Spray forming atomizer process parameter optimization method
By combining fluid mechanics simulation and neural network prediction model with NSGAⅡ algorithm to optimize the parameters of the jet forming atomizer, the problem of low efficiency in the existing technology was solved, multi-objective optimization and efficient process parameter control were achieved, and the forming effect was improved.
Patent Information
- Application Number
- CN202510424993.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-07
- Publication Date
- 2025-09-05
AI Technical Summary
Existing parameter optimization methods for jet forming atomizers are inefficient and have low applicability, making it difficult to meet actual production needs, especially in terms of multi-objective optimization and complex process parameter control.
By establishing a fluid mechanics simulation model, combining the BP neural network prediction model and the NSGAⅡ algorithm, the parameters of the spray forming atomizer are optimized. The atomizer model is established using CFD simulation software, the droplet data is collected, the neural network is trained, and a multi-objective optimization algorithm framework is built, which is finally verified experimentally.
It achieves efficient optimization of the parameters of the spray forming atomizer, improves the feasibility of process control, improves the prediction accuracy of the droplet diameter and temperature, and ensures the grain size and density of the finished ingot.
Smart Images
Figure CN120597741A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of injection molding, and in particular to an atomizer parameter optimization method based on fluid mechanics simulation, neural network prediction and genetic algorithm. Background Art
[0002] At present, the research on the optimization of spray forming atomizer parameters faces the technical difficulties of numerous input parameters and many optimization targets that are difficult to optimize. Spray forming technology, as a near-net forming method that combines rapid solidification and atomization forming, is widely used in the preparation of light metal materials such as aluminum, titanium, and magnesium. However, due to the complexity of the structural parameters and process parameters of the spray forming atomizer, the morphology, density, and grain size of the formed ingot are affected by different parameters, making the optimization of process parameters particularly difficult. In order to achieve precise control of the atomization effect of the molten alloy, researchers have proposed a variety of methods.
[0003] Currently, the optimization research of atomizer parameters mainly focuses on single-objective optimization. Usually, parameters such as droplet diameter and droplet temperature are adjusted independently, and the optimization effect of a single optimization target is good. However, considering the complexity of the actual production process, a single optimization target obviously cannot fully meet production needs.
[0004] In addition, some researchers use computational fluid dynamics (CFD) simulation software to perform numerical simulation of the melt atomization process. By systematically adjusting the atomizer parameters and conducting a large number of simulation experiments to approach the optimization target, this method can optimize the atomizer process parameter combination close to the target. However, this method is often time-consuming in calculation, which limits its application efficiency in actual production.
[0005] Although these methods have achieved certain results in optimizing the process parameters of spray forming atomizers, they still suffer from problems such as low efficiency and low applicability.
[0006] In summary, while existing spray forming atomizer parameter optimization methods have their own strengths, they still face challenges in practical application. Therefore, there is an urgent need to develop more efficient atomizer process parameter methods to improve the feasibility of spray forming process control. Summary of the Invention
[0007] The purpose of the present invention is to provide a method for optimizing atomizer parameters. A large amount of simulation data is obtained by establishing a fluid mechanics atomization simulation model, and a BP neural network prediction model of the droplet diameter and temperature during deposition is obtained based on the simulation data. In combination with the prediction model, a multi-objective optimization algorithm framework is built based on the NSGAⅡ algorithm to obtain the optimal parameter solution set, which is finally verified experimentally.
[0008] To achieve the above object, the present invention provides the following technical solution: a method for optimizing process parameters of a spray forming atomizer, comprising the following steps:
[0009] S1. Using CFD simulation software, establish a spray forming atomization simulation model according to the atomizer size used in actual production. Based on the established atomization simulation model, construct a sampling plane on the deposition surface to collect droplet diameter and temperature data during deposition; obtain droplet data during deposition under different input atomizer parameters;
[0010] S2. Based on the large amount of data obtained from the atomization simulation model in S1, a prediction model for droplet diameter and temperature based on BP neural network is established;
[0011] S3. Based on the prediction model in S2, the NSGAⅡ algorithm optimization framework is built, and the ideal droplet diameter and temperature are specified and iterated multiple times to obtain the ideal optimized parameters;
[0012] S4. By analyzing the parameters optimized in S2 and based on actual production conditions, several sets of parameters are selected to conduct actual spray forming experiments, the grain size and density of the finished ingots are tested, and the final ideal parameters are selected.
