Methods and Systems for Collaborative Safety Monitoring and Risk Early Warning of Equipment and Personnel in Coal Mines

Abstract

This invention discloses a method for collaborative safety monitoring and risk early warning of personnel and equipment in coal mines, including: deploying a sensor network to collect and process personnel and equipment data; assessing the health status of equipment based on equipment data; constructing a trajectory model after cleaning personnel data; obtaining a comprehensive risk value through risk coupling analysis based on a three-dimensional risk identification model; and responding in a graded manner based on the comprehensive risk value and determining whether model optimization is necessary. This invention also discloses a system for collaborative safety monitoring and risk early warning of personnel and equipment in coal mines, including a perception layer connected to an analysis layer via a network layer, an analysis layer connected to an execution layer, and an execution layer connected to the perception layer. This method and system for collaborative safety monitoring and risk early warning of personnel and equipment in coal mines overcomes the technical bottlenecks of traditional monitoring systems, such as data fragmentation, delayed response, and poor environmental adaptability, and reduces the false alarm rate of accidents by more than 35% through spatial coupling analysis.

Description

Technical Field

[0001] This invention belongs to the field of mine safety monitoring technology, and relates to a method for collaborative safety monitoring and risk warning of underground equipment and personnel in coal mines, as well as a system for collaborative safety monitoring and risk warning of underground equipment and personnel in coal mines. Background Technology

[0002] Current underground safety monitoring technology in coal mines suffers from three major deficiencies: First, equipment condition monitoring and personnel positioning systems operate independently. Equipment sensor networks only collect physical parameters such as temperature and vibration, while personnel positioning relies on UWB (Ultra-Wide Band) or RFID (Radio Frequency Identification) technologies. Neither type of data has a correlation model established at the transmission or analysis layers, making it impossible to simultaneously assess the threat level to surrounding personnel in the event of a sudden equipment failure. Actual cases show that in scraper conveyor chain breakage accidents, approximately 73% of secondary injuries stem from the system's failure to promptly evacuate workers within a 5-meter radius. Second, the high dust and low illumination environment underground severely interferes with monitoring accuracy. Existing binocular visual positioning systems are ineffective at dust concentrations >200 mg / m³. 3 The positioning error exceeds 0.5 meters, and traditional filtering algorithms cannot effectively handle the intermittent obstruction caused by the movement of hydraulic supports, resulting in personnel trajectory breaks or drift. Furthermore, the allocation of network transmission resources is rigid. When vibration data (safety level λ=2.2) and temperature data (λ=1.8) share the transmission channel, no differentiated protection is implemented. Critical alarm information is delayed by up to 300ms when the mining industrial ring network is congested, far exceeding the 100ms response threshold stipulated in the "Coal Mine Safety Regulations".

[0003] In summary, existing technologies suffer from several problems in traditional underground coal mine safety monitoring systems, including the disconnect between equipment status monitoring and personnel positioning data, insufficient monitoring accuracy in high dust and low light environments, and rigid network transmission resource allocation leading to delayed responses. Consequently, these technologies are unable to effectively assess the spatial coupling risks between equipment failure evolution and personnel activity trajectories. Summary of the Invention

[0004] The purpose of this invention is to provide a collaborative safety monitoring and risk warning method for equipment and personnel in underground coal mines. This method solves the problems in existing technologies, such as the disconnect between equipment status monitoring and personnel positioning data in traditional underground coal mine safety monitoring systems, insufficient monitoring accuracy in high dust and low light environments, and rigid network transmission resource allocation leading to delayed response, which in turn makes it impossible to effectively assess the spatial coupling risk between equipment failure evolution and personnel activity trajectories.

[0005] The purpose of this invention is to provide a collaborative safety monitoring and risk early warning system for underground equipment and personnel in coal mines.

[0006] The technical solution adopted in this invention is a method for collaborative safety monitoring and risk early warning of equipment and personnel in underground coal mines, comprising:

[0007] Step 1: Deploy a sensor network to collect and process personnel and equipment data;

[0008] Step 2: Assess the health status of the equipment based on equipment data;

[0009] Step 3: After cleaning the personnel data, construct a trajectory model;

[0010] Step 4: Obtain the comprehensive risk value by performing risk coupling analysis based on the three-dimensional risk identification model;

[0011] Step 5: Classify responses based on comprehensive risk values ​​and determine whether model optimization is necessary.

[0012] The invention is further characterized by:

[0013] Step one includes:

[0014] Step 1.1: Deploy intrinsically safe mining sensor networks at key equipment locations in the mining face to collect personnel movement data and equipment operating parameters in real time;

[0015] Step 1.2: Generate a time series dataset of equipment parameters based on the equipment operating parameters;

[0016] Step 1.3: Transmit the personnel movement data and equipment parameter time series dataset to the coal mine underground equipment and personnel collaborative safety monitoring and risk early warning system for processing via 5G URLLC slicing.

[0017] The device parameter time series dataset is shown below:

[0018] D equip ={( t k , t k , v k , p k , l k )∣ k =1, 2, ..., N},

[0019] in, D equip This is a time series dataset of device parameters; t k The acquisition time is recorded in UNIX timestamp format; t k This is the temperature value;v k The magnitude of the vibration vector is expressed in grams (g). p k Real-time power, in kW; l k Load factor, in percentages (%)

[0020] In 5G URLLC slicing transmission, the network resource dynamic allocation factor P net The formula for calculation is:

[0021]

[0022] in, P net This represents the dynamic allocation factor of network resources; i Indicates data type index, i =1 indicates temperature. i =2 indicates vibration. i =3 indicates power. i =4 indicates a load; c i Indicates transmission priority weight; F i This indicates the data packet size, in MB. l i Indicates the safety level coefficient; B i This represents network slice bandwidth resources, in MHz; x j This represents the real-time utilization rate of the j-th slice; or j Indicates the performance parameters of the slicing technique. k =0.75 represents a normal value for network topology repair.

