Method and system for regulating residual stress through cooperation of prestress and thermal vibration aging

Abstract

The invention relates to the field of part machining, and discloses a method and system for cooperatively regulating residual stress through prestress and thermal vibration aging, and the method comprises the following steps: preparing a part, and obtaining the Poisson's ratio and the elastic modulus of the part; measuring the initial residual stress of the part; the initial residual stress of the part is compiled into an input data set, the part is placed on a fixed heating furnace table through a clamp, a prestress field is formed through stretching of the clamp, and thermal aging is conducted on the part; thermal property factors of the material are mainly considered in the thermal aging process; meanwhile, the vibration exciter arranged on the workbench starts to work, vibration aging is conducted on the part, by combining the prestress technology and the thermal vibration aging technology, integration and uniformization of residual stress regulation and control of the part are achieved, the machining quality and machining precision of the part are improved, relevant parameters of the part which is not machined can be predicted, and the machining precision of the part is improved. The problem that the residual stress of the high-strength composite material is difficult to regulate and control is solved, and the method has high economic value.

Description

Technical Field

[0001] The present invention relates to the field of parts processing, and in particular to a method and system for cooperatively regulating residual stress by prestressing and thermal vibration aging. Background Art

[0002] At present, alloy parts have large residual stress due to deformation during forging and heat treatment, and uneven cooling. They are prone to deformation during final processing, resulting in scrapped forgings. Through a large number of studies, it was found that the main factor causing this processing deformation is the initial residual stress inside the blank. Solving the processing deformation problem becomes regulating the initial residual stress. The manufacturing accuracy of high-performance alloy parts has become a major problem that needs to be solved in my country's aviation strategic projects;

[0003] Commonly used residual stress control methods include thermal aging and vibration aging. Both methods are relatively mature. Thermal aging will coarsen the grains of the ring and reduce the strength, while vibration aging can reduce the residual stress on the end face of the part. However, both methods have their own defects, and the range of action on large parts is small and the accuracy is not high. It is difficult to control the residual stress of high-strength composite materials and the accuracy is low. Prestressing is a commonly used technology in engineering. By applying prestress, the residual stress generated during the manufacturing process of the component can be partially offset, thereby improving the overall performance of the component. The coordinated control of residual stress by prestressing and thermal vibration aging requires a new type of intelligent processing method because of the large number of controllable parameters and the difficulty in ensuring the control accuracy. Therefore, in order to solve the above problems, a method and system for coordinated control of residual stress by prestressing and thermal vibration aging is very important. Summary of the invention

[0004] 1. Technical issues to be solved

[0005] In view of the deficiencies in the prior art, the present invention provides a method and system for cooperatively regulating residual stress by prestressing and thermal vibration aging, which solves the problems existing in the above-mentioned background technology.

[0006] (II) Technical solution

[0007] To achieve the above object, the present invention provides the following technical solution: a method for synergistically regulating residual stress by prestressing and thermal vibration aging, comprising the following steps:

[0008] S1. Prepare parts, obtain Poisson's ratio and elastic modulus of the parts; measure the initial residual stress of the parts; compile the initial residual stress of the parts as an input data set;

[0009] S2. Place the part on a fixed heating furnace table through a fixture, form a prestress field through the stretching of the fixture, and perform thermal aging on the part; the thermal aging process mainly considers the thermal performance factors of the material; the factors include the heating rate of the part, the insulation temperature, and the cooling rate; at the same time, start the exciter arranged on the workbench to perform vibration aging on the part; the vibration aging process mainly considers the excitation frequency, excitation time, and excitation amplitude;

[0010] S3. Record the residual stress value, prestress value, heating rate, holding temperature, cooling rate, excitation frequency, excitation time and excitation amplitude of the parts after aging;

[0011] S4, compile the above data into an output data set; merge the input and output data sets into a data set, and divide the test set into a data set according to a 4:1 data ratio, and perform an RBF neural network training model on the data set;

[0012] S5. Based on machine learning technology, the RBF neural network model is optimized; mainly including: using the sparrow search algorithm (SSA) to optimize the weight, center value and width value of the RBF neural network algorithm; then assigning the global optimal parameters to the RBF neural network to improve the prediction performance of the RBF neural network;

[0013] S6. According to the prediction model, the production parameters of the parts are optimized through the improved slime mold optimization algorithm.

