JP7146127B1 - Manufacturing method of high-strength steel sheet - Google Patents

Abstract

A method for producing a high-strength steel sheet having high levels of tensile strength, elongation and hole expansion ratio without containing a large amount of manganese and without undergoing a complicated production process. A step of soaking a rolled material having a predetermined composition at a predetermined soaking temperature, cooling the rolled material at a predetermined average cooling rate to a temperature of 451° C. or more and 560° C. A first heat treatment step of holding for a holding time of 44.3 seconds or more, an intermediate cooling step of cooling the rolled material to a predetermined intermediate cooling temperature at a predetermined average cooling rate, and a predetermined average heating rate of the rolled material. A second heat treatment step of heating to a temperature of 350 ° C. or more and 450 ° C. or less, holding for a holding time of 100.0 seconds or more and 990.0 seconds or less, and then cooling to room temperature at a predetermined average cooling rate. A method for manufacturing a high-strength steel plate. [Selection drawing] Fig. 1

Description

TECHNICAL FIELD The present invention relates to a method for producing a high-strength steel sheet, and more particularly to a method for producing a high-strength steel sheet that can be used in various applications including automobile parts, a method for generating a prediction model, and a method for designing a steel sheet.

High-strength steel sheets with a tensile strength (TS) of 780 MPa class or higher are being applied to frame parts and seat parts of automobiles due to the growing need to reduce the weight of automobile bodies. However, when high-strength steel sheets are applied to automobile parts, cracks are likely to occur during processing due to reduced ductility and reduced formability. In addition, ductility such as elongation is required in consideration of various uses.

The steel sheets described in Patent Literature 1 and Patent Literature 2 are known as steel sheets capable of achieving such high levels of strength, formability and ductility. Patent Literature 1 examines the balance between strength and hole expansibility on the premise of higher ductility. Moreover, in Patent Document 2, the strength, ductility and stretch flangeability are improved by optimizing the conditions of the continuous annealing process.
Further, Patent Document 3 discloses a technique for constructing a machine learning model for predicting the mechanical properties (strength) of low alloy steel. This machine learning model is a combination of a deep neural network and a genetic algorithm. Using this model, the chemical composition and heat treatment process of new low-alloy steel are designed, and the designed alloy is manufactured and tested. .

Japanese Patent No. 6439903 International publication WO2012-36269

Z. Zhu et al. , Materials 2020, 13(23), 5316

In recent years, due to the further increase in the need for weight reduction as described above, in many applications including automobile parts, it is necessary to achieve three higher levels of high tensile strength, high elongation and high hole expansion ratio. is required.
Specifically, the tensile strength (TS), elongation at break (EL) and hole expansion ratio (λ) are required to satisfy the following.
The tensile strength is required to be 950 MPa or more so that sufficient strength of the parts can be obtained even if the weight is reduced.
In order to ensure moldability during part molding, it is also required that the hole expansion rate, which indicates hole expandability, is 20% or more.
In addition, the elongation at break is required to be 22% or more.

Furthermore, from the viewpoint of reducing manufacturing costs and simplifying process management, it is required to reduce the amount of additive components as much as possible. It is required to reduce the content of manganese, which is a rare metal. Specifically, for example, the amount of Mn is required to be 2.3% by mass or less. In addition, from the viewpoint of manufacturing cost reduction, there is also a demand not to require complicated production processes in manufacturing.

In the production methods described in Patent Documents 1 and 2, there is provided a method for producing a steel material that satisfies the above-mentioned tensile strength, elongation at break and hole expansion ratio and has a reduced Mn content without complicated production processes. was difficult to provide.

The development of a steel manufacturing method that satisfies various properties including strength such as tensile strength and ductility such as elongation at break and hole expansion ratio usually requires a long development period. The social demand for reduction is rapidly increasing, and there is a demand for further acceleration and efficiency of development than ever before. As one of the methods for accelerating and improving the efficiency of development, a method based on data science, which is also used in Non-Patent Document 1, has been attracting attention in recent years, and is being applied to steel materials as well.
However, in the conventional method using data science disclosed by Non-Patent Document 1, etc., steel materials that satisfy tensile strength, elongation at break and hole expansion ratio and have a reduced Mn content are produced through complicated production processes. It has been difficult to provide a method of manufacturing without

The present invention has been made in view of such circumstances, and more specifically, to efficiently obtain practical and rational design conditions or manufacturing methods for high-strength steel sheets that achieve target conditions from a data science approach. Specifically, from a data science approach, it is possible to reduce alloying, especially without containing a large amount of manganese and without undergoing a complicated production process, tensile strength (TS), elongation at break (EL) and hole expansion ratio ( It is an object of the present invention to find and provide a method for producing a high-strength steel sheet, a method for generating a predictive model, and a method for designing a steel sheet, all of which have a high level of λ).

Aspect 1 of the present invention is
C: 0.10% by mass or more and 0.40% by mass or less Si: 0.50% by mass or more and 1.80% by mass or less,
Mn: 1.60% by mass or more and 2.20% by mass or less,
Al: 0.01% by mass or more and 0.90% by mass or less,
Cr: 0.30% by mass or more and 0.80% by mass or less,
and the balance being iron and inevitable impurities, a soaking step of soaking the rolled material at a soaking temperature T1 that satisfies the following formula (1);
The rolled material is cooled to a temperature T _preAT of 451° C. or more and 560° C. or less at a first average cooling rate C1 of 2° C./s or more and 60° C./s or less, and the temperature T _preAT is 4.0 seconds or more and 44.3 seconds. a first heat treatment step held for a holding time t _preAT of:
an intermediate cooling step of cooling the rolled material to an intermediate cooling temperature T2 that satisfies the following formula (3) at a second average cooling rate C2 of 10° C./s or more and 60° C./s or less;
The rolled material is heated to a temperature T _postAT of 350° C. or more and 450° C. or less at a first average heating rate H1 of 5° C./sec or more and 60° C./sec or less, and the temperature T _postAT is 100.0 seconds or more and 990.0 seconds. A second heat treatment step of cooling to room temperature at a third average cooling rate C3 of 10° C./s or more and 60° C./s or less after holding for the following holding time t _postAT ;
A method for producing a high-strength steel sheet having a tensile strength of 950 MPa or more, a breaking elongation of 22% or more, and a hole expansion ratio of 20% or more.

-30.0°C ≤ T1-Ae3 ≤ 40.0°C (1)
Here, Ae3 is based on the following formula (2).

Ae3 (°C) = 896.5414 - 236.22 x [C] + 26.18998 x [Si] - 28.4882 x [Mn] + 106.9042 x [Al] + 1.284594 x [Ti] + 12.66673 x [ Nb]+42.66673×[V]−39.9466×[Ni]+2.467166×[Mo]−17.8592×[Cr] (2)
Here, [ ] is the content of each element in mass %.

Ms-215≤T2≤Ms-118 (3)
Here, Ms is based on the following formula (4).

Ms (°C) = 561 - 474 x [C] - 33 x [Mn] (4)
Here, [ ] is the content of each element in mass %.

Aspect 2 of the present invention is the method for producing a high-strength steel sheet according to aspect 1, wherein the temperature T _preAT is 500° C. or higher and 560° C. or lower.

Aspect 3 of the present invention is the method for producing a high-strength steel sheet according to Aspect 1 or 2, wherein the holding time t _preAT at the temperature T _preAT in the first heat treatment step is 4.0 seconds or more and 35.0 seconds or less. is.

In a fourth aspect of the present invention, the rolled material contains Cu: 0.50% by mass or less (excluding 0% by mass), Ni: 0.50% by mass or less (excluding 0% by mass), Mo: 0.50% by mass or less (excluding 0% by mass). 50% by mass or less (excluding 0% by mass), B: 0.01% by mass or less (excluding 0% by mass), V: 0.05% by mass or less (excluding 0% by mass), Nb: 0 .05% by mass or less (excluding 0% by mass), Ti: 0.05% by mass or less (excluding 0% by mass), Ca: 0.05% by mass or less (excluding 0% by mass), REM: A method for producing a high-strength steel sheet according to any one of aspects 1 to 3, further containing one or more selected from the group consisting of 0.01% by mass or less (not including 0% by mass).

