CN113198835B - AH 36-grade hot-rolled flat-bulb steel preparation method based on Adam-SVM model - Google Patents

Abstract

The invention discloses an AH 36-grade hot-rolled flat-bulb steel preparation method based on an Adam-SVM model, which comprises the following steps: optimizing the component performance of the hot-rolled flat-bulb steel by utilizing an Adam-SVM algorithm to obtain the alloy components of the AH 36-grade hot-rolled flat-bulb steel; according to the obtained alloy components, converter smelting, external refining and protective casting are adopted to obtain continuous casting billets; after soaking the obtained continuous casting billet, hot rolling the continuous casting billet by a two-roller reversing mill, a three-roller mill and a universal mill in sequence according to the hole shape to obtain flat bulb steel; and after the hot saw is used for saw cutting, the hot saw is put on a cooling bed for air cooling, and then spray cooling is carried out after the cooling is carried out to a certain temperature. The technical scheme of the invention formulates an optimized alloy design scheme based on a machine learning algorithm, reasonably reduces the content of the added elements in the steel, further reduces the cost and shortens the research and development period. Meanwhile, the flat-bulb steel with low cost and high performance is successfully developed through the optimized design of a heating system and a rolling process.

Description

A preparation method of AH36 hot-rolled flat bulb steel based on Adam-SVM model

technical field

The invention belongs to the technical field of flat bulb steel manufacturing, and in particular relates to a method for preparing AH36-grade hot-rolled flat bulb steel based on an Adaptive Moment Estimation (Adam)-Support Vector Machine (SVM) model.

Background technique

Entering the 21st century, the field of shipbuilding and marine engineering equipment in my country has ushered in a new period of rapid growth. The rapid development of the field of shipbuilding and marine engineering equipment has put forward the urgent needs of high strength, high toughness and corrosion resistance for shipbuilding and marine engineering steel. , and also need to meet the needs of large thickness and large size specifications. The average annual demand for ship steel in the international new shipbuilding market is about 80 million tons. In 2019, the domestic demand for ship steel reached about 12 million tons, and the trend is increasing year by year. As a special structural steel for ships, the bulb flat steel is an indispensable special profile for the construction of large ships and various ocean-going, coastal, inland water ships and various ships, and has an important impact on the bearing capacity and safety of the hull structure. Among the hull profiles, bulb flat steel accounts for about 80%, the largest share, and the annual demand exceeds 2 million tons. The development of a series of high-strength and high-toughness bulb steel for ships and the use of existing ship steel to achieve performance matching and meet the needs of new surface ships for high-strength and tough materials is to ensure high-tech ships, ultra-large submarine hulls, and high-tech surface operations An important material basis for the construction of ships, etc.

The related prior art includes a high-strength and low-temperature-resistant flat bulb steel and a production process. The plan discloses the rolling process of the flat bulb steel in detail, and its final structure is acicular ferrite, tempered martensite and a small amount of austenite. Tensite structure, but its properties have not been measured.

The related prior art also includes a large-sized and high-strength marine flat bulb steel and a production process thereof. The flat bulb steel is designed for EH36 according to the laws of physical metallurgy and experience, and a flat bulb steel with higher low-temperature impact toughness has been developed. The composition design process is relatively cumbersome, and it needs to be verified by a large number of experiments, which requires relatively high research and development costs and a long research and development cycle.

SUMMARY OF THE INVENTION

In order to overcome the defects of the prior art, the present invention provides a method for preparing AH36-grade hot-rolled flat bulb steel based on Adam-SVM model. The prepared hot-rolled flat bulb steel has larger specifications and lower cost.

According to the first aspect of the present invention, a method for preparing AH36 hot-rolled flat bulb steel based on the Adam-SVM model is provided. The method is operated based on machine learning theory and includes the following steps:

(1) Use Adam-SVM algorithm to optimize the composition and properties of hot-rolled flat bulb steel, and obtain the alloy composition of the AH36 grade hot-rolled flat bulb steel;

(2) according to the alloy composition obtained, adopt converter smelting, out-furnace refining, protective casting to obtain the continuous casting billet with a cross section of 280mm×165mm;

(3) after the continuous casting billet obtained is put into the heating furnace for heating, it passes through the 850BD two-roll reversing mill, the 800TRM three-roll mill, and the 950SC universal rolling mill in turn according to the pass shape, and obtains the spherical flat steel with a specification of HP370×13;

(4) After sawing with hot saw, it is cooled by air cooling on the cooling bed, and cooled to 200 ℃ by spray cooling.

