A method and device for optimizing parameters of stress-strain calculation formula for hot-press shear test

Abstract

The present invention belongs to the technical field related to plastic forming, and discloses a method and device for optimizing the parameters of a stress-strain calculation formula for a hot press shear test. The method and device comprise the following steps: (1) inputting the load-displacement curve data corresponding to the hot press shear test under different deformation conditions into a Mises equivalent equation to convert the data into an equivalent stress-strain curve; (2) correcting the temperature rise caused by deformation heat during the hot press shear test; (3) inputting a neural network model into a hot press shear deformation finite element model, and then calculating the Mises parameter corresponding to the minimum mean square error between the load-displacement curve simulated by the hot press shear deformation finite element model and the load-displacement curve obtained by the test, i.e., the optimal parameter; the input of the neural network model is strain, strain rate, and correction temperature; and the output is stress. The present invention improves accuracy and has strong applicability.

Description

Technical Field

[0001] The present invention belongs to the technical field related to plastic forming, and more specifically, relates to a method and equipment for optimizing parameters of a stress-strain calculation formula for a hot pressing shear test. Background Art

[0002] The stress-strain curve of metal materials under high temperature conditions is an important way to reflect the thermal deformation behavior of metals. Physical simulation tests are often used to obtain stress-strain curves of metals under different deformation conditions because they have deformation states similar to actual forming processes, in order to guide the design of forming processes. Currently, the most commonly used physical simulation tests include uniaxial tensile tests, uniaxial compression tests (hereinafter referred to as compression tests), plane strain compression tests, etc. Among them, hot compression tests are widely used to study the thermal deformation behavior of metals because of their characteristics of large deformation and uniform deformation, which can obtain the stress-strain relationship of metals within a large strain range.

[0003] However, the metal plastic forming process often involves a combination of compression and shearing, which makes it impossible for conventional hot compression tests to accurately obtain stress-strain curves that reflect actual working conditions. In addition, shear deformation of metal materials dominates in forming processes such as stamping, bending, shear spinning, and numerical simulation based on the stress-strain curves obtained from conventional hot compression tests will inevitably produce large errors, which is not conducive to the formulation of production processes. In order to accurately describe the constitutive relationship of metal materials under shear deformation behavior, domestic and foreign scholars have designed a new hot pressing and shearing specimen whose deformation mode is closer to the real physical process.

[0004] The hot pressing and shearing specimen is made by opening two oblique grooves on the conventional hot compression specimen, so that the deformation area is concentrated inside the oblique grooves, forming a unique deformation mode with extremely small deformation at both ends and shear deformation as the main part in the middle section. But at the same time, how to convert the load and displacement curve of the hot pressing and shearing specimen becomes a difficult problem. The stress-strain conversion formula of the conventional hot compression specimen is only applicable to the "drum-shaped" deformation mode, and it is impossible to calculate the stress-strain relationship of the hot pressing and shearing specimen. The existing stress and strain calculation formulas for hot pressing and shearing specimens are mainly determined based on empirical values ​​and have poor applicability to different metal materials. For example, the document "Research on the Composite Thermal Deformation Behavior of High-Strength Aluminum Alloy Compression and Shearing" discloses a stress and strain calculation formula for hot pressing and shearing tests applicable to 7075 aluminum alloy, and the results are directly obtained by substituting the first-order equivalent stress and strain formula. When facing different metal materials, this calculation method is difficult to quickly calculate the empirical parameters, lacks a reasonable calculation process, and requires a lot of time for trial and error. Summary of the Invention

[0005] In response to the above defects or improvement needs of the prior art, the present invention provides a method and equipment for optimizing the parameters of the stress-strain calculation formula of the hot pressing shear test, which aims to solve the problems of inaccuracy and poor applicability of the existing hot pressing shear stress-strain calculation formula.

[0006] To achieve the above object, according to one aspect of the present invention, a method for optimizing the parameters of a stress-strain calculation formula for a hot compression shear test is provided, the method comprising the following steps:

[0007] (1) The load-displacement curve data corresponding to the hot compression shear test under different deformation conditions are input into the Mises equivalent equation to convert it into the equivalent stress-strain curve;

[0008] (2) Correct the temperature rise caused by deformation heat during the hot compression shear test;

[0009] (3) The neural network model is input into the hot pressing and shearing deformation finite element model, and then the Mises parameter corresponding to the minimum mean square error between the load-displacement curve simulated by the hot pressing and shearing deformation finite element model and the load-displacement curve obtained by the test is calculated, that is, the optimal parameter; the input of the neural network model is strain, strain rate, and corrected temperature; the output is stress.

