A construction holographic monitoring and early warning method and system based on digital model

By constructing a data flow path map of the entire construction process and extracting features from real-time video data, combined with graph convolutional neural networks, the problem of delayed identification of changes in data flow paths during construction was solved, realizing dynamic perception and accurate early warning of the entire construction process, and improving the scientific nature and predictability of construction safety management.

CN123072153APending Publication Date: 2026-11-10HONGYUAN CONSTRUCTION GROUP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611159223.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-01
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing monitoring systems struggle to capture dynamic topological changes in data flow paths during construction and cannot integrate and compare visual features in video streams with backend management data, resulting in delayed location of abnormal processes and difficulty in assessing their chain reaction impact on subsequent processes.

Method used

By acquiring multi-source heterogeneous monitoring data, a data flow path map of the entire construction process is constructed. The data transmission sequence is extracted by combining real-time video data acquisition and spatiotemporal attention mechanism. Abnormal data flow paths are identified by graph convolutional neural networks, and the source and scope of construction disturbance are located based on graph convolutional neural network algorithm.

Benefits of technology

It enables dynamic perception and precise early warning throughout the entire construction process, improving the efficiency and scientific nature of construction safety management. It can keenly identify subtle deviations in data transmission direction and delay, accurately locate disturbance sources and delineate the scope of impact, and realize the transformation from passive response to proactive prevention and control.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123072153A_ABST
    Figure CN123072153A_ABST
Patent Text Reader

Abstract

This invention discloses a construction holographic monitoring and early warning method and system based on a digital model. The method first collects multi-source heterogeneous data from the target construction site, including resource allocation and process sequence, to construct a data flow path map of the entire construction process. It then extracts features from real-time video data using a spatiotemporal attention mechanism to generate the actual data flow path of the current construction scene. The actual path is then compared topologically with the pre-constructed map to identify abnormal data flow paths. A graph convolutional neural network is used to encode the features of the abnormal paths, accurately locating the source of construction disturbance and defining its impact range. Finally, based on the location results and impact data, a tiered early warning is triggered. This invention solves the problems of data silos, delayed anomaly identification, and difficulty in tracing the source in traditional monitoring, achieving dynamic perception, accurate early warning, and proactive risk control throughout the entire construction process, significantly improving the efficiency and scientific nature of construction safety management and decision-making.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for building construction, and in particular to a method and system for holographic monitoring and early warning of construction based on digital models. Background Technology

[0002] With the development of industrialized and digitalized construction, BIM, IoT, and video surveillance technologies have been widely applied to construction site management. However, existing monitoring systems often focus on threshold alarms for single-point data (such as stress exceeding limits or displacement exceeding limits), lacking a holistic understanding of the data flow logic throughout the entire construction process. During construction, each process is not isolated but forms a tightly coupled data flow network through the interaction of personnel, materials, and machinery. Theoretically, normal construction logic should establish a stable data flow path within this network (e.g., data from completed rebar tying should flow to the formwork support process on time).

[0003] Therefore, once a construction process experiences violations, resource shortages, or operational delays, the direction, timing, or content of its data output will inevitably deviate, causing the actual data flow path to deviate from the preset normal pattern. This "path deviation" is a deep-seated manifestation of process anomalies, reflecting systemic risks earlier than changes in a single physical quantity. However, current technology struggles to capture such dynamic topological changes and cannot fuse and compare visual features in video streams with backend management data at the flow path level, resulting in delayed location of abnormal processes and difficulty in assessing their cascading impact on subsequent processes. Therefore, a method capable of holographic comparison and tracing based on data flow paths is urgently needed to achieve accurate early warning of construction risks. Summary of the Invention

[0004] To address at least one of the aforementioned technical problems, this invention proposes a construction holographic monitoring and early warning method and system based on digital models.

[0005] The first aspect of this invention provides a construction holographic monitoring and early warning method based on a digital model, comprising: Acquire multi-source heterogeneous monitoring data of the target construction site, and construct a data flow path map of the entire construction process based on the multi-source heterogeneous monitoring data. The multi-source heterogeneous monitoring data includes resource allocation data and process connection sequence data for each construction process. Real-time video data is collected from the target construction site. The spatiotemporal attention mechanism algorithm is used to extract features of construction process nodes from the real-time video data to obtain the data transmission sequence of the current construction process. The actual data flow path in the current construction scenario is determined based on the data transmission sequence. The actual data flow path is compared with the data flow path map to identify abnormal data flow paths that deviate from the data flow path map. The abnormal data flow path is feature-encoded based on the graph convolutional neural network algorithm to determine the construction disturbance source and the scope of influence corresponding to the abnormal data flow path, and to obtain the disturbance source location result and the scope of influence data. Based on the location results of the disturbance source and the data on the scope of influence, a construction early warning operation is performed on the target construction site.

[0006] In this solution, the acquisition of multi-source heterogeneous monitoring data from the target construction site, and the construction of a data flow path map of the entire construction process based on the multi-source heterogeneous monitoring data, wherein the multi-source heterogeneous monitoring data includes resource allocation data, process connection sequence data, and physical field monitoring data for each construction procedure, specifically: The resource allocation data for each construction process is obtained by retrieving the number of personnel involved, the mechanical equipment scheduling records, and the material consumption ledgers from the process management system of the target construction site. Extract the interaction and occurrence time of adjacent processes, the duration of each process, and the dependency relationship identifiers between processes from the construction progress control platform to obtain process connection sequence data; Resource allocation data and sequence data of each construction process are associated and matched according to the unique identifier of the construction process, with the construction process as the node and the data interaction relationship between the processes as the directed edge; The weights of the directed edges are set to the information transmission delay time in the corresponding process connection time sequence data, and the attribute values ​​of the nodes are configured to the resource configuration data of the corresponding construction process. All nodes are connected to the directed edges in sequence according to the process advancement order of the entire construction process to construct the data flow path graph of the entire construction process.

