A method for identifying and predicting parameters during the growth phase of highway tunnel fires

Abstract

This invention relates to a method for identifying and predicting parameters during the growth phase of a highway tunnel fire, comprising the following steps: S1, selecting a fire source model for the tunnel fire growth phase; S2, determining the parameters of the fire source model for the tunnel fire growth phase; S3, establishing and setting the operating conditions of an FDS numerical simulation model for the tunnel fire growth phase; S4, establishing a database for observing the tunnel fire growth phase; S5, labeling and normalizing the tunnel fire growth phase data; S6, establishing an LSTM neural network model for identifying and predicting parameters during the tunnel fire growth phase; S7, training the LSTM neural network model for identifying and predicting parameters during the tunnel fire growth phase; and S8, analyzing the performance of the parameter identification and prediction methods for the tunnel fire growth phase. This method addresses the problem that existing tunnel fire prediction methods do not fully analyze and utilize the large amount of data generated by various monitoring devices within the tunnel under fire conditions, cannot promptly identify fire source information and predict its growth trend, and cannot provide decision-making services for personnel safety evacuation and fire rescue.

Description

Technical Field

[0001] This invention belongs to the field of tunnel fire safety technology, and relates to a method for identifying and predicting parameters during the growth phase of highway tunnel fires. Background Technology

[0002] Highway tunnel fires are extremely dangerous, easily causing traffic paralysis, disruption of commerce and logistics, and even mass casualties, leading to severe social repercussions. Current fire prevention methods in highway tunnels mainly include remote monitoring, alarm systems, ventilation and smoke extraction, and firefighters entering the tunnel to extinguish the fire. However, these methods do not adequately consider the fire's growth phase, focusing primarily on the stable development stage. The growth phase of a highway tunnel fire, considered a "golden period" for evacuation and fire rescue, has not received sufficient attention.

[0003] Due to the relatively enclosed structure of tunnels and the limitations of fire monitoring equipment within them, the information available to the outside world during a fire is extremely limited. It is difficult for those inside the tunnel and firefighters to obtain timely information about the real-time fire situation and its development, leading to escape and rescue actions largely relying on experiential judgments based on visual perception. However, current prediction methods mostly require on-site information as input data, but the high temperatures and dense smoke inside tunnels prevent firefighters and equipment from reaching the fire source to obtain information. Furthermore, existing technologies cannot directly measure the heat release rate in actual fire scenarios. Simultaneously, various monitoring devices within tunnels generate a large amount of data during fire conditions, such as temperature, concentrations of various gas components, and wind speed; however, the analysis and utilization of this data are often limited. There is an urgent need to find a method for efficiently identifying, and even predicting, the growth phase of highway tunnel fires using available data, to provide decision-making support for safe evacuation and fire rescue. Summary of the Invention

[0004] In view of this, in order to solve the problem that existing tunnel fire prediction methods do not fully analyze and utilize the large amount of data generated by various monitoring devices in the tunnel under fire conditions, and cannot timely identify fire source information and predict its growth trend, this invention provides a method for identifying and predicting parameters of the growth period of highway tunnel fires, providing decision-making services for personnel safety evacuation and fire rescue.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for identifying and predicting parameters during the growth phase of a highway tunnel fire includes the following steps:

[0007] S1. Selection of ignition source model during the growth phase of a tunnel fire

[0008] Based on the heat release rate growth models for tunnel fires established by scholars at home and abroad, the quadratic growth model, which can better characterize the growth pattern of the heat release rate of fire sources caused by vehicles and goods in tunnels and is often used in computer simulations for establishing unsteady fire sources due to its simple parameter adjustment, is selected as the basic model to describe the change of the heat release rate of the fire source over time. The relationship is as follows:

[0009] Q = αt 2

[0010] In the formula: Q is the heat release rate (kW); t is the time required to reach that heat release rate (s); α is the fire growth coefficient (kW / s). 2 ).

