Machine learning-based intelligent microstructure prediction method for special steel forging process

Abstract

The invention provides a special steel forging process microstructure intelligent prediction method based on machine learning, and relates to the technical field of microstructures. Comprising the steps of collecting process parameter data, microstructure data and dislocation motion data. And obtaining an influence coefficient according to the influence degree of the dislocation motion data on the dynamic recrystallization behavior, processing the microstructure data according to the influence coefficient to obtain optimized dynamic recrystallization behavior data, and adjusting the process parameter data according to the optimized dynamic recrystallization behavior data to obtain structure type data. And for the optimized generalized regression neural network model, inputting tissue type data, and outputting to obtain a microstructure prediction result. According to the method, the regression neural network model is optimized by using the fruit fly optimization algorithm, the microstructure prediction result can be accurately output, the influence of the dynamic recrystallization behavior on microstructure formation can be predicted, better microstructure and performance can be obtained, and the production risk is reduced for special steel forging production.

Description

Technical Field

[0001] The present invention relates to the technical field of microstructure, and particularly to an intelligent prediction method for the microstructure of special steel forging process based on machine learning. Background Art

[0002] Special steel has excellent properties such as high strength, high toughness, and corrosion resistance, and plays an irreplaceable role in key fields such as aerospace, national defense, and energy. Its forging process is crucial for the formation of the microstructure and properties of the material, directly affecting the quality and reliability of the final product. However, the forging of special steel is a complex physical metallurgical process, affected by the interaction of multiple factors such as forging temperature, strain rate, and deformation amount. These factors determine the dislocation movement, dynamic recrystallization and other microscopic phenomena inside the material, and thus determine the morphology and properties of the microstructure.

[0003] With the rapid development of computer technology and data science, the application of machine learning in the field of materials science has become increasingly widespread. Machine learning algorithms can automatically learn and extract rules from a large amount of data, and have powerful non-linear mapping ability and data processing ability, providing new ideas and methods for solving the complex problem of predicting the microstructure of special steel forging process.

[0004] In the field of materials, machine learning has achieved certain success in material property prediction, material design, etc., showing advantages that cannot be compared with traditional methods, such as improving prediction accuracy, shortening the R & D cycle, and reducing costs, providing a theoretical and practical basis for the research of an intelligent prediction method for the microstructure of special steel forging process based on machine learning.

[0005] In the production of special steel, accurately predicting the microstructure is of great significance for optimizing the forging process, improving product quality, and reducing production costs. By predicting the microstructure in advance, the process parameters can be reasonably adjusted, defects can be avoided, production efficiency and product consistency can be improved, and the high-quality requirements for special steel materials in fields such as high-end equipment manufacturing can be met.

[0006] At the same time, with the intensification of market competition and the continuous improvement of product performance requirements, enterprises are urgently in need of an efficient and accurate microstructure prediction method to enhance their core competitiveness and realize the intelligent and digital transformation of special steel production. Summary of the Invention

[0007] The present invention provides an intelligent prediction method for the microstructure of special steel forging process based on machine learning, so as to solve the defects in the prior art that it is difficult to predict the influence of dynamic recrystallization behavior on the formation of the microstructure and lack a basis for adjusting process parameters.

[0008] The present invention provides an intelligent prediction method for the microstructure in the forging process of special steel based on machine learning, including: Collect data on special steel in previous forging processes to obtain historical experimental data, and extract process parameter data, microstructure data, and dislocation movement data from the historical experimental data.

[0009] Extract key dislocation data from the dislocation movement data, establish a dislocation kinetics model based on the dynamic recrystallization theory, calculate the index change amount, then analyze to obtain the influence coefficient, analyze the microstructure data, determine the parameters in combination with the influence coefficient, input them into the dislocation kinetics model to obtain adjustment data, compare with the actual data to obtain the comparison difference data, adjust the microstructure data according to the comparison difference data to obtain optimized dynamic recrystallization behavior data, and adjust the process parameter data according to the optimized dynamic recrystallization behavior data to obtain the tissue type data.

[0010] Construct a generalized regression neural network model, input the microstructure data for training, and use the fruit fly optimization algorithm to optimize the hyperparameters of the generalized regression neural network model to obtain an optimized generalized regression neural network model, input the tissue type data, and output the microstructure prediction result.

