A BIM-based construction progress intelligent analysis method and system

Abstract

The application provides a BIM-based construction progress intelligent analysis method and system, relates to the technical field of intelligent analysis, acquires a BIM design model of a target construction project, constructs a construction progress planning network graph based on the BIM design model, collects multi-source monitoring data of a construction site in real time, performs fusion processing on the multi-source monitoring data, and generates a real-time construction scene digital twin; the real-time construction scene digital twin is registered with the BIM design model, actual construction progress features are extracted, the progress deviation amount is determined according to the actual construction progress features, lag risk task nodes are identified based on the progress deviation amount, the influence on subsequent construction progress is predicted, progress warning information is generated, and a construction progress adjustment strategy scheme is generated according to the progress warning information. The application can realize real-time and accurate perception of the construction progress state and early prediction of lag risks, and improve the fine level and response timeliness of construction progress management.

Description

Technical Field

[0001] This application relates to the field of intelligent analysis technology, and more specifically, to a BIM-based intelligent analysis method and system for construction progress. Background Technology

[0002] Currently, construction progress management in building projects mainly relies on construction management personnel conducting regular on-site inspections and manual measurements. The actual completion status is compared with the planned construction schedule, and the completion status of each construction process is recorded using Gantt charts or tables. Experience is then used to determine if there are any schedule deviations. Some existing solutions attempt to simply overlay the collected data with the BIM design model, observing the differences between the actual construction progress and the design model through manual visual comparison. Other solutions utilize image recognition technology to automatically determine the installation status of construction components, using whether a component has been installed as the criterion for judging progress completion, and presenting the progress comparison results to management personnel in the form of lists or charts.

[0003] However, current technologies rely on the superficial criterion of whether components have been installed to assess construction progress. This neglects the comprehensive impact of deeper, unobservable construction drivers on the actual progress during component installation, such as actual geometric deviations, installation resistance, accessibility of the work surface, and interference from multiple parallel trades. This results in insufficient accuracy in determining progress deviations. Therefore, achieving real-time and accurate perception of construction progress status and early prediction of delay risks, thereby improving the refinement and timeliness of construction progress management, is a challenge facing the industry. Summary of the Invention

[0004] This application provides a BIM-based intelligent analysis method and system for construction progress, which can realize real-time and accurate perception of construction progress status and advance prediction of delay risks, thereby improving the refinement level and responsiveness of construction progress management.

[0005] Firstly, this application provides a BIM-based intelligent analysis method for construction progress, the intelligent analysis method comprising the following steps: Obtain the BIM design model of the target building project, and construct a construction schedule network diagram based on the BIM design model; Real-time collection of multi-source monitoring data from the construction site; fusion processing of the multi-source monitoring data; generation of a digital twin of the real-time construction scene. The real-time construction scene digital twin is registered with the BIM design model, the actual construction progress characteristics of each construction task node are extracted, and the progress deviation of each construction task node is determined based on the actual construction progress characteristics. Based on the aforementioned schedule deviation, delayed risk task nodes are identified, their impact on subsequent construction progress is predicted, schedule early warning information is generated, and a construction schedule adjustment strategy is generated based on the schedule early warning information.

[0006] In this embodiment, the construction schedule network diagram is generated based on the component hierarchy structure in the BIM design model. Each construction task node in the construction schedule network diagram is uniquely bound to a component or a combination of components in the BIM design model. The task attributes of the construction task nodes are automatically extracted from the component attribute fields of the BIM design model and mapped to the node parameters in the construction schedule network diagram. The node parameters include at least the planned start time, planned end time, planned duration, and a list of prerequisite dependent components.

[0007] In this embodiment, the multi-source monitoring data includes: UAV oblique photogrammetry data, video stream data from ground-fixed panoramic cameras, vehicle-mounted positioning and attitude sensing data of construction machinery, data from wearable positioning terminals of construction personnel, fixed environmental sensing data, and data on the arrival and consumption records of construction materials.

[0008] In this embodiment, registering the real-time construction scene digital twin with the BIM design model and extracting the actual construction progress features of each construction task node specifically includes: The temporal variation feature field is extracted from the digital twin of the real-time construction scene. The temporal variation feature field is generated by determining the spatial gradient difference tensor between the digital twin at the current moment and the digital twin at the previous moment. The temporal variation feature field is used as the driving prior for dynamic registration, guiding the registration algorithm to focus on regions with deformation rates greater than preset deformation rates. A first registration sampling point is set in the region and a second registration sampling point is set in the region with a deformation rate of zero, thereby constructing a non-uniform adaptive registration point cloud. Extract the curvature extremum lines and edge topology of the surfaces of each component in the BIM design model, and construct the parametric feature manifold corresponding to each component; The non-uniform adaptive registration point cloud is used to perform manifold mapping matching with the parameterized feature manifolds corresponding to each component, thereby determining the six-degree-of-freedom pose drift of each component region; The non-uniform adaptive registration point cloud is projected onto the surface of the BIM design model corresponding to the parameterized feature manifold of each component to determine the actual surface coverage of each component. Then, the six-degree-of-freedom pose drift and the actual surface coverage are used as the actual construction progress features of each construction task node.

[0009] In this embodiment, determining the progress deviation of each construction task node based on the actual construction progress characteristics specifically includes: An implicit progress state space is constructed, wherein each dimension of the implicit progress state space corresponds to a construction driving force that cannot be directly observed at the construction site; The actual construction progress features are taken as observable explicit variables. The implicit variables in the implicit progress state space are used as an intermediate bridge to establish a hidden-explicit bidirectional mapping generative adversarial network. The discriminator of the generative adversarial network is used to determine whether the progress features generated under the current combination of implicit variables are consistent with the actual observation. The generator is used to infer the posterior distribution of the implicit variables from the actual observation. The posterior distribution of the latent variables is sampled, and the latent variable combination obtained from each sample is input into a construction progress forward simulator based on physical constraints to obtain the progress simulation results under each sample. The progress simulation results under all samples are statistically analyzed to obtain the progress completion probability cloud of each construction task node, and then the progress deviation of each construction task node is determined.

[0010] In this embodiment, the construction driving force corresponding to each dimension in the implicit progress state space includes at least: the actual installation impedance of the component, the effective accessibility coefficient of the on-site work surface, the interference intensity of parallel operations of multiple trades, the attenuation coefficient of the actual working efficiency of construction machinery, and the material supply delay coefficient.

