Construction progress deviation early warning method and system based on data mining

Abstract

The invention relates to the technical field of big data analysis and mining, in particular to a construction progress deviation early warning method and system based on data mining, and the method comprises the following steps: integrating a construction task list, the spatial position of a component in a BIM, the number of operators in a construction log, and the operation state of a tower crane, and generating a multi-dimensional constraint construction map node set; according to the method, a multi-dimensional node set with time, space and resource attributes is constructed by integrating a construction task list, the spatial position of a BIM component, the number of operators in a construction log and the operation state of a tower crane, and structural expression of a construction site is achieved; on the basis, a causal association relationship among construction logic, resource occupation and space proximity is identified, and a time sequence causal conduction map with a deviation conduction coefficient value is further established, so that a risk conduction path between construction events has quantitative calculation capability.

Description

Technical Field

[0001] This invention relates to the field of big data analysis and mining technology, and in particular to a method and system for early warning of construction progress deviations based on data mining. Background Technology

[0002] The field of big data analysis and mining technology refers to a systematic approach that uses massive, complex, multi-source, and heterogeneous data as a foundation, and through data collection, storage, processing, modeling, and visualization, to uncover the potential relationships, evolutionary patterns, and decision-making information behind the data.

[0003] Current technologies lack the ability to model the causal transmission chains in construction data. In actual engineering projects, schedule deviations are often not triggered by a single node, but are caused by the combined effects of factors such as uneven resource allocation, spatial interference, or mismatched task sequences. Current technologies only establish weak correlations between these factors and fail to form a logical closed loop. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a construction progress deviation early warning method and system based on data mining.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a construction progress deviation early warning method based on data mining, comprising the following steps:

[0006] Integrate the construction task list, the spatial location of components in BIM, and the number of workers and tower crane operation status in the construction log to generate a multi-dimensional constrained construction drawing node set;

[0007] Based on the multidimensional constrained construction map node set, the construction logic, resource sharing and spatial proximity relationships between nodes are retrieved, causal relationship links between nodes are established, and deviation transmission coefficient values ​​are calculated and assigned to links based on the causal relationship links between nodes to construct a time-series causal transmission map.

[0008] Based on the aforementioned time-series causal transmission graph, the number of workers, tower crane load, and current project volume of each construction task node are extracted periodically to generate a construction status time-series data stream. Based on the construction status time-series data stream, the status sequence in which progress deviation occurs is marked, and a high-risk sequence pattern library of construction deviation is established.

[0009] Based on the high-risk sequence pattern library of construction deviations, the real-time status sequence of the current construction task is matched with the patterns in the library to obtain a set of high-risk status trigger nodes. Based on the set of high-risk status trigger nodes, the corresponding nodes are located in the temporal causal transmission graph, and a construction progress deviation warning instruction is output.

[0010] Preferably, the steps for obtaining the multidimensional constrained construction drawing node set are as follows:

[0011] The data of the construction task list, the spatial location of components in BIM, the number of workers and the tower crane operation status in the construction log are analyzed item by item. A unique identifier code is assigned to each construction task, resource and component. The corresponding time attribute, spatial attribute and resource requirement attribute are attached to each identifier code to obtain a set of construction entity codes containing all attribute information.

[0012] Based on the construction entity code set containing all attribute information, each identified code entity is traversed, and the attribute attached to the code entity is used as the basis for judgment to determine the attribute association between construction entity codes with the same construction task attribute, spatial attribute or resource requirement attribute, and an attribute association matrix is ​​established to form a construction entity attribute association matrix.

[0013] Based on the construction entity attribute association matrix, for each construction entity code in the matrix, the construction entity code is defined as an independent node in the graph according to the attribute association, and the nodes are mapped to the construction graph one by one to generate a multi-dimensional constrained construction graph node set.

[0014] Preferably, the step of obtaining the causal relationship links between nodes is as follows:

[0015] Based on the multidimensional constraint construction map node set, the construction task attributes of each independent node in the node set are analyzed one by one. According to the construction sequence rules corresponding to the construction task attributes, the construction execution order relationship between each independent node is matched one by one to establish a construction execution order relationship table and form a set of construction logic relationships.

[0016] Based on the set of construction logic relationships, the resource requirement attributes attached to each independent node are traversed. By comparing the number of workers and tower crane operation status in the resource requirement attributes attached to different independent nodes, resource types with common or conflicting needs are identified. Based on the consistency or conflict of resource types, a resource sharing relationship table between nodes is established to form a set of resource sharing relationships.

[0017] Based on the resource sharing relationship set, the spatial attribute coordinate information of each independent node is called. By calculating the Euclidean distance between the spatial attributes of different nodes, and using whether the distance between spatial coordinates is less than a preset spatial proximity threshold as a judgment condition, it is determined whether the nodes meet the spatial proximity constraint, a spatial proximity relationship table is established, and causal relationship links between nodes are formed.

[0018] Preferably, the steps for obtaining the temporal causal transmission map are as follows:

[0019] Based on the causal links between the nodes, the starting and ending nodes connected by each causal link are traversed one by one. The peak value of the resource demand attached to the starting node and the available resource capacity in the corresponding time period are extracted. The difference between the peak value of the resource demand and the available resource capacity is calculated and extracted as the resource gap value. At the same time, the time interval between the earliest start time of the ending node and the latest completion time of the starting node is calculated and extracted as the buffer time value. The planned duration of the starting node is also recorded to form a combined parameter set of resource gap value, buffer time value and planned duration.

[0020] Calculate the deviation transmission coefficient value based on the combined parameter set;

[0021] Based on the deviation transmission coefficient value, the deviation transmission coefficient value corresponding to each causal relationship link is bound to the attribute parameter of the link edge, the edge attribute of the causal relationship link is updated, and a time-series causal transmission graph is formed.

[0022] Preferably, the steps for acquiring the construction status time-series data stream are as follows:

[0023] Based on the aforementioned temporal causal transmission graph, the number of workers, tower crane load, and actual project volume for each construction task node are extracted sequentially according to a fixed time period. At the same time, the upper limit of the number of workers, the upper limit of the tower crane load, and the target planned project volume corresponding to the node are called to generate a temporal data stream of construction status.

[0024] Preferably, the steps for obtaining the high-risk sequence pattern library of construction deviations are as follows:

[0025] Based on the construction status time-series data stream, calculate the comprehensive risk index value for each consecutive time segment;

[0026] Based on the comprehensive risk index value, all continuous time segments are screened, and time segments with comprehensive risk index values ​​higher than the risk trigger threshold are extracted. According to the aggregation characteristics of the over-limit rate of workers, the over-limit rate of tower crane load, and the output deviation rate, adjacent risk segments are merged and uniformly marked as high-risk time period sequences to form a high-risk sequence pattern library of construction deviations.

[0027] Preferably, the steps for obtaining the high-risk state trigger node set are as follows:

[0028] Based on the high-risk sequence pattern library for construction deviations, the real-time status sequence of the current construction task is called, and each real-time status sequence is compared with the existing high-risk time period sequences in the pattern library. The start time, end time and construction task node identifier of the successfully matched real-time status sequence are recorded, and a high-risk status trigger node set is generated.

[0029] Preferably, the step of obtaining the construction progress deviation early warning instruction is as follows:

[0030] Based on the set of high-risk state triggering nodes, all construction task nodes in the set of high-risk state triggering nodes are retrieved one by one from the temporal causal transmission graph. The position of each construction task node in the graph is determined. Starting from each construction task node, the adjacent downstream nodes are traversed level by level along the directed edge. The deviation transmission coefficient value of each causal relationship link on the path is extracted and accumulated in sequence to obtain the transmission risk value of the downstream node.

[0031] Based on the transmission risk value of the downstream nodes, it is determined whether the transmission risk value of each downstream node exceeds the set risk warning threshold. Downstream nodes that exceed the risk warning threshold are marked as high-risk warning nodes, and a construction progress deviation warning instruction is output.

[0032] This invention also provides a construction progress deviation early warning system, including:

[0033] The map generation module integrates the construction task list, the spatial location of components in BIM, the number of workers and tower crane operation status in the construction log, and generates a multi-dimensional constrained construction map node set.