[0011] S2. Determination of ignition source model parameters during the growth phase of a tunnel fire.

[0012] Based on the analysis of full-scale tunnel fire test data from both domestic and international sources, the maximum heat release rate (HRR) for different vehicle types (cars, buses, trucks, tankers) and vehicle-related materials (gasoline, diesel, and automotive polymers) was obtained. max Fire growth coefficient α, time t required to reach maximum heat release rate max The range of values ​​for is used to further verify the correctness of the fire source model during the growth period of tunnel fires.

[0013] S3. Establishment and Working Condition Setting of FDS Numerical Simulation Model during the Growth Phase of Tunnel Fire

[0014] A numerical simulation model of FDS during the growth period of tunnel fires was established. The monitoring locations and densities of temperature and wind speed sensors were set, and various simulation conditions were set, taking into account different fire source locations, longitudinal wind speeds, maximum heat release rates and their growth coefficients.

[0015] S4. Establishment of a database for observing the growth phase of tunnel fires.

[0016] The calculation results were obtained through multi-condition numerical simulation, and the fire growth coefficient and HRR of the tunnel fire growth stage were extracted from the calculation results of each condition. max Data samples of HRR, time, sensor location, fire source location, temperature, and longitudinal wind speed were collected and stored in the same time series to establish an observation database for the growth stage of tunnel fires.

[0017] S5. Data labeling and normalization during the tunnel fire growth period.

[0018] Extract data labels from the database: HRR, fire growth coefficient, fire source location, and longitudinal wind speed, and then label and normalize the data.

[0019] S6. Establishment of an LSTM neural network model for parameter identification and prediction during the tunnel ignition source growth period.

[0020] An LSTM neural network model was built, a regression prediction model was established with HRR as the output variable, and a multi-label classification model was established with fire growth coefficient, fire source location, and longitudinal wind speed as output variables.

[0021] S7. Training of an LSTM neural network model for parameter identification and prediction during the tunnel ignition source growth period.

[0022] The model is trained on data in a specific training environment, and the training effect of the LSTM neural model is evaluated by using the loss function MSE. The parameters in the network are continuously updated by minimizing the loss function to optimize the model.

[0023] S8. Parameter Identification and Predictive Performance Analysis of Tunnel Ignition Growth Period

[0024] The performance of the identification and prediction of parameters during the fire source growth period was analyzed using test set data, including generalization, timeliness, rationality and practical applicability.

[0025] Furthermore, in step S3, an FDS numerical simulation model for the tunnel fire growth period is established. The tunnel model cross section is based on the standard inner contour of a two-lane tunnel of a highway or first-class road, and the line type is selected as a straight line type with a length of 100m.

[0026] The sensor monitoring locations and densities were set, and 11 ceiling temperature monitoring devices were set up at 5m intervals within a 25m range upstream and downstream of the fire source. This can be used to simulate the sensor arrangement density at 5m and 10m intervals. Two wind speed monitoring devices were set up 10 meters upstream and downstream of the fire source.

[0027] Further, in step S5, data labels are extracted from the database: HRR, fire growth coefficient, fire source location, and longitudinal wind speed. Each data table is divided into multiple data samples according to a time length of 30 seconds. The Min-Max normalization method is used to standardize the data, removing differences in units and orders of magnitude between different types of data, and transforming them to the range of [0, 1]. The normalization formula is:

[0028]

[0029] In the formula: x i Initialize the data of type i; x i * represents the normalized value; max x i The maximum value of data of type i; minx i It represents the minimum value of data of type i.

[0030] Furthermore, in step S7, the network training optimizer is Adam (Adaptive Moment Estimation), the initial learning rate is set to 0.005, and the initial number of hidden layer neurons is set to 32. MSE refers to the expected value of the squared difference between the predicted parameter value and the true parameter value, as shown in the following formula. The smaller the MSE value, the higher the prediction accuracy of the neural network.