[0011] According to the intelligent prediction method for the microstructure in the forging process of special steel based on machine learning provided by the present invention, the steps of obtaining the influence coefficient include: Extract dislocation density, dislocation velocity, and dislocation interaction frequency data from the dislocation movement data for normalization processing, unify data with different dimensions to the same numerical range to obtain key dislocation data.

[0012] Based on the key dislocation data and the dynamic recrystallization theory, establish a mathematical model between the two to obtain a dislocation kinetics model.

[0013] When using the dislocation kinetics model to calculate the change data of key dislocation data, calculate the index change amount of the dynamic recrystallization behavior, and measure the sensitivity of different key dislocation data to dynamic recrystallization to obtain the analysis result.

[0014] According to the analysis result and in combination with the concept of the relative influence coefficient, compare the influence degree of dislocation movement on the dynamic recrystallization behavior to obtain the influence coefficient.

[0015] According to the intelligent prediction method for the microstructure in the forging process of special steel based on machine learning provided by the present invention, the steps of obtaining the dislocation kinetics model include: According to the dislocation theory, the increase in dislocation density will promote the nucleation of dynamic recrystallization, establish the relationship between dislocation density and nucleation to obtain the nucleation change.

[0016] Obtain the growth change according to the influence of dislocation velocity and temperature on grain growth.

[0017] By affecting the effective density of dislocations, the nucleation change and growth change are influenced, the relationship between the dislocation interaction frequency and dynamic recrystallization is established, and the nucleation rate and growth rate are obtained.

[0018] Based on the nucleation rate and growth rate and combined with the Avrami model, a dislocation dynamics model is obtained.

[0019] According to the intelligent prediction method for the microstructure of special steel forging process based on machine learning provided by the present invention, based on the dislocation dynamics model, the formula is expressed as:

[0020]

[0021] In the formula, is the recrystallized volume fraction, is a constant related to the material and temperature, is a constant related to the material properties, is the dislocation density, is a constant related to the material and geometry, is the empirical exponent, is the gas constant, is the absolute temperature, is the activation energy, is the nucleation and growth situation, is the correction coefficient, is the time, is the dislocation velocity, is the dislocation interaction frequency, is the natural exponential function.

[0022] According to the intelligent prediction method for the microstructure of special steel forging process based on machine learning provided by the present invention, the steps for obtaining the optimized dynamic recrystallization behavior data include: Analyze the microstructure data in terms of distribution characteristics and correlation to obtain the analyzed tissue data, and determine the parameters for adjusting the forging process to optimize the dynamic recrystallization behavior by combining the analyzed tissue data with the influence coefficient.

[0023] Input the parameters and key dislocation data into the dislocation dynamics model, simulate the changes in the dynamic recrystallization behavior, and output the adjusted data.

[0024] Conduct a forging experiment on special steel according to the parameters, then obtain the actual microstructure data through experimental means, and compare the adjusted data with the actual microstructure data to obtain the comparison difference data.

[0025] Adjust the microstructure data according to the contrast difference data to optimize the dynamic recrystallization behavior, thereby obtaining optimized dynamic recrystallization behavior data.

[0026] According to the intelligent prediction method for the microstructure of special steel forging process based on machine learning provided by the present invention, the steps of obtaining the tissue type data include: Based on the internal relationship between the nucleation rate, growth rate and microstructure properties, obtain the characteristic manifestations of the microstructure under different dynamic recrystallization states.

[0027] According to existing experience, sort out the action rules of process parameter data on the dynamic recrystallization behavior during special steel forging to obtain qualitative connection data.

[0028] According to the requirements of optimizing the dynamic recrystallization behavior data, combine the characteristic manifestations and qualitative connection data, and consider the interaction and synergistic effect between process parameters to formulate a process parameter adjustment plan.

[0029] Modify the process parameter data according to the process parameter adjustment plan to obtain the tissue type data.

[0030] According to the intelligent prediction method for the microstructure of special steel forging process based on machine learning provided by the present invention, the steps of obtaining the optimized generalized regression neural network model include: Drosophila generation movement, Drosophila algorithm optimization and iterative optimization to construct the model.

