Roll shape optimization design method for three-roller planetary rotary pipe rolling mill

Abstract

The invention belongs to the technical field of three-roller planetary rotary rolling pipe design, and particularly relates to a three-roller planetary rotary rolling pipe mill roller shape optimization design method which specifically comprises the steps that performance parameters of a finished pipe and target parameters of all sections of a roller shape curve are determined; determining a target parameter value range of each section of the roller type curve, and designing a plurality of groups of experimental schemes based on the determined value range; multiple groups of experimental schemes are imported into a three-roller planetary rolling three-dimensional model, and the quantitative relation between different roller type curves and finished pipe performance parameters is obtained after finite element simulation; and based on a finite element simulation result, performing iterative optimization on the target parameter of each section of the roller type curve through an optimization algorithm to obtain an optimal value of the target parameter of each section of the roller type curve. According to the method, the numerical simulation technology and the intelligent algorithm are combined, rapid and intelligent roller shape optimization design of the three-roller planetary rotary pipe rolling mill is achieved, the production cost is effectively saved, and the preparation period is shortened.

Description

Technical Field

[0001] The invention belongs to the technical field of three-roller planetary spinning tube design, and in particular relates to a roller profile optimization design method for a three-roller planetary spinning tube machine. Background Art

[0002] Three-roll planetary rolling, a type of three-roll rolling technology, is a highly efficient and high-precision metal plastic processing technique primarily used for the roll forming of metal products such as pipes and bars. Its core feature is that the rolls simultaneously revolve and rotate during the rolling process, forming a complex planetary motion trajectory. This subjects the metal material to continuous compression and shear deformation in three dimensions, achieving uniform plastic flow and ensuring product precision and performance stability.

[0003] Traditional roll profile design relies heavily on empirical formulas, from which the roll profile parameters are calculated. However, three-roll planetary rolling involves numerous process parameters and roll profile parameters. The resulting parameter combinations using traditional design methods not only suffer from long calculation cycles and inflexible parameter optimization, but also result in unstable product quality and waste of raw materials and energy.

[0004] Therefore, there is an urgent need for a roll profile optimization design method for a three-roll planetary spinning tube mill with short calculation cycle, flexible parameter optimization configuration and stable product quality. Summary of the Invention

[0005] In response to the shortcomings of the above-mentioned existing technologies, the present invention provides a roller profile optimization design method for a three-roll planetary spinning tube mill to solve the problems of the existing design methods, such as long calculation cycle, inflexible parameter optimization configuration, unstable product quality, and easy waste of raw materials and energy.

[0006] The present invention determines the initial values of the target parameters of each section of the roll curve based on the new three-roll planetary rolling roll calculation equation, and then designs and optimizes the target parameters of each section of the roll curve by integrating finite element simulation, neural network, and optimization algorithm. This not only shortens the R&D cycle while reducing costs and improving production efficiency, but also significantly improves the comprehensive performance of the roll curve, providing solid technical support for the efficient spin-rolling of high-quality pipes.

[0007] The present invention provides a method for optimizing the roll profile of a three-roll planetary spinning tube mill, wherein the three-roll planetary rollers are segmented rollers, which are divided into a rounding section, a leveling section, a concentrated deformation section, and a diameter reducing section. The specific steps are as follows: S1. Determine the basic parameters of the tube blank, the performance parameters of the finished tube and the target parameters of each section of the roller profile curve, where the target parameters of each section of the roller profile curve include the length of the circular segment , averaged arc chord slope , the slope of the average straight line , the slope of the arc chord of the concentrated deformation segment and the slope of the concentrated deformation segment line , , , , , ; S2. Determine the initial values of target parameters of each section of the roll profile curve based on the three-roll planetary rolling process parameters and the new three-roll planetary rolling roll profile calculation equation; S3. Determine the value range of the target parameter of each section of the roller profile curve based on the initial value of the target parameter of each section of the roller profile curve; S4. Design multiple experimental schemes based on the target parameter value range of each section of the roller profile curve; S5. Establish a three-roll planetary rolling 3D model, and import multiple experimental schemes into the three-roll planetary rolling 3D model. After finite element simulation, obtain the quantitative relationship between different roll profile curves and the performance parameters of the finished pipe; S6. Constructing a neural network-based prediction model for performance parameters of finished pipes, wherein the prediction model uses target parameters of each section of the roll profile curve as an input layer and the performance parameters of the finished pipe as an output layer; S7. Based on the finite element simulation results in step S5, the performance parameters of the finished pipe are used as the optimization target, and the target parameters of each section of the roll profile curve are used as the optimization variables. The target parameters of each section of the roll profile curve are iteratively optimized through an optimization algorithm to obtain the optimal value of the target parameters of each section of the roll profile curve. The fitness value of the individual in the iterative optimization process is obtained by prediction using the finished pipe performance parameter prediction model based on the neural network constructed in step S6.

[0008] Preferably, the basic parameters of the tube blank include the tube blank type, tube blank diameter, tube blank wall thickness, target diameter of the finished tube, and target wall thickness of the finished tube; the performance parameters of the finished tube include roundness and recrystallization temperature, wherein the roundness is calculated according to national standards, and the calculation formula is as follows: , The recrystallization temperature is calculated based on the physical properties of the tube type. The calculation formula is as follows: , Where D is the roundness, It is the outer diameter size deviation of the pipe specified in the national standard. is the recrystallization temperature of the tube material, It is the melting point of the tube material.

