A bim-based construction resource real-time management method and system

Abstract

The application discloses a kind of based on BIM's construction resource real-time management method and system, it is related to building construction technical field.The method includes: obtaining the real-time internet of things data stream of construction site, BIM model data, the physical resource in internet of things data stream is dynamically associated to the task component of BIM model, generates the context consistency degree of fusion perception data and current task;Dynamic context model is built;The current state of dynamic context model is input into the preset construction knowledge graph, identifies abnormal event, and carries out cause-effect path backtracking to abnormal event, and outputs at least the abnormal diagnosis report containing diagnostic confidence;Abnormal diagnosis report is pushed to management personnel, receives feedback data, utilizes feedback data and abnormal diagnosis report, and optimizes the inference rule of construction knowledge graph.The application effectively improves the accuracy and adaptability of construction resource management intelligent decision-making.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of building construction, in particular to a construction resource real-time management method and system based on BIM. BACKGROUND

[0002] With the rapid advancement of building industrialization and intelligentization, the scale and complexity of construction projects continue to improve, and the demand for real-time deployment and dynamic control of construction resources is increasingly urgent. The existing construction resource management methods rely on manual inspection and report statistics, which are inefficient and prone to information lag. Although data state monitoring technology based on the Internet of Things has gradually developed, it can achieve multi-dimensional data collection and visual display of personnel, equipment and environment, but it is still at the data collection level.

[0003] However, the traditional method can gather massive Internet of Things data streams and BIM model data, achieve simple threshold alarm and state presentation, but cannot deeply analyze the on-site business logic behind the data, lacks correlation analysis capability, and cannot accurately locate the root cause of resource abnormalities, and cannot provide effective construction resource real-time management intelligent decision support for managers. SUMMARY

[0004] The present application provides a construction resource real-time management method and system based on BIM, which aims to solve the technical problem that the existing technology cannot provide effective construction resource real-time management intelligent decision support for managers.

[0005] In view of the above problems, the present application provides a construction resource real-time management method and system based on BIM.

[0006] In a first aspect, the present application provides a construction resource real-time management method based on BIM, comprising:

[0007] Obtaining real-time Internet of Things data streams and BIM model data of the construction site, dynamically associating physical resources in the Internet of Things data streams to task components of the BIM model, and generating context consistency degrees of fusion perception data and current tasks;

[0008] Based on the fusion perception data and its context consistency degree, a dynamic context model is constructed in real time;

[0009] Inputting the current state of the dynamic context model into a preset construction knowledge graph, identifying abnormal events, and performing causal path tracing on the abnormal events, and outputting an abnormal diagnosis report containing at least a diagnosis confidence;

[0010] Pushing the abnormal diagnosis report to the manager, receiving feedback data, and optimizing the inference rules of the construction knowledge graph using the feedback data and the abnormal diagnosis report.

[0011] In a second aspect, the present application provides a BIM-based construction resource real-time management system, comprising:

[0012] a data perception fusion module configured to acquire real-time Internet of Things data streams and BIM model data of a construction site, dynamically associate physical resources in the Internet of Things data streams to task components of the BIM model, and generate a context consistency degree of the fusion perception data and the current task;

[0013] a dynamic context modeling module configured to construct a dynamic context model in real time based on the fusion perception data and the context consistency degree thereof;

[0014] an abnormality diagnosis and tracing module configured to input a current state of the dynamic context model into a preset construction knowledge graph, identify an abnormal event, trace a cause-effect path of the abnormal event, and output an abnormality diagnosis report containing at least a diagnosis confidence degree;

[0015] a graph feedback optimization module configured to push the abnormality diagnosis report to a manager, receive feedback data, and optimize inference rules of the construction knowledge graph by using the feedback data and the abnormality diagnosis report.

[0016] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0017] The present application provides a BIM-based construction resource real-time management method and system, which realizes accurate dynamic association of physical resources and task components and generates a context consistency degree by deeply fusing Internet of Things data streams of a construction site with BIM model data, effectively solving the problem of disconnection between data and business logic; a multi-layer dynamic context model is constructed based on the fused data, which comprehensively integrates information in the physical, business and social dimensions, providing panoramic site state support for abnormality analysis; the construction knowledge graph is used to realize accurate identification and cause-effect path tracing of abnormal events, outputting a structured report with a diagnosis confidence degree, filling the gap in the existing technology in terms of root cause diagnosis capability; the inference rules of the knowledge graph are continuously optimized by receiving feedback data from managers, forming a closed-loop mechanism of data, analysis, diagnosis and optimization, and continuously improving the accuracy and adaptability of intelligent decision-making, and ultimately realizing intelligent upgrading of construction resource management from passive monitoring to active early warning and from experience-based decision-making to data-driven decision-making. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0019] Figure 1A flowchart illustrating a BIM-based real-time construction resource management method provided in an embodiment of the present invention;

[0020] Figure 2 A schematic diagram of the structure of a BIM-based real-time construction resource management system provided in an embodiment of the present invention;

[0021] The components represented by each number in the attached diagram are explained below:

[0022] Data perception and fusion module 11, dynamic context modeling module 12, anomaly diagnosis and tracing module 13, and graph feedback optimization module 14. Detailed Implementation

[0023] This invention provides a BIM-based real-time construction resource management method and system, which addresses the technical problem that existing technologies cannot provide managers with effective intelligent decision support for real-time construction resource management.

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0025] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0026] Example 1, as Figure 1 As shown, this invention provides a BIM-based real-time construction resource management method, the method comprising:

[0027] S100: Acquire real-time IoT data streams and BIM model data from the construction site, dynamically associate physical resources in the IoT data streams with task components in the BIM model, and generate contextual consistency between fused sensing data and the current task.

[0028] In the embodiment of the application, real-time Internet of Things data streams and BIM model data of the construction site are acquired, physical resources in the Internet of Things data streams are dynamically associated with task components of the BIM model, and a context consistency degree of fused perception data and a current task is generated. In a construction resource management scenario, real-time data collected by Internet of Things devices and design / planning data carried by a BIM model are naturally in a data island state: the former can only reflect the original state of physical resources, and the latter can only present preset information of a task, and the lack of effective association between the two leads to the inability of data to be mapped to a specific construction business scenario. Meanwhile, the prior art does not quantitatively evaluate the matching degree of data and a task, and subsequent abnormal diagnosis is prone to misjudgment due to the disconnection between data and business. Therefore, the step needs to realize the business fusion of multi-source data by data acquisition, space-time alignment, resource-task association and consistency degree calculation, and provide a high-quality and scenario-based data basis for subsequent dynamic modeling and intelligent diagnosis.

[0029] The step S100 in the method provided by the embodiment of the application comprises:

[0030] The Internet of Things data streams of the construction site are collected, and the Internet of Things data streams at least comprise: real-time coordinates of personnel and Internet of Things devices acquired through positioning tags, operating parameters of mechanical equipment acquired through state sensors, and temperature and humidity data acquired through environmental sensors;

[0031] The space boundary and the task planning time window of the corresponding task are extracted from the BIM model;

[0032] A unified space-time coordinate system of the construction site is established, the real-time coordinates of the Internet of Things devices, the BIM model coordinates and the actual coordinates of the construction site are converted and aligned, and a space-time mapping relationship is formed;

[0033] The resource positions of various physical resources in the Internet of Things data streams are compared with the space boundary of the activated task in the BIM model in real time, and when the physical resources enter the space boundary of the corresponding task and are within the task planning time window, the dynamic association between the physical resources and the BIM model is established;

[0034] Based on the resource positions of the physical resources and the space boundary, the overlap ratio of the resource positions of the physical resources and the space boundary is calculated as a spatial consistency degree;

[0035] Based on the time when the physical resources enter the space boundary and the task planning time window, a time coincidence degree is calculated;

[0036] Based on the spatial consistency degree and the time coincidence degree, a context consistency degree of the fused perception data is calculated.

[0037] Firstly, the Internet of Things data stream of the construction site is collected, which at least includes: real-time coordinates of personnel and Internet of Things equipment obtained through positioning tags, running parameters of mechanical equipment obtained through state sensors, and temperature and humidity data obtained through environmental sensors. The Internet of Things data stream refers to the continuous data sequence reflecting the state of physical resources and environmental conditions collected in real time by various Internet of Things terminals deployed in the construction site, including positioning tags, state sensors, environmental sensors, etc. UWB positioning tags are deployed in the construction site, worn by construction personnel, installed on mechanical equipment, and device state sensors are added to tower cranes, steel cutting machines, etc. Temperature and humidity sensors are arranged in the work area, and real-time coordinates, mechanical equipment running parameters, and temperature and humidity data are collected in real time through wireless transmission. For example, the collected data includes: positioning tag data of construction personnel A, real-time coordinates: X=35.2m, Y=18.7m, Z=12.3m; state sensor data of tower crane B, running parameters: lifting capacity 3.5t, rotation speed 0.8r / min, hydraulic oil temperature 42℃; environmental sensor data of the work area, temperature: 25℃, relative humidity 60%RH.

