An intelligent safety analysis system for construction sites

Abstract

The application discloses a kind of construction site intelligent safety analysis systems, it is related to construction safety analysis technical field, by the nonlinear coupling fusion of environmental disturbance mutation index GHT and operation stability space function Ψ, system forms the space-time dynamic risk intensity value Irik of unified expression construction risk intensity, significantly break through the bottleneck that environment risk and personnel risk are each in traditional model and cannot be linked quantization.Fusion result not only has continuity, also can dynamically respond the rapid change of environment and personnel state, support the dynamic adjustment of construction safety prevention and control.Color layer and transparency layer are used as weight factor source, and risk weight factor Fqa is constructed, to strengthen the correlation of visual perception and actual expression of risk intensity value.This design significantly improves the quantifiable and adjustable ability of system to risk space atlas, makes up the functional vacancy that only shows in traditional BIM visual layer expression and does not participate in judgment.

Description

Technical Field

[0001] The present invention relates to the technical field of construction safety analysis, and in particular to an intelligent safety analysis system for a construction site. Background Art

[0002] During construction, especially in high-risk jobs like those involving confined spaces, working at height, and operating enclosed equipment, accidents often occur due to a failure to promptly detect environmental anomalies or delayed recognition of worker behavior. In this context, environmental disturbance recognition mechanisms based on environmental parameters such as dust concentration, gas leakage, and sudden changes in temperature and humidity, and operational stability monitoring mechanisms based on inertial sensor data such as micro-jitter, attitude angle, and center of gravity offset, have become two core components of on-site intelligent risk identification systems, crucial for early risk perception and decision-making support.

[0003] Currently, construction sites rely heavily on single-dimensional risk perception and fixed rule-based triggering mechanisms for early warning. For example, evacuation prompts are only triggered when dust concentration exceeds a certain threshold or a gas sensor alarm sounds. These systems lack cross-dimensional fusion recognition capabilities and dynamic risk assessment models. Furthermore, most monitoring of worker safety status remains at the passive level, focusing on whether workers are wearing sensors or entering hazardous areas. This makes it difficult to assess fatigue trends, postural imbalances, or continuous microseismic instability in real time.

[0004] The fundamental reason for this situation is that construction site environmental factors are highly dynamic and personnel behavior is highly nonlinear. If the system cannot integrate this multi-source data in real time for correlational judgment, risk identification will be delayed and behavioral analysis data will be isolated. For example, in a confined pipe corridor, if dust concentration gradually increases and gas concentration becomes abnormal, but no response is triggered because the "single warning threshold" is not reached, and the operator exhibits continuous micro-shakes and posture deviations, if the system fails to combine these two and identify them as a "high-voltage coupling risk," secondary accidents such as explosions, falls, and hypoxia and fainting can easily occur. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides an intelligent safety analysis system for a construction site, which solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a construction site intelligent safety analysis system, including a multi-source data acquisition and preprocessing module, a disturbance and stability index extraction module, a spatial risk fusion assessment module, a BIM model risk mapping and visualization module, a risk level determination and early warning module, and a scheduling optimization and feedback learning module;

[0007] The multi-source data acquisition and preprocessing module collects environmental data and work behavior data from the construction site through sensors, fits them into the original data set YW, and performs preprocessing to obtain the safety data set AW;

[0008] The disturbance and stability index extraction module extracts features from the safety dataset AW, including the environmental disturbance mutation index GHT and the operation stability index HZW;

[0009] The spatial risk fusion assessment module fuses the acquired environmental disturbance mutation index GHT and operation stability index HZW to obtain the spatiotemporal dynamic risk intensity value Irik;

[0010] The BIM model risk mapping and visualization module maps the spatiotemporal dynamic risk intensity value Irik to the three-dimensional building space in the BIM model to obtain the risk visualization field map Frks;

[0011] The risk level determination and warning module determines the warning level A based on the acquired spatiotemporal dynamic risk intensity value Irik and the risk visualization field map Frks, and generates a warning response strategy;

[0012] The scheduling optimization and feedback learning module obtains the job optimization scheduling recommendation table Ωopt based on the warning level A and the warning response strategy.

[0013] Preferably, the multi-source data acquisition and pre-processing module includes a multi-parameter intelligent acquisition unit and a data processing and structuring unit;

[0014] The multi-parameter intelligent acquisition unit collects environmental data and work behavior data from the construction site through sensors installed at the construction site, including air dust concentration Adc, gas concentration Agc, ambient temperature AT, head vertical vibration amplitude Bso, work posture angle Bsz, and work center of gravity lateral offset rate Bxp, and fits them into the original data set YW;

[0015] Among them, the air dust concentration Adc is collected and obtained by the laser dust concentration sensor deployed in the high-altitude truss;

[0016] The gas concentration Agc is collected and obtained by a micro electrochemical multi-component gas sensor;

[0017] The ambient temperature AT is collected and obtained through distributed temperature sensors;

[0018] The vertical head vibration amplitude Bso is acquired through the three-axis IMU on the helmet worn by the operator;

[0019] The vertical head vibration amplitude Bso is obtained by the following formula:

[0020] ;

[0021] Where n represents the sliding window length, az(tk,i) represents the Z-axis acceleration of the head of the i-th worker at time tk, and paz(t,i) represents the mean Z-axis acceleration of the i-th worker at time t;

[0022] The operating attitude angle Bsz is obtained by calculating the direction angle change rate through the IMU worn by the operator;

[0023] The working posture angle Bsz is obtained by the following formula:

[0024] ;

[0025] Where a(t, i) represents the acceleration vector of the i-th worker in the direction of the main axis of the body at time t, g represents the unit vector in the direction of gravity, and arccos represents the inverse cosine function;

[0026] The lateral deviation rate of the working center of gravity, Bxp, is obtained by calculating the center of gravity trajectory estimated by the foot locator and gait modeling;

[0027] The lateral deviation rate of the working center of gravity Bxp is obtained by the following formula:

[0028] ;

[0029] Where xcg(t, i) represents the actual X-axis position of the i-th operator at time t, xref(t, i) represents the reference center axis position, and Wsafe represents the width of the safe working zone.

[0030] The data processing and structuring unit cleans and normalizes the original dataset YW to obtain the secure dataset AW;

[0031] Cleaning includes data denoising and missing value interpolation;

[0032] Data denoising is performed by processing the data in the original dataset YW using the sliding average method and edge anomaly filtering method to remove the noise in the data;

[0033] Missing value interpolation is performed by using linear interpolation to fill in the missing data in the original dataset YW;

[0034] Normalization processing is performed on the original data set YW by using maximum and minimum normalization to obtain the secure data set AW;

[0035] The security dataset AW is obtained using the following formula:

[0036] ;

[0037] Where AWu represents the u-th data item in the security dataset AW, YWu represents the u-th data item in the original dataset YW, minYWu represents the valley value of the u-th data item in the original dataset YW, and maxYWu represents the peak value of the u-th data item in the original dataset YW.

