A method for dynamic solidification control of high-purity carbon structural steel based on four-dimensional regulation

By constructing a four-dimensional dynamic solidification control method for high-purity carbon structural steel, and combining multi-physics field coupling simulation and intelligent closed-loop control, the problems of coarse grains, compositional segregation, and inclusion aggregation in traditional processes have been solved. This has enabled the production of high-purity carbon structural steel with high efficiency and low energy consumption, meeting the quality and performance requirements of high-end equipment manufacturing.

CN123102950APending Publication Date: 2026-11-17SHANDONG IRON & STEEL GRP YONGFENG LINGANG CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611292552.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-25
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

The existing high-purity carbon structural steel continuous casting solidification process lacks four-dimensional coordinated control, has low parameter control accuracy, and cannot achieve online real-time monitoring. This results in coarse grains, severe compositional segregation, inclusion aggregation, high energy consumption, low efficiency, and an inability to meet the quality requirements of high-end equipment manufacturing.

Method used

A multiphysics coupling simulation model based on ANSYS Fluent and cellular automata is adopted, combined with FactSage thermodynamic calculations, to construct a four-dimensional control system of time-space-temperature-composition. Through intelligent closed-loop control, the cooling rate, stirring parameters and modifier addition amount are monitored and dynamically adjusted in real time to achieve high-cleanliness and homogeneous production of billets.

Benefits of technology

The results achieved a billet grain size of ≤5μm, a compositional segregation index of ≤1.05, and an inclusion size of ≤5μm, significantly improving the mechanical properties of the product, increasing production efficiency by 20%, reducing energy consumption by 30%, and reducing the scrap rate to below 2%, thus meeting the requirements for high-end equipment steel.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

This invention belongs to the field of continuous casting technology in iron and steel metallurgy, and particularly relates to a dynamic solidification control method for high-purity carbon structural steel based on four-dimensional regulation. The method first constructs a dynamic solidification model. After FactSage thermodynamic calculations and calibration with actual data, the solidification process is divided into three stages according to the time dimension: strong cooling, slow cooling, and gradient cooling. According to the spatial dimension, the continuous casting machine is divided into four sections: the crystallizer, bending section, straightening section, and cooling section, and the cooling intensity, electromagnetic stirring parameters, and light reduction are adjusted accordingly. Simultaneously, the cooling intensity, stirring parameters, and modifier addition amount are dynamically adjusted. Finally, through an intelligent decision-making system, with the target of grain size ≤5μm, segregation index ≤1.05, and inclusion size ≤5μm, control commands are output and feedback is iterated to form a closed-loop control. This invention achieves homogeneous solidification of the cast billet, reduces energy consumption, and improves the yield.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of continuous casting technology in iron and steel metallurgy, and particularly relates to a dynamic solidification control method for high-purity carbon structural steel based on four-dimensional regulation. Background Technology

[0002] High-purity carbon structural steel is a key basic material for the manufacturing of high-end equipment such as automotive main shafts, aerospace bearings, and nuclear power pressure vessels. The grain refinement, compositional uniformity, and inclusion control level of the cast billet directly determine the service performance and service life of the final product. The continuous casting solidification process is the core link in the production of carbon structural steel. During this process, the multi-field coupling behavior of temperature field, flow field, stress field, and solute field directly affects the solidification structure and quality of the cast billet.

[0003] Currently, the traditional continuous casting solidification process for high-purity carbon structural steel mainly faces the following technical problems:

[0004] (I) Lack of Four-Dimensional Coordinated Control. Traditional processes often employ fixed cooling parameters and a single stirring mode, failing to achieve coordinated linkage between time, space, temperature, and composition. The disconnect between cooling rate, stirring parameters, and composition control leads to coarse grains (average size > 15 μm), severe central segregation (segregation index > 1.4), and inclusion aggregation (size > 20 μm), making it impossible to stably achieve the production of high-cleanliness, homogeneous cast billets. Existing technologies mostly focus on single-dimensional parameter optimization, without forming a multi-dimensional coordinated control mechanism.