[0031] Drosophila generation movement: Randomly generate a preset number of Drosophila individuals, each Drosophila individual represents a set of hyperparameter combinations of the generalized regression neural network model, and define the initial position range of the Drosophila individuals in the search space, so that each Drosophila individual moves randomly within the initial position range.

[0032] Drosophila algorithm optimization: Calculate the taste concentration determination value and perform model training and error calculation, then select the current optimal Drosophila individual, and finally update the positions of all Drosophila individuals to obtain the corresponding updated positions.

[0033] Iterative optimization to construct the model: Repeat the Drosophila algorithm optimization until the preset number of iterations is reached, and select the hyperparameter combination corresponding to the Drosophila individual with the smallest concentration value to construct the generalized regression neural network model to obtain the optimized generalized regression neural network model.

[0034] According to the intelligent prediction method for the microstructure of special steel forging process based on machine learning provided by the present invention, the steps of obtaining the updated position include: For each Drosophila individual, calculate the taste concentration determination value according to the current position to obtain the concentration value, split the microstructure data into a training set and a test set, bring the concentration value into the generalized regression neural network model, use the training set for training, and then use the test set to calculate the prediction error at the concentration value.

[0035] Calculate the concentration values of all Drosophila individuals, and find the Drosophila individual with the smallest concentration value among them as the current optimal Drosophila individual.

[0036] Make other Drosophila individuals fly towards the position of the current optimal Drosophila individual, and use the update formula to update their own positions to obtain the corresponding updated positions.

[0037] According to the intelligent prediction method for the microstructure of special steel forging process based on machine learning provided by the present invention, the update formula is expressed as:

[0038] In the formula, is the step length of Drosophila individual flight, is the position of the current optimal Drosophila individual, is the th Drosophila individual's position in the search space at the

[0039] According to the intelligent prediction method for the microstructure of special steel forging process based on machine learning provided by the present invention, the steps for obtaining the microstructure prediction result include: Correct the missing values in the tissue type data by using the mean filling method, and for the existing non-numerical variables, use the one-hot encoding method to convert the non-numerical variables into numerical data, so as to obtain the processed tissue type data.

[0040] Organize the processed tissue type data into a matrix form that conforms to the optimized generalized regression neural network model, and input the matrix form into the optimized generalized regression neural network model.

[0041] According to the matrix form and the training set, calculate the distance metric between the samples of the matrix form and the samples of the training set to obtain the output layer data.

[0042] Perform weighted summation on the output layer data to obtain the weighted data, perform direct summation on the output layer data to obtain the summation data, and divide the weighted data by the summation data to obtain the microstructure prediction result.

[0043] The intelligent prediction method for the microstructure in the special steel forging process based on machine learning provided by the present invention quantifies the influence degree of the dynamic recrystallization behavior by introducing a dislocation dynamics model and combining dislocation movement data, solving the problem that it is difficult to accurately predict and optimize the influence of the dynamic recrystallization behavior on the formation of the microstructure in the special steel forging process, resulting in inaccurate control of the microstructure. Thus, the complexity of the dynamic recrystallization behavior is more accurately handled. Key data such as dislocation density, dislocation velocity, and dislocation interaction frequency are extracted from the dislocation movement data and normalized, and a dislocation dynamics model is established in combination with the dynamic recrystallization theory, so as to make more comprehensive use of these data. And the fruit fly optimization algorithm is used to optimize the hyperparameters of the generalized regression neural network model to obtain an optimized generalized regression neural network model, improving the prediction accuracy and generalization ability of the model, helping to better predict the microstructure, and thus achieving the effect of reducing production risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0045] Figure 1 FIG. 1 is one of the flow diagrams of the intelligent prediction method for the microstructure in the special steel forging process based on machine learning provided by the embodiments of the present invention; Figure 2 FIG. 2 is another flow diagram of the intelligent prediction method for the microstructure in the special steel forging process based on machine learning provided by the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention in conjunction with the drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0047] The following will describe Figure 1 - Figure 2 the intelligent prediction method for the microstructure in the special steel forging process based on machine learning of the present invention.

[0048] As Figure 1 shown, the intelligent prediction method for the microstructure in the special steel forging process based on machine learning provided by the embodiments of the present invention includes: Collect the historical experimental data by gathering the data of special steel during previous forging processes, and extract the process parameter data, microstructure data, and dislocation movement data from the historical experimental data.

