Steel thermophysical parameter online calculation method and system based on classification and regression

Abstract

The invention discloses a steel thermophysical parameter online calculation method based on classification and regression. The method comprises the following steps that S1, steel types are classified according to components of the steel types through a steel type classification model; and S2, on the basis of a component and physical property parameter model established by applying a machine learning algorithm under each steel type category, the thermophysical property parameters of the steel types are calculated. The invention discloses a steel thermophysical parameter online calculation system based on classification and regression. According to the method, the steel types are classified according to the steel type classification rule, the learning data of the model is established, and it is guaranteed that enough learning data exist in each type, so that the wide adaptability of the established model is guaranteed, and the method can be used for new steel types of subsequent development.

Description

Technical Field

[0001] The invention relates to a steelmaking continuous casting technology, and more specifically to an online calculation method and system for steel thermophysical property parameters based on classification and regression. Background Art

[0002] The continuous casting process is a typical high-temperature process. During the operation, there are complex changes in thermal and mechanical states and changes in microstructure and magnetic state, which have a decisive influence on the performance of the ingot. The performance evolution of steel during the continuous casting process is closely related to the quality of the ingot. Only on the basis of accurate understanding of the changes in high-temperature thermophysical properties of continuous casting can the correct process operation or alloying method be applied to regulate the performance of the ingot at high temperature and achieve the production of high-quality, high-performance ingots and final steel products. From a process perspective, if the relationship between composition and thermophysical properties can be obtained through experiments or theoretical calculations, then the continuous casting process can be accurately designed to obtain defect-free or less-defective ingots.

[0003] Among them, the thermophysical parameters that are highly relevant to the design of continuous casting process parameters include the following three:

[0004] (1) Liquidus: determines the temperature control range of the tundish, i.e., the superheat control during casting.

[0005] (2) The thermoplasticity of steel, especially the determination of the temperature range of the third brittle zone of steel. The third brittle zone needs to be avoided in the secondary cooling and straightening zone. Otherwise, cracks may easily appear in the ingot due to the straightening force during the straightening stage.

[0006] (3) The A3 temperature of steel, that is, the low-temperature ferrite precipitation temperature of steel, determines the critical upper limit temperature of hot charging of steel.

[0007] On the other hand, the research methods of continuous casting process include field test, pilot study and numerical simulation. The most accurate data often need to be obtained through field test or full-scale equipment, so that the operation under the same conditions can be predicted. However, the continuous casting process is carried out under high temperature conditions, the equipment is very large, the cost is high and the time cycle is long, so it is impossible to study it through field test and full-scale test. Therefore, the study of continuous casting process through numerical simulation has become the main means. Accurate numerical simulation of continuous casting process requires accurate input of thermophysical parameters. A continuous casting numerical simulation that can participate in online control is necessary, and its realization is based on the online and fast calculation of thermophysical parameters.

[0008] In summary, in order to better design the continuous casting process of steel grades and quickly and accurately perform numerical simulation of continuous casting, it is necessary to provide a method that can quickly and accurately calculate the thermophysical properties of steel, especially the liquidus, thermoplastic properties, density, thermal conductivity, etc.

[0009] At present, there are many literatures at home and abroad that have conducted relevant research on the thermophysical parameters of steel. The methods for obtaining physical parameters include laboratory testing, thermodynamic commercial software calculation, and empirical formula calculation. The experimental test results are usually closest to the real results, but they are often limited by the experimental temperature, and are forced to use linear extrapolation methods to obtain physical parameters in a higher temperature range, and the measured steel types are limited. The physical parameters provided by commercial software are increasingly used in continuous casting model research. Commonly used commercial software that can be used to calculate thermophysical parameters include JMatPro, Thermo-Calc, ProCAST, etc. The applicability of the thermophysical parameters of these software depends on the accuracy and scope of application of the database. They are often insufficient for new steel types and cannot be applied online. The empirical formula is currently widely used in production bases. The advantage of this method is that the physical parameter value or the physical parameter-temperature change curve can be quickly calculated by inputting the steel composition and temperature. The disadvantage is that the source and accuracy of the empirical formula are unclear; the empirical formula is often linearly related to the element content, which is inconsistent with the complexity of the relationship between thermophysical properties and composition.

