Building construction progress intelligent tracking and pushing method and system based on BIM

Abstract

The invention relates to the technical field of construction progress management, in particular to a building construction progress intelligent tracking and pushing method and system based on BIM, and the method comprises the following steps: S1, constructing a 3D-BIM model, building a mapping relation, and forming a progress plan model; s2, collecting dynamic construction progress data of a construction site, and generating a construction real spot cloud model; s3, based on the generated construction real spot cloud model, analyzing the construction progress monitoring data difference between the Ti time period and the Ti + 1 time period, and evaluating the actual construction speed; s4, comparing the construction progress deviation between the actual construction speed and the planned construction speed, and judging whether progress delay exists or not; and S5, when progress delay exists, identifying and classifying delay reasons, generating a subentry progress deviation report, and pushing a deviation correction scheme. According to the invention, dynamic and real-time visual progress tracking and control are realized, the real-time requirement of project design file progress dynamic tracking is met, and the fine management level of construction progress tracking is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of construction progress management, and in particular to a method and system for intelligently tracking and pushing construction progress based on BIM. Background Art

[0002] Prefabricated buildings are developing rapidly in my country. However, they involve multiple parties and links, and data integration and sharing are key to progress management. Through information and intelligent technologies, the efficiency and quality of construction progress management can be effectively improved, and resource waste and environmental impact can be reduced. BIM technology, as an emerging information technology, can create and manage digital information models of construction projects, enabling collaborative work among multiple parties and offering significant advantages in project progress visualization and planning. While existing prefabricated building construction progress management systems utilize this new technology and management approach, their application is superficial and unable to accurately describe and reflect the dynamic progress of the project. Problems exist, such as untimely and inaccurate information transmission, which results in irrational resource allocation and, in turn, project delays. Summary of the Invention

[0003] In view of this, the present invention proposes a method and system for intelligent tracking and pushing of construction progress based on BIM, aiming to solve problems such as poor information communication and insufficient effectiveness of corrective measures during the construction process, thereby further improving the level of refined project progress management.

[0004] The technical solution of the present invention is implemented as follows: The present invention provides a method for intelligent tracking and pushing of construction progress based on BIM, comprising the following steps:

[0005] S1. Build a 3D-BIM model based on the project design documents, and map the construction schedule information to the 3D-BIM model components through the component coding system to form a schedule model;

[0006] S2. Collect dynamic construction progress data from the construction site, including point cloud data and text information. Process the construction progress data through cross-modal feature fusion and multimodal self-attention mechanism to generate a construction site point cloud model.

[0007] S3. Analyze T based on the generated construction site cloud model. i Time period and T i+1 Differences in construction progress monitoring data during different time periods to assess actual construction speed;

[0008] S4. Extracting the planned construction speed for the corresponding time period based on the progress planning model, comparing the actual construction speed with the planned construction speed to see whether the construction progress deviation exceeds a preset threshold; if not, continuing construction as planned; if so, determining that there is a progress delay;

[0009] S5. When there is a progress delay, identify and classify the cause of the delay, generate a sub-item progress deviation report, and automatically push targeted correction plans to the relevant responsible parties based on different delay types.

[0010] Based on the above technical solution, preferably, the formation of the schedule model in step S1 includes:

[0011] S11. Divide the project design documents into component levels through the work breakdown structure and establish the relationship between component tasks and resource requirements;

[0012] S12. Use standardized coding rules to identify tasks in the work breakdown structure and components in the BIM model, and establish a corresponding mapping relationship between the two;

[0013] S13. Based on the established mapping relationship, the time sequence progress information in the work breakdown structure is associated with the 3D-BIM model components to generate a schedule model.

[0014] Based on the above technical solution, preferably, the text information includes:

[0015] Basic project information, including project name, zone number, collection date, collection personnel, and weather information;

[0016] Component information, including component number, component type, design size, production batch number, installation status, installation location, installation positioning deviation and elevation deviation;

[0017] Construction progress tracking information, including construction task number, task name, planned start time, planned completion time, actual completion time and task progress status.

[0018] Based on the above technical solution, preferably, the point cloud data includes point cloud data representing large-scale construction progress information collected by the lidar and point cloud data representing small-scale construction progress information collected by the depth camera, wherein the depth camera is calibrated using a camera calibration method based on multivariate quadratic regression, and the multivariate quadratic regression calibration equation is as follows:

[0019] X=X uc +C v X uc Y uc

[0020] Y=Y uc +C v X uc Y uc

[0021] C v =a(q,0)+a(q,1)Xuc +a(q,2)Y uc +a(q,3)X 2 uc +a(q,4)Y 2 uc

[0022] Among them C v represents the calibration coefficient, q represents the coordinate quadrant, a represents a set of calibration coefficients, (X uc ,Y uc ) are uncalibrated pixel coordinates.

