CN111861041B - Method for predicting dynamic recrystallization type rheological stress of Nb microalloyed steel - Google Patents
Method for predicting dynamic recrystallization type rheological stress of Nb microalloyed steel Download PDFInfo
- Publication number
- CN111861041B CN111861041B CN202010767229.3A CN202010767229A CN111861041B CN 111861041 B CN111861041 B CN 111861041B CN 202010767229 A CN202010767229 A CN 202010767229A CN 111861041 B CN111861041 B CN 111861041B
- Authority
- CN
- China
- Prior art keywords
- stress
- strain
- rheological
- rheological stress
- curve
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 68
- 238000001953 recrystallisation Methods 0.000 title claims abstract description 58
- 229910000742 Microalloyed steel Inorganic materials 0.000 title claims abstract description 18
- 238000013178 mathematical model Methods 0.000 claims abstract description 44
- 229910000831 Steel Inorganic materials 0.000 claims abstract description 28
- 239000010959 steel Substances 0.000 claims abstract description 28
- 230000002068 genetic effect Effects 0.000 claims abstract description 15
- 238000013528 artificial neural network Methods 0.000 claims abstract description 11
- 238000003062 neural network model Methods 0.000 claims description 13
- 238000005272 metallurgy Methods 0.000 claims description 11
- 238000012216 screening Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 9
- 238000010438 heat treatment Methods 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 6
- 238000005482 strain hardening Methods 0.000 claims description 6
- 230000035772 mutation Effects 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 4
- 210000002569 neuron Anatomy 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 abstract description 10
- 238000002474 experimental method Methods 0.000 abstract description 5
- 230000006835 compression Effects 0.000 abstract description 4
- 238000007906 compression Methods 0.000 abstract description 4
- 238000011160 research Methods 0.000 abstract description 2
- 238000004364 calculation method Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 229910001566 austenite Inorganic materials 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 241000490494 Arabis Species 0.000 description 1
- 229910045601 alloy Inorganic materials 0.000 description 1
- 239000000956 alloy Substances 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013400 design of experiment Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 239000011343 solid material Substances 0.000 description 1
- 239000002436 steel type Substances 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Abstract
A method for predicting dynamic recrystallization type rheological stress of Nb microalloyed steel belongs to the technical field of steel research and machine learning intersection. Based on experimental data of dynamic recrystallization type rheological stress curves and steel grade information of series Nb microalloy steel, a genetic algorithm is adopted to learn parameters in mathematical models corresponding to each rheological stress curve, a Bayesian regularized BP neural network is used to establish a network relation model between the steel grade information and rheological stress curve characteristics, and then the mathematical models corresponding to the rheological stress curves are combined to predict dynamic recrystallization type rheological stress. The model established by the method can predict the rheological stress curve of the series of steels under various components and process conditions with high precision, obviously reduces the workload of single compression experiments, and improves the prediction efficiency and precision of dynamic recrystallization type rheological stress curves.
Description
Technical Field
The invention belongs to the technical field of steel research and machine learning, and particularly relates to a method for predicting dynamic recrystallization type rheological stress of Nb microalloyed steel.
Background
The Nb microalloyed high strength steel is widely applied to the aspects of pipelines, bridge buildings and the like. Nb microalloyed high strength steels require high strength and good toughness. The fine grain strengthening can improve the toughness of the steel at the same time, and the dynamic recrystallization is one of important ways for refining the austenite grain size, so that the final mechanical property of the Nb microalloyed high-strength steel is greatly influenced. At present, two main methods for researching the rheological stress of austenite dynamic recrystallization are available, one is to directly obtain the rheological stress curve of experimental steel by adopting a single compression experiment; the other is to build a mathematical model according to the existing rheological stress curve. The rheological stress curve can be obtained according to a single compression experiment, but the experiment workload is larger. The dynamic recrystallization type rheological stress curve can be predicted according to the rheological stress mathematical model established by the existing rheological stress curve, but the dynamic recrystallization type rheological stress curve is only suitable for a single steel grade and process conditions, and the precision is still to be improved. The machine learning theory and method can automatically learn the characteristics in the data set according to large-scale data, and has the characteristics of strong universality and high precision, and the machine learning theory and method have application in many aspects of material science at present, such as prediction of novel solid materials, calculation of material properties and the like. Aiming at the problems of precision and generalization of dynamic recrystallization type rheological stress prediction, machine learning can be well solved, so that the development of the work in the aspect has important significance.