[0023] Step two includes: calculating the equipment degradation index based on equipment data and an adaptive feature fusion algorithm to assess the equipment health status; if the equipment degradation index... H fail A device alarm code is generated when the value is ≥1.15;

[0024] The formula for calculating the equipment degradation index is as follows:

[0025] ,

[0026] in, Indicates the equipment degradation index; α This indicates the correction density factor, the value of which is preset according to the equipment type; β Indicates the deviation of operating conditions. β =1- e-δ（ΔS-μ）2 ,in, d The fatigue coefficient of a material is represented by Δ. S Indicates regional synchronization degree. m This represents the median synchronization degree. C Indicates the deviation from operating conditions;

[0027] The formula for calculating the deviation of operating conditions is as follows:

[0028]

[0029] in, i m : Indicates the current parameter value, m =1~4 correspond to temperature, vibration, power, and load, respectively; m m This represents the mean of the sliding window, and the window length. L =50; s m Indicates the window standard deviation; u n （m） Indicates adjacent devices n Window mean, n =1,2; s n （m） Indicates adjacent devices n The window standard deviation; Represents the normalization function. , x min , x max Determined by historical extreme values;

[0030] The device alarm codes are as follows:

[0031] ALM equip = (Device MAC, Failure Code, t alert ),

[0032] in, ALM equip This indicates the device alarm code. The Failure Code is encoded according to the national standard GB / T27930. t alert The alarm timestamp is accurate to milliseconds, and the MAC address represents the physical address of the device's network interface card.

[0033] Step three includes:

[0034] Step 3.1: Clean the personnel movement data;

[0035] Calculate frame sharpness at a resolution of 2560×1440 pixels, and then discard frames with poor sharpness. Q clarity Blurred frames with a resolution of <95 are used to clean up the data;

[0036] Step 3.2: The cleaned personnel movement data is processed by a multi-level feature fusion module to obtain enhanced features. F enh ;

[0037] Step 3.3: Enhance features F enh Construct a trajectory model.

[0038] Frame resolution Q clarity The calculation formula is as follows:

[0039]

[0040] Where W represents the number of pixels in the horizontal direction of the video frame; H represents the number of pixels in the vertical direction of the video frame; and x and y represent the horizontal and vertical coordinates of the pixel in the video frame, respectively. Represents the gradient vector The modulus;

[0041] The multi-level feature fusion module includes a 3×3 convolution module, a bottleneck layer with stacked residual structures, a feature enhancement module + spatial attention module, a layer normalization module, a channel splicing feature fusion module, and a 1×1 convolution module;

[0042] The feature enhancement module includes a deep feature extraction unit, which extracts spatial features through a ResNet-18 backbone network; the spatial attention module includes a spatiotemporal attention mechanism that generates Q, K, and V vectors through depthwise separable convolutions, as shown below:

[0043]

[0044] in, This represents the enhanced features output after computation through the attention mechanism; Q represents the Softmax activation function; Q represents the query matrix; K represents the key matrix. Represents the dimension of the key vector. d k =64; V represents the value matrix Value;

[0045] The trajectory model is shown below:

[0046] T raj ={( xn , y n , z n , t n )∣ n =1, 2, ..., T},

[0047] in, T raj Represents a trajectory model; x n , y n , z n (The coordinates are in the WGS-84 coordinate system, with an accuracy of ±0.1m;) t n Indicates the time.

[0048] The formula for calculating the overall risk value is as follows:

[0049]

[0050] in, This represents the overall risk value; Indicates the equipment degradation index; The weighting coefficient representing the risk of the k-th type of personnel; An indicator function representing the risk of the k-th type of personnel;

[0051] Personnel risks include individual abnormality risk, collaborative following risk, and group gathering risk;

[0052] The trigger threshold for individual abnormal risk is v p ≥1.8m / s∩Δ z ≤0.5m, where, v p Δ represents the speed of movement of an individual person. z This represents the vertical displacement change of an individual; the criteria for determining the risk of individual abnormalities are as follows:

[0053]

[0054] in, Indicates the movement speed of an individual person; Indicates time The position vector of the personnel at that time; This represents the position vector of a person at time t; Indicates a time interval;

[0055] The trigger threshold for collaborative following risk is ,in, This represents the rate of change of the relative distance between people over time. This indicates the duration of the cooperative following state; the criteria for determining the risk of cooperative following are as follows:

[0056]

[0057] in, Indicates the relative distance between people; Represents the position vector of person A; This represents the position vector of person B;

[0058] The trigger threshold for the risk of group gathering is r area ≥0.8 people / m 2 ∩ t stay ≥110s, where, r area Indicates the population density of the area. t stay This indicates the time people spend in densely populated areas; the criteria for determining the risk of group gatherings are as follows:

[0059]

[0060] in, Indicates the population density of the area; Indicates the number of people in the area; Indicates the area of ​​the region.

[0061] Step five includes: classifying response levels based on comprehensive risk values ​​and executing actions; if the real-time coordinate detection accuracy of the personnel's movement is lower than the set value, the model closed-loop optimization mechanism is activated; if the real-time coordinate detection accuracy of the personnel's movement is not lower than the set value, the model closed-loop optimization mechanism is not activated.

[0062] The response levels include Level I, Level II, and Level III; Level I response has a comprehensive risk value of [0, 0.35] and the action performed is data recording; Level II response has a comprehensive risk value of [0.35, 0.65] and the action performed is audible and visual warning; Level III response has a comprehensive risk value of [0.65, 1.0] and the action performed is equipment shutdown and personnel evacuation.

[0063] The closed-loop optimization mechanism of the model is as follows:

[0064]

[0065] in, Indicates the number of times the model needs to be retrained; Indicates the number of false alarms; Indicates the total number of tests; Indicates the number of underreported items; This indicates the number of safe samples.

[0066] Another technical solution adopted in this invention is a mine equipment and personnel collaborative safety monitoring and risk early warning system, including a perception layer. The perception layer is connected to the analysis layer through the network layer, the analysis layer is connected to the execution layer, and the execution layer is connected to the perception layer. The perception layer is responsible for multi-source data front-end acquisition; the network layer is responsible for data transmission and resource scheduling; the analysis layer is responsible for intelligent data analysis, risk coupling calculation, and early warning decision-making; and the execution layer is responsible for risk terminal handling and response.

[0067] Another feature of the technical solution of this invention is that:

[0068] The perception layer includes an intrinsically safe sensor array for mining and an intrinsically safe binocular camera; the intrinsically safe sensor array for mining includes a vibration sensor array for mining, an intrinsically safe infrared thermal imager, and several sensors associated with the equipment; the network layer includes a 5G+TSN converged gateway and a mining ring network switch; the analysis layer includes a risk coupling analysis engine, which is connected to the equipment health assessment module, the personnel behavior analysis engine, and the hierarchical early warning decision module; the personnel behavior analysis engine includes a multi-level feature fusion module; and the execution layer includes an explosion-proof audible and visual alarm and an intrinsically safe emergency stop controller.