[0038] Preferably, the disturbance and stability index extraction module includes an environmental disturbance identification unit and an operation posture stability modeling unit;

[0039] The environmental disturbance recognition unit extracts features from the safety data set AW, including air dust concentration Adc, gas concentration Agc and ambient temperature AT, and obtains the short-term disturbance change rate Env;

[0040] The short-term disturbance change rate Env is obtained by the following formula:

[0041] ;

[0042] Where, represents the disturbance contribution coefficient of dust concentration, represents the disturbance contribution coefficient of gas concentration, represents the disturbance contribution coefficient of temperature, Adc(t) represents the air dust concentration at time t, Agc(t) represents the gas concentration at time t, Var[AT] represents the variance of temperature, and d represents the integral sign;

[0043] By using the nonlinear amplification disturbance function to analyze the short-term disturbance change rate Env, the environmental disturbance mutation index GHT is calculated and the environmental state is judged;

[0044] The environmental disturbance mutation index GHT is obtained by the following formula:

[0045] ;

[0046] Where tanh represents the hyperbolic tangent function, θ represents the upper limit adjustment factor of the response, e represents a constant, and λ represents the exponential enhancement coefficient;

[0047] The environment status is obtained by matching:

[0048] When the environmental disturbance mutation index GHT is less than 0.6, it means that the environmental state is stable and the construction is safe;

[0049] When 0.6≤Environmental disturbance mutation index GHT<1.0, it means that the environmental status is abnormal and there are risks at the construction site.

[0050] Preferably, the work posture stability modeling unit extracts features from the safety data set AW, including the vertical head vibration amplitude Bso, the work posture angle Bsz, and the lateral offset rate of the work center of gravity Bxp, and performs feature coupling to calculate and obtain a feature fusion index POS;

[0051] The feature fusion index POS is obtained by the following formula:

[0052] ;

[0053] Where, They represent the preset weight values ​​of the vertical vibration amplitude Bso, the working posture angle Bsz and the lateral deviation rate of the working center of gravity Bxp respectively, and Bsezref represents the reference value of the posture angle of the normal standing posture;

[0054] Among them, the reference value of the posture angle of normal standing posture Bsezref is obtained by collecting the posture angles Bsz of 50 workers under low-risk and standard conditions and calculating their average.

[0055] The acquired feature fusion index POS is analyzed, and the operation stability index HZW is calculated through the nonlinear saturation stability function to judge the personnel status;

[0056] The operation stability index HZW is obtained by the following formula:

[0057] ;

[0058] In the formula, ln represents the logarithmic function, and F represents the behavior amplification factor;

[0059] The personnel status is obtained by matching in the following ways:

[0060] When the operation stability index HZW is less than 0.4, it means that the personnel status is stable;

[0061] When 0.4≤operation stability index HZW<1.0, it indicates that the personnel status is abnormal and triggers a behavioral reminder.

[0062] Preferably, the spatial risk fusion assessment module includes an operation stability space mapping unit and a multi-source risk fusion calculation unit;

[0063] The work stability space mapping unit maps the individual work stability index HZW to the corresponding spatial position (x, y), and then constructs the work stability space function Ψ;

[0064] The operation stability space function Ψ is obtained by the following formula:

[0065] ;

[0066] Where Ψ(t, x, y) represents the spatial function of the operation stability at the spatial position (x, y) at time t, N represents the total number of operators, HZW(t, i) represents the operation stability index of the i-th operator at time t, exp represents the exponential function, sk represents the spatial diffusion control parameter, and (xi, yi) represents the spatial position of the i-th operator;

[0067] The multi-source risk fusion calculation unit nonlinearly couples the obtained operation stability spatial function Ψ with the environmental disturbance mutation index GHT to obtain the spatiotemporal dynamic risk intensity value Irik;

[0068] The spatiotemporal dynamic risk intensity value Irik is obtained by the following formula:

[0069] ;

[0070] Where c1 represents the environmental disturbance influence coefficient, and c2 represents the attitude stability influence coefficient.

[0071] The environmental disturbance influence coefficient c1 is obtained by counting the proportion of accidents dominated by environmental factors in the historical accident database of the construction site;

[0072] The attitude stability influence coefficient c2 is obtained by statistically analyzing the discrete values ​​of the working attitude angle Bsz and the lateral deviation rate Bxp of the working center of gravity of multiple workers in the same task, and integrating the deviation behavior density distribution using the deviation density kernel estimation method;

[0073] Preferably, the BIM model risk mapping and visualization module includes a risk space mapping and level classification unit and a visualization coding and layer rendering unit;

[0074] The risk space mapping and classification unit interpolates the obtained spatiotemporal dynamic risk intensity value Irik into the three-dimensional building model and performs highly layered projection to form a three-dimensional risk value 3DIrik, and converts it into a grade label Yt according to the risk grade standard;

[0075] The three-dimensional risk value 3DIrik is obtained by the following formula:

[0076] ;

[0077] Where 3DIrik(t, x, y, z) represents the three-dimensional risk value at the spatial position (x, y, z) at time t, Irik(t, x, y) represents the spatiotemporal dynamic risk intensity value at the spatial position (x, y) at time t, and η(z) represents the height risk adjustment factor.

[0078] The level label Yt is obtained by matching:

[0079] When the three-dimensional risk value 3DIrik is less than 0.3, it indicates the first level, the level label Yt=1, low risk, safe area, no response required;

[0080] When 0.3≤3D risk value 3DIrik<0.6, it indicates the second level, level label Yt=2, medium risk, controllable risk, continuous attention;

[0081] When 0.6≤3D risk value 3DIrik<0.8, it indicates the third level, level label Yt=3, high risk, potential accident area exists, and early warning is issued;

[0082] When 0.8≤3D risk value 3DIrik, it indicates the fourth level, level label Yt=4, extremely high risk, emergency response, and processing.

[0083] Preferably, the visualization coding and layer rendering unit encodes the obtained three-dimensional risk value 3DIrik and grade label Yt into multiple types of layers, dynamically renders them to the BIM model, and calculates and obtains the visualization color value Color and the transparency value Opacity through the color mapping function and the transparency adjustment function;

[0084] The visual color value Color is obtained by the following formula:

[0085] ;

[0086] Where Color(t, x, y, z) represents the visual color value at the spatial position (x, y, z) at time t, and fc represents the color mapping function;

[0087] The transparency value Opacity is obtained by the following formula:

[0088] ;

[0089] Where Ox represents the transparency adjustment coefficient;

[0090] The obtained visualization color value Color and transparency value Opacity are layered and combined with elements to obtain the risk visualization field map Frks;

[0091] ;

[0092] Where Frks(t, x, y, z) represents the risk visualization field map at the spatial location (x, y, z) at time t, Trajectory represents the personnel trajectory layer, and EscapeRoute represents the evacuation route layer.