[0010] Furthermore, the load-displacement curve data corresponding to the hot compression shear test under different deformation conditions are collected and preprocessed, and then the preprocessed load-displacement curve data are input into the Mises equivalent equation; the preprocessing includes removing the preload part before deformation and the part after fracture.

[0011] Furthermore, the Mises equivalent equation is:

[0012]

[0013]

[0014] Where P is the axial loading pressure, d is the vertical displacement of deformation, and d y is the yield displacement, and k1~k5 are the Mises coefficients related to the size and material of the compression-shear specimen.

[0015] Furthermore, the temperature rise correction formula used is:

[0016]

[0017] Among them, the adiabatic correction coefficient is usually related to the strain rate, and its specific expression is as follows:

[0018]

[0019] Where ρ is the material density, C Pis the specific heat capacity of the material, is the strain rate of the compression shear specimen; ε is the strain of the compression shear specimen.

[0020] Furthermore, the strain rates are 0.01 and 0.1s -1 When the strain rate is 1, 10s -1 When , the adiabatic correction factor is taken as 0.95.

[0021] Furthermore, the activation function of the neural network model is a natural exponential function.

[0022] Furthermore, the subroutine VUHARD in ABAQUS software was used for secondary development to embed the neural network model into the hot pressing and shear deformation finite element model.

[0023] Furthermore, the Mises parameter corresponding to the minimum mean square error between the load-displacement curve obtained by simulating the hot pressing shear deformation finite element model and the load-displacement curve obtained by the test is obtained through cyclic calculation and genetic algorithm.

[0024] The present invention also provides a system for optimizing the parameters of a hot pressing shear stress and strain calculation formula. The system includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it executes the above-mentioned method for optimizing the parameters of the hot pressing shear stress and strain calculation formula.

[0025] The present invention also provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the above-mentioned method for optimizing the parameters of the hot-pressing shear stress-strain calculation formula.

[0026] In general, compared with the prior art, the above technical solutions conceived by the present invention have the following beneficial effects:

[0027] 1. The present invention combines secondary development with a neural network model, introduces the load-displacement curve of the compression-shear specimen into the Mises equivalent equation, compares the mean square error between the simulated results and the experimental load-displacement curve, and minimizes the mean square error between the simulated and experimental values ​​through cyclic iteration, thereby obtaining the Mises parameters under optimal conditions.

[0028] 2. The stress-strain conversion formula for the compression-shear specimen mentioned in the present invention includes adiabatic temperature rise correction in the process of solving parameters. The traditional stress-strain conversion formula for compression-shear tests does not take this into account. In fact, the deformation area of ​​the compression-shear specimen is small and the strain rate is large. A large amount of deformation heat is generated during deformation. The temperature in the deformation area does not conform to the isothermal thermal compression law. Therefore, the temperature needs to be corrected to improve accuracy.

[0029] 3. The present invention uses finite element software to carry out secondary development of the neural network model, and then calculates and optimizes the Mises parameters, and performs multiple rounds of calculations until the solved Mises equivalent equation meets the requirements. Compared with the traditional one-round solution method, it is closer to reality, the solved results are more reasonable, and the cost of manual calculation is greatly reduced.

[0030] 4. The present invention performs an adiabatic temperature rise correction. Considering that temperature is a nonlinear data set, the general constitutive model requires constant temperature and constant strain rate, which is not applicable to the situation described in the present invention. Therefore, a neural network model is adopted; the neural network model provides a method for learning nonlinear data, and a good constitutive model can be trained by adjusting parameters.

[0031] 5. For some complex alloy materials, conventional compression shear specimen stress-strain conversion methods cannot solve the problem. The present invention uses a neural network model to continuously approximate the calculated value to the experimental result, thereby obtaining the compression shear specimen stress-strain conversion formula for this material. This method only requires the load-displacement curve of the isothermal hot compression of the compression shear specimen, and is widely applicable to a variety of different materials with strong applicability.