[0069] The beneficial effects of this invention are as follows: This invention relates to dynamic collaborative monitoring of equipment operation status and personnel behavior in fully mechanized coal mining faces based on multi-source data fusion, as well as the hardware system architecture for implementing this method. In particular, it addresses the spatial coupling risk of mechanical equipment failure evolution and personnel activity trajectories within the mining area by constructing a hierarchical early warning mechanism. This solves the technical bottlenecks of traditional monitoring systems, such as data fragmentation, delayed response, and poor environmental adaptability. Through spatial coupling analysis, it reduces the false alarm rate of accidents by more than 35%, improves the positioning accuracy in high-dust environments to ±0.1m, and reduces the key alarm transmission delay to <10ms, providing core technical support for the construction of smart mines. Attached Figure Description

[0070] Figure 1 This is a flowchart illustrating the collaborative safety monitoring and risk warning method for underground equipment and personnel in coal mines according to the present invention.

[0071] Figure 2 This is a schematic diagram of the processing flow of the multi-level feature fusion module in this invention;

[0072] Figure 3 This is a schematic diagram of the processing flow of the depth feature extraction unit in this invention;

[0073] Figure 4 This is a schematic diagram of the structure of the coal mine underground equipment and personnel collaborative safety monitoring and risk early warning system of the present invention. Detailed Implementation

[0074] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0075] Methods for collaborative safety monitoring and risk early warning of equipment and personnel in underground coal mines, such as... Figure 1 As shown, it includes:

[0076] Step 1: Deploy a sensor network to collect and process personnel and equipment data;

[0077] Step 1.1: Deploy intrinsically safe mining sensor networks at key equipment locations in the mining face to collect personnel movement data and equipment operating parameters in real time;

[0078] Step 1.2: Generate a time series dataset of equipment parameters based on the equipment operating parameters;

[0079] The device parameter time series dataset is shown below:

[0080] D equip ={( t k , t k , v k , p k , l k )∣ k =1, 2, ..., N},

[0081] in, D equip This is a time series dataset of device parameters; t k The acquisition time is recorded in UNIX timestamp format; t k This is the temperature value; v k The magnitude of the vibration vector is expressed in grams (g). p k Real-time power, in kW; l k Load factor, in percentages (%)

[0082] Step 1.3: Transmit the personnel movement data and equipment parameter time series dataset to the coal mine underground equipment and personnel collaborative safety monitoring and risk early warning system for processing via 5G URLLC slicing;

[0083] In 5G URLLC slicing transmission, the network resource dynamic allocation factor Pnet The formula for calculation is:

[0084]

[0085] in, P net This represents the dynamic allocation factor of network resources; i Indicates data type index, i =1 indicates temperature. i =2 indicates vibration. i =3 indicates power. i =4 indicates a load; c i Indicates transmission priority weight; F i This indicates the data packet size, in MB. l i Indicates the safety level coefficient; B i This represents network slice bandwidth resources, in MHz; x j This represents the real-time utilization rate of the j-th slice; or j Indicates the performance parameters of the slicing technique. k =0.75 indicates a normal value for network topology repair;

[0086] Step 2: Assess the health status of the equipment based on equipment data;

[0087] The equipment health status is assessed by calculating the equipment degradation index based on equipment data and an adaptive feature fusion algorithm. If the equipment degradation index... H fail A device alarm code is generated when the value is ≥1.15;

[0088] The formula for calculating the equipment degradation index is as follows:

[0089] ,

[0090] in, Indicates the equipment degradation index; α This indicates the correction density factor, the value of which is preset according to the equipment type; β Indicates the deviation of operating conditions. β =1- e -δ（ΔS-μ）2 ,in, d The fatigue coefficient of a material is represented by Δ. S Indicates regional synchronization degree. m This represents the median synchronization degree. C Indicates the deviation from operating conditions;

[0091] The formula for calculating the deviation of operating conditions is as follows:

[0092]

[0093] in, i m : Indicates the current parameter value, m =1~4 correspond to temperature, vibration, power, and load, respectively; m m This represents the mean of the sliding window, and the window length. L =50; s m Indicates the window standard deviation; u n （m） Indicates adjacent devices n Window mean, n =1,2; s n （m） Indicates adjacent devices n The window standard deviation; Represents the normalization function. , x min , x max Determined by historical extreme values;

[0094] The device alarm codes are as follows:

[0095] ALM equip = (Device MAC, Failure Code, t alert ),

[0096] in, ALM equip This indicates the device alarm code. The Failure Code is encoded according to the national standard GB / T27930. t alert The alarm timestamp is accurate to milliseconds, and the MAC address represents the physical address of the device's network interface card.

[0097] Step 3: After cleaning the personnel data, construct a trajectory model;

[0098] Step 3.1: Clean the personnel movement data;

[0099] Calculate frame sharpness at a resolution of 2560×1440 pixels, and then discard frames with poor sharpness. Q clarity Blurred frames with a resolution of <95 are used to clean up the data;

[0100] Frame resolutionQ clarity The calculation formula is as follows:

[0101]

[0102] Where W represents the number of pixels in the horizontal direction of the video frame; H represents the number of pixels in the vertical direction of the video frame; and x and y represent the horizontal and vertical coordinates of the pixel in the video frame, respectively. Represents the gradient vector The modulus;

[0103] Step 3.2: The cleaned personnel movement data is processed by a multi-level feature fusion module to obtain enhanced features. F enh ;

[0104] like Figure 2 As shown, the multi-level feature fusion module includes a 3×3 convolution module, a bottleneck layer with stacked residual structures, a feature enhancement module + spatial attention module, a layer normalization module, a channel splicing feature fusion module, and a 1×1 convolution module.

[0105] The feature enhancement module includes a deep feature extraction unit, such as... Figure 3 As shown, the deep feature extraction unit extracts spatial features through the ResNet-18 backbone network; the spatial attention module includes a spatiotemporal attention mechanism that generates Q, K, and V vectors through depthwise separable convolutions, as shown below:

[0106]

[0107] in, This represents the enhanced features output after computation through the attention mechanism; Q represents the Softmax activation function; Q represents the query matrix; K represents the key matrix. Represents the dimension of the key vector. d k =64; V represents the value matrix Value;

[0108] Step 3.3: Enhance features F enh Construct a trajectory model;

[0109] The trajectory model is shown below:

[0110] T raj ={( x n , y n , z n ,t n )∣ n =1, 2, ..., T},

[0111] in, T raj Represents a trajectory model; x n , y n , z n (The coordinates are in the WGS-84 coordinate system, with an accuracy of ±0.1m;) t n Indicates time;

[0112] Step 4: Obtain the comprehensive risk value by performing risk coupling analysis based on the three-dimensional risk identification model;

[0113] The formula for calculating the overall risk value is as follows:

[0114]

[0115] in, This represents the overall risk value; Indicates the equipment degradation index; The weighting coefficient representing the risk of the k-th type of personnel; An indicator function representing the risk of the k-th type of personnel;

[0116] Personnel risks include individual abnormality risk, collaborative following risk, and group gathering risk;

[0117] The trigger threshold for individual abnormal risk is v p ≥1.8m / s∩Δ z ≤0.5m, where, v p Δ represents the speed of movement of an individual person. z This represents the amount of vertical displacement change that is abnormal in an individual.

[0118] The criteria for determining individual abnormal risk are as follows:

[0119]

[0120] in, Indicates the movement speed of an individual person; Indicates time The position vector of the personnel at that time; This represents the position vector of a person at time t; Indicates a time interval;

[0121] The trigger threshold for collaborative following risk is ,in, This represents the rate of change of the relative distance between people over time. Indicates the duration of the cooperative following state;

[0122] The criteria for determining the risk of collaborative following are as follows:

[0123]

[0124] in, Indicates the relative distance between people; Represents the position vector of person A; This represents the position vector of person B;

[0125] The trigger threshold for the risk of group gathering is r area ≥0.8 people / m 2 ∩ t stay ≥110s, where, r area Indicates the population density of the area. t stay Indicates the time people spend in a densely populated area;

[0126] The criteria for determining the risk of group gatherings are as follows:

[0127]

[0128] in, Indicates the population density of the area; Indicates the number of people in the area; Indicates the area of ​​the region;

[0129] Step 5: Classify responses based on comprehensive risk values ​​and determine whether model optimization is necessary;

[0130] Based on the comprehensive risk value, response levels are divided and actions are executed. When the real-time coordinate detection accuracy of the inspector's movement is lower than the set value, the model closed-loop optimization mechanism is activated; if the real-time coordinate detection accuracy of the inspector's movement is not lower than the set value, the model closed-loop optimization mechanism is not activated.

[0131] The response levels include Level I, Level II, and Level III; Level I response has a comprehensive risk value of [0, 0.35] and the action performed is data recording; Level II response has a comprehensive risk value of [0.35, 0.65] and the action performed is audible and visual warning; Level III response has a comprehensive risk value of [0.65, 1.0] and the action performed is equipment shutdown and personnel evacuation.

[0132] The closed-loop optimization mechanism of the model is as follows:

[0133]

[0134] in, Indicates the number of times the model needs to be retrained; Indicates the number of false alarms; Indicates the total number of tests; Indicates the number of underreported items; This indicates the number of safe samples.

[0135] Coal mine underground equipment and personnel collaborative safety monitoring and risk early warning system, such as Figure 4 As shown, it includes a perception layer, which is connected to the analysis layer through the network layer, the analysis layer is connected to the execution layer, and the execution layer is connected to the perception layer. The perception layer is responsible for multi-source data front-end acquisition; the network layer is responsible for data transmission and resource scheduling; the analysis layer is responsible for intelligent data analysis, risk coupling calculation, and early warning decision-making; and the execution layer is responsible for risk terminal handling and response.

[0136] The perception layer includes an intrinsically safe sensor array for mining and an intrinsically safe binocular camera; the intrinsically safe sensor array for mining includes a vibration sensor array for mining, an intrinsically safe infrared thermal imager, and several sensors associated with the equipment; the network layer includes a 5G+TSN converged gateway and a mining ring network switch; the analysis layer includes a risk coupling analysis engine, which is connected to the equipment health assessment module, the personnel behavior analysis engine, and the hierarchical early warning decision module; the personnel behavior analysis engine includes a multi-level feature fusion module; and the execution layer includes an explosion-proof audible and visual alarm and an intrinsically safe emergency stop controller.

[0137] The key equipment in step one of this invention includes a coal mining machine, a scraper conveyor, and hydraulic supports; temperature value t k Its measurement range is 0-150℃ with an accuracy of ±0.5℃; transmission priority weight. c i The possible values ​​are as follows: c 2 = 0.45 c 3 = 0.25 c 1 = 0.2, c 4 = 0.1; Safety level coefficient l i The possible values ​​are as follows: l 2=2.2, the rest are 1.8; Slicing technology performance parameters during URLLC slicing. or urllc =1.2; Network topology correction constant k The value is 0.75. In step two of this invention, the material fatigue coefficient is obtained by fitting historical equipment failure data, and the regional synchronization degree represents the degree of difference in the state of adjacent equipment. The median synchronization degree is dynamically updated based on historical data.

[0138] This invention relates to a time series dataset of device parameters. D equip Various parameters t k , t k , v k , p k , l k It is the original input for equipment health assessment, and the quality of its collection directly affects the equipment degradation index. Calculation accuracy. Network resource dynamic allocation factor. Safety level coefficient l i Data type precedence c i There is a positive correlation: l i The higher the value, the corresponding c i Higher weights ensure priority transmission of critical data; network slicing bandwidth. B i With data packet size F i Together they determine transmission efficiency, when F i Dynamic enhancement is required when increasing size. B i Resource allocation ratio; k As a normal quantity for network topology repair, and related to real-time utilization x j Negative correlation: when k When it rises, P net As the denominator increases, the amount of low-priority data transmission is automatically reduced.

[0139] Deviation of operating conditions in step two of this invention C Calculation depends on neighboring device parameters , To achieve coordinated spatial state monitoring of scraper conveyors and hydraulic supports, etc.; material fatigue coefficient d Generated by fitting historical fault data, with a synchronization degree Δ with the current region. S Together they constitute an exponential decay relationship: when the regional synchronicity Δ S When increased, operating condition offset β A value approaching 1 indicates that the equipment degradation index... More sensitive to equipment malfunctions; corrected density factor α Strongly correlated with equipment type: such as coal mining machines α =1.2, hydraulic support α=0.9 (preset value), reflecting the difference in risk tolerance among different devices;

[0140] Frame clarity in step three of this invention Q clarity With gradient magnitude Linear correlation, when the underground dust concentration is >200 mg / m³ 3 At that time, through W × H Resolution ensures basic feature quantity; enhanced features F enh The generation depends on the spatiotemporal attention mechanism: d k =64 key vector dimension forms a cascaded feature extraction with the ResNet-18 backbone network; coordinates ( x n , y n , z n The ±0.1m accuracy requirement Q clarity If the value is less than 95, the condition is always true; otherwise, the model retraining mechanism is triggered. ;

[0141] The weighting coefficient for personnel risk in step four of this invention can be determined as follows: w 1 = 0.5 w 2 = 0.3, w 3 = 0.2; Equipment risk item 0.6 Use the output of step two directly, when When the value is ≥1.15, the system automatically compensates and increases the weight of this item to 0.7; Personnel Risk Item China: Individual Abnormal Risk I 1. Calculation depends on step three T raj Position sequence, velocity v p Activated when speed exceeds 1.8 m / s w 1 = 0.5; Collaborative following risk I 2% of the distance change rate With duration Composition and logical relationship; spatial coupling is manifested as: when the device Rising and personnel r area >0.8 people / m 2 hour, The calculation uses a double weighting method, which is the original formula value multiplied by 1.3.

[0142] The technical parameters of some equipment in the coal mine underground equipment and personnel collaborative safety monitoring and risk early warning system of this invention are shown in Table 1:

[0143] Table 1

[0144]

[0145] This invention has the following innovative features: A pioneering equipment-personnel risk coupling mechanism. R total = f ( H fail , I k The risk fusion model breaks through the limitations of traditional single-dimensional monitoring, realizing comprehensive risk assessment of the mining environment; adaptive feature enhancement technology: based on spatial feature enhancement modules and bottleneck layer structure, it improves personnel positioning accuracy in low-light and high-dust environments; dynamic resource allocation algorithm: network resource factors P net It achieves differentiated transmission guarantee of four-dimensional parameters (temperature, vibration, power, load); a three-level closed-loop optimization system: establishes a complete closed loop of "data acquisition → risk analysis → execution feedback → model update", reducing the false alarm rate by more than 35%.

[0146] Example 1

[0147] This embodiment proposes a collaborative safety monitoring and risk early warning method for underground equipment and personnel in coal mines, including:

[0148] Step 1: Deploy a sensor network to collect and process personnel and equipment data;

[0149] Step 2: Assess the health status of the equipment based on equipment data;

[0150] Step 3: After cleaning the personnel data, construct a trajectory model;

[0151] Step 4: Obtain the comprehensive risk value by performing risk coupling analysis based on the three-dimensional risk identification model;

[0152] Step 5: Grade the response based on the comprehensive risk value and optimize the model.

[0153] Example 2

[0154] This embodiment proposes a collaborative safety monitoring and risk early warning method for underground equipment and personnel in coal mines, including:

[0155] Step 1: Deploy a sensor network to collect and process personnel and equipment data;

[0156] Step 1.1: Deploy intrinsically safe mining sensor networks at key equipment locations in the mining face to collect personnel movement data and equipment operating parameters in real time;

[0157] Step 1.2: Generate a time series dataset of equipment parameters based on the equipment operating parameters;

[0158] Step 1.3: Transmit the personnel movement data and equipment parameter time series dataset to the coal mine underground equipment and personnel collaborative safety monitoring and risk early warning system for processing via 5G URLLC slicing;

[0159] Step 2: Assess the health status of the equipment based on equipment data;

[0160] Step 3: After cleaning the personnel data, construct a trajectory model;

[0161] Step 4: Obtain the comprehensive risk value by performing risk coupling analysis based on the three-dimensional risk identification model;

[0162] Step 5: Grade the response based on the comprehensive risk value and optimize the model.

[0163] Example 3

[0164] This embodiment proposes a collaborative safety monitoring and risk early warning method for underground equipment and personnel in coal mines, including:

[0165] Step 1: Deploy a sensor network to collect and process personnel and equipment data;

[0166] Step 2: Assess the health status of the equipment based on equipment data;

[0167] The equipment health status is assessed by calculating the equipment degradation index based on equipment data and an adaptive feature fusion algorithm. If the equipment degradation index... H fail A device alarm code is generated when the value is ≥1.15;

[0168] Step 3: After cleaning the personnel data, construct a trajectory model;

[0169] Step 4: Obtain the comprehensive risk value by performing risk coupling analysis based on the three-dimensional risk identification model;

[0170] Step 5: Classify responses based on comprehensive risk values ​​and determine whether model optimization is necessary.

[0171] In this embodiment, the scraper conveyor faces a risk of chain breakage: the system collects abnormal data through intrinsically safe vibration sensors and temperature sensors, including vibration vector magnitude. v k Reaching 4.2g (continuously exceeding the 3.5g threshold for 5 seconds), temperature t k The temperature suddenly rose to 98°C (normal upper limit 75°C), and the load rate... l k The rate dropped to 15% (normal range 40%-60%), while the binocular camera detected two maintenance personnel within a 3-meter radius of the equipment, with their trajectory coordinates... Ttraj =( x 12.3, y 5.7 z :0), speed v p =0.8m / s is not exceeding the speed limit; the network layer prioritizes the transmission of vibration data through 5G URLLC slicing (safety level). l 2=2.2), dynamic allocation factor P net A bandwidth of B2 = 20MHz is calculated to ensure real-time performance;

[0172] During the risk analysis phase, the equipment health assessment module calculates the degradation index. H fail =1.24 (based on correction density factor) α =0.9, operating condition offset β =0.93 and operating condition deviation C =0.78), exceeding the 1.15 threshold, generates alarm code ALM. equip = (MAC: 00-1A-3F, GB / T27930-Code: 307, talert: 1722889200123), Coupled Risk Model R total Combined with equipment risk item 0.7× H fail and personnel risk items (individual abnormalities) I 1 Not activated), due to personnel density r area =0.85>0.8 triggers a second weighting, ultimately R total =1.128, which corresponds to a Level III response. The system immediately executes the following: the explosion-proof audible and visual alarm outputs a 110 dB buzzer and flashes a red light; the emergency stop controller cuts off the power to the scraper conveyor; and the personnel positioning system sends evacuation instructions to the workers' handheld terminals, guiding them to the coordinates of the safe zone. x： (25, y: 15, z: 0), effectively avoiding secondary damage.

[0173] Example 4

[0174] This embodiment proposes a collaborative safety monitoring and risk early warning method for underground equipment and personnel in coal mines, including:

[0175] Step 1: Deploy a sensor network to collect and process personnel and equipment data;

[0176] Step 2: Assess the health status of the equipment based on equipment data;

[0177] Step 3: After cleaning the personnel data, construct a trajectory model;

[0178] Step 3.1: Clean the personnel movement data;

[0179] Calculate frame sharpness at a resolution of 2560×1440 pixels, and then discard frames with poor sharpness. Q clarity Blurred frames with a resolution of <95 are used to clean up the data;

[0180] Step 3.2: The cleaned personnel movement data is processed by a multi-level feature fusion module to obtain enhanced features. F enh ;

[0181] The multi-level feature fusion module includes a 3×3 convolution module, a bottleneck layer with stacked residual structures, a feature enhancement module + spatial attention module, a layer normalization module, a channel splicing feature fusion module, and a 1×1 convolution module;

[0182] The feature enhancement module includes a deep feature extraction unit, which extracts spatial features through a ResNet-18 backbone network; the spatial attention module includes a spatiotemporal attention mechanism that generates Q, K, and V vectors through depthwise separable convolution.

[0183] Step 3.3: Enhance features F enh Construct a trajectory model;

[0184] Step 4: Obtain the comprehensive risk value by performing risk coupling analysis based on the three-dimensional risk identification model;

[0185] Step 5: Classify responses based on comprehensive risk values ​​and determine whether model optimization is necessary.

[0186] In this embodiment, the hydraulic support malfunction and personnel violation occurred: the equipment-side pressure sensor returned a real-time pressure of 0 MPa (below the normal lower limit of 28 MPa), and the temperature... t k =45℃ is normal; the personnel-side binocular camera captures video streams in a high-dust environment, and the clarity is good. Q clarity =102>95 ensures data validity; the feature enhancement module outputs enhanced features through a deep feature extraction unit and a spatiotemporal attention mechanism. F enh The worker's movement across the scaffolding was successfully detected, and the trajectory point was recorded. z n From 0 to 1.2m, the network layer contains power data ( l 3=1.8) Allocate default bandwidth B 3 = 10MHz;

[0187] In risk analysis, equipment health index Hfail =0.86 (based on) α =0.9 and Δ S =0.4 Calculated operating condition deviation β =0.7), personnel risk item activated. I 2 (illegal climbing behavior) and I 3 (Duration of Action) t dur =8s>6s threshold), coupling risk R total Due to abnormal equipment status, the automatic compensation weight was adjusted to 0.7, resulting in a calculated value of 1.02, triggering a Level III response. The system implemented self-locking of the support to prevent tipping, activated the audible and visual alarm, and broadcast a directional voice warning, "Do not cross the support." The violator's employee number was also recorded on the safety platform, thus strengthening behavioral constraints.

[0188] Example 5

[0189] This embodiment proposes a collaborative safety monitoring and risk early warning method for underground equipment and personnel in coal mines, including:

[0190] Step 1: Deploy a sensor network to collect and process personnel and equipment data;

[0191] Step 2: Assess the health status of the equipment based on equipment data;

[0192] Step 3: After cleaning the personnel data, construct a trajectory model;

[0193] Step 4: Obtain the comprehensive risk value by performing risk coupling analysis based on the three-dimensional risk identification model;

[0194] Step 5: Classify responses based on comprehensive risk values ​​and determine whether model optimization is necessary;

[0195] Based on the comprehensive risk value, the response level is divided and the action is executed. When the real-time coordinate detection accuracy of the detection personnel's movement is lower than the set value, the model closed-loop optimization mechanism is activated.

[0196] The response levels include Level I, Level II, and Level III; Level I response has a comprehensive risk value of [0, 0.35] and the action performed is data recording; Level II response has a comprehensive risk value of [0.35, 0.65] and the action performed is audible and visual warning; Level III response has a comprehensive risk value of [0.65, 1.0] and the action performed is equipment shutdown and personnel evacuation.

[0197] This embodiment addresses positioning drift caused by dust interference: after a blasting operation, the dust concentration exceeds 300 mg / m³. 3 Video stream resolution Q clarity=82<95 triggers culling for 3 consecutive frames, causing a jump in the original personnel trajectory, from ( x 10, y :8) to ( x 13, y :11), speed false alarm v p =2.1m / s, dynamic optimization mechanism activated, based on historical false alarm data. N FP =3 and total number of tests N det =100 Calculate the number of retraining iterations N retrain =⌈120×(3 / 100+0.5×0 / 50)⌉=4, load dust mode parameters; feature enhancement module corrects coordinates to ( x 11.2, y :8.9) and will v p The value was revised to 0.9 m / s, and the risk value was adjusted from the misjudgment level. R total =0.58 (Activation) I 1) The response level was reduced to 0.21, downgrading the response level from Level II audible and visual warning to Level I data recording. Only events are recorded to avoid accidental shutdowns, thus optimizing resource usage.

[0198] Example 6

[0199] This embodiment proposes a collaborative safety monitoring and risk early warning system for equipment and personnel in underground coal mines. The system includes a perception layer, which is connected to the analysis layer via a network layer, the execution layer, and the perception layer itself. The perception layer is responsible for multi-source data acquisition; the network layer is responsible for data transmission and resource scheduling; the analysis layer is responsible for intelligent data analysis, risk coupling calculation, and early warning decision-making; and the execution layer is responsible for risk terminal handling and response. The perception layer includes an intrinsically safe sensor array and an intrinsically safe binocular camera. The intrinsically safe sensor array includes a vibration sensor array, an intrinsically safe infrared thermal imager, and several sensors associated with the equipment. The network layer includes a 5G+TSN fusion gateway and a mining ring network switch. The analysis layer includes a risk coupling analysis engine, which is connected to an equipment health assessment module, a personnel behavior analysis engine, and a graded early warning decision-making module. The personnel behavior analysis engine includes a multi-level feature fusion module. The execution layer includes an explosion-proof audible and visual alarm and an intrinsically safe emergency stop controller.