[0011] In this embodiment, the schedule deviation is used to characterize the degree of difference between the actual progress status and the planned progress status of each construction task node.

[0012] In this embodiment, identifying delayed risk task nodes based on the progress deviation specifically includes: The progress deviation of each construction task node is compared with a preset lag judgment threshold, and the construction task node whose progress deviation is greater than the preset lag judgment threshold is marked as a lag candidate node. Obtain the topological location parameters of each lagging candidate node in the construction schedule network diagram. The topological location parameters include at least the out-degree, in-degree, and longest path length from the node to the project endpoint. Based on the progress deviation and topology location parameters of each lagging candidate node, the risk priority score of each lagging candidate node is determined, and the lagging candidate node with a risk priority score greater than a preset risk score threshold is designated as a lagging risk task node.

[0013] In this embodiment, the preset risk scoring threshold is a pre-calibrated fixed threshold, which is determined based on the 75th percentile of the statistical distribution of the risk priority scores of lagging candidate nodes of at least three historical construction projects.

[0014] Secondly, this application provides a BIM-based intelligent construction progress analysis system for executing a BIM-based intelligent construction progress analysis method, the intelligent analysis system comprising: The network construction module is used to acquire the BIM design model of the target building project and construct a construction schedule network diagram based on the BIM design model. The data fusion module is used to collect multi-source monitoring data from the construction site in real time, fuse the multi-source monitoring data, and generate a digital twin of the real-time construction scene. The schedule deviation module is used to register the digital twin of the real-time construction scene with the BIM design model, extract the actual construction progress characteristics of each construction task node, and determine the schedule deviation of each construction task node based on the actual construction progress characteristics. The scheme generation module is used to identify delayed risk task nodes based on the progress deviation, predict the impact on subsequent construction progress, generate progress early warning information, and generate construction progress adjustment strategy schemes based on the progress early warning information.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: Obtain the BIM design model of the target building project and construct a construction schedule network diagram based on the BIM design model; collect multi-source monitoring data from the construction site in real time, fuse the multi-source monitoring data, and generate a digital twin of the real-time construction scene; register the digital twin of the real-time construction scene with the BIM design model, extract the actual construction progress characteristics of each construction task node, determine the progress deviation of each construction task node based on the actual construction progress characteristics; based on the progress deviation, identify the lagging risk task nodes, predict the impact on subsequent construction progress, generate progress early warning information, and generate a construction progress adjustment strategy based on the progress early warning information.

[0016] Therefore, this application firstly constructs a construction schedule network diagram based on the BIM design model, realizing the unique binding between construction task nodes and BIM components and the automatic extraction and mapping of task attributes, providing a planning benchmark that corresponds to the design model for subsequent progress monitoring; then, by collecting multi-source monitoring data in real time and performing fusion processing to generate a digital twin of the real-time construction scene, a digital mapping of the physical space and geometric shape of the construction site is realized, providing a data foundation that is updated synchronously with the actual construction for subsequent progress feature extraction; furthermore, by registering the digital twin of the real-time construction scene with the BIM design model and extracting actual construction progress features, progress deviation quantification based on the actual construction geometry is realized, so that progress assessment no longer relies on manual reporting, but is directly derived from the real spatial state data of the construction site; finally, by identifying lagging risk task nodes based on the progress deviation, predicting the subsequent progress impact, and generating early warning information and adjustment strategy schemes, a complete control closed loop from lagging identification to forward warning and then to strategy output is realized, improving the refinement level and response timeliness of construction progress management.

[0017] In summary, the technical solution adopted in this application can achieve real-time and accurate perception of construction progress status and advance prediction of delay risks, thereby improving the refinement level and response timeliness of construction progress management. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this embodiment of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an exemplary flowchart of a BIM-based intelligent analysis method for construction progress provided in this application; Figure 2 This is a schematic diagram of a device frame for intelligent analysis of construction progress based on BIM, provided in this application. Figure 3 This is a module structure diagram of the BIM-based intelligent construction progress analysis system provided in this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0021] This application provides a BIM-based intelligent construction progress analysis method and system. Its core is to acquire the BIM design model of the target building project and construct a construction progress schedule network diagram based on the BIM design model; collect multi-source monitoring data from the construction site in real time, fuse the multi-source monitoring data to generate a digital twin of the real-time construction scene; register the digital twin of the real-time construction scene with the BIM design model, extract the actual construction progress characteristics of each construction task node, determine the progress deviation of each construction task node based on the actual construction progress characteristics; based on the progress deviation, identify lagging risk task nodes, predict their impact on subsequent construction progress, generate progress warning information, and generate a construction progress adjustment strategy based on the progress warning information.

[0022] Example 1: To better understand the above technical solution, the following will provide a detailed description of the technical solution in conjunction with the accompanying drawings and specific implementation methods. (Refer to...) Figure 1 As shown in the figure, this is an exemplary flowchart of a BIM-based intelligent analysis method for construction progress according to this embodiment of the present application. The intelligent analysis method includes the following steps: In step S1, the BIM design model of the target building project is obtained, and a construction schedule network diagram is constructed based on the BIM design model.

[0023] In practical implementation, the BIM design model of the target building project is acquired, and a construction schedule network diagram is constructed based on the BIM design model. Specifically, the BIM design model uses a common data format such as IFC or Revit, and contains information on all components of the target building project. Each component carries at least a unique identifier, component name, component type, material grade, geometric dimensions, spatial coordinates, and bounding box information. The BIM design model is directly imported from the design unit through the construction project management platform. The component hierarchy of the BIM design model is then parsed, traversing the component tree. Based on the compositional relationships, spatial inclusion relationships, and subcontracting affiliations between components, all components are organized into a three-layer structure: unit project layer, sub-project layer, and itemized project layer. The components or component combinations corresponding to the itemized project layer serve as the smallest binding unit for subsequent construction task nodes. For multiple closely related components within the same itemized project that must be constructed concurrently, they are merged into a single component combination, and a unified combination code is assigned. For each component or component combination corresponding to each sub-project layer, the construction process code in its attribute field is read. This construction process code corresponds to the process number in the current construction specifications. Simultaneously, the planned installation start date, planned installation end date, planned installation duration, and list of prerequisite dependent components for that component are read. If any of the above parameters are missing in the attribute field, the industry average duration parameter of the project category to which the component belongs is used as the default value to complete the data. Each component or component combination corresponding to each sub-project layer is treated as a construction task node. A unique node number is generated for each construction task node, and the planned start time, planned end time, planned duration, and list of prerequisite dependent components are used as the initial parameters of that node and stored in the node attribute table. The construction schedule network diagram is generated based on the component hierarchy structure in the BIM design model. Each construction task node in the construction schedule network diagram is uniquely bound to a component or component combination in the BIM design model. The task attributes of the construction task node are automatically extracted from the component attribute fields of the BIM design model and mapped to the node parameters in the construction schedule network diagram. The node parameters include at least the planned start time, planned end time, planned duration, and list of prerequisite dependent components.