Aspect 5 of the present invention is a method for generating a prediction model,
A manufacturing parameter set prepared based on a plurality of types of manufacturing parameters for manufacturing a steel sheet corresponds to test results relating to at least one predetermined characteristic of a steel sheet actually manufactured using the manufacturing parameter set. a step of acquiring the attached training data set;
a step of generating a prediction model with the manufacturing parameter set as an explanatory variable and the test result as an objective variable by machine learning using the learning data set;
The manufacturing parameter set is prepared based on the following manufacturing parameters,
The test results relate to at least one property selected from tensile strength, breaking elongation, and hole expansion ratio of the steel sheet,
How to generate the predictive model.
The rolled material used to manufacture the steel plate is
C: 0.10% by mass or more and 0.40% by mass or less Si: 0.50% by mass or more and 1.80% by mass or less,
Mn: 1.60% by mass or more and 2.20% by mass or less,
Al: 0.01% by mass or more and 0.90% by mass or less,
Cr: 0.30% by mass or more and 0.80% by mass or less,
with the balance consisting of iron and unavoidable impurities.
A soaking heat treatment step, a first heat treatment step, an intermediate cooling step, and a second heat treatment step are sequentially performed on the rolled material.
The soaking step is performed at a soaking temperature T1 that satisfies the following formula (1).
In the first heat treatment step, the rolled material is cooled to a temperature T _preAT of 451° C. or higher and 560° C. or lower at a first average cooling rate C1 of 2° C./second or higher and 60° C./second or lower, and 4. at the temperature T _preAT . It must be held for a holding time t _preAT of 0 seconds or more and 44.3 seconds or less.
The intermediate cooling step is performed by cooling the rolled material to an intermediate cooling temperature T2 that satisfies the following formula (3) at a second average cooling rate C2 of 10° C./second or more and 60° C./second or less.
In the second heat treatment step, the rolled material is cooled to a temperature T _postAT of 350° C. or more and 450° C. or less at a first average heating rate H1 of 5° C./sec or more and 60° C./sec or less, and the temperature T _postAT is 100. After holding for a holding time t _postAT of 0 to 990.0 seconds, cooling to room temperature at a third average cooling rate C3 of 10° C./s to 60° C./s.
-30.0°C ≤ T1-Ae3 ≤ 40.0°C (1)
Here, Ae3 is based on the following formula (2).
Ae3 (°C) = 896.5414 - 236.22 x [C] + 26.18998 x [Si] - 28.4882 x [Mn] + 106.9042 x [Al] + 1.284594 x [Ti] + 12.66673 x [ Nb]+42.66673×[V]−39.9466×[Ni]+2.467166×[Mo]−17.8592×[Cr] (2)
Here, [ ] is the content of each element in mass %.
Ms-215≤T2≤Ms-118 (3)
Here, Ms is based on the following formula (4).
Ms (°C) = 561 - 474 x [C] - 33 x [Mn] (4)
Here, [ ] is the content of each element in mass %.

A sixth aspect of the present invention is a steel plate design method using a prediction model,
(a) acquiring a manufacturing parameter set reflecting multiple types of manufacturing parameters;
(b) calculating a predicted value using the prediction model generated by the generation method according to aspect 5 and the manufacturing parameter set;
(c) varying a predetermined manufacturing parameter in the current manufacturing parameter set to generate a new manufacturing parameter set in which manufacturing parameters other than said predetermined manufacturing parameter are maintained;
(d) calculating new predicted values using the predictive model and the new set of manufacturing parameters;
(e) determining the range of the predetermined manufacturing parameter using the plurality of predicted values calculated in the steps (b) and (d).

Aspect 7 of the present invention is the design method of Aspect 6, further comprising the step of (f) repeating steps (c) and (d) until the predetermined manufacturing parameter varies by a predetermined range. do.

Aspect 8 of the present invention is the design method according to aspect 6 or 7, wherein the plurality of predicted values are at least one type selected from tensile strength, breaking elongation, and hole expansion ratio of the steel plate is a value for the property of

Aspect 9 of the present invention is the design method according to aspect 8, wherein the target requirements for the tensile strength, breaking elongation, and hole expansion ratio of the steel plate are 950 MPa or more, 22% or more, and 20% or more, respectively. and the prediction model has performance that satisfies the following conditions.

According to the present invention, it is possible to efficiently obtain practical and rational thin plate design conditions or manufacturing methods that achieve target conditions from a data science approach. More specifically, from a data science approach, tensile strength (TS), elongation at break (EL) and hole expansion ratio (λ ) can find and provide a method for manufacturing a high-strength steel sheet, a method for generating a predictive model, and a method for designing a steel sheet, all of which are at a high level.

FIG. 2 is a schematic diagram showing a heat treatment pattern (heat treatment history) of a rolled material in the method for manufacturing a high-strength steel sheet according to the present invention; It is a flow chart which shows the procedure of learning and prediction concerning the present invention. 1 is a hardware block diagram showing the configuration of a computer system according to an exemplary embodiment; FIG. It is a flowchart which shows the procedure of a prediction model acquisition process. FIG. 10 is a flowchart showing a detailed procedure of processing for generating a learning data set; FIG. FIG. 4 is a diagram showing an example of a table in which a certain manufacturing parameter set and test results regarding tensile strength (TS) of steel sheets manufactured under the manufacturing parameter set are associated with each other; FIG. 4 is a diagram showing an example of a training data set _Tk ; FIG. 4 is a diagram showing an example of constructed n sets of learning data sets T _i (i=1, . . . , n). FIG. 10 is a diagram showing a kernel function when the variable _xi in Equation 4 is the variable x; FIG. 10 is a schematic diagram showing the relationship between the profile of the prediction model f(z) and the normal distribution represented by each term on the right side of Equation 6 that constitutes the prediction model f(z). 4 is a flowchart showing the procedure of model learning processing using a learning data set; 4 is a flow chart showing a procedure of prediction processing based on an unknown manufacturing parameter set z; FIG. 10 is a flow chart showing a procedure of manufacturing parameter determination processing in consideration of a plurality of predicted performance values; FIG. FIG. 4 is a diagram showing an example of a known manufacturing parameter set used for prediction and the range of each manufacturing parameter; It is the cross-validation figure which compared the predicted value and actual value of tensile strength (TS). It is the cross-validation figure which compared the predicted value and actual value of breaking elongation (EL). It is the cross-validation figure which compared the predicted value and actual value of a hole expansion rate ((lambda)). FIG. 10 is a diagram showing results showing composition examples that satisfy the target characteristics and composition examples that do not satisfy the target characteristics extracted based on the predictive model; FIG. 4 is a diagram showing examples of manufacturing parameter sets and ranges of each manufacturing parameter that are adjusted to satisfy requirements for high-strength steel sheets.

The inventors of the present application used a rolled material with a predetermined composition in which the manganese (Mn) content was suppressed, and as a result of earnestly studying a heat treatment pattern that can be performed using ordinary continuous annealing equipment, the details will be described later. , soaking process, first heat treatment process, intermediate cooling process and second heat treatment process, tensile strength (TS), elongation at break (EL) and hole expansion ratio (λ) by optimizing the conditions for each of the four processes. have reached the present invention capable of producing a high-strength steel sheet at a high level.