[0023] Based on the above technical solution, preferably, the processing of the construction progress data by cross-modal feature fusion and multimodal self-attention mechanism in step S2 specifically includes:

[0024] S21, respectively analyzing the text information and point cloud data of different scales to obtain text features and point cloud features;

[0025] S22, performing multi-scale fusion of the extracted point cloud features of different scales through a spatial alignment feature pyramid network, and converting the text features and the multi-scale fused point cloud features into vectors of the same dimension through linear mapping;

[0026] S23. Add the position coding information of the component to the vector of the same dimension, input the vector with the position coding into the cross-modal Transformer model for feature fusion, and generate a construction site cloud model.

[0027] Based on the above technical solution, preferably, the cross-modal Transformer model includes:

[0028] The encoder part is composed of multiple identical layers stacked together. Each layer contains a multi-head self-attention mechanism and a feedforward neural network to process the fusion features of the input.

[0029] The decoder part is composed of multiple identical layers stacked together, each of which contains a masked multi-head self-attention mechanism, an encoder-decoder attention mechanism, and a feedforward neural network;

[0030] Each attention mechanism module is followed by a residual connection and layer normalization structure;

[0031] The encoder generates a context representation after processing the input feature sequence, and the decoder generates a final construction site point cloud model based on the context representation generated by the encoder.

[0032] On the basis of the above technical solution, preferably, in step S3, the construction progress monitoring data includes construction progress characteristics, machine group utilization rate and prefabricated component installation quantity, and the construction progress characteristics include component positioning deviation and elevation deviation.

[0033] On the basis of the above technical solution, preferably, in step S4, the preset thresholds include component positioning deviation not exceeding ±3m, elevation deviation not exceeding ±3m, T i Time period and T i+1 The deviation of component installation quantity during the period shall not exceed 10%, T i Time period and T i+1 The deviation of machinery utilization rate during each period shall not exceed 10%.

[0034] Based on the above technical solution, preferably, the determination of whether the construction progress deviation exceeds a preset threshold in step S4 includes the following judgment criteria:

[0035] Completion status: When all quantitative indicators do not exceed the preset threshold, it is determined to be completed;

[0036] Partially Completed: When no more than 30% of the quantitative indicators exceed the preset threshold but none of them exceeds 1.2 times the preset threshold, it is considered partially completed.

[0037] Unfinished status: When more than 30% of the quantitative indicators exceed the preset threshold, or any quantitative indicator exceeds 1.2 times the preset threshold, it is determined to be in an unfinished state;

[0038] When it is determined to be in a partially completed state or an uncompleted state, there is a progress delay, and the process goes to step S5 to identify the cause of the delay and push a correction plan.

[0039] The present invention also provides a BIM-based intelligent construction progress tracking and push system, which is used to implement the method described in any one of the above items, including:

[0040] The progress information management module is used to build 3D-BIM models, establish mapping relationships to form a progress plan model, and collect dynamic construction progress data at the construction site;

[0041] The information analysis module is used to process the collected point cloud data and text information to generate a construction site point cloud model;

[0042] The progress analysis module is used to analyze the differences in construction progress monitoring data at different time periods, evaluate the actual construction speed, and compare the deviation between the actual construction speed and the planned construction speed to determine whether there are any progress delays;

[0043] The progress investigation and push module is used to identify and classify the causes of delays when there are progress delays, generate item-by-item progress deviation reports, and push targeted correction plans to the relevant responsible parties.

[0044] The BIM-based intelligent construction progress tracking and push method and system of the present invention have the following beneficial effects compared with the prior art:

[0045] (1) The present invention realizes the intelligent management of the whole process of prefabricated building construction progress by organically combining technical means such as constructing a BIM schedule model integrating time dimensions, collecting multi-source construction site data, cross-modal feature fusion processing, and progress monitoring based on preset thresholds. This method breaks through the limitation of the single data source of traditional construction progress management. By comparing and analyzing construction data in different time periods, it can accurately identify and quantify construction progress deviations and automatically push targeted correction plans, significantly improving the refinement level of construction progress management and construction efficiency, and reducing the risk of progress delays.

[0046] (2) This invention adopts a technical solution combining a spatially aligned feature pyramid network with a cross-modal Transformer model to solve the technical problem of fusing point cloud data of different scales with text information. Through a multi-head self-attention mechanism and a residual connection layer normalization structure, it achieves effective integration of large-scale lidar point clouds, small-scale depth camera point clouds, and text features, significantly improving the integrity and accuracy of construction progress monitoring data and providing high-quality basic data support for construction progress analysis.

