Comprehensive management system and method based on hazardous gas detection alarm and personnel positioning

Abstract

The invention belongs to the technical field of industrial safety monitoring, and discloses a comprehensive management system and method based on hazardous gas detection alarm and personnel positioning, and the system comprises a gas detection module, a personnel positioning module, a data processing center and a dynamic response system. The gas detection module collects gas concentration data in real time through a point, line and plane multi-source sensor; the personnel positioning module realizes centimeter-level three-dimensional positioning based on the UWB technology; the data processing center associates the gas concentration with the personnel position through a space-time fusion analysis engine, dynamically constructs a three-dimensional gas concentration field and evaluates the risk; and the dynamic response system executes graded alarm, ventilation control and AR path guidance according to the risk grade. According to the method, closed-loop management from data acquisition, fusion analysis and risk assessment to intelligent response is realized, and the real-time performance, the accuracy and the intelligent level of gas leakage emergency response in a high-risk environment are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of industrial safety monitoring technology in chemical industrial parks, and in particular relates to a comprehensive management system and method based on hazardous gas detection alarm and personnel positioning. Background Technology

[0002] In high-risk industrial environments such as chemical industrial parks, petrochemical plants, and mines, the leakage and accumulation of hazardous gases are among the main causes of fires, explosions, and poisoning accidents. Therefore, real-time detection and alarm of hazardous gases, as well as precise location and dynamic management of personnel, have become core requirements in the field of industrial safety.

[0003] Currently, there are several independent systems for gas detection or personnel positioning in existing technologies, but they still have obvious technical limitations and system fragmentation problems, which are manifested as follows:

[0004] Gas detection systems often have limited functionality: Traditional gas detection relies on fixed or portable gas sensors, enabling real-time monitoring of gas concentration and audible and visual alarms for exceeding limits. However, such systems typically only provide concentration information at local points, lacking the ability to perceive three-dimensional spatial trends in gas diffusion, and the alarm information is not correlated with the actual location of personnel, resulting in a lack of targeted emergency response.

[0005] Personnel positioning systems operate independently: Existing personnel positioning technologies (such as UWB, Bluetooth, ZigBee, etc.) can achieve indoor and outdoor positioning accuracy from centimeter to meter level, and are widely used in scenarios such as personnel dispatching, attendance management, and electronic fences. However, these systems usually operate independently of the gas monitoring network, and there is a lack of effective spatiotemporal fusion and linkage analysis between positioning data and gas concentration data, which cannot provide personnel with risk avoidance guidance based on actual gas distribution.

[0006] Emergency response relies on human decision-making: In emergency situations such as gas leaks, existing systems often only provide raw data or simple alarms, requiring safety personnel to manually assess the severity of the danger, determine the affected area, and direct evacuation routes. This results in long response delays and a risk of missing the best opportunity to take action.

[0007] Lack of intelligent evacuation and coordinated control mechanisms: Although some advanced systems have attempted to introduce evacuation guidance functions, these are mostly based on static maps or preset paths, failing to combine real-time gas diffusion models for dynamic risk calculation and path optimization. Furthermore, the systems rarely achieve closed-loop coordinated control with on-site actuators such as ventilation equipment and isolation devices, making it difficult to achieve truly intelligent emergency management.

[0008] In summary, although gas detection and personnel positioning systems have been developed in the existing technologies, there are significant gaps between the two at the data, decision-making, and execution levels. They have failed to form an integrated closed loop of "monitoring-positioning-assessment-response," making it difficult to meet the higher requirements of real-time, accurate, and intelligent safety management in modern high-risk industrial environments.

[0009] Therefore, there is an urgent need for a comprehensive management system and method that can integrate three-dimensional monitoring of hazardous gases, precise personnel positioning, dynamic risk assessment, and intelligent response linkage to improve the inherent safety level of industrial sites. Summary of the Invention

[0010] The purpose of this invention is to provide a comprehensive management system and method based on hazardous gas detection alarm and personnel location, so as to solve the above-mentioned technical problems.

[0011] To solve the above-mentioned technical problems, the specific technical solution of the integrated management system and method based on hazardous gas detection alarm and personnel positioning of the present invention is as follows:

[0012] A comprehensive management system based on hazardous gas detection alarm and personnel positioning includes a gas detection module, a personnel positioning module, a data processing center, and a dynamic response system. The gas detection module includes a multi-type sensor array, a laser gas telemetry instrument, and an infrared gas cloud imager, which collects concentration data in real time in a point-line-area manner and transmits it to edge computing nodes to form a three-dimensional image of the hazardous gas. The personnel positioning module includes UWB positioning tags and a base station cluster for centimeter-level three-dimensional positioning. The data processing center includes a spatiotemporal fusion analysis engine for correlating gas concentration with personnel location coordinates. The dynamic response system includes a graded alarm, ventilation control, and an AR guidance terminal for executing evacuation route planning and equipment linkage control.

[0013] This invention also discloses a comprehensive management method based on a comprehensive management system for hazardous gas detection alarms and personnel location, comprising the following steps:

[0014] S1: Multi-source gas data acquisition and synchronization: Deploy three types of gas detection equipment: point, line, and area, and use a precise time protocol to achieve time synchronization of all equipment; collect gas concentration data of the monitoring area in real time at different sampling frequencies, and all data are accompanied by timestamps and spatial coordinate information;

[0015] S2: Dynamic construction of three-dimensional gas concentration field: preprocess the multi-source data collected in S1 to eliminate outliers and noise; unify all data to the same spatiotemporal reference; use a data fusion algorithm to fuse point, line and surface data to generate a high-resolution three-dimensional gridded gas concentration field, and update it regularly;

[0016] S3: Precise personnel positioning and coordinate acquisition: By measuring the propagation time or phase difference of UWB signals and combining the spatial coordinates of multiple base stations, the real-time location of the target tag is calculated;

[0017] S4: Personnel-Gas Risk Correlation and Dynamic Assessment: The coordinates of each person obtained in S3 are matched in real time with the three-dimensional gas concentration field constructed in S2, and the gas concentration at their location is calculated by interpolation; combined with the toxicity weight and safety threshold of the gas, the immediate risk value and cumulative risk value are dynamically calculated, and the short-term predicted risk value is predicted based on the gas diffusion model and personnel movement trend; the risk level is classified according to the risk value results.

[0018] S5: Tiered Early Warning and Dynamic Response: Based on the "correlation between 3D images of hazardous gases and UWB personnel positioning" in S4, an algorithm for a dynamic response system for real-time risk assessment and early warning is created, and a system is constructed.

[0019] A closed-loop process of "real-time data correlation - dynamic risk calculation - graded early warning triggering - intelligent response execution";

[0020] S6: Feedback and System Optimization: Record the complete data chain and response effect of each event, and use machine learning algorithms to continuously optimize the risk assessment model, early warning threshold and path planning algorithm to reduce the false alarm rate and improve the system's adaptability.

[0021] Furthermore, step S1 includes the following steps:

[0022] S1.1: Initialize the parameter configuration of the sensor array, laser telemetry instrument and infrared imager;

[0023] S1.2: Establish a time synchronization mechanism to ensure that the sampling time deviation of the three types of devices is <10ms;

[0024] S1.3: Set the sampling frequency:

[0025] Point sensor: 10Hz continuous sampling, unit is ppm;

[0026] Laser telemetry: 5Hz, column concentration along the laser path, in ppm·m;

[0027] Infrared imaging: 2Hz, resolution 320×256 or 640×512 pixels, each pixel is a "column density";

[0028] S1.4: Receive data from each device in real time, including:

[0029] Point data: [Device ID, timestamp, coordinates (x, y, z), gas type, concentration value, confidence level];

[0030] Line data: [Device ID, timestamp, laser line coordinate set, gas type, concentration distribution array]; Area data: [Device ID, timestamp, imaging area coordinate range, gas type, concentration matrix]. Further, step S2 includes the following steps:

[0031] S2.1: Data Preprocessing:

[0032] a) Outlier detection and correction: Outliers from point sensors are removed using the 3σ criterion; sliding window filtering is used for laser telemetry data; and median filtering is used to remove noise from infrared images.

