A construction safety equipment state monitoring system based on data analysis processing

Abstract

This invention relates to the field of data processing and construction safety monitoring technology, and discloses a construction safety equipment status monitoring system based on data analysis and processing. The system includes: a data acquisition and preprocessing module for converting multi-source data from the site into standardized state vectors; a dynamic state causal graph construction module for constructing and updating a dynamic state causal graph based on state vectors and entity information; a risk potential energy calculation engine for performing iterative calculations based on a graph structure to quantify the risk potential energy and time gradient of nodes; a task risk simulation module for simulating construction task plans and identifying potential risks; an adaptive scheduling and intervention module for monitoring risk indicators, tracing risk sources and generating intervention instructions when thresholds are exceeded; and a visualization and interactive terminal for presenting risk information. This invention, by constructing a causal graph model and a risk potential energy algorithm, achieves quantitative assessment, pre-simulation, and intervention of risks associated with entities at the construction site.

Description

Technical Field

[0001] This invention relates to the field of data processing and construction safety monitoring technology, specifically to a construction safety equipment status monitoring system based on data analysis and processing. Background Technology

[0002] With the expansion and increasing complexity of modern construction projects, safety monitoring at construction sites has become a crucial aspect of ensuring the smooth progress of projects. The construction site environment involves a large number of heavy machinery and equipment, a dense workforce, and a dynamically changing physical environment, with close coordination and constraints among these elements.

[0003] Currently, common construction safety monitoring systems primarily utilize various sensors installed on equipment to collect operational data. Alarm control is achieved by comparing the collected physical quantities such as load, displacement, and temperature with preset static thresholds. While this monitoring mode can detect instantaneous anomalies in a single piece of equipment or location, its limitation lies in treating each monitored object as an isolated entity, ignoring the physical connections, spatial proximity, and task coordination among entities on the construction site. When risks propagate among multiple related entities or experience cumulative effects due to spatial congestion, relying solely on single-point threshold detection is insufficient to effectively perceive and quantify such dynamically correlated risks, easily leading to missed reports of complex safety hazards caused by the interaction between equipment.

[0004] Furthermore, existing monitoring methods primarily focus on monitoring and recording the current real-time status of the construction site, lacking deep integration with the construction schedule. Before the actual work begins, existing technologies struggle to effectively simulate and extrapolate the site conditions and equipment operating trajectories for future periods based on the established task plan, making it impossible to predict spatiotemporal conflicts or accumulated risks arising from overlapping operations of different processes. This lagging management model often leaves safety control at the stage of post-event alarms and passive responses, failing to identify potential high-risk paths in advance before task execution, and hindering proactive optimization and adjustments to the construction plan.

[0005] Meanwhile, in complex construction scenarios with densely packed and interconnected equipment, existing systems struggle to quickly and accurately identify the specific source of risk when monitoring indicators show anomalies. Due to a lack of quantitative analysis of risk transmission paths and contributions, managers find it difficult to determine whether alarms are caused by component failures within the equipment itself or by external risks transmitted from the surrounding environment or other upstream equipment. This ambiguity in risk attribution makes it difficult to quickly pinpoint the root cause of risks in emergencies, and the system cannot automatically generate intervention commands targeting specific operational objects and actions, reducing the efficiency and relevance of on-site risk investigation and handling. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a construction safety equipment status monitoring system based on data analysis and processing. This system solves the problems of traditional construction safety monitoring, which mainly relies on single-point threshold alarms, has difficulty in quantifying the dynamic correlation risks between equipment and the environment, and cannot perform risk simulations and targeted interventions based on construction plans.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A construction safety equipment status monitoring system based on data analysis and processing includes: a data acquisition and preprocessing module, a dynamic status cause-effect graph construction module, a risk potential energy calculation engine, a task risk simulation module, an adaptive scheduling and intervention module, and a visualization and interactive terminal.

[0008] The data acquisition and preprocessing module is used to acquire raw data from various data sources at the construction site and process it into standardized state vectors. The dynamic state cause-effect graph construction module is electrically connected to the data acquisition and preprocessing module. It is used to receive standardized state vectors and construct and update the dynamic state cause-effect graph in real time based on the standardized state vectors and the location and task information of the entities. The risk potential calculation engine is electrically connected to the dynamic state causal graph construction module. It receives the dynamic state causal graph and performs iterative calculations based on the topology and node states of the dynamic state causal graph to quantify the risk potential and temporal gradient of the graph nodes. The task risk simulation module is electrically connected to the risk potential energy calculation engine. After receiving the construction task plan, it calls the risk potential energy calculation engine to perform simulation, identify potential risks, and generate an optimized construction plan. The adaptive scheduling and intervention module is electrically connected to the risk potential energy calculation engine. It is used to receive risk potential energy and time gradient and perform real-time monitoring. When the risk exceeds the preset threshold, it executes risk source tracing and generates structured intervention instructions. The visualization and interactive terminal is electrically connected to the risk potential calculation engine, the task risk pre-simulation module, and the adaptive scheduling and intervention module, respectively, to receive and present the risk potential, the optimized construction plan, and the structured intervention instructions to the user.

[0009] Preferably, when the data acquisition and preprocessing module generates a standardized state vector from raw data, it is configured to: acquire multi-source heterogeneous raw data; perform denoising and completion processing on the raw data to obtain preprocessed data; perform state vectorization processing on the preprocessed data, map the physical quantity values ​​to a dimensionless interval through calculation to obtain normalized state components, and combine all normalized state components of the same entity to form a standardized state vector.

[0010] Preferably, the dynamic state cause-effect graph construction module abstracts and defines the monitored entities at the construction site as graph nodes. Each graph node is uniquely identified by a triple containing an entity identifier, a task identifier, and a location identifier. Based on the received standardized state vectors and the entity's location and task information, this module performs real-time management of creating, updating, or deleting graph nodes in the dynamic state cause-effect graph, realizing dynamic mapping of construction site entities and their attribute states.