Specifically, in the step (1), the step of using the Adam-SVM algorithm component performance optimization includes:

(1.1) Define the optimization category: define the category of bulb flat steel according to the impact energy. Category A: 70~100J; Category B: 100~130J; Category C: 130~160J; Category D: 160~190J; Category E: Above 190J;

(1.2) Data processing: standard score normalization processing is performed on the original data, wherein the original experimental data includes C, Si, Mn, P, S, Al, Nb, N, impact energy;

(1.3) Optimal design of impact performance: take the components (C, Si, Mn, P, S, Al, Nb, N) as the input and the impact energy category as the output, use the Adam-SVM algorithm model to predict the impact energy, select The training set and the test set, with impact performance class E as the prediction target, reversely optimize the composition design of bulb flat steel.

Specifically, the method for preparing the AH36-grade hot-rolled flat bulb steel based on the Adam-SVM model described in step (1.2) adopts the following formula:

Among them, μ is the mean of all sample data; σ is the standard deviation of all sample data, and x is the component performance data set;

Specifically, in the method for preparing the AH36 hot-rolled flat bulb steel based on the Adam-SVM model described in step (1.3), the steps of reversely optimizing the composition of impact properties include:

(1.3.1): Import sample data (components, performance data), initialize the maximum mean square error and cross-validation folds, and randomly select validation test samples and training samples;

(1.3.2): Initialize the maximum number of iterations, acceleration factor, population size SVM cross-validation parameters, etc.;

(1.3.3): Use the Adam algorithm to optimize the SVM parameters, train successively, and establish the Adam-SVM model for the reverse optimization component of performance;

(1.3.4): Use the test sample to verify the feasibility of the model, obtain the sample output value, calculate the error, and judge whether the termination condition is satisfied, if satisfied: output the result; if not: return to step 5.1 to re-optimize the training.

Specifically, the method for preparing AH36 hot-rolled flat bulb steel based on Adam-SVM model described in step (1.3.3) adopts Adam to optimize the SVM algorithm and the steps are as follows:

(1.3.3.1) Determine m _t , v _t update rules

m _t =β ₁ m _t-1 +(1-β ₁ )g _t

Among them, m _t is the estimated value of the first moment (mean) of the gradient; v _t is the estimated value of the second moment of the gradient (without central variance); β ₁ is the first-order moment decay coefficient; β ₂ is the second-order moment decay coefficient; g _t is the gradient obtained from the derivation of the objective function.

(1.3.3.2) Perform bias correction on the estimated values of the first and second moments

为m_t的偏置矫正；

为v_t的偏置矫正；in,

is the offset correction of m _t ;

is the offset correction of v _t ;

(1.3.3.3) Determine Adam update rules

where θ solves (updates) the parameters.

Specifically, the AH36 grade hot-rolled flat bulb steel is composed of the following components by mass percentage: C: 0.14-0.16%, Si: 0.25-0.40%, Mn: 1.40-1.55%, P: ≤ 0.01%, S: ≤ 0.004 %, Al: 0.02-0.04%, Nb: 0.015-0.020%, N: 0.005-0.009%, the balance of Fe and impurity elements.

Wherein, in the step (3), the recrystallization of grains is realized in the hot rolling process, wherein the grain size of the original billet of the flat bulb steel at room temperature is ≤70 μm, the average grain size of the finished ball head position of the flat bulb steel is ≤32 μm, The average grain size at the web position of the finished flat steel is less than or equal to 21 μm.

Specifically, in the step (3), the heating furnace adopts a three-stage heating system, which are respectively a preheating section, a heating section, and a soaking section. 1200℃, the temperature of soaking section is 1200～1250℃, and the rolling temperature is 1140～1160℃.