[0005] (ii) Low precision of parameter control. The setting of solidification parameters relies on manual experience and lacks a precise control model based on multi-physics coupling. It is impossible to dynamically adjust according to the real-time solidification state of the billet, resulting in large fluctuations in billet quality, grain size standard deviation >1.2μm, and scrap rate as high as 8% or more.

[0006] (III) Delayed online monitoring and feedback. Existing technologies mostly use offline sampling and testing, which cannot monitor grain growth, compositional segregation and inclusion distribution during the solidification process in real time. The parameter adjustment response time is >5s, making it difficult to suppress solidification defects in time. The steel leakage rate is as high as 0.5 times / 10,000 tons, and the yield is only about 92%.

[0007] (iv) Imbalance between energy consumption and efficiency. In pursuit of grain refinement, traditional processes blindly increase cooling intensity or extend stirring time, resulting in high production energy consumption (≥350kgce / t), low continuous casting speed (≤1.0m / min), and low production efficiency, which does not meet the needs of green and intelligent upgrading in the steel industry.

[0008] In summary, there is an urgent need for a solidification control method for high-purity carbon steel that can achieve four-dimensional coordinated control of time, space, temperature, and composition, and dynamically optimize solidification parameters based on a multi-physics coupling model and online real-time monitoring data. Summary of the Invention

[0009] The purpose of this invention is to provide a dynamic solidification control method for high-purity carbon structural steel based on four-dimensional regulation, so as to solve the problems existing in the prior art.

[0010] The technical solution adopted by this invention to solve its technical problem is:

[0011] A method for dynamic solidification control of high-purity carbon structural steel based on four-dimensional regulation includes the following steps:

[0012] S1. Construct a coordinated control system of four-dimensional parameters of time, space, temperature and composition, and determine the quantitative correlation of parameters in each dimension based on a multi-physics field coupled simulation model.

[0013] Furthermore, in step S1, the multiphysics coupling simulation model is a dynamic solidification model based on the ANSYS Fluent fluid simulation software and cellular automata, which couples the temperature field, flow field, stress field and solute field, and is calibrated by combining FactSage thermodynamic calculations.

[0014] Specifically: ANSYS Fluent 19.0 fluid simulation software was used to establish the macroscopic control equations for the flow and heat transfer of molten steel during continuous casting. A cellular automaton (CA) was used to establish a microscopic evolution model of grain nucleation and growth during solidification. The macroscopic temperature field calculation results were used as the input boundary conditions for the microscopic model, achieving macroscopic-microscopic cross-scale coupling. The temperature field was solved based on the Fourier heat transfer differential equation, comprehensively considering the latent heat release during the solidification process of molten steel, the convective heat transfer between the billet surface and cooling water, and the radiative heat transfer between the billet and the mold. The flow field was solved based on the Navier-Stokes equations and the standard k-ε turbulence model, simulating the flow behavior of molten steel in the mold. The stress field was solved based on a thermo-elastoplastic constitutive model, using the finite element method to calculate the thermal and mechanical stresses of the billet during cooling and straightening. The solute field was solved based on the solute transport equation, considering the solute redistribution coefficient between the solid and liquid phases. The thermophysical parameters in the model, including liquidus temperature, solidus temperature, solute partition coefficient, and diffusion coefficient, were optimized using FactSage thermodynamic calculation software. Infrared thermal imager and X-ray tomography were used to acquire actual solidification process data. The multiphysics coupled simulation model was calibrated, and key parameters such as temperature conductivity and solute diffusion coefficient were adjusted to ensure that the deviation between model predictions and actual measured values ​​was ≤5%, and the model prediction accuracy was ≥96%.

[0015] S2. Time-dimension segmented control: The solidification process of carbon structural steel is divided into initial solidification stage, intermediate solidification stage and final solidification stage according to time sequence, and a differentiated cooling rate is set and executed for each stage.

[0016] Furthermore, in step S2, the time-dimension segmented control adopts the following strategy:

[0017] In the initial solidification stage, a strong cooling mode was adopted, with a cooling rate set at 3-5℃ / s to suppress solute element aggregation and reduce central segregation. The grain nucleation rate was increased by 40% in this stage.