In particular, among these four processes, the first heat treatment process, which greatly contributes to improving the balance between elongation at break and hole expansion rate, was tested on various steel materials for strength, ductility, and hole expansion rate, and the results Based on the regression calculation (Gaussian process) based on this, the relationship between the production parameters and these characteristics was found, and appropriate production conditions were derived therefrom. Steel sheets manufactured according to the manufacturing conditions have at least tensile strength (TS), elongation at break (EL) and hole expansion ratio (λ) meeting the aforementioned requirements:
TS≧950MPa, EL≧22%, λ≧20%
can be estimated to satisfy
The details of the present invention are described below.

1. Chemical Composition The high-strength steel sheet obtained by the method for producing a high-strength steel sheet according to the present invention has the following chemical composition.
(1) Carbon (C): 0.10% by mass or more and 0.40% by mass or less C is an element necessary for enhancing hardenability and ensuring high strength, and effectively exerts such an action. Therefore, it is necessary to contain 0.10% by mass or more. However, if the amount of C is too large, coarse MA (Martensite-Austenite constituent) is formed and the hole expansion rate is lowered. It is preferably 0.14% by mass or more and 0.34% by mass or less.

(2) Silicon (Si): 0.50% by Mass or More and 1.80% by Mass or Less Si is an element effective in improving temper softening resistance. In addition, it is an element that is also effective in improving strength through solid-solution strengthening. From the viewpoint of exhibiting these effects, 0.50% by mass or more of Si is contained. The content is preferably 0.65% by mass or more, more preferably 0.70% or more. However, since Si is a ferrite-forming element, if it is included in a large amount, the hardenability is impaired, making it difficult to ensure high strength. Therefore, the amount of Si is set to 1.80% by mass or less. The content is preferably 1.60% by mass or less, more preferably 1.40% by mass or less, even more preferably 1.20% by mass or less, and even more preferably 1.00% by mass or less.

(3) Manganese (Mn): 1.60% by Mass or More and 2.20% by Mass or Less Mn suppresses the formation of ferrite. In order to exhibit such action effectively, it is necessary to contain 1.60% by mass or more. However, if it exceeds 2.20% by mass, the bainite transformation is suppressed, so the retained γ cannot be formed, and the hole expansibility cannot be improved. It is preferably 1.70% by mass or more, more preferably 1.80% by mass or more. Moreover, it is preferably 2.10% by mass or less.
By setting the Mn content to 1.60% by mass or more and 2.20% by mass or less in this way, the requirement that a large amount of Mn content is not required (e.g., Mn: 2.3% by mass or less) is satisfied. be able to.

(4) Aluminum (Al): 0.01% by Mass to 0.90% by Mass Al acts as a deoxidizing agent and also has the effect of improving the corrosion resistance of steel. In order to fully exhibit these effects, it is necessary to contain 0.01% by mass or more. The content of Al is preferably 0.04% by mass or more, more preferably 0.06% by mass or more. However, if Al is contained excessively, a large amount of inclusions are formed and cause surface defects, so the upper limit is made 0.90% by mass. The Al content is preferably 0.60% by mass or less, more preferably 0.50% by mass or less.

(5) Chromium (Cr): 0.30% by Mass or More and 0.80% by Mass or Less Cr is useful as an element that contributes to improving the strength of steel. In order to fully exhibit such an effect, it is necessary to contain 0.30% by mass or more. On the other hand, when the content exceeds 0.80% by mass, the excessive increase in strength may promote embrittlement. Moreover, it may become economically disadvantageous. The Cr content is preferably 0.40% by mass or more. Also, the Cr content is preferably 0.65% by mass or less.

(6) Balance The basic ingredients are as described above, and in one preferred embodiment the balance is iron and inevitable impurities. Unavoidable impurities include trace elements (for example, As, Sb, Sn, N, O, etc.) brought in depending on the conditions of raw materials, materials, manufacturing facilities, and the like.
As for phosphorus (P) and sulfur (S), P: 0.05% by mass or less and S: 0.01% by mass or less are guidelines, respectively. As for nitrogen (N) and oxygen (O), N: 10 to 60 ppm and O: 10 ppm or less are guidelines, respectively.

(7) Selective Additive Elements In another preferred embodiment of the present invention, elements other than those mentioned above may be contained as long as the effects of the present invention are not impaired. The high-strength steel sheet according to the present invention, for example, if necessary, copper (Cu): 0.50% by mass or less (not including 0% by mass), nickel (Ni): 0.50% by mass or less (0 mass% %), molybdenum (Mo): 0.50% by mass or less (excluding 0% by mass), boron (B): 0.01% by mass or less (excluding 0% by mass), vanadium (V) : 0.05% by mass or less (excluding 0% by mass), Niobium (Nb): 0.05% by mass or less (excluding 0% by mass), Titanium (Ti): 0.05% by mass or less (0% by mass %), calcium (Ca): 0.05% by mass or less (excluding 0% by mass), rare earth element (REM): from the group consisting of 0.01% by mass or less (excluding 0% by mass) It may contain one or more selected types.
In the present specification, "not including 0% by mass" means that the content exceeds the impurity level. In addition, the content of rare earth elements (REM) means the total content of 17 elements including scandium (Sc) and yttrium (Y) plus lanthanoids, and may include only some elements instead of all 17 elements. .

Cu, Ni, Mo, and B improve the hardenability, prevent the formation of ferrite, and contribute to the stabilization of austenite and the refinement of bainite, thereby improving the strength-ductility balance.
V, Nb, and Ti precipitate and strengthen the parent phase, thereby increasing the strength without significantly deteriorating the ductility and improving the strength-ductility balance.
Ca and REM finely disperse inclusions typified by MnS, thereby contributing to strength-ductility balance and improvement of hole expansibility.
In the present invention, the content of the selective elements, which are relatively expensive alloy components, is preferably suppressed to a small amount from the viewpoint of cost.

2. 1. Method for Manufacturing High-Strength Steel Plate FIG. 1 is a schematic diagram showing a heat treatment pattern (heat treatment history) of a rolled material in a method for manufacturing a high-strength steel plate according to the present invention. In the heat treatment pattern shown in FIG. 1, the rolled material having the above chemical composition is subjected to (a) a soaking step of soaking at a soaking temperature T1 that satisfies predetermined conditions, and then (b) 2 ° C./sec. At a first average cooling rate C1 of 60 ° C./sec or less, cooling from the soaking temperature T1 to a temperature T _preAT of 451 ° C. or more and 560 ° C. or less, and at the temperature T _preAT 4.0 seconds or more and 44.3 seconds or less. Then, (c) a second average cooling rate C2 of 10° C./s or more and 60° _C./s or less to satisfy a predetermined condition from the temperature T _preAT An intermediate cooling step of cooling to an intermediate cooling temperature T2 is performed, followed by (d) a temperature T _postAT of 350 ° C. or higher and 450 ° C. or lower from the temperature T2 at a first average heating rate H1 of 5 ° C./sec or more and 60 ° C./sec or less. After heating to temperature T _postAT for a holding time t _postAT of 100.0 seconds or more and 990.0 seconds or less, cooling to room temperature at a third average cooling rate C3 of 10 ° C./s or more and 60 ° C./s or less A second heat treatment step is sequentially included.

Steps (a) to (d) are performed sequentially as described above. One preferred form of such a heat treatment method is continuous using a Continuous Annealing Facility (CAL). That is, it is preferable to perform all of the above steps (a) to (d) during the process of winding the rolled steel sheet (rolled material) unwound from the coil into the coil again. As a result, high productivity can be obtained, and a large amount of steel sheets can be processed efficiently.
However, the present invention is not limited to this, and for example, each step may be continuously performed in batch processing. Batch processing allows efficient processing of small or short strips.
Steps (a) to (d) are detailed below.

