Gas mixer size optimization method based on deep learning and gas mixer

Abstract

The invention relates to a gas mixing technology, and discloses a gas mixer size optimization method based on deep learning, which comprises the following steps: acquiring design parameters of a gas mixer, and establishing a gas mixer model; performing flow field simulation analysis on the model to obtain performance parameters of the gas mixer under different process parameters; establishing a database for the design parameters, the process parameters and the performance parameters; constructing a neural network prediction model based on the database, updating hyper-parameters of the prediction model based on a reverse update algorithm, obtaining an optimal hidden parameter value, and then obtaining an optimal performance parameter prediction model; and inputting to-be-optimized design parameters and process parameters of the gas mixer into the optimal performance parameter prediction model, and further obtaining optimal design parameters under the optimal performance parameters of the gas mixer based on the model. The invention also discloses a gas mixer. The optimization method is high in accuracy and optimization efficiency, and the gas mixer is simple in structure and good in mixing effect.

Description

Technical Field

[0001] The present invention relates to a gas mixing technology, and in particular to a gas mixer size optimization method based on deep learning and a gas mixer. Background Art

[0002] Gas mixing is a common and critical operation unit in industrial processes. Gas mixing relies on the unique flow characteristics of gas, and uses mechanical stirring and other operation methods or fluid dynamics methods to make the concentration and temperature of the mixed gas system composed of two or more gases uniformly distributed, so that its inhomogeneity is continuously reduced. As a key step in the industrial production process, gas mixing not only affects the smooth progress of the production process, but sometimes even determines the compliance and safety of the exhaust gas. The traditional gas mixer device has a complex internal structure and a relatively large volume, which cannot meet some special requirements. For example, highly corrosive gases and high dust content gases cause great damage to the equipment, causing the equipment to fail to operate normally, accelerating depreciation, and the complex structure of the gas mixer is difficult to maintain. In order to achieve the best mixing effect of the gas mixer under different process parameters, the gas mixer needs to be subjected to flow field simulation analysis and then the various size parameters of the gas mixer are continuously adjusted and optimized according to the simulation results. During this period, multiple iterations are usually required, and each iteration needs to be recalculated, which has low optimization efficiency. Summary of the invention

[0003] The technical problem to be solved by the present invention on the one hand is to provide a gas mixer size optimization method based on deep learning, which has fast optimization speed and high prediction accuracy.

[0004] In addition, another technical problem to be solved by the present invention is to provide a gas mixer, which has a simple structure, simple equipment maintenance and a long service life.

[0005] In order to solve the above technical problems, the present invention provides a gas mixer size optimization method based on deep learning, comprising obtaining design parameters of the gas mixer, and establishing the gas mixer model based on the design parameters; performing flow field simulation analysis on the gas mixer model to obtain performance parameters of the gas mixer under different process parameters; establishing a database for the design parameters, process parameters and performance parameters obtained from simulation tests and actual production tests; constructing a neural network prediction model based on the database, updating the neural network prediction model hyperparameters based on a reverse update algorithm, obtaining the optimal hidden parameter value of the neural network, and then obtaining the optimal performance parameter prediction model, wherein the input parameters of the prediction model include the design parameters and process parameters of the gas mixer, and the output parameters include the performance parameters of the gas mixer; inputting the design parameters and process parameters to be optimized of the gas mixer into the optimal performance parameter prediction model, and further obtaining the optimal design parameters under the optimal performance parameters of the gas mixer based on the model.

[0006] Preferably, the gas mixer model is subjected to flow field simulation analysis to obtain performance parameters of the gas mixer under different process parameters, including: dividing the gas mixer model with a hybrid grid to obtain a grid model; determining a turbulence model for numerical simulation; performing flow field simulation calculations on the grid model using the turbulence model based on Fluent; and visualizing the calculation results to obtain flow field information at different positions of the gas mixer.

[0007] Preferably, establishing a database for the design parameters, process parameters and performance parameters obtained from simulation tests and actual production tests includes standardizing the database and includes the following steps: a. Standardize the original data X in the database to obtain standardized matrix data :

[0008] in, is the mean corresponding to each sample data, is the standard deviation corresponding to each sample data; b. Calculate the standardized matrix data The correlation matrix R is:

[0009] Wherein, N represents the total number of samples in the database; c. Perform eigenvalue decomposition on the correlation matrix R to obtain the eigenvalue and the corresponding eigenvector , where the eigenvalue The length of the vector is consistent with the original data, which is M. The eigenvector yes ×M vector, take the eigenvalue The first m values ​​of the vector make the sum of the first m eigenvalues ​​account for more than 90% of the total variance, and the corresponding ×m feature vector P; d. Calculate the principal components to obtain the transformed new database : .

