Method and system for automatically checking potential safety hazards on highway construction site

Abstract

The invention discloses a method and a system for automatically checking potential safety hazards on a highway construction site, and belongs to the field of data processing. The method for automatically checking the potential safety hazards on the highway construction site comprises the following steps: continuously collecting original data from various intelligent sensing units, the system has the advantages that global real-time sensing is achieved through data collection, lagged manual inspection is replaced, and second-level capture of parameters such as surrounding rock deformation and harmful gas is achieved; information islands are broken through multi-source data fusion, and automatic association and standardization of heterogeneous information such as geology and monitoring are achieved; through multi-engine intelligent diagnosis and analysis result dynamic fusion early warning, active and advanced recognition and millisecond-level decision making of complex risks (such as water inrush precursor) are achieved, and a traditional postmortem reaction type management mode is converted into an active management and control mode of beforehand early warning and automatic disposal.

Description

Technical Field

[0001] This invention belongs to the field of data processing, and in particular relates to an automatic method and system for identifying safety hazards at highway construction sites. Background Technology

[0002] Current highway tunnel construction presents complex geological environments, including issues such as unstable surrounding rock, sudden water and mud inrushes, and hazardous gases. Unstable surrounding rock: During tunnel excavation, uncertain geological conditions, such as faults, fracture zones, and weak surrounding rock, can lead to sudden rock collapses or large deformations. Sudden water and mud inrushes: When tunnels pass through water-rich strata, karst areas, or fault zones, they may suddenly encounter high-pressure water inrushes or mud and sand inrushes, flooding equipment and personnel. Hazardous gases: Tunnels (especially in coal-bearing strata or oil and gas areas) may accumulate flammable or toxic gases such as methane (gas) and hydrogen sulfide, potentially causing explosions or poisoning. Faced with these complex geological environments, traditional safety hazard investigation methods are significantly untimely.

[0003] This lag manifests itself in three main aspects: First, it relies on manual inspections, with inspections occurring daily or even weekly, making it impossible to capture and grasp instantaneous changes in the surrounding rock ahead or sudden increases in gas concentration; warnings and emergencies occur almost simultaneously. Second, information silos are severe; geological forecasts, monitoring measurements, and gas detection data are scattered across different work teams, relying on manual recording and transmission, resulting in lengthy comprehensive analysis and judgment, often missing the optimal window for intervention. Third, decision-making and response are passive; after a potential hazard is discovered, it requires reporting at each level, holding meetings for analysis, and then issuing disposal instructions. Under dynamically changing tunnel geological conditions, this procedural delay often causes small problems to escalate into major accidents. Essentially, this is a reactive management model, unable to proactively control highly concealed and sudden geological risks, and needs improvement. Summary of the Invention

[0004] Therefore, it is necessary to provide an automatic method and system for identifying safety hazards at highway construction sites to address the aforementioned issues.

[0005] The present invention is implemented as follows: an automatic method for identifying safety hazards at highway construction sites includes the following steps:

[0006] It continuously collects raw data from various intelligent sensing units (such as displacement gauges, piezometers, gas sensors, and geological prediction systems), and aggregates diverse and heterogeneous raw data (such as surrounding rock stress, convergence displacement, water inflow, harmful gas concentration, and ground radar waveforms) into a unified data receiving platform to complete the automatic mapping from the physical world to the digital world.

[0007] Based on preset rules and algorithms, the system automatically performs data cleaning (outlier removal), format standardization, timestamp alignment, and spatial location binding on the raw data. It associates displacement data, stress data, and geological sketch information of the same monitoring area (such as a cross section) to generate a standardized fusion dataset with unified spatiotemporal coordinates.

[0008] The fused dataset is input into the risk diagnosis engine to identify potential hazards. First, a real-time logic based on explicit rules (i.e., the first analysis engine) is executed to judge the data. The real-time data is automatically compared with preset safety thresholds (e.g., if the crown settlement rate and the horizontal convergence rate both exceed the limit simultaneously, an early warning of accelerated surrounding rock deformation is immediately triggered), obtaining the first analysis result. Second, a correlation analysis based on a multi-parameter model (i.e., the second analysis engine) is performed. By calling built-in specialized engineering models (such as the water inrush risk index model and the surrounding rock stability grading model), the coupling relationship and changing trend of multiple key parameters are calculated (e.g., by comprehensively considering the abnormal waveforms in the advanced geological forecast, the sudden rise in borehole water pressure, and the trend of displacement softening, the risk level of water inrush and mudslide is automatically calculated and judged), obtaining the second analysis result. Finally, deep pattern recognition based on machine learning (i.e., the third analysis engine) is used to perform similarity matching between the real-time data stream and the historical accident case database (e.g., to identify the implicit correlation between the current microseismic event sequence, strain distribution pattern, and historical pre-collapse data patterns), obtaining the third analysis result.

[0009] The three analysis results are automatically merged to generate a structured early warning instruction that includes the type, level, location, and confidence level of the hazard.

[0010] In one embodiment, the present invention provides an automatic method for identifying safety hazards at highway construction sites. The step of automatically fusing three analysis results to generate a structured early warning instruction containing hazard type, level, location, and confidence level specifically includes:

[0011] Determine the specific environmental state of the current tunnel construction (e.g., conventional tunneling, crossing a fault zone, or being in a water-rich karst area). Based on the preset strategies for different environmental states, assign an initial decision influence benchmark weight to the three analysis engines (rule judgment, multi-parameter model, and machine learning). The decision influence benchmark weight defines which type of analysis logic is more heavily relied upon in the current environment.

[0012] After obtaining the initial risk assessment of each analysis engine, the reliability of each analysis result is quantitatively evaluated. The evaluation criteria include the quality and completeness of the data used to generate the analysis result, the certainty of the analysis process itself (e.g., model calculation error, the confidence of machine learning prediction), and the recent warning accuracy history of the analysis engine. A dynamic confidence score is calculated for each analysis result.

