Smelting rotational flow dry powder spray gun design method and system based on machine learning

Abstract

The invention relates to the technical field of immersed top-blowing spray guns, and discloses a machine learning-based smelting rotational flow dry powder spray gun design method and system, and the method comprises the steps: optimizing the structural parameters of a rotational flow chaos dry powder spray gun through the combination of CFD-DEM and ANN technologies; performing simulation calculation on the rotational flow chaos dry powder spray gun under different structural parameter configurations, and extracting feature data; training an artificial neural network model according to the feature data; predicting a target parameter by the artificial neural network model, and performing selection, intersection and variation by applying a genetic algorithm (GA) according to the difference between a predicted value and a target parameter value to obtain a parameter with optimal fitness; and carrying out analogue simulation according to the optimal fitness parameter. According to the method, the influences of structural parameters such as the blade height, the blade rotation angle, the blade installation height, the CH4 pipe diameter and the mixing area length of the spray gun on the rotational flow intensity and the turbulence intensity are clear, the optimal parameter combination is found on the basis, the performance of the spray gun is improved, and the service life of the spray gun is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of submerged top-blowing spray guns, and in particular to a design method and system for a smelting cyclone dry powder spray gun based on machine learning. Background Art

[0002] The top-submerged lance (TSL) process is an immersion injection method that injects a high-flow jet or jet-particles from a top-blown lance into a high-temperature molten pool through a specially designed vertically mounted lance. The jet-particle coaxial swirl has attracted much attention in various industrial processes due to its advantages such as low load and good momentum / heat transfer efficiency. In the smelting process, the submerged top-blown lance equipped with a cyclone converts the axial jet into a chaotic swirl and then injects the entrained particles into the high-temperature molten pool. The particles in the reactor move in a spiral motion under the disturbance of the swirl, which enhances the momentum and mass transfer performance between phases and improves the melting and reaction rate of the particles. However, in actual application, the tangential entrainment of the gas phase on the solid particles and the four-way coupling between the phases make this instantaneous complex flow have strong anisotropic behavior, and the chaotic nonlinear characteristics of the spiral motion gas-solid two-phase flow lead to the design of industrial equipment and process optimization relying largely on practical experience. Therefore, a comprehensive understanding of the motion mechanism, flow pattern, gas phase cyclone intensity, particle dispersion, and turbulent flow structure of coaxial gas-solid cyclonic flow is beneficial to improving the momentum, heat, and mass transfer efficiency between the gas and solid phases and optimizing the reactor device.

[0003] At present, the study of coaxial jet-particle swirl mainly relies on experimental measurements to explore its flow shape and particle behavior, and analyze the factors affecting the swirl intensity and particle dispersion characteristics. However, due to the complex local and instantaneous details of multiphase information in industrial-scale reactors, high measurement costs and long cycles, many scholars have begun to use computational fluid dynamics (CFD) and discrete element method (DEM) to couple particle-particle or particle-wall interactions. Studies have found that swirl entrains particles to accelerate motion and improve the operating performance of the reactor, while the swirl intensity directly affects the motion and dispersion characteristics of particles in the reactor. The nonlinear motion characteristics of the swirl lead to multi-scale turbulent motion characteristics in the reactor, and changing the cyclone structure will make this turbulent motion more complicated. This method provides in-depth and comprehensive insights into the basic physical laws of gas-solid swirl systems.

[0004] In the above description, although CFD-DEM coupling technology can simulate particle-fluid interaction and particle motion trajectory in detail, when it deals with highly nonlinear and complex multiphase flow problems, CFD-DEM simulation may need to consume a large amount of computing resources and time, which limits its application efficiency in design optimization and still has limitations. Therefore, the present invention introduces the artificial neural network (ANN) technology in machine learning and combines it with CFD-DEM to deeply understand the influence of the geometric structure of the swirl chaotic dry powder spray gun on the swirl intensity and turbulence intensity of the spray gun, and establishes a nonlinear mapping relationship between the two, so as to optimize the structure of the spray gun and improve its performance and service life.

[0005] Prior art 1, Chinese patent, application number: CN201910751432.9 discloses a Laval jet submerged top blowing spray gun, which relates to the field of non-ferrous metal smelting technology, including a hollow gun body, the gun body is cylindrical, one end of the gun body is a closed end, and the other end is an open end, characterized in that the inner cavity of the gun body is divided into a mixing section, a conveying section, a swirl section and a turbulent section from the closed end to the open end, a fuel pipe coaxially penetrates into the gun body from the center of the closed end and extends into the turbulent section, a pressure pipe is arranged on the outer sleeve of the fuel pipe in the gun body, an end pressure sampler is installed outside the gun body at the closed end, the end pressure sampler is inserted into the gun body and connected to the pressure pipe; at least two fluid interfaces are opened on the side wall of the gun body in the mixing section; at least one Laval injection device is installed in the gun body, and the Laval injection device is located between any two adjacent fluid interfaces; a swirler is installed in the gun body in the swirl section. Although it has a simple structure, uniform fluid mixing, safety and reliability, and a long service life, it consumes a lot of computing resources and time when dealing with highly nonlinear and complex multiphase flow problems.

[0006] Prior art 2, Chinese patent, application number: CN202110223827.9 discloses an immersed top-blown converter, including a top-blown converter body, a material inlet and outlet are provided on the top of the top-blown converter body, and an immersed top-blown lance is inserted into the material inlet and outlet. The furnace has high reaction efficiency. By smelting in an immersed top-blown molten pool, materials can be continuously added to the furnace, reducing the shutdown time and improving the smelting operation rate. Since the immersed top-blown lance can be directly inserted into the molten pool, and a cyclone is provided in the immersed top-blown lance, the immersed top-blown lance stirs the molten pool violently, the materials are mixed more evenly, and the reaction efficiency in the furnace is improved. The top-blown converter is provided with a detection rod, which can sample and analyze the melt at any time, reduce the number of shutdown sampling times, keep the furnace temperature constant, and extend the furnace life. The support of the top-blown converter adopts a combination structure of a support ring and a roller, and adopts a large and small gear transmission. Although it makes the transmission more stable and overcomes the vibration impact of the furnace body, it also weakens the stirring ability of the liquid phase area.

[0007] Prior art three, Chinese patent, application number: CN202010588955.9 discloses a modification method of a top-blown smelting facility, a continuous copper smelting facility and a method. The modification method of the top-blown smelting facility includes: canceling the converter or converting furnace of the top-blown converting facility; reducing the furnace height of the top-blown smelting furnace of the top-blown smelting facility; canceling the single submerged top-blown main lance of the top-blown smelting furnace; setting at least three top-blown smelting lances spaced from each other on the top of the top-blown smelting furnace; adding a copper matte discharge port to the top-blown smelting furnace; canceling the electrodes and electrode supporting devices of the sedimentation electric furnace of the top-blown smelting facility, and setting at least three top-blown converting lances spaced from each other on the top of the sedimentation electric furnace to transform the sedimentation electric furnace into a top-blown converting furnace; connecting the copper matte discharge port of the top-blown smelting furnace with the copper matte inlet of the top-blown converting furnace through a flow channel. Although the transformation method of the top-blowing smelting facility can make full use of the existing top-blowing smelting facilities, it can also be beneficial to the output of high-grade copper matte, create conditions for continuous blowing, and reduce the investment cost of continuous copper smelting, but it will cause the friction coefficient between the annular tube wall of the spray gun and the outer wall of the center tube to increase rapidly, thereby resulting in a short life of the spray gun.

[0008] At present, when dealing with highly nonlinear and complex multiphase flow problems, the prior art one, the prior art two and the prior art three need to consume a lot of computing resources and time, and also weaken the stirring ability of the liquid phase area and cause the friction coefficient between the annular tube wall and the outer wall of the center tube of the spray gun to increase rapidly, thereby resulting in a short life of the spray gun. In order to solve the above problems, the present invention provides a smelting cyclone dry powder spray gun design method and system based on machine learning. Summary of the invention

[0009] The main purpose of the present invention is to provide a smelting cyclone dry powder spray gun design method and system based on machine learning, so as to solve the problem that when dealing with highly nonlinear and complex multiphase flow problems in the prior art, a large amount of computing resources and time are consumed, the stirring ability of the liquid phase area is weakened, and the friction coefficient between the annular tube wall and the outer wall of the center tube of the spray gun increases rapidly.

[0010] To achieve the above object, the present invention provides the following technical solutions:

[0011] A method for designing a smelting cyclone dry powder spray gun based on machine learning, wherein the method for optimizing structural parameters of a molten pool smelting cyclone chaotic dry powder spray gun comprises:

[0012] Use CFD-DEM coupling software to simulate and calculate the swirl chaotic dry powder spray gun under different structural parameter configurations; extract the key characteristic parameters of the flow field in the CFD-DEM coupling software, calculate the ratio of the tangential momentum to the axial momentum of the fluid through the key characteristic parameters, and form the target parameters;

[0013] Match the flow field characteristic parameters with the target parameters, construct a data set, and clean the data set; divide the data set into a training set and a validation set, train the artificial neural network model based on the training set; and use the validation set to evaluate the trained artificial neural network model;

[0014] Analyze the impact of each hyperparameter in the artificial neural network model on the model performance; determine the structural parameters that have the greatest impact on the target parameters and set the parameter combination; calculate the target parameters of the parameter combination, select individuals for crossover and mutation based on the calculation results, and form the optimal fitness parameter combination;

[0015] Calculate the maximum number of iterations or fitness function convergence. If so, perform simulation verification on the new model combination. If not, continue with crossover and mutation.

[0016] As a further improvement of the present invention, the process of forming the target parameters includes:

[0017] The swirl chaotic dry powder spray gun is locally encrypted, and the key positions such as the swirl blade boundary and the spray gun outlet are simulated and calculated using CFD-DEM coupling software for the swirl chaotic dry powder spray gun under different structural parameter configurations;

[0018] Extract the key characteristic parameters of the flow field in CFD-DEM software;

[0019] Among them, the key characteristic parameters include the density, coordinates, and tangential velocity U of each data point. tan , axial speed U axi , the inner diameter of the annular tube of the spray gun R, the range from the annular zone to the mixing zone is from R0 to R1, the characteristic length of the spray gun L, the velocity u of the jet and the integrated volume V;

[0020] The ratio of the fluid's tangential momentum flux to its axial momentum flux, SN, and the turbulence intensity of the jet in the spray gun are calculated by using the parameters.

[0021] As a further improvement of the present invention, the process of calculating the ratio SN of the tangential momentum flux to the axial momentum flux of the fluid and the turbulence intensity of the jet in the spray gun includes:

[0022] Extract the key characteristic parameters of the flow field in CFD-DEM software;

[0023] Calculate the ratio of the tangential momentum flux to the axial momentum flux of the fluid, SN; the equation for SN is:

[0024]

[0025] Where, SN local It represents the ratio of tangential momentum flux to axial momentum flux, and it represents the momentum distribution characteristics of the local area. angular,localIt represents the tangential momentum flux in the local area, which means the momentum transfer of the fluid in the tangential direction. linear,local It represents the axial momentum flux of the local area, which means the momentum transfer of the fluid in the axial direction. R represents the radius, which usually refers to the characteristic radius of the flow field or the radius of a specific area. ρ represents the density of the fluid, which means the mass of the fluid per unit volume. r represents the radial distance, which means the distance from the center of the flow field to a certain point. U axi Indicates the axial velocity, which indicates the velocity component of the fluid in the axial direction, U tan represents the tangential velocity, which represents the velocity component of the fluid in the tangential direction, R0 represents the lower limit of integration, which usually represents the inner boundary radius of the flow field, and R1 represents the upper limit of integration, which usually represents the outer boundary radius of the flow field;

[0026] The turbulence intensity of the jet in the spray gun introduced into the gas phase Reynolds number Re and turbulence Reynolds TRe are used to evaluate the turbulence intensity of the jet in the spray gun; where Re is defined as the ratio of inertia force to friction force, and TRe is defined as the turbulent kinetic energy k of the fluid. g The ratio of the frictional force to the work done W is expressed as:

[0027]

[0028]

[0029] In the formula, Re is the Reynolds number, which represents the ratio of inertial force to viscous force and is used to evaluate the turbulence intensity of the fluid, and ρ g Indicates the density of the gas phase fluid, which indicates the mass per unit volume of the gas phase fluid, u g It represents the characteristic velocity of the gas phase fluid, which means the velocity of the gas phase fluid at a certain characteristic length. L represents the characteristic length, which means the characteristic size of the flow field, usually the width or height of the flow field. μ g It represents the dynamic viscosity of gas phase fluid, which represents the viscous characteristics of gas phase fluid. TRe represents the turbulent Reynolds number, which represents the ratio of the turbulent kinetic energy of fluid to the work done by friction, and is used to evaluate the intensity of turbulence. g represents the turbulent kinetic energy of the gas phase fluid, represents the energy of turbulent motion in the gas phase fluid, V represents the volume of the fluid, represents the total volume of the fluid in the flow field, represents the mean velocity square, represents the mean square value of the fluid velocity, represents the square of the average velocity, which represents the square value of the average velocity of the fluid, W represents the work done by friction, which represents the work done by friction in the fluid, μ represents the dynamic viscosity of the fluid, which represents the viscous characteristics of the fluid, represents the velocity gradient, which indicates the rate of change of fluid velocity in a certain direction, n v Represents the normal vector, which represents the unit vector in a certain direction in the flow field. i ′u j ​′> represents the Reynolds stress tensor, which indicates the correlation of velocity pulsation in the fluid.

[0030] As a further improvement of the present invention, the equation for evaluating the trained artificial neural network model using the validation set includes:

[0031] Match the flow field characteristic parameters with the target parameters, construct the data set, clean the data set, etc.; divide the data set into 80% training set and 20% validation set;

[0032] Use the grid search method to adjust the artificial neural network model and select the artificial neural network model with the smallest loss function for training; train the artificial neural network model according to the training set, adjust the network parameters through the back propagation algorithm, and use the early stopping strategy to avoid overfitting;

[0033] Use the validation set to evaluate the trained artificial neural network and calculate the error index of the artificial neural network model on the validation set; set a threshold, and if it exceeds the threshold, readjust the artificial neural network model; if it is below the threshold, perform feature extraction and analysis.

[0034] As a further improvement of the present invention, the process of constructing a data set includes:

[0035] Check the missing values ​​in the data set; if missing values ​​occur, use mean filling as follows:

[0036] j′=1

[0037]

[0038] In the formula, x i′ refers to the value at the location of the missing value, n′ represents the total number of non-missing values ​​in the data set, and x i′j′ represents the j′th eigenvalue of the i′th sample or data point;

[0039] Draw box plots of each institution's parameters and process characteristic values; calculate the upper quartile, lower quartile and interquartile range of each parameter, and determine the upper and lower thresholds; mark values ​​below the lower limit or above the upper limit as outliers; analyze outliers by methods such as replacement or deletion;

[0040] The extracted structural parameters and flow field characteristics are standardized; the standardization formula is as follows:

[0041]

[0042] Among them, x norm is the standardized data, x is the original data, μ is the mean of the data, and σ is the standard deviation of the data.

[0043] As a further improvement of the present invention, the process of training the artificial neural network model with a training set includes:

[0044] Design the network structure of the artificial neural network model, select ReLU as the activation function, and select Huner loss as the loss function; input the training set into the artificial neural network model, and use the Adam optimizer to adjust the network weights and biases of the artificial neural network model;

[0045] Among them, the network structure of the artificial neural network model includes an input layer (including 7 input features: blade length, blade rotation angle, blade installation height, blade tip clearance, blade thickness, CH4 tube diameter, and mixing zone length), several hidden layers, and an output layer (including 3 output features: SN, Re, and TRe);

[0046] Determine the range of values ​​of grass parameters such as learning rate, number of hidden layers, number of nodes, number of batches and batch size; form all hyperparameters into a grid; for each hyperparameter combination in the grid, use the training set to train the artificial neural network model;

[0047] The validation set is used to verify the mean square error, accuracy, etc. of the artificial neural network model after each training, and a threshold is set. The artificial neural network model corresponding to the hyperparameter combination with the smallest error compared to the threshold is selected as the final artificial neural network model.

[0048] As a further improvement of the present invention, the process of forming the optimal fitness parameter combination includes:

[0049] The SHAP importance evaluation method is used to analyze the influence of each structural parameter in the artificial neural network model on the target parameter; through the analysis of the importance of the target feature, the key structural parameters with the greatest influence on the target parameter are determined, and the parameter combination is set. The combination is input into the trained artificial neural network model to predict the corresponding target parameter;

[0050] The gap between the target parameters and the actual parameters predicted by the artificial neural network model is adjusted, and the parameter combination is randomly generated by the genetic algorithm. The optimization is continuously iterated until the maximum number of iterations is reached or the function converges, and finally the optimal parameter combination for fitness is obtained.

[0051] The optimal fitness parameter combination is selected for CFD-DEM simulation verification, and the verified performance is tested and corrected.

[0052] As a further improvement of the present invention, the process of predicting the corresponding target parameter includes:

[0053] Suppose that a training data set of an artificial neural network is collected, and the data includes input features and corresponding target parameters; after the model training is completed, the SHAP method is used to calculate the contribution value of each input feature to the target parameter; a contribution value standard is set, and a comparison is performed to select features with a contribution value higher than the standard;

[0054] According to the contribution values ​​obtained by comparison, the importance of the features is sorted; the higher the contribution value, the greater the influence of the feature on the target parameter, and the higher the relative importance; according to the sorting results, the key feature ranked first is selected;

[0055] The selected key feature combination is input into the trained artificial neural model to predict the target parameters; the actual parameters are compared with the predicted parameters, and if they are higher than a certain threshold, retraining and prediction are performed; the above steps are repeated until the maximum number of iterations or the preset standard is reached.

[0056] As a further improvement of the present invention, the process of randomly generating an initial structural parameter population using a genetic algorithm includes:

[0057] Through feature importance analysis, the key structural parameters that have the greatest impact on the target parameters are determined, and the parameter combination is set. The parameters are input into the trained artificial neural network model to predict the corresponding target parameters;

[0058] The gap between the output target parameters and the actual target parameters is predicted by the artificial neural model, and the structural parameters are adjusted to minimize the gap;

[0059] Randomly generate the initial structural parameter population, and predict the SN, Re and TRe of each parameter combination through the artificial intelligence network model; calculate the fitness function of each combination, that is, the difference between the predicted SN, Re and TRe values ​​and the target SN, Re and TRe; select individuals with better performance according to the fitness function for crossover and mutation to generate a new parameter population;

[0060] The process of selection, crossover and mutation is repeated, and the optimization is continuously iterated until the maximum number of iterations is reached or the fitness function converges; the parameter combination with the best fitness is finally verified by CFD-DEM simulation; a verification threshold is set, and if it is higher than the threshold, individuals will continue to be selected according to the fitness function for crossover and mutation to generate the parameter combination with the best fitness.

[0061] To achieve the above object, the present invention also provides the following technical solutions:

[0062] A smelting cyclone dry powder spray gun design system based on machine learning, wherein the molten pool smelting cyclone chaotic dry powder spray gun structural parameter optimization method comprises:

[0063] The flow field simulation and data acquisition module is used to simulate the swirl chaotic dry powder spray gun under different architecture parameter configurations using CFD-DEM coupling software, encrypt the grids at the swirl blade boundary and the spray gun outlet before calculation; extract the characteristic parameters of the flow field in the CFD-DEM coupling software; calculate the ratio of the tangential momentum flux to the axial momentum flux of the fluid and the turbulence intensity of the jet in the spray gun based on the density, coordinates, tangential velocity and axial velocity of each data point, and transmit the calculated data to the data cleaning and model training module;

[0064] The data cleaning and model training module is used to match the process thermal parameters with the calculated data to construct a data set; check the missing values ​​in the data set, use the box plot method of data distribution to describe the data distribution and identify outliers; and standardize the extracted structural parameters and flow field characteristics; output the processed training set to the artificial neural network model, and use the grid search method to adjust the model; adjust the network parameters through the back propagation algorithm to reduce the loss function value; use the validation set to evaluate the trained artificial neural network model;

[0065] The model training and verification module is used to determine the key structural parameters with the greatest impact on the target parameters through feature importance analysis, and set parameter combinations; input the set parameter combinations into the trained artificial neural network model to predict the corresponding target parameters; compare the target parameters predicted and output by the artificial neural network model with the actual target parameters to minimize the gap in the structural parameters; randomly generate the initial structural parameter population using a genetic algorithm, and predict the SN, Re and TRe of each parameter combination through an artificial intelligence network model; calculate the fitness function of each combination, that is, the gap between the predicted SN, Re and TRe values ​​and the target SN, Re and TRe; set a fitness function threshold, select the individuals with the smallest difference from the threshold for crossover and mutation, and form a new parameter population; repeat the process of selection, crossover and mutation, and continuously iterate and optimize until the maximum number of iterations is reached or the fitness function converges; perform CFD-DM simulation verification on the best parameter combination with the best fitness; set a verification threshold, and if it is higher than the threshold, continue to select individuals according to the fitness function for crossover and mutation to generate the best parameter combination with the best fitness.

[0066] The present invention combines computational fluid dynamics (CFD) with discrete element method (DEM) coupling technology and artificial neural network (ANN) technology to quickly predict the swirl number (SN), turbulence intensity Reynolds number (Re) and turbulence Reynolds number (TRe) of different structural parameter combinations, effectively shortening the design cycle; then the parameter combination is iteratively optimized through a genetic algorithm (GA), combined with the target parameter value predicted and output by an ANN model, individuals with excellent performance are selected for crossover and mutation based on a fitness function, and the gap between the predicted value and the target value is gradually reduced; not only the optimization efficiency is improved, but also the accuracy and reliability of the result are improved; finally, the optimized parameter combination is subjected to CFD-DEM simulation verification to prove its stability and efficiency, ensuring that the expected performance indicators can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A schematic diagram of the steps of an embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning of the present invention;

[0068] Figure 2 A schematic flow chart of steps for forming target parameters in an embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning of the present invention;

[0069] Figure 3 A schematic flow chart of the steps of calculating the ratio SN of the tangential momentum flux to the axial momentum flux of the fluid and the turbulence intensity of the jet in the spray gun according to an embodiment of the method for designing a smelting cyclone dry powder spray gun based on machine learning of the present invention;

[0070] Figure 4 A schematic flow chart of the steps of evaluating a trained artificial neural network model using a validation set in one embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning according to the present invention;

[0071] Figure 5 A schematic diagram of an embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning of the present invention using a validation set to evaluate a trained artificial neural network model;

[0072] Figure 6 A schematic flow chart of the steps of constructing a data set for an embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning of the present invention;

[0073] Figure 7 A schematic flow chart of the steps of training an artificial neural network model using a training set in one embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning of the present invention;

[0074] Figure 8 A schematic flow chart of the steps of forming an optimal parameter combination for fitness in an embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning of the present invention;

[0075] Fig. 9 A schematic diagram of forming an optimal parameter combination for fitness in an embodiment of a smelting cyclone dry powder spray gun design method based on machine learning of the present invention;

[0076] Fig.10 A schematic flow chart of the steps for predicting corresponding target parameters in an embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning of the present invention;

[0077] Fig.11 A schematic flow chart of the steps of randomly generating an initial structural parameter population using a genetic algorithm in an embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning according to the present invention;

[0078] Fig.12 A schematic flow chart of the steps of performing CFD-DEM simulation verification on the optimal parameters of an embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning of the present invention;

[0079] Fig.13 A schematic flow chart of the steps of running CFD-DEM simulation for an embodiment of a method for designing a smelting cyclone dry powder spray gun based on machine learning of the present invention;

[0080] Fig.14 It is a functional module schematic diagram of an embodiment of a smelting cyclone dry powder spray gun design system based on machine learning of the present invention;

[0081] Fig.15 It is a structural schematic diagram of an embodiment of an electronic device of the present invention;

[0082] Fig.16 It is a schematic structural diagram of an embodiment of the storage medium of the present invention. DETAILED DESCRIPTION

[0083] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0084] The terms "first", "second" and "third" in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0085] Reference to an "embodiment" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0086] like Figure 1 As shown, this embodiment provides an embodiment of a smelting cyclone dry powder spray gun design method based on machine learning. In this embodiment, the smelting cyclone dry powder spray gun design method based on machine learning specifically includes the following steps:

[0087] Step S1: using CFD-DEM coupling software to simulate and calculate the swirl chaotic dry powder spray gun under different structural parameter configurations; extracting key characteristic parameters of the flow field in the CFD-DEM coupling software, calculating the ratio of the tangential momentum to the axial momentum of the fluid through the key characteristic parameters, and forming the target parameters;

[0088] Step S2: Match the flow field characteristic parameters with the target parameters, construct a data set, and clean the data set; divide the data set into a training set and a validation set, train the artificial neural network model according to the training set; and use the validation set to evaluate the trained artificial neural network model;

[0089] Step S3: Analyze the influence of each structural hyperparameter on the target parameter in the artificial neural network model, determine the structural parameter with the greatest influence on the target parameter, and set the parameter combination; calculate the target parameter of the parameter combination, select individuals for crossover and mutation according to the calculation results, and form the optimal parameter combination for fitness;

[0090] Calculate the maximum number of iterations or fitness function convergence. If so, perform simulation verification on the new model combination. If not, continue with crossover and mutation.

[0091] Preferably, steps S1 to S3 of this embodiment can quickly predict the swirl number (SN), turbulence intensity Reynolds number (Re) and turbulence Reynolds number (TRe) of different structural parameter combinations by combining computational fluid dynamics (CFD) and discrete element method (DEM) coupling technology and artificial neural network (ANN) technology, thereby effectively shortening the design cycle; the parameter combination is then iteratively optimized through a genetic algorithm (GA), and the target parameter value predicted and output by the ANN model is combined, and individuals with excellent performance are selected based on the fitness function for crossover and mutation, so as to gradually reduce the gap between the predicted value and the target value; not only the optimization efficiency is improved, but also the accuracy and reliability of the results are improved; finally, the optimized parameter combination is subjected to CFD-DEM simulation verification to prove its stability and efficiency, thereby ensuring that the expected performance indicators can be achieved.

[0092] Therefore, this embodiment significantly improves the design and optimization efficiency of the spray gun, reduces the test cost, and effectively optimizes the performance and durability of the swirl chaotic dry powder spray gun, realizing an efficient and stable smelting process, which has important application value to the smelting industry.

[0093] Furthermore, if Figure 2 As shown, the process of forming the target parameters in step S1 specifically includes the following steps:

[0094] Step S11: locally encrypt the swirl chaotic dry powder spray gun, especially at key locations such as the swirl blade boundary and the spray outlet, and use CFD-DEM coupling software to simulate and calculate the swirl chaotic dry powder spray gun under different structural parameter configurations;

[0095] Step S12: extracting key characteristic parameters of the flow field in CFD-DEM software;

[0096] Among them, the key characteristic parameters include the density of each data point, the coordinates (calculating the distance r between the rotation point and the velocity vector), the tangential velocity U tan , axial speed U axi , the inner diameter R of the annular tube of the spray gun, the range from the annular zone to the mixing zone is from R0 to R1, the characteristic length L of the spray gun, the velocity u of the jet, the integrated volume V, etc.;

[0097] Step S13: Calculate the ratio SN of the tangential momentum flux to the axial momentum flux of the fluid and the turbulence intensity of the jet in the spray gun by using the above parameters.

[0098] Preferably, step S11 of this embodiment can accurately capture the complex flow behavior of the fluid in these areas and improve the accuracy of the simulation by performing local encryption at key positions. The use of CFD-DEM coupling software can simulate the interaction between fluid and particles, which is crucial for understanding the working mechanism of the swirl chaotic dry powder spray gun. Step S12 extracts key characteristic parameters such as density, coordinates, tangential velocity, axial velocity, etc., which can deeply analyze the dynamic characteristics of the flow field and provide basic data for momentum flux calculation and turbulence intensity analysis. Step S13 calculates the ratio SN of the tangential momentum flux to the axial momentum flux to evaluate the influence of the swirl intensity on the fluid motion; and the calculation of turbulence intensity understands the instability of the fluid, which is of great significance for optimizing the spray gun design and improving the injection efficiency.

[0099] Therefore, the significance of this embodiment is that by combining CFD-DEM coupling simulation and extraction of key characteristic parameters, the design of the swirl chaotic dry powder spray gun can be effectively analyzed and optimized to improve its performance and efficiency.

[0100] Furthermore, if Figure 3 As shown, the process of calculating the ratio SN of the tangential momentum flux to the axial momentum flux of the fluid and the turbulence intensity of the jet in the spray gun in step S13 specifically includes the following steps:

[0101] Step S131: extracting key characteristic parameters of the flow field in CFD-DEM software;

[0102] Step S132: Calculate the ratio SN of the tangential momentum flux to the axial momentum flux of the fluid; the equation of SN can be expressed as:

[0103]

[0104]

[0105] Where, SN local It represents the ratio of tangential momentum flux to axial momentum flux, and indicates the momentum distribution characteristics of the local area. angular,local It represents the tangential momentum flux in the local area, which means the momentum transfer of the fluid in the tangential direction. linear,local It represents the axial momentum flux of the local area, which means the momentum transfer of the fluid in the axial direction. R represents the radius, which usually refers to the characteristic radius of the flow field or the radius of a specific area. ρ represents the density of the fluid, which means the mass of the fluid per unit volume. r represents the radial distance, which means the distance from the center of the flow field to a certain point. U axiIndicates the axial velocity, which indicates the velocity component of the fluid in the axial direction, U tan represents the tangential velocity, which represents the velocity component of the fluid in the tangential direction, R0 represents the lower limit of integration, which usually represents the inner boundary radius of the flow field, and R1 represents the upper limit of integration, which usually represents the outer boundary radius of the flow field;

[0106] Step S133: the turbulence intensity Reynolds number (Re) and turbulence Reynolds number (TRe) of the jet in the spray gun into which the gas phase is introduced are used to evaluate the turbulence intensity of the jet in the spray gun;

[0107] Among them, Re is defined as the ratio of inertial force to friction force, and TRe is defined as the turbulent kinetic energy of the fluid (k g ) to the work done by friction (W), which can be expressed as:

[0108]

[0109] In the formula, Re is the Reynolds number, which represents the ratio of inertial force to viscous force and is used to evaluate the turbulence intensity of the fluid, and ρ g Indicates the density of the gas phase fluid, which indicates the mass per unit volume of the gas phase fluid, u g It represents the characteristic velocity of the gas phase fluid, which means the velocity of the gas phase fluid at a certain characteristic length. L represents the characteristic length, which means the characteristic size of the flow field, usually the width or height of the flow field. μ g It represents the dynamic viscosity of gas phase fluid, which represents the viscous characteristics of gas phase fluid. TRe represents the turbulent Reynolds number, which represents the ratio of the turbulent kinetic energy of fluid to the work done by friction, and is used to evaluate the intensity of turbulence. g represents the turbulent kinetic energy of the gas phase fluid, represents the energy of turbulent motion in the gas phase fluid, V represents the volume of the fluid, represents the total volume of the fluid in the flow field, represents the mean velocity square, represents the mean square value of the fluid velocity, represents the square of the average velocity, which represents the square value of the average velocity of the fluid, W represents the work done by friction, which represents the work done by friction in the fluid, μ represents the dynamic viscosity of the fluid, which represents the viscous characteristics of the fluid, represents the velocity gradient, which indicates the rate of change of fluid velocity in a certain direction, n v Represents the normal vector, which represents the unit vector in a certain direction in the flow field. i ′u j ′> represents the Reynolds stress tensor, which indicates the correlation of velocity pulsation in the fluid.

[0110] ​Preferably, in step S131, the key characteristic parameters of the flow field are extracted in the CFD-DEM software. By combining computational fluid dynamics (CFD) and discrete element method (DEM), the behavior of the fluid in a complex geometric structure can be simulated and analyzed more accurately; the microscopic and macroscopic characteristics of the fluid can be captured. In step S132, the ratio SN of the tangential momentum flux to the axial momentum flux of the fluid is calculated. By calculating the momentum flux of the fluid in different directions, the flow characteristics and vortex structure of the fluid can be better understood; it is helpful to evaluate the turbulence intensity and flow stability of the fluid, which is of great significance for optimizing the design of the fluid system. In step S133, the turbulence intensity Reynolds number (Re) and turbulent Reynolds number (TRe) of the jet in the spray gun of the gas phase are introduced to evaluate the turbulence intensity of the jet in the spray gun. By introducing a turbulence model, the turbulence characteristics of the jet in the spray gun can be accurately predicted and analyzed. The selection and parameter setting of the turbulence model are crucial to the accuracy of the simulation results. Different models show different performances under different conditions.

[0111] Therefore, the significance of this embodiment lies in that by combining CFD and DEM technology, the behavior of fluids in complex geometric structures can be analyzed and predicted more comprehensively and accurately, thereby providing a scientific basis for engineering design and optimization.

[0112] Furthermore, if Figure 4 As shown, the process of evaluating the trained artificial neural network model with the validation set in step S2 specifically includes the following steps:

[0113] Step S21: matching the flow field characteristic parameters with the target parameters, constructing a data set, and performing operations such as cleaning the data set; dividing the data set into 80% training set and 20% validation set;

[0114] Step S22: using a grid search method to adjust the artificial neural network model, selecting the artificial neural network model with the smallest loss function for training; training the artificial neural network model according to the training set, adjusting the network parameters through a back propagation algorithm, and using an early stopping strategy to avoid overfitting;

[0115] Step S23: Use the validation set to evaluate the trained artificial neural network and calculate the error index of the artificial neural network model on the validation set; set a threshold, if it exceeds the threshold, readjust the artificial neural network model; if it is lower than the threshold, perform feature extraction and analysis.

[0116] Preferably, in step S21 of this embodiment, by matching the flow field characteristic parameters with the target parameters, a data set is constructed to ensure the accuracy and consistency of the data. The data cleaning operation can remove noise and outliers, improve data quality, and thus improve the training effect of the model. In step S22, the grid search method is used to adjust the artificial neural network model, and different hyperparameter combinations can be systematically explored. The model with the smallest loss function is selected for training, which can effectively improve the performance of the model. By adjusting the network parameters through the back propagation algorithm, the weights and biases of the model can be optimized to better fit the training data. An early stopping strategy is adopted to avoid overfitting, that is, training is stopped when the error on the validation set no longer decreases, so as to prevent the model from overfitting on the training set, thereby improving the generalization ability of the model. In step S23, the validation set is used to evaluate the trained artificial neural network, and the error index is calculated, so as to objectively evaluate the performance of the model and ensure the performance of the model on unseen data. A threshold is set. If it exceeds the threshold, the artificial neural network model is readjusted. If it is lower than the threshold, feature extraction and analysis are performed. The adaptability and stability of the model on different data sets can be ensured through a dynamic adjustment mechanism. (For specific principles, please refer to the attached Figure 5 ).

[0117] Therefore, the significance of this embodiment is to ensure the accuracy and generalization ability of the artificial neural network model in the prediction of flow field characteristics through means such as data preprocessing, model optimization, verification and adjustment.

[0118] Furthermore, if Figure 6 As shown, the process of constructing the data set in step S21 specifically includes the following steps:

[0119] Step S211: Check the missing values ​​in the data set; if missing values ​​are found, fill them with the mean value as follows:

[0120] j′=1

[0121]

[0122] In the formula, x i′ refers to the value at the location of the missing value, n′ represents the total number of non-missing values ​​in the data set, and x i′j′ represents the j′th eigenvalue of the i′th sample or data point;

[0123] Step S212: draw a box plot of each mechanism parameter and process characteristic value; calculate the upper quartile, lower quartile and interquartile range of each parameter, and determine the upper and lower thresholds; mark the values ​​below the lower limit or above the upper limit as outliers; analyze the outliers by methods such as replacement or deletion;

[0124] Step S213: Standardize the extracted structural parameters and flow field characteristics; the standardization formula is as follows:

[0125]

[0126] Among them, x norm is the standardized data, x is the original data, μ is the mean of the data, and σ is the standard deviation of the data.

[0127] Preferably, in step S121 of this embodiment, missing values ​​in the data set are processed to ensure the integrity of the data and the accuracy of subsequent analysis. The presence of missing values ​​will affect the accuracy of statistical estimation and the effectiveness of subsequent work. Mean filling is a simple and commonly used method for processing missing values, which is suitable for situations with low missing rates. By filling with mean values, the information loss caused by deleting missing tuples can be avoided while keeping the size of the data set unchanged. In step S122, outliers in the data are identified and processed to ensure the quality of the data and the accuracy of the analysis. The box plot is an effective visualization tool that can intuitively display the distribution of data and outliers. By calculating the quartiles and interquartile ranges, the upper and lower thresholds of the data can be determined to identify outliers that deviate from the normal range. The processing method of outliers includes replacement or deletion, depending on the nature of the outliers and the impact on the overall data. In step S123, data of different dimensions are converted to the same dimension to facilitate subsequent comparison and analysis. Standardization processing can eliminate the influence of different data indicator dimensions, so that data can be compared on the same scale, thereby improving the accuracy of data analysis. Normalized data is more suitable for training and evaluating machine learning models, as many models are sensitive to the scale of the data.

[0128] Therefore, the significance of this embodiment lies in that it plays a key role in the data preprocessing process, ensuring the quality of the data and the effectiveness of the analysis.

[0129] Furthermore, if Figure 7 As shown, the process of training the artificial neural network model with the training set in step S22 specifically includes the following steps:

[0130] Step S221: designing the network structure of the artificial neural network model, selecting the activation function and the loss function; inputting the training set into the artificial neural network model, and using the Adam optimizer to adjust the network weights and biases of the artificial neural network model;

[0131] Among them, the network structure of the artificial neural network model includes an input layer (including 7 input features: blade length, blade rotation angle, blade installation height, blade tip clearance, blade thickness, CH4 tube diameter, and mixing zone length), several hidden layers, and an output layer (including 3 output features: SN, Re, and TRe);

[0132] Activation function expression:

[0133]

[0134] In the formula, x is the input value, α is the adaptive leakage coefficient, whose value is dynamically adjusted according to the gradient during the training process, and the initial value is set to 0.01;

[0135] The expression of the loss function is:

[0136] Weighted Huber Lossmulti-scale=∑i=1 n

[0137]

[0138] In the formula, n represents the number of samples, y i represents the actual value of the i-th sample, represents the predicted value of the i-th sample, δ represents the threshold of Huber loss, which is set to 1, K represents the number of scales of multi-scale regularization, and λ k represents the regularization coefficient of the kth scale, which controls the influence of the regularization term, w j represents the jth weight in the model, and m represents the total number of weights in the model;

[0139] Step S222: Determine the value range of hyperparameters such as learning rate, number of hidden layers, number of nodes, number of batches and batch size; form a grid with all hyperparameters; for each hyperparameter combination in the grid, use the training set to train the artificial neural network model;

[0140] Step S223: Use the validation set to verify the mean square error, accuracy, etc. of the artificial neural network model after each training, set a threshold, and select the artificial neural network model corresponding to the hyperparameter combination with the smallest error compared to the threshold as the final artificial neural network model.

[0141] Preferably, in step S221 of the present embodiment, the network structure of the artificial neural network model is designed, the appropriate number of hidden layers and activation function (ReLU) are selected, and Huber loss is used as the loss function. It is ensured that the network can effectively learn and represent complex nonlinear relationships, and the weights and biases of the network are adjusted by the Adam optimizer to optimize the performance of the model. In step S222, hyperparameters such as learning rate, number of hidden layers, number of nodes, number of batches and batch size are determined and optimized. By forming a grid of hyperparameters and training each hyperparameter combination, the optimal hyperparameter combination can be systematically explored and selected, thereby improving the generalization ability and training efficiency of the model. In step S223, the trained model is verified using a validation set, and the artificial neural network model corresponding to the hyperparameter combination with the smallest error is selected as the final model by setting a threshold. This step ensures the performance of the model on unseen data, thereby improving the reliability and practicality of the model.

[0142] Therefore, the significance of this embodiment is to ensure that the artificial neural network model can learn and predict efficiently and accurately through a systematic design, optimization and verification process, so as to perform well in practical applications.

[0143] Furthermore, if Figure 8 As shown, the process of forming the optimal fitness parameter combination in step S3 specifically includes the following steps:

[0144] Step S31: using the SHAP importance evaluation method to analyze the influence of each structural parameter in the artificial neural network model on the target parameter; through the analysis of the importance of the target feature, determine the key structural parameters that have the greatest influence on the target parameter, set the parameter combination, input the combination into the trained artificial neural network model, and predict the corresponding target parameter;

[0145] Step S32: using the artificial neural network model to predict the gap between the target parameters and the actual parameters, adjusting the parameter combination, and using a genetic algorithm to randomly generate an initial structural parameter population; continuously iterating and optimizing until the maximum number of iterations is reached or the function converges;

[0146] Step S33: Select the optimal fitness parameters for CFD-DEM simulation verification, and test and correct the verified performance.

[0147] Preferably, step S31 of this embodiment identifies the key structural parameters that have the greatest impact on the target parameters through SHAP importance analysis. This helps to streamline the model input and improve the interpretability and prediction efficiency of the model. By setting a key parameter combination and inputting it into a trained artificial neural network model, the target parameters can be predicted more accurately, thereby optimizing the performance of the model. In step S32, an initial structural parameter population is randomly generated by a genetic algorithm, and iterative optimization is continuously performed until the maximum number of iterations is reached or the function converges. It can effectively explore the parameter space and find the optimal or near-optimal parameter combination, thereby improving the prediction accuracy and generalization ability of the model. In step S33, CFD-DEM simulation verification and performance test correction are performed to ensure the effectiveness of the optimized parameter combination in practical applications. This embodiment combines theoretical analysis and experimental verification to ensure the reliability and practicality of the model. (For specific principles, refer to the attached Fig. 9 )

[0148] Therefore, the significance of this embodiment lies in improving the prediction accuracy of the artificial neural network model, optimizing parameter selection, enhancing the robustness and adaptability of the model, and promoting technological progress and application expansion, etc., which has important technical significance.

[0149] Furthermore, if Fig.10 As shown, the process of predicting the corresponding target parameter in step S31 specifically includes the following steps:

[0150] Step S311: Collect training artificial neural network data sets, the data includes input features and corresponding target parameters; after model training is completed, use the SHAP method to calculate the contribution value of each input feature to the target parameter; set a contribution value standard, compare, and select features with contribution values ​​higher than the standard;

[0151] Step S312: sorting the contribution values ​​obtained by comparison according to their importance; the higher the contribution value, the greater the influence of the feature on the target parameter, and the higher the relative importance; according to the sorting result, selecting the key feature ranked first;

[0152] Step S313: Input the selected key feature combination into the trained artificial neural model to predict the target parameters; compare the actual parameters with the predicted parameters, and if they are higher than a certain threshold, re-train and predict; repeat the above steps until the maximum number of iterations or the preset standard is reached.

[0153] Preferably, step S311 of this embodiment provides basic data for model training by collecting a training data set, including input features and corresponding target parameters. The SHAP method is used to calculate the contribution value of each input feature to the target parameter, which can quantify the degree of influence of each feature on the model prediction result. By setting the contribution value standard and comparing it, selecting features with contribution values ​​higher than the standard helps to screen out key features that have a greater impact on the target parameter, thereby improving the prediction accuracy and efficiency of the model. Step S312 sorts the features according to the contribution value. The higher the contribution value, the greater the impact of the feature on the target parameter, and the higher the relative importance. Selecting the key features ranked first can focus on the most important features for analysis and prediction. By sorting the importance, the features that have the greatest impact on the target parameter can be identified, and by understanding the decision-making process of the model, a basis for feature engineering and model optimization can be provided. Step S313 inputs the selected key feature combination into the trained artificial neural model to predict the target parameter, and compares the actual parameter with the predicted parameter. If it is higher than a certain threshold, retraining and prediction are performed. Repeat the above steps until the maximum number of iterations or the preset standard is reached. Through iterative training and prediction, the model can be continuously optimized to improve the accuracy and stability of the prediction; at the same time, by setting thresholds and maximum number of iterations, the complexity and time cost of model training can be controlled.

[0154] Therefore, this embodiment provides basic data for model training, and quantifies the importance of features through the SHAP method to screen out key features; importance sorting identifies the features that have the greatest impact on the target parameters, providing a basis for subsequent analysis and prediction; and through iterative training and prediction, continuously optimizes the model, improves the accuracy and stability of the prediction, and controls the complexity and time cost of model training.

[0155] Furthermore, if Fig.11 As shown, the process of randomly generating an initial structural parameter population using a genetic algorithm in step S32 specifically includes the following steps:

[0156] Step S321: determining the key structural parameters that have the greatest impact on the target parameters through feature importance analysis, setting parameter combinations, inputting the parameters into the trained artificial neural network model, and predicting the corresponding target parameters;

[0157] Step S322: predicting the gap between the output target parameter and the actual target parameter through the artificial neural model, and adjusting the structural parameters to minimize the gap;

[0158] Randomly generate the initial structural parameter population, and predict the SN, Re and TRe of each parameter combination through the artificial intelligence network model; calculate the fitness function of each combination, that is, the difference between the predicted SN, Re and TRe values ​​and the target SN, Re and TRe; select individuals with better performance according to the fitness function for crossover and mutation to generate a new parameter population;

[0159] Step S323: Repeat the process of selection, crossover and mutation, and continuously iterate and optimize until the maximum number of iterations is reached or the fitness function converges; finally, the parameter combination with the best fitness is verified by CFD-DEM simulation; set the verification threshold, if it is higher than the threshold, continue to select individuals according to the fitness function for crossover and mutation, and generate the parameter combination with the best fitness.

[0160] Preferably, step S321 of this embodiment determines the key structural parameters that have the greatest impact on the target parameters through the analysis of feature importance, sets the parameter combination, inputs the parameters into the trained artificial neural network model, and predicts the corresponding target parameters. The residual learning framework in deep learning is used to optimize the network by learning the residual function, so that the network is easier to optimize and improve the accuracy. In addition, it reflects the powerful ability of neural networks in dealing with complex problems, such as the application in image recognition tasks. Step S322 predicts the gap between the output target parameters and the actual target parameters through the artificial neural model, and adjusts the structural parameters to minimize the gap. The Adam optimization algorithm is used, which performs well in processing large-scale data and parameters, and has good convergence and computational efficiency. In addition, the application of genetic algorithms in optimization problems, such as the real-valued genetic algorithm in PID parameter optimization, finds the optimal solution through real number coding and genetic operations. Step S323 randomly generates an initial structural parameter population, and predicts the SN, Re, and TRe of each parameter combination through an artificial intelligence network model; calculates the fitness function of each combination, that is, the gap between the predicted SN, Re, and TRe values ​​and the target SN, Re, and TRe, and selects individuals with better performance for crossover and mutation according to the fitness function to generate a new parameter population. The combination of differential evolution algorithm and genetic algorithm is used to optimize the parameter combination by adaptively adjusting parameters and selecting the best individuals, thereby improving the prediction accuracy.

[0161] Therefore, the significance of this embodiment lies in that it realizes efficient optimization and prediction of complex system parameters through the combination of deep learning, optimization algorithm and genetic algorithm, reflecting the wide application and technical advantages of artificial intelligence in engineering and scientific computing.

[0162] Furthermore, if Fig.12 As shown, the process of performing CFD-DEM simulation verification of the optimal parameters in step S33 specifically includes the following steps:

[0163] Step S331: setting the initial flow field and particle distribution, running CFD-DEM simulation according to the artificial neural network model simulation results, and recording key variables; the key variables include flow velocity, pressure, particle concentration, collision frequency, etc.;

[0164] Step S332: Calculate target parameters according to the simulation results, including SN, Re, TRe, etc.; set an optimization target value, compare the simulation result with the optimization target value, and if the simulation result exceeds the optimization target value, re-divide the grid; repeat the above steps until the simulation result is less than the optimization target value;

[0165] Step S333: Evaluate the performance of the optimized parameter combination according to the simulation results, set performance indicators, and continue to adjust the simulation if the indicators are not met; the performance of the optimized parameter combination includes but is not limited to constraints such as the service life of the spray gun and the manufacturing cost.

[0166] Preferably, step S331 of this embodiment provides basic data for subsequent CFD-DEM simulation by setting the initial flow field and particle distribution. According to the simulation results of the artificial neural network model, the accuracy and reliability of the simulation results are ensured. By recording key variables such as flow rate, pressure, particle concentration, collision frequency, etc., important data support can be provided for subsequent analysis and optimization. Step S332 calculates the target parameters and optimizes them by analyzing the simulation results. Compare the simulation results with the optimization target value. If the simulation results exceed the optimization target value, the grid is re-divided and the above steps are repeated until the simulation results are less than the optimization target value. The accuracy and optimization effect of the simulation results are ensured, and a reliable basis is provided for subsequent performance evaluation. Step S333 evaluates the performance in the optimization parameter combination according to the simulation results and sets the performance index. If it does not meet the standard, the simulation is adjusted continuously. The performance in the optimization parameter combination includes but is not limited to constraints such as the service life of the spray gun and the manufacturing cost; it ensures that the performance of the optimization parameter combination meets expectations and provides technical support for practical applications.

[0167] Therefore, the significance of this embodiment is to ensure the accuracy of simulation results and optimization effect by setting initial conditions, analyzing simulation results, and evaluating optimization parameter combinations, so as to provide reliable technical support for practical applications.

[0168] Furthermore, if Fig.13 As shown, the process of running the CFD-DEM simulation in step S331 specifically includes the following steps:

[0169] Step S3311: inputting the optimal parameter combination into the CFD software, performing flow field simulation according to the optimal parameter combination, and optimizing flow field characteristics;

[0170] Step S3312: by coupling DEM with CFD, numerical simulation is performed on the movement of the swirl chaotic dry powder spray gun during the flow field operation process, the characteristics and performance of the swirl chaotic dry powder spray gun are obtained, and key variables are recorded;

[0171] Among them, key variables include flow rate, pressure, particle concentration, collision frequency, etc.

[0172] Step S3313: Calculate the target parameters according to the characteristics, performance and key variables of the swirl chaotic dry powder spray gun, set the optimization target value, compare the simulation results with the optimization target value, and re-divide the grid if the simulation results exceed the optimization target value.

[0173] Preferably, step S3311 of this embodiment can ensure that the initial conditions of the flow field simulation are optimal by inputting the optimal parameter combination, thereby improving the accuracy and reliability of the simulation results. The flow field characteristics are optimized to make the flow of fluid in a specific area more stable and efficient. Step S3312 can simultaneously consider the movement of particles and fluids through the DEM-CFD coupling method to accurately simulate the complex behavior of the swirl chaotic dry powder spray gun in the flow field. By recording key variables, the performance and characteristics of the spray gun can be fully understood to provide data support for optimization. Step S3313 can evaluate the accuracy of the simulation results by calculating the target parameters and comparing them with the optimized target values, and adjust the grid division as needed to improve the simulation accuracy.

[0174] Therefore, the experimental cost and time can be reduced and the design efficiency can be improved, especially when multiple parameter adjustments are required to achieve optimal performance. This embodiment comprehensively analyzes the performance and characteristics of the swirl chaotic dry powder spray gun through the DEM-CFD coupling method. The simulation-based optimization method can significantly reduce the number of experiments, save resources, and improve the accuracy and efficiency of the design.

[0175] like Figure 4 As shown, this embodiment also provides an embodiment of a molten pool smelting swirl chaotic dry powder spray gun structural parameter optimization system. In this embodiment, the molten pool smelting swirl chaotic dry powder spray gun structural parameter optimization system is applied to the smelting swirl dry powder spray gun design method based on machine learning as in the above embodiment. The molten pool smelting swirl chaotic dry powder spray gun structural parameter optimization system includes a flow field simulation and data acquisition module 1, a data cleaning and model training module 2, and a model training and verification module 3 which are electrically connected in sequence.

[0176] Among them, the flow field simulation and data acquisition module 1 is used to use CFD-DEM coupling software to simulate the swirl chaotic dry powder spray gun under different architecture parameter configurations, and encrypt the grids at the swirl blade boundary and the spray gun outlet before calculation; extract the characteristic parameters of the flow field in the CFD-DEM coupling software; calculate the ratio of the tangential momentum flux to the axial momentum flux of the fluid and the turbulence intensity of the jet in the spray gun for the density, coordinates, tangential velocity and axial velocity of each data point, and transmit the calculated data to the data cleaning and model training module 2; the data cleaning and model training module 2 is used to match the process thermal parameters with the calculated data to construct a data set; check the missing values ​​in the data set, use the box plot method of data distribution to describe the distribution of the data and identify outliers; and standardize the extracted structural parameters and flow field characteristics; output the processed training set to the artificial neural network model, and use the grid search method to adjust the model; adjust the network parameters through the back propagation algorithm to reduce the loss function value; use the validation set to evaluate the trained artificial neural network model; model training and validation Verification module 3 is used to determine the key structural parameters with the greatest impact on the target parameters through feature importance analysis, and set parameter combinations; input the set parameter combinations into the trained artificial neural network model to predict the corresponding target parameters; compare the target parameters predicted and output by the artificial neural network model with the actual target parameters to minimize the gap between the structural parameters; randomly generate the initial structural parameter population using a genetic algorithm, and predict the SN, Re and TRe of each parameter combination through the artificial neural network model; calculate the fitness function of each combination, that is, the gap between the predicted values ​​of SN, Re and TRe and the target SN, Re and TRe; set the fitness function threshold, select the individual with the smallest difference from the threshold for crossover and mutation, and form a new parameter population; repeat the process of selection, crossover and mutation, and continuously iterate and optimize until the maximum number of iterations is reached or the fitness function converges; perform CFD-DEM simulation verification on the best parameter combination with the best fitness; set a verification threshold, if it is higher than the threshold, continue to select individuals according to the fitness function for crossover and mutation to generate the best parameter combination with the best fitness;

[0177] Preferably, the use of the CFD-DEM coupling software of the present embodiment is key, can efficiently simulate flow field and analyze performance, and can significantly improve prediction accuracy and save calculation time in combination with neural network technology. In addition, the application of neural network in CFD core processes such as grid generation and numerical solution also shows its high efficiency and intelligent advantages. The present embodiment carries out flow field modeling by deep learning, can not only utilize the inherent variation law of the flow field system learned, make simulation and prediction to flow field, but also can reduce prediction time and computing resources. In practical applications, CFD-DEM coupling software can provide guidance for experiment, to save a lot of manpower, material resources and time, and allow researchers to be freed from heavy programming and experiment.

[0178] Therefore, the significance of this embodiment lies in the combination of CFD-DEM coupling software and neural network technology, which not only improves the efficiency and accuracy of flow field simulation, but also provides new ideas and methods for the optimal design of complex fluid systems.

[0179] like Fig.15 As shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 4 includes a processor 41 and a memory 42 coupled to the processor 41.

[0180] The memory 42 stores program instructions for implementing the smelting cyclone dry powder spray gun design method based on machine learning of any of the above embodiments.

[0181] The processor 41 is used to execute the program instructions stored in the memory 42 to optimize the structural parameters of the molten pool smelting swirl chaotic dry powder spray gun.

[0182] The processor 41 may also be referred to as a CPU (Central Processing Unit). The processor 41 may be an integrated circuit chip having signal processing capabilities. The processor 41 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0183] Further, Fig.16 The schematic diagram of the structure of the storage medium of an embodiment of the present application is that the storage medium 5 of the embodiment of the present application stores program instructions 51 that can implement all the above methods, wherein the program instructions 51 can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0184] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0185] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units. The above is only an implementation mode of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present invention.

[0186] The specific implementation methods of the invention are described in detail above, but they are only examples, and the invention is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the invention, therefore, the equalization, modification, improvement, etc. made without departing from the spirit and principle of the invention should be included in the scope of the invention.

Claims

1. A design method for a smelting cyclone dry powder spray gun based on machine learning, characterized in that: The method for designing a smelting cyclone dry powder spray gun based on machine learning includes: Use CFD-DEM coupling software to simulate and calculate the swirl chaotic dry powder spray gun under different structural parameter configurations; extract the key characteristic parameters of the flow field in the CFD-DEM coupling software, calculate the ratio of the tangential momentum to the axial momentum of the fluid through the key characteristic parameters, and form the target parameters; Match the flow field characteristic parameters with the target parameters, construct a data set, and clean the data set; divide the data set into a training set and a validation set, train the artificial neural network model based on the training set; and use the validation set to evaluate the trained artificial neural network model; Analyze the impact of each hyperparameter in the artificial neural network model on the model performance; determine the structural parameters that have the greatest impact on the target parameters and set the parameter combination; calculate the target parameters of the parameter combination, select individuals for crossover and mutation based on the calculation results, and form the optimal fitness parameter combination; Calculate the maximum number of iterations or fitness function convergence. If so, perform simulation verification on the new model combination. If not, continue with crossover and mutation.

2. The method for designing a smelting cyclone dry powder spray gun based on machine learning according to claim 1 is characterized in that: The process of forming target parameters includes: The swirl chaotic dry powder spray gun is locally encrypted, and the key positions at the swirl blade boundary and the spray gun outlet are simulated and calculated using CFD-DEM coupling software for the swirl chaotic dry powder spray gun under different structural parameter configurations; Extract the key characteristic parameters of the flow field in CFD-DEM software; Among them, the key characteristic parameters include the density, coordinates, and tangential velocity U of each data point. tan , axial speed U axi , the inner diameter of the annular tube of the spray gun R, the range from the annular zone to the mixing zone is from R0 to R1, the characteristic length of the spray gun L, the velocity u of the jet and the integrated volume V; The ratio of the fluid's tangential momentum flux to its axial momentum flux, SN, and the turbulence intensity of the jet in the spray gun are calculated by using the parameters.

3. The method for designing a smelting cyclone dry powder spray gun based on machine learning according to claim 2 is characterized in that: The process of calculating the ratio SN of the tangential momentum flux to the axial momentum flux of the fluid and the turbulence intensity of the jet in the spray gun includes: Extract the key characteristic parameters of the flow field in CFD-DEM software; Calculate the ratio of the tangential momentum flux to the axial momentum flux of the fluid, SN; the equation for SN is: Where, SN local It represents the ratio of tangential momentum flux to axial momentum flux, and it represents the momentum distribution characteristics of the local area. angular，local It represents the tangential momentum flux in the local area, which means the momentum transfer of the fluid in the tangential direction. linear，local It represents the axial momentum flux of the local area, which means the momentum transfer of the fluid in the axial direction. R represents the radius, which refers to the characteristic radius of the flow field or the radius of a specific area. ρ represents the density of the fluid, which means the mass of the fluid per unit volume. r represents the radial distance, which means the distance from the center of the flow field to a certain point. U axi Indicates the axial velocity, which indicates the velocity component of the fluid in the axial direction, U tan represents the tangential velocity, which represents the velocity component of the fluid in the tangential direction, R0 represents the lower limit of integration, which represents the inner boundary radius of the flow field, and R1 represents the upper limit of integration, which represents the outer boundary radius of the flow field; The turbulence intensity Reynolds number Re and turbulence Reynolds number TRe of the jet in the spray gun introduced into the gas phase are used to evaluate the turbulence intensity of the jet in the spray gun; among them, Re is defined as the ratio of the inertial force to the friction force, and TRe is defined as the ratio of the turbulent kinetic energy L of the fluid to the work done by the friction force W, and the expressions are listed as follows: In the formula, Re is the Reynolds number, which represents the ratio of inertial force to viscous force and is used to evaluate the turbulence intensity of the fluid, and ρ g Indicates the density of the gas phase fluid, which indicates the mass per unit volume of the gas phase fluid, u g It represents the characteristic velocity of the gas phase fluid, which means the velocity of the gas phase fluid at a certain characteristic length. L represents the characteristic length, which means the characteristic size of the flow field, which is the width or height of the flow field. μ g It represents the dynamic viscosity of gas phase fluid, which represents the viscous characteristics of gas phase fluid. TRe represents the turbulent Reynolds number, which represents the ratio of the turbulent kinetic energy of fluid to the work done by friction, and is used to evaluate the intensity of turbulence. g represents the turbulent kinetic energy of the gas phase fluid, represents the energy of turbulent motion in the gas phase fluid, V represents the volume of the fluid, represents the total volume of the fluid in the flow field, represents the mean velocity square, represents the mean square value of the fluid velocity, represents the square of the average velocity, which represents the square value of the average velocity of the fluid, W represents the work done by friction, which represents the work done by friction in the fluid, μ represents the dynamic viscosity of the fluid, which represents the viscous characteristics of the fluid, represents the velocity gradient, which indicates the rate of change of fluid velocity in a certain direction, n v Represents the normal vector, which represents the unit vector in a certain direction in the flow field. i ′u j ′> represents the Reynolds stress tensor, which indicates the correlation of velocity pulsation in the fluid.​ 4. The method for designing a smelting cyclone dry powder spray gun based on machine learning according to claim 3 is characterized in that: The equations for evaluating the trained ANN model using the validation set include: Match the flow field characteristic parameters with the target parameters, construct a data set, and perform a cleaning operation on the data set; divide the data set into 80% training set and 20% validation set; Use the grid search method to adjust the artificial neural network model and select the artificial neural network model with the smallest loss function for training; train the artificial neural network model according to the training set, adjust the network parameters through the back propagation algorithm, and use the early stopping strategy to avoid overfitting; Use the validation set to evaluate the trained artificial neural network and calculate the error index of the artificial neural network model on the validation set; set a threshold, and if it exceeds the threshold, readjust the artificial neural network model; if it is below the threshold, perform feature extraction and analysis.