[0010] Preferably, the design parameters include the inlet air duct diameter, outlet air duct diameter, static mixer side length, mixed gas intake duct diameter, mixed gas intake duct installation angle and mixed gas intake duct insertion depth of the gas mixer.

[0011] Preferably, the process parameters include the components of the gas to be mixed, the air inlet flow rate, the gas inlet flow rate to be mixed and the concentration of each component in the gas to be mixed.

[0012] Preferably, the performance parameter includes mixing inhomogeneity S, and the mixing inhomogeneity S is:

[0013] in, represents the volume fraction of the gas to be mixed at the sampling point, represents the average volume fraction of the gas to be mixed at all sampling points, Indicates the number of sampling points.

[0014] Preferably, the minimum judgment condition for the optimal performance parameter is that the mixing unevenness S<1 .

[0015] Preferably, based on the model, optimal design parameters under the optimal performance parameters of the gas mixer are further obtained, including finding design parameters that satisfy the minimum mixing heterogeneity S through an optimization algorithm.

[0016] The present invention also provides a gas mixer, comprising an inlet air duct, an outlet air duct, a static mixer duct arranged between the inlet air duct and the outlet air duct, and a plurality of gas intake ducts to be mixed, wherein the gas intake ducts to be mixed are suitable for being obliquely inserted into the interior of the static mixer duct and being centrally symmetrically distributed on the static mixer duct.

[0017] Preferably, the central axes of the inlet air duct, the outlet air duct and the static mixer duct coincide, and the cross-sections of the inlet air duct and the outlet air duct are circular, the cross-section of the static mixer duct is square, and the cross-section of the gas intake duct to be mixed is circular.

[0018] Through the above basic technical scheme, the original data in the database of the gas mixer size optimization method based on deep learning provided by the present invention is representative of the data obtained from the flow field simulation test and the actual production test of the gas mixer, and the prediction accuracy of the neural network prediction model constructed by the database is relatively high. The hyperparameters of the prediction model are updated based on the reverse update algorithm, and the prediction model with high precision and high stability can predict the performance parameters of the gas mixer under different process parameters, and then the design parameters of the gas mixer that meet the optimal performance parameters under different process parameters can be obtained, which reduces the time of obtaining the design parameters of the gas mixer through a large number of artificial simulation tests and actual tests, and the obtained test parameters of the gas mixer are accurate and effective.

[0019] Other features and advantages of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The following drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but the protection scope of the present invention is not limited to the following drawings and specific embodiments. In the drawings: Figure 1 is a step diagram of a gas mixer size optimization method based on deep learning of the present invention; Figure 2 It is one of the structural schematic diagrams of a specific embodiment of the gas mixer of the present invention; Figure 3 This is a second structural schematic diagram of a specific embodiment of the gas mixer of the present invention; Figure 4 This is the third structural schematic diagram of a specific implementation of the gas mixer of the present invention.

[0021] Description of Reference Numerals 1- inlet air duct; 2- outlet air duct; 3- static mixer duct; 4- air inlet duct for gas to be mixed. DETAILED DESCRIPTION

[0022] The specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention, and the protection scope of the present invention is not limited to the following specific embodiments.