[0031]

[0032] In the formula: y i Let y' be the initial value for the i-th data; i is the predicted value obtained by the model; n is the number of data points.

[0033] Furthermore, the prediction accuracy in step S8 is calculated as follows:

[0034]

[0035] In the formula: R yc To improve prediction accuracy.

[0036] The beneficial effects of this invention are as follows:

[0037] The method for identifying and predicting parameters during the growth phase of highway tunnel fires disclosed in this invention has good reliability and feasibility. It can be used to optimize the arrangement of sensors (distributed fiber optic linear heat detectors, fiber optic grating linear heat detectors, etc.); it can be used for early fire control and extinguishing during the fire development stage based on the fire source location (it can be combined with intelligent fire extinguishers); and it can be used for early fire identification, personnel evacuation warnings and guidance based on the fire source location, fire scale, and longitudinal wind speed, providing real-time disaster alerts and evacuation decision support for drivers and passengers. Given the limited data acquisition channels for the actual growth phase of tunnel fires, this method can promptly identify fire source information and predict its growth trend, improving disaster prevention and mitigation capabilities.

[0038] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0039] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0040] Figure 1 This is a schematic diagram of the process for identifying and predicting parameters during the growth phase of a highway tunnel fire, as described in this invention.

[0041] Figure 2 These are actual test images of combustion tests on different vehicle-related materials according to the present invention;

[0042] Figure 3 These are actual test images of fire combustion tests on different vehicle types according to the present invention;

[0043] Figure 4 This is a schematic diagram of the arrangement of temperature sensors and wind speed sensors inside a highway tunnel according to the present invention.

[0044] Figure 5 This is a schematic diagram of the tunnel fire growth stage observation database of the present invention;

[0045] Figure 6 (a) is a schematic diagram of the regression prediction model in the LSTM neural network model of this invention;

[0046] Figure 6 (b) is a schematic diagram of the multi-label classification model in the LSTM neural network model of the present invention;

[0047] Figure 7 (a) is a graph showing the analysis results of the longitudinal wind speed recognition accuracy of the training and test datasets in the LSTM neural network model when the sensor spacing is 10m.

[0048] Figure 7 (b) is a graph showing the accuracy analysis results of fire source location coordinate identification in the training and test datasets of the LSTM neural network model when the sensor spacing is 10m.

[0049] Figure 7 (c) is a graph showing the results of the analysis of the accuracy of fire growth coefficient identification in the training and test datasets of the LSTM neural network model when the sensor spacing is 10m.

[0050] Figure 7 (d) is a graph showing the average recognition accuracy of each key parameter in the training and testing datasets of the LSTM neural network model when the sensor spacing is 10m.

[0051] Figure 8 (a) is a graph showing the analysis results of the longitudinal wind speed recognition accuracy of the training and test datasets in the LSTM neural network model under different sensor densities in this invention;

[0052] Figure 8 (b) is a graph showing the analysis results of the accuracy of fire source location coordinate recognition in the training and test datasets of the LSTM neural network model under different sensor densities in this invention;

[0053] Figure 8(c) is a graph showing the results of the analysis of the accuracy of fire growth coefficient identification on the training and test datasets of the LSTM neural network model under different sensor densities in this invention.

[0054] Figure 8 (d) is a graph showing the average recognition accuracy of each key parameter in the training and testing datasets of the LSTM neural network model under different sensor densities in this invention.

[0055] Figure 9 (a) is a graph showing the predicted HRR growth of car combustion when the present invention is applied to a full-size highway tunnel fire test;

[0056] Figure 9 (b) is a graph showing the predicted HRR growth of a bus fire when the present invention was applied to a full-size highway tunnel fire test. Detailed Implementation

[0057] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. In the absence of conflict, the following embodiments and features can be combined with each other. The accompanying drawings are for illustrative purposes only, representing schematic diagrams only, not actual physical images, and should not be construed as limiting the present invention. To better illustrate the embodiments of the present invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable that those skilled in the art may omit certain well-known structures and their descriptions in the drawings.