[0013] Furthermore, the establishment of the spray forming atomization simulation model in step S1 is specifically as follows:
[0014] Based on the VOF-DPM model in Fluent, the VOF method is used to track the gas-liquid interface, the DPM method is used to track discrete particles, the Euler method describes the fluid distribution, and the Lagrangian method simulates the droplet motion.
[0015] Furthermore, in step S1, after establishing the flow field simulation model, when adjusting the atomizer process parameters, the main atomizer pressure adjustment range is 0.7~2.2Mpa, the main atomizer top angle adjustment range is 3~35°, and the main atomizer nozzle diameter adjustment range is 3~6mm. By analyzing the distribution of the argon flow field in the atomization area and the flow field velocity cloud map, it is found that when the main atomizer acts alone, a backflow field will be generated to hinder the flow of molten droplets, and when the auxiliary atomizer pressure is 0.3Mpa, the influence of the backflow field can be eliminated.
[0016] Furthermore, in step S2, the steps of establishing the BP neural network are specifically as follows:
[0017] A prediction model for droplet diameter and temperature was established. The input parameters were atomizer pressure, apex angle and pore diameter. The number of hidden layer nodes was 6. The Tanh function was selected as the activation function in the input layer, the trainlm training function was selected in the hidden layer, and the linear transfer function purelin was used in the output layer.
[0018] Furthermore, in step S2, the steps of establishing the BP neural network are specifically as follows:
[0019] (1) Determine the model topology structure, where the process parameter layer consists of three nodes: pressure, vertex angle, and pore diameter;
[0020] The process effect layer consists of two nodes: the droplet diameter during deposition and the droplet temperature during deposition;
[0021] (2) According to the model topology, the number of nodes is calculated to be in the range of
[0022] [2,16], trained the models with different numbers of hidden layer nodes and obtained the mean square error of the validation set. It was found that the number of hidden layer nodes was [4,10] with the smallest error. The number of nodes 6 with the smallest error was selected as the number of hidden layer nodes.
[0023] (3) Selection of activation functions: Tanh function is selected as the activation function for the input layer, trainlm training function is selected for the hidden layer, and purelin linear transfer function is used for the output layer;
[0024] (4) Set the number of training rounds to 100, the minimum error to 0.0001, and the learning rate to 0.01.
[0025] Furthermore, in step S3, the steps of building the NSGAⅡ algorithm optimization framework are specifically as follows:
[0026] (3) Determine the optimization objective function as the absolute value of the difference between the predicted value and the target value;
[0027] (4) Determine the elite retention strategy as non-dominated sorting combined with congestion sorting.
[0028] Furthermore, the NSGAⅡ algorithm framework is established in step S3 as follows:
[0029] The objective function is selected as f(x)=abs(y pred (x)-y target ), a fast non-dominated sort combined with congestion calculation is used as the elite retention strategy.
[0030] Further, in step S4, in step S4: the droplet diameter is measured using the detected ingot grain size.
[0031] Furthermore, in step S4, the experimental scheme for spray forming ingots is specifically as follows:
[0032] The solidification process of aluminum-lithium alloy droplets is in-situ solidification. The grain size formed after solidification is closely related to the diameter of the droplets after atomization. The method of making an ingot and detecting its grain size can more effectively verify the atomization effect, and the density test can verify the solidification effect of the ingot. The experimental sampling plan is to take samples from the tail, middle, center of the head, 1 / 2 radius, and edge of the ingot, for a total of 9 samples.
