Jet diffusion flame size and radiation characteristic prediction method and system based on machine learning method

Abstract

The invention relates to the technical field of jet flow flame disaster-causing parameter prediction, and discloses a jet flow diffusion flame size and radiation characteristic prediction method and system based on a machine learning method. The method comprises the following steps: constructing a flame size empirical model based on experimental data characteristics; constructing a machine learning model to predict flame height, width and lifting height, and combining the empirical model as a loss function with the empirical model; mass flow is determined according to the gas parameters, and a flame dwell time and radiation fraction calculation model is established in combination with a flame width prediction model; and establishing a flame total radiation power calculation model and a radiation source weighted distribution model, and solving the radiation heat flux by combining a flame height and lifting height prediction model. The prediction method provided by the invention can be used for predicting the sizes and radiation characteristics of jet diffusion flames with different scales under various working conditions, and has the characteristics of high prediction efficiency, high prediction precision, wide application range and the like.

Description

Technical Field

[0001] This invention relates to the field of jet flame disaster parameter prediction technology, specifically to a method and system for predicting the size and radiation characteristics of jet diffusion flames based on machine learning. Background Technology

[0002] During transportation, storage, and use, gaseous fuels are prone to gas leaks, which can lead to jet fires and pose a significant threat to public safety.

[0003] Jet-borne diffused flames are characterized by high initial momentum, long jet distance, and high thermal radiation, which can lead to serious consequences. Therefore, accurate and efficient prediction of their key disaster-causing parameters is of great significance. Flame size is an important characteristic parameter of jet flames, used to indicate the dangerous area of ​​the flame and predict its thermal radiation and heat release rate, providing a reference for fire safety and process design. Flame radiation poses a great hazard to humans and the environment, threatening not only surrounding buildings and combustible materials but also rescue personnel, causing skin ulcers, disability, and even death. Therefore, accurate prediction of flame size and radiation characteristics can effectively prevent casualties and provide a reliable basis for calculating fire compartments and fire separation distances in fire prevention work. Furthermore, in industrial applications, accurate prediction of flame size and radiation characteristics is crucial for the thermal protection design of related equipment, helping to improve the durability and service life of the combustion chamber. Therefore, developing methods for predicting the size and radiation characteristics of jet-borne diffused flames is of great significance for guiding the design and safe and efficient operation of industrial combustion equipment and addressing related fire safety issues.

[0004] Currently, some scholars have dedicated themselves to the research of flame size parameters and proposed some prediction models. However, existing prediction models are often designed for specific conditions or scenarios, which has significant limitations and makes it difficult to directly apply them to other conditions or scenarios, thus limiting their wide applicability. Furthermore, the prediction of flame radiation distribution depends on flame size; therefore, the accuracy and applicability of existing radiation prediction methods often fall short of requirements. For example, for raised flames, existing prediction methods struggle to efficiently predict their flame profile, let alone accurately predict their flame radiation distribution. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method and system for predicting the size and radiation characteristics of jet-diffused flames based on machine learning. This method offers multiple functions, including predicting flame size parameters such as height, width, and rise height, as well as flame radiation distribution characteristics. It boasts high accuracy, high prediction efficiency, and wide applicability. This invention combines empirical models with machine learning, enabling the prediction of flame size parameters with high accuracy and faster computation speed. Based on a rich database of jet-diffused flame parameters, this invention significantly improves the accuracy and applicability of the prediction model. By combining a flame radiation model with high-precision prediction models for flame height, width, and rise height, this invention significantly improves the accuracy of flame radiation distribution prediction.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a method for predicting the size and radiation characteristics of a jet diffusion flame based on machine learning, comprising the following steps:

[0008] Conduct jet diffusion flame experiments, collect experimental data, and construct a jet diffusion flame parameter library;

[0009] Analyze the characteristics of flame size parameters and construct an empirical model of jet diffusion flame size;

[0010] A machine learning model is constructed, and the constructed machine learning model is trained by combining the flame parameters in the jet diffusion flame parameter library and the jet diffusion flame size empirical model to obtain the optimal flame size prediction model and save the corresponding weights.

[0011] Determine the initial parameters of the jet, and then use the weights to directly predict the flame size parameters;

[0012] By combining the predicted results of flame size parameters, the flame residence time and flame radiation fraction are solved, and a calculation model for total flame radiation power and a weighted distribution model of radiation sources are established.

[0013] The radiative heat flux is solved based on the total radiative power calculation model of the flame and the weighted distribution model of the radiation source.

[0014] In one embodiment, the flame parameters in the jet diffusion flame parameter library include jet initial parameters and flame size parameters; the jet initial parameters include nozzle diameter, fuel mixing ratio, outlet pressure and outlet velocity, and the flame size parameters include flame height, flame width and flame rise height.