[0007] In this solution, the real-time video data acquisition of the target construction site is performed, and the spatiotemporal attention mechanism algorithm is used to extract features of construction process nodes from the real-time video data to obtain the data transmission sequence of the current construction process. Based on the data transmission sequence, the actual data flow path in the current construction scenario is determined, specifically as follows: By collecting real-time video data streams covering the entire construction site using multi-view high-definition camera equipment deployed at the target construction site, the real-time video data streams are divided into continuous video frame sequences at a fixed frame rate. Feature point detection and matching operations are performed frame by frame on the continuous video frame sequences to extract the operating posture features of construction machinery, the movement trajectory features of personnel, and the transfer position features of materials in each video frame. The extracted features are used to construct a spatiotemporal feature tensor according to the time sequence of video frames. The spatiotemporal feature tensor is then aligned and mapped with the node attributes in the data flow path graph using the unique identifier of the construction process as an index. This results in the construction of a holographic digital model that includes the nodes of the entire construction process and the connection relationships between the processes. The spatiotemporal feature tensor is input into the spatiotemporal attention mechanism module of the construction holographic digital model. The spatiotemporal attention mechanism module sequentially calculates the spatial attention weight of different construction process nodes in the same video frame and the temporal attention weight of the same construction process node in different video frames. Based on the spatial attention weight, the process node features in the same frame are weighted and fused to obtain spatial aggregated features. The features of the same process node across frames are weighted and fused according to the temporal attention weight to obtain the temporal aggregated features. The spatial aggregated features and the temporal aggregated features are concatenated and input into the feature decoding layer. The data activation intensity of each construction process node in the current construction scenario is output. The construction process nodes are sorted in order of data activation intensity from high to low. The data transmission direction of each construction process node is concatenated according to the sorting result to obtain the data transmission sequence of the current construction process. By taking adjacent data transmission directions in the data transmission sequence as directed edges and the construction process nodes corresponding to the data transmission directions as endpoints, the topology of directed edges is constructed sequentially according to the connection relationship of the data transmission directions to obtain the actual data flow path in the current construction scenario.

[0008] In this solution, the step of performing a topological comparison between the actual data flow path and the data flow path map to identify abnormal data flow paths that deviate from the data flow path map specifically involves: The directed edges of the actual data flow path are matched one by one with the directed edges of the corresponding process connection relationship in the data flow path graph. The data transmission direction of each adjacent process node in the actual data flow path is extracted and compared with the direction of the directed edge of the corresponding process connection relationship in the data flow path graph. Adjacent process nodes with inconsistent directions are marked as direction deviation edges. The data transmission delay of each adjacent process node in the actual data flow path is extracted and compared with the weight of the corresponding directed edge in the data flow path graph. The directed edge of the adjacent process node whose delay exceeds the set range of the corresponding weight is marked as a delay deviation edge. The abnormal data flow path in the actual construction process is determined based on the directional deviation edge and the delay deviation edge.

[0009] In this solution, the graph convolutional neural network algorithm is used to perform feature encoding on the abnormal data flow path to determine the construction disturbance source and its impact range corresponding to the abnormal data flow path, thereby obtaining the disturbance source location result and impact range data. Specifically: Align and map the process nodes in the abnormal data flow path with the corresponding process nodes in the data flow path graph. Use the resource configuration data of each process node in the abnormal data flow path as the initial feature vector of the node, and use the deviation type of the directional deviation edge and the delay deviation edge in the abnormal data flow path as the edge feature vector to construct the abnormal data flow path graph structure data containing node features and edge features. The abnormal data flow path graph structure data is input into a pre-constructed graph convolutional neural network. The neighborhood features of each process node are aggregated layer by layer through multi-layer graph convolution operations. In each layer of graph convolution operation, the feature difference between the central process node and the adjacent process nodes is first calculated by weighting the edge feature vectors. The feature difference is concatenated with the current feature vector of the central process node and then mapped through a non-linear activation function to obtain the updated node feature vector. After layer-by-layer iteration, the high-order semantic feature vector of each process node is output. Calculate the cosine similarity between the high-order semantic feature vector of each process node and the baseline feature vector of the corresponding process node under normal working conditions. Mark process nodes with cosine similarity below a set threshold as candidate disturbance nodes. Statistically analyze the connected subgraphs of all candidate disturbance nodes in the abnormal data flow path graph structure data. The core candidate disturbance nodes in the connected subgraph are used as construction disturbance sources, and the spatial distribution range of all process nodes in the connected subgraph is extracted as the initial influence range. Based on the dependency relationship identifiers between processes in the data flow path map, the dependent process nodes are traversed layer by layer along the downstream dependency direction of each process node within the initial influence range. The dependent process nodes obtained from the traversal are included in the influence range. The spatial coordinates of the construction disturbance source are associated and encapsulated with the spatial boundary of the influence range to obtain the disturbance source location result and the influence range data.

[0010] In this solution, the step of performing a construction early warning operation on the target construction site based on the disturbance source location results and the impact range data specifically includes: Based on the disturbance source location results, the construction process nodes corresponding to the construction disturbance source are determined. Based on the influence range data, all construction process nodes within the influence range are extracted. Combined with the dependency relationships between each construction process node in the data flow path map, a disturbance propagation association chain is constructed, and the cumulative propagation influence value of each construction process node in the disturbance propagation association chain is calculated. The cumulative propagation impact value of each construction process node is correlated with the data activation intensity of the corresponding construction process node to obtain the comprehensive disturbance index of each construction process node. When the comprehensive disturbance index is greater than the preset disturbance index, the corresponding construction process node is marked as a key early warning node. When the comprehensive disturbance index is less than or equal to the preset disturbance index, the corresponding construction process node is marked as a continuous monitoring node. The proportion of key early warning nodes within the affected area is statistically analyzed, and the coverage ratio of the abnormal data flow path corresponding to the key early warning nodes is calculated. When the proportion of key early warning nodes is greater than the node proportion threshold and the coverage ratio of the abnormal data flow path is greater than the coverage ratio threshold, a construction interruption early warning instruction is generated and sent to the digital model. The corresponding construction process nodes within the affected area are marked as suspended, and the disturbance source location results and affected area data are output simultaneously. When the proportion of key early warning nodes is less than or equal to the node proportion threshold and the coverage ratio of abnormal data flow path is less than or equal to the coverage ratio threshold, a construction adjustment early warning instruction is generated, and the resource allocation order and process connection order of the construction process corresponding to the key early warning node are adjusted according to the disturbance propagation association chain. The construction interruption warning command or construction adjustment warning command is associated with the unique identifier of the corresponding construction process node, and the warning status identifier of the corresponding construction process node in the digital model is updated simultaneously to complete the construction warning operation of the target construction site.