[0033] b) Data standardization: Convert all concentration values ​​to a unified unit; perform error compensation based on equipment calibration parameters;

[0034] c) Coordinate unification: Convert all device data to the same spatial coordinate system; perform distortion correction and perspective transformation on infrared images;

[0035] S2.2: Spatiotemporal Registration: i.e., the spatiotemporal registration module, which is divided into time registration and spatial registration; a) Time Registration: based on the system master clock; interpolation processing is performed on data with different timestamps; b) Spatial Registration: coordinate calibration is performed based on preset control points; spatial correlation between point-line-surface data is established; the assignment of each data point in the three-dimensional mesh is calculated;

[0036] S2.3: Fusion modeling, i.e., multi-source data fusion modeling;

[0037] a) Constructing a 3D mesh model: Setting the mesh resolution; Initializing the 3D concentration array;

[0038] b) Layered data fusion: Point data: fills the corresponding grid using a weighted average method; Line data: interpolates and fills along the laser path, then merges with the point data; Surface data: diffuses into three-dimensional space using an inverse distance weighted algorithm.

[0039] c) Consistency check: Calculate the consistency error of different data sources in the overlapping area; if the error > threshold, start weighted iterative optimization;

[0040] d) Dynamic updates: Dynamic prediction based on time series analysis; the 3D model is updated every 0.5 seconds;

[0041] S2.4: 3D visualization and output;

[0042] a) Three-dimensional concentration field generation: Concentration distribution visualization is achieved using volume rendering technology; a concentration threshold is set, and dangerous areas are automatically marked.

[0043] b) Output results: Generate 3D mesh data; provide concentration profile and arbitrary section query functions; c) Anomaly alarm: Trigger an alarm when the concentration exceeds the safety threshold; locate the highest concentration area and analyze the diffusion path.

[0044] Furthermore, step S3 includes the following steps:

[0045] S3.1: Signal preprocessing and synchronization:

[0046] a) UWB tags periodically send nanosecond-level ultrashort pulse signals, which are received by multiple fixed base stations;

[0047] b) Filter the received signal and synchronize the tag with the base station clock through timestamps to ensure time measurement accuracy;

[0048] S3.2: Distance Measurement

[0049] a) Time-of-flight (TOF) method: Calculates the propagation time of the signal from the tag to the base station, and directly converts the distance using the speed of light;

[0050] b) Time Difference of Arrival (TDoA) method: This method does not require strict synchronization between the tag and the base station clock. It calculates the distance difference by using the time difference of the same signal arriving at different base stations.

[0051] S3.3: Location calculation, based on distance or distance difference information from multiple base stations, uses a geometric algorithm to solve for the coordinates:

[0052] a) Multi-directional positioning method: Draw a circle or sphere with each base station as the center and the measured distance as the radius. The intersection point is the tag location.

[0053] b) Hyperbolic positioning method: A hyperbola is formed with two base stations as the focal points and the distance difference between them as a constant. The intersection of multiple hyperbolas is the location.

[0054] S3.4: Optimization and Correction:

[0055] a) Use Kalman filtering or particle filtering to process the raw position data, smooth the motion trajectory, and reduce the impact of measurement noise;

[0056] b) To address the multipath effect, direct wave signals can be filtered by signal strength, or error compensation can be performed by combining environmental maps.

[0057] Furthermore, step S4 includes the following steps:

[0058] S4.1: Unification of Spatiotemporal Reference:

[0059] a) Time synchronization: Align the timestamps of UWB positioning data with those of gas 3D image data to ensure a time deviation of <50ms;

[0060] b) Spatial registration: The UWB positioning coordinate system and the gas 3D image coordinate system are unified through a coordinate transformation matrix, and the error is controlled within the positioning accuracy range;

[0061] S4.2: Real-time location-concentration correlation:

[0062] a) When the UWB system outputs the real-time coordinates (x, y, z) of the personnel, the corresponding grid data in the gas 3D model can be quickly queried through the spatial index;

[0063] b) If the personnel are not located on a grid node, the gas concentration at that point is calculated using inverse distance weighted interpolation.

[0064] Furthermore, step S5 includes the following steps:

[0065] S5.1: Risk level assessment, based on real-time personnel exposure status, constructs a multi-dimensional risk assessment model through a spatiotemporal correlation algorithm, and employs improved DBSCAN clustering regression analysis to achieve:

[0066] Risk value = f(gas concentration, diffusion rate, personnel density),

[0067] By using the improved DBSCAN clustering algorithm, the dynamically changing spatiotemporal data is first clustered into "spatiotemporal units" with similar characteristics. Then, regression analysis is combined to establish a quantitative relationship between these units and risk values, ultimately realizing that the risk value is jointly determined by gas concentration, diffusion rate and personnel density.

[0068] S5.2: Early Warning Classification and Decision-Making Level;

[0069] The path planning algorithm based on the three-dimensional distribution of hazardous gases, real-time personnel positioning, and risk assessment results calculates the optimal path from the starting point to the destination while ensuring safety. It can also dynamically adapt to changes in gas diffusion and personnel movement, including an improved scheme based on the A* algorithm: Path = minΣ(hazard coefficient × distance) + avoidance of high concentration areas.

[0070] S5.3: Dynamic Response Execution Layer:

[0071] a) Personnel response: Real-time push of risk level, evacuation route, and location of refuge point via UWB tag terminal or mobile APP;

[0072] b) System response: If multiple people in a single area are at risk of level 3 or above, the warning range will be automatically expanded to 10m around the area. If the gas concentration continues to rise, the safety threshold will be dynamically adjusted and the monitoring equipment will be linked to focus on the high-risk area to assist in emergency command.

[0073] S5.4: Feedback Optimization Layer:

[0074] a) Record the response time, personnel evacuation efficiency, and gas concentration change curve for each warning to form an event log;

[0075] b) Based on historical data, reinforcement learning is used to optimize risk weights, early warning thresholds, and path planning algorithms; the accuracy of gas diffusion prediction models and personnel trajectory predictions is calibrated regularly to reduce false alarm rates.

[0076] Furthermore, step S5.1 includes the following steps:

[0077] S5.1.1: Spatiotemporal data preprocessing and improved DBSCAN clustering;

[0078] First, collect 3D data within a continuous time window:

[0079] Spatial dimensions: gas concentration, gas diffusion rate, and population density in each area;

[0080] Time dimension: The collection timestamp of each data point ensures the time synchronization of data within the same window; Note: The improved DBSCAN adjusts the clustering rules based on this.

[0081] In addition to spatial distance, temporal proximity is also taken into account. Only regions that simultaneously satisfy spatial proximity and temporal synchronization can be classified into the same cluster.

[0082] Dynamically adjust clustering parameters: for areas with high gas concentrations or densely populated areas, reduce the spatial neighborhood radius; for gases with fast diffusion rates, shorten the time window.

[0083] This clustering method divides the entire monitoring area into multiple "spatiotemporal clusters," each cluster representing a unit with similar gas concentration, diffusion rate, and personnel density characteristics within a certain time and a certain continuous space.

[0084] S5.1.2: Regression analysis based on clustering results:

[0085] For each "spatiotemporal cluster", three core feature values ​​are extracted:

[0086] a) Average gas concentration within the cluster;

[0087] b) Average gas diffusion rate within the cluster;

[0088] c) Crowd density within the cluster;

[0089] Using these characteristic values ​​as independent variables and the actual risk level of the cluster as the dependent variable, a function relationship was fitted through multiple regression analysis:

[0090] Risk value = f(gas concentration, diffusion rate, personnel density);

[0091] The form of function f is dynamically optimized.

