CN117727387A - Method for designing components of high-strength low-alloy steel - Google Patents
Method for designing components of high-strength low-alloy steel Download PDFInfo
- Publication number
- CN117727387A CN117727387A CN202211021355.XA CN202211021355A CN117727387A CN 117727387 A CN117727387 A CN 117727387A CN 202211021355 A CN202211021355 A CN 202211021355A CN 117727387 A CN117727387 A CN 117727387A
- Authority
- CN
- China
- Prior art keywords
- alloy steel
- strength low
- machine learning
- data
- model
- 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.)
- Pending
Links
- 229910000922 High-strength low-alloy steel Inorganic materials 0.000 title claims abstract description 50
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000010801 machine learning Methods 0.000 claims abstract description 60
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 29
- 238000012795 verification Methods 0.000 claims abstract description 22
- 230000002068 genetic effect Effects 0.000 claims abstract description 21
- 238000004364 calculation method Methods 0.000 claims abstract description 20
- 238000012549 training Methods 0.000 claims description 26
- 239000000203 mixture Substances 0.000 claims description 18
- 229910045601 alloy Inorganic materials 0.000 claims description 10
- 239000000956 alloy Substances 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000007637 random forest analysis Methods 0.000 claims description 8
- 238000010972 statistical evaluation Methods 0.000 claims description 7
- 238000012706 support-vector machine Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 5
- 238000009864 tensile test Methods 0.000 claims description 5
- 238000007550 Rockwell hardness test Methods 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 210000000349 chromosome Anatomy 0.000 claims description 3
- 238000005275 alloying Methods 0.000 claims description 2
- 239000000463 material Substances 0.000 abstract description 11
- 238000005457 optimization Methods 0.000 abstract description 4
- 108700003861 Dominant Genes Proteins 0.000 abstract description 3
- 238000004458 analytical method Methods 0.000 abstract description 3
- 238000012512 characterization method Methods 0.000 abstract description 3
- 230000008030 elimination Effects 0.000 abstract description 3
- 238000003379 elimination reaction Methods 0.000 abstract description 3
- 230000007786 learning performance Effects 0.000 abstract description 3
- 238000012216 screening Methods 0.000 abstract description 3
- 229910000831 Steel Inorganic materials 0.000 description 14
- 239000010959 steel Substances 0.000 description 14
- 238000013461 design Methods 0.000 description 8
- 229910052799 carbon Inorganic materials 0.000 description 7
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 6
- 238000010438 heat treatment Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 238000012827 research and development Methods 0.000 description 3
- 239000000126 substance Substances 0.000 description 3
- 238000001816 cooling Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 229910000851 Alloy steel Inorganic materials 0.000 description 1
- 201000004569 Blindness Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 229910052804 chromium Inorganic materials 0.000 description 1
- 239000011365 complex material Substances 0.000 description 1
- 239000013078 crystal Substances 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 229910052750 molybdenum Inorganic materials 0.000 description 1
- 229910052758 niobium Inorganic materials 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 229910052720 vanadium Inorganic materials 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
- 238000003466 welding Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Investigating And Analyzing Materials By Characteristic Methods (AREA)
Abstract
The invention discloses a method for designing high-strength low-alloy steel components, which is characterized in that a machine learning performance prediction model is established by Jmatpro calculation data based on machine learning and genetic algorithm, the relation between non-dominant components and performances in a material complex system is found, a global optimal solution is searched through genetic algorithm multi-objective optimization, rapid screening of dominant genes is carried out based on a biological superior/inferior elimination principle, and the optimal component process final experimental characterization verification analysis is obtained in a very large component process search range through high-throughput calculation, so that a high-strength low-alloy steel component scheme meeting the needs can be rapidly found.
Description
Technical Field
The invention relates to alloy steel component design technology, in particular to a high-strength low-alloy steel component design method.