[0014] Furthermore, in the process of continuous application of the prestress field in S2, thermal vibration aging is performed to regulate the residual stress, which is different from the conventional residual stress regulation method.

[0015] Furthermore, in order to improve the stability and prediction performance of the network model, the S4 performs missing value processing, outlier processing and data value standardization on the data set.

[0016] Furthermore, in S5, the weight, center value and width value of the RBF neural network algorithm are adjusted by a sparrow search algorithm (SSA); and the global optimal parameters are assigned to the RBF neural network to improve the prediction performance of the RBF neural network. The weight, center value and width value of the basis function of the RBF (radial basis function) neural network directly affect the prediction performance of the neural network. The weight and basis function width of the traditional RBF neural network are randomly generated, which can easily lead to various errors in BP neural network training and make it difficult to obtain a global optimal solution. The sparrow search algorithm has the advantages of strong global search capability, simple and fast parameter setting, excellent convergence performance, etc., and can improve the fitting degree between the predicted value and the actual value of the RBF neural network.

[0017] Furthermore, the steps of optimizing the RBF neural network model based on machine learning technology in S5 are: S5.1, establishing an RBF neural network, inputting the input data set and the output data set into the RBF neural network, and establishing the RBF prediction model, S5.2, initializing the weight, center value and width value of the RBF neural network, using the weight, center value and width value in the network as variables in the search space, and setting the upper and lower bounds of the search space, S5.3, initializing the sparrow population, and dividing it into two categories of discoverers and joiners according to the fitness values ​​of the individuals, S5.4, calculating the fitness value of the population, updating the optimal position of the discoverer, S5.5, updating the fitness value of the population, updating the positions of the discoverer and joiner, and recording the global optimal solution, S5.6, judging whether the stopping condition is met, if the stopping condition is not met, repeating the above steps; otherwise, returning the global optimal solution as the optimal weight and basis function width value of the RBF neural network.

[0018] Furthermore, in S6, the production parameters of the output data are optimized by an improved slime mold algorithm (SMA). SMA has the advantages of fast convergence speed, strong global search capability, and simple and easy algorithm implementation, but it also has the defects of low initial population quality and slow convergence speed. The refractive reverse learning algorithm can find the reverse solution according to the current solution. The probability that the reverse solution is close to the optimal solution is higher than that of the current solution, thereby improving the population quality, thereby improving the quality of the algorithm and the optimization results.

[0019] Furthermore, the steps of optimizing the production parameters by refractive reverse learning optimization SMA algorithm in S6 are: S6.1, initializing the population through the reverse learning strategy, calculating the fitness value, S6.2, updating the slime mold weight parameters, S6.3, judging the position of the offspring, updating the position of the offspring, S6.4, calculating the fitness value, updating the global optimal solution, S6.5, judging whether the end condition is met, if not, repeating the above steps; in step S6.3, r is a random number [0,1], z is the proportion of randomly distributed slime mold individuals in the total, p = tanh|S(i)-DF|, S(i) is the fitness, and DF represents the best fitness in all iterations.

[0020] According to another aspect of the present invention, there is provided a thermal vibration workbench system, comprising a power supply, a workpiece, a workbench, a heating system, a vibration system, and a clamping system, wherein the heating system comprises a heating couple, a heating furnace, and a temperature controller, the vibration system comprises an exciter, and a frequency controller, and the clamping device comprises a clamp, a guide rail, and a prestress control device longitudinally arranged on the workbench, and the prestress control device applies a prestress field to the workpiece.

[0021] (III) Beneficial effects

[0022] The present invention provides a method and system for coordinating residual stress by prestressing and thermal vibration aging, which has the following beneficial effects:

[0023] By combining prestressing technology with thermal vibration aging technology, the overall and uniform regulation of residual stress of parts is achieved, the processing quality and accuracy of parts are improved, and the relevant parameters of parts that have not yet been processed can be predicted, solving the problem of the difficulty of regulating residual stress of high-strength composite materials in existing technologies, which has great economic value.

