A mine belt conveyor fire hazard grade determination and intelligent prevention and control method

Abstract

The application discloses a kind of mine belt conveyor fire hazard grade determination and intelligent prevention and control method, comprising the following steps: one, the fire monitoring data set of mine belt conveyor roadway is collected;Two, dimensionless processing is carried out to fire monitoring data set;Three, the training of fire hazard grade determination model;Four, the fire hazard grade determination of subsequent mine belt conveyor roadway is carried out;Five, two-level smoldering stage fire prevention and control;Six, three-stage open fire stage fire prevention and control.The application builds LSTM-DAE feature network, uses improved DAE algorithm to replace encoding and decoding neurons with LSTM neurons, introduces attention mechanism to improve the effectiveness of feature extraction, and then uses two fully connected layer networks and a Softmax classifier to output the corresponding fire hazard level of the current input monitoring data, which is more accurate and faster in determining the hazard level. Different fire prevention and control measures are taken according to different levels to make the fire prevention and control more intelligent.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of mine fire monitoring and prevention, and particularly relates to a method for determining fire hazard grade and intelligently preventing and controlling mine belt conveyor. BACKGROUND

[0002] With the rapid development of China's economy and the exhaustion of shallow coal seams, coal mining has gradually extended downward, leading to frequent coal and gas outburst, coal spontaneous combustion fire, and secondary disaster accidents induced by external fire, increasing the risk of safety production. In recent years, mine conveyor belt fires, electrical equipment fires, and other external fires have occurred frequently, causing a large number of casualties and property losses, and even mass casualty accidents have attracted considerable social attention, causing considerable adverse impact on social stability and China's political image in the world.

[0003] Mine external fires, especially mine conveyor belt fires, are the most prominent. Currently, there is less research on mine conveyor belt fire monitoring and early warning, and the corresponding detection and early warning technology is not mature. Mine conveyor belt fires differ from coal spontaneous combustion fires in many ways, with long fire lines, difficult fire hazard identification, and strong toxic smoke. Due to the suddenness and rapid development, there is a lack of comprehensive early warning and prevention technology, the existing monitoring system has low coverage, early warning is not timely, and prevention is lagging. After a fire occurs, a large amount of toxic smoke is generated, and it is easy to ignite other underground structures, thereby causing numerous secondary disasters. The rubber material of the mine flame-retardant belt conveyor is flammable, producing a large amount of toxic gas and causing serious harm. Therefore, in order to improve the intelligent control level of coal mine safety production and prevent mine conveyor belt fire accidents, it is necessary and urgent to research relevant monitoring and prevention methods. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a method for determining fire hazard grade and intelligently preventing and controlling mine conveyor belt fires, which addresses the shortcomings of the prior art. By constructing an LSTM-DAE feature network, the improved denoising autoencoder DAE replaces the neurons in the hidden layer with long short-term memory neural network LSTM, improves the encoder through the attention mechanism to improve the effectiveness of belt combustion feature extraction, and then uses two fully connected layer networks and a Softmax classifier to output the corresponding fire hazard grade of the current input monitoring data according to the characteristics of belt combustion. The hazard grade determination accuracy is higher, the early warning speed is faster, and different fire prevention measures are taken according to different fire hazard grades, making fire prevention more intelligent.

[0005] To solve the above technical problems, the technical solution adopted by the present application is as follows: a method for determining fire hazard grade and intelligently preventing and controlling mine conveyor belt fires, characterized in that the method comprises the following steps:

[0006] Step one, collect the fire monitoring data set of the mine belt conveyor roadway:

[0007] Collect the fire smoke concentration in the mine belt conveyor roadway and the temperature of the mine belt conveyor roadway as the fire monitoring data set; wherein the fire smoke includes smoke, oxygen, methane, hydrogen chloride, sulfur dioxide, hydrogen cyanide, carbon dioxide and carbon monoxide;

[0008] Step two, dimensionless processing of the fire monitoring data set: after data cleaning and noise value elimination of the fire monitoring data set, the maximum and minimum standardization method is used for dimensionless processing of the fire monitoring data set;

[0009] Step three, training of the fire hazard level determination model;

[0010] Step 301, establishing a fire hazard level determination model, which includes an LSTM-DAE feature network, a two-layer fully connected neural network and a Softmax classifier;