[0047] (3) The present invention introduces a multi-dimensional progress monitoring indicator system including construction progress characteristics, machine group utilization rate and prefabricated component installation volume. It uses an automatic registration method based on point cloud block features to accurately match point cloud data of different time periods, which can fully reflect the actual progress status of the construction site, realize comprehensive monitoring of construction progress from macro to micro, and reduce blind spots and information islands in progress management;

[0048] (4) The present invention establishes a three-level progress status judgment standard (completed status, partially completed status, and unfinished status) based on preset thresholds. Through clear quantitative indicators and judgment boundaries, the construction progress assessment is transformed from empirical judgment to data-driven precise judgment, eliminating the subjectivity and uncertainty of traditional progress assessment, improving the objectivity and reliability of progress assessment, and providing a scientific basis for timely discovery and correction of construction deviations. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 This is a flow chart of the BIM-based construction progress intelligent tracking and push method of the present invention;

[0051] Figure 2 This is a structural diagram of the BIM-based intelligent construction progress tracking and push system of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] like Figure 1 As shown, the present invention provides a method for intelligent tracking and pushing of building construction progress based on BIM, comprising the following steps:

[0054] S1. Build a 3D-BIM model based on the project design documents, and map the construction schedule information to the 3D-BIM model components through the component coding system to form a schedule model;

[0055] S2. Collect dynamic construction progress data from the construction site, including point cloud data and text information. Process the construction progress data through cross-modal feature fusion and multimodal self-attention mechanism to generate a construction site point cloud model.

[0056] S3. Analyze T based on the generated construction site cloud model. i Time period and T i+1 Differences in construction progress monitoring data during different time periods to assess actual construction speed;

[0057] S4. Extracting the planned construction speed for the corresponding time period based on the progress planning model, comparing the actual construction speed with the planned construction speed to see whether the construction progress deviation exceeds a preset threshold; if not, continuing construction as planned; if so, determining that there is a progress delay;

[0058] S5. When there is a progress delay, identify and classify the cause of the delay, generate a sub-item progress deviation report, and automatically push targeted correction plans to the relevant responsible parties based on different delay types.

[0059] The present invention realizes the accurate monitoring and quantitative analysis of construction progress by constructing a progress planning model and organically combining it with the collected real-time point cloud data. The cross-modal feature fusion and multi-modal self-attention mechanism are used to process point cloud and text data, which overcomes the limitation of incomplete information of traditional single data source and improves the accuracy and comprehensiveness of construction progress monitoring. By comparing the construction progress characteristics, machinery group utilization rate and prefabricated component installation volume changes in different time periods, the system can automatically identify construction delays and quantify the degree of deviation, realizing the transition from subjective experience judgment to objective data analysis. In particular, the function of automatically pushing targeted correction solutions for different delay types greatly shortens the response time from problem discovery to solution, and improves construction management efficiency. This method organically combines BIM technology, point cloud recognition technology and artificial intelligence algorithm to realize intelligent full-process management of construction progress, effectively reduce the cost of manual inspections, reduce the risk of construction delays, and provide a more scientific and efficient technical means for project management.

[0060] In step S1, a 3D-BIM model is first constructed based on the project design documents. Then, the construction schedule information is associated with the model to form a schedule model with a time dimension, including:

[0061] S11. Divide the project design documents into component levels through the work breakdown structure and establish the relationship between component tasks and resource requirements;

[0062] S12. Use standardized coding rules to identify tasks in the work breakdown structure and components in the BIM model, and establish a corresponding mapping relationship between the two;

[0063] S13. Based on the established mapping relationship, the time sequence progress information in the work breakdown structure is associated with the 3D-BIM model components to generate a schedule model.

[0064] During the construction preparation phase, a precise 3D-BIM model is constructed based on the design and construction drawings of the prefabricated building project. This model not only includes building geometry information but also key parameters such as component material properties, component type, and location information. To effectively link the BIM model with the construction schedule, a work breakdown structure (WBS) is used to break down the entire project design document into component-level components. Through the work breakdown structure (WBS), the project is broken down into its smallest manageable components, with each component's production, transportation, and installation tasks and corresponding resource allocation requirements clearly defined. This decomposition approach ensures refined management of the construction schedule. During the development of the component coding system, each component is standardized using the international standard OmniClass classification system, establishing a one-to-one correspondence between WBS task IDs and BIM component IDs. This standardized coding approach not only improves the standardization of information management but also lays the foundation for subsequent automated linkage. The WBS data is imported into the BIM software platform to extract key progress information, such as task ID, task name, planned start time, and planned end time. This progress information is then parametrically linked to the corresponding component elements in the BIM model, forming a linkage mapping table. Finally, combined with the construction organization design plan, the construction schedule information was linked to the 3D BIM model to generate a 4D BIM schedule model (4D-BIM) that integrates the time dimension. This model organically combines the construction schedule with 3D spatial information, providing basic data support for subsequent progress monitoring and analysis.

