction module updates the virtual model through state estimation algorithm based on the latest monitoring data.

[0220] Step S7: Dynamic scheduling optimization and plan adjustment

[0221] When actual execution deviates from the plan, the system starts the dynamic adjustment mechanism. This step is dominated by the intelligent production scheduling module, receives deviation information from the real-time monitoring and tracking module, adjustment request from the intelligent lifting planning module, and optimization suggestions from the 4D construction simulation module. A rolling time domain strategy is adopted to re-optimize the plan within a fixed time window in the future. The adjustment content includes component production sequence, transportation batch arrangement, and lifting operation timing. The optimized plan is notified to each execution unit through message push.

[0222] Step S8: Mixed reality quality acceptance and deviation analysis

[0223] The mixed reality acceptance module performs quality inspection after the completion of component installation. This step superimposes the BIM model on the physical component through the MR device, uses computer vision technology to identify ArUco markers for coarse registration, and uses point cloud registration algorithm for accurate alignment. The system automatically calculates the deviation between the physical component and the design model, and generates a three-dimensional deviation distribution map. Acceptance personnel can view detailed deviation information of each part through gesture interaction. Acceptance data is uploaded to the cloud in real time, generating a digital acceptance report containing deviation statistics, quality assessment, and rectification suggestions.

[0224] Step S9: Data archiving and knowledge extraction

[0225] The system stores sensor monitoring data, execution deviation records, optimization adjustment history, and quality acceptance results in a structured manner during project execution. This step extracts construction experience and optimization patterns through data mining technology, and identifies key factors affecting construction period and quality. The extracted knowledge is updated to the system knowledge base, used to optimize algorithm parameters, improve prediction models, and perfect decision rules.

[0226] Step S10: Full-process digital delivery and archive generation

[0227] After the project is completed, the system generates a complete digital delivery result. This step integrates the design model, construction records, quality archives, and acceptance reports to form a structured digital archive. The delivery content adopts the IFC standard format to ensure compatibility with other systems. The digital archive contains the full life cycle information of the components, supporting later operation and maintenance management and traceability inquiries.

[0228] The above workflow realizes collaboration between modules through an event-driven mechanism, and asynchronous communication based on a message queue ensures stable system operation. Data sharing and linkage optimization between modules improve construction efficiency, ensure engineering quality, and reduce safety risks.

[0229] In summary, the present application realizes virtual mapping of the construction site through digital twinning, breaks down data barriers through BIM-GIS integration, optimizes hoisting and production scheduling through intelligent algorithms, realizes full-process tracking through multiple sensors, and innovates quality acceptance through mixed reality. The system realizes intelligent decision-making through virtual and real fusion, efficient resource utilization, complete quality traceability, and standardized interface integration, providing a complete technical solution for the intelligentization of fabricated buildings.

[0230] The above description is only the preferred embodiment of the present application, and any modifications, equivalent replacements, and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A BIM-driven full-process digital delivery system for fabricated concrete structures, characterized in that, The method comprises the following steps: A digital twin construction module is used to establish a real-time digital twin model of the construction site through an edge-cloud collaborative architecture. An edge computing node collects data from temperature and humidity sensors, an anemometer, an inclination sensor, a vibration sensor, and a video camera. The discrete observation data is fused into a continuous state description through a state estimation method based on an extended Kalman filter, and dynamic mapping of the physical construction environment and the virtual model is realized. A BIM-GIS fusion module is used to realize conversion of a BIM local coordinate system to a geographic coordinate system through a seven-parameter conversion model, to construct an R-tree index to manage two-dimensional geographic elements, to construct an octree index to manage three-dimensional BIM components, and to provide a unified spatial query interface. An intelligent hoisting planning module is used to generate a hoisting path based on an improved path search algorithm. The improvement includes the introduction of dynamic penalty terms for comprehensive wind load, obstacle distance, and load rate. A four-dimensional space-time occupancy graph is constructed for anti-collision control during collaborative operation of multiple tower cranes. A 4D construction simulation module is used to evaluate schedule risk through discrete event simulation and the Monte Carlo method, and to optimize resource balance using a genetic algorithm to minimize the mean square error of resource usage. An intelligent production scheduling module is used to model the production scheduling problem as a constraint satisfaction problem, taking into account production capacity constraints, transportation capacity constraints, timing constraints, and storage space constraints. Production scheduling decisions are made based on the comprehensive priority of components in the critical path, storage cost, and time window. A real-time monitoring and tracking module is used to identify component identity through RFID tags, track transportation trajectories using GPS locators and Kalman filter smoothing, monitor hoisting posture using inclination sensors, and detect installation quality through spectral analysis of vibration sensors. A hybrid physical experiment and acceptance module is used to perform coarse registration by identifying ArUco markers, perform point cloud fine registration using the iterative closest point algorithm, calculate the deviation distribution of physical components and BIM models, and generate a digital acceptance report. The state update information of the digital twin construction module is synchronized to the intelligent hoisting planning module, the 4D construction simulation module, and the real-time monitoring and tracking module through an event-driven mechanism. The intelligent hoisting planning module obtains real-time environmental data from the digital twin construction module and queries spatial constraints from the BIM-GIS fusion module. The risk assessment results of the 4D construction simulation module are fed back to the intelligent production scheduling module for plan adjustment. The deviation information of the real-time monitoring and tracking module triggers dynamic rescheduling of the intelligent production scheduling module.