[0024] In addition, in the specific implementation, the list of prerequisite components for all construction task nodes is traversed. The construction task node corresponding to the prerequisite component is taken as the direct predecessor node of the current node. A directed edge is established between the current node and each of its direct predecessor nodes, with the direction of the directed edge pointing from the direct predecessor node to the current node, indicating that the current node can only start after the direct predecessor node is completed. If a construction task node has no prerequisite components, then this node is the starting node; if a construction task node is not a prerequisite node for any other node, then this node is the ending node. All directed edges constitute the directed acyclic graph skeleton of the construction schedule network diagram. Then, the spatial position coordinates and bounding box information of each component in the BIM design model are traversed to determine whether there is a spatial collision intersection or sharing of the same working surface between the components corresponding to any two construction task nodes. If the above situations exist, a soft constraint association edge is established between the two construction task nodes. The soft constraint association edge is stored as an implicit logical dependency in a separate soft constraint list, which is different from the explicitly defined directed edges. The associated edges in the soft constraint list do not participate in the basic topology calculation of the construction schedule network diagram, but they are used as weighted adjustment factors in subsequent schedule deviation analysis and risk propagation prediction. The system checks whether all construction task nodes satisfy the directed acyclic graph constraints; if a directed cycle is detected, an alarm is triggered, prompting the user to adjust the list of prerequisite dependent components. The system also checks whether the planned duration of all construction task nodes is positive; if a negative or zero value is found, it is replaced with a preset minimum duration value.

[0025] In step S2, multi-source monitoring data of the construction site are collected in real time, and the multi-source monitoring data is fused to generate a digital twin of the real-time construction scene.

[0026] In practical implementation, multi-source monitoring data is collected in real time at the construction site. Specifically, before construction begins, various data acquisition systems are deployed at the site, including a drone oblique photogrammetry system. This system utilizes a drone platform equipped with a five-lens oblique photogrammetry camera and a real-time differential positioning module, planning a cruise route covering the entire construction area with an 80% forward overlap rate and a 70% lateral overlap rate. Data is collected twice daily during the main structure construction phase and once daily during the decoration and finishing phase. A ground-based fixed panoramic camera system continuously collects video streams using panoramic cameras installed on the standard sections of tower cranes and at the edges of high-rise work surfaces. All cameras are synchronized via a unified clock through a network time protocol server. A vehicle-mounted positioning and attitude sensing system for construction machinery is implemented, installing a global navigation satellite system receiver, inertial measurement unit, and hydraulic sensors on each piece of machinery. This system transmits the machinery's three-dimensional coordinates, heading angle, pitch angle, roll angle, and boom extension status once per second. A wearable positioning terminal system for construction personnel is also included. Each worker is equipped with an ultra-wideband positioning smart safety helmet, which, in conjunction with on-site ultra-wideband positioning base stations, enables real-time positioning and categorizes and labels the positioning data according to job type. A fixed environmental sensing system is implemented by deploying environmental monitoring stations with integrated temperature, humidity, wind speed, and illuminance sensors at the work site and material storage area to collect environmental parameters. A construction material arrival and consumption recording system uses a weighbridge, RFID readers at the material storage yard, and an automatic metering system at the concrete mixing plant to collect real-time data on material arrival quantities, remaining inventory, and consumption.

[0027] In practical implementation, multi-source monitoring data is fused to generate a digital twin of the real-time construction scene. First, all multi-source monitoring data is projected onto a dynamic adaptive spatiotemporal coordinate system. The coordinate system uses the origin of the BIM design model as the initial reference point. Every hour, the actual construction bounding box is calculated based on the spatial coordinate extreme values ​​of all construction machinery positioning data, construction personnel positioning data, and UAV point cloud data at the current moment. The geometric center of this bounding box is used as the coordinate origin, and the direction of the principal inertial axis is used as the coordinate axis direction. The origin and axis of the coordinate system synchronously follow the offset of the bounding box. Cubic spline interpolation is used for smooth transition during the offset process, with a transition time of five minutes. Second, construction semantic tags for each component are generated. The attribute field text of each component in the BIM design model is extracted and concatenated into a component description string. This string is input into a pre-trained language model in the construction domain to obtain a 768-dimensional semantic embedding vector for each component. Density clustering is performed on the semantic embedding vectors of all components, and each cluster in the clustering result is assigned a unique construction semantic tag. Components within the same cluster share the same semantic tag in subsequent processing. Next, heterogeneous data association mapping based on construction semantic tags is performed on each data source after projection. UAV point clouds are associated with the nearest component surface according to spatial coordinates to obtain semantic tags; video stream data identifies components and matches and binds semantic tags using object detection algorithms; mechanical positioning data and personnel positioning data are assigned to corresponding semantic tags based on the spatial inclusion relationship between positioning points and component bounding boxes. After mapping, monitoring data from multiple heterogeneous data sources are aggregated under each semantic tag. Then, a three-layer association constraint relationship is established: data layer constraints, feature layer constraints, and semantic layer constraints. Data layer constraints normalize the values ​​of each data source to the same dimension range and unify the sampling frequency to a fusion time step of sixty seconds through linear interpolation resampling; feature layer constraints optimize the coordinate offset compensation parameters of each data source by minimizing the weighted sum of squared spatial distance residuals between the geometric feature vectors and visual feature vectors of each data source; semantic layer constraints forcibly correct inconsistent semantic tags in the digital twin based on the semantic attributes of the BIM design model.