[0032] 6. In the present invention, a genetic algorithm is used to quickly search for optimization in the process of solving the neural network model. The constitutive model based on the current Mises equivalent equation is brought into the finite element model to obtain the mean square error of the calculated and experimental values ​​as the optimization target, and the fitness function is determined. Through operations such as selection, crossover, and mutation, the optimal network structure and hyperparameters can be automatically searched to improve the generalization ability and performance of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a flow chart of a method for optimizing parameters of a stress-strain calculation formula for a hot-press shear test provided by the present invention;

[0034] Figure 2 (a), (b), and (c) are schematic diagrams of compression-shear specimens at three different angles. DETAILED DESCRIPTION

[0035] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0036] See also Figure 1 and Figure 2 The present invention provides a method for optimizing the parameters of the hot pressing shear stress and strain calculation formula, which uses secondary development to embed a neural network model into a finite element model, and then uses cyclic calculation and genetic algorithm to obtain the optimal parameters of the stress and strain calculation formula, thereby obtaining a reasonable conversion formula. It is applicable to different materials and can reasonably and efficiently realize the solution of the parameters of the hot pressing shear test stress and strain calculation formula.

[0037] The optimization method mainly comprises the following steps:

[0038] S1, collect the load-displacement curve data corresponding to the hot compression shear test under different deformation conditions and perform preprocessing.

[0039] The design of hot pressing shear specimen is to open two grooves with 45° angle in the middle of the conventional cylindrical specimen, such as Figure 2 The preprocessing includes removing the preloaded part before deformation and the part after fracture; and smoothing the curve to remove the error caused by the experimental device.

[0040] In one embodiment, the parameters of the hot compression test are determined based on the metal type of the shear specimen, generally including strain rate, temperature, holding time, and cooling method. The specimen is fixed between two indenters using the mechanical axis of a thermal simulation tester. A thermocouple is connected to measure real-time temperature changes in the deformation area. Pre-tightening is performed to prevent poor contact. The specimen is then heated to the desired temperature at a set heating rate, held for a period of time, and then compression begins. After compression is complete, the machine automatically collects the specimen's load-displacement curve, removes the pre-tightening and fracture stages, and smoothes the curve using data processing software to obtain the initial load-displacement curve.

[0041] S2, inputting the obtained load-displacement curve data into the Mises equivalent equation to convert it into an equivalent stress-strain curve.

[0042] The Mises equivalent equation is:

[0043]

[0044]

[0045] Where P is the axial loading pressure, d is the vertical displacement of deformation, and d y is the yield displacement, while k1–k5 are Mises coefficients related to the compression-shear specimen size and material. This conversion yields an equivalent stress-equivalent strain curve. The Mises coefficients k1–k5 in the equation are pre-determined to random values ​​within an appropriate range. Specifically, a set of initial values ​​for the Mises coefficients k1–k5 is selected based on the literature.

[0046] In one embodiment, the Mises equivalent equation is used to convert the load-displacement curve into a stress-strain curve. The initial conversion of the equivalent stress-equivalent strain curve is not very accurate. This constitutive relationship must be incorporated into a finite element model for calculation. The calculated results from the finite element model are compared with the actual load-displacement curve to assess the validity of the Mises equivalent equation.

[0047] S3, correct the temperature rise caused by deformation heat during the hot press shear test. The temperature rise correction formula used is:

[0048]

[0049] Among them, the adiabatic correction coefficient is usually related to the strain rate, and its specific expression is as follows:

[0050]

[0051] Where ρ is the material density, C P is the specific heat capacity of the material, is the strain rate of the compression shear specimen; ε is the strain of the compression shear specimen. The adiabatic correction factor is determined by the strain rate parameter of the hot compression test.

[0052] The temperature curve obtained from the isothermal hot compression test cannot collect the temperature rise value caused by the local large deformation of the compression-shear specimen. Therefore, the theoretical formula is used to perform adiabatic temperature rise correction on the experimental temperature. The strain rates are 0.01 and 0.1s -1 When the strain rate is 1, 10s -1 When , the adiabatic correction factor is taken as 0.95.