[0065] This application integrates the progress information of different stages in the construction of prefabricated buildings by constructing a 4D-BIM progress model of prefabricated buildings, based on digital information simulation of real building information. This integration supports visual and collaborative construction progress management, allowing project participants to understand the construction progress more intuitively, and collaborate on a unified platform to reduce communication barriers. The 4D-BIM model associates progress information with the BIM model, so that progress information can be delivered to relevant personnel more promptly and accurately, improving the effectiveness and accuracy of information transmission. It can refine the construction progress to the specific construction process level and support the tracking and management of resource inputs for key processes. This refined management helps project managers to promptly identify and resolve progress delays, optimize resource allocation, and ensure that construction proceeds as planned.

[0066] In step S2, it is necessary to collect dynamic construction progress data of the construction site, including point cloud data and text information, and process the construction progress data through cross-modal feature fusion and multimodal self-attention mechanism to generate a construction site point cloud model. The specific implementation method is as follows:

[0067] In an embodiment of the present invention, the collection of dynamic construction progress data includes two aspects: point cloud data collection and text information collection. The point cloud data includes point cloud data representing large-scale construction progress information collected by lidar and point cloud data representing small-scale construction progress information collected by depth camera:

[0068] Large-scale point cloud data acquisition based on lidar: A drone equipped with lidar conducts cruise scans of the construction site. By emitting laser pulses and receiving reflected signals, it acquires characteristic information such as the position and velocity of target objects, generating a point cloud model representing large-scale visual progress information. This method is suitable for macro-progress monitoring covering the entire construction site and can quickly capture large-scale visual progress information. A spatial partitioning method (Voronoi) is used to divide the target area into multiple grid cells, each representing an independent acquisition unit. K-means clustering is used to ensure that each drone's scanning area is similar. A 32-line lidar with a pulse repetition frequency (PRF) of 50kHz is used. The drones fly at an altitude of H = 100m, with a flight distance of D = 50m, a flight time of 20 minutes, and a speed of 10m / s. Real-time kinematic positioning (RTK) is used to enhance absolute accuracy, achieving a positioning error of ±2cm. Each drone uses the ant colony optimization (ACO) algorithm for path planning.

[0069] Small-scale point cloud data acquisition based on depth cameras: According to the scale and complexity of the construction site, multiple depth cameras are reasonably deployed to obtain depth images and color images from different angles in real time. Dynamic Fusion technology is used to generate a point cloud model that represents the progress information of small-scale construction images. A camera calibration method based on multivariate quadratic regression is used to achieve positioning calibration of the depth camera and optimize measurement accuracy. Camera calibration method based on multivariate quadratic regression: A high-precision calibration plate chessboard is used to move at multiple angles within the public field of view of the depth camera, and the calibration plate image is collected synchronously to ensure coverage of different angles and positions. The intersection of the vertical and horizontal lines in the calibration plate image is detected and its pixel coordinates (X uc ,Y uc ). The uncalibrated pixel coordinates (X uc ,Y uc ) Input the Multivariate Quadratic Regression (MQR) model, which directly corrects the pixel coordinate error caused by lens distortion and compensates the relative position deviation of the camera (calculates the compensation angle Comp, corrects the relative direction error of the camera, and optimizes the triangulation accuracy). The calibration equation is as follows:

[0070] X=X uc +C v X uc Yuc

[0071] Y=Y uc +C v X uc Y uc

[0072] C v Is a dynamically changing calibration coefficient used to adjust the uncalibrated coordinate value X uc and Y uc , making it closer to the real calibration coordinate values X and Y, which change according to the quadrant and pixel coordinate values, thereby obtaining better compensated coordinate values.

[0073] C v =a(q,0)+a(q,1)X uc +a(q,2)Y uc +a(q,3)X 2 uc +a(q,4)Y 2 uc

[0074] Where q represents the coordinate quadrant and a represents a set of calibration coefficients used to describe the uncalibrated coordinate value X uc and Y uc The nonlinear relationship between the calibrated coordinate values X and Y. The coefficient a is obtained by fitting a multivariate quadratic regression model.

[0075] Text information collection is mainly achieved through handheld PDAs: During the construction process, construction workers use handheld PDAs to directly input construction text information. At the same time, sensors are installed on transport vehicles, cranes, and prefabricated components. The handheld PDAs connect to different types of sensors through an industrial intelligent gateway, collect data from each sensor, and convert the collected data into XML text format through the gateway. The collected text information mainly includes: (1) basic project information: including project name, partition number, collection date, collection personnel, and weather information; (2) component information: including component number, component type, design size, production batch number, installation status, installation location, installation positioning deviation, and elevation deviation; (3) construction progress tracking information: including construction task number, task name, planned start time, planned completion time, actual completion time, and task progress status. See Table 1-3 for details.