[0023] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "setting", "connection" and "installation" should be understood in a broad sense. For example, the connection can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection, or an indirect connection through an intermediate medium, or the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0024] In the description of the present invention, it is necessary to understand that the directions or positional relationships indicated by the terms "upper", "lower", "inside" and "outside" are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0025] refer to Figure 1 The present invention provides a gas mixer size optimization method based on deep learning, the method comprising obtaining design parameters of the gas mixer, and establishing a gas mixer model based on the design parameters; performing flow field simulation analysis on the gas mixer model to obtain performance parameters of the gas mixer under different process parameters; establishing a database for the design parameters, process parameters and performance parameters obtained from simulation tests and actual production tests; constructing a neural network prediction model based on the database, updating the neural network prediction model hyperparameters based on a reverse update algorithm, obtaining the optimal hidden parameter value of the neural network, and then obtaining the optimal performance parameter prediction model, wherein the input parameters of the prediction model include the design parameters and process parameters of the gas mixer, and the output parameters include the performance parameters of the gas mixer. Specifically, the optimal performance parameters can be obtained by training the neural network prediction model. The prediction model training steps include setting the number of iterations and the learning rate; in each round of iteration, the training set is traversed batch by batch, the gradient is cleared, forward propagation is performed, the prediction value of the prediction model is obtained, the loss of each batch is calculated, back propagation is performed, the gradient is calculated, and the model parameters are optimized by random gradient descent; after one round of training, the total loss of the iteration is integrated and calculated to evaluate the accuracy of the prediction model on the training set; after repeating the above steps to complete the training of the set number of iterations, the optimal performance parameter prediction model is obtained, the verification set is imported into the prediction model, the generalization performance of the model is verified, and the performance parameter prediction model with the best generalization performance is obtained; the design parameters and process parameters to be optimized of the gas mixer are input into the optimal performance parameter prediction model, and based on the model, the optimal design parameters under the optimal performance parameters of the gas mixer are further obtained.

[0026] Specifically, a flow field simulation analysis is performed on the gas mixer model to obtain the performance parameters of the gas mixer under different process parameters, including: using structured grids and unstructured grids to divide the gas mixer model to obtain a grid model. Specifically, structured grid division can be used in the inlet air duct 1 and the outlet air duct 2, and unstructured grid division can be used in the static mixer duct 3 and the multiple gas inlet ducts 4 to be mixed. In numerical calculations, the number of grids has a key impact on the calculation accuracy and efficiency. Sparse grids often have low calculation accuracy, and the results obtained have a large deviation from the actual situation. However, too fine grids will increase the calculation time and reduce the calculation efficiency. Therefore, in this embodiment, a grid division scheme with a number of about 200,000 grids is selected, taking into account the accuracy and calculation efficiency of numerical simulation calculations. This embodiment is used for gas-gas mixing, and the flow of gas in the mixer is regarded as incompressible flow. The flow process follows the conservation of mass, conservation of energy, conservation of momentum, and the mass conservation equations of the main components in the gas, and adopts the standard turbulent kinetic energy dissipation rate (k- ) model to describe the turbulent flow state and establish a turbulent model. The computational fluid dynamics software Fluent is used to simulate the flow field of the grid model using the turbulence model. In terms of boundary condition setting, the air inlet and the mixed gas inlet are both velocity inlet boundary conditions, the boundary of the mixed gas outlet is the pressure outlet boundary condition, and the inner surface of the static mixer is the no-slip boundary condition. A pressure-based solver is used to solve the problem through the pressure correction algorithm, and the residual convergence accuracy is set to , that is, when the residuals in the calculation process are lower than When the solution converges, the calculation can be regarded as convergent. After the solution converges, the calculation results can be imported into the post-processing software CFD for visualization to obtain the flow field information at different positions of the gas mixer, including the gas velocity, flow rate, gas flow pressure, etc.

[0027] In the step of establishing a database for the design parameters, process parameters and performance parameters obtained from the simulation test and the actual production test, in order to improve the efficiency of subsequent calculations and avoid the collinearity problem of parameter variables in the subsequent regression algorithm, the original data X in the above database can be simplified. In this embodiment, the principal component analysis method is used to standardize the original data X in the database, and the calculation steps are mainly as follows: (1) Standardize the original data X in the database to obtain the standardized matrix data :

[0028] in, is the mean corresponding to each sample data, For the standard deviation corresponding to each sample data, through step (1), each original parameter variable can be standardized so that the mean value of each original parameter variable is 0 and the standard deviation is 1; (2) Calculate the standardized matrix data The correlation matrix R is:

[0029] Where N represents the total number of samples in the database; (3) Perform eigenvalue decomposition on the correlation matrix R to obtain the eigenvalue and the corresponding eigenvector , where the eigenvalue The length of the vector is consistent with the original data, which is M. The eigenvector yes ×M vector, take the eigenvalue The first m values ​​of the vector make the sum of the first m eigenvalues ​​account for more than 90% of the total variance, and the corresponding ×m feature vector P; (4) Calculate the principal components and obtain the transformed new database : .