Background
High strength low alloy steels (HSLA: high Strength Low Alloy Steels) are a class of weldable low carbon engineering structural steels. The main characteristics are low carbon content, good weldability (carbon content is generally lower than 0.45%, pcm is smaller than or equal to 0.3%), fine crystal grain, high yield strength, and commonly adopting Nb, V, ti and other alloy elements for strengthening and toughening. It is produced by advanced refining and heat treatment processes. The high-strength low-alloy steel can achieve good comprehensive use performance of combination of strength, toughness, formability and weldability at quite low cost, so that the high-strength low-alloy steel is widely applied. Development of high-strength low-alloy steel is a key measure for changing from the national steel country to the national steel country in China, so that development and development of high-strength low-alloy steel suitable for welding and heat treatment are necessary to meet the requirements of different structures on steel.
At present, how to design high-strength low-alloy steel efficiently and quickly to meet the current requirements remains a great challenge. The traditional trial-and-error method is difficult to quickly and accurately find out the influence rule of elements on the alloy structure and performance, difficult to quickly locate components according to requirements, and greatly consumes manpower and material resources.
Currently, machine learning is applied in the field of materials primarily to derive predicted performance from compositional processes.
For example: chinese patent application No. 202010693602.5 entitled "thermoelectric material performance prediction based on artificial intelligence data analysis" discloses a method for predicting thermoelectric material performance by building a thermoelectric material parameter database and training a model. However, this technique does not allow for component design by deriving the corresponding components from the performance requirements.
Therefore, there is a need to devise a new approach that can combine machine learning models and genetic algorithms to derive components towards performance requirements to solve or ameliorate the above problems.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a component design method for high-strength low-alloy steel, which is characterized in that a machine learning performance prediction model is established for key performances of the high-strength low-alloy steel, the relation between non-dominant components and performances in a complex system of materials is found, a global better solution is searched for through multi-objective optimization of a genetic algorithm, rapid screening of dominant genes is carried out based on a biological superior/inferior elimination principle, and an optimal component process is obtained in a very large component process searching range through high-throughput calculation, and finally experimental characterization verification analysis is carried out, so that a component scheme meeting the needs can be rapidly found.
A method for designing components of high-strength low-alloy steel comprises the following steps.
S1, establishing a high-strength low-alloy steel database;
s2, preprocessing data, namely normalizing and scaling the database to the same interval, and dividing the database into a training set and a verification set;
s3, establishing a machine learning prediction model, and training a training set in a database by adopting a plurality of different machine learning algorithms;
s4, model evaluation, namely selecting a proper statistical evaluation index, comprehensively comparing various performances of different machine learning models, and selecting a machine learning model which is most suitable for a high-strength low-alloy steel mechanical property database;
s5, experimental verification, namely performing prediction verification on data which do not participate in training by adopting an optimal machine learning model;
s6, optimizing the model through a genetic algorithm on the basis of a machine learning model, and reversing the components by utilizing the expected performance within a limited component range.
In step S1, the data source of the database is JmatPro software, and the high-strength low alloy steel calculation data is obtained through calculation.
The high-strength low-alloy steel calculation data comprise input data and output data, wherein the input data comprise alloy elements, austenitizing temperature and grain size; the output data includes yield strength, tensile strength and Rockwell hardness.
In step S2, the data preprocessing adopts a minmaxscaler module of a sklearn library in python, scales the data to be within the [ -1,1] interval, sets 70% of the normalized data as a training set and sets 30% as a verification set.
In step S3, the machine learning algorithm includes: support vector machine, gradient lifting regression machine, neural network and random forest.
Prior to the model comparison, a mesh search module using the sklearn library in python found hyper-parameters that optimized the model within the given range of model parameters.
In step S4, the statistical evaluation index is to quantitatively represent the prediction performance of the machine learning model by using the mean absolute error and the mean square error.
In step S5, the predictive verification includes performing a room temperature tensile test and a rockwell hardness test on a universal tensile tester according to GB/T228.1-2010.
The step S6 specifically includes the following steps:
s61, initializing a parent population N;
s62, performing non-dominant ranking and fitness calculation based on a machine learning model;
s63, performing chromosome coding on the components, and simultaneously selecting, crossing and mutating;
s64, generating a subgroup and merging 2N with the father group;
s65, adopting elite strategy to win and lose the rank in non-dominant order;
s66, generating a new population N.
In S61, initializing parent population N to 100 according to the defined ranges of alloying elements, austenitizing temperature and grain size, and predicting corresponding mechanical properties using a machine learning model; in S63, the coding mode is binary coding; the selected mode is roulette; the crossing mode is multi-point crossing, and the crossing probability Pc is 0.7; the probability of variation Pm is 0.3 and the maximum iteration number time is 400.
The method for designing the components of the high-strength low-alloy steel has the following advantages:
1. the machine learning model and the genetic algorithm can be combined to realize reverse deduction from performance to composition;
2. the blindness of the research and development of the traditional trial-and-error method material can be avoided, the component research and development efficiency of the high-strength low-alloy steel is improved, and the research and development period and components are greatly reduced.
Drawings
FIG. 1 is a flow chart of a method of designing a composition of a high strength low alloy steel in accordance with the present invention;
fig. 2 is a flowchart showing the specific process of S6 in fig. 1.
Detailed Description
The method is further described below, and based on machine learning and genetic algorithm, a machine learning performance prediction model is established through Jmatpro calculation data, the relation between non-dominant components and performances in a complex material system is found, a global better solution is searched through multi-objective optimization search of the genetic algorithm, rapid screening of dominant genes is performed based on a biological superior/inferior elimination principle, and an optimal component process is obtained in a very large component process search range through high-throughput calculation, and finally experimental characterization verification analysis is performed, so that a component scheme meeting the needs can be rapidly found. As shown in fig. 1, it comprises the steps of:
s1, establishing a high-strength low-alloy steel database; the data source of the database is JMATPro software, and high-strength low-alloy steel calculation data are obtained through calculation; the high-strength low-alloy steel calculation data comprise input data and output data, wherein the input data comprise alloy elements, austenitizing temperature and grain size; the output data includes yield strength, tensile strength and Rockwell hardness.
S2, preprocessing data, namely normalizing and scaling the database to the same interval, and dividing the database into a training set and a verification set; the data preprocessing adopts a minmaxscaler module of a sklearn library in python, the data is scaled into a [ -1,1] interval, 70% of normalized data is set as a training set, and 30% is set as a verification set.
S3, establishing a machine learning prediction model, and training a training set in a database by adopting a plurality of different machine learning algorithms; the machine learning algorithm includes: support vector machine, gradient lifting regression machine, neural network and random forest.
S4, model evaluation, namely selecting a proper statistical evaluation index, comprehensively comparing various performances of different machine learning models, and selecting a machine learning model which is most suitable for a high-strength low-alloy steel mechanical property database; before the model comparison, a grid search module of a sklearn library in python is adopted to find out the hyper-parameters optimizing the model in the given range of model parameters; the statistical evaluation index is to quantitatively represent the prediction performance of the machine learning model by using average absolute error and mean square error.
S5, experimental verification, namely performing prediction verification on data which do not participate in training by adopting an optimal machine learning model; the prediction verification uses a model to calculate the performance of components which are not in a database, and then the components are subjected to experimental detection performance, and compared with the model calculation, the performance comprises room temperature tensile test and Rockwell hardness test on a universal tensile testing machine according to GB/T228.1-2010.
S6, optimizing the model through a genetic algorithm on the basis of a machine learning model, and reversely pushing the components by utilizing the expected performance within a limited component range; as shown in fig. 2, the step S6 specifically includes the following steps:
s61, initializing a parent population N;
s62, performing non-dominant ranking and fitness calculation based on a machine learning model;
s63, performing chromosome coding on the components, and simultaneously selecting, crossing and mutating;
s64, generating a subgroup and merging 2N with the father group;
s65, adopting elite strategy to win and lose the rank in non-dominant order;
s66, generating a new population N.
In S61, initializing parent population N to 100 according to the defined range of austenitizing temperature and grain size of the alloy element, and predicting corresponding mechanical properties by using a machine learning model; in S63, the coding mode is binary coding; the selected mode is roulette; the crossing mode is multi-point crossing, and the crossing probability Pc is 0.7; the probability of variation Pm is 0.3 and the maximum iteration number time is 400.
According to the invention, the mechanical properties corresponding to all possible component schemes in the element range can be calculated through the machine learning model, then the performance evaluation condition of each component in the element range can be calculated through the genetic algorithm through the given mechanical property requirement, and finally the optimal component scheme meeting the performance requirement is output, so that a large amount of manpower and material resources are saved, the cost is reduced, the efficiency and the accuracy are improved, and the method can be expanded to the design of other material components.
The following embodiments take a method for designing a 50mm thick high strength low alloy steel composition based on machine learning and genetic algorithm as an example, and specifically describe the application of the method. The method specifically comprises the following steps:
s1: establishing a database of high strength low alloy steels
Taking a high-strength low-alloy steel plate with the thickness of 50mm as an example, mechanical property data of the high-strength low-alloy steel used for modeling come from JMatPro software, 660 groups of data are obtained through the JMatPro software, the austenitizing temperature is uniformly set to 900 ℃, and the grain size is set to 9. Each set of data includes seven input variables C, si, mn, ni, cr, mo, B and yield strength, tensile strength, and rockwell hardness at a position 25mm from the surface thickness direction as output variables. Three properties of yield strength, tensile strength and Rockwell hardness at a core position (25 mm from the surface) of a high-strength low-alloy steel plate with a thickness of 50mm were taken as output variables predicted by a machine learning model, wherein the chemical composition property data distribution of the steel in the database is as shown in Table 1 (mass fraction):
table 1 chemical composition and mechanical Properties of steels
S2: and (3) preprocessing data, namely normalizing and scaling the database to the same interval, and dividing the database into a training set and a verification set.
Because the input parts such as components and processes are different in units like the conventional element C and the microalloy element B, the prediction result of the model can be greatly influenced, and in order to eliminate the unit problem of the input parts of the model, unified normalization processing is required to be carried out on input data, and data samples are scaled to be within the same interval of [ -1,1 ].
The data preprocessing uses the minmaxscaler function of the sklearn library in python, and the input components are normalized. By data normalization, the data can be scaled to a specific range, and the calculation formula is as follows:
where u is the characteristic column mean, the raw data values in the xi database, N is the data sample size 660, and Z is the value after normalization of each data.
70% of the normalized data were set as training set and 30% were set as validation set. The validation set does not participate in the training model.
S3: establishing a machine learning prediction model, and training a training set in a database by adopting a plurality of different machine learning algorithms
Because the method is the calculation data of the JMAPro, when a machine learning algorithm is selected in comparison, only the optimal yield strength prediction model is selected, and the machine learning model is adopted for the subsequent prediction of the tensile strength resistance and the Rockwell hardness.
The adopted machine learning method comprises the following steps: support vector machines (SVR), gradient lifting regression machines (GBR), neural networks (ANN) and Random Forests (RF).
Parameters of four machine learning models were optimized using grid search (gridsearch cv) in the Sklearn library of python, the specific key parameter search ranges for the four machine learning methods are shown in table 2, the search was completed, and the optimal parameters for each machine learning method given by gridsearch cv are shown in table 3:
table 2 search scope of grid search method
TABLE 3 optimal parameters for machine learning models
S4: model evaluation, namely selecting a proper statistical evaluation index, comprehensively comparing various performances of different machine learning models, and selecting a machine learning model which is most suitable for a high-strength low-alloy steel mechanical property database;
for quantitative description and comparison of prediction errors of machine learning models. A Mean Absolute Error (MAE) and a Mean Square Error (MSE) are chosen to quantitatively represent the predictive performance of the machine learning model. As can be seen from the formula, the lower the Mean Absolute Error (MAE) and Mean Square Error (MSE) values, the more accurate the model is predicted. The calculation formula is as follows:
yi-the original value in the database,-machine learning the calculated values.
The comparison of the four machine-learned statistical indexes is shown in tables 4 and 5, and it can be seen that the MAE and MSE of the RF model on the training set are the lowest, but the MAE and MSE of the RF model on the test set are very high, and the overfitting phenomenon occurs. And the MAE and MSE of the ANN model on the training set and the test set are low, which indicates that the ANN model has good prediction performance on training data for modeling or verification data which does not participate in modeling. The ANN model is thus ultimately selected.
Table 4 statistical index MAE comparison results for four machine learns
| Testing set | Training set | |
| SVR | 17.2 | 12.5 |
| GBR | 23.6 | 11.3 |
| RF | 22.4 | 4.9 |
| ANN | 6.89 | 5.6 |
Table 5 results of comparison of four machine learned statistical indices MSE
And then an ANN model is adopted, and a machine learning prediction model of tensile strength and Rockwell hardness is also established.
S5: experimental verification, namely carrying out prediction verification on data which do not participate in training by adopting an optimal machine learning model:
three steel plates with different components and thicknesses of 50mm are adopted for mechanical property experiments, and the components are shown in table 6. The three experimental steels are normalized and quenched, and the specific process comprises heating to 900 ℃, preserving heat for 30min, air cooling, heating to 900 ℃ again, preserving heat for 30min, and water cooling. And cutting a standard tensile sample with phi 5 at 1/2 of the thickness direction of the experimental steel, carrying out room-temperature tensile experiments on a universal tensile testing machine according to GB/T228.1-2010, detecting Rockwell hardness on a durometer, and testing the mechanical properties of the sample at room temperature. The yield strength, tensile strength and Rockwell hardness test results are shown in Table 7.
TABLE 6 Experimental chemical composition
| No. | C | Si | Mn | Cr | Ni | Mo | V | Ti | Nb | B |
| 1 | 0.16 | 0.20 | 1.0 | 0.50 | 1.0 | 0.65 | 0.10 | 0.015 | 0 | 0.0015 |
| 2 | 0.16 | 0.20 | 1.0 | 0.50 | 2.0 | 0.65 | 0.10 | 0.015 | 0 | 0.0015 |
| 3 | 0.16 | 0.20 | 1.0 | 0.50 | 2.0 | 0.65 | 0.10 | 0.015 | 0 | 0.0015 |
TABLE 7 comparison of experimental values, calculated and predicted values for yield and tensile Strength
S6: based on the machine learning model, optimizing the model by a genetic algorithm, and reversely deducing components by utilizing the wanted performance within a limited component range:
after the machine learning model is built, the excellent characteristics of the genetic algorithm and the association of components and performances built by the machine learning model are needed to be utilized later, and a material component scheme meeting the performance requirements of people is designed.
Taking the design of 890MPa high-strength low-alloy steel as an example, when the genetic algorithm is used for carrying out optimal solution search, the element search range of 890MPa high-strength low-alloy steel is shown in table 8, and the mechanical property requirements of 890MPa high-strength low-alloy steel at different plate thicknesses are shown in table 9. The optimal solution will be generated in both of these ranges.
Table 8 genetic algorithm element search Range
Table 9 genetic Algorithm Performance Requirements
As can be seen from tables 8 and 9, designing a high strength low alloy steel composition of 50mm thickness requires comprehensive consideration in terms of Rockwell hardness, yield strength, tensile strength, and carbon equivalent, and thus involves a multi-objective optimization process in genetic algorithm. If the Rockwell hardness, the yield strength, the tensile strength and the carbon equivalent are required to meet the mechanical property requirement of the steel plate at the same time, the carbon equivalent and the performance requirement can be excessively deviated and eliminated in the optimizing process of the Rockwell hardness, however, the eliminated solution can be the optimal solution of the yield strength, so that the only optimal solution for meeting the four performance requirements at the same time is impossible to find, and multiple-aspect weighing selection is needed in the solving process, so that a Pareto optimal solution set is solved, wherein the Pareto optimal solution set is a non-inferior optimal solution.
In the genetic algorithm, the solving method of the multi-objective optimizing process is divided into: the selection method, the mixing method and the parallel selection method are selected because the optimized performances are that the yield strength, the tensile strength and the Rockwell hardness are all mechanical properties and are the same as the strength and the weakness. By means of a multi-objective genetic algorithm, a high strength low alloy steel of 50mm thickness was subjected to the performance requirements of Table 9 within the composition range of Table 8, giving a total of 32 composition results, as shown in Table 10. It can be seen that the Ni content is generally low and the Mo and Cr contents are generally high in the components satisfying the properties of the high-strength low-alloy steel having a thickness of 50 mm.
Table 10 results of genetic algorithm design for 50mm thick high strength low alloy steels
However, it will be appreciated by persons skilled in the art that the above embodiments are provided for illustration of the invention and not for limitation thereof, and that changes and modifications to the above described embodiments are intended to fall within the scope of the appended claims as long as they fall within the true spirit of the invention.
Claims (10)
1. The method for designing the components of the high-strength low-alloy steel is characterized by comprising the following steps of:
s1, establishing a high-strength low-alloy steel database;
s2, preprocessing data, namely normalizing and scaling the database to the same interval, and dividing the database into a training set and a verification set;
s3, establishing a machine learning prediction model, and training a training set in a database by adopting a plurality of different machine learning algorithms;
s4, model evaluation, namely selecting a proper statistical evaluation index, comprehensively comparing various performances of different machine learning models, and selecting a machine learning model which is most suitable for a high-strength low-alloy steel mechanical property database;
s5, experimental verification, namely performing prediction verification on data which do not participate in training by adopting an optimal machine learning model;
s6, optimizing the model through a genetic algorithm on the basis of a machine learning model, and reversing the components by utilizing the expected performance within a limited component range.
2. A method of designing a high strength low alloy steel composition according to claim 1, wherein: in step S1, the data source of the database is JmatPro software, and the high-strength low alloy steel calculation data is obtained through calculation.
3. A method of designing a high strength low alloy steel composition according to claim 2, wherein: the high-strength low-alloy steel calculation data comprise input data and output data, wherein the input data comprise alloy elements, austenitizing temperature and grain size; the output data includes yield strength, tensile strength and Rockwell hardness.
4. A method of designing a high strength low alloy steel composition according to claim 1, wherein: in step S2, the data preprocessing adopts a minmaxscaler module of a sklearn library in python, scales the data to be within the [ -1,1] interval, sets 70% of the normalized data as a training set and sets 30% as a verification set.
5. The method for designing the composition of the high-strength low-alloy steel according to claim 1, wherein: in step S3, the machine learning algorithm includes: support vector machine, gradient lifting regression machine, neural network and random forest.
6. The method for designing the composition of the high-strength low-alloy steel according to claim 1, wherein: in step S4, prior to the model comparison, the mesh search module using the sklearn library in python finds hyper-parameters that optimize the model within the given range of model parameters.
7. The method for designing the composition of the high-strength low-alloy steel according to claim 1, wherein: in step S4, the statistical evaluation index is to quantitatively represent the prediction performance of the machine learning model by using the mean absolute error and the mean square error.
8. The method for designing the composition of the high-strength low-alloy steel according to claim 1, wherein: in step S5, the predictive verification includes performing a room temperature tensile test and a rockwell hardness test on a universal tensile tester according to GB/T228.1-2010.
9. The method for designing a high-strength low-alloy steel composition according to claim 1, wherein the step S6 comprises the steps of:
s61, initializing a parent population N;
s62, performing non-dominant ranking and fitness calculation based on a machine learning model;
s63, performing chromosome coding on the components, and simultaneously selecting, crossing and mutating;
s64, generating a subgroup and merging 2N with the father group;
s65, adopting elite strategy to win and lose the rank in non-dominant order;
s66, generating a new population N.
