A method for predicting tunnel fire self-extinguishing based on water curtain shielding effect

Abstract

The present invention discloses a method for predicting self-extinguishing of tunnel fire based on water curtain shielding effect, and relates to the field of fire safety technology. Compared with the previous single-sided closed tunnel fire prevention method, the present invention solves the problem of how to carry out the spread of smoke from a single-sided closed tunnel fire, and combine the shielding effect of a water curtain to suppress the development of the fire and finally achieve smoke control and fire extinguishing; by detecting the fire situation in the single-sided closed tunnel, linking and controlling related devices to form a water curtain, and separating the fire source and fire smoke through the shielding effect of the water curtain, it can effectively prevent the spread of fire smoke in the single-sided closed tunnel, suppress the development of the fire and finally achieve the effect of smoke control and fire extinguishing; and combined with theoretical analysis, simulate and analyze the effect of different spraying intensities on the self-extinguishing effect of water curtain smoke blocking, and build a time prediction model for the space to achieve self-extinguishing; propose a set of effective design schemes for water curtain smoke blocking and even self-extinguishing in single-sided closed tunnels, thereby providing a reference for tunnel fire protection design.

Description

Technical Field

[0001] The invention relates to the technical field of fire safety, and in particular to a tunnel fire self-extinguishing prediction method based on water curtain shielding effect. Background Art

[0002] Tunnels are important infrastructure in urban transportation systems. They facilitate water and land transportation, make full use of underground space, and relieve the pressure of land transportation. In recent years, the number of highway tunnels in my country has continued to rise. This growing trend has brought new challenges due to the various shapes and designs when building these tunnels, as well as the need to be compatible with local terrain. In the past few years, tunnels with closed or similar structures have gradually become tunnel warehouses, highway tunnels, tunnels under construction, subway stations, and corridors. People are increasingly concerned about the fire risk of this type of tunnel with one end closed. Compared with traditional tunnels with two open entrances, the smoke and heat released by the fire can only be discharged through the tunnel with one closed end, resulting in a large amount of heat and smoke gathering near the closed end, and the risk of tunnel structure damage is increased. In addition, the evacuation route of personnel in these tunnels is in the same direction as the smoke movement. Therefore, personnel are likely to be exposed to high temperatures and suffocating gases. Therefore, effective fire prevention, protection, and smoke control strategies are of great significance.

[0003] In recent years, many scholars have used the soft-closed smoke-proof method of water curtains to block smoke in confined spaces and control flames to a certain extent. Existing technicians have found through experiments in small-scale water curtain tunnels that water curtains not only achieve a certain effect of smoke prevention, but also suppress the development of flames and even achieve self-extinguishing of flames. Existing technicians have also found through a 1:10 scale model tunnel that the pressure of the water curtain, the number of rows arranged, and the heat release rate of the fire source all affect the smoke-blocking effect of the water curtain system. This effect decreases with the increase of the heat release rate of the fire source, and increases significantly with the increase of the water curtain pressure and the number of rows. Existing technicians have also conducted experimental verification through the LES solver-Fire FOAM and a 1 / 10 scale model tunnel. The study found that the water curtain delayed the time for smoke to flow to the entrance, providing additional time for evacuation. The water curtain has a good effect of controlling smoke and suppressing flames.

[0004] Therefore, how to carry out the smoke spread of a fire in a single-side closed tunnel and combine it with the shielding effect of a water curtain to suppress the development of the fire and ultimately achieve the purpose of smoke control and fire extinguishing has become an urgent problem that needs to be solved for the safe operation of the tunnel.