[0013] Based on the confidence scores of each analysis result, the actual weight of each analysis result in the final decision is dynamically adjusted. Then, a comprehensive risk value is obtained through weighted fusion, which generates a structured early warning instruction containing the type, level, location, and confidence level of the hazard. During this process, a set of hierarchical adjudication logic is executed: if any analysis engine triggers the highest level of safety red line (such as immediate evacuation), the highest level alarm is triggered directly and the weighted fusion is skipped; if the comprehensive risk value exceeds the preset first risk threshold, and the conclusions of most analysis engines (more than half) indicate a high risk level or above, a clear early warning is issued; if the comprehensive risk value exceeds the preset first risk threshold, but the number of analysis engines indicating a high risk level or above does not exceed half, an enhanced monitoring instruction is generated and pushed instead of a direct alarm to avoid false alarms.

[0014] In one embodiment, the present invention provides an automatic method for identifying safety hazards at highway construction sites, further comprising:

[0015] Based on the generated structured early warning instructions, the system automatically executes multi-channel response and closed-loop management: pushes early warning information to the terminal devices of relevant responsible personnel and simultaneously activates the audible and visual alarm devices inside the tunnel; automatically creates corresponding emergency response work orders, assigns them to preset response personnel, and triggers configured linkage control instructions (such as starting emergency ventilation and suspending tunneling operations); at the same time, it automatically tracks feedback data of the response process (such as on-site response records and review monitoring data), and automatically archives the entire process data of this case into the learning library for subsequent continuous optimization of the risk diagnosis engine.

[0016] In one embodiment, the present invention provides an automatic method for identifying safety hazards at highway construction sites, further comprising:

[0017] Based on a fusion dataset, the system dynamically monitors chronic hazard parameters (dust concentration, harmful gas concentration, noise, temperature, and humidity, etc.) and sets different hazard level thresholds according to occupational health exposure limit standards. It tracks the identity information, real-time location, and cumulative exposure time in specific hazardous environments of each worker in real time, automatically calculates individualized exposure doses, and automatically pushes graded warnings (such as "rest recommended" or "immediate evacuation and rotation") to the worker's personal smart terminal (such as safety helmet integrated device or wristband) and the management terminal of the team leader. It also records the warning and evacuation execution status to form a health exposure file.

[0018] In one embodiment, the present invention provides an automatic method for identifying safety hazards at highway construction sites, further comprising:

[0019] Based on a fusion dataset (which requires monitoring of vibration acceleration in the equipment cab / operator position during data acquisition), vibration dose profiles are established for operators whose work environment involves whole-body vibration or hand-transmitted vibration. According to the vibration dose limits set by international standards (such as ISO 2631 and ISO 5349), the cumulative vibration exposure dose of each operator is calculated in real time. When the dose approaches the limit, a graded warning will be automatically pushed to the operator's personal smart terminal and the management terminal of the team leader, and the warning and evacuation execution status will be recorded to form a vibration exposure profile.

[0020] In one embodiment, the present invention provides an automatic safety hazard detection system for highway construction sites, comprising:

[0021] The data acquisition module is used to continuously collect raw data from various intelligent sensing units (such as displacement gauges, piezometers, gas sensors, and geological prediction systems), and to aggregate diverse and heterogeneous raw data (such as surrounding rock stress, convergence displacement, water inflow, harmful gas concentration, and ground radar waveforms) to a unified data receiving platform, thus completing the automatic mapping from the physical world to the digital world.

[0022] The data processing module is used to automatically perform data cleaning (removing outliers), format standardization, timestamp alignment and spatial location binding on the raw data according to preset rules and algorithms. It associates displacement data, stress data and geological sketch information of the same monitoring area (such as cross section) to generate a standardized fusion dataset with unified spatiotemporal coordinates.

[0023] The hazard identification module is used to input the fused dataset into the risk diagnosis engine to identify hazards. First, it executes real-time logic based on explicit rules (i.e., the first analysis engine) to judge and automatically compare the real-time data with preset safety thresholds (e.g., if the crown settlement rate and the horizontal convergence rate both continuously exceed the limit, an early warning of accelerated surrounding rock deformation is immediately triggered), obtaining the first analysis result. Second, it performs correlation analysis based on a multi-parameter model (i.e., the second analysis engine), by calling built-in specialized engineering models (such as the water inrush risk index model and the surrounding rock stability grading model) to calculate the coupling relationship and change trend of multiple key parameters (e.g., by comprehensively considering the abnormal waveforms in advanced geological predictions, the sudden increase in borehole water pressure, and the trend of displacement softening, it automatically calculates and judges the risk level of water inrush and mudslide), obtaining the second analysis result. Finally, it uses deep pattern recognition based on machine learning (i.e., the third analysis engine) to perform similarity matching between the real-time data stream and the historical accident case database (e.g., to identify the implicit correlation between the current microseismic event sequence, strain distribution pattern, and historical pre-collapse data patterns), obtaining the third analysis result.

[0024] The analysis result fusion and judgment module is used to automatically merge the three analysis results and finally generate a structured early warning instruction that includes the type, level, location and confidence level of the hidden danger.

[0025] In one embodiment, the present invention provides an automatic safety hazard detection system for highway construction sites, wherein the analysis result fusion and judgment module includes:

[0026] The benchmark weight adjustment unit is used to determine the specific environmental state of the current tunnel construction (e.g., conventional tunneling, crossing fault zones, or being in a water-rich karst area). Based on the preset strategies for different environmental states, it assigns an initial decision influence benchmark weight to the three analysis engines (rule judgment, multi-parameter model, and machine learning). The decision influence benchmark weight defines which type of analysis logic is more heavily relied upon in the current environment.

[0027] The confidence score calculation unit is used to quantitatively evaluate the reliability of each analysis result after obtaining the preliminary risk assessment of each analysis engine. The evaluation criteria include the quality and completeness of the data used to generate the analysis result, the certainty of the analysis process itself (e.g., model calculation error, the confidence of machine learning prediction), and the recent warning accuracy history of the analysis engine. A dynamic confidence score is calculated for each analysis result.

[0028] The actual weight adjustment unit dynamically adjusts the actual weight of each analysis result in the final decision based on the confidence score of each analysis result. Then, a comprehensive risk value is obtained through weighted fusion, which generates a structured early warning instruction containing the type, level, location, and confidence level of the hazard. In this process, a set of hierarchical adjudication logic is executed: if any analysis engine triggers the highest level of safety red line (such as immediate evacuation), the highest level alarm is directly triggered and the weighted fusion is skipped; if the comprehensive risk value exceeds the preset first risk threshold, and the conclusions of most analysis engines (more than half) indicate a high risk level or above, a clear early warning is issued; if the comprehensive risk value exceeds the preset first risk threshold, but the number of analysis engines indicating a high risk level or above does not exceed half, an enhanced monitoring instruction is generated and pushed instead of a direct alarm to avoid false alarms.

[0029] In one embodiment, the present invention provides an automatic safety hazard detection system for highway construction sites, further comprising:

[0030] The multi-channel response and management module is used to automatically execute multi-channel response and closed-loop management based on the generated structured early warning instructions: push early warning information to the terminal devices of relevant responsible personnel and simultaneously activate the audible and visual alarm devices in the tunnel; automatically create corresponding emergency response work orders, assign them to preset response personnel, and trigger the configured linkage control instructions (such as starting emergency ventilation and suspending tunneling operations); at the same time, automatically track the feedback data of the response process (such as on-site response records and review monitoring data), and automatically archive the entire process data of this case to the learning library for subsequent continuous optimization of the risk diagnosis engine.

[0031] In one embodiment, the present invention provides an automatic safety hazard detection system for highway construction sites, further comprising:

[0032] The health exposure early warning module is used to dynamically monitor chronic hazard parameters (dust concentration, harmful gas concentration, noise, temperature and humidity, etc.) based on a fusion dataset. It sets different hazard level thresholds according to occupational health exposure limit standards, tracks the identity information, real-time location and cumulative exposure time in specific hazardous environments of each worker in real time, automatically calculates individualized exposure doses, and automatically pushes graded warnings (such as "rest recommended" or "immediate evacuation and rotation") to the worker's personal smart terminal (such as safety helmet integrated device or wristband) and the management terminal of the team leader when the cumulative exposure dose or time of a worker is determined to be close to the preset safety limit. It also records the warning and evacuation execution status to form a health exposure file.

[0033] In one embodiment, the present invention provides an automatic safety hazard detection system for highway construction sites, further comprising:

[0034] The vibration exposure early warning module is used to establish vibration dose profiles for operators whose working environment is subject to whole-body vibration or hand-transmitted vibration, based on a fusion dataset (which requires monitoring of vibration acceleration in the equipment cab / operator position during data acquisition). According to vibration dose limits set by international standards (such as ISO 2631 and ISO 5349), it calculates the cumulative vibration exposure dose for each operator in real time. When the dose approaches the limit, it automatically pushes tiered warnings to the operator's personal smart terminal and the management terminal of their team leader, and records the execution of warnings and evacuations, thus forming a vibration exposure profile.

[0035] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention achieves real-time perception across the entire domain through data acquisition, replacing the lagging manual inspection and realizing the second-level capture of parameters such as surrounding rock deformation and harmful gases; it breaks down information silos through multi-source data fusion, realizing the automatic association and standardization of heterogeneous information such as geology and monitoring; and it achieves proactive and advanced identification and millisecond-level decision-making for complex risks (such as precursors to water inrush) through multi-engine intelligent diagnosis and dynamic fusion of analysis results, transforming the traditional post-event reactive management model into a proactive control model of pre-event warning and automatic handling. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the first part of a method for automatically identifying safety hazards at highway construction sites, provided as an embodiment of the present invention.

[0037] Figure 2 This is a schematic diagram of the process for fusing three analysis results provided in an embodiment of the present invention.

[0038] Figure 3 This is a schematic diagram of the second part of an automatic method for identifying safety hazards at highway construction sites, provided as an embodiment of the present invention.

[0039] Figure 4 This is a flowchart of the third part of an automatic method for identifying safety hazards at highway construction sites, provided as an embodiment of the present invention.

[0040] Figure 5 This is a flowchart of the fourth part of an automatic method for identifying safety hazards at highway construction sites, provided as an embodiment of the present invention.

[0041] Figure 6 This is a schematic diagram of the first part of an automatic safety hazard detection system for highway construction sites provided in an embodiment of the present invention.

[0042] Figure 7 This is a schematic diagram of the analysis result fusion and judgment module provided in an embodiment of the present invention.

[0043] Figure 8 This is a schematic diagram of the second part of an automatic safety hazard detection system for highway construction sites provided in an embodiment of the present invention.

[0044] Figure 9 This is a schematic diagram of the third part of an automatic safety hazard detection system for highway construction sites provided in an embodiment of the present invention.

[0045] Figure 10 This is a schematic diagram of the fourth part of an automatic safety hazard detection system for highway construction sites provided in an embodiment of the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0047] It is understood that the terms "first," "second," etc., used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are used only to distinguish one element from another. For example, without departing from the scope of this application, a first script may be referred to as a second script, and similarly, a second script may be referred to as a first script.

[0048] In one embodiment, such as Figure 1 As shown, an automatic method for identifying safety hazards at highway construction sites includes the following steps:

[0049] Step S1: Continuously collect raw data from various intelligent sensing units (such as displacement gauges, piezometers, gas sensors, and geological prediction systems), and aggregate diverse and heterogeneous raw data (such as surrounding rock stress, convergence displacement, water inflow, harmful gas concentration, and ground radar waveforms) to a unified data receiving platform to complete the automatic mapping from the physical world to the digital world.

[0050] Step S2: Based on preset rules and algorithms, the raw data is automatically cleaned (outliers are removed), format is standardized, timestamps are aligned and spatial locations are bound. Displacement data, stress data and geological sketch information of the same monitoring area (such as cross section) are associated to generate a standardized fusion dataset with unified spatiotemporal coordinates.

[0051] Step S3 involves inputting the fused dataset into the risk diagnosis engine to identify potential hazards. First, a real-time logic based on explicit rules (i.e., the first analysis engine) is executed to judge the data. The real-time data is automatically compared with preset safety thresholds (e.g., if the crown settlement rate and the horizontal convergence rate both exceed the limit simultaneously, an early warning of accelerated surrounding rock deformation is immediately triggered), obtaining the first analysis result. Second, a correlation analysis based on a multi-parameter model (i.e., the second analysis engine) is performed. By calling built-in specialized engineering models (such as the water inrush risk index model and the surrounding rock stability grading model), the coupling relationship and changing trend of multiple key parameters are calculated (e.g., by comprehensively considering the abnormal waveforms in the advanced geological forecast, the sudden rise in borehole water pressure, and the trend of displacement softening, the risk level of water inrush and mudslide is automatically calculated and judged), obtaining the second analysis result. Finally, deep pattern recognition based on machine learning (i.e., the third analysis engine) is used to perform similarity matching between the real-time data stream and the historical accident case database (e.g., identifying the implicit correlation between the current microseismic event sequence, strain distribution pattern, and historical pre-collapse data patterns), obtaining the third analysis result.

[0052] Step S4 automatically merges the three analysis results to generate a structured early warning instruction that includes the type, level, location, and confidence level of the hazard.

[0053] Addressing the core pain points of traditional construction site safety hazard investigation methods, such as the lag in manual inspections and information silos, step S1 achieves real-time and automatic collection of risk parameters through a global sensor network, replacing discrete manual inspections; step S2 breaks down professional data barriers and builds a unified analysis foundation through automatic fusion and standardization of multi-source data; step S3 introduces a three-engine intelligent diagnosis to simulate comprehensive expert judgment and achieve proactive identification of complex geological risks (such as precursors to water inrush); step S4 transforms multi-engine conclusions into reliable and timely early warning instructions through dynamic fusion and hierarchical adjudication, thereby shifting from a reactive to a proactive management model of pre-emptive warning.

[0054] In one embodiment, such as Figure 2 As shown, an automatic safety hazard detection method for highway construction sites includes step S4, which automatically merges three analysis results to generate a structured early warning instruction containing hazard type, level, location, and confidence level. Specifically, this step includes:

[0055] Step S41: Determine the specific environmental state of the current tunnel construction (e.g., conventional tunneling, crossing a fault zone, or being in a water-rich karst area). Based on the preset strategies for different environmental states, assign an initial decision influence benchmark weight to the three analysis engines (rule judgment, multi-parameter model, and machine learning). The decision influence benchmark weight defines which type of analysis logic is more heavily relied upon in the current environment.

[0056] Step S42: After obtaining the preliminary risk assessment of each analysis engine, the reliability of each analysis result is quantitatively evaluated. The evaluation criteria include the quality and completeness of the data used to generate the analysis result, the certainty of the analysis process itself (e.g., model calculation error, the confidence of machine learning prediction), and the recent warning accuracy history of the analysis engine. A dynamic confidence score is calculated for each analysis result.

[0057] Step S43: Based on the confidence scores of each analysis result, dynamically adjust the actual weight of each analysis result in the final decision. Then, obtain a comprehensive risk value through weighted fusion, and generate a structured early warning instruction containing the type, level, location, and confidence level of the hazard. During this process, a set of hierarchical decision-making logic is executed: if any analysis engine triggers the highest level of safety red line (such as immediate evacuation), the highest level alarm is directly triggered and the weighted fusion is skipped; if the comprehensive risk value exceeds the preset first risk threshold, and the conclusions of most analysis engines (more than half) indicate a high risk level or above, a clear early warning is issued; if the comprehensive risk value exceeds the preset first risk threshold, but the number of analysis engines indicating a high risk level or above does not exceed half, an enhanced monitoring instruction is generated and pushed instead of a direct alarm to avoid false alarms.

[0058] To improve the accuracy and environmental adaptability of risk warning, steps S41 to S43 are designed. Step S41 presets differentiated analysis strategies based on the geological characteristics of different tunnel sections to ensure that decisions are aligned with actual working conditions. Step S42 performs real-time confidence assessment on the output of each analysis engine to identify and reduce the risk of misjudgment due to data quality or model limitations. Step S43 dynamically adjusts weights and performs graded adjudication based on confidence levels to ensure that safety red lines are prioritized in emergency situations and conservative strategies (such as enhanced monitoring) are adopted in uncertain situations, thereby achieving both sensitive and robust automated decision-making in complex environments.

[0059] In one embodiment, such as Figure 3 As shown, an automatic method for identifying safety hazards at highway construction sites also includes:

[0060] Step S5: Based on the generated structured early warning instructions, automatically execute multi-channel response and closed-loop management: push early warning information to the terminal devices of relevant responsible personnel and simultaneously activate the audible and visual alarm devices in the tunnel; automatically create corresponding emergency response work orders, assign them to preset response personnel, and trigger the configured linkage control instructions (such as starting emergency ventilation and suspending tunneling operations); at the same time, automatically track the feedback data of the response process (such as on-site response records and review monitoring data), and automatically archive the entire process data of this case into the learning library for subsequent continuous optimization of the risk diagnosis engine.

[0061] To ensure that early warnings can be transformed into effective actions, step S5 was designed. Through multi-channel push and device linkage, early warning information is ensured to be received and executed in a timely manner. Work orders are automatically created and dispatched to achieve rapid implementation of handling responsibilities. Through tracking feedback and data archiving, the experience of each handling is accumulated into the learning library, driving the continuous optimization of the risk diagnosis engine, thereby fundamentally solving the problems of passive decision-making response and inability to accumulate knowledge.

[0062] In one embodiment, such as Figure 4 As shown, an automatic method for identifying safety hazards at highway construction sites also includes:

[0063] Step S6: Based on the fused dataset, dynamically monitor chronic hazard parameters (dust concentration, harmful gas concentration, noise, temperature and humidity, etc.), and set different hazard level thresholds according to occupational health exposure limit standards. Track the identity information, real-time location and cumulative exposure time in specific hazardous environments of each worker in real time, automatically calculate individualized exposure doses, and when it is determined that the cumulative exposure dose or time of a worker is close to the preset safety limit, automatically push graded warnings (such as "recommend rest" or "immediate evacuation and rotation") to the worker's personal smart terminal (such as safety helmet integrated device or wristband) and the management terminal of the team leader, and record the warning and evacuation execution status to form a health exposure file.

[0064] To address chronic occupational health hazards such as dust and harmful gases within tunnels, step S6 expands the dimensions of monitoring and early warning. By tracking the cumulative exposure dose of individuals in hazardous environments and linking it with personnel location and identity information, a leap from environmental monitoring to personalized health risk early warning is achieved. The data is simultaneously archived into the learning database of step S5, providing a scientific basis for optimizing ventilation, adjusting work schedules, and allocating protective resources.

[0065] In one embodiment, such as Figure 5 As shown, an automatic method for identifying safety hazards at highway construction sites also includes:

[0066] Step S7: Based on the fused dataset (data acquisition requires monitoring of vibration acceleration in the equipment cab / operating position); establish vibration dose profiles for operators whose working environment experiences whole-body vibration or hand-transmitted vibration. Calculate the cumulative vibration exposure dose for each operator in real time according to vibration dose limits set by international standards (such as ISO 2631 and ISO 5349). When the dose approaches the limit, a graded warning will be automatically pushed to their personal smart terminal and the management terminal of their team leader, and the warning and evacuation execution status will be recorded to form a vibration exposure profile.

[0067] To address the hidden hazard of occupational disease caused by long-term operation of heavy equipment (such as rock drilling rigs and shotcrete robots), step S7 reuses the individual exposure tracking logic of step S6 to establish a dedicated vibration exposure file. By monitoring equipment vibration data and combining it with international standards, the cumulative exposure of operators is quantitatively assessed and early warnings are issued, achieving precise control over physical chronic hazards. This data is also fed back into a learning database, providing support for equipment improvement, process optimization, and occupational health management, thus perfecting the protection system for all hazard factors affecting personnel.

[0068] In one embodiment, such as Figure 6 As shown, an automatic safety hazard detection system for highway construction sites includes:

[0069] Data acquisition module 1 is used to continuously collect raw data from various intelligent sensing units (such as displacement gauges, piezometers, gas sensors, and geological prediction systems), and to aggregate diverse and heterogeneous raw data (such as surrounding rock stress, convergence displacement, water inflow, harmful gas concentration, and ground radar waveforms) to a unified data receiving platform, thus completing the automatic mapping from the physical world to the digital world.

[0070] Data processing module 2 is used to automatically perform data cleaning (removing outliers), format standardization, timestamp alignment and spatial location binding on the raw data according to preset rules and algorithms. It associates displacement data, stress data and geological sketch information of the same monitoring area (such as cross section) to generate a standardized fusion dataset with unified spatiotemporal coordinates.

[0071] The hazard identification module 3 is used to input the fused dataset into the risk diagnosis engine to identify hazards. First, it executes real-time logic based on explicit rules (i.e., the first analysis engine) to judge and automatically compare the real-time data with preset safety thresholds (e.g., if the crown settlement rate and the horizontal convergence rate both continuously exceed the limit, an early warning of accelerated surrounding rock deformation is immediately triggered), obtaining the first analysis result. Second, it performs correlation analysis based on a multi-parameter model (i.e., the second analysis engine), by calling built-in specialized engineering models (such as the water inrush risk index model and the surrounding rock stability grading model) to calculate the coupling relationship and change trend of multiple key parameters (e.g., by comprehensively considering the abnormal waveforms in the advanced geological forecast, the sudden rise in borehole water pressure, and the trend of displacement softening, it automatically calculates and judges the risk level of water inrush and mudslide), obtaining the second analysis result. Finally, it uses deep pattern recognition based on machine learning (i.e., the third analysis engine) to perform similarity matching between the real-time data stream and the historical accident case library (e.g., to identify the implicit correlation between the current microseismic event sequence, strain distribution pattern, and historical pre-collapse data patterns), obtaining the third analysis result.

[0072] The analysis result fusion and judgment module 4 is used to automatically fuse the three analysis results and finally generate a structured early warning instruction that includes the type, level, location and confidence level of the hidden danger.

[0073] The prerequisite for implementing data acquisition module 1 is to deploy various intelligent sensing units in key sections, working faces, and high-risk areas within the tunnel. These include micro-strain and displacement sensors installed at the contact surface between the initial support and the surrounding rock (for real-time monitoring of surrounding rock convergence and initial support stress), piezometers and flow meters (for monitoring changes in seepage pressure and flow), multi-parameter gas detectors (for continuous monitoring of multiple gas concentrations), and ground-penetrating radar and TSP receivers (for routine advanced geological exploration). All sensors transmit data in real-time via the tunnel industrial ring network or a 5G private network.

[0074] In one embodiment, such as Figure 7 As shown, an automatic safety hazard detection system for highway construction sites includes an analysis result fusion and judgment module 4, which comprises:

[0075] The benchmark weight adjustment unit 41 is used to determine the specific environmental state of the current tunnel construction (e.g., conventional tunneling, crossing fault zones, or being in a water-rich karst area). Based on the preset strategies for different environmental states, it assigns an initial decision influence benchmark weight to the three analysis engines (rule judgment, multi-parameter model, and machine learning). The decision influence benchmark weight defines which type of analysis logic is more heavily relied upon in the current environment.

[0076] The confidence calculation unit 42 is used to quantitatively evaluate the reliability of each analysis result after obtaining the preliminary risk judgment of each analysis engine. The evaluation basis includes the quality and completeness of the data used to generate the analysis result, the certainty of the analysis process itself (e.g., model calculation error, the confidence of machine learning prediction), and the recent warning accuracy history of the analysis engine, and calculates a dynamic confidence score for each analysis result.

[0077] The actual weight adjustment unit 43 is used to dynamically adjust the actual weight of each analysis result in the final decision based on the confidence score of each analysis result. Subsequently, a comprehensive risk value is obtained through weighted fusion, which in turn generates a structured early warning instruction containing the type, level, location, and confidence level of the hazard. In this process, a set of hierarchical adjudication logic is executed: if any analysis engine triggers the highest level safety red line (such as immediate evacuation), the highest level alarm is directly triggered and the weighted fusion is skipped; if the comprehensive risk value exceeds the preset first risk threshold, and the conclusions of most analysis engines (more than half) indicate high risk... If the risk level is at or above the preset first risk threshold, a clear warning will be issued. If the overall risk value exceeds the preset first risk threshold, but the number of analysis engines indicating a high risk level or above is less than half, an enhanced monitoring instruction will be generated and pushed instead of a direct alarm to avoid false alarms. (Specifically, if the weighted overall risk index R is greater than the preset first risk threshold and the risk consensus S is greater than or equal to the preset second consensus threshold, a clear warning will be issued. If the weighted overall risk index R is greater than the preset first risk threshold, but the risk consensus S is less than the preset second consensus threshold, an enhanced monitoring instruction will be generated and pushed instead of a direct alarm to avoid false alarms.)

[0078] The fusion of the three analysis results is achieved through an environment-adaptive dynamic decision fusion model. The core mechanism of this model is to automatically assess the confidence level of each analysis engine's output based on real-time operating conditions, dynamically adjust their contribution weights in the final decision, and ultimately output a structured early warning through hierarchical adjudication logic. Its specific process and formulaic description are as follows:

[0079] 1. Environmental perception and baseline weight mapping

[0080] First, identify the current construction environment mode (M), such as M∈{“Conventional Tunneling”, “Crossing Fault Zones”, “Water-Rich Karst Zones”, “Stress Anomaly Zones”}. Each mode is mapped to a pre-defined baseline trust vector W0:

[0081]

[0082] in, These represent the initial weight allocations of the first, second, and third analysis engines under mode M, respectively, and satisfy the following conditions: =1. For example, in the "Crossing a known fault zone" (M=fault zone) mode, the parameter may be set to (0.5, 0.4, 0.1), emphasizing the rule and mechanism model; in the "Abnormal and unprecedented" (M=abnormal) mode, it may be set to (0.2, 0.3, 0.5), increasing the weight of the data-driven model.

[0083] 2. Real-time confidence assessment and dynamic weight adjustment

[0084] The confidence score for each engine's output is calculated in real time. For the i-th analysis engine (i can be 1, 2, or 3), its confidence score Ci is determined by the following factors:

[0085] Data Quality Score (Qi): Based on sensor status, data integrity, and signal-to-noise ratio.

[0086] Output deterministic score (Di): accuracy of rule matching, model fit residuals, and probability entropy of machine learning predictions.

[0087] Recent Accuracy (Ai): The historical accuracy of the engine's alerts over the past N periods.

[0088] The overall confidence level can be quantified as: Ci=f(Qi, Di, Ai), a weighted average function.

[0089] Subsequently, the baseline weights are dynamically adjusted using confidence levels to obtain the actual weights W=(α,β,γ) for this fusion. A typical adjustment method is confidence-weighted normalization:

[0090] The calculation of β and γ follows the same principle. This formula ensures that when the confidence of an engine decreases, its weight in the actual decision-making process will automatically decrease.

[0091] 3. Layered Conflict Resolution and Final Decision

[0092] After obtaining the dynamic weights and output values ​​of each engine, a hierarchical decision-making process is performed:

[0093] First Layer: Safety Red Line Arbitration: If V1 outputs "Emergency Danger," the highest-level alarm is immediately triggered, skipping weighted fusion. V1, V2, and V3 are the outputs (i.e., analysis results) of the first, second, and third analysis engines, respectively. The first analysis engine is designed as a hard switch for the system to strictly adhere to legal and safety bottom lines. It directly compares real-time data with preset, uncompromising absolute safety thresholds (such as the lower explosion limit of gas, emergency evacuation concentration), and its output is a binary and explicit "yes / no" logic (e.g., "whether the lower explosion limit has been reached"). Therefore, only it can output the highest-level alarm status, "Emergency Danger."

[0094] The core functions of the second analysis engine (multi-parameter model) and the third analysis engine (machine learning) are risk assessment and probability prediction. They output a continuous risk probability value or risk index (e.g., between 0 and 1) to measure the level and trend of risk. Their purpose is to provide risk grading (e.g., high, medium, low) and early warnings, rather than making uncompromising binary safety decisions. Therefore, in the system design, only the first analysis engine (rule engine) is given the authority to output "urgent danger" and directly trigger the highest-level emergency response.

[0095] Second layer: Weighted consensus fusion: Calculating the weighted comprehensive risk index R and the risk consensus degree S:

[0096] R = α*N(V1) + β*V2 + γ*V3;

[0097] S=

[0098] Where N() is the normalization function and I() is the indicator function. The threshold is set as the high-risk threshold. The final warning is determined jointly by R and S. If the weighted comprehensive risk index R is greater than the preset first risk threshold and the risk consensus S is greater than or equal to the preset second consensus threshold, a clear warning will be issued; if the weighted comprehensive risk index R is greater than the preset first risk threshold, but the risk consensus S is less than the preset second consensus threshold, an enhanced monitoring instruction will be generated and pushed instead of a direct alarm to avoid false alarms.

[0099] The third layer: Learning and optimization: Closed-loop data is used to optimize the internal models of each engine, and environmental patterns and baseline weight vectors are adjusted through machine learning. The mapping relationship enables system evolution.

[0100] In one embodiment, such as Figure 8 As shown, an automatic safety hazard detection system for highway construction sites also includes:

[0101] The multi-channel response and management module 5 is used to automatically execute multi-channel response and closed-loop management based on the generated structured early warning instructions: push early warning information to the terminal devices of relevant responsible personnel and simultaneously activate the audible and visual alarm devices in the tunnel; automatically create corresponding emergency response work orders, assign them to preset response personnel, and trigger the configured linkage control instructions (such as starting emergency ventilation and suspending tunneling operations); at the same time, automatically track the feedback data of the response process (such as on-site response records and review monitoring data), and automatically archive the entire process data of this case to the learning library for subsequent continuous optimization of the risk diagnosis engine.

[0102] For example, when the system determines through the hazard identification module 3 and the analysis result fusion judgment module 4 that "the arch subsidence rate continues to exceed the limit, and the comprehensive risk level is red," the multi-channel response and management module 5 automatically executes: 1. Push and alarm: The warning message "At XXX, the surrounding rock deformation is accelerating (red), it is recommended to immediately suspend tunneling and conduct support verification" is simultaneously pushed to the terminals of the project manager, chief engineer, and site foreman, and triggers the audible and visual alarms in that section. 2. Work order creation and linkage control: The system automatically creates a disposal work order, assigns it to the safety officer, and automatically executes the linkage commands of "suspend the tunneling trolley operation at this working face" and "start emergency lighting." 3. Tracking and archiving: The safety officer provides feedback through the terminal that "temporary support has been implemented and monitoring retesting has been arranged." After the verification data is uploaded, the system automatically marks the work order as completed and archives the warning, decision basis, disposal measures, and subsequent deformation convergence data completely to the learning library.