[0010] However, there are few existing patents involving the calculation of thermophysical parameters of steel. For example, the authorization announcement number CN103192048B discloses a simulation method for the solidification and cooling process of continuous casting billets based on accurate thermophysical parameters, which includes a method for accurately calculating the thermophysical parameters of steel to form a thermophysical parameter database of steel, and provide basic data for basic simulation and simulation in the cooling and solidification simulation process of continuous casting billets, thereby improving the accuracy of the simulation results of the temperature field of the billet. In this patent technology, the thermophysical property calculation of steel focuses on the calculation method of high-temperature plasticity of steel related to cracks. This high-temperature plasticity model uses a multivariate linear regression analysis method to form a high-temperature plasticity prediction method for steel with chemical composition as an independent variable, and establishes a high-temperature plasticity prediction model, wherein the data used to establish the high-temperature thermoplasticity model of the multivariate linear regression is obtained by experimental methods. Although the method described in this patent technology establishes a high-temperature thermoplasticity model based on experimental data and a multivariate linear regression method, it can make up for the lack of unclear sources of existing empirical formulas, and can be applied online, but there are obvious limitations. First, the limited experimental data has a limited scope of application and cannot be applied to new steel types with increasingly complex alloy components. Second, like the shortcomings of empirical formulas, it is linearly related to the element content, which is inconsistent with the complexity of the relationship between thermophysical properties and composition. Thermophysical properties change discontinuously with the organization, and the organization changes discontinuously with the composition and temperature. Therefore, the thermophysical properties must have a discontinuous relationship with the composition, while linear correlation means that the changes in thermophysical properties and composition are continuous. Summary of the invention

[0011] In view of the defects existing in the prior art, the purpose of the present invention is to provide an online calculation method and system for the thermal physical properties of steel based on classification and regression. The steel types are classified according to the steel type classification rules, and the learning data of the model is established to ensure that there is sufficient learning data in each category, thereby ensuring the wide adaptability of the model, which can be used for new steel types developed subsequently.

[0012] To achieve the above object, the present invention adopts the following technical solution:

[0013] The first aspect of the present invention provides an online calculation method for steel thermophysical property parameters based on classification and regression, comprising the following steps:

[0014] S1. Classify the steel grades according to their composition through a steel grade classification model;

[0015] S2. Based on the composition and physical parameter model established by applying machine learning algorithm under each steel grade category, the thermophysical parameters of the steel grade are calculated.

[0016] Preferably, in step S1, the input of the steel classification model is the carbon content and various alloy elements, and the output is the steel category code.

[0017] Preferably, the steel grade classification model is a steel grade classification rule.

[0018] Preferably, the steel classification rule adopts a three-level classification rule, wherein the first-level classification rule classifies according to the carbon element, the second-level classification rule classifies according to the sum of alloy elements, and the third-level classification rule classifies according to the nitride metal element.

[0019] Preferably, the primary classification rules are as follows:

[0020] When the mass percentage of carbon content is between 0.01 and 0.025, the primary category is Category 1;

[0021] When the carbon content mass percentage is between 0.025 and 0.065, the primary category is category 2;

[0022] When the carbon content mass percentage is between 0.065 and 0.18, the primary category is category 3;

[0023] When the carbon content mass percentage is between 0.18 and 0.25, the primary category is category 4;

[0024] When the carbon content is greater than 0.25 mass percent, the primary category is 5.

[0025] Preferably, the secondary classification rules are as follows:

[0026] Determine whether the sum of the mass percentages of the alloying elements is less than A1. If so, the secondary category of the steel grade is 0; if not, the secondary category of the steel grade is 1.

[0027] The value of A1 is between 0.006 and 0.010.

[0028] Preferably, the three-level classification rules are as follows:

[0029] Determine whether the product of Al and N in the steel grade is less than A2, whether the product of B and N is less than A3, and whether the product of Nb, N and C is less than A4. If so, the third-level category of the steel grade is 0, if not, the third-level category of the steel grade is 1;

[0030] Al, N, B, Bb, and C are all mass percentage contents;

[0031] The value of A2 is between 0.00002 and 0.00005, the value of A3 is between 0.000001 and 0.000003, and the value of A4 is between 0.00000002 and 0.00000005.

