Intelligent hot working quality prediction method for establishing coupling tissue and deformation condition influence

Abstract

The invention discloses a hot working quality intelligent prediction method based on establishment of coupling structure and deformation condition influences, and belongs to the field of metal material plastic forming processes. According to the method, the coupling influence of the microstructure state and the deformation condition is considered, the metal material hot working quality prediction method suitable for the actual machining process is established, and the problem that the hot working quality under the constant deformation condition and the non-constant deformation condition cannot be accurately predicted through an existing universal model is solved. According to the method, the dislocation density, the grain size, the dynamic recrystallization degree, the strain rate, the deformation temperature and the deformation degree serve as input variables, the power dissipation rate and the deformation instability factor serve as output variables, a BP neural network prediction structure is established, training is conducted according to the experimental result under the constant deformation condition, and the dynamic recrystallization degree is obtained. And a hot working quality intelligent prediction model influenced by the coupling tissue and the deformation condition is obtained. The prediction of the hot working quality of the metal material under the complex deformation condition in the single-pass thermal deformation process is realized, and the prediction accords with the actual engineering processing condition.

Description

Technical Field

[0001] The present invention belongs to the field of plastic forming processes of metallic materials, and more particularly, relates to an intelligent prediction method for hot working quality that couples the effects of microstructure and deformation conditions. Background Art

[0002] Hot plastic forming is an important processing method for metallic materials. During the hot working process, the material undergoes plastic deformation at a certain strain rate in a high-temperature state. Changes in the strain rate, deformation temperature, and degree of deformation cause changes in the deformation mechanism of the material, thereby affecting the forming quality. Establishing a quantitative relationship between the deformation conditions and the forming quality during the hot deformation process is of great significance. The hot working map is one of the powerful tools that reflects the direct relationship between the deformation conditions and the forming quality. It is a superposition of the power dissipation rate map and the instability factor map. Under the dynamic material model theory, the power dissipation rate and the instability factor are functions of strain, strain rate, and deformation temperature. The existing hot working map theory obtains the direct relationship between the hot working quality and the steady-state deformation conditions. Through the hot working map, the power dissipation rate and the instability factor can be obtained when the material deforms to a certain extent at a constant deformation temperature and strain rate, so as to judge whether good forming quality can be obtained under the current deformation conditions.

[0003] During the actual processing process, the deformation conditions undergo complex changes. During the production process, the deformation conditions of the material, namely the deformation temperature and the strain rate, change continuously with time. For example: during the rolling process, the rolled piece is bitten in and withdrawn under the driving action of the roll and undergoes heat exchange with the roll and the external environment. The strain rate of the rolled piece in the rolling deformation zone continuously increases and decreases, and is accompanied by a continuous decrease in temperature. The existing hot working map theory defaults to the assumption of constant deformation conditions during the application process, resulting in inaccuracies in the actual hot working quality assessment process.

[0004] The difficulty in evaluating the hot working quality under non-constant deformation conditions is that the continuous change of the deformation conditions brings about real-time changes in the microstructure state. The hot working quality is the superposition result of the deformation conditions acting on the microstructure state. The same deformation conditions acting on different microstructure states often result in different forming qualities. Therefore, there is an urgent need to establish an intelligent prediction method for hot working quality that couples the effects of microstructure and deformation conditions to accurately predict the hot working quality of the material under the state of continuous non-constant change of the deformation conditions during the actual processing process. This is of great significance and practical value for controlling the shape and properties of metallic materials during the plastic forming process. Summary of the Invention

[0005] In view of the above defects or upgrade and optimization requirements of the prior art, the present invention provides an intelligent prediction method for hot working quality considering the influence of coupled microstructure and deformation conditions, aiming to provide an intelligent prediction method and idea for the hot working quality of metal materials to address the aforementioned problems, making the existing hot working quality evaluation theory more applicable to the actual processing process of metal materials and solving the problem that it is difficult to accurately predict the hot working quality evaluation of metal materials in the actual processing process.