By searching a national intellectual property office database and an SOOPAT database, no related patent is issued for the dynamic recrystallization type rheological stress of Nb microalloy steel at present; the prediction of dynamic recrystallization rheological stress in the related literature is only aimed at a single steel grade or a single process condition, and the precision is low and the application range is limited. Machine learning models were built for steady state rheological stresses during thermal deformation of X70 steel, such as abarghoooei et al, which predicts steady state rheological stresses, but which not only do not relate to the effect of chemical components on rheological stresses, but also do not predict the rheological stress curves of the whole deformation process [ abarghoooei H, arabi H, seyedein S H, et al modeling of Steady State Hot Flow Behavior of API-X70Microalloyed Steel using Genetic Algorithm and Design of Experiments J Applied Soft Computing,2017, 52:471-477 ].
Disclosure of Invention
Aiming at the defects existing in the prior art, the invention provides a method for predicting dynamic recrystallization type rheological stress, which takes a mathematical model as a guide, adopts a machine learning method to establish a model for predicting the dynamic recrystallization type rheological stress of Nb microalloy steel, is suitable for series Nb microalloy steel, has higher precision and applicability on the premise of ensuring the traditional physical metallurgy law, consumes shorter time, and is suitable for predicting the dynamic recrystallization type rheological stress of any steel grade or alloy.
A method for predicting dynamic recrystallization type rheological stress of Nb microalloyed steel, comprising the steps of:
step 1, constructing an initial data set based on experimental data of a dynamic recrystallization type rheological stress curve, steel grade information and technological parameters of the existing Nb microalloyed steel, wherein: the steel grade information comprises: c content, mn content, and Nb content; the technological parameters include: heating temperature, deformation temperature, maximum strain and strain rate;
step 2, judging whether the rheological stress curve in the initial data set accords with the physical metallurgy law, and reserving the rheological stress curve which accords with the physical metallurgy law to form a screening data set;
step 3, determining the actual measured peak strain epsilon of each rheological stress curve in the screening data set according to the rheological stress curves in the screening data set p Peak stress sigma p Steady state strain ε s And steady state stress sigma s ；
And 4, determining a mathematical model form of a dynamic recrystallization type rheological stress curve, wherein the method comprises the following specific steps of:
step 4-1, dividing rheological stress into two parts, namely a part before peak stress and a part after peak stress;
step 4-2, selecting a mathematical model A suitable for rheological stress before peak stress, and selecting a mathematical model B suitable for peak stress to steady-state stress, wherein:
the mathematical model form of the dynamic recrystallization type rheological stress curve is as follows:
wherein sigma is stress, epsilon is strain, sigma p For peak stress, ε p For peak strain, sigma s Is steady state stress, C and C 1 Is a constant;
step 5, respectively learning parameters in the mathematical model A and the mathematical model B according to the actual measurement rheological stress curve by adopting a genetic algorithm according to the dynamic recrystallization rheological stress mathematical model form determined in the step 4;
step 6, establishing a nonlinear mapping network relation model between steel grade information and dynamic recrystallization type rheological stress characteristics by adopting a Bayesian regularized BP neural network, and then carrying out model training to obtain a trained BP neural network model;
step 7, selecting at least one group of components and processes according to the trained BP neural network model, and predicting dynamic recrystallization type rheological stress characteristics;
and 8, combining the dynamic recrystallization rheological stress characteristics predicted by the step 7 with the rheological stress mathematical model A and the mathematical model B determined by the step 4 to obtain a dynamic recrystallization rheological stress curve.
In the step 3, the peak strain epsilon of each rheological stress curve in the screening data set is determined p Peak stress sigma p Steady state strain ε s And steady state stress sigma s The specific process of (2) is as follows:
determining peak strain epsilon from peaks on a rheological stress curve (i.e., stress sigma-strain epsilon curve) p And peak stress sigma p The method comprises the steps of carrying out a first treatment on the surface of the Definition of Strain hardening Rate(Δσ is stress increment, Δε is strain increment), and θ is first calculated from the strain hardening rate θ -strain ε curveThe strain at recovery to 0 value is taken as steady state strain epsilon s Steady state stress sigma s Determined from the stress sigma-strain epsilon curve.