10. The method for designing the high-strength low-alloy steel composition according to claim 9, wherein: in S61, initializing parent population N to 100 according to the defined ranges of alloying elements, austenitizing temperature and grain size, and predicting corresponding mechanical properties using a machine learning model; in S63, the coding mode is binary coding; the selected mode is roulette; the crossing mode is multi-point crossing, and the crossing probability Pc is 0.7; the probability of variation Pm is 0.3 and the maximum iteration number time is 400.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211021355.XA CN117727387A (en) | 2022-08-24 | 2022-08-24 | Method for designing components of high-strength low-alloy steel |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211021355.XA CN117727387A (en) | 2022-08-24 | 2022-08-24 | Method for designing components of high-strength low-alloy steel |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117727387A true CN117727387A (en) | 2024-03-19 |
Family
ID=90207445
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211021355.XA Pending CN117727387A (en) | 2022-08-24 | 2022-08-24 | Method for designing components of high-strength low-alloy steel |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117727387A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118643334A (en) * | 2024-06-19 | 2024-09-13 | 北京科技大学 | Method and device for optimizing low alloy steel performance based on differential evolution and machine learning algorithm |
-
2022
- 2022-08-24 CN CN202211021355.XA patent/CN117727387A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118643334A (en) * | 2024-06-19 | 2024-09-13 | 北京科技大学 | Method and device for optimizing low alloy steel performance based on differential evolution and machine learning algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pan et al. | Just-in-time learning based soft sensor with variable selection and weighting optimized by evolutionary optimization for quality prediction of nonlinear processes | |
| Li et al. | Fault-relevant optimal ensemble ICA model for non-Gaussian process monitoring | |
| Dai et al. | Energetic, economic and environmental (3E) multi-objective optimization of the back-end separation of ethylene plant based on adaptive surrogate model | |
| CN110428876A (en) | A kind of steel material design method of the machine learning algorithm based on physical instruction | |
| CN116825253B (en) | Method for establishing mechanical properties prediction model of hot-rolled strip based on feature selection | |
| CN113073267A (en) | High-strength high-temperature-resistant RAFM steel and machine learning-based design method thereof | |
| CN117727387A (en) | Method for designing components of high-strength low-alloy steel | |
| CN108268979B (en) | Medium plate quality prediction method based on evolutionary fuzzy association rule | |
| CN116844673A (en) | Reverse design method of high-performance magnesium alloy based on machine learning | |
| CN110046377A (en) | A kind of selective ensemble instant learning soft-measuring modeling method based on isomery similarity | |
| CN107894710A (en) | A kind of principal component analysis modeling method of cracking reaction furnace temperature | |
| CN110989522B (en) | Multi-steel-coil-oriented optimal setting method for technological parameters in continuous annealing production process | |
| Hanai et al. | Analysis of initial conditions for polymerization reaction using fuzzy neural network and genetic algorithm | |
| Kanumuri et al. | A hybrid neural network-genetic algorithm for prediction of mechanical properties of ASS-304 at elevated temperatures | |
| CN111797574B (en) | Integrated Gaussian process regression model method for polymer molecular weight distribution | |
| Dong et al. | Just-in-time learning-based soft sensor for mechanical properties of strip steel via multi-block weighted semisupervised models | |
| Wang et al. | Evolutionary modeling approach based on multiobjective genetic programming for strip quality prediction | |
| CN112786120B (en) | Method for synthesizing chemical material with assistance of neural network | |
| CN119227782B (en) | Continuous annealing band steel performance multi-index prediction method based on multi-objective evolutionary deep learning | |
| CN110442953B (en) | Q & P steel design method based on machine learning under guidance of physical metallurgy | |
| CN113658646A (en) | A prediction method of molecular weight distribution in polymerization process based on ensemble probability modeling | |
| Kulina et al. | Ensemble Learning for Predicting the Tensile Strength of Alloy Steels from Chemical Composition and Processing Parameters | |
| CN111178627A (en) | Neural network hybrid optimization prediction method based on SPCA | |
| Liu et al. | A Multi-Objective Evolutionary Machine Learning Method for the Yield Strength Prediction of Strip Steel in Continuous Annealing | |
| Wu et al. | Machine Learning-based Investigation of the Influencing Factors on the Hardness of Steel Materials |
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 |