[0024] The workpiece is heated to a certain temperature in a heating furnace, then kept warm for a period of time, and finally cooled in a controlled manner. Its basic principle is that the yield strength of the material will decrease with the increase of temperature. When the temperature rises and the residual stress inside the workpiece exceeds the yield limit of the material, local plastic deformation occurs inside the workpiece, the residual strain is released, and the residual stress will be reduced. At the same time, a vibration device is used to apply cyclic load vibration to the workpiece at a specific frequency, and the vibration is continued for a period of time to apply additional alternating stress or deformation to the workpiece. When the additional alternating stress is superimposed on the residual stress, the energy is absorbed through the friction inside the material, and when it reaches or exceeds a certain threshold of the material, the workpiece undergoes microscopic or macroscopic viscoelastic-plastic mechanical changes, thereby reducing and homogenizing the residual stress inside the workpiece and stabilizing its dimensional accuracy;

[0025] The weight, center value and width value of the RBF neural network algorithm are adjusted by the sparrow search algorithm (SSA); then the global optimal parameters are assigned to the RBF neural network to improve the prediction performance of the RBF neural network. The weight, center value and width value of the basis function of the RBF (radial basis function) neural network directly affect the prediction performance of the neural network. The weight and basis function width of the traditional RBF neural network are randomly generated, which can easily lead to various errors in the BP neural network training and make it difficult to obtain the global optimal solution. The sparrow search algorithm has the advantages of strong global search ability, simple and fast parameter setting, excellent convergence performance, etc., which can improve the fitting degree between the predicted value and the actual value of the RBF neural network.

[0026] The improved Slime Mould Algorithm (SMA) is used to optimize the production parameters of the output data. SMA has the advantages of fast convergence speed, strong global search capability, and simple and easy algorithm implementation. However, it also has the defects of low initial population quality and slow convergence speed. The refracted reverse learning algorithm can find the reverse solution based on the current solution. The probability that the reverse solution is close to the optimal solution is higher than that of the current solution, thereby improving the population quality, thereby improving the quality of the algorithm and the optimization results. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1It is a method and system operation flow chart of coordinating residual stress control by prestressing and thermal vibration aging;

[0028] Figure 2 It is a method for coordinating residual stress by prestressing and thermal vibration aging, and a flowchart of an optimization algorithm for the RBF neural network model based on machine learning technology in the system;

[0029] Figure 3 A method for coordinating residual stress by prestressing and thermal vibration aging and a flow chart of the improved slime mold algorithm in the system;

[0030] Figure 4 The present invention is a method for cooperatively regulating residual stress by prestressing and thermal vibration aging and a structural schematic diagram of the thermal vibration workbench system in the system.

[0031] In the figure: 1. Heating couple; 2. Heating furnace; 3. Temperature controller; 4. Power supply; 5. Prestress control device; 6. Workbench; 7. Guide rail; 8. Fixture; 9. Workpiece; 10. Vibration controller; 11. Vibrator. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] The present invention provides a technical solution: a method for cooperatively regulating residual stress by prestressing and thermal vibration aging, comprising the following steps:

[0034] S1. Prepare parts, obtain Poisson's ratio and elastic modulus of the parts; measure the initial residual stress of the parts; and compile the initial residual stress of the parts as an input data set.

[0035] S2. Place the parts on a fixed heating furnace table through a fixture, form a prestressed field through the stretching of the fixture, and perform thermal aging on the parts; the thermal aging process mainly considers the thermal performance factors of the material; the factors include the heating rate of the parts, the insulation temperature, and the cooling rate; at the same time, the exciter arranged on the workbench starts working to perform vibration aging on the parts; the vibration aging process mainly considers the excitation frequency, excitation time, and excitation amplitude. During the continuous application of the prestressed field, thermal vibration aging is performed to regulate the residual stress, which is different from the conventional residual stress regulation method.

[0036] S3. Record the residual stress value, prestress value, heating rate, holding temperature, cooling rate, excitation frequency, excitation time and excitation amplitude of the parts after aging.

[0037] S4. Compile the above data into an output data set; merge the input and output data sets into a data set, and divide the test set into a data set according to a data ratio of 4:1. Perform RBF neural network training model on the data set. In order to improve the stability and prediction performance of the network model, the data set is processed for missing values, outliers and data value standardization.