(a) Soaking Process A rolled material (rolled steel sheet) having the chemical composition described above is subjected to a soaking process.
The rolled material to be used may be obtained, for example, by subjecting a cast material such as a slab to hot rolling in a conventional manner and further subjecting the hot rolled steel sheet to cold rolling in a conventional manner. Hot rolling may be carried out under arbitrary conditions, such as a method of directly rolling the continuously cast slab without heating it, or a method of subjecting the continuously cast slab to heat treatment for a short time before rolling.
Conditions for cold rolling are not particularly limited, either. For example, the rolling reduction may be adjusted so that the finally obtained high-strength steel sheet has a tensile strength of 950 MPa or more.
In addition, although the case where both hot rolling and cold rolling are performed is illustrated above, it is not limited to this, and either one of hot rolling and cold rolling may be performed to obtain a rolled material.

The obtained rolled material is soaked at a soaking temperature T1 that satisfies the following formula (1).

-30.0°C ≤ T1-Ae3 ≤ 40.0°C (1)
Here, Ae3 is based on the following formula (2).

Ae3 (°C) = 896.5414 - 236.22 x [C] + 26.18998 x [Si] - 28.4882 x [Mn] + 106.9042 x [Al] + 1.284594 x [Ti] + 12.66673 x [ Nb]+42.66673×[V]−39.9466×[Ni]+2.467166×[Mo]−17.8592×[Cr] (2)
Here, [ ] is the content of each element in mass %.

If the soaking temperature T1 is lower than Ae3-30.0° C., the amount of ferrite becomes excessive and sufficient tensile strength cannot be obtained. The Ae3 point is the equilibrium transformation temperature between the α phase and the γ phase, which can be obtained from the equation (2).
If the soaking temperature T1 is higher than Ae3+40.0° C., the amount of ferrite becomes too small, and sufficient elongation at break cannot be ensured. Moreover, the heat treatment temperature is high, which is not preferable from the viewpoint of economy and productivity.
The holding time t _preAT at the soaking temperature T1 may be appropriately selected within a range in which an appropriate amount of ferrite can be obtained. A preferred holding time t _preAT is 10 seconds to 1800 seconds.

(b) First heat treatment step Subsequently, as shown in FIG. 1, the rolled material after the soaking is cooled from the soaking temperature T1 to the temperature T _preAT at an average cooling rate C1 and held for the holding time t _preAT . 1 heat treatment.
The first heat treatment is performed by holding the temperature T _preAT for the time t _preAT , thereby introducing bainitic ferrite to finely divide the austenite, and adding carbon to the austenite remaining around the bainitic ferrite. can be thickened. As a result, the retained austenite produced in the intermediate cooling step and the second heat treatment step, which are subsequent steps, can be made fine and in a stable state with a high carbon concentration, which greatly contributes to improving the strength-breaking elongation balance. In addition, unlike ferrite that is commonly used, bainitic ferrite itself has high strength, so it is possible to reduce the difference in strength with the surrounding hard martensite and retained austenite, resulting in improved elongation at break. It is also effective in improving the hole expansion ratio, which tends to deteriorate with aging, and improving the balance between elongation at break and hole expansion ratio.

As described above, the first heat treatment step is a particularly important step among the steps (a) to (d). Based on the regression calculation (Gaussian process) based on the results, the relationship between the holding temperature T _preAT and the holding time t _preAT and these characteristics was found, and from there, the proper holding temperature T _preAT was 451 ° C. as the proper first heat treatment step. 560° C. or more, and the appropriate holding time t _preAT was found to be 4.0 seconds or more and 44.3 seconds or less.

In the following, the procedure for determining the holding temperature and holding time t _preAT based on the Gaussian process implemented by the inventors will be described.

FIG. 2 is a flow chart showing a procedure for performing learning using a known manufacturing parameter set and using the learning result to predict the result when using a new manufacturing parameter set.

This procedure includes four steps. That is, the data preparation process in step S2, the data processing process in step S4, the model learning process in step S6, and the prediction calculation process in step S8.

These steps can be roughly divided into two processes. The first is the process of obtaining a prediction model f by a Gaussian process in steps S2, S4 and S6. The second is a process of predicting one or more manufacturing parameters that are estimated to be appropriate for the first heat treatment step using the prediction model f in step S8. In this specification, the above-described first processing may be referred to as "prediction model acquisition processing", and the second processing may be referred to as "prediction processing".

In addition, in FIG. 2, "prediction model acquisition processing" and "prediction processing" are shown as a series of procedures, but this description is merely an example for convenience of understanding. The prediction model acquisition process and the prediction process can originally be executed independently. For example, a computer system may execute a prediction model acquisition process to acquire a prediction model, and another computer system may use the prediction model to perform the prediction process.

Next, a computer system that executes the above procedure will be described.
FIG. 3 is a hardware block diagram illustrating the configuration of computer system 200 according to an exemplary embodiment.

The computer system 200 has an arithmetic device 100 , an input device 110 and a monitor 120 . Arithmetic device 100 is connected to each of input device 110 and monitor 120 . However, the input device 110 and the monitor 120 need not always be connected to the computing device 100, and may be connected as needed.

Arithmetic device 100 is, for example, a general-purpose PC. The arithmetic device 100 receives data input from the input device 110 and performs processing described later. The processing result of the arithmetic unit 100 is transmitted to the monitor 120 and displayed on the monitor 120, for example. The configuration of the arithmetic device 100 will be described later.

The input device 110 is a device that receives input from a user and provides it to the arithmetic circuit 10 . An example of input device 110 is a touch panel, mouse and/or keyboard. Alternatively, the input device 110 may be a storage device that stores pre-existing data and outputs it on demand.

The monitor 120 is a display device having a display panel such as a liquid crystal display panel or an organic EL panel. The monitor 120 visually outputs, for example, the result predicted by the image processing device 100 using the prediction model f.

The arithmetic device 100 has an arithmetic circuit 10, a memory 12, a storage device 14, an image processing circuit 16, a bus 18, and interface devices 20a to 20c. The components are communicatively connected to each other by a bus 18 .

Arithmetic circuit 10 may be, for example, an integrated circuit (IC) chip such as a central processing unit (CPU) or a digital signal processor. The arithmetic circuit 10 acquires a prediction model by performing model learning processing, which will be described later, and executes prediction processing using the acquired prediction model.

The arithmetic circuit 10 does not need to perform both the model learning process of acquiring the prediction model and the prediction process using the prediction model. The computing device 100 may perform only model learning processing, or may perform only prediction processing using a prediction model acquired using another computing device.

The memory 12 is a general term for semiconductor volatile memory such as RAM and/or semiconductor non-volatile memory such as flash ROM, which have a plurality of storage elements. At least part of the memory 12 may be a removable recording medium.

The memory 12 stores various learning data sets during model learning processing. After obtaining the prediction model f, the memory 12 stores a parameter group 12m that defines the prediction model f. Further, memory 12 may store a computer program that controls the operation of arithmetic circuit 10 .

The storage device 14 is, for example, a storage device having a larger storage capacity than the memory 12 . The storage device 14 can store various parameters for model learning received from the input device 110 .

The image processing circuit 16 may be an integrated circuit (IC) chip, also called a GPU (Graphics Processing Unit). The image processing circuit 16 generates image data to be displayed on the monitor 120 .

The interface device 20 a receives data from the input device 110 . The interface device 20 b outputs video data to the monitor 120 . The interface device 20c inputs or outputs data such as a learning data set, a prediction model f, and prediction results.

Examples of the interface devices 20a and 20c are USB terminals and Ethernet terminals. An example of the interface device 20b is an HDMI (registered trademark) terminal. Instead of these examples, other terminals may be used as the interface devices 20a-20c. One or more of the interface devices 20a to 20c may be wireless interfaces for wireless communication. As such a wireless interface, a standard conforming to the Wi-Fi (registered trademark) standard is adopted, which performs wireless communication using frequencies such as 2.4 GHz/5.2 GHz/5.3 GHz/5.6 GHz. obtain.

Next, the "prediction model acquisition process" and the "prediction process" shown in FIG. 2 will be described in detail. Both processes are assumed to be executed by the arithmetic circuit 10 .