[0076] Table 1 Basic information of the project

[0077] Field Name Input method Remark Project Name Drop-down selection Select from the system project library Partition number RFID scanning Select from preset numbers Collection date and time Automatically obtain GPS time System automatic recording Collector ID System account automatic binding Login account automatically associated Weather conditions Automatic sensor collection Sunny / Rainy / Fog / Snowy

[0078] Table 2 Component information

[0079] Field Name Input method Remark Component number RFID scanning Must comply with Omni Class coding standards Component Type Drop-down selection Select from preset types Design size Automatically associate BIM models Extracting data from BIM models Production batch number RFID scanning Must be consistent with the quality inspection report Installation Status Drop-down selection Not installed / located / fixed, with timestamp Installation location Autofill GNSS positioning Installation positioning deviation Automatic sensor collection Threshold ≤±3m Elevation deviation Automatic sensor collection Threshold ≤±3m

[0080] Table 3 Construction progress tracking information

[0081] Field Name Input method Remark Construction task number Drop-down selection Task number in the associated project plan Task Name Autofill Automatically populate based on task number Planned start time Automatically link project plans Extract from project plan Planned completion time Automatically link project plans Extract from project plan Actual start time Scan code to automatically record Record the actual start time of the task Actual completion time Scan code to automatically record Record the actual completion time of the task Task progress status Drop-down selection Not started / In progress / Completed

[0082] Synchronous data acquisition: Use timestamps to synchronize data collected by multiple cameras to ensure consistent frame rates.

[0083] The collected point cloud data and text information need to be processed through cross-modal feature fusion and multimodal self-attention mechanism, which includes the following steps:

[0084] S21. Analyze the text information and point cloud data of different scales to obtain text features and point cloud features.

[0085] Specifically, information analysis of point cloud data includes noise reduction, feature extraction, and registration. A bilateral filtering algorithm is used for noise reduction preprocessing. Voxel grid downsampling is used to divide the point cloud into a voxel grid, and the average value within each voxel is taken to reduce the data volume while preserving geometric features. A Mask Region-based Convolutional Neural Network (Mask R-CNN) is used to identify and segment the machinery, materials, and prefabricated components in the generated construction site point cloud model, extracting point cloud features. An automatic registration method based on point cloud block features is used to register and align point cloud data at different timestamps, enabling the integration of dynamic progress information at different timestamps.

[0086] Text parsing involves preprocessing and feature extraction: Regular expressions are used to clean the text, matching and removing irrelevant characters, numbers, and extraneous spaces to extract the pure text content. The natural language processing tool NLTK (Natural Language Toolkit) uses a tokenizer to segment the text, breaking sentences into words or subwords. The tokenizer identifies punctuation, spaces, and other delimiters within a sentence, breaking the text into meaningful units. NLTK's stop word list is used to filter out common but non-semantic words. The pos_tag function in NLTK performs part-of-speech tagging on words, analyzing their context within the sentence to determine their corresponding part of speech (e.g., noun, verb, adjective, etc.) to aid in understanding sentence structure and semantics. Semantic features of the text are extracted using the pretrained language model RoBERTa. The preprocessed text is encoded into an input format acceptable to the RoBERTa model and then fed into the model to generate context-sensitive embeddings for each word, thereby extracting the text's semantic features.

[0087] S22. The extracted point cloud features of different scales are multi-scale fused through a spatial alignment feature pyramid network, and the text features and the multi-scale fused point cloud features are converted into vectors of the same dimension through linear mapping.

[0088] The extracted point cloud features at different scales (large-scale LiDAR point cloud features and small-scale depth camera point cloud features) are fused at multiple scales using the Spatially Aligned Feature Pyramid Network (SAFPN). This network can fuse feature information at different scales while maintaining spatial alignment. The text features and the multi-scale fused point cloud features are converted into vectors of the same dimension through linear mapping, allowing features from different modalities to be represented in the same feature space.

[0089] The SAFPN architecture consists of a basic feature extraction layer, a feature pyramid layer, and a spatially aware alignment and fusion module (SAFM). It achieves feature dimensionality reduction through segmentation and accumulation, integrating channel attention and spatial attention mechanisms to enhance the network's focus on important information and capture fine-grained semantic details. It uses a convolutional neural network (ResNet) to extract feature maps of different scales. Through lateral connections, it fuses low-resolution feature maps with underlying feature maps of the same resolution. SAFM aligns and fuses feature maps of adjacent scales to address spatial misalignment and extract richer, fine-grained features.

[0090] S23. Add component position coding information to the same-dimensional vector, enhance the spatial perception of features through position coding, and input the vector with position coding into the cross-modal Transformer model for feature fusion to generate a construction site cloud model. The position coding information refers to the three-dimensional spatial coordinates (x, y, z) of the component in the construction scene and its directional attributes. The spatial position is converted into a vector using sine and cosine functions. These codes enable the model to identify the relative spatial relationships and geometric positions of components, preserve the topological structure between components, enhance the spatial perception of features, and thus improve the accuracy of the cross-modal Transformer model in processing component spatial relationships.