[0028] Furthermore, in the specific implementation, a graph neural network is used for joint encoding of structural semantics. Each component is treated as a node, and the node features are 384-dimensional fusion feature vectors generated by fusing all monitoring data associated with the component. Logical dependency edges and spatial adjacency edges are used as graph edges. A message passing mechanism updates the node features at each fusion time step. All updated node features are concatenated in component number order to form a construction state tensor. Finally, the real-time construction scene digital twin is iteratively updated. The construction state tensor of the current time step is input into a time series predictor based on gated recurrent units, and the predictor outputs the predicted construction state tensor for the next time step. The predicted tensor is compared with the actual construction state tensor of the next time step, and the Frobenius norm is calculated as the tensor residual. If the residual is less than a preset threshold, the digital twin is updated with the predicted tensor; if the residual is greater than or equal to the preset threshold, the digital twin is updated with the actual tensor, and the residual is backpropagated to the graph neural network for parameter adjustment. At each fusion time step, digital twin data containing the three-dimensional geometry, spatial pose, and percentage of process completion of all components is output.

[0029] In step S3, the real-time construction scene digital twin is registered with the BIM design model, the actual construction progress characteristics of each construction task node are extracted, and the progress deviation of each construction task node is determined based on the actual construction progress characteristics.

[0030] In this embodiment, the digital twin of the real-time construction scene is registered with the BIM design model to extract the actual construction progress features of each construction task node. This can be achieved through the following steps: The temporal variation feature field is extracted from the digital twin of the real-time construction scene. The temporal variation feature field is generated by determining the spatial gradient difference tensor between the digital twin at the current moment and the digital twin at the previous moment. The temporal variation feature field is used as the driving prior for dynamic registration, guiding the registration algorithm to focus on regions with deformation rates greater than preset deformation rates. A first registration sampling point is set in the region and a second registration sampling point is set in the region with a deformation rate of zero, thereby constructing a non-uniform adaptive registration point cloud. Extract the curvature extremum lines and edge topology of the surfaces of each component in the BIM design model, and construct the parametric feature manifold corresponding to each component; The non-uniform adaptive registration point cloud is used to perform manifold mapping matching with the parameterized feature manifolds corresponding to each component, thereby determining the six-degree-of-freedom pose drift of each component region; The non-uniform adaptive registration point cloud is projected onto the surface of the BIM design model corresponding to the parameterized feature manifold of each component to determine the actual surface coverage of each component. Then, the six-degree-of-freedom pose drift and the actual surface coverage are used as the actual construction progress features of each construction task node.

[0031] In specific implementation, firstly, the temporal variation feature field of the real-time construction scene digital twin is extracted. Specifically, the temporal variation feature field is generated by determining the spatial gradient difference tensor between the current digital twin and the previous digital twin. This involves obtaining the spatial position offset fields of the current and previous digital twins along the three coordinate axes of the dynamic adaptive spatiotemporal coordinate system, calculating the partial derivatives of the spatial position offset fields along the three coordinate axes, combining them into a 3x3 Jacobian matrix, and taking the symmetric gradient part of the Jacobian matrix as the spatial gradient difference tensor. The spatial gradient difference tensor is then calculated. The Frobenius norm is used to characterize the comprehensive deformation rate at each spatial location, with units of millimeters per day. Secondly, the temporal variation feature field is used as the driving prior for dynamic registration, guiding the registration algorithm to prioritize regions with deformation rates greater than a preset deformation rate. First registration sampling points are set in these regions, and second registration sampling points are set in regions with zero deformation rates. This constructs a non-uniform adaptive registration point cloud. Specifically, the vertices of the triangular mesh of each component in the BIM design model are used as candidate registration points, and the probability of each candidate registration point being selected into the registration point cloud is calculated based on its comprehensive deformation rate. Candidate registration points with deformation rates greater than the first preset threshold (5 mm / day) have a sampling probability of 0.8 to 1.0; candidate registration points with deformation rates less than the second preset threshold (0.5 mm / day) have a sampling probability of 0.1 to 0.3; and candidate registration points with deformation rates between these two thresholds have their sampling probability determined by linear interpolation. Based on the above sampling probability, all candidate registration points are randomly sampled to generate a non-uniform adaptive registration point cloud, so that the registration sampling point density in the severely deformed area is three to ten times that in the area with zero deformation. The preset threshold can be set by expert experience.

[0032] Furthermore, in practical implementation, the curvature extrema lines and edge topology maps of each component surface in the BIM design model are extracted to construct the parametric feature manifold corresponding to each component. That is, the curvature extrema lines, edge topology maps, and texture feature lines of each component surface are extracted. The curvature extrema lines are used as the main ridge lines of the manifold, the edge topology maps are used as the boundary constraints of the manifold, and the texture feature lines are used as the internal latitude and longitude grid of the manifold. The parametric feature manifold is constructed using a non-uniform rational B-spline surface fitting method. This manifold carries the design dimension parameters and design space of the corresponding component. Pose parameters; then, the non-uniform adaptive registration point cloud is used to perform manifold mapping matching with the parameterized feature manifolds corresponding to each component, thereby determining the six-degree-of-freedom pose drift of each component region, that is, searching for mapping points on the parameterized feature manifold corresponding to each registration point in the non-uniform adaptive registration point cloud, so that the sum of the geodesic distances of all mapping points on the parameterized feature manifold is minimized. The constraints of the objective function include the parameter domain boundary constraints of the mapping points on the parameterized feature manifold and the relative topological constraints between the mapping points determined by the edge topology graph. The solution process introduces Lie group representations of the six-DOF pose drift, transforming the objective function solution into a nonlinear optimization problem on a Lie group. An iterative solution is performed using the Lie algebra local linearization method. In each iteration, a first-order Taylor expansion of the objective function is performed near the current estimate to linearize the nonlinear problem, updating the step size. Iteration stops when the difference between the objective function values ​​of two adjacent iterations is less than a preset convergence threshold (10 to the power of -6). This obtains the mapping relationship from the non-uniform adaptive registration point cloud to the parameterized feature manifold, and then the six-DOF pose drift of each component region is calculated. The six-DOF pose drift includes drifts in three translational degrees of freedom and three rotational degrees of freedom. The drift units for translational degrees of freedom are millimeters, and the drift units for rotational degrees of freedom are degrees.