[0058] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0059] like Figure 1 The method for identifying and predicting parameters during the growth phase of a highway tunnel fire, as shown, includes the following steps:

[0060] S1. Selection of ignition source model during the growth phase of a tunnel fire

[0061] Based on the heat release rate growth models for tunnel fires established by scholars at home and abroad, the quadratic growth model, which can better characterize the growth pattern of the heat release rate of fire sources caused by vehicles and goods in tunnels and is often used in computer simulations for establishing unsteady fire sources due to its simple parameter adjustment, is selected as the basic model to describe the change of the heat release rate of the fire source over time. The relationship is as follows:

[0062] Q = αt 2

[0063] In the formula: Q is the heat release rate (kW); t is the time required to reach that heat release rate (s); α is the fire growth coefficient (kW / s). 2 );

[0064] In actual tunnel fire scenarios, it is extremely difficult to describe the change of the heat release rate of the fire source over time. Most of them use three growth models, namely linear growth, square growth and exponential growth, to quantitatively describe the growth and decay patterns of fire. Among them, the square growth model can better characterize the growth pattern of the heat release rate of the fire source when the fire is caused by vehicles and goods in the tunnel. Moreover, due to its simple parameter adjustment, it is often used to establish unsteady fire sources in computer simulation. Therefore, the square growth model is selected as the basic model for the growth stage of tunnel fire.

[0065] S2. Determination of ignition source model parameters during the growth phase of a tunnel fire.

[0066] To realistically depict the growth of ignition sources in tunnel fire scenarios, measured data from fire tests on different vehicle types (cars, buses, and trucks) and combustion tests on vehicle-related materials (gasoline, diesel, and other oil pool fires, as well as automotive polymers) in full-scale tunnels both domestically and internationally were collected, organized, and analyzed. The maximum heat release rate (HRR) was extracted and analyzed. max and the time t required to reach the maximum heat release rate max Relationships, such as Figure 2 , Figure 3 As shown. Figure 2 Statistical images of combustion tests on materials related to different vehicles; Figure 3 These are actual test images from fire combustion tests on different vehicle types. The results show that HRR... max With t max A non-one-to-one correspondence cannot be expressed in terms of growth time t. max The size of the HRR can be used to predict the HRR that can be achieved during the combustion process. max The key to establishing a fire source model for the growth phase of a tunnel fire lies in determining the rate and peak of the heat release rate from the fire source, i.e., the fire growth coefficient α and the peak heat release rate Q in the squared growth model. max .

[0067] Note: The peak heat release rate (Q) of the combustible material under a certain combustion condition. max This is called the maximum heat release rate (HRR). max Q max With HRR max The difference is that Q max HRR is a state quantity at a certain moment during the combustion process, representing the heat released by the substance through combustion per unit time. max This indicates the maximum heat release capacity of the combustible material under a certain combustion condition, and the two are equal in value.

[0068] Based on the squared growth model, to obtain the growth coefficient α and the peak value of the maximum heat release rate Q during the growth period of a tunnel fire. max Relationship Further analysis based on measured data from full-scale tunnel fire tests both domestically and internationally shows that the HRR (Heat Rate) of different types of vehicles burning in tunnel fires varies. max Value range: 2-10MW for cars, 10MW-50MW for buses, 50MW-200MW for trucks, and 200MW-300MW for tankers; Value range of combustion growth coefficient α for different types of vehicles in tunnel fires: 0.003kW / s for cars. 2 -0.013kW / s 2 Bus 0.05kW / s 2 -0.15kW / s 2 0.2kW / s for heavy-duty trucks 2 -0.3kW / s 2 Tanker truck 0.4kW / s 2 -0.6kW / s 2 .

[0069] Therefore, the parameter values ​​of the squared growth model for combustion of different vehicle types during the growth period of tunnel fires were determined, as shown in Table 1.

[0070] Table 1. Parameter values ​​for the quadratic growth model of combustion of different vehicle types in tunnel fires.