[0009] Preferably, the three-roll planetary rolling process parameters in step S2 include the tilt angle, the axial forward displacement of each roll in the primary deformation zone, the speed at the tube inlet, and the main plate revolution speed. The specific steps of determining the initial values of the target parameters of each segment of the roll profile curve based on the three-roll planetary rolling process parameters and the new three-roll planetary rolling roll profile calculation equation are as follows: S21, circle segment S211, the circle segment curve is as follows: , Where, is the tilt angle, is the minimum radius of the circle segment; S212, the length of the circle segment, as shown in the following formula: , Where, is the length of the compass circle segment; is the roundness coefficient, ranging from 1 to 2.5, and increases with the gradual increase of the tube diameter; S22, leveling section S221, even segment arc curve Slope of the chord of the averaged arc segment: , The midpoint of the arc chord B( , ): , Arc center ( , ): , The equation of the uniform arc curve: , , Where, is the angle between the chord of the uniform arc and the perpendicular line, The angle set for artificial experience, is the uniform length, is the arc radius of the uniform segment; S222, even straight line Arc endpoint C( , ): , Slope of the average straight line: , Equation of the average straight line: , Where, is the angle between the straight line and the perpendicular line, The perspective set for artificial experience; S223, the length of the equalized segment is as follows: , , Where, is the total extension coefficient, The distance that each roller moves forward in the axial direction during one deformation zone rolling; is the velocity at the pipe inlet, m / s; is the revolution speed of the disk, / rpm; is the number of rollers, N=3; S23, concentrated deformation section S231, concentrated deformation segment arc curve Arc endpoint D( , ): , Slope of arc chord of concentrated deformation segment: , Coordinates of the midpoint of the arc chord E( , ): , Arc center ( , ): , The arc curve equation of the concentrated deformation segment: , , Where, is the angle between the chord of the arc of the concentrated deformation segment and the vertical line, The angle set for artificial experience, is the length of the concentrated deformation section, is the arc radius of the concentrated deformation segment; S232, concentrated deformation segment straight line Arc endpoint F( , ): , Slope of the concentrated deformation segment line: , The equation of the straight line of the concentrated deformation segment: , Where, is the angle between the concentrated deformation segment straight line and the vertical line, The perspective set for artificial experience; S24, reducing section Slope of the straight line of the reducing section: , End point of reducing section : , The equation of the reducing section: .

[0010] Preferably, the range of the length of the rule circle segment is selected on both sides of its initial value; the slope of the chord of the uniform segment arc, the slope of the straight line of the uniform segment arc, the slope of the chord of the concentrated deformation segment arc and the slope of the straight line of the concentrated deformation segment are satisfied. 、 Under the conditions, The value range is selected from both sides of its initial value.

[0011] Preferably, step S5 is specifically as follows: S51. Establish a three-roll planetary rolling three-dimensional model and set simulation parameters, wherein the simulation parameters include an inclination angle, a deflection angle, a friction coefficient between the rolls and the tube, a roll speed, a thermal conductivity coefficient of heat conduction between the rolls and the tube, and a thermal conductivity coefficient of heat conduction between the tube and air; S52. Multiple experimental schemes are introduced into the three-roll planetary rolling 3D model, and the quantitative relationship between different roller profile curves and the performance parameters of the finished pipe is obtained through finite element simulation.

[0012] Preferably, step S6 is specifically as follows: S61. Arrange multiple groups of experimental schemes and finite element simulation output corresponding to each group of experimental schemes, and perform normalization processing on them; S62: Divide the normalized data into training set, validation set and test set; S63: Inputting the training set data into the prediction model based on the neural network for training to obtain an initial prediction model; S64: Input the validation set data into the initial prediction model for testing. If the test criteria are met, a trained prediction model is obtained; otherwise, return to step S63 to continue training until the test criteria are met and training is stopped. S65: Input the test set data into the trained prediction model, and compare its finite element simulation results with the results predicted by the trained prediction model. If the predicted results fit the finite element simulation results well, the test ends and a neural network-based prediction model for finished pipe performance parameters is obtained; otherwise, steps S63-S65 are executed again.

[0013] Preferably, in step S4, uniform experimental design and orthogonal experimental methods are used to design multiple experimental schemes; In addition, the neural network in step S6 is a BP neural network based on the Levenberg-Marquardt optimization algorithm, the hidden layer is a Sigmoid activation function, and the output layer is a two-layer feedforward network with linear output neurons. The neural network threshold and weight matrix of the Purelin activation function are selected as the neural network matrix file and stored in the database.

[0014] Preferably, in step S7, the target parameters of each section of the roll profile curve are iteratively optimized by a genetic algorithm. Step S7 is specifically as follows: S71. Determine optimization variables, value ranges of optimization variables, and optimization targets, wherein the optimization variables are target parameters of each section of the roll profile curve, and the optimization targets are performance parameters of the finished pipe; S72. Set the population size to N P , and initialize to generate the initial population; S73, encoding each individual in the population and calculating the fitness value of each individual, wherein each individual includes all optimization variables, and the fitness value of each individual is predicted by the neural network model; S74. After performing selection, crossover, and mutation operations on individuals in the population, a new generation of population is generated; S75. Determine whether the termination condition is met. If so, terminate the evolution and output the best individual obtained as the optimal solution; if not, return to step S73 to continue optimization.