[0011] Step 302, dividing the dimensionless processed fire monitoring data set into a training set and a test set;

[0012] Step 303, using the training set to train the LSTM-DAE feature network, wherein the attention mechanism is introduced to filter the feature information and adjust the weight of the feature information;

[0013] Step 304, using the trained LSTM-DAE feature network to extract the feature information of the test set data as input, using a two-layer fully connected neural network and a Softmax classifier to realize the determination of the fire hazard level;

[0014] Step 305, by minimizing the error between the output result and the actual labeled fire hazard level of the test set, using the cross entropy loss function to update the parameters of the two-layer fully connected neural network;

[0015] Step four, using the trained fire hazard level determination model to determine the fire hazard level of the mine belt conveyor roadway in the subsequent stage; the fire hazard level includes the first normal stage, the second smoldering stage and the third open fire stage;

[0016] When it is determined as the first normal stage, step four is executed cyclically;

[0017] When it is determined as the second smoldering stage, step five is executed;

[0018] When it is determined as the third open fire stage, step six is executed;

[0019] Step five, fire prevention and control in the second smoldering stage;

[0020] Step six, three-stage open fire stage fire prevention and control.

[0021] The method for determining and intelligently preventing and controlling the fire hazard level of the mine belt conveyor has the characteristics that, in step one, the fire monitoring data set of the mine belt conveyor tunnel is collected by a fire monitoring and early warning system, the fire monitoring and early warning system comprises a central server and a plurality of fire monitoring units in communication with the central server, and the plurality of fire monitoring units are respectively arranged above the head of the belt conveyor, above the tail of the belt conveyor and above the power transformation chamber; the fire monitoring unit comprises an environmental gas sampling pipeline and a fire monitoring and early warning device connected with the environmental gas sampling pipeline;

[0022] The fire monitoring and early warning device comprises a shell, a gas chamber arranged in the shell, an air extraction pump for extracting environmental gas into the gas chamber, a gas sensor module arranged in the gas chamber and a control circuit board arranged in the shell; the control circuit board is integrated with a microcontroller and a wireless communication module connected with the microcontroller and used for remote communication with the central server;

[0023] The gas sensor module comprises a gas detection circuit board, and a smoke sensor, an oxygen sensor, a methane sensor, a hydrogen chloride sensor, a sulfur dioxide sensor, a hydrogen cyanide sensor, a carbon dioxide sensor, a carbon monoxide sensor and a temperature sensor arranged on the gas detection circuit board;

[0024] The environmental gas sampling pipeline comprises a sampling ring pipe and a plurality of air extraction nozzles uniformly arranged along the circumference of the sampling ring pipe and opening downward, and the sampling ring pipe is provided with mounting holes for mounting the air extraction nozzles;

[0025] The air inlet end of the air extraction pump is communicated with the sampling ring pipe through an air extraction pipe, the air extraction pipe penetrates through the shell, the bottom of the gas chamber is provided with an air inlet pipe connected with the air outlet end of the air extraction pump, and the top of the gas chamber is provided with an air outlet pipe, one end of the air outlet pipe away from the gas chamber is communicated with the outside of the shell.

[0026] The one end of the air extraction pipe communicated with the sampling ring pipe is provided with a first sintered filter and a water-proof air-permeable membrane, and the one end of the air outlet pipe away from the gas chamber is provided with a second sintered filter.

[0027] The one end of the air extraction pipe communicated with the sampling ring pipe is provided with a first sintered filter and a water-proof air-permeable membrane, and the one end of the air outlet pipe away from the gas chamber is provided with a second sintered filter.

[0028] The one end of the air extraction pipe communicated with the sampling ring pipe is provided with a first sintered filter and a water-proof air-permeable membrane, and the one end of the air outlet pipe away from the gas chamber is provided with a second sintered filter.

[0029] The mine belt conveyor fire hazard grade determination and intelligent prevention and control method has the characteristics that an audible and visual alarm is arranged outside the shell, and the audible and visual alarm is connected with the microcontroller.

[0030] Compared with the prior art, the present application has the following advantages:

[0031] The present application has the following advantages compared with the prior art:

[0032] The technical solutions of the present application will be further described in detail below with reference to the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 The structure diagram of the fire monitoring and early warning system used in the present application.

[0034] Figure 2 The structure diagram of the fire monitoring unit used in the present application.