[0018] During the intermediate solidification stage, a slow cooling mode is adopted, with the cooling rate set at 1.5-2.5℃ / s to promote the growth of equiaxed crystals and inhibit the coarsening of columnar crystals.

[0019] The final solidification stage adopts a gradient cooling mode with a cooling rate set at 0.5-1.0℃ / s to release solidification stress and prevent the billet from cracking.

[0020] S3. Spatial Dimension Zoning Control: Differentiated cooling intensity, electromagnetic stirring parameters, and light reduction amount are set and executed according to the crystallizer section, bending section, straightening section, and cooling section of the continuous casting machine.

[0021] Furthermore, in step S3, the spatial dimension partitioning control adopts the following strategy:

[0022] Crystallizer section: Cooling intensity 0.6-0.8 L / kg, combined with rotary electromagnetic stirring, used to break up initial columnar crystals;

[0023] Bending section: Cooling intensity 0.8-0.85 L / kg, using linear electromagnetic stirring to promote equiaxed crystal growth;

[0024] Straightening section: Cooling intensity 1.2-1.25L / kg, dynamic light compression is implemented, compression amount 10-11mm, solid phase ratio ≥0.6, used to promote interdendritic liquid phase flow and reduce component segregation;

[0025] Cooling section: Gradual cooling is adopted, with the cooling intensity gradually decreasing from 1.0L / kg to 0.5L / kg to ensure uniform cooling of the billet and avoid stress concentration.

[0026] S4. Dynamic temperature control: Real-time monitoring of billet temperature, and dynamic adjustment of cooling intensity in corresponding sections based on the monitored temperature deviation.

[0027] Furthermore, in step S4, the dynamic control strategy for the temperature dimension is as follows: control the cooling gradient ≤15℃ / cm, and the surface temperature fluctuation of the billet ≤±3℃; when the local temperature of the billet is detected to be higher than the set value, automatically increase the cooling intensity of the corresponding section by 0.2L / kg; when the temperature is lower than the set value, reduce the cooling intensity by 0.1L / kg.

[0028] S5. Dynamic control of composition: Online monitoring of solute composition in molten steel and calculation of composition segregation index; dynamic adjustment of electromagnetic stirring parameters and light reduction amount based on changes in segregation index; and addition of modifier in the crystallizer.

[0029] Furthermore, in step S5, the dynamic control strategy for composition dimension is as follows: control the composition segregation index ≤ 1.05 and the inclusion size ≤ 5μm; when the segregation index > 1.05, automatically increase the electromagnetic stirring frequency by 2-3Hz and increase the light reduction by 2mm; the modifier is an Al-Ti-B alloy, with an addition amount of 0.02%-0.03%, used to refine grains and inhibit inclusion growth.

[0030] S6. Intelligent closed-loop control: Real-time acquisition of continuous casting process parameters, dynamic adjustment of electromagnetic stirring intensity and modifier addition amount based on acquisition results, and feedback of control results to multi-physics coupling model for iterative optimization until billet quality meets standards.

[0031] Furthermore, in step S6, the intelligent closed-loop control method is as follows: real-time acquisition of billet geometric dimension data, billet internal defect data, chemical composition data, and temperature field data. The acquired data is input into an intelligent decision-making system based on an LSTM neural network. The intelligent decision-making system calculates the current grain size and inclusion status by combining a multi-physics coupling model and compares it with the target value. Based on the comparison result, the electromagnetic stirring intensity and the amount of modifier added are dynamically adjusted, and the control result is fed back to the multi-physics coupling model for iterative optimization until the billet quality meets the standard.

[0032] Furthermore, the dynamic adjustment of the electromagnetic stirring intensity and the amount of modifier added is as follows: when the grain size of the billet is detected to be >5μm, the electromagnetic stirring current is automatically increased by 50-100A and the strong cooling time is extended by 5-10s; when the size of the inclusions is detected to be >5μm, the amount of modifier added is increased by 0.005%-0.01%.

[0033] Specifically, sensors such as laser diameter gauges, ultrasonic flaw detectors, laser spectrometers, and infrared thermometers are deployed to collect real-time data on: the geometric dimensions of the cast billet, such as width, thickness, profile, and bulge; internal defects, such as the location and equivalent size of cracks, shrinkage cavities, and porosity; chemical composition, such as the content of elements like C, Mn, S, and P; and temperature field data, such as surface temperature, internal temperature, and cooling gradient. The collected data serves as input, with cooling intensity, electromagnetic stirring parameters, light reduction, and modifier addition as outputs. The control targets are grain size ≤ 5 μm, compositional segregation index ≤ 1.05, and inclusion size ≤ 5 μm. An intelligent decision-making system based on an LSTM neural network was constructed. The LSTM neural network has a three-layer structure, including an input layer, a hidden layer (128 neurons), and an output layer. Supervised learning training was performed using historical data training set until the prediction error converged. When the grain size of the billet is detected to be >5μm, the electromagnetic stirring current is automatically increased by 50-100A and the strong cooling time is extended by 5-10s. When the inclusion size is detected to be >5μm, the amount of modifier added is increased by 0.005%-0.01%. The system response time is ≤0.5s. At the same time, the control results are fed back to the multiphysics coupling model for iterative optimization until the billet quality meets the standard.

[0034] Furthermore, as an alternative, if the Al-Ti-B modifier is unavailable on-site, Nb-V microalloying can be used to replace the Al-Ti-B modifier, with an Nb-V addition amount of 0.03%-0.04%. At the same time, the cooling rate of the initial intensive cooling stage is increased to 5℃ / s, the intensive cooling time is extended to 35s, and the electromagnetic stirring current is increased by 50A.

[0035] Furthermore, as an alternative, if the LSTM neural network intelligent decision-making system cannot be deployed on-site, a PLC control system can be used to replace the LSTM neural network intelligent decision-making system, with three sets of four-dimensional control parameter templates for different steel grades preset, combined with manual adjustment.

[0036] The present invention has the following beneficial effects:

[0037] 1. Through four-dimensional collaborative control and multi-physics field coupled simulation, the defects of coarse grains, compositional segregation, and inclusion aggregation in traditional processes are completely solved, achieving a billet grain size ≤5μm, compositional segregation index ≤1.05, and inclusion size ≤5μm. This meets the ultra-pure and homogeneous requirements of high-end equipment steel, and significantly improves the mechanical properties of the product (tensile strength ≥850MPa, impact toughness ≥80J / cm). 2 ).

[0038] 2. By replacing manual experience-based regulation with intelligent closed-loop control, the parameter response time is ≤0.5s, the endpoint quality hit rate is increased from 65% to over 98%, the steel leakage rate is reduced from 0.5 times / 10,000 tons to ≤0.1 times / 10,000 tons, and the scrap rate is reduced to <2%; the continuous casting speed is increased to ≥1.2m / min, and the production efficiency is increased by more than 20%.

[0039] 3. Optimize solidification process parameters to reduce energy consumption per ton of steel to ≤210kgce, which is more than 30% lower than the industry average, thus reducing energy consumption; at the same time, reduce billet scrap rate, reduce raw material waste, and achieve green and efficient production of high-cleanliness carbon structural steel.

[0040] 4. This invention can be directly adapted to existing continuous casting production lines without large-scale equipment modifications. Industrial application can be achieved through process parameter optimization and sensor deployment. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0042] Example 1:

[0043] This embodiment describes the production of high-purity carbon structural steel (45# steel) for automotive main shafts.

[0044] (1) Parameter settings:

[0045] The initial steel composition was: C 0.45%, Mn 0.85%, S 0.012%, P 0.020%.

[0046] The four-dimensional control parameters are set as follows:

[0047] Time dimension: 0-30s cooling rate 4℃ / s, 30-60s cooling rate 2℃ / s, after 60s cooling rate 0.8℃ / s;

[0048] Spatial dimensions: Crystallizer cooling intensity 0.7L / kg, rotary electromagnetic stirring 15Hz / 350A; Bending section cooling intensity 0.8L / kg, linear electromagnetic stirring 5mm / 450A; Straightening section cooling intensity 1.2L / kg, light reduction 10mm; Cooling section gradient cooling 0.8-0.5L / kg;

[0049] Temperature dimension: Temperature gradient 12℃ / cm;

[0050] Compositional dimension: Modifier (Al-Ti-B alloy) addition amount 0.025%.

[0051] (2) Implementation process:

[0052] A multiphysics coupled ANSYS Fluent+ cellular automata model was established according to S1. Actual solidification process data were acquired using an infrared thermal imager (accuracy ±0.1℃) and X-ray tomography (resolution 1μm). The temperature conductivity and solute diffusion coefficient in the model were calibrated to ensure that the deviation between the model's predicted values ​​and the actual measured values ​​was ≤5%. After calibration, the model was put into operation. Four-dimensional coordinated regulation and online intelligent closed-loop control of time, space, temperature, and composition were implemented according to S2 to S6.

[0053] (3) Effect verification:

[0054] The average grain size of the cast billet is 4.2 μm, the standard deviation of grain size is 0.25 μm, the compositional segregation index is 1.03, and the average size of inclusions is 4.1 μm.

[0055] Mechanical properties: tensile strength 865MPa, yield strength 660MPa, elongation 19%, impact toughness (-20℃) 85J / cm²;

[0056] Production efficiency: continuous casting speed 1.25m / min, yield 96.5%, energy consumption per ton of steel 205kgce, and steel leakage rate 0.08 times / 10,000 tons;

[0057] The continuous production of 60 heats achieved a billet quality qualification rate of 99.8%, meeting the requirements for steel used in automotive main shafts.

[0058] Example 2:

[0059] This embodiment describes the production of high-purity carbon structural steel (20MnNiMo steel) for nuclear power pressure vessels.

[0060] (1) Parameter settings:

[0061] The initial steel composition is: C 0.20%, Mn 1.50%, Ni 0.80%, Mo 0.25%, S 0.010%.

[0062] The four-dimensional control parameters are set as follows:

[0063] Time dimension: 0-30s cooling rate 4.5℃ / s, 30-60s cooling rate 2.2℃ / s, after 60s cooling rate 0.7℃ / s;

[0064] Spatial dimensions: Crystallizer cooling intensity 0.75L / kg, rotary electromagnetic stirring 16Hz / 380A; Bending section cooling intensity 0.85L / kg, linear electromagnetic stirring 5mm / 480A; Straightening section cooling intensity 1.25L / kg, light reduction 11mm; Cooling section gradient cooling 0.9-0.5L / kg;

[0065] Temperature dimension: Temperature gradient 11℃ / cm;

[0066] Ingredient information: 0.028% added deteriorator.

[0067] (2) Effect verification:

[0068] The average grain size of the cast billet is 3.8 μm, the standard deviation of grain size is 0.22 μm, the compositional segregation index is 1.02, and the average size of inclusions is 3.8 μm.

[0069] Mechanical properties: tensile strength 880MPa, yield strength 675MPa, elongation 20%, impact toughness (-20℃) 90J / cm 2 ;

[0070] Production efficiency: continuous casting speed 1.2m / min, yield 97%, energy consumption per ton of steel 202kgce, steel leakage rate 0.07 times / 10,000 tons;

[0071] The steel was produced in 50 consecutive furnaces with a billet quality qualification rate of 99.9%, meeting the stringent requirements for steel used in nuclear power and pressure vessels.

[0072] Comparative Example 1:

[0073] Using traditional continuous casting solidification process (fixed cooling parameters, single electromagnetic stirring, no online closed-loop control), 45# steel billets of the same specification were produced with an average grain size of 16.2μm, a compositional segregation index of 1.42, an average inclusion size of 18.5μm, a tensile strength of 785MPa, a continuous casting speed of 0.95m / min, an energy consumption of 358kgce per ton of steel, and a yield of 91.2%.

[0074] As can be seen from the comparison, in Example 1 of the present invention, the grain size is refined by about 74%, the segregation index is reduced by about 27%, the tensile strength is increased by about 10%, the pulling speed is increased by about 32%, the energy consumption is reduced by about 43%, and the yield is increased by about 5.8 percentage points.

[0075] Example 3:

[0076] Alternative Solution 1 (Modifier Substitution): If Al-Ti-B modifier is unavailable on-site, Nb-V microalloying can be used to replace the Al-Ti-B modifier. The Nb-V addition amount is 0.03%-0.04%. Simultaneously, the initial forced cooling rate is increased to 5℃ / s, the forced cooling time is extended to 35s, and the electromagnetic stirring current is increased by 50A. Under this alternative solution, the average grain size is ≤5.5μm, the compositional segregation index is ≤1.08, the inclusion size is ≤5.5μm, and the mechanical properties meet the requirements for mid-to-high-end carbon structural steel (tensile strength ≥850MPa, yield strength ≥650MPa), making it suitable for scenarios where modifier supply is limited.

[0077] Alternative Solution 2 (Control System Replacement): If an LSTM neural network intelligent decision-making system cannot be deployed on-site, a PLC control system can be used instead. Three sets of four-dimensional control parameter templates for different steel grades (e.g., 45# steel, 20MnNiMo steel, Q345 steel) are preset, and combined with manual adjustment, the response time is ≤1s. Under this alternative solution, the model prediction deviation is ≤8%, the grain size is ≤6μm, and the compositional segregation index is ≤1.10, still meeting the basic requirements for high-end carbon structural steel and suitable for small- to medium-scale production line upgrades.

[0078] The working principle of this invention is:

[0079] This invention first constructs a dynamic solidification model coupling four fields—temperature, flow, stress, and solute—based on ANSYS Fluent fluid simulation software and cellular automata. ANSYS Fluent calculates the macroscopic temperature distribution and flow state of the molten steel during continuous casting, while the cellular automata simulates grain nucleation and growth at the microscale. The two models achieve cross-scale coupling through boundary conditions in the temperature field. The stress field calculates the thermal and mechanical stresses of the cast billet during cooling and straightening based on a thermo-elastoplastic constitutive model; the solute field calculates the redistribution behavior of elements during solidification based on the solute transport equation. The model is calibrated using FactSage thermodynamic calculations and actual monitoring data from infrared thermal imaging and X-ray tomography, achieving a prediction accuracy of ≥96%, providing theoretical support for four-dimensional control. Based on this, time-dimension segmented control (0-30s strong cooling, 30-60s slow cooling, and gradient cooling after 60s) and spatial-dimension zonal control (rotary stirring in the crystallizer, linear stirring in the bending section, light reduction in the straightening section, and gradient cooling in the cooling section) are implemented to achieve initial control of the solidification process. During solidification, the billet temperature is monitored in real time using an infrared thermometer and thermocouples, and the cooling intensity is dynamically adjusted according to the temperature deviation. The solute composition of the molten steel is monitored online using a laser spectrometer, the composition segregation index is calculated, and the electromagnetic stirring parameters and light reduction are dynamically adjusted accordingly. At the same time, Al-Ti-B modifiers are added to the crystallizer. The system uses multiple sensors, such as diameter gauges, ultrasonic flaw detectors, spectrometers, and thermometers, to collect real-time data on the geometric dimensions, internal defects, chemical composition, and temperature field of the billet. Using this data as input and cooling intensity, electromagnetic stirring parameters, light reduction, and modifier addition as output, an intelligent decision-making system built with an LSTM neural network dynamically optimizes and controls the parameters and feeds them back to each execution module, forming a closed-loop operation of monitoring-analysis-control-feedback until the billet quality meets the standards.

[0080] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the concept and scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the design concept of the present invention should fall within the protection scope of the present invention.

[0081] The technologies, shapes, and structures not described in detail in this invention are all known technologies.

Claims

1. A method for dynamic solidification control of high-purity carbon structural steel based on four-dimensional regulation, characterized in that, Includes the following steps: S1. Construct a coordinated control system of four-dimensional parameters of time, space, temperature and composition, and determine the quantitative correlation of parameters in each dimension based on a multi-physics field coupled simulation model; S2. Time-dimensional segmented control: The solidification process of carbon structural steel is divided into the initial solidification stage, the intermediate solidification stage and the final solidification stage according to the time sequence, and a differentiated cooling rate is set and executed in each stage. S3. Spatial Dimension Zoning Control: Differentiated cooling intensity, electromagnetic stirring parameters, and light reduction amount are set and executed according to the crystallizer section, bending section, straightening section, and cooling section of the continuous casting machine. S4. Dynamic temperature control: Real-time monitoring of billet temperature, and dynamic adjustment of cooling intensity in corresponding sections based on the monitored temperature deviation. S5. Dynamic control of composition dimension: Online monitoring of solute composition in molten steel and calculation of composition segregation index; dynamic adjustment of electromagnetic stirring parameters and light reduction amount based on changes in segregation index; and addition of modifier in crystallizer. S6. Intelligent closed-loop control: Real-time acquisition of continuous casting process parameters, dynamic adjustment of electromagnetic stirring intensity and modifier addition amount based on acquisition results, and feedback of control results to multi-physics coupling model for iterative optimization until billet quality meets standards.

2. The method according to claim 1, characterized in that, In step S1, the multiphysics coupling simulation model is a dynamic solidification model based on the coupling of four fields: temperature field, flow field, stress field and solute field, constructed using ANSYS Fluent fluid simulation software and cellular automata, and calibrated using FactSage thermodynamic calculations.

3. The method according to claim 1, characterized in that, In step S2, the time-dimension segmented control adopts the following strategy: The initial solidification stage adopts a strong cooling mode with a cooling rate set at 3-5℃ / s. During the intermediate solidification stage, a slow cooling mode is adopted, with the cooling rate set at 1.5-2.5℃ / s; The final solidification stage adopts a gradient cooling mode with a cooling rate set at 0.5-1.0℃ / s.

4. The method according to claim 1, characterized in that, In step S3, the spatial dimension partitioning control adopts the following strategy: Crystallizer section: cooling intensity 0.6-0.8 L / kg, combined with rotary electromagnetic stirring; Bending section: Cooling intensity 0.8-0.85 L / kg, using linear electromagnetic stirring; Straightening section: Cooling intensity 1.2-1.25L / kg, dynamic light compression, compression amount 10-11mm, solid fraction ≥0.6%; Cooling section: Gradient cooling is adopted, with the cooling intensity gradually decreasing from 1.0L / kg to 0.5L / kg.

5. The method according to claim 1, characterized in that, In step S4, the dynamic control strategy for the temperature dimension is as follows: control the cooling gradient ≤15℃ / cm, and the surface temperature fluctuation of the billet ≤±3℃; when the local temperature of the billet is detected to be higher than the set value, automatically increase the cooling intensity of the corresponding section by 0.2L / kg; when the temperature is lower than the set value, reduce the cooling intensity by 0.1L / kg.

6. The method according to claim 1, characterized in that, In step S5, the dynamic control strategy for the composition dimension is as follows: control the composition segregation index ≤ 1.05 and the inclusion size ≤ 5μm; when the segregation index > 1.05, automatically increase the electromagnetic stirring frequency by 2-3Hz and increase the light reduction by 2mm; the modifier is an Al-Ti-B alloy, and the addition amount is 0.02%-0.03%.

7. The method according to claim 1, characterized in that, In step S6, the intelligent closed-loop control method is as follows: real-time acquisition of billet geometric dimension data, billet internal defect data, chemical composition data, and temperature field data. The acquired data is input into an intelligent decision-making system based on an LSTM neural network. The intelligent decision-making system calculates the current grain size and inclusion status using a multi-physics coupling model, compares it with the target value, dynamically adjusts the electromagnetic stirring intensity and the amount of modifier added based on the comparison result, and feeds the control result back to the multi-physics coupling model for iterative optimization until the billet quality meets the standard.

8. The method according to claim 7, characterized in that, The specific steps for dynamically adjusting the electromagnetic stirring intensity and the amount of modifier added are as follows: when the grain size of the billet is detected to be >5μm, the electromagnetic stirring current is automatically increased by 50-100A and the strong cooling time is extended by 5-10s; when the size of the inclusions is detected to be >5μm, the amount of modifier added is increased by 0.005%-0.01%.