[0033] In addition, in specific implementation, the non-uniform adaptive registration point cloud is projected onto the surface of the BIM design model corresponding to the parametric feature manifold of each component, and the actual surface coverage of each component is determined. Then, the six-degree-of-freedom pose drift and the actual surface coverage are used as the actual construction progress features of each construction task node. That is, the registration points belonging to the same component in the non-uniform adaptive registration point cloud are orthogonally projected along the normal direction of the BIM design model surface to which the parametric feature manifold of the component is attached. The ratio of the number of projection points falling within the boundary of the component's design surface to the total number of registration points of the component is counted. This ratio is used as the actual surface coverage, with a value range of zero to one. When the actual surface coverage is greater than or equal to the first preset coverage threshold, the construction task node corresponding to the component is determined to be in a completed state; when the actual surface coverage is less than or equal to the second preset coverage threshold, the construction task node corresponding to the component is determined to be in a non-started state; when the actual surface coverage is between the two, the construction task node corresponding to the component is determined to be in progress. The percentage of process completion is calculated by dividing the actual surface coverage by 0.95 and then multiplying by 100%. The obtained six-degree-of-freedom pose drift and the actual surface coverage are used together as the actual construction progress characteristics of each construction task node.

[0034] It should be noted that the preset values ​​in this step can be set through experimental data analysis and expert experience.

[0035] In this embodiment, the progress deviation of each construction task node is determined based on the actual construction progress characteristics. This can be achieved through the following steps: An implicit progress state space is constructed, wherein each dimension of the implicit progress state space corresponds to a construction driving force that cannot be directly observed at the construction site; The actual construction progress features are taken as observable explicit variables. The implicit variables in the implicit progress state space are used as an intermediate bridge to establish a hidden-explicit bidirectional mapping generative adversarial network. The discriminator of the generative adversarial network is used to determine whether the progress features generated under the current combination of implicit variables are consistent with the actual observation. The generator is used to infer the posterior distribution of the implicit variables from the actual observation. The posterior distribution of the latent variables is sampled, and the latent variable combination obtained from each sample is input into a construction progress forward simulator based on physical constraints to obtain the progress simulation results under each sample. The progress simulation results under all samples are statistically analyzed to obtain the progress completion probability cloud of each construction task node, and then the progress deviation of each construction task node is determined.

[0036] In practical implementation, firstly, an implicit progress state space is constructed. Each dimension of this implicit progress state space corresponds to a construction driving force that cannot be directly observed at the construction site. The implicit progress state space is a multi-dimensional continuous vector space, and its number of dimensions equals the total number of preset construction driving force types. In this embodiment, the number of dimensions is set to five, corresponding to the actual installation impedance of components, the effective accessibility coefficient of the work surface, the interference intensity of parallel operations of multiple trades, the attenuation coefficient of the actual working efficiency of construction machinery, and the material supply delay coefficient. These five construction driving forces are all physical quantities that objectively exist at the construction site but cannot be directly observed by sensors. The value of each dimension of the implicit progress state space is normalized to the interval between zero and one, where 0 indicates that the driving force does not hinder the construction progress, and 1 indicates that the driving force has the maximum hindering effect on the construction progress. Initially, each dimension is set to 0.5 by default, representing an unknown state. Then, the actual construction progress features are used as observable explicit variables. Using the latent variables in the implicit progress state space as an intermediary bridge, a bidirectional mapping generative adversarial network (GAN) is established. The discriminator of the GAN is used to determine whether the progress features generated under the current combination of latent variables conform to the actual observations. The generator is used to infer the posterior distribution of the latent variables from the actual observations. That is, the adversarial network consists of two components: a generator and a discriminator. The generator adopts an encoder-decoder architecture. The encoder consists of four fully connected neural networks. The input layer dimension is equal to the total number of dimensions of the actual construction progress features. In this embodiment, it is six values ​​of six degrees of freedom pose drift plus one value of actual surface coverage, for a total of seven dimensions. The first hidden layer dimension is 256, the second hidden layer dimension is 128, and the output layer dimension is five, which is the number of dimensions of the implicit progress state space. The output value is used as the mean vector of the posterior distribution of the latent variables. At the same time, the encoder also outputs a five-dimensional variance vector. The decoder consists of a four-layer fully connected neural network: an input layer with a dimension of five, a first hidden layer with a dimension of 128, a second hidden layer with a dimension of 256, and an output layer with a dimension of seven, outputting a reconstruction progress feature value. The discriminator uses a three-layer multilayer perceptron architecture: an input layer with a dimension of seven, a hidden layer with a dimension of 128, and an output layer with a dimension of one, outputting a discrimination probability value between zero and one.The training samples are the manifest variable sequences and corresponding latent variable labeled data recorded in historical construction projects. The latent variable labeled data is obtained through construction process retrospective analysis. Specifically, the actual installation impedance of components is calculated by extracting jack pressure sensor data and bolt torque data from historical projects during component installation; the effective accessibility coefficient of the work surface is obtained by analyzing the density of construction personnel and the smoothness of movement paths within the work surface through video image analysis; the interference intensity of multi-trade parallel operations is obtained by the frequency of cross-trade conflict events recorded in the construction log; the efficiency attenuation coefficient of construction machinery is obtained by comparing the ratio of machinery fuel consumption to completed work volume with the ratio of rated fuel consumption; and the material supply delay coefficient is obtained by the difference between the actual arrival time and the planned arrival time of materials. The total number of training samples is no less than 10,000. During training, the encoder of the generator maps the input actual progress features to the mean vector and variance vector of the latent variable posterior distribution, samples the latent variables, and reconstructs them into progress features by the decoder. The discriminator receives the reconstructed progress features and the actual progress features and outputs the discrimination result. The loss function consists of three terms: the generative adversarial loss uses binary cross-entropy loss to make the reconstructed features approximate the true feature distribution; the latent variable regularization loss uses KL divergence to make the latent variable distribution of the encoder output approximate the standard normal distribution; and the manifest variable reconstruction loss uses mean squared error loss to minimize the error between the reconstructed features and the input features. The discriminator and generator are trained alternately, with the generator trained after each training iteration of the discriminator, for a total of two thousand training epochs, until the discriminator's discrimination accuracy converges to within the range of 50% ± 2%, meaning the discriminator can no longer effectively distinguish between the reconstructed features and the true features.