[0032] Preferably, the steel classification model obtains high-temperature structures of no less than 10 steels obtained by quenching in the solid-liquid two-phase region and the structures after reaching room temperature from each category of steel calibrated according to the three-level classification rule, and uses characterization means to obtain the organizational characteristics of the steel.

[0033] Preferably, the organizational characteristics include whether there is a peritectic reaction, whether there is proeutectoid ferrite, whether there is cementite, and whether there are three types of nitrides, namely, AlN, NbCN, and BN.

[0034] Preferably, in step S2, the physical property parameters of the composition and physical property parameter model are the liquidus temperature that affects the design of continuous casting process parameters, and the temperature-density curve and temperature-thermal conductivity curve that need to be input into the continuous casting numerical simulation model.

[0035] Preferably, the thermophysical property parameter is a thermoplasticity curve.

[0036] Preferably, the machine learning algorithm is a multivariate linear and nonlinear regression algorithm.

[0037] Preferably, the liquidus temperature, the temperature-density curve, and the temperature-thermal conductivity curve are all subjected to composition-performance modeling using a multivariate linear regression method.

[0038] Preferably, the thermoplasticity curve extracts the temperature range of the third brittle zone, specifically as follows:

[0039] Apply random forest machine learning method to obtain composition-low temperature ferrite precipitation temperature, and obtain composition-nitride precipitation temperature for steel types with nitride precipitation;

[0040] According to the third brittle temperature range of steel grades, the relationship model between low-temperature ferrite precipitation temperature, nitride precipitation temperature and the third brittle temperature range was established, and then the composition-third brittle temperature range model was established through these two models.

[0041] The second aspect of the present invention provides an online calculation system for steel thermal physical property parameters based on classification and regression, comprising:

[0042] Steel grade physical property parameter case library, used to store the collected steel grade physical property parameter data;

[0043] A calculation module, used for calculating the physical property parameter data of the steel grade physical property parameter case library;

[0044] A physical property parameter retrieval module, used for retrieving the physical property parameter data;

[0045] The online calculation system for thermal properties of steel based on classification and regression realizes the online calculation method for thermal properties of steel based on classification and regression provided in the first aspect of the present invention.

[0046] Preferably, the steel grade physical property parameter case library includes a steel grade composition table, a steel grade physical property parameter table, and a steel grade physical property parameter folder;

[0047] The steel grade composition table is the actual average composition of the steel grade;

[0048] The steel grade physical property parameter table lists the physical property parameter names of the steel grade in the steel grade physical property parameter case library and the address of the file where the parameter data is located;

[0049] The steel grade physical property parameter folder contains the physical property parameter data of the steel grade, and one physical property parameter data corresponds to one file.

[0050] Preferably, the calculation module includes:

[0051] Steel classification model, which classifies steel grades according to their composition;

[0052] The calculation model of steel grade physical property parameters adopts the following process for calculation:

[0053] If the input tapping mark can be found in the physical property parameter case library, the data used is the data in the physical property parameter case library; if the input tapping mark cannot be found in the physical property parameter case library, the data used is the calculation data of the steel grade physical property parameter calculation model;

[0054] The data input model, after the newly measured physical property parameters are confirmed, the data storage model enters the new physical property data case into the physical property parameter case library.

[0055] Preferably, the physical property parameter retrieval module is written using a B / S architecture, and the retrieval content includes steel-making marks, component input, user-input components, and output physical property parameters.