[0038] S5. Based on machine learning technology, the RBF neural network model is optimized; mainly including: using the sparrow search algorithm (SSA) to optimize the weight, center value and width value of the RBF neural network algorithm; then assigning the global optimal parameters to the RBF neural network to improve the prediction performance of the RBF neural network, using the sparrow search algorithm (SSA) to optimize the weight, center value and width value of the RBF neural network algorithm; then assigning the global optimal parameters to the RBF neural network to improve the prediction performance of the RBF neural network. The weights, center value and width value of the basis function of the RBF (radial basis function) neural network directly affect the prediction performance of the neural network. The weights and basis function width of the traditional RBF neural network are randomly generated, which can easily lead to various errors in BP neural network training and make it difficult to obtain the global optimal solution. The sparrow search algorithm has the advantages of strong global search ability, simple and fast parameter setting, and excellent convergence performance. It can improve the fitting degree between the predicted value and the actual value of the RBF neural network. Based on machine learning technology, the steps for optimizing the RBF neural network model are as follows: S5.1. Establish an RBF neural network, input the input data set and the output data set into the RBF neural network, and establish the RBF prediction model , S5.2, initialize the weights, center values ​​and width values ​​of the RBF neural network, use the weights, center values ​​and width values ​​in the network as variables in the search space, and set the upper and lower bounds of the search space, S5.3, initialize the sparrow population, and divide it into two categories: discoverers and joiners according to the fitness values ​​of the individuals, S5.4, calculate the fitness value of the population, update the optimal position of the discoverer, S5.5, update the fitness value of the population, update the positions of the discoverer and joiner, and record the global optimal solution, S5.6, determine whether the stopping condition is met, if the stopping condition is not met, repeat the above steps; otherwise, return the global optimal solution as the optimal weight and basis function width value of the RBF neural network.

[0039] S6. According to the prediction model, the production parameters of the parts are optimized by the improved slime mold optimization algorithm. The SMA algorithm is used to optimize the production parameters of the output data. The SMA has the advantages of fast convergence speed, strong global search ability, and simple and easy algorithm implementation. However, it also has the defects of low initial population quality and slow convergence speed. The refraction reverse learning algorithm can find the reverse solution according to the current solution. The probability that the reverse solution is close to the optimal solution is higher than that of the current solution, thereby improving the population quality, thereby improving the quality of the algorithm and the optimization results. The steps of optimizing the production parameters by refraction reverse learning SMA algorithm are as follows: S6.1, initialize the population through the reverse learning strategy, calculate the fitness value, S6.2, update the slime mold weight parameters, S6.3, determine the position of the daughter, update the position of the daughter, S6.4, calculate the fitness value, update the global optimal solution, S6.5, determine whether the end condition is met, if not, repeat the above steps; in step S6.3, r is a random number [0,1], z is the proportion of randomly distributed slime mold individuals in the total, p = tanh|S(i)-DF|, S(i) is the fitness, and DF represents the best fitness in all iterations.

[0040] The method also sets up a thermal vibration workbench system, which includes a power supply 4, a workpiece 9, a workbench 6, a heating system, a vibration system, and a clamping system. The heating system includes a heating couple 1, a heating furnace 2, and a temperature controller 3. The vibration system includes an exciter 11 and a frequency controller 10. The clamping device includes a clamp 8, a guide rail 7, and a prestress control device 5 arranged longitudinally on the workbench 6. The prestress control device 5 applies a prestress field to the workpiece.

[0041] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0042] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for synergistically regulating residual stress by prestressing and thermal vibration aging, characterized in that: The following steps are involved: S1. Prepare parts, obtain Poisson's ratio and elastic modulus of the parts; measure the initial residual stress of the parts; compile the initial residual stress of the parts as an input data set; S2. Place the part on a fixed heating furnace table through a fixture, form a prestress field through the stretching of the fixture, and perform thermal aging on the part; the thermal aging process mainly considers the thermal performance factors of the material; the factors include the heating rate of the part, the insulation temperature, and the cooling rate; at the same time, start the exciter arranged on the workbench to perform vibration aging on the part; the vibration aging process mainly considers the excitation frequency, excitation time, and excitation amplitude; S3. Record the residual stress value, prestress value, heating rate, holding temperature, cooling rate, excitation frequency, excitation time and excitation amplitude of the parts after aging; S4, compile the above data into an output data set; merge the input and output data sets into a data set, and divide the test set into a data set according to a 4:1 data ratio, and perform an RBF neural network training model on the data set; S5. Based on machine learning technology, the RBF neural network model is optimized; mainly including: using the sparrow search algorithm (SSA) to optimize the weight, center value and width value of the RBF neural network algorithm; then assigning the global optimal parameters to the RBF neural network to improve the prediction performance of the RBF neural network; S6. According to the prediction model, the production parameters of the parts are optimized through the improved slime mold optimization algorithm.