[0015] Preferably, the value range of the optimized variable in step S7 is obtained by the designer after adjusting the value range of the target parameter of each section of the roll profile determined in step S3 based on the quantitative relationship between different roll profile curves and the performance parameters of the finished pipe obtained in step S5; the termination condition is that the iterative search tends to be stable and the optimal solution is found.

[0016] Preferably, a database is also designed to store the initial design data, standard data, finite element simulation design and result data of the three-roll planetary spinning process, the design and result data of the prediction model established based on the neural network, and the result data iterated by the optimization algorithm, wherein the standard data includes national standards for products of various specifications.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention adopts a method that combines numerical simulation technology and intelligent algorithms, which not only solves the problems of long parameter combination calculation cycle, inflexible parameter optimization configuration, unstable product quality, and easy waste of raw materials and energy obtained by traditional design methods, but also realizes the rapid and intelligent optimization design of the roll shape of the three-roll planetary spinning tube mill, effectively saves production costs, and shortens the preparation cycle.

[0018] 2. Based on a novel three-roll planetary rolling roll profile calculation equation, the present invention can achieve efficient and precise design of target parameters for each section of the three-roll planetary rolling roll profile curve according to the finished tube specification parameters and the three-roll planetary rolling process parameters. This not only reduces the complexity and difficulty of design and shortens the design cycle, but also lays a solid foundation for subsequent optimization.

[0019] 3. The present invention not only effectively handles the complex boundary conditions and nonlinear problems in the three-roll planetary spinning process through finite element simulation, but also reduces costs and improves production efficiency while shortening the R&D cycle.

[0020] 4. The prediction model established based on the neural network of the present invention, with its powerful nonlinear solving ability, effectively maps the complex linear relationship between the target parameters of each section of the roll profile and the performance parameters of the finished pipe, significantly enhancing the accuracy and reliability of predicting the performance parameters of different roll profiles.

[0021] 5. The present invention simulates the biological natural selection process through genetic algorithms, quickly finds the optimal solution from the target parameters of each section of the complex roll curve, with fast convergence speed and high calculation efficiency. At the same time, in the process of multiple rounds of iterative evolution of the population, it innovatively introduces a prediction model based on neural network construction to efficiently calculate the fitness value of each individual in the genetic algorithm. Through continuous iterative optimization, it finally accurately determines the optimal value of the target parameters of each section of the roll curve of the three-roll planetary spinning tube mill, significantly improving the comprehensive performance of the roll curve and providing solid technical support for the efficient spinning of high-quality pipes.

[0022] 6. The present invention designs a database and continuously accumulates data sets, so that the database presents a dynamic and stable optimization trend, effectively ensuring the continuous improvement of production efficiency and product quality.

[0023] 7. The present invention optimizes the design of the roll profile of a three-roll planetary tube spinning mill based on a new three-roll planetary rolling roll profile calculation equation and finite element simulation + neural network + optimization algorithm. It not only overcomes the defect that the traditional roll profile theoretical model of three-roll planetary rolling does not perform specific analysis on the rolled material, greatly enhances the adaptability of the roll profile design formula, but also has better comprehensive performance than the design of the roll profile of a three-roll planetary tube spinning mill using only the new three-roll planetary rolling roll profile calculation equation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 This is a flow chart of a method for optimizing the roll profile of a three-roll planetary tube spinning mill according to an embodiment of the present invention; Figure 2 Schematic diagram of roller profile parameters of three-roller planetary spinning rollers in an embodiment of the present invention; Figure 3Schematic diagram of the three-roller planetary spinning roller segmentation in an embodiment of the present invention; Figure 4 This is a partial enlarged view of the circle in the schematic diagram of roller profile parameters of the three-roller planetary spinning roller in an embodiment of the present invention; Figure 5 It is a partial enlarged view of the uniform section of the three-roller planetary spinning roller in an embodiment of the present invention; Figure 6 It is a partial enlarged view of the concentrated deformation section of the three-roller planetary spinning roller in an embodiment of the present invention; Figure 7 1 is a finite element model diagram of a three-roller planetary spinning roller in an embodiment of the present invention; Figure 8 A graph showing the neural network training results in an embodiment of the present invention; Figure 9 4 is a diagram of the iterative process of the genetic algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0027] See Figures 1-9 The present invention provides a method for optimizing the roll profile of a three-roll planetary spinning tube mill, wherein the three-roll planetary rollers are segmented rollers, which are divided into a round section, a leveling section, a concentrated deformation section, and a diameter reduction section. The specific steps are as follows: S1. Determine the basic parameters of the tube blank, the performance parameters of the finished tube and the target parameters of each section of the roller profile curve.

[0028] In this application, the basic parameters of the tube blank include the tube blank type, tube blank diameter, tube blank wall thickness, target diameter of the finished tube and target wall thickness of the finished tube.

[0029] In this application, the performance parameters of the finished tube include roundness and recrystallization temperature.

[0030] In the embodiment of the present application, the roundness is calculated according to the national standard, and the calculation formula is as follows: , Where D is the roundness, It is the outer diameter size deviation of the pipe specified in the national standard.

[0031] In the embodiment of the present application, the recrystallization temperature is calculated based on the physical properties of the tube type, and the calculation formula is as follows: , Where, is the recrystallization temperature of the tube material, It is the melting point of the tube material.

[0032] In the embodiment of the present application, the roundness is obtained according to the national standard GBT 3639-2021 cold-rolled precision seamless steel pipe.