[0011] Preferably, the dynamic state causal graph construction module establishes directed edges between graph nodes in the dynamic state causal graph to represent the relationships between entities. The types of directed edges include physical connection edges, task-related edges, spatial proximity edges, and inferential causal edges. For spatial proximity edges, the system calculates the edge weight based on the spatial distance between two graph nodes, thereby quantifying the impact of spatial distance on risk propagation.

[0012] Preferably, the risk potential calculation engine employs a graph theory iterative algorithm to quantify the risk potential of graph nodes. Specifically, firstly, based on the state vectors associated with the graph nodes in the dynamic state causal graph, the weighted sum of each component of the state vector is performed to calculate the self-risk potential representing the inherent risk state of the node. Subsequently, based on the calculated self-risk potential and combined with the topological structure of the dynamic state causal graph, multiple rounds of iterative calculations are performed to calculate the transmission and superposition of risk along directed edges between graph nodes, ultimately obtaining the total risk potential of the graph node, including its own risk and externally introduced risks.

[0013] Preferably, when performing iterative calculations, the risk potential energy calculation engine uses its own risk potential energy as the initial condition for the iterative calculation, and introduces a nonlinear activation function during the propagation and convergence process to limit the cumulative risk value and prevent the calculation results from diverging.

[0014] Preferably, the task risk simulation module achieves pre-emptive risk control by calculating future states. This module parses the construction task plan to generate a time-ordered sequence of future events; obtains a copy of the dynamic state causal graph at the current moment, and simulates the occurrence of the future event sequence on the copy in chronological order, updating the topology or node state of the copy; after each simulation update, it calls the risk potential energy calculation engine to calculate the predicted risk potential energy, identifies high-risk paths based on the predicted risk potential energy, adjusts the task arrangement according to the optimization rule base, and generates an optimized construction plan.

[0015] Preferably, the adaptive scheduling and intervention module employs a dual-threshold triggering mechanism to determine whether to initiate the intervention process. The triggering condition is set as follows: when the total risk potential energy of any graph node exceeds a preset risk potential energy threshold, or the absolute value of the risk potential energy gradient exceeds a preset risk gradient threshold, risk source tracing is initiated. This mechanism simultaneously considers the absolute magnitude and rate of change of the risk, enabling it to respond to sudden changes in risk.

[0016] Preferably, when performing risk source tracing, the adaptive scheduling and intervention module first calculates the total input risk of the graph node that triggered the alarm; then, based on the total input risk, it calculates its own risk contribution and the incoming risk contribution of the upstream neighbor graph node to obtain a risk contribution list; finally, based on the risk contribution list, it identifies the risk sources whose contribution meets the preset conditions as the main risk contribution sources, and generates structured intervention instructions based on the type and attributes of the main risk contribution sources to achieve targeted risk handling.

[0017] Preferably, the visualization and interactive terminal loads the building information model of the construction scene and maps the graph node information in the dynamic state cause-effect graph to the 3D scene to form a digital twin environment. A spatial risk heat map is superimposed on the horizontal cross-section of the digital twin environment. Based on the risk potential energy and spatial distance influence of each graph node, the heat map calculates the risk density value of any coordinate point in the 3D scene, intuitively presenting the regional overall risk distribution.

[0018] This invention provides a construction safety equipment status monitoring system based on data analysis and processing. It has the following beneficial effects: 1. This invention, by constructing a dynamic state causal graph and introducing a risk potential energy calculation engine, changes the traditional monitoring system's reliance on single-point data threshold alarms. The system utilizes graph topology and iterative algorithms to quantify the propagation and convergence process of risk among different entities. It can calculate the specific impact of equipment failure or environmental anomalies on surrounding related entities, thereby achieving a quantitative assessment of the dynamic correlation risk between equipment and the environment at the construction site, reducing the risk of missed reports due to neglecting correlation effects.

[0019] 2. This invention utilizes a task risk simulation module to create a copy of a dynamic state causal graph and simulate event sequences before the actual construction task is executed. This mechanism can identify potential high-risk paths and automatically generate adjusted construction plans before the task begins by calculating the predicted risk potential energy of key nodes at future moments. This solves the problem that existing technologies cannot avoid risks in advance based on the construction progress, and realizes the transformation of safety management from passive response after the fact to proactive prevention before the fact.

[0020] 3. This invention employs an adaptive scheduling and intervention module, combined with a dual-threshold triggering mechanism and a risk source tracing algorithm based on contribution. When an anomaly is detected, the system calculates the contribution ratio of the node's own risk and the risk transmitted from its neighbors, enabling it to quickly distinguish whether the risk is caused by the node's own failure or by external environmental transmission. Based on this, it generates structured intervention instructions containing specific operational objects, solving the problem of difficulty in quickly locating the root cause of risks and formulating targeted disposal measures in complex construction environments. Attached Figure Description

[0021] Figure 1 This is a system functional module block diagram provided in an embodiment of the present invention; Figure 2 The main flowchart of the monitoring method provided in the embodiments of the present invention.

[0022] The module includes: 100, Data Acquisition and Preprocessing Module; 200, Dynamic State Cause-Effect Graph Construction Module; 300, Risk Potential Calculation Engine; 400, Task Risk Pre-Drilling Module; 500, Adaptive Scheduling and Intervention Module; and 600, Visualization and Interactive Terminal. Detailed Implementation

[0023] The technical solutions in 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] See attached document Figure 1 The system provided by this invention, in one embodiment, may include: a data acquisition and preprocessing module 100, a dynamic state causal graph construction module 200, a risk potential calculation engine 300, a task risk pre-simulation module 400, an adaptive scheduling and intervention module 500, and a visualization and interactive terminal 600. Each module exchanges data with the central data storage unit 700 via an internal bus.

[0025] The data acquisition and preprocessing module 100 is used to acquire raw data from various data sources at the construction site and process it into standardized state vectors. Data sources include, but are not limited to, equipment sensors, personnel wearable devices, environmental monitoring stations, and Building Information Modeling (BIM) systems.

