A construction safety prompting method and system

Abstract

The present invention provides a construction safety prompt method and system, the method comprising: collecting and fusing multi-dimensional environmental data of a construction site to construct a real-time updated environmental digital twin model. According to the model and the individual physiological data of the workers, the safety status of each worker is adaptively evaluated to obtain an individualized dynamic safety threshold. Combining the environmental digital twin model, the individualized dynamic safety threshold and the real-time status data, the graph neural network reasoning technology is used to comprehensively evaluate the safety situation and predict the potential risk trend. According to the safety situation assessment results and the risk prediction trend, a situation-aware hierarchical processing method is used to generate adaptive early warning and intervention plans for different risk levels, and provide accurate safety prompts. The method monitors and predicts the safety status of workers in real time according to the dynamic changes of the construction site, improves the intelligence level and accuracy of construction safety management, and reduces the risk of accidents.

Description

Technical Field

[0001] The present invention relates to the technical field of construction safety, and in particular to a construction safety prompting method and system. Background Art

[0002] As the scale of construction and the complexity of the construction environment continue to increase, worker safety management and environmental monitoring of construction sites have become important components of ensuring construction safety. At present, most traditional safety monitoring systems rely on single environmental monitoring or personnel monitoring, which has problems of information islands and poor real-time performance. The environmental factors at the construction site are changeable, such as climate, temperature and humidity, gas concentration, etc., and the impact of these factors on the health and safety of workers is time-sensitive and dynamically changing. At the same time, the complex structure and spatial distribution of the construction site make the instability of positioning and communication signals a common problem. Although the smart devices worn by workers can collect certain vital signs data, the accuracy and real-time performance of vital signs data are often difficult to guarantee due to factors such as environmental noise, interference and transmission distance.

[0003] Existing construction safety monitoring methods usually divide environmental monitoring, worker monitoring and risk assessment into independent modules, lacking an effective fusion mechanism, resulting in the inability to comprehensively and accurately assess workers' safety status and environmental risks. In order to overcome the above problems, intelligent and data-driven safety management methods have gradually become a research hotspot in recent years. In particular, the application of digital twin technology in the field of construction safety has shown great potential due to its ability to update and dynamically simulate the construction site environment in real time. By integrating environmental data and individual physiological data of workers, individual safety assessments can be provided for each worker, further improving the accuracy of safety monitoring.

[0004] However, existing technologies still face many problems when facing the challenges of changing construction environments, individual differences among workers, and data complexity. For example, how to accurately assess workers’ safety status and predict possible risks in real time under dynamically changing environmental conditions, and how to obtain more reliable safety reminder solutions based on the fusion of multi-dimensional environmental and physiological data, etc. These problems need to be solved urgently. Summary of the invention

[0005] The main purpose of the present invention is to solve the technical problem that the existing construction safety monitoring system cannot accurately and real-time assess the safety status of workers and environmental risks when facing a complex construction environment;

[0006] A first aspect of the present invention provides a construction safety prompting method, the construction safety prompting method comprising:

[0007] Collect and integrate multi-dimensional environmental data of the construction site to obtain a real-time updated digital twin model of the environment;

[0008] Based on the digital twin model of the environment and the individual physiological data of the workers, the safety status of each worker is adaptively evaluated to obtain an individualized dynamic safety threshold;

[0009] Based on the environmental digital twin model, the individualized dynamic safety threshold and the real-time status data of the construction site, the safety situation is processed by graph neural network reasoning to obtain the current safety situation assessment results and risk prediction trends;

[0010] According to the security situation assessment results and risk prediction trends, the warning information is processed in a context-aware hierarchical manner to obtain adaptive warning and intervention plans for different risk levels, and safety prompts are given based on the adaptive warning and intervention plans.

[0011] Optionally, in a first implementation of the first aspect of the present invention, the collecting and fusing multi-dimensional environmental data of the construction site to obtain a real-time updated environmental digital twin model includes:

[0012] Edge computing is used to process laser scanning data, infrared thermal imaging data, and multi-parameter gas sensor data collected by multiple environmental monitoring stations to obtain a preliminary environmental situation map.

[0013] Perform multi-source fusion processing on the foundation settlement data of the pier construction area in the construction site, the hoisting equipment load data at the main beam hoisting position, the inclination data and the wind speed data of the aerial work platform to obtain the safety status index of the construction site;

[0014] According to the environmental situation map and safety status indicators, the construction site is three-dimensionally reconstructed and parameter-mapped to obtain a multi-level environmental digital twin model;

[0015] The environmental digital twin model is subjected to time series evolution analysis and processing to obtain a dynamically updated environmental change prediction model.

[0016] Optionally, in a second implementation of the first aspect of the present invention, the adaptively evaluating the safety status of each worker based on the environmental digital twin model and the individual physiological data of the worker to obtain an individualized dynamic safety threshold includes:

[0017] Perform multi-dimensional feature extraction processing on the individual physiological data of the workers to obtain the individual feature vectors of the workers;

[0018] Performing correlation analysis on the individual feature vectors and the environmental parameters in the environmental digital twin model to obtain a sensitivity matrix of the workers to different environmental factors;

[0019] Based on the sensitivity matrix and the environmental parameters, the potential risks of the workers are quantitatively calculated to obtain a preliminary safety status assessment value;

[0020] Performing time series analysis on the preliminary safety status assessment values ​​to obtain short-term fluctuations and long-term trend characteristics of the workers' safety status;

[0021] According to the short-term fluctuation and long-term trend characteristics, combined with the work intensity and duration information, the safety threshold is dynamically adjusted to obtain an individualized dynamic safety threshold.

[0022] Optionally, in a third implementation of the first aspect of the present invention, the quantitative calculation and processing of the potential risk of the worker based on the sensitivity matrix and the environmental parameters to obtain a preliminary safety status assessment value includes:

[0023] Performing singular value decomposition processing on the sensitivity matrix to obtain the main sensitive eigenvectors of the workers to various environmental factors;

[0024] Normalizing each environmental factor according to the environmental parameters to obtain a standardized environmental state vector;

[0025] Performing convolution operation on the main sensitive feature vector and the standardized environmental state vector to obtain a risk exposure index of the worker in the current environment;

[0026] Based on the risk exposure index, a nonlinear mapping process is performed on the worker's physiological limit threshold to obtain the worker's current safety tolerance;

[0027] According to the current safety tolerance and the individual physiological data of the workers, a preliminary safety status assessment value is obtained through fuzzy logic reasoning processing.

