A method for adding alloy to a converter process based on dynamic adjustment of production conditions

Abstract

The present invention provides a method for adding alloys to a converter process based on dynamic adjustment of production conditions, and belongs to the technical field of converter production in the iron and steel metallurgical industry. The alloy adding method comprises collecting alloy detection data, establishing an alloy database based on alloy composition; collecting converter production data, establishing a prediction model database; performing data cleaning on the collected converter production data; establishing a neural network prediction model for the yield of each element; training each element yield prediction model separately; calculating the amount of alloy addition according to the alloy detection composition and the predicted alloy yield of each element prediction model. The present invention solves the problem of adding converter alloys according to operating experience at the production site through a prediction model based on a neural network, improves the efficiency of alloy use, and reduces alloy consumption to a certain extent. At the same time, the use of a fully automatic prediction model is conducive to the solidification of the production process and provides a basis for future intelligentization.

Description

Technical Field

[0001] The invention relates to the field of converter production in steel mills, and in particular to a converter process alloy adding method based on dynamic adjustment of production conditions, which reduces the amount of alloy added in the converter production process while meeting production requirements. Background Art

[0002] Alloying in the converter process of steelmaking is an important measure to ensure that the product composition meets the requirements. Generally, the amount of alloy added in each furnace is determined based on the converter experience. However, due to the differences in molten steel weight, molten steel composition, molten steel oxygen content, nitrogen content, alloy quality, etc., the alloy yield of different furnaces in the same batch of production fluctuates greatly. In order to ensure the minimum requirements for product composition, the amount of alloy added must be increased to a certain extent, which not only causes a large consumption of alloy in the converter process, but also causes pressure on the alloying operation in the refining process.

[0003] Chinese patent CN 107217120 B discloses a converter alloy addition control method, including the following steps: establishing a database; setting the standard number of reference furnaces; searching for a reference furnace with the same index parameters as the current converter from the database, and selecting the alloy addition scheme based on the reference furnace data or the operator's experience; after the product is qualified, the data of this converter is added to the database as a reference furnace. The data of the reference furnace of the present invention are all successful experiences, which supplement the field data that may be lacking in actual performance, eliminate the impact of the missing blow stop component and the missing free oxygen data on the algorithm during the LP algorithm calculation, and when the LP algorithm fails, the final alloy addition scheme is obtained by directly using the reference furnace data weighting, which can also guarantee the quality of molten steel to a certain extent, and greatly avoid the situation where the alloy calculation model has no solution.

[0004] Chinese patent CN 109897931 B discloses an alloy addition optimization method for converter process production of steel with large alloy content of easily oxidized elements. The technical solution is: based on the requirements of special steel composition, easily oxidized alloys are alloyed in batches in the converter production process, and the converter double slag control technology and the selective oxidation characteristics in the furnace are used to realize the molten steel temperature compensation mechanism, the alloy preheating in the ladle, the alloying of steel tapping, and the optimization and adjustment of the refined alloy. The large alloy addition process in the converter smelting process is combined with alloying to prevent the easily oxidized element alloying converter rephosphorization, compensate for the temperature loss of batch alloy molten steel, and weaken the influence of batch alloy on process connection and refining treatment. Summary of the invention

[0005] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art and provide a method for adding alloys to a converter process based on dynamic adjustment of production conditions, so as to solve the problem that there are many factors affecting the alloy yield, the alloy yield is low and unstable, improve the alloy utilization efficiency of the converter process, and reduce the amount of alloy added.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] A method for adding alloy to a converter process based on dynamic adjustment of production conditions, the method comprising the following steps:

[0008] 1) Collect alloy test data and establish an alloy database based on alloy composition. The alloy database is constructed according to the alloy type, the alloy composition is inspected according to the batch, and the inspection data is used as the basis for calculating the subsequent alloy addition amount. Among them, the alloy composition is divided into effective element composition and invalid element composition. The effective element composition of the alloy test data is the element involved in alloying in the alloy, that is, the effective element of high manganese alloy and medium manganese alloy is manganese, the effective element of medium chromium alloy is chromium, the effective elements of vanadium nitrogen alloy are vanadium and nitrogen, the effective element of vanadium-iron alloy is vanadium, the effective element of ferrosilicon alloy is silicon, the effective element of niobium-iron alloy is niobium, the effective element of nickel plate alloy is nickel, and the effective element of metal manganese alloy is manganese; the invalid element composition of the alloy includes phosphorus content, sulfur content, oxygen content, and nitrogen content (except vanadium-nitrogen alloy).