[0092] Furthermore, the specific algorithm and calculation process of S5.2 are as follows:

[0093] S5.2.1: Environmental Modeling: Constructing a 3D Risk Cost Mesh;

[0094] S5.2.1.1: Spatial discretization: The monitoring area is divided into a three-dimensional grid, and each grid corresponds to a unique coordinate (x, y, z);

[0095] S5.2.1.2: Risk Cost Assignment: Base Cost: Physical travel cost of the grid;

[0096] Risk cost: The risk value output by the spatiotemporal correlation algorithm is converted into a cost coefficient according to the formula: Risk cost = Basic cost × (1 + k × Risk value), where k is the risk weight;

[0097] S5.2.1.3: Dynamic update: Every 0.5-1 second, the risk cost of each grid is recalculated based on the latest gas concentration, diffusion rate and personnel density data;

[0098] S5.2.2: Path Search: Improved safe path calculation of the A* algorithm;

[0099] Based on the classic A* algorithm, risk constraints are introduced, and the steps are as follows:

[0100] S5.2.2.1: Parameter Settings:

[0101] Starting point: The person's current location;

[0102] End point: target location;

[0103] Heuristic function: h(n) = α × straight-line distance (n, destination) + β × average risk value of the region (n, destination);

[0104] S5.2.2.2: Search process:

[0105] Starting from the starting point, traverse adjacent grids and calculate the total cost for each grid f(n) = g(n) + h(n), where g(n) is the actual cost from the starting point to the current grid.

[0106] Prioritize the grid with the lowest total cost as the next step, while filtering grids with a risk value ≥ 1.0; if a local high-risk area is encountered, an automatic detour mechanism is triggered: search for feasible paths again within 3-5 meters on both sides of the original path;

[0107] S5.2.2.3: Optimal Path Output: When the destination is reached, backtrack the path nodes to generate the optimal path consisting of a grid coordinate sequence;

[0108] S5.2.3: Dynamic Adjustment: Path Update Based on Real-Time Data

[0109] S5.2.3.1: Triggering conditions:

[0110] When the risk value of a grid on the path increases by more than 0.3 within 1 second;

[0111] The population density within 30 meters in front exceeds the threshold.

[0112] The destination location has changed;

[0113] S5.2.3.2: Local replanning:

[0114] There is no need to recalculate the entire path; only a local search is performed on the path 10-20 meters in front of the trigger point to maintain the continuity with the original path.

[0115] S5.2.3.3: Multiple Path Alternatives:

[0116] Simultaneously calculate 2-3 differentiated paths, and switch to the alternative path within 0.5 seconds when the primary path is unavailable;

[0117] S5.2.4: Special Scenario Handling:

[0118] S5.2.4.1: Emergency Evacuation: When the risk value is ≥1.0, ignore the distance cost and prioritize the path with the "lowest risk value + highest passability", while pushing the location of emergency shelters along the route; S5.2.4.2: Multi-person Collaboration: When the distance between multiple people is ≤5 meters, merge the path planning to avoid cross congestion and ensure that the paths complement each other when the group evacuates to the same safe area;

[0119] S5.2.4.3: Equipment linkage: If the path passes through a ventilation opening, a signal is automatically sent to the control system to temporarily enhance ventilation in the area and reduce the local risk value.

[0120] Furthermore, S5 divides the risk value into 4 levels of early warning and dynamically triggers the corresponding strategies: Level 1: No early warning, continuous monitoring;

[0121] Level 2: A pop-up window on the terminal displays "Low concentration of gas exists in the current area," and simultaneously pushes the location of surrounding safe areas;

[0122] Level 3: Terminal audible and visual alarm + background notification to administrator, calculate and push the optimal evacuation route;

[0123] Level 4: Triggers a full-area audible and visual alarm, locks personnel locations, automatically dispatches emergency teams, and simultaneously activates ventilation equipment to disperse gases in a directional manner.

[0124] The integrated management system and method based on hazardous gas detection alarm and personnel positioning of the present invention have the following advantages:

[0125] 1. It achieves real-time and dynamic coupling of the three-dimensional distribution of hazardous gas concentration with the precise location of personnel, solving the pain point of separation between monitoring and positioning in traditional systems.

[0126] 2. By fusing multi-source data to construct a high-precision three-dimensional concentration field, the limitation of single-point sensors in being unable to perceive the overall diffusion situation is overcome.

[0127] 3. A dynamic assessment model that includes real-time, cumulative, and predictive risks has been introduced, making risk assessment more comprehensive and forward-looking.

[0128] 4. Path planning based on the improved A* algorithm and real-time risk cost map ensures the scientific nature and safety of evacuation and rescue routes.

[0129] 5. It has formed a complete automated closed loop from perception, analysis, decision-making to execution, greatly improving efficiency.

[0130] It has improved the speed and intelligence of emergency response, effectively protecting the lives of personnel. Attached Figure Description

[0131] Figure 1 This is a flowchart of the integrated management method based on hazardous gas detection alarm and personnel location according to the present invention. Detailed Implementation

[0132] To better understand the purpose, structure, and function of this invention, the integrated management system and method based on hazardous gas detection alarm and personnel positioning of this invention will be described in further detail below with reference to the accompanying drawings.

[0133] The present invention relates to a comprehensive management system based on hazardous gas detection alarm and personnel positioning, comprising a gas detection module, a personnel positioning module, a data processing center, and a dynamic response system. The gas detection module includes a multi-type sensor array (for detecting CH4 / H2S / CO, etc.), a laser gas telemetry instrument, and an infrared gas cloud imager, which collects concentration data in real time in a point-line-surface manner and transmits it to edge computing nodes to form a three-dimensional image of the hazardous gas. The personnel positioning module includes UWB positioning tags and a base station cluster for centimeter-level three-dimensional positioning (error <30cm). The data processing center includes a spatiotemporal fusion analysis engine for correlating gas concentration (3D) with personnel location coordinates (3D). The dynamic response system includes a graded alarm, ventilation control, and an AR guidance terminal for executing evacuation route planning and equipment linkage control.

[0134] The gas detection module includes:

[0135] Point-type detection unit: It consists of an array of multiple types of gas sensors deployed in the monitoring area, used to collect concentration data of various hazardous gases (CH4 / H2S / CO, etc.) at specific coordinate points;

[0136] Linear detection unit: includes at least one laser gas telemetry instrument to acquire integrated data of gas column concentration along a laser path;

[0137] Area-type detection unit: includes at least one infrared gas cloud imager, used to acquire two-dimensional gas concentration distribution data over a certain area of ​​the monitoring region;

[0138] Each detection unit achieves time synchronization through a precision time protocol, and the sampled data is accompanied by device ID, high-precision timestamp, and spatial coordinate information.

[0139] The personnel positioning module includes:

[0140] Multiple ultra-wideband (UWB) positioning base stations deployed within the monitoring area;

[0141] UWB positioning tags worn by personnel;

[0142] The personnel positioning module calculates the real-time three-dimensional coordinates of the tag by measuring the wireless signal transmission time or time difference, thereby achieving centimeter-level precise positioning of personnel.

[0143] The data processing center is the core computing unit of the system, including:

[0144] Spatiotemporal fusion analysis engine: used to perform spatiotemporal registration, coordinate unification, filtering and denoising, and data fusion on point, line, and surface multi-source heterogeneous data output by the gas detection layer, and to construct and dynamically update a three-dimensional gas concentration field model of the monitoring area;

[0145] Risk dynamic assessment module: It is used to associate the real-time location information obtained by the personnel positioning layer with the three-dimensional gas concentration field, and calculate the immediate risk, cumulative risk and predicted risk of each person based on the preset gas toxicity weight, safety threshold and exposure time, and classify the risk level.

[0146] The dynamic response system includes:

[0147] Tiered alarm system: Triggers different levels of sound, light, and vibration alarms based on the risk level;

[0148] Ventilation control system: Receives instructions from the data processing center and automatically adjusts fan speed, opens and closes isolation doors or dampers;

[0149] AR guidance terminal: Receives the optimal evacuation or rescue route generated by the route planning module and displays it to personnel in a graphical or augmented reality (AR) manner;

[0150] The path planning module is based on the improved A* algorithm and uses a three-dimensional risk cost grid as the environmental map to calculate a safe path with the lowest overall risk and shortest distance.