[0028] Optionally, in a fourth implementation of the first aspect of the present invention, the security situation is subjected to graph neural network reasoning processing based on the environmental digital twin model, the individualized dynamic security threshold, and the real-time status data of the construction site, and the current security situation assessment result and risk prediction trend are obtained, including:

[0029] According to the digital twin model of the environment, the positions of the workers and the distribution of the equipment, a heterogeneous graph construction process is performed on the construction site to obtain a dynamic graph structure representing the construction site;

[0030] The individualized dynamic safety threshold is used as the attribute of the worker node in the graph, and the node and edge attributes in the dynamic graph structure are encoded to obtain the corresponding graph embedding representation;

[0031] Based on the graph embedding representation and the real-time updated real-time state data, message passing network processing is performed on the state propagation between nodes to obtain updated node states;

[0032] Compare and analyze the node status of high-risk areas and individual dynamic safety thresholds, and strengthen the processing through the attention mechanism to obtain the safety risk factors that are of particular concern;

[0033] Calculate the safety score of the construction site according to the updated node status and the key safety risk factors to obtain the current safety situation assessment result;

[0034] Based on the current security situation assessment results, updated node status, key security risk factors and individualized dynamic security thresholds, the overall security situation is predicted in time and space to obtain short-term and medium- to long-term risk evolution trends.

[0035] Optionally, in a fifth implementation of the first aspect of the present invention, the node status of the high-risk area is compared and analyzed with the individualized dynamic safety threshold, and the safety risk factors that are focused on are enhanced through the attention mechanism, including:

[0036] Perform multi-dimensional feature decomposition on the node status in high-risk areas to obtain a status feature matrix that includes environmental, equipment, and personnel factors;

[0037] Performing element-level difference calculation processing on the state characteristic matrix and the corresponding individualized dynamic safety threshold to obtain a risk deviation matrix;

[0038] Performing singular value decomposition on the risk deviation matrix to obtain the eigenvectors of the main risk factors;

[0039] Based on the feature vector, the original node state is reconstructed to obtain the risk contribution ranking;

[0040] By utilizing the multi-head attention mechanism, the factors with higher risk contribution are weighted and processed to obtain a set of security risk factors that are the focus of attention.

[0041] Optionally, in a sixth implementation of the first aspect of the present invention, performing context-aware hierarchical processing on the warning information according to the security situation assessment result and the risk prediction trend to obtain adaptive warning and intervention plans for different risk levels, and providing security prompts based on the adaptive warning and intervention plans includes:

[0042] Perform multi-dimensional risk quantification on the security situation assessment results to obtain a risk level matrix that includes severity, urgency, and impact scope;

[0043] According to the risk level matrix and the real-time status data, the warning information is adaptively coded to obtain multi-modal warning information for different receiving objects;

[0044] Based on the multimodal warning information and the location of the worker, the safety reminder content is contextualized to obtain a multi-dimensional safety warning including visual, auditory and tactile feedback;

[0045] According to the multi-dimensional safety warnings and risk prediction trends, predictive analysis is performed on machinery and equipment and worker behaviors in high-risk areas to obtain adaptive warning and intervention plans, and safety prompts are given based on the adaptive warning and intervention plans.

[0046] A second aspect of the present invention provides a construction safety prompt system, the construction safety prompt system comprising:

[0047] The environmental perception module is used to collect and integrate multi-dimensional environmental data of the construction site to obtain a real-time updated environmental digital twin model;

[0048] An individual assessment module, used to perform adaptive assessment processing on the safety status of each worker based on the environmental digital twin model and the individual physiological data of the worker, and obtain an individualized dynamic safety threshold;

[0049] A situation reasoning module is used to perform graph neural network reasoning processing on the safety situation based on the environmental digital twin model, the individualized dynamic safety threshold and the real-time status data of the construction site to obtain the current safety situation assessment result and risk prediction trend;

[0050] The early warning and intervention module is used to perform context-aware hierarchical processing of the early warning information according to the security situation assessment results and risk prediction trends, obtain adaptive early warning and intervention plans for different risk levels, and provide safety prompts based on the adaptive early warning and intervention plans.

[0051] The above-mentioned construction safety prompt method and system collect and integrate multi-dimensional environmental data of the construction site to build a real-time updated environmental digital twin model. According to the model and the individual physiological data of the workers, the safety status of each worker is adaptively evaluated to obtain an individualized dynamic safety threshold. Combining the environmental digital twin model, individualized dynamic safety threshold and real-time status data, the graph neural network reasoning technology is used to comprehensively evaluate the safety situation and predict potential risk trends. According to the safety situation assessment results and risk prediction trends, a situation-aware hierarchical processing method is used to generate adaptive warning and intervention plans for different risk levels, providing accurate safety prompts. This method monitors the safety status of workers in real time and predicts and warns them according to the dynamic changes of the construction site, improves the intelligence level and accuracy of construction safety management, and reduces the risk of accidents.

[0052] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a schematic diagram of a first embodiment of a construction safety prompting method according to an embodiment of the present invention;

[0055] Figure 2 It is a schematic diagram of an embodiment of a construction safety reminder system in an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0058] To facilitate understanding of this embodiment, a construction safety prompting method disclosed in an embodiment of the present invention is first described in detail. Figure 1 As shown, the method comprises the following steps:

[0059] 101. Collect and integrate multi-dimensional environmental data of the construction site to obtain a real-time updated environmental digital twin model;

[0060] In one embodiment of the present invention, the multi-dimensional environmental data of the construction site is collected and fused to obtain a real-time updated environmental digital twin model, including: edge computing processing of laser scanning data, infrared thermal imaging data and multi-parameter gas sensor data collected by multiple environmental monitoring stations to obtain a preliminary environmental situation map; multi-source fusion processing of foundation settlement data in the pier construction area of ​​the construction site, load data of the hoisting equipment at the main beam hoisting position, inclination data and wind speed data of the aerial work platform to obtain safety status indicators of the construction site; three-dimensional reconstruction and parameter mapping of the construction site according to the environmental situation map and safety status indicators to obtain a multi-level environmental digital twin model; time series evolution analysis and processing of the environmental digital twin model to obtain a dynamically updated environmental change prediction model.

[0061] Specifically, multiple environmental monitoring stations at the construction site collect environmental data through laser scanning, infrared thermal imaging and gas sensors. Laser scanners use laser beams to measure the surface geometry and spatial position of the construction area, providing high-precision three-dimensional coordinate data for obtaining detailed structural information of the construction area. Infrared thermal imagers detect temperature changes and find abnormal temperature areas, which is essential for discovering equipment overheating or potential fire hazards. Gas sensors detect the concentration of harmful gases in the air in real time, such as carbon monoxide, methane, ammonia, etc. These data are pre-processed by edge computing technology, which refers to transferring data processing tasks to devices closer to the data source for processing instead of transmitting data to a remote server. This can reduce data transmission delays and improve processing efficiency. On the edge computing device, the raw sensor data is denoised, pre-processed and preliminarily analyzed to generate a preliminary environmental situation map. This picture shows the spatial temperature distribution, gas concentration, and spatial structural changes at the construction site, providing a basis for subsequent analysis.