FIG. 4 is a flowchart showing the procedure of prediction model acquisition processing.
In step S12, the arithmetic circuit 10 executes processing for generating a learning data set. A “learning data set” is a set of various manufacturing parameters for manufacturing a steel plate and test results related to predetermined characteristics of a steel plate actually manufactured using the manufacturing parameter set. Say a combination. Note that the process of generating the learning data set includes preparing data actually used in manufacturing and processing the data into a format suitable for the subsequent model learning process. In step S14, the arithmetic circuit 10 executes model learning processing using the learning data set. More specific contents of steps S12 and S14 described above will be described later in detail with reference to FIGS. Finally, in step S16, the arithmetic circuit 10 acquires the learned prediction model f.

FIG. 5 is a flowchart showing detailed procedures of the learning data set generation process (step S12 in FIG. 4). Of the steps in FIG. 5, steps S22 to S28 are steps executed by a steel plate manufacturer, for example, and do not require the computer system 200 to be used. The computer system 200 can be used after step S30.

The manufacturer determines the steel plate manufacturing parameter set _x0i (step S22), manufactures the steel plate _Si using the manufacturing parameter set _x0i (step S24), and performs various tests on the manufactured steel plate _Si . (Step S26). This obtains the test result _yi . The "various tests" in step S26 are tests for obtaining tensile strength (TS), elongation at break (EL) and hole expansion ratio (λ) in this embodiment.

FIG. 6 shows an example of a table in which a certain production parameter set is associated with test results regarding the tensile strength (TS) of steel sheets produced under the production parameter set. In this example, the numerical values actually used are shown as they are as the manufacturing parameter set. Other test results than tensile strength (TS) may be included, such as elongation at break (EL) and hole expansion ratio (λ).

In addition, the manufacturer usually changes the manufacturing parameters, actually manufactures the steel sheet in the laboratory test equipment or the actual production line, and has carried out the test on the steel sheet. Therefore, the manufacturer may already have accumulated a large amount of associated manufacturing parameter sets and test result data. In such a case, it can be said that steps S22 to S28 have already been performed.

Next, in step S30 of FIG. 5, the arithmetic circuit 10 processes the manufacturing parameter set _x0i to generate the processed manufacturing parameter set _xi . The "processing" referred to here refers to a process of generating new parameters from the numerical values of the manufacturing parameters determined in step S22 so as to strengthen the relationship with the test results. For example, in the present embodiment, the "soaking temperature T1 _{- Ae3} , “Ms-intermediate cooling temperature T2,” “log(retention time t _preAT ),” “log(retention time t _postAT) ” are generated and included in the processed manufacturing parameter set. A manufacturer or the like can predetermine a rule as to which manufacturing parameter of the manufacturing parameter set _x0i is to be processed and how to generate a processed manufacturing parameter. The arithmetic circuit 10 can automatically generate a processed manufacturing parameter set from the manufacturing parameter set according to the rule.

In step S32, the arithmetic circuit 10 associates the processed manufacturing parameter set x _i with the test result y _i to generate and store the learning data set T _i . FIG. 7 shows an example of a training data set _Tk . As is clear from comparing FIG. 6 and FIG. 7, the processed manufacturing parameter set x _k includes “soaking temperature T1−Ae3”, “Ms−intermediate cooling temperature T2”, “log (retention time t _preAT ) ', 'log(retention time t _postAT )'. The present embodiment treats the processed manufacturing parameter set x _i as a row vector.

In step S34, the arithmetic circuit 10 constructs n sets of learning data sets T _i (i=1, . . . , n). The n sets of learning data sets T _i include sets that satisfy requirements such as tensile strength (TS), elongation at break (EL) and / or hole expansion ratio (λ) required for high-strength steel sheets. , unsatisfactory tuples are also included. By performing learning using these, it is possible to obtain a prediction result as to whether the above requirements are satisfied or not satisfied from an unknown manufacturing parameter set after the completion of learning.

FIG. 8 shows an example of constructed n sets of training data sets T _i (i=1, . . . , n). In this embodiment, a set of training data sets is a set of row vector processed manufacturing parameter set x _i and vector y _i . FIG. 8 clearly shows the learning data set T _k corresponding to FIG. 7 as an example. The arithmetic circuit 10 stores the constructed n sets of learning data sets T _i (i=1, . . . , n) in the storage device 14 . Alternatively, the arithmetic circuit 10 may transmit the learning data set Ti from the interface device _20c to another computer system.

The obtained learning data set T _i is used for the model learning process described below. As shown in FIG. 8, in the present embodiment, a processed manufacturing parameter set is used as an explanatory variable X, and data relating to test results as an objective variable y for model learning processing. The explanatory variable X is a matrix and the objective variable y is a column vector. The number of rows of the explanatory variable X represents the number of manufacturing parameter data when the steel plate was manufactured, and the number of columns represents the number of processed manufacturing parameters.

ここで、ｋ（ｚ）^Ｔ及び行列Ｃは以下のように定義される。

ｘ_ｉ（ｉ＝１，・・・，ｎ）は、行列Ｘのｉ行目の要素ベクトルを表し、ｎはデータ数を表す。また、ｋ（ｘ_ｉ，ｘ_ｊ）はカーネル関数を表す。本明細書では、代表的なカーネル関数として以下のガウスカーネルを採用して説明するが、他の周知のカーネル関数を採用してもよい。

As described above, the inventors performed model learning using a Gaussian process and obtained a trained prediction model. It is known that the prediction model f(z) obtained by the Gaussian process finally becomes the following formula. Note that z represents an unknown manufacturing parameter set.

where k(z) ^T and matrix C are defined as follows.

x _i (i=1, . . . , n) represents the i-th row element vector of the matrix X, and n represents the number of data. Also, k(x _i , x _j ) represents a kernel function. In this specification, the following Gaussian kernel is adopted as a representative kernel function for explanation, but other well-known kernel functions may be adopted.

FIG. 9 shows the kernel function when the variable x _i in Equation 4 is the variable x. A Gaussian kernel has the same shape as a normal distribution with mean x _j and variance 1/(θ ₁ ) ^1/2 .
First, regarding the matrix C in Equation 3, since _{x i} (i= ₁ , . . . , n) is known (FIG. 8), substituting Equation 4 yields Matrix C can be calculated. As a result, the inverse matrix C ⁻¹ can also be calculated. Note that the inverse matrix is assumed to exist. The parameters θ ₀ , θ ₁ and σ are so-called hyperparameters and play a role in adjusting the output of the prediction model shown in Equation 1 above. Details of the parameters θ ₀ , θ ₁ and σ will be described later.

Furthermore, since the objective variable y is also known, the "C ⁻¹ ·y" term in Equation 1 can be computed as a constant column vector, including the parameters θ ₀ , θ ₁ and σ. Let C ⁻¹ ·y be a column vector a.

Then, Equation 1 can be transformed using Equation 4 as follows.

The function f(z) at this time is expressed as the sum of normal distributions represented by the terms on the right side. FIG. 10 schematically shows the relationship between the profile of the prediction model f(z) and the normal distribution expressed by each term on the right side of Equation 6 that constitutes the prediction model f(z). As can be seen from FIG. 10, given an unknown manufacturing parameter set z, the function f(z) yields a predicted value corresponding to a known manufacturing parameter set _xk close to the manufacturing parameter set z, e.g. (TS) predicted value. _Intuitively , therefore, the predictive model f(z) obtained by the _Gaussian process is the tensile force It can be said to output a predicted value close to strength (TS).

Next, how to set the parameters θ ₀ , θ ₁ and σ necessary for actually outputting the calculation result of Equation 6 as a predicted value will be examined.
In general, hyperparameters are parameters that are manually tuned by humans. Therefore, the parameters θ ₀ , θ ₁ and σ may be set manually.