[0015] In one embodiment, the analysis of flame size parameter characteristics and the construction of an empirical model of jet diffusion flame size specifically includes:

[0016] The constructed empirical model for the size of the jet diffusion flame is expressed as follows:

[0017]

[0018] Where H and d represent flame height and nozzle diameter, respectively; ρ0 and ρ∞ represent fuel density and air density, respectively; and u0 and u∞ represent fuel jet velocity and local speed of sound, respectively.

[0019] In one embodiment, the construction of the machine learning model involves training the constructed machine learning model by combining flame parameters from the jet diffusion flame parameter library and the jet diffusion flame size empirical model to obtain the optimal flame size prediction model and saving the corresponding weights. Specifically, this includes:

[0020] The sample data in the jet diffusion flame parameter library are standardized; a neural network model is constructed; the neural network parameters are determined; a loss function is constructed based on the empirical model of jet diffusion flame size; the standardized sample data is trained, and the prediction error is calculated using the loss function; when the prediction error meets the requirements, the optimal flame size prediction model is obtained, and the optimal weights are saved.

[0021] In one embodiment, the flame size parameters include flame height, flame width, and flame rise height.

[0022] In one embodiment, the process of calculating the flame dwell time and flame radiation fraction based on the prediction results of the flame size parameters specifically includes:

[0023] The flame dwell time t is:

[0024]

[0025] Where, ρ f The flame density is represented by W, the flame width by H, and the flame height by f. s This indicates the mass fraction of fuel under stoichiometric conditions. This indicates the fuel mass flow rate; flame width and flame height are obtained by calling the optimal flame size prediction model.

[0026] The flame radiation fraction X rad for:

[0027]

[0028] Among them, a p and T ad represents Planck's average absorptivity and the temperature of the adiabatic flame, respectively.

[0029] In one embodiment, establishing the total flame radiation power calculation model and the radiation source weighted distribution model specifically includes:

[0030] The total radiant power S of the flame rad The calculation model is as follows:

[0031]

[0032] Where, ΔH c Indicates the heat of combustion of fuel;

[0033] The weighted distribution model of the radiation source is as follows:

[0034]

[0035] Among them, w i The weight of the i-th radiation source is represented by N, and the number of radiation sources is represented by 1≤n≤N. The radiation sources are uniformly distributed along the center line of the flame. The position of the first radiation source is determined based on the predicted value of the flame rise height, and the position of the N-th radiation source is determined based on the predicted value of the flame height.

[0036] In one embodiment, the step of solving for the radiative heat flux based on the total flame radiant power calculation model and the radiation source weighted distribution model specifically includes:

[0037] The radiative heat flux q is:

[0038]

[0039] Among them, L i τ represents the distance between the monitoring point and the i-th radiation source. i α represents transmittance. i This represents the angle between the normal of the monitoring point and the line connecting the i-th radiation source and the monitoring point.

[0040] In a second aspect, the present invention provides a computer system including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method of any embodiment of the first aspect.

[0041] Compared with the prior art, the beneficial technical effects of the present invention are:

[0042] (1) Wide range of applications: Based on a large database of jet diffusion flame parameters, this invention has rich sample diversity. By establishing a combination of machine learning methods and empirical models, it can accurately predict the size and radiation characteristics of jet diffusion flame under various working conditions such as different nozzle diameters, fuel compositions, jet velocities and outlet pressures.

[0043] (2) High prediction efficiency: This invention combines the empirical model as a loss function with the machine learning model, which greatly improves the model training efficiency and accuracy; when making predictions, the saved weight file is called to directly calculate the target parameters, which greatly improves the prediction efficiency.

[0044] (3) High prediction accuracy: This invention combines the empirical model of flame size with machine learning methods, which greatly improves the prediction accuracy of the model, and thus improves the prediction accuracy of radiation distribution based on flame size parameters; applying the flame lift height to the flame radiation model can more accurately predict the radiation distribution of the lift flame. Attached Figure Description

[0045] Figure 1 This is a flowchart of the method in an embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram illustrating the process of constructing, training, and inferring the optimal flame size prediction model in an embodiment of the present invention.

[0047] Figure 3 This is a regression diagram of the flame height prediction results for the test set in this embodiment of the invention;

[0048] Figure 4 The above are the predicted radiation distribution results of the hydrogen-doped methane jet diffusion flame in the embodiments of the present invention; wherein, (a) is the longitudinal profile cloud map with y = 0m, and (b) is the transverse profile cloud map with z = 0.99m.

[0049] Figure 5 This is a comparison between the predicted and experimental results of the radiation distribution of the hydrogen-doped methane jet diffusion flame at a location 0.7m from the flame centerline in this embodiment of the invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0051] The invention will now be described using a diffusion flame of a hydrogen-methane mixed fuel jet as an example.