Specifically, in the step (3), 10 passes (K10-K1 passes) are required for rolling, K10-K6 passes are rolled with an 850BD two-roller reversing mill, and the K10 inlet temperature is 1140-1160°C, The outlet temperature of K6 is 1030～1050℃, K5～K3 pass is rolled by 800TRM three-roll mill, and K2～K1 pass is rolled by 950SC universal rolling mill. The starting temperature of K10 is 1150℃, and the final rolling temperature of K1 is 925℃.

According to the second aspect of the present invention, there is provided an AH36-grade hot-rolled flat bulb steel, wherein the AH36-grade hot-rolled flat bulb steel is prepared by the method according to any one of the above aspects, and the AH36-grade hot-rolled flat bulb steel is Steel requires yield strength ≥380MPa, tensile strength ≥550MPa, elongation after fracture ≥27%, and longitudinal V-notch impact absorption energy at 0°C ≥190J.

Beneficial effects: Compared with the prior art, the method for preparing AH36 hot-rolled flat bulb steel based on the Adam-SVM model has the following advantages:

1) The SVM model is a small sample learning method with a solid theoretical foundation, which realizes efficient "transduction reasoning" from training samples to forecast samples, and greatly simplifies the usual problems of classification and regression. At the same time, the computational complexity of the model depends on the number of support vectors, not the dimension of the sample space, which avoids the "dimension disaster" in a sense, and has better performance in the calculation of AH36 grade hot-rolled flat bulbs. "robustness".

2) The Adam algorithm is used to optimize the SVM model, and the accuracy of the algorithm classification and fitting and the accuracy of the bulb flat steel composition design are improved through the optimization of the SVM parameter selection process;

3) The Adam-SVM composite model is used to optimize the composition of the bulb flat steel, which realizes the organic coupling of different chemical compositions and properties, and reduces the material cost of the bulb flat steel while improving the mechanical properties of the material;

4) The AH36 hot-rolled flat bulb steel designed by the Adam-SVM composite model has good performance, its yield strength ≥ 380MPa, tensile strength ≥ 550MPa, elongation after fracture ≥ 27%, 0 ℃ longitudinal V-notch impact Absorption work≥190J.

Description of drawings

Fig. 1 is the flow chart of the preparation method of AH36 grade hot-rolled flat bulb steel of the present invention;

2 is a flow chart of the rolling process of the AH36-grade hot-rolled flat bulb steel of the present invention.

Detailed ways

Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. Where the following description refers to the drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the illustrative examples below are not intended to represent all implementations consistent with this disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as recited in the appended claims.

The terms "first", "second", etc. in the description and claims of the present disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances such that the embodiments of the disclosure described herein can, for example, be practiced in sequences other than those illustrated or described herein.

Furthermore, the terms "comprising" and "having" and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those expressly listed Rather, those steps or units may include other steps or units not expressly listed or inherent to these processes, methods, products or devices.

Multiple, including two or more.

And/or, it should be understood that the term "and/or" used in the present disclosure is merely an association relationship for describing associated objects, indicating that three kinds of relationships may exist. For example, A and/or B can mean that A exists alone, A and B exist at the same time, and B exists alone.

The invention designs and optimizes the bulb flat steel alloy system that meets the requirements of the relevant performance indicators through the Adam-SVM composition-performance classification prediction model, and obtains that the microstructure at room temperature is composed of pearlite and ferrite through appropriate processing technology, and has a good structure. High-strength and tough bulb flat steel with matching properties.

The following examples are all produced by the method of the present invention, as shown in Figure 1 and Figure 2, that is, the SVM model is used to optimize the component properties, and the production is carried out in strict accordance with the process specification of the present invention. The optimized AH36 hot-rolled flat bulb is composed of the following components by mass: C: 0.14-0.16%, Si: 0.30-0.40%, Mn: 1.20-1.30%, P: ≤ 0.02%, S: ≤ 0.004 %, Al: 0.02-0.03%, Nb: 0.012-0.015%, N: 0.004-0.007%, the balance of Fe and impurity elements.

Example 1: After using the Adam-SVM model to optimize the element composition of AH36 hot-rolled flat bulb steel, smelting and casting according to the above composition range to obtain a billet with a cross-section of 280mm × 165mm, and then detect the composition of the billet, as shown in the table 1.