[0034] The causal graph construction module, based on the multidimensional constrained construction graph node set, retrieves the construction logic, resource sharing and spatial proximity relationships between nodes, establishes causal relationship links between nodes, calculates deviation transmission coefficient values ​​based on the causal relationship links between nodes and assigns them to links, and constructs a time-series causal transmission graph.

[0035] The high-risk pattern recognition module, based on the time-series causal transmission graph, extracts the number of workers, tower crane load and current project volume for each construction task node on a periodic basis, generates a construction status time-series data stream, and marks the status sequence in which progress deviation occurs based on the construction status time-series data stream, and establishes a high-risk sequence pattern library of construction deviation.

[0036] The deviation warning output module, based on the high-risk sequence pattern library of construction deviations, matches the real-time status sequence of the current construction task with the patterns in the library to obtain a set of high-risk status trigger nodes. Based on the set of high-risk status trigger nodes, it locates the corresponding node in the time-series causal transmission graph and outputs a construction progress deviation warning command.

[0037] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0038] This invention integrates construction task lists, BIM component spatial locations, and worker numbers and tower crane operating status from construction logs to construct a multi-dimensional node set with temporal, spatial, and resource attributes, achieving a structured representation of the construction site. Based on this, it identifies the causal relationships between construction logic, resource occupancy, and spatial proximity, further establishing a temporal causal transmission graph with deviation transmission coefficient values, enabling quantitative calculation of risk transmission paths between construction events. By periodically extracting real-time status data of construction nodes and combining them into a continuous sequence in chronological order, it identifies deviation segments from the plan, achieving a correlation between construction status and historical deviation risks. Using high-risk pattern matching, it locates potential high-risk nodes and performs risk accumulation extrapolation along the deviation path propagation direction, assessing the degree of impact on downstream nodes, ultimately achieving early warning and intervention for key processes. This quantifies implicit influencing factors into parameter indicators on the transmission path, improving the accuracy and response speed of deviation identification. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the steps of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0041] Please see Figure 1 This invention provides a technical solution: a construction progress deviation early warning method based on data mining, comprising the following steps:

[0042] Integrate the construction task list, the spatial location of components in BIM, and the number of workers and tower crane operation status in the construction log to generate a multi-dimensional constrained construction drawing node set;

[0043] Based on the multidimensional constrained construction map node set, the construction logic, resource sharing and spatial proximity relationships between nodes are retrieved, causal relationship links between nodes are established, deviation transmission coefficient values ​​are calculated and assigned to links based on the causal relationship links between nodes, and a time-series causal transmission map is constructed.

[0044] Based on the temporal causal transmission graph, the number of workers, tower crane load and current project volume of each construction task node are extracted periodically to generate a construction status temporal data stream. Based on the construction status temporal data stream, the status sequence in which progress deviation occurs is marked, and a high-risk sequence pattern library of construction deviation is established.

[0045] Based on the high-risk sequence pattern library of construction deviations, the real-time status sequence of the current construction task is matched with the patterns in the library to obtain the set of high-risk status trigger nodes. Based on the set of high-risk status trigger nodes, the corresponding nodes are located in the temporal causal transmission graph, and a construction progress deviation warning instruction is output.

[0046] The steps for obtaining the node set of the multidimensional constrained construction drawing are as follows:

[0047] The data of the construction task list, the spatial location of components in BIM, the number of workers and the tower crane operation status in the construction log are analyzed item by item. A unique identifier code is assigned to each construction task, resource and component. The corresponding time attribute, spatial attribute and resource requirement attribute are attached to each identifier code to obtain a set of construction entity codes containing all attribute information.

[0048] Based on the set of construction entity codes containing all attribute information, each identified code entity is traversed. The attributes attached to the code entity are used as the basis for judgment to determine the attribute association between construction entity codes that have the same construction task attribute, spatial attribute, or resource requirement attribute, and an attribute association matrix is ​​established to form a construction entity attribute association matrix.

[0049] Based on the construction entity attribute association matrix, for each construction entity code in the matrix, the construction entity code is defined as an independent node in the graph according to the attribute association, and the nodes are mapped to the construction graph one by one to generate a multi-dimensional constrained construction graph node set.

[0050] Specifically, the system first performs item-by-item data parsing on the construction task list, the spatial location of components in the BIM, the number of workers in the construction log, and the tower crane operating status. Specifically, the system interface first reads the construction task list stored in tabular form, extracting the task name, unique WBS (Work Breakdown Structure) code, planned start time, planned finish time, and the WBS code of the preceding tasks for each row. Then, it calls the BIM model interface to traverse all building components in the model. For each component, it extracts its unique IFCGUID (Globally Unique Identifier), component type (e.g., beam, slab, column), and the three-dimensional coordinates (X, Y, F, G) of its geometric center calculated from its 3D model. The system calculates the minimum and maximum vertex coordinates of the bounding box (Y, Z) and its circumbound box. Simultaneously, it synchronously or periodically reads construction logs in text format. Using a regular expression-based matching pattern, it parses the log entries to extract the record date, associated task WBS code, the work teams and personnel involved that day, the tower crane equipment number used, and the tower crane's operating status at different time periods. For example, "8:00-10:00, Tower Crane #1, hoisting precast slab PC-01-03" is parsed into time period, equipment number, work content, and associated components. Based on the parsed data, it assigns each independent construction task, BIM component, and resource (here, resources mainly refer to...) to... Each work team and tower crane is assigned a globally unique identifier code, generated in the format of "entity type-original ID". For example, a task with WBS code "A-01-01" has a unique identifier code of "TASK-A-01-01", a component with IFCGUID "1a2b3c" has an identifier code of "COMP-1a2b3c", and a tower crane numbered "TC-01" has an identifier code of "RES-TC-01". Next, standardized attribute information is appended to each generated identifier code, where the time attribute is obtained from the construction task list and is formatted to include the planned start time and planned finish time. The spatial attributes of the tuples are obtained from the BIM model and recorded as the (X, Y, Z) coordinates of the geometric center of the component. The resource requirement attributes are determined based on the construction task list and construction log. For example, for an entity "TASK-A-01-01", its resource requirement attributes will be recorded as needing a "RES-TEAM-CONCRETE-01" work team and a "RES-TC-01" tower crane, with the required quantity specified, such as 10 workers and 1 tower crane. Finally, all entities carrying unique identification codes and their corresponding time, space, and resource requirement attributes are organized into a structured list to obtain a set of construction entity codes containing all attribute information.

[0051] Based on the construction entity code set containing all attribute information obtained in the previous steps, the system initiates a nested loop to traverse each pair of identified code entities in the set. Using the attributes attached to the code entities as the criterion, in each iteration, the attribute information of two entities (entity i and entity j) is extracted and compared to determine whether there is an attribute relationship between them. Specifically, the determination of attribute relationship is processed in parallel across three dimensions. First, in the construction task attribute dimension, if the code type of both entity i and entity j is "TASK", the system checks whether their WBS codes have a hierarchical relationship (such as "A-01" and "A-01-01") or are explicitly defined as having a direct sequence in the construction task list. Logically, if two entities are both of type "COMP", the system will trace back to their respective construction tasks. If they serve the same WBS task code, it is determined that they have a construction task attribute association. Secondly, in the spatial attribute dimension, this judgment is only triggered when both entity i and entity j are component entities of type "COMP". The system extracts the geometric center coordinates in their spatial attributes and calculates the three-dimensional Euclidean distance between them. At the same time, a dynamic spatial proximity threshold is set for the judgment. This threshold is not a fixed value, but is related to the type of construction activity it is associated with. For example, for the "structural hoisting" task that requires the operation of large equipment, its spatial proximity threshold is set to 1 of the average size of the components involved. Five times, if one component is 8 meters and the other is 6 meters, with an average size of 7 meters, then the threshold is set to 10.5 meters. For delicate tasks such as "interior wall plastering," the threshold is tightened to 1.0 times the average size plus one standard deviation to avoid work surface conflicts. If the calculated Euclidean distance is less than the spatial proximity threshold corresponding to the current construction task, then the two are determined to have a spatial attribute relationship. Third, in the resource demand attribute dimension, this judgment mainly targets two "TASK" type entities. The system checks the resource identifier code requested in their resource demand attributes. If both requested the same unique resource, such as the same "RES-TC-01" tower crane, and their time attributes (planned start and completion) are consistent, then the threshold is set to 10.5 meters. If there is an overlap in the construction time, it is determined that the two entities have a resource requirement attribute association. After completing the association judgment of all entity pairs, the system constructs an N×N square matrix, where N is the total number of entities in the construction entity code set. The row and column indices of this matrix correspond to the unique identifier code of each entity. The element M(i,j) in the matrix is ​​not a single value, but a vector containing three Boolean values ​​[task_assoc, spatial_assoc, resource_assoc]. The value of each bit (1 or 0) indicates whether there is a construction task, spatial or resource-level association between entity i and entity j. Finally, the entire matrix is ​​filled to form the construction entity attribute association matrix.

[0052] Based on the construction entity attribute association matrix generated in the previous process, the system begins to map the abstract entity codes into graphically represented nodes. Specifically, the system traverses each row of the construction entity attribute association matrix. Since the rows and columns of the matrix are symmetrical and correspond one-to-one with all construction entities, traversing the rows can cover all entities. For the i-th row of the matrix, the system extracts its corresponding construction entity code and uses this code as a unique identifier for a new node in the graph. Subsequently, the system retrieves the complete attribute record matching the current construction entity code from the original "construction entity code set containing all attribute information," including its detailed time attributes (planned start and finish times), spatial attributes (three-dimensional coordinates), and resource requirement attributes (required personnel, equipment, and their quantities). This structured attribute information is stored as the internal attributes of the newly created node. This means that each generated node is not only an identity identifier but also an information-rich data container. It fully inherits the multi-dimensional information carried by the original construction entity. For example, when processing the node representing "..." When the entity code for "Pouring the shear wall in Area A of the first floor" is generated, the system creates a node with the ID "TASK-A-01-01". Its internal attributes record the planned construction period, required concrete volume, number of workers, and the spatial location information of all wall components corresponding to it in the BIM model. This mapping process is essentially a transformation of tabular entity data into node objects oriented towards a graph structure. Although the construction entity attribute association matrix is ​​used as an index for traversal in this step, the relationship information between entities recorded in the matrix (i.e., the M(i,j) vector) is not used to build the connection edges between nodes at this time, but is temporarily reserved as the direct basis for building graph edges (i.e., causal relationship links) later. By processing each construction entity code in the matrix one by one, the system converts them all into independent nodes carrying complete multidimensional constraint information and stores these node objects in a set, thus finally generating a multidimensional constraint construction graph node set that contains all tasks, components, and key resources in the project, and each element is ready for relational linking.

[0053] The steps to obtain the causal links between nodes are as follows:

[0054] Based on the multidimensional constrained construction diagram node set, the construction task attributes of each independent node in the node set are analyzed one by one. According to the construction sequence rules corresponding to the construction task attributes, the construction execution order relationship between each independent node is matched one by one to establish a construction execution order relationship table and form a set of construction logic relationships.

[0055] Based on the set of construction logic relationships, the resource demand attributes attached to each independent node are traversed. By comparing the number of workers and tower crane operation status in the resource demand attributes attached to different independent nodes, resource types with common or conflicting needs are identified. Based on the consistency or conflict of resource types, a resource sharing relationship table between nodes is established to form a set of resource sharing relationships.

[0056] Based on the resource sharing relationship set, the spatial attribute coordinate information of each independent node is called. By calculating the Euclidean distance between the spatial attributes of different nodes, and using whether the distance between spatial coordinates is less than the preset spatial proximity threshold as the judgment condition, it is determined whether the nodes meet the spatial proximity constraint, a spatial proximity relationship table is established, and causal relationship links between nodes are formed.

[0057] Specifically, based on the multi-dimensional constrained construction drawing node set, the system initiates a traversal process, parsing the internal attributes of each independent node marked as "TASK" in the node set. Specifically, the system first extracts the node's "construction task attribute," which includes the WBS code parsed from the original construction task list and the explicitly stated predecessor task WBS codes. Then, the system matches any two independent nodes according to a built-in construction sequence rule base. This rule base consists of two parts: the first part is hard rules based on general project management logic, such as the "complete-start" (FS) relationship, meaning a node (task) can only begin after its predecessor node (task) is completed; the second part is process sequence rules based on common sense in architectural engineering, such as "foundation construction" must precede "main structure construction," and in the "column construction" task, "formwork installation" must precede "reinforcement binding," and "reinforcement binding" must precede "concrete pouring." The matching process... Implemented through a double loop, for nodes A and B, the system first checks whether the WBS code of node A is in the "preceding task WBS code" list of node B. If so, a "complete-start" relationship is established from node A to node B. Next, the system queries the task types of the two nodes and compares them with the process sequence rule base. For example, if node A is "C1 column formwork installation" and node B is "C1 column rebar tying", the system establishes a logical link from A to B according to the rule base. By performing the above matching logic on all possible task node pairs in the node set, the system generates a list containing all direct dependencies. Each item in this list is a tuple in the format (preceding task node identifier, subsequent task node identifier, relationship type), such as (TASK-A-01-01, TASK-A-01-02, FS). This list is the construction execution sequence relationship table. Finally, all these relationship entries are summarized to form a construction logical relationship set.

[0058] Based on the set of construction logic relationships, the system further traverses each independent node in the multi-dimensional constraint construction diagram node set, extracting its additional "resource requirement attributes." These attributes record in detail the specific resources required by the construction task represented by the node during execution, including the type and number of work teams, as well as the number and planned usage period of the specific tower crane equipment. The system identifies potential resource sharing and conflicts by comparing the resource requirement attributes of different independent nodes. The specific comparison process is as follows: For any two construction task nodes (node ​​A and node B) that overlap in time, the system first checks whether their "resource requirement attributes" contain the same unique resource identifier. For example, if node A's requirement is "{Tower Crane: RES-TC-01, Time: D1-D3}", and node B's requirement is "{Tower Crane: RES-TC-01, Time: D2-D4}", since both require the same tower crane during the period from the second day to the third day (D2-D3),... When the system identifies "RES-TC-01", it recognizes it as a resource conflict with shared needs. Next, it checks if the total demand for similar non-unique resources exceeds the available limit. For example, if the project has a daily available concrete worker team of 50, and node A needs 30 concrete workers on a certain day, while node B needs 25 on the same day, the total demand of 55 exceeds the available limit of 50. The system then determines this is a resource pool-based conflict. Based on these identified shared needs or conflicts, the system begins to build a resource sharing relationship table between nodes. This table records all node pairs with resource associations. Each record is formatted as (Node Identifier 1, Node Identifier 2, Associated Resource Identifier, Relationship Type). The relationship type can be categorized as "Exclusive Resource Conflict" or "Total Resource Insufficiency". By comparing all node pairs with overlapping time periods, the system finally summarizes all identified relationship entries to form a resource sharing relationship set.

[0059] Based on the resource-sharing relationship set, the system continues to call the "spatial attribute coordinate information" of each independent node in the multi-dimensional constraint construction drawing node set to determine the spatial proximity constraints between nodes. This process mainly targets nodes representing specific components (COMP type) or task (TASK type) nodes directly associated with components. The system traverses all possible node pairs (node ​​i, node j) in the node set, extracting their respective spatial attributes. These attributes are usually the (X, Y, Z) three-dimensional coordinates of the component's geometric center. Then, the system quantifies their spatial proximity by calculating the three-dimensional Euclidean distance between these two coordinate points. The calculation process involves solving for the square root of the sum of the squares of the differences between the two points on the X, Y, and Z coordinate axes. Next, the calculated distance value is compared with a preset spatial proximity threshold. This threshold is not fixed but dynamically set according to the construction activity type of the nodes involved. For example, for tasks such as "large precast component hoisting," the safety operating space requirement is large, and spatial proximity is crucial. The method for setting the proximity threshold is to take the average of the diagonal lengths of the outer bounding boxes of the two components, and then multiply it by a safety factor of 1.2. For example, if the diagonal length of component A is 10 meters and that of component B is 8 meters, then the threshold is ((10+8) / 2)*1.2=10.8 meters. If the center distance between the two components is less than 10.8 meters, then it is determined that there is a spatial proximity constraint. For manual activities such as "interior decoration", the working space is small, and the threshold can be set to an empirical value, such as 3 meters, which represents the minimum working surface of a standard work team. If the calculated Euclidean distance is less than the spatial proximity threshold in the current scenario, the system establishes a spatial proximity relationship between the two nodes and records the relationship in the spatial proximity relationship table. Each item in the table is a tuple containing the identifiers of two nodes. Finally, the system merges the spatial proximity relationship table generated in this step with the previously established set of construction logic relationships and resource sharing relationships to form a complete set describing the mutual influence between nodes, that is, the causal relationship link between nodes.

[0060] The steps for obtaining a temporal causal transmission map are as follows:

[0061] Based on the causal links between nodes, the starting node and the ending node connected by each causal link are traversed one by one. The peak value of the resource demand attached to the starting node and the available resource capacity in the corresponding time period are extracted. The difference between the peak value of the resource demand and the available resource capacity is calculated and extracted as the resource gap value. At the same time, the time interval between the earliest start time of the ending node and the latest completion time of the starting node is calculated and extracted as the buffer time value. The planned duration of the starting node is also recorded to form a combined parameter set of resource gap value, buffer time value and planned duration.

[0062] Based on the combined parameter set, the deviation transmission coefficient value is calculated using the following formula:

[0063]

[0064] Among them, D c U is the bias transmission coefficient value of the c-th causal link. c Let A be the peak resource requirement of the starting node in the c-th causal link. c Let T be the available resource capacity of the starting node in the c-th causal link, where ∈ is a very small positive constant with a denominator of zero. c T is the buffer time value for the c-th causal link. dur The planned duration of the starting node in the c-th causal link;

[0065] Based on the deviation transmission coefficient value, the deviation transmission coefficient value corresponding to each causal relationship link is bound to the attribute parameter of the link edge, the edge attribute of the causal relationship link is updated, and a time-series causal transmission graph is formed.

[0066] Specifically, based on the causal links between nodes, the system traverses each link one by one, taking the starting and ending nodes as the processing objects. First, the system focuses on the starting node and extracts its internal resource requirement attributes. These attributes record the daily resource requirements of the task within the planning period in the form of a time series. For example, for the starting node "pouring three-story slabs", its resource requirement attributes may be recorded as "{number of workers: [D1: 10, D2: 15, D3: 12], tower crane: [D1: TC-02(4h), D2: TC-02(4h)], tower crane: [D1: TC-02(4h)], tower crane: [D ... (8h), D3: TC-02(6h)]}”, the system identifies the maximum demand value of each type of resource during the entire task duration from this sequence, that is, the peak demand of the resource quantity. In this example, the peak demand for the number of workers is 15 people, and the peak demand for the tower crane is 8 hours of daily use. Subsequently, the system queries the project-level resource pool database to obtain the total available capacity of the corresponding type of resource during the initial task execution period. For example, the project department has a total of 50 concrete workers registered, and the daily available time of tower crane No. 1 is 10 hours. The system will then subtract all the resources that are related to the maximum demand value of each type of resource during the entire task execution period. The system calculates the available resource capacity for the current task by considering the resources allocated to other tasks running concurrently with it. If other tasks occupy 38 concrete workers, the available resource capacity for this task is 12 workers. The system then calculates the difference between the peak resource demand and the available resource capacity. If the peak is greater than the capacity, the difference is the resource gap value. In this example, the resource gap value for the number of workers is 3 workers. If the peak is less than or equal to the capacity, the resource gap value is 0. Simultaneously, the system calls the time parameters of each node obtained by performing forward and backward calculations on the entire project network plan using the Critical Path Method (CPM). It extracts the earliest start time of the termination node and the latest finish time of the start node, calculates the time difference between the two, and this difference is the buffer time value. It reflects the maximum leeway that the start task can be delayed without affecting the timely start of the termination task. Finally, the system directly reads the planned duration from the attributes of the start node and combines the calculated resource gap value, buffer time value, and recorded planned duration into a structured set of combined parameters, providing quantitative input for subsequent deviation propagation analysis.

[0067] formula: The advantage of formulas is that they can... This item relativizes the resource gap, accurately measuring the pressure that resource scarcity puts on the task, avoiding misjudgments caused by absolute gap values, while the exponential item... This introduces the time dimension, which causes the risk reduction effect of the buffer time to decrease exponentially. That is, when the buffer time is short, the risk decays very slowly, reflecting the severity of the tight schedule, while when the buffer time is sufficient, the risk is rapidly reduced.

[0068] Uc The peak resource demand in the c-th causal link is the starting node. This parameter represents the maximum demand for a specific resource (such as workers of a specific trade or equipment of a specific type) within the entire planned duration of the construction task represented by the starting node. Obtaining this value requires first parsing the construction task list and resource plan, extracting the detailed resource demand plan for each task on a daily or hourly basis, forming a time series of resource demands. For example, for a 5-day "exterior stone cladding" task, the planned demand for stone installers is [Day 1: 8 people, Day 2: 12 people, Day 3: 12 people, Day 4: 10 people, Day 5: 6 people]. The system will iterate through this sequence to find the maximum value; therefore, the peak resource demand U for this task is the peak value. c That is, 12 people. This parameter directly reflects the maximum instantaneous pressure on resource supply during the execution of the task.

[0069] A c This parameter represents the available resource capacity of the starting node in the c-th causal link. It indicates the total amount of that resource that can be allocated to this task at the specific time point when the initial task requires peak resources. The process of obtaining this resource is relatively complex. First, the system needs to query the total amount of that resource from the project-level resource library. For example, if the project team has signed contracts for 50 stone installers, this is the theoretical maximum available amount. Next, the system needs to calculate the total resource requirements of all other ongoing parallel tasks that require the same resource on the days when the initial task's peak demand occurs (i.e., the 2nd and 3rd days). For example, if the statistics show that other parallel tasks need to call 35 stone installers on the 2nd day, then the available resource capacity A available for this task on the 2nd day is... c This is the total amount of resources minus the amount occupied by other tasks, i.e., 50-35=15 people. This value dynamically reflects the actual resource supply capacity under the consideration of resource competition for the entire project.

[0070] To avoid extremely small positive numbers with a denominator of zero, this parameter is only used to ensure numerical stability in calculations. In floating-point arithmetic, a divisor of zero will lead to program errors or results in infinity. In practical engineering, the available resource capacity A... c In some extreme cases, such as when resources are completely occupied by other higher-priority tasks, the theoretical value might be zero. To prevent this from causing the formula to fail, ∈ is introduced as an additional term in the denominator. Its value is set to a very small positive number, much smaller than any actual amount of resources, such as 10. -6 This value is set on the premise that it must be small enough that its existence will not have a meaningful impact on the normal calculation results (for example, for resources measured in "people", adding 10...). -6 The impact can be ignored, but at the same time, the denominator A can be guaranteed.c +∈ is always positive, thus ensuring the robustness of the algorithm. In all calculations, ∈ is always a fixed value of 10. -6 value.

[0071] T c The buffer time value for the c-th causal link, also known as the total float time, represents the slack time that the starting task can have without affecting the overall project duration or the latest start time of its immediate successor tasks. This value is determined by analyzing the entire project's construction network plan. Specifically, the system uses the Critical Path Method (CPM). First, it performs forward calculations on all task nodes to obtain the earliest start time (ES) and earliest finish time (EF) for each task. Then, it performs backward calculations to obtain the latest start time (LS) and latest finish time (LF) for each task. For a "finish-start" (FS) logical relationship link from starting node A to ending node B, its buffer time T... c The calculation formula is: T c =LS B -EF A or T c =LF B -EF A -Duration B To illustrate with a specific example, if the earliest completion time of the initial task A (masonry work) is day 20, and the latest start time of its immediate successor task B (plastering work) is day 23, then the buffer time T between these two tasks... c That is, 23 - 20 = 3 days.

[0072] T dur The planned duration of the starting node in the c-th causal link is the standard completion period specified in the initial construction plan for the starting task. This parameter is static basic data, extracted directly from the construction task list or project scheduling software (such as P6, MS Project). This data is determined during the project planning phase. For example, in the project WBS decomposition, if task "A-2-3: Second Floor Interior Wall Masonry" is assigned a planned duration of 10 working days, then when calculating the deviation transmission coefficient for all causal links starting with "A-2-3", its T... dur All parameter values ​​will be uniformly set to 10 days. This parameter will be used as the denominator in the formula to normalize the buffer time, so that the relative length of the buffer time (relative to the task duration itself) can be evaluated, thereby more reasonably measuring its risk reduction effect.

[0073] Calculation process:

[0074] Based on the aforementioned parameter acquisition steps, the following parameter values ​​are obtained:

[0075] Peak resource requirement U of the starting node c =12 people;

[0076] The available resource capacity A of the starting node c =10 people;

[0077] Avoid extremely small positive numbers with a denominator of zero ∈=10 -6 ;

[0078] Buffer time value T of causal link c = 3 days;

[0079] The planned duration T of the starting node dur = 10 days.

[0080] The calculation process is as follows:

[0081] Calculate the resource stress level within the parentheses:

[0082]

[0083] Calculate the buffer capacity decay factor for the exponential part:

[0084]

[0085] Exponenterate the two parts mentioned above:

[0086] (1.2) 0.7408 ≈1.1465;

[0087] Finally, subtract 1 to normalize:

[0088] D c =1.1465-1=0.1465;

[0089] The results show that, after considering the resource gap of 2 people and the buffer time of 3 days, the deviation transmission coefficient of the causal link from the "exterior wall stone dry hanging" task to the "exterior wall sealant construction" task is 0.1465. This is a dimensionless risk indicator. A value greater than 0 indicates the existence of deviation transmission risk. The magnitude of the value (0.1465) quantifies the intensity of this risk. In the early warning system, a graded threshold can be set. For example, 0.1 can be set as the low-risk threshold and 0.3 as the medium-risk threshold. If the calculated deviation transmission coefficient is 0.1465, the system can mark this link as low-risk.

[0090] Based on the deviation transmission coefficient values, the system binds these values, which quantify the intensity of risk transmission, to the graph edges corresponding to the causal links between nodes. Specifically, the system first traverses all calculated deviation transmission coefficient values ​​and their corresponding causal links. Each causal link is represented in the graph as a directed edge from the starting node to the ending node. In the graph's data structure, the system predefines an attribute field named "transmission coefficient" for each edge. During the traversal, the system retrieves the deviation transmission coefficient value of a link, for example, 0.1465, and then finds the directed edge corresponding to this link, pointing from the "exterior wall stone dry hanging" node to the "exterior wall sealant construction" node. The floating-point number 0.1465 is written into the "transmission coefficient" attribute field of the edge. This process is performed sequentially on all causal links until the deviation transmission coefficient values ​​of all links are assigned to their corresponding edges in the graph. Through this series of binding and updating operations, the unweighted edges that originally only represented logical, resource, or spatial relationships are given weights that can accurately measure the probability and intensity of the transmission of delay risk. The original static topology network has thus evolved into a dynamic, complex network with weighted directed edges. Finally, when the attributes of all edges have been updated, a complete temporal causal transmission graph that can simulate and predict how construction progress deviations are transmitted between different tasks is constructed.

[0091] The steps for obtaining the construction status time-series data stream are as follows:

[0092] Based on the temporal causal transmission graph, the number of workers, tower crane load and current actual project volume of each construction task node are extracted sequentially according to a fixed time period. At the same time, the upper limit of the number of workers, the upper limit of the tower crane load and the target planned project volume corresponding to the node are called to generate a temporal data stream of construction status.

[0093] Specifically, based on the time-series causal transmission graph, the system sets a fixed data collection period. For example, at the end of each workday, for each ongoing construction task node in the graph, the data extraction program is initiated. First, the system obtains the total number of workers actually involved in the task that day through interfaces with the site management system, such as personnel entry and exit records from access control gates or sign-in data from mobile attendance apps. Next, the system connects to the tower crane monitoring system, which is equipped with sensors that can monitor the tower crane boom angle, lifting capacity, and running time in real time. The system then extracts... The total effective working hours and average lifting weight of the tower crane associated with this task are converted into a standardized tower crane load rate using the tower crane performance curve. For example, if a tower crane with a rated lifting moment of 100 ton-meters completes a total of 800 ton-meters of lifting operations on a given day, its load rate is equivalent to 8 standard shifts. Simultaneously, the system calls the interface with the BIM model and compares the site point cloud data obtained through technologies such as laser scanning or UAV photogrammetry with the BIM model to automatically calculate the actual amount of work completed that day. For example, comparing the planned pouring volume in the BIM model... For a 100-cubic-meter concrete wall under construction, point cloud data shows that 70 cubic meters have been completed. Therefore, the actual project volume for the current period is 70 cubic meters. While extracting this real-time data, the system retrieves preset baseline values ​​from the static attributes of this construction task node. These include the maximum number of workers set according to the construction organization design and safety specifications (e.g., 15 people), the maximum tower crane load set according to the tower crane equipment manual and operating procedures (e.g., no more than 8 standard shifts per day), and the target planned project volume broken down from the project schedule to the current day (e.g., 20 cubic meters). These real-time collected values ​​are paired with preset baseline values ​​at each time point. The system generates a time-series record containing multi-dimensional status information for each construction task node, such as "[Date: 2023-10-26, Task ID: TASK-A-01-01, Actual number of people: 18, Maximum number of people: 15, Actual load: 9.2, Maximum load: 8, Actual amount of work: 18, Planned amount of work: 20]". Finally, the records of all task nodes in the continuous time period are collected to form a structured construction status time-series data stream.

[0094] The steps for obtaining the high-risk sequence pattern library of construction deviations are as follows:

[0095] Based on the construction status time-series data stream, the comprehensive risk index value for each consecutive time segment is calculated using the following formula:

[0096]

[0097] Among them, R′ m Let P be the comprehensive risk index value of the m-th continuous time slice. j Let P be the number of workers at time j. maxL represents the maximum number of workers allowed for this task. j Let L be the tower crane load at time j. max E represents the maximum allowable tower crane load for this task. j E represents the actual amount of work completed at time j. target,j Let β be the target planned workload at time j, β be the risk sensitivity coefficient, and k be the number of consecutive time points contained in the m-th time slice.

[0098] Based on the comprehensive risk index value, all continuous time segments are screened, and time segments with comprehensive risk index values ​​higher than the risk trigger threshold are extracted. According to the aggregation characteristics of the over-limit rate of workers, the over-limit rate of tower crane load, and the output deviation rate, adjacent risk segments are merged and uniformly marked as high-risk time period sequences, forming a high-risk sequence pattern library of construction deviations.

[0099] Specifically, the formula:

[0100] The advantages of the formula are that it integrates the risk of exceeding resource input limits (number of workers, tower crane load) by using the square root of the sum of squares. This structure is equivalent to calculating the Euclidean distance, which can comprehensively reflect the overall imbalance of resource allocation, rather than simply adding up the various risks. Secondly, the formula introduces the exponential function exp(·) to amplify the impact of output deviation (work volume not meeting the standard). This means that even if the resource input is only slightly exceeded, if the output is seriously insufficient, the overall risk will be amplified sharply. Finally, the introduction of the risk sensitivity coefficient β provides the model with adjustable flexibility, allowing managers to dynamically adjust their focus on output efficiency according to the criticality of the project or the risk tolerance of a specific stage.

[0101] P j Let P be the number of workers at time j. This parameter refers to the number of on-site workers actually participating in a specific construction task at time j within a data collection period (e.g., a day). This data is collected automatically, for example, by deploying intelligent access control systems with RFID or facial recognition functions at the construction site entrance and key work areas. When workers wearing identification badges enter or leave the work area of ​​a specific task, the system automatically records their ID and timestamp. At the end of each collection period (e.g., 18:00 daily), the system backend will count the number of unique personnel IDs who have stayed in the task work area for more than the preset minimum working time (e.g., 4 hours), thus obtaining the total number of workers involved that day. For example, on October 27th, the system recorded that 22 different workers were effectively working in the "three-story beam and slab reinforcement binding" task area. Then, the P of this task at time j on October 27th is P. j The value is 22 people.

[0102] P max This parameter represents the maximum number of workers allowed for the task. It is a threshold determined based on the construction organization design, safety management regulations, and work space capacity. It signifies the maximum number of simultaneous workers that can be accommodated to ensure safety, efficiency, and quality in a specific construction task. The determination process is as follows: First, the construction technical department calculates the theoretically required number of workers based on the task's workload and standard labor quotas. Second, combining this with the actual work space area of ​​the task in the BIM model, the maximum number of workers that the space can accommodate is calculated. Finally, the project manager combines both factors, considering the impact of overlapping work, and approves a final maximum number of workers. For example, for the task of "three-story beam and slab reinforcement binding," the work area is 200 square meters, and according to regulations, it can accommodate a maximum of 133 people. However, considering efficiency and management, the final determined maximum number of workers is... max There are 20 people.

[0103] L j Let L be the tower crane load at time j. This parameter quantifies the actual working intensity of the tower crane serving the task within a specific data collection period. Its acquisition relies on Internet of Things (IoT) sensors installed on the tower crane, including lifting torque sensors, slewing angle sensors, and hook height sensors. The system collects raw data from these sensors in real time and calculates the actual torque and working time for each lifting operation based on the tower crane's performance parameters (rated lifting torque, maximum boom length, etc.). At the end of the data collection period, the system sums the products of torque and time for all lifting operations serving the task to obtain the total lifting torque work. This sum is then divided by the rated lifting torque work corresponding to a "standard shift" for that type of tower crane to obtain a dimensionless load value. For example, if the lifting torque work of a tower crane in a standard shift is 1000 ton-meter-hours, and it completed 1100 ton-meter-hours of work for a certain task on October 27th, then its load L at that time... j The value is 1.1.

[0104] L maxThe maximum allowable tower crane load for this task is set based on the tower crane's performance manual, safety operating procedures, and the overall project lifting plan. This parameter represents the maximum daily working load and is designed to prevent mechanical fatigue, performance degradation, and even safety accidents caused by overuse. The determination process is as follows: First, consult the tower crane manufacturer's equipment manual to obtain their recommended maximum daily continuous working hours and the number of cycles under rated load. Second, the project equipment management department reduces the theoretical upper limit based on maintenance cycles and equipment aging; for example, stipulating that equipment utilization should not exceed 85%. Finally, considering the day's weather conditions (such as wind speed restrictions) and the coordination plan among multiple tower cranes, a final load limit is determined. For example, for a tower crane with a standard 8-hour shift, considering nighttime closures and midday breaks, and applying an 85% utilization factor, its L... max It may be set as one standard shift.

[0105] E j Let E represent the actual completed work volume at time j. This parameter indicates the actual physical work volume produced by the construction task within a specific data collection period. Its acquisition primarily relies on automated on-site measurement technology. For example, for volume-related tasks such as concrete pouring or earthwork excavation, a drone equipped with LiDAR can be used to scan the work area twice. By comparing the point cloud models generated from the two scans, the volume change can be accurately calculated, which is the actual completed work volume. For tasks measured by quantity or area, such as rebar tying or formwork installation, quality inspectors can use mobile terminals equipped with image recognition capabilities to photograph and count the completed portions. The background AI algorithm automatically identifies and counts the quantity. For example, on October 27th, through drone scanning and comparison, the earthwork excavation volume of "Area C of the foundation pit" was calculated to be 450 cubic meters. Therefore, the E of this task on that day... j The value is 450.

[0106] E target,j Let E be the target planned quantity of work at time j. This parameter is derived from the overall project schedule (Level 1 schedule) through decomposition and refinement, representing the quantity of work that should be completed within a specific data collection period. It serves as a benchmark for measuring whether the schedule is progressing as planned. The acquisition process is as follows: The project planning engineer first uses project management software to create a detailed network plan that includes all construction tasks, their durations, and quantities. Then, the software evenly distributes the total quantity of work to each workday based on the planned start and end dates of the tasks and a preset work calendar, thereby generating the daily planned quantity of work. For example, for an earthwork excavation task with a total quantity of 5000 cubic meters and a planned duration of 10 days, if distributed linearly, the daily target planned quantity of work E would be... target,j That is 500 cubic meters.

[0107] β is the risk sensitivity coefficient, an adjustable weighting parameter used to control the contribution of output deviation to the overall risk index. Its value reflects the project manager's emphasis on "efficiency" risk. The β value is not fixed but dynamically adjusted based on project characteristics and current management priorities. The setting process is as follows: The project management team first conducts a criticality assessment of all construction tasks in the project. The assessment dimensions include whether they are on the critical path, the degree of impact on subsequent work, and the milestone requirements in the contract. Then, based on the assessment results, the tasks are divided into three critical levels: "high," "medium," and "low." For tasks of the "high" critical level, such as the main structure construction on the critical path, the manager has a very low tolerance for schedule delays, so a high β value is set, such as β = 2.0, to amplify the risk signal caused by any insufficient output. For "medium" critical tasks, a moderate β value is set, such as β = 1.5. For auxiliary work on non-critical paths of the "low" critical level, a low β value is set, such as β = 1.0.

[0108] k represents the number of consecutive moments contained in the m-th time slice. This parameter defines the size of the time window for calculating the comprehensive risk index, i.e., how many consecutive data collection cycles (moments) are considered as an independent analysis unit "time slice". The setting of its value needs to balance the immediacy of risk identification with the stability of trend judgment. If the value of k is too small (e.g., k=1), occasional fluctuations on a single day may be misjudged as high risk, leading to excessively frequent warnings. If the value of k is too large, it will delay the discovery of persistent risks. Usually, this value is determined according to the nature and duration of the construction task. For short-cycle, rapidly changing tasks, such as sporadic operations that can be completed in a single day, k=1 can be set. For regular construction tasks that last for several weeks, such as floor structure construction, k=5 is usually set, i.e., a complete work week (5 days) is used as a time slice to smooth out daily fluctuations and observe the overall trend of the week.

[0109] Calculation process:

[0110] Based on the construction status time-series data stream, calculate the comprehensive risk index value for a continuous 5-day time slice (k=5), where the task corresponding to this time slice is "reinforcement binding of three-story beams and slabs," and its P... max For 20 people, L max For one standard shift, β is set to 1.5.

[0111] The 5-day time series data is as follows:

[0112] Day 1: P1 = 18, L1 = 0.9, E1 = 3.8 tons, E target,1 = 4 tons;

[0113] Day 2: P2 = 22, L2 = 1.1, E2 = 4.2 tons, E target,2= 4 tons;

[0114] Day3: P3=21, L3=1.2, E3=3.5 tons, E target,3 = 4 tons;

[0115] Day 4: P4 = 19, L4 = 0.8, E4 = 4.1 tons, E target,4 = 4 tons;

[0116] Day 5: P5 = 23, L5 = 1.0, E5 = 3.0 tons, E target,5 = 4 tons;

[0117] The calculation process is as follows:

[0118] Day 1:

[0119] Resource exceeding limits:

[0120] Output deviation:

[0121] Daily risk value: 0 * 1.0779 = 0;

[0122] Day 2:

[0123] Resource exceeding limits:

[0124]

[0125] Output deviation:

[0126] Daily risk value: 0.1414·1 = 0.1414;

[0127] Day 3:

[0128] Resource exceeding limits:

[0129]

[0130] Output deviation:

[0131] Daily risk value: 0.2062·1.2062≈0.2487;

[0132] Day 4:

[0133] Resource exceeding limits:

[0134] Output deviation:

[0135] Daily risk value: 0.1 = 0;

[0136] Day 5:

[0137] Resource exceeding limits: Output deviation:

[0138] Daily risk value: 0.15·1.4550≈0.2183;

[0139] Finally, the daily risk values ​​are summed to obtain the overall risk index value R′ for that time slice. m :

[0140] R′ m =0 + 0.1414 + 0.2487 + 0 + 0.2183 = 0.6084;

[0141] The results show that the comprehensive risk index value for this 5-day time slice is 0.6084. This value is a cumulative risk measure, and its absolute value can be compared with the preset risk trigger threshold to determine whether there is a risk of schedule deviation during this period. For example, if the risk trigger threshold is set to 0.5, then 0.6084 exceeds the threshold, indicating that the task execution status is abnormal during this period and needs to be monitored.

[0142] Based on the comprehensive risk index value, the system automatically filters all calculated continuous time segments. Its core judgment criterion is a preset risk trigger threshold, which is determined based on statistical analysis of historical project data and expert experience. Specifically, by analyzing the complete construction status time-series data streams of several similar past projects, the system calculates the comprehensive risk index value of all time segments, forming a distribution. The 90th percentile of this distribution is selected as the initial benchmark, for example, 1.2. Experienced project managers then fine-tune this threshold according to risk management strategies. For example, for a demanding contract, the threshold might be lowered to 1.0 to improve the sensitivity of the warning. The system iterates through all time segments, comparing their comprehensive risk index values ​​with this risk trigger threshold (e.g., 1.0). Time segments with index values ​​higher than this threshold are initially marked as high-risk candidate segments. Subsequently, the system further analyzes these... Candidate segments undergo aggregation analysis. This involves examining the raw data within each segment, calculating the over-limit rate of workers (number of days exceeding limits / total number of days), the over-limit rate of tower crane load, and the average output deviation rate (daily average of (1-E / E_target)). Based on the similarity of these characteristics, high-risk candidate segments that are temporally adjacent or close are merged. For example, if two adjacent 5-day time segments both have a risk index exceeding 1.0 and their main risk characteristics are "high over-limit rate of workers and high output deviation rate," the system merges these two 10-day time segments into a unified high-risk time segment sequence and assigns it a pattern label, such as "high manpower, low work risk." Through the above-mentioned filtering, extraction, aggregation, and labeling processes on all historical and current project data, the system continuously builds and expands a structured database containing various typical high-risk state evolution patterns, namely, the construction deviation high-risk sequence pattern library.

[0143] The steps to obtain the high-risk state trigger node set are as follows:

[0144] Based on the high-risk sequence pattern library for construction deviations, the real-time status sequence of the current construction task is called, and each real-time status sequence is compared with the existing high-risk time period sequences in the pattern library. The start time, end time and construction task node identifier of the successfully matched real-time status sequence are recorded, and a high-risk status trigger node set is generated.

[0145] Specifically, based on the high-risk construction deviation sequence pattern library, the system calls the real-time status sequence of all currently ongoing construction tasks. This real-time status sequence is a multi-dimensional time series consisting of the number of workers, tower crane load, and actual project volume for the most recent N days (the value of N is determined according to the typical length of the patterns in the pattern library, for example, 5 days) on a daily basis. Subsequently, the real-time status sequence of each task is compared with each high-risk time period sequence pattern stored in the pattern library. The specific matching process is as follows: First, the various dimensions of the real-time status sequence and the pattern sequence in the library are normalized. For example, the actual number of workers P on day j is normalized. j Through formula P norm,j =P j / P max Convert to a relative value between 0 and 1, where P max This represents the upper limit of the number of workers allowed for the task. Similarly, the tower crane load and workload are normalized relative to their upper limits or targets, forming a standardized multidimensional feature vector sequence. Then, for a real-time sequence S of a task and a pattern P in the library, the system calculates the DTW distance between them. This distance quantifies the morphological similarity between the two sequences, allowing for non-linear distortion on the time axis and enabling the identification of risk patterns that are essentially the same but have slightly different rhythms. After completing the distance calculation, the system compares the distance value with a preset matching similarity threshold. This threshold is dynamically set for each pattern in the library. The method for setting it is to calculate the similarity between the pattern and all historical instance sequences that have been identified by experts as having the same risk level when the pattern is added to the library. The DTW distance is calculated by taking the average of these distance values ​​plus 1.2 times the standard deviation, which serves as a loose matching boundary. For example, for the "high-person-low-work-risk" pattern, the average historical matching distance is 0.25 and the standard deviation is 0.05. Therefore, the matching similarity threshold is set to 0.25 + 1.2 * 0.05 = 0.31. If the currently calculated DTW distance is less than 0.31, the match is considered successful. Once a match is successful, the system immediately records the start and end times of the real-time state sequence, as well as the unique identifier of the construction task node to which the sequence belongs, and adds this record to a temporary list. After traversing all combinations of current tasks and all patterns in the library, this temporary list becomes the set of high-risk state trigger nodes.

[0146] The steps to obtain the construction progress deviation early warning instruction are as follows:

[0147] Based on the set of high-risk state triggering nodes, all construction task nodes in the set of high-risk state triggering nodes are retrieved one by one from the temporal causal transmission graph. The position of each construction task node in the graph is determined. Starting from each construction task node, the adjacent downstream nodes are traversed level by level along the directed edge. The deviation transmission coefficient value of each causal relationship link on the path is extracted and accumulated in sequence to obtain the transmission risk value of the downstream node.

[0148] Based on the transmission risk value of downstream nodes, it is determined whether the transmission risk value of each downstream node exceeds the set risk warning threshold. Downstream nodes that exceed the risk warning threshold are marked as high-risk warning nodes, and construction progress deviation warning instructions are output.

[0149] Specifically, based on the set of high-risk state trigger nodes, the system retrieves all construction task node identifiers recorded within this set from the temporal causal transmission graph. These nodes are precisely located in the graph as the starting point for risk propagation. For each located trigger node, the system uses it as the root and employs a depth-limited breadth-first search (BFS) algorithm to traverse downstream along the directed edges in the graph level by level to trace the risk propagation path. The traversal depth is set as a configurable parameter, typically three levels. This means the system only analyzes downstream tasks that are no more than two intermediate tasks away from the trigger node, focusing on the most direct and tightly linked impact chains. During the process, whenever moving from a parent node U to a child node V, the system extracts the pre-calculated and stored deviation transmission coefficient value from the causal link connecting U and V. This coefficient value reflects the strength of risk transmission from U to V. The system calculates the transmission risk value of each downstream node using a path risk accumulation method. Specifically, there may be multiple paths from a trigger node S to any downstream node N. The risk transmission strength of each path is defined as the product of all deviation transmission coefficient values ​​on that path. The final transmission risk value of node N is the value of the path with the highest risk transmission strength among all paths from S to N. The calculation formula is as follows: Where D c (e) is the deviation transmission coefficient value of edge e on the path. If a downstream node can be affected by multiple different triggering nodes, its final transmission risk value is the maximum value calculated from all possible sources. The system maintains a transmission risk value attribute for each node in the graph and continuously updates it during the traversal calculation process until the downstream impact analysis of all triggering nodes is completed. Finally, a set containing all affected downstream nodes and their corresponding maximum transmission risk values ​​is obtained, which is the transmission risk value of the downstream node.

[0150] Based on the transmission risk value of downstream nodes, the system assesses each downstream node with a calculated transmission risk value, comparing its risk value with a dynamically set risk warning threshold. This threshold is not static but is correlated with the criticality level of the downstream node itself. The specific setting process is as follows: First, the system categorizes each construction task node into three criticality levels based on its position in the project network plan. Tasks on the critical path with a total float of zero or negative are marked as "high" criticality; tasks with a total float between 1 and 5 days are marked as "medium" criticality; and tasks with a total float greater than 5 days are marked as "low" criticality. Then, differentiated risk warning thresholds are set for different levels. These thresholds are derived by calibrating historical project data, analyzing the probability of actual delays occurring at nodes with different transmission risk values ​​in historical data, and setting thresholds to achieve the highest possible warning accuracy (e.g., 85%) within an acceptable false alarm rate (e.g., 15%). For example... The risk warning threshold for "high" critical level nodes is 0.35, for "medium" critical level it is 0.50, and for "low" critical level it is 0.65. During the judgment process, the system extracts the transmission risk value and criticality level of a downstream node and compares it with the corresponding threshold. For example, a node on the critical path with a transmission risk value of 0.42 is marked as a high-risk warning node because it exceeds the threshold of 0.35. Conversely, if a non-critical node has a transmission risk value of 0.48, it is not processed because it does not exceed the threshold of 0.65. For all tasks marked as high-risk warning nodes, the system will generate a structured construction progress deviation warning instruction. This instruction includes the following fields: warning node identifier, task name of the warning node, calculated transmission risk value, the corresponding risk warning threshold, the main source of risk (i.e., the initial trigger node identifier), and the strongest path of risk transmission (listing the node sequence on the path and the deviation transmission coefficient value of each link). Finally, all generated instructions are output uniformly.