[0103] In one embodiment, such as Figure 9 As shown, an automatic safety hazard detection system for highway construction sites also includes:

[0104] The health exposure early warning module 6 is used to dynamically monitor chronic hazard parameters (dust concentration, harmful gas concentration, noise, temperature and humidity, etc.) based on a fusion dataset. It sets different hazard level thresholds according to occupational health exposure limit standards, tracks the identity information, real-time location and cumulative exposure time in specific hazardous environments of each worker in real time, automatically calculates individualized exposure doses, and automatically pushes graded warnings (such as "rest recommended" and "immediate evacuation and rotation") to the worker's personal smart terminal (such as safety helmet integrated device, wristband) and the management terminal of the team leader when the cumulative exposure dose or time of a worker is determined to be close to the preset safety limit. It also records the warning and evacuation execution status to form a health exposure file.

[0105] For example, at the tunneling face, dust sensors detect that the concentration remains within the "permissible concentration" range but does not exceed the standard. The health exposure early warning module 6 tracks the cumulative exposure time of shotcrete workers (e.g., ID: A23) working in this area in real time. When their individual cumulative exposure dose (concentration × time) reaches 85% of the preset safety limit after 4 hours of work, the system automatically triggers a level-two warning: on the one hand, it vibrates the worker's smart bracelet and displays "Dust exposure is approaching the safety limit, a 20-minute rest is recommended"; on the other hand, it sends a message to their team leader's terminal "Employee A23 needs to take a break, it is recommended to adjust to a low-dust work area." This warning and subsequent rest records are simultaneously recorded in the employee's health exposure file for analysis of the rationality of the team's shift scheduling.

[0106] In one embodiment, such as Figure 10 As shown, an automatic safety hazard detection system for highway construction sites also includes:

[0107] Vibration exposure early warning module 7 is used to establish vibration dose files for operators whose working environment is subject to whole-body vibration or hand-transmitted vibration based on a fusion dataset (which requires monitoring of vibration acceleration in the equipment cab / operating position during data acquisition). According to the vibration dose limits set by international standards (such as ISO 2631 and ISO 5349), it calculates the cumulative vibration exposure dose of each operator in real time. When the dose approaches the limit, it will automatically push graded warnings to the operator's personal smart terminal and the management terminal of the team leader, and record the warning and evacuation execution status to form a vibration exposure file.

[0108] For example, the vibration sensor built into the operating position of the rock drilling rig detects a hand-transmitted vibration acceleration value of 8.5 m / s². The vibration exposure early warning module 7 calculates the operator's 8-hour energy equivalent vibration exposure value in real time based on this value, the current work duration (2 hours), and the operator's (ID: A31) historical cumulative dose record, according to the ISO 5349 standard model. When the system predicts that the vibration dose limit will be exceeded at the end of the shift, it immediately sends an early warning to the operator's cab terminal and safety officer terminal: "Vibration exposure is expected to exceed the vibration dose limit; it is recommended that subsequent work be halved or a 2-hour low-vibration task be arranged." The system records this early warning and subsequent task adjustments for assessing equipment vibration reduction performance and developing job rotation plans.

[0109] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0110] Those skilled in the art will understand that all or part of the processes in the systems described in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0111] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0113] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0114] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for automatically identifying safety hazards at highway construction sites, characterized in that, The automatic safety hazard detection method at highway construction sites includes the following steps: It continuously collects raw data from various intelligent sensing units, gathers diverse and heterogeneous raw data into a unified data receiving platform, and completes the automatic mapping from the physical world to the digital world. Based on preset rules and algorithms, the system automatically performs data cleaning, format standardization, timestamp alignment, and spatial location binding on the raw data. It associates displacement data, stress data, and geological sketch information in the same monitoring area to generate a standardized fusion dataset with unified spatiotemporal coordinates. The fused dataset is input into the risk diagnosis engine to identify potential hazards. First, real-time logical judgment based on explicit rules is executed, and the real-time data is automatically compared with preset safety thresholds to obtain the first analysis result. Second, correlation analysis based on a multi-parameter model is performed, and the coupling relationship and changing trend of multiple key parameters are calculated by calling the built-in special engineering model to obtain the second analysis result. Finally, deep pattern recognition based on machine learning is used to perform similarity matching between the real-time data stream and the historical accident case database to obtain the third analysis result. The three analysis results are automatically merged to generate a structured early warning instruction that includes the type, level, location, and confidence level of the hazard.

2. The automatic method for identifying safety hazards at highway construction sites according to claim 1, characterized in that, The step of automatically integrating the three analysis results to generate a structured early warning instruction containing the type, level, location, and confidence level of the hazard specifically includes: Determine the specific environmental state of the current tunnel construction, and assign an initial decision influence benchmark weight to the three analysis engines according to the preset strategies for different environmental states. The decision influence benchmark weight defines which type of analysis logic is more relied upon in the current environment. After obtaining the initial risk assessment of each analysis engine, the reliability of each analysis result is quantitatively evaluated. The evaluation criteria include the quality and completeness of the data used to generate the analysis result, the certainty of the analysis process itself, and the recent warning accuracy history of the analysis engine. A dynamic confidence score is calculated for each analysis result. Based on the confidence scores of each analysis result, the actual weight of each analysis result in the final decision is dynamically adjusted. Then, a comprehensive risk value is obtained through weighted fusion, which generates a structured early warning instruction containing the type, level, location, and confidence level of the hazard. During this process, a set of hierarchical adjudication logic is executed: if any analysis engine triggers the highest level of safety red line, the highest level alarm is triggered directly and the weighted fusion is skipped; if the comprehensive risk value exceeds the preset first risk threshold, and the conclusions of most analysis engines indicate a high risk level or above, a clear early warning is issued; if the comprehensive risk value exceeds the preset first risk threshold, but the number of analysis engines indicating a high risk level or above does not exceed half, an enhanced monitoring instruction is generated and pushed instead of a direct alarm to avoid false alarms.