[0091] Input point cloud features and text features into the cross-modal Transformer model. Through the self-attention mechanism and encoder-decoder structure, the Transformer automatically learns the correspondence between point cloud and text, achieves feature alignment and fusion, generates a high-quality 3D point cloud model, and improves the accuracy and completeness of information. The Transformer model includes:

[0092] The encoder is composed of multiple identical layers (preferably six layers), each layer including a multi-head self-attention mechanism and a feedforward neural network to process the fused input features. Each module is followed by a residual connection and layer normalization. The decoder is composed of multiple identical layers (preferably six layers), each layer including a masked multi-head self-attention mechanism, an encoder-decoder attention mechanism, and a feedforward neural network. Each module is also followed by a residual connection and layer normalization. The encoder is responsible for processing the input sequence and converting it into a context-rich representation. The decoder generates the target language sentence based on the context vector generated by the encoder.

[0093] Through the above processing steps, the present invention effectively integrates point cloud data and text information, generating a construction site point cloud model containing rich semantic information, providing high-quality basic data for subsequent construction progress analysis. This method utilizes lidar, a depth camera, and a handheld PDA (Personal Digital Assistant) to continuously acquire construction scene progress information. For the lidar, a point cloud model representing large-scale visual progress information is generated using unmanned aerial vehicle (UAV)-mounted lidar technology; for the depth camera, a point cloud model representing small-scale visual progress information is generated using Dynamic Fusion technology; and for the PDA, construction workers use the handheld PDA to record text information about construction progress in real time. Simultaneously, an industrial intelligent gateway connects different types of sensors to collect data from each sensor. Subsequently, a Transformer-based spatially aligned feature pyramid network (SAFPN) is used to fuse the lidar-generated point cloud model representing large-scale visual progress information with the depth camera-based dynamic fusion technology and text information collected by the PDA to generate a high-quality 3D point cloud model, improving the accuracy and completeness of information acquisition.

[0094] In step S3, based on the construction site cloud model generated in step S2, the different time periods (T i Time period and T i+1 The actual construction speed is evaluated by comparing the construction progress monitoring data of different time periods. The specific implementation is as follows:

[0095] For different time periods (T i Time period and T i+1To analyze discrepancies in construction progress monitoring data from different time periods (e.g., different time periods), it is first necessary to register the point cloud data from different time periods. In an embodiment of the present invention, an automatic registration method based on point cloud block features is used to register point cloud data from different time periods. This method extracts key point features from the point cloud and employs the RANSAC random sampling consensus algorithm and the iterative closest point (ICP) algorithm to achieve precise matching of point cloud data from different time periods, providing a foundation for subsequent analysis.

[0096] After the registration is completed, by comparing T i Time period and T i+1 The point cloud model of the time period is used to extract the following three types of monitoring data: construction progress characteristics, machine group utilization rate, and prefabricated component installation volume. Construction progress characteristics include component positioning deviation and elevation deviation. Based on the extracted monitoring data, the process of evaluating the actual construction speed in this invention is as follows: T i The visual progress characteristics of key processes in a period and T i+1 Comparison of visual progress characteristics of key processes during the period, including changes in component positioning deviation and elevation deviation, as a basis for evaluating construction quality progress; i The utilization rate of various machines during the period and T i+1 Compare the utilization rates of various types of machinery during each period to evaluate the efficiency of construction resource utilization; i The installation quantity of prefabricated components in a certain period and T i+1 Compare the installation volume of prefabricated components during each period to quantify the actual progress of the installation work. Comprehensively analyze the differences and changes in the above three types of monitoring data to evaluate T i to T i+1 The actual construction speed during a period is determined by taking into account the comprehensive changes in project progress, resource utilization efficiency, and actual installation volume, thereby obtaining an objective assessment of the actual construction speed and providing a basis for subsequent comparison with the planned construction speed.

[0097] In step S4, the planned construction speed for the corresponding time period is extracted based on the progress planning model formed in step S1, and compared with the actual construction speed evaluated in step S3 to determine whether the construction progress deviation exceeds a preset threshold, thereby determining whether there is a progress delay. The specific implementation method is as follows:

[0098] Analyze and judge T i Time period and T i+1 Whether the component positioning deviation and elevation deviation of the key process in the time period are greater than the preset threshold; analyze and judge T i The utilization rate of various machines during the period and T i+1 Whether the difference in the utilization rate of various types of machinery during the period is greater than the preset threshold; analyze and judge T i The installation quantity of prefabricated components in a certain period and T i+1 Whether the difference in the installation quantity of prefabricated components in each period is greater than the preset threshold.

[0099] Judgment threshold: The plane position deviation of the component generally does not exceed ±3m, the elevation deviation generally does not exceed ±3m, T i Time period and T i+1 The deviation of component installation quantity during the period shall not exceed 10%, T i Time period and T i+1 The deviation of machinery utilization rate during each period shall not exceed 10%.