5. The method for designing a smelting cyclone dry powder spray gun based on machine learning according to claim 4 is characterized in that: The process of building a dataset includes: Check the missing values ​​in the data set; if missing values ​​occur, use mean filling as follows: In the formula, x i ′ refers to the value at the location of the missing value, n′ represents the total number of non-missing values ​​in the data set, and x i 'j' represents the j'th eigenvalue of the i'th sample or data point; Draw box plots of each institution's parameters and process characteristic values; calculate the upper quartile, lower quartile and interquartile range of each parameter, and determine the upper and lower thresholds; mark values ​​below the lower limit or above the upper limit as outliers; analyze outliers by methods such as replacement or deletion; The extracted structural parameters and flow field characteristics are standardized; the standardization formula is as follows: Among them, x norm is the standardized data, x is the original data, μ is the mean of the data, and σ is the standard deviation of the data.

6. The method for designing a smelting cyclone dry powder spray gun based on machine learning according to claim 4 is characterized in that: The process of training the artificial neural network model with the training set includes: Design the network structure of the artificial neural network model, select ReLU as the activation function, and select Hunerloss as the loss function; input the training set into the artificial neural network model, and use the Adam optimizer to adjust the network weights and biases of the artificial neural network model; Among them, the network structure of the artificial neural network model includes an input layer (including 7 input features: blade length, blade rotation angle, blade installation height, blade tip clearance, blade thickness, CH4 tube diameter, and mixing zone length), several hidden layers, and an output layer (including 3 output features: SN, Re, and TRe); Determine the range of values ​​for the learning rate, number of hidden layers, number of nodes, number of batches, and batch size parameters; form a grid with all hyperparameters; for each hyperparameter combination in the grid, use the training set to train the artificial neural network model; The validation set is used to verify the mean square error, accuracy, etc. of the artificial neural network model after each training, and a threshold is set. The artificial neural network model corresponding to the hyperparameter combination with the smallest error compared to the threshold is selected as the final artificial neural network model.

7. The method for designing a smelting cyclone dry powder spray gun based on machine learning according to claim 1, characterized in that: The process of forming the optimal fitness parameter combination includes: The SHAP importance evaluation method is used to analyze the influence of each structural parameter in the artificial neural network model on the target parameter; through the analysis of the importance of the target feature, the key structural parameters with the greatest influence on the target parameter are determined, and the parameter combination is set. The combination is input into the trained artificial neural network model to predict the corresponding target parameter; The gap between the target parameters and the actual parameters predicted by the artificial neural network model is adjusted, and the parameter combination is randomly generated by the genetic algorithm. The optimization is continuously iterated until the maximum number of iterations is reached or the function converges, and finally the optimal parameter combination for fitness is obtained. The optimal fitness parameter combination is selected for CFD-DEM simulation verification, and the verified performance is tested and corrected.

8. The method for designing a smelting cyclone dry powder spray gun based on machine learning according to claim 7 is characterized in that: The process of predicting the corresponding target parameters includes: Collect training data sets for artificial neural networks, including input features and corresponding target parameters; after model training is completed, use the SHAP method to calculate the contribution value of each input feature to the target parameter; set contribution value standards, make comparisons, and select features with contribution values ​​higher than the standards; According to the contribution values ​​obtained by comparison, the importance of the features is sorted; the higher the contribution value, the greater the influence of the feature on the target parameter, and the higher the relative importance; according to the sorting results, the key feature ranked first is selected; The selected key feature combination is input into the trained artificial neural model to predict the target parameters; the actual parameters are compared with the predicted parameters, and if they are higher than a certain threshold, retraining and prediction are performed; the above steps are repeated until the maximum number of iterations or the preset standard is reached.

9. The method for designing a smelting cyclone dry powder spray gun based on machine learning according to claim 7, characterized in that: The process of randomly generating the initial structural parameter population using a genetic algorithm includes: Through feature importance analysis, the key structural parameters that have the greatest impact on the target parameters are determined, and the parameter combination is set. The parameters are input into the trained artificial neural network model to predict the corresponding target parameters; The gap between the output target parameters and the actual target parameters is predicted by the artificial neural model, and the structural parameters are adjusted to minimize the gap; Randomly generate the initial structural parameter population, and predict the SN, Re and TRe of each parameter combination through the artificial intelligence network model; calculate the fitness function of each combination, that is, the difference between the predicted SN, Re and TRe values ​​and the target SN, Re and TRe; select individuals with better performance according to the fitness function for crossover and mutation to generate a new parameter population; The process of selection, crossover and mutation is repeated, and the optimization is continuously iterated until the maximum number of iterations is reached or the fitness function converges; the parameter combination with the best fitness is finally verified by CFD-DEM simulation; a verification threshold is set, and if it is higher than the threshold, individuals will continue to be selected according to the fitness function for crossover and mutation to generate the parameter combination with the best fitness.

10. A smelting cyclone dry powder spray gun design system based on machine learning, which is applied to the smelting cyclone dry powder spray gun design method based on machine learning as described in claims 1 to 9, characterized in that: include: The flow field simulation and data acquisition module is used to simulate the swirl chaotic dry powder spray gun under different architecture parameter configurations using CFD-DEM coupling software, encrypt the grids at the swirl blade boundary and the spray gun outlet before calculation; extract the characteristic parameters of the flow field in the CFD-DEM coupling software; calculate the ratio of the tangential momentum flux to the axial momentum flux of the fluid and the turbulence intensity of the jet in the spray gun based on the density, coordinates, tangential velocity and axial velocity of each data point, and transmit the calculated data to the data cleaning and model training module; The data cleaning and model training module is used to match the process characteristic parameters with the calculated data to construct a data set; check the missing values ​​in the data set, use the box plot method of data distribution to describe the data distribution and identify outliers; and standardize the extracted structural parameters and flow field characteristics; Output the processed training set to the artificial neural network model, and use the grid search method to adjust the model; adjust the network parameters through the back propagation algorithm to reduce the loss function value; use the validation set to evaluate the trained artificial neural network model; The model training and verification module is used to determine the key structural parameters that have the greatest impact on the target parameters through feature importance analysis, and set parameter combinations; input the set parameter combinations into the trained artificial neural network model to predict the corresponding target parameters; compare the target parameters predicted and output by the artificial neural network model with the actual target parameters to minimize the gap between the structural parameters; randomly generate the initial structural parameter population using a genetic algorithm, and predict the SN, Re and TRe of each parameter combination through the artificial neural network model; calculate the fitness function of each combination, that is, the gap between the predicted SN, Re and TRe values ​​and the target SN, Re and TRe; set the fitness function threshold, select the individuals with the smallest difference from the threshold for crossover and mutation, and form a new parameter population; repeat the selection, crossover and mutation process, and continuously iterate and optimize until the maximum number of iterations is reached or the fitness function converges; perform CFD-DM simulation verification on the best parameter combination with the best fitness; Set the verification threshold. If it is higher than the threshold, individuals will be selected for crossover and mutation according to the fitness function to generate the optimal fitness parameter combination.

Images