[0056] The present invention provides an online calculation method and system for steel thermal property parameters based on classification and regression, which has the following beneficial effects:

[0057] (1) The calculation model is simple and can quickly calculate the thermophysical properties of steel for continuous casting process design and as input for continuous casting numerical simulation;

[0058] (2) The establishment of the composition-thermophysical property model fully considers the discontinuity of composition-structure-thermophysical properties through steel grade classification, that is, the sudden change of thermophysical properties when the composition changes. It is more accurate than the linear regression method currently used;

[0059] (3) Compared with simply predicting the high-temperature thermoplastic properties of a steel grade, the present invention proposes predicting the third brittle temperature range of the steel grade, which is more conducive to the design of the secondary cooling straightening temperature in continuous casting production. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a schematic diagram of the framework structure of the online calculation system of steel thermal physical property parameters based on classification and regression of the present invention. DETAILED DESCRIPTION

[0061] In order to better understand the above technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0062] The present invention also provides an online calculation method for steel thermal property parameters based on classification and regression, comprising the following steps:

[0063] S1. Classify the steel grades according to their composition through a steel grade classification model;

[0064] The input of the steel grade classification model is carbon content and various alloying elements, and the output is the steel grade category code.

[0065] S2. Calculate the thermophysical parameters of each steel grade based on the composition and physical parameter model established by applying the machine learning algorithm under each steel grade category;

[0066] The physical property parameters in the composition and physical property parameter model are the liquidus temperature that affects the design of continuous casting process parameters, that is, the thermoplasticity curve, the low-temperature ferrite precipitation temperature, and the temperature-density curve and temperature-thermal conductivity curve that need to be input into the continuous casting numerical simulation model;

[0067] The machine learning algorithms are multivariate linear and nonlinear regression algorithms.

[0068] In the above step S1, the steel grade classification model is a steel grade classification rule.

[0069] The steel classification rules adopt a three-level classification rule, among which the first-level classification rule is classified according to the carbon element, the second-level classification rule is classified according to the sum of alloy elements, and the third-level classification rule is classified according to the nitride metal element. Specifically:

[0070] The first-level classification rules are as follows:

[0071] When the mass percentage of carbon content is between 0.01 and 0.025, the primary category is Category 1;

[0072] When the carbon content mass percentage is between 0.025 and 0.065, the primary category is category 2;

[0073] When the carbon content mass percentage is between 0.065 and 0.18, the primary category is category 3;

[0074] When the carbon content mass percentage is between 0.18 and 0.25, the primary category is category 4;

[0075] When the carbon content is greater than 0.25 mass percent, the primary category is 5.

[0076] The secondary classification rules are calibrated according to the sum of alloying elements, because the sum of alloying elements affects the effect of carbon content on the structure of steel grades, especially its effect on the solid-liquid two-phase region, thus affecting the thermal properties of steel grades in the solid-liquid two-phase region. The details are as follows:

[0077] Determine whether the sum of the mass percentages of the alloying elements is less than A1. If so, the secondary category of the steel grade is 0; if not, the secondary category of the steel grade is 1.

[0078] The value of A1 is between 0.006 and 0.010.

[0079] The three-level classification rules are as follows:

[0080] Determine whether the product of Al and N in the steel grade is less than A2, whether the product of B and N is less than A3, and whether the product of Nb, N and C is less than A4. If so, the third-level category of the steel grade is 0, if not, the third-level category of the steel grade is 1;

[0081] Al, N, B, Bb, and C are all mass percentage contents;

[0082] The value of A2 is between 0.00002 and 0.00005, the value of A3 is between 0.000001 and 0.000003, and the value of A4 is between 0.00000002 and 0.00000005.

[0083] In order to classify steel types more accurately, the steel type classification model in the above step S1 is a steel type classification model obtained by a classification machine learning algorithm.

[0084] The steel classification model is calibrated according to the three-level classification rules. The high-temperature microstructures of no less than 10 steel types obtained by quenching in the solid-liquid two-phase region and the microstructures after reaching room temperature are obtained, and the microstructural characteristics of the steel type are obtained by characterization means.

[0085] The organizational characteristics include whether there is peritectic reaction, whether there is proeutectoid ferrite, whether there is cementite, and whether there are three types of nitrides, namely, ALN, NbCN, and BN. According to the combination of these judgment conditions, the above test steel grades are calibrated with category labels. Machine learning classification algorithms such as random forest algorithms are applied to obtain steel classification models.