[0049] The process parameter data mainly includes forging temperature, deformation amount, strain rate, forging time, etc. These parameters directly affect the deformation behavior and microstructure evolution of special steel during forging.

[0050] The microstructure data is an important indicator reflecting the properties of special steel. Through advanced analytical and testing methods such as metallographic microscope, scanning electron microscope (SEM), transmission electron microscope (TEM), etc., detailed information about the microstructure can be obtained, such as grain size, shape, orientation, type, content, distribution of phases, and size, morphology, and position of the second-phase particles, etc.

[0051] The dislocation movement data reveals the plastic deformation mechanism of special steel during forging from the microscopic level. Dislocations, as an important type of defect in crystals, their behaviors such as movement, multiplication, and interaction play a key role in the deformation and strengthening of metals. Through experimental techniques such as dislocation density measurement and dislocation movement trajectory observation, relevant data on dislocation movement can be obtained, such as changes in dislocation density, dislocation velocity, and dislocation interaction frequency.

[0052] Such as Figure 2 As shown, obtain the influence coefficient according to the influence degree of the dislocation movement data on the dynamic recrystallization behavior, process the microstructure data according to the influence coefficient to obtain the optimized dynamic recrystallization behavior data, and adjust the process parameter data according to the optimized dynamic recrystallization behavior data to obtain the tissue type data.

[0053] The steps to obtain the influence coefficient include: Extract the dislocation density, dislocation velocity, and dislocation interaction frequency data from the dislocation movement data for normalization processing, unify the data with different dimensions to the same numerical range, and obtain the key dislocation data.

[0054] Based on the key dislocation data and the dynamic recrystallization theory, establish a mathematical model between the two to obtain the dislocation-based kinetic model.

[0055] The steps to obtain the dislocation-based kinetic model include: According to the dislocation theory, the increase in dislocation density promotes the nucleation of dynamic recrystallization. Establish the relationship between dislocation density and nucleation to obtain the nucleation change, and the formula is expressed as:

[0056] In the formula, is a constant related to the material and temperature, is the empirical exponent, is the nucleation, is the dislocation density.

[0057] The growth change is obtained based on the influence of dislocation velocity and temperature on grain growth, and the formula is expressed as:

[0058] In the formula, is a constant related to material properties, is the influence of temperature on the diffusion process, is the dislocation velocity, is the growth change.

[0059] By affecting the effective density of dislocations, the nucleation change and growth change are affected, the relationship between the dislocation interaction frequency and dynamic recrystallization is established, and the nucleation rate and growth rate are obtained. The formula is expressed as:

[0060]

[0061] In the formula, is the nucleation rate, is the growth rate, is the correction coefficient.

[0062] Based on the nucleation rate and growth rate and combined with the Avrami model, the dislocation dynamics model is obtained. The formula is expressed as:

[0063]

[0064] In the formula, is the recrystallized volume fraction, is a constant related to material and temperature, is a constant related to material properties, is the dislocation density, is a constant related to material and geometry, is the empirical exponent, is the gas constant, is the absolute temperature, is the activation energy, is the nucleation and growth situation, is the correction coefficient, is the time, is the dislocation velocity, is the dislocation interaction frequency, is the natural exponential function.

[0065] When using a dislocation dynamics model to calculate the change data of key dislocation data, the change amount of the index of the dynamic recrystallization behavior is used to measure the sensitivity of different key dislocation data to dynamic recrystallization, and the analysis results are obtained.

[0066] According to the analysis results and combined with the concept of the relative influence coefficient, the influence degree of dislocation movement on the dynamic recrystallization behavior is compared to obtain the influence coefficient.

[0067] The steps to obtain the data for optimizing the dynamic recrystallization behavior include: Analyze the microstructure data in terms of distribution characteristics and correlation to obtain the analyzed tissue data. Combine the influence coefficient with the analyzed tissue data to determine the parameters for adjusting the forging process to optimize the dynamic recrystallization behavior.

[0068] Input the parameters and key dislocation data into the dislocation dynamics model to simulate the change of the dynamic recrystallization behavior, and output the adjusted data.

[0069] Conduct a forging experiment on special steel according to the parameters, and then obtain the actual microstructure data through experimental means. Compare the adjusted data with the actual microstructure data to obtain the comparison difference data.

[0070] Adjust the microstructure data according to the comparison difference data to optimize the dynamic recrystallization behavior, thereby obtaining the data for optimizing the dynamic recrystallization behavior.

[0071] The steps to obtain the tissue type data include: Based on the internal relationship between the nucleation rate, growth rate and microstructure properties, obtain the characteristic manifestations of the microstructure under different dynamic recrystallization states.

[0072] According to the existing experience, sort out the action rules of the process parameter data on the dynamic recrystallization behavior during the forging of special steel to obtain the qualitative connection data.

[0073] According to the requirements of the data for optimizing the dynamic recrystallization behavior, combine the characteristic manifestations and the qualitative connection data, and consider the interaction and synergistic effect among the process parameters to formulate a process parameter adjustment plan.

[0074] Modify the process parameter data according to the process parameter adjustment plan to obtain the tissue type data.

[0075] Construct a generalized regression neural network model, input the microstructure data for training, and use the fruit fly optimization algorithm to optimize the hyperparameters of the generalized regression neural network model to obtain the optimized generalized regression neural network model. Input the tissue type data and output the microstructure prediction result.

[0076] The steps to obtain an optimized generalized regression neural network model include: fruit fly generation and movement, fruit fly algorithm optimization, and iterative optimization to construct the model.

[0077] Fruit fly generation and movement: Randomly generate a preset number of fruit fly individuals. Each fruit fly individual represents a set of hyperparameter combinations of the generalized regression neural network model, and define the initial position range of the fruit fly individuals in the search space, so that each fruit fly individual moves randomly within the initial position range.

[0078] Fruit fly algorithm optimization: Calculate the taste concentration determination value and perform model training and error calculation. Then, select the current optimal fruit fly individual. Finally, update the positions of all fruit fly individuals to obtain the corresponding updated positions.

[0079] The steps to obtain the updated positions include: For each fruit fly individual, calculate the taste concentration determination value based on the current position to obtain the concentration value. Split the microstructure data into a training set and a test set. Substitute the concentration value into the generalized regression neural network model, use the training set for training, and then use the test set to calculate the prediction error under the concentration value. The formula is expressed as:

[0080] In the formula, is the number of the test set, is the th true value of the sample, is the predicted value of the th sample by the generalized regression neural network model.

[0081] Calculate the concentration values of all fruit fly individuals, and find the fruit fly individual with the smallest concentration value as the current optimal fruit fly individual.

[0082] Let other fruit fly individuals fly towards the position of the current optimal fruit fly individual, and use the update formula to update their own positions to obtain the corresponding updated positions. The formula of the update formula is expressed as:

[0083] In the formula, is the step size of the fruit fly individual's flight, is the position of the current optimal fruit fly individual, is the th fruit fly's position in the search space at the th iteration.

[0084] Iterative optimization to construct the model: Repeat the fruit fly algorithm optimization until the preset number of iterations is reached. Select the hyperparameter combination corresponding to the fruit fly individual with the smallest concentration value to construct the generalized regression neural network model to obtain the optimized generalized regression neural network model.

[0085] The steps to obtain the prediction results of the microstructure include: The missing values in the tissue type data are corrected by the mean filling method, and for the existing non-numerical variables, the one-hot encoding method is used to convert the non-numerical variables into numerical data, so as to obtain the processed tissue type data.

[0086] The processed tissue type data is organized into a matrix form that conforms to the optimized generalized regression neural network model, and the matrix form is input into the optimized generalized regression neural network model.

[0087] According to the matrix form and the training set, the distance metric between the samples of the matrix form and the samples of the training set is calculated to obtain the output layer data.

[0088] The weighted sum of the output layer data is calculated to obtain the weighted data, the direct sum of the output layer data is calculated to obtain the sum data, and the weighted data is divided by the sum data to obtain the prediction result of the microstructure.