[0100] Judgment criteria: The completion status of the process is divided into completed, partially completed, and uncompleted; if completed, continue construction as planned; if partially completed or uncompleted, proceed to step S5 to adjust the construction organization and resource allocation.

[0101] Completion: When all quantitative indicators do not exceed the preset threshold, the component installation position is accurate and the elevation meets the requirements, T i Time period and T i+1 The deviation between the component installation quantity and the machinery utilization rate during the period is within the threshold range, and it is determined to be completed. The progress is not delayed and construction continues as planned.

[0102] Partial completion: When no more than 30% of the quantitative indicators exceed the preset threshold but none of them exceeds 1.2 times the preset threshold, it is determined to be partially completed and the progress is delayed. Step S5 is entered to adjust the construction organization and resource allocation.

[0103] Unfinished: When more than 30% of the quantitative indicators exceed the preset threshold, or any quantitative indicator exceeds 1.2 times the preset threshold, it is determined to be in an unfinished state, the progress is delayed, and step S5 is entered to adjust the construction organization and resource allocation.

[0104] This invention addresses the numerous uncertainties and schedule adjustments inherent in the construction process. By identifying and quantifying the differences in point clouds of construction visual progress tracking elements acquired at fixed intervals, it tracks the progress of key processes and enables dynamic tracking of construction progress. This enables project managers to keep abreast of construction progress, promptly identify and correct deviations, and improve the refinement and efficiency of progress management.

[0105] In step S5, when the progress is delayed, a detailed progress report is generated based on the completion rate of the key processes at different timestamps. The reasons for the delay (such as staff shortage, insufficient supply of prefabricated components and mechanical equipment) are analyzed, checked and determined based on the detailed progress report. Finally, targeted corrective measures and plans are pushed based on the reasons for the delay, and management personnel adjust the construction organization and resource allocation accordingly.

[0106] Based on the BIM-based intelligent tracking and pushing method for prefabricated building construction progress of the present invention, Figure 2As shown, the present invention also provides a BIM-based intelligent construction progress tracking and push system, including:

[0107] The progress information management module is used to build 3D-BIM models, establish mapping relationships to form a progress plan model, and collect dynamic construction progress data from the construction site. This module is used to collect, store, and manage system data, including a resource information library and a dynamic progress information library. The resource information library includes the progress plan model, schedule, and checklist information, while the dynamic progress information library includes point cloud information, text information, and a construction site point cloud model.

[0108] In the examples of the present invention, a drone equipped with a laser radar is used to obtain real-time point cloud information representing large scales; a depth camera is combined with Dynamic Fusion technology to obtain real-time point cloud information representing small scales; a handheld terminal PDA is connected to a sensor to obtain dynamic construction progress text information, integrating multiple data sources to ensure data integrity and real-time performance.

[0109] The information analysis module is used to process the collected point cloud data and text information to generate a construction site point cloud model; point cloud information analysis includes data preprocessing, feature extraction, and data registration; text information analysis includes text preprocessing and text feature extraction.

[0110] The progress analysis module is used to analyze the differences in construction progress monitoring data at different time periods, evaluate the actual construction speed, and compare the deviation between the actual construction speed and the planned construction speed to determine whether there are any progress delays;

[0111] The progress tracking and notification module identifies and categorizes the causes of delays, generates itemized progress deviation reports, and delivers targeted corrective actions to the responsible parties. This module provides real-time notifications, ensuring managers receive timely progress updates and issue warnings.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for intelligent tracking and pushing of construction progress based on BIM, characterized in that: The following steps are involved: S1. Build a 3D-BIM model based on the project design documents, and establish a mapping relationship between the construction schedule information and the 3D-BIM model components through the component coding system to form a schedule model; S2. Collect dynamic construction progress data from the construction site, including point cloud data and text information. Process the construction progress data through cross-modal feature fusion and multimodal self-attention mechanism to generate a construction site point cloud model. S3. Analyze T based on the generated construction site cloud model. i Time period and T i+1 Differences in construction progress monitoring data during different time periods to assess actual construction speed; S4. Extracting the planned construction speed for the corresponding time period based on the progress planning model, comparing the actual construction speed with the planned construction speed to see whether the construction progress deviation exceeds a preset threshold; if not, continuing construction as planned; if so, determining that there is a progress delay; S5. When there is a progress delay, identify and classify the cause of the delay, generate a sub-item progress deviation report, and automatically push targeted correction plans to the relevant responsible parties based on different delay types.

2. The method for intelligent tracking and pushing of construction progress based on BIM as claimed in claim 1, characterized in that: The formation of the schedule model in step S1 includes: S11. Decompose the project design documents into the smallest manageable component level through the work breakdown structure, and clarify the production, transportation, installation tasks and resource allocation requirements of each component; S12. Use standardized coding rules to identify tasks in the work breakdown structure and components in the BIM model, and establish corresponding mapping relationships between tasks and components; S13. Based on the established mapping relationship, the time schedule information in the work breakdown structure, including the production, transportation, installation and resource allocation requirements of the tasks, is associated with the corresponding components of the 3D-BIM model to generate a schedule model.