[0086] The advantages of adopting the three-level classification rule and further adopting the steel grade classification model are:

[0087] The influence of organization on thermophysical properties, especially the influence on high-temperature thermoplastic properties, is fully considered. The organization that affects high-temperature thermoplasticity includes whether there is peritectic reaction in the solid-liquid phase region and the subsequent precipitation of nitrides. Whether there is peritectic reaction is determined by the sum of carbon content and alloying elements. Therefore, the first-level classification rule and the second-level classification rule of steel grades are determined by the sum of carbon content and alloying elements respectively, and the rule threshold is the experience obtained by combining the organization observed in the experiment. Considering that not all alloying elements have an effect on high-temperature thermoplastic properties, the influence of high-temperature thermoplastic properties is mainly low-temperature ferrite precipitation and nitrides. Among them, low-temperature ferrite precipitation is mainly related to carbon content. In the first-level classification rule of steel grades, the classification of carbon content has taken into account the influence of carbon content on low-temperature ferrite precipitation, especially on the precipitation temperature of low-temperature ferrite. Therefore, the third-level rule of steel grade classification mainly considers the influence of nitrides on high-temperature plastic properties. However, not all nitrides have an effect on the high-temperature thermoplasticity of steel. According to previous test data, the nitrides that affect high-temperature thermoplastic properties are mainly AlN, BN and (NbC)N. The three nitrides that affect the precipitation are AlN product, BN product and NbNC product. When they are not lower than a certain threshold, the corresponding nitrides will precipitate. Compared with the simple threshold setting in the three-level classification rule, the machine learning algorithm will be more able to consider the mutual influence between different elements. Therefore, the three-level classification rule of steel grades was further improved based on the data measured in the experiment according to the machine learning classification algorithm.

[0088] The liquidus temperature, temperature-density curve, and temperature-thermal conductivity curve were all modeled using the multivariate linear regression method for composition-performance modeling.

[0089] Thermoplasticity curve extracts the temperature range of the third brittle zone, which is related to two temperatures, namely, the low-temperature ferrite precipitation temperature and the nitride precipitation temperature. The details are as follows:

[0090] Apply random forest machine learning method to obtain composition-low temperature ferrite precipitation temperature, and obtain composition-nitride precipitation temperature for steel types with nitride precipitation;

[0091] According to the third brittle temperature range of steel grades, the relationship model between low-temperature ferrite precipitation temperature, nitride precipitation temperature and the third brittle temperature range was established, and then the composition-third brittle temperature range model was established through these two models.

[0092] The advantage / effect of using a machine learning regression algorithm on each steel grade category is that the steel grade classification takes into account the discontinuity of composition-thermal properties caused by the discontinuity of composition-structure, and the prediction accuracy is further improved due to the support of the machine learning algorithm rather than just the linear regression algorithm. In addition, the temperature range of the third brittleness range is used to specify a suitable secondary cooling straightening temperature for the production of continuous casting billets, and the present invention also provides a prediction method for this temperature range.

[0093] Combination Figure 1 As shown, the present invention also provides an online calculation system for steel thermal physical property parameters based on classification and regression, comprising:

[0094] Steel type physical property parameter case library 1, used to store the collected steel type physical property parameter data;

[0095] Calculation module 2, used for calculating the physical property parameter data of the steel physical property parameter case library 1;

[0096] Physical property parameter retrieval module 3, used for remote retrieval of physical property parameter data;

[0097] The online calculation method of steel thermal physical property parameters based on classification and regression of the present invention is realized through the online calculation system of steel thermal physical property parameters based on classification and regression of the present invention.

[0098] Steel grade physical property parameter case library 1 includes steel grade composition table, steel grade physical property parameter table, and steel grade physical property parameter folder;

[0099] The steel grade composition table shows the actual average composition of the steel grade;

[0100] The steel grade physical property parameter table lists the physical property parameter names of the steel grade in the steel grade physical property parameter case library and the address of the file where the parameter data is located;

[0101] The steel grade physical property parameter folder contains the physical property parameter data of the steel grade, and one physical property parameter data corresponds to one file.

[0102] The computing module 2 includes:

[0103] Steel classification model, which classifies steel grades according to their composition;

[0104] The calculation model of steel grade physical property parameters adopts the following process for calculation:

[0105] If the input tapping symbol can be found in the physical property parameter case library, the data used is the data in the physical property parameter case library; if the input tapping symbol cannot be found in the physical property parameter case library, the data used is the calculation data of the steel grade physical property parameter calculation model;

[0106] The data input model, after the newly measured physical property parameters are confirmed, the data storage model will enter the new physical property data case into the physical property parameter case library.