[0089] The intelligent prediction method for the microstructure in the special steel forging process based on machine learning provided in this embodiment collects the historical experimental data of the previous forging process of special steel, extracts multi-dimensional data, and establishes a dislocation dynamics model, reducing the dependence on a large amount of repetitive test data, and solving the problems of large demand for test data, long cycle, and high cost existing in the process trial method and the empirical formula derivation method. And considering the relationship between dislocation movement and dynamic recrystallization, multi-step processing and optimization are carried out on the microstructure data, and the fruit fly optimization algorithm is combined to optimize the generalized regression neural network model, improving the accuracy of microstructure prediction. It can also formulate a reasonable process parameter adjustment plan according to the influence coefficient of dislocation movement data on dynamic recrystallization and the analysis of microstructure data, providing a scientific basis for the optimization of the special steel forging process, and helping to obtain better microstructure and performance. The fruit fly optimization algorithm is also used to optimize the regression neural network model, which can accurately output the prediction results of the microstructure. While accurately predicting the microstructure, it also provides reliable prediction support for the special steel forging production and reduces the production risk.

[0090] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative work.

[0091] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent prediction method for the microstructure of special steel forging process based on machine learning, characterized in that, Including: Collect historical experimental data by collecting data of special steel in previous forging processes, and extract process parameter data, microstructure data, and dislocation movement data from the historical experimental data; Extract key dislocation data from the dislocation movement data, establish a dislocation kinetics model based on the dynamic recrystallization theory, calculate the index change amount, then analyze to obtain the influence coefficient, analyze the microstructure data, determine the parameters in combination with the influence coefficient, input the parameters into the dislocation kinetics model to obtain adjustment data, compare with the actual data to obtain comparison difference data, adjust the microstructure data according to the comparison difference data to obtain optimized dynamic recrystallization behavior data, and adjust the process parameter data according to the optimized dynamic recrystallization behavior data to obtain tissue type data; Construct a generalized regression neural network model, input the microstructure data for training, and use the fruit fly optimization algorithm to optimize the hyperparameters of the generalized regression neural network model to obtain an optimized generalized regression neural network model, input the tissue type data, and output the microstructure prediction result.

2. The intelligent prediction method for the microstructure of special steel forging process based on machine learning according to claim 1, wherein The steps for obtaining the influence coefficient include: Extract dislocation density, dislocation velocity, and dislocation interaction frequency from the dislocation movement data for normalization processing, unify data with different dimensions to the same numerical range to obtain the key dislocation data; Based on the key dislocation data and the dynamic recrystallization theory, establish a mathematical model between the two to obtain the dislocation kinetics model; Using the dislocation kinetics model, when calculating the change data of the key dislocation data, calculate the index change amount of the dynamic recrystallization behavior, and measure the sensitivity of different key dislocation data to dynamic recrystallization to obtain an analysis result; According to the analysis result and combined with the concept of relative influence coefficient, compare the influence degree of dislocation movement on the dynamic recrystallization behavior to obtain the influence coefficient.

3. The intelligent prediction method for the microstructure of special steel forging process based on machine learning according to claim 2, characterized in that, The steps for obtaining the dislocation kinetics model include: According to dislocation theory, the increase in dislocation density promotes the nucleation of dynamic recrystallization, establish the relationship between dislocation density and nucleation to obtain the nucleation change; Obtain the growth change according to the influence of dislocation velocity and temperature on grain growth; Influence the nucleation change and the growth change by affecting the effective density of dislocations, establish the relationship between the dislocation interaction frequency and the dynamic recrystallization to obtain the nucleation rate and growth rate; Based on the nucleation rate and the growth rate and combined with the Avrami model to obtain the dislocation kinetics model.

4. The intelligent prediction method for the microstructure of special steel forging process based on machine learning according to claim 3, wherein The dislocation kinetics model is expressed by the formula: Wherein, is the recrystallized volume fraction, is a constant related to the material and temperature, is a constant related to the material properties, is the dislocation density, is a constant related to the material and geometry, is the empirical exponent, is the gas constant, is the absolute temperature, is the activation energy, is the nucleation and growth situation, is the correction factor, is the time, is the dislocation velocity, is the dislocation interaction frequency, is the natural exponential function.