[0151] like Figure 1 As shown, the integrated management method based on hazardous gas detection alarm and personnel location of the present invention includes the following steps:

[0152] S1: Multi-source gas data acquisition and synchronization:

[0153] Three types of gas detection equipment—point, line, and area—are deployed, and a precise time protocol is used to achieve time synchronization of all equipment; gas concentration data of the monitoring area is collected in real time at different sampling frequencies, and all data is accompanied by timestamps and spatial coordinate information.

[0154] S1.1: Initialize the parameter configuration of the sensor array (point), laser telemetry instrument (line), and infrared imager (area);

[0155] S1.2: Establish a time synchronization mechanism to ensure that the sampling time deviation of the three types of devices is <10ms;

[0156] S1.3: Set the sampling frequency:

[0157] Point sensor: 10Hz continuous sampling, unit is ppm;

[0158] Laser telemetry: 5Hz, column concentration along the laser path, in ppm·m;

[0159] Infrared imaging: 2Hz, resolution 320×256 or 640×512 pixels, each pixel is a "column density";

[0160] S1.4: Receive data from each device in real time, including:

[0161] Point data: [Device ID, timestamp, coordinates (x, y, z), gas type, concentration value, confidence level];

[0162] Line data: [Device ID, timestamp, laser line coordinate set, gas type, concentration distribution array];

[0163] Surface data: [Device ID, timestamp, imaging area coordinate range, gas type, concentration matrix].

[0164] S2: Dynamic construction of three-dimensional gas concentration field:

[0165] The multi-source data collected in S1 are preprocessed to eliminate outliers and noise; all data are unified to the same spatiotemporal reference; a data fusion algorithm is used to fuse point, line and surface data to generate a high-resolution three-dimensional gridded gas concentration field, which is updated regularly.

[0166] S2.1: Data Preprocessing:

[0167] a) Outlier Detection and Correction

[0168] The 3σ criterion is used to eliminate outlier values ​​from point sensors;

[0169] Laser telemetry data were filtered using a sliding window (window size 5).

[0170] Infrared images are filtered using median filtering to remove noise;

[0171] b) Data standardization:

[0172] Convert all concentration values ​​to a uniform unit (ppm);

[0173] Error compensation is performed based on equipment calibration parameters;

[0174] c) Coordinate unification:

[0175] Convert all device data to the same spatial coordinate system (such as the UTM coordinate system);

[0176] Perform distortion correction and perspective transformation on infrared images;

[0177] S2.2: Spatiotemporal Registration: i.e., the spatiotemporal registration module, which is divided into temporal registration and spatial registration; a) Temporal Registration:

[0178] Based on the system master clock;

[0179] Interpolate data from different timestamps;

[0180] b) Spatial registration:

[0181] Coordinate calibration based on preset control points

[0182] Establish spatial relationships between point-line-area data

[0183] S2.3: Fusion modeling, i.e., multi-source data fusion modeling;

[0184] a) Constructing a 3D mesh model:

[0185] Set the grid resolution (e.g., 0.5m × 0.5m × 0.5m);

[0186] Initialize the three-dimensional concentration array;

[0187] b) Layered data fusion:

[0188] Point data: The corresponding grid is filled using a weighted average method (weights are based on confidence level);

[0189] Line data: Interpolation and filling are performed along the laser path, and then fused with point data;

[0190] Surface data: Diffusion into three-dimensional space using the inverse distance weighted (IDW) algorithm;

[0191] c) Consistency check:

[0192] Calculate the consistency error of different data sources in overlapping areas;

[0193] If the error exceeds the threshold, then weighted iterative optimization is initiated.

[0194] d) Dynamic updates:

[0195] Dynamic forecasting based on time series analysis;

[0196] The 3D model is updated every 0.5 seconds;

[0197] S2.4: 3D visualization and output;

[0198] a) Generation of three-dimensional concentration field:

[0199] Concentration distribution visualization is achieved using volume rendering technology;

[0200] Set concentration thresholds and automatically mark hazardous areas;

[0201] b) Output of results:

[0202] Generate 3D mesh data;

[0203] Provides concentration profile and arbitrary cross-section query functions;

[0204] c) Abnormal alarm:

[0205] An alarm is triggered when the detected concentration exceeds the safety threshold.

[0206] Locate the area of ​​highest concentration and analyze the diffusion path.

[0207] S3: Precise personnel positioning and coordinate acquisition:

[0208] Its basic algorithm calculates the real-time location of the target tag (carried on a person) by measuring the propagation time or phase difference of the UWB signal and combining the spatial coordinates of multiple base stations.

[0209] S3.1: Signal preprocessing and synchronization:

[0210] a) The UWB tag periodically sends nanosecond-level ultra-short pulse signals, and multiple fixed base stations (at least 3) receive the signals.

[0211] b) Filter the received signal (remove multipath interference and noise) and synchronize the tag with the base station clock through timestamps to ensure time measurement accuracy (down to sub-nanosecond level).

[0212] S3.2: Distance Measurement (Two Methods)

[0213] a) TOF (Time-of-Flight) method: Calculates the propagation time of the signal from the tag to the base station, combined with the speed of light (3 × 10⁻⁶). 8 Directly convert distance to m / s (distance = time × speed of light / 2, divided by 2 because round-trip distance measurement is usually used).

[0214] b) TDoA (Time Difference of Arrival): This method does not require strict synchronization between the tag and the base station clock. It calculates the distance difference by measuring the time difference between the arrival times of the same signal at different base stations (distance difference = time difference × speed of light).

[0215] S3.3: Location calculation, based on the distance (or distance difference) information of multiple base stations, uses a geometric algorithm to solve for the coordinates:

[0216] a) Multilateral positioning method (applicable to TOF): Draw a circle (2D) or a sphere (3D) with each base station as the center and the measured distance as the radius. The intersection point is the tag location.

[0217] b) Hyperbolic positioning method (applicable to TDoA): Two base stations are used as the focal points, and the distance difference is constant to form a hyperbola. The intersection of multiple hyperbolas is the location.

[0218] S3.4: Optimization and Correction:

[0219] a) Use Kalman filtering or particle filtering to process the raw position data, smooth the motion trajectory, and reduce the impact of measurement noise.

[0220] b) To address the multipath effect (errors caused by signal reflection from walls, etc.), direct wave signals are filtered by signal strength, or error compensation is performed by combining environmental maps.

[0221] S4: Personnel-Gas Risk Correlation and Dynamic Assessment

[0222] The coordinates of each person obtained by S3 are matched in real time with the three-dimensional gas concentration field constructed by S2, and the gas concentration at their location is calculated by interpolation. Combining the toxicity weight of the gas and the safety threshold, the immediate risk value and cumulative risk value are dynamically calculated, and the short-term predicted risk value is predicted based on the gas diffusion model and the movement trend of the personnel. The risk level is classified according to the risk value results.

[0223] S4.1: Unification of Spatiotemporal Reference:

[0224] a) Time synchronization: Align the timestamps of UWB positioning data (update frequency typically 10-20Hz) with those of gas 3D image data (update frequency 0.5-2Hz) to ensure a time deviation of <50ms.

[0225] b) Spatial registration: The UWB positioning coordinate system (usually based on local coordinates of the plant / indoor area) and the gas three-dimensional image coordinate system (such as UTM or local grid coordinates) are unified through a coordinate transformation matrix, and the error is controlled within the positioning accuracy range (<30cm).

[0226] S4.2: Real-time location-concentration correlation:

[0227] a) When the UWB system outputs the real-time coordinates (x, y, z) of the personnel, the corresponding grid data in the gas 3D model can be quickly queried through the spatial index.

[0228] b) If the personnel are not located on a grid node, inverse distance weighted (IDW) interpolation is used to calculate the gas concentration (such as CH4, H2S, etc.) at that point.