[0107] The physical property parameter retrieval module 3 is written in C / S architecture, and the retrieval content includes steel-making marks, component input, user-input components, and output physical property parameters.

[0108] In summary, in the physical property calculation model of the present invention, steel types are classified according to the steel type classification rules, especially considering the great influence of trace elements on physical property parameters, and at the same time, a model between composition and physical property established by a machine learning algorithm is applied in each category, thereby considering the nonlinear relationship between elements and thermal physical property parameters. The learning data for establishing the model comes from the mutual verification of experiments and calculation methods, ensuring that there is sufficient learning data in each category, thereby ensuring the wide adaptability of the established model, which can be used for new steel types developed subsequently.

[0109] Example

[0110] In this embodiment, the online calculation method of steel thermophysical property parameters based on classification and regression is used to calculate the temperature range of the third brittle zone of thermophysical properties for 12 steel types with secondary cooling straightening cracks in a continuous casting machine. The steel type composition, calculation results and actual secondary cooling straightening temperature are compared as shown in Table 1.

[0111] It was found that the straightening temperatures of 11 steel grades were in the low brittleness zone, and the calculated hit rate reached 90%. The physical properties of the one steel grade that was not hit will be measured and added to the physical property parameter case library to further improve the accuracy of the system.

[0112] Table 1

[0113]

[0114]

[0115] Those skilled in the art should recognize that the above embodiments are only used to illustrate the present invention, and are not intended to limit the present invention. As long as they are within the spirit of the present invention, any changes or modifications to the above embodiments will fall within the scope of the claims of the present invention.

Claims

1. An online calculation method for steel thermophysical property parameters based on classification and regression, characterized in that: The following steps are involved: S1. Classify the steel grades according to their composition through a steel grade classification model; S2. Based on the composition and physical parameter model established by applying machine learning algorithm under each steel grade category, the thermophysical parameters of the steel grade are calculated.

2. The online calculation method for steel thermophysical property parameters based on classification and regression according to claim 1 is characterized in that: In step S1, the input of the steel classification model is the carbon content and various alloy elements, and the output is the steel category code.

3. The online calculation method of steel thermophysical property parameters based on classification and regression according to claim 2 is characterized in that: The steel grade classification model is a steel grade classification rule.

4. The online calculation method of steel thermophysical property parameters based on classification and regression according to claim 3 is characterized in that: The steel classification rule adopts a three-level classification rule, wherein the first-level classification rule classifies according to the carbon element, the second-level classification rule classifies according to the sum of alloy elements, and the third-level classification rule classifies according to the nitride metal element.

5. The online calculation method for steel thermophysical property parameters based on classification and regression according to claim 4 is characterized in that: The first-level classification rules are as follows: When the mass percentage of carbon content is between 0.01 and 0.025, the primary category is Category 1; When the carbon content mass percentage is between 0.025 and 0.065, the primary category is category 2; When the carbon content mass percentage is between 0.065 and 0.18, the primary category is category 3; When the carbon content mass percentage is between 0.18 and 0.25, the primary category is category 4; When the carbon content is greater than 0.25 mass percent, the primary category is 5.

6. The online calculation method of steel thermophysical property parameters based on classification and regression according to claim 5 is characterized in that: The secondary classification rules are as follows: Determine whether the sum of the mass percentages of the alloying elements is less than A1. If so, the secondary category of the steel grade is 0; if not, the secondary category of the steel grade is 1. The value of A1 is between 0.006 and 0.

010.

7. The online calculation method for steel thermophysical property parameters based on classification and regression according to claim 6 is characterized in that: The three-level classification rules are as follows: Determine whether the product of Al and N in the steel grade is less than A2, whether the product of B and N is less than A3, and whether the product of Nb, N and C is less than A4. If so, the third-level category of the steel grade is 0, if not, the third-level category of the steel grade is 1; Al, N, B, Bb, and C are all mass percentage contents; The value of A2 is between 0.00002 and 0.00005, the value of A3 is between 0.000001 and 0.000003, and the value of A4 is between 0.00000002 and 0.00000005.