[0071]

[0072]

[0073] S3. Establishment and Working Condition Setting of FDS Numerical Simulation Model during the Growth Phase of Tunnel Fire

[0074] A numerical simulation model of FDS for the growth period of tunnel fires was established. The tunnel model cross section was based on the standard inner contour of a two-lane tunnel of a highway or first-class road. The line type was selected as the most common straight line type for tunnels, with a length of 100m.

[0075] The locations and densities of the sensors were set. Eleven ceiling temperature monitoring devices (temperature sensors) were installed at 5m intervals within a 25m range upstream and downstream of the fire source. This can be used to simulate sensor placement densities at 5m and 10m intervals. Two wind speed monitoring devices (wind speed sensors) were installed 10 meters upstream and downstream of the fire source. The device layout is as follows. Figure 4 As shown.

[0076] The fire source parameters were set based on the square growth model, taking into account different fire source locations, longitudinal wind speeds, maximum heat release rates and their growth coefficients, and various simulation conditions (240 conditions) were set, as shown in Table 2.

[0077] Table 2 Numerical Simulation Operating Conditions Table

[0078]

[0079] S4. Establishment of a database for observing the growth phase of tunnel fires.

[0080] The calculation results were obtained through multi-condition numerical simulation, and the fire growth coefficient and HRR of the tunnel fire growth stage were extracted from the calculation results of each condition. max Data samples including HRR, time, sensor location, fire source location, temperature, and longitudinal wind speed are organized and stored in the same time series to establish a database for observing the growth stages of tunnel fires. Figure 5 As shown.

[0081] S5. Data labeling and normalization during the tunnel fire growth period.

[0082] Data labels were extracted from the database: HRR, fire growth factor, fire source location, and longitudinal wind speed. Each data table was divided into multiple data samples with a time length of 30 seconds, ultimately forming a labeled sample set containing 7030 data points.

[0083] The data is standardized using the Min-Max normalization method (as shown in the following formula) to remove the differences in units and orders of magnitude between different types of data and transform them to the range of [0, 1].

[0084]

[0085] In the formula: x i Initialize the data of type i; x i * represents the normalized value; max x i The maximum value of data of type i; minx i It represents the minimum value of data of type i.

[0086] All labeled samples were divided into a training dataset and a test dataset, with sizes of 80% and 20%, respectively. That is, the training dataset of 5590 data samples was used to train the model, and the test dataset of 1440 data samples was used to evaluate the quality of the trained model.

[0087] S6. Establishment of an LSTM neural network model for parameter identification and prediction during the tunnel ignition source growth period.

[0088] The structure of the LSTM neural network model is as follows: Figure 6 (a) and Figure 6 As shown in (b), a regression prediction model is established with HRR as the output variable, and a multi-label classification model is established with fire growth coefficient, fire source location, and longitudinal wind speed as output variables. That is, the regression prediction model has 3 input neurons and 1 output neuron; the multi-label classification model has 3 input neurons and 3 output neurons.

[0089] S7. Training of an LSTM neural network model for parameter identification and prediction during the tunnel ignition source growth period.

[0090] The LSTM neural model was trained on data in a specific training environment, and the training effect was evaluated using the loss function MSE. The parameters in the network were continuously updated by minimizing the loss function to optimize the model. The network training optimizer was Adam (Adaptive Moment Estimation), with an initial learning rate of 0.005 and an initial number of hidden layer neurons of 32. The neural network training environment is shown in Table 3, and the final parameters of the LSTM neural network are shown in Table 4.

[0091] MSE refers to the expected value of the squared difference between the predicted and actual values ​​of the parameters, as shown in the following formula. The smaller the MSE value, the higher the accuracy of the neural network prediction, and it is the most commonly used loss function for regression prediction models.

[0092]

[0093] In the formula: y i Let y' be the initial value for the i-th data; i is the predicted value obtained by the model; n is the number of data points.