In the step 5, according to the form of the mathematical model of the rheological stress determined in the step 4, a genetic algorithm is adopted, and parameters in the mathematical model A and the mathematical model B are learned according to the actually measured rheological stress curve, and the specific process is as follows:
setting parameters such as crossover rate, mutation rate, maximum iteration number and the like during genetic algorithm learning according to the actually measured rheological stress curve and the mathematical model form determined in the step 4, and learning parameters C and C in the mathematical model corresponding to each rheological stress curve 1 。
In the step 6, a Bayesian regularized BP neural network is adopted to establish a nonlinear mapping network relation model between steel grade information, technological parameters and dynamic recrystallization type rheological stress characteristics, and then model training is carried out, wherein the specific process is as follows:
establishing a three-layer neural network model by adopting a BP neural network based on Bayesian regularization, wherein input parameters of an input layer are C content, mn content, nb content, heating temperature, deformation temperature, maximum strain and strain rate; the output parameters of the output layer are peak strain, peak stress, steady state strain, steady state stress, C and C 1 The method comprises the steps of carrying out a first treatment on the surface of the The hidden layer is 5 neurons; the training of the neural network is then performed.
In the step 7, at least one group of components and processes are selected according to the trained BP neural network model, and the dynamic recrystallization type rheological stress characteristics are predicted, wherein the specific process is as follows:
for each component and the process thereof, predicting peak stress, peak strain, steady state stress, C and C under the conditions of the component and the process by using a trained BP neural network model 1 。
Compared with the prior art, the invention has the advantages that:
(1) The applicability is wide. According to the invention, 300 pieces of dynamic recrystallization type rheological stress curve of the Nb microalloy steel are collected, a data set is constructed, the data set contains more comprehensive components and technological parameter information of the Nb microalloy steel, the defect of less rheological stress information under single steel grade or technological conditions is avoided, and meanwhile, the selected mathematical model is ensured to have the characteristic of conforming to the physical metallurgy law, so that the comprehensive model has wider applicability;
(2) The precision is higher. The invention adopts a machine learning method to learn information in the rheological stress curve, adopts the machine learning method to construct a network model between steel grade information and rheological stress characteristics, overcomes the defect of low precision of the traditional physical metallurgy mathematical model, and has the characteristic of higher precision;
(3) Rheological stresses under a variety of ingredients and process conditions can be predicted. According to the invention, the network model established by machine learning is adopted, so that rheological stress characteristics in any range and under process conditions can be predicted, and a rheological stress curve is drawn according to the rheological stress mathematical model, so that the workload of a single compression experiment is greatly reduced, and the efficiency and the accuracy of predicting dynamic recrystallization type rheological stress are improved.
Drawings
FIG. 1 is a flow chart of a method for predicting dynamic recrystallization rheological stress of Nb microalloyed steel in accordance with example 1 of the present invention;
FIG. 2 is a graph comparing predicted rheological stress curves with measured curves for example 1 of the present invention, wherein:
FIG. 2 (a) is a graph showing the comparison of the predicted dynamic recrystallization rheological stress curve and the measured curve in the A-component process;
FIG. 2 (B) is a graph showing the comparison of the predicted dynamic recrystallization rheological stress curve and the measured curve in the B-component process;
FIG. 3 is a graph of the accuracy of the predicted rheological stress curve versus the measured curve for example 1 of the present invention, wherein:
FIG. 3 (a) is a graph of the accuracy of the predicted dynamic recrystallization rheological stress curve versus the measured curve for the A-component process;
fig. 3 (B) is a graph of the accuracy of the predicted dynamic recrystallization rheological stress curve versus the measured curve for the B-component process.
Detailed Description
Embodiments of the present invention are further described below with reference to the accompanying drawings.