8. The online calculation method of steel thermophysical property parameters based on classification and regression according to claim 7 is characterized in that: The steel classification model obtains high-temperature structures of no less than 10 steels obtained by quenching in the solid-liquid two-phase region and the structures after reaching room temperature from each category of steels calibrated according to the three-level classification rules, and characterization means are used to obtain the organizational characteristics of the steel.

9. The online calculation method of steel thermophysical property parameters based on classification and regression according to claim 8, characterized in that: The organizational characteristics include whether there is a peritectic reaction, whether there is proeutectoid ferrite, whether there is cementite, and whether there are three types of nitrides, namely, AlN, NbCN, and BN.

10. The online calculation method of steel thermophysical property parameters based on classification and regression according to claim 1, characterized in that: In the step S2, the physical property parameters of the composition and physical property parameter model are the liquidus temperature that affects the design of continuous casting process parameters and the temperature-density curve and temperature-thermal conductivity curve that need to be input into the continuous casting numerical simulation model.

11. The online calculation method of steel thermophysical property parameters based on classification and regression according to claim 10, characterized in that: The thermophysical property parameter is a thermoplasticity curve.

12. The online calculation method of steel thermophysical property parameters based on classification and regression according to claim 10, characterized in that: The machine learning algorithms are multivariate linear and nonlinear regression algorithms.

13. The online calculation method of steel thermophysical property parameters based on classification and regression according to claim 11, characterized in that: The liquidus temperature, the temperature-density curve, and the temperature-thermal conductivity curve are all subjected to component-performance modeling using a multivariate linear regression method.

14. The online calculation method for steel thermophysical property parameters based on classification and regression according to claim 11, characterized in that: The thermoplasticity curve extracts the temperature range of the third brittle zone, which is as follows: Apply random forest machine learning method to obtain composition-low temperature ferrite precipitation temperature, and obtain composition-nitride precipitation temperature for steel types with nitride precipitation; According to the third brittle temperature range of steel grades, the relationship model between low-temperature ferrite precipitation temperature, nitride precipitation temperature and the third brittle temperature range was established, and then the composition-third brittle temperature range model was established through these two models.

15. An online calculation system for steel thermophysical property parameters based on classification and regression, characterized in that: include: Steel grade physical property parameter case library, used to store the collected steel grade physical property parameter data; A calculation module, used for calculating the physical property parameter data of the steel grade physical property parameter case library; A physical property parameter retrieval module, used for retrieving the physical property parameter data; The online calculation system for thermal properties of steel based on classification and regression implements the online calculation method for thermal properties of steel based on classification and regression as described in any one of claims 1-14.

16. The online calculation system for steel thermophysical property parameters based on classification and regression according to claim 15, characterized in that: The steel grade physical property parameter case library includes a steel grade composition table, a steel grade physical property parameter table, and a steel grade physical property parameter folder; The steel grade composition table is the actual average composition of the steel grade; The steel grade physical property parameter table lists the physical property parameter names of the steel grade in the steel grade physical property parameter case library and the address of the file where the parameter data is located; The steel grade physical property parameter folder contains the physical property parameter data of the steel grade, and one physical property parameter data corresponds to one file.

17. The online calculation system for steel thermophysical property parameters based on classification and regression according to claim 16, characterized in that: The calculation module comprises: Steel classification model, which classifies steel grades according to their composition; The calculation model of steel grade physical property parameters adopts the following process for calculation: If the input tapping mark can be found in the physical property parameter case library, the data used is the data in the physical property parameter case library; if the input tapping mark cannot be found in the physical property parameter case library, the data used is the calculation data of the steel grade physical property parameter calculation model; The data input model, after the newly measured physical property parameters are confirmed, the data storage model enters the new physical property data case into the physical property parameter case library.

18. The online calculation system for steel thermophysical property parameters based on classification and regression according to claim 15, characterized in that: The physical property parameter retrieval module is written in B / S architecture, and the retrieval content includes steel-making marks, component input, user-input components, and output physical property parameters.

Images