Table 1 Composition of steel billets (wt.%)

Rolling was carried out according to the order of 850BD/800TRM/950SCI/950SCⅡ rolling mill, and the rolling passes were 5/3/1/1 respectively. After rolling, HP370×13 specification bulb flat steel was obtained.

Table 2 Mechanical properties

Example 2: After using the Adam-SVM model to optimize the element composition of AH36 hot-rolled flat bulb steel, smelting and casting according to the above composition range to obtain a billet with a cross-section of 280mm × 165mm, and then detect the composition of the billet, as shown in the table 3.

Table 3 Composition of steel billets (wt.%)

Rolling was carried out in the order of 850BD/800TRM/950SCI/950SCⅡ rolling mills, and the rolling passes were 5/3/1/1 respectively. After rolling, HP370×13 specification bulb flat steel was obtained.

Table 4 Mechanical properties

Example 3: After using the Adam-SVM model to optimize the element composition of AH36 hot-rolled flat bulb steel, smelting and casting according to the above composition range to obtain a billet with a cross-section of 280mm × 165mm, and then detect the composition of the billet, as shown in the table 5.

Table 5 Composition of steel billets (wt.%)

Rolling was carried out in the order of 850BD/800TRM/950SCI/950SCⅡ rolling mill, and the rolling passes were 5/3/1/1 respectively. After rolling, HP370×13 specification bulb flat steel was obtained.

Table 6 Mechanical properties

In order to further highlight the effect of the present invention, the following two groups of comparisons are provided. When the Adam-SVM algorithm is not used to classify and optimize the composition properties, and the process optimization is not performed (the rolling temperature is above 1200°C), the composition properties are as shown in the following table. Show:

Contrast 1:

Table a Composition of billet (wt.%)

Table b Mechanical properties

Contrast 2:

Table c. Composition of billets (wt.%)

Table d Mechanical Properties

Through comparison 1 and 2, it can be seen that before the optimization of the composition and rolling process, its performance is relatively poor. Excellent mechanical properties.

The embodiments of the present invention have been described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific embodiments, which are merely illustrative rather than restrictive. Under the inspiration of the present invention, without departing from the scope of protection of the present invention and the claims, many forms can be made, which all belong to the protection of the present invention.

Claims (9)

1. A preparation method of AH 36-grade hot-rolled flat bulb steel based on an Adam-SVM model is characterized by comprising the following steps:

(1) optimizing the component performance of the hot-rolled flat-bulb steel by using an adaptive moment estimation (Adam) -Support Vector Machine (SVM) algorithm to obtain the alloy components of the AH 36-grade hot-rolled flat-bulb steel;

(2) according to the obtained alloy components, converter smelting, external refining and protective casting are adopted to obtain continuous casting billets;

(3) Putting the obtained continuous casting billet into a heating furnace for heating, and then carrying out hot rolling on the continuous casting billet by a two-roller reversing mill, a three-roller mill and a universal mill in sequence according to the hole shape to obtain flat-bulb steel;

(4) after the hot saw is sawed, the material is cooled in air by a cooling bed, and then is cooled to a certain temperature and then is sprayed and cooled,

in the step (1), the step of optimizing the composition performance of the hot-rolled flat-bulb steel by using the Adam-SVM algorithm comprises the following steps:

(1.1) defining an optimization category: and (3) defining the class of the flat-bulb steel according to the impact energy, wherein the class A is as follows: 70-100J; b type: 100 to 130J; class C: 130-160J; and D type: 160-190J; and E type: 190J or more;

(1.2) data processing: performing standard fraction normalization processing on original experimental data, wherein the original experimental data comprise C, Si, Mn, P, S, Al, Nb, N and impact energy;

(1.3) impact performance optimization design: the components C, Si, Mn, P, S, Al, Nb and N are used as input, the impact energy category is used as output, an Adam-SVM algorithm model is used for predicting the impact energy, a training set and a testing set are selected, the impact performance E category is used as a performance design target, and the design of the components of the flat-bulb steel is reversely optimized.