[0229] S5: Tiered Early Warning and Dynamic Response

[0230] Based on the correlation between S4's "3D imagery of hazardous gases and UWB personnel positioning," a dynamic response system algorithm for real-time risk assessment and early warning was created. Its core is to construct a closed-loop process of "real-time data correlation - dynamic risk calculation - graded early warning triggering - intelligent response execution," as detailed below:

[0231] S5.1: Risk level assessment, based on real-time personnel exposure status, constructs a multi-dimensional risk assessment model through a spatiotemporal correlation algorithm, and employs improved DBSCAN clustering regression analysis to achieve:

[0232] Risk value = f(gas concentration, diffusion rate, personnel density)

[0233] The core logic is as follows: Using an improved DBSCAN clustering algorithm, dynamically changing spatiotemporal data (including gas distribution, personnel location, and time information) are first clustered into "spatiotemporal units" with similar characteristics. Then, regression analysis is combined to establish a quantitative relationship between these units and risk values, ultimately realizing that the risk value is jointly determined by gas concentration, diffusion rate, and personnel density. The specific process is as follows:

[0234] S5.1.1: Spatiotemporal data preprocessing and improved DBSCAN clustering;

[0235] First, collect 3D data within a continuous time window (e.g., every 10 seconds):

[0236] Spatial dimensions: gas concentration in each area (e.g., ppm values ​​of CH4 and H2S), gas diffusion rate (concentration change rate and direction per unit time), and personnel density (number of people per unit area, based on UWB positioning statistics).

[0237] Time dimension: The collection timestamp of each data point ensures the time synchronization of data within the same window.

[0238] Note: The improved DBSCAN adjusts the clustering rules based on this:

[0239] In addition to considering spatial distance (such as whether two areas are within 5 meters), temporal proximity (whether they belong to the same time window) is also taken into account. Only areas that simultaneously satisfy spatial proximity and temporal synchronization can be classified into the same cluster.

[0240] Dynamically adjust clustering parameters: for areas with high concentrations of gas or densely populated areas, reduce the spatial neighborhood radius (e.g., from 5 meters to 3 meters) to avoid clustering ambiguity due to rapid changes in risk characteristics; for gases with fast diffusion rates, shorten the time window (e.g., from 10 seconds to 5 seconds) to more sensitively capture risk dynamics.

[0241] This clustering method divides the entire monitoring area into multiple "spatiotemporal clusters," each cluster representing a unit with similar gas concentration, diffusion rate, and population density characteristics within a specific time and continuous space.

[0242] S5.1.2: Regression analysis based on clustering results:

[0243] For each "spatiotemporal cluster", three core feature values ​​are extracted:

[0244] a) Average gas concentration within the cluster (weighted average, with higher-risk gases given higher weights);

[0245] b) Average gas diffusion rate within the cluster (combined with the diffusion direction to determine whether it is moving towards densely populated areas);

[0246] c) Cluster density (peak density, rather than average density, is a better indicator of cluster risk).

[0247] Using these characteristic values ​​as independent variables and the actual risk level of the cluster (such as accident cases in historical data or risk scores assessed by experts) as the dependent variable, a functional relationship is fitted through multiple regression analysis (such as stepwise regression or random forest regression):

[0248] Risk value = f(gas concentration, diffusion rate, personnel density)

[0249] The form of function f will be dynamically optimized: for example, when the gas concentration exceeds the safety threshold, its weight on the risk value will increase non-linearly; when the diffusion velocity direction is towards densely populated areas, the coefficient of diffusion velocity will increase significantly.

[0250] This algorithm can transform monitored gas and personnel data into "feature units" through spatiotemporal clustering in real time, and then quickly calculate the risk value of each unit using a regression function f. Compared with traditional methods, it retains DBSCAN's ability to capture spatial clustering features, and adapts to the dynamic changes in gas diffusion and personnel movement through improvements in the time dimension and dynamic parameter adjustments, making the risk value calculation more closely reflect the linkage risk of "concentration-diffusion-personnel" in real-world scenarios.

[0251] For example:

[0252] a) Immediate risk, i.e., the risk situation under the current circumstances;

[0253] Risk value = Σ(current gas concentration / safety threshold × toxicity weight) (The toxicity weight is set according to the gas toxicity, such as H2S toxicity weight is set to 1.5, CH4 to 0.8, and CO to 1.2).

[0254] b) Cumulative risk, i.e., the risk value that accumulates slowly;

[0255] Cumulative risk = Immediate risk × Exposure time (time starts after the safety threshold is exceeded, unit:

[0256] minute)

[0257] c) Predicting risk, i.e., how risk conditions change over time. This can be achieved by combining gas diffusion models (such as the Gaussian Plume model) with predictions of human movement trajectories (based on historical speed and direction).

[0258] Calculate the maximum concentration along the path of people in the next 3-5 seconds and output the "potential risk value";

[0259] S5.2: Early Warning Classification and Decision-Making Level;

[0260] The path planning algorithm, based on the three-dimensional distribution of hazardous gases, real-time personnel location, and risk assessment results, focuses on calculating the optimal path (shortest distance or least time) from the starting point to the destination while ensuring safety (risk value below a threshold), and dynamically adapting to changes in gas diffusion and personnel movement. An improved scheme based on the A* algorithm is: Path = minΣ(hazard coefficient × distance) + avoidance of high-concentration areas. The specific algorithm and calculation process are as follows:

[0261] S5.2.1: Environmental Modeling: Constructing a 3D Risk Cost Mesh;

[0262] 1. Spatial discretization: The monitoring area (such as factory area, mine) is divided into a three-dimensional grid (such as 1m×1m×0.5m), and each grid corresponds to a unique coordinate (x,y,z).

[0263] 2. Risk cost assignment:

[0264] Basic cost: The physical cost of traversing the grid (e.g., 1 for flat ground, ∞ for obstacles, 1.5 for stairs). Risk cost: The risk value output by the spatiotemporal correlation algorithm, converted into a cost coefficient according to the formula:

[0265] Risk cost = Basic cost × (1 + k × Risk value)

[0266] (k is the risk weight, such as k=5 for highly toxic gases and k=2 for low-toxic gases; when the risk value is ≥1.0, the cost is set to ∞, that is, passage is prohibited).

[0267] 3. Dynamic updates: Every 0.5-1 second, the risk cost of each grid is recalculated based on the latest gas concentration, diffusion rate and personnel density data (with a focus on updating high-risk areas and their surroundings).

[0268] S5.2.2: Path Search: Safe Path Calculation of an Improved A* Algorithm

[0269] Based on the classic A* algorithm, risk constraints are introduced, and the steps are as follows:

[0270] 1. Parameter settings:

[0271] Starting point: Person's current location (UWB real-time positioning coordinates).

[0272] End point: Target location (such as safe zone, work site).

[0273] Heuristic function: h(n) = α × straight-line distance (n, destination) + β × average risk value of the region (n, destination)

[0274] (α and β are weights; under normal circumstances, α = 0.7 and β = 0.3; in emergency situations, α = 0.9 and β = 0.1, prioritizing shortening the distance).

[0275] 2. Search process:

[0276] Starting from the starting point, traverse adjacent grids (in the six directions: up, down, left, right, front, and back) and calculate the total cost for each grid f(n) = g(n) + h(n) (where g(n) is the actual cost from the starting point to the current grid).

[0277] Prioritize the grid with the lowest total cost for the next step, while filtering out grids with a risk value ≥ 1.0 (prohibit them from entering).

[0278] If a local high-risk area is encountered (such as a sudden increase in the risk value of a certain grid on the path), an automatic detour mechanism will be triggered: a feasible path will be searched again within 3-5 meters on both sides of the original path.

[0279] 3. Optimal path output: When the destination is reached, backtrack the path nodes to generate the optimal path (including 3D coordinates and estimated travel time) consisting of a grid coordinate sequence.

[0280] S5.2.3: Dynamic Adjustment: Path Update Based on Real-Time Data

[0281] 1. Triggering conditions:

[0282] When the risk value of a grid on the path rises by more than 0.3 within 1 second (e.g., due to sudden gas diffusion);

[0283] If the population density within 30 meters in front exceeds the threshold (e.g., ≥5 people / ㎡, avoid crowding);

[0284] Change of destination (e.g., adjustment of target by emergency command).