[0037] Furthermore, in the specific implementation, the posterior distribution of the latent variables is sampled. Each sampled latent variable combination is input into a physically constrained forward construction schedule simulator to obtain the progress simulation results under each sample. The progress simulation results under all samples are statistically analyzed to obtain the progress completion probability cloud for each construction task node, thereby determining the progress deviation of each construction task node. That is, for the actual construction progress features of each construction task node obtained from the registration step at the current moment, they are input into the encoder of the trained generator to obtain the mean and variance of the posterior distribution of each dimension in the implicit progress state space. Samples are independently drawn from this posterior distribution, with the sampling number set to two thousand. Each sampling yields a five-dimensional latent variable combination, which represents a possible current construction driving force state. Then, the physically constrained forward construction schedule simulator uses the logical dependencies and planned duration parameters in the construction schedule network diagram as fixed constraint boundaries, and the input latent variable combinations as variable driving parameters, to simulate the progress evolution process of each construction task node under given driving force conditions through a discrete event simulation mechanism. The physical constraints of the construction progress forward simulator include at least the following: the actual construction duration of each construction task node is equal to the planned duration multiplied by the weighted influence coefficient of each driving force, and the weighted influence coefficient is calculated according to a preset influence factor matrix; multiple components in the same component combination must be completed before the start of subsequent nodes can be triggered; in the end-start dependency relationship between construction task nodes, the earliest start time of a subsequent node must not be earlier than the latest completion time of all its predecessor nodes. After each sampling, the sampled five-dimensional latent variables are input into the construction progress forward simulator. The construction progress forward simulator advances from the current moment, calculates the estimated completion time of each incomplete construction task node in turn, and outputs the simulated completion time of all construction task nodes.

[0038] Furthermore, in practical implementation, the simulated completion time of each construction task node under two thousand samples is compared with the planned completion time. The proportion of samples where the simulated completion time of each construction task node is less than or equal to the planned completion time out of the total two thousand samples is used as the probability of progress completion for that construction task node. The probability of progress completion for all construction task nodes is arranged in the order of node numbers in the construction schedule network diagram, forming a probability cloud of progress completion. For each construction task node, the arithmetic mean of its simulated completion time under two thousand samples is calculated as the expected completion time. The difference between the expected completion time and the planned completion time is used as the absolute deviation, in days. A positive value indicates lag, a negative value indicates lead, and zero value indicates consistency with the plan. The absolute deviation of all construction task nodes is used as the progress deviation for each construction task node. The progress deviation is used to characterize the degree of difference between the actual progress status and the planned progress status of each construction task node.

[0039] In step S4, based on the progress deviation, delayed risk task nodes are identified, their impact on subsequent construction progress is predicted, progress warning information is generated, and a construction progress adjustment strategy is generated based on the progress warning information.

[0040] In this embodiment, based on the progress deviation, the delayed risk task nodes are identified, which can be achieved through the following steps: The progress deviation of each construction task node is compared with a preset lag judgment threshold, and the construction task node whose progress deviation is greater than the preset lag judgment threshold is marked as a lag candidate node. Obtain the topological location parameters of each lagging candidate node in the construction schedule network diagram. The topological location parameters include at least the out-degree, in-degree, and longest path length from the node to the project endpoint. Based on the progress deviation and topology location parameters of each lagging candidate node, the risk priority score of each lagging candidate node is determined, and the lagging candidate node with a risk priority score greater than a preset risk score threshold is designated as a lagging risk task node.

[0041] In practical implementation, firstly, the schedule deviation of each construction task node is compared with a preset lag threshold. Construction task nodes with schedule deviations greater than the preset lag threshold are marked as lag candidate nodes. That is, the preset lag threshold is first determined by reading the schedule deviations of all construction task nodes in the system database, calculating their arithmetic mean and standard deviation, and using the sum of the mean and 1.5 times the standard deviation as the preset lag threshold. If the calculated threshold is less than 0.5 days, then 0.5 days is used as the preset lag threshold. Then, all construction task nodes are traversed, and the schedule deviation of each construction task node is compared with the above-mentioned preset lag threshold. If the schedule deviation of a construction task node is greater than the preset lag threshold, then the node is marked as a lag candidate node; if the schedule deviation is less than or equal to the preset lag threshold, then it is considered a normal schedule node and does not enter the subsequent risk assessment process. Then, the topological location parameters of each lag candidate node in the construction schedule network diagram are obtained. That is, the topological location parameters include at least the node's out-degree, in-degree, and the longest path length from the node to the project endpoint. Using the directed acyclic graph structure of the construction schedule network diagram as the data source, for each candidate lagging node, the following three topological location parameters are calculated: Out-degree, which is the number of directed edges from this node to its direct successor nodes. In the construction schedule network diagram, a larger out-degree indicates that the lagging of this node will directly affect more subsequent tasks, and its risk propagation range is wider. In-degree, which is the number of directed edges from the direct predecessor nodes to this node. Nodes with a larger in-degree are usually located at the junction of work processes, and their lag may be caused by the cumulative delays of multiple predecessor nodes. Longest path length from the node to the project endpoint, which is calculated by traversing all possible paths to the project endpoint node along the direction of the directed edges, starting from the corresponding candidate lagging node, calculating the sum of the planned durations of all construction task nodes passed along each path, and taking the maximum value among all paths as the longest path length of the node, in days. The longest path length reflects the potential impact of this node on the overall project completion date. The larger the longest path length, the closer the node is to the critical path, and the more significant the impact of its lag on the total project duration.

[0042] Furthermore, in practical implementation, based on the schedule deviation and topological location parameters of each lagging candidate node, a risk priority score is determined for each lagging candidate node. Lagging candidate nodes with a risk priority score greater than a preset risk score threshold are designated as lagging risk task nodes. Specifically, the risk priority score is calculated using a weighted summation formula: the risk priority score equals α multiplied by the normalized schedule deviation plus β multiplied by the normalized out-degree plus γ multiplied by the normalized longest path length. Here, the normalized schedule deviation is the absolute value of the schedule deviation of the corresponding lagging candidate node divided by the maximum absolute value of the schedule deviation among all lagging candidate nodes; the normalized out-degree is the out-degree of the lagging candidate node divided by the maximum out-degree among all lagging candidate nodes; and the normalized longest path length is the longest path length of the lagging candidate node divided by the maximum longest path length among all lagging candidate nodes. The values ​​of the three normalization indicators all fall within the range of zero to one. α, β, and γ are preset weighting coefficients, and their sum equals one. These preset weighting coefficients can be set through experimental data analysis. The preset risk scoring threshold is a pre-defined fixed threshold, determined based on the 75th percentile of the statistical distribution of risk priority scores for at least three historical construction projects' lagging candidate nodes. Finally, the risk priority score of each lagging candidate node is compared with the preset risk scoring threshold. If the risk priority score of a lagging candidate node is greater than the preset risk scoring threshold, the node is identified as a lagging risk task node; if the risk priority score is less than or equal to the preset risk scoring threshold, the node is marked as a low-risk lagging node, recorded only for future reference, and no risk warning is triggered.