[0053] S4, using secondary development to input the neural network model into the hot pressing and shearing deformation finite element model, and then through cyclic calculation and genetic algorithm to obtain the Mises parameter corresponding to the minimum mean square error between the load-displacement curve simulated by the hot pressing and shearing deformation finite element model and the load-displacement curve obtained by the test, that is, the optimal parameter; the input of the neural network model is strain, strain rate, and corrected temperature; the output is stress.

[0054] The activation function of the neural network model is a natural exponential function; the number of layers and neurons of the neural network model are the variables to be explored. Several different numbers of layers and neurons are set to explore the rules. The training results are adjusted by repeatedly modifying the weights and biases of each layer, and the group with the smallest calculation error value of the final result is selected as the final parameter.

[0055] Secondary development was performed using the VUHARD subroutine in ABAQUS software, embedding the neural network model into the hot-pressing and shearing deformation finite element model. The hot-pressing and shearing finite element model was established using ABAQUS, with the compression and shearing specimen set as a deformable body and the two indenters as rigid bodies. Parameters such as load and boundary conditions were set, and calculations began after meshing. The error minimization process included setting a threshold. If the mean square error between the current calculation result and the experimental value was higher than the threshold, the Mises parameter was modified, and steps S2 to S4 were repeated until the mean square error was lower than the selected threshold. A genetic algorithm was used for rapid optimization to automatically search for the optimal neural network structure and parameters.

[0056] In one embodiment, a BP neural network model is used to construct the constitutive relationship of the above-mentioned modified equivalent stress-strain curves under different process parameter conditions, with strain, modified temperature, and pre-twinned crystal volume fraction as the input layer and stress as the output layer; the number of hidden layers and the number of neurons are also calculated as variables, and the value with the smaller mean absolute error (MAE) is selected as the neural network parameter.

[0057] Through secondary development, the neural network model was embedded in the hot-pressing and shearing finite element model, and the VUHARD subroutine was used to convert the neural network code into a Fortran format file in ABAQUS, which is the polynomial of the weight threshold combination after the final optimization of the neural network. The hot-pressing and shearing finite element model of the compression and shearing specimen was run for simulation calculation, and the load and displacement curves in the simulation results were extracted. The optimal Mises parameters were obtained by minimizing the mean square error between the load and displacement curves of the simulation results and the load and displacement curves obtained by converting the Mises equivalent equation.

[0058] The present invention is further described in detail below with reference to specific embodiments.

[0059] Example 1

[0060] Taking the AZ31 magnesium alloy compression shear specimen as an example, the stress-strain conversion formula of the compression shear specimen is calculated by embedding the neural network model into the hot compression shear finite element model. The specific steps include the following:

[0061] Step (1): The dimensions of the AZ31 magnesium alloy compression shear specimen are as follows: Figure 2As shown in Figure 1, the isothermal hot compression test parameters are set as follows: temperature 100°C to 400°C, strain rate 0.01°C to 10°C, compression amount 25%, heating rate 10°C / s, heating to the deformation temperature and holding for 3 minutes, then starting compression.

[0062] The load and displacement curves after compression were collected, and the origin software was used to delete the preload part and the part where the sample was compressed, delete the duplicate points, and perform smoothing.

[0063] Step (2): Convert the load and displacement curve of the compression-shear specimen obtained in the previous step into a stress-strain curve, and substitute it into the Mises equivalent equation to obtain a preliminary conversion result. The Mises parameters are taken according to the literature as follows: k1 = 4.12, k2 = -5.43, k3 = 2.17, k4 = 0.91, k5 = 0.21.

[0064] Step (3): Use the temperature rise correction formula The local temperature rise in the deformation area of ​​the compression shear specimen during the isothermal hot compression test was corrected. The density of AZ31 magnesium alloy is ρ = 1.78 g / cm 3 , specific heat capacity c p =1.02J / (g·℃), stress σ and strain ε take the stress and strain values ​​of the previous conversion results; when the strain rate is 0.01 and 0.1s -1 When the adiabatic correction factor When the strain rate is 1, 10s -1 When , the adiabatic correction coefficient δ = 0.95.

[0065] Step (4): In the BP neural network model, the input layer is strain, corrected temperature and pre-twin crystal volume fraction, the output layer is stress, the pre-twin crystal volume fraction is 2% to 10%, and the activation function is the natural exponential function e x ; Calculate the calculation results of the BP neural network model with different numbers of hidden layers and neurons, and select the value with the smaller mean absolute error (MAE) as the neural network parameter.