[0009] 2) Collect converter production data and establish a prediction model database. The converter production data includes converter production input data and refining production output data, wherein the converter production input data includes converter endpoint manganese content, carbon content, phosphorus content, temperature, oxygen content, and converter age; the refining production output data is the content of each alloy element in molten steel.

[0010] 3) Clean the collected converter production data to remove abnormal data caused by human error, equipment failure, etc. The abnormal data evaluation is shown in formula (1):

[0011] or

[0012] Among them, X i is the i-th data, X max is the maximum value in this type of data, X min The minimum value in this type of data.

[0013] 4) The alloy components are divided into effective element components and ineffective element components, and the converter production input data are used as the prediction model input nodes, and the content of each alloy element in the molten steel is used as the output node, and a prediction model for the yield of each element based on a radial basis function neural network is established. The prediction model based on the radial basis function neural network has a three-layer structure, and the number of nodes in the middle layer is determined by the size of the mean square error between the predicted value and the measured value.

[0014] 5) The yield prediction model of each element is trained separately.

[0015] 6) Collecting on-site converter production data and alloy inspection data, comparing the predicted data with the actual data, and adjusting the original prediction model. The on-site converter production data and alloy inspection data are automatically captured directly from the secondary production system and alloy inspection system through an open database connection, and are cleaned by the method described in (3).

[0016] 7) According to the alloy test composition, combined with the predicted alloy yield of each element prediction model, calculate the amount of alloy added and guide the actual production on site. The alloy addition amount is calculated as shown in formula (2):

[0017]

[0018] Among them, M i is the amount of element added, wt i The content of the i-th element in the converter process is increased, R i is the alloy yield, R Ni is the recovery rate of invalid element components in the ith alloy, n is the number of invalid element component types, which is 4 in this case, and wtNi is the increase in invalid element components in the ith alloy.

[0019] 8) By taking samples to inspect the composition of molten steel during the refining process, determine whether the amount of alloy added is accurate. If it is accurate, proceed to the next round of production. If the deviation is large, perform model optimization.

[0020] The alloy addition method of the present invention comprises a series of processes including collecting alloy detection data, establishing an alloy database based on alloy composition; collecting converter production data, establishing a prediction model database; performing data cleaning on the collected converter production data; establishing a neural network element yield prediction model; training each element yield prediction model separately; calculating the alloy addition amount according to the alloy detection composition and the predicted alloy yield of each element prediction model, and the like.

[0021] Compared with the prior art, the beneficial effects of the above technical solution are:

[0022] (1) Provide a converter production alloy addition prediction model that can meet dynamic adjustment requirements and provide technical support for alloy addition at the production site.

[0023] (2) Through the prediction model based on neural network, the problem of adding converter alloys according to operating experience at the production site is solved, the efficiency of alloy use is improved, and the alloy consumption is reduced to a certain extent.

[0024] (3) The above method can improve the stability of the element content at the end point of the molten steel, providing technical guarantee for the performance stability of subsequent rolling production.

[0025] (4) The use of a fully automatic prediction model is conducive to solidifying the production process and providing a foundation for future intelligentization. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0027] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] Example 1

[0029] A steel mill produces normalized rolled weldable fine-grained structural steel plates (European standard S355NL). In order to meet the requirements for the content of various elements in the product, ferrosilicon, ferroniobium, nickel plates, and metallic manganese are added with the steel flow when the steel is tapped from the converter. The composition of S355NL can be found in relevant standards (such as EN10025-3). In the original production, the production staff calculated the amount of alloy added according to production experience and added it. However, it was found in actual production that this method has the problem of large fluctuations in the content of various elements after the steel is tapped from the converter. In order to ensure that all batches meet the requirements, the amount of various alloys added must be increased, which will not only increase the alloy consumption in the converter process, but also increase the operating pressure of alloying in the refining process.