2. The method for preparing AH36 grade hot-rolled flat bulb steel based on the Adam-SVM model according to claim 1, wherein in step (1.2), the standard fraction normalization processing is performed on the original experimental data by using the following formula:

Where μ is the mean of all sample data, σ is the standard deviation of all sample data, and x is the constituent performance data set.

3. The method for preparing AH36 hot-rolled flat bulb steel based on the Adam-SVM model is characterized in that in the step (1.3), with the impact performance class E as a performance design target, the step of reversely optimizing the design of the composition of the flat bulb steel comprises the following steps:

(1.3.1) importing sample data: the method comprises the following steps of (1) initializing a maximum mean square error value and a fold number of cross validation according to component and performance data, and randomly selecting a validation test sample and a training sample;

(1.3.2) initializing maximum iteration times, acceleration factors, population scale and SVM cross validation parameters;

(1.3.3) optimizing SVM parameters by using an Adam algorithm, performing successive training, and establishing an Adam-SVM model of a performance reverse optimization component;

(1.3.4) verifying the feasibility of the model by adopting the test sample to obtain a sample output value, calculating an error, and judging whether a termination condition is met or not, wherein the conditions are as follows: outputting a result; does not satisfy: and returning to the step (1.3.1) to re-optimize the training.

4. The method for preparing AH 36-grade hot-rolled flat bulb steel based on the Adam-SVM model according to claim 3, wherein in the step (1.3.3), the step of optimizing SVM parameters by using the Adam algorithm is as follows:

(1.3.3.1) determining m _t 、v _t Update the rule, wherein m _t Is an estimate of a first moment of the gradient; v. of _t Is a gradient second moment estimate;

(1.3.3.2) bias correcting the first and second moment estimates;

(1.3.3.3) determining an Adam update rule.

5. The method for preparing the AH36 grade hot-rolled flat bulb steel based on the Adam-SVM model according to any one of claims 1-4, characterized in that the mass percentages of alloy components in the steel are C: 0.14 to 0.16%, Si: 0.25 to 0.40%, Mn: 1.40-1.55%, P: less than or equal to 0.01 percent, S: less than or equal to 0.004%, Al: 0.02 to 0.04%, Nb: 0.015-0.020%, N: 0.005-0.009%, and the balance Fe and impurity elements.

6. The method for preparing the AH36 hot-rolled flat-bulb steel based on the Adam-SVM model as claimed in claim 5, characterized in that in step (3), the hot rolling process realizes the recrystallization of crystal grains, wherein the normal-temperature crystal grain size of the original square billet of the flat-bulb steel is less than or equal to 70 μm, the average crystal grain size of the bulb position of the finished flat-bulb steel is less than or equal to 32 μm, and the average crystal grain size of the web position of the finished flat-bulb steel is less than or equal to 21 μm.

7. The method for preparing the AH36 hot-rolled flat bulb steel based on the Adam-SVM model is characterized in that in the step (3), the heating furnace adopts a three-stage heating system which comprises a preheating section, a heating section and a soaking section, wherein the temperature of the preheating section is 1000-1100 ℃, the temperature of the heating section is 1100-1200 ℃, the temperature of the soaking section is 1200-1250 ℃, and the start rolling temperature is 1140-1160 ℃.

8. The method for preparing AH 36-grade hot-rolled flat bulb steel based on the Adam-SVM model according to claim 5, wherein in the step (3), 10 passes are adopted: rolling in K10-K1 passes, rolling in K10-K6 passes by using an 850BD two-roller reversible rolling mill, rolling in K10 at 1140-1160 ℃, rolling in K6 at 1030-1050 ℃, rolling in K5-K3 passes by using an 800TRM three-roller rolling mill, and rolling in K2-K1 passes by using a 950SC universal rolling mill; the initial rolling temperature of K10 is 1150 ℃, and the final rolling temperature of K1 is 925 ℃.

9. An AH36 grade hot-rolled flat bulb steel, characterized in that the AH36 grade hot-rolled flat bulb steel is prepared by the method according to any one of claims 1 to 8, and the AH36 grade hot-rolled flat bulb steel has yield strength of not less than 380MPa, tensile strength of not less than 550MPa, elongation after fracture of not less than 27%, and longitudinal V-notch impact absorption power at 0 ℃ of not less than 190J.

Images