[0035] Figure 3 The Figure 2 The bottom view of the ambient gas sampling pipeline.

[0036] Figure 4 The circuit principle block diagram of the fire monitoring and early warning device used in the present application.

[0037] Figure 5 The flow chart of the method of the present application.

[0038] Explanation of reference signs:

[0039] 1-belt conveyor roadway; 2-transformer chamber; 3-central server;

[0040] 4-fire monitoring unit; 5-belt conveyor; 6-shell;

[0041] 7-gas chamber; 8-pump; 9-control circuit board;

[0042] 10-microcontroller; 11-wireless communication module; 12-gas sensor module;

[0043] 13-smoke sensor; 14-oxygen sensor; 15-methane sensor;

[0044] 16-hydrogen chloride sensor; 17-sulfur dioxide sensor; 18-hydrogen cyanide sensor;

[0045] 19-carbon dioxide sensor; 20-carbon monoxide sensor; 21-temperature sensor;

[0046] 22-sampling ring; 23-exhaust nozzle; 24-exhaust pipe;

[0047] 25-inlet pipe; 26-outlet pipe; 27-first sintered filter;

[0048] 28-waterproof air permeable membrane; 29-second sintered filter; 30-third sintered filter;

[0049] 31-communication antenna; 32-acoustic light alarm. DETAILED DESCRIPTION

[0050] As shown in Figure 1 A mine belt conveyor fire hazard level determination and intelligent prevention and control method, the method comprises the following steps:

[0051] Step one, collect the fire monitoring data set of the mine belt conveyor roadway:

[0052] Collect the fire smoke concentration in the mine belt conveyor roadway 1 and the temperature of the mine belt conveyor roadway 1 as the fire monitoring data set; wherein the fire smoke includes smoke, oxygen, methane, hydrogen chloride, sulfur dioxide, hydrogen cyanide, carbon dioxide and carbon monoxide;

[0053] Step two, dimensionless processing of the fire monitoring data set: after data cleaning and noise value elimination of the fire monitoring data set, the maximum and minimum standardization method is used for dimensionless processing of the fire monitoring data set;

[0054] Specifically, data cleaning can use the average, median and mode to fill in the missing data, and if it is a periodic continuous variable, the average value of each period at that time can be used to fill in. Interpolation method can also be selected, which is to average the two points before and after the missing point, and the calculation result is the interpolation value.

[0055] Step three, training of the fire hazard level determination model;

[0056] Step 301, establishing a fire hazard level determination model, the fire hazard level determination model comprises an LSTM-DAE feature network, a two-layer fully connected neural network and a Softmax classifier;

[0057] It should be noted that the LSTM-DAE feature network is improved by selecting the DAE algorithm, replacing the encoding and decoding neurons in the structure with LSTM neurons, and making the reconstructed data consistent with the original data after the input noisy sequence data is trained by the model. In the LSTM-DAE architecture, the entire input sequence is first read, and Gaussian noise is added after reading the entire input sequence; then, the features of the input sequence with noise are extracted and reduced to a specified size vector; finally, the vector is output through the decoder model, thereby generating the output sequence.

[0058] Step 302, divide the dimensionless processed fire monitoring data set into training set and test set;

[0059] In this embodiment, 70% of the fire monitoring data set is used as the training set, and the remaining 30% of the data set is used as the test set for model testing and performance evaluation;

[0060] It should be noted that the fire hazard level corresponding to each group of data in the fire monitoring data set can be artificially given according to the actual situation in the belt conveyor tunnel 1 or the belt conveyor tunnel simulation test room; the fire monitoring data set needs to include multiple groups of first normal stage data groups, second smoldering stage data groups and third open fire stage data groups.

[0061] Step 303, train the LSTM-DAE feature network using the training set, wherein the attention mechanism is introduced to filter the feature information and adjust the weight of the feature information; Attention checks the similarity between the current Decoder output and the input Encoder, calculates the importance of each input encoding, normalizes all input Encoders to obtain an Attention vector, converts it to a probability through Softmax, and finally forms a background vector by multiplying it with the input Encoder;

[0062] Step 304, use the trained LSTM-DAE feature network to extract the feature information of the test set data as input, and use a two-layer fully connected neural network and a Softmax classifier to realize the judgment of the fire hazard level;

[0063] Specifically, the original data is input into the LSTM-DAE model to obtain the feature representation of the intermediate hidden layer, and the feature representation is taken as the input data of the next layer for training; then, different weights are given according to the importance of the hidden layer features by using the attention mechanism. The Softmax classifier is used to convert the input multi-dimensional vector into a multi-dimensional output value with a total sum of 1. Given a series of categories, the Softmax classifier can give the probability distribution of the division of a certain input into each category. According to the characteristics of the belt combustion, a two-layer fully connected layer network and a Softmax classifier are used to output the probabilities of the current input monitoring data being in the first normal stage, the second smoldering stage or the third open fire stage, and the output probability of the maximum danger level is selected as the danger level of the current belt conveyor.