[0285] 2. Local replanning:

[0286] There is no need to recalculate the entire path; only a local search is performed on the path 10-20 meters in front of the trigger point to maintain the continuity with the original path.

[0287] For example, if a sudden increase in H2S concentration is detected 5 meters ahead, a new sub-path is searched between the current position and the node 10 meters behind the original path to bypass the high-concentration area.

[0288] 3. Multiple path alternatives:

[0289] Simultaneously calculate 2-3 differentiated paths (such as detouring to the left, detouring to the right, and a slightly longer but safer path), and switch to the alternative path within 0.5 seconds when the main path is unavailable.

[0290] S5.2.4: Special Scenario Handling:

[0291] 1. Emergency evacuation: When the risk value is ≥1.0, ignore the distance cost and prioritize the route with the "lowest risk value + highest passability", while pushing the location of emergency shelters along the route.

[0292] 2. Multi-person collaboration: When the distance between multiple people is ≤5 meters, merge the path planning to avoid cross congestion and ensure that the paths complement each other when the group evacuates to the same safe zone.

[0293] 3. Equipment linkage: If the path passes through a ventilation opening, it automatically sends a signal to the control system to temporarily enhance ventilation in that area and reduce the local risk value.

[0294] This algorithm, through a dual-objective optimization and dynamic adjustment mechanism based on "risk-distance", can ensure that personnel avoid high-risk areas while also preventing excessively long detours. It is suitable for personnel navigation and emergency evacuation in high-risk scenarios such as chemical plants and mines.

[0295] This embodiment uses the above method to divide the risk value into 4 levels of early warning and dynamically triggers the corresponding strategies:

[0296] Level 1 (Safety): No early warning, continuous monitoring;

[0297] Level 2 (Prompt): A pop-up window on the terminal displays "Low concentration of gas exists in the current area," and simultaneously pushes the location of surrounding safe areas;

[0298] Level 3 (Warning): Terminal audible and visual alarm + background notification to administrator, calculate and push the optimal evacuation route (avoiding high concentration areas);

[0299] Level 4 (Emergency): Triggers a full-area audible and visual alarm, locates personnel, automatically dispatches an emergency team, and simultaneously activates ventilation equipment to disperse gases in a directional manner.

[0300] S5.3: Dynamic Response Execution Layer:

[0301] a) Personnel response: Real-time push of risk level, evacuation route (overlaid with gas concentration heat map), and location of refuge point via UWB tag terminal or mobile APP;

[0302] b) System response:

[0303] If multiple people in a single area are at risk of level three or above, the warning range will be automatically expanded to 10m around the area. If the gas concentration continues to rise, the safety threshold will be dynamically adjusted (e.g., the upper limit of exposure time will be shortened) and monitoring equipment (e.g., cameras) will be linked to focus on high-risk areas to assist in emergency command.

[0304] S5.4: Feedback Optimization Layer:

[0305] a) Record the response time, personnel evacuation efficiency, and gas concentration change curve for each warning to form an event log;

[0306] b) Based on historical data, reinforcement learning is used to optimize risk weights, early warning thresholds, and path planning algorithms (such as reducing the overlap between evacuation routes and high-concentration areas);

[0307] Regularly calibrate the gas diffusion prediction model and the accuracy of personnel trajectory prediction to reduce the false alarm rate.

[0308] S6: Feedback and System Optimization

[0309] Record the complete data chain and response effect of each event, and use machine learning algorithms to continuously optimize the risk assessment model, early warning threshold and path planning algorithm to reduce the false alarm rate and improve the system's adaptability.

[0310] Example 1: Emergency Response to Hazardous Gas Leakage in Chemical Plant Tank Area

[0311] This embodiment uses a chemical plant tank farm as an application scenario to explain the implementation process of this system in detail.

[0312] 1. System Deployment

[0313] Gas detection layer deployment: A multi-gas sensor node (monitoring CH4, H2S, CO) is deployed every 20 meters in the tank area to form a sensor network; an infrared gas cloud imager (IRCAM) is installed at the highest point in the area to monitor the gas cloud in the entire tank area; an open-type laser methane telemetry instrument is deployed above the key passage to scan the main process pipeline area.

[0314] Personnel positioning layer deployment: UWB positioning base stations are deployed throughout the factory area according to the line-of-sight principle to form a fully covered positioning network. All operators and inspectors entering hazardous areas wear explosion-proof smart bracelets or badges with built-in UWB tags.

[0315] Data Processing and Response Center Deployment: An edge computing server and system main control platform are deployed in the central control room of the plant, integrating a spatiotemporal fusion analysis engine, a dynamic risk assessment model, and a path planning algorithm.

[0316] 2. Operation Process

[0317] Data Acquisition and Synchronization: Point sensors acquire concentration data at a frequency of 10 Hz; laser telemetry acquires path integral concentration at a frequency of 5 Hz; and infrared imagers output gas concentration matrix images at a frequency of 2 Hz. All devices achieve time synchronization via the IEEE 1588 Precision Time Protocol (PTP), with a deviation of less than 10 ms.

[0318] Construction of three-dimensional gas concentration field:

[0319] a. Data preprocessing: Outliers are removed from point sensor data using the 3σ criterion; moving average filtering is used for laser telemetry data; and non-uniformity correction and geometric correction are performed on infrared images.

[0320] b. Spatiotemporal registration: Convert all data collected by all devices to a unified coordinate system of the plant area (with the center of the plant area as the origin), and interpolate data with different timestamps to the same time reference.

[0321] c. Data Fusion: The processed point, line, and surface data are fused into a 0.5m × 0.5m × 0.5m 3D grid. Point data directly fills the corresponding grid; line data is filled along the laser path through linear interpolation; surface data is inverted from the 2D image to 3D space using the inverse distance weighted (IDW) algorithm.

[0322] d. Dynamic update: The entire three-dimensional concentration field is updated every 0.5 seconds and visualized using volume rendering technology, and displayed in real time on the large screen in the central control room.

[0323] Personnel positioning and risk correlation:

[0324] a. The UWB tag worn by the personnel transmits signals at a frequency of 10Hz. The base station calculates its three-dimensional coordinates using the TDoA algorithm, and the positioning accuracy is better than 30cm.

[0325] b. The system associates each person's real-time coordinates with a three-dimensional gas concentration field, and calculates their exposure risk in real time by querying the concentration value of their grid.

[0326] Risk warning and dynamic response:

[0327] a. At a certain moment, the system detected that the methane concentration near storage tank No. 15 continued to rise to 25% LEL (lower explosive limit) within 10 seconds.

[0328] b. The spatiotemporal fusion engine immediately identifies the area with excessive concentration and calls the risk model to calculate: Risk value = (25 / 100) * 0.8 (CH4 toxicity weight) = 0.2 (immediate risk). Since the concentration has not spread rapidly, the predicted risk is low.

[0329] c. The system automatically designates a 10-meter radius area as a Level 2 risk zone, triggering a "prompt" level warning.

[0330] d. System Response:

[0331] Vibration alarms and text messages were sent to the wristbands of three workers in the area: "Attention! Methane concentration in tank area No. 15 ahead has increased. It is recommended to evacuate upwind."

[0332] Meanwhile, the system automatically generates the optimal evacuation route: based on the improved A* algorithm, starting from the current location of the personnel and ending at the nearest safe assembly point, with the risk weight k in the cost function set to 2, a safe route that avoids downwind and potential spread paths is calculated and pushed to the personnel's wristband.

[0333] The system sends a command to the control system to increase the speed of the two explosion-proof fans adjacent to tank No. 15 from 800 rpm to 1200 rpm to enhance local ventilation and dilute the leaked gas.

[0334] This embodiment achieves full automation and intelligence from gas leak detection, personnel location, risk calculation to linkage control, reducing the response time from minutes to seconds, which is traditionally dependent on manual discovery, and significantly improving the emergency response capability and intrinsic safety level of chemical plant areas.

[0335] Example 2: Application in guiding mine accident rescue

[0336] This embodiment uses a coal mine fire followed by a CO leak as a case study to illustrate the application of this system in rescue operations.