[0033] In summary, due to the adoption of the above technology, the beneficial effects of the present invention are:
[0034] The method proposed in the present invention targets the spray forming atomization process, which has many influencing parameters. A fluid simulation model is established for the atomization process. By analyzing the influence of the flow field dynamic pressure and the recirculation zone on the atomization effect, it is found that the three key parameters of the atomizer apex angle, pore diameter and pressure have the greatest influence on the atomization effect. By establishing a BP neural network prediction model to replace the time-consuming simulation process, an NSGAⅡ algorithm framework is built on the basis of the prediction model to optimize the atomizer parameters. Finally, experiments are carried out to verify the optimization effect. This method can efficiently optimize the atomizer parameters at the target grain diameter temperature. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 The overall flow chart of the method proposed in the present invention;
[0036] Figure 2 Schematic diagram of the structure of the flow field simulation model in the method proposed by the present invention;
[0037] Figure 3 This is a flow chart of the NSGAⅡ algorithm in the method proposed in the present invention;
[0038] Figure 4 Schematic diagram of sampling position for ingot forming experiment.
[0039] Description of reference numerals:
[0040] Main atomizer 1, auxiliary atomizer 2, deposition chamber wall 3, deposition substrate 4, flow guide tube 5, symmetry axis 6. DETAILED DESCRIPTION
[0041] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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 making creative work are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention for which protection is sought, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0042] In this embodiment, the CFD simulation software is Computational Fluid Dynamics simulation software, specifically ANSYS Fluent. The BP is Back Propagation.
[0043] like Figures 1 to 4 , a specific embodiment of the present invention is given:
[0044] like Figure 1 As shown, the idea of the parameter optimization method of the spray forming atomizer proposed by the present invention is given, which specifically includes the following steps:
[0045] S1. Using the CFD simulation software ANSYS Fluent, a spray forming atomization simulation model was established based on the actual atomizer dimensions used in production. Based on this atomization simulation model, a sampling plane was constructed on the deposition surface to collect droplet diameter and temperature data during deposition. Droplet deposition data was obtained for different input atomizer parameters.
[0046] As a preferred implementation of step S1, step S1 specifically includes:
[0047] An efficient way to build atomization simulation models is based on the fact that the fluid domain of the spray atomization process presents a centrally symmetrical cylindrical region. This characteristic allows the simulation of the entire fluid domain to be converted into a two-dimensional simulation on the axial section. Since the axial section is an axisymmetric figure, the problem can be simplified to consider only half of the region during modeling and calculation, such as Figure 2 This method effectively reduces computational complexity and improves simulation efficiency; a sampling plane is established at the pressure outlet to collect the diameter and temperature of the droplet falling there during the simulation process.
[0048] S2. Based on the large amount of data obtained from the atomization simulation model in S1, a prediction model for droplet diameter and temperature based on BP neural network is established;
[0049] More specifically, in step S2, the process of establishing the BP neural network is as follows:
[0050] (5) Determine the model topology structure, where the process parameter layer consists of three nodes: pressure, vertex angle, and pore diameter;
[0051] The process effect layer consists of two nodes: the droplet diameter during deposition and the droplet temperature during deposition.
[0052] (6) According to the model topology, the number of nodes is calculated to be in the range of
[0053] [2,16], trained the models with different numbers of hidden layer nodes and obtained the mean square error of the validation set. It was found that the error was smaller when the number of hidden layer nodes was [4,10]. The number of nodes with the smallest error, 6, was selected as the number of hidden layer nodes.
[0054] (7) Selection of activation functions: Tanh function is selected as the activation function for the input layer, trainlm training function is selected for the hidden layer, and purelin linear transfer function is used for the output layer.
[0055] (8) Set the number of training rounds to 100, the minimum error to 0.0001, and the learning rate to 0.01.
[0056] S3. Based on the prediction model in S2, the NSGAⅡ algorithm optimization framework is built, and the ideal droplet diameter and temperature are specified and iterated multiple times to obtain the ideal optimized parameters.