[0043] In practical implementation, the impact on subsequent construction progress is predicted, and progress warning information is generated. Based on the progress warning information, a construction progress adjustment strategy is generated. Specifically, the delay propagation path of lagging risk task nodes is determined. Starting from each identified lagging risk task node, all direct and indirect successor nodes are traversed along the directed edges in the construction progress network diagram. All paths from the lagging risk task node to each successor node and the node sequence along each path are recorded. For each lagging risk task node, its progress deviation is used as the initial delay value. For any direct successor node of this node, its expected delay is equal to the initial delay value of the lagging risk task node multiplied by a preset propagation coefficient. The preset propagation coefficient is determined by expert experience based on the logical dependency relationship between the two nodes. For indirect successor nodes, their expected delay is calculated layer by layer according to the product of the propagation coefficients along the path. When a successor node is simultaneously affected by the delays of multiple predecessor nodes, the maximum value of the delay values ​​propagated from each predecessor node is taken as the expected delay of the successor node. Starting from all delayed risk task nodes, the delay is passed layer by layer to the project endpoint according to the propagation rules. The expected delay of the project endpoint is the expected delay of the overall project completion date. Based on the above calculations, progress warning information is generated: for each unworked task node, a yellow warning is generated when its expected delay exceeds one-third of the planned duration of that node; an orange warning is generated when it exceeds one-half; and a red warning is generated when it exceeds three-quarters. For the overall project completion date, a project-level yellow warning is generated when its expected delay exceeds 5% of the total planned project duration; a project-level orange warning is generated when it exceeds 10%; and a project-level red warning is generated when it exceeds 15%. Warning information is pushed according to the warning level. Red warnings are pushed to construction management personnel and supervision units in real time, orange warnings are pushed hourly, and yellow warnings are pushed daily.

[0044] In addition, in practical implementation, based on the early warning information and the current availability of construction resources, one or more combinations of the following adjustment measures are selected, such as: increasing the input of construction resources, including increasing the number of construction personnel and construction machinery shifts; adjusting the operation time of work processes, compressing the planned duration of subsequent work processes, but the compressed duration shall not be less than the preset minimum construction period; adjusting the construction sequence, advancing the start of subsequent work processes under the premise of satisfying logical dependencies, but not earlier than the earliest completion time of all its predecessor nodes. According to the combination of adjustment measures and combined with the current availability of resources, multiple different construction progress adjustment strategy schemes are generated, each scheme specifying the specific implementation objects and implementation parameters of the adjustment measures. For each adjustment scheme, with the current actual progress as the initial state, the progress calculation of the construction progress schedule network diagram is re-executed according to the adjusted parameters to obtain the estimated completion time of each un-constructed task node under the scheme and the overall estimated completion date of the project. The results of each scheme are compared with the case of no adjustment measures to calculate the improvement effect of each scheme on the overall project duration. Based on the simulation results, the scheme with the smallest overall project duration delay is selected as the preferred scheme. If multiple schemes have the same duration improvement effect, the scheme with the lowest adjustment cost is selected. The optimal solution is output to the construction management platform in a structured data format for construction management personnel to execute.

[0045] like Figure 2 The diagram shows a device framework for intelligent construction progress analysis based on BIM. It takes the BIM design model of the target building project and real-time multi-source monitoring data collected from the construction site as dual inputs. After fusion processing, a digital twin of the real-time construction scene is generated. Based on the registration of the digital twin of the real-time construction scene with the BIM design model, the construction progress features are extracted and deviations are analyzed. Finally, the construction progress control instructions are output to the construction management platform, realizing a closed-loop data process for intelligent construction progress analysis.

[0046] Therefore, this application firstly constructs a construction schedule network diagram based on the BIM design model, realizing the unique binding between construction task nodes and BIM components and the automatic extraction and mapping of task attributes, providing a planning benchmark that corresponds to the design model for subsequent progress monitoring; then, by collecting multi-source monitoring data in real time and performing fusion processing to generate a digital twin of the real-time construction scene, a digital mapping of the physical space and geometric shape of the construction site is realized, providing a data foundation that is updated synchronously with the actual construction for subsequent progress feature extraction; furthermore, by registering the digital twin of the real-time construction scene with the BIM design model and extracting actual construction progress features, progress deviation quantification based on the actual construction geometry is realized, so that progress assessment no longer relies on manual reporting, but is directly derived from the real spatial state data of the construction site; finally, by identifying lagging risk task nodes based on the progress deviation, predicting the subsequent progress impact, and generating early warning information and adjustment strategy schemes, a complete control closed loop from lagging identification to forward warning and then to strategy output is realized, improving the refinement level and response timeliness of construction progress management.

[0047] In summary, the technical solution adopted in this application can achieve real-time and accurate perception of construction progress status and advance prediction of delay risks, thereby improving the refinement level and response timeliness of construction progress management.

[0048] Example 2: This application provides a BIM-based intelligent construction progress analysis system, referencing... Figure 3 As shown in the figure, this is a module structure diagram of a BIM-based intelligent construction progress analysis system according to this embodiment of the present application. The intelligent analysis system includes: The network construction module 100 is used to acquire the BIM design model of the target building project and construct a construction schedule network diagram based on the BIM design model. The data fusion module 200 is used to collect multi-source monitoring data from the construction site in real time, perform fusion processing on the multi-source monitoring data, and generate a digital twin of the real-time construction scene. The schedule deviation module 300 is used to register the real-time construction scene digital twin with the BIM design model, extract the actual construction progress characteristics of each construction task node, and determine the schedule deviation of each construction task node based on the actual construction progress characteristics. The scheme generation module 400 is used to identify delayed risk task nodes based on the progress deviation, predict the impact on subsequent construction progress, generate progress early warning information, and generate construction progress adjustment strategy schemes based on the progress early warning information.