[0030] Specifically, the design parameters of the gas mixer include the diameter of the inlet air duct, the diameter of the outlet air duct, the side length of the static mixer, the diameter of the gas inlet duct to be mixed, the installation angle of the gas inlet duct to be mixed, the insertion depth of the gas inlet duct to be mixed, the length of the inlet air duct, the length of the outlet air duct, the length of the static mixer, the length of the gas inlet duct to be mixed, etc. The process parameters used in gas mixing include the gas components to be mixed, the air inlet flow rate, the gas inlet flow rate to be mixed, the concentration of each component in the gas to be mixed, the pressure in the static mixer, etc. The above sample data can be normalized, that is, the collected physical quantities are dimensionless and converted into dimensionless index evaluation values. The index values ​​are at the same quantitative level, and comprehensive evaluation analysis can be performed. The above sample data can also be correlated and abnormal data can be eliminated.

[0031] Specifically, a main indicator for measuring the mixing performance of a gas mixer is the mixing uniformity of the gas. In this embodiment, the coefficient of variation of the volume fraction of the gas concentration sample (i.e., mixing non-uniformity S) is used to quantify this indicator. The calculation formula of the mixing non-uniformity S is:

[0032] in, represents the volume fraction of the gas to be mixed at the sampling point, represents the average volume fraction of the gas to be mixed at all sampling points, It represents the number of sampling points, which may be all grid nodes on the outlet cross section of the outlet air duct 2.

[0033] In some embodiments, if the mixing non-uniformity S of the gas mixer is less than 1%, it can be considered that the mixing effect of the gas mixer is excellent, and the design parameters of the gas mixer can be put into use. In the gas mixer size optimization method based on deep learning of the present invention, the mixing non-uniformity S<1% is the minimum judgment condition for the optimal performance parameters, that is, the size parameters and process parameters to be optimized of the gas mixer are input into the prediction model to obtain the predicted performance parameters. Even if the mixing non-uniformity of the predicted performance parameters is lower than the minimum judgment condition of 1%, the gas mixer design parameters when the mixing non-uniformity S is the lowest can be further obtained based on the prediction model. The gas mixer design parameters belong to the design parameters with the best mixing effect. Furthermore, the mixing uniformity C can also be used as a performance parameter indicator. The mixing uniformity C=100%-S, then in the process of optimization, the gas mixer design parameters under the highest mixing uniformity C are obtained.

[0034] In the embodiment where the performance parameter index is the mixing non-uniformity S, the optimization algorithm can be used to find the gas mixer design parameters that meet the minimum mixing non-uniformity S. Specifically, in the process of finding the best performance through the optimization algorithm, it is first determined whether the difference between the current process parameters and the input initial process parameters is within the set error range, and then the corresponding gas mixer design parameters are found when the mixing non-uniformity S is as small as possible. The specific steps are as follows: 1. Define the loss function In this embodiment, in order to obtain the design parameters for the minimum mixing inhomogeneity, the loss function can be defined as: , where n represents the number of sample data in the database, It represents the predicted mixing inhomogeneity based on the optimal performance parameter prediction model and the initial process parameters.

[0035] 2. Initializing the design parameters In this embodiment, the initial values ​​of the design parameters are set as the values ​​of the design parameters to be optimized.

[0036] 3. Set up the Adam optimizer: Initialize the Adam optimizer, set the design parameters to be optimized as the variables to be optimized, and set hyperparameters such as the learning rate.

[0037] 4. Iterative optimization. In each iteration, the current design parameters and initial process parameters are input into the optimal performance prediction model to obtain the predicted mixing inhomogeneity, calculate the gradient of the loss function with respect to the current design parameters, and use the Adam optimization algorithm to update the design parameters. Repeat the iteration until the loss function value is less than the set threshold or the number of iterations reaches the upper limit.

[0038] 5. After the optimization process is completed, the design parameters that minimize the loss function (i.e., mixing inhomogeneity) are obtained.

[0039] Based on the above optimization process, the design parameters of the gas mixer with the best mixing effect under different process parameter conditions can be obtained. There is no need to frequently model the gas mixer and simulate the flow field to obtain the design parameters of the optimal mixing effect, which saves time and cost, and the design parameter optimization effect of the gas mixer is good.

[0040] refer to Figures 2 to 4 The present invention also provides a gas mixer, which includes an inlet air duct 1, an outlet air duct 2, a static mixer duct 3 disposed between the inlet air duct 1 and the outlet air duct 2, and a plurality of gas inlet ducts 4 to be mixed, wherein the gas inlet ducts 4 to be mixed are suitable for being obliquely inserted into the static mixer duct 3 and being centrally symmetrically distributed on the static mixer duct 3. Preferably, the plurality of gas inlet ducts 4 to be mixed are axially inclined 45° forward and inserted into the static mixer duct 3 at an angle of 45° clockwise in the radial direction, and a part of the gas inlet ducts 4 to be mixed is located in the static mixer duct 3, so that when the gas to be mixed is provided to the gas inlet duct 4 to be mixed and enters the static mixer duct 3, it can generate a counterflow and swirl with the incoming air in the horizontal direction, thereby being fully and evenly mixed with the air, and the gas inlet duct 4 to be mixed inserted into the static mixer duct 3 also has a turbulent effect on the incoming air coming from the inlet air duct 1. All of the above-mentioned pipelines can be formed by fillet welding using nickel-based welding rods. Furthermore, the sizes of the inlet air pipeline 1 and the outlet air pipeline 2 can be set to be the same as the size of the delivery pipeline, thereby replacing the original delivery pipeline and saving project investment costs.

[0041] In some embodiments, the central axes of the inlet air duct 1, the outlet air duct 2 and the static mixer duct 3 coincide, and the cross-sections of the inlet air duct 1 and the outlet air duct 2 are circular, the cross-section of the static mixer duct 3 is square, and the cross-section of the gas inlet duct 4 to be mixed is circular. Preferably, the length of the inlet air duct 1 may be three times the diameter of the inlet air duct 1, the length of the outlet air duct 2 may be three times the diameter of the outlet air duct 2, the length of the static mixer duct 3 may be ten times the side length of the static mixer 3, the length of the gas inlet duct 4 to be mixed may be three times the diameter of the gas inlet duct 4 to be mixed, and the ratio of the length of the gas inlet duct 4 to be mixed that penetrates into the static mixer duct 3 to the diameter of the gas inlet duct 4 to be mixed is 1 / 1.1.

[0042] As a specific embodiment of the present invention, the gas mixer is used to mix gas and air, the air inlet flow rate is 15m / s, the gas inlet flow rate is 15m / s, the temperature is room temperature, and the methane in the gas is The concentration is 5%. The above process parameters and the gas mixer design parameters to be optimized are input into the optimal performance parameter prediction model, and the gas mixer design parameters under the optimal performance parameters are further obtained based on the model. The gas mixer design parameters are preferably: the inlet air duct diameter D=1.9m, length 3D=5.7m; the outlet air duct diameter D=1.9m, length 3D=5.7m; the static mixer duct side length L=2.2m, length 10L=22m; the inlet duct diameter d=1.1m, length 3d=3.3m of the gas to be mixed, and the inlet duct 4 of the gas to be mixed is inserted into the static mixer duct 3 at an angle of 45° axially forward and 45° radially clockwise.

[0043] Based on the above gas mixer design parameters and process parameters, flow field simulation verification was carried out to obtain the gas mixer flow field simulation calculation results shown in Table 1. It can be seen from the table that the outlet methane The non-uniformity of concentration distribution is 0.91%, which meets the minimum judgment condition for the optimal mixing effect. The optimization method has high accuracy.

[0044] Table 1 Calculation results of gas mixer flow field simulation

[0045] The gas mixer of the present invention is not limited to the mixing of the above-mentioned gaseous gas and air, but can also be used for the mixing of methane gas and hydrogen, the mixing of nitrogen and hydrogen, etc. The gas mixer of the present invention has a simple structure, a small volume, simple equipment maintenance and a long service life. When the gas mixer of the present invention is used for gas mixing, the resistance in the tube is small, the mixing efficiency is high and the mixing uniformity is high.

[0046] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited thereto. Within the technical concept of the present invention, the technical solution of the present invention can be subjected to a variety of simple modifications, including the combination of various specific technical features in any appropriate manner. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations. However, these simple modifications and combinations should also be regarded as the contents disclosed by the present invention and belong to the protection scope of the present invention.

[0047] In addition, various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.