[0038] Secondly, the spatial boundary and task plan time window corresponding to the task are extracted from the BIM model. BIM model, which stands for Building Information Model, is a digital model containing multi-dimensional information such as building engineering geometric size, task division, time plan, etc. Spatial boundary refers to the range of physical work area corresponding to a construction task in three-dimensional space. Task plan time window refers to the preset start and end time interval of the task. For example, through BIM model visualization software such as Revit, Navisworks, open project BIM model, locate to TZ-001 (10th floor steel binding) task node, extract the three-dimensional spatial coordinate range and preset time plan of the task. Extracted from the BIM model: spatial boundary: , Y∈[15.0m,20.0m],Z∈[12.0m,14.0m];Task plan time window: May 10, 2025 08:00-18:00, a total of 10 hours.

[0039] Further, a unified space-time coordinate system is established to align the real-time coordinates of the Internet of Things devices, the BIM model coordinates, and the actual coordinates of the construction site, forming a space-time mapping relationship. The unified space-time coordinate system refers to a unified reference system that includes three-dimensional space coordinates and time dimension, with a fixed reference point in the construction site as the origin, used to eliminate the coordinate deviation of different data sources. Select the fixed reference point O in the construction site, the actual coordinates: X0=0m, Y0=0m, Z0=0m; convert the design coordinates of the BIM model and the collection coordinates of the Internet of Things devices with the actual coordinates of the reference point O, respectively, to establish the mapping relationship between the design coordinates and the actual coordinates, and the collection coordinates and the actual coordinates. For example, the design coordinates of the TZ-001 task in the BIM model are X=30.0m, corresponding to the actual coordinates X=30.5m in the construction site, the conversion formula is: actual X=design X+0.5m; the positioning tag collection coordinates of tower crane B are X=4.7m, corresponding to the actual coordinates X=30.5m, the conversion formula is: actual X=collection X+25.8m; finally, the mapping relationship under the unified space-time coordinate system is formed: the BIM design coordinates and the device collection coordinates are all converted to the actual coordinates with O as the origin.

[0040] Subsequently, the resource locations of various physical resources in the Internet of Things data stream are compared with the spatial boundaries of the activated tasks in the BIM model in real time, and when the physical resources enter the spatial boundaries of the corresponding tasks and are within the task planning time window, the dynamic association between the physical resources and the BIM model is established. Physical resources refer to personnel, mechanical equipment, tools, and other entities participating in construction tasks on the construction site. Activated tasks refer to construction tasks that are currently within the task planning time window, are being performed, or are to be performed. Dynamic association refers to automatically establishing the binding relationship between resources and tasks when the actual state of the physical resources meets the task requirements, and dynamically updating with the change of resource state. Based on the unified space-time coordinate system, the actual location of the physical resources and the spatial boundary of the activated tasks, the current time of the physical resources and the task planning time window are compared in real time; if both the location is within the spatial boundary and the time is within the time window, the association is automatically established. For example, time judgment: the current time is May 10, 2025, 10:00, which is within the time window 08:00-18:00 of the TZ-001 task; location judgment: the actual coordinates of tower crane B (35.2m, 18.7m, 12.3m) fall within the spatial boundary [30.0-40.0m, 15.0-20.0m, 12.0-14.0m]; the dynamic association between tower crane B and the TZ-001 task is automatically established, and the association record is [resource ID: TD-002 (tower crane B) - task ID: TZ-001 (10th floor steel binding)].

[0041] Further, based on the resource location of the physical resource and the space boundary, the coincidence ratio of the resource location of the physical resource and the space boundary is calculated as the spatial consistency degree. The spatial consistency degree indicates the coincidence degree of the actual location of the physical resource and the task space boundary, and quantitatively reflects whether the resource is accurately located in the task operation area. The value range is [0, 1], 0 represents complete non-coincidence, and 1 represents complete coincidence. The area coincidence ratio method is used for calculation. Taking the core working range of the physical resource as the benchmark, the proportion of the overlapping volume of the range and the task space boundary to the total volume of the task space is calculated as the spatial consistency degree. For example, the core operation range of tower crane B is a cylindrical region with its actual coordinates as the center and a radius of 5 m, with a volume of ; the space boundary of the TZ-001 task is a cuboid region, with a volume of ; the overlapping volume of the two is = 90 ; the spatial consistency degree is / = 90 / 100 = 0.9.

[0042] In addition, based on the time when the physical resource enters the space boundary and the task planning time window, the time coincidence degree is calculated. The time coincidence degree indicates the matching degree of the time when the physical resource enters the task space boundary and the task planning time window, and quantitatively reflects whether the resource participates in the operation within the task preset time period. The value range is [0, 1], 0 represents complete non-coincidence, and 1 represents complete coincidence. If the resource enters within the time window, the time coincidence degree = (current operation time + remaining planned operation time) / total task time; if the resource enters in advance or delays, the coincidence degree is deducted according to the deviation time, and 0.1 is deducted for every 1 hour of deviation, with a minimum of 0. For example, tower crane B enters the TZ-001 task space at 09:00 on May 10, 2025, and the current time is 10:00, with 1 hour of operation; the total task time is 10 hours, the remaining planned operation time is 8 hours, and the planned end time is 18:00; the time coincidence degree = (1 + 8) / 10 = 0.9; if the tower crane enters at 19:00, it deviates by 1 hour, and the time coincidence degree = 0.8.

[0043] Finally, based on the spatial consistency and the time coincidence, the context consistency of the corresponding fusion perception data is calculated. The context consistency refers to the matching degree of the fusion perception data and the current construction task context, is the core index for quantifying the data reliability and the business relevance, and has a value range of [0, 1]. The higher the value is, the more the data can support the subsequent business analysis. The weighted average method is adopted, the spatial accuracy has a greater impact on the work, the weight of the spatial consistency is set to 0.6, the weight of the time coincidence is set to 0.4, and the formula is: context consistency = spatial consistency * 0.6 + time coincidence * 0.4. For example, if the spatial consistency is 0.9 and the time coincidence is 0.9, the context consistency = 0.9 * 0.6 + 0.9 * 0.4 = 0.9, which indicates that the fusion perception data of the tower crane B is highly matched with the context of the TZ-001 task.

[0044] In the embodiment of the application, through multi-source data acquisition, space-time coordinate alignment, resource-task dynamic association and three-level consistency calculation, the island effect of the Internet of Things data and the BIM data is broken, the original data of the physical resource is converted into the scenario data strongly associated with the specific task, the spatial, time and context three-level consistency is quantified, the explicit evaluation standard for the data reliability is provided, the dynamic association mechanism ensures the real-time binding of the resource and the task, the context consistency provides the weight basis for the subsequent dynamic modeling and abnormal diagnosis, and the analysis error caused by the disconnection between the data and the business is effectively avoided, thereby laying a data foundation for the precision of the whole method.

[0045] S200: Real-time construction of a dynamic context model based on the fusion perception data and the context consistency thereof.

[0046] In the embodiment of the application, a dynamic context model is constructed in real time based on the fusion perception data and the context consistency thereof. After the data fusion and the consistency quantification in S100, the scenario data strongly associated with the specific task is obtained, but these data are still scattered in the independent dimensions of the resource state, the task information and the personnel cooperation, and have not formed the overall scenario cognition covering all elements of the construction. In the prior art, the single-dimensional data cannot support the deep analysis of the resource, the task and the personnel cooperation relationship, and the subsequent abnormal diagnosis is easy to ignore the cross-dimension correlation factors. Therefore, this step needs to construct a dynamic context model including three layers of the physical, the business and the society, integrate the scattered data into the panoramic and associated scenario information, and provide a multi-dimensional and structured analysis basis for the abnormal identification and the root cause tracing.

[0047] The step S200 in the method provided in the embodiment of the application includes:

[0048] A three-layer dynamic context model is constructed, and the dynamic context model includes a physical context, a business context and a social context.