[0064] Step 305, updating the parameters of the two-layer fully connected neural network by minimizing the error between the output result and the actual labeled fire danger level of the test set using the cross-entropy loss function;

[0065] Step four, using the trained fire danger level determination model to determine the fire danger level of the subsequent mine belt conveyor tunnel; the fire danger level includes the first normal stage, the second smoldering stage and the third open fire stage;

[0066] When it is determined to be the first normal stage, step four is executed cyclically;

[0067] When it is determined to be the second smoldering stage, step five is executed;

[0068] When it is determined to be the third open fire stage, step six is executed;

[0069] Step five, fire prevention and control in the second smoldering stage;

[0070] Step six, fire prevention and control in the third open fire stage.

[0071] It should be noted that the third open fire stage generally has smoke, and the temperature is greater than or equal to 45 DEG C, the carbon monoxide is greater than or equal to 100 ppm, and the hydrogen chloride is greater than or equal to 20 ppm.

[0072] In this embodiment, the fire prevention and control in the second smoldering stage and the fire prevention and control in the third open fire stage both use the roadway fire extinguishing device to eliminate fire hazards or fires, and the roadway fire extinguishing device comprises a water supply pipeline system arranged in the belt conveyor tunnel 1 and a plurality of groups of high-pressure fine water mist nozzles connected to the water supply pipeline system, and the plurality of groups of high-pressure fine water mist nozzles are evenly arranged along the belt conveyor tunnel 1, wherein the high-pressure fine water mist nozzle group must be arranged in the key fireproof areas such as the head, tail, transfer point, downwind side of the driving roller and air inlet shaft of the belt conveyor tunnel 1.

[0073] The high-pressure water mist nozzle group comprises a plurality of high-pressure water mist nozzles, each of which is connected with the water supply pipeline system through an electric ball valve, the electric ball valve is controlled by the fire monitoring unit 4 adjacent thereto, and the electric ball valve can be automatically controlled to be opened according to the detection result of the fire monitoring unit 4, or can be remotely controlled to be opened by the central server 3.

[0074] In this embodiment, the specific steps of fire prevention and control in the secondary smoldering stage in step five are as follows: when the fire monitoring data collected by a fire monitoring unit 4 determines that the current fire danger level is in the secondary smoldering stage, spray the high-pressure water mist nozzles corresponding to the fire monitoring unit 4 in half of the high-pressure water mist nozzle group to solve the initial fire hazard, and at the same time, the sound and light alarm 32 alarms to remind the surrounding staff of the existing fire hazard and reduce personnel casualties.

[0075] The specific steps of fire prevention and control in the tertiary open fire stage in step six are as follows: when the fire monitoring data collected by a fire monitoring unit 4 determines that the current fire danger level is in the tertiary open fire stage, spray all the high-pressure water mist nozzles corresponding to the fire monitoring unit 4 in the high-pressure water mist nozzle group to carry out fire extinguishing operation, and at the same time, the sound and light alarm 32 alarms to remind personnel to evacuate.

[0076] In this embodiment, before step four is executed, the fire danger level determination model needs to be optimized, and the specific optimization process is as follows: use AdaGrad, RMSProp and Adam optimization algorithms to optimize the danger determination model, analyze and compare the results, and select the best optimization algorithm.

[0077] The same steps are taken as the BP, SVM, PCA and DAE models to obtain the control results and perform performance evaluation. In order to better measure the performance of the fire danger level determination model, precision, recall and F1 score are introduced as performance indicators. The precision is the proportion of the number of samples accurately predicted by the model to the total number of samples obtained by the fire monitoring system in the belt tunnel. The recall is the number of samples whose danger level is identified by the model as the predicted danger level of the fire monitoring model in the belt tunnel, divided by the actual number of samples whose danger level is marked. The F1 score is the balance point of the precision and recall of the model. The larger the ratio of the above three performance indicators, the better the performance of the belt fire danger level determination model.