Example 1
A method for predicting dynamic recrystallization rheological stress of Nb microalloyed steel, the flow chart of which is shown in figure 1, comprising the following steps:
step 1, constructing an initial data set based on experimental data of current 500 Nb microalloyed steel dynamic recrystallization rheological stress curves, steel grade information and process parameter information, wherein the steel grade information comprises: c content, mn content, and Nb content; the technological parameters include: heating temperature, deformation temperature, maximum strain and strain rate;
step 2, judging whether the collected 500 rheological stress curves accord with the physical metallurgy law, wherein the specific judgment standard is as follows: (1) judging whether the rheological stress curve accords with the physical metallurgy law under the same component and different deformation conditions. If the deformation temperature is reduced under the condition of different deformation temperatures of the same component, the rheological stress is gradually increased under the same strain quantity; under the condition of different strain rates of the same component, as the strain rate increases, the rheological stress gradually increases at the same strain quantity; (2) and judging whether the rheological stress curve accords with the physical metallurgy law or not when different components are in the same deformation condition. If the same deformation condition is different in Nb content, the rheological stress is gradually increased with the increase of the Nb content at the same strain amount; when the same deformation condition has different Mn contents, the rheological stress gradually increases with the increase of the Mn content and the same strain quantity; when the same deformation condition is different in C content, the rheological stress gradually decreases with the increase of the C content at the same strain amount. The data screened by the physical metallurgy principle form a screening data set which comprises 310 rheological stress curves, and table 1 shows steel grade and process parameter information corresponding to the screened rheological stress curves;
table 1 information on the steel types and process parameters selected
Step 3, according to the rheological stress curve in the screening data set, determining the actual measured peak strain epsilon of each curve in the data set p Peak stress sigma p Steady state strain ε s And steady state stress sigma s ；
In the embodiment of the invention, the peak strain epsilon is determined according to the peak value on the rheological stress curve (namely the stress sigma-strain epsilon curve) p And peak stress sigma p The method comprises the steps of carrying out a first treatment on the surface of the Definition of Strain hardening Rate(Δσ is the stress increment, Δε is the strain increment), and the strain when θ is recovered to 0 value for the first time is defined as the steady state strain ε from the strain hardening rate θ -strain ε curve s Steady state stress sigma s Determined from the stress sigma-strain epsilon curve; table 2 is information contained in the determined rheological stresses.
TABLE 2 rheological stress containing information
And 4, determining a mathematical model form of a dynamic recrystallization type rheological stress curve, wherein the mathematical model form is as follows:
wherein sigma is stress, epsilon is strain, sigma p For peak stress, ε p For peak strain, sigma s Is steady state stress, C and C 1 Is constant.
Step 5, according to the dynamic recrystallization type rheological stress mathematical model form determined in the step 4, adopting a genetic algorithm to respectively learn parameters C and C in the mathematical model according to the actually measured rheological stress curve 1 ；
In the embodiment of the invention, a genetic algorithm is adopted, and a mathematical model A and a mathematical model corresponding to each rheological stress curve are respectively learned according to the actually measured rheological stress curveParameters C and C in form B 1 For a 0 to peak strain, i.e., the parameter to be learned in mathematical model a is C; for peak strain to steady state strain, i.e. the parameter to be learned in mathematical model B is C 1 The method comprises the steps of carrying out a first treatment on the surface of the The population number of the genetic algorithm is set to be 50, the maximum evolution algebra is 100 generations, the selection operation adopts a roulette mode, the crossover operation adopts a single-point crossover algorithm, the crossover probability is set to be 0.85, the mutation operation adopts random uniform mutation, the mutation probability is set to be 0.01, and the maximum iteration is 5000 times.
(1) When the parameter C is learned, the parameter range is set to be 0.2-2.0, and the fitness function is as follows:
in the method, in the process of the invention,and->Respectively the ith strain epsilon i Corresponding stress calculation value and stress actual measurement value sigma p For peak stress, ε p Is the peak strain. Optimizing and obtaining optimal parameter C through genetic algorithm to enable FUN stress1 The minimum value is obtained.
(2) Learning parameter C 1 When the parameter range is set to be 1.0-250, the fitness function is as follows:
in the method, in the process of the invention,and->Respectively the ith strain epsilon i Corresponding stress calculation value and stress actual measurement value sigma s Is steady state stress, sigma p For peak stress, ε p Is the peak strain. Optimizing by genetic algorithm to obtain optimal parameter C 1 Make FUN stress2 The minimum value is obtained.