5. The intelligent prediction method for the microstructure of special steel forging process based on machine learning according to claim 1, characterized in that The steps for obtaining the optimized dynamic recrystallization behavior data include: Analyze the microstructure data from the aspects of distribution characteristics and correlation to obtain analyzed tissue data, and determine the parameters for adjusting the forging process to optimize the dynamic recrystallization behavior in combination with the influence coefficient according to the analyzed tissue data; Input the parameters and the key dislocation data into the dislocation kinetics model, simulate the change of the dynamic recrystallization behavior, and output to obtain the adjustment data; Conduct forging experiments on special steel according to the parameters, and then obtain actual microstructure data through experimental means. Compare the adjusted data with the actual microstructure data to obtain the comparison difference data; Adjust the microstructure data according to the comparison difference data to optimize the dynamic recrystallization behavior, thereby obtaining the optimized dynamic recrystallization behavior data.

6. The intelligent prediction method for the microstructure of special steel forging process based on machine learning according to claim 3, wherein, The steps for obtaining the tissue type data include: Based on the internal relationship between the nucleation rate, growth rate and microstructure properties, obtain the characteristic manifestations of the microstructure under different dynamic recrystallization states; According to existing experience, sort out the action rules of the process parameter data on the dynamic recrystallization behavior during the forging of special steel to obtain qualitative connection data; According to the requirements of the optimized dynamic recrystallization behavior data, combine the characteristic manifestations and the qualitative connection data, and consider the interaction and synergistic effect between process parameters to formulate a process parameter adjustment plan; Modify the process parameter data according to the process parameter adjustment plan to obtain the tissue type data.

7. The intelligent prediction method for the microstructure of special steel forging process based on machine learning according to claim 1, characterized in that The steps for obtaining the optimized generalized regression neural network model include: fruit fly generation movement, fruit fly algorithm optimization and iterative optimization to construct the model; Fruit fly generation movement: Randomly generate a preset number of fruit fly individuals. Each fruit fly individual represents a set of hyperparameter combinations of the generalized regression neural network model, and define the initial position range of the fruit fly individual in the search space, so that each fruit fly individual randomly moves within the initial position range; Fruit fly algorithm optimization: Calculate the taste concentration determination value and perform model training and error calculation, then screen the current optimal fruit fly individual, and finally update the positions of all fruit fly individuals to obtain the corresponding updated positions; Iterative optimization to construct the model: Repeat the fruit fly algorithm optimization until the preset number of iterations is reached, and select the hyperparameter combination corresponding to the fruit fly individual with the smallest concentration value to construct the generalized regression neural network model to obtain the optimized generalized regression neural network model.

8. The intelligent prediction method for the microstructure of special steel forging process based on machine learning according to claim 7, characterized in that, The steps for obtaining the updated position include: For each fruit fly individual, calculate the taste concentration determination value based on the current position to obtain the concentration value. Split the microstructure data into a training set and a test set, substitute the concentration value into the generalized regression neural network model, use the training set for training, and then use the test set to calculate the prediction error at the concentration value; Calculate the concentration values of all fruit fly individuals, and find the fruit fly individual with the smallest concentration value among them as the current optimal fruit fly individual; Make other fruit fly individuals fly towards the position of the current optimal fruit fly individual, and update their own positions using the update formula to obtain the corresponding updated positions.

9. The intelligent prediction method for the microstructure of special steel forging process based on machine learning according to claim 8, characterized in that, The update formula is expressed as: In the formula, is the step length of the flight of a Drosophila individual, is the position of the current optimal Drosophila individual, is the th Drosophila individual at the th iteration in the search space.

10. The intelligent prediction method for the microstructure of special steel forging process based on machine learning according to claim 8, characterized in that, The steps for obtaining the microstructure prediction result include: Correct the missing values in the tissue type data using the mean filling method, and for the existing non-numeric variables, use the one-hot encoding method to convert the non-numeric variables into numeric data, thereby obtaining the processed tissue type data; Organize the processed tissue type data into a matrix form that conforms to the optimized generalized regression neural network model, and input the matrix form into the optimized generalized regression neural network model; Calculate the distance metric between the samples in the matrix form and the samples in the training set according to the matrix form and the training set to obtain the output layer data; Perform weighted summation on the output layer data to obtain weighted data, perform direct summation on the output layer data to obtain summation data, and divide the weighted data by the summation data to obtain the prediction result of the microstructure.

Images