[0026] The dynamic state cause-effect graph construction module 200 is electrically connected to the data acquisition and preprocessing module 100. This module receives standardized state vectors and entity location and task information to construct and update in real time a dynamic state cause-effect graph (DSCG) that can characterize the dynamic relationships between entities on the construction site.

[0027] The risk potential calculation engine 300 is electrically connected to the dynamic state causal graph construction module 200. Based on the topology and node states of the DSCG, this engine performs iterative calculations to quantify the risk potential of each node and its temporal gradient.

[0028] The task risk simulation module 400 is electrically connected to the risk potential energy calculation engine 300 and the central data storage unit 700. After receiving the construction task plan, the task risk simulation module 400 calls the risk potential energy calculation engine 300 to simulate the future task execution process in order to identify and avoid potential risks.

[0029] The adaptive scheduling and intervention module 500 is electrically connected to the risk potential energy calculation engine 300. The adaptive scheduling and intervention module 500 monitors the risk potential energy and its gradient of each node in real time. When the risk exceeds a preset threshold, it performs risk source tracing and generates structured intervention instructions.

[0030] The visualization and interactive terminal 600 is electrically connected to the risk potential calculation engine 300, the task risk simulation module 400, and the adaptive scheduling and intervention module 500, respectively. This terminal is used to present users with real-time risk situations, simulated optimization plans, and generated intervention instructions.

[0031] The central data storage unit 700 provides persistent data storage services for all modules of the system, including raw data, state vectors, graph structure snapshots, risk calculation results, and historical instruction records.

[0032] See attached document Figure 2 The method provided by the present invention may include the following steps in some embodiments: S201 collects heterogeneous data from multiple sources and preprocesses the data to generate standardized state vectors; S202, based on the state vector and the position and task information of the entity, constructs and updates a dynamic state causal graph in real time; S203, on a dynamic state causal graph, quantifies the risk potential energy and its temporal gradient of each node through iterative calculation; S204, Upon receiving the construction task plan, the plan is parsed to generate a sequence of future events, and the event sequence is simulated on a copy of the dynamic state cause-effect graph to predict future risks and generate an optimized construction plan; S205 monitors the risk potential and its time gradient of each node in real time. When any indicator exceeds the preset threshold, it performs risk source tracing to locate the main risk contributor and generates structured intervention instructions. S206 presents the risk potential, optimized construction plan, and intervention instructions through a visualization and interactive terminal, and sends the intervention instructions to the external execution system.

[0033] The technical implementation details of each step in the above method process will be described in detail below so that those skilled in the art can understand and implement the present invention.

[0034] The implementation method of step S201 (collecting multi-source heterogeneous data and preprocessing the data to generate a standardized state vector) is described in detail. This step is executed by the data acquisition and preprocessing module 100, and can be broken down into the following sub-steps: Step S201-1: Acquire multi-source heterogeneous data. The data acquisition and preprocessing module 100 communicates with various data sources deployed at the construction site through its built-in multiple data interfaces to obtain raw data. For example, for low-power sensors integrated into safety equipment such as safety helmets and scaffolding fasteners, data periodically reported by them can be received via the MQTT protocol or LoRaWAN gateway; for UWB positioning tags or vital sign monitoring devices worn by construction workers, data can be received via proprietary wireless protocols or Bluetooth gateways; for environmental monitoring stations, data can be obtained via the Modbus protocol or HTTP interface; and for construction task plans stored in BIM systems or project management software, they can be retrieved by calling standard RESTful API interfaces.

[0035] Step S201-2: Perform data preprocessing. The data acquisition and preprocessing module 100 performs preprocessing on the acquired raw data to address issues such as noise, data loss, or inconsistent timestamps. For noisy data generated by sensors, algorithms such as Kalman filtering or moving average filtering can be used for smoothing. For data point loss caused by network interruptions, methods such as linear interpolation or pre-value padding can be used for data completion. For heterogeneous data from different systems, clocks from each data source can be synchronized based on network time protocols to ensure alignment of all data in the time dimension. The specific implementations of these preprocessing methods are well-known technologies in the field and will not be elaborated upon here.

[0036] Step S201-3: Perform state vectorization processing. The data acquisition and preprocessing module 100, based on a preset dimension conversion logic, converts the preprocessed entity attribute values ​​into a unified, standardized state vector, aiming to eliminate the influence of different physical dimensions on subsequent calculation weights. This process is based on the relative position of the current attribute value within its safe operating range. Before normalization, for negative indicators (such as oxygen concentration and power supply voltage) where "lower values ​​indicate higher risk"), the module first reverses their polarity or takes their reciprocal to convert them into positive indicators where "higher values ​​indicate higher risk," ensuring the consistency of subsequent risk calculation logic. Subsequently, the state component normalization formula is used for calculation. The state component normalization formula is: ; in: , indicating the entity at time The The normalized state component corresponding to each state attribute is a dimensionless numerical value whose range is limited to 1 / 2. between; , indicating the entity at time The The preprocessed values ​​of each state attribute; and , respectively represent the first Each status attribute has a preset lower and upper limit value for the safe operating range. In this embodiment, the upper and lower limits of the safe operating range are not fixed, but are based on the technical specifications provided by the equipment manufacturer and the statistical distribution of historical operating data (e.g., taking 3). (Interval) dynamically set.

[0037] This is a preset numerical stability constant to ensure that the denominator is never zero.

[0038] Functions used to truncate calculation results to... Within the range. Physically speaking, if the calculated value is truncated to 1, it means that the state attribute has exceeded the safety limit and is in an extremely high-risk state; if the calculated value is 0, it means that the attribute is within the absolutely safe range. This processing method maps the offset of the physical state to a dimensionless scalar in the risk probability space, laying the data foundation for subsequent weighted fusion.

[0039] Then, an entity The combination of all normalized state components constitutes the entity at time [time]. state vector The state vector can be represented as: ; in, Representing entities At any moment The state vector; Representing entities respectively The 1st to the 1st One normalized state component; This represents the total number of state attributes monitored for the entity, i.e., the dimension of the state vector; This represents the matrix transpose operation, specifying... It is a column vector.