Optimizing the obtained C and C 1 Respectively putting the obtained products into mathematical models A and B until the correlation coefficient R of the optimized rheological stress and the measured rheological stress is more than 0.9 and the mean square error RMSE is less than 10MPa 2 (the calculation methods of the correlation coefficient R and the mean square error RMSE are shown in formulas (5) and (6) respectively), and the correspondingly obtained learning parameters are model parameters C and C after genetic algorithm learning 1 The range of model parameters after genetic algorithm learning is shown in table 3;
wherein M is i Is the ith strain epsilon i Corresponding measured value of rheological stress, P i Is the ith strain epsilon i The corresponding predicted value of the rheological stress model,is the average of the measured values of the rheological stress, +.>Is the average value of the predicted values of the rheological stress, and N is the total number of strains corresponding to each rheological stress curve;
table 3 model parameter ranges after genetic algorithm learning
Step 6, establishing a nonlinear mapping network relation model between steel grade information, process parameter information and dynamic recrystallization type rheological stress characteristics by adopting a BP neural network regularized by Bayes, and then carrying out model training;
in the embodiment of the invention, the specific model training process comprises the following steps: establishing a three-layer neural network model by adopting a BP neural network based on Bayesian regularization, wherein input parameters of an input layer are C content, mn content, nb content, heating temperature, deformation temperature, maximum strain and strain rate; the output parameters of the output layer are peak strain, peak stress, steady state strain, steady state stress, C and C1; the hidden layer is 5 neurons. 310 pieces of data information in the screening data set are processed according to 8:2 is divided into two parts, 80% is used for training and 20% is used for testing; and then training the neural network to obtain a trained BP neural network model.
Step 7, selecting at least one group of components and processes according to the trained BP neural network model, and predicting dynamic recrystallization type rheological stress characteristics;
in the embodiment of the invention, the specific prediction process is as follows: two groups of components are selected respectively, including:
component A: 0.1C-1.42Mn-0.035Nb;
and the component B comprises the following components: 0.104C-0.4Mn-0.05Nb;
the corresponding process is as follows:
a process: deformation temperature 1100 ℃ and strain rate 0.2s -1 Heating at 1400 deg.c to maximum strain of 3.0;
and B, technology: deformation temperature 1000 ℃ and strain rate 1s -1 Heating to 1200 ℃ and maximum strain of 1.0;
for each component and each process thereof, predicting dynamic recrystallization type rheological stress characteristics of the component and the process under the condition by a trained BP neural network model, wherein the dynamic recrystallization type rheological stress characteristics specifically comprise: peak stress sigma p Peak strain epsilon p Steady state stress sigma s C and C 1 The result is:
the component A comprises the following processes: epsilon p ＝0.6211，σ p ＝89.2581，σ s ＝72.7970，C＝0.6443，C 1 ＝3.1211；
The component B comprises the following processes: epsilon p ＝0.3998，σ p ＝161.686，σ s ＝147.7749，C＝0.5825，C 1 ＝21.2994。
Step 8, combining the dynamic recrystallization rheological stress characteristics predicted in the step 7 with the rheological stress curve mathematical model A and the mathematical model B determined in the step 4 to obtain a dynamic recrystallization rheological stress curve, wherein a comparison chart of the predicted dynamic recrystallization rheological stress curve and the actual measurement curve in the A component process is shown in fig. 2 (a), and a comparison chart of the predicted dynamic recrystallization rheological stress curve and the actual measurement curve in the B component process is shown in fig. 2 (B); in this example, the precision comparison graphs of the predicted dynamic recrystallization rheological stress curves under the a-component process and under the B-component process with the respective measured curves are shown in fig. 3 (a) and 3 (B), respectively.