[0005] In order to solve the above problems, the present invention proposes a tunnel fire self-extinguishing prediction method based on the shielding effect of water curtain. Summary of the invention

[0006] The purpose of the present invention is to propose a tunnel fire self-extinguishing prediction method based on water curtain shielding effect to solve the problems raised in the background technology:

[0007] How to investigate the spread of smoke from a fire in a single-sided closed tunnel, and combine it with the shielding effect of a water curtain to suppress the development of the fire and ultimately achieve the effect of smoke control and fire extinguishing.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] A method for predicting tunnel fire self-extinguishing based on water curtain shielding effect, the prediction method is implemented based on a fire self-extinguishing system, the fire self-extinguishing system comprises: a fire detector, a fire alarm controller and a water curtain fire fighting system, the fire detector is respectively connected to the fire alarm controller and the water curtain fire fighting system in communication; the water curtain fire fighting system comprises a water supply system, a plurality of groups of longitudinally arranged water curtain nozzle groups arranged at reserved positions, sensors arranged between adjacent water curtain nozzle groups and a plurality of detection devices, the water supply system comprises a correspondingly connected A spray group, wherein the water curtain nozzle group includes a plurality of rows of high-pressure water curtain nozzles with adjustable spray angles arranged along the longitudinal section of the tunnel, wherein the high-pressure water curtain nozzles are arranged in a reserved area at the top of the tunnel, wherein the reserved area is preset with a plurality of rows of spray slots arranged in a staggered manner, wherein the distance between adjacent rows is 0.5 m, and the high-pressure water curtain nozzles are arranged corresponding to the spray slots; the sensor is connected to the fire alarm controller; the detection equipment is arranged longitudinally along the tunnel ceiling and is used to measure the tunnel temperature; the detection equipment is specifically a thermocouple, and the thermocouples are arranged longitudinally along the tunnel at a spacing of 0.5 m;

[0010] The prediction method comprises the following steps:

[0011] S1: When a fire occurs in a tunnel, the fire detector collects fire information in the tunnel;

[0012] S2: The fire alarm controller is turned on through the sensor, and the water curtain sprinkler group is turned on in conjunction to form a full-section separation water curtain on the right side of the fire source;

[0013] S3: After the water curtain is formed, the fire source and fire smoke are blocked in the water curtain separation section, the collected fire information is summarized by computer, and the self-extinguishing time is predicted by oxygen consumption method:

[0014] The oxygen consumption method predicts the self-extinguishing time t based on the self-extinguishing prediction formula:

[0015]

[0016] in, represents the initial volume fraction of oxygen in the space; represents the critical oxygen volume fraction; M represents the relative molecular mass of oxygen; V represents the spatial volume; ΔH c (O 2 ) represents activation energy, i.e. the heat consumed by unit mass of oxygen; V m represents the molar volume of gas; HRR represents the heat release rate.

[0017] Preferably, the fire information in S1 includes fire scale, temperature distribution, and smoke spread range.

[0018] Preferably, the pressure of the water curtain nozzles in the water curtain spray group opened in S2 is greater than 0.4 MPa.

[0019] Preferably, the sensor in S3 collects spatial information within the water curtain separation section after the water curtain is formed, and builds a three-dimensional model within the water curtain separation section based on the digital twin technology equipped with a graph convolutional network, and calculates the spatial volume V within the water curtain separation section.

[0020] Preferably, in S3, the corresponding critical oxygen volume fraction is generated according to the calculated spatial volume V in the water curtain separation section.

[0021] Preferably, in S3, the sensor also collects environmental status information of the fire area, and simulates it based on fluid mechanics, while combining the improved LSTM network and graph neural network to capture the time series characteristics, spatial correlation and dynamic evolution process of the fire, and calculates the oxygen volume fraction in the enclosed space. Make predictions;

[0022] The improved LSTM network consists of a forward LSTM and a reverse LSTM, which can be expressed as:

[0023] r ijk =LSTM(l ijk ),k∈[1,K]

[0024] Among them, l ijk The vector of the kth position point of the jth parameter of the i-th segment sequence; K is the total number of position points; r ijk The vector after LSTM encoding;

[0025] The attention mechanism is introduced as follows:

[0026] w ijk =tanh(ω w r ijk +b w )

[0027]

[0028] Among them, wijk is the position point after conversion; ω w 、w w and b w is a learnable parameter; is the attention weight matrix; T represents the transpose of the vector; W ijk It is the feature representation of the position point after weighting using the global attention mechanism;

[0029] Add the gating mechanism as follows:

[0030] Y=σ(μ*X+b)·X

[0031] Among them, σ is the sigmoid activation function; μ and b are learnable parameters; X is the input feature; Y is the new feature space;

[0032] At the same time, the parameters are optimized based on the butterfly optimization algorithm; the butterfly optimization algorithm is as follows:

[0033] The fragrance concentrations are as follows:

[0034] d=uv α

[0035] Where d is the concentration of the scent emitted by the butterfly; u is the perception form; v is the stimulus intensity; α is the power exponent;

[0036] Randomly generate butterfly positions representing weight values ​​within the weight allowable value, calculate and store the butterfly positions and fitness, where the fitness function is the inverse of the LSTM output error value;

[0037] When the butterfly moves based on a scent greater than its own, this phase is a global search, as follows:

[0038]

[0039] in, are the positions of the ath butterfly at the t+1th iteration and the tth iteration respectively; e is a random number between [0,1]; g t is the best butterfly in the tth iteration; d a is the fragrance concentration of the ath butterfly;

[0040] When a butterfly cannot sense the scent concentration of other butterflies, it sends a random move. This phase is the local search phase, as follows:

[0041]

[0042] in, and are the positions of the bth butterfly and the cth butterfly at the tth iteration respectively;

[0043] The global search and local search phases occur during the butterfly foraging process. Phase transitions are performed based on the transition probability p. The transition probability p is compared with the random number e to determine whether global search or local search occurs.

[0044] After reaching the maximum number of iterations, if the preset conditions are met, the iteration is terminated and the optimal value is output; otherwise, the iteration continues until the optimal value is obtained.

[0045] Compared with the prior art, the present invention provides a tunnel fire self-extinguishing prediction method based on the shielding effect of a water curtain, which has the following beneficial effects:

[0046] The present invention detects the fire situation in a single-sided closed tunnel, controls related devices to form a water curtain, and isolates the fire source and fire smoke through the shielding effect of the water curtain, which can effectively prevent the spread of fire smoke in the single-sided closed tunnel, inhibit the development of the fire and ultimately achieve the effect of smoke control and fire extinguishing; and combined with theoretical analysis, simulates and analyzes the effect of different spray intensities on the self-extinguishing effect of water curtain smoke blocking, and builds a time prediction model for space to achieve self-extinguishing; a set of effective design solutions for water curtain smoke blocking and even self-extinguishing in single-sided closed tunnels are proposed, thereby providing a reference for tunnel fire protection design. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a system block diagram mentioned in Embodiment 1 of the present invention;

[0048] Figure 2 This is a simulation schematic diagram of the fire extinguishing and smoke prevention system of a single-side closed tunnel mentioned in Example 1 of the present invention;

[0049] Figure 3 This is a schematic diagram of the arrangement of the detection equipment mentioned in Example 1 of the present invention;

[0050] Figure 4 This is a schematic diagram of the water curtain blocking the airflow mentioned in Example 1 of the present invention;

[0051] Figure 5 This is a schematic diagram of the fire source system mentioned in Example 2 of the present invention;

[0052] Figure 6 This is a schematic diagram of the 10MW fire source power in the water curtain open state mentioned in Example 2 of the present invention;

[0053] Figure 7 This is a schematic diagram of the ceiling temperature corresponding to the 10MW fire source power in the water curtain open state mentioned in Example 2 of the present invention;

[0054] Figure 8 This is a schematic diagram of flame and oxygen concentration distribution corresponding to a 10MW fire source power in the water curtain open state mentioned in Example 2 of the present invention;