[0040] The implementation of step S202 (constructing and updating a dynamic state causal graph in real time based on the state vector and the entity's position and task information) is described in detail. This step is executed by the dynamic state causal graph construction module 200, and can be broken down into the following sub-steps: Step S202-1, define and manage execution nodes. The dynamic state cause-effect graph construction module 200 abstracts and defines each monitored entity on the construction site (such as a safety helmet, a worker, or a scaffolding unit) as a graph node. In one specific embodiment, each node is uniquely identified by a triple containing an entity identifier, a task identifier, and a location identifier, with the following structure: ; in, It is a unique identifier for a physical entity, such as an asset number for equipment or an employee's employee number; It is the identifier of the construction task currently associated with this entity, which comes from the project management system; It is the identifier or coordinates of the current three-dimensional spatial location of the entity.

[0041] This node definition method ensures that each node contains both the entity's static identity and dynamic spatiotemporal context information. The dynamic state cause-effect graph construction module 200 manages the nodes in the graph in real time based on the data stream received from the data acquisition and preprocessing module 100. For example, it creates a new node when a new entity enters the monitoring range, or updates the corresponding node identification information when the location of an entity or the task it performs changes.

[0042] Step S202-2: Define edges and calculate dynamic weights. The dynamic state causal graph construction module 200 establishes directed edges between nodes in the graph to represent multi-dimensional relationships between entities. In one embodiment, the types of edges include, but are not limited to, physical connection edges, spatial proximity edges, task-related edges, and inferential causal edges.

[0043] For physically connected edges, when two entities have a physically subordinate or fixed connection relationship (e.g., workers), Wearing a safety helmet ), establish an edge between their corresponding nodes, and set its weight to a constant 1.

[0044] For task-related edges, when multiple entities serve the same construction task (i.e., their node task identifiers) (Same as above), establish edges between the nodes corresponding to these entities, and set their weights to a constant 1.

[0045] For spatially adjacent edges, their weights This is used to quantify the spatial proximity of two nodes, essentially reflecting the attenuation characteristics of physical risk propagation in the spatial medium. The spatial proximity edge weight calculation formula is used for this purpose. The formula is as follows: ; in, Indicates from node arrive The weight of the spatially adjacent edge is a dimensionless numerical value with a range between (0,1]. and Representing nodes respectively and Location marker; Represents a node and Euclidean distance between corresponding positions; This represents the square of the distance; , is a preset distance influence factor, whose physical meaning corresponds to the standard deviation of a Gaussian distribution, determining the spatial decay rate of the risk influence. In this embodiment, The value is strictly set according to the type of operation: for the risk of falling objects from heights, The value is taken as the fall radius specified in the relevant safety regulations (e.g., 1 / 2 of the height); for hot work, The value is set to the limiting radius of the spark splash. This parameter establishes the effective domain of the risk in the spatial dimension. It is twice the variance of the distance influence factor and is used to scale the square of the distance; It is a dimensionless negative value, and its absolute value increases with increasing distance; It is the natural exponential function, used to map distance to non-linear weight values, where the closer the distance, the closer the weight is to 1.

[0046] For inferential causal edges, these edges represent potential causal relationships discovered through offline data analysis and not directly observed. The dynamic state causal graph construction module 200 periodically interacts with a causal inference model based on a Bayesian network. This model is not a black-box model, but a directed acyclic graph (DAG), where the nodes correspond to various risk factors in the historical accident database (such as "humidity > 80%", "fastener batch A", "stress anomaly"), and the edges correspond to conditional probability tables. During the model construction phase, the K2 scoring algorithm is used to perform structure learning on the historical accident report data, and the maximum likelihood estimation method is used for parameter learning, thereby quantifying the conditional probabilities between various risk factors. When the system detects a node In state and conditional probability ACCIDENT When the confidence threshold is exceeded (e.g., 0.75), the dynamic state causal graph construction module 200 establishes inferential causal edges between entity nodes that meet the conditions, with their weights... The value is directly taken as the conditional probability value. This mechanism enables the system to capture non-intuitive, implicit risk associations such as "a surge in failure rate of a specific batch of material under specific environmental conditions".

[0047] Step S202-3: Perform dynamic updates to the graph. A dynamic state cause-effect graph is a graph structure that changes over time. The dynamic state cause-effect graph construction module 200 uses an event-driven mechanism to update the graph. When the dynamic state cause-effect graph construction module 200 detects a specific change in the data input from upstream, such as the position coordinates of an entity... Updates, and related tasks When a rule condition for a change or inferential causal edge is triggered, the dynamic state causal graph construction module 200 will re-evaluate the relevant nodes and edges, performing operations such as adding, deleting, or updating nodes, and adding, deleting, or updating edges and their weights to ensure the graph structure is stable. It can accurately reflect the construction site at any time. The latest status.

[0048] The implementation of step S203 (quantifying the risk potential energy and its temporal gradient of each node through iterative calculation on the dynamic state causal graph) is explained in detail. This step is executed by the risk potential energy calculation engine 300, and can be broken down into the following sub-steps: Step S203-1: Calculate the risk potential energy of each node. The risk potential energy calculation engine 300 first calculates the risk potential energy of each node. The state vector of the associated entity Calculate the inherent risk of this node, i.e., its own risk potential energy. The calculation is performed using the node's own risk potential energy calculation formula, which is as follows: ; in, Represents a node At any moment Its own risk potential energy is a dimensionless scalar. It is a node The associated entity at time The state vector has been defined in step S201-3; It is a state vector A vector of weight coefficients of the same dimension, where each element Indicates the first The contribution weight of each state component to its own risk. Weight coefficient vector. It can be configured based on the knowledge base of domain experts, or trained using machine learning methods such as multivariate regression analysis on historical security data; It is a weight coefficient vector Transpose of; express and The dot product operation, that is, the weighted summation of each component of the state vector, yields the inherent risk value of the node when considering external influences.