[0033] For example, taking 304 stainless steel as an example, the tube diameter is 32mm, the tube wall thickness is 4mm, the target diameter of the finished tube is 21mm, and the target wall thickness of the finished tube is 2mm. A new rolling model is established. According to the national standard GBT 3639-2021 cold-rolled precision seamless steel tube, the outer diameter size deviation of the tube is ≤ 0.08mm, and the out-of-roundness of the steel tube is ≤ 80% of the outer diameter tolerance. Therefore, in the embodiment of this application, the roundness of the finished tube is ≤ , that is, the roundness of the finished tube is ≤0.064. In addition, the melting point of 304 stainless steel is 1440℃. , it can be seen that the recrystallization temperature of 304 stainless steel needs to be ≥432℃~576℃, that is, the recrystallization temperature of 304 stainless steel needs to be ≥705K~850K.

[0034] In this application, the target parameters of each section of the roller profile curve include the length of the circular segment , averaged arc chord slope , the slope of the average straight line , the slope of the arc chord of the concentrated deformation segment and the slope of the concentrated deformation segment line ,in , , , , .

[0035] S2. Based on the three-roll planetary rolling process parameters and the new three-roll planetary rolling roll profile calculation equation, determine the initial values of the target parameters of each section of the roll profile curve.

[0036] In the present application, the three-roll planetary rolling process parameters include but are not limited to the tilt angle, the axial forward displacement distance of each roller in rolling the deformation zone, the speed at the pipe inlet and the rotation speed of the large plate.

[0037] See Figure 2-Figure 6 The specific steps of determining the initial values of the target parameters of each section of the roll profile curve based on the three-roll planetary rolling process parameters and the new three-roll planetary rolling roll profile calculation equation in step S2 are as follows: S21, circle segment S211, the circle segment curve is as follows: , Where, is the tilt angle, is the minimum radius of the circle segment; S212, the length of the circle segment, as shown in the following formula: , Where, is the length of the compass circle segment; is the roundness coefficient, ranging from 1 to 2.5, and increases with the gradual increase of the tube diameter; S22, leveling section S221, even segment arc curve Slope of the chord of the averaged arc segment: , The midpoint of the arc chord B( , ): , Arc center ( , ): , The equation of the uniform arc curve: , , Where, is the angle between the chord of the uniform arc and the perpendicular line, The angle set for artificial experience, is the uniform length, is the arc radius of the uniform segment; S222, even straight line Arc endpoint C( , ): , Slope of the average straight line: , Equation of the average straight line: , Where, is the angle between the straight line and the perpendicular line, The perspective set for artificial experience; S223, the length of the equalized segment is as follows: , , Where, is the total extension coefficient, The distance that each roller moves forward in the axial direction during one deformation zone rolling; is the velocity at the pipe inlet, m / s; is the revolution speed of the disk, / rpm; is the number of rollers, N=3; S23, concentrated deformation section S231, concentrated deformation segment arc curve Arc endpoint D( , ): , Slope of arc chord of concentrated deformation segment: , Coordinates of the midpoint of the arc chord E( , ): , Arc center ( , ): , The arc curve equation of the concentrated deformation segment: , , Where, is the angle between the chord of the arc of the concentrated deformation segment and the vertical line, The angle set for artificial experience, is the length of the concentrated deformation section, is the arc radius of the concentrated deformation segment; S232, concentrated deformation segment straight line Arc endpoint F( , ): , Slope of the concentrated deformation segment line: , The equation of the straight line of the concentrated deformation segment: , Where, is the angle between the concentrated deformation segment straight line and the vertical line, The perspective set for artificial experience; S24, reducing section Slope of the straight line of the reducing section: , End point of reducing section : , The equation of the reducing section: .

[0038] It should be noted that, except for the target parameters of each section of the roller profile curve in step S2, the rest are known parameters. Figure 5-Figure 6 The H in is a horizontal line and has no special meaning.

[0039] In this application, the slope of the straight line of the reducing section It is known that it is equal to the tangent of the cone base angle. The total height of the roller is L = 55.43 mm, and L = + + + .

[0040] In the embodiment of the present application, the initial values of the target parameters of each section of the roller profile curve are: =0.8mm, , , , .

[0041] Traditional roller shape design mostly designs planetary rollers into three sections: wall reduction section, leveling section and rounding section, while the new three-roller planetary rolling roller shape calculation equation used in the present invention designs it into four sections: rounding section, leveling section, concentrated deformation section and diameter reduction section.

[0042] It should be noted that, since the slope of the straight line equation of the diameter reduction section is a known parameter, which is equal to the tangent of the cone base angle, the target parameters of each section of the roller profile curve in this application do not include the diameter reduction section parameters.

[0043] S3. Based on the initial values of the target parameters of each section of the roller profile curve, determine the value range of the target parameters of each section of the roller profile curve.

[0044] In this application, the value range of the target parameters of each section of the roller profile curve is determined through conventional experience.

[0045] Preferably, the range of the length of the rule circle segment is selected on both sides of its initial value; the slope of the chord of the uniform segment arc, the slope of the straight line of the uniform segment arc, the slope of the chord of the concentrated deformation segment arc and the slope of the straight line of the concentrated deformation segment are satisfied. 、 Under the conditions, The value range is selected from both sides of its initial value.