[0049] In the physical context layer, based on the fusion perception data and its context consistency, a real-time state snapshot table of all monitored physical resources is established, which at least records resource ID, type, current location, running state, currently associated task ID and context consistency of the fusion perception data;

[0050] In the business context layer, based on the task profile of the BIM model and the real-time construction progress data, a current active task queue is dynamically generated, and each task node in the current active task queue at least contains task ID, required resource type and quantity, process logic constraint and planned duration;

[0051] In the social context layer, based on historical collaboration data, a collaboration relationship graph representing the collaboration strength and mode between different roles on site is implicitly constructed through a social group discovery algorithm, and the historical collaboration data includes historical personnel positioning clustering data, historical communication records and historical task collaboration records.

[0052] Firstly, a three-layer dynamic context model is constructed, which includes physical context, business context and social context. The dynamic context model refers to a structured model that integrates three types of core information in the construction scene in real time, including physical resource state, business task requirements and personnel collaboration relationship. The physical context layer focuses on the actual state of resources, the business context layer focuses on the preset requirements of tasks, and the social context layer focuses on the collaboration rules of personnel, and the three layers of data are updated in real time. A layered architecture design is adopted, and the fusion perception data and context consistency output by S100, the task profile of the BIM model and the historical collaboration data are respectively accessed through data interfaces to build the association mapping rules of the three layers of data, ensuring real-time synchronization of data updates. For example, the real-time state data of tower crane B and construction personnel A is accessed in the physical context layer, the preset requirement data of task TZ-001 is accessed in the business context layer, and the historical collaboration data of construction personnel A, tower crane driver C and the like is accessed in the social context layer, and the three layers of data are associated through task ID (TZ-001), resource ID (TD-002, RY-001).

[0053] Secondly, in the physical context layer, based on the fusion perception data and its context consistency, a real-time state snapshot table of all monitored physical resources is established, which at least records resource ID, type, current position, running state, currently associated task ID and context consistency of the fusion perception data relied on. The real-time state snapshot table refers to a structured data table that records the key state of all monitored physical resources at fixed time intervals, and the data is dynamically refreshed with the change of resource state. The running state refers to the working state of the physical resource, such as the running / standby / failure of the device, the work / rest / leave of the personnel. With the fusion perception data and context consistency of S100 as input, the resource core state field is extracted according to the unique identification principle of resource ID, and snapshot records are generated at fixed time intervals and stored synchronously to the model database.

[0054] For example, on May 10, 2025, 10:01, the generated real-time state snapshot table contains the following key records: the construction personnel with resource ID RY-001 is currently located at (35.5m, 16.2m, 12.5m) in the unified coordinate system, the running state is work, the associated task ID is TZ-001, and the corresponding context consistency is 0.88; the tower crane with resource ID TD-002 is currently located at (35.2m, 18.7m, 12.3m), the running state is running, the associated task ID is TZ-001, and the context consistency is 0.90; the steel bar cutting machine with resource ID SB-003 is currently located at (32.1m, 17.3m, 12.0m), the running state is standby, the associated task ID is TZ-001, and the context consistency is 0.85.

[0055] Further, in the business context layer, based on the task profile of the BIM model and the real-time construction progress data, a current active task queue is dynamically generated, and each task node in the current active task queue at least contains task ID, required resource type and quantity, process logic constraint and planned duration. The task profile refers to the complete attribute information of a task stored in the BIM model, including resource demand, process requirement, time plan, etc. The current active task queue refers to a set of tasks that are currently in progress or in standby state, sorted by process logic. The process logic constraint refers to the construction sequence requirement that needs to be followed in task execution, such as steel binding needs to be performed after formwork completion. The profile information of all tasks is extracted from the BIM model, combined with real-time progress data of the construction site, and the tasks currently to be executed are selected to generate an active task queue sorted by process logic, and the task state is updated in real time.

[0056] For example, based on the real-time progress, the current 10-layer formwork task has been completed, and the generated current active task queue is sorted according to the process logic as follows: the first task ID is TZ-001, the task name is 10-layer steel bar binding, the required resource type and quantity are 3 construction personnel, 1 tower crane and 1 steel bar cutting machine, the process logic constraint is to be carried out after the completion of the 10-layer formwork, the planned construction period is May 10, 2025 08:00-18:00, and the current state is in progress; the second task ID is TZ-002, the task name is 10-layer formwork removal, the required resources are 2 construction personnel and 2 pry bars, the process logic constraint is to be carried out after the completion of the 10-layer steel bar binding, the planned construction period is May 11, 2025 08:00-12:00, and the current state is to be started; the third task ID is TZ-003, the task name is 10-layer concrete pouring, the required resources are 1 concrete pump truck and 4 construction personnel, the process logic constraint is to be carried out after the completion of the 10-layer steel bar binding acceptance, the planned construction period is May 11, 2025 14:00-17:00, and the current state is to be started.

[0057] Finally, in the social context layer, based on historical collaboration data, including historical personnel positioning clustering data, historical communication records and historical task collaboration records, a collaboration relationship graph representing the collaboration strength and collaboration mode between different roles on site is implicitly constructed through a community discovery algorithm. Historical collaboration data refers to the collaboration records of personnel participating in tasks on site over a period of time, including positioning clustering, communication interaction, task division and cooperation, etc. Positioning clustering refers to the same task area working together. The community discovery algorithm refers to an algorithm that identifies close collaboration personnel groups and collaboration modes by analyzing data such as interaction frequency and collaboration duration, such as the Louvain algorithm. The collaboration relationship graph refers to a visualization graph with personnel as nodes and collaboration strength as edges, with the thickness of the edges representing the collaboration strength, with thicker edges indicating closer collaboration. Collecting historical collaboration data for the past 3 months, analyzing the collaboration frequency, collaboration duration and number of tasks worked together between personnel through the community discovery algorithm, calculating the collaboration strength with a value range of [0, 1], and constructing a weighted and undirected collaboration relationship graph, which is updated in real time with new collaboration data.

[0058] For example, the collected historical collaboration data includes that construction personnel A (RY-001) and tower crane driver C (RY-004) have worked together on 8 steel bar binding tasks for the past 3 months, with a total collaboration duration of 120 hours; construction personnel A and steel worker D (RY-005) have worked together 6 times, with a collaboration duration of 80 hours. According to the Louvain algorithm, the collaboration strength between A and C is 0.92, and the collaboration strength between A and D is 0.78. In the constructed collaboration relationship graph, the connection between nodes RY-001 and RY-004 is the thickest, followed by the connection between RY-001 and RY-005, intuitively presenting the core collaboration relationship.

[0059] In the embodiment of the present application, the scattered resource state is structured and visualized through the real-time state snapshot table of the physical context layer, so that the real-time binding of resources and tasks can be quickly and conveniently queried; the process logic and resource requirement of the task are clarified through the activated task queue of the business context layer, so as to provide a basis for judging whether the resource meets the requirements of the task; the implicit collaboration rules among personnel are mined through the collaboration relationship graph of the social context layer, so as to fill the blank of ignoring the personnel collaboration factor in the prior art. The linkage integration of the three-layer data converts the isolated information of resources, tasks and personnel into panoramic scene cognition, provides a complete and structured analysis basis for subsequent cross-dimension correlation analysis and root cause tracing of abnormal events, and improves the comprehensiveness and accuracy of abnormal diagnosis.

[0060] S300: input the current state of the dynamic context model into a preset construction knowledge graph, identify an abnormal event, perform causal path tracing on the abnormal event, and output an abnormal diagnosis report containing at least a diagnosis confidence.

[0061] In the embodiment of the present application, the current state of the dynamic context model is input into a preset construction knowledge graph, an abnormal event is identified, and causal path tracing is performed on the abnormal event, and an abnormal diagnosis report containing at least a diagnosis confidence is output. The three-layer dynamic context model constructed through S200 has formed a panoramic scene snapshot of resources, tasks and personnel, but the snapshot only presents the current state of the scene and lacks the ability of correlation analysis with the knowledge in the construction field. The prior art can only trigger an alarm based on a single threshold, cannot judge the essence of the abnormality in combination with the knowledge in the fields of process logic and collaboration rules, and is more difficult to locate the root cause in a cross-dimension. Therefore, the preset construction knowledge graph needs to be introduced in this step, the abnormality is identified through state and knowledge comparison, the root cause is traced by means of causal reasoning, a structured report with confidence is output, and the technical pain point of only knowing the abnormality but not knowing the reason is solved.