[0049] Step S203-2: Perform the propagation and convergence calculation of the total risk potential energy. After calculating the risk potential energy of each node, the risk potential energy calculation engine 300 simulates the propagation and convergence of risk among nodes through iterative calculation based on the topology of the dynamic state causal graph, ultimately obtaining the total risk potential energy of each node. The total risk potential energy is calculated using the propagation and convergence formula, which is as follows: ; in, Indicates at time The In the round of iteration, nodes Total risk potential energy; Indicates at time The In the round of iteration, nodes in-degree neighbor nodes Total risk potential energy; Indicates at time From neighboring nodes Pointing to node The weight of the edge; This indicates that for all pointed-to nodes The neighboring nodes (i.e.) in-degree neighbor set Summation Representing neighboring nodes To the node The risk of transmission; , representing a node Total input risk received from all neighbors; Represents a node The risk inherent to the entity is a linear superposition of the risks it receives from external inputs. It is a non-linear activation function, preferably a logistic function (Sigmoid function). This function is used to simulate the saturation effect of risk, that is, when the input risk accumulates to a certain level, the growth rate of the total risk will slow down, so that the risk value is kept within a reasonable range, such as [0,1].

[0050] The initial condition for this iterative calculation is: The termination condition for iteration can be reaching a preset maximum number of iterations. Alternatively, when the risk potential energy changes of all nodes calculated in two consecutive iterations are both less than a preset convergence threshold. After the iteration terminates, the result obtained from the last iteration is... The final total risk potential energy at that moment is denoted as... .

[0051] Step S203-3: Perform risk potential energy gradient analysis. To further assess the changing trend of risk, the risk potential energy calculation engine 300, after obtaining the total risk potential energy at the current moment, calculates its rate of change relative to time, i.e., the risk potential energy gradient. The risk potential energy gradient calculation formula is used for the calculation. The risk potential energy gradient calculation formula is as follows: ; in, Represents a node At any moment The risk potential gradient is expressed in units of the reciprocal of a unit of time (e.g., seconds - 1). It is a node At the present moment Total risk potential energy; It is a node In the previous calculation cycle (time point) The total risk potential energy stored; It is the time interval between two risk calculations; numerator Indicates in The gradient represents the change in total risk potential energy over a given time period; the entire fraction represents the average rate of change of total risk potential energy. The sign of the gradient value indicates the direction of risk change: a positive value indicates that the risk is increasing, and a negative value indicates that the risk is decreasing. The magnitude of its absolute value reflects the drasticness of the risk change; the larger the absolute value, the faster the risk status changes.

[0052] The implementation of step S204 (upon receiving a construction task plan, parsing the plan to generate a sequence of future events, simulating the event sequence on a copy of a dynamic state cause-effect graph to predict future risks and generate an optimized construction plan) is described in detail. This step is executed by the task risk simulation module 400 and can be divided into the following steps: Step S204-1: Execute the parsing and eventification of the construction task plan. After the task risk simulation module 400 receives a construction task plan from an external system (such as BIM or project management software), it first parses the plan. The construction task plan can be in the form of a Gantt chart, task list, or work breakdown structure (WBS). The task risk simulation module 400 converts each task step in the plan into one or more standardized future events with specific timestamps. Thus, the entire construction task plan is transformed into a time-ordered sequence of future events. In one embodiment, each future event can be defined as a data structure, such as a tuple containing the event subject, action, time, and related parameters. , , , ).in, It is the entity identifier that performs the action. It refers to the specific type of action (such as "moving", "starting", "lifting"). It is the future moment when the event occurs. These are parameters related to the action (such as the coordinates of the target position to be moved).

[0053] Step S204-2: Perform hypothesis deduction based on dynamic state causal graph. The task risk simulation module 400 first obtains the dynamic state causal graph at the current moment. A complete copy is used as the initial state for the simulation. Subsequently, the task risk simulation module 400 traverses the sequence of future events generated in step S204-1 in chronological order. For each event in the sequence, the task risk simulation module 400 simulates the occurrence of the event on the graph copy, that is, updates the state of the corresponding node in the graph according to the event definition. For example, for a "move" event, the task risk simulation module 400 will update the corresponding node. Location marker For a "task change" event, update its task identifier. .

[0054] After each simulation update of the graph replica's state, especially when key attributes such as node positions change, causing the graph topology or edge weights to need recalculation, the task risk pre-simulation module 400 invokes the risk potential calculation engine 300. By executing the risk potential calculation process described in step S203 on the updated graph replica, the future moment is obtained. Below, the predicted risk potential of all nodes. By extrapolating the entire event sequence, the task risk simulation module 400 ultimately obtains a series of snapshots of the global risk distribution at key future time points.

[0055] Step S204-3: Perform risk path identification and planning optimization. The task risk simulation module 400 analyzes the predicted risk potential data generated in step S204-2 to identify potential high-risk links. The task risk simulation module 400 compares the predicted risk potential of each node with a preset "simulation risk threshold". The simulation risk threshold can be configured according to the project's risk tolerance; for example, it can be set to 50% to 80% of the real-time alarm threshold in step S205 to provide sufficient early warning margin. When it is found that the predicted risk potential of one or more nodes continuously or peakably exceeds the threshold within a certain period of time in the future, the sequence of task steps corresponding to that period is identified as a "high-risk path".

[0056] After identifying high-risk paths, the task risk simulation module 400 generates an optimized construction plan based on an internal optimization rule base. The optimization rule base may include, but is not limited to, the following rules: Time avoidance rule: If two tasks with high risk potential are scheduled in the same time period and their execution areas overlap, the system suggests that the time of one of the tasks be adjusted to avoid peak hours.

[0057] Spatial avoidance rules: If the planned route of a task (such as material transportation) needs to pass through an area identified as a risk hotspot, the system suggests replanning a route that detours around the area.

[0058] Resource replacement rules: If a specific entity (such as an aging device) is found to be the main source of significantly increased risk in the simulation, the system recommends replacing it with a spare or higher-performance device in the task.

[0059] Finally, the task risk simulation module 400 outputs the generated optimization suggestions. These suggestions can be the original plan with risk warnings and mitigation measures, or alternative plans with adjustments to time, space, or resources. These suggestions are then sent to the visualization and interactive terminal 600 for management decision-making.

[0060] The implementation method of step S205 (real-time monitoring of the risk potential energy and its temporal gradient of each node; when any indicator exceeds a preset threshold, performing risk source tracing to locate the main risk contributing source and generating structured intervention instructions) is described in detail. This step is executed by the adaptive scheduling and intervention module 500, and can be divided into the following sub-steps: Step S205-1: Execute the dual-threshold triggering of risk events. The adaptive scheduling and intervention module 500 continuously receives and monitors the real-time total risk potential of each node output by the risk potential calculation engine 300. and its risk potential gradient The adaptive scheduling and intervention module 500 employs a dual-threshold triggering mechanism to determine whether an intervention process needs to be initiated. The triggering condition is: when the total risk potential energy of any node exceeds a preset risk potential energy threshold. Or, the absolute value of its risk potential gradient exceeds a preset risk potential gradient threshold. Risk tracing will be initiated immediately upon such occurrence. Thresholds can be set based on historical security data statistics; for example, [the threshold can be set as follows]. Set to the 95th percentile of the risk potential under normal operating conditions; or configure according to the benchmark value in industry safety standards.

[0061] Step S205-2: Perform source tracing based on risk contribution. After triggering the intervention process, the adaptive scheduling and intervention module 500 will target the nodes that triggered the alarm. Perform risk source tracing to locate the main sources of risk contribution. The tracing process needs to distinguish whether the risk originates from the node itself (inherent risk) or is propagated from its upstream neighbor nodes (introduced risk).

[0062] First, the adaptive scheduling and intervention module has 500 computing nodes. Total input risk This value is the sum of all risk components before the activation function takes effect: ; Subsequently, their respective risk contribution rates were calculated. Contribution of incoming risk to each upstream neighbor node : Self-contribution to risk This represents the proportion of inherent risk of a node to the total input risk: ; Contribution of Ingress Risk , indicating that by neighboring nodes The proportion of incoming risk to total input risk: ; in, For nodes Total input risk prior to activation; For nodes Its own risk potential energy; Representative node Any in-degree neighbor node; the definitions of other symbols are the same as above.

[0063] The adaptive scheduling and intervention module 500 will determine its own risk contribution. Contribution of ingress risk to all upstream neighbors Perform a unified sorting. Identify the "risk sources" (which can be the node itself or its neighboring nodes) whose contribution meets the preset conditions as the main risk contribution sources.

[0064] The preset conditions could be: risk sources whose risk contribution exceeds a specific threshold (e.g., 20%); or several top-ranking risk sources whose cumulative contribution accounts for 80% of the total input risk. This method allows for accurate determination that high risk is an internal factor. High) or external factors ( (Dissemination) takes the lead, thereby achieving more precise intervention.

[0065] Step S205-3: Generate structured intervention instructions. After locating the main risk contributors, the adaptive scheduling and intervention module 500 generates structured intervention instructions based on an internal rule engine. This rule engine stores a series of "IF-THEN" rules, which match and generate corresponding instructions based on the type and attributes of the risk sources. In one embodiment, the rules may include: If the risk source is a certain device node, and its high risk is mainly caused by its own condition (such as excessive stress), then a maintenance work order instruction is generated for that device.

[0066] If the risk source is a personnel node, and its high risk is due to entering a high-risk area (i.e., the weight of its spatial neighbor edges increases), then an evacuation alarm command is sent to the personnel's wearable device.

[0067] If the risk source is an environmental monitoring node (such as excessive toxic gas concentration), then evacuation instructions will be broadcast to all personnel nodes in the affected area.

[0068] Intervention instructions are formatted into a standard data structure to facilitate interface communication with external systems (such as work order management systems and on-site broadcasting systems). For example, a JSON-formatted instruction can be represented as: "{"target_id":"device_123","action":"MAINTENANCE_REQUEST","priority":"HIGH","details":"Stress value exceeded threshold"}", which defines the intervention target, the action to be performed, the priority, and the specific reason.

[0069] The implementation method of step S206 (visualizing and interacting with risk potential energy, time gradient, risk contribution sources, and optimized construction plan in a 3D digital twin scene) is explained in detail. This step is executed by the visualization and interaction terminal 600, and can be broken down into the following sub-steps: Step S206-1: Construct and map the data of the 3D digital twin scene. The visualization and interactive terminal 600 loads the Building Information Model (BIM) of the construction scene and uses it as the geometric basis of the 3D scene. This is achieved by receiving dynamic state graphs in real time. Node information (especially location identifiers) This method maps virtual models representing various entities (such as people and equipment) to their corresponding locations in a 3D scene, forming a digital twin environment synchronized with the physical site. The virtual models can be general 3D models corresponding to the entity type (such as humanoid models or block models), or refined models from an equipment library, allowing for visual differentiation between different entities. For scene rendering, those skilled in the art can use graphics engines such as WebGL, Unity3D, or Unreal Engine, which are well-known technologies in the field and will not be elaborated upon here.

[0070] Step S206-2: Perform the visualization of multi-dimensional risk information. The visualization and interaction terminal 600 visualizes the data output by the risk potential calculation engine 300 and the adaptive scheduling and intervention module 500 in the following ways: Risk potential for a single entity This is achieved by changing the color of the risk potential energy on the 3D model. Specifically, the value of the risk potential energy (e.g., within the range of [0,1]) can be linearly or non-linearly mapped to a preset color gradient (e.g., from green to 0, to yellow to 0.5, and then to red to 1).

[0071] For the overall risk distribution in a region, a spatial risk heatmap is overlaid on a specific horizontal section of the 3D scene. The risk density at any coordinate point in the scene is calculated using the spatial risk density calculation formula, which is: ; in, Indicates at time Spatial coordinates Risk density value at the location; It is a node At any moment Total risk potential energy; It is a node The location identifier or coordinates are consistent with the definition in step S202-1; Represents coordinate points With nodes Location The Euclidean distance between them; It is a bandwidth parameter used to control the spatial range of the impact of risk potential energy. Its value can be associated with the safety impact radius of a specific operation (such as high altitude or hot work). It is a kernel function, such as the Gaussian kernel function, used to smoothly distribute the risk potential energy of a node to its surrounding space; the entire summation term represents the weighted summation of the risk potential energy of all nodes in the scene according to their distance from the target point, forming the total risk density of that point.

[0072] For the risk potential gradient A dynamic icon can be attached next to the 3D model of the corresponding entity. For example, an upward flashing arrow indicates that the risk is increasing rapidly, and a downward moving arrow indicates that the risk is decreasing. The flashing frequency or size of the arrow can be proportional to the absolute value of the gradient.

[0073] Step S206-3: Perform interactive analysis and intervention confirmation. The visualization and interactive terminal 600 provides interactive functions, allowing managers to conduct in-depth risk analysis. When a user clicks on a high-risk entity in the 3D scene, an information panel pops up. This panel lists in detail the current risk potential of the entity. Risk gradient And a risk contribution level The system displays a list of risk contribution sources arranged in descending order. When a user clicks on a risk contribution source in the list, the system highlights the causal relationship line from that risk source to the current entity in the 3D scene. The highlighting can be achieved by changing the color of the line, increasing its width, or adding glowing or dynamic flowing effects, visually indicating the risk transmission path. For intervention commands automatically generated by the system, a pop-up window is presented to the administrator, providing options for "Confirm Execution," "Modify," or "Ignore," thus achieving closed-loop management through human-machine collaboration.

[0074] Step S206-4: Visualize the results of the risk simulation. The visualization and interactive terminal 600 provides a dedicated display interface for the output of the task risk simulation module 400. This interface includes a timeline control, allowing users to drag the timeline slider to view the predicted risk heatmap and risk status of each entity at different future times when executed according to the original plan or the optimized plan. The system can display the risk evolution of the original plan and the risk evolution of the optimized plan side-by-side or overlaid, thus clearly comparing the effect of planning optimization.

[0075] See attached document Figure 1 The construction safety risk dynamic assessment and simulation system of the present invention can be deployed in a hardware environment that integrates cloud server, edge computing node and field terminal equipment.

[0076] The system's core backend services, including a data acquisition and preprocessing module 100, a dynamic state causal graph construction module 200, a risk potential calculation engine 300, a task risk pre-simulation module 400, and an adaptive scheduling and intervention module 500, can be deployed on one or more servers. The servers can be virtual server instances deployed on cloud platforms (e.g., Alibaba Cloud, Amazon AWS) or physical servers deployed in local data centers. In one specific embodiment, the server nodes are configured with high-performance multi-core central processing units (CPUs, such as Intel Xeon series), large-capacity memory (e.g., 256GB or more), high-speed solid-state drives (SSDs), and graphics processing units (GPUs, such as NVIDIA A100) for accelerating graph computation and machine learning models.

[0077] At the software architecture level, the backend core modules can be implemented as a set of independent microservices, running in a containerized environment (e.g., Docker) and managed by a container orchestration system (e.g., Kubernetes). This architecture allows for differentiated deployment of services with different characteristics; for example, more GPU resources can be allocated to the compute-intensive risk potential calculation engine 300, while better network and storage resources can be configured for the I / O-intensive data acquisition and preprocessing module 100. Data exchange between services can be performed through a high-performance message queuing system (e.g., Apache Kafka or RabbitMQ) to handle large amounts of real-time data streams from the field. For data storage, a hybrid database architecture can be adopted: a time-series database (e.g., InfluxDB) can be used to store time-series data generated by sensors; a graph database (e.g., Neo4j) can be used to store and query the structure of dynamic state causal graphs; and a relational database (e.g., PostgreSQL) can be used to store structured data such as project configurations and user information.

[0078] At the construction site, various IoT sensing and positioning devices are deployed, serving as the data source for the data acquisition and preprocessing module 100. These devices may include: smart safety helmets integrated with ultra-wideband or GPS positioning modules; stress / strain sensors mounted on scaffolding or supporting structures; temperature, humidity, and hazardous gas sensors monitoring the construction environment; and high-definition cameras for visual analysis. These devices transmit data to the backend server via wireless network technologies such as 5G, Wi-Fi 6, or LoRa.

[0079] In an optional embodiment, an edge computing gateway can be deployed in-situ. For example, for video streams captured by high-definition cameras, the edge computing gateway incorporates a lightweight convolutional neural network (CNN) model based on an improved YOLOv5 architecture. This model is designed to extract structured entity state information from unstructured video data. Specifically, the input layer receiver size of this model is adjusted to... The network uses RGB image frames; the backbone network employs a CSPDarknet53 structure for feature extraction and introduces a Ghost Module to reduce computational cost; the neck network uses a PA Net structure to achieve multi-scale feature fusion; the output layer outputs the target bounding box coordinates. The model is constructed using tensors representing the probability of target categories (e.g., personnel, safety helmets, construction vehicles) and their confidence scores. Before deployment, supervised training is performed on an image dataset collected from the construction site, encompassing multiple working conditions (e.g., nighttime, rainy / foggy days). During training, CIoU Loss (or GIoU Loss / DIoU Loss; depending on the algorithm used, CIoU is commonly used in improved versions of YOLOv5) is employed as the bounding box regression loss function, and binary cross-entropy is used as the classification loss function. After large-scale iterative training on a cloud server and quantization pruning using TensorRT, the generated inference engine is deployed to the NPU (Neural Processing Unit) of the edge gateway. The edge gateway only uploads the structured data output by the model inference (i.e., the identified entity categories and location coordinates) to the backend, thereby achieving real-time perception of on-site entities while ensuring privacy and bandwidth efficiency.

[0080] The visualization and interaction terminal 600 is typically a high-performance workstation, personal computer, or tablet computer equipped with a dedicated graphics card that supports 3D graphics acceleration (e.g., NVIDIA GeForce RTX series) to smoothly render digital twin scenes. The client applications running on it can be web applications accessed through a web browser, developed using technologies such as WebGL and Three.js; or standalone desktop applications developed using game engines (e.g., Unity3D or Unreal Engine) to provide richer interactions and visual effects.