カーネル関数ｋから計算されるカーネル行列をＫとおく。カーネル行列Ｋは、目的変数（観測値）ｙがガウス過程に従うとしたときの、説明変数Ｘの要素ｘ_ｉ，ｘ_ｊ間の共分散をまとめた共分散行列に相当する。つまり、カーネル行列Ｋは、加工済み製造パラメータｘを用いて算出され得る。カーネル関数がθに依存するためカーネル行列Ｋもθに依存する。 On the other hand, the parameters θ ₀ , θ ₁ and σ can also be set by estimation. As a method therefor, the present inventors estimated by learning from the data X and y using the second-class maximum likelihood estimation method.
Now treat the hyperparameters θ ₀ , θ ₁ and σ as a parameter vector θ=(θ ₀ , θ _{1 ,} σ). Then, the kernel function of Equation 4 depends on θ.

Let K be a kernel matrix calculated from a kernel function k. The kernel matrix K corresponds to a covariance matrix summarizing the covariances between the elements x _i and x _j of the explanatory variable X when the objective variable (observed value) y follows a Gaussian process. That is, the kernel matrix K can be calculated using the processed fabrication parameter x. Since the kernel function depends on θ, the kernel matrix K also depends on θ.

数８は尤度関数と呼ばれる。ハイパーパラメータθの尤度関数ｐ（ｙ｜θ）を最大化するθを求めることを、第２種最尤推定法と呼ぶ。 Let K _θ be the kernel matrix K that depends on θ. At this time, the learning data probability is represented by the following Equation 8. Note that N( ) indicates that the observed value y follows a normal distribution with an average of 0 and a covariance matrix _Kθ .

Equation 8 is called the likelihood function. Finding θ that maximizes the likelihood function p(y|θ) of the hyperparameter θ is called a second-class maximum likelihood estimation method.

As described above, the parameter vector θ described above is obtained, and as a result, a specific prediction model can be obtained.

FIG. 11 is a flowchart showing a procedure of model learning processing using a learning data set. This procedure is the details of step S14 in FIG.
At step S42, the arithmetic circuit 10 acquires a learning data set Ti (i=1, . . . , n). When the learning data set Ti is stored in the storage device 14 , the arithmetic circuit 10 acquires the learning data set Ti by reading from the storage device 14 . When the learning data set Ti is generated by an external computer system, the arithmetic circuit 10 acquires the learning data set Ti through the interface device 20c.
In step S44, the arithmetic circuit 10 calculates the vector C ⁻¹ ·y (equation 5) of the prediction model f(z)=k(z) ^T ·C ⁻¹ ·y. The resulting vector C ⁻¹ ·y contains the above hyperparameters.
In step S46, arithmetic circuit 10 determines parameters θ ₀ , θ ₁ and σ in the model by learning using processed manufacturing parameter x and observed value y.

Through the above processing, the parameters θ, θ, and σ can be set to f(z) shown in Equation (6). Equation 6 and the hyperparameters in Equation 6 that define the derived prediction model f(z) are stored in storage device 14, for example.

In the above description, an example using the second-class maximum likelihood estimation method was given, but the hyperparameter θ may be estimated by other methods. For example, a plurality of candidates for the hyperparameter θ may be prepared, and the one with the highest prediction accuracy may be selected by a grid search method.

The above prediction model f(z) is derived for each characteristic of the steel sheet to be tested. In other words, although the tensile strength (TS) was illustrated and explained in FIGS. .

Next, prediction processing based on an unknown manufacturing parameter set z will be described with reference to FIG.
FIG. 12 is a flow chart showing the procedure of prediction processing based on an unknown manufacturing parameter set z.
In step S<b>52 , the arithmetic circuit 10 acquires a trained prediction model y=f(z) by, for example, reading Equation 6 and hyperparameters applied to Equation 6 from the storage device 14 .

In step S54, the arithmetic circuit 10 calculates and stores the output value _yi using the manufacturing parameter set _zi to be predicted.
At step S56, the arithmetic circuit 10 determines whether or not a specific manufacturing parameter in the manufacturing parameter set has been changed within a predetermined range. For example, if the specific manufacturing parameter is "temperature T _preAT ", the arithmetic circuit 10 determines whether the temperature T _preAT is changed from 200°C to 600°C. If it is determined that the predetermined range has not yet been changed, the process proceeds to step S58, and if it is determined that the predetermined range has been changed, the process proceeds to step S60.

At step S58, the arithmetic circuit 10 changes the predetermined manufacturing parameter. At this time, the rest of the manufacturing parameters are kept unchanged. This means that the arithmetic circuit 10 generates the manufacturing parameter set z _i+1 from the manufacturing parameter set z _i used when calculating the predicted value y _i . Arithmetic circuit 10 holds the value of the manufacturing parameter after the change. For example, when the temperature T _preAT is changed, the arithmetic circuit 10 increments or decrements the value of the temperature T _preAT by a predetermined unit (for example, 1° C.) and holds the current value of the temperature T _preAT in the memory 12 . This results in a production parameter set z _i+1 in which the value of the temperature T _preAT is incremented or decremented while the other production parameters are kept constant. After that, the process returns to step S54. Arithmetic circuit 10 calculates a predicted value using each of a plurality of manufacturing parameter sets in which a specific manufacturing parameter is varied over a predetermined range and other manufacturing parameters are kept constant in step S56.
In step S60, the arithmetic circuit 10 determines a suitable range for a given manufacturing parameter from changes in predicted values due to changes in the given manufacturing parameter.
As described above, in the present embodiment, the processing of FIG. 12 is performed for each of tensile strength (TS), elongation at break (EL) and hole expansion ratio (λ).

FIG. 13 is a flow chart showing the procedure of the manufacturing parameter determination process considering a plurality of performance prediction values.
In step S72, the arithmetic circuit 10 acquires k types of predicted value data related to characteristics. In the present embodiment, arithmetic circuit 10 acquires predicted value data relating to three types of properties (tensile strength (TS), elongation at break (EL), and hole expansion ratio (λ)).
In step S74, the arithmetic circuit 10 determines ranges R ₁ to Rk of each manufacturing parameter that satisfy the requirements of the first to kth characteristics.

FIG. 16 shows a specific example of obtaining predicted value data according to the flow of FIG.
FIG. 16 shows a composition example that satisfies the target characteristics and a composition example that does not, extracted based on the predictive model. The colored squares indicate any of the desired range, the entry number (leftmost column) determined to be outside the result, the composition example, and the predicted result. Among the predicted values of tensile strength (TS), elongation at break (EL), and hole expansion ratio (λ) (three lines on the far right in FIG. 16), "○" indicates an example in which the target conditions were achieved. , "x" indicates an example in which the target condition is not achieved. "E2" and "E-2" in FIG. 16 represent 10 squared and 10 minus squared, respectively.
As shown in FIG. 16, it is possible to determine the range to be satisfied by T _preAT and t _preAT based on the characteristic prediction values when a plurality of combinations of manufacturing parameters are prepared. For example, T _preAT is excluded from the scope of the present invention because EL is not satisfied at least at 400° C. or lower. Further, if t _preAT is at least 50 seconds or more, the EL is not satisfied, so it is excluded from the scope of the present invention.

In step S76, the arithmetic circuit 10 determines the range R overlapping all of the ranges R ₁ to Rk as the range in which the manufacturing parameter is adopted.

In the present invention, a temperature range of 451° C. or more and 560° C. or less is adopted as the temperature T _preAT by preparing a large number of combinations of manufacturing parameters and verifying their characteristic prediction values. However, it is more preferable to adopt a range narrower than this range, for example, a range of 451° C. or higher and 530° C. or lower. Alternatively, it is also suitable to adopt a range of, for example, 500° C. or higher and 560° C. or lower.