[0078] In step one, the fire monitoring data set of the mine belt conveyor roadway 1 is collected by a fire monitoring and early warning system, the fire monitoring and early warning system comprises a center server 3 and a plurality of fire monitoring units 4 in communication with the center server 3, and the plurality of fire monitoring units 4 are respectively arranged above the head of the belt conveyor 5, above the tail of the belt conveyor 5 and above the power transformation chamber 2; the fire monitoring unit 4 comprises an ambient gas sampling pipeline and a fire monitoring and early warning device connected with the ambient gas sampling pipeline;

[0079] The fire monitoring and early warning device comprises a shell 6, an air chamber 7 arranged in the shell 6, an air pump 8 for extracting ambient gas into the air chamber 7, a gas sensor module 12 arranged in the air chamber 7 and a control circuit board 9 arranged in the shell 6; the control circuit board 9 is integrated with a microcontroller 10 and a wireless communication module 11 connected with the microcontroller 10 and used for remote communication with the center server 3;

[0080] The gas sensor module 12 comprises a gas detection circuit board, and a smoke sensor 13, an oxygen sensor 14, a methane sensor 15, a hydrogen chloride sensor 16, a sulfur dioxide sensor 17, a hydrogen cyanide sensor 18, a carbon dioxide sensor 19, a carbon monoxide sensor 20 and a temperature sensor 21 arranged on the gas detection circuit board and wirelessly communicating with the gas detection circuit board;

[0081] In the embodiment, the temperature sensor 21 is a mine intrinsic safety type self-power generation wireless temperature sensor, which comprises an NST112 type temperature sensor and an EH301A type vibration energy harvesting module, and the EH301A type vibration energy harvesting module converts vibration energy and rotation energy of the belt roller into electric energy to supply power for the NST112 type temperature sensor;

[0082] In the embodiment, the smoke sensor 13 is an MP-2 type smoke sensor, which is a mine intrinsic safety type device, and when smoke appears in the running process of the belt conveyor, the smoke sensor 13 sends a signal to the center server 3 for data processing.

[0083] In the embodiment, the methane sensor 15 and the carbon dioxide sensor 19, the oxygen sensor 14 is a MIX8 type oxygen sensor, the hydrogen chloride sensor 16 is a ME3-HCl type hydrogen chloride sensor, the MSHia-DP-HC4-CO2 infrared carbon dioxide and methane two-in-one sensor, the sulfur dioxide sensor 17 is a ME3-SO2 type sulfur dioxide sensor, and the hydrogen cyanide sensor 18 is a HCN-50 type hydrogen cyanide sensor.

[0084] The environment gas sampling pipeline comprises a sampling ring pipe 22 and a plurality of suction nozzles 23 which are uniformly arranged along the circumference of the sampling ring pipe 22 and open downward, and mounting holes are formed in the sampling ring pipe 22 for mounting the suction nozzles 23;

[0085] The air inlet end of the suction pump 8 is communicated with the sampling ring pipe 22 through a suction pipe 24, the suction pipe 24 penetrates through the shell 6, the bottom of the air chamber 7 is provided with an air inlet pipe 25 connected with the air outlet end of the suction pump 8, and the top of the air chamber 7 is provided with an air outlet pipe 26, one end of the air outlet pipe 26 away from the air chamber 7 is communicated with the outside of the shell 6.

[0086] In the embodiment, the sampling ring pipe 22 is an elliptical ring pipe.

[0087] In the embodiment, before the sampling ring pipe 22 is mounted, an anchor rod is first punched into the coal bank, the anchor rod extends horizontally to the center of the roadway with a length of 0.5 m, and then the sampling ring pipe 22 is mounted at the end of the anchor rod away from the coal bank. Because the wind speed is small near the coal wall, it is not conducive to real-time detection of gas values, and if it is in the center of the roadway, a large amount of coal dust in the air will block the sampling ring pipe 22, so the sampling ring pipe 22 is slightly away from the coal bank, thereby prolonging the service life of the equipment.

[0088] In the embodiment, the air inlet pipe 25 and the air outlet pipe 26 are arranged on two opposite corners of the air chamber 7, so that the gas entering the air chamber 7 can fill the air chamber 7 from the bottom of the air chamber 7, and finally flow out from the air outlet pipe 26, thereby ensuring the concentration of the measured environment gas in the air chamber 7 when the gas sensor module detects, and ensuring the accuracy of the detection.

[0089] In the embodiment, the sensors in the gas sensor module are connected with the microcontroller 10, and the suction pump 8 is controlled by the microcontroller 10.

[0090] In the embodiment, the central server 3 is a computer, and the wireless communication module 11 can be an NB-IoT Internet of Things communication module with a model of M5311 or a LORA wireless ad hoc network module with a model of ASR6601.

[0091] It should be noted that by arranging the fire monitoring unit 4 above the head of the belt conveyor 5, above the tail of the belt conveyor 5 and above the power chamber 2 in the belt conveyor roadway 1 where fire is prone to occur, the monitoring object is clear, the pertinence is strong, and the monitoring effect is good;

[0092] By arranging the wireless communication module 11, the numerous belt combustion index gas detection data monitored in the belt conveyor roadway 1 are transmitted to the central server for subsequent data processing, and an alarm is given immediately when an abnormality is found, so that the fire monitoring and early warning of the belt conveyor roadway 1 can be performed in time and effectively;

[0093] According to the composition of various special gases generated by the belt combustion in the belt conveyor tunnel fire, the multiple sensors in the gas sensor module are arranged to monitor the state of the belt conveyor tunnel, which is targeted and accurate in detection and early warning results.

[0094] The gas sensor module is isolated from other devices by the gas chamber 7, which ensures the concentration of the measured environmental gas in the detection area of the gas sensor module, and the measurement result is more accurate. At the same time, the control circuit board 9 and components installed in the space between the shell 6 and the gas chamber 7 are not affected by the harmful environmental gas in the fire, which prolongs the service life of the components.

[0095] The sampling ring pipe 22 is arranged to increase the gas sampling range, which is beneficial to the large-scale fire monitoring and early warning of the belt conveyor tunnel, reduces the number of fire monitoring and early warning devices, and saves costs.

[0096] In this embodiment, the first sintered filter 27 and the water-resistant air-permeable membrane 28 are installed at one end of the air extraction pipe 24 communicating with the sampling ring pipe 22, and the second sintered filter 29 is installed at one end of the air outlet pipe 26 away from the gas chamber 7.

[0097] In this embodiment, the third sintered filter 30 is arranged on the air extraction nozzle 23.

[0098] In this embodiment, the first sintered filter 27, the second sintered filter 29 and the third sintered filter 30 are all stainless steel five-layer sintered filters.

[0099] It should be noted that the sintered filter and the water-resistant air-permeable membrane 28 are arranged to filter the gas extracted into the gas chamber 7, preventing impurities such as coal cinder, dust and water vapor from entering the gas chamber 7, thereby ensuring the accuracy of the gas sensor module detection.

[0100] In this embodiment, a communication antenna 31 is arranged on the outside of the shell 6, and the communication antenna 31 is connected with the wireless communication module 11.

[0101] In this embodiment, an audible and visual alarm 32 is arranged on the outside of the shell 6, and the audible and visual alarm 32 is connected with the microcontroller 10.

[0102] The above is only a preferred embodiment of the present application, and does not limit the present application. Any simple modification, change and equivalent structural change made according to the technical essence of the present application to the above embodiment are still within the protection scope of the technical solution of the present application.