[0200] This invention achieves precise, real-time, and intelligent safety monitoring and risk early warning in coal mines through collaborative innovation of a multi-level technical architecture. At the data sensing level, the combination of a mining vibration sensor array (frequency response range 0-5kHz) and an intrinsically safe infrared thermal imager (temperature resolution 0.1℃) can accurately capture key parameters such as equipment vibration and temperature. Combined with high-definition video acquisition (2560×1440 pixels) from an intrinsically safe binocular camera, and quantitative filtering through frame clarity... Q clarity ≥95) Effectively filters fuzzy data in high-dust environments, providing high-quality raw input for subsequent analysis. Its data accuracy reaches industry-leading levels. For example, the detection error of the laser gas monitoring module for CO is ≤0.1ppm, laying a data foundation for early risk identification.

[0201] The network transmission stage employs a collaborative architecture combining 5G URLLC slicing technology (air interface latency ≤8ms) and a 5G+TSN converged gateway. Drawing inspiration from the "dual-transmitter, selective receiver" scheme of Madiliang Coal Mine and the dual-frequency private network design of Buliangou Coal Mine, a ring-redundant transmission link is constructed to achieve low-latency, high-reliability transmission of equipment monitoring data and personnel video streams. Compared to traditional fiber optic communication, this network architecture offers 1-2 levels of improved anti-interference capability, effectively resolving communication interruptions caused by lightning strikes and equipment start-up / shutdown, ensuring "zero-interruption" operation of critical services such as remote control of coal mining machines. Furthermore, edge computing (MEC) enables local data processing, further reducing transmission pressure and response latency.

[0202] The analysis layer achieves accurate risk assessment through multi-level feature fusion and intelligent algorithm coupling. Specifically, the feature fusion module based on the ResNet-18 backbone network and spatiotemporal attention mechanism enhances the extraction accuracy of personnel contours and motion features, significantly improving the 3D coordinate accuracy of trajectory modeling (±0.1m). The risk coupling analysis engine employs the "equipment degradation index (H... fail The weighting model is "0.6 × 0.6 + personnel risk × 0.4", combined with abnormal personnel speed (≥1.8m / s) and group density (≥0.8 people / m²). 2 Dynamic thresholds, such as those for equipment and personnel risks, are used to achieve quantitative coupling between these risks. This is achieved through a model retraining mechanism (N...). retrain ) for false alarms (N FP ), underreporting (N) FN Through continuous learning of data, the system's early warning accuracy has increased by 27.3% and the false alarm rate has decreased by 41.6%, making it more adaptable to complex working conditions than traditional single-indicator early warning methods.

[0203] The rapid response mechanism at the execution layer forms a closed-loop security protection. When the comprehensive risk value ( R totalWhen an alarm is triggered, the explosion-proof audible and visual alarm (110dB@1m) can issue an immediate warning, and the intrinsically safe emergency stop controller (response time ≤100ms) can quickly cut off the power supply to the hazardous equipment, significantly improving the response speed compared to the industry standard requirement of "local control execution time ≤2s". This closed-loop "monitoring-analysis-decision-execution" system, combined with a five-level graded early warning mechanism, can achieve multi-scenario risk linkage and handling from gas over-limit, equipment abnormality to personnel violations. For example, the early warning response time for the risk of personnel gathering is shortened to the millisecond level, and potential hazards can be discovered 2-3 days earlier than the traditional manual inspection mode, significantly reducing the accident rate.

[0204] The system as a whole has achieved a synergistic improvement in safety and efficiency through technological innovation. On the one hand, the combination of the equipment health assessment module (processing latency <50ms) and the personnel behavior analysis engine (supporting 200 video streams) has reduced the workforce in the fully mechanized mining face to 3-5 people, enabling unmanned operation of the substation and significantly reducing labor costs. On the other hand, through the dynamic coupling and hierarchical handling of "equipment-personnel" risks, the accuracy rate of identifying "three violations" in coal mines has reached 92%, and the early warning period for major risks has been extended to 4.8 hours, realizing a transformation from passive defense to proactive early warning and providing a scalable technological paradigm for the construction of smart mines.

Claims

1. A method for collaborative safety monitoring and risk early warning of equipment and personnel in coal mines, characterized in that, include: Step 1: Deploy a sensor network to collect and process personnel and equipment data; Step 1.1: Deploy intrinsically safe mining sensor networks at key equipment locations in the mining face to collect personnel movement data and equipment operating parameters in real time; Step 1.2: Generate a time series dataset of equipment parameters based on the equipment operating parameters; Step 1.3: Transmit the personnel movement data and equipment parameter time series dataset to the coal mine underground equipment and personnel collaborative safety monitoring and risk early warning system for processing via 5G URLLC slicing; Step 2: Assess the health status of the equipment based on equipment data; The equipment health status is assessed by calculating the equipment degradation index based on equipment data and an adaptive feature fusion algorithm. If the equipment degradation index... H fail A device alarm code is generated when the value is ≥1.15; The formula for calculating the equipment degradation index is as follows: ， in, Indicates the equipment degradation index; α This indicates the correction density factor, the value of which is preset according to the equipment type; β Indicates the deviation of operating conditions. ,in, δ The fatigue coefficient of a material is represented by Δ. S Indicates regional synchronization degree. μ This represents the median synchronization degree. Γ Indicates the deviation from operating conditions; The formula for calculating the deviation of the operating condition is as follows: in, θ m : Indicates the current parameter value, m =1~4 correspond to temperature, vibration, power, and load, respectively; μ m This represents the mean of the sliding window, and the window length. L =50; σ m Indicates the standard deviation of the window; u n （m） Indicates adjacent devices n The window mean, n =1,2; s n （m） Indicates adjacent devices n The window standard deviation; Represents the normalization function. , x min , x max Determined by historical extreme values; The device alarm codes are as follows: ALM equip = (Device MAC, Failure Code, t alert ), in, ALM equip This indicates the device alarm code. The Failure Code is encoded according to the national standard GB / T27930. t alert The alarm timestamp is accurate to milliseconds, and the MAC address represents the physical address of the device's network interface card. Step 3: After cleaning the personnel data, construct a trajectory model; Step 3.1: Clean the personnel movement data; Calculate frame sharpness at a resolution of 2560×1440 pixels, and then discard frames with poor sharpness. Q clarity Blurred frames with a resolution of <95 are used to clean up the data; Step 3.2: The cleaned personnel movement data is processed by a multi-level feature fusion module to obtain enhanced features. F enh ; Step 3.3: Enhance features F enh Construct a trajectory model; Step 4: Obtain the comprehensive risk value by performing risk coupling analysis based on the three-dimensional risk identification model; The formula for calculating the comprehensive risk value is as follows: in, This represents the overall risk value; Indicates the equipment degradation index; The weighting coefficient representing the risk of the k-th type of personnel; An indicator function representing the risk of the k-th type of personnel; The personnel risks mentioned include individual abnormality risk, collaborative following risk, and group gathering risk; The trigger threshold for the individual abnormal risk is v p ≥1.8m / s∩Δ z ≤0.5m, where, v p Δ represents the speed of movement of an individual person. z This represents the vertical displacement change of an individual; the criteria for determining the risk of such an individual anomaly are as follows: in, Indicates the movement speed of an individual person; Indicates time The position vector of the personnel at that time; This represents the position vector of a person at time t; Indicates a time interval; The trigger threshold for the collaborative following risk is: ,in, This represents the rate of change of the relative distance between people over time. This indicates the duration of the cooperative following state; the criteria for determining the cooperative following risk are as follows: in, Indicates the relative distance between people; Represents the position vector of person A; This represents the position vector of person B; The trigger threshold for the risk of group gathering is ρ area ≥0.8 people / m 2 ∩ t stay ≥110s, where, ρ area Indicates the population density of the area. t stay This indicates the time people spend in densely populated areas; the criteria for determining the risk of group gatherings are as follows: in, Indicates the population density of the area; Indicates the number of people in the area; Indicates the area of ​​the region; Step 5: Classify responses based on comprehensive risk values ​​and determine whether model optimization is necessary.