3. The automatic method for identifying safety hazards at highway construction sites according to claim 1, characterized in that, Also includes: Based on the generated structured early warning instructions, the system automatically executes multi-channel response and closed-loop management: pushes early warning information to the terminal devices of relevant responsible personnel and simultaneously activates the audible and visual alarm devices in the tunnel; automatically creates corresponding emergency response work orders, assigns them to preset response personnel, and triggers the configured linkage control instructions; at the same time, it automatically tracks the feedback data of the response process and automatically archives the entire process data of this case into the learning library for subsequent continuous optimization of the risk diagnosis engine.

4. The automatic method for identifying safety hazards at highway construction sites according to any one of claims 1 to 3, characterized in that, Also includes: Based on the fusion dataset, chronic hazard parameters are dynamically monitored, and different hazard level thresholds are set according to the occupational health exposure limit standards. The system tracks the identity information, real-time location, and cumulative exposure time in specific hazard environments of each worker in real time, automatically calculates individualized exposure doses, and automatically pushes graded warnings to the worker's personal smart terminal and the management terminal of the team leader when it is determined that the cumulative exposure dose or time of a worker is close to the preset safety limit. The system also records the warning and evacuation execution status, forming a health exposure file.

5. The automatic method for identifying safety hazards at highway construction sites according to claim 4, characterized in that, Also includes: Based on the fusion dataset, vibration dose profiles are established for operators whose work environment involves whole-body vibration or hand-transmitted vibration. According to the vibration dose limits set by international standards, the cumulative vibration exposure dose of each operator is calculated in real time. When the dose approaches the limit, a graded warning will be automatically pushed to the operator's personal smart terminal and the management terminal of the team leader, and the warning and evacuation execution status will be recorded to form a vibration exposure profile.

6. An automatic safety hazard detection system for highway construction sites, characterized in that, include: The data acquisition module is used to continuously collect raw data from various intelligent sensing units, and to aggregate diverse and heterogeneous raw data to a unified data receiving platform to complete the automatic mapping from the physical world to the digital world. The data processing module is used to automatically perform data cleaning, format standardization, timestamp alignment and spatial location binding on the raw data according to preset rules and algorithms. It associates displacement data, stress data and geological sketch information in the same monitoring area to generate a standardized fusion dataset with unified spatiotemporal coordinates. The hazard identification module is used to input the fused dataset into the risk diagnosis engine to identify hazards. First, it performs real-time logical judgment based on explicit rules, automatically comparing the real-time data with preset safety thresholds to obtain the first analysis result. Second, it performs correlation analysis based on a multi-parameter model, calculating the coupling relationship and changing trend of multiple key parameters by calling the built-in special engineering model to obtain the second analysis result. Finally, it uses deep pattern recognition based on machine learning to perform similarity matching between the real-time data stream and the historical accident case database to obtain the third analysis result. The analysis result fusion and judgment module is used to automatically merge the three analysis results and finally generate a structured early warning instruction that includes the type, level, location and confidence level of the hidden danger.

7. The automatic safety hazard detection system for highway construction sites according to claim 6, characterized in that, The analysis result fusion and judgment module includes: The benchmark weight adjustment unit is used to determine the specific environmental state of the current tunnel construction. Based on the preset strategies for different environmental states, it assigns an initial decision influence benchmark weight to the three analysis engines. The decision influence benchmark weight defines which type of analysis logic is more important in the current environment. The confidence score calculation unit is used to quantitatively evaluate the reliability of each analysis result after obtaining the preliminary risk assessment of each analysis engine. The evaluation criteria include the quality and completeness of the data used to generate the analysis result, the certainty of the analysis process itself, and the recent warning accuracy history of the analysis engine. A dynamic confidence score is calculated for each analysis result. The actual weight adjustment unit dynamically adjusts the actual weight of each analysis result in the final decision based on the confidence score of each analysis result. Then, a comprehensive risk value is obtained through weighted fusion, which generates a structured early warning instruction containing the type, level, location, and confidence level of the hazard. In this process, a set of hierarchical adjudication logic is executed: if any analysis engine triggers the highest level of safety red line, the highest level alarm is triggered directly and the weighted fusion is skipped; if the comprehensive risk value exceeds the preset first risk threshold and the conclusions of most analysis engines indicate a high risk level or above, a clear early warning is issued; if the comprehensive risk value exceeds the preset first risk threshold, but the number of analysis engines indicating a high risk level or above does not exceed half, an enhanced monitoring instruction is generated and pushed instead of a direct alarm to avoid false alarms.

8. The automatic safety hazard detection system for highway construction sites according to claim 6, characterized in that, Also includes: The multi-channel response and management module is used to automatically execute multi-channel response and closed-loop management based on the generated structured early warning instructions: push early warning information to the terminal devices of relevant responsible personnel and simultaneously activate the audible and visual alarm devices in the tunnel; automatically create corresponding emergency response work orders, assign them to preset response personnel, and trigger the configured linkage control instructions; at the same time, automatically track the feedback data of the response process, and automatically archive the entire process data of this case into the learning library for subsequent continuous optimization of the risk diagnosis engine.

9. The automatic safety hazard detection system for highway construction sites according to any one of claims 6 to 8, characterized in that, Also includes: The health exposure early warning module is used to dynamically monitor chronic hazard parameters based on a fusion dataset, and set different hazard level thresholds according to occupational health exposure limit standards. It tracks the identity information, real-time location, and cumulative exposure time in specific hazard environments of each worker in real time, automatically calculates individualized exposure doses, and automatically pushes graded warnings to the worker's personal smart terminal and the management terminal of the team leader when it is determined that the cumulative exposure dose or time of a worker is close to the preset safety limit. It also records the warning and evacuation execution status and forms a health exposure file.

10. The automatic safety hazard detection system for highway construction sites according to claim 9, characterized in that, Also includes: The vibration exposure early warning module is used to create vibration dose profiles for operators whose work environment is subject to whole-body vibration or hand-transmitted vibration based on a fusion dataset. According to the vibration dose limits set by international standards, it calculates the cumulative vibration exposure dose of each operator in real time. When the dose approaches the limit, it will automatically push graded warnings to the operator's personal smart terminal and the management terminal of the team leader, and record the warning and evacuation execution status to form a vibration exposure profile.

Images