[0046] It should be noted that since the function of the rounding segment is to round the tube blank, the length of the rounding segment should not be too short. In this application, the range of the length of the rounding segment should be selected above its initial value, that is, above the initial value of 0.8mm. However, in the examples of this application, in order to explore the specific extent to which "not too short" in the standard cannot be too short, the examples of this application also designed two groups of 0.4mm and 0.6mm below its initial value, that is, on the left side.

[0047] In the embodiment of the present application, in step S3 The value range is 0.4mm~2mm, The value range is 34°~40°. The value range is 26°~36°, The value range is 22°~30°. The value range is -5°~0°.

[0048] It should be emphasized that the slope of the gage segment and the tangent of the inclination angle are equal, so as to ensure that the gage segment and the tube are parallel. Since the slope of the uniform segment cannot exceed the slope of the gage segment, the angle between the chord of the uniform segment and the vertical line cannot exceed the inclination angle. .

[0049] In the embodiment of this application, ,but .

[0050] S4. Design multiple groups of experimental schemes based on the value range of target parameters in each section of the roller profile curve.

[0051] It should be noted that due to are slopes. In finite element simulation, the slope is converted into an angle for the convenience of calculation. 、 、 、 .

[0052] In this application, 44 groups of experimental schemes were designed using the uniform experimental design and orthogonal experimental methods, as shown in Table 1.

[0053] Table 1. 44 experimental plans

[0054]

[0055] S5. Establish a three-roll planetary rolling three-dimensional model, and import multiple groups of experimental schemes into the three-roll planetary rolling three-dimensional model. After finite element simulation, obtain the quantitative relationship between different roller profile curves and the performance parameters of the finished pipe.

[0056] The finite element model of the three-roll planetary spinning roller constructed in this application is shown in FIG. Figure 7 shown.

[0057] Because the core characteristic of three-roll planetary rolling is that the rollers simultaneously revolve and rotate during the rolling process, forming complex planetary motion trajectories, the three-roll planetary rolling process is subject to complex boundary conditions and nonlinear issues. This application establishes a three-dimensional model of three-roll planetary rolling. Through finite element simulation, it not only effectively addresses the complex boundary conditions and nonlinear issues in the three-roll planetary rolling process, but also shortens the R&D cycle, reduces costs, and improves production efficiency.

[0058] In this application, step S5 is specifically as follows: S51. Establish a three-roll planetary rolling three-dimensional model and set simulation parameters, wherein the simulation parameters include the inclination angle, the deflection angle, the friction coefficient between the rolls and the tube, the roll speed, the thermal conductivity of heat conduction between the rolls and the tube, and the thermal conductivity of heat conduction between the tube and air.

[0059] In the embodiment of the present application, the inclination angle is 41°, the deflection angle is 15°, the friction coefficient between the roller and the tube is 0.3, the roller rotation speed is 18.85 rad / s, the thermal conductivity of heat conduction between the roller and the tube is 21.9 W / m•K, and the thermal conductivity of heat conduction between the tube and the air is 0.024 W / m•K.

[0060] S52. Multiple experimental schemes are introduced into the three-roll planetary rolling 3D model, and the quantitative relationship between different roller profile curves and the performance parameters of the finished pipe is obtained through finite element simulation.

[0061] In the embodiments of the present application, the quantitative relationship between different roller profile curves and the roundness of the finished tube after rolling is shown in Table 2, and the quantitative relationship between different roller profile curves and the recrystallization temperature of the deformation zone of the finished tube after rolling is shown in Table 3.

[0062]

[0063]

[0064] Table 2 clearly shows that the roundness of the finished tubes after rolling is within a range of 0.0155 mm to 0.4187 mm, with many groups of finished tubes exhibiting roundness greater than 0.064. Table 3 clearly shows that the recrystallization temperature of the deformation zone of the finished tubes after rolling is within a range of 915 K to 1221 K, all above 705 K to 850 K. Therefore, in this embodiment, only the roundness of the finished tubes after rolling is optimized.

[0065] It should be noted that the three-roll planetary rolling mill adopts segmented rolling mill rolls, which can meet the recrystallization temperature requirements of most rolling processes. The recrystallization temperatures obtained from the 44 groups of experimental schemes in the embodiments of this application also further confirm this principle. Therefore, the optimization of the recrystallization temperature is no longer considered in the subsequent steps. Therefore, the performance parameters of the finished pipe in steps S6-S7 in this application all refer to roundness.

[0066] S6. Construct a finished tube performance parameter prediction model based on a neural network, wherein the prediction model uses the target parameters of each section of the roller profile curve as an input layer and the finished tube performance parameters as an output layer.

[0067] In this application, step S6 is specifically as follows: S61. Arrange multiple groups of experimental schemes and the finite element simulation output corresponding to each group of experimental schemes, and normalize them.

[0068] S62: Divide the normalized data into a training set, a validation set, and a test set.

[0069] In the embodiment of the present application, 44 groups of experimental schemes and their corresponding result outputs are randomly divided into 33 training sets, 6 validation sets and 5 test sets.

[0070] It should be noted that if there are a large number of experimental data sets, those skilled in the art may also divide the data into training sets, validation sets, and test sets in proportion.

[0071] S63: Input the training set data into the prediction model based on the neural network for training to obtain an initial prediction model.

[0072] In this application, the neural network in step S6 is a BP neural network based on the Levenberg-Marquardt optimization algorithm, the hidden layer is a Sigmoid activation function, and the output layer is a two-layer feedforward network with linear output neurons. The neural network threshold and weight matrix of the Purelin activation function are selected as the neural network matrix file and stored in the database.