[0049] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0050] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compactdisc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0051] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

Claims

1. A BIM-based intelligent analysis method for construction progress, characterized in that, The intelligent analysis method includes the following steps: Obtain the BIM design model of the target building project, and construct a construction schedule network diagram based on the BIM design model; Real-time collection of multi-source monitoring data from the construction site; fusion processing of the multi-source monitoring data; generation of a digital twin of the real-time construction scene. The real-time construction scene digital twin is registered with the BIM design model, the actual construction progress characteristics of each construction task node are extracted, and the progress deviation of each construction task node is determined based on the actual construction progress characteristics. Based on the aforementioned schedule deviation, delayed risk task nodes are identified, their impact on subsequent construction progress is predicted, schedule early warning information is generated, and a construction schedule adjustment strategy is generated based on the schedule early warning information.

2. The BIM-based intelligent construction progress analysis method as described in claim 1, characterized in that, The construction schedule network diagram is generated based on the component hierarchy structure in the BIM design model. Each construction task node in the construction schedule network diagram is uniquely bound to a component or a combination of components in the BIM design model. The task attributes of the construction task node are automatically extracted from the component attribute fields of the BIM design model and mapped to the node parameters in the construction schedule network diagram. The node parameters include at least the planned start time, planned end time, planned duration, and a list of prerequisite dependent components.

3. The intelligent construction progress analysis method based on BIM as described in claim 1, characterized in that, The multi-source monitoring data includes: UAV oblique photogrammetry data, video stream data from ground-fixed panoramic cameras, vehicle-mounted positioning and attitude sensing data of construction machinery, data from wearable positioning terminals of construction personnel, fixed environmental sensing data, and data on the arrival and consumption records of construction materials.

4. The BIM-based intelligent construction progress analysis method as described in claim 1, characterized in that, Registering the real-time construction scene digital twin with the BIM design model and extracting the actual construction progress features of each construction task node specifically includes: The temporal variation feature field is extracted from the digital twin of the real-time construction scene. The temporal variation feature field is generated by determining the spatial gradient difference tensor between the digital twin at the current moment and the digital twin at the previous moment. The temporal variation feature field is used as the driving prior for dynamic registration, guiding the registration algorithm to focus on regions with deformation rates greater than preset deformation rates. A first registration sampling point is set in the region and a second registration sampling point is set in the region with a deformation rate of zero, thereby constructing a non-uniform adaptive registration point cloud. Extract the curvature extremum lines and edge topology of the surfaces of each component in the BIM design model, and construct the parametric feature manifold corresponding to each component; The non-uniform adaptive registration point cloud is used to perform manifold mapping matching with the parameterized feature manifolds corresponding to each component, thereby determining the six-degree-of-freedom pose drift of each component region; The non-uniform adaptive registration point cloud is projected onto the surface of the BIM design model corresponding to the parameterized feature manifold of each component to determine the actual surface coverage of each component. Then, the six-degree-of-freedom pose drift and the actual surface coverage are used as the actual construction progress features of each construction task node.

5. The intelligent construction progress analysis method based on BIM as described in claim 1, characterized in that, Based on the actual construction progress characteristics, the specific progress deviations for each construction task node are determined as follows: An implicit progress state space is constructed, wherein each dimension of the implicit progress state space corresponds to a construction driving force that cannot be directly observed at the construction site; The actual construction progress features are taken as observable explicit variables. The implicit variables in the implicit progress state space are used as an intermediate bridge to establish a hidden-explicit bidirectional mapping generative adversarial network. The discriminator of the generative adversarial network is used to determine whether the progress features generated under the current combination of implicit variables are consistent with the actual observation. The generator is used to infer the posterior distribution of the implicit variables from the actual observation. The posterior distribution of the latent variables is sampled, and the latent variable combination obtained from each sample is input into a construction progress forward simulator based on physical constraints to obtain the progress simulation results under each sample. The progress simulation results under all samples are statistically analyzed to obtain the progress completion probability cloud of each construction task node, and then the progress deviation of each construction task node is determined.

6. The intelligent construction progress analysis method based on BIM as described in claim 5, characterized in that, The construction driving forces corresponding to each dimension in the implicit progress state space include at least: the actual installation impedance of components, the effective accessibility coefficient of the on-site work surface, the interference intensity of parallel operations of multiple trades, the attenuation coefficient of the actual working efficiency of construction machinery, and the material supply delay coefficient.

7. The BIM-based intelligent construction progress analysis method as described in claim 6, characterized in that, The schedule deviation is used to characterize the degree of difference between the actual progress status and the planned progress status of each construction task node.

8. The BIM-based intelligent construction progress analysis method as described in claim 1, characterized in that, Based on the aforementioned schedule deviation, identifying delayed risk task nodes specifically includes: The progress deviation of each construction task node is compared with a preset lag judgment threshold, and the construction task node whose progress deviation is greater than the preset lag judgment threshold is marked as a lag candidate node. Obtain the topological location parameters of each lagging candidate node in the construction schedule network diagram. The topological location parameters include at least the out-degree, in-degree, and longest path length from the node to the project endpoint. Based on the progress deviation and topology location parameters of each lagging candidate node, the risk priority score of each lagging candidate node is determined, and the lagging candidate node with a risk priority score greater than a preset risk score threshold is designated as a lagging risk task node.

9. The intelligent construction progress analysis method based on BIM as described in claim 1, characterized in that, The preset risk scoring threshold is a pre-calibrated fixed threshold, which is determined based on the 75th percentile of the statistical distribution of the risk priority scores of lagging candidate nodes of at least three historical construction projects.

10. A BIM-based intelligent construction progress analysis system, used to execute a BIM-based intelligent construction progress analysis method as described in any one of claims 1 to 9, characterized in that, The intelligent analysis system includes: The network construction module is used to acquire the BIM design model of the target building project and construct a construction schedule network diagram based on the BIM design model. The data fusion module is used to collect multi-source monitoring data from the construction site in real time, fuse the multi-source monitoring data, and generate a digital twin of the real-time construction scene. The schedule deviation module is used to register the digital twin of the real-time construction scene with the BIM design model, extract the actual construction progress characteristics of each construction task node, and determine the schedule deviation of each construction task node based on the actual construction progress characteristics. The scheme generation module is used to identify delayed risk task nodes based on the progress deviation, predict the impact on subsequent construction progress, generate progress early warning information, and generate construction progress adjustment strategy schemes based on the progress early warning information.

Images