[0030] To solve the above problems, the present invention provides a method for adding alloys to a converter process based on dynamic adjustment of production conditions, which solves the problem of many factors affecting alloy yield, low and unstable alloy yield, improves the alloy utilization efficiency of the converter process, and reduces the amount of alloy added. Figure 1 As shown in the process, the implementation plan is as follows:

[0031] (1) Collect alloy test data and establish an alloy database based on alloy composition. The alloy database is constructed according to the alloy type, the alloy composition is tested according to the batch, and the test data is used as the basis for calculating the subsequent alloy addition amount. Among them, the alloy composition is divided into effective element composition and ineffective element composition. The effective element composition of the alloy test data is the element involved in alloying in the alloy, that is, the effective element of ferrosilicon alloy is silicon, the effective element of ferroniobium alloy is niobium, the effective element of nickel plate alloy is nickel, and the effective element of metal manganese alloy is manganese; the ineffective element composition of the alloy includes phosphorus content, sulfur content, oxygen content, and nitrogen content.

[0032] (2) Collecting converter production data and establishing a prediction model database. The converter production data includes converter production input data and refining production output data, wherein the converter production input data includes converter endpoint manganese content, carbon content, phosphorus content, temperature, oxygen content, and converter age; the refining production output data is the content of each alloy element in molten steel.

[0033] (3) Clean the collected converter production data to remove abnormal data caused by human error, equipment failure, etc. The abnormal data evaluation is shown in formula (1):

[0034] or

[0035] Among them, X i is the i-th data, X max is the maximum value in this type of data, X min The minimum value in this type of data.

[0036] (4) The alloy components are divided into effective element components and ineffective element components, and the converter production input data are used as the prediction model input nodes, and the content of each alloy element in the molten steel is used as the output node to establish a prediction model for the yield of each element based on a radial basis function neural network. The prediction model based on the radial basis function neural network has a three-layer structure, and the number of nodes in the middle layer is determined by the size of the mean square error between the predicted value and the measured value.

[0037] (5) The prediction models for the yield of each element were trained separately. The production data of 12583 120t converters in the plant were collected.

[0038] (6) Collect on-site converter production data and alloy inspection data, compare the predicted data with the actual data, and adjust the original prediction model. The on-site converter production data and alloy inspection data are automatically captured directly from the secondary production system and alloy inspection system through an open database connection, and are cleaned by the method described in (3). After cleaning, 10342 furnace data are obtained.

[0039] (7) According to the alloy test composition, combined with the predicted alloy yield of each element prediction model, the alloy addition amount is calculated to guide the actual production on site. The alloy addition amount is calculated as shown in formula (2):

[0040]

[0041] Among them, M i is the amount of element added, wt i The content of the i-th element in the converter process is increased, R i is the alloy yield, R Ni is the recovery rate of invalid element components of the i-th alloy, n is the number of invalid element component types, which is 4 in this case. wtNi is the increase in invalid element components in the i-th alloy.

[0042] (8) By taking samples to inspect the composition of molten steel during the refining process, it is determined whether the amount of alloy added is accurate. If it is accurate, the next round of production will be carried out. If the deviation is large, the model will be optimized.

[0043] Of the 10342 furnace data obtained after cleaning, 60%, i.e., 6205 furnace data, were selected for model training, and 40% of the cleaned data, i.e., 4137 furnace data, were used to verify and optimize the model to improve the prediction effect.

[0044] To evaluate the prediction accuracy of different models, the root mean square error and correlation coefficient, which are the most widely used performance evaluation indicators, are selected as evaluation indicators.

[0045] The root mean square error represents the deviation between the predicted value and the true value. The smaller the value, the closer the predicted value is to the true value, indicating that the model prediction effect is better. The correlation coefficient represents the ratio of the regression sum of squares to the total sum of squares. The value range is [0, 1]. The closer the value is to 1, the better the model prediction effect is. Table 1 shows the evaluation of the prediction model values ​​and verification values ​​of the four alloys.

[0046] Table 1 Evaluation of prediction model values ​​and verification values ​​for four alloys

[0047]

[0048] It can be seen from Table 1 that the predicted values ​​of each alloy have good prediction performance compared with the measured values.

[0049] In order to verify the effect of the method of the present invention in the actual S355NL steel converter production, 1 group (serial number 0) of alloy additions in the original process and 10 groups (serial numbers 1-10) of alloy additions and costs per ton of steel using the method of the present invention were selected for comparison. The specific data are shown in Table 2.

[0050] Table 2 Comparison of alloy addition amount and ton steel cost between the original process and the method of the present invention

[0051]

[0052] From the data comparison in Table 2, it can be seen that the average addition amount of ferrosilicon per ton of steel is reduced by 0.20kg, a reduction of 11.48%, the average addition amount of ferroniobium per ton of steel is reduced by 0.07kg, a reduction of 12.63%, the average addition amount of nickel plate is reduced by 0.18kg, a reduction of 10.34%, the average addition amount of metallic manganese is reduced by 0.40kg, a reduction of 3.24%, the total alloying alloy addition amount is reduced by 0.85kg, and the cost per ton of steel is reduced by RMB 59.89 / t, indicating that the method of the present invention can reduce the alloy addition amount of converter alloying, and has good economic benefits.

[0053] Example 2

[0054] A steel mill produces Q550MD steel plates. In order to meet the requirements for the content of various elements in the product, medium manganese, medium chromium, ferroniobium, ferrovanadium and metallic manganese are added with the steel flow when the converter is tapped. The composition of Q550MD can be found in GB / T 1591-2018. In the original production, the production staff calculated the amount of alloy added according to production experience and added it. However, it was found in actual production that this method has the problem of large fluctuations in the content of various elements after the converter is tapped. In order to ensure that all batches meet the requirements, the amount of various alloys added must be increased, which will not only increase the alloy consumption in the converter process, but also increase the operating pressure of alloying in the refining process.

[0055] To solve the above problems, the present invention provides a method for adding alloys to a converter process based on dynamic adjustment of production conditions, which solves the problem of many factors affecting alloy yield, low and unstable alloy yield, improves the alloy utilization efficiency of the converter process, and reduces the amount of alloy added. Figure 1 As shown in the process, the implementation plan is as follows:

[0056] (1) Collect alloy test data and establish an alloy database based on alloy composition. The alloy database is constructed according to the alloy type, the alloy composition is tested according to the batch, and the test data is used as the basis for calculating the subsequent alloy addition amount. Among them, the alloy composition is divided into effective element composition and invalid element composition. The effective element composition of the alloy test data is the element involved in alloying in the alloy, that is, the effective element of the medium manganese alloy is manganese, the effective element of the medium chromium alloy is chromium, the effective element of the niobium-iron alloy is niobium and iron, and the effective element of the vanadium-iron alloy is vanadium and iron; the effective element of the metal manganese alloy is manganese, and the invalid element composition of the alloy includes phosphorus content, sulfur content, oxygen content, and nitrogen content.

[0057] (2) Collecting converter production data and establishing a prediction model database. The converter production data includes converter production input data and refining production output data, wherein the converter production input data includes converter endpoint manganese content, carbon content, phosphorus content, temperature, oxygen content, and converter age; the refining production output data is the content of each alloy element in molten steel.

[0058] (3) Clean the collected converter production data to remove abnormal data caused by human error, equipment failure, etc. The abnormal data evaluation is shown in formula (1):

[0059] or

[0060] Among them, X i is the i-th data, X max is the maximum value in this type of data, X min The minimum value in this type of data.

[0061] (4) The alloy components are divided into effective element components and ineffective element components, and the converter production input data are used as the prediction model input nodes, and the content of each alloy element in the molten steel is used as the output node to establish a prediction model for the yield of each element based on a radial basis function neural network. The prediction model based on the radial basis function neural network has a three-layer structure, and the number of nodes in the middle layer is determined by the size of the mean square error between the predicted value and the measured value.

[0062] (5) The prediction models for the yield of each element were trained separately. The production data of 214 120t converters in the plant were collected.

[0063] (6) Collect on-site converter production data and alloy inspection data, compare the predicted data with the actual data, and adjust the original prediction model. The on-site converter production data and alloy inspection data are automatically captured directly from the secondary production system and alloy inspection system through an open database connection, and are cleaned by the method described in (3). After cleaning, 197 furnace data are obtained.

[0064] (7) According to the alloy test composition, combined with the predicted alloy yield of each element prediction model, the alloy addition amount is calculated to guide the actual production on site. The alloy addition amount is calculated as shown in formula (2):

[0065]

[0066] Among them, M i is the amount of element added, wt i The content of the i-th element in the converter process is increased, R i is the alloy yield, R Ni is the recovery rate of invalid element components of the i-th alloy, n is the number of invalid element component types, which is 4 in this case. wtNi is the increase in invalid element components in the i-th alloy.

[0067] (8) By taking samples to inspect the composition of molten steel during the refining process, it is determined whether the amount of alloy added is accurate. If it is accurate, the next round of production will be carried out. If the deviation is large, the model will be optimized.

[0068] Of the 197 furnace data obtained after cleaning, 60%, i.e., 118 furnace data, were selected for model training, and 40% of the cleaned data, i.e., 79 furnace data, were used to verify and optimize the model to improve the prediction effect.

[0069] To evaluate the prediction accuracy of different models, the root mean square error and correlation coefficient, which are the most widely used performance evaluation indicators, are selected as evaluation indicators.

[0070] The root mean square error represents the deviation between the predicted value and the true value. The smaller the value, the closer the predicted value is to the true value, indicating that the model prediction effect is better. The correlation coefficient represents the ratio of the regression sum of squares to the total sum of squares. The value range is [0, 1]. The closer the value is to 1, the better the model prediction effect is. Table 3 shows the evaluation of the prediction model values ​​and verification values ​​of the five alloys.

[0071] Table 3 Evaluation of prediction model values ​​and verification values ​​for five alloys

[0072]

[0073] It can be seen from Table 3 that the predicted values ​​of each alloy have good prediction performance compared with the measured values.

[0074] In order to verify the effect of the method of the present invention in the actual Q550MD steel converter production, 1 group (serial number 0) of original process alloy addition and 10 groups (serial numbers 1-10) of alloy addition and cost per ton of steel using the method of the present invention were selected for comparison. The specific data are shown in Table 4.

[0075] Table 4 Comparison of alloy addition amount and steel cost per ton between the original process and the method of the present invention

[0076]

[0077]

[0078] From the data comparison in Table 4, it can be seen that the average addition amount of medium manganese per ton of steel is reduced by 0.63kg, a reduction of 7.60%; the average addition amount of medium chromium per ton of steel is reduced by 0.32kg, a reduction of 9.00%; the average addition amount of ferroniobium is reduced by 0.01kg, a reduction of 5.40%; the average addition amount of ferrovanadium is reduced by 0.03kg, a reduction of 9.30%; the average addition amount of metallic manganese is reduced by 0.52kg, a reduction of 7.40%, the total alloying alloy addition amount is reduced by 1.51kg, and the cost per ton of steel is reduced by 24.09 yuan / t, indicating that the method of the present invention can reduce the alloy addition amount of converter alloying, and has good economic benefits.

[0079] The process parameters (such as temperature, time, etc.) of the present invention can realize the method by taking upper and lower limits and interval values, and the embodiments are not listed one by one here.

[0080] Any contents not described in detail in the present invention can be based on the conventional technical knowledge in the art.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention is described in detail with reference to the embodiments, it should be understood by those skilled in the art that any modification or equivalent replacement of the technical solutions of the present invention does not depart from the spirit and scope of the technical solutions of the present invention and should be included in the scope of the claims of the present invention.

Claims

1. A method for adding alloys to a converter process based on dynamic adjustment of production conditions, the method comprising the following steps: 1) Collect alloy test data and establish an alloy database based on alloy composition; 2) Collect converter production data and establish a prediction model database; 3) Clean the collected converter production data and remove abnormal data; The abnormal data evaluation is shown in formula (1): or Among them, X i is the i-th data, X max is the maximum value in this type of data, X min is the minimum value in this type of data; 4) Taking the effective element composition and ineffective element composition of the alloy composition and the converter production input data as the input nodes of the prediction model, taking the content of each alloy element in the molten steel as the output node, a prediction model for the yield rate of each element based on the radial basis function neural network is established; 5) The yield prediction models of each element are trained separately; 6) Collect on-site converter production data and alloy inspection data, compare the predicted data with the actual data, and adjust the original element yield prediction model; 7) According to the alloy detection composition, combined with the predicted alloy yield of each element prediction model, the alloy addition amount is calculated, and the actual production on site is guided; the alloy addition amount is calculated as shown in formula (2): Among them, M i is the amount of element added, wt i The content of the i-th element in the converter process is increased, R i is the alloy yield, R Ni is the recovery rate of invalid element components in the i-th alloy, n is the number of invalid element component types, and wtNi is the increase in invalid element components in the i-th alloy; 8) By taking samples to inspect the composition of molten steel during the refining process, determine whether the amount of alloy added is accurate. If it is accurate, proceed to the next round of production. If the deviation is large, perform model optimization.

2. The method for adding alloy to a converter process based on dynamic adjustment of production conditions according to claim 1, characterized in that: In the step 1), an alloy database is constructed according to the alloy types, alloy composition is tested according to batches, and the test data is used as a basis for calculating the subsequent alloy addition amount.

3. The method for adding alloy to a converter process based on dynamic adjustment of production conditions according to claim 1, characterized in that: In step 1), the alloy composition is divided into effective element composition and ineffective element composition; the effective element composition of the alloy detection data is the element involved in alloying in the alloy; the ineffective element composition of the alloy includes phosphorus content, sulfur content, oxygen content, and nitrogen content.

4. The method for adding alloy to a converter process based on dynamic adjustment of production conditions according to claim 1, characterized in that: In step 1), the alloy is one or more of high manganese alloy, medium manganese alloy, medium chromium alloy, vanadium nitrogen alloy, vanadium iron alloy, ferrosilicon alloy, ferroniobium alloy, nickel plate alloy and metal manganese alloy.

5. The method for adding alloy to a converter process based on dynamic adjustment of production conditions according to claim 4, characterized in that: The effective element of high manganese alloy and medium manganese alloy is manganese, the effective element of medium chromium alloy is chromium, the effective element of vanadium-nitrogen alloy is vanadium and nitrogen, the effective element of ferrovanadium alloy is vanadium, the effective element of ferrosilicon alloy is silicon, the effective element of ferroniobium alloy is niobium, the effective element of nickel plate alloy is nickel, and the effective element of metallic manganese alloy is manganese.

6. The method for adding alloy to a converter process based on dynamic adjustment of production conditions according to claim 1, characterized in that: In step 2), the converter production data includes converter production input data and refining production output data.

7. The method for adding alloy to a converter process based on dynamic adjustment of production conditions according to claim 6, characterized in that: The converter production input data include converter endpoint manganese content, carbon content, phosphorus content, temperature, oxygen content, and converter age; the refining production output data are the contents of various alloy elements in molten steel.

8. The method for adding alloy to a converter process based on dynamic adjustment of production conditions according to claim 1, characterized in that: In step 3), the prediction model based on radial basis function neural network has a three-layer structure, and the number of nodes in the middle layer is determined by the size of the mean square error between the predicted value and the measured value.

9. The method for adding alloy to a converter process based on dynamic adjustment of production conditions according to claim 1, characterized in that: The on-site converter production data and alloy inspection data are automatically captured directly from the secondary production system and alloy inspection system through an open database connection, and data cleaning is performed using the method described in step 3).

10. The method for adding alloy to a converter process based on dynamic adjustment of production conditions according to claim 1, characterized in that: To evaluate the prediction accuracy of different models, root mean square error and correlation coefficient were selected as evaluation indicators.

Images