Refining slagging method based on big data steelmaking model under low molten iron ratio condition

Abstract

The invention discloses a refining slagging method based on a big data steelmaking model under a low molten iron ratio condition. The method comprises the following steps: training an initial model by utilizing historical data to obtain a prediction model; collecting real-time data in the molten steel refining process; inputting the real-time data into the prediction model, and obtaining the total refining slagging amount output by the prediction model; calculating the amount of slag added in the molten steel refining process; subtracting the added slag amount from the total refining slag amount output by the prediction model, and determining the slag amount to be added in the molten steel refining process; calculating a CaO-Al2O3-SiO2-MgO quaternary phase diagram by utilizing Factsample, and calculating the to-be-added amount of each component in the slag charge according to the total refining slag amount, the added slag amount, the to-be-added slag amount and a phase diagram calculation result; and adding the slag charge according to the calculated to-be-added amount of each component in the slag charge in the molten steel refining process. According to the scheme provided by the invention, intelligentization of the refining process can be realized, the refining time is shortened, and the product quality is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of iron and steel smelting, and particularly to a refining slag-making method based on a big data steelmaking model under the condition of a low hot metal ratio. Background Art

[0002] The LF refining furnace is an important equipment in the iron and steel smelting process. Through refining treatment, impurities in the molten steel can be removed, thereby improving the quality of the molten steel. The slag-making technology is a key link in the operation of the LF refining furnace and has a crucial impact on the chemical reactions in the furnace and the quality of the molten steel. In recent years, with the continuous improvement of the requirements for the quality of molten steel and production efficiency in the iron and steel industry, the research and application of high-efficiency slag-making technology for LF refining furnaces have received extensive attention. Iron and steel enterprises are urgently in need of optimizing the slag-making technology to improve production efficiency, reduce production costs, and meet the market demand for high-quality steel.

[0003] The traditional refining process relies on manual judgment and operation, and there are problems such as large fluctuations in process control, high material consumption and energy consumption, and poor product quality stability. Due to the limitations of manual experience and operation, the slag-making effect of the traditional slag-making technology is often not ideal, and the impurities in the molten steel are not removed thoroughly. And affected by the slag from the converter, it is often difficult to make white slag during the LF refining process, the refining time is long, the control effects of desulfurization and inclusions are poor, the product quality is unstable, and the production cost is high.

[0004] Technologies such as the Internet of Things, big data, and artificial intelligence will promote the development of iron and steel production towards intelligence and automation, improving efficiency and product quality. How to optimize the refining slag-making process by using advanced means such as big data models is a technical problem that urgently needs to be solved at present. Summary of the Invention

[0005] To solve the technical problems of long refining time and unstable product quality existing in the existing refining slag-making process relying on manual experience judgment, an embodiment of the present invention provides a refining slag-making method based on a big data steelmaking model under the condition of a low hot metal ratio.

[0006] The technical solution of the embodiment of the present invention is realized as follows: An embodiment of the present invention provides a refining slag-making method based on a big data steelmaking model under the condition of a low hot metal ratio. The method includes: Obtaining refining historical data; the historical data includes the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted, the amount of refining scrap steel, the type of refining scrap steel, the temperature of the refined scrap steel after baking, the weight of the slag from the converter, the composition of the converter slag, the temperature of the molten steel entering the station, the composition of the molten steel tapped from the converter, and the amount of refining slag-making. Train an initial model using the historical data to obtain a prediction model; the input of the prediction model is the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted, the amount of refining scrap steel, the type of refining scrap steel, the temperature of the refined scrap steel after baking, the weight of the converter slag flowing into the ladle, the composition of the converter slag, the temperature of the molten steel entering the station, and the composition of the molten steel tapped from the converter, and the output of the prediction model is the amount of refining slag making; Collect real-time data during the molten steel refining process; the real-time data includes the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted, the amount of refining scrap steel, the type of refining scrap steel, the temperature of the refined scrap steel after baking, the weight of the converter slag flowing into the ladle, the composition of the converter slag flowing into the ladle, the temperature of the molten steel entering the station, and the composition of the molten steel entering the station; Input the real-time data into the prediction model to obtain the total amount of refining slag making output by the prediction model; Calculate the amount of slag already added during the molten steel refining process; Subtract the amount of slag already added from the total amount of refining slag making output by the prediction model to determine the amount of slag to be added during the molten steel refining process; Use Factsage to calculate the CaO-Al2O3-SiO2-MgO quaternary phase diagram, and calculate the amount of each component to be added in the slag material according to the total amount of refining slag making, the amount of slag already added, the amount of slag to be added, and the calculation results of the phase diagram; Add the slag material in the molten steel refining process in the amount of each component to be added calculated in the slag material.

[0007] In one embodiment, calculating the amount of slag already added during the molten steel refining process includes: Obtain the oxide content after alloy oxidation during the molten steel refining process; Obtain the amount of slag material directly added after tapping from the converter; Obtain the estimated ladle erosion amount for this time based on historical data; Obtain the amount of converter slag flowing into the ladle detected after tapping from the converter; Determine the sum of the oxide content, the content of the directly added slag material, the ladle erosion amount for this time, and the amount of converter slag flowing into the ladle as the amount of slag already added.

[0008] In one embodiment, use Factsage to calculate the CaO-Al2O3-SiO2-MgO quaternary phase diagram, and calculate the amount of each component to be added in the slag material according to the total amount of refining slag making, the amount of slag already added, the amount of slag to be added, and the calculation results of the phase diagram; Use Factsage to calculate the CaO-Al2O3-SiO2-MgO quaternary phase diagram; Find the area where the slag melting point ≤ the tapping temperature of the steel grade - 25 °C in the CaO-Al2O3-SiO2-MgO quaternary phase diagram; Determine the amount of each component to be added in the slag material according to the maximum principle of binary basicity (CaO / SiO2) in the said region.

[0009] In one embodiment, the initial model is a gradient boosting classifier algorithm model.

[0010] In one embodiment, the ranges of the parameters in the gradient boosting classifier algorithm model are as follows: the range of the number of weak classifiers is 800 - 1200, the range of the learning rate is 0.01 - 0.02, the range of the maximum depth of the decision tree is 4, the range of the minimum number of samples for decision tree splitting is 2, and the range of the minimum number of samples for decision tree leaf nodes is 1.

[0011] In one embodiment, the process of determining the ranges of the parameters in the gradient boosting classifier algorithm model is as follows: Obtain the initial value ranges of each parameter of the gradient boosting classifier algorithm model; Use grid search or random search to traverse the initial value ranges of each parameter to form multiple parameter combinations; Perform cross - validation on each parameter combination to obtain the model performance evaluation results; Determine the optimal parameter combination according to the model performance evaluation results; use the optimal parameter combination as the final value range of the parameters of the gradient boosting classifier algorithm model.

[0012] In one embodiment, after adding the slag material with the calculated amount of each component in the slag material during the refining process of the steel grade, the method further includes: After the refining of the steel grade is completed, determine whether the refining data of this time meets the preset requirements; If it meets the preset requirements, update the refining data of this time to the refining historical data and retrain the model using the updated historical data.

[0013] In one embodiment, the preset requirements are: 1400 °C ≤ target temperature of the steel grade ≤ 1650 °C, 0 < target C of the steel grade ≤ 1%, 0 < target Si of the steel grade ≤ 1%, 0 < target Mn of the steel grade ≤ 3%, 0 < target P of the steel grade ≤ 0.1%, 0 < target S of the steel grade ≤ 0.06%, 0 < target Al of the steel grade ≤ 0.6%, refining scrap steel amount ≤ 60 t, 0 °C < temperature of refined scrap steel after baking ≤ 900 °C, 0 t < weight of converter slag flowing down ≤ 5 t, 20% ≤ CaO in converter slag ≤ 60%, 10% ≤ SiO2 in converter slag ≤ 40%, 0% < Al2O3 in converter slag ≤ 10%, 0% < MgO in converter slag ≤ 6%, 0% < MnO in converter slag ≤ 6%, 1540 °C ≤ temperature of molten steel entering the station ≤ 1650 °C, 0 < C in molten steel tapped from the converter ≤ 0.08%, 0.03% ≤ O in molten steel tapped from the converter ≤ 0.13%, 0.002% ≤ P in molten steel tapped from the converter ≤ 0.08%, S in molten steel tapped from the converter ≤ 0.15%.

[0014] In one embodiment, an initial model is trained using the historical data to obtain a prediction model, including: Dividing the historical data into multiple levels according to the target silicon content of the steel grade; Using the historical data of each level to train the initial model respectively to obtain multiple corresponding prediction models.

[0015] The solution of this embodiment has the following beneficial effects: In this embodiment, a big data model is used to predict refining slag making, and then the amount of each component to be added in the slag material is determined according to the predicted refining slag making, which can realize the intelligence of the refining process, reduce the refining time, and improve the product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic flow chart of a refining slag making method based on a big data steelmaking model under the condition of low hot metal ratio according to an embodiment of the present invention; Figure 2 It is a schematic diagram of a CaO - Al2O3 - SiO2 - MgO quaternary phase diagram calculated by Factsage according to an embodiment of the present invention; Figure 3 It is a schematic diagram of a multiple linear programming of the total slag making amount of all steel grades in LF according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The present invention will be further described in detail below in conjunction with the drawings and embodiments.

[0018] An embodiment of the present invention provides a refining slag making method based on a big data steelmaking model under the condition of low hot metal ratio, as Figure 1 shown, the method includes: Step 101: Obtain refining historical data; the historical data includes the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted, the amount of refining scrap steel, the type of refining scrap steel, the temperature of the refined scrap steel after baking, the weight of the converter slag coming down, the composition of the converter slag, the temperature of the molten steel entering the station, the composition of the molten steel tapped from the converter, and the amount of refining slag making; Step 102: Use the historical data to train an initial model to obtain a prediction model; the input of the prediction model is the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted, the amount of refining scrap steel, the type of refining scrap steel, the temperature of the refined scrap steel after baking, the weight of the converter slag coming down, the composition of the converter slag, the temperature of the molten steel entering the station, and the composition of the molten steel tapped from the converter, and the output of the prediction model is the amount of refining slag making; Step 103: Collect real - time data during the molten steel refining process; the real - time data includes the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted, the amount of refining scrap steel, the type of refining scrap steel, the temperature of the refined scrap steel after baking, the weight of the converter slag coming down, the composition of the converter slag coming down, the temperature of the molten steel entering the station, and the composition of the molten steel entering the station; Step 104: Input the real-time data into the prediction model to obtain the total refined slag amount output by the prediction model; Step 105: Calculate the slag amount already added during the molten steel refining process; Step 106: Subtract the already added slag amount from the total refined slag amount output by the prediction model to determine the slag amount to be added during the molten steel refining process; Step 107: Use Factsage to calculate the CaO - Al2O3 - SiO2 - MgO quaternary phase diagram, and calculate the amount to be added for each component in the slag material according to the total refined slag amount, the already added slag amount, the slag amount to be added, and the phase diagram calculation results; Step 108: Add the slag material in the molten steel refining process according to the calculated amount to be added for each component in the slag material.

[0019] The initial model in this embodiment is a gradient boosting classifier algorithm model (GBM).

[0020] Specifically, the gradient boosting classifier algorithm model can determine the model parameters through tuning to obtain the prediction model.

[0021] Here, the value range of each parameter of the gradient boosting classifier algorithm model can be tuned by the following steps; First, determine the value range of each parameter according to experience and data characteristics; then, use grid search (GridSearchCV) or random search (RandomizedSearchCV) to traverse the parameter space; after that, perform cross - validation on each parameter combination to evaluate the model performance; finally, select the optimal parameter combination according to the cross - validation results.

[0022] Considering that the model dataset is small, to prevent overfitting, this embodiment simultaneously performs cross - validation on 2000 groups of historical smelting data, and obtains the following optimal parameter combination through the verification results: Set random_state = 46 to control the randomness parameter and ensure the reproducibility of the experiment; to maintain the model complexity and reduce the risk of overfitting, the minimum number of samples for decision tree splitting (min_samples_split) takes the value of 2; similar to min_samples_split, to maintain the model complexity and reduce the risk of overfitting, the minimum number of samples at the decision tree leaf node (min_samples_leaf) takes the value of 1; the sample ratio (subsample) takes the value of 1, and the number of weak classifiers (n_estimators) can take values from 800 to 1200; the learning rate (learning_rate) can take values from 0.01 to 0.02; the maximum depth of the decision tree (max_depth) can take the value of 4.

[0023] In this embodiment, a big data model is used to predict the total amount of refining slag required during the refining process. Furthermore, the composition of each component in the slag material to be added can be determined based on the predicted total amount of refining slag and the amount of slag already added. As a result, intelligent control of the refining process can be achieved, reducing the refining time and improving product quality.

[0024] Specifically, the technological process of the iron and steel smelting process is generally as follows: BOF (converter) - argon station - LF (refining) - CC (continuous casting). After the converter smelting process is completed, the converter taps steel, and the molten steel enters the argon station (where the temperature of the molten steel and the location for taking composition samples will be measured). In the argon station process, a part of the slag material and alloy will be added in advance. Then the molten steel will enter the refining process. At this time, scrap steel will be added first, and then the slag material and alloy will be added continuously. After the molten steel composition is qualified, it will enter the continuous casting process.

[0025] The amount of refining slag is one of the core control parameters in the iron and steel refining process. It directly affects the metallurgical effects such as desulfurization and inclusion removal during the iron and steel refining process. Insufficient slag addition will severely limit the desulfurization efficiency, resulting in an excessive sulfur content in the finished steel. During the iron and steel refining process, the amount of slag mainly comes from the following parts: 1. Oxides formed after alloy oxidation enter the slag material.

[0026] The content of this part can be calculated by multiplying the amount of alloy already added by the proportion of alloy oxidation (which can be obtained by averaging according to experience).

[0027] 2. The amount of slag from the converter bottom tapping.

[0028] This part can be obtained by detecting the molten steel when it enters the argon station after the converter process is completed.

[0029] 3. The slag material added in the argon station and the refining process after the converter taps steel.

[0030] This part can be calculated based on the amount and composition directly added in the argon station and the refining process.

[0031] 4. Ladle erosion.

[0032] The composition of the ladle refractory can become components of the slag. Since the composition of the ladle refractory is fixed, the erosion amount of this part can be obtained by averaging according to experience.

[0033] Therefore, in practical applications, the amount of slag already added during the iron and steel refining process can be calculated in the following way: Obtain the content of oxides formed after alloy oxidation during the molten steel refining process; obtain the amount of slag material directly added after the converter taps steel; obtain the predicted amount of ladle erosion for this time based on historical data; obtain the amount of converter bottom slag detected after the converter taps steel; determine the sum of the oxide content, the directly added slag material content, the predicted amount of ladle erosion for this time, and the amount of converter bottom slag as the amount of slag already added.

[0034] The refining slag materials mainly include four component elements: CaO, Al2O3, SiO2, and MgO. Therefore, after obtaining the predicted total amount of refining slag formation using the big data model, and calculating the amount to be supplemented in refining by subtracting the amount of slag already added from the predicted total amount of refining slag formation, it is still necessary to further determine the components of each element in the slag material to be added, that is, the respective contents of CaO, Al2O3, SiO2, and MgO.

[0035] Here, a CaO-Al2O3-SiO2-MgO quaternary phase diagram can be used for determination.

[0036] Specifically, it can be determined in the following manner: Refer to Figure 2 , the Factsage software can be used to calculate the CaO-Al2O3-SiO2-MgO quaternary phase diagram (where ); find the region in the phase diagram where the slag melting point ≤ the tapping temperature of the steel grade - 25°C, , determine the mass percentage of CaO in the slag material based on the principle of the maximum binary basicity (CaO / SiO2), and the remaining is Al2O3; calculate the amount to be added for each element in the slag.

[0037] If MgO% > 8%, appropriately increase the amount of refining slag formation according to MgO = 8%; if SiO2% is greater than the maximum value of SiO2 in the target region, appropriately increase the amount of refining slag formation according to SiO2% = the maximum value of SiO2 in the target region; if SiO2% is lower than the minimum value of SiO2 in the target region, appropriately reduce the amount of refining slag formation according to SiO2% = the minimum value of SiO2 in the target region.

[0038] This embodiment uses a big data model to predict the total amount of slag formation required during the refining process, and then determines the components of each element in the slag material to be added based on the predicted total amount of refining slag formation and the current amount of slag already added. Thus, the intelligent control of the refining process can be realized, the refining time can be reduced, and the product quality can be improved.

[0039] Next, a specific embodiment will be used to illustrate this solution.

[0040] Specifically, in a scenario, the steps of this embodiment include: S1: Collect effective historical data and establish a database. The data includes the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted (C, Si, Mn, Al, P, S), the amount of refining scrap, the type of refining scrap, the temperature of the refined scrap after baking, the weight of the converter slag flowing into the ladle, the composition of the converter slag (CaO, SiO2, Al2O3, MnO, MgO, FeTOT), the temperature of the molten steel entering the station, the composition of the molten steel tapped from the converter (C, Si, Mn, Al, P, S), and the amount of refining slag formation.

[0041] S2: Historical data of the training database. Use the parameters of the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted (C, Si, Mn, Al, P, S), the amount of refining scrap steel, the type of refining scrap steel, the temperature of the refined scrap steel after baking, whether to pour the remaining slag, the weight of the converter slag, the composition of the converter slag (CaO, SiO2, Al2O3, MnO, MgO, FeTOT), the temperature of the molten steel entering the station, the composition of the molten steel entering the station (C, Si, Mn, Al, P, S) for multi-linear programming of the amount of refining slag making. After the data training is completed, a model is obtained.

[0042] S3: Production application. Input the collected real-time parameters: the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted (C, Si, Mn, Al, P, S), the amount of refining scrap steel, the type of refining scrap steel, the temperature of the refined scrap steel after baking, the weight of the converter slag, the composition of the converter slag (CaO, SiO2, Al2O3, MnO, MgO, FeTOT), the temperature of the molten steel entering the station, the composition of the molten steel entering the station (C, Si, Mn, Al, P, S) into the model, and multi-linearly program the amount of refining slag making.

[0043] S4: Obtain the existing amounts of CaO, Al2O3, SiO2, and MgO in all material sources during the steel refining process. All material sources include converter slag, added alloy oxidation, added slag materials, and ladle erosion; S5: Use Factsage software to calculate the CaO-Al2O3-SiO2-MgO quaternary phase diagram (where ); find the region in the phase diagram where the slag melting point ≤ the tapping temperature of the steel grade - 25°C, , and determine the mass percentage of CaO in the slag based on the principle of the maximum binary basicity (CaO / SiO2), with Al2O3 as the remainder; calculate the addition amounts of each component in the slag.

[0044] S6: Based on the calculation results of the slag components and combined with the amount of refining slag making obtained by multi-linear programming, calculate the addition amount of each component = the total amount to be added - the existing amount, and add the corresponding slag-making materials to make slag; S7: If there is no abnormality in the smelting and the data meets the requirements of the database, then import the smelting data of this furnace into the database by classification as historical data and conduct training regularly.

[0045] The converter specifications in this embodiment are 210 t, and the requirements for the warehousing data are as follows: 1400 °C ≤ target temperature of steel grade ≤ 1650 °C, 0 < target C of steel grade ≤ 1%, 0 < target Si of steel grade ≤ 1%, 0 < target Mn of steel grade ≤ 3%, 0 < target P of steel grade ≤ 0.1%, 0 < target S of steel grade ≤ 0.06%, 0 < target Al of steel grade ≤ 0.6%, refined scrap steel amount ≤ 60 t, 0 °C < temperature of refined scrap steel after baking ≤ 900 °C, 0 t < weight of converter slag falling ≤ 5 t, 20% ≤ CaO in converter slag ≤ 60%, 10% ≤ SiO2 in converter slag ≤ 40%, 0% < Al2O3 in converter slag ≤ 10%, 0% < MgO in converter slag ≤ 6%, 0% < MnO in converter slag ≤ 6%, 1540 °C ≤ temperature of molten steel entering the station ≤ 1650 °C, 0 < C in molten steel tapped from the converter ≤ 0.08%, 0.03% ≤ O in molten steel tapped from the converter ≤ 0.13%, 0.002% ≤ P in molten steel tapped from the converter ≤ 0.08%, S in molten steel tapped from the converter ≤ 0.15%.

[0046] In addition, the target silicon content of the steel grade in this embodiment can be divided into 3 categories: Si ≤ 0.1%, 0.1% < Si ≤ 0.3%, Si > 0.3%. The data of each classification are stored separately and linearly programmed.

[0047] See Figure 3 , which is a schematic diagram of the effect of multiple linear programming of the total slag-making amount of all steel grades in the refining process. It can be known from the figure that the sulfur content of all the furnace charges using the model for slag-making is qualified, the smelting time is shortened by 5 minutes, and the hit rate of the total slag amount of the model for slag-making ≥ 95%.

[0048] This embodiment uses a big data model to predict the refining slag-making, and then determines the amount of each component to be added in the slag materials according to the predicted refining slag-making, which can realize the intelligence of the refining process, reduce the refining time, and improve the product quality.

[0049] It should also be noted that the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or equipment including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, commodity or equipment. Without further limitation, the element defined by the statement "including a..." does not exclude the existence of another identical element in the process, method, commodity or equipment including the element.

[0050] The above are only the embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A refining slag-making method based on a big data steelmaking model under the condition of a low molten iron ratio, characterized in that The method includes: Obtaining refined historical data; the historical data includes the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted, the amount of refined scrap steel, the type of refined scrap steel, the temperature of the refined scrap steel after baking, the weight of the converter slag flowing into the ladle, the composition of the converter slag, the temperature of the molten steel entering the station, the composition of the molten steel tapped from the converter, and the amount of slag making in refining. Training an initial model using the historical data to obtain a prediction model; the input of the prediction model is the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted, the amount of refined scrap steel, the type of refined scrap steel, the temperature of the refined scrap steel after baking, the weight of the converter slag flowing into the ladle, the composition of the converter slag, the temperature of the molten steel entering the station, and the composition of the molten steel tapped from the converter, and the output of the prediction model is the amount of slag making in refining. Collecting real-time data during the refining process of molten steel; the real-time data includes the target temperature of the steel grade to be smelted, the target composition of the steel grade to be smelted, the amount of refined scrap steel, the type of refined scrap steel, the temperature of the refined scrap steel after baking, the weight of the converter slag flowing into the ladle, the composition of the converter slag flowing into the ladle, the temperature of the molten steel entering the station, and the composition of the molten steel entering the station. Inputting the real-time data into the prediction model to obtain the total amount of slag making in refining output by the prediction model. Calculating the amount of slag already added during the refining process of molten steel. Subtracting the amount of slag already added from the total amount of slag making in refining output by the prediction model to determine the amount of slag to be added during the refining process of molten steel. Calculating the amount of each component to be added to the slag material using Factsage to calculate the CaO-Al2O3-SiO2-MgO quaternary phase diagram, based on the total amount of slag making in refining, the amount of slag already added, the amount of slag to be added, and the calculation results of the phase diagram. Adding the slag material in the refining process of molten steel in the amount of each component to be added to the slag material calculated.

2. The refining slag-making method based on the big data steelmaking model under the condition of low molten iron ratio according to claim 1, wherein Calculating the amount of slag already added during the refining process of molten steel, including: Obtaining the oxide content after alloy oxidation during the refining process of molten steel. Obtaining the amount of slag material directly added after tapping from the converter. Obtaining the estimated amount of ladle erosion for this time based on historical data. Obtaining the amount of converter slag flowing into the ladle detected after tapping from the converter. Determining the sum of the oxide content, the content of the directly added slag material, the amount of ladle erosion for this time, and the amount of converter slag flowing into the ladle as the amount of slag already added.

3. The refining slag-making method based on the big data steelmaking model under the condition of low molten iron ratio according to claim 1, characterized in that, Calculating the amount of each component to be added to the slag material using Factsage to calculate the CaO-Al2O3-SiO2-MgO quaternary phase diagram, based on the total amount of slag making in refining, the amount of slag already added, the amount of slag to be added, and the calculation results of the phase diagram. Calculating the CaO-Al2O3-SiO2-MgO quaternary phase diagram using Factsage. Finding the region in the CaO-Al2O3-SiO2-MgO quaternary phase diagram where the slag melting point ≤ the tapping temperature of the steel grade - 25 °C. Determining the amount of each component to be added to the slag material based on the principle of the maximum binary basicity (CaO / SiO2) in the region.

4. The refining slag-making method based on the big data steelmaking model under the condition of low molten iron ratio according to claim 1, characterized in that, The initial model is a gradient boosting classifier algorithm model.

5. The refining slag-making method based on the big data steelmaking model under the condition of low iron-water ratio according to claim 4, characterized in that, The range of parameters in the gradient boosting classifier algorithm model is: the range of the number of weak classifiers is 800 - 1200, the range of the learning rate is 0.01 - 0.02, the range of the maximum depth of the decision tree is 4, the range of the minimum number of samples for decision tree splitting is 2, and the range of the minimum number of samples for decision tree leaf nodes is 1.

6. The refining slag-making method based on the big data steelmaking model under the condition of low iron-water ratio according to claim 4, characterized in that The process for determining the parameter range in the gradient boosting classifier algorithm model is as follows: Obtain the initial value ranges of the parameters of the gradient boosting classifier algorithm model; Use grid search or random search to traverse the initial value ranges of the parameters to form multiple parameter combinations; Perform cross-validation on each parameter combination to obtain the model performance evaluation results; Determine the optimal parameter combination according to the model performance evaluation results; use the optimal parameter combination as the final value range of the parameters of the gradient boosting classifier algorithm model.

7. The refining slag-making method based on the big data steelmaking model under the condition of low iron-water ratio according to claim 1, characterized in that After adding the slag materials with the calculated addition amounts of each component in the slag materials during the refining process of the steel grade, the method further includes: After the refining of the steel grade is completed, determine whether the refining data this time meets the preset requirements; If it meets the preset requirements, update the refining data this time to the refining historical data, and use the updated historical data to train the model again.

8. The refining slag-making method based on the big data steelmaking model under the condition of low iron-to-water ratio according to claim 7, characterized in that, The preset requirements are: 1400 °C ≤ target temperature of the steel grade ≤ 1650 °C, 0 < target C of the steel grade ≤ 1%, 0 < target Si of the steel grade ≤ 1%, 0 < target Mn of the steel grade ≤ 3%, 0 < target P of the steel grade ≤ 0.1%, 0 < target S of the steel grade ≤ 0.06%, 0 < target Al of the steel grade ≤ 0.6%, refining scrap steel amount ≤ 60 t, 0 °C < temperature of the refined scrap steel after baking ≤ 900 °C, 0 t < weight of the converter bottom slag ≤ 5 t, 20% ≤ CaO in the converter slag ≤ 60%, 10% ≤ SiO2 in the converter slag ≤ 40%, 0% < Al2O3 in the converter slag ≤ 10%, 0% < MgO in the converter slag ≤ 6%, 0% < MnO in the converter slag ≤ 6%, 1540 °C ≤ tapping temperature of the molten steel in the converter ≤ 1650 °C, 0 < C in the molten steel tapped from the converter ≤ 0.08%, 0.03% ≤ O in the molten steel tapped from the converter ≤ 0.13%, 0.002% ≤ P in the molten steel tapped from the converter ≤ 0.08%, S in the molten steel tapped from the converter ≤ 0.15%.

9. The refining slag-making method based on the big data steelmaking model under the condition of low iron-water ratio according to claim 1, characterized in that Training the initial model using the historical data to obtain a prediction model, including: Divide the historical data into multiple levels according to the target silicon content of the steel grade; Use the historical data of each level to train the initial model respectively to obtain multiple corresponding prediction models.

Images