[0011] A second aspect of the present invention also provides a construction holographic monitoring and early warning system based on a digital model. The system includes a memory and a processor. The memory includes a construction holographic monitoring and early warning method program based on a digital model. When the processor executes the construction holographic monitoring and early warning method program based on a digital model, it implements the steps of the construction holographic monitoring and early warning method based on a digital model as described in any of the above claims.

[0012] This invention discloses a construction holographic monitoring and early warning method and system based on a digital model. The method first collects multi-source heterogeneous data from the target construction site, including resource allocation and process sequence, to construct a data flow path map of the entire construction process. It then extracts features from real-time video data using a spatiotemporal attention mechanism to generate the actual data flow path of the current construction scene. The actual path is then compared topologically with the pre-constructed map to identify abnormal data flow paths. A graph convolutional neural network is used to encode the features of the abnormal paths, accurately locating the source of construction disturbance and defining its impact range. Finally, based on the location results and impact data, a tiered early warning is triggered. This invention solves the problems of data silos, delayed anomaly identification, and difficulty in tracing the source in traditional monitoring, achieving dynamic perception, accurate early warning, and proactive risk control throughout the entire construction process, significantly improving the efficiency and scientific nature of construction safety management and decision-making. Attached Figure Description

[0013] Figure 1 A flowchart of a construction holographic monitoring and early warning method based on a digital model according to the present invention is shown; Figure 2 This invention illustrates a flowchart of the data flow path map for constructing the entire construction process. Figure 3This invention illustrates a flowchart of how the present invention determines the actual data flow path in the current construction scenario; Figure 4 A block diagram of a construction holographic monitoring and early warning system based on a digital model according to the present invention is shown. Detailed Implementation

[0014] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0015] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0016] Figure 1 The flowchart of a construction holographic monitoring and early warning method based on a digital model according to the present invention is shown.

[0017] like Figure 1 As shown, the first aspect of the present invention provides a construction holographic monitoring and early warning method based on a digital model, comprising: Acquire multi-source heterogeneous monitoring data of the target construction site, and construct a data flow path map of the entire construction process based on the multi-source heterogeneous monitoring data. The multi-source heterogeneous monitoring data includes resource allocation data and process connection sequence data for each construction process. Real-time video data is collected from the target construction site. The spatiotemporal attention mechanism algorithm is used to extract features of construction process nodes from the real-time video data to obtain the data transmission sequence of the current construction process. The actual data flow path in the current construction scenario is determined based on the data transmission sequence. The actual data flow path is compared with the data flow path map to identify abnormal data flow paths that deviate from the data flow path map. The abnormal data flow path is feature-encoded based on the graph convolutional neural network algorithm to determine the construction disturbance source and the scope of influence corresponding to the abnormal data flow path, and to obtain the disturbance source location result and the scope of influence data. Based on the location results of the disturbance source and the data on the scope of influence, a construction early warning operation is performed on the target construction site.

[0018] It should be noted that by integrating resource allocation and timing data of construction procedures to construct a full-process data flow path map, discrete management information is transformed into a visualized topological network, providing a standardized data flow benchmark for the construction process. A spatiotemporal attention mechanism is used to extract features from real-time video data, capturing dynamic changes in the operating posture of construction machinery, personnel movement trajectories, and material transfer locations, generating a data transmission sequence that highly matches the actual construction process, solving the problem of the disconnect between visual features and business logic. By comparing the actual data flow path with the pre-constructed map topologically, subtle deviations in data transmission direction and delay can be accurately identified, locking abnormal flow paths from a global topological perspective, overcoming the limitations of single-point monitoring. Based on graph convolutional neural networks, abnormal paths are feature-encoded, fully exploring the deep dependencies between nodes. This not only accurately locates the source procedure causing the disturbance but also scientifically delineates the scope of impact based on the procedure dependency chain, avoiding errors from experience-based judgment. Finally, a tiered early warning system is implemented based on the location of the disturbance source and the scope of impact, adopting procedures to block or adjust resources for different levels of risk, achieving a shift from passive response to proactive prevention and control, significantly improving the predictability and scientific nature of construction safety management.

[0019] Figure 2 A flowchart illustrating the data flow path map of the entire construction process constructed according to the present invention is shown.

[0020] According to an embodiment of the present invention, the acquisition of multi-source heterogeneous monitoring data of the target construction site, and the construction of a data flow path map of the entire construction process based on the multi-source heterogeneous monitoring data, wherein the multi-source heterogeneous monitoring data includes resource allocation data, process connection timing data, and physical field monitoring data for each construction process, specifically: The resource allocation data for each construction process is obtained by retrieving the number of personnel involved, the mechanical equipment scheduling records, and the material consumption ledgers from the process management system of the target construction site. Extract the interaction and occurrence time of adjacent processes, the duration of each process, and the dependency relationship identifiers between processes from the construction progress control platform to obtain process connection sequence data; Resource allocation data and sequence data of each construction process are associated and matched according to the unique identifier of the construction process, with the construction process as the node and the data interaction relationship between the processes as the directed edge; The weights of the directed edges are set to the information transmission delay time in the corresponding process connection time sequence data, and the attribute values ​​of the nodes are configured to the resource configuration data of the corresponding construction process. All nodes are connected to the directed edges in sequence according to the process advancement order of the entire construction process to construct the data flow path graph of the entire construction process.

[0021] It should be noted that by unifying and associating resource allocation information and process connection information that were originally scattered across the construction site in the process management system and progress control platform, the status of resources such as personnel, machinery, and materials, and the information transmission relationships between processes are mapped to the same topological space. This transforms the construction process from traditional linear progress management to a dynamic network expression oriented towards data flow relationships. By using process nodes to carry resource allocation status, directed edges to represent data transmission relationships between processes, and information transmission delay time to quantify the association strength between nodes, the constructed data flow path graph can not only reflect the spatial organization relationships in the construction process but also characterize the data interaction efficiency between different processes. This provides a unified data benchmark for identifying deviations between the actual construction status and the standard construction process, avoiding the problem of delayed detection of construction anomalies caused by traditional methods that rely solely on a single progress node or manual inspection. The dependencies include the sequential execution dependencies between construction processes, resource supply dependencies, spatial operation dependencies, equipment coordination dependencies, and data transmission dependencies.

[0022] Figure 3 The flowchart illustrating how the present invention determines the actual data flow path in the current construction scenario is shown.

[0023] According to an embodiment of the present invention, the step of real-time video data acquisition of the target construction site, feature extraction of construction process nodes from the real-time video data using a spatiotemporal attention mechanism algorithm to obtain the data transmission sequence of the current construction process, and determination of the actual data flow path in the current construction scenario based on the data transmission sequence, specifically includes: By collecting real-time video data streams covering the entire construction site using multi-view high-definition camera equipment deployed at the target construction site, the real-time video data streams are divided into continuous video frame sequences at a fixed frame rate. Feature point detection and matching operations are performed frame by frame on the continuous video frame sequences to extract the operating posture features of construction machinery, the movement trajectory features of personnel, and the transfer position features of materials in each video frame. The extracted features are used to construct a spatiotemporal feature tensor according to the time sequence of video frames. The spatiotemporal feature tensor is then aligned and mapped with the node attributes in the data flow path graph using the unique identifier of the construction process as an index. This results in the construction of a holographic digital model that includes the nodes of the entire construction process and the connection relationships between the processes. The spatiotemporal feature tensor is input into the spatiotemporal attention mechanism module of the construction holographic digital model. The spatiotemporal attention mechanism module sequentially calculates the spatial attention weight of different construction process nodes in the same video frame and the temporal attention weight of the same construction process node in different video frames. Based on the spatial attention weight, the process node features in the same frame are weighted and fused to obtain spatial aggregated features. The features of the same process node across frames are weighted and fused according to the temporal attention weight to obtain the temporal aggregated features. The spatial aggregated features and the temporal aggregated features are concatenated and input into the feature decoding layer. The data activation intensity of each construction process node in the current construction scenario is output. The construction process nodes are sorted in order of data activation intensity from high to low. The data transmission direction of each construction process node is concatenated according to the sorting result to obtain the data transmission sequence of the current construction process. By taking adjacent data transmission directions in the data transmission sequence as directed edges and the construction process nodes corresponding to the data transmission directions as endpoints, the topology of directed edges is constructed sequentially according to the connection relationship of the data transmission directions to obtain the actual data flow path in the current construction scenario.

[0024] It should be noted that the execution of different construction procedures on a construction site is not isolated, but rather a dynamic process with sequential relationships and resource flow characteristics formed by the spatiotemporal interaction between personnel, machinery, and materials. Changes in the working posture of construction machinery can reflect the current procedure status of the equipment; personnel movement trajectories can characterize the participation and transfer relationships of workers between different procedure nodes; and changes in the material transfer location can reflect the flow direction of construction resources between procedures. Therefore, by extracting spatiotemporal features from continuous video frames, visual changes that originally had no business meaning can be transformed into temporal features that can characterize the state changes of construction nodes. Furthermore, by associating and mapping the above spatiotemporal features with construction procedure nodes in the data flow path map, the video perception results at each moment correspond to specific construction procedure nodes. Then, a spatiotemporal attention mechanism is used to capture the spatial correlation between different procedure nodes at the same moment and the temporal continuity of the state changes of the same procedure node at different moments. Spatial attention enhances the features of nodes with collaborative relationships within the current construction area, while temporal attention strengthens the continuous node change features that conform to the construction progress pattern, thereby calculating the data activation intensity of each construction procedure node in the current construction scenario. Since data activation intensity reflects the degree of data interaction actually participated in the construction process by a construction process node, a higher activation intensity indicates that the process node is undertaking more resource interaction and process connection functions. Therefore, sorting by data activation intensity can restore the actual flow sequence of each process node under the current construction state, and further construct the actual data flow path based on the transmission direction between nodes, realizing the transformation from visually perceived information to the construction process data transmission sequence. The spatial aggregation feature refers to the comprehensive feature reflecting the current construction spatial organization state obtained by fusing the positional relationship, collaborative operation relationship, and resource interaction relationship between different construction process nodes at the same moment through a spatial attention mechanism. The temporal aggregation feature refers to the comprehensive feature reflecting the dynamic evolution state of the construction process obtained by fusing the state change pattern and process advancement trend of the same construction process node under different time frames through a temporal attention mechanism.

[0025] According to an embodiment of the present invention, the step of performing a topological comparison between the actual data flow path and the data flow path map to identify abnormal data flow paths that deviate from the data flow path map specifically includes: The directed edges of the actual data flow path are matched one by one with the directed edges of the corresponding process connection relationship in the data flow path graph. The data transmission direction of each adjacent process node in the actual data flow path is extracted and compared with the direction of the directed edge of the corresponding process connection relationship in the data flow path graph. Adjacent process nodes with inconsistent directions are marked as direction deviation edges. The data transmission delay of each adjacent process node in the actual data flow path is extracted and compared with the weight of the corresponding directed edge in the data flow path graph. The directed edge of the adjacent process node whose delay exceeds the set range of the corresponding weight is marked as a delay deviation edge. The abnormal data flow path in the actual construction process is determined based on the directional deviation edge and the delay deviation edge.

[0026] It should be noted that directional deviation edges are used to characterize errors in the direction of data transmission between construction processes. This means that during actual construction, the flow of information, resources, or tasks between one process node and another deviates from the dependencies in the preset construction process. For example, in a normal construction process, after the rebar binding is completed, the acceptance data should be transmitted to the concrete pouring process node, which will then initiate the pouring operation based on the acceptance results. However, if the actual monitoring shows that the data flow is transmitted from the concrete pouring node back to the rebar binding node, it indicates that the sequence of processes has become abnormal, forming a directional deviation edge. Delay deviation edges are used to characterize that the data transmission time between construction processes exceeds the normal construction cycle time. This means that although the data interaction direction between processes remains consistent, the information transmission efficiency is reduced or there is an abnormal lag. For example, under normal circumstances, after the formwork is installed, the installation completion status should be transmitted to the rebar acceptance node within a preset time to trigger the subsequent acceptance process. However, in actual monitoring, the formwork installation status has not been synchronized to the acceptance node for a long time, causing the information transmission delay between the two processes to exceed the normal delay threshold in the data flow path graph, thus forming a delay deviation edge. Therefore, the directional deviation reflects whether the "transmission relationship is correct" in the construction process, while the delay deviation reflects whether the "transmission process is timely" in the construction process. Together, they describe the deviation of the actual construction data flow path from the standard construction path in terms of topology and time rhythm.

[0027] According to an embodiment of the present invention, the step of using a graph convolutional neural network algorithm to perform feature encoding on the abnormal data flow path, determining the construction disturbance source and the scope of influence corresponding to the abnormal data flow path, and obtaining the disturbance source location result and the scope of influence data, specifically involves: Align and map the process nodes in the abnormal data flow path with the corresponding process nodes in the data flow path graph. Use the resource configuration data of each process node in the abnormal data flow path as the initial feature vector of the node, and use the deviation type of the directional deviation edge and the delay deviation edge in the abnormal data flow path as the edge feature vector to construct the abnormal data flow path graph structure data containing node features and edge features. The abnormal data flow path graph structure data is input into a pre-constructed graph convolutional neural network. The neighborhood features of each process node are aggregated layer by layer through multi-layer graph convolution operations. In each layer of graph convolution operation, the feature difference between the central process node and the adjacent process nodes is first calculated by weighting the edge feature vectors. The feature difference is concatenated with the current feature vector of the central process node and then mapped through a non-linear activation function to obtain the updated node feature vector. After layer-by-layer iteration, the high-order semantic feature vector of each process node is output. It should be noted that the graph convolutional neural network represents the abnormal data flow path as a graph structure, where each process node has initial features such as corresponding resource configuration status and abnormal deviation type. In each convolutional layer, the graph convolutional network takes a central process node as the computation object, obtains the neighboring process nodes that have data flow relationships with it, adjusts the influence of the neighboring node features on the central node through the edge feature weights corresponding to the directional deviation edge and the delay deviation edge, and calculates the feature differences between the neighboring nodes and the central node to extract the difference information between the node and the surrounding nodes in terms of resource status, process connection, and abnormal propagation relationship. Then, this difference information is fused with the central process node's own features, and an updated node feature is generated through nonlinear mapping. As the number of graph convolutional layers increases, the central process node can gradually aggregate the information of related nodes in a larger range, so that the low-order features that originally only represent the status of a single process are gradually transformed into high-order semantic feature vectors that simultaneously contain their own abnormality degree, the influence relationship of adjacent processes, the direction of abnormal propagation, and the process dependency relationship. A central process node refers to a construction process node that is the object of feature update during the current graph convolution calculation. In other words, it is a specific construction stage currently being analyzed within the abnormal data flow path graph structure. For example, when analyzing the "concrete pouring" process node, this node is the central process node. A higher-order semantic feature vector is a multi-dimensional numerical representation formed after multi-layer graph convolution feature aggregation. This vector comprehensively describes the state of the construction process node and its role in the anomaly propagation network. It no longer only contains basic information such as the amount of resources invested in a single process and the status of equipment, but also integrates deep correlation information such as the node's own state, the degree of anomaly of adjacent processes, process dependencies, and anomaly propagation paths. This information is used to determine whether the process node deviates from the normal construction state and whether it may become a source of construction disturbance.

[0028] Calculate the cosine similarity between the high-order semantic feature vector of each process node and the baseline feature vector of the corresponding process node under normal working conditions. Mark process nodes with cosine similarity below a set threshold as candidate disturbance nodes. Statistically analyze the connected subgraphs of all candidate disturbance nodes in the abnormal data flow path graph structure data. The core candidate disturbance nodes in the connected subgraph are used as construction disturbance sources, and the spatial distribution range of all process nodes in the connected subgraph is extracted as the initial influence range. Based on the dependency relationship identifiers between processes in the data flow path map, the dependent process nodes are traversed layer by layer along the downstream dependency direction of each process node within the initial influence range. The dependent process nodes obtained from the traversal are included in the influence range. The spatial coordinates of the construction disturbance source are associated and encapsulated with the spatial boundary of the influence range to obtain the disturbance source location result and the influence range data.

[0029] It should be noted that by mapping the process nodes in the abnormal data flow path to the corresponding nodes in the standard data flow path graph, and using the process resource configuration status as node features and the directional deviation edges and delay deviation edges as edge features, the abnormal construction process is transformed into graph structure data that simultaneously contains "node state differences" and "node association anomalies". Among them, resource configuration anomalies reflect changes in the execution capability of the process itself, while directional deviations and delay deviations reflect imbalances in the collaborative relationship between processes. Furthermore, a graph convolutional neural network is used to propagate features of this graph structure, so that each process node not only retains its own resource status information, but also integrates the abnormal impact information of adjacent process nodes, thereby obtaining high-order semantic features that can characterize the degree of local anomalies of nodes and their role in the overall construction process. Since true construction disturbances typically manifest as deviations from normal operating conditions and can trigger continuous anomalies in surrounding nodes through process relationships, candidate disturbance nodes deviating from normal construction states can be screened by calculating the similarity between the high-order semantic features of each process node and the baseline features of normal operating conditions. Then, an anomaly association subgraph is formed using the connectivity relationships between these candidate disturbance nodes. The core node, located at the core of the anomaly propagation chain and capable of connecting multiple anomaly nodes, is identified as the construction disturbance source, rather than simply classifying all anomaly nodes as risk points. This allows for the tracing and identification of the anomaly's origin. Finally, downstream propagation analysis is performed along the dependencies between construction processes—including execution dependencies, resource supply dependencies, spatial operation dependencies, equipment coordination dependencies, and data transmission dependencies—to expand the affected process nodes layer by layer, obtaining the potential construction scope impacted by the disturbance. This achieves the transformation from "anomaly detection" to "disturbance source location" and "impact range prediction." Normal operating conditions refer to the standard operating state in which the resource allocation, data transmission direction, sequence of process connections and dependencies of each construction process node are within a stable range when the construction site is running according to the preset construction organization plan. Core candidate disturbance nodes refer to candidate process nodes that are associated with multiple abnormal nodes in the abnormal data flow path, have the largest deviation of high-order semantic features from normal operating conditions, and play a major role in the propagation of abnormalities.

[0030] According to an embodiment of the present invention, the construction early warning operation performed on the target construction site based on the disturbance source location result and the influence range data specifically includes: Based on the disturbance source location results, the construction process nodes corresponding to the construction disturbance source are determined. Based on the influence range data, all construction process nodes within the influence range are extracted. Combined with the dependency relationships between each construction process node in the data flow path map, a disturbance propagation association chain is constructed, and the cumulative propagation influence value of each construction process node in the disturbance propagation association chain is calculated. The cumulative propagation impact value of each construction process node is correlated with the data activation intensity of the corresponding construction process node to obtain the comprehensive disturbance index of each construction process node. When the comprehensive disturbance index is greater than the preset disturbance index, the corresponding construction process node is marked as a key early warning node. When the comprehensive disturbance index is less than or equal to the preset disturbance index, the corresponding construction process node is marked as a continuous monitoring node. The proportion of key early warning nodes within the affected area is statistically analyzed, and the coverage ratio of the abnormal data flow path corresponding to the key early warning nodes is calculated. When the proportion of key early warning nodes is greater than the node proportion threshold and the coverage ratio of the abnormal data flow path is greater than the coverage ratio threshold, a construction interruption early warning instruction is generated and sent to the digital model. The corresponding construction process nodes within the affected area are marked as suspended, and the disturbance source location results and affected area data are output simultaneously. When the proportion of key early warning nodes is less than or equal to the node proportion threshold and the coverage ratio of abnormal data flow path is less than or equal to the coverage ratio threshold, a construction adjustment early warning instruction is generated, and the resource allocation order and process connection order of the construction process corresponding to the key early warning node are adjusted according to the disturbance propagation association chain. The construction interruption warning command or construction adjustment warning command is associated with the unique identifier of the corresponding construction process node, and the warning status identifier of the corresponding construction process node in the digital model is updated simultaneously to complete the construction warning operation of the target construction site.

[0031] It should be noted that by combining the disturbance source location results and the impact range data, a disturbance propagation correlation chain is constructed along the dependencies between construction procedures, and the cumulative propagation impact value of each procedure node is calculated. This can quantify the degree of impact of construction disturbances spreading from the source to downstream procedures, avoiding the problem of inaccurate risk range judgment caused by relying solely on the status of a single abnormal node for early warning. Furthermore, by fusing the cumulative propagation impact value with the data activation intensity of construction procedure nodes to obtain a comprehensive disturbance index, key nodes in high-frequency data interaction areas and those significantly affected by disturbances are prioritized as key early warning nodes, achieving precise locking of potential risk core areas. When the proportion of key warning nodes exceeds the node proportion threshold and the coverage ratio of abnormal data flow paths exceeds the coverage ratio threshold, it indicates that the current abnormal data flow path has covered a large area of ​​the construction process, and most key process nodes are affected by the disturbance. This suggests that the anomaly has evolved from a localized process anomaly into a systemic construction risk with a spreading trend. At this time, generating a construction interruption warning command and marking the relevant processes as paused can promptly cut off the anomaly propagation chain and prevent the disturbance from further affecting subsequent construction stages. When the proportion of key warning nodes is less than or equal to the node proportion threshold and the coverage ratio of abnormal data flow paths is less than or equal to the coverage ratio threshold, it indicates that the anomaly is mainly... When anomalies are concentrated at a few construction nodes and do not spread widely, representing only localized construction disturbances, generating construction adjustment early warning commands and optimizing resource allocation and process sequence based on the disturbance propagation chain can eliminate the impact of localized anomalies without interrupting the overall construction process, thus improving construction continuity. Finally, by associating early warning commands with unique identifiers of construction process nodes and synchronously updating the early warning status identifiers in the digital model, real-time synchronization between the risk status at the construction site and the digital model is achieved. This transforms construction management from passive handling after anomalies occur to proactive intervention based on risk propagation trends, improving the accuracy of construction safety monitoring and the effectiveness of early warning decisions. The coverage ratio of the abnormal data flow path refers to the proportion of construction process nodes included in the abnormal data flow path to the total number of construction process nodes within the target impact area, used to characterize the current anomaly's spread range and path occupancy degree in the construction process.

[0032] According to an embodiment of the present invention, it further includes: Before extracting features of construction process nodes from the real-time video data using the spatiotemporal attention mechanism algorithm, the real-time video data stream is divided into spatial regions by spatial region masking based on the spatial distribution coordinates of each construction process node in the data flow path map, thereby obtaining process interest regions corresponding to each construction process node. The process interest regions are used to limit the range of visual feature extraction for each construction process node. When the process interest region is used as a spatial prior constraint and input into the spatial attention weight calculation process of the spatiotemporal attention mechanism algorithm, the spatial attention weight is allocated to prioritize the construction machinery operation posture features and personnel movement trajectory features within the corresponding process interest region, and suppress cross-work type visual feature responses within non-corresponding process interest regions. Based on the spatial attention weights adjusted according to the spatial prior constraints, the features of process nodes within the same frame are weighted and fused to obtain spatial aggregate features. Then, based on the spatial aggregate features, the data transmission sequence of the current construction process is determined, effectively eliminating visual feature crosstalk caused by cross-operations of multiple trades.

[0033] According to an embodiment of the present invention, the step of dividing the real-time video data stream into spatial regions using a mask based on the spatial distribution coordinates of each construction process node in the data flow path map to obtain a process interest region corresponding to each construction process node specifically involves: Based on the unique identifier of the construction process in the data flow path map, the theoretical operation coordinate range of each construction process node in the construction plan layout is retrieved, and the theoretical operation coordinate range is mapped to the pixel coordinate system of the real-time video data stream to generate the initial process interest region. Extract the semantic segmentation results from the real-time video data stream, identify temporary material piles and mechanical equipment obstructions in the interest areas of each process in the current construction scene, and dynamically trim the initial process interest area according to the outline boundary of the obstruction to obtain an optimized process interest area after deducting the physical obstruction area. The optimized process interest region is used as the dynamic spatial attribute of the corresponding construction process node and updated to the node attribute value of the data flow path map. This allows the subsequent spatiotemporal attention mechanism algorithm to use the dynamic spatial attribute as a benchmark to define the effective spatial boundary of feature extraction when calculating the spatial attention weight of different construction process nodes in the same video frame, thus avoiding misassociation of visual features of cross-work operations in overlapping areas.

[0034] It should be noted that in construction sites where multiple trades operate in parallel, such as high-rise steel structure hoisting, civil engineering teams and steel structure teams often carry out overlapping operations such as tying rebar and welding steel columns in the same core tube area. The simultaneous operation of multiple tower cranes leads to a high degree of spatial overlap and frequent temporal intersection of construction actions of different processes in the video monitoring screen. Traditional spatiotemporal attention mechanisms, due to the lack of spatial prior constraints, are prone to misassociating visual features of non-target process areas with the current process node, resulting in abnormal back-and-forth jumps in the data transmission sequence of "rebar tying, steel column welding, rebar tying," thereby generating a large number of invalid and abnormal data flow paths. This seriously interferes with the efficiency of managers in judging the actual construction status. However, this invention addresses this by using the spatial distribution coordinates of each construction process node in the data flow path map to analyze real-time video data. The algorithm performs spatial region masking to generate process interest regions (POGs) bound to unique process identifiers. These POGs are then used as spatial prior constraints in the calculation of spatiotemporal attention weights, forcing the algorithm to prioritize the mechanical operation postures and personnel trajectory features within the corresponding POGs. Simultaneously, it dynamically identifies and trims interference factors such as temporary material stacks and equipment obstructions through semantic segmentation results to optimize the boundaries of the POGs. The optimized POG attributes are then fed back to the data flow path map to update the node features. This achieves physical isolation of cross-job visual signals at the source of feature extraction, effectively eliminating crosstalk between multi-source features on overlapping construction surfaces. It significantly improves the recognition accuracy and stability of the actual data flow path topology in complex cross-operation scenarios, avoiding frequent false alarms and ineffective management interventions caused by misjudgment of visual features.

[0035] Figure 4 A block diagram of a construction holographic monitoring and early warning system based on a digital model according to the present invention is shown.

[0036] A second aspect of the present invention also provides a construction holographic monitoring and early warning system based on a digital model. The system includes a memory 401, a processor 402, and a communication interface 403. The memory includes a construction holographic monitoring and early warning method program based on a digital model. The communication interface is used for data connection and communication between the memory and the processor. When the construction holographic monitoring and early warning method program based on a digital model is executed by the processor, it implements the steps of the construction holographic monitoring and early warning method based on a digital model as described in any of the above claims.

[0037] This invention discloses a construction holographic monitoring and early warning method and system based on a digital model. The method first collects multi-source heterogeneous data from the target construction site, including resource allocation and process sequence, to construct a data flow path map of the entire construction process. It then extracts features from real-time video data using a spatiotemporal attention mechanism to generate the actual data flow path of the current construction scene. The actual path is then compared topologically with the pre-constructed map to identify abnormal data flow paths. A graph convolutional neural network is used to encode the features of the abnormal paths, accurately locating the source of construction disturbance and defining its impact range. Finally, based on the location results and impact data, a tiered early warning is triggered. This invention solves the problems of data silos, delayed anomaly identification, and difficulty in tracing the source in traditional monitoring, achieving dynamic perception, accurate early warning, and proactive risk control throughout the entire construction process, significantly improving the efficiency and scientific nature of construction safety management and decision-making.

[0038] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0039] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0040] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A construction holographic monitoring and early warning method based on a digital model, characterized in that, Includes the following steps: Acquire multi-source heterogeneous monitoring data of the target construction site, and construct a data flow path map of the entire construction process based on the multi-source heterogeneous monitoring data. The multi-source heterogeneous monitoring data includes resource allocation data and process connection sequence data for each construction process. Real-time video data is collected from the target construction site. The spatiotemporal attention mechanism algorithm is used to extract features of construction process nodes from the real-time video data to obtain the data transmission sequence of the current construction process. The actual data flow path in the current construction scenario is determined based on the data transmission sequence. The actual data flow path is compared with the data flow path map to identify abnormal data flow paths that deviate from the data flow path map. The abnormal data flow path is feature-encoded based on the graph convolutional neural network algorithm to determine the construction disturbance source and the scope of influence corresponding to the abnormal data flow path, and to obtain the disturbance source location result and the scope of influence data. Based on the location results of the disturbance source and the data on the scope of influence, a construction early warning operation is performed on the target construction site.

2. The construction holographic monitoring and early warning method based on a digital model according to claim 1, characterized in that, The process involves acquiring multi-source heterogeneous monitoring data from the target construction site and constructing a data flow path map of the entire construction process based on this data. Specifically, the multi-source heterogeneous monitoring data includes resource allocation data for each construction procedure, procedure connection timing data, and physical field monitoring data. The resource allocation data for each construction process is obtained by retrieving the number of personnel involved, the mechanical equipment scheduling records, and the material consumption ledgers from the process management system of the target construction site. Extract the interaction and occurrence time of adjacent processes, the duration of each process, and the dependency relationship identifiers between processes from the construction progress control platform to obtain process connection sequence data; Resource allocation data and sequence data of each construction process are associated and matched according to the unique identifier of the construction process, with the construction process as the node and the data interaction relationship between the processes as the directed edge; The weights of the directed edges are set to the information transmission delay time in the corresponding process connection time sequence data, and the attribute values ​​of the nodes are configured to the resource configuration data of the corresponding construction process. All nodes are connected to the directed edges in sequence according to the process advancement order of the entire construction process to construct the data flow path graph of the entire construction process.

3. The construction holographic monitoring and early warning method based on a digital model according to claim 1, characterized in that, The process involves real-time video data acquisition from the target construction site, feature extraction of construction process nodes from the real-time video data using a spatiotemporal attention mechanism algorithm to obtain the data transmission sequence of the current construction process, and determining the actual data flow path in the current construction scenario based on the data transmission sequence. Specifically: By collecting real-time video data streams covering the entire construction site using multi-view high-definition camera equipment deployed at the target construction site, the real-time video data streams are divided into continuous video frame sequences at a fixed frame rate. Feature point detection and matching operations are performed frame by frame on the continuous video frame sequences to extract the operating posture features of construction machinery, the movement trajectory features of personnel, and the transfer position features of materials in each video frame. The extracted features are used to construct a spatiotemporal feature tensor according to the time sequence of video frames. The spatiotemporal feature tensor is then aligned and mapped with the node attributes in the data flow path graph using the unique identifier of the construction process as an index. This results in the construction of a holographic digital model that includes the nodes of the entire construction process and the connection relationships between the processes. The spatiotemporal feature tensor is input into the spatiotemporal attention mechanism module of the construction holographic digital model. The spatiotemporal attention mechanism module sequentially calculates the spatial attention weight of different construction process nodes in the same video frame and the temporal attention weight of the same construction process node in different video frames. Based on the spatial attention weight, the process node features in the same frame are weighted and fused to obtain spatial aggregated features. The features of the same process node across frames are weighted and fused according to the temporal attention weight to obtain the temporal aggregated features. The spatial aggregated features and the temporal aggregated features are concatenated and input into the feature decoding layer. The data activation intensity of each construction process node in the current construction scenario is output. The construction process nodes are sorted in order of data activation intensity from high to low. The data transmission direction of each construction process node is concatenated according to the sorting result to obtain the data transmission sequence of the current construction process. By taking adjacent data transmission directions in the data transmission sequence as directed edges and the construction process nodes corresponding to the data transmission directions as endpoints, the topology of directed edges is constructed sequentially according to the connection relationship of the data transmission directions to obtain the actual data flow path in the current construction scenario.

4. The construction holographic monitoring and early warning method based on a digital model according to claim 1, characterized in that, The step of performing a topological comparison between the actual data flow path and the data flow path map to identify abnormal data flow paths that deviate from the data flow path map specifically involves: The directed edges of the actual data flow path are matched one by one with the directed edges of the corresponding process connection relationship in the data flow path graph. The data transmission direction of each adjacent process node in the actual data flow path is extracted and compared with the direction of the directed edge of the corresponding process connection relationship in the data flow path graph. Adjacent process nodes with inconsistent directions are marked as direction deviation edges. The data transmission delay of each adjacent process node in the actual data flow path is extracted and compared with the weight of the corresponding directed edge in the data flow path graph. The directed edge of the adjacent process node whose delay exceeds the set range of the corresponding weight is marked as a delay deviation edge. The abnormal data flow path in the actual construction process is determined based on the directional deviation edge and the delay deviation edge.

5. The construction holographic monitoring and early warning method based on a digital model according to claim 1, characterized in that, The graph convolutional neural network algorithm is used to encode the features of the abnormal data flow path, determine the construction disturbance source and the scope of influence corresponding to the abnormal data flow path, and obtain the disturbance source location result and the scope of influence data, specifically: Align and map the process nodes in the abnormal data flow path with the corresponding process nodes in the data flow path graph. Use the resource configuration data of each process node in the abnormal data flow path as the initial feature vector of the node, and use the deviation type of the directional deviation edge and the delay deviation edge in the abnormal data flow path as the edge feature vector to construct the abnormal data flow path graph structure data containing node features and edge features. The abnormal data flow path graph structure data is input into a pre-constructed graph convolutional neural network. The neighborhood features of each process node are aggregated layer by layer through multi-layer graph convolution operations. In each layer of graph convolution operation, the feature difference between the central process node and the adjacent process nodes is first calculated by weighting the edge feature vectors. The feature difference is concatenated with the current feature vector of the central process node and then mapped through a non-linear activation function to obtain the updated node feature vector. After layer-by-layer iteration, the high-order semantic feature vector of each process node is output. Calculate the cosine similarity between the high-order semantic feature vector of each process node and the baseline feature vector of the corresponding process node under normal working conditions. Mark process nodes with cosine similarity below a set threshold as candidate disturbance nodes. Statistically analyze the connected subgraphs of all candidate disturbance nodes in the abnormal data flow path graph structure data. The core candidate disturbance nodes in the connected subgraph are used as construction disturbance sources, and the spatial distribution range of all process nodes in the connected subgraph is extracted as the initial influence range. Based on the dependency relationship identifiers between processes in the data flow path map, the dependent process nodes are traversed layer by layer along the downstream dependency direction of each process node within the initial influence range. The dependent process nodes obtained from the traversal are included in the influence range. The spatial coordinates of the construction disturbance source are associated and encapsulated with the spatial boundary of the influence range to obtain the disturbance source location result and the influence range data.

6. The construction holographic monitoring and early warning method based on a digital model according to claim 1, characterized in that, The construction early warning operation based on the disturbance source location results and the impact range data for the target construction site is specifically as follows: Based on the disturbance source location results, the construction process nodes corresponding to the construction disturbance source are determined. Based on the influence range data, all construction process nodes within the influence range are extracted. Combined with the dependency relationships between each construction process node in the data flow path map, a disturbance propagation association chain is constructed, and the cumulative propagation influence value of each construction process node in the disturbance propagation association chain is calculated. The cumulative propagation impact value of each construction process node is correlated with the data activation intensity of the corresponding construction process node to obtain the comprehensive disturbance index of each construction process node. When the comprehensive disturbance index is greater than the preset disturbance index, the corresponding construction process node is marked as a key early warning node. When the comprehensive disturbance index is less than or equal to the preset disturbance index, the corresponding construction process node is marked as a continuous monitoring node. The proportion of key early warning nodes within the affected area is statistically analyzed, and the coverage ratio of the abnormal data flow path corresponding to the key early warning nodes is calculated. When the proportion of key early warning nodes is greater than the node proportion threshold and the coverage ratio of the abnormal data flow path is greater than the coverage ratio threshold, a construction interruption early warning instruction is generated and sent to the digital model. The corresponding construction process nodes within the affected area are marked as suspended, and the disturbance source location results and affected area data are output simultaneously. When the proportion of key early warning nodes is less than or equal to the node proportion threshold and the coverage ratio of abnormal data flow path is less than or equal to the coverage ratio threshold, a construction adjustment early warning instruction is generated, and the resource allocation order and process connection order of the construction process corresponding to the key early warning node are adjusted according to the disturbance propagation association chain. The construction interruption warning command or construction adjustment warning command is associated with the unique identifier of the corresponding construction process node, and the warning status identifier of the corresponding construction process node in the digital model is updated simultaneously to complete the construction warning operation of the target construction site.

7. A construction holographic monitoring and early warning system based on a digital model, characterized in that, The construction holographic monitoring and early warning system based on digital models includes a memory and a processor. The memory includes a construction holographic monitoring and early warning method program based on digital models. When the construction holographic monitoring and early warning method program based on digital models is executed by the processor, it implements the steps of the construction holographic monitoring and early warning method based on digital models as described in any one of claims 1 to 6.