[0062] Specifically, at the construction site, in addition to environmental data, there are other important safety data that need to be monitored and analyzed in real time, such as foundation settlement data, hoisting equipment load data, aerial work platform inclination data and wind speed data. In order to effectively process these data from different sources, multi-source data fusion technology is needed. First, the foundation settlement data is monitored in real time by sensors, which can show the settlement of the foundation in the construction area. If the settlement exceeds the predetermined threshold, it means that there is an unstable risk in the foundation. The hoisting equipment load data provides the load of the equipment during the hoisting process. Overload may cause equipment failure or structural damage. The inclination data of the aerial work platform reflects the stability of the platform. Excessive inclination angle may cause aerial work accidents. Wind speed data provides real-time wind conditions at the construction site. Especially during hoisting or aerial work, excessive wind speed may bring additional risks. When fusing these data, the synchronous timestamps of the sensors are first used to align the time points corresponding to each data source. Then, data fusion algorithms such as weighted average and Kalman filtering are used to integrate these multi-source data into a comprehensive safety status indicator. This indicator can comprehensively reflect the safety status of the construction site and be used for subsequent decision support and risk assessment.

[0063] Specifically, after having the environmental situation map and safety status indicators, this information needs to be converted into a three-dimensional digital twin model. The three-dimensional reconstruction process first uses laser scanning data to generate three-dimensional point cloud data of the construction site. These point cloud data contain the spatial geometry information of the construction site. Then, the point cloud data is processed using three-dimensional modeling software (such as AutoCAD or Revit) to convert it into a real three-dimensional model. Then, the data in the environmental situation map (such as temperature and gas concentration) is mapped to the three-dimensional model to ensure that the environmental data at each location corresponds to the spatial geometry. This process usually uses parameter mapping technology to map the temperature, gas concentration and other information in the environmental data to each corresponding point in the three-dimensional space to form a dynamic, real-time updated environmental digital twin model. To ensure real-time performance, the model is updated synchronously with the data acquisition frequency of the sensor, so that the digital twin model of the construction site can dynamically reflect environmental changes and accurately display the environmental status of each spatial location.

[0064] Specifically, time series evolution analysis and processing is to predict environmental changes based on the trend of environmental data changing over time at the construction site. Time series analysis techniques such as ARIMA (autoregressive integrated moving average model) or LSTM (long short-term memory neural network) are usually used to achieve this goal. First, historical data and real-time data are collected to form a time series data set. For environmental data such as temperature, humidity, and gas concentration, time series analysis can help analyze the changing patterns of these data and identify potential periodic or trend changes. By using the ARIMA model, the changing trends of these data in the future can be predicted. For example, an increase in temperature or an upward trend in gas concentration may indicate an increase in safety risks. LSTM is a deep learning model that is particularly suitable for capturing data change trends over a long span of time. By learning from historical data, LSTM can predict environmental changes in the next few hours or days. Time series evolution analysis can not only predict short-term environmental changes, but also dynamically adjust the environmental change prediction model according to the changing trend of environmental data to ensure that environmental changes at the construction site are responded to in a timely manner. By continuously updating the prediction model, the accuracy of the real-time monitoring system can be ensured, and real-time warnings can be provided for worker safety.

[0065] 102. Based on the digital twin model of the environment and the individual physiological data of the workers, the safety status of each worker is adaptively evaluated and processed to obtain an individualized dynamic safety threshold;

[0066] In one embodiment of the present invention, the adaptive evaluation of the safety status of each worker based on the environmental digital twin model and the individual physiological data of the workers to obtain an individualized dynamic safety threshold includes: performing multi-dimensional feature extraction on the individual physiological data of the workers to obtain an individual feature vector of the workers; correlating the individual feature vector with the environmental parameters in the environmental digital twin model to obtain a sensitivity matrix of the workers to different environmental factors; based on the sensitivity matrix and the environmental parameters, quantitatively calculating the potential risks of the workers to obtain a preliminary safety status assessment value; performing time series analysis on the preliminary safety status assessment value to obtain short-term fluctuations and long-term trend characteristics of the workers' safety status; and dynamically adjusting the safety threshold based on the short-term fluctuations and long-term trend characteristics in combination with work intensity and duration information to obtain an individualized dynamic safety threshold.

[0067] Specifically, the individual physiological data of workers will be collected in real time through various wearable devices, including vital signs data such as heart rate, blood pressure, body temperature, skin galvanic response, and respiratory rate. Through these devices, the physiological state data of workers can be obtained for subsequent analysis. In order to convert these physiological data into meaningful information, it is first necessary to perform multi-dimensional feature extraction processing on these data. For example, by statistically analyzing the continuous values ​​of heart rate and respiratory rate data, the characteristics of heart rate volatility, breathing pattern, and whether there are abnormal fluctuations in a specific time period can be extracted. These characteristics reflect the physical condition of workers and can be directly associated with changes in environmental factors and workload, thereby affecting their safety status. At the same time, by analyzing body temperature and skin galvanic response, it can be understood whether the workers are overworked, dehydrated, or have other physical discomfort caused by the harsh environment. Through this series of feature extraction processing, the individual feature vector finally obtained is a comprehensive expression of the worker's physiological state, which contains information in multiple dimensions and can accurately reflect the worker's physiological response under the current environment and working conditions.

[0068] Specifically, the individual feature vectors are analyzed and processed in association with the environmental parameters in the environmental digital twin model. In this process, it is first necessary to match the various environmental factors in the environmental model (such as temperature, humidity, gas concentration, wind speed, etc.) with the physiological feature vectors of the workers. By analyzing the physiological responses of workers under different environmental conditions, a sensitivity matrix between workers and environmental factors can be established. For example, some workers may be more sensitive to high temperature environments, and their heart rate or body temperature may fluctuate greatly, while some workers may be more sensitive to changes in gas concentration, resulting in an abnormal increase in breathing rate. Through multi-dimensional association analysis, the relationship between the worker's physiological data and environmental change factors is quantified to obtain a sensitivity matrix that describes the sensitivity of workers under different environmental conditions. This matrix is ​​the key to understanding the safety status of workers in a specific environment, and can reveal which environmental factors have a greater impact on worker safety and which factors have a smaller impact.