[0006] The technical solution adopted by the present invention is as follows: An intelligent prediction method for hot working quality considering the influence of coupled microstructure and deformation conditions, comprising the following steps: S1: Obtain the stress-strain relationship during a single-pass hot deformation process through high-temperature physical simulation experiments and perform preprocessing; S2: Based on the dynamic material model theory, solve the power dissipation rate and deformation instability factor at different strain rates, deformation temperatures, and deformation degrees; S3: Apply a macro-micro coupled hot deformation constitutive model to solve the dislocation density, grain size, and dynamic recrystallization degree at different strain rates, deformation temperatures, and deformation degrees; S4: Establish a BP neural network structure, and use the strain rate, deformation temperature, deformation degree, and the corresponding microstructure under corresponding conditions as input variables, and the power dissipation rate and deformation instability factor under corresponding conditions as output variables to form a training set, and train the BP neural network structure to obtain an intelligent prediction model for hot working quality considering the influence of coupled microstructure and deformation conditions.

[0007] Preferably, the acquisition and preprocessing of the stress-strain relationship during a single-pass hot deformation process in S1 include the following sub-steps: S1-1: Select cylindrical compression specimens with dimensions of Φ8×12 mm or Φ10×15 mm, conduct single-pass high-temperature compression tests at different strain rates and deformation temperatures, obtain the engineering stress-strain relationship, and at least 3 groups of strain rates and 3 groups of deformation temperatures should be set in the tests; S1-2: Based on the strain conversion relationship, convert it into the true stress-strain relationship during a single pass; S1-3: Correct the obtained true stress-strain relationship for friction and adiabatic temperature rise; Preferably, the macro-micro coupled hot deformation constitutive model in S3 should be able to predict the changes in dislocation density, grain size, and dynamic recrystallization degree with deformation conditions.

[0008] Preferably, the number of neurons in the input layer of the BP neural network structure in S4 is 6, the number of neurons in the hidden layer should be 3-8, and the number of neurons in the output layer is 2.

[0009] Preferably, in S4, the sigmod function is selected for the hidden layer of the BP neural network structure, the purelin function is selected for the output layer, and the training function is the Levenberg-Marquardt algorithm: ; wherein, is the input variable, is the hidden layer value, is the output value of the output layer.

[0010] Preferably, the training set of the BP neural network structure in S4 should cover all the deformation conditions of the experiment and contain at least the number of test-set strain rates × the number of test-set deformation temperatures × the test-set maximum deformation degree / 0.1 sample quantities.

[0011] Preferably, in S4, the input variables of the BP neural network structure except the strain rate are normalized by the following formula: ; wherein, is the normalized input variable, is the input variable except the strain rate, is the minimum value of the input variables, is the maximum value of the input variables.

[0012] Preferably, in S4, the strain rate in the input variables of the BP neural network structure is normalized by the following formula: ; wherein, is the normalized strain rate, is the logarithmized strain rate, is the minimum value of the logarithmized strain rate, is the maximum value of the logarithmized strain rate.

[0013] Generally speaking, compared with the prior art by the above technical solution conceived by the present invention, the following beneficial effects can be achieved.

[0014] (1) The intelligent prediction method of hot working quality that couples the influence of microstructure and deformation conditions is consistent with the actual processing situation. With dislocation density, degree of dynamic recrystallization, grain size, strain rate, deformation temperature, and deformation degree as input variables, the accurate prediction of hot working quality under complex deformation in the actual processing process is realized through the BP neural network; (2) The intelligent prediction method of hot working quality of the present invention has a wide application range, and is applicable to both the constant state of deformation temperature and strain rate and the continuously changing state of deformation temperature and strain rate, which is more in line with the real situation of the material processing process; (3) The intelligent prediction method for hot working quality of the present invention has accurate prediction results. When compared with the experimental results under sampling conditions, the correlation coefficients between the intelligent prediction method for hot working quality of the present invention and the power dissipation rate and deformation instability factor obtained from the steady-state single-pass compression experiment are 0.98569 and 0.97474 respectively, and the average relative errors are 3.8144% and 3.3848% respectively, and the average absolute errors are 0.0109 and 0.0289 respectively. Therefore, the present invention is a reliable numerical method, and the intelligent prediction method for hot working quality is relatively accurate. Brief Description of the Drawings

[0015] Figure 1 Shows the flowchart for constructing the intelligent prediction method for hot working quality of the present invention.