[0066] Step (5): The neural network code is converted into a Fortran format file through the ABAQUS subroutine VUHARD to obtain the final optimized weight threshold combination polynomial; the finite element simulation calculation of the compression shear specimen is run, the load and displacement curves in the simulation results are extracted, and the average absolute relative error (AARE) between the load and displacement curves in the finite element simulation results and the experimental values ​​is calculated: If the AARE value is less than 10%, the current Mises parameter is the optimal parameter;

[0067] Otherwise, extract the equivalent stress of the neutral surface and equivalent strain Repeat steps (2) to (5) until the AARE value is less than 10%.

[0068] The present invention also provides a system for optimizing the parameters of a hot pressing shear stress and strain calculation formula. The system includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it executes the above-mentioned method for optimizing the parameters of the hot pressing shear stress and strain calculation formula.

[0069] The present invention also provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the above-mentioned method for optimizing the parameters of the hot-pressing shear stress-strain calculation formula.

[0070] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements 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 parameters of a stress-strain calculation formula for a hot pressing shear test, characterized in that: The method comprises the following steps: (1) The load-displacement curve data corresponding to the hot compression shear test under different deformation conditions are input into the Mises equivalent equation to convert it into the equivalent stress-strain curve; (2) Correct the temperature rise caused by deformation heat during the hot compression shear test; (3) The neural network model is input into the hot pressing and shearing deformation finite element model, and then the Mises parameter corresponding to the minimum mean square error between the load-displacement curve simulated by the hot pressing and shearing deformation finite element model and the load-displacement curve obtained by the test is calculated, that is, the optimal parameter; the input of the neural network model is strain, strain rate, and corrected temperature; the output is stress.

2. The method for optimizing the parameters of the stress-strain calculation formula for the hot compression shear test according to claim 1, characterized in that: Collect the load-displacement curve data corresponding to the hot compression shear test under different deformation conditions, perform preprocessing, and then input the preprocessed load-displacement curve data into the Mises equivalent equation; Pretreatment includes removing the preloaded part before deformation and the part after fracture.

3. The method for optimizing parameters of the stress-strain calculation formula for hot pressing and shearing tests according to claim 1, wherein: The Mises equivalent equation is: Where P is the axial loading pressure, d is the vertical displacement of deformation, and d y is the yield displacement, and k1~k5 are the Mises coefficients related to the size and material of the compression-shear specimen.

4. The method for optimizing the parameters of the stress-strain calculation formula for the hot compression shear test according to claim 3, wherein: The temperature rise correction formula used is: Among them, the adiabatic correction coefficient is usually related to the strain rate, and its specific expression is: Where ρ is the material density, C P is the specific heat capacity of the material, is the strain rate of the compression shear specimen; ε is the strain of the compression shear specimen.

5. The method for optimizing the parameters of the stress-strain calculation formula for the hot-press shear test according to claim 4, characterized in that: The strain rate is 0.01s and 0.1s -1 When the strain rate is 1, 10s -1 When , the adiabatic correction factor is taken as 0.

95.

6. The method for optimizing parameters of the stress-strain calculation formula for hot-press shear testing according to claim 1, wherein: The activation function of the neural network model is a natural exponential function.

7. The method for optimizing parameters of the stress-strain calculation formula for hot-press shear testing according to claim 1, wherein: Secondary development was performed using the subroutine VUHARD in ABAQUS software, and the neural network model was embedded into the hot pressing and shear deformation finite element model.

8. The method for optimizing parameters of a stress-strain calculation formula for a hot compression shear test according to any one of claims 1 to 7, wherein: The Mises parameter corresponding to the minimum mean square error between the load-displacement curve obtained by simulating the hot pressing shear deformation finite element model and the load-displacement curve obtained by the test is obtained through cyclic calculation and genetic algorithm.

9. A system for optimizing parameters of a hot pressing shear stress and strain calculation formula, characterized by: The system includes a memory and a processor, the memory stores a computer program, and the processor executes the method for optimizing the parameters of the hot pressing shear stress and strain calculation formula according to any one of claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by the processor, the machine-executable instructions prompt the processor to implement the method for optimizing the parameters of the hot pressing shear stress and strain calculation formula according to any one of claims 1 to 8.

Images