[0062] The step S300 in the method provided by the embodiment of the present application comprises:

[0063] A construction knowledge graph is constructed in advance, the construction knowledge graph stores the knowledge in the construction field in the form of entity-entity relationship-attribute, and the entities contained in the construction knowledge graph at least include resource entities, state entities and task entities, and the relationships between the entities at least include causal relationships, process logic relationships and constraint relationships;

[0064] The current state of the dynamic context model is compared with the task portrait in the business context layer, a threshold-based rule engine is triggered, an abnormal event is identified, and the abnormal event includes resource loss, state deviation and progress deviation.

[0065] First, a construction knowledge graph is constructed in advance, which stores construction field knowledge in the form of entity-entity relationship-attribute, and contains at least resource entities, state entities and task entities, and the relationships between entities include at least causal relationships, process logic relationships and constraint relationships. The construction knowledge graph refers to a semantic network that stores construction field knowledge in the form of entity-entity relationship-attribute triple, and is the core knowledge base for intelligent reasoning. The resource entity refers to various physical entities participating in the operation on the construction site, such as tower cranes, construction personnel, steel cutting machines and concrete pump trucks. The state entity refers to the specific situation of an entity or a task, such as the running / failure / oil temperature of a device being too high, the progress of a task being normal / lagging, the on-site personnel being on duty / collaboration being smooth, etc. The task entity refers to a specific work unit divided in the construction process, such as steel binding, formwork removal and concrete pouring. The causal relationship refers to the causal association between entities, such as a cooling system failure causing high hydraulic oil temperature. The process logic relationship refers to the sequence requirement of task execution, such as steel binding, steel acceptance and concrete pouring. The constraint relationship refers to the threshold standard of entity operation or task execution, such as tower crane hydraulic oil temperature ≤ 45℃, steel binding progress ≥ 60% every 6 hours. The attributes include entity parameter thresholds and relationship base probabilities. The core knowledge elements are sorted by collecting construction industry specifications, historical cases of similar residential building projects in the past 5 years and the experience of 3 construction field experts; the collected knowledge is decomposed into entity-entity relationship-attribute triples, and the classification of each entity, the type of relationship and the attribute parameters are determined; the triple data is input into the knowledge graph construction tool Neo4j to establish the association index between entities and form a structured knowledge base that can support traversal reasoning.

[0066] For example, the core knowledge directly related to the TZ-001 task in the constructed construction knowledge graph includes: entity set: resource entity: tower crane TD, construction personnel RY, steel cutting machine SB; state entity: cooling system failure S4, high hydraulic oil temperature S1, progress lag S3, lifting capacity over limit S2; task entity: steel binding TZ, concrete pouring HZ; relationship set: causal relationship: S4-S1, base probability 0.92; S1-S3, base probability 0.90; S2-S1, base probability 0.85; process logic relationship: TZ-HZ; constraint relationship: TD hydraulic oil temperature ≤ 45℃, TZ task 6-hour progress ≥ 60%, TD lifting capacity ≤ 5t, TZ requires at least 3 construction personnel; attribute parameter: causal relationship S2-S1 base probability 0.92, constraint relationship TD hydraulic oil temperature threshold 45℃, etc.

[0067] Secondly, the current state of the dynamic context model is compared with the task portrait in the business context layer, triggering a threshold-based rule engine to identify abnormal events, including resource deficiency, state deviation, and progress deviation. The current state refers to the digital snapshot of the construction site at a specific time slice, which is not static data and is continuously updated with sensing data, and is composed of the fusion of physical, business, and social contexts. The rule engine is an automated judgment module that incorporates constraint relationships and process logic rules from the construction knowledge graph, and triggers abnormal identification through threshold comparison and logic verification. Abnormal events refer to situations where the current state violates the constraint relationships or process logic in the knowledge graph, including resource deficiency, state deviation, and progress deviation. Resource deficiency refers to the failure of required resources to be in place as required, state deviation refers to resource operating parameters exceeding threshold standards, and progress deviation refers to actual task progress falling behind plan requirements. The fusion data at a specific time slice is extracted from the dynamic context model, integrating physical, business, and social contexts. The extracted current state is compared item by item with the constraint relationships and task portrait in the construction knowledge graph, focusing on verifying whether resources meet requirements, parameters meet threshold values, and progress meets standards. The rule engine automatically determines the corresponding type of abnormal event based on the comparison results if there are situations that do not meet, exceed threshold values, or do not meet standards, and assigns a unique event ID.

[0068] For example, the current state is extracted: the current time is May 10, 2025 14:00, the TZ-001 task has been executed for 6 hours, and the planned progress needs to reach 60%. The current state of integration is as follows: physical context: tower crane B (TD-002) current position (35.2m, 18.7m, 12.3m), running state is running, hydraulic oil temperature is 52°C, context consistency is 0.90; 2 construction workers (RY-001, RY-006) are on duty, both are in working state, context consistency is 0.88 and 0.86 respectively; steel bar cutting machine (SB-003) is standby, context consistency is 0.85; business context: TZ-001 task plan progress 6 hours ≥ 60%, actual completed steel bar binding amount 2.5t, actual progress 30%; task required resources are 3 construction workers, 1 tower crane and 1 steel bar cutting machine; social context: construction worker RY-001 and tower crane driver RY-004 cooperation intensity 0.92, no cooperation conflict; rule matching comparison: constraint relationship comparison: tower crane B hydraulic oil temperature 52°C > knowledge graph threshold 45°C; TZ-001 actual progress 30% < plan threshold 60%; 2 construction workers on duty < requirement 3; process logic comparison: TZ-001 task has no pre-completion process, process execution has no deviation; abnormal trigger judgment: tower crane B hydraulic oil temperature exceeds threshold, judged as state deviation, abnormal event ID: YC-001; TZ-001 task progress is not up to standard, judged as progress deviation, abnormal event ID: YC-002; 1 construction worker is absent, but it does not affect the core work efficiency, rule engine judges it as non-key abnormality, not included in the formal abnormal event list.

[0069] After identifying the abnormal event, further root cause reasoning is performed:

[0070] Taking one or more abnormal events as starting nodes, the causal network is traversed and reasoned in the construction knowledge graph to find the most consistent causal path that can connect multiple abnormal events and meet the current physical context and social context as the graph reasoning result, wherein the context consistency of associated data is used as a weight adjustment factor for path search in the reasoning process, and the associated data is part of the data identified by the construction knowledge graph and used to obtain the reasoning result;

[0071] Based on the graph reasoning result, a structured abnormal diagnosis report is generated, which at least includes: abnormal description, associated abnormal event list, inferred root cause, visual display of causal evidence chain, inferred diagnosis confidence and context consistency of key support data.

[0072] First, one or more abnormal events are taken as starting nodes, and causal network traversal and reasoning are performed in the construction knowledge graph to find the most consistent causal path that can connect multiple abnormal events and meet the current physical context and social context, as the graph reasoning result. During the reasoning process, the context consistency degree of the associated data is taken as the weight adjustment factor of the path search. The associated data is part of the data identified by the construction knowledge graph and used to derive the reasoning result. Causal network traversal refers to the process of searching for cause-effect association paths from the starting state entity node along the causal relationship edges in the knowledge graph. The context constraint condition refers to the physical context and social context in the current dynamic context model, which is used to filter invalid paths that do not match the actual site. The weight adjustment factor, i.e., the context consistency degree of the associated data, is the higher the consistency degree, the greater the credibility weight of the corresponding path. The associated data is defined according to user notes, and here refers to the dynamically selected data set that supports causal hypothesis during the reasoning process, such as device sensor data and progress statistical data. The current physical context and social context are extracted from the dynamic context model as path filtering conditions; the context consistency degree corresponding to the associated data is extracted as the weight adjustment factor of the path search; the traversal rule is set: only paths that meet the context constraint and have a consistency degree of associated data ≥0.6 are retained.

[0073] For example, the context constraint condition: physical context: the current lifting capacity of tower crane B is 3.5t, which does not exceed the 5t threshold; the cooling system sensor feedback indicates that the heat dissipation efficiency is decreasing; social context: the collaboration intensity between construction personnel A and tower crane driver C is 0.92, and there is no scheduling conflict; weight adjustment factor: the consistency degree of tower crane B operating state data is 0.90, the consistency degree of cooling system sensor data is 0.88, and the consistency degree of progress statistical data is 0.85; associated data set: dynamically selected tower crane cooling system sensor data, tower crane operating parameter data, and TZ-001 task progress statistical data.

[0074] The causal network traversal and reasoning in the construction knowledge graph specifically includes:

[0075] Mapping the abnormal event to a state entity node in the construction knowledge graph;

[0076] Starting from the state entity node, bidirectional search is performed along the causal relationship edges on the potential multiple causal paths, and the current physical context and social context are introduced as filtering and weighting conditions for causal path search;

[0077] The confidence score of the multiple causal paths is calculated, and the causal path with the highest confidence score is selected as the root cause. The normalized confidence score is taken as the diagnostic confidence of this diagnosis.

[0078] First, abnormal events are mapped to state entity nodes in the construction knowledge graph. State entity nodes are core nodes in the construction knowledge graph used to represent abnormal states, and they have a one-to-one semantic correspondence with abnormal events, serving as the starting anchor points for causal path tracing. The core semantics of each abnormal event are extracted, such as "tower crane hydraulic oil temperature too high" or "rebar tying progress delayed." Based on a semantic matching algorithm, the corresponding state entity nodes are found in the construction knowledge graph, completing the semantic alignment between the abnormal events and the knowledge graph. For example, abnormal event YC-001 (tower crane B hydraulic oil temperature too high) is mapped to state entity node S1 in the knowledge graph; abnormal event YC-002 (TZ-001 task progress delayed) is mapped to state entity node S3 in the knowledge graph, establishing the starting points for subsequent causal traversal.

[0079] Secondly, starting from the state entity nodes, a bidirectional search is performed along the causal relationship edges to explore multiple potential causal paths. The current physical and social contexts are introduced as filtering and weighting conditions for the causal path search. The bidirectional search means starting from the initial state entity node, tracing back the preceding causes while simultaneously verifying subsequent effects, ensuring that the path covers multiple anomaly associations. Candidate causal paths are the set of potential causal paths selected after the bidirectional search that can connect multiple anomalous events and meet contextual constraints. Starting from the initial nodes S1 and S3, a bidirectional search is performed along the causal relationship edges of the knowledge graph; invalid paths are filtered using contextual constraints; and paths that can simultaneously connect two anomalous events are retained, forming a candidate causal path set. For example, forward tracing: trace from S1 (excessive oil temperature) to the preceding cause nodes S4 (cooling system failure) and S2 (excessive lifting capacity), filter S2 and retain S4; reverse verification: verify the subsequent impact from S1 (excessive oil temperature), find that S1 (excessive oil temperature) - S3 (schedule deviation), excessive oil temperature leads to a decrease in tower crane operating efficiency, which in turn drags down the schedule, and is connected to another abnormal event node S3; candidate causal path: finally select the only candidate path S4-S1-S3, which connects two abnormal events and meets all context constraints.

[0080] Furthermore, the confidence scores of multiple causal paths are calculated, and the causal path with the highest confidence score is selected as the root cause. The confidence scores are then normalized and used as the diagnostic confidence output for this diagnosis.

[0081] The calculation of confidence scores for multiple causal paths includes:

[0082] The prior confidence of the path is obtained by multiplying the preset basic probabilities of all causal edges on the causal path.

[0083] The evidence support degree coefficient is calculated by accumulating the weighted contribution values of all the associated data supporting each causal node on a certain causal path in the current dynamic context model, wherein the weighted contribution value is the product of the context consistency degree of the corresponding associated data and the preset relevance weight of the associated data to the causal node;

[0084] The confidence score of the causal path is calculated based on the path prior confidence and the evidence support degree coefficient.

[0085] First, the path prior confidence is calculated by multiplying all the preset base probabilities of the causal relationship edges on the causal path. The path prior confidence refers to the inherent credibility of the path calculated based on the preset base probabilities of the causal relationship edges in the construction knowledge graph, reflecting the universal occurrence probability of the causal logic in the construction field. The preset base probability refers to a value stored in advance in the construction knowledge graph, representing the possibility of occurrence of a certain causal relationship, with a value range of [0, 1], determined by historical project data statistics and expert evaluation. For example, the base probability of the causal relationship of cooling system failure-oil temperature being too high is 0.92, representing a 92% probability of occurrence of this causal relationship in past cases. A causal relationship edge is an associated link connecting two causal nodes, and each edge corresponds to a preset base probability. All the causal relationship edges on the target causal path are identified, the preset base probability of each causal relationship edge is extracted, and all the base probabilities are multiplied to obtain the path prior confidence. The smaller the product, the less likely the causal logic is to occur in the universal scenario.

[0086] For example, the target path 1 (core path): S4-S1-S3, contains two causal relationship edges: S4-S1: preset base probability 0.92, universal probability of cooling system failure leading to oil temperature being too high; S1-S3: preset base probability 0.90, universal probability of oil temperature being too high leading to schedule deviation; path prior confidence = 0.92 x 0.90 = 0.828; target path 2 (virtual comparison path): S5-S6-S3, contains two causal relationship edges: S5-S6: preset base probability 0.75, universal probability of poor personnel coordination leading to scheduling delay; S6-S3: preset base probability 0.80, universal probability of scheduling delay leading to schedule deviation; path prior confidence = 0.75 x 0.80 = 0.60.

[0087] Secondly, the evidence support degree coefficient is calculated by accumulating the weighted contribution values of all the association data supporting each causal node on a certain causal path in the current dynamic context model. The weighted contribution value is the product of the context consistency degree of the corresponding association data and the preset relevance weight of the association data to the causal node. The evidence support degree coefficient refers to the quantitative support strength of the association data to each node in the causal path based on the current field data. The greater the value, the more the current field data can confirm the rationality of the causal path. The weighted contribution value refers to the support strength of a single association data to a certain causal node, which is obtained by multiplying the context consistency degree and the relevance weight. The context consistency degree refers to the matching degree of the association data and the current task context calculated in step S100, with a value range of [0, 1], such as the consistency degree of 0.88 of the cooling system sensor data. The preset relevance weight is a value representing the strength of the association between the association data and the causal node, with a value range of [0, 1], and the stronger the association, the higher the weight, such as the weight of 0.8 of the cooling system data to S4 and the weight of 0.3 of the cooling system data to S1. The association data refers to all the data sets dynamically selected by the knowledge graph reasoning algorithm to support the causal path, such as the cooling system sensor data and the tower crane running state data. All the causal nodes on the target causal path are screened; the corresponding association data is matched for each node; the context consistency degree of each association data and the preset relevance weight of the corresponding node are extracted, and the product of the two is calculated, i.e. the weighted contribution value; the weighted contribution values of all the association data are accumulated, and the result is the evidence support degree coefficient of the path.

[0088] For example, the nodes of the target path 1 are matched with the association data: node S4: association data = cooling system sensor data, context consistency degree = 0.88, relevance weight = 0.8, weighted contribution value = 0.88 x 0.8 = 0.704; node S1: association data = tower crane running state data, context consistency degree = 0.90, relevance weight = 0.9, weighted contribution value = 0.90 x 0.9 = 0.81; node S3: association data = task progress statistical data, context consistency degree = 0.85, relevance weight = 0.85, weighted contribution value = 0.85 x 0.85 = 0.7225; evidence support degree coefficient = 0.704 + 0.81 + 0.7225 = 2.2365. Similarly, the evidence support degree coefficient of the target path 2 can be obtained, for example, 1.5375.

[0089] Then, the confidence score of the causal path is calculated based on the path prior confidence and the evidence support coefficient. The confidence score is a final credibility index that integrates the path prior confidence and the evidence support coefficient. The higher the score, the more credible the path. Confidence score = prior probability x (1 + evidence support coefficient). For example, target path 1 confidence score = 0.828 x (1 + 2.2365) ≈ 2.68, and target path 2 confidence score = 0.60 x (1 + 1.5375) ≈ 1.5225.

[0090] On this basis, the causal path with the highest confidence score is selected as the root cause, and the normalized confidence score is output as the diagnostic confidence of this diagnosis. Normalization means mapping the original confidence score to the 0-1 interval, which is convenient for intuitive judgment of credibility. For example, taking the maximum confidence score 2.68 as the benchmark, target path 1 confidence score = 2.68 / 2.68 ≈ 1.0, and after threshold correction combined with the actual scene, it is 0.94, and target path 2 confidence score = 1.5225 / 2.68 ≈ 0.568. Therefore, the diagnostic confidence output of this diagnosis is 0.94.

[0091] Finally, based on the reasoning result of the graph, a structured abnormal diagnosis report is generated, which at least includes: abnormal description, associated abnormal event list, inferred root cause, visual display of causal evidence chain, inferred diagnosis confidence and context consistency of key support data. The structured abnormal diagnosis report is a standardized report integrating reasoning results, evidence data and credibility evaluation, providing clear basis for decision making. The visual display of the causal evidence chain is a logical link between the root cause, intermediate abnormality and target abnormality presented in simple text. The key support data refers to the core evidence subset supporting the diagnosis conclusion. The context consistency of the key support data refers to the matching degree of the core evidence and the current task context, with a value of [0, 1]. The abnormal diagnosis report is, for example: abnormal description: TZ-001 (10-layer steel binding) started at 08:00 on May 10, 2025, and executed for 6 hours by 14:00, with actual progress of 30% (plan 60%); tower crane B (TD-002) hydraulic oil temperature 52℃ (threshold value 45℃). Associated abnormal event list: YC-001 (state deviation): tower crane B hydraulic oil temperature 52℃>45℃; YC-002 (progress deviation): TZ-001 task 6 hours progress 30% not up to plan. Inferred root cause: tower crane B cooling system failure leads to oil temperature over limit, equipment efficiency drops, and TZ-001 task progress is delayed. Causal evidence chain visualization: cooling system failure-tower crane oil temperature too high (YC-001)-tower crane efficiency decreases-TZ-001 progress lags (YC-002). Diagnosis confidence: 0.94. Key support data context consistency: tower crane B cooling system sensor data: 0.88; tower crane B running state data: 0.90; TZ-001 task progress statistical data: 0.85.

[0092] In the embodiments of the present application, by deeply integrating the field panoramic state of the dynamic context model with the construction field knowledge graph, not only the accurate identification of multiple types of abnormal events such as resources, progress and state is realized, but also the limitation of single abnormal independent judgment is broken through, the root cause is located by penetrating the surface phenomenon through causal path tracing, and the structured report containing abnormal description, causal evidence chain and key support data is generated, so that the association logic between abnormality and root cause is intuitive and traceable, providing a scientific and operable decision basis for construction management personnel, effectively improving the comprehensiveness, accuracy and intelligent level of construction site abnormal diagnosis, and helping to efficiently promote construction management and problem rectification.

[0093] S400: push the abnormal diagnosis report to the management personnel, receive feedback data, and optimize the reasoning rules of the construction knowledge graph using the feedback data and the abnormal diagnosis report.

[0094] In the embodiment of the present application, the abnormal diagnosis report is pushed to the management personnel, the feedback data is received, and the feedback data and the abnormal diagnosis report are used to optimize the reasoning rules of the construction knowledge graph. After the abnormal diagnosis report is generated through S300, the reasoning rules of the construction knowledge graph are still set based on the initial domain knowledge and historical data, and lack of dynamic adaptation of actual application scenarios. If long-term optimization is not combined with artificial feedback, it may lead to rule lag or misjudgment accumulation in subsequent diagnosis, affecting the continuous improvement of diagnosis accuracy. Therefore, S400 realizes the closed-loop mechanism of report pushing, feedback collecting and rule optimizing, so that the knowledge graph can be continuously iterated in actual application, and the reasoning rules are highly matched with the actual situation of the construction site.

[0095] The method provided in the embodiment of the present application comprises the following steps S400:

[0096] According to the abnormal diagnosis report, the push content comprises a reasoning process abstract and a diagnosis confidence prompt;

[0097] A diagnosis result feedback page is provided on the management terminal, and artificial confirmation information is collected as feedback data;

[0098] The feedback data and the diagnosis confidence of the corresponding abnormal diagnosis report are used to dynamically adjust the weight of the cause-effect relationship edge in the knowledge graph.

[0099] First, according to the abnormal diagnosis report, the push content comprises a reasoning process abstract and a diagnosis confidence prompt. The management terminal refers to the office equipment used by the management personnel in daily life, such as a mobile phone APP and a computer client. The reasoning process abstract refers to the core logic of the simplified cause-effect path, which avoids being lengthy and is convenient for quick browsing. The diagnosis confidence prompt refers to the highlighted normalized diagnosis confidence, which assists the management personnel in judging the priority. The corresponding management personnel is matched according to a preset rule; the core information of the abnormal diagnosis report is automatically extracted, and push content comprising a reasoning abstract, a confidence prompt and a complete report entry is generated; the push is realized through the message notification function of the management terminal, so that the management personnel can receive the push in time. For example, the abnormal diagnosis report of the task TZ-001 is pushed to the mobile phone APP of the project mechanical and electrical manager, and the push content is: “high-confidence warning. The progress of the 10-layer steel binding task (TZ-001) is lagging behind. The diagnosis conclusion is that the cooling system failure of the tower crane B (TD-002) causes the oil temperature to exceed the limit. The reasoning abstract is: cooling system failure-oil temperature too high-progress lag, diagnosis confidence: 0.94 (high confidence), click to view the complete report and feedback”.

[0100] Secondly, a diagnostic result feedback page is provided in the management terminal to collect manual confirmation information as feedback data. The diagnostic result feedback page refers to an interactive page built in the management terminal, which provides standardized feedback options. The feedback data refers to the manual confirmation or correction information submitted by the management personnel based on the actual on-site verification results. The manual confirmation information includes three core options: correct diagnosis, incorrect diagnosis, and partial correctness. The management personnel clicks the push message to enter the feedback page and view the complete abnormal diagnosis report; combined with the on-site verification results, the corresponding feedback options are selected: if correct, it is directly submitted; if the reason is corrected, the correct root cause is selected from the preset options or customized; the feedback data is automatically collected and stored in association with the corresponding abnormal diagnosis report. For example, after the mechanical and electrical supervisor inspects the site, he finds that tower crane B indeed has a cooling system failure, which is consistent with the diagnostic conclusion, and selects “diagnosis correct” on the feedback page and submits it; if the actual root cause is found to be that the tower crane driver violated the operation rules, which led to the high oil temperature, not a cooling system failure, then select “diagnosis error” and check the corresponding option of “violation of operation rules-high oil temperature” from the preset reason list, and submit the corrected feedback data.

[0101] Further, the weights of the causal relationship edges in the knowledge graph are dynamically adjusted using the feedback data and the diagnostic confidence of the corresponding abnormal diagnosis report.

[0102] Among them, the weights of the causal relationship edges in the knowledge graph are dynamically adjusted using the feedback data and the diagnostic confidence of the corresponding abnormal diagnosis report, including:

[0103] A feedback case library is established to store the abnormal diagnosis report, feedback data and diagnostic confidence corresponding to each diagnosis;

[0104] When the feedback data confirms the correct diagnosis, the weights of each causal relationship edge in the causal path on which the correct diagnosis depends are increased;

[0105] When the feedback data corrects other reasons, the weights of each causal relationship edge in the causal path corresponding to the corrected reason are increased, and the amplitude of the weight adjustment of each causal relationship edge is differentially controlled according to the diagnostic confidence.

[0106] First, a feedback case library is established to store the abnormal diagnosis report, feedback data and diagnostic confidence corresponding to each diagnosis. The feedback case library is a structured database that stores diagnostic full-link data, which is used for rule optimization traceability. The abnormal diagnosis report of each diagnosis, the feedback data submitted by the management personnel, and the corresponding diagnostic confidence are stored in association according to the principle of one case per record, forming a reusable optimization data source. For example, the cooling system failure caused by progress lag diagnosis report of task TZ-001, the feedback data of the mechanical and electrical supervisor, and the diagnostic confidence of 0.94 are stored to establish a complete case record.

[0107] Secondly, when the feedback data confirms that the diagnosis is correct, the weight of each causal relationship edge in the causal path on which the diagnosis depends is increased. The causal path on which the diagnosis depends is extracted, and the weight of each causal relationship edge in the path is increased by a fixed proportion to strengthen the reasoning priority of the path. For example, feedback confirms that S4-S1-S3 is correct, and the causal relationship edge weights of S4-S1 and S1-S3 in the path are increased.

[0108] Finally, when the feedback data is corrected to other reasons, the weight of each causal relationship edge in the causal path corresponding to the corrected reason is increased, and the amplitude of the weight adjustment of each causal relationship edge is differentially controlled according to the diagnosis confidence. Differential control refers to adjusting the weight variation amplitude according to the original diagnosis confidence. The higher the confidence of the error case, the greater the amplitude of adjustment. The new causal path corresponding to the corrected reason is identified, and the weight of each causal relationship edge in the path is increased. The causal relationship edge weight of the original error path is lowered, and the higher the original diagnosis confidence, the greater the error, and the greater the lowering amplitude. For example, feedback corrects the reason as illegal operation-oil temperature too high-progress deviation, and the causal relationship edge weight of the new path is increased. The original diagnosis confidence is 0.94 (high confidence) but error, and the weights of S4-S1 and S1-S3 in the original path are lowered.

[0109] In the embodiments of the present application, a closed-loop mechanism of diagnosis report pushing, feedback collection, and knowledge graph optimization is constructed. First, abnormal diagnosis reports are accurately pushed to management personnel, and a convenient feedback entrance is provided to collect manual confirmation or correction information. Then, a feedback case library is established to store complete diagnosis link data. For the diagnosis results confirmed to be correct, the reasoning weight of the corresponding causal path is strengthened. For the case where the diagnosis is corrected to other reasons, the weight of the new causal path is increased, and the original diagnosis confidence is differentially adjusted. For the case where the high confidence is error, the weight is corrected and the rule is reviewed with a greater amplitude. The artificial experience and the actual situation on site are continuously fed back to the construction knowledge graph, and the reasoning rules are continuously optimized to make the subsequent abnormal diagnosis more consistent with the actual situation on site, and the self-adaptation ability and the intelligent level are gradually improved.

[0110] The embodiments of the present application achieve the following technical effects through the specific implementation manner described above.

[0111] The application provides a BIM-based construction resource real-time management method and system. Through the whole-process link of data processing and context matching, panoramic dynamic model construction, intelligent abnormal diagnosis and root cause tracing, knowledge graph closed-loop optimization, the diagnostic foundation is first built based on multi-dimensional data, then the panoramic state of the construction site is presented through the dynamic context model, then the precise identification of abnormal events, multi-abnormal correlation analysis and root cause penetrating tracing are realized combined with the construction field knowledge graph, a structured diagnostic report with clear logic and sufficient evidence is generated, and finally the reasoning rules are continuously optimized relying on artificial feedback closed loop to continuously strengthen the adaptation ability of the system to the actual scene of the construction site. The whole process realizes the transformation from passive investigation to active early warning, from subjective judgment to data-driven, from static rules to dynamic iteration, and comprehensively improves the accuracy, comprehensiveness and intelligent level of construction abnormal diagnosis, and provides efficient and reliable technical support for construction management.

[0112] In an embodiment, as shown in Figure 2 The application provides a BIM-based construction resource real-time management system, which comprises:

[0113] A data perception fusion module 11 is configured to acquire real-time Internet of Things data streams and BIM model data of a construction site, dynamically associate physical resources in the Internet of Things data streams to task components of the BIM model, and generate a context consistency degree of the fusion perception data and the current task;

[0114] A dynamic context modeling module 12 is configured to construct a dynamic context model in real time based on the fusion perception data and the context consistency degree thereof;

[0115] An abnormal diagnosis and tracing module 13 is configured to input a current state of the dynamic context model into a preset construction knowledge graph, identify abnormal events, perform causal path tracing on the abnormal events, and output an abnormal diagnosis report containing at least a diagnosis confidence degree;

[0116] A graph feedback optimization module 14 is configured to push the abnormal diagnosis report to a management personnel, receive feedback data, and optimize reasoning rules of the construction knowledge graph by using the feedback data and the abnormal diagnosis report.

[0117] In an embodiment, the data perception fusion module 11 is further configured to:

[0118] Acquire Internet of Things data streams of the construction site, wherein the Internet of Things data streams at least include real-time coordinates of personnel and Internet of Things devices acquired through positioning tags, operating parameters of mechanical equipment acquired through state sensors, and temperature and humidity data acquired through environmental sensors;

[0119] Extract the spatial boundary and the task plan time window corresponding to the task from the BIM model;

[0120] establishing a unified space-time coordinate system at the construction site, converting and aligning the real-time coordinates of the Internet of Things devices, the BIM model coordinates and the actual coordinates of the construction site to form a space-time mapping relationship;

[0121] comparing the resource locations of various physical resources in the Internet of Things data stream with the space boundaries of the activated tasks in the BIM model in real time, when the physical resource enters the space boundary of the corresponding task and is within the task planning time window, a dynamic association between the physical resource and the BIM model is established;

[0122] based on the resource location of the physical resource and the space boundary, calculating the coincidence ratio of the resource location of the physical resource and the space boundary as the spatial consistency degree;

[0123] based on the time when the physical resource enters the space boundary and the task planning time window, calculating the time coincidence degree;

[0124] based on the spatial consistency degree and the time coincidence degree, calculating the context consistency degree of the corresponding fusion perception data.

[0125] In one embodiment, the dynamic context modeling module 12 is further configured to:

[0126] constructing a three-layer dynamic context model, the dynamic context model comprising a physical context, a business context and a social context;

[0127] at the physical context layer, based on the fusion perception data and its context consistency degree, establishing a real-time state snapshot table of all monitored physical resources, the real-time state snapshot table recording at least resource ID, type, current location, running state, currently associated task ID and context consistency degree of the fusion perception data;

[0128] at the business context layer, based on the task profile of the BIM model and the real-time construction progress data, dynamically generating a current activated task queue, each task node in the current activated task queue at least containing task ID, required resource type and quantity, process logic constraint and planned duration;

[0129] at the social context layer, based on historical collaboration data, constructing a collaboration relationship graph representing the collaboration strength and mode between different roles on site through a social group discovery algorithm, the historical collaboration data including historical personnel positioning clustering data, historical communication records and historical task collaboration records.

[0130] In one embodiment, the anomaly diagnosis and tracing module 13 is further configured to:

[0131] A construction knowledge graph is pre-constructed, which stores knowledge in the construction field in the form of entity-entity relationship-attribute, and contains at least resource entities, state entities and task entities, and the relationships between entities include at least causal relationships, process logic relationships and constraint relationships.

[0132] The current state of the dynamic context model is compared with the task portrait in the business context layer, a threshold-based rule engine is triggered to identify abnormal events, and the abnormal events include resource loss, state deviation and progress deviation.

[0133] After identifying abnormal events, further root cause reasoning is performed.

[0134] Taking one or more abnormal events as starting nodes, causal network traversal and reasoning are performed in the construction knowledge graph to find a causal path that can connect multiple abnormal events and meet the current physical context and social context, as a graph reasoning result, wherein the context consistency degree of associated data is used as a weight adjustment factor for path search during reasoning, and the associated data is part of the data identified by the construction knowledge graph and used to obtain the reasoning result.

[0135] Based on the graph reasoning result, a structured abnormal diagnosis report is generated, which includes at least abnormal description, associated abnormal event list, inferred root cause, visualized display of causal evidence chain, inferred diagnosis confidence and context consistency degree of key support data.

[0136] In the construction knowledge graph, causal network traversal and reasoning are performed, specifically including:

[0137] Map the abnormal event to a state entity node in the construction knowledge graph;

[0138] Starting from the state entity node, bidirectional search is performed along the causal relationship edges for potential multiple causal paths, and the current physical context and social context are introduced as filtering and weighting conditions for causal path search;

[0139] Calculate the confidence score of multiple causal paths, select the causal path with the highest confidence score as the root cause, and normalize the confidence score as the diagnosis confidence of this diagnosis.

[0140] Wherein, the confidence score of multiple causal paths is calculated, including:

[0141] The path prior confidence is obtained by multiplying the preset base probability of all causal relationship edges in the causal path;

[0142] The evidence support degree coefficient is calculated by accumulating the weighted contribution values of all the associated data supporting each causal node on a certain causal path in the current dynamic context model, the weighted contribution value being the product of the context consistency degree of the corresponding associated data and the preset correlation degree weight of the associated data to the causal node;

[0143] The confidence score of the causal path is calculated based on the path prior confidence and the evidence support degree coefficient.

[0144] In one embodiment, the knowledge graph feedback optimization module 14 is further configured to:

[0145] The abnormality diagnosis report is automatically pushed to the management terminal, and the push content includes a reasoning process summary and a diagnosis confidence prompt;

[0146] A diagnosis result feedback page is provided on the management terminal, and manual confirmation information is collected as feedback data;

[0147] The feedback data and the diagnosis confidence of the corresponding abnormality diagnosis report are used to dynamically adjust the weights of the causal relationship edges in the knowledge graph.

[0148] The feedback data and the diagnosis confidence of the corresponding abnormality diagnosis report are used to dynamically adjust the weights of the causal relationship edges in the knowledge graph, including:

[0149] A feedback case library is established to store the abnormality diagnosis report, the feedback data, and the diagnosis confidence corresponding to each diagnosis;

[0150] When the feedback data confirms that the diagnosis is correct, the weights of the causal relationship edges in the causal path on which the diagnosis is correct are increased;

[0151] When the feedback data is corrected to other reasons, the weights of the causal relationship edges in the causal path corresponding to the corrected reasons are increased, and the weights of the causal relationship edges are differentially controlled according to the diagnosis confidence.