2. The system of claim 1, wherein, The extended Kalman filter of the digital twin construction module comprises: A prediction step that predicts the current state based on a nonlinear state transition function and control input, and calculates a prediction error covariance matrix through the Jacobian matrix of the state transition function. An update step that calculates a Kalman gain matrix through the Jacobian matrix of the observation function, corrects the predicted state using the observation residual, and updates the estimated error covariance matrix.

3. The system of claim 1, wherein, The dynamic penalty terms of the intelligent hoisting planning module include: A wind load penalty term, which is zero when the wind speed is below a safe operation threshold, is proportional to the square of the wind speed exceeding amount and the effective windward area of the component between the safe operation threshold and a stop operation threshold, and is set to infinity when the wind speed reaches the stop operation threshold; An obstacle distance penalty term, which is in the form of a Gaussian function, exponentially decays with the square of the distance between the node and the obstacle; A load rate penalty term, which is proportional to the square of the ratio of the component weight to the maximum lifting capacity of the tower crane and the horizontal distance between the hook and the center of the tower crane.

4. The system of claim 1, wherein, The 4D construction simulation module determines the probability distribution of the activity duration by using a three-point estimation method, constructs a beta distribution model based on the optimistic time, the most likely time and the pessimistic time, generates random values of the activity duration by Monte Carlo sampling, and identifies the critical activity sequence affecting the total project duration by using the critical path method.

5. The system of claim 1, wherein, The component priority of the intelligent production scheduling module is calculated by the weighted sum of the critical path priority, the normalized storage cost priority and the time window urgency, wherein the critical path priority is a binary variable, the storage cost priority is the normalized value of the product of the unit storage cost and the expected storage duration, and the time window urgency is the reciprocal of the difference between the latest installation time and the earliest installation time.

6. The system of claim 1, wherein, The vibration signal quality detection of the real-time monitoring and tracking module includes: Performing fast Fourier transform on the collected vibration acceleration signal to obtain a frequency domain signal; Extracting three characteristic parameters of the peak frequency, the total energy and the high-to-low frequency energy ratio, wherein the demarcation frequency between the high and low frequencies is set to 100 Hz; Using a support vector machine classifier with a radial basis kernel function to determine the qualified or unqualified state of the installation quality based on the characteristic parameters.

7. The system of claim 1, wherein, The hybrid experiment and model recovery module solves the camera extrinsic parameter matrix including the rotation matrix and the translation vector by using the perspective-n-point algorithm; solves the optimal rigid transformation matrix by minimizing the sum of the squared Euclidean distances between the point cloud and the corresponding points of the model by using the iterative closest point algorithm; and the deviation statistical indicators include the arithmetic mean, the standard deviation and the maximum value of the deviations of all sampling points.

8. The system of claim 1, wherein, The seven-parameter conversion model of the BIM-GIS fusion module includes three translation parameters, three rotation parameters and one scale parameter, and the conversion parameters are solved by least squares fitting of the control points; the R-tree index is constructed by using the STR algorithm, is sorted by x coordinates after grouping, and is sorted by y coordinates to form leaf nodes in the group; the octree index recursively divides the three-dimensional space until the number of components contained in each leaf node is less than a preset threshold.

Images