[0057] As a preferred embodiment of step S3, the process is as follows Figure 3 As shown, specifically including:
[0058] S31 defines the number of input variables, the number of objective functions, the size of the initialization population, the maximum number of iterations, the crossover probability, the mutation probability, and the upper and lower limits of the input variables.
[0059] S32, generate the initial population through a loop, randomly generate 100 sets of input parameters under the upper and lower limits of the parameters, call the BP neural network prediction model in S2, and calculate the objective function value. The objective function is:
[0060] f(x)=abs(y pred (x)-y target )
[0061] S33. Perform non-dominated sorting on the initial population to obtain a non-dominated set; calculate the aggregation distance of each individual to evaluate the diversity of individuals in the selection process.
[0062] S34. Start the main iteration loop, use the tournament selection method to select parent individuals, generate offspring individuals through single-point crossover, mutate the offspring individuals and calculate their objective function values, merge the parent and offspring individuals into a new population, and perform non-dominated sorting and aggregation distance calculation again to select the excellent population to enter the next generation of iteration.
[0063] S35. Visualize the results at a specific number of iterations (such as the 20th, 50th, 100th, 150th, and 200th), plot the objective function values of different non-dominated sets, and use different color markers. The error range between the optimized parameters and the target values can be observed from the graph.
[0064] S4. By analyzing the parameters optimized in S2 and based on actual production conditions, several sets of parameters are selected to conduct actual spray forming experiments, the grain size and density of the finished ingots are tested, and the final ideal parameters are selected.
[0065] As a preferred implementation of step S4, a specific sampling scheme is:
[0066] like Figure 4 As shown, samples were taken from the tail, middle, center, 1 / 2 radius, and edge of the ingot, for a total of 9 samples, and the density and grain size of the 9 samples were tested.
[0067] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.
[0068] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.
Claims
1. A method for optimizing process parameters of a spray forming atomizer, characterized in that: The following steps are involved: S1. Establish a spray forming atomization simulation model based on the CFD simulation software according to the atomizer dimensions used in actual production. Based on the established atomization simulation model, construct a sampling plane on the deposition surface to collect droplet diameter and temperature data during deposition. Obtain droplet data during deposition under different input atomizer parameters. S2. Based on the large amount of data obtained from the atomization simulation model in S1, a prediction model for droplet diameter and temperature based on BP neural network is established; S3. Based on the prediction model in S2, the NSGAⅡ algorithm optimization framework is built, and the ideal droplet diameter and temperature are specified and iterated multiple times to obtain the ideal optimized parameters; S4. By analyzing the parameters optimized in S2 and based on actual production conditions, several sets of parameters are selected to conduct actual spray forming experiments, the grain size and density of the finished ingots are tested, and the final ideal parameters are selected.
2. The method for optimizing process parameters of a spray forming atomizer according to claim 1, wherein: The specific steps of establishing the spray forming atomization simulation model in step S1 are as follows: Based on the VOF-DPM model in Fluent, the VOF method is used to track the gas-liquid interface, the DPM method is used to track discrete particles, the Euler method describes the fluid distribution, and the Lagrangian method simulates the droplet motion.
3. The method for optimizing process parameters of a spray forming atomizer according to claim 1, wherein: In step S2, the steps of establishing the BP neural network are specifically as follows: A prediction model for droplet diameter and temperature was established. The input parameters were atomizer pressure, apex angle and pore diameter. The number of hidden layer nodes was 6. The Tanh function was selected as the activation function in the input layer, the trainlm training function was selected in the hidden layer, and the linear transfer function purelin was used in the output layer.
4. The method for optimizing process parameters of a spray forming atomizer according to claim 1, wherein: In step S3, the steps for building the NSGAⅡ algorithm optimization framework are as follows: (1) Determine the optimization objective function as the absolute value of the difference between the predicted value and the target value; (2) Determine the elite retention strategy as non-dominated sorting combined with congestion sorting.
5. The method for optimizing process parameters of a spray forming atomizer according to claim 1, wherein: In step S4, in step S4: the droplet diameter is measured using the detected ingot grain size.