[0151] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A construction progress deviation early warning method based on data mining, characterized in that, Includes the following steps: Integrate the construction task list, the spatial location of components in BIM, and the number of workers and tower crane operation status in the construction log to generate a multi-dimensional constrained construction drawing node set; Based on the multidimensional constrained construction map node set, the construction logic, resource sharing and spatial proximity relationships between nodes are retrieved, causal relationship links between nodes are established, and deviation transmission coefficient values ​​are calculated and assigned to links based on the causal relationship links between nodes to construct a time-series causal transmission map. Based on the aforementioned time-series causal transmission graph, the number of workers, tower crane load, and current project volume of each construction task node are extracted periodically to generate a construction status time-series data stream. Based on the construction status time-series data stream, the status sequence in which progress deviation occurs is marked, and a high-risk sequence pattern library of construction deviation is established. Based on the high-risk sequence pattern library of construction deviations, the real-time status sequence of the current construction task is matched with the patterns in the library to obtain a set of high-risk status trigger nodes. Based on the set of high-risk status trigger nodes, the corresponding nodes are located in the temporal causal transmission graph, and a construction progress deviation warning instruction is output.

2. The construction progress deviation early warning method based on data mining according to claim 1, characterized in that, The steps for obtaining the multidimensional constrained construction drawing node set are as follows: The data of the construction task list, the spatial location of components in BIM, the number of workers and the tower crane operation status in the construction log are analyzed item by item. A unique identifier code is assigned to each construction task, resource and component. The corresponding time attribute, spatial attribute and resource requirement attribute are attached to each identifier code to obtain a set of construction entity codes containing all attribute information. Based on the construction entity code set containing all attribute information, each identified code entity is traversed, and the attribute attached to the code entity is used as the basis for judgment to determine the attribute association between construction entity codes with the same construction task attribute, spatial attribute or resource requirement attribute, and an attribute association matrix is ​​established to form a construction entity attribute association matrix. Based on the construction entity attribute association matrix, for each construction entity code in the matrix, the construction entity code is defined as an independent node in the graph according to the attribute association, and the nodes are mapped to the construction graph one by one to generate a multi-dimensional constrained construction graph node set.

3. The construction progress deviation early warning method based on data mining according to claim 1, characterized in that, The steps for obtaining the causal relationship links between the nodes are as follows: Based on the multidimensional constraint construction map node set, the construction task attributes of each independent node in the node set are analyzed one by one. According to the construction sequence rules corresponding to the construction task attributes, the construction execution order relationship between each independent node is matched one by one to establish a construction execution order relationship table and form a set of construction logic relationships. Based on the set of construction logic relationships, the resource requirement attributes attached to each independent node are traversed. By comparing the number of workers and tower crane operation status in the resource requirement attributes attached to different independent nodes, resource types with common or conflicting needs are identified. Based on the consistency or conflict of resource types, a resource sharing relationship table between nodes is established to form a set of resource sharing relationships. Based on the resource sharing relationship set, the spatial attribute coordinate information of each independent node is called. By calculating the Euclidean distance between the spatial attributes of different nodes, and using whether the distance between spatial coordinates is less than a preset spatial proximity threshold as a judgment condition, it is determined whether the nodes meet the spatial proximity constraint, a spatial proximity relationship table is established, and causal relationship links between nodes are formed.

4. The construction progress deviation early warning method based on data mining according to claim 1, characterized in that, The steps for obtaining the temporal causal transmission map are as follows: Based on the causal links between the nodes, the starting and ending nodes connected by each causal link are traversed one by one. The peak value of the resource demand attached to the starting node and the available resource capacity in the corresponding time period are extracted. The difference between the peak value of the resource demand and the available resource capacity is calculated and extracted as the resource gap value. At the same time, the time interval between the earliest start time of the ending node and the latest completion time of the starting node is calculated and extracted as the buffer time value. The planned duration of the starting node is also recorded to form a combined parameter set of resource gap value, buffer time value and planned duration. Calculate the deviation transmission coefficient value based on the combined parameter set; Based on the deviation transmission coefficient value, the deviation transmission coefficient value corresponding to each causal relationship link is bound to the attribute parameter of the link edge, the edge attribute of the causal relationship link is updated, and a time-series causal transmission graph is formed.

5. The construction progress deviation early warning method based on data mining according to claim 1, characterized in that, The steps for obtaining the construction status time-series data stream are as follows: Based on the aforementioned temporal causal transmission graph, the number of workers, tower crane load, and actual project volume for each construction task node are extracted sequentially according to a fixed time period. At the same time, the upper limit of the number of workers, the upper limit of the tower crane load, and the target planned project volume corresponding to the node are called to generate a temporal data stream of construction status.

6. The construction progress deviation early warning method based on data mining according to claim 1, characterized in that, The steps for obtaining the high-risk sequence pattern library of construction deviations are as follows: Based on the construction status time-series data stream, calculate the comprehensive risk index value for each consecutive time segment; Based on the comprehensive risk index value, all continuous time segments are screened, and time segments with comprehensive risk index values ​​higher than the risk trigger threshold are extracted. According to the aggregation characteristics of the over-limit rate of workers, the over-limit rate of tower crane load, and the output deviation rate, adjacent risk segments are merged and uniformly marked as high-risk time period sequences to form a high-risk sequence pattern library of construction deviations.

7. The construction progress deviation early warning method based on data mining according to claim 1, characterized in that, The steps for obtaining the set of high-risk state trigger nodes are as follows: Based on the high-risk sequence pattern library for construction deviations, the real-time status sequence of the current construction task is called, and each real-time status sequence is compared with the existing high-risk time period sequences in the pattern library. The start time, end time and construction task node identifier of the successfully matched real-time status sequence are recorded, and a high-risk status trigger node set is generated.

8. The construction progress deviation early warning method based on data mining according to claim 1, characterized in that, The steps for obtaining the construction progress deviation early warning instruction are as follows: Based on the set of high-risk state triggering nodes, all construction task nodes in the set of high-risk state triggering nodes are retrieved one by one from the temporal causal transmission graph. The position of each construction task node in the graph is determined. Starting from each construction task node, the adjacent downstream nodes are traversed level by level along the directed edge. The deviation transmission coefficient value of each causal relationship link on the path is extracted and accumulated in sequence to obtain the transmission risk value of the downstream node. Based on the transmission risk value of the downstream nodes, it is determined whether the transmission risk value of each downstream node exceeds the set risk warning threshold. Downstream nodes that exceed the risk warning threshold are marked as high-risk warning nodes, and a construction progress deviation warning instruction is output.

9. The construction schedule deviation early warning system according to any one of claims 1-8, characterized in that, include: The map generation module integrates the construction task list, the spatial location of components in BIM, and the number of workers and tower crane operation status in the construction log to generate a multi-dimensional constrained construction map node set. The causal graph construction module, based on the multidimensional constrained construction graph node set, retrieves the construction logic, resource sharing and spatial proximity relationships between nodes, establishes causal relationship links between nodes, calculates deviation transmission coefficient values ​​based on the causal relationship links between nodes and assigns them to links, and constructs a time-series causal transmission graph. The high-risk pattern recognition module, based on the time-series causal transmission graph, extracts the number of workers, tower crane load and current project volume for each construction task node on a periodic basis, generates a construction status time-series data stream, and marks the status sequence in which progress deviation occurs based on the construction status time-series data stream, and establishes a high-risk sequence pattern library of construction deviation. The deviation warning output module, based on the high-risk sequence pattern library of construction deviations, matches the real-time status sequence of the current construction task with the patterns in the library to obtain a set of high-risk status trigger nodes. Based on the set of high-risk status trigger nodes, it locates the corresponding node in the time-series causal transmission graph and outputs a construction progress deviation warning command.

Images