A method for predicting thermal radiation of aviation kerosene pool fire on flight deck under action of ambient wind
By establishing a numerical simulation model and performing dimensionless processing, combined with BP neural network training, the problem of predicting the thermal radiation of multi-scale aviation kerosene pools was solved, achieving low-cost and high-efficiency prediction results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA UNIV OF GEOSCIENCES (WUHAN)
- Filing Date
- 2023-03-23
- Publication Date
- 2026-05-19
AI Technical Summary
Existing technologies are difficult to effectively predict the thermal radiation of aviation kerosene pools at multiple scales. Numerical simulation methods are costly and require high-dimensional training data. Artificial neural networks lack full-scale research and are difficult to validate.
By establishing a numerical simulation model, performing dimensionless processing and normalization, and combining it with BP neural network training, the model uses dimensionless wind speed distance and radiative heat flux density for prediction, thereby reducing data dimensionality and optimizing neural network training.
It enables low-cost, rapid, multi-scale prediction of thermal radiation from aviation kerosene pools, improves neural network training efficiency, reduces prediction costs, and enhances prediction accuracy.
Smart Images
Figure CN116822041B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of digital computing or data processing methods specifically applicable to particular applications, and specifically relates to the field of predicting thermal radiation from aviation kerosene pools on flight decks under the influence of ambient wind. Background Technology
[0002] An aircraft carrier is a large surface combat vessel whose primary weapon is carrier-based aircraft, and it is an indispensable and crucial piece of military equipment in modern naval operations. Aviation kerosene, due to its high calorific value and rapid combustion rate, has a wide range of applications in aerospace and military fields. However, these characteristics also make aviation kerosene a unique source of danger. Oil pool fires caused by aviation kerosene leaks on aircraft carrier flight decks can result in heat radiation that harms personnel, aircraft, weapons, and equipment. Compared to ordinary fires, aviation kerosene pool fires in open spaces are more dangerous because the oil layer is thinner, the burning area is relatively larger, and the fire plume is more easily affected by the surrounding environment. Therefore, studying heat radiation can not only better protect the lives and health of personnel but also provide valuable reference data for fire and rescue operations.
[0003] Currently, many scholars have conducted experimental studies on the thermal radiation of aviation kerosene pools. However, due to limitations such as high economic costs and difficulties in controlling experimental parameters, previous studies have often been limited to single scales such as small and medium scales. Therefore, traditional experimental methods are insufficient to meet the current needs for multi-scale research on the thermal radiation of aviation kerosene pools. In recent years, numerical simulation methods have gained popularity due to their lower economic costs; for example, DOI is [DOI missing].
[0004] The paper "Simulation Analysis of the Consequences of Kerosene Leakage Accidents in Aviation Kerosene Tank Areas" (10.3969 / j.issn.1006-7906.2021.03.016) uses numerical simulation to simulate the combustion and explosion phenomena caused by aviation kerosene leakage, and analyzes the range of its thermal radiation impact. However, numerical simulation methods also have their drawbacks: simulation results need to be verified with corresponding experimental results; numerical simulation parameters need to be constantly adjusted, such as the thermal radiation fraction, which is affected not only by the scale of the oil pool but also by the wind speed; and the simulation time cost is high, potentially requiring tens of hours or even more for a single scenario. Therefore, for multi-scale aviation kerosene pool fire research, numerical simulation methods are insufficient to obtain a large amount of reliable data in a short period.
[0005] Artificial neural networks (ANNs) are widely used in regression prediction due to their powerful nonlinear regression capabilities. In the field of oil pool fires, ANNs have also demonstrated their superiority, predicting the dangerous consequences of jet fires, early-stage oil pool fires, and late-stage oil pool fires through ANN analysis. However, the validation data in existing ANN prediction methods only cover the thermal radiation of fuel pool fires at specific scales, without further research on the thermal radiation of fuel pool fires at all scales. Training data often includes multiple dimensions such as fire source diameter, wind speed, mass loss rate, and distance, and this high data dimensionality makes training neural networks quite difficult. Summary of the Invention
[0006] In view of this, the present invention provides a method for predicting the thermal radiation of an aviation kerosene pool on a flight deck under the action of ambient wind, comprising the following steps:
[0007] S1. Collect existing experimental data;
[0008] S2. Establish a numerical simulation model of the aviation kerosene pool fire on the flight deck, and verify the established numerical simulation model using the experimental data collected in step S1, until the simulation data simulated by the numerical simulation model and the experimental data in step S1 are within the error threshold range.
[0009] S3. The numerical simulation model outputs numerical simulation data; the numerical simulation data includes multiple data sets, each data set including the diameter of the fire source, the ambient wind speed, and the radiative heat flux density at different distances from the center of the fire source.
[0010] S4. Perform dimensionless processing and normalization on the numerical simulation data to obtain processed data sets. Each processed data set includes dimensionless wind speed distance and dimensionless radiative heat flux density at different distances from the center of the fire source.
[0011] S5. Build a neural network model using the data set processed in step S4. Use the dimensionless wind speed distance in the processed data set as the input parameter and the dimensionless radiative heat flux density in the processed data set as the output parameter to build the neural network model. Use the data set processed in S4 as data samples to train the neural network model and obtain the optimal neural network model.
[0012] S6. Using the optimal neural network model obtained in step S5, predict the radiative heat flux density at different distances from the center of the fire source under different fire source diameters, wind speeds, and distances.
[0013] Furthermore, the process of establishing a numerical simulation model of the aviation kerosene pool fire on the flight deck in step S2 is as follows: selecting a fire model, determining the combustion rate, and adjusting the mesh size and radiation fraction.
[0014] Furthermore, the combustion rate is determined in step S2 according to the following formula:
[0015]
[0016] In the formula: u w Ambient wind speed, m / s; m" represents combustion rate, kg / (m³). 2 ·s); D is the diameter of the ignition source, in meters;
[0017] Furthermore, in step S2, the mesh size ranges from [D* / 16, D* / 4]. Here, D* is the characteristic diameter of the fire source, calculated using the following formula:
[0018]
[0019] In the formula: Q is the heat release rate, kW; ρ ∞ air density, kg / m³ 3 c p The specific heat capacity of air at constant pressure is given by T, in kJ / (kg·K). ∞ The ambient temperature is K; g is the acceleration due to gravity, m / s². 2 ;
[0020] Furthermore, in step S2, the heat release rate Q is calculated using the following formula:
[0021] Q = m″ΔH c A
[0022] Where: Q is the heat release rate, kW; m" is the combustion rate, kg / (m³) 2 ·s); ΔH c Heat of combustion, kJ / kg; A is the area of the oil tank, m². 2 ;
[0023] Furthermore, the fire model in step S2 includes a fluid dynamics sub-model, a turbulence sub-model, a combustion sub-model, and a thermal radiation sub-model.
[0024] Furthermore, the equations of the thermal radiation sub-model are:
[0025]
[0026] In the formula: It is a position vector; It is the direction vector; σ is the scattering direction vector; a is the absorption coefficient; n is the refractive index; σ s σ is the scattering coefficient; σ is the Stefan-Boltzmann constant, with a value of 5.672 × 10⁻⁶. -8 W / (m 2 ·K 4 I represents radiation intensity, W / m².2 Φ is the heat dissipation phase function of the condensed phase; Ω′ is the solid angle, in rad.
[0027] Furthermore, the process of dimensionless processing and normalization of the numerical simulation data in step S4 is as follows:
[0028] The data set of fire source diameter, ambient wind speed, and radiative heat flux density at different distances from the fire source center were dimensionless, and then the dimensionless data were normalized.
[0029] The expression for the dimensionless treatment of the radiative heat flux density is:
[0030]
[0031] In the formula, q * q is the dimensionless radiative heat flux density; kW / m³ is the radiative heat flux density. 2 ; m" represents the combustion rate, kg / (m 2 ·s); c p T represents the specific heat capacity of air at constant pressure, in kJ / (kg·K); ∞ The ambient temperature, in K;
[0032] The expression for combining the fire source diameter and ambient wind speed into a dimensionless wind speed distance is as follows:
[0033]
[0034] In the formula, u * L is a dimensionless ambient wind speed. * The distance is dimensionless; g is the acceleration due to gravity, in m / s². 2 ; m" represents the combustion rate, kg / (m 2 ·s); D is the diameter of the ignition source, in meters; ρ ∞ air density, kg / m³ 3 ;u w The ambient wind speed is in m / s; L is the horizontal distance to the center of the oil tank in m.
[0035] The expression for normalizing dimensionless data is:
[0036]
[0037] In the formula, The data is after normalization; x i The data is dimensionless; x min Let x be the minimum value in the sample data sequence. max It is the maximum value;
[0038] Furthermore, the training process for the neural network model in step S4 is as follows:
[0039] The data sample was divided into a training set (90%) and a test set (10%). Using a trial-and-error method, the optimal number of hidden layer neurons was determined based on the mean squared error of the training set under different numbers of hidden layer neurons. The formula for determining the optimal number of hidden layer neurons is as follows:
[0040]
[0041] Where h is the optimal number of hidden layer neurons; s is the number of input layer neurons; t is the number of output layer neurons; and b is a constant, ranging from 1 to 10.
[0042] The advantages of this invention are:
[0043] 1) The present invention provides a method for predicting the thermal radiation of an aviation kerosene pool on a flight deck under the action of environmental wind. This method overcomes the difficulty of obtaining sample data of thermal radiation from aviation kerosene pools by using numerical simulation.
[0044] 2) The present invention provides a method for predicting the thermal radiation of a multi-scale aviation kerosene pool fire on a flight deck under the action of ambient wind. This method uses a BP neural network to avoid the experimental and numerical simulation methods of aviation kerosene pool fire, thereby reducing the prediction cost.
[0045] 3) The present invention provides a method for predicting the thermal radiation of aviation kerosene pools on flight decks under the action of environmental wind. This method reduces the dimensionality of the dataset through dimensionless processing and improves the training efficiency of neural networks.
[0046] 4) The present invention provides a method for predicting the thermal radiation of aviation kerosene pool fire on flight deck under the action of ambient wind. This method only needs to change two inputs (ambient wind speed and horizontal distance from the fire source) to predict the thermal radiation of aviation kerosene pool fire on multi-scale under the action of ambient wind. Attached Figure Description
[0047] Figure 1 This is a flowchart illustrating a method for predicting thermal radiation from an aviation kerosene pool on a flight deck under environmental wind conditions, as described in this invention.
[0048] Figure 2 A comparison of the mesh size used in numerical simulations of aviation kerosene pool fires at different scales with experimental data;
[0049] Figure 3 A graph showing the comparison between radiation fractions obtained from numerical simulations of aviation kerosene pool fires at different scales and experimental data;
[0050] Figure 4 Dimensionless data from numerical simulations and their linear fit plots;
[0051] Figure 5 This is a performance graph of the best BP neural network model based on sample data;
[0052] Figure 6 A linear fit plot of the data predicted by the coupled neural network. Detailed Implementation
[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0054] Please refer to Figure 1 This is a flowchart illustrating a method for predicting thermal radiation from an aviation kerosene pool on a flight deck under environmental wind conditions, as described in this invention. The method includes the following steps:
[0055] S1. Collect existing experimental data; search existing aviation kerosene pool fire experiments and collect experimental data on temperature and thermal radiation under specific operating conditions of aviation kerosene pool fire.
[0056] S2. Establish a numerical simulation model of the aviation kerosene pool fire on the flight deck, and verify and adjust the established numerical simulation model using the experimental data collected in step S1, until the simulated data simulated by the numerical simulation model and the experimental data in step S1 are within the error threshold range.
[0057] The process of establishing a numerical simulation model for an aviation kerosene pool fire on a flight deck is as follows: selecting a fire model, determining the combustion rate, and adjusting the mesh size and radiation fraction.
[0058] The combustion rate in the model is determined based on the following formula:
[0059]
[0060] In the formula: u w Ambient wind speed, m / s; m" represents combustion rate, kg / (m³). 2 ·s); D is the diameter of the ignition source, in meters;
[0061] The mesh size of the numerical simulation model determines the accuracy and time cost of the numerical simulation calculation, and it is advisable to choose [D* / 16, D* / 4].
[0062] Where D* is the characteristic diameter of the fire source, calculated using the following formula:
[0063]
[0064] In the formula: Q is the heat release rate, kW; ρ ∞ air density, kg / m³ 3 c p The specific heat capacity of air at constant pressure is given by T, in kJ / (kg·K). ∞ The ambient temperature is K; g is the acceleration due to gravity, m / s². 2 ;
[0065] The formula for calculating the heat release rate Q is as follows:
[0066] Q = m″ΔH c A
[0067] Where: Q is the heat release rate, kW; m" is the combustion rate, kg / (m³) 2 ·s); ΔH c Heat of combustion, kJ / kg; A is the area of the oil tank, m². 2 ;
[0068] The selected fire model mainly includes a fluid dynamics sub-model, a turbulence sub-model, a combustion sub-model, and a thermal radiation sub-model. The fluid dynamics sub-model uses an approximation of the Navier-Stokes equations, suitable for describing thermally driven low-speed flows. The turbulence sub-model employs Large Eddy Simulation (LES). The combustion sub-model uses a mixture fractional combustion model. The thermal radiation sub-model primarily calculates radiative heat conduction by solving the radiative transport equation (RTE) for gray volumes.
[0069] Among them, the equations for the thermal radiation sub-model are:
[0070]
[0071] In the formula: It is a position vector; It is the direction vector; σ is the scattering direction vector; a is the absorption coefficient; n is the refractive index; σ s σ is the scattering coefficient; σ is the Stefan-Boltzmann constant, with a value of 5.672 × 10⁻⁶. -8 W / (m 2 ·K 4 I represents radiation intensity, W / m². 2 Φ is the heat dissipation phase function of the condensed phase; Ω′ is the solid angle, in rad.
[0072] The fire heat release rate Q, especially the convective heat component (Qc = (1-X)Q), is a key technical parameter that directly affects the determination of smoke production and critical wind speed. Here, X is the fire radiation fraction, which is often taken as a constant of 0.2-0.4, referencing oil pool fire data under windless conditions in engineering design. As ventilation speed increases, the flame shape, size, and brightness change significantly, and the value of X may change accordingly. Therefore, different radiation fractions need to be set in numerical simulations. Based on the experimental data collected in step S1, the mesh size and radiation fraction of the numerical simulation model are determined.
[0073] S3. Output numerical simulation data; Based on the numerical simulation model established in step S2, obtain thermal radiation data (radiative heat flux density) at specific locations under different working conditions.
[0074] S4. Dimensionless and Normalization Processing: Based on the numerical simulation data of different operating conditions in step S3, the radiative heat flux density is processed to be dimensionless, and its expression is as follows:
[0075]
[0076] Where, q * q is the dimensionless radiative heat flux density; kW / m³ is the radiative heat flux density. 2 ; m" represents the combustion rate, kg / (m 2 ·s); c p T represents the specific heat capacity of air at constant pressure, in kJ / (kg·K); ∞ The ambient temperature, in K;
[0077] The expression for combining the fire source diameter and ambient wind speed into a dimensionless wind speed distance is:
[0078]
[0079] Where, u * L is a dimensionless ambient wind speed. * The distance is dimensionless; g is the acceleration due to gravity, in m / s². 2 ; m" represents the combustion rate, kg / (m 2 ·s); D is the diameter of the ignition source, in meters; ρ ∞ air density, kg / m³ 3 ;u w The ambient wind speed is in m / s; L is the horizontal distance to the center of the oil tank in m.
[0080] The dimensionless data is normalized as shown in the following formula:
[0081]
[0082] In the formula, The data is after normalization; xi The data is dimensionless; x min Let x be the minimum value in the sample data sequence. max It is the maximum value;
[0083] S5. Establish a neural network model; using dimensionless wind speed and distance as input and dimensionless radiative heat flux density as output, divide the sample data into a training set (90%) and a test set (10%); using a trial-and-error method, based on the mean square error of the training set under different numbers of hidden layer neurons, obtain the optimal number of hidden layer neurons. The formula for determining the optimal number of hidden layer neurons is:
[0084]
[0085] Where h is the optimal number of hidden layer neurons; s is the number of input layer neurons; t is the number of output layer neurons; and b is a constant, ranging from 1 to 10.
[0086] As a preferred approach, the optimal number of hidden layer neurons in the neural network model (BP neural network) is determined to be 3 training iterations. The mean squared error (MSE) of the 3 training iterations is compared, and the minimum value is taken as the optimal parameter. A smaller MSE indicates a smaller error between the output value and the actual value, and its expression is as follows:
[0087]
[0088] In the formula, x i and y i These represent the output data and training data, respectively; N is the number of samples. The remaining training parameters of the BP neural network remain at their default values.
[0089] S6. Predict the thermal radiation of aviation kerosene pool fire at multiple scales; based on the optimal BP neural network model in step S5, adjust the dimensionless wind speed distance input to predict the thermal radiation of aviation kerosene pool fire at multiple scales and wind speeds.
[0090] The following example, using a certain type of aviation kerosene, further illustrates the aforementioned steps:
[0091] In step S1, existing literature on aviation kerosene pool fire experiments is searched, mainly focusing on experiments at different pool sizes (large, medium, and small) and ambient wind speeds. Experimental data such as temperature and thermal radiation of aviation kerosene pool fire under specific operating conditions are collected. The collected experimental conditions are shown in Table 1.
[0092] Table 1. Experimental conditions for flight deck kerosene pool fire under ambient wind conditions.
[0093]
[0094] In step S2, a three-dimensional numerical simulation model of an aviation kerosene pool fire is established.
[0095] 2-1: Establish a three-dimensional numerical simulation model;
[0096] 2-2: Determine the combustion rate based on the retrieved experimental data:
[0097] At a small scale (D=0.08m), the combustion rate is:
[0098] m″=0.0122u w +0.0065
[0099] The combustion rate at the mesoscale (D = 0.6 m) is:
[0100] m″=0.083u w +0.0333
[0101] At large scales (D = 5m), the combustion rate is:
[0102] m″=0.0016u w +0.054
[0103] In the formula: u w Ambient wind speed, m / s; m" represents combustion rate, kg / (m³). 2 ·s);
[0104] 2-3: Adjust the mesh size and radiation fraction in the numerical simulation model, and compare the simulation results with the collected experimental data to obtain the optimal parameters. The mesh size comparison results are shown below. Figure 2 As shown, the comparison results of radiation fraction X are as follows: Figure 3 As shown in Table 3 (where the large-scale value D = 5m and X = 0.35), the radiation fraction X corresponding to each wind speed at multiple scales has been verified.
[0105] Table 2. Radiation fraction settings at different wind speeds across multiple scales.
[0106]
[0107] 2-4: The fire model mainly includes a fluid dynamics sub-model, a turbulence sub-model, a combustion sub-model, and a thermal radiation sub-model. The fluid dynamics model adopts an approximate form of the Navier-Stokes equations, suitable for describing thermally driven low-speed flows. The turbulence model uses the Large Eddy Simulation (LES) method. The combustion model selects the Mixture Fraction Combustion Model. The radiation model mainly uses the Radiative Transport Equation (RTE) to calculate radiative heat conduction.
[0108] The thermal radiation model equation is as follows:
[0109]
[0110] In the formula: It is a position vector; It is the direction vector; σ is the scattering direction vector; a is the absorption coefficient; n is the refractive index; σ s σ is the scattering coefficient; σ is the Stefan-Boltzmann constant, with a value of 5.672 × 10⁻⁶. -8 W / (m 2 ·K 4 I represents radiation intensity, W / m². 2 Φ is the heat dissipation phase function of the condensed phase; Ω′ is the solid angle, in rad.
[0111] In step S3, based on the validated aviation kerosene numerical simulation model, the diameter of the fire source and the ambient wind speed are adjusted to obtain thermal radiation simulation data under different working conditions, mainly including the radiative heat flux density at different distances from the center of the fire source.
[0112] In step S4, the data obtained in step S3, including the fire source diameter, ambient wind speed, and radiative heat flux density at distance from the fire source center, are subjected to dimensionless and normalized processing to obtain 969 sample data points with different dimensionless radiative heat flux densities and dimensionless wind speed distances, such as... Figure 4 As shown.
[0113] In step S5, based on the sample data from step S4, using dimensionless wind speed and distance as input and dimensionless radiative heat flux density as output, 90% of the sample data, totaling 872 data points, is divided into a training set, and the remaining data is divided into a test set; a trial-and-error method is adopted, based on the formula... In this example, the number of input layers is 2, the number of output layers is 1, and b takes values from 1 to 10. The remaining parameters of the BP neural network are set to default values. The mean squared error (×10) of the training set with 2-11 hidden layer neurons is compared. -3 The results are shown in Table 3:
[0114] Table 3. Mean squared error of training sets with different numbers of neurons in hidden layers (×10) -3 )
[0115]
[0116] In three training iterations, the mean squared error was lowest when the number of hidden layer neurons was 8, indicating that 8 neurons were the optimal number for the hidden layer. The remaining parameters were set to their default values. The regression performance of the BP neural network is as follows: Figure 5 As shown.
[0117] In step S6, based on the aforementioned optimal BP neural network model, dimensionless wind speed distances (0:0.2:2) and fire source diameters (0.04, 1.5, 2.5, and 3.5 m) different from the sample data are input to predict the thermal radiation of aviation kerosene pool fires at multiple scales and wind speeds. The prediction results are as follows: Figure 6 As shown, the fitting effect (R) 2 =0.901) compared to (R = 0.901) based solely on numerical simulation results 2 =0.883) is better, which demonstrates the excellent predictive performance of the present invention.
[0118] The beneficial effects of the technical solution provided by this invention are: overcoming the difficulty of obtaining sample data of thermal radiation from aviation kerosene pool fires; avoiding the experimental and numerical simulation methods of aviation kerosene pool fires, thus reducing prediction costs; reducing the dimensionality of the dataset, thus improving the training efficiency of neural networks; and realizing multi-scale prediction of thermal radiation from aviation kerosene pool fires.
[0119] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting the thermal radiation of an aviation kerosene pool on a flight deck under environmental wind conditions, characterized in that, It includes the following steps: S1. Collect existing experimental data; S2. Establish a numerical simulation model of the aviation kerosene pool fire on the flight deck, and verify the established numerical simulation model using the experimental data collected in step S1, until the simulation data simulated by the numerical simulation model and the experimental data in step S1 are within the error threshold range. S3. The numerical simulation model outputs numerical simulation data; the numerical simulation data includes multiple data sets, each data set including the diameter of the fire source, the ambient wind speed, and the radiative heat flux density at different distances from the center of the fire source. S4. Perform dimensionless processing and normalization on the numerical simulation data to obtain the processed data set. The process of performing dimensionless processing and normalization on the numerical simulation data is as follows: The diameter of the fire source, the ambient wind speed, and the radiative heat flux density at different distances from the center of the fire source in the data set were dimensionless, and then the dimensionless data were normalized. The expression for the dimensionless treatment of the radiative heat flux density is: In the formula, q * The dimensionless radiative heat flux density; q Radiative heat flux density, kW / m 2 ; m" The combustion rate is expressed in kg / (m³). 2 ·s); c p is the specific heat capacity of air at constant pressure, kJ / (kg·K); T ∞ The ambient temperature, in K; Combining the fire source diameter and ambient wind speed to form a dimensionless wind speed distance, its expression is: In the formula, u * Dimensionless ambient wind speed; L * The distance is dimensionless; g The acceleration due to gravity is m / s². 2 ; m" The combustion rate is expressed in kg / (m³). 2 ·s); D The diameter of the fire source is in meters (m). ρ ∞ air density, kg / m³ 3 ; u w The ambient wind speed is in m / s. L The horizontal distance from the center of the oil tank is in meters (m). The expression for normalizing dimensionless data is: In the formula, The data has been normalized. x i Dimensionless data; x min The minimum value in the sample data sequence. x max It is the maximum value; Each processed data set includes dimensionless wind speed distance and dimensionless radiative heat flux density at different distances from the center of the fire source; S5. Build a neural network model using the data set processed in step S4. Use the dimensionless wind speed distance in the processed data set as the input parameter and the dimensionless radiative heat flux density in the processed data set as the output parameter to build the neural network model. Use the data set processed in S4 as data samples to train the neural network model and obtain the optimal neural network model. S6. Using the optimal neural network model obtained in step S5, predict the radiative heat flux density at different distances from the center of the fire source under different fire source diameters, wind speeds, and distances.
2. The method for predicting thermal radiation from an aviation kerosene pool on a flight deck under environmental wind conditions, as described in claim 1, is characterized in that... The process of establishing a numerical simulation model of an aviation kerosene pool fire on a flight deck in step S2 is as follows: select a fire model, determine the combustion rate, and adjust the mesh size and radiation fraction.
3. The method for predicting thermal radiation from an aviation kerosene pool on a flight deck under environmental wind conditions, as described in claim 2, is characterized in that... The combustion rate is determined in step S2 according to the following formula: In the formula: u w For ambient wind speed; m" The combustion rate; D The diameter of the fire source.
4. The method for predicting thermal radiation from an aviation kerosene pool on a flight deck under environmental wind conditions, as described in claim 2, is characterized in that... In step S2, the range of grid size values is [ D* / 16, D* / 4], where, D* The characteristic diameter of the fire source is calculated using the following formula: In the formula: Q The rate of heat release; air density; c p The specific heat capacity of air at constant pressure; Ambient temperature; g This is the acceleration due to gravity.
5. The method for predicting thermal radiation from an aviation kerosene pool on a flight deck under ambient wind conditions, as described in claim 4, is characterized in that... In step S2, the heat release rate Q The calculation formula is as follows: In the formula: Q The rate of heat release; m" The combustion rate; H c Heat of combustion; A The area of the oil tank.
6. The method for predicting thermal radiation from an aviation kerosene pool on a flight deck under environmental wind conditions, as described in claim 2, is characterized in that... The fire model in step S2 includes a fluid dynamics sub-model, a turbulence sub-model, a combustion sub-model, and a thermal radiation sub-model.
7. The method for predicting thermal radiation from an aviation kerosene pool on a flight deck under environmental wind conditions, as described in claim 6, is characterized in that: The equations for the thermal radiation sub-model are: In the formula: It is a position vector; It is the direction vector; The scattering direction vector; The absorption coefficient; The refractive index is 1. The scattering coefficient; Here is the Stefan-Boltzmann constant, with a value of 5.672 × 10⁻⁶. -8 W / (m 2 ·K 4 ), Radiation intensity; For the heat dissipation phase function of the condensed phase; It is a solid angle.
8. The method for predicting thermal radiation from an aviation kerosene pool on a flight deck under environmental wind conditions, as described in claim 1, is characterized in that: The training process for the neural network model in step S4 is as follows: The data sample was divided into a training set (90%) and a test set (10%). Using a trial-and-error method, the optimal number of hidden layer neurons was determined based on the mean squared error of the training set under different numbers of hidden layer neurons. The formula for determining the optimal number of hidden layer neurons is as follows: in, h The optimal number of neurons in the hidden layer; s This represents the number of neurons in the input layer. t This represents the number of neurons in the output layer. b It is a constant, ranging from 1 to 10.