Through similar verification, a range of 4.0 seconds or more and 44.3 seconds or less is adopted as the retention time t _preAT . However, it is more preferable to adopt a range narrower than this range, for example, a range of 4.0 seconds or more and 35.0 seconds or less.

As described above, a prediction model f(z) based on a Gaussian process is obtained by using the manufacturing parameter sets accumulated so far and the test results of the steel sheets actually manufactured with the manufacturing parameter sets, The predictive model f(z) can be used to predict test results when using an unknown manufacturing parameter set.

The present inventors verified the predictive model used in the present embodiment by cross-validation and confirmed its effectiveness. Verification will be described later in "3. Relationship between material properties and manufacturing conditions".

The holding temperature T _preAT obtained in this manner is 451° C. or higher and 560° C. or lower, preferably 451° C. or higher and 530° C. or lower, and the holding time t _preAT is 4.0 seconds or longer and 44.3 seconds or shorter, preferably 4.0° C. or higher. It is 0 seconds or more and 35.0 seconds or less. For the holding temperature T _preAT , it is also suitable to adopt a range of, for example, 500° C. or higher and 560° C. or lower.
Metallurgically examining the obtained results, when the temperature T _preAT is less than 451 ° C., the Θ (cementite) phase precipitation driving force at the interface between the bainitic ferrite and the austenite generated in the first heat treatment step increases and is maintained. , the carbon concentration of untransformed austenite decreases. For this reason, it is considered that retained austenite with a high degree of stability cannot be finally built in, and a high elongation at break cannot be ensured. Moreover, when the temperature T _preAT exceeds 580°C, bainite transformation does not occur, polygonal ferrite is generated, and coarse MA is formed around the generated polygonal ferrite, so a sufficient hole expansion ratio cannot be secured. it is conceivable that. If the time t _preAT exceeds 44.3 seconds, it is considered that bainite is excessively generated and sufficient tensile strength cannot be secured.

The first average cooling rate C1 is not a parameter adjusted to control the properties of the steel sheet in the present invention, but parameters such as the temperature T _preAT and the holding time t _preAT set to achieve the purpose of the first heat treatment step described above. It is set for proper adjustment, and when using continuous annealing equipment, it is a condition that can realize stable operation. Therefore, the information of C1 is not included in the training data set above. As a specific numerical range, the first average cooling rate C1 is 2° C./second or more and 60° C./second or less.
In this specification, the "average cooling rate" is a value obtained by dividing the temperature difference between the cooling start temperature and the cooling end temperature by the time required for cooling from the cooling start temperature to the cooling end temperature.

(c) Intermediate Cooling Step Subsequently, as shown in FIG. 1, an intermediate cooling step is performed in which the rolled material after the first heat treatment is cooled from the temperature T _preAT to the intermediate cooling temperature T2 at the second average cooling rate C2. The intermediate cooling temperature T2 satisfies the following formula (3). Ms (martensitic transformation start temperature: Ms point) included in the formula (3) can be obtained by the following formula (4).

Ms-215≤T2≤Ms-118 (3)
Here, Ms is based on the following formula (4).

Ms (°C) = 561 - 474 x [C] - 33 x [Mn] (4)
Here, [ ] is the content of each element in mass %.

If the intermediate cooling temperature T2 is lower than Ms-215° C., most of the untransformed austenite transforms into martensite, and sufficient retained austenite that contributes to elongation at break cannot be secured, resulting in insufficient elongation at break. On the other hand, if the intermediate cooling temperature T2 is as high as Ms-118° C., the degree of supercooling is small and the amount of martensite cannot be secured sufficiently, resulting in insufficient tensile strength. The intermediate cooling temperature T2 is preferably Ms-230.0°C or more and Ms-110.0°C or less. More preferably, the intermediate cooling temperature T2 is Ms-210°C or more and Ms-150°C or less.

The second average cooling rate C2 is not a parameter adjusted to control the properties of the steel sheet in the present invention, but is set to appropriately adjust the intermediate cooling temperature T2 set to achieve the purpose of the above-described intermediate cooling process. In addition, when using continuous annealing equipment, the conditions are such that stable operation can be realized. Therefore, the information of C2 is not included in the training data set above. As for a specific numerical range, the second average cooling rate C2 is 10° C./second or more and 60° C./second or less.

(d) Second heat treatment step Subsequently, the rolled material that has _undergone the intermediate cooling step as shown in FIG. 1 is heated from the intermediate cooling temperature T2 to the temperature T _postAT at a first average heating rate H1, After that, the second heat treatment of cooling from the temperature T _postAT to room temperature at the third average cooling rate C3 is performed.
The temperature T _postAT is 350°C or higher and 450°C or lower, preferably 380°C or higher and 420°C or lower. The holding time t _postAT is 100.0 seconds or more and 990.0 seconds or less, preferably 150.0 seconds or more and 400.0 seconds or less.

If the temperature T _postAT is lower than 350° C., the distribution of C from the martensite generated during supercooling in the intermediate cooling step to the untransformed austenite does not occur sufficiently, and stable retained γ cannot be incorporated. Breaking elongation cannot be secured. On the other hand, if the temperature T _postAT is higher than 450° C., carbides are formed in the untransformed austenite, lowering the carbon concentration in the untransformed austenite, making it impossible to ensure sufficient elongation at break.

If the holding time t _postAT is shorter than 100.0 seconds, the distribution of C from martensite generated during supercooling in the intermediate cooling process to untransformed austenite does not occur sufficiently, and stable retained γ cannot be sufficiently secured. Sufficient elongation at break cannot be secured. On the other hand, if the holding time t _postAT is longer than 990.0 seconds, carbides are formed in the untransformed austenite, lowering the carbon concentration in the untransformed austenite, making it impossible to ensure sufficient elongation at break. In addition, if the holding time t _postAT becomes longer, the productivity decreases, so from this point of view, holding for 990.0 seconds or more is not preferable.

The first average heating rate H1 is not a parameter adjusted to control the properties of the steel sheet in the present invention, but parameters such as the temperature T _postAT and the holding time t _postAT set to achieve the purpose of the second heat treatment step described above. It is set for proper adjustment, and when using continuous annealing equipment, it is a condition that can realize stable operation. Therefore, the above learning data set does not include the information of H1. As a specific numerical range, the first average heating rate H1 is 5° C./second or more and 60° C./second or less.
In this specification, the "average heating rate" is a value obtained by dividing the absolute value of the temperature difference between the heating start temperature and the heating end temperature by the time required to heat from the heating start temperature to the heating end temperature. .

The third average cooling rate C3 is not a parameter adjusted for controlling the properties of the steel sheet in the present invention, but is set to achieve the purpose of the second heat treatment step described above, and when using continuous annealing equipment conditions to realize stable operation. Therefore, we did not include the information of C3 in the training data set above. As a specific numerical range, the third average cooling rate C3 is 10° C./second or more and 60° C./second or less.

As described above, in the method for manufacturing a high-strength steel sheet according to the present invention, a rolled material having a predetermined composition is subjected to the above-described (a) soaking step, (b) first heat treatment step, and (c) intermediate cooling step. and (d) by sequentially performing the second heat treatment step, it has excellent properties such as tensile strength (TS) of 950 MPa or more, elongation at break (EL) of 22% or more, and hole expansion ratio (λ) of 20% or more. A high-strength steel sheet can be obtained.
Moreover, the Mn content of the rolled material to be used is 2.20% by mass or less, and the excellent properties described above can be obtained without requiring a large amount of Mn.
Furthermore, (a) the soaking step, (b) the first heat treatment step, (c) the intermediate cooling step and (d) the second heat treatment step are performed continuously using a single general-purpose continuous annealing facility (CAL). It does not require a complicated process such as being able to be performed on

In the steps (a) to (d) described above, the temperatures T1, T _preAT , T2, T _postAT , the first average cooling rate C1, the second average cooling rate C2, the third average cooling rate C2 and the first The average heating rate H1 can be measured by bringing a thermocouple into contact with the work (rolled material), such as by attaching the thermocouple to the work (rolled material). Further, for example, when the equipment is large and it is difficult to measure using a thermocouple, a non-contact thermometer such as a radiation thermometer may be used for measurement as a simple method.