3. The method for intelligent tracking and pushing construction progress based on BIM as claimed in claim 1, characterized in that: The text information includes: Basic information of the project design document, including the project design document name, zone number, collection date, collection personnel and weather information; Component information, including component number, component type, design size, production batch number, installation status, installation location, installation positioning deviation and elevation deviation; Construction progress tracking information, including construction task number, task name, planned start time, planned completion time, actual completion time and task progress status.

4. The method for intelligent tracking and pushing of construction progress based on BIM according to claim 1, characterized in that: The point cloud data includes point cloud data representing large-scale construction progress information collected by lidar and point cloud data representing small-scale construction progress information collected by depth camera. The depth camera is calibrated using a camera calibration method based on multivariate quadratic regression. The multivariate quadratic regression calibration equation is as follows: X=X uc +C v X uc Y uc Y=Y uc +C v X uc AND uc C v =a(q,0)+a(q,1)X uc +a(q,2)Y uc +a(q,3)X 2 uc +a(q,4)Y 2 uc Among them C v represents the calibration coefficient, q represents the coordinate quadrant, a represents a set of calibration coefficients, (X uc ,Y uc ) are uncalibrated pixel coordinates.

5. The method for intelligent tracking and pushing of construction progress based on BIM as claimed in claim 4, characterized in that: In step S2, the construction progress data is processed through cross-modal feature fusion and multimodal self-attention mechanism, specifically including: S21, respectively analyzing the text information and point cloud data of different scales to obtain text features and point cloud features; S22, performing multi-scale fusion of the extracted point cloud features of different scales through a spatial alignment feature pyramid network, and converting the text features and the multi-scale fused point cloud features into vectors of the same dimension through linear mapping; S23. Add the position coding information of the component to the vector of the same dimension, input the vector with the position coding into the cross-modal Transformer model for feature fusion, and generate a construction site cloud model.

6. The method for intelligent tracking and pushing of construction progress based on BIM as claimed in claim 5, characterized in that: The cross-modal Transformer model includes: The encoder part is composed of multiple identical layers stacked together. Each layer contains a multi-head self-attention mechanism and a feedforward neural network to process the fusion features of the input. The decoder part is composed of multiple identical layers stacked together, each of which contains a masked multi-head self-attention mechanism, an encoder-decoder attention mechanism, and a feedforward neural network; Each attention mechanism module is followed by a residual connection and layer normalization structure; The encoder generates a context representation after processing the input feature sequence, and the decoder generates a final construction site point cloud model based on the context representation generated by the encoder.

7. The method for intelligent tracking and pushing construction progress based on BIM as claimed in claim 1, characterized in that: In step S3, the construction progress monitoring data includes construction progress characteristics, machine group utilization rate and prefabricated component installation quantity, and the construction progress characteristics include component positioning deviation and elevation deviation.

8. The method for intelligent tracking and pushing of construction progress based on BIM as claimed in claim 7, characterized in that: In step S4, the preset thresholds include component positioning deviation not exceeding ±3m, elevation deviation not exceeding ±3m, T i Time period and T i+1 The deviation of component installation quantity during the period shall not exceed 10%, T i Time period and T i+1 The deviation of machinery utilization rate during each period shall not exceed 10%.

9. The method for intelligent tracking and pushing of construction progress based on BIM according to claim 1, characterized in that: In step S4, determining whether the construction progress deviation exceeds a preset threshold includes the following criteria: Completion status: When all quantitative indicators do not exceed the preset threshold, it is determined to be completed; Partially Completed: When no more than 30% of the quantitative indicators exceed the preset threshold but none of them exceeds 1.2 times the preset threshold, it is considered partially completed. Unfinished status: When more than 30% of the quantitative indicators exceed the preset threshold, or any quantitative indicator exceeds 1.2 times the preset threshold, it is determined to be in an unfinished state; When it is determined to be in a partially completed state or an uncompleted state, there is a progress delay, and the process goes to step S5 to identify the cause of the delay and push a correction plan.

10. A BIM-based intelligent construction progress tracking and push system, characterized by: The system is used to implement the method according to any one of claims 1 to 9, comprising: The progress information management module is used to build 3D-BIM models, establish mapping relationships to form a progress plan model, and collect dynamic construction progress data at the construction site; The information analysis module is used to process the collected point cloud data and text information to generate a construction site point cloud model; The progress analysis module is used to analyze the differences in construction progress monitoring data at different time periods, evaluate the actual construction speed, and compare the deviation between the actual construction speed and the planned construction speed to determine whether there are any progress delays; The progress investigation and push module is used to identify and classify the causes of delays when there are progress delays, generate item-by-item progress deviation reports, and push targeted correction plans to the relevant responsible parties.

Images