[0016] Figure 2 Shows the comparison between the experimental value of the power dissipation rate in randomly selected single-pass compression and the predicted value obtained by using the method of the present invention.

[0017] Figure 3 Shows the comparison between the experimental value of the deformation instability factor in randomly selected single-pass compression and the predicted value obtained by using the method of the present invention.

[0018] Figure 4 Shows the predicted value of the power dissipation rate under the condition of continuously changing deformation conditions.

[0019] Figure 5 Shows the predicted value of the deformation instability factor under the condition of continuously changing deformation conditions. Detailed Embodiments

[0020] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0021] Taking the establishment of the intelligent prediction method for hot working quality in the single-pass compression test of 316L austenitic stainless steel as an example to illustrate the present invention. The specific process is as Figure 1 shown.

[0022] The 316L austenitic stainless steel was processed into Φ8×12mm cylindrical compression specimens, and the single-pass compression test was carried out on a Gleeble3800 physical simulation testing machine. During the single-pass hot compression test, the specimens were first heated to 1150°C at a rate of 10°C / s and held for 240 s to ensure complete austenitization of the internal microstructure of the specimens. Secondly, under the control of the thermocouple, the temperature was reduced to 1000°C, 1050°C, and 1100°C at a cooling rate of 5°C / s and held for 120 s to ensure uniform temperature inside the specimens. Subsequently, isothermal compression experiments were carried out at strain rates of 0.01, 0.1, and 1 s -1 respectively. When the compression amount of the specimens reached 60%, the specimens were taken out of the equipment and rapidly water quenched. Nine engineering stress-engineering strain curves were obtained from the above experiments, and all the curves were transformed into true stress-strain curves. And the true stress-strain curves were corrected considering the influence of friction and adiabatic temperature rise phenomena.

[0023] Based on the dynamic material model theory, the power dissipation rate and deformation instability factor were solved for 153 cases where the true strain was 0.1, 0.15, 0.2, …, 0.9 under the above deformation temperatures and strain rates.

[0024] A macro-microscopic coupled thermal deformation constitutive model was established, and the dislocation density, degree of dynamic recrystallization, and grain size were solved for 153 cases where the true strain was 0.1, 0.15, 0.2, …, 0.9 under the above deformation temperatures and strain rates.

[0025] Furthermore, the following normalization treatments were carried out on the deformation temperature, true strain, dislocation density, degree of dynamic recrystallization, and grain size:

[0026] where, is the normalized input variable, is the input variable except the strain rate, is the minimum value of the input variables, is the maximum value of the input variables.

[0027] The following normalization treatment was carried out on the strain rate:

[0028] where, is the normalized strain rate, is the logarithmized strain rate, is the minimum value of the logarithmized strain rate, is the maximum value of the logarithmized strain rate.

[0029] Finally, taking the strain rate, deformation temperature, degree of deformation, and the dislocation density, grain size, and degree of dynamic recrystallization under corresponding conditions as input variables, and the power dissipation rate and deformation instability factor under corresponding conditions as output variables, a training set is formed to train the BP neural network structure, and an intelligent prediction model of hot processing quality considering the influence of coupled microstructure and deformation conditions is obtained.