[0093] Preferably, the risk level determination and early warning module includes a dynamic risk level generation unit and an early warning response strategy generation unit;

[0094] The risk level generation unit analyzes the risk visualization field map Frks and constructs the risk weight factor Fqa;

[0095] The risk weight factor Fqa is obtained by the following formula:

[0096] ;

[0097] Wherein, B1 represents the color coating adjustment coefficient, and B2 represents the transparency coating adjustment coefficient;

[0098] The color coating adjustment coefficient B1 is calculated by setting up multiple risk visualization interfaces with different color contrasts, recruiting target user groups (such as on-site construction workers and dispatch monitors) for testing, and recording the risk identification accuracy under different color changes;

[0099] The transparency coating adjustment coefficient B2 is obtained by testing the user's layer recognition accuracy under different transparency configurations when multiple layers (such as personnel trajectories, risk hotspots, and evacuation paths) are superimposed simultaneously.

[0100] The obtained risk weight factor Fqa is integrated with the spatiotemporal dynamic risk intensity value Irik to obtain the risk correction intensity value Rcv, which is then compared with the first risk threshold Tfh1 and the second risk threshold Tfh2 to obtain the warning level A;

[0101] The risk-corrected intensity value Rcv is obtained by multiplying the risk weight factor Fqa and the spatiotemporal dynamic risk intensity value Irik;

[0102] The warning level A is obtained by the following formula:

[0103] .

[0104] Preferably, the early warning response strategy generating unit generates an early warning response strategy according to the early warning level A;

[0105] The early warning response strategy is obtained by matching in the following ways:

[0106] When the warning level A=1, it indicates the first warning level, and the warning response strategy is to pop up a reminder and deepen the layer brightness;

[0107] When the warning level A=2, it indicates the second warning level, and the warning response strategy is red light warning, controlled entry and dispatch command linkage;

[0108] When the warning level A=3, it indicates the third warning level, and the warning response strategy is voice broadcast, notification to team leaders and restricted access.

[0109] When the warning level A=4, it indicates the fourth warning level, the warning response strategy is to activate the evacuation signal and turn off the main power switch.

[0110] Preferably, the scheduling optimization and feedback learning module records the warning response success rate Svr, the personnel risk exposure density μR and the average response time Art according to the warning level A and the warning response strategy, and obtains the operation optimization scheduling recommendation table Ωopt;

[0111] The warning response success rate Svr is obtained by the ratio of the number of successful responses Nsu(t) at time t to the total number of warnings triggered Nto(t) at time t;

[0112] The personnel risk exposure density μR is obtained by the following formula:

[0113] ;

[0114] Where R represents the designated area, including the working surface, platform and tunnel. represents the indicator function, where the condition is met and recorded as 1, and the condition is not met and recorded as 0;

[0115] The average response time Art is obtained by the following formula:

[0116] ;

[0117] Where, tac(ai) represents the ai-th warning triggering time, and twa(ai) represents the ai-th response completion time;

[0118] The job optimization scheduling recommendation table Ωopt is obtained by the following formula:

[0119] ;

[0120] Where argmin represents a minimization operation, which means finding the one that minimizes the objective function among all possible scheduling schemes Ω. μR(t,Ω) represents the risk exposure density of personnel under a given scheduling scheme Ω at time t. A higher value indicates that the personnel are more densely packed in the danger zone. dr represents the scheduling trade-off factor.

[0121] Among them, the scheduling trade-off factor dr is obtained by analyzing the impact trend of μR changes on response efficiency in multiple task scheduling samples, constructing a quadratic loss function and solving it.

[0122] The present invention provides a construction site intelligent safety analysis system, which has the following beneficial effects:

[0123] (1) During system operation, the system not only collects environmental data but also integrates operator behavior data through the multi-source data acquisition and preprocessing module to form a unified raw data set YW. It then performs standardization processing such as denoising, synchronization, and normalization, breaking down the barriers between different data sources and solving the perception fragmentation problem of scattered sensor data and lack of behavioral state perception in traditional systems. The system introduces the environmental disturbance mutation index GHT and the operation stability index HZW as core fusion feature indicators, which can simultaneously characterize environmental anomalies and unstable trends of operator behavior.

[0124] (2) By jointly constructing the short-term disturbance change rate Env through the three types of data: air dust concentration Adc, gas concentration Agc, and ambient temperature AT, the original static indicator is converted into a dynamic disturbance indicator in a continuously changing process, breaking the static lag bottleneck of "only looking at the absolute threshold and ignoring the change trend" in traditional safety judgment; in particular, by introducing the disturbance contribution coefficient mechanism, the unequal role of various environmental factors in the total disturbance is effectively quantified, improving the resolution and sensitivity of risk identification. By using exponential enhancement and hyperbolic tangent function processing mechanisms on the disturbance change rate, the system amplifies small disturbances into a highly sensitive response indicator, the environmental disturbance mutation index GHT, which can realize the early identification of hidden risks such as potential leakage and concentration accumulation.

[0125] Through a dimensional fusion modeling approach, three typical work posture indicators—vertical head vibration amplitude (Bso), work posture angle (Bsz), and work center of gravity lateral offset rate (Bxp)—are coupled to form a characteristic fusion index (POS). This solves the problem of previous systems relying solely on displacement or a single posture angle, which fails to fully reflect abnormal work trends. This fusion model can effectively identify signs of instability caused by fatigue, improper posture, or stance deviation, improving the ability to accurately perceive individual work risks.

[0126] (3) By nonlinearly coupling and fusing the environmental disturbance mutation index GHT with the spatial function Ψ of operational stability, the system forms a unified spatiotemporal dynamic risk intensity value Irik that expresses the intensity of construction risk. This significantly overcomes the bottleneck of the traditional model where environmental risk and personnel risk are independent and cannot be quantified in a coordinated manner. The fusion result not only has the ability to perceive continuity and intensity, but also can dynamically respond to rapid changes in environmental and personnel status, supporting dynamic adjustments to construction safety control.