[0152] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes a specific embodiment of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0153] The above only describes the preferred embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0154] The specification and drawings are, of course, to be regarded in an illustrative rather than a restrictive sense. It is to be understood that the application is capable of various modifications and alternative constructions and that certain of the above described elements are capable of substitutions therefor. As such, the application is not to be taken as limited to the drawings and description provided herein, but is capable of numerous modifications and adaptations.

Claims

1. A BIM-based real-time management method for construction resources, characterized in that, The method includes: Acquire real-time IoT data streams and BIM model data from the construction site, dynamically associate physical resources in the IoT data streams with task components in the BIM model, and generate contextual consistency between fused sensing data and the current task. Based on the fused sensing data and its contextual consistency, a dynamic contextual model is constructed in real time. The current state of the dynamic context model is input into a preset construction knowledge graph to identify abnormal events, trace the causal path of the abnormal events, and output an abnormal diagnosis report that includes at least the diagnostic confidence level. The anomaly diagnosis report is pushed to the management personnel, feedback data is received, and the reasoning rules of the construction knowledge graph are optimized using the feedback data and the anomaly diagnosis report. This includes acquiring real-time IoT data streams and BIM model data from the construction site, dynamically associating physical resources from the IoT data streams with task components of the BIM model, and generating contextual consistency between the fused sensing data and the current task, including: The IoT data stream collected at the construction site includes at least: real-time coordinates of personnel and IoT devices obtained through positioning tags, operating parameters of mechanical equipment obtained through status sensors, and temperature and humidity data obtained through environmental sensors. Extract the spatial boundaries and task planning time windows of the corresponding tasks from the BIM model; Establish a unified spatiotemporal coordinate system for the construction site, and convert and align the real-time coordinates of IoT devices, BIM model coordinates, and actual coordinates of the construction site to form a spatiotemporal mapping relationship; Real-time comparison of the resource locations of various physical resources in the IoT data stream with the spatial boundaries of activated tasks in the BIM model. When a physical resource enters the spatial boundary of the corresponding task and is within the task's planned time window, a dynamic association between the physical resource and the BIM model is established. Based on the resource location of the physical resource and the spatial boundary, the overlap ratio between the resource location of the physical resource and the spatial boundary is calculated as the spatial consistency. Calculate the time fit based on the time when physical resources enter the space boundary and the time window of the mission plan; Based on spatial consistency and temporal coherence, the contextual consistency of the corresponding fused sensing data is calculated.

2. The BIM-based real-time construction resource management method according to claim 1, characterized in that, Based on the fused sensing data and its contextual consistency, a dynamic contextual model is constructed in real time, including: A three-layer dynamic context model is constructed, which includes physical context, business context, and social context. At the physical context layer, based on the fused sensing data and its context consistency, a real-time status snapshot table of all monitored physical resources is established. The real-time status snapshot table records at least the resource ID, type, current location, running status, currently associated task ID, and the context consistency of the fused sensing data on which it is based. At the business context layer, based on the task profile of the BIM model and real-time construction progress data, the currently active task queue is dynamically generated. Each task node in the currently active task queue contains at least the task ID, the required resource type and quantity, process logic constraints, and the planned duration. At the social context layer, based on historical collaboration data, a collaboration relationship graph representing the collaboration intensity and collaboration mode among different roles on site is implicitly constructed through community discovery algorithms. The historical collaboration data includes historical personnel location clustering data, historical communication records, and historical task collaboration records.

3. The BIM-based real-time construction resource management method according to claim 1, characterized in that, The current state of the dynamic context model is input into a preset construction knowledge graph to identify abnormal events, trace the causal paths of these events, and output an anomaly diagnosis report that includes at least diagnostic confidence levels, including: A construction knowledge graph is pre-constructed. The construction knowledge graph stores construction domain knowledge in the form of entity-entity relationship-attribute. The entities it contains include at least resource entities, state entities, and task entities. The relationships between entities include at least causal relationships, process logic relationships, and constraint relationships. The current state of the dynamic context model is compared with the task profile in the business context layer, triggering a threshold-based rule engine to identify abnormal events, including resource missing, state deviation, and progress deviation.

4. The BIM-based real-time construction resource management method according to claim 3, characterized in that, After identifying the anomalous event, further root cause reasoning is performed: Starting with one or more abnormal events, causal network traversal and reasoning are performed in the construction knowledge graph to find the most consistent causal path that can connect multiple abnormal events and conforms to the current physical and social context. This path is used as the graph reasoning result. In the reasoning process, the contextual consistency of the associated data is used as a weight adjustment factor for the path search. The associated data is the part of the data that is identified by the construction knowledge graph and used to derive the reasoning result. Based on the graph reasoning results, a structured anomaly diagnosis report is generated. The anomaly diagnosis report includes at least: anomaly description, a list of associated anomaly events, the inferred root cause, a visual representation of the causal evidence chain, the diagnostic confidence of the inference, and the contextual consistency of key supporting data.

5. A BIM-based real-time construction resource management method according to claim 4, characterized in that, The causal network traversal and reasoning in the construction knowledge graph specifically includes: Map abnormal events to state entity nodes in the construction knowledge graph; Starting from the state entity node, a bidirectional search is performed along the causal relationship edge for multiple potential causal paths, and the current physical context and social context are introduced as filtering and weighting conditions for the causal path search. Calculate the confidence scores of multiple causal paths, select the causal path with the highest confidence score as the root cause, and normalize the confidence scores as the diagnostic confidence output for this diagnosis.

6. The BIM-based real-time construction resource management method according to claim 5, characterized in that, Calculate confidence scores for multiple causal paths, including: The prior confidence of the path is obtained by multiplying the preset basic probabilities of all causal edges on the causal path. The evidence support coefficient is calculated by accumulating the weighted contribution values ​​of all associated data supporting each causal node on a certain causal path in the current dynamic context model. The weighted contribution value is the product of the contextual consistency of the corresponding associated data and the preset relevance weight of the associated data to the causal node. The confidence score of the causal path is calculated based on the prior confidence of the path and the evidence support coefficient.

7. The BIM-based real-time construction resource management method according to claim 1, characterized in that, The anomaly diagnosis report is pushed to management personnel, feedback data is received, and the inference rules of the construction knowledge graph are optimized using the feedback data and the anomaly diagnosis report, including: The abnormal diagnosis report is automatically pushed to the management terminal. The pushed content includes a summary of the reasoning process and a diagnostic confidence level prompt. The management terminal provides a diagnostic result feedback page and collects manual confirmation information as feedback data. By utilizing feedback data and the diagnostic confidence of corresponding anomaly diagnostic reports, the weights of causal relationship edges in the knowledge graph are dynamically adjusted.

8. The BIM-based real-time construction resource management method according to claim 1, characterized in that, By utilizing feedback data and the diagnostic confidence levels of corresponding anomaly diagnostic reports, the weights of causal relationship edges in the knowledge graph are dynamically adjusted, including: Establish a feedback case library to store abnormal diagnosis reports, feedback data, and diagnostic confidence levels for each diagnosis. When feedback data confirms that the diagnosis is correct, increase the weight of each causal relationship edge in the causal path on which the diagnosis depends. When the feedback data is corrected to other causes, the weight of each causal relationship edge in the causal path corresponding to the corrected cause is increased, and the magnitude of the weight adjustment of each causal relationship edge is differentiated according to the diagnostic confidence level.

9. A BIM-based real-time construction resource management system, characterized in that, The system is used to implement the BIM-based real-time construction resource management method according to any one of claims 1-8, the system comprising: The data perception and fusion module is used to acquire real-time IoT data streams and BIM model data from the construction site, dynamically associate physical resources in the IoT data streams with task components in the BIM model, and generate the contextual consistency between the fused perception data and the current task. The dynamic context modeling module is used to construct a dynamic context model in real time based on the fused sensing data and its context consistency. The anomaly diagnosis and tracing module is used to input the current state of the dynamic context model into a preset construction knowledge graph, identify abnormal events, trace the causal path of the abnormal events, and output an anomaly diagnosis report that includes at least the diagnostic confidence level. The graph feedback optimization module is used to push anomaly diagnosis reports to management personnel, receive feedback data, and optimize the inference rules of the construction knowledge graph using the feedback data and anomaly diagnosis reports.

Images