[0337] 1. System Deployment

[0338] Rescue robots equipped with the system's mobile unit were deployed into the disaster-stricken alleyways. The robots integrate: a laser CO telemetry device, a portable CO point sensor, an infrared imager, a UWB positioning beacon, and a mobile edge computing unit. Rescue team members wore UWB tags and vital sign sensors.

[0339] 2. Operation Process

[0340] The rescue robot travels along the main tunnel, using its own sensors to "scan" and construct a three-dimensional CO concentration field in the tunnel.

[0341] The system accurately locates the positions of all team members and trapped miners using tags worn by team members.

[0342] The data processing center combined gas distribution and personnel location information and found that the CO concentration at the corner of the alley 50 meters ahead was as high as 800 ppm (exceeding the safety threshold), and there were two trapped people.

[0343] The system activates emergency rescue mode:

[0344] a. Risk Calculation: Calculate the cumulative exposure risk of team members on their way to the rescue route.

[0345] b. Path planning: An improved A* algorithm is used to generate the shortest safe path that avoids the high concentration of CO accumulation (>500ppm) and dynamically marks the positions of the dampers that can be opened along the way.

[0346] c. Linkage control: The system sends instructions to the mine ventilation control system via the robot to remotely open specific air doors, change the ventilation path, and direct fresh airflow to the team members' travel route and high-concentration areas to forcibly dilute CO.

[0347] d. Real-time guidance: The optimal path and real-time gas concentration heat map are pushed to the explosion-proof tablet computer held by the rescue team leader to realize AR augmented reality visualization guidance.

[0348] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A comprehensive management system based on hazardous gas detection alarm and personnel positioning, characterized in that, The system includes a gas detection module, a personnel positioning module, a data processing center, and a dynamic response system. The gas detection module comprises a multi-type sensor array, a laser gas telemetry instrument, and an infrared gas cloud imager, which collects concentration data in real time in a point-line-area manner and transmits it to edge computing nodes to form a three-dimensional image of the hazardous gas. The personnel positioning module includes UWB positioning tags and a base station cluster for centimeter-level three-dimensional positioning. The data processing center includes a spatiotemporal fusion analysis engine for correlating gas concentration with personnel location coordinates. The dynamic response system includes graded alarms, ventilation control, and AR guidance terminals for executing evacuation route planning and equipment linkage control.

2. A comprehensive management method for a comprehensive management system based on hazardous gas detection alarm and personnel positioning as described in claim 1, characterized in that, Includes the following steps: S1: Multi-source gas data acquisition and synchronization: Deploy three types of gas detection equipment: point, line, and area, and use a precise time protocol to achieve time synchronization of all equipment; collect gas concentration data of the monitoring area in real time at different sampling frequencies, and all data are accompanied by timestamps and spatial coordinate information; S2: Dynamic construction of three-dimensional gas concentration field: preprocess the multi-source data collected in S1 to eliminate outliers and noise; unify all data to the same spatiotemporal reference; use a data fusion algorithm to fuse point, line and surface data to generate a high-resolution three-dimensional gridded gas concentration field, and update it regularly; S3: Precise personnel positioning and coordinate acquisition: By measuring the propagation time or phase difference of UWB signals and combining the spatial coordinates of multiple base stations, the real-time location of the target tag is calculated; S4: Personnel-Gas Risk Correlation and Dynamic Assessment: The coordinates of each person obtained in S3 are matched in real time with the three-dimensional gas concentration field constructed in S2, and the gas concentration at their location is calculated by interpolation; combined with the toxicity weight and safety threshold of the gas, the immediate risk value and cumulative risk value are dynamically calculated, and the short-term predicted risk value is predicted based on the gas diffusion model and personnel movement trend; the risk level is classified according to the risk value results. S5: Graded Early Warning and Dynamic Response: Based on the "correlation between three-dimensional images of hazardous gases and UWB personnel positioning" in S4, a dynamic response system algorithm for real-time risk assessment and early warning is created, and a closed-loop process of "real-time data correlation - dynamic risk calculation - graded early warning triggering - intelligent response execution" is constructed. S6: Feedback and System Optimization: Record the complete data chain and response effect of each event, and use machine learning algorithms to continuously optimize the risk assessment model, early warning threshold and path planning algorithm to reduce the false alarm rate and improve the system's adaptability.

3. The integrated management method according to claim 2, characterized in that, S1 includes the following steps: S1.1: Initialize the parameter configuration of the sensor array, laser telemetry instrument and infrared imager; S1.2: Establish a time synchronization mechanism to ensure that the sampling time deviation of the three types of devices is <10ms; S1.3: Set the sampling frequency: Point sensor: 10Hz continuous sampling, unit is ppm; Laser telemetry: 5Hz, column concentration along the laser path, in ppm·m; Infrared imaging: 2Hz, resolution 320×256 or 640×512 pixels, each pixel is a "column density"; S1.4: Receive data from each device in real time, including: Point data: [Device ID, timestamp, coordinates (x, y, z), gas type, concentration value, confidence level]; Line data: [Device ID, timestamp, laser line coordinate set, gas type, concentration distribution array]; Area data: [Device ID, timestamp, imaging area coordinate range, gas type, concentration matrix].

4. The integrated management method according to claim 2, characterized in that, S2 includes the following steps: S2.1: Data Preprocessing: a) Outlier detection and correction: Outliers from point sensors are removed using the 3σ criterion; sliding window filtering is used for laser telemetry data; and median filtering is used to remove noise from infrared images. b) Data standardization: Convert all concentration values ​​to a unified unit; perform error compensation based on equipment calibration parameters; c) Coordinate unification: Convert all device data to the same spatial coordinate system; perform distortion correction and perspective transformation on infrared images; S2.2: Spatiotemporal registration: i.e., the spatiotemporal registration module, which is divided into temporal registration and spatial registration; a) Time registration: Based on the system master clock, interpolate data with different timestamps; b) Spatial registration: Coordinate calibration is performed based on preset control points; Establish spatial relationships between point-line-area data; calculate the assignment of each data point in the 3D mesh; S2.3: Fusion modeling, i.e., multi-source data fusion modeling; a) Constructing a 3D mesh model: Setting the mesh resolution; Initialize the three-dimensional concentration array; b) Layered data fusion: Point data: fills the corresponding grid using a weighted average method; Line data: interpolates and fills along the laser path, then merges with the point data. Surface data: diffused into three-dimensional space using an inverse distance weighted algorithm; c) Consistency check: Calculate the consistency error of different data sources in the overlapping area; if the error > threshold, start weighted iterative optimization; d) Dynamic updates: Dynamic prediction based on time series analysis; the 3D model is updated every 0.5 seconds; S2.4: 3D visualization and output; a) Three-dimensional concentration field generation: Concentration distribution visualization is achieved using volume rendering technology; a concentration threshold is set, and dangerous areas are automatically marked. b) Output results: Generates 3D mesh data; provides concentration profile and arbitrary section query functions; c) Abnormal alarm: When the concentration exceeds the safety threshold, an alarm is triggered; the area with the highest concentration is located and the diffusion path is analyzed.

5. The integrated management method according to claim 2, characterized in that, S3 includes the following steps: S3.1: Signal preprocessing and synchronization: a) UWB tags periodically send nanosecond-level ultrashort pulse signals, which are received by multiple fixed base stations; b) Filter the received signal and synchronize the tag with the base station clock through timestamps to ensure time measurement accuracy; S3.2: Distance Measurement a) Time-of-flight (TOF) method: Calculates the propagation time of the signal from the tag to the base station, and directly converts the distance using the speed of light; b) Time Difference of Arrival (TDoA) method: This method does not require strict synchronization between the tag and the base station clock. It calculates the distance difference by using the time difference of the same signal arriving at different base stations. S3.3: Location calculation, based on distance or distance difference information from multiple base stations, uses a geometric algorithm to solve for the coordinates: a) Multi-directional positioning method: Draw a circle or sphere with each base station as the center and the measured distance as the radius. The intersection point is the tag location. b) Hyperbolic positioning method: A hyperbola is formed with two base stations as the focal points and the distance difference between them as a constant. The intersection of multiple hyperbolas is the location. S3.4: Optimization and Correction: a) Use Kalman filtering or particle filtering to process the raw position data, smooth the motion trajectory, and reduce the impact of measurement noise; b) To address the multipath effect, direct wave signals can be filtered by signal strength, or error compensation can be performed by combining environmental maps.