2. The method for collaborative safety monitoring and risk early warning of underground equipment and personnel in coal mines according to claim 1, characterized in that, The time series dataset of the device parameters is shown below: D equip ={（ t k ， τ k ， v k ， p k ， l k ）∣ k =1，2，…， N}， in, D equip This is a time series dataset of device parameters; t k The acquisition time is recorded in UNIX timestamp format; τ k This is the temperature value; v k The magnitude of the vibration vector is expressed in grams (g). p k Real-time power, in kW; l k Load factor, in percentages (%) In the 5G URLLC slice transmission, the network resource dynamic allocation factor Ψ net The formula for calculation is: in, Ψ net This represents the dynamic allocation factor of network resources; i Indicates data type index, i =1 indicates temperature. i =2 indicates vibration. i =3 indicates power. i =4 indicates a load; γ i Indicates transmission priority weight; Φ i This indicates the data packet size, in MB. λ i Indicates the safety level coefficient; B i This represents network slice bandwidth resources, in MHz; ξ j This represents the real-time utilization rate of the j-th slice; η j Indicates the performance parameters of the slicing technique. κ =0.75 represents a normal value for network topology repair.

3. The method for collaborative safety monitoring and risk early warning of underground equipment and personnel in coal mines according to claim 1, characterized in that, The frame clarity Q clarity The calculation formula is as follows: Where W represents the number of pixels in the horizontal direction of the video frame; H represents the number of pixels in the vertical direction of the video frame; and x and y represent the horizontal and vertical coordinates of the pixel in the video frame, respectively. Represents the gradient vector The modulus; The multi-level feature fusion module includes a 3×3 convolution module, a bottleneck layer with stacked residual structures, a feature enhancement module, a spatial attention module, a layer normalization module, a channel splicing feature fusion module, and a 1×1 convolution module. The feature enhancement module includes a deep feature extraction unit, which extracts spatial features through a ResNet-18 backbone network; the spatial attention module includes a spatiotemporal attention mechanism that generates Q, K, and V vectors via depthwise separable convolution, as shown below: in, This represents the enhanced features output after computation through the attention mechanism; Q represents the Softmax activation function; Q represents the query matrix; K represents the key matrix. Represents the dimension of the key vector. d k =64; V represents the value matrix Value; The trajectory model is shown below: T raj ={（ x n ， y n ， z n ， t n ）∣ n =1，2，…， T}， in, T raj Represents a trajectory model; x n , y n , z n (The coordinates are in the WGS-84 coordinate system, with an accuracy of ±0.1m;) t n Indicates the time.

4. The method for collaborative safety monitoring and risk early warning of underground equipment and personnel in coal mines according to claim 1, characterized in that, Step five includes: classifying response levels based on comprehensive risk values ​​and executing actions; if the real-time coordinate detection accuracy of the personnel movement is lower than a set value, the model closed-loop optimization mechanism is activated; if the real-time coordinate detection accuracy of the personnel movement is not lower than the set value, the model closed-loop optimization mechanism is not activated. The response levels include Level I, Level II, and Level III; Level I response has a comprehensive risk value of [0, 0.35] and the action performed is data recording; Level II response has a comprehensive risk value of [0.35, 0.65] and the action performed is audible and visual warning; Level III response has a comprehensive risk value of [0.65, 1.0] and the action performed is equipment shutdown and personnel evacuation. The closed-loop optimization mechanism of the model is as follows: in, Indicates the number of times the model needs to be retrained; Indicates the number of false alarms; Indicates the total number of tests; Indicates the number of underreported items; This indicates the number of safe samples.

5. A collaborative safety monitoring and risk early warning system for underground equipment and personnel in coal mines, characterized in that: The method for collaborative safety monitoring and risk warning of underground equipment and personnel in coal mines as described in any one of claims 1-4 includes a perception layer, which is connected to an analysis layer via a network layer, the analysis layer is connected to an execution layer, and the execution layer is connected to the perception layer; the perception layer is responsible for multi-source data front-end acquisition; and the network layer is responsible for data transmission and resource scheduling. The analysis layer is responsible for intelligent data analysis, risk coupling calculation, and early warning decision-making. The execution layer is responsible for handling and responding to risk terminals.

6. The coal mine underground equipment and personnel collaborative safety monitoring and risk early warning system according to claim 5, characterized in that, The perception layer includes an intrinsically safe sensor array for mining and an intrinsically safe binocular camera; the intrinsically safe sensor array for mining includes a vibration sensor array for mining, an intrinsically safe infrared thermal imager, and several sensors associated with the equipment; the network layer includes a 5G+TSN converged gateway and a mining ring network switch; the analysis layer includes a risk coupling analysis engine, which is connected to the equipment health assessment module, the personnel behavior analysis engine, and the hierarchical early warning decision module; the personnel behavior analysis engine includes a multi-level feature fusion module; The execution layer includes an explosion-proof audible and visual alarm and an intrinsically safe emergency stop controller.

Images