[0127] (4) The color layer and the transparency layer are taken as weight factor sources to construct a risk weight factor Fqa, which is used to strengthen the correlation between the visual perception and the actual expression of the risk intensity value. This design significantly improves the quantifiable and adjustable ability of the system to the risk space map, makes up for the functional vacancy of the traditional BIM visual layer expression which is only used for display and does not participate in judgment, so that the layer is no longer a static additional vision but a core reference factor for early warning judgment. By introducing the risk correction intensity value Rcv and establishing a double-threshold risk level judgment system, the system can dynamically adjust the original risk value Irik according to the real-time space field characteristics, and enhance the robustness and flexibility of risk assessment. BRIEF DESCRIPTION OF DRAWINGS

[0128] Figure 1 It is a flowchart process schematic diagram of the construction site intelligent safety analysis system of the present application;

[0129] Figure 2 It is a flowchart process of the work optimization scheduling suggestion table of the present application;

[0130] Figure 3 It is a column chart of the short-term disturbance change rate of the present application;

[0131] Figure 4 It is a broken line chart of the feature fusion index of the present application. DETAILED DESCRIPTION

[0132] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0133] Embodiment 1, the present application provides a construction site intelligent safety analysis system, please refer to Figures 1 to 4 , including a multi-source data acquisition and preprocessing module, a disturbance and stability index extraction module, a spatial risk fusion evaluation module, a BIM model risk mapping and visualization module, a risk level judgment and early warning module, and a scheduling optimization and feedback learning module;

[0134] The multi-source data acquisition and preprocessing module collects the environmental data and the work behavior data of the construction site through sensors, fits the original data set YW, and performs preprocessing to obtain the safety data set AW;

[0135] The disturbance and stability index extraction module extracts features from the safety data set AW, including the environmental disturbance mutation index GHT and the work stability index HZW;

[0136] The spatial risk fusion assessment module fuses the acquired environmental disturbance mutation index GHT and operation stability index HZW to obtain the spatiotemporal dynamic risk intensity value Irik;

[0137] The BIM model risk mapping and visualization module maps the spatiotemporal dynamic risk intensity value Irik to the three-dimensional building space in the BIM model to obtain the risk visualization field map Frks;

[0138] The risk level determination and warning module determines the warning level A based on the acquired spatiotemporal dynamic risk intensity value Irik and the risk visualization field map Frks, and generates a warning response strategy;

[0139] The scheduling optimization and feedback learning module obtains the job optimization scheduling recommendation table Ωopt based on the warning level A and the warning response strategy.

[0140] In this embodiment, through a multi-source data acquisition and preprocessing module, the system not only collects environmental data but also integrates operator behavior data to form a unified raw data set YW. This data is then subjected to standardized processing, including denoising, synchronization, and normalization. This breaks down the barriers between different data sources and addresses the fragmented perception issues of traditional systems, where sensor data is dispersed and behavioral state perception is lacking. The system introduces the environmental disturbance mutation index (GHT) and the operation stability index (HZW) as core fusion feature indicators, which can simultaneously characterize environmental anomalies and unstable trends in operator behavior.

[0141] The Spatial Risk Fusion Assessment Module couples the disturbance index with the behavioral index to analyze the dynamic spatiotemporal risk intensity value Irik. This value is then rendered into the 3D building model via the BIM Model Risk Mapping and Visualization Module, forming a visual risk map Frks. The Risk Level Determination and Early Warning Module intelligently determines the risk level and generates a response plan based on the dynamic spatiotemporal risk intensity value Irik and the visual risk map Frks, ensuring differentiated response mechanisms for different risk levels. This system is suitable for high-risk operations such as underground construction, pipeline corridors, bridges, high-rise formwork, and tunnels, ensuring stable operation despite high data complexity, frequent personnel movement, and severe environmental disturbances.

[0142] Example 2: This example is explained in Example 1. Figure 1 ,Specifically: the multi-source data acquisition and pre-processing module includes a multi-parameter intelligent acquisition unit and a data ,processing and structuring unit;

[0143] The multi-parameter intelligent acquisition unit collects environmental data and work behavior data from the construction site through sensors installed at the construction site, including air dust concentration Adc, gas concentration Agc, ambient temperature AT, head vertical vibration amplitude Bso, work posture angle Bsz, and work center of gravity lateral offset rate Bxp, and fits them into the original data set YW;

[0144] Among them, the air dust concentration Adc is collected and obtained by the laser dust concentration sensor deployed in the high-altitude truss;

[0145] The gas concentration Agc is collected and obtained by a micro electrochemical multi-component gas sensor;

[0146] The ambient temperature AT is collected and obtained through distributed temperature sensors;

[0147] The vertical head vibration amplitude Bso is acquired through the three-axis IMU on the helmet worn by the operator;

[0148] The operating attitude angle Bsz is obtained by calculating the direction angle change rate through the IMU worn by the operator;

[0149] The lateral deviation rate of the working center of gravity, Bxp, is obtained by calculating the center of gravity trajectory estimated by the foot locator and gait modeling;

[0150] The data processing and structuring unit cleans and normalizes the original dataset YW to obtain the secure dataset AW;

[0151] Cleaning includes data denoising and missing value interpolation; data denoising processes the data in the original dataset YW through the sliding average method and edge anomaly filtering method to remove noise in the data;

[0152] Missing value interpolation is performed by using linear interpolation to fill in the missing data in the original dataset YW;

[0153] Normalization processing is performed on the original data set YW by using maximum and minimum normalization to obtain the secure data set AW;

[0154] The security dataset AW is obtained using the following formula:

[0155] ;

[0156] Where AWu represents the u-th data item in the security dataset AW, YWu represents the u-th data item in the original dataset YW, minYWu represents the valley value of the u-th data item in the original dataset YW, and maxYWu represents the peak value of the u-th data item in the original dataset YW.

[0157] This embodiment explicitly constructs a collaborative collection system for environmental data and work behavior data. The collection parameters cover six key indicators, including air dust concentration Adc, gas concentration Agc, ambient temperature AT, vertical head vibration amplitude Bso, work posture angle Bsz, and lateral offset rate of work center of gravity Bxp. These cover the environmental disturbance sources and work behavior inducements in the high-incidence accident factors at construction sites, breaking through the limitations of the traditional system's single sensing path and providing a more comprehensive and detailed perception basis for subsequent risk judgment.

[0158] The deployment of various sensors fully combines the characteristics of the construction scene. For example, laser dust sensors are deployed on high-altitude trusses to adapt to the law of dust rise; electrochemical gas sensors are deployed in a miniaturized, multi-component structure, suitable for embedding in confined spaces or ventilation blind spots; IMU sensors are embedded in helmets and foot devices to realize work motion analysis under high-frequency sampling; foot positioning combined with gait modeling is used to estimate the center of gravity offset, which effectively makes up for the shortcomings of traditional cameras or RFID positioning in capturing human posture details, demonstrating the system's high on-site adaptability and practicality.

[0159] This embodiment, through data processing and structuring units, achieves denoising, completion, and normalization of raw data, addressing common issues in construction environments such as severe data fluctuations, high acquisition noise, and missing time periods. In particular, the combined application of a sliding average method and edge anomaly filtering effectively enhances data stability and improves the robustness of subsequent risk modeling. Linear interpolation and maximum and minimum normalization ensure dimensional uniformity and relative comparability between different physical quantities, guaranteeing the accuracy of cross-parameter feature extraction.

[0160] Through the above-mentioned multi-parameter collection and high-quality processing process, the safety data set AW generated by the system is comprehensive, timely and stable after cleaning, becoming the core input for the subsequent extraction of environmental disturbance index and operation stability index, effectively supporting multi-dimensional data fusion analysis.

