A construction risk management system

Abstract

This application discloses a construction risk management system, belonging to the field of construction risk management technology. It includes: a risk point pre-setting module for pre-setting project risk points and obtaining a risk database; an information processing module for acquiring target image information corresponding to the construction object and establishing a geometric construction model of the construction object; the geometric construction model and the construction object have a mapping relationship; the geometric construction model is semantically segmented, dividing it into multiple regional feature units, each associated with material properties, construction time windows, and equipment deployment information; a spatiotemporal correlation matching algorithm is used to match each regional feature unit with the risk database, matching risk point coordinate deviation thresholds in the spatial dimension and construction progress deviation thresholds in the temporal dimension. When the two-dimensional matching is successful, it is determined to be a valid risk point. This invention enhances the safety and controllability of construction.

Description

Technical Field

[0001] This application relates to a construction risk management system, belonging to the field of construction risk management technology. Background Technology

[0002] Currently, during the construction preparation phase, insufficient risk identification based on construction drawings and the construction environment, coupled with reliance on traditional experience-based judgment, easily leads to gaps in risk identification and inadequate construction preparation. During construction, risk monitoring personnel conduct risk assessments through monitoring or on-site inspections, which has limited coverage. The risk identification process is inefficient due to limitations in professional technical capabilities. After risks are identified, the process of issuing instructions and communicating to mitigate them is lengthy and inefficient. Project construction supervision is challenging and prone to accidents due to inadequate oversight. Existing solutions involve deploying monitoring systems during construction, with supervisors observing the site and conducting manual inspections to visually assess risks. This approach is highly subjective and limited by the supervisors' professional technical capabilities and experience, making it difficult to efficiently complete the risk identification and rectification process. Summary of the Invention

[0003] According to one aspect of this application, a construction risk management system is provided that significantly enhances the safety and controllability of construction.

[0004] A construction risk management system, characterized in that it includes:

[0005] The risk point pre-setting module uses project construction drawings, construction plans and construction schemes to identify safety risks during the construction process, and combines them with the system database to pre-set project risk points, thus obtaining a risk database.

[0006] The information processing module acquires target image information corresponding to the building construction object and establishes a geometric construction model of the building construction object. The geometric construction model has a mapping relationship with the building construction object. The geometric construction model is semantically segmented and divided into multiple regional feature units. Each regional feature unit is associated with material properties, construction time windows, and equipment deployment information. A spatiotemporal correlation matching algorithm is used to match each regional feature unit with a risk database. The risk point coordinate deviation threshold is matched in the spatial dimension, and the construction progress deviation threshold is matched in the temporal dimension. When the two-dimensional matching is successful, it is determined to be a valid risk point.

[0007] The controller receives data from the risk point preset module and the information processing module, and automatically configures corresponding risk management decisions based on the preset information in the risk database and the identification and judgment results of the risk points.

[0008] The risk management module, coupled with the controller, enables automated control of the entire risk management process, executes risk management decisions, and completes risk management.

[0009] Specifically, the risk identification device performs risk point matching between the regional feature unit and the risk database. When the regional feature unit fails to match the risk database, the sample reinforcement learning process is initiated. The feature vector of the unidentified scene and the expert manual annotation result are input into the risk database to generate identification rules adapted to the new risk type. After successful matching, the spatiotemporal evolution features of the risk point are automatically extracted, and the meta-learning parameters of the risk identification process are updated.

[0010] Furthermore, before performing risk point matching on the regional feature units and the risk database, the process also includes:

[0011] The regional feature units are classified by type, and the priority of risk point identification is set;

[0012] According to different security levels, the regional feature units are divided into: security risk units, behavioral risk units, and environmental risk units. The security risk units are set as the first risk priority, the behavioral risk units are set as the second risk priority, and the environmental risk units are set as the third risk priority.

[0013] Furthermore, prioritize risk identification, including:

[0014] Based on the security levels of historical data in the risk database, the security levels of the risk points in the regional feature units are obtained.

[0015] The historical data have a set of associated attributes, which records the attributes associated with the historical data. Based on the attributes and the attributes of the historical data, the association between the risk point and the historical data is obtained.

[0016] Complete the identification and matching of the priority of the aforementioned risk points.

[0017] Furthermore, obtaining the correlation between the risk points and the historical data includes:

[0018] When the attribute of the risk point is in the set of associated attributes, calculate the correlation degree between the risk point and the historical data;

[0019] When the correlation is higher than a preset threshold, it is determined that the risk point is correlated with the corresponding historical data.

[0020] Furthermore, a spatiotemporal correlation matching algorithm is used to match the feature units of each region with the risk database, including:

[0021] Spatial dimension matching:

[0022] Use spatial indexes to index coordinates in the risk database to speed up proximity queries;

[0023] For the center point of each regional feature unit, calculate the distance between the center point and the risk point:

[0024]

[0025] Where (x1,y1,z1) are the spatial coordinates of the center point of the regional feature unit; (x2,y2,z2) are the actual spatial coordinates of the risk point;

[0026] Risk points exceeding the distance deviation threshold are selected to form a spatial candidate set;

[0027] Time dimension matching:

[0028] Obtain the current construction stage and actual time progress;

[0029] For each risk point's time window, calculate the deviation between the current time and the planned time:

[0030] ΔT=|Current time -Plan time |

[0031] Among them, Current time Current time; Plan time For planned time;

[0032] Risk points exceeding the time deviation threshold are selected to form a time candidate set;

[0033] Two-dimensional joint matching:

[0034] The intersection of the spatial candidate set and the temporal candidate set is used to obtain the risk points that simultaneously meet the conditions.

[0035] If the intersection is not empty, it is determined to be a valid risk point;

[0036] Among them, the matched risk points are sorted by spatial distance and time deviation, and the urgent risk points are dealt with first.

[0037] Furthermore, the risk identification device has a built-in risk identification model;

[0038] The target image information corresponding to the construction object is input into the risk identification model for risk identification, and the risk identification result is generated and sent to the risk handling module. At the same time, when the probability of risk determination exceeds 80% during risk identification, the risk is determined to have occurred and a risk warning is issued.

[0039] When a risk point matching is completed, the identification results and identification data of the identified risk points are input into the risk identification model as training samples to optimize the risk identification model.

[0040] Furthermore, the mapping relationship set includes: at least one of the regional feature units, and the correspondence between the regional feature units and the building construction objects.

[0041] Furthermore, the risk database also allows for the addition or modification of project risk points, enabling real-time adjustments to risk conditions.

[0042] Furthermore, the correspondence between the regional feature units and the building construction object includes:

[0043] Based on the project risk points and the geometric construction model, the pre-defined risk point locations {M = M1, M2, M3, ..., Mi, ..., Mn} and the corresponding project risk point locations {N = N1, N2, N3, ..., Ni, ..., Nn} on the geometric construction model are obtained respectively; among them, the pre-defined risk point location M is clear and specific when it is set, and the corresponding project risk point location N can be found in the geometric construction model.

[0044] Furthermore, the security risk unit identifies risk points according to the first identification rule;

[0045] The behavioral risk unit identifies risk points according to the second identification rule;

[0046] The environmental risk unit identifies risk points according to the third identification rule.

[0047] The beneficial effects that this application can produce include:

[0048] This application provides a construction risk management system that integrates construction drawings, plans, and schemes through a risk point pre-setting module, and constructs a risk database based on historical data to achieve systematic prediction of risks throughout the construction cycle. The information processing module, relying on geometric construction models and spatiotemporal correlation matching algorithms, accurately identifies effective risk points in both spatial and temporal dimensions, effectively avoiding the limitations of single-dimensional judgment and significantly improving the accuracy and real-time performance of risk identification. The controller automatically generates risk management decisions based on the risk database and real-time identification results. The risk handling module, through deep coupling with the controller, achieves fully automated execution of the entire process, forming a closed-loop management of "identification-decision-handling," significantly reducing the cost of manual intervention, improving risk response speed, and ensuring the standardization and efficiency of construction safety management. Simultaneously, a sample reinforcement learning mechanism is introduced. When unidentified risk scenarios occur, dynamic identification rules are generated through collaborative learning of feature vectors and expert annotations, and the spatiotemporal evolution features of risks are automatically extracted to optimize meta-learning parameters. This enables the system to adapt to risk evolution, continuously expand the coverage of risk types, and significantly improve the generalization of risk identification in complex construction scenarios. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the structure of a construction risk management system according to one embodiment of this application. Detailed Implementation

[0050] The present application is described in detail below with reference to the embodiments, but the present application is not limited to these embodiments.

[0051] See Figure 1 A construction risk management system, characterized in that it includes:

[0052] The risk point pre-setting module uses project construction drawings, construction plans and construction schemes to identify safety risks during the construction process, and combines them with the system database to pre-set project risk points, thus obtaining a risk database.

[0053] This process involves using project construction drawings to conduct a detailed analysis of the project's construction structure, technological processes, and equipment layout. Combined with the construction plan, key information such as construction timelines, personnel allocation, and material supply is identified. Through this comprehensive analysis of construction drawings and the construction plan, potential safety risks during construction are preliminarily identified. Simultaneously, based on the risk identification results from the construction drawings and plan, and risk data in the system database, project risk points are pre-defined. The pre-defined risk points should consider multiple aspects, including the type, location, and scope of impact, to ensure accuracy and comprehensiveness. These pre-defined risk points are then entered into the system database, forming a risk database. The risk database should contain detailed information about each risk point, such as risk type, location, scope of impact, and control measures. The construction of this risk database enables comprehensive management and monitoring of project risks.

[0054] The information processing module acquires target image information corresponding to the building construction object and establishes a geometric construction model of the building construction object. The geometric construction model has a mapping relationship with the building construction object. The geometric construction model is semantically segmented and divided into multiple regional feature units. Each regional feature unit is associated with material properties, construction time windows, and equipment deployment information. A spatiotemporal correlation matching algorithm is used to match each regional feature unit with a risk database. The risk point coordinate deviation threshold is matched in the spatial dimension, and the construction progress deviation threshold is matched in the temporal dimension. When the two-dimensional matching is successful, it is determined to be a valid risk point.

[0055] The controller receives data from the risk point preset module and the information processing module, and automatically configures corresponding risk management decisions based on the preset information in the risk database and the identification and judgment results of the risk points.

[0056] As the core control unit of the entire system, the controller receives data from the risk point pre-setting module and the information processing module. Based on the pre-set information in the risk database and the identification and judgment results of the risk points, the controller automatically configures corresponding risk management decisions. The controller executes risk handling instructions, such as initiating preventive measures, adjusting construction plans, and allocating resources, to address potential safety risks. Throughout the process, the controller continuously monitors the execution of risk management decisions and makes flexible adjustments based on actual conditions to ensure the effectiveness and timeliness of risk management.

[0057] The risk response module, coupled with the controller, automates the entire risk management process, executes risk management decisions, and completes risk response. Under the unified scheduling of the controller, the risk response module is responsible for the specific execution of risk management decisions. Based on the controller's instructions, the risk response module may be involved in multiple aspects, including the implementation of preventative measures, the initiation of emergency responses, and the adjustment of resource allocation. Through tight integration with the controller, the risk response module achieves automated control of the entire risk management process, improving management efficiency and accuracy.

[0058] Specifically, high-precision image acquisition can be achieved through technologies such as high-resolution cameras, drone aerial photography, or satellite remote sensing. A drone equipped with a Sony α7R IV (61MP) can be used for oblique photography, with a forward overlap of ≥80% and a lateral overlap of ≥70%. Ground-based supplementary photography uses a 360° panoramic camera (Insta360 Pro 2) to capture details in blind spots. Simultaneously, laser scanning fusion is performed, acquiring millimeter-level point clouds (point density ≥500pt / m²) through ground-based laser scanning (Leica RTC360). 2The system utilizes a vehicle-mounted mobile scanning system to supplement surrounding data and acquire high-definition image information of the construction object. This high-definition image information includes key elements such as the overall structure of the construction object, construction progress, and surrounding environment. Advanced image processing technology and 3D modeling software are used to process and analyze the acquired target image information, thereby establishing a geometric construction model of the construction object. In establishing the geometric construction model, the collected data is preprocessed and then fused into a hybrid point cloud. Structural features are then extracted, and parametric modeling is performed to obtain a 3D construction model. This model can be established using existing algorithms. The model should accurately reflect the shape, location, and other geometric features of the construction object, providing a foundation for subsequent risk point matching. Semantic segmentation divides the geometric construction model into multiple regional feature units according to their different functions and characteristics. These regional feature units can represent different parts or different construction stages in the construction process. This division allows for more detailed analysis of various aspects of the construction, providing a more accurate basis for subsequent attribute association and risk matching. For example, for a bridge, it can be divided into pier areas, bridge deck areas, bridge bearing areas, etc. Each regional feature unit is associated with material properties, construction time windows, and equipment deployment information. Material properties, construction time windows, and equipment deployment information are crucial factors in building construction. Linking these attributes allows for a more comprehensive understanding of the construction requirements and risks of each regional feature unit. For example, the bridge pier area might be associated with concrete material properties, specific construction time windows (such as the foundation construction phase), and equipment deployment information such as cranes. A spatiotemporal correlation matching algorithm is used to match each regional feature unit with a risk database. Spatially, the algorithm matches risk point coordinate deviation thresholds, checking if the spatial location of the regional feature unit is close to the location of risk points in the risk database; temporally, it matches construction progress deviation thresholds, checking if the construction progress of the regional feature unit matches the time of occurrence of risk points in the risk database. This two-dimensional matching allows for more accurate identification of valid risk points, improving the accuracy of risk identification. When the two-dimensional match is successful—that is, when the regional feature unit matches a risk point in the risk database both spatially and temporally—it is determined to be a valid risk point. This ensures that the identified risk points are genuine and valid, rather than being misjudged or missed. For example, if the construction progress of the bridge pier area matches the time of a certain risk point in the risk database, and the spatial location is also similar, then the risk point can be determined as a valid risk point.

[0059] Furthermore, the controller receives the identification and judgment results from the information processing module, namely, the risk point information of the construction object. This information is presented to staff in an intuitive and easy-to-understand manner (such as charts and reports) so they can quickly understand the location and nature of the risk points. Based on the identification and judgment results, a targeted risk management plan is matched. The plan should include risk control measures, emergency plans, and division of responsibilities to ensure a rapid and effective response when a risk occurs. Risk control measures are implemented according to the risk management plan, such as strengthening safety protection and adjusting the construction schedule. Simultaneously, the risk management module should continuously monitor changes in the risk points to ensure the effectiveness of risk control measures. If a risk point is found to be expanding or worsening, the risk management plan should be adjusted in coordination with the controller, and corresponding remedial measures should be taken. After the risk management is completed, the entire process is summarized and evaluated. The effectiveness of the risk management plan, problems encountered during implementation, and improvement measures are analyzed. The summary results are fed back to the information processing module and the risk database for continuous optimization and improvement in future risk management.

[0060] Specifically, the risk identification device performs risk point matching between the regional feature unit and the risk database. When the regional feature unit fails to match the risk database, the sample reinforcement learning process is initiated. The feature vector of the unidentified scene and the expert manual annotation result are input into the risk database to generate identification rules adapted to the new risk type. After successful matching, the spatiotemporal evolution features of the risk point are automatically extracted, and the meta-learning parameters of the risk identification process are updated.

[0061] It's worth noting that, using risk identification equipment, the pre-defined regional feature units are matched against a pre-built risk database to identify risk points. The risk database stores information on known risk types, characteristics, and corresponding response strategies. The matching process aims to quickly determine whether a known risk exists in the current regional feature unit. If a regional feature unit cannot be successfully matched with the content in the risk database, it means a new risk scenario not covered by the database has emerged. In this case, the system automatically initiates a sample reinforcement learning process to address this unknown situation.

[0062] The sample reinforcement learning process includes: extracting feature vectors for unidentified scenarios. Feature vectors are digital representations of the essential characteristics of a scenario, potentially encompassing multiple dimensions such as spatial location, construction stage, material properties, and construction techniques. For example, in a construction site, the feature vector for a specific construction area might include the area's coordinates, the current construction process (e.g., concrete pouring, rebar tying), and the type of building materials used. Experts in relevant fields are invited to manually annotate the unidentified scenarios. Based on their extensive experience and expertise, experts determine whether the scenario carries risk, its type, and severity. The annotation results provide accurate labeling information for sample reinforcement learning. The feature vectors of the unidentified scenarios and the expert annotations are input into a risk database as new learning samples. Based on these new learning samples, the system uses machine learning algorithms (e.g., decision trees, neural networks) to generate identification rules adapted to new risk types. These rules can identify risks similar to the new scenarios and incorporate them into the risk database's identification system. For example, by analyzing the feature vectors of a large number of new scenarios and the expert annotation results, the system can summarize some general feature patterns to determine whether similar risks exist.

[0063] Spatiotemporal evolution feature extraction involves the system automatically extracting the spatiotemporal evolution features of a risk point after a regional feature unit successfully matches with the risk database. These features describe the changing patterns of the risk point in time and space. In the time dimension, this may include the time of occurrence, duration, and trend of change of the risk point; in the spatial dimension, it may include changes in the location of the risk point and its range of influence. For example, in building construction, a risk point may gradually move in location and its range of influence may gradually expand as construction progresses.

[0064] Meta-learning parameter updates involve updating the meta-learning parameters of the risk identification process based on the extracted spatiotemporal evolution features. These parameters are key factors influencing the performance and generalization ability of the risk identification model. By continuously updating these parameters, the risk identification model can better adapt to different construction scenarios and risk types, improving the accuracy and reliability of risk identification. For example, if the spatiotemporal evolution features of a certain risk point are found to be inconsistent with previous assumptions, the system can adjust the relevant parameters to make the model more accurate in future identifications.

[0065] Before performing risk point matching on the regional feature units and the risk database, the method further includes:

[0066] The regional feature units are classified by type, and the priority of risk point identification is set;

[0067] According to different security levels, the regional feature units are divided into: security risk units, behavioral risk units, and environmental risk units. The security risk units are set as the first risk priority, the behavioral risk units are set as the second risk priority, and the environmental risk units are set as the third risk priority.

[0068] Specifically, before matching risk points to several regional characteristic units, they are first classified according to their characteristics. The classification criteria are based on multiple dimensions, including risk source, scope of impact, and controllability. Based on the classification results, regional characteristic units are divided into safety risk units, behavioral risk units, and environmental risk units. Safety risk units primarily focus on safety hazards related to building structures and equipment. Behavioral risk units focus on whether construction workers' actions are standardized and whether there are any violations. Environmental risk units consider the impact of the natural environment, climate conditions, and surrounding social environment on construction safety. After completing the classification, regional characteristic units are assigned different risk point identification priorities based on their safety levels. Safety risk units, because they are directly related to the safety of building structures and equipment, are assigned the highest priority and require priority identification and handling. Behavioral risk units, although related to the operational behavior of construction workers, can be controlled and improved through training and management, and are therefore assigned the second highest priority. Environmental risk units are typically affected by various factors such as natural conditions and the social environment, and their controllability is relatively weak; therefore, they are assigned the third highest priority.

[0069] During risk point matching, the system processes regional feature units according to a predefined priority order. For regional feature units with the highest risk priority, the system prioritizes risk point identification and judgment, and promptly relays the results to relevant personnel for processing. For regional feature units with second and third risk priorities, the system processes them sequentially after completing the tasks assigned to the highest priority units. Each type of risk is assessed to determine its nature, probability of occurrence, and potential consequences. Through risk assessment, the priority and urgency of risks are determined, providing a basis for subsequent risk control.

[0070] Prioritize risk identification, including:

[0071] Based on the security levels of historical data in the risk database, the security levels of the risk points in the regional feature units are obtained.

[0072] The historical data have a set of associated attributes, which records the attributes associated with the historical data. Based on the attributes and the attributes of the historical data, the association between the risk point and the historical data is obtained.

[0073] Complete the identification and matching of the priority of the aforementioned risk points.

[0074] Specifically, prioritizing risk identification helps allocate resources and attention more effectively, addressing risks with potentially more severe consequences first. The system first extracts historical data similar to the current project from the risk database. This data includes risk events that occurred in past projects, the safety level of the risk points, the frequency of occurrence, and the consequences. Analyzing this historical data focuses particularly on historical events similar to risk points in the current project's regional feature units. By analyzing the safety levels of these events, the system can preliminarily assess the potential severity of the current risk point. The association attribute set is a collection recording the association attributes between historical data. These attributes include the type of risk point, its specific location, the personnel or equipment involved, and environmental factors. Based on the attributes of the current risk point, the system searches for historical events with similar attributes in the historical data. By comparing the attributes of these events, a correlation is established between the current risk point and the historical data. Once the correlation is established, the system also assesses the strength of the relationship by calculating the similarity or correlation between attributes. A higher correlation strength indicates a greater similarity between the current risk point and the historical data, allowing for more effective use of historical data to predict and assess the safety level of the current risk point. After determining the security level of the current risk points and their correlation with historical data, the priority of these risk points is set based on this information. Generally, risk points with higher security levels and stronger correlations are assigned higher priority. As the project progresses and new risk points emerge, the system continuously updates the risk database, including new historical data and sets of related attributes. This helps the system to more accurately prioritize risk points in future projects.

[0075] Obtaining the correlation between the risk points and the historical data includes:

[0076] When the attribute of the risk point is in the set of associated attributes, calculate the correlation degree between the risk point and the historical data;

[0077] When the correlation is higher than a preset threshold, it is determined that the risk point is correlated with the corresponding historical data.

[0078] Specifically, the first step is to identify the attributes of the current risk point and compare them with a predefined set of associated attributes. The associated attribute set is a collection of multiple attributes that typically describe the risk point's characteristics, type, location, involved personnel, or equipment. If one or more attributes of the risk point appear in the associated attribute set, these attributes are considered potential bridges between the risk point and historical data. Once the intersection of the risk point and the associated attribute set is determined, the system needs to calculate the correlation degree between the risk point and each potentially relevant historical data point. Correlation degree is a quantitative metric used to measure the similarity or relevance between the risk point and historical data.

[0079] The methods for calculating correlation include: calculating the degree of matching between risk point attributes and historical data attributes; for example, if two risk points involve the same type of construction equipment, their attribute matching degree may be high; considering the proximity between the time of occurrence of the risk point and the time of historical data recording; if the time interval is short, the correlation degree may be high; if the risk point and historical data are geographically close, their correlation degree may also be high. One or more methods may be used to calculate the correlation degree, and the results of these methods are combined to obtain a comprehensive correlation score. Furthermore, a preset threshold is used to determine whether there is a correlation between the risk point and historical data. This threshold is a value set based on experience or system requirements, used to determine at what degree of correlation a correlation can be considered between the risk point and historical data. If the correlation degree between a risk point and a piece of historical data is higher than the preset threshold, the system can determine that there is a correlation between the two objects. This also means that information about the risk point in the historical data (such as safety level, handling measures, etc.) may be valuable for current risk management decisions.

[0080] A spatiotemporal correlation matching algorithm is used to match feature units of each region with a risk database, including:

[0081] Spatial dimension matching:

[0082] Use spatial indexes to index coordinates in the risk database to speed up proximity queries;

[0083] For the center point of each regional feature unit, calculate the distance between the center point and the risk point:

[0084]

[0085] Where (x1,y1,z1) are the spatial coordinates of the center point of the regional feature unit; (x2,y2,z2) are the actual spatial coordinates of the risk point;

[0086] Risk points exceeding the distance deviation threshold are selected to form a spatial candidate set;

[0087] Specifically, risk databases contain coordinate information for a large number of risk points, making direct proximity queries inefficient. By establishing spatial indexes, proximity queries on coordinates in the risk database can be accelerated, improving matching efficiency. Common spatial index structures include R-trees and quadtrees. Taking an R-tree as an example, spatial objects are organized hierarchically according to their spatial extent, with each node representing a spatial region, and leaf nodes storing the actual spatial objects (risk point coordinates). During proximity queries, by traversing the nodes of the R-tree, spatial regions that may contain the center point of the target region's feature unit can be quickly located, thereby reducing the number of risk points that need to be compared. A distance deviation threshold is set, and risk points with a distance *d* exceeding this threshold are filtered out, forming a spatial candidate set. These risk points are spatially relatively close to the center point of the region's feature unit and may constitute potential risks.

[0088] Time dimension matching:

[0089] Obtain the current construction stage and actual time progress;

[0090] For each risk point's time window, calculate the deviation between the current time and the planned time:

[0091] ΔT=|Current time -Plan time |

[0092] Among them, Current time Current time; Plan time For planned time;

[0093] Risk points exceeding the time deviation threshold are selected to form a time candidate set;

[0094] Specifically, determine the current construction stage of the building project, such as the foundation construction stage, main structure construction stage, and decoration and finishing construction stage. Different construction stages may face different types of risks and time requirements. Obtain the current actual time and the planned time corresponding to each risk point. The planned time can be determined according to the construction schedule, which specifies the time range within which each risk point should occur or be addressed. Set a time deviation threshold and filter out risk points whose time deviation ΔT exceeds the threshold, forming a candidate set of time risks. These risk points have a large time deviation from the current construction progress and may pose a time risk.

[0095] Two-dimensional joint matching:

[0096] The intersection of the spatial candidate set and the temporal candidate set is used to obtain the risk points that simultaneously meet the conditions.

[0097] If the intersection is not empty, it is determined to be a valid risk point;

[0098] Among them, the matched risk points are sorted by spatial distance and time deviation, and the urgent risk points are dealt with first.

[0099] Specifically, the intersection of the spatial and temporal candidate sets is taken to obtain risk points that simultaneously satisfy both spatial and temporal conditions. These risk points are spatially close to the center point of the regional feature unit and temporally deviate from the current construction progress, making them potential valid risk points. If the intersection is not empty, the risk points in the intersection are determined to be valid risk points. The matched risk points are sorted according to spatial distance and temporal deviation, and different weights can be set according to the actual situation, for example, the weight of spatial distance can be w. d The weight of the time deviation is w t Calculate the overall score S for each risk point;

[0100]

[0101] Where, d max and ΔT max These represent the maximum values ​​of spatial distance and temporal deviation, respectively. Risk points are ranked according to their overall score S, with higher-scoring risk points (those with higher urgency) being prioritized for processing.

[0102] The risk identification device has a built-in risk identification model;

[0103] The target image information corresponding to the construction object is input into the risk identification model for risk identification, and the risk identification result is generated and sent to the risk handling module. At the same time, when the probability of risk determination exceeds 80% during risk identification, the risk is determined to have occurred and a risk warning is issued.

[0104] When a risk point matching is completed, the identification results and identification data of the identified risk points are input into the risk identification model as training samples to optimize the risk identification model.

[0105] Specifically, the risk identification device receives target image information corresponding to the construction object transmitted by the information processing module and inputs the received target image information into the risk identification model. The risk identification model utilizes advanced technologies such as deep learning and computer vision to extract features and recognize patterns from the input information. By comparing and matching with a preset risk feature library, the model can identify potential risk points in the target image. After completing the risk identification process, the risk identification model generates detailed risk identification results. The identification results include the location, nature, potential impact, and suggested risk control measures of the risk points. The risk identification device sends the generated risk identification results to the risk management module for further analysis and processing by relevant personnel. During the risk identification process, the risk identification model also performs a probability assessment of the identified risks. When the probability of risk determination exceeds a set threshold (e.g., 80%), the model determines that a risk has occurred and triggers a risk warning mechanism. The risk warning mechanism includes measures such as sending warning information to relevant personnel, activating emergency plans, or adjusting construction schedules to ensure that risks are controlled in a timely and effective manner. Upon completion of a risk point matching process, the risk identification device inputs the identified risk points and related data into the risk identification model. This data serves as training samples to optimize and improve the model's algorithms and parameters. Through continuous learning and iteration, the accuracy and efficiency of the risk identification model will be continuously improved, thus better adapting to the risk identification needs of different construction environments and conditions.

[0106] The mapping relationship set includes: at least one of the regional feature units, and the correspondence between the regional feature units and the building construction objects.

[0107] Specifically, the mapping relationship set is a dataset containing at least one regional feature unit, explicitly indicating the correspondence between each regional feature unit and the building construction object. This correspondence is bidirectional, meaning that a regional feature unit can be traced back to a specific part of the building construction object, and a corresponding regional feature unit can be found from a part of the building construction object. Regional feature units are the basic elements of the mapping relationship set, representing the independent regions after the building construction object has been divided. Each regional feature unit has unique attributes and characteristics, such as location, shape, and construction activities. The correspondence describes the connection between the regional feature unit and the building construction object. This relationship can be based on geographical location (e.g., a regional feature unit is located on a specific floor or in a specific area of ​​the building construction object) or on function or construction activity (e.g., a regional feature unit involves a specific construction process or equipment). The mapping relationship set provides the foundation for risk point matching. By comparing the attributes of regional feature units with risk characteristics in the risk database, the system can identify potential risk points. Once a risk point is identified, the mapping relationship set helps the system quickly locate the corresponding part of the building construction object, thereby triggering a risk warning mechanism and ensuring that relevant personnel can take timely intervention measures. In the risk management process, mapping relationships help relevant personnel understand the specific location and scope of impact of risk points, thereby enabling the development of targeted risk control measures. Simultaneously, it can also be used to monitor changes in risk points, ensuring the effectiveness of risk control measures.

[0108] The risk database also allows for the addition or modification of project risk points, enabling real-time adjustments to risk conditions.

[0109] Specifically, the risk database is designed with the complexity and dynamism of construction projects in mind. It's not merely a static data repository, but a dynamic system that continuously updates as the project progresses and the external environment changes. Users can add newly discovered risk information to the database based on the actual project situation. This information may include the risk's location, nature, potential impact, and suggested control measures. Adding risk points is typically done through the system's user interface. Users simply fill out a form, submit it, and the system adds the new risk information to the database. After a risk point is added, the system immediately updates the risk database, ensuring all relevant personnel have timely access to the latest risk information. As the project progresses and the external environment changes, existing risk information may need to be updated or corrected. The risk database allows users to modify these risk points to reflect the latest risk status.

[0110] By allowing users to add or modify project risk points, the risk database enables real-time adjustments to risk conditions. This means the system can continuously update risk information as the project progresses and the external environment changes, thus maintaining accurate risk identification and effective management. The ability to adjust risk conditions in real-time is extremely useful at every stage of a construction project. For example, in the early stages of construction, users may need to add new geological risk points based on geological survey results; in the middle stages, as construction activities progress, users may need to modify the nature or potential impact of certain risk points; and in the later stages, users may need to delete or update completed risk points based on the final acceptance results.

[0111] The correspondence between regional feature units and the building construction object includes:

[0112] Based on the project risk points and the geometric construction model, the pre-defined risk point locations {M = M1, M2, M3, ..., Mi, ..., Mn} and the corresponding project risk point locations {N = N1, N2, N3, ..., Ni, ..., Nn} on the geometric construction model are obtained respectively; among them, the pre-defined risk point location M is clear and specific when it is set, and the corresponding project risk point location N can be found in the geometric construction model.

[0113] Specifically, regional characteristic units refer to areas within a construction site that possess specific attributes or functions, such as foundation construction areas, main structure construction areas, and finishing construction areas. These areas vary depending on the construction content, difficulty, and environmental factors, and therefore may face different risks. Construction objects refer to specific buildings or structures, and the related construction activities. During the construction process, risk management and control of these objects are necessary. In the risk identification phase, some common risk points are pre-defined based on experience or historical data; these points are clearly defined during the pre-determining process. During actual construction, the actual project risk points corresponding to the pre-defined risk points need to be found using a geometric construction model. The pre-defined risk points are mapped to their corresponding positions on the geometric construction model, thus determining the specific locations of the project risk points. Correspondence establishment: Through the risk point mapping process, a correspondence is established between the pre-defined risk points and the project risk points. This correspondence helps to quickly locate and identify risk points in subsequent risk management, thereby enabling the implementation of appropriate risk control measures.

[0114] The security risk unit identifies risk points according to the first identification rule;

[0115] The behavioral risk unit identifies risk points according to the second identification rule;

[0116] The environmental risk unit identifies risk points according to the third identification rule.

[0117] It's worth noting that the safety risk unit is primarily responsible for identifying risks directly related to construction safety. The first identification rule is specifically formulated based on construction safety standards, historical accident data, and expert experience. It covers multiple aspects, including the operational safety of construction equipment, the safety of materials used, and the compliance of construction methods. Simultaneously, the safety risk unit analyzes various operations, materials, and equipment used during the construction process, as well as the on-site safety management status, to determine whether potential safety risks exist based on the first identification rule.

[0118] The Behavioral Risk Unit focuses on identifying risks that may arise from the behavior of construction workers. The second identification rule specifically focuses on personnel operational behavior, safety training, and the wearing of personal protective equipment. These may include adherence to operating procedures, proficiency in emergency response, and potential risk factors such as fatigue driving or operation. The Behavioral Risk Unit monitors the behavior of construction workers, using video surveillance, sensor data, and other means, combined with the second identification rule, to assess the safety of the behavior and identify potential behavioral risks.

[0119] The environmental risk unit identifies potential risks from the construction environment. The third identification rule specifically considers factors such as climate conditions, geological conditions, the layout of the construction site, and the mutual influence of the surrounding environment. For example, it considers the risk of flooding due to heavy rain, structural damage caused by earthquakes, and traffic congestion caused by an unreasonable construction site layout. The environmental risk unit analyzes data from the construction site and its surrounding environment, including meteorological data, geological survey reports, and site layout maps, and assesses environmental safety according to the third identification rule, identifying potential environmental risks.

[0120] Therefore, in a construction risk management system, the safety risk unit, behavioral risk unit, and environmental risk unit are complementary. They operate independently but together constitute a comprehensive risk identification system. By integrating risk information from these three units, the system can provide a more comprehensive and accurate risk assessment, thereby helping project managers develop effective risk control strategies.

[0121] The above description is merely a few embodiments of this application and is not intended to limit this application in any way. Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any changes or modifications made by those skilled in the art without departing from the scope of the technical solution of this application using the disclosed technical content are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A construction risk management system, characterized in that, include: The risk point pre-setting module uses project construction drawings, construction plans and construction schemes to identify safety risks during the construction process, and combines them with the system database to pre-set project risk points, thus obtaining a risk database. The information processing module acquires target image information corresponding to the building construction object and establishes a geometric construction model of the building construction object. The geometric construction model has a mapping relationship with the building construction object. The geometric construction model is semantically segmented and divided into multiple regional feature units. Each regional feature unit is associated with material properties, construction time windows, and equipment deployment information. A spatiotemporal correlation matching algorithm is used to match each regional feature unit with a risk database. The risk point coordinate deviation threshold is matched in the spatial dimension, and the construction progress deviation threshold is matched in the temporal dimension. When the two-dimensional matching is successful, it is determined to be a valid risk point. The controller receives data from the risk point preset module and the information processing module, and automatically configures corresponding risk management decisions based on the preset information in the risk database and the identification and judgment results of the risk points. The risk management module, coupled with the controller, enables automated control of the entire risk management process, executes risk management decisions, and completes risk management. In this process, risk points are matched between the regional feature units and the risk database using a risk identification device. When the regional feature units fail to match the risk database, a sample reinforcement learning process is initiated. The feature vectors of the unidentified scenarios and the results of expert manual annotation are input into the risk database to generate identification rules that are adapted to the new risk type. After successful matching, the spatiotemporal evolution features of the risk point are automatically extracted, and the meta-learning parameters of the risk identification process are updated. Before performing risk point matching on the regional feature units and the risk database, the method further includes: The regional feature units are classified by type, and the priority of risk point identification is set; According to different security levels, the regional feature units are divided into: security risk units, behavioral risk units, and environmental risk units. The security risk units are set as the first risk priority, the behavioral risk units are set as the second risk priority, and the environmental risk units are set as the third risk priority. A spatiotemporal correlation matching algorithm is used to match feature units of each region with a risk database, including: Spatial dimension matching: Use spatial indexes to index coordinates in the risk database to speed up proximity queries; For the center point of each regional feature unit, calculate the distance between the center point and the risk point: ； Where (x1,y1,z1) are the spatial coordinates of the center point of the regional feature unit; (x2,y2,z2) are the actual spatial coordinates of the risk point; Risk points exceeding the distance deviation threshold are selected to form a spatial candidate set; Time dimension matching: Obtain the current construction stage and actual time progress; For each risk point's time window, calculate the deviation between the current time and the planned time: ΔT=|Current time -Plan time | Among them, Current time The current time; Plant ime For planned time; Risk points exceeding the time deviation threshold are selected to form a time candidate set; Two-dimensional joint matching: The intersection of the spatial candidate set and the temporal candidate set is used to obtain the risk points that simultaneously meet the conditions. If the intersection is not empty, it is determined to be a valid risk point; Among them, the matched risk points are sorted by spatial distance and time deviation, and the urgent risk points are dealt with first.

2. The construction risk management system according to claim 1, characterized in that, Prioritize risk identification, including: Based on the security levels of historical data in the risk database, the security levels of the risk points in the regional feature units are obtained. The historical data have a set of associated attributes, which records the attributes associated with the historical data. Based on the attributes and the attributes of the historical data, the association between the risk point and the historical data is obtained. Complete the identification and matching of the priority of the aforementioned risk points.

3. The construction risk management system according to claim 2, characterized in that, Obtaining the correlation between the risk points and the historical data includes: When the attribute of the risk point is in the set of associated attributes, calculate the correlation degree between the risk point and the historical data; When the correlation is higher than a preset threshold, it is determined that the risk point is correlated with the corresponding historical data.

4. A construction risk management system according to claim 1, characterized in that, The risk identification device has a built-in risk identification model; The target image information corresponding to the construction object is input into the risk identification model for risk identification, and the risk identification result is generated and sent to the risk handling module. At the same time, when the probability of risk determination exceeds 80% during risk identification, the risk is determined to have occurred and a risk warning is issued. When a risk point matching is completed, the identification results and identification data of the identified risk points are input into the risk identification model as training samples to optimize the risk identification model.

5. A construction risk management system according to claim 1, characterized in that, The mapping relationship set includes: at least one of the regional feature units, and the correspondence between the regional feature units and the building construction objects.

6. A construction risk management system according to claim 1, characterized in that, The risk database also allows for the addition or modification of project risk points, enabling real-time adjustments to risk conditions.

7. A construction risk management system according to claim 5, characterized in that, The correspondence between regional feature units and the building construction object includes: Based on the project risk points and the geometric construction model, the preset risk point locations {M = M1, M2, M3, ..., Mi, ..., Mn} and the corresponding project risk point locations {N = N1, N2, N3, ..., Ni, ..., Nn} on the geometric construction model are obtained respectively; Among them, the preset risk point M is clearly defined during the preset process, and the corresponding project risk point N can be found in the geometric construction model.

8. A construction risk management system according to claim 1, characterized in that, The security risk unit identifies risk points according to the first identification rule; The behavioral risk unit identifies risk points according to the second identification rule; The environmental risk unit identifies risk points according to the third identification rule.

Images