Claims

1. A construction safety equipment state monitoring system based on data analysis processing, characterized by, include: The data acquisition and preprocessing module is used to acquire raw data from various data sources at the construction site and process it into standardized state vectors. The dynamic state cause-effect graph construction module is electrically connected to the data acquisition and preprocessing module. It is used to receive the standardized state vector and construct and update the dynamic state cause-effect graph in real time based on the standardized state vector and the location and task information of the entity. The risk potential calculation engine is electrically connected to the dynamic state causal graph construction module. It is used to receive the dynamic state causal graph and perform iterative calculations based on the topology and node states of the dynamic state causal graph to quantify the risk potential and time gradient of the graph nodes. The task risk simulation module is electrically connected to the risk potential energy calculation engine. After receiving the construction task plan, it calls the risk potential energy calculation engine to perform simulation, identify potential risks, and generate an optimized construction plan. The adaptive scheduling and intervention module is electrically connected to the risk potential energy calculation engine. It is used to receive the risk potential energy and the time gradient and monitor them in real time. When the risk exceeds a preset threshold, it performs risk source tracing and generates structured intervention instructions. The visualization and interactive terminal is electrically connected to the risk potential calculation engine, the task risk pre-simulation module, and the adaptive scheduling and intervention module, respectively, and is used to receive and present the risk potential, the optimized construction plan, and the structured intervention instructions to the user.

2. The construction safety equipment state monitoring system based on data analysis processing according to claim 1, characterized in that, When the data acquisition and preprocessing module processes the raw data to generate the standardized state vector, it is specifically used for: Acquire raw data from multiple heterogeneous sources; The original data is preprocessed to obtain preprocessed data; The preprocessed data is processed into state vectors, and normalized state components are calculated using the state component normalization formula. All the normalized state components of the entity are then combined to form the standardized state vector.

3. The construction safety equipment state monitoring system based on data analysis processing according to claim 1, characterized in that, The dynamic state cause-effect graph construction module is specifically used for: The monitored entities at the construction site are abstracted and defined as graph nodes, wherein each graph node is uniquely determined by a triple containing an entity identifier, a task identifier, and a location identifier. Based on the received standardized state vector and the entity's position and task information, the graph nodes in the dynamic state causal graph are created, updated, or deleted in real time.

4. The construction safety equipment status monitoring system based on data analysis and processing according to claim 1, characterized in that, The dynamic state cause-effect graph construction module is also specifically used for: Directed edges are established between the graph nodes in the dynamic state causal graph to represent the relationships between entities. The types of directed edges include physical connection edges, task association edges, spatial proximity edges, and inferential causal edges. Furthermore, for the spatially adjacent edges, the weight of the spatially adjacent edges is calculated using the spatially adjacent edge weight calculation formula, thereby quantifying the spatial proximity of two graph nodes.

5. The construction safety equipment status monitoring system based on data analysis and processing according to claim 1, characterized in that, The risk potential calculation engine, when quantifying the risk potential of the graph nodes, is specifically used for: Based on the state vector associated with the graph node in the dynamic state causal graph, the self-risk potential of the graph node is calculated by weighted summation of each component of the state vector using the node's own risk potential calculation formula. Based on the calculated self-risk potential energy and combined with the topology of the dynamic state causal graph, the total risk potential energy propagation and convergence formula is used for iterative calculation to simulate the propagation and convergence of risk between graph nodes, and finally obtain the total risk potential energy of the graph node.

6. The construction safety equipment status monitoring system based on data analysis and processing according to claim 5, characterized in that, When the risk potential energy calculation engine performs iterative calculations using the total risk potential energy propagation and convergence formula, the initial condition for the iterative calculation is its own risk potential energy, and the saturation effect of risk is simulated through a nonlinear activation function.

7. A construction safety equipment status monitoring system based on data analysis and processing according to claim 1, characterized in that, When generating the optimized construction plan, the task risk simulation module is specifically used for: The construction task plan is parsed to generate a time-ordered sequence of future events; Obtain a copy of the dynamic state causal graph at the current moment, and simulate the occurrence of the future event sequence on the copy in chronological order, and update the state of the copy; After each simulation update, the risk potential energy calculation engine is invoked to calculate the predicted risk potential energy, and high-risk paths are identified based on the predicted risk potential energy. The optimized construction plan is then generated according to the optimization rule base.

8. A construction safety equipment status monitoring system based on data analysis and processing according to claim 1, characterized in that, The adaptive scheduling and intervention module, when monitoring the risk potential energy and the time gradient in real time, is specifically used for: A dual-threshold triggering mechanism is used to determine whether to initiate the intervention process. The triggering condition is: when the total risk potential energy of any graph node exceeds the preset risk potential energy threshold, or the absolute value of the risk potential energy gradient exceeds the preset risk gradient threshold, the risk source tracing is initiated.

9. A construction safety equipment status monitoring system based on data analysis and processing according to claim 1, characterized in that, When the adaptive scheduling and intervention module executes the risk source tracing to generate the structured intervention instructions, it is specifically used for: Calculate the total input risk of the graph node that triggered the alarm; Based on the total input risk, calculate the risk contribution of the node itself and the input risk contribution of the upstream neighbor graph nodes respectively to obtain a risk contribution list; Based on the risk contribution list, risk sources that meet the preset conditions are identified as major risk contribution sources, and the structured intervention instructions are generated according to the type and attributes of the major risk contribution sources.

10. A construction safety equipment status monitoring system based on data analysis and processing according to claim 1, characterized in that, When presenting the risk potential to the user, the visualization and interactive terminal is specifically used for: Load the building information model of the construction scene and map the graph node information in the dynamic state cause-effect graph to the three-dimensional scene to form a digital twin environment; A spatial risk heatmap is superimposed on the horizontal cross-section of the digital twin environment. The spatial risk heatmap uses a spatial risk density calculation formula to calculate the risk density value of any coordinate point in the three-dimensional scene based on the risk potential energy of each graph node, presenting a regional overall risk distribution.

Images