[0069] Specifically, based on the sensitivity matrix and environmental parameters, the potential risks of the workers are then quantified and calculated to obtain a preliminary safety status assessment value. In this process, a certain algorithm is applied to quantify the potential risks faced by the workers in combination with the workers' physiological state and environmental conditions. For example, if a worker is in a high temperature environment and engages in high-intensity labor for a long time, it may increase their physical exhaustion, thereby increasing the burden on the heart. At this time, it is necessary to combine factors such as temperature, humidity, and work intensity with the worker's individual sensitivity matrix to quantify a risk exposure index, and use this index to represent the worker's potential risk level. Through this calculation process, the current safety status assessment value of each worker can be obtained, reflecting their safety tolerance in the current environment. This value will serve as the basis for subsequent safety assessments and provide data support for workers' individual safety prompts and intervention measures.

[0070] Specifically, the preliminary safety status assessment values ​​will be processed by time series analysis to understand the short-term fluctuations and long-term trend characteristics of the workers' safety status. Time series analysis is to continuously track the workers' safety status assessment values ​​in order to analyze their trends over time. Through time series analysis, it is possible to identify short-term fluctuations in workers' safety status over a period of time, such as factors such as instantaneous fatigue and fluctuations in health status, as well as long-term trends, which reflect changes in the workers' overall physiological state, such as the continuous consumption of workers' physical strength due to long-term high-temperature work. By using common time series analysis methods, such as sliding average, exponential smoothing, or more complex LSTM (long short-term memory neural network), the safety status of workers in the future can be predicted and potential safety risks can be identified. The results of this analysis can reveal the fluctuations in workers' health status and make predictions and adjustments to subsequent work intensity and rest cycles.

[0071] Specifically, according to the short-term fluctuation and long-term trend characteristics of the workers' safety status, combined with their work intensity and work duration information, the safety threshold is dynamically adjusted to obtain an individualized dynamic safety threshold. In this step, the workers' work intensity and duration are key factors. For example, for workers who work in a high temperature environment for more than the prescribed time, the safety threshold needs to be gradually reduced as the work duration increases, because the risks faced by the workers gradually increase. According to the individual physiological state, work intensity and working hours of the workers, by dynamically adjusting the safety threshold, a personalized safety early warning system can be obtained, so that the safety threshold of each worker can match his current working environment and physical condition. In this way, the individualized dynamic safety threshold can reflect the actual safety risks faced by the workers in real time, thereby helping managers to take timely measures to ensure the safety of the workers. Throughout the process, the dynamic safety threshold can not only be adjusted in real time, but also be intelligently optimized based on historical data and individual changes of workers to ensure the accuracy and effectiveness of safety management.

[0072] Furthermore, the potential risks of the workers are quantitatively calculated based on the sensitivity matrix and the environmental parameters to obtain a preliminary safety status assessment value, including: performing singular value decomposition on the sensitivity matrix to obtain the main sensitive feature vectors of the workers to each environmental factor; normalizing each environmental factor according to the environmental parameters to obtain a standardized environmental state vector; performing convolution operation on the main sensitive feature vector and the standardized environmental state vector to obtain the risk exposure index of the workers in the current environment; based on the risk exposure index, performing nonlinear mapping on the physiological limit threshold of the workers to obtain the current safety tolerance of the workers; and obtaining a preliminary safety status assessment value through fuzzy logic reasoning based on the current safety tolerance and the individual physiological data of the workers.

[0073] Specifically, singular value decomposition (SVD) of the sensitivity matrix is ​​the core step of this technical solution. Through SVD, a matrix can be decomposed into the product of three matrices, thereby extracting the sensitive characteristics of workers to different environmental factors. It represents the workers' response to various environmental factors. Each column of the matrix represents the impact of an environmental factor on different workers, and each row represents the physiological response of workers under different environmental factors. By performing singular value decomposition on the sensitivity matrix, three matrices are obtained: , and ,in, is a left singular matrix containing the characteristic information of workers. Each column represents the potential sensitive characteristics of workers, reflecting the overall response pattern of workers to different environmental factors. It is a diagonal matrix containing singular values. The larger the singular value, the greater the proportion of the corresponding eigenvector in describing the workers' response to environmental factors. is a right singular matrix, which contains the characteristic information of environmental factors. Each row represents the potential characteristics of an environmental factor, reflecting how the environmental factor affects the worker's response. Through singular value decomposition, the main sensitive feature vectors of workers to each environmental factor can be extracted. These eigenvectors are the most important associations between environmental factors and individual physiological responses of workers, helping to identify which environmental factors should be given higher attention in safety management. Next, each environmental factor is normalized according to the environmental parameters to obtain a standardized environmental state vector. This is to ensure that different environmental factors have consistent dimensions so that their impact on the safety status of workers can be fairly evaluated. In the normalization process, each environmental parameter needs to be normalized first. After decentralizing the mean, divide it by the standard deviation, and the formula is as follows:

[0074]

[0075] in, is the environmental parameter The mean of is the standard deviation, is the standardized environmental state vector. This standardization step eliminates the imbalance between environmental factors due to differences in units and magnitudes, so that the impact of each environmental factor on the worker's safety status is treated uniformly in the calculation. The standardized environmental state vector provides a clear picture of the environmental changes faced by workers, which facilitates the subsequent risk assessment and safety warning calculation. Subsequently, the standardized environmental state vector The sensitive feature vector of the worker Perform convolution operation to obtain the risk exposure index of workers in the current environment The convolution operation can capture the interaction between workers' sensitive features and environmental changes, and thus quantify the risk exposure faced by workers. The convolution process is expressed as:

[0076]

[0077] in, is the worker’s sensitive feature vector to various environmental factors, is the normalized environment state vector, is the worker’s risk exposure index in the current environment. The convolution result reflects the comprehensive effect of how workers respond to different environmental factors. The higher the risk exposure index, the greater the potential risk faced by the worker. Therefore, the convolution operation can effectively assess the worker’s risk exposure in the current environment and provide key data for subsequent safety assessments. Next, based on the worker’s risk exposure index , the safety tolerance of workers is obtained by nonlinear mapping of their physiological limit thresholds The physiological limit threshold is the maximum physiological load that a worker can withstand under specific environmental conditions, such as maximum heart rate, respiratory rate, and body temperature. Due to changes in the environment and working conditions, the worker's safety tolerance needs to be dynamically adjusted according to the risk exposure index. The nonlinear mapping function associates the risk exposure index with the physiological limit threshold, and the formula is as follows:

[0078] ;

[0079] in, and is the adjustment parameter, is the risk exposure index, is the worker's safety tolerance. The characteristic of this nonlinear function is that when the risk exposure index When it is higher, the worker's safety tolerance The safety tolerance of workers can be adjusted flexibly to accurately reflect their physiological limits in different environments and work intensities.