[0052] like Figure 1 As shown, the present invention provides a method for predicting the size and radiation characteristics of a jet diffusion flame based on machine learning, which specifically includes the following steps:

[0053] S1. Conduct experiments on hydrogen-doped methane jet diffusion flames and collect experimental data from other researchers to construct a parameter library for hydrogen-doped methane jet diffusion flames. Specifically, the flame parameters in the parameter library include initial jet parameters such as nozzle diameter, hydrogen-doped methane mixing ratio, outlet pressure, and outlet velocity, as well as flame size parameters such as flame height, width, and lift height.

[0054] S2, systematically analyze the characteristics of flame size parameters and construct an empirical model of jet diffusion flame size.

[0055] Specifically, the jet diffusion flame height model can be expressed as:

[0056]

[0057] Where H and d represent the flame height and nozzle diameter, respectively; ρ0 and ρ∞ represent the density of hydrogen-doped methane and the density of air, respectively; and u0 and u∞ represent the jet velocity of hydrogen-doped methane and the local speed of sound, respectively.

[0058] S3. Construct a machine learning model. Combine the flame parameters in the hydrogen-doped methane jet diffusion flame parameter library established in step S1 and the jet diffusion flame size empirical model established in step S2 to train the constructed machine learning model, obtain the optimal flame size prediction model, and save the corresponding weights.

[0059] Specifically, the sample data from the parameter library of hydrogen-doped methane jet diffusion flame established in step S1 is input; the sample data is standardized to accelerate convergence, improve model stability, and prevent overfitting; a backpropagation neural network (BPNN) model is constructed, with hidden and output layers built based on the ReLU function; the maximum number of iterations of the neural network is set to 5000, and the learning rate is set to 0.001, with flame height, width, and rise height as target values ​​for training; a loss function is constructed based on the empirical model of jet diffusion flame size established in step S2 for flame height prediction, and the mean squared error loss is used as the loss function for flame width and rise height prediction; the prediction error is calculated using the loss function, and the optimal flame size prediction model is obtained when the error meets the requirements, and the optimal weights are saved. Figure 3 The results of flame height prediction for the test set data are shown, and it can be found that the prediction results are in good agreement with the sample data.

[0060] S4. Determine the initial parameters of the jet: the hydrogen doping ratio of the hydrogen-doped methane is 0.6, the outlet velocity is 609 m / s, the outlet pressure is 0.64 MPa, and the nozzle diameter is 0.006 m. Call the weights saved in step S3 to directly predict the flame size parameters, and obtain the flame height as 1.32 m, the width as 0.19 m, and the lift height as 0.27 m.

[0061] S5. Combine the predicted results of the flame size parameters in step S4 to solve for the flame residence time and flame radiation fraction.

[0062] Specifically, the flame dwell time can be calculated as follows:

[0063]

[0064] Where, ρ f The flame density is represented by W, the flame width by f. s This indicates the mass fraction of hydrogen-doped methane under stoichiometric conditions. This represents the mass flow rate of hydrogen-doped methane; the flame width and flame height are predicted by calling a machine learning model in step S4.

[0065] The flame radiation fraction can be calculated as follows:

[0066]

[0067] Among them, a p and T ad represents Planck's average absorptivity and the temperature of the adiabatic flame, respectively.

[0068] S6. Establish a calculation model for the total radiant power of the flame and a weighted distribution model for the radiation sources.

[0069] Specifically, the total flame radiation power S rad The computational model can be represented as:

[0070]

[0071] Where, ΔH c This indicates the heat of combustion of hydrogen-doped methane.

[0072] The weighted distribution model of the radiation sources can be expressed as:

[0073]

[0074] Among them, w i The weight of the i-th radiation source is represented by N, which represents the number of radiation sources. In this embodiment, N is 50, and 1 ≤ n ≤ N. The radiation sources are evenly distributed along the center line of the flame. The position of the first radiation source is determined based on the predicted value of the flame rise height, and the position of the 50th radiation source is determined based on the predicted value of the flame height.

[0075] S7, Solve for the radiative heat flux, which can be calculated as follows:

[0076]

[0077] Among them, L iτ represents the distance between the monitoring point and the i-th radiation source. i α represents transmittance. i This represents the angle between the normal of the monitoring point and the line connecting the i-th radiation source and the monitoring point.

[0078] Figure 2 The process of constructing, training, and inference of the optimal flame size prediction model of the present invention is demonstrated.

[0079] Figure 4 The study demonstrates the flame radiation distribution predicted using current technology, and the results show that the jet diffusion flame size and radiation characteristics prediction method based on machine learning provided in this invention can accurately predict flame rise. Figure 5 The study presents a comparison between the predicted and experimentally measured values ​​of flame radiation from a hydrogen-doped methane jet at a distance of 0.7 m from the flame centerline. The results show that the flame radiation predicted using current technology is in high agreement with the experimental results.

[0080] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0081] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple steps or stages, which are not necessarily completed at the same time, but may be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but may be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0082] In one embodiment, the present invention provides a computer system, which may be a server. The computer system includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores data used in the methods described above. The network interface communicates with external terminals via a network connection. The computer program is executed by the processor to implement the methods described above.

[0083] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0084] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.

[0085] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for predicting the size and radiation characteristics of a jet diffusion flame based on machine learning, characterized in that, Includes the following steps: Conduct jet diffusion flame experiments, collect experimental data, and construct a jet diffusion flame parameter library; Analyze the characteristics of flame size parameters and construct an empirical model of jet diffusion flame size; A machine learning model is constructed, and the constructed machine learning model is trained by combining the flame parameters in the jet diffusion flame parameter library and the jet diffusion flame size empirical model to obtain the optimal flame size prediction model and save the corresponding weights. Determine the initial parameters of the jet, and then use the weights to directly predict the flame size parameters; By combining the predicted results of flame size parameters, the flame residence time and flame radiation fraction are solved, and a calculation model for total flame radiation power and a weighted distribution model of radiation sources are established. The radiative heat flux is solved based on the total radiative power calculation model of the flame and the weighted distribution model of the radiation source.

2. The method for predicting the size and radiation characteristics of a jet diffusion flame based on machine learning, as described in claim 1, is characterized in that... The flame parameters in the jet diffusion flame parameter library include initial jet parameters and flame size parameters; the initial jet parameters include nozzle diameter, fuel mixing ratio, outlet pressure and outlet velocity, and the flame size parameters include flame height, flame width and flame rise height.

3. The method for predicting the size and radiation characteristics of a jet diffusion flame based on machine learning, as described in claim 1, is characterized in that... The analysis of flame size parameter characteristics and the construction of an empirical model for jet diffusion flame size specifically include: The constructed empirical model for the size of the jet diffusion flame is expressed as follows: Where H and d represent flame height and nozzle diameter, respectively; ρ0 and ρ∞ represent fuel density and air density, respectively; and u0 and u∞ represent fuel jet velocity and local speed of sound, respectively.

4. The method for predicting the size and radiation characteristics of a jet diffusion flame based on machine learning, as described in claim 1, is characterized in that... The construction of the machine learning model involves combining flame parameters from the jet diffusion flame parameter library with the jet diffusion flame size empirical model, training the constructed machine learning model to obtain the optimal flame size prediction model, and saving the corresponding weights. Specifically, this includes: The sample data in the jet diffusion flame parameter library are standardized; a neural network model is constructed; the neural network parameters are determined; a loss function is constructed based on the empirical model of jet diffusion flame size; the standardized sample data is trained, and the prediction error is calculated using the loss function; when the prediction error meets the requirements, the optimal flame size prediction model is obtained, and the optimal weights are saved.

5. The method for predicting the size and radiation characteristics of a jet diffusion flame based on machine learning, as described in claim 1, is characterized in that... The flame size parameters include flame height, flame width, and flame rise height.

6. The method for predicting the size and radiation characteristics of a jet diffusion flame based on machine learning, as described in claim 1, is characterized in that... The calculation of flame dwell time and flame radiation fraction based on the prediction results of flame size parameters specifically includes: The flame dwell time t is: Where, ρ f The flame density is represented by W, the flame width by H, and the flame height by f. s This indicates the mass fraction of fuel under stoichiometric conditions. This indicates the fuel mass flow rate; flame width and flame height are obtained by calling the optimal flame size prediction model. The flame radiation fraction X rad for: Among them, a p and T ad represents Planck's average absorptivity and the temperature of the adiabatic flame, respectively.

7. The method for predicting the size and radiation characteristics of a jet diffusion flame based on machine learning, as described in claim 6, is characterized in that... The establishment of the total flame radiation power calculation model and the radiation source weighted distribution model specifically includes: The total radiant power S of the flame rad The calculation model is as follows: Where, ΔH c Indicates the heat of combustion of fuel; The weighted distribution model of the radiation source is as follows: Among them, w i The weight of the i-th radiation source is represented by N, and the number of radiation sources is represented by 1≤n≤N. The radiation sources are uniformly distributed along the center line of the flame. The position of the first radiation source is determined based on the predicted value of the flame rise height, and the position of the N-th radiation source is determined based on the predicted value of the flame height.

8. The method for predicting the size and radiation characteristics of a jet diffusion flame based on machine learning, as described in claim 7, is characterized in that... The method for solving the radiative heat flux based on the total flame radiant power calculation model and the radiation source weighted distribution model specifically includes: The radiative heat flux q is: Among them, L i τ represents the distance between the monitoring point and the i-th radiation source. i α represents transmittance. i This represents the angle between the normal of the monitoring point and the line connecting the i-th radiation source and the monitoring point.

9. A computer system comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

Images