[0030] The comparison between the experimental values of the power dissipation rate in single-pass compression randomly selected and the predicted values obtained by using the method of the present invention is as Figure 2 shown. The correlation coefficient is 0.98569, the average relative error is 3.8144%, and the average absolute error is 0.0109. The comparison between the experimental values of the deformation instability factor in single-pass compression randomly selected and the predicted values obtained by using the method of the present invention is as Figure 3 shown. The correlation coefficient is 0.97474, the average relative error is 3.3848%, and the average absolute error is 0.0289. Figure 4 、 5 The predicted values of the power dissipation rate and the deformation instability factor under the continuously changing state of deformation conditions are shown. It can be seen that as the strain rate continuously increases, the power dissipation rate and the deformation instability factor decrease, and the hot processing quality deteriorates. As the strain rate continuously decreases, the power dissipation rate and the deformation instability factor increase, and the hot processing quality improves. The prediction situation is consistent with the actual processing situation.

[0031] In summary, the intelligent prediction method of hot processing quality considering the influence of coupled microstructure and deformation conditions established by the present invention is a reliable and feasible numerical analysis method, and this method is applicable to fields such as numerical simulation and optimization of forming processes in the material processing process.

[0032] It is easy for those skilled in the art to understand that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent prediction method for hot processing quality that takes into account the influence of coupling organization and deformation conditions, characterized in that, It includes the following steps: S1: Obtain the stress-strain relationship of the single-pass hot deformation process through high-temperature physical simulation experiments and perform preprocessing; S2: Based on the continuous medium mechanics of large plastic deformation and the theory of irreversible thermodynamics, solve the power dissipation rate and deformation instability factor at different strain rates, deformation temperatures, and deformation degrees; S3: Apply the macro-micro coupled hot deformation constitutive model to solve the dislocation density, grain size, and dynamic recrystallization degree at different strain rates, deformation temperatures, and deformation degrees; S4: Establish a BP neural network structure, and use the strain rate, deformation temperature, deformation degree, and the dislocation density, grain size, and dynamic recrystallization degree under the corresponding conditions as input variables, and use the power dissipation rate and deformation instability factor under the corresponding conditions as output variables to form a training set. Normalize the input variables, and further train the BP neural network structure to obtain an intelligent prediction model of hot processing quality that couples the influence of microstructure and deformation conditions.

2. The method according to claim 1, characterized in that The acquisition and preprocessing of the stress-strain relationship of the single-pass hot deformation process in S1 include the following sub-steps: S1-1: Select cylindrical compression specimens with dimensions of Φ8×12mm or Φ10×15mm, conduct single-pass high-temperature compression tests at different strain rates and deformation temperatures, and obtain the engineering stress-strain relationship. The test should set at least 3 groups of strain rates and 3 groups of deformation temperatures; S1-2: Based on the strain conversion relationship, convert it into the single-pass true stress-strain relationship; S1-3: Correct the obtained true stress-strain relationship for friction and adiabatic temperature rise.

3. The method according to claim 1, characterized in that The macro-micro coupled hot deformation constitutive model in S3 should be able to predict the changes in dislocation density, grain size, and dynamic recrystallization degree with deformation conditions.

4. The method according to claim 1, wherein The number of neurons in the input layer of the BP neural network structure in S4 is 6, the number of neurons in the hidden layer should be 3-8, and the number of neurons in the output layer is 2.

5. The method according to claim 1, characterized in that In S4, the tansig type function is selected for the hidden layer of the BP neural network structure, the logsig type function is selected for the output layer, and the training function is the Levenberg-Marquardt algorithm: ； Among them, is the input variable, is the hidden layer value, is the output value of the output layer.

6. The method according to claim 1, wherein The training set of the BP neural network structure in S4 should cover all the deformation conditions of the experiment and contain at least the number of set strain rates in the experiment × the number of set deformation temperatures in the experiment × the set maximum deformation degree / 0.1 sample quantities.

7. The method according to claim 1, characterized in that, In S4, the following normalization process is performed on the input variables of the BP neural network structure except for the strain rate: ； Among them, is the normalized input variable, is the input variable except the strain rate, is the minimum value among the input variables, is the maximum value among the input variables.

8. The method according to claim 1, wherein In S4, the following normalization process is performed on the strain rate in the input variables of the BP neural network structure: ； Among them, is the normalized strain rate, is the logarithmized strain rate, is the minimum value of the logarithmized strain rate, is the maximum value of the logarithmized strain rate.

Images