6. The integrated management method according to claim 2, characterized in that, S4 includes the following steps: S4.1: Unification of Spatiotemporal Reference: a) Time synchronization: Align the timestamps of UWB positioning data with those of gas 3D image data to ensure a time deviation of <50ms; b) Spatial registration: The UWB positioning coordinate system and the gas 3D image coordinate system are unified through a coordinate transformation matrix, and the error is controlled within the positioning accuracy range; S4.2: Real-time location-concentration correlation: a) When the UWB system outputs the real-time coordinates (x, y, z) of the personnel, the corresponding grid data in the gas 3D model can be quickly queried through the spatial index; b) If the personnel are not located on a grid node, the gas concentration at that point is calculated using inverse distance weighted interpolation.

7. The integrated management method according to claim 2, characterized in that, S5 includes the following steps: S5.1: Risk level assessment, based on real-time personnel exposure status, constructs a multi-dimensional risk assessment model through a spatiotemporal correlation algorithm, and employs improved DBSCAN clustering regression analysis to achieve: Risk value = f(gas concentration, diffusion rate, personnel density), By using the improved DBSCAN clustering algorithm, the dynamically changing spatiotemporal data are first clustered into "spatiotemporal units" with similar characteristics. Then, regression analysis is combined to establish a quantitative relationship between these units and risk values, ultimately realizing that the risk value is jointly determined by gas concentration, diffusion rate and personnel density. S5.2: Early Warning Classification and Decision-Making Level; The path planning algorithm based on the three-dimensional distribution of hazardous gases, real-time personnel positioning, and risk assessment results calculates the optimal path from the starting point to the destination while ensuring safety. It can also dynamically adapt to changes in gas diffusion and personnel movement, including an improved scheme based on the A* algorithm: Path = minΣ(hazard coefficient × distance) + avoidance of high concentration areas. S5.3: Dynamic Response Execution Layer: a) Personnel response: Real-time push of risk level, evacuation route, and location of refuge point via UWB tag terminal or mobile APP; b) System response: If multiple people in a single area are at risk of level 3 or above, the warning range will be automatically expanded to 10m around the area. If the gas concentration continues to rise, the safety threshold will be dynamically adjusted and the monitoring equipment will be linked to focus on the high-risk area to assist in emergency command. S5.4: Feedback Optimization Layer: a) Record the response time, personnel evacuation efficiency, and gas concentration change curve for each warning to form an event log; b) Based on historical data, reinforcement learning is used to optimize risk weights, early warning thresholds, and path planning algorithms; Regularly calibrate the gas diffusion prediction model and the accuracy of personnel trajectory prediction to reduce the false alarm rate.

8. The integrated management method according to claim 7, characterized in that, S5.1 includes the following steps: S5.1.1: Spatiotemporal data preprocessing and improved DBSCAN clustering; First, collect 3D data within a continuous time window: Spatial dimensions: gas concentration, gas diffusion rate, and population density in each area; Time dimension: The collection timestamp of each data point ensures the time synchronization of data within the same window; Note: The improved DBSCAN adjusts the clustering rules based on this: In addition to spatial distance, temporal proximity is also taken into account. Only regions that simultaneously satisfy spatial proximity and temporal synchronization can be classified into the same cluster. Dynamically adjust clustering parameters: for areas with high gas concentrations or densely populated areas, reduce the spatial neighborhood radius; For gases with fast diffusion rates, shorten the time window; This clustering method divides the entire monitoring area into multiple "spatiotemporal clusters," each cluster representing a unit with similar gas concentration, diffusion rate, and personnel density characteristics within a certain time and continuous space. S5.1.2: Regression analysis based on clustering results: For each "spatiotemporal cluster", three core feature values ​​are extracted: a) Average gas concentration within the cluster; b) Average gas diffusion rate within the cluster; c) Crowd density within the cluster; Using these characteristic values ​​as independent variables and the actual risk level of the cluster as the dependent variable, a function relationship was fitted through multiple regression analysis: Risk value = f(gas concentration, diffusion rate, personnel density); The form of function f is dynamically optimized.

9. The integrated management method according to claim 7, characterized in that, The specific algorithm and calculation process of S5.2 are as follows: S5.2.1: Environmental Modeling: Constructing a 3D Risk Cost Mesh; S5.2.1.1: Spatial discretization: The monitoring area is divided into a three-dimensional grid, and each grid corresponds to a unique coordinate (x, y, z); S5.2.1.2: Risk Cost Assignment: Base Cost: Physical travel cost of the grid; Risk cost: The risk value output by the spatiotemporal correlation algorithm is converted into a cost coefficient according to the formula: Risk cost = Basic cost × (1 + k × Risk value), where k is the risk weight; S5.2.1.3: Dynamic update: Every 0.5-1 second, the risk cost of each grid is recalculated based on the latest gas concentration, diffusion rate and personnel density data; S5.2.2: Path Search: Improved safe path calculation of the A* algorithm; Based on the classic A* algorithm, risk constraints are introduced, and the steps are as follows: S5.2.2.1: Parameter Settings: Starting point: The person's current location; End point: target location; Heuristic function: h(n) = α × straight-line distance (n, destination) + β × average risk value of the region (n, destination); S5.2.2.2: Search process: Starting from the starting point, traverse adjacent grids and calculate the total cost for each grid f(n) = g(n) + h(n), where g(n) is the actual cost from the starting point to the current grid. Prioritize the grid with the lowest total cost as the next step, while filtering grids with a risk value ≥ 1.0; if a local high-risk area is encountered, an automatic detour mechanism is triggered: search for feasible paths again within 3-5 meters on both sides of the original path; S5.2.2.3: Optimal Path Output: When the destination is reached, backtrack the path nodes to generate the optimal path consisting of a grid coordinate sequence; S5.2.3: Dynamic Adjustment: Path Update Based on Real-Time Data S5.2.3.1: Triggering conditions: When the risk value of a grid on the path increases by more than 0.3 within 1 second; The population density within 30 meters in front exceeds the threshold. The destination location has changed; S5.2.3.2: Local replanning: There is no need to recalculate the entire path; only a local search is performed on the path 10-20 meters in front of the trigger point to maintain the continuity with the original path. S5.2.3.3: Multiple Path Alternatives: Simultaneously calculate 2-3 differentiated paths, and switch to the alternative path within 0.5 seconds when the primary path is unavailable; S5.2.4: Special Scenario Handling: S5.2.4.1: Emergency evacuation: When the risk value is ≥1.0, ignore the distance cost and prioritize the path with the "lowest risk value + highest passability", while pushing the location of emergency shelters along the route; S5.2.4.2: Multi-person collaboration: When the distance between multiple people is ≤5 meters, merge the path planning to avoid cross congestion and ensure that the paths are complementary when the group evacuates to the same safe zone; S5.2.4.3: Equipment linkage: If the path passes through a ventilation opening, a signal is automatically sent to the control system to temporarily enhance ventilation in the area and reduce the local risk value.

10. The integrated management method according to claim 7, characterized in that, The S5 classifies warnings into four levels based on risk values ​​and dynamically triggers corresponding strategies: Level 1: No early warning, continuous monitoring; Level 2: A pop-up window on the terminal displays "Low concentration of gas exists in the current area," and simultaneously pushes the location of surrounding safe areas; Level 3: Terminal audible and visual alarm + background notification to administrator, calculate and push the optimal evacuation route; Level 4: Triggers a full-area audible and visual alarm, locks personnel locations, automatically dispatches emergency teams, and simultaneously activates ventilation equipment to disperse gases in a directional manner.

Images