[0080] Finally, through fuzzy logic reasoning, combined with the workers' individual physiological data and safety tolerance, a preliminary safety status assessment value is obtained. Fuzzy logic reasoning can handle complex situations with uncertainty and ambiguity, and is suitable for assessing the safety status of workers. Through fuzzy reasoning rules, taking into account the worker's physiological data (such as heart rate, body temperature, respiratory rate, etc.) and environmental risk exposure index and safety tolerance , derive the worker's safety status assessment value. Fuzzy rules, such as "if the worker's safety tolerance is low and the heart rate is high, the safety status is poor", and generate corresponding outputs based on these rules. Through defuzzification processing, the worker's current safety status assessment value is obtained ,This value can be used to monitor the safety status of workers in real time and provide ,intervention measures in time.

[0081] 103. Based on the environmental digital twin model, individualized dynamic safety thresholds and real-time status data of the construction site, the safety situation is processed by graph neural network reasoning to obtain the current safety situation assessment results and risk prediction trends;

[0082] In one embodiment of the present invention, the security situation is subjected to graph neural network reasoning processing based on the environmental digital twin model, the individualized dynamic safety threshold and the real-time status data of the construction site to obtain the current security situation assessment result and risk prediction trend, including: performing heterogeneous graph construction processing on the construction site according to the environmental digital twin model, the worker location and the equipment distribution to obtain a dynamic graph structure representing the construction site; using the individualized dynamic safety threshold as the attribute of the worker node in the graph, encoding the node and edge attributes in the dynamic graph structure to obtain the corresponding graph embedding representation; based on the graph embedding representation and the real-time updated real-time status data, The state propagation between nodes is processed through a message passing network to obtain an updated node state; the node state and the individualized dynamic safety threshold in the high-risk area are compared and analyzed, and the key safety risk factors are obtained through enhanced processing through the attention mechanism; according to the updated node state and the key safety risk factors, the safety score of the construction site is calculated to obtain the current safety situation assessment result; based on the current safety situation assessment result, the updated node state, the key safety risk factors and the individualized dynamic safety threshold, the overall safety situation is predicted and processed in a spatiotemporal sequence to obtain the short-term and medium-term risk evolution trends.

[0083] Specifically, a heterogeneous graph is constructed for the construction site based on the environmental digital twin model, worker location, and equipment distribution. The core of this step is to establish a dynamic graph structure based on the physical structure and safety data of the construction site. Each worker and equipment on the construction site can be regarded as a node in the graph, and the relationship between the nodes is represented by an edge, and the weight of the edge represents the interaction between the worker and the equipment or the environmental impact. The environmental digital twin model provides detailed information about each location on the construction site, including factors such as terrain, temperature and humidity, and gas concentration, and the location distribution of workers and equipment gives spatial information to the nodes in the graph. By constructing a heterogeneous graph, multi-dimensional information such as environmental data, personnel location, equipment distribution, and risk status can be combined to form a dynamically updated graph structure, which can not only show the static layout of the construction site, but also reflect dynamic environmental changes and worker activities. This graph structure has strong flexibility and can be updated in real time according to the actual changes on the construction site to ensure the dynamicity and accuracy of the graph.

[0084] Specifically, the individualized dynamic safety threshold is used as the attribute of the worker node in the graph, and the node and edge attributes in the dynamic graph structure are encoded to obtain the corresponding graph embedding representation. In this process, the individualized dynamic safety threshold is calculated based on the worker's current safety tolerance, work intensity and environmental factors, and is provided to the graph neural network as an attribute of the worker node. In addition to the spatial information of the worker's location, each node also needs to carry the safety threshold information related to the individual worker. The edge attributes represent the interactive effects between workers or between workers and equipment, such as the distance between workers, the load conditions of workers and equipment, etc. These attributes are converted into graph embedding representations through the encoding processing of the graph neural network. These representations not only retain the original information of the nodes, but also can fuse the relationship between nodes and edges to form a multi-dimensional representation with contextual information. These embedded representations enable the graph to fully reflect the safety situation of the construction site and provide structured input for subsequent state propagation and risk prediction.

[0085] Specifically, after obtaining the graph embedding representation, the state propagation between nodes is processed by a message passing network based on these representations and the real-time updated state data. This step involves the message passing mechanism in the graph neural network, in which the node updates its state based on its attributes and the information of the neighboring nodes. The state of each node depends not only on its own attributes (such as safety thresholds and real-time physiological data), but also on the neighboring nodes connected to it. For example, a node (worker) will update its own safety state to form a new node state after receiving the influence of the state of its neighboring nodes (equipment or other workers). This process is implemented through a graph convolutional network (GCN) or a graph attention network (GAT) to ensure the effective propagation of information in the graph. The update of node status can reflect the real-time changes in the construction site environment and safety status, and further provide data support for the safety situation assessment of the entire construction site.

[0086] Specifically, as the node status is updated, it is necessary to compare and analyze the node status of high-risk areas and the individualized dynamic safety threshold, and strengthen the processing through the attention mechanism. The goal of this process is to identify those areas with higher safety risks through comparative analysis, and further focus on the risk factors in these areas through the attention mechanism. The node status of high-risk areas usually includes some nodes with extreme environmental conditions (such as excessive gas concentration or extreme temperature) or large physiological loads (such as high heart rate of workers), and the safety status of these nodes is often fragile. Through the attention mechanism, the network can automatically identify and increase the attention of these high-risk area nodes, strengthen the safety risks of these areas, and give priority to their impact. The introduction of the attention mechanism not only improves the accuracy of the model, but also enhances the flexibility and sensitivity of the network in processing different node and edge attributes, thereby more accurately identifying potential high-risk factors.

[0087] Specifically, based on the updated node status and the safety risk factors of focus, the safety score of the construction site is calculated to obtain the current safety situation assessment result. The calculation of the safety score is based on the safety status of each node and its mutual influence with the adjacent nodes. Through the comprehensive evaluation of the node status and the enhanced attention to the high-risk area, the network can give the current overall safety status score of the construction site. This score reflects the overall safety situation of the construction site and can provide specific safety situation feedback to managers. The safety score can not only help assess the current risk level, but also provide targeted intervention suggestions for decision makers, such as increasing monitoring in high-risk areas or adjusting the work arrangements of workers. Based on the current safety situation assessment results, the updated node status, the safety risk factors of focus and the individualized dynamic safety threshold, the overall safety situation is processed by spatiotemporal sequence prediction to obtain the short-term and medium-term risk evolution trends. This process uses the spatiotemporal sequence prediction model, such as the time series model in the neural network (such as the time series graph convolutional network TGCN), combined with the safety assessment results of the current construction site and the real-time updated data, to predict the evolution of the safety situation in the future. By predicting the short-term and medium-term risk evolution trends, possible safety hazards can be identified in advance and countermeasures can be taken in advance. For example, as construction progresses, certain areas may gradually become dangerous due to increased equipment load. The predictive model can provide early warning of these potential risks and help the construction site take preventive measures.

[0088] Furthermore, the node status of the high-risk area and the individualized dynamic safety threshold are compared and analyzed, and the safety risk factors that are focused on are obtained through enhanced processing by the attention mechanism, including: multi-dimensional feature decomposition processing of the node status of the high-risk area to obtain a state feature matrix containing environmental, equipment and personnel factors; element-level difference calculation processing of the state feature matrix and the corresponding individualized dynamic safety threshold to obtain a risk deviation matrix; singular value decomposition processing of the risk deviation matrix to obtain eigenvectors of major risk factors; based on the eigenvector, the original node status is reconstructed to obtain a risk contribution ranking; using a multi-head attention mechanism, weight enhancement processing is performed on factors with higher risk contribution to obtain a set of safety risk factors that are focused on.

[0089] Specifically, first, the node status in the high-risk area is subjected to multi-dimensional feature decomposition processing, with the goal of subdividing the status information of each node so as to more clearly understand how factors such as environment, equipment and personnel affect the overall security situation. The node status in the high-risk area is usually composed of multiple dimensions, such as environmental factors (temperature and humidity, gas concentration), equipment factors (load, operating status) and personnel factors (physiological status, worker behavior, etc.). These different dimensions need to be feature decomposed to obtain a state feature matrix ,Each row of the matrix represents a node, which contains the characteristic information of the node in each dimension, and each column represents a specific dimension. Through decomposition, it is possible to quantify and describe in detail the various factors that affect node safety, thereby providing more accurate data support for subsequent risk calculation and analysis. Next, the state characteristic matrix and the corresponding individualized dynamic safety threshold are calculated at the element level to obtain the risk deviation matrix Individualized dynamic safety thresholds take into account the uniqueness of each node, such as physiological differences of workers or workload of equipment. These thresholds are usually set based on the maximum safe load of the node under specific environmental conditions. By calculating the difference between the state characteristics of each node and its safety threshold at the element level, the risk deviation matrix can be obtained. , represents the risk exposure of each node in each dimension. Specifically, the elements of the risk deviation matrix Indicates The node in The state deviation under the dimension is larger. The larger the deviation, the greater the security threat faced by the node. Then, singular value decomposition (SVD) is performed to extract the main risk factor eigenvectors from the risk deviation matrix. Singular value decomposition is a commonly used dimensionality reduction technique that can decompose the risk deviation matrix into three matrices: (indicates node characteristics), (including singular values, indicating the importance of risk factors) and (Indicates environmental characteristics). By using the risk deviation matrix SVD decomposition can extract the main risk factors that affect worker safety and quantify the impact of these factors. The eigenvector contains risk characteristics in various dimensions such as environment, equipment, and personnel, and can identify which factors have the greatest impact on the safety status of the node. The singular value reflects the weight of these risk factors. The larger the singular value, the greater the proportion of the corresponding risk factor in the safety assessment. Based on the eigenvector obtained from SVD, the original node state is reconstructed to obtain the risk contribution ranking of the node. The reconstruction process uses the extracted eigenvector and performs a weighted summation on it with the original node state data to restore the safety risk contribution of each node. The risk contribution of each node It can be calculated by the following formula

[0090] ;

[0091] in and is the matrix element obtained by SVD decomposition, is a singular value, indicating The importance of each risk factor. According to this formula, the risk contribution of each node can be calculated. The higher the contribution of the node, the greater the impact of the node on the overall safety status. In this way, a safety risk level can be assigned to each node, and the nodes can be sorted according to the contribution to determine which nodes need more attention. Finally, the multi-head attention mechanism is used to perform weight enhancement processing on factors with higher risk contributions to obtain a set of safety risk factors that are focused on. The multi-head attention mechanism plays an important role in this process. By calculating multiple attention heads in parallel, different risk dimensions (such as environment, equipment, and personnel factors) are weighted. Each attention head independently calculates the weight of the risk factor it focuses on, and then merges the results. In this way, the weight can be automatically adjusted according to the contribution of each risk factor to ensure that the key risk factors are strengthened, so that they occupy a larger proportion in the safety assessment. Specifically, the attention mechanism is implemented through the following formula:

[0092] ;

[0093] in, ,and are query, key, and value matrices respectively, is the dimension of the key, and the softmax function is used to calculate the attention weight. Through the multi-head attention mechanism, a set of safety risk factors is finally obtained. These factors are the high-risk factors that need special attention at the construction site, ensuring that the risk management system can respond in real time and give priority to the most urgent safety hazards.

[0094] 104. Based on the security situation assessment results and risk prediction trends, the warning information is processed in a context-aware hierarchical manner to obtain adaptive warning and intervention plans for different risk levels, and security prompts are provided based on the adaptive warning and intervention plans.

[0095] In one embodiment of the present invention, the warning information is subjected to context-aware hierarchical processing according to the security situation assessment results and risk prediction trends to obtain adaptive warning and intervention plans for different risk levels, and safety prompts are provided based on the adaptive warning and intervention plans, including: multi-dimensional risk quantification processing of the security situation assessment results to obtain a risk level matrix including severity, urgency and scope of impact; adaptive coding processing of the warning information is performed according to the risk level matrix and real-time status data to obtain multimodal warning information for different receiving objects; based on the multimodal warning information and the location of the workers, the safety prompt content is contextualized to obtain multi-dimensional safety warnings including visual, auditory and tactile feedback; based on the multi-dimensional safety warnings and risk prediction trends, predictive analysis is performed on mechanical equipment and worker behaviors in high-risk areas to obtain adaptive warning and intervention plans, and safety prompts are provided based on the adaptive warning and intervention plans.

[0096] Specifically, when performing situational awareness classification processing of early warning information based on the security situation assessment results and risk prediction trends, it is first necessary to perform multi-dimensional risk quantification processing on the security situation assessment results. At this point, the security situation assessment results reflect the current risk level of the construction site, including the safety status of workers, environmental changes, and equipment operation status. In order to accurately reflect the risks at different levels, the security situation needs to be subdivided according to the three dimensions of severity, urgency, and impact range, and a risk level matrix is ​​constructed based on this. Severity indicates the degree of threat to personnel safety from potential risks, urgency reflects the urgency of the risk, and the impact range describes the breadth of the impact that the risk may have on other areas, equipment, or personnel at the construction site. Through quantitative analysis of these dimensions, the security situation can be converted into a hierarchical risk level matrix, in which each node (i.e., each risk source in the construction site) is assigned a risk level, and these levels correspond to different risk response strategies. The matrix provides clear standards for subsequent early warning and intervention decisions, helping managers understand which risks are the most serious and which risks require immediate response.

[0097] Specifically, according to the risk level matrix and real-time status data, the warning information will be adaptively coded to generate multimodal warning information for different recipients. In this process, real-time status data, such as worker location, equipment status, and environmental monitoring data, will be used as input to dynamically adjust each warning in combination with the risk level matrix. The key to adaptive coding is the ability to customize warning information based on the characteristics of the recipient (such as managers, on-site workers, or equipment). For example, for managers far away from the source of risk, the system can display the overall risk level through short text or charts, while for workers in high-risk areas, more intuitive and specific voice or visual warnings are used to remind them to take appropriate protective measures. In this way, the warning information can provide feedback that meets the needs of different users while ensuring accuracy, ensuring that the information can be delivered to all relevant parties quickly and effectively.

[0098] Specifically, based on the multimodal warning information and the location of the workers, the safety prompt content is contextualized to enhance the operability and effectiveness of the warning information. The core of contextual processing is to automatically adjust the warning content according to the specific location of the worker, the current task and the changes in the working environment. For example, when a worker is in an extremely dangerous area, the system will not only warn the dangerous factors in the surrounding environment through voice, but also combine visual display devices such as display screens or helmet displays to present specific safe paths or emergency evacuation routes. At the same time, tactile feedback devices (such as vibrating bracelets) are used to remind workers to stop the work at hand immediately or perform necessary protective operations. Contextual processing ensures the real-time and targeted nature of safety prompts, allowing workers to understand and respond to warnings in the shortest time, thereby avoiding potential safety accidents.

[0099] Specifically, based on multi-dimensional safety warnings and risk prediction trends, predictive analysis is performed on mechanical equipment and worker behavior in high-risk areas to provide adaptive early warning and intervention plans. In this process, by analyzing historical data and real-time monitoring data, combined with environmental changes and the state of equipment load, the risks that may occur in the future are predicted. Through machine learning and data-driven analysis, the system can provide early warnings for overload or failure of mechanical equipment, and predict the potential physiological burden of workers under certain operations (such as excessive physical exertion). These analysis results will be used to generate adaptive early warning plans and provide personalized safety intervention measures based on the results of risk assessment. For example, if it is predicted that a certain equipment may fail in a short period of time, the system will notify the operator in advance to adjust the equipment parameters, or notify the maintenance personnel to check. At the same time, based on the real-time physiological state of the worker (such as an excessively high heart rate), the system can suggest that the worker suspend high-intensity work, rest or change shifts. Finally, based on these adaptive early warning and intervention plans, the system will generate safety prompts in real time to guide workers to take appropriate protective measures to ensure the safety of the construction site.

[0100] In this embodiment, a real-time updated environmental digital twin model is constructed by collecting and fusion processing of multi-dimensional environmental data of the construction site. According to the model and the individual physiological data of the workers, the safety status of each worker is adaptively evaluated to obtain an individualized dynamic safety threshold. Combining the environmental digital twin model, individualized dynamic safety threshold and real-time status data, the graph neural network reasoning technology is used to comprehensively evaluate the safety situation and predict potential risk trends. According to the safety situation assessment results and risk prediction trends, a situation-aware hierarchical processing method is used to generate adaptive early warning and intervention plans for different risk levels, providing accurate safety prompts. This method monitors the safety status of workers in real time and predicts early warning according to the dynamic changes of the construction site, improves the intelligence level and accuracy of construction safety management, and reduces the risk of accidents.

[0101] The above describes the construction safety prompting method in the embodiment of the present invention. The following describes the construction safety prompting system in the embodiment of the present invention. Figure 2 An embodiment of the construction safety reminder system in the embodiment of the present invention includes:

[0102] The environmental perception module 201 is used to collect and integrate multi-dimensional environmental data of the construction site to obtain a real-time updated environmental digital twin model;

[0103] An individual assessment module 202 is used to perform adaptive assessment processing on the safety status of each worker based on the environmental digital twin model and the individual physiological data of the worker to obtain an individualized dynamic safety threshold;

[0104] The situation reasoning module 203 is used to perform graph neural network reasoning processing on the safety situation based on the environmental digital twin model, the individualized dynamic safety threshold and the real-time status data of the construction site to obtain the current safety situation assessment result and risk prediction trend;

[0105] The early warning and intervention module 204 is used to perform context-aware hierarchical processing on the early warning information according to the security situation assessment results and risk prediction trends, obtain adaptive early warning and intervention plans for different risk levels, and provide safety prompts based on the adaptive early warning and intervention plans.

[0106] In an embodiment of the present invention, the construction safety reminder system runs the above-mentioned construction safety reminder method, and the construction safety reminder system collects and integrates the multi-dimensional environmental data of the construction site to build a real-time updated environmental digital twin model. According to the model and the individual physiological data of the workers, the safety status of each worker is adaptively evaluated to obtain an individualized dynamic safety threshold. Combining the environmental digital twin model, the individualized dynamic safety threshold and the real-time status data, the graph neural network reasoning technology is used to comprehensively evaluate the safety situation and predict the potential risk trend. According to the safety situation assessment results and risk prediction trends, a situation-aware hierarchical processing method is used to generate adaptive early warning and intervention plans for different risk levels, and provide accurate safety reminders. The method monitors and predicts the safety status of workers in real time according to the dynamic changes of the construction site, improves the intelligence level and accuracy of construction safety management, and reduces the risk of accidents.