2. The method for cooperatively regulating residual stress by prestressing and thermal vibration aging according to claim 1 is characterized in that: In the S2, during the continuous application of the prestress field, thermal vibration aging is performed to regulate the residual stress, which is different from the conventional residual stress regulation method.

3. The method for cooperatively regulating residual stress by prestressing and thermal vibration aging according to claim 1 is characterized in that: In S4, in order to improve the stability and prediction performance of the network model, missing value processing, outlier processing and data value standardization are performed on the data set.

4. The method for cooperatively regulating residual stress by prestressing and thermal vibration aging according to claim 1 is characterized in that: In the S5, the weight, center value and width value of the RBF neural network algorithm are adjusted by the sparrow search algorithm (SSA); and the global optimal parameters are assigned to the RBF neural network to improve the prediction performance of the RBF neural network. The weight, center value and width value of the basis function of the RBF (radial basis function) neural network directly affect the prediction performance of the neural network. The weight and basis function width of the traditional RBF neural network are randomly generated, which can easily lead to various errors in the BP neural network training and make it difficult to obtain the global optimal solution. The sparrow search algorithm has the advantages of strong global search capability, simple and fast parameter setting, excellent convergence performance, etc., and can improve the fitting degree between the predicted value and the actual value of the RBF neural network.

5. The method for cooperatively regulating residual stress by prestressing and thermal vibration aging according to claim 4 is characterized in that: The steps of optimizing the RBF neural network model based on machine learning technology in S5 are: S5.1, establishing an RBF neural network, inputting the input data set and the output data set into the RBF neural network, and establishing the RBF prediction model, S5.2, initializing the weight, center value and width value of the RBF neural network, using the weight, center value and width value in the network as variables in the search space, and setting the upper and lower bounds of the search space, S5.3, initializing the sparrow population, and dividing it into two categories of discoverers and joiners according to the fitness values ​​of the individuals, S5.4, calculating the fitness value of the population, and updating the optimal position of the discoverer, S5.5, updating the fitness value of the population, updating the positions of the discoverer and joiner, and recording the global optimal solution, S5.6, judging whether the stopping condition is met, if the stopping condition is not met, repeating the above steps; otherwise, returning the global optimal solution as the optimal weight and basis function width value of the RBF neural network.

6. The method for cooperatively regulating residual stress by prestressing and thermal vibration aging according to claim 1 is characterized in that: In the S6, the production parameters of the output data are optimized by using an improved slime mold algorithm (SMA). SMA has the advantages of fast convergence speed, strong global search capability, and simple and easy algorithm implementation, but it also has the defects of low initial population quality and slow convergence speed. The refracted reverse learning algorithm can find the reverse solution according to the current solution. The probability that the reverse solution is close to the optimal solution is higher than that of the current solution, thereby improving the population quality, thereby improving the quality of the algorithm and the optimization results.

7. The method for cooperatively regulating residual stress by prestressing and thermal vibration aging according to claim 6 is characterized in that: The steps of optimizing the production parameters by refractive reverse learning optimization SMA algorithm in S6 are: S6.1, initializing the population by reverse learning strategy, calculating the fitness value, S6.2, updating the slime mold weight parameters, S6.3, judging the position of the offspring, updating the position of the offspring, S6.4, calculating the fitness value, updating the global optimal solution, S6.5, judging whether the end condition is met, if not, repeating the above steps; in step S6.3, r is a random number [0,1], z is the proportion of randomly distributed slime mold individuals in the total, p = tanh|S(i)-DF|, S(i) is the fitness, and DF represents the best fitness in all iterations.

8. A thermal vibration workbench system, comprising the method for synergistically regulating residual stress by prestressing and thermal vibration aging as described in any one of claims 1 to 7, characterized in that: The invention comprises a power source (4), a workpiece (9), a workbench (6), a heating system, a vibration system, and a clamping system. The heating system comprises a heating couple (1), a heating furnace (2), and a temperature controller (3). The vibration system comprises an exciter (11) and a frequency controller (10). The clamping device comprises a clamp (8) longitudinally arranged on the workbench (6), a guide rail (7), and a prestress control device (5). The prestress control device (5) applies a prestress field to the workpiece.

Images