[0073] In the embodiment of the present application, the neural network structure for predicting the roundness of the finished tube by three-roll planetary rolling is as follows: the input layer is 5 nodes, including the length of the rule circle segment, the slope of the arc chord of the uniform segment, the slope of the straight line of the uniform segment, the slope of the arc chord of the concentrated deformation segment, and the slope of the straight line of the concentrated deformation segment; the output layer is 1 node, which is the roundness; and the hidden layer is 10 layers.

[0074] S64: Input the validation set data into the initial prediction model for testing. If the test criteria are met, a trained prediction model is obtained; otherwise, return to step S63 to continue training until the test criteria are met and the training is stopped.

[0075] In the embodiment of the present application, the test standard is to verify the fitting factor , those skilled in the art can set it themselves according to their needs in actual applications.

[0076] S65: Input the test set data into the trained prediction model, and compare its finite element simulation results with the results predicted by the trained prediction model. If the predicted results fit the finite element simulation results well, the test ends and a neural network-based prediction model for finished pipe performance parameters is obtained; otherwise, steps S63-S65 are executed again.

[0077] In the embodiment of the present application, the criterion for good fitting effect between the prediction result and the finite element simulation result is the test fitting factor , those skilled in the art can set it themselves according to their needs in actual applications.

[0078] like Figure 8As shown, the training fitting factor of the trained neural network , verify the fitting factor , test fitting factor , the total fitting factor of 44 experimental schemes ,visible Both are greater than 0.9, so the neural network trained in this application has a good fitting effect.

[0079] In the embodiment of the present application, a group of experimental schemes were randomly set up to input the finished pipe performance parameter prediction model based on the neural network to detect whether it can operate normally. The roundness predicted by this group of experimental schemes is 0.1230, which shows that the prediction model can operate normally.

[0080] It should be noted that since all 44 experimental protocols are used throughout the training, validation, and testing of the neural network, if the imported 44 experimental protocols are used when testing the program for proper operation, it is impossible to determine whether the results are data predicted after neural network training or directly derived from the input data program. Therefore, in the examples of this application, another set of experimental protocols is randomly set to test whether the program can operate properly.

[0081] Preferably, each time a new finite element simulation result is obtained, it is normalized and integrated into the existing training sample file. Subsequently, the neural network is retrained using the updated sample file. If the training result meets the error requirements, the newly generated neural network threshold and weight matrix are stored in the database as a neural network matrix file.

[0082] In this way, the present application enables the training sample files and the neural network matrix files to be continuously updated as the finite element simulation data continues to increase, thereby ensuring the accuracy and adaptability of the model.

[0083] S7. Based on the finite element simulation results in step S5, the performance parameters of the finished pipe are used as the optimization target, and the target parameters of each section of the roll profile curve are used as the optimization variables. The target parameters of each section of the roll profile curve are iteratively optimized through an optimization algorithm to obtain the optimal value of the target parameters of each section of the roll profile curve. The fitness value of the individual in the iterative optimization process is obtained by prediction using the finished pipe performance parameter prediction model based on the neural network constructed in step S6.

[0084] Preferably, in step S7, the target parameters of each section of the roller profile curve are iteratively optimized by a genetic algorithm.

[0085] Genetic algorithms (GAs) are a type of randomized search method derived from the laws of evolution in the biological world, employing a genetic mechanism of survival of the fittest. First proposed by American Professor J. Holland in 1975, GAs operate directly on structural objects, eliminating the need for derivatives and function continuity. They possess inherent implicit parallelism and enhanced global optimization capabilities. They employ a probabilistic optimization approach, automatically acquiring and guiding the optimized search space and adaptively adjusting the search direction without the need for fixed rules. These properties of GAs have been widely applied in fields such as combinatorial optimization, machine learning, signal processing, adaptive control, and artificial life, and represent a key technology in modern intelligent computing. Therefore, this algorithm is well known to those skilled in the art and will not be described in detail in this application.

[0086] In this application, in step S7, the target parameters of each section of the roll profile curve are iteratively optimized by a genetic algorithm. Step S7 is specifically as follows: S71. Determine optimization variables, optimization variable value ranges, and optimization targets, wherein the optimization variables are target parameters of each section of the roller profile curve, and the optimization targets are performance parameters of the finished pipe.

[0087] Preferably, the optimized variable value range in step S7 is obtained by the designer adjusting the target parameter value range of each section of the roll profile determined in step S3 based on the quantitative relationship between different roll profile curves and finished pipe performance parameters obtained in step S5.

[0088] In the embodiment of the present application, in step S7 The value range is 0.4mm~2mm, The value range is 34°~40°. The value range is 29°~32°, The value range is 20°~30°. The value range is -5°~0°.

[0089] S72. Set the population size to N P , and initialize to generate the initial population.

[0090] S73. Encode each individual in the population and calculate the fitness value of each individual, wherein each individual includes all optimized variables, and the fitness value of each individual is predicted by the neural network model.

[0091] S74. After performing selection, crossover and mutation operations on individuals in the population, a new generation of population is generated.

[0092] S75. Determine whether the termination condition is met. If so, terminate the evolution and output the best individual obtained as the optimal solution; if not, return to step S73 to continue optimization.