Claims

1. A method for mine belt conveyor fire hazard grade determination and intelligent prevention and control, characterized in that, The method comprises the following steps: Step one, collecting the fire monitoring data set of the mine belt conveyor roadway: Collecting the fire smoke concentration in the mine belt conveyor roadway (1) and the temperature of the mine belt conveyor roadway (1) as the fire monitoring data set; wherein the fire smoke includes smoke, oxygen, methane, hydrogen chloride, sulfur dioxide, hydrogen cyanide, carbon dioxide and carbon monoxide; Step two, dimensionless processing of the fire monitoring data set: after data cleaning and noise value elimination of the fire monitoring data set, the maximum and minimum standardization method is used for dimensionless processing of the fire monitoring data set; Step three, training of the fire hazard level determination model; Step 301, establishing a fire hazard level determination model, which comprises an LSTM-DAE feature network, a two-layer fully connected neural network and a Softmax classifier; Step 302, dividing the dimensionless processed fire monitoring data set into a training set and a test set; Step 303, training the LSTM-DAE feature network using the training set, wherein the attention mechanism is introduced to filter the feature information and adjust the weight of the feature information; Step 304, using the trained LSTM-DAE feature network to extract the test set data feature information as input, using the two-layer fully connected neural network and the Softmax classifier to realize the determination of the fire hazard level; Step 305, by minimizing the error between the output result and the actual labeled fire hazard level of the test set, using the cross entropy loss function to update the two-layer fully connected neural network parameters; Step four, using the trained fire hazard level determination model to determine the fire hazard level of the subsequent mine belt conveyor roadway; the fire hazard level comprises a first normal stage, a second smoldering stage and a third open fire stage; When it is determined as the first normal stage, step four is executed cyclically; When it is determined as the second smoldering stage, step five is executed; When it is determined as the third open fire stage, step six is executed; Step five, fire prevention and control in the second smoldering stage; Step six, fire prevention and control in the third open fire stage.

2. The mine belt conveyor fire hazard grade determination and intelligent prevention and control method according to claim 1, characterized in that: In step one, the fire monitoring data set of the mine belt conveyor roadway (1) is collected by using a fire monitoring and early warning system, which comprises a center server (3) and a plurality of fire monitoring units (4) in communication with the center server (3), the plurality of fire monitoring units (4) are respectively arranged above the head of the belt conveyor (5), above the tail of the belt conveyor (5) and above the power chamber (2); the fire monitoring unit (4) comprises an environmental gas sampling pipeline and a fire monitoring and early warning device connected with the environmental gas sampling pipeline; The fire monitoring and early warning device comprises a shell (6), an air chamber (7) arranged in the shell (6), an air extraction pump (8) for extracting ambient air into the air chamber (7), a gas sensor module (12) arranged in the air chamber (7), and a control circuit board (9) arranged in the shell (6); the control circuit board (9) is integrated with a microcontroller (10) and a wireless communication module (11) connected with the microcontroller (10) and used for remote communication with a central server (3); The gas sensor module (12) comprises a gas detection circuit board, and a smoke sensor (13), an oxygen sensor (14), a methane sensor (15), a hydrogen chloride sensor (16), a sulfur dioxide sensor (17), a hydrogen cyanide sensor (18), a carbon dioxide sensor (19), a carbon monoxide sensor (20), and a temperature sensor (21) wirelessly communicating with the gas detection circuit board are all arranged on the gas detection circuit board; The ambient air sampling pipeline comprises a sampling ring pipe (22) and a plurality of air extraction nozzles (23) uniformly arranged along the circumference of the sampling ring pipe (22) and opening downward, and the sampling ring pipe (22) is provided with mounting holes for mounting the air extraction nozzles (23); The air inlet end of the air extraction pump (8) is communicated with the sampling ring pipe (22) through an air extraction pipe (24), the air extraction pipe (24) penetrates through the shell (6), the bottom of the air chamber (7) is provided with an air inlet pipe (25) connected with the air outlet end of the air extraction pump (8), and the top of the air chamber (7) is provided with an air outlet pipe (26), one end of the air outlet pipe (26) away from the air chamber (7) is communicated with the outside of the shell (6).

3. The method for judging and intelligently preventing and controlling fire hazard grade of mine belt conveyor according to claim 2, characterized in that: The end of the air extraction pipe (24) communicated with the sampling ring pipe (22) is mounted with a first sintered filter (27) and a water-proof and air-permeable film (28), and the end of the air outlet pipe (26) away from the air chamber (7) is mounted with a second sintered filter (29).

4. The method for judging and intelligently preventing and controlling fire hazard grade of mine belt conveyor according to claim 2, characterized in that: The air extraction nozzle (23) is provided with a third sintered filter (30).

5. The method for judging and intelligently preventing and controlling fire hazard grade of mine belt conveyor according to claim 2, characterized in that: The outer side of the shell (6) is provided with a communication antenna (31), and the communication antenna (31) is connected with the wireless communication module (11).

6. The method for judging and intelligently preventing and controlling fire hazard grade of mine belt conveyor according to claim 2, characterized in that: The outer side of the shell (6) is provided with an audible and visual alarm (32), and the audible and visual alarm (32) is connected with the microcontroller (10).

Images