3. Relationship Between Material Properties and Manufacturing Conditions The present inventors verified the prediction accuracy using the prediction model described above, and confirmed the effectiveness of the prediction model.
FIG. 14 shows an example of a known manufacturing parameter set used for prediction and the range of each manufacturing parameter. The inventors recursively obtained the relational expression from the data group satisfying the range shown in FIG. 14 using a Gaussian process. The prediction accuracy of the regression equation was confirmed using a cross-validation chart that compares actual results and predictions.

15A to 15C are cross-validation (CV) charts comparing predicted and actual values of tensile strength (TS), elongation at break (EL) and hole expansion ratio (λ). Cross-validation is a technique in statistics that divides sample data, first analyzes a part of it, tests the analysis on the remaining part, and applies it to verify and confirm the validity of the analysis itself. In the cross-validation, the ratio of the amount of data used for creating the prediction model and the amount of data used for verification was 9:1.

As a result of verification by cross-validation, it was confirmed that the predicted and actual results show good correspondence in each of tensile strength (TS), elongation at break (EL) and hole expansion ratio (λ). In particular, as shown by the dashed circles in each of FIGS. 15A to 15C, the tensile strength (TS), which is the characteristic standard of the present invention material, is 950 MPa, the elongation at break (EL) is 22%, and the hole expansion ratio (λ). In the vicinity of 20%, it was found that these characteristics can be sufficiently predicted by the learning model generated by the above-described procedure because the prediction and the actual result tend to match well.

Note that FIG. 14 does not include the first average cooling rate C1, the second average cooling rate C2, the third average cooling rate C2, and the first average heating rate H1 as manufacturing parameters. This is because the inventors made performance predictions without including these as manufacturing parameters, and as a result of verifying the correspondence between the predicted values and the actual values, the conclusion was obtained that the prediction results were good as described above. However, some or all of these may be included as manufacturing parameters for performance prediction.

1. Preparation of Samples Steel containing the components shown in Table 1, with the balance being iron and unavoidable impurities, was smelted to obtain cast materials. Example 1 shown in Table 1 corresponds to sample No. 1 in FIG. 1, and Comparative Example 2 corresponds to sample No. 1 in FIG. Equivalent to 10. The hot-rolled material obtained by hot-rolling the obtained cast material was further cold-rolled to obtain a rolled material (cold-rolled material) having a plate thickness of 1.4 mm.

Using a continuous annealing facility, the obtained rolled material is subjected to (a) a soaking step, (b) a first heat treatment step, (c) an intermediate cooling step, and (d) a second heat treatment according to the heat treatment pattern shown in FIG. The steps were performed in order to prepare an example sample (Example 1) and a comparative example sample (Comparative Example 1). The holding time t _preAT at the soaking temperature T1 was set to 180 seconds.
(2) "Soaking temperature T1-Ae3" in formula (1) obtained using Ae3 calculated using formula and soaking temperature T1, temperature T _preAT and holding time t _preAT in the first heat treatment step, ( 4) "Ms-T2" obtained using the martensitic transformation start temperature Ms and the intermediate cooling temperature T2 calculated using the formula (if this value is 118 or more and 215 or less, it means that the formula (3) is satisfied) , the temperature T _postAT and the holding time t _postAT of the second heat treatment step are shown in Table 3. The first average cooling rate C1 was 30°C/second, the second average cooling rate C2 was 10°C/second, and the third average cooling rate C3 was 30°C/second. Also, the first average heating rate H1 was set to 10° C./sec.
Each temperature was measured by attaching a thermocouple to the sample.

2. Evaluation of Mechanical Properties JIS No. 5 tensile test pieces were taken from the obtained Example 1 sample and Comparative Example 1 sample. Using this tensile test, a tensile test (in accordance with JIS Z2241) was performed to measure tensile strength (TS) and elongation at break (EL). Tensile strength of 950 MPa or more was considered acceptable, and breaking elongation of 22% or more was considered acceptable.

Further, the sample of Example 1 and the sample of Comparative Example 1 were subjected to a hole expansion test in accordance with the Japan Iron and Steel Federation standard JFST 1001 to evaluate the hole expansion rate (λ). A case where the hole expansion ratio was 20% or more was regarded as a pass.
Table 4 shows the evaluation results of tensile strength, elongation at break and hole expansion ratio.

The sample of Example 1, which satisfies all the requirements defined by the method of manufacturing a high-strength steel sheet according to the present invention, passed all the results of tensile strength, elongation at break and hole expansion ratio.
On the other hand, the comparative example 1 sample which has an excessive Mn content and a small Cr content and does not satisfy the formulas (1) and (3), the temperature T _preAT , the holding time t _preAT , the temperature T _postAT and the holding time t _postAT Although the tensile strength was comparable to that of the Example 1 sample and the hole expansion ratio was high, the elongation at break was very low and was rejected.
FIG. 17 is a diagram showing examples of manufacturing parameter sets and ranges of each manufacturing parameter that are adjusted to satisfy requirements for high-strength steel sheets. Suitable conditions for manufacturing parameters can be determined based on the manufacturing parameters derived using Gaussian process regression. High-strength steel sheets produced under such conditions have been confirmed to satisfy predetermined property requirements. It should be noted that the average heating rate and average cooling rate are within the numerical ranges described above regardless of the values of other parameters.

Claims (1)

C: 0.10% by mass or more and 0.40% by mass or less Si: 0.50% by mass or more and 1.80% by mass or less,
Mn: 1.60% by mass or more and 2.20% by mass or less,
Al: 0.01% by mass or more and 0.90% by mass or less,
Cr: 0.30% by mass or more and 0.80% by mass or less,
and the balance being iron and inevitable impurities, a soaking step of soaking the rolled material at a soaking temperature T1 that satisfies the following formula (1);
The rolled material is cooled to a temperature T _preAT of 451° C. or more and 560° C. or less at a first average cooling rate C1 of 2° C./s or more and 60° C./s or less, and the temperature T _preAT is 4.0 seconds or more and 44.3 seconds. a first heat treatment step held for a holding time t _preAT of:
an intermediate cooling step of cooling the rolled material to an intermediate cooling temperature T2 that satisfies the following formula (3) at a second average cooling rate C2 of 10° C./s or more and 60° C./s or less;
The rolled material is heated to a temperature T _postAT of 350° C. or more and 450° C. or less at a first average heating rate H1 of 5° C./sec or more and 60° C./sec or less, and the temperature T _postAT is 100.0 seconds or more and 990.0 seconds. A second heat treatment step of cooling to room temperature at a third average cooling rate C3 of 10° C./s or more and 60° C./s or less after holding for the following holding time t _postAT ;
A method for producing a high-strength steel sheet having a tensile strength of 950 MPa or more, a breaking elongation of 22% or more, and a hole expansion ratio of 20% or more, comprising in order.

-30.0°C ≤ T1-Ae3 ≤ 40.0°C (1)
Here, Ae3 is based on the following formula (2).

Ae3 (°C) = 896.5414 - 236.22 x [C] + 26.18998 x [Si] - 28.4882 x [Mn] + 106.9042 x [Al] + 1.284594 x [Ti] + 12.66673 x [ Nb]+42.66673×[V]−39.9466×[Ni]+2.467166×[Mo]−17.8592×[Cr] (2)
Here, [ ] is the content of each element in mass %.

Ms-215≤T2≤Ms-118 (3)
Here, Ms is based on the following formula (4).

Ms (°C) = 561 - 474 x [C] - 33 x [Mn] (4)
Here, [ ] is the content of each element in mass %.

Images