[0093] In the embodiment of this application, the population size N P The value ranges from 50 to 200, the hybridization probability is 0.75, the mutation probability is 0.08, the maximum evolution iteration is 200, and the termination condition is that the iteration tends to be stable and the optimal solution is found.

[0094] It should be noted that, in actual use, those skilled in the art can set the parameters and termination conditions according to their actual needs. In addition, those skilled in the art can also select a suitable optimization algorithm in actual application.

[0095] like Figure 9 As shown, in the embodiment of the present application, the population evolves from generation to generation until the 129th generation tends to stabilize and search for the optimal solution, thereby determining the optimal value of each optimization variable, as shown in Table 4.

[0096]

[0097] The 、 、 、 After converting the angle to the slope, the conversion results are shown in Table 5.

[0098]

[0099] To verify the accuracy of the optimization results, this application used a three-roll planetary rolling 3D model and the optimal values of each optimization variable to perform finite element simulations. The simulation results showed that the finished tube diameter was 20.9852 mm (approximately 21 mm), the target diameter of the finished tube was 21 mm, the wall thickness was 2 mm (approximately 2 mm), the roundness was 0.0486 (less than 0.064), and the recrystallization temperature was 1041 K to 1207 K, all above the 705 K to 850 K range. This indicates that the simulation results are generally consistent with the expected results and meet the performance parameter requirements for the finished tube. Therefore, the optimization method designed in this application is effective.

[0100] In this application, a database is also designed to store the initial design data, standard data, finite element simulation design and result data of the three-roll planetary spinning process, the design and result data of the prediction model established based on the neural network, and the result data iterated by the optimization algorithm, among which the standard data includes national standards for products of various specifications.

[0101] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for optimizing the roll profile of a three-roll planetary spinning mill, wherein the three-roll planetary rollers are segmented rollers, which are divided into a rounding section, a leveling section, a concentrated deformation section, and a diameter-reducing section, characterized in that: The specific steps are: S1. Determine the basic parameters of the tube blank, the performance parameters of the finished tube and the target parameters of each section of the roller profile curve, where the target parameters of each section of the roller profile curve include the length of the circular segment , averaged arc chord slope , the slope of the average straight line , the slope of the arc chord of the concentrated deformation segment and the slope of the concentrated deformation segment line , , , , , ; S2. Determine the initial values of target parameters of each section of the roll profile curve based on the three-roll planetary rolling process parameters and the new three-roll planetary rolling roll profile calculation equation; S3. Determine the value range of the target parameter of each section of the roller profile curve based on the initial value of the target parameter of each section of the roller profile curve; S4. Design multiple experimental schemes based on the target parameter value range of each section of the roller profile curve; S5. Establish a three-roll planetary rolling 3D model, and import multiple experimental schemes into the three-roll planetary rolling 3D model. After finite element simulation, obtain the quantitative relationship between different roll profile curves and the performance parameters of the finished pipe; S6. Constructing a neural network-based prediction model for performance parameters of finished pipes, wherein the prediction model uses target parameters of each section of the roll profile curve as an input layer and the performance parameters of the finished pipe as an output layer; S7. Based on the finite element simulation results in step S5, the performance parameters of the finished pipe are used as the optimization target, and the target parameters of each section of the roll profile curve are used as the optimization variables. The target parameters of each section of the roll profile curve are iteratively optimized through an optimization algorithm to obtain the optimal value of the target parameters of each section of the roll profile curve. The fitness value of the individual in the iterative optimization process is obtained by prediction using the finished pipe performance parameter prediction model based on the neural network constructed in step S6.

2. The method for optimizing the roll profile of a three-roll planetary spinning tube mill according to claim 1, characterized in that: The basic parameters of the tube blank include tube blank type, tube blank diameter, tube blank wall thickness, target diameter of the finished tube, and target wall thickness of the finished tube. The performance parameters of the finished tube include roundness and recrystallization temperature. The roundness is calculated according to national standards. The calculation formula is as follows: ， The recrystallization temperature is calculated based on the physical properties of the tube type. The calculation formula is as follows: ， Where D is the roundness, It is the outer diameter size deviation of the pipe specified in the national standard. is the recrystallization temperature of the tube material, It is the melting point of the tube material.

3. The method for optimizing the roll profile of a three-roll planetary spinning tube mill according to claim 1, characterized in that: In step S2, the three-roll planetary rolling process parameters include the tilt angle, the axial forward displacement of each roll in the primary deformation zone, the speed at the tube inlet, and the main plate revolution speed. The specific steps for determining the initial values of the target parameters of each section of the roll profile curve based on the three-roll planetary rolling process parameters and the new three-roll planetary rolling roll profile calculation equation are as follows: S21, circle segment S211, the circle segment curve is as follows: ， Where, is the tilt angle, is the minimum radius of the circle segment; S212, the length of the circle segment, as shown in the following formula: ， Where, is the length of the compass circle segment; is the roundness coefficient, ranging from 1 to 2.5, and increases with the gradual increase of the tube diameter; S22, leveling section S221, even segment arc curve Slope of the chord of the averaged arc segment: , The midpoint of the arc chord B( , ): , Arc center ( , ): , The equation of the uniform arc curve: , , Where, is the angle between the chord of the uniform arc and the perpendicular line, The angle set for artificial experience, is the uniform length, is the arc radius of the uniform segment; S222, even straight line Arc endpoint C( , ): , Slope of the average straight line: , Equation of the average straight line: , Where, is the angle between the straight line and the perpendicular line, The perspective set for artificial experience; S223, the length of the equalized segment is as follows: ， ， Where, is the total extension coefficient, The distance that each roller moves forward in the axial direction during one deformation zone rolling; is the velocity at the pipe inlet, m / s; is the revolution speed of the large plate, / rpm; is the number of rollers, N=3; S23, concentrated deformation section S231, concentrated deformation segment arc curve Arc endpoint D( , ): , Slope of arc chord of concentrated deformation segment: , Coordinates of the midpoint of the arc chord E( , ): , Arc center ( , ): , The arc curve equation of the concentrated deformation segment: , , Where, is the angle between the chord of the arc of the concentrated deformation segment and the vertical line, The angle set for artificial experience, is the length of the concentrated deformation section, is the arc radius of the concentrated deformation segment; S232, concentrated deformation segment straight line Arc endpoint F( , ): , Slope of the concentrated deformation segment line: , The equation of the straight line of the concentrated deformation segment: , Where, is the angle between the concentrated deformation segment straight line and the vertical line, The perspective set for artificial experience; S24, reducing section Slope of the straight line of the reducing section: , End point of reducing section : , The equation of the reducing section: .