Claims (2)
1. A method for predicting dynamic recrystallization rheological stress of Nb microalloyed steel, comprising the steps of:
step 1, constructing an initial data set based on experimental data of a dynamic recrystallization type rheological stress curve, steel grade information and technological parameters of the existing Nb microalloyed steel, wherein: the steel grade information comprises: c content, mn content, and Nb content; the technological parameters include: heating temperature, deformation temperature, maximum strain and strain rate;
step 2, judging whether the rheological stress curve in the initial data set accords with the physical metallurgy law, and reserving the rheological stress curve which accords with the physical metallurgy law to form a screening data set;
step 3, determining the actual measured peak strain epsilon of each rheological stress curve in the screening data set according to the rheological stress curves in the screening data set p Peak stress sigma p Steady state strain ε s And steady state stress sigma s ；
And 4, determining a mathematical model form of a dynamic recrystallization type rheological stress curve, wherein the method comprises the following specific steps of:
step 4-1, dividing rheological stress into two parts, namely a part before peak stress and a part after peak stress;
step 4-2, selecting a mathematical model A suitable for rheological stress before peak stress, and selecting a mathematical model B suitable for peak stress to steady-state stress, wherein:
the mathematical model form of the dynamic recrystallization type rheological stress curve is as follows:
wherein sigma is stress, epsilon is strain, sigma p For peak stress, ε p For peak strain, sigma s Is steady state stress, C and C 1 Is a constant;
and 5, respectively learning parameters in the mathematical model A and the mathematical model B according to the actual measurement rheological stress curve by adopting a genetic algorithm according to the dynamic recrystallization rheological stress mathematical model form determined in the step 4, wherein the specific process is as follows:
setting parameters of the crossover rate, the mutation rate and the maximum iteration number during the learning of a genetic algorithm according to the actually measured rheological stress curve and the mathematical model form determined in the step 4, and learning parameters C and C in the mathematical model corresponding to each rheological stress curve 1 ；
Step 6, a Bayesian regularized BP neural network is adopted to establish a nonlinear mapping network relation model between steel grade information and dynamic recrystallization type rheological stress characteristics, and then model training is carried out to obtain a trained BP neural network model, wherein the specific process is as follows:
establishing a three-layer neural network model by adopting a BP neural network based on Bayesian regularization, wherein the input parameters of an input layer are CContent, mn content, nb content, heating temperature, deformation temperature, maximum strain amount and strain rate; the output parameters of the output layer are peak strain, peak stress, steady state strain, steady state stress, C and C 1 The method comprises the steps of carrying out a first treatment on the surface of the The hidden layer is 5 neurons; then training the neural network;
step 7, selecting at least one group of components and processes according to the trained BP neural network model, and predicting dynamic recrystallization type rheological stress characteristics, wherein the specific process is as follows:
for each component and the process thereof, predicting peak stress, peak strain, steady state stress, C and C under the conditions of the component and the process by using a trained BP neural network model 1 ；
And 8, combining the dynamic recrystallization rheological stress characteristics predicted by the step 7 with the rheological stress mathematical model A and the mathematical model B determined by the step 4 to obtain a dynamic recrystallization rheological stress curve.
2. The method for predicting dynamic recrystallization type rheological stress of Nb microalloyed steel of claim 1 wherein in step 3, the peak strain ε of each rheological stress curve in the screening dataset is determined p Peak stress sigma p Steady state strain ε s And steady state stress sigma s The specific process of (2) is as follows:
determining peak strain epsilon from peaks on a rheological stress curve p And peak stress sigma p The method comprises the steps of carrying out a first treatment on the surface of the Definition of Strain hardening RateWherein Δσ is the stress increment and Δε is the strain increment; the strain when θ is recovered to 0 value for the first time is regarded as steady state strain ε according to the θ -strain ε curve of strain hardening rate s Steady state stress sigma s Determined from the stress sigma-strain epsilon curve.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010767229.3A CN111861041B (en) | 2020-08-03 | 2020-08-03 | Method for predicting dynamic recrystallization type rheological stress of Nb microalloyed steel |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010767229.3A CN111861041B (en) | 2020-08-03 | 2020-08-03 | Method for predicting dynamic recrystallization type rheological stress of Nb microalloyed steel |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111861041A CN111861041A (en) | 2020-10-30 |