[0107] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0108] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0109] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A construction safety reminder method, characterized in that: The construction safety reminder method includes: Collect and integrate multi-dimensional environmental data of the construction site to obtain a real-time updated digital twin model of the environment; Perform multi-dimensional feature extraction processing on the individual physiological data of the workers to obtain the individual feature vector of the workers; perform correlation analysis processing on the individual feature vector and the environmental parameters in the environmental digital twin model to obtain the sensitivity matrix of the workers to different environmental factors; based on the sensitivity matrix and the environmental parameters, perform quantitative calculation processing on the potential risks of the workers to obtain a preliminary safety status assessment value; perform time series analysis processing on the preliminary safety status assessment value to obtain the short-term fluctuation and long-term trend characteristics of the workers' safety status; according to the short-term fluctuation and long-term trend characteristics, combined with the work intensity and duration information, dynamically adjust the safety threshold to obtain an individualized dynamic safety threshold; According to the digital twin model of the environment, the positions of the workers and the distribution of the equipment, a heterogeneous graph is constructed for the construction site to obtain a dynamic graph structure representing the construction site; the individualized dynamic safety threshold is used as the attribute of the worker node in the dynamic graph structure, and the node and edge attributes in the dynamic graph structure are encoded to obtain the corresponding graph embedding representation; based on the graph embedding representation and the real-time updated real-time status data, the state propagation between the nodes is processed by a message passing network to obtain an updated node state; the node state and the individualized dynamic safety threshold in the high-risk area are compared and analyzed, and the safety risk factors of key concern are obtained through the enhanced processing of the attention mechanism; according to the updated node state and the safety risk factors of key concern, a safety score is calculated for the construction site to obtain the current safety situation assessment result; based on the current safety situation assessment result, the updated node state, the safety risk factors of key concern and the individualized dynamic safety threshold, the overall safety situation is predicted in a spatiotemporal sequence to obtain the short-term and medium-term risk evolution trends; According to the security situation assessment results and risk prediction trends, the warning information is processed in a context-aware hierarchical manner to obtain adaptive warning and intervention plans for different risk levels, and security prompts are given based on the adaptive warning and intervention plans.

2. The construction safety prompting method according to claim 1, characterized in that: The multi-dimensional environmental data of the construction site is collected and integrated to obtain a real-time updated environmental digital twin model, including: Edge computing is used to process laser scanning data, infrared thermal imaging data, and multi-parameter gas sensor data collected by multiple environmental monitoring stations to obtain a preliminary environmental situation map. Perform multi-source fusion processing on the foundation settlement data of the pier construction area in the construction site, the hoisting equipment load data at the main beam hoisting position, the inclination data and the wind speed data of the aerial work platform to obtain the safety status index of the construction site; According to the environmental situation map and safety status indicators, the construction site is three-dimensionally reconstructed and parameter-mapped to obtain a multi-level environmental digital twin model; The environmental digital twin model is subjected to time series evolution analysis and processing to obtain a dynamically updated environmental change prediction model.

3. The construction safety prompting method according to claim 1, characterized in that: The quantitative calculation of the potential risk of the worker based on the sensitivity matrix and the environmental parameters to obtain a preliminary safety status assessment value includes: Performing singular value decomposition processing on the sensitivity matrix to obtain the main sensitive eigenvectors of the workers to various environmental factors; Normalizing each environmental factor according to the environmental parameters to obtain a standardized environmental state vector; Performing convolution operation on the main sensitive feature vector and the standardized environmental state vector to obtain a risk exposure index of the worker in the current environment; Based on the risk exposure index, a nonlinear mapping process is performed on the worker's physiological limit threshold to obtain the worker's current safety tolerance; According to the current safety tolerance and the individual physiological data of the workers, a preliminary safety status assessment value is obtained through fuzzy logic reasoning processing.

4. The construction safety prompting method according to claim 1, characterized in that: The node status of the high-risk area is compared and analyzed with the individual dynamic security threshold, and the security risk factors that are focused on include: Perform multi-dimensional feature decomposition on the node status in high-risk areas to obtain a status feature matrix that includes environmental, equipment, and personnel factors; Performing element-level difference calculation processing on the state characteristic matrix and the corresponding individualized dynamic safety threshold to obtain a risk deviation matrix; Performing singular value decomposition on the risk deviation matrix to obtain the eigenvectors of the main risk factors; Based on the feature vector, the original node state is reconstructed to obtain the risk contribution ranking; By utilizing the multi-head attention mechanism, the factors with higher risk contribution are weighted and processed to obtain a set of security risk factors that are the focus of attention.

5. The construction safety prompting method according to claim 1, characterized in that: The step of performing situation-aware hierarchical processing on the warning information according to the security situation assessment results and risk prediction trends to obtain adaptive warning and intervention plans for different risk levels, and providing safety prompts based on the adaptive warning and intervention plans includes: Perform multi-dimensional risk quantification on the security situation assessment results to obtain a risk level matrix that includes severity, urgency, and impact scope; According to the risk level matrix and the real-time status data, the warning information is adaptively coded to obtain multi-modal warning information for different receiving objects; Based on the multimodal warning information and the location of the worker, the safety reminder content is contextualized to obtain a multi-dimensional safety warning including visual, auditory and tactile feedback; According to the multi-dimensional safety warnings and risk prediction trends, predictive analysis is performed on machinery and equipment and worker behaviors in high-risk areas to obtain adaptive warning and intervention plans, and safety prompts are given based on the adaptive warning and intervention plans.

6. A construction safety reminder system, characterized in that: The construction safety reminder system includes: The environmental perception module is used to collect and integrate multi-dimensional environmental data of the construction site to obtain a real-time updated environmental digital twin model; An individual assessment module is used to perform multi-dimensional feature extraction processing on the individual physiological data of the workers to obtain the individual feature vectors of the workers; perform correlation analysis processing on the individual feature vectors and the environmental parameters in the environmental digital twin model to obtain the sensitivity matrix of the workers to different environmental factors; based on the sensitivity matrix and the environmental parameters, perform quantitative calculation processing on the potential risks of the workers to obtain a preliminary safety status assessment value; perform time series analysis processing on the preliminary safety status assessment value to obtain the short-term fluctuation and long-term trend characteristics of the workers' safety status; according to the short-term fluctuation and long-term trend characteristics, combined with the work intensity and duration information, dynamically adjust the safety threshold to obtain an individualized dynamic safety threshold; A situation reasoning module is used to construct a heterogeneous graph for the construction site according to the digital twin model of the environment, the location of the workers and the distribution of the equipment, so as to obtain a dynamic graph structure representing the construction site; the individualized dynamic safety threshold is used as the attribute of the worker node in the dynamic graph structure, and the node and edge attributes in the dynamic graph structure are encoded to obtain the corresponding graph embedding representation; based on the graph embedding representation and the real-time updated real-time status data, the state propagation between the nodes is processed by a message passing network to obtain an updated node state; the node state and the individualized dynamic safety threshold in the high-risk area are compared and analyzed, and the safety risk factors of key concern are obtained through the enhanced processing of the attention mechanism; according to the updated node state and the safety risk factors of key concern, the safety score of the construction site is calculated to obtain the current safety situation assessment result; based on the current safety situation assessment result, the updated node state, the safety risk factors of key concern and the individualized dynamic safety threshold, the overall safety situation is predicted in a spatiotemporal sequence to obtain the short-term and medium-term risk evolution trend; The early warning and intervention module is used to perform context-aware hierarchical processing of the early warning information according to the security situation assessment results and risk prediction trends, obtain adaptive early warning and intervention plans for different risk levels, and provide safety prompts based on the adaptive early warning and intervention plans.

Images