[0094] Table 3 LSTM Neural Network Training Environment

[0095]

[0096] Table 4 LSTM Model Parameter Table

[0097]

[0098] S8. Parameter Identification and Predictive Performance Analysis of Tunnel Ignition Growth Period

[0099] The performance of the identification and prediction of parameters during the fire source growth period was analyzed using test set data, including generalization, timeliness, rationality and practical applicability.

[0100] Prediction accuracy is calculated as follows:

[0101]

[0102] In the formula: R yc To improve prediction accuracy.

[0103] ① Generalization: The model has strong generalization ability. For example... Figure 7 As shown, where Figure 7 (a) is a graph showing the analysis results of the longitudinal wind speed recognition accuracy of the training and test datasets in the LSTM neural network model when the sensor spacing is 10m. Figure 7 (b) is a graph showing the accuracy analysis results of fire source location coordinate identification in the training and test datasets of the LSTM neural network model when the sensor spacing is 10m. Figure 7 (c) is a graph showing the results of the analysis of the accuracy of fire growth coefficient identification in the training and test datasets of the LSTM neural network model when the sensor spacing is 10m. Figure 7 (d) is a graph showing the average recognition accuracy of key parameters in the training and testing datasets of the LSTM neural network model when the sensor spacing is 10m. The accuracy of predicting the heat release rate during the tunnel fire source growth period reaches 92%, the average recognition accuracy of longitudinal wind speed reaches 95%, the average recognition accuracy of fire source location coordinates reaches 94%, and the average recognition accuracy of fire growth coefficient reaches 93%.

[0104] ② Timeliness: The model has good timeliness, and can generate a response within 10 seconds and output recognition and prediction results with an accuracy of over 90% within 70 seconds.

[0105] ③ Reasonableness: Figure 8 (a) is a graph showing the analysis results of the longitudinal wind speed recognition accuracy of the training and test datasets in the LSTM neural network model under different sensor densities in this invention; Figure 8 (b) is a graph showing the analysis results of the accuracy of fire source location coordinate recognition in the training and test datasets of the LSTM neural network model under different sensor densities in this invention; Figure 8 (c) is a graph showing the results of the analysis of the accuracy of fire growth coefficient identification on the training and test datasets of the LSTM neural network model under different sensor densities in this invention. Figure 8 (d) is a graph showing the average recognition accuracy of each key parameter in the training and testing datasets of the LSTM neural network model under different sensor densities in this invention.

[0106] Increasing sensor density by 50% only improves parameter identification and prediction accuracy by 2%, while reducing effective prediction time loss by 20 seconds. For accuracy requirements above 90%, a sensor density of 10m / unit is reasonable.