[0055] Fig. 9 This is a schematic diagram of the 20MW fire source power in the water curtain open state mentioned in Example 2 of the present invention;

[0056] Fig.10 This is a schematic diagram of the ceiling temperature corresponding to the 20MW fire source power in the water curtain open state mentioned in Example 2 of the present invention;

[0057] Fig.11 This is a schematic diagram of flame and oxygen concentration distribution corresponding to a 20MW fire source power in the water curtain open state mentioned in Example 2 of the present invention;

[0058] Fig.12 This is a schematic diagram of the 30MW fire source power in the water curtain open state mentioned in Example 2 of the present invention;

[0059] Fig.13 This is a schematic diagram of the ceiling temperature corresponding to the 30MW fire source power in the water curtain open state mentioned in Example 2 of the present invention;

[0060] Fig.14 This is a schematic diagram of the flame and oxygen concentration distribution corresponding to the 30MW fire source power in the water curtain open state mentioned in Example 2 of the present invention.

[0061] Meaning of the symbols in the figure:

[0062] 1. Fire detector; 2. Fire alarm controller; 3. Water curtain fire fighting system; 31. Water supply system; 32. Water curtain sprinkler group; 33. Sensor; 34. Detection equipment. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0064] The present invention detects the fire situation in a single-sided closed tunnel, controls related devices to form a water curtain, and isolates the fire source and fire smoke through the shielding effect of the water curtain, which can effectively prevent the spread of fire smoke in the single-sided closed tunnel, inhibit the development of the fire and ultimately achieve the effect of smoke control and fire extinguishing; and combined with theoretical analysis, simulates and analyzes the effect of different spraying intensities on the self-extinguishing effect of water curtain smoke blocking, and builds a time prediction model for space to achieve self-extinguishing; proposes a set of effective design solutions for water curtain smoke blocking and even self-extinguishing in single-sided closed tunnels, thereby providing a reference for tunnel fire protection design. Specifically including the following contents.

[0065] Embodiment 1:

[0066] See also Figure 1-4The present invention provides a tunnel fire self-extinguishing prediction method based on the water curtain shielding effect, which is implemented based on the fire self-extinguishing system. The construction of the fire self-extinguishing system can refer to Figure 1-3 , specifically including a fire detector 1, a fire alarm controller 2 and a water curtain fire fighting system 3. The fire detector 1 is respectively connected to the fire alarm controller 2 and the water curtain fire fighting system 3 in communication; the water curtain fire fighting system 3 includes a water supply system 31 and a plurality of groups of longitudinally arranged water curtain nozzle groups 32 set in reserved positions. The water curtain nozzle group 32 can be arranged in the longitudinal position of the single-sided closed tunnel. The position needs to reserve a spray trough in advance and should be set more than 20m away from the area where the fire is most likely to occur. Adjacent water curtain nozzles in the water curtain nozzle group 32 are staggered, and the distance between adjacent rows is 0.5m. The water supply system 31 includes a spray group corresponding to each group of water curtain nozzle groups 32. The water curtain nozzle group 32 includes several rows of water curtain nozzles arranged along the longitudinal section of the single-sided closed tunnel. The spray angle of the water curtain nozzle is adjustable, and the water curtain nozzle is set in the reserved area at the top of the single-sided closed tunnel. The water curtain fire fighting system 3 also includes a sensor 33 arranged between adjacent water curtain nozzle groups 32. The sensor 33 is connected to the fire alarm controller 2. Refer to Figure 4 The water curtain fire fighting system 3 also includes a plurality of detection devices 34, which are arranged longitudinally along the ceiling of the single-sided closed tunnel and are used to measure the temperature of the single-sided closed tunnel; the detection devices 34 are specifically thermocouples, and the thermocouples are arranged longitudinally along the single-sided closed tunnel with a spacing of 0.5m.

[0067] The prediction method includes the following steps:

[0068] S1: When a fire occurs in a single-side closed tunnel, the fire detector 1 collects and obtains the fire information in the single-side closed tunnel; the details are as follows:

[0069] The fire detector 1 collects and obtains the fire information in the single-side closed tunnel, including the fire scale, temperature distribution and smoke spreading range.

[0070] S2: The fire alarm controller 2 is turned on through the sensor 33, and the water curtain nozzle group 32 is turned on in conjunction to form a full-section separation water curtain on the right side of the fire source; the details are as follows:

[0071] The sensor 33 turns on the fire alarm controller 2 and the water curtain nozzle group 32 based on the fire information detected and collected by the fire detector 1, forming a full-section separation water curtain on the right side of the fire source. Figure 4 Due to the effect of the water curtain, a certain pressure is generated, causing the fresh air to return, and the low-oxygen gas is blocked from flowing back to the fire source area.

[0072] S3: After the water curtain is formed, the fire source and the fire smoke are blocked in the water curtain separation section, the collected fire information is summarized by computer, and the self-extinguishing time is predicted by oxygen consumption method: the details are as follows:

[0073] After the water curtain is formed, the fire source and fire smoke are blocked in the water curtain separation section, preventing the smoke from spreading further. The water curtain is used to lower the section temperature, dissolve some toxic and harmful gases, and block the fresh gas from the outside from flowing to the fire source, thereby reducing the oxygen concentration in the left space.

[0074] In this process, a three-dimensional model of the water curtain separation section is built based on the digital twin technology equipped with a graph convolutional network, and the spatial volume V in the water curtain separation section is calculated. The corresponding critical oxygen volume fraction is also generated based on the calculated spatial volume V in the water curtain separation section. The environmental status information of the fire area is also collected and simulated based on fluid mechanics. At the same time, the improved LSTM network and graph neural network are combined to capture the time series characteristics, spatial correlation and dynamic evolution process of the fire, and the initial volume fraction of oxygen in the enclosed space is calculated. Make predictions.

[0075] Assuming that the fire source power is constant, the heat released by consuming 1kg of oxygen is 13.1MJ, and the volume of the space on the left side of the water curtain is constant, the initial smoke gas volume fraction is 21%, and the oxygen volume fraction during self-extinguishing is 12%. The oxygen consumption method predicts the self-extinguishing time t based on the self-extinguishing prediction formula according to the simulated predicted value:

[0076]

[0077] in, represents the initial volume fraction of oxygen in the space; represents the critical oxygen volume fraction; M represents the relative molecular mass of oxygen; V represents the spatial volume; ΔH c (O 2 ) represents activation energy, i.e. the heat consumed by unit mass of oxygen; V m represents the molar volume of gas; HRR represents the heat release rate.

[0078] In the simulation scene, the volume of the space on the left side of the water curtain is 24.2m*8m*4.2m, or 813.12m 3 The initial volume fraction of oxygen in the space is 21%, and the volume fraction of oxygen is 12% when it self-extinguishes. The heat released by the consumption of oxygen per unit mass is an approximately constant value, that is, 13.1 kJ / g. The relative molecular mass of oxygen is 32*10- 3 kg / mol, Vm=22.4×10-3L / mol.

[0079] Embodiment 2:

[0080] Please refer to Figure 5The simulated fire source system simulates a fire situation in a single-enclosed single-side enclosed tunnel. This embodiment uses heptane fire with a self-ignition temperature of 215°C, and the automatic sprinkler uses the FDS default Generic storage Spray.

[0081] In the actual simulation process, the critical oxygen concentration will change to a certain extent with the change of HRR, that is, the increase in ambient temperature and external heating will lead to a decrease in the critical oxygen concentration. The air around the fire is constantly heated, and a larger HRR will generate more heat, causing the temperature around the fire to rise. As the smoke continues to heat up, the amount of oxygen that can be consumed in the area around the fire source will also increase. Therefore, the fire source can continuously obtain oxygen from the surrounding air to maintain its own combustion. Therefore, there are certain differences in the prediction models of different fire source powers in this embodiment. The results of the prediction based on the prediction method of Example 1 can be specifically referred to Table 1:

[0082] Table 1 Prediction results of fire self-extinguishing at different critical oxygen concentrations

[0083]

[0084] Also refer to Figure 6-14 ,It can be seen from the simulation results that the water curtain has a good smoke-blocking effect in the scenario with one end closed, and effectively blocks the diffusion of high-temperature smoke. It is worth noting that when the fire source power is 10MW, the smoke becomes lighter after 120s, when the fire source power is 20MW, the smoke becomes lighter after 65s, and when the fire source power is 30MW, the smoke becomes lighter after 50s. This is because the flame is self-extinguished due to lack of oxygen. In addition, under the barrier of the water curtain, when the fire source power is 10MW, the ceiling temperature of the space on the left is controlled below 1000℃. The average ceiling temperature of the area on the right side of the water curtain is only 26.95℃, which is slightly higher than the ambient temperature by 6.95℃. When the fire source power is 20MW, the ceiling temperature of the space on the left is controlled below 1000℃. The average ceiling temperature of the area on the right side of the water curtain is only 32.75℃, which is slightly higher than the ambient temperature by 12.75℃. When the fire source power is 30MW, the ceiling temperature of the left space is controlled below 1100℃. The average ceiling temperature of the area on the right side of the water curtain is only 38.68℃, slightly higher than the ambient temperature of 18.68℃. Obviously, the water curtain can effectively reduce the temperature in the single-side closed tunnel, which can provide a certain guarantee for the safety of the single-side closed tunnel structure.

[0085] In summary, for flames with low fire power, the critical oxygen concentration can be appropriately increased to improve the prediction accuracy, but it should be kept between 10% and 15%.

[0086] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for predicting tunnel fire self-extinguishing based on water curtain shielding effect, characterized in that: The prediction method is implemented based on a fire self-extinguishing system, the fire self-extinguishing system comprising: a fire detector (1), a fire alarm controller (2) and a water curtain fire fighting system (3), the fire detector (1) being respectively connected to the fire alarm controller (2) and the water curtain fire fighting system (3); the water curtain fire fighting system (3) comprising a water supply system (31), a plurality of groups of longitudinally arranged water curtain nozzle groups (32) arranged at reserved positions, sensors (33) arranged between adjacent water curtain nozzle groups (32) and a plurality of detection devices (34), the water supply system (31) comprising a sensor (33) connected to each group of water curtain nozzle groups (32) The water curtain nozzle group (32) includes a plurality of rows of high-pressure water curtain nozzles with adjustable spray angles arranged along the longitudinal section of the tunnel, the high-pressure water curtain nozzles are arranged in a reserved area at the top of the tunnel, the reserved area is preset with a plurality of rows of spray slots arranged in a staggered manner, the distance between adjacent rows is 0.5 m, and the high-pressure water curtain nozzles are arranged corresponding to the spray slots; the sensor (33) is connected to the fire alarm controller (2); the detection device (34) is arranged along the longitudinal direction of the tunnel ceiling and is used to measure the tunnel temperature; the detection device (34) is specifically a thermocouple, and the thermocouples are arranged along the longitudinal direction of the tunnel with a spacing of 0.5 m; The prediction method comprises the following steps: S1: When a fire occurs in a tunnel, the fire detector (1) collects and obtains fire information in the tunnel; S2: The fire alarm controller (2) is turned on through the sensor (33), and the water curtain nozzle group (32) is turned on in conjunction to form a full-section separation water curtain on the right side of the fire source; S3: After the water curtain is formed, the fire source and fire smoke are blocked in the water curtain separation section, the collected fire information is summarized by computer, and the self-extinguishing time is predicted by oxygen consumption method: The oxygen consumption method predicts the self-extinguishing time t based on the self-extinguishing prediction formula: in, represents the initial volume fraction of oxygen in the space; represents the critical oxygen volume fraction; M represents the relative molecular mass of oxygen; V represents the spatial volume; ΔH c (O2) represents activation energy, i.e. the heat consumed by unit mass of oxygen; V m represents the molar volume of gas; HRR represents the heat release rate.

2. The method for predicting tunnel fire self-extinguishing based on water curtain shielding effect according to claim 1 is characterized in that: The fire information in S1 includes fire scale, temperature distribution, and smoke spread range.

3. The method for predicting tunnel fire self-extinguishing based on water curtain shielding effect according to claim 1 is characterized in that: The pressure of the water curtain nozzles in the water curtain nozzle group (32) opened in S2 is greater than 0.4 MPa.

4. The method for predicting tunnel fire self-extinguishing based on water curtain shielding effect according to claim 1 is characterized in that: The sensor (33) in S3 collects spatial information within the water curtain separation section after the water curtain is formed, and builds a three-dimensional model within the water curtain separation section based on the digital twin technology equipped with a graph convolutional network, and calculates the spatial volume V within the water curtain separation section.

5. The method for predicting tunnel fire self-extinguishing based on water curtain shielding effect according to claim 4 is characterized in that: In S3, the corresponding critical oxygen volume fraction is generated according to the calculated spatial volume V in the water curtain separation section.

6. The method for predicting tunnel fire self-extinguishing based on water curtain shielding effect according to claim 2 is characterized in that: In S3, the sensor (33) also collects environmental status information of the fire area and simulates it based on fluid mechanics. At the same time, it combines the improved LSTM network and graph neural network to capture the time series characteristics, spatial correlation and dynamic evolution process of the fire, and calculates the initial volume fraction of oxygen in the closed space. Make predictions; The improved LSTM network consists of a forward LSTM and a reverse LSTM, which can be expressed as: r ijk =LSTM(l ijk ),k∈[1,K] Among them, l ijk The vector of the kth position point of the jth parameter of the i-th segment sequence; K is the total number of position points; r ijk The vector after LSTM encoding; The attention mechanism is introduced as follows: w ijk =tanh(ω w r ijk +b w ) Among them, w ijk is the position point after conversion; ω w 、w w and b w is a learnable parameter; is the attention weight matrix; T represents the transpose of the vector; W ijk It is the feature representation of the position point after weighting using the global attention mechanism; Add the gating mechanism as follows: Y=σ(μ*X+b)·X Among them, σ is the sigmoid activation function; μ and b are learnable parameters; X is the input feature; Y is the new feature space; At the same time, the parameters are optimized based on the butterfly optimization algorithm; the butterfly optimization algorithm is as follows: The fragrance concentrations are as follows: d=uv α Where d is the concentration of the scent emitted by the butterfly; u is the perception form; v is the stimulus intensity; α is the power exponent; Randomly generate butterfly positions representing weight values ​​within the weight allowable value, calculate and store the butterfly positions and fitness, where the fitness function is the inverse of the LSTM output error value; When the butterfly moves based on a scent greater than its own, this phase is a global search, as follows: in, are the positions of the ath butterfly at the t+1th iteration and the tth iteration respectively; e is a random number between [0,1]; g t is the best butterfly in the tth iteration; d a is the fragrance concentration of the ath butterfly; When a butterfly cannot sense the scent concentration of other butterflies, it sends a random move. This phase is the local search phase, as follows: in, and are the positions of the bth butterfly and the cth butterfly at the tth iteration respectively; The global search and local search phases occur during the butterfly foraging process. Phase transitions are performed based on the transition probability p. The transition probability p is compared with the random number e to determine whether global search or local search occurs. After reaching the maximum number of iterations, if the preset conditions are met, the iteration is terminated and the optimal value is output; otherwise, the iteration continues until the optimal value is obtained.

Images