Claims

1. A gas mixer size optimization method based on deep learning, characterized in that: include: Acquiring design parameters of a gas mixer, and establishing a model of the gas mixer based on the design parameters; Performing flow field simulation analysis on the gas mixer model to obtain performance parameters of the gas mixer under different process parameters; Establishing a database for the design parameters, process parameters and performance parameters obtained from simulation tests and actual production tests; A neural network prediction model is constructed based on the database, and hyperparameters of the neural network prediction model are updated based on a reverse update algorithm to obtain optimal hidden parameter values ​​of the neural network, thereby obtaining an optimal performance parameter prediction model, wherein input parameters of the prediction model include design parameters and process parameters of the gas mixer, and output parameters include performance parameters of the gas mixer; The design parameters and process parameters to be optimized of the gas mixer are input into the optimal performance parameter prediction model, and based on the model, the optimal design parameters under the optimal performance parameters of the gas mixer are further obtained.

2. The gas mixer size optimization method based on deep learning according to claim 1, characterized in that: Perform flow field simulation analysis on the gas mixer model to obtain performance parameters of the gas mixer under different process parameters, including: Dividing the gas mixer model by hybrid mesh to obtain a mesh model; Determine the turbulence model for numerical simulation; Based on Fluent, the grid model is used to perform flow field simulation calculation using the turbulence model; The calculation results are processed visually to obtain flow field information at different positions of the gas mixer.

3. The gas mixer size optimization method based on deep learning according to claim 1, characterized in that: Establishing a database for the design parameters, process parameters and performance parameters obtained from the simulation test and the actual production test, including standardizing the database, and including the following steps: a. Standardize the original data X in the database to obtain standardized matrix data : in, is the mean corresponding to each sample data, is the standard deviation corresponding to each sample data; b. Calculate the standardized matrix data The correlation matrix R is: Wherein, N represents the total number of samples in the database; c. Perform eigenvalue decomposition on the correlation matrix R to obtain the eigenvalue and the corresponding eigenvector , where the eigenvalue The length of the vector is consistent with the original data, which is M. The eigenvector yes ×M vector, take the eigenvalue The first m values ​​of the vector make the sum of the first m eigenvalues ​​account for more than 90% of the total variance, and the corresponding ×m feature vector P; d. Calculate the principal components to obtain the transformed new database : 。 4. The gas mixer size optimization method based on deep learning according to any one of claims 1 to 3, characterized in that: The design parameters include the inlet air duct diameter, outlet air duct diameter, side length of the static mixer, diameter of the gas inlet duct to be mixed, installation angle of the gas inlet duct to be mixed, and insertion depth of the gas inlet duct to be mixed.

5. The gas mixer size optimization method based on deep learning according to any one of claims 1 to 3, characterized in that: The process parameters include the components of the gas to be mixed, the air inlet flow rate, the inlet flow rate of the gas to be mixed, and the concentration of each component in the gas to be mixed.

6. The gas mixer size optimization method based on deep learning according to any one of claims 1 to 3, characterized in that: The performance parameters include mixing non-uniformity S, which is: in, represents the volume fraction of the gas to be mixed at the sampling point, represents the average volume fraction of the gas to be mixed at all sampling points, Indicates the number of sampling points.

7. The gas mixer size optimization method based on deep learning according to claim 6, characterized in that: The minimum judgment condition for the optimal performance parameter is that the mixing non-uniformity S is less than 1%.

8. The gas mixer size optimization method based on deep learning according to claim 7, characterized in that: Based on the model, the optimal design parameters under the optimal performance parameters of the gas mixer are further obtained, including finding the design parameters that meet the minimum mixing non-uniformity S through an optimization algorithm.

9. A gas mixer, characterized in that: The invention comprises an inlet air duct (1), an outlet air duct (2), a static mixer duct (3) arranged between the inlet air duct (1) and the outlet air duct (2), and a plurality of to-be-mixed gas intake ducts (4), wherein the to-be-mixed gas intake ducts (4) are suitable for being obliquely inserted into the interior of the static mixer duct (3) and being centrally symmetrically distributed on the static mixer duct (3).

10. The gas mixer according to claim 9, characterized in that The central axes of the inlet air duct (1), the outlet air duct (2) and the static mixer duct (3) coincide with each other, and the cross-sections of the inlet air duct (1) and the outlet air duct (2) are circular, the cross-section of the static mixer duct (3) is square, and the cross-section of the to-be-mixed gas intake duct (4) is circular.

Images