| CN111861041B true CN111861041B (en) | 2023-09-05 |
Family
ID=72952790
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010767229.3A Active CN111861041B (en) | 2020-08-03 | 2020-08-03 | Method for predicting dynamic recrystallization type rheological stress of Nb microalloyed steel |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111861041B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112507542B (en) * | 2020-12-02 | 2022-04-26 | 中南大学 | Method for accurately predicting stress relaxation behavior under tensile stress and compressive stress states |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102339344A (en) * | 2011-05-25 | 2012-02-01 | 深圳大学 | Back analysis identification method for parameters of dynamic re-crystallizing model |
| CN105652666A (en) * | 2016-03-09 | 2016-06-08 | 中南大学 | Large die forging press beam feeding speed predictive control method based on BP neural networks |
| CN111079309A (en) * | 2019-12-30 | 2020-04-28 | 湖北汽车工业学院 | Multi-pass compression rheological stress model establishment method for coupled recrystallization dynamics |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11488010B2 (en) * | 2018-12-29 | 2022-11-01 | Northeastern University | Intelligent analysis system using magnetic flux leakage data in pipeline inner inspection |
-
2020
- 2020-08-03 CN CN202010767229.3A patent/CN111861041B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102339344A (en) * | 2011-05-25 | 2012-02-01 | 深圳大学 | Back analysis identification method for parameters of dynamic re-crystallizing model |
| CN105652666A (en) * | 2016-03-09 | 2016-06-08 | 中南大学 | Large die forging press beam feeding speed predictive control method based on BP neural networks |
| CN111079309A (en) * | 2019-12-30 | 2020-04-28 | 湖北汽车工业学院 | Multi-pass compression rheological stress model establishment method for coupled recrystallization dynamics |
Non-Patent Citations (1)
| Title |
|---|
| 高铌微合金钢高温变形流变应力预测模型;肖宝亮等;《钢铁研究学报》;第20卷(第9期);第28-31页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111861041A (en) | 2020-10-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jie et al. | Flow behavior of Al–6.2 Zn–0.70 Mg–0.30 Mn–0.17 Zr alloy during hot compressive deformation based on Arrhenius and ANN models | |
| CN111310348A (en) | Material constitutive model prediction method based on PSO-LSSVM | |
| CN114611844B (en) | Method and system for determining alloy addition amount in converter tapping process | |
| CN111933221B (en) | Method for predicting dynamic recrystallization fraction of Nb microalloyed steel | |
| CN112651562B (en) | Laser cladding process optimization method based on signal-to-noise ratio and improved gray correlation degree | |
| CN111861041B (en) | Method for predicting dynamic recrystallization type rheological stress of Nb microalloyed steel | |
| CN110472349A (en) | A kind of hot-rolled steel performance prediction method based on EEMD and depth convolutional network | |
| CN114638164B (en) | Method for predicting high-temperature creep life of power station pressure pipeline | |
| CN108062583A (en) | A kind of heating furnace technique parameter optimization method towards energy-saving and emission-reduction | |
| CN114662406A (en) | Method for establishing rock joint peak shear strength prediction model with assistance of machine learning | |
| CN113191092A (en) | Industrial process product quality soft measurement method based on orthogonal increment random configuration network | |
| CN116882585A (en) | Genetic algorithm and neural network coupled aluminum alloy hub low-pressure casting process optimization method | |
| Jimenez-Martinez et al. | Fatigue Life Prediction of Aluminum Using Artificial Neural Network. | |
| Wen et al. | Machine learning-assisted constitutive modeling of a novel powder metallurgy superalloy | |
| CN113128124B (en) | Multi-grade C-Mn steel mechanical property prediction method based on improved neural network | |
| CN113033105A (en) | Steel material fatigue performance prediction method based on mechanical theory guide transfer learning | |
| CN116484745A (en) | Design method of hot-rolled low-alloy steel for physical metallurgy guidance industry big data mining | |
| CN116611348A (en) | Unified parameter-free fatigue life prediction method for metal materials based on machine learning | |
| Wei et al. | High-cycle fatigue SN curve prediction of steels based on a transfer learning-guided convolutional neural network | |
| CN115938496A (en) | Quality Estimation Method Based on XGBoost Algorithm | |
| CN115394381A (en) | High-entropy alloy hardness prediction method and device based on machine learning and two-step data expansion | |
| CN110442953B (en) | Q & P steel design method based on machine learning under guidance of physical metallurgy | |
| CN117494531B (en) | Medium carbon steel decarburization depth prediction method based on finite element and XGBoost algorithm | |
| CN113343566B (en) | Nickel-based alloy fracture toughness prediction method and system based on deep learning | |
| CN115659688B (en) | Process determination method for improving reliability of additive aluminum alloy and composite material thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |