Method for predicting residual carbon content of molten pool of steelmaking converter

Abstract

The invention relates to a method for predicting the residual carbon content of a molten pool of a steelmaking converter, which comprises the following steps of: firstly, calculating theoretical oxygen consumption, acquiring actual oxygen consumption, and setting a decarburization stage dynamic division mechanism; secondly, corresponding dynamic correction coefficients K are generated according to the difference of decarburization modes of the technological processes of TSC and TSO in different stages of the smelting period; and finally, performing dynamic correction optimization on the K value by using an optimization iterative algorithm, and performing real-time prediction application. According to the method, non-contact continuous inversion of the global carbon content of the molten pool is realized by monitoring the converter mouth flue gas component parameters at the converter blowing last stage in real time and combining a carbon-oxygen reaction mechanism model and a machine learning algorithm. The method is suitable for accurate control of the molten steel carbon content in the converter smelting link, can replace or assist a traditional sublance spot measurement technology, remarkably improves the end point carbon hit rate, reduces the over-blowing / under-blowing risk, and provides powerful support for accurate smelting control.

Description

Technical Field

[0001] This patent application belongs to the field of intelligent detection technology in the iron and steel metallurgical process, and more specifically, it relates to a method for predicting the residual carbon content in the molten pool of a steelmaking converter. Background Technology

[0002] In the converter smelting process of iron and steel, the final carbon content of the molten pool is a key process parameter affecting steel quality, alloy yield, furnace lining life, and product qualification rate. Achieving precise control of the final carbon content plays a decisive role in reducing production costs, improving production efficiency, ensuring steel purity, and meeting the performance indicators of high-end steel products. Currently, the main methods for controlling the carbon content at the converter smelting endpoint include secondary lance detection technology, bomb-type measurement, and flue gas analysis.

[0003] Mechanical probe insertion measurement (secondary gun technology) is a commonly used measurement method that obtains point-based data such as temperature and carbon content by directly inserting the probe into the molten pool. Its advantage lies in the intuitiveness of the measured values, allowing direct reading of the chemical composition at a specific location in the molten pool. However, this technology has significant temporal and spatial limitations; a single measurement only reflects transient, localized data and cannot capture the dynamic changes of the entire molten pool. Operationally, it relies heavily on experience to determine the timing of insertion; inserting too early or too late will lead to a decrease in endpoint hit rate, and frequent probe contact with molten steel will accelerate equipment wear and tear.

[0004] The bomb-type probe measurement uses an automatic throwing device to drop the probe into the molten pool, acquiring data such as temperature and carbon-oxygen content within seconds, with an effective measurement success rate of over 95%. Compared to mechanical probes, its advantage lies in not requiring furnace shutdown or reduction of oxygen flow, thus shortening the smelting cycle. However, it is essentially still a point-measurement technology, with limitations similar to that of the secondary probe. The data still represents the instantaneous state, and the probe must be completely immersed in the molten pool to obtain accurate readings. Intense turbulence in the molten pool may affect measurement stability. Furthermore, the probe is a disposable item, increasing the cost per measurement.

[0005] In recent years, the steel smelting industry has increasingly urgent needs for high real-time performance, continuous stability, and low-cost operation in its production processes. Against this backdrop, flue gas analysis, which accurately addresses these demands, is seeing its application adoption rate continuously increase, and related technological research has led to a series of optimized application solutions. The core logic of these solutions lies in deploying flue gas analyzers at key locations in the converter flue to monitor the composition of the furnace gas in real time and continuously, thereby inferring the dynamic reaction state inside the molten pool. As a typical non-contact measurement technology, flue gas analysis can accurately capture core process parameters such as the CO / CO2 ratio throughout the entire process, providing key technical support for achieving continuous monitoring of the entire steel smelting process and effectively helping the industry balance production efficiency and cost control requirements. However, certain technical shortcomings still exist in practical application scenarios.

[0006] Chinese patent "A Method for Determining the Decarburization Endpoint in Steelmaking Using Converter Gas" (application number: CN202411272108.6) discloses a method for determining the decarburization endpoint in steelmaking based on flue gas composition calculations. Its core is to dynamically calculate and predict the carbon content of molten steel by real-time monitoring of the volume fraction molars of CO2 and CO in the converter gas during the steelmaking process, using the linear functional relationship between this data and the carbon content of the molten steel, thereby determining the decarburization endpoint. The core logic of this method relies solely on the theoretical principle of mass conservation, namely, the total molar mass of carbon-containing gases (mainly CO and CO2) in the flue gas per unit time is equal to the molar mass of carbon reduced in the molten steel during the same period. However, because it does not consider the fluctuations in flue gas flow rate due to various process conditions, equipment acquisition errors, and detection time lags in actual production, the final calculation results will have significant deviations, affecting the actual production performance.

[0007] Chinese patent "Method and Apparatus, Storage Medium and Electronic Equipment for Carbon Content Detection in Converters" (application number: CN202211064964.3) discloses a method that utilizes a secondary lance and a flue gas analysis system in synergy. By calibrating the time lag between the two, a high-frequency, dynamic carbon content calculation is performed starting at a specific time point. While this method considers the time lag issue, it lacks consideration for nonlinear variations in the smelting process and fails to effectively balance the prediction accuracy deviation caused by non-time lag.

[0008] Chinese patent "A Correction Method and System for a Converter Endpoint Carbon Content Prediction Model" (application number: CN202010328795.4) discloses a correction scheme for a converter endpoint carbon content prediction model based on flue gas analysis: the ratio of the difference between the carbon content of molten iron entering the furnace and the carbon content measured by the auxiliary lance to the total decarburization amount is used as the correction coefficient. Combined with the actual data measured by the auxiliary lance during the TSC stage, the carbon content at the TSO endpoint is predicted. However, this method has obvious limitations: on the one hand, in actual production, the time from the TSC auxiliary lance measurement to the TSO endpoint is short, and there are few variables affecting the carbon content. The main carbon reduction period in the converter is concentrated in the TSC stage, and this method does not predict the carbon content of the molten pool for this main decarburization process; on the other hand, the calculation method of its correction coefficient has the possibility of accidental equality, and it does not have the ability to decide the correction coefficient based on characteristic changes.

[0009] Given the many shortcomings of traditional methods in predicting the remaining carbon content in the converter molten pool, it is necessary to develop a method for predicting the remaining carbon content in the converter molten pool based on flue gas analysis and the principle of dynamic carbon conservation. Summary of the Invention

[0010] The technical problem to be solved by this invention is to provide a method for predicting the residual carbon content in the molten pool of a steelmaking converter. Traditional flue gas analysis methods are limited by flow fluctuations, hysteresis errors, and the lack of a dynamic correction mechanism in predicting the end point of converter decarburization, which affects the accuracy of carbon content prediction and is not conducive to the refined control of carbon content in molten steel.

[0011] To solve the above problems, the technical solution adopted by the present invention is as follows:

[0012] A method for predicting the residual carbon content in the molten pool of a steelmaking converter includes the following steps:

[0013] S1. Calculate the theoretical oxygen consumption, obtain the actual oxygen consumption, and set up a dynamic division mechanism for the decarbonization stage.

[0014] S2. Based on the differences in decarburization methods between TSC and TSO processes at different stages of the smelting cycle, corresponding dynamic correction coefficients K are generated.

[0015] S3. Use an optimization iterative algorithm to dynamically adjust and optimize the K value, and predict the application in real time.

[0016] Furthermore, step S1 specifically includes the following:

[0017] S11. Let the current accumulated actual oxygen consumption be Qactual O2, and the theoretical oxygen consumption be Qtheory O2. The theoretical oxygen consumption Qtheory O2 is determined by the initial carbon content [C]0 of the molten iron and the weight W of the molten iron. HM Scrap steel weight W FG , Carbon contribution of scrap steel ΔC scrap The results were obtained through material balance (MB) and heat balance (HB) calculations:

[0018] Qtheory O2=f([C]0, W HM W FG ,ΔC scrap (Iron temperature T0 and iron composition)

[0019] S12. Based on the ratio of actual oxygen consumption to theoretical oxygen consumption To define the decarburization stage in the smelting process:

[0020]

[0021] Based on this classification, in the early stage of smelting when η < 30%, silicon / manganese oxidation is the dominant process. At this time, the decarburization reaction is weak and the carbon reduction rate is low but gradually increases.

[0022] When the carbon content is 30%≤η≤80% during the middle stage of smelting, the decarburization reaction is dominant. At this time, the reaction is intense and the carbon reduction rate is fast, which is the main stage of the decarburization reaction.

[0023] When η > 80%, it is already the decarbonization period of slowing down. By slowly reducing carbon emissions, the timing of stopping the blowing should be determined.

[0024] Furthermore, the ratio η = (30%, 80%) can be dynamically adjusted by ±5% to ±10% according to actual needs to clarify the time range of the main decarbonization period.

[0025] Furthermore, step S2 specifically includes the following:

[0026] S21, the stage of the smelting cycle is TSC, and a dynamic correction coefficient K is constructed. TSC The smelting cycle stage is TSO, and a dynamic correction coefficient K is constructed. TSO K TSC With K TSO The processing procedure is the same;

[0027] S22, with K TSC For example, K TSC The computational optimization logic is as follows:

[0028] This can be achieved by obtaining the time series data of historical furnace cycles: flue gas flow rate Q. g (t), carbon monoxide concentration C CO (t), carbon dioxide concentration C CO2 (t), oxygen flow rate F O2 (t) and static parameters of molten iron (including initial carbon content [C]0, molten iron weight W) HM Scrap steel weight W FG 1. Overall carbon content of scrap steel [C] FG Using this basic information in conjunction with static parameter material conservation calculations, the initial carbon content [C]0 of the molten pool can be calculated:

[0029] [C]0=((W HM ·[C]0) + (W FG ·[C] FG )) ·10 -2 ·10 6

[0030] Throughout the entire smelting process of a single furnace, the concentration of carbon monoxide (C) varies. CO (t), carbon dioxide concentration C CO2 From the time-series data of (t), through mass conservation calculations, the cumulative decarburization amount ΔC at each data sampling frequency can be obtained from the start to the end of smelting. decarb :

[0031]

[0032] At the moment of the secondary gun under TSC, a cumulative decarburization amount ΔC can be obtained through the above method. decarb Also called total decarbonization, here ΔC decarb This represents the mass of carbon contained in carbon monoxide and carbon dioxide in the flue gas during converter smelting, measured in tons. The remaining carbon content [C] in the molten pool can be calculated from the difference between the initial carbon content [C]0 in the molten pool and the total decarburization amount. pred :

[0033]

[0034] Here The calculated result is a percentage content, representing the carbon content removed relative to the total carbon content in the molten steel.

[0035] The calculation at this point is entirely based on the law of conservation of mass. However, in actual production, there are data lags and nonlinear interference factors. Therefore, a dynamic correction coefficient K needs to be introduced to balance the data anomalies caused by errors.

[0036]

[0037] By iteratively calculating historical time-series data and using the bisection method to solve for the dynamic correction coefficient K, the remaining carbon content [C]K pred in the molten pool after equilibrium can be used to measure the true carbon content [C] in the molten pool using the secondary lance. target Within ±5% (the target offset accuracy can be adjusted according to the actual situation; 5% is used here to consider the computational complexity and number of iterations), this is the optimal dynamic correction coefficient K for the current smelting furnace:

[0038]

[0039] Since this occurs during the TSC stage of the smelting cycle, which is the period from the start of oxygen blowing to the first measurement at the lower lance in converter smelting, the dynamic correction coefficient K calculated during this stage is referred to here as K. TSC .

[0040] Furthermore, in step S3, the dynamic adjustment and optimization mechanism for the K value and the real-time prediction mechanism are as follows:

[0041] S31. When the blowing begins, start collecting real-time measurement data from the flue gas analyzer, such as carbon monoxide, carbon dioxide, and flue gas flow rate. Combine this with the decarbonization stage division method and closely monitor the changes in the oxygen blowing stage.

[0042] S32. When the oxygen blowing progress reaches 75%, start the online optimization program to extract feature vector values; find the optimal value through Euclidean distance matching and optimize the dynamic correction coefficient K.

[0043] When η=75%, the online optimization program is initiated; the curve characteristics within the current furnace primary decarburization time range are extracted, and ΔC is analyzed. decarb The t-curve is piecewise fitted, and the slope β and intercept ΔC0 are extracted; the feature vector V of the current curve is calculated. current =[β,ΔC0] and the i-th furnace V in the historical feature database i Euclidean distance:

[0044]

[0045] Select D min K corresponding to the furnace number opt As the dynamic correction coefficient for this furnace cycle, D min K represents the minimum Euclidean distance. opt This is a correction factor used to compensate for or correct system measurement errors and process dynamic control errors in online monitoring equipment during normal smelting. In normal steel smelting, the difference between the initial carbon mass of molten iron and the final carbon mass of molten steel is theoretically approximately equal to the total decarburization mass of carbon monoxide and carbon dioxide in the flue gas. Based on practical industrial production experience, the error between this theoretical calculation and the actual decarburization amount is usually controlled within ±5%. The sources of error mainly include: system measurement errors of online monitoring equipment (such as drift in flue gas composition analyzer accuracy) and dynamic control errors in the smelting process (such as fluctuations in oxygen supply intensity, lag in slag formation adjustments, and uneven initial composition of molten iron). It can be expressed by the formula: Initial molten iron carbon mass – Final molten steel carbon mass ≈ K opt * (The total amount of decarbonization actually measured by the flue gas analyzer) is used as the dynamic correction coefficient for this furnace cycle. Here, * means multiplication, and the "total amount of decarbonization actually measured by the flue gas analyzer" in parentheses is actually a single value.

[0046] When the oxygen blowing rate η ≥ 80%, the TSC carbon content at the end of decarbonization is corrected and predicted, using K. opt Cumulative decarbonization C'decard during the calibration η < 80% stage:

[0047] C'decard=K opt ·ΔC decarb

[0048] At this point, the corrected instantaneous carbon content of the molten pool can be calculated. :

[0049]

[0050] Among them, the initial carbon content of molten iron [C] is 0, and the weight of molten iron is W. HM Scrap steel weight W FG ;

[0051] By fitting the carbon decrease curve of carbon content during the main decarbonization period and using the current d[C] / dt as the slope, the carbon content trajectory in the next Δt seconds is predicted:

[0052] .

[0053] Due to the adoption of the above technical solution, the beneficial effects achieved by this invention are:

[0054] This invention addresses the problem of insufficient prediction accuracy caused by lag, fluctuations, and nonlinear interference factors in traditional flue gas analysis methods for measuring carbon content in molten pools. It proposes a solution based on flue gas analysis methods and introducing a dynamic correction coefficient K. The core of this solution is to first dynamically divide the smelting process into silicon / manganese oxidation, main decarburization, and decreasing-rate decarburization periods using the ratio η of actual to theoretical oxygen consumption. Then, for different smelting stages, combining historical flue gas data and static parameters, the initial carbon content and cumulative decarburization are calculated using material conservation principles. The optimal K is then determined iteratively using a bisection method to correct for errors.

[0055] The method of this invention collects real-time data on multiple factors in flue gas, including carbon monoxide concentration, carbon dioxide concentration, oxygen, flow rate, and temperature, using a flue gas analyzer. It combines this with big data machine learning algorithms and introduces a dynamic correction coefficient K to balance the lag, fluctuations, and nonlinear interference factors inherent in traditional flue gas analysis methods, achieving second-level updates and global predictions of carbon content. Compared to traditional methods, this technology offers advantages such as continuous monitoring, strong anti-interference capabilities, and adaptive optimization. It provides operators with more accurate and timely information on carbon content evolution trends, enabling precise timing of shutdown, minimizing over-oxidation and under-oxidation, improving endpoint hit rate and process stability, and ultimately achieving significant economic benefits.

[0056] As can be seen, the technology of this invention achieves non-contact continuous inversion of the global carbon content in the molten pool by real-time monitoring of the flue gas composition parameters at the furnace mouth during the final stage of converter blowing, combined with a carbon-oxygen reaction mechanism model and machine learning algorithms. This invention is applicable to the precise control of the carbon content in molten steel during converter smelting, and can replace or assist traditional secondary lance point measurement technology, significantly improving the endpoint carbon hit rate, reducing the risk of over-blowing / under-blowing, and providing strong support for precise smelting control. Attached Figure Description

[0057] Figure 1 This is a flowchart of the calculation process of the present invention.

[0058] Figure 2 This is a graph showing the changes and predictions of the remaining carbon content in the molten pool after the present invention. Detailed Implementation

[0059] The present invention will be further described in detail below with reference to the embodiments.

[0060] This invention addresses the problem of insufficient prediction accuracy caused by lag, fluctuations, and nonlinear interference factors in traditional flue gas analysis methods for measuring carbon content in molten pools. It proposes a solution based on flue gas analysis methods and introducing a dynamic correction coefficient K. The core of this solution is to first dynamically divide the smelting process into silicon / manganese oxidation, main decarburization, and decreasing-rate decarburization periods using the ratio η of actual to theoretical oxygen consumption. Then, for different smelting stages, combining historical flue gas data and static parameters, the initial carbon content and cumulative decarburization are calculated using material conservation principles. The optimal K is then determined iteratively using a bisection method to correct the error. The specific technical solution adopted in this invention includes the following steps:

[0061] S1. Calculate the theoretical oxygen consumption, obtain the actual oxygen consumption, and set up a dynamic division mechanism for the decarbonization stage.

[0062] S2. Based on the differences in decarburization methods between TSC and TSO processes at different stages of the smelting cycle, corresponding dynamic correction coefficients K are generated.

[0063] S3. Use an optimization iterative algorithm to dynamically adjust and optimize the K value, and predict the application in real time.

[0064] Step S1 specifically includes the following:

[0065] S11. Let the current accumulated actual oxygen consumption be Qactual O2, and the theoretical oxygen consumption be Qtheory O2. The theoretical oxygen consumption Qtheory O2 is determined by the initial carbon content [C]0 of the molten iron and the weight W of the molten iron. HM Scrap steel weight W FG , Carbon contribution of scrap steel ΔC scrap The results were obtained through material balance (MB) and heat balance (HB) calculations:

[0066] Qtheory O2=f([C]0, W HM W FG ,ΔC scrap (Iron temperature T0 and iron composition)

[0067] S12. Based on the ratio of actual oxygen consumption to theoretical oxygen consumption To define the decarburization stage in the smelting process:

[0068]

[0069] Based on this classification, in the early stage of smelting when η < 30%, silicon / manganese oxidation is the dominant process. At this time, the decarburization reaction is weak and the carbon reduction rate is low but gradually increases.

[0070] When the carbon content is 30%≤η≤80% during the middle stage of smelting, the decarburization reaction is dominant. At this time, the reaction is intense and the carbon reduction rate is fast, which is the main stage of the decarburization reaction.

[0071] When η > 80%, it is already the decarbonization period of slowing down. By slowly reducing carbon emissions, the timing of stopping the blowing should be determined.

[0072] The ratio η = (30%, 80%) can be dynamically adjusted by ±5% to ±10% according to actual needs to clarify the time range of the main decarbonization period.

[0073] Step S2 specifically includes the following:

[0074] S21. The smelting cycle stage is TSC, and a dynamic correction coefficient K is constructed during this stage. TSC The TSC (Temperature Sampling Carbon) stage in the smelting process mainly controls the carbon content, temperature, and impurity elements of the molten steel through oxygen blowing and slag formation, ensuring that the steel composition and temperature meet the requirements of subsequent processes. This stage achieves precise control of the molten steel by adjusting parameters such as oxygen flow rate, slag composition, and operating time.

[0075] The smelting cycle stage is TSO, and a dynamic correction coefficient K is constructed during this stage. TSO K TSC With K TSO The processing procedure is the same; the TSO (Temperature Sampling Oxygen) stage in the smelting process is carried out after TSC, by stopping oxygen blowing and adjusting the slag state to make the oxygen and carbon content of the molten steel reach the target values.

[0076] S22, with K TSC For example, K TSC The computational optimization logic is as follows:

[0077] This can be achieved by obtaining the time series data of historical furnace cycles: flue gas flow rate Q. g (t), carbon monoxide concentration C CO (t), carbon dioxide concentration C CO2 (t), oxygen flow rate F O2 (t) and static parameters of molten iron (including initial carbon content [C]0, molten iron weight W) HM Scrap steel weight W FG 1. Overall carbon content of scrap steel [C] FG Using this basic information in conjunction with static parameter material conservation calculations, the initial carbon content [C]0 of the molten pool can be calculated:

[0078] [C]0=((W HM ·[C]0) + (W FG ·[C] FG )) ·10 -2 ·10 6

[0079] Throughout the entire smelting process of a single furnace, the concentration of carbon monoxide (C) varies. CO (t), carbon dioxide concentration C CO2 From the time-series data of (t), through mass conservation calculations, the cumulative decarburization amount ΔC at each data sampling frequency can be obtained from the start to the end of smelting. decarb :

[0080]

[0081] At the moment of the secondary gun under TSC, a cumulative decarburization amount ΔC can be obtained through the above method. decarb Also called total decarbonization, here ΔC decarb This represents the mass of carbon contained in carbon monoxide and carbon dioxide in the flue gas during converter smelting, measured in tons. The remaining carbon content [C] in the molten pool can be calculated from the difference between the initial carbon content [C]0 in the molten pool and the total decarburization amount. pred :

[0082]

[0083] Here The calculated result is a percentage content, representing the carbon content removed relative to the total carbon content in the molten steel.

[0084] The calculation at this point is entirely based on the law of conservation of mass. However, in actual production, there are data lags and nonlinear interference factors. Therefore, a dynamic correction coefficient K needs to be introduced to balance the data anomalies caused by errors.

[0085]

[0086] By iteratively calculating historical time-series data and using the bisection method to solve for the dynamic correction coefficient K, the remaining carbon content [C]K pred in the molten pool after equilibrium can be used to measure the true carbon content [C] in the molten pool using the secondary lance. target Within ±5% (the target offset accuracy can be adjusted according to the actual situation; 5% is used here to consider the computational complexity and number of iterations), this is the optimal dynamic correction coefficient K for the current smelting furnace:

[0087]

[0088] Since this occurs during the TSC stage of the smelting cycle, which is the period from the start of oxygen blowing to the first measurement at the lower lance in converter smelting, the dynamic correction coefficient K calculated during this stage is referred to here as K. TSC .

[0089] In step S3, the dynamic adjustment and real-time prediction mechanism for the K value is as follows:

[0090] S31. When the blowing begins, start collecting real-time measurement data from the flue gas analyzer, such as carbon monoxide, carbon dioxide, and flue gas flow rate. Combine this with the decarbonization stage division method and closely monitor the changes in the oxygen blowing stage.

[0091] S32. When the oxygen blowing progress reaches 75%, start the online optimization program to extract feature vector values; find the optimal value through Euclidean distance matching and optimize the dynamic correction coefficient K.

[0092] When η=75%, the online optimization program is initiated; the curve characteristics within the current furnace primary decarburization time range are extracted, and ΔC is analyzed. decarb The t-curve is piecewise fitted, and the slope β and intercept ΔC0 are extracted; the feature vector V of the current curve is calculated. current =[β,ΔC0] and the i-th furnace V in the historical feature database i Euclidean distance:

[0093]

[0094] Select D min K corresponding to the furnace number opt As the dynamic correction coefficient for this furnace run; D min K represents the minimum Euclidean distance; opt It is a correction factor used to compensate for or correct system measurement errors and process dynamic control errors of online detection equipment during normal smelting.

[0095] In the normal steelmaking process, the difference between the initial carbon mass of molten iron and the final carbon mass of molten steel is theoretically approximately equal to the total decarburization mass of carbon monoxide and carbon dioxide in the flue gas. Based on practical industrial production experience, the error between this theoretical calculation and the actual decarburization amount is usually controlled within ±5%. The sources of error mainly include: system measurement errors of online detection equipment (such as the accuracy drift of flue gas composition analyzers) and dynamic control errors in the smelting process (such as fluctuations in oxygen supply intensity, lag in slag formation adjustment, and uneven initial composition of molten iron).

[0096] This can be expressed by the formula: Initial molten iron carbon mass – Final molten steel carbon mass ≈ K opt * (Total decarbonization actually measured by the flue gas analyzer) is used as the dynamic correction coefficient for this furnace cycle; here, * means multiplication, and the "total decarbonization actually measured by the flue gas analyzer" in parentheses is actually a single value.

[0097] When the oxygen blowing rate η ≥ 80%, the TSC carbon content at the end of decarbonization is corrected and predicted, using K. opt Cumulative decarbonization C'decard during the calibration η < 80% stage:

[0098] C'decard=K opt ·ΔC decarb

[0099] At this point, the corrected instantaneous carbon content [C] of the molten pool can be calculated. t :

[0100]

[0101] Among them, the initial carbon content of molten iron [C] is 0, and the weight of molten iron is W. HM Scrap steel weight W FG ;

[0102] By fitting the carbon decrease curve of carbon content during the main decarbonization period and using the current d[C] / dt as the slope, the carbon content trajectory in the next Δt seconds is predicted:

[0103] . Specific Implementation

[0104] Initial conditions:

[0105] Weight of molten iron: W HM =257.37t

[0106] Molten iron temperature: T0 = 1333℃

[0107] Scrap steel weight: W FG =40.96t

[0108] Actual carbon content at the TSC stage: [C] target =0.42%

[0109] The composition of molten iron is shown in Table 1:

[0110] C Si Mn P S 4.73665 0.37001 0.17804 0.1424 0.04376

[0111] 1. Calculation of theoretical oxygen consumption:

[0112] Carbon consumes oxygen:

[0113]

[0114] Silicon consumes oxygen:

[0115]

[0116] Manganese consumes oxygen:

[0117]

[0118] Phosphorus consumes oxygen:

[0119]

[0120] Iron in scrap steel consumes oxygen:

[0121]

[0122] Here, we don't need to consider the oxygen consumption by sulfur (S). Desulfurization is primarily a reduction reaction, while converter smelting mainly involves oxidation. Therefore, the oxidation of sulfur during converter smelting can be ignored. Furthermore, if the sulfur content in the molten iron is too high, a pre-desulfurization process is usually arranged on-site before converter smelting. By the time of converter smelting, the sulfur content is usually already low, and the oxidation of sulfur can be disregarded. Therefore, we only calculate the reactions of the main elements with oxygen during the converter process.

[0123] Theoretical oxygen consumption during the TSC phase:

[0124]

[0125] 2. Optimization calculation of dynamic correction coefficient K:

[0126] Calculate the cumulative decarburization rate per second from the start of oxygen blowing to the time of TSC (Transmission Control Center) at the secondary lance:

[0127]

[0128] The optimization factor K is calibrated using carbon measured by TSC. opt :K opt This is a correction for the unmeasurable decarburization during converter smelting, as explained above regarding K. opt The specific meaning is described in detail below. It can be expressed by a formula: Initial molten iron carbon mass – Final molten steel carbon mass ≈ K opt * (Total carbon removal amount actually measured by the flue gas analyzer). W Metal It is the total mass of the molten steel, that is: W Metal =W HM +W FG .

[0129]

[0130] Find a suitable K value using the binary search method:

[0131] A 5% margin of error is allowed between the prediction and the actual value, i.e., the remaining carbon content in the molten pool [C]. pred The solution should be within the range [0.399, 0.441]. This method requires solving an approximate solution for the continuous function equation on the closed interval [a, b], as shown in Table 2.

[0132] Serial Number interval Random K <![CDATA[[C] pred ]]> 1 1.332 0.2445 2 0.454 0.8373 3 0.832 0.3256 4 0.537 0.4832 5 0.667 0.4145

[0133] At this point, K = 0.667, which is the optimization coefficient under the current circumstances.

[0134] 3. Feature extraction:

[0135] Decarbonization stage definition: Based on the theoretical oxygen consumption Qtheory O2, the main decarbonization range is precisely defined.

[0136] TSC starting point (t0): η=30% (oxygen blowing volume reaches 3986 Nm) 3 )

[0137] TSC endpoint (t1): η=80% (oxygen blowing volume reaches 10629 Nm³) 3 )

[0138] Extracting flue gas from the interval [t0, t1] and For time-series data, feature vectors are constructed through linear fitting:

[0139] Feature vector: V current =[a co , b co ,a co2 , b co2 ]

[0140] 4. Euclidean distance optimization:

[0141] For example: V current =[14740.12, -42171.12, 4536.99, 60192.89]

[0142] Table 3 Historical Characteristics and Optimization Coefficients

[0143] Serial Number K Euclidean distance 1 14707.79 -371545.70 5172.58 114555.95 0.942 333831.3365 2 15989.01 -323015.85 4962.45 56821.26 0.627 280868.0669 3 17505.91 222867.17 4928.93 -29175.78 0.845 279713.8852 4 16233.63 -74831.22 4728.68 87375.78 1.16 42518.9247 5 18057.73 -186351.27 4884.73 112138.46 0.938 153288.5689

[0144] Take the K with the smallest Euclidean distance opt =1.16 is used as the current furnace optimization coefficient for calculation.

[0145] The same applies to the TSO phase, so I won't go into details.

[0146] Specific processes combined Figure 1 To showcase, Figure 2 The graph shows the trend of carbon content change from the start of converter smelting to the TSC sub-lance measurement. From this graph, we can see that:

[0147] 1. The initial carbon content is approximately 4.2.

[0148] 2. The blue curve is relatively stable in the early stage of smelting, and the decarburization rate increases significantly in the middle and late stages. This indicates that the initial stage of converter smelting is a weak decarburization reaction, and the main reaction in the early stage is the oxidation reaction of silicon and manganese. The main period of decarburization is in the middle and late stages.

[0149] 3. The red line in the figure has a longer prediction period than the blue line. This is because the flue gas analyzer is usually installed at the end of the flue, and there is a certain delay in the measurement of carbon monoxide and carbon dioxide. Therefore, the prediction time point needs to be postponed. The predicted carbon result for the TSC stage is about 0.3.

Claims

1. A method for predicting the residual carbon content in the molten pool of a steelmaking converter, characterized in that, Includes the following steps: S1. Calculate the theoretical oxygen consumption, obtain the actual oxygen consumption, and set up a dynamic division mechanism for the decarbonization stage. S2. Based on the differences in decarburization methods between TSC and TSO processes at different stages of the smelting cycle, corresponding dynamic correction coefficients K are generated. S3. Use an optimization iterative algorithm to dynamically adjust and optimize the K value, and predict the application in real time.

2. The method for predicting the residual carbon content in the molten pool of a steelmaking converter according to claim 1, characterized in that, Step S1 specifically includes the following: S11. Let the current accumulated actual oxygen consumption be Qactual O2, and the theoretical oxygen consumption be Qtheory O2. The theoretical oxygen consumption Qtheory O2 is determined by the initial carbon content [C]0 of the molten iron and the weight W of the molten iron. HM Scrap steel weight W FG , Carbon contribution of scrap steel ΔC scrap The results were obtained through material balance (MB) and heat balance (HB) calculations: Qtheory O2=f([C]0, W HM W FG ,ΔC scrap (Iron temperature T0 and iron composition) S12. The decarburization stage in the smelting process is defined based on the ratio η between actual oxygen consumption and theoretical oxygen consumption: ； Based on this classification, when η < 30% in the early stage of smelting, silicon / manganese oxidation is dominant. At this time, the decarburization reaction is weak and the carbon reduction rate is low but gradually increases. When the carbon content is 30%≤η≤80% during the middle stage of smelting, the decarburization reaction is dominant. At this time, the reaction is intense and the carbon reduction rate is fast, which is the main stage of the decarburization reaction. When η > 80%, it is already the decarbonization period of slowing down. By slowly reducing carbon emissions, the timing of stopping the blowing should be determined.

3. The method for predicting the residual carbon content in the molten pool of a steelmaking converter according to claim 2, characterized in that, The ratio η = (30%, 80%) can be dynamically adjusted by ±5% to ±10% according to actual needs to clarify the time range of the main decarbonization period.

4. The method for predicting the residual carbon content in the molten pool of a steelmaking converter according to claim 2, characterized in that, Step S2 specifically includes the following: S21, the stage of the smelting cycle is TSC, and a dynamic correction coefficient K is constructed. TSC The smelting cycle stage is TSO, and a dynamic correction coefficient K is constructed. TSO K TSC With K TSO The processing procedure is the same; S22, with K TSC For example, K TSC The computational optimization logic is as follows: Historical furnace time series data obtained: flue gas flow rate Q g (t), carbon monoxide concentration C CO (t), carbon dioxide concentration C CO2 (t), oxygen flow rate F O2 (t) and static parameters of molten iron, using this basic information combined with static parameter material conservation calculations, the initial carbon content [C]0 of the molten pool can be calculated: [C]0=((W HM ·[C]0) + (W FG ·[C] FG )) ·10 -2 ·10 6 Throughout the entire smelting process of a single furnace, the concentration of carbon monoxide (C) varies. CO (t), carbon dioxide concentration C CO2 From the time-series data of (t), through mass conservation calculations, the cumulative decarburization amount ΔC at each data sampling frequency can be obtained from the start to the end of smelting. decarb : ； At the moment of the secondary gun under TSC, a cumulative decarburization amount ΔC can be obtained through the above method. decarb Here ΔC decarb This indicates the mass of carbon contained in carbon monoxide and carbon dioxide in the flue gas during converter smelting, and the unit is tons. The remaining carbon content [C] in the molten pool can be calculated by the difference between the initial carbon content [C]0 and the total decarburization amount. pred : ； Here The calculated result is a percentage content, representing the carbon content removed relative to the total carbon content in the molten steel. The calculation at this point is entirely based on the law of conservation of mass. However, in actual production, there are data lags and nonlinear interference factors. Therefore, a dynamic correction coefficient K needs to be introduced to balance the data anomalies caused by errors. ； By iteratively calculating historical time-series data and using the bisection method to solve for the dynamic correction coefficient K, the remaining carbon content [C]K pred in the molten pool after equilibrium can be used to measure the true carbon content [C] in the molten pool using the secondary lance. target Within ±5%, this is the optimal dynamic correction coefficient K for the current smelting furnace. ； Since this occurs during the TSC stage of the smelting cycle, the dynamic correction coefficient K calculated in this stage is referred to here as K. TSC .

5. The method for predicting the residual carbon content in the molten pool of a steelmaking converter according to claim 4, characterized in that, In step S22, the static parameters of molten iron include the initial carbon content [C]0 and the weight of molten iron W. HM Scrap steel weight W FG Combined carbon content [C] of scrap steel FG .

6. The method for predicting the residual carbon content in the molten pool of a steelmaking converter according to claim 1, characterized in that, In step S3, the dynamic adjustment and optimization mechanism for the K value and the real-time prediction mechanism are as follows: S31. When the blowing begins, start collecting real-time measurement data from the flue gas analyzer, and closely monitor the changes in the oxygen blowing stage in conjunction with the decarbonization stage division method. S32. When the oxygen blowing progress reaches 75%, start the online optimization program to extract feature vector values; find the optimal value through Euclidean distance matching and optimize the dynamic correction coefficient K. When η=75%, the online optimization program is initiated; the curve characteristics within the current furnace primary decarburization time range are extracted, and ΔC is analyzed. decarb The t-curve is piecewise fitted, and the slope β and intercept ΔC0 are extracted; the feature vector V of the current curve is calculated. current =[β,ΔC0] and the i-th furnace V in the historical feature database i Euclidean distance: ； Select D min K corresponding to the furnace number opt As the dynamic correction coefficient for this furnace cycle, D min K represents the minimum Euclidean distance; opt It is a correction factor used to compensate for or correct system measurement errors and process dynamic control errors of online detection equipment during normal smelting; When the oxygen blowing rate η ≥ 80%, the TSC carbon content at the end of decarbonization is corrected and predicted, using K. opt Cumulative decarbonization C'decard during the calibration η < 80% stage: C'decard=K opt ·ΔC decarb At this point, the corrected instantaneous carbon content [C] of the molten pool can be calculated. t : ； Among them, the initial carbon content of molten iron [C] is 0, and the weight of molten iron is W. HM Scrap steel weight W FG ; By fitting the carbon decrease curve of carbon content during the main decarbonization period and using the current d[C] / dt as the slope, the carbon content trajectory in the next Δt seconds is predicted: 。

Images