4. The method for optimizing the roll profile of a three-roll planetary spinning tube mill according to claim 1, wherein: The range of the length of the rule circle segment is selected on both sides of its initial value; the slope of the chord of the uniform segment arc, the slope of the straight line of the uniform segment arc, the slope of the chord of the concentrated deformation segment arc and the slope of the straight line of the concentrated deformation segment are selected in the range of the length of the rule circle segment. 、 Under the conditions, The value range is selected from both sides of its initial value.

5. The method for optimizing the roll profile of a three-roll planetary spinning tube mill according to claim 1, characterized in that: Step S5 is specifically as follows: S51. Establish a three-roll planetary rolling three-dimensional model and set simulation parameters, wherein the simulation parameters include an inclination angle, a deflection angle, a friction coefficient between the rolls and the tube, a roll speed, a thermal conductivity coefficient of heat conduction between the rolls and the tube, and a thermal conductivity coefficient of heat conduction between the tube and air; S52. Multiple experimental schemes are introduced into the three-roll planetary rolling 3D model, and the quantitative relationship between different roller profile curves and the performance parameters of the finished pipe is obtained through finite element simulation.

6. The method for optimizing the roll profile of a three-roll planetary spinning tube mill according to claim 1, characterized in that: Step S6 is specifically as follows: S61. Arrange multiple groups of experimental schemes and finite element simulation output corresponding to each group of experimental schemes, and perform normalization processing on them; S62: Divide the normalized data into training set, validation set and test set; S63: Inputting the training set data into the prediction model based on the neural network for training to obtain an initial prediction model; S64: Inputting the validation set data into the initial prediction model for testing, and if the test criteria are met, obtaining a trained prediction model; Otherwise, return to step S63 and continue training until the test criteria are met; S65: Input the test set data into the trained prediction model, and compare its finite element simulation results with the results predicted by the trained prediction model. If the predicted results fit the finite element simulation results well, the test ends and a neural network-based prediction model for finished pipe performance parameters is obtained; otherwise, steps S63-S65 are executed again.

7. The method for optimizing the roll profile of a three-roll planetary spinning tube mill according to claim 1, characterized in that: In step S4, uniform experimental design and orthogonal experimental methods are used to design multiple experimental schemes; In addition, the neural network in step S6 is a BP neural network based on the Levenberg-Marquardt optimization algorithm, the hidden layer is a Sigmoid activation function, and the output layer is a two-layer feedforward network with linear output neurons. The neural network threshold and weight matrix of the Purelin activation function are selected as the neural network matrix file and stored in the database.

8. The method for optimizing the roll profile of a three-roll planetary spinning tube mill according to claim 1, characterized in that: In step S7, the target parameters of each section of the roll profile curve are iteratively optimized by a genetic algorithm. Step S7 is specifically as follows: S71. Determine optimization variables, value ranges of optimization variables, and optimization targets, wherein the optimization variables are target parameters of each section of the roll profile curve, and the optimization targets are performance parameters of the finished pipe; S72. Set the population size to N P , and initialize to generate the initial population; S73, encoding each individual in the population and calculating the fitness value of each individual, wherein each individual includes all optimization variables, and the fitness value of each individual is predicted by the neural network model; S74. After performing selection, crossover, and mutation operations on individuals in the population, a new generation of population is generated; S75. Determine whether the termination condition is met. If so, terminate the evolution and output the best individual obtained as the optimal solution; if not, return to step S73 to continue optimization.

9. The method for optimizing the roll profile of a three-roll planetary spinning tube mill according to claim 8, characterized in that: The optimization variable value range in step S7 is obtained by the designer based on the quantitative relationship between different roller profiles and the performance parameters of the finished pipe obtained in step S5, and by adjusting the value range of the target parameters of each section of the roller profile determined in step S3. The termination condition is that the iterative search tends to be stable and the optimal solution is found.

10. The method for optimizing the roll profile of a three-roll planetary spinning tube mill according to claim 1, wherein: A database is also designed to store the initial design data, standard data, finite element simulation design and result data of the three-roll planetary spinning process, the design and result data of the prediction model established based on the neural network, and the result data iterated by the optimization algorithm. The standard data includes national standards for products of various specifications.

Images