[0161] Example 3, this example is explained in Example 2, please refer to Figure 3 and Figure 4 ,Specifically: the disturbance and stability index extraction module includes the ,environmental disturbance recognition unit and the operation posture ,stability modeling unit;

[0162] The environmental disturbance recognition unit extracts features from the safety data set AW, including air dust concentration Adc, gas concentration Agc and ambient temperature AT, and obtains the short-term disturbance change rate Env;

[0163] The short-term disturbance change rate Env is obtained by the following formula:

[0164] ;

[0165] Where, represents the disturbance contribution coefficient of dust concentration, represents the disturbance contribution coefficient of gas concentration, represents the disturbance contribution coefficient of temperature, Adc(t) represents the air dust concentration at time t, Agc(t) represents the gas concentration at time t, Var[AT] represents the variance of temperature, and d represents the integral sign;

[0166] Specific examples:

[0167] Table 1: Calculation table of short-term disturbance change rate;

[0168]

[0169] By using the nonlinear amplification disturbance function to analyze the short-term disturbance change rate Env, the environmental disturbance mutation index GHT is calculated and the environmental state is judged;

[0170] The environmental disturbance mutation index GHT is obtained by the following formula:

[0171] ;

[0172] Where tanh represents the hyperbolic tangent function, θ represents the upper limit adjustment factor of the response, e represents a constant, and λ represents the exponential enhancement coefficient;

[0173] The environment status is obtained by matching:

[0174] When the environmental disturbance mutation index GHT is less than 0.6, it means that the environmental state is stable and the construction is safe;

[0175] When 0.6≤Environmental disturbance mutation index GHT<1.0, it means that the environmental status is abnormal and there are risks at the construction site.

[0176] The work posture stability modeling unit extracts features from the safety dataset AW, including the vertical head vibration amplitude Bso, the work posture angle Bsz, and the lateral offset rate of the work center of gravity Bxp, and performs feature coupling to calculate the feature fusion index POS;

[0177] The feature fusion index POS is obtained by the following formula:

[0178] ;

[0179] Where, They represent the preset weight values ​​of the vertical vibration amplitude Bso, the working posture angle Bsz and the lateral deviation rate of the working center of gravity Bxp respectively, and Bsezref represents the reference value of the posture angle of the normal standing posture;

[0180] Specific examples:

[0181] Table 2: Calculation table of operation posture feature fusion index;

[0182]

[0183] The acquired feature fusion index POS is analyzed, and the operation stability index HZW is calculated through the nonlinear saturation stability function to judge the personnel status;

[0184] The operation stability index HZW is obtained by the following formula:

[0185] ;

[0186] In the formula, ln represents the logarithmic function, and F represents the behavior amplification factor;

[0187] The personnel status is obtained by matching in the following ways:

[0188] When the operation stability index HZW is less than 0.4, it means that the personnel status is stable;

[0189] When 0.4≤operation stability index HZW<1.0, it indicates that the personnel status is abnormal and triggers a behavioral reminder.

[0190] This embodiment constructs the short-term disturbance change rate (Env) by combining three types of data: air dust concentration Adc, gas concentration Agc, and ambient temperature AT. This transforms the original static indicator into a dynamic disturbance indicator in a continuously changing process, breaking the static lag bottleneck of traditional safety judgments that "only considers absolute thresholds and ignores changing trends." In particular, by introducing a disturbance contribution coefficient mechanism, it effectively quantifies the unequal contributions of various environmental factors to the total disturbance, improving the resolution and sensitivity of risk identification. By applying exponential enhancement and hyperbolic tangent function processing mechanisms to the disturbance change rate, the system amplifies small disturbances into a highly sensitive response indicator, the environmental disturbance mutation index (GHT), which can achieve early identification of hidden risks such as potential leaks and concentration accumulation.

[0191] This module employs a fusion modeling approach to the behavioral dimension, coupling three typical work posture indicators—vertical head vibration amplitude (Bso), work posture angle (Bsz), and work center of gravity lateral offset rate (Bxp)—to form a feature fusion index (POS). This addresses the problem of previous systems relying solely on displacement or a single posture angle, which failed to fully reflect abnormal work trends. This fusion model effectively identifies precursors to instability caused by fatigue, improper posture, or stance deviation, improving the ability to accurately perceive individual work risks.

[0192] By mapping the fusion indicators using a nonlinear saturation function, the system generates the operational stability index (HZW). This not only achieves a quantitative transformation from feature fusion to behavioral assessment, but also establishes an automated judgment and response mechanism through a hierarchical threshold system. Compared to traditional methods that require manual identification or video inspections, the system can identify and trigger early warning alerts in real time during the operation, creating a truly dynamic closed-loop control system for operational risks.

[0193] Example 4: This example is explained in Example 3. Figure 2 ,Specifically: the spatial risk fusion assessment module includes an ,operation stability spatial mapping unit and a multi-source risk fusion calculation ,unit;

[0194] The work stability space mapping unit maps the individual work stability index HZW to the corresponding spatial position (x, y), and then constructs the work stability space function Ψ;

[0195] The operation stability space function Ψ is obtained by the following formula:

[0196] ;

[0197] Where Ψ(t, x, y) represents the spatial function of the operation stability at the spatial position (x, y) at time t, N represents the total number of operators, HZW(t, i) represents the operation stability index of the i-th operator at time t, exp represents the exponential function, sk represents the spatial diffusion control parameter, and (xi, yi) represents the spatial position of the i-th operator;

[0198] The multi-source risk fusion calculation unit nonlinearly couples the obtained operation stability spatial function Ψ with the environmental disturbance mutation index GHT to obtain the spatiotemporal dynamic risk intensity value Irik;

[0199] The spatiotemporal dynamic risk intensity value Irik is obtained by the following formula:

[0200] ;

[0201] Where c1 represents the environmental disturbance influence coefficient, and c2 represents the attitude stability influence coefficient.

[0202] The BIM model risk mapping and visualization module includes a risk space mapping and level classification unit and a visualization coding and layer rendering unit;

[0203] The risk space mapping and classification unit interpolates the obtained spatiotemporal dynamic risk intensity value Irik into the three-dimensional building model and performs highly layered projection to form a three-dimensional risk value 3DIrik, and converts it into a grade label Yt according to the risk grade standard;

[0204] The three-dimensional risk value 3DIrik is obtained by the following formula:

[0205] ;

[0206] Where 3DIrik(t, x, y, z) represents the three-dimensional risk value at the spatial position (x, y, z) at time t, Irik(t, x, y) represents the spatiotemporal dynamic risk intensity value at the spatial position (x, y) at time t, and η(z) represents the height risk adjustment factor.

[0207] The level label Yt is obtained by matching:

[0208] When the three-dimensional risk value 3DIrik is less than 0.3, it indicates the first level, the level label Yt=1, safe area, and no response is required;

[0209] When 0.3≤3D risk value 3DIrik<0.6, it indicates the second level, level label Yt=2, controllable risk, continuous attention;

[0210] When 0.6≤3D risk value 3DIrik<0.8, it indicates the third level, level label Yt=3, there is a potential accident area, and an early warning is issued;

[0211] When 0.8≤3D risk value 3DIrik, it indicates the fourth level, level label Yt=4, emergency response, and processing.

[0212] The visualization coding and layer rendering unit encodes the obtained three-dimensional risk value 3DIrik and grade label Yt into multi-type layers, dynamically renders them to the BIM model, and calculates the visualization color value Color and transparency value Opacity through the color mapping function and transparency adjustment function;

[0213] The visual color value Color is obtained by the following formula:

[0214] ;

[0215] Where Color(t, x, y, z) represents the visual color value at the spatial position (x, y, z) at time t, and fc represents the color mapping function;

[0216] The transparency value Opacity is obtained by the following formula:

[0217] ;

[0218] Where Ox represents the transparency adjustment coefficient;

[0219] The obtained visualization color value Color and transparency value Opacity are layered and combined with elements to obtain the risk visualization field map Frks;

[0220] ;

[0221] Where Frks(t, x, y, z) represents the risk visualization field map at the spatial location (x, y, z) at time t, Trajectory represents the personnel trajectory layer, and EscapeRoute represents the evacuation route layer.

[0222] This embodiment, for the first time, precisely maps each operator's operational stability index (HZW) to their spatial location coordinates, constructing a spatial operational stability function. This mapping breaks the limitations of previous behavioral indicators, which were limited to analyzing only individual levels. It enables the identification of regional trends in the accumulation, diffusion, and overlap of unstable behaviors in space. This allows for the early detection of localized risk hotspots such as "high-risk imbalance zones" and "multiple anomaly overlap zones," enhancing the system's forward-looking perception of potential accident locations.

[0223] By nonlinearly coupling and fusing the environmental disturbance mutation index (GHT) with the spatial function Ψ of operational stability, the system generates a unified spatiotemporal dynamic risk intensity value (Irik) that expresses the intensity of construction risk. This significantly overcomes the bottleneck of traditional models in which environmental and personnel risks operate independently and cannot be quantified in a coordinated manner. The fusion result not only possesses continuity and intensity perception capabilities, but also dynamically responds to rapid changes in environmental and personnel conditions, supporting dynamic adjustments to construction safety control measures.

[0224] This implementation further projects spatiotemporal risk intensity values ​​into the three-dimensional BIM model space using the height dimension, constructing a "structural risk volume map" that includes location, floor, and spatial stratification, and categorizes it into four levels of risk response areas using hierarchical labels. This approach addresses the problem of traditional systems where risk information cannot be modeled and structured for management, enabling risk analysis to move from the "data abstraction layer" to the "building structure layer," providing a technical foundation for refined management and control within construction scenarios.

[0225] This system dynamically encodes three-dimensional risk levels into visual layers through color mapping and transparency adjustment functions, overlaying personnel trajectories and evacuation routes to achieve a truly integrated display of risk, personnel, and emergency response. This layered approach enhances managers' intuitive understanding of on-site risk structures and improves operational decision-making efficiency, supporting intuitive identification of risk areas, rapid assessment of evacuation routes, and assisted planning of response strategies.

[0226] Example 5: This example is explained in Example 4. Please refer to Figure 1 and Figure 2,Specifically: the risk level determination and early warning module includes a dynamic risk level ,generating unit and an early warning response strategy generating unit;

[0227] The risk level generation unit analyzes the risk visualization field map Frks and constructs the risk weight factor Fqa;

[0228] The risk weight factor Fqa is obtained by the following formula:

[0229] ;

[0230] Wherein, B1 represents the color coating adjustment coefficient, and B2 represents the transparency coating adjustment coefficient;

[0231] The obtained risk weight factor Fqa is integrated with the spatiotemporal dynamic risk intensity value Irik to obtain the risk correction intensity value Rcv, which is then compared with the first risk threshold Tfh1 and the second risk threshold Tfh2 to obtain the warning level A;

[0232] The risk-corrected intensity value Rcv is obtained by multiplying the risk weight factor Fqa and the spatiotemporal dynamic risk intensity value Irik;

[0233] The warning level A is obtained by the following formula:

[0234] .

[0235] The early warning response strategy generation unit generates an early warning response strategy according to the early warning level A;

[0236] The early warning response strategy is obtained by matching in the following ways:

[0237] When the warning level A=1, it indicates the first warning level, and the warning response strategy is to pop up a reminder and deepen the layer brightness;

[0238] When the warning level A=2, it indicates the second warning level, and the warning response strategy is red light warning, controlled entry and dispatch command linkage;

[0239] When the warning level A=3, it indicates the third warning level, and the warning response strategy is voice broadcast, notification to team leaders and restricted access.

[0240] When the warning level A=4, it indicates the fourth warning level, the warning response strategy is to activate the evacuation signal and turn off the main power switch.

[0241] In this embodiment, the color layer and the transparency layer are taken as the weight factor sources to construct the risk weight factor Fqa, which is used to strengthen the correlation between the visual perception and the actual expression of the risk intensity value. This design significantly improves the quantifiable and adjustable ability of the system to the risk space map, makes up for the functional vacancy of the traditional BIM visual layer expression which is only for display and does not participate in judgment, and makes the layer not only a static additional vision but also a core reference factor for early warning judgment. By introducing the risk correction intensity value Rcv and establishing a double-threshold risk level judgment system, the system can dynamically adjust the original risk value Irik according to the real-time space field characteristics, and enhance the robustness and flexibility of risk assessment.

[0242] The system sets four levels of risk response mechanisms from the first level to the fourth level, and sets differentiated and multi-channel early warning linkage strategies for each level, effectively solving the problem of single early warning means and generalized response strategy in traditional systems. The response content of each level is highly consistent with the actual construction situation, not only meets the progressive response mode of “reminder-guidance-control-enforcement”, but also can be flexibly expanded through configuration strategy. Through multi-modal response strategies including layer highlighting, visual enhancement, red light control and voice broadcast, the system significantly enhances the intuitive cognition of the risk state of the site personnel, which helps to reduce the delayed response and wrong judgment. The early warning level directly links to the system entrance control and operation permission setting, improves the real-time of dispatching coordination and the matching degree of strategy, and provides a full-process automatic risk response chain for the construction site.

[0243] Embodiment 6, this embodiment is an explanation and description in embodiment 5, please refer to Figure 1 , in particular: the dispatching optimization and feedback learning module records the early warning response success rate Svr, the personnel risk exposure density μR and the average response time Art according to the early warning level A and the early warning response strategy, and obtains the operation optimization scheduling suggestion table Ωopt;

[0244] The early warning response success rate Svr is obtained by the ratio of the number of successful responses Nsu(t) at time t to the total number of early warnings Nto(t) triggered at time t;

[0245] The personnel risk exposure density μR is obtained by the following formula:

[0246] ;

[0247] In the formula, R represents the specified area, indicates the index function, and the condition is recorded as 1, and 0 is recorded if it is not satisfied;

[0248] The average response time Art is obtained by the following formula:

[0249] ;

[0250] Where, tac(ai) represents the ai-th warning triggering time, and twa(ai) represents the ai-th response completion time;

[0251] The job optimization scheduling recommendation table Ωopt is obtained by the following formula:

[0252] ;

[0253] Where argmin represents the minimization operation, μR(t,Ω) represents the personnel risk exposure density under the given scheduling scheme Ω at time t, and dr represents the scheduling trade-off factor.

[0254] This implementation systematically incorporates three quantitative metrics: early warning response success rate (Svr), personnel risk exposure density (μR), and average response time (Art). This not only enables measurable and traceable results for each risk response, but also provides a precise data foundation for optimizing subsequent scheduling strategies. This mechanism extends the system's focus from "pre-event identification and in-event response" to "post-event review and continuous learning," forming a self-feedback and self-iterative security management and control model, significantly enhancing the overall system's intelligence and autonomous learning capabilities.

[0255] The introduction of personnel risk exposure density μR as a key scheduling optimization metric is a major technical highlight of this embodiment. By quantifying the time-space density of individual exposure within high-risk areas, the system can accurately assess the degree of intersection between current work arrangements and risk areas, identifying problems such as personnel concentration and localized high-risk exposure caused by irrational scheduling. By dynamically calculating the response time and response success rate for each warning, the system not only monitors the effectiveness of emergency response execution but also identifies structural issues such as delays in the scheduling mechanism and breaks in the organizational response chain.

[0256] By integrating various risk assessment data and constructing an optimization function aimed at minimizing the density of personnel risk exposure, the system automatically generates a recommended table for optimal job scheduling, Ωopt, which directly guides subsequent personnel placement and job path setting. This mechanism effectively avoids typical scheduling errors such as repeated exposure, high-density work, and path conflicts, and improves the scientific nature and dynamic adaptability of construction site workspace configuration.

[0257] Compared to traditional systems that focus on responding to single risk events, this module aims for global optimization, systematically integrating early warning information, personnel spatial distribution, and temporal behavior trajectories. Through a feedback learning mechanism, it forms a sustainable and evolving scheduling optimization chain. Ultimately, while ensuring safety at the construction site, it also optimizes operational collaboration, efficiency, and rational staffing, thereby enhancing the overall effectiveness and controllability of on-site management.

[0258] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A construction site intelligent safety analysis system, characterized by: It includes multi-source data acquisition and preprocessing module, disturbance and stability index extraction module, spatial risk fusion assessment module, BIM model risk mapping and visualization module, risk level determination and early warning module, and scheduling optimization and feedback learning module; The multi-source data acquisition and preprocessing module collects environmental data and work behavior data from the construction site through sensors, fits them into the original data set YW, and performs preprocessing to obtain the safety data set AW; The disturbance and stability index extraction module extracts features from the safety dataset AW, including the environmental disturbance mutation index GHT and the operation stability index HZW; The spatial risk fusion assessment module fuses the acquired environmental disturbance mutation index GHT and operation stability index HZW to obtain the spatiotemporal dynamic risk intensity value Irik; The spatial risk fusion assessment module includes an operation stability spatial mapping unit and a multi-source risk fusion calculation unit; The work stability space mapping unit maps the individual work stability index HZW to the corresponding spatial position (x, y), and then constructs the work stability space function Ψ; The operation stability space function Ψ is obtained by the following formula: ； Where Ψ(t, x, y) represents the spatial function of the operation stability at the spatial position (x, y) at time t, N represents the total number of operators, HZW(t, i) represents the operation stability index of the i-th operator at time t, exp represents the exponential function, sk represents the spatial diffusion control parameter, and (xi, yi) represents the spatial position of the i-th operator; The multi-source risk fusion calculation unit nonlinearly couples the obtained operation stability spatial function Ψ with the environmental disturbance mutation index GHT to obtain the spatiotemporal dynamic risk intensity value Irik; The spatiotemporal dynamic risk intensity value Irik is obtained by the following formula: ； Where c1 represents the environmental disturbance influence coefficient, c2 represents the attitude stability influence coefficient; The BIM model risk mapping and visualization module maps the spatiotemporal dynamic risk intensity value Irik to the three-dimensional building space in the BIM model to obtain the risk visualization field map Frks; The risk level determination and warning module determines the warning level A based on the acquired spatiotemporal dynamic risk intensity value Irik and the risk visualization field map Frks, and generates a warning response strategy; The scheduling optimization and feedback learning module obtains the job optimization scheduling recommendation table Ωopt based on the warning level A and the warning response strategy.

2. The construction site intelligent safety analysis system according to claim 1, characterized in that: The multi-source data acquisition and pre-processing module includes a multi-parameter intelligent acquisition unit and a data processing and structuring unit; The multi-parameter intelligent acquisition unit collects environmental data and work behavior data from the construction site through sensors installed at the construction site, including air dust concentration Adc, gas concentration Agc, ambient temperature AT, head vertical vibration amplitude Bso, work posture angle Bsz, and work center of gravity lateral offset rate Bxp, and fits them into the original data set YW; Among them, the air dust concentration Adc is collected and obtained by the laser dust concentration sensor deployed in the high-altitude truss; The gas concentration Agc is collected and obtained by a micro electrochemical multi-component gas sensor; The ambient temperature AT is collected and obtained through distributed temperature sensors; The vertical head vibration amplitude Bso is acquired through the three-axis IMU on the helmet worn by the operator; The operating attitude angle Bsz is obtained by calculating the direction angle change rate through the IMU worn by the operator; The lateral deviation rate of the working center of gravity, Bxp, is obtained by calculating the center of gravity trajectory estimated by the foot locator and gait modeling; The data processing and structuring unit cleans and normalizes the original dataset YW to obtain the secure dataset AW; Cleaning includes data denoising and missing value interpolation; data denoising processes the data in the original dataset YW through the sliding average method and edge anomaly filtering method to remove noise in the data; Missing value interpolation is performed by using linear interpolation to fill in the missing data in the original dataset YW; Normalization processing is performed on the original data set YW by using maximum and minimum normalization to obtain the secure data set AW; The security dataset AW is obtained by the following formula: ； Where AWu represents the u-th data item in the security dataset AW, YWu represents the u-th data item in the original dataset YW, minYWu represents the valley value of the u-th data item in the original dataset YW, and maxYWu represents the peak value of the u-th data item in the original dataset YW.

3. The construction site intelligent safety analysis system according to claim 2, characterized in that: The disturbance and stability index extraction module includes an environmental disturbance recognition unit and an operation posture stability modeling unit; The environmental disturbance recognition unit extracts features from the safety data set AW, including air dust concentration Adc, gas concentration Agc and ambient temperature AT, and obtains the short-term disturbance change rate Env; The short-term disturbance change rate Env is obtained by the following formula: ； Where, represents the disturbance contribution coefficient of dust concentration, represents the disturbance contribution coefficient of gas concentration, represents the disturbance contribution coefficient of temperature, Adc(t) represents the air dust concentration at time t, Agc(t) represents the gas concentration at time t, Var[AT] represents the variance of temperature, and d represents the integral sign; By using the nonlinear amplification disturbance function to analyze the short-term disturbance change rate Env, the environmental disturbance mutation index GHT is calculated and the environmental state is judged; The environmental disturbance mutation index GHT is obtained by the following formula: ； Where tanh represents the hyperbolic tangent function, θ represents the upper limit adjustment factor of the response, e represents a constant, and λ represents the exponential enhancement coefficient; The environment status is obtained by matching: When the environmental disturbance mutation index GHT is less than 0.6, it means that the environmental state is stable and the construction is safe; When 0.6≤Environmental disturbance mutation index GHT<1.0, it means that the environmental status is abnormal and there are risks at the construction site.

4. The construction site intelligent safety analysis system according to claim 3, characterized in that: The work posture stability modeling unit extracts features from the safety dataset AW, including the vertical head vibration amplitude Bso, the work posture angle Bsz, and the lateral offset rate of the work center of gravity Bxp, and performs feature coupling to calculate the feature fusion index POS; The feature fusion index POS is obtained by the following formula: ； Where, They represent the preset weight values ​​of the head vertical vibration amplitude Bso, the working posture angle Bsz and the lateral deviation rate of the working center of gravity Bxp respectively, and Bszref represents the reference value of the posture angle of the normal standing posture; The acquired feature fusion index POS is analyzed, and the operation stability index HZW is calculated through the nonlinear saturation stability function to judge the personnel status; The operation stability index HZW is obtained by the following formula: ； In the formula, ln represents the logarithmic function, and F represents the behavior amplification factor; The personnel status is obtained by matching in the following ways: When the operation stability index HZW is less than 0.4, it means that the personnel status is stable; When 0.4≤operation stability index HZW<1.0, it indicates that the personnel status is abnormal and triggers a behavioral reminder.

5. The construction site intelligent safety analysis system according to claim 4, characterized in that: The BIM model risk mapping and visualization module includes a risk space mapping and level classification unit and a visualization coding and layer rendering unit; The risk space mapping and classification unit interpolates the obtained spatiotemporal dynamic risk intensity value Irik into the three-dimensional building model and performs highly layered projection to form a three-dimensional risk value 3DIrik, and converts it into a grade label Yt according to the risk grade standard; The three-dimensional risk value 3DIrik is obtained by the following formula: ； Where 3DIrik(t, x, y, z) represents the three-dimensional risk value at the spatial position (x, y, z) at time t, Irik(t, x, y) represents the spatiotemporal dynamic risk intensity value at the spatial position (x, y) at time t, and η(z) represents the height risk adjustment factor. The level label Yt is obtained by matching: When the three-dimensional risk value 3DIrik is less than 0.3, it indicates the first level, the level label Yt=1, the safe area, and no response is required; When 0.3≤3D risk value 3DIrik<0.6, it indicates the second level, level label Yt=2, controllable risk, continuous attention; When 0.6≤3D risk value 3DIrik<0.8, it indicates the third level, level label Yt=3, there is a potential accident area, and an early warning is issued; When 0.8≤3D risk value 3DIrik, it indicates the fourth level, level label Yt=4, emergency response, and processing.

6. The construction site intelligent safety analysis system according to claim 5, characterized in that: The visualization coding and layer rendering unit encodes the obtained three-dimensional risk value 3DIrik and grade label Yt into multi-type layers, dynamically renders them to the BIM model, and calculates the visualization color value Color and transparency value Opacity through the color mapping function and transparency adjustment function; The visual color value Color is obtained by the following formula: ； Where Color(t, x, y, z) represents the visual color value at the spatial position (x, y, z) at time t, and fc represents the color mapping function; The transparency value Opacity is obtained by the following formula: ； Where Ox represents the transparency adjustment coefficient; The obtained visualization color value Color and transparency value Opacity are layered and combined with elements to obtain the risk visualization field map Frks; ； Where Frks(t, x, y, z) represents the risk visualization field map at the spatial location (x, y, z) at time t, Trajectory represents the personnel trajectory layer, and EscapeRoute represents the evacuation route layer.

7. The construction site intelligent safety analysis system according to claim 6, characterized in that: The risk level determination and early warning module includes a dynamic risk level generation unit and an early warning response strategy generation unit; The risk level generation unit analyzes the risk visualization field map Frks and constructs the risk weight factor Fqa; The risk weight factor Fqa is obtained by the following formula: ； Wherein, B1 represents the color coating adjustment coefficient, and B2 represents the transparency coating adjustment coefficient; The obtained risk weight factor Fqa is integrated with the spatiotemporal dynamic risk intensity value Irik to obtain the risk correction intensity value Rcv, which is then compared with the first risk threshold Tfh1 and the second risk threshold Tfh2 to obtain the warning level A; The risk-corrected intensity value Rcv is obtained by multiplying the risk weight factor Fqa and the spatiotemporal dynamic risk intensity value Irik; The warning level A is obtained by the following formula: 。 8. The construction site intelligent safety analysis system according to claim 7, characterized in that: The early warning response strategy generation unit generates an early warning response strategy according to the early warning level A; The early warning response strategy is obtained by matching in the following ways: When the warning level A=1, it indicates the first warning level, and the warning response strategy is to pop up a reminder and deepen the layer brightness; When the warning level A=2, it indicates the second warning level, and the warning response strategy is red light warning, controlled entry and dispatch command linkage; When the warning level A=3, it indicates the third warning level. The warning response strategy is voice broadcast, notification to team leader and restriction of entry. When the warning level A=4, it indicates the fourth warning level, the warning response strategy is to activate the evacuation signal and turn off the main power switch.

9. The construction site intelligent safety analysis system according to claim 8, characterized in that: The scheduling optimization and feedback learning module records the warning response success rate Svr, personnel risk exposure density μR and average response time Art according to the warning level A and the warning response strategy, and obtains the operation optimization scheduling recommendation table Ωopt; The warning response success rate Svr is obtained by the ratio of the number of successful responses Nsu(t) at time t to the total number of warnings triggered Nto(t) at time t; The personnel risk exposure density μR is obtained by the following formula: ； In the formula, R represents the specified area, represents the indicator function, where the condition is met and recorded as 1, and the condition is not met and recorded as 0; The average response time Art is obtained by the following formula: ； Where, tac(ai) represents the ai-th warning triggering time, and twa(ai) represents the ai-th response completion time; The job optimization scheduling recommendation table Ωopt is obtained by the following formula: ； Where argmin represents the minimization operation, μR(t,Ω) represents the personnel risk exposure density under the given scheduling scheme Ω at time t, and dr represents the scheduling trade-off factor.

Images