[0107] ④ Practical applicability: Based on the actual vehicle combustion test data of a full-size highway tunnel, the model has certain practical applicability in real tunnel fire scenarios under non-standard working conditions, and its performance in seeking optimal solutions to ill-posed problems is excellent. The average accuracy of HRR growth process prediction reaches 70%, the average error of growth coefficient identification is within 21%, the accuracy of longitudinal wind speed identification is 100%, and the error of fire source coordinates is within 2.5m. Figure 9 (a) is a graph showing the predicted HRR growth of car combustion when the present invention is applied to a full-size highway tunnel fire test; Figure 9 (b) is a graph showing the predicted HRR growth of a bus fire when the present invention was applied to a full-size highway tunnel fire test.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A highway tunnel fire growth period parameter identification and prediction method, characterized in that, The method comprises the following steps: S1, selection of a tunnel fire growth period fire source model According to the tunnel fire heat release rate growth model established by domestic and foreign scholars, the square growth model established by adjusting the parameters simply and commonly used in computer simulation simulation for non-steady-state fire is selected as the basic model to describe the change of the heat release rate of the fire source with time, and the relationship is as follows: wherein: Q is the heat release rate, kW; t is the time to rise to this heat release rate, s; a is the fire growth coefficient, kW / s 2 ; S2, determination of the tunnel fire growth period fire source model parameters According to the analysis of the measured data of full-scale tunnel fire tests at home and abroad, the measured data of different vehicle type fire tests and vehicle related material combustion tests are obtained, and the value range of the maximum heat release rate HRR max , fire growth coefficient α, and time t max required to reach the maximum heat release rate are analyzed to further verify the correctness of the tunnel fire growth period fire source model. S3, establishment of a tunnel fire growth period FDS numerical simulation model and working condition setting The tunnel fire growth period FDS numerical simulation model is established, the monitoring positions and densities of the temperature sensor and the wind speed sensor are set, different fire source positions, longitudinal wind speeds, maximum heat release rates and growth coefficients are considered, and multiple simulation working conditions are set; S4, establishment of a tunnel fire growth period observation database The calculation results are obtained by multi-condition numerical simulation, and the data samples of fire growth coefficient, HRR max , HRR, time, sensor position, fire source position, temperature and longitudinal wind speed in the fire growth stage of each condition calculation result are extracted, and the data is stored in the same time sequence, so as to establish an observation database of the fire growth stage of the tunnel. S5, data labeling and normalization processing of the tunnel fire growth period The data labels HRR, fire growth coefficient, fire source position and longitudinal wind speed are extracted from the database, and the data is labeled and normalized; S6, establishment of an LSTM neural network model for tunnel fire growth period parameter identification and prediction An LSTM neural network model is built, a regression prediction model is established with HRR as the output variable, and a multi-label classification model is established with the fire growth coefficient, the fire source position and the longitudinal wind speed as the output variables; S7, training of the LSTM neural network model for tunnel fire growth period parameter identification and prediction The data is trained in the LSTM neural network training environment, the loss function MSE is selected to evaluate the training effect of the LSTM neural model, the parameters in the network are continuously updated through the minimization of the loss function, and the purpose of optimizing the model is achieved; S8, performance analysis of tunnel fire growth period parameter identification and prediction The performance of the tunnel fire growth period parameter identification and prediction is analyzed by using the test set data, including generalization, timeliness, rationality and practicality.

2. The highway tunnel fire growth period parameter identification and prediction method according to claim 1, characterized in that, In step S3, the tunnel fire growth period FDS numerical simulation model is established, the tunnel model section is based on the standard inner contour of a two-lane tunnel of an expressway and a first-class highway, the linear type is selected as a straight linear type, and the length is 100 m; the sensor monitoring positions and densities are set, 11 ceiling temperature monitoring devices are arranged at an interval of 5 m in a range of 25 m upstream and downstream of the fire source, which can simulate the sensor arrangement density at an interval of 5 m and 10 m, and 2 wind speed monitoring devices are arranged 10 m upstream and downstream of the fire source.

3. The highway tunnel fire growth period parameter identification and prediction method according to claim 2, characterized in that, In step S5, the data labels HRR, fire growth coefficient, fire source position and longitudinal wind speed are extracted from the database, each data table is divided into multiple data samples according to a time length of 30 s; the Min-Max normalization method is used for standardization processing of the data to remove the differences in dimensions and orders of magnitude between different types of data, and convert them to the range of [0, 1]; the normalization formula is as follows: wherein: x i is the initial value of the i-type data; is the normalized value; max x i is the maximum value of the i-type data; min x i is the minimum value of the i-type data.

4. The highway tunnel fire growth period parameter identification and prediction method of claim 3, wherein, In step S7, MSE refers to the expected value of the square difference between the parameter predicted value and the parameter true value, as follows, and the smaller the MSE value, the higher the prediction accuracy of the neural network; wherein: y i is the initial value of the i-th data; is the predicted value obtained from the model; n is the number of data.

5. The highway tunnel fire growth period parameter identification and prediction method of claim 4, wherein, In step S8, the prediction accuracy is calculated as follows: wherein: R yc is the prediction accuracy.

Images