A Method for Establishing the Relationship between Continuous Casting Process, Microstructure, and Properties Using Interpretable Machine Learning

Abstract

This invention provides a method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning, comprising: Step 1, conducting laboratory-scale experiments to produce continuous casting billet samples under different process conditions; Step 2, sampling the center position of the continuous casting billet and preparing EBSD samples to observe their microstructure morphology; Step 3, extracting different microstructure features and measuring the Vickers hardness values ​​of samples under different process conditions; Step 4, analyzing the correlation between microstructure features and hardness; Step 5, applying a grid method to divide the microstructure feature cloud map and selecting the four items with the highest correlation for contribution superposition to establish the relationship between continuous casting process, microstructure, and properties. This invention can significantly reduce the amount of experimental data required, and can use microstructure morphology to connect the relationship between continuous casting process and properties, establishing the relationship between continuous casting process, microstructure, and properties, thus significantly reducing the cost compared to industrial experiments.

Description

Technical Field

[0001] This invention relates to the field of steel casting technology, and in particular to an interpretable machine learning method for establishing the relationship between continuous casting process, microstructure, and properties. Background Technology

[0002] Continuous casting is a crucial step in steel production, producing slabs, billets, and round billets that can be further processed. These products are used in shipbuilding, aerospace, bridge construction, and other industries. As a key step in the cooling and shaping of molten steel, the quality stability and performance of continuously cast billets have a critical impact on the performance of subsequently processed products.

[0003] Smart factories and intelligent manufacturing technologies are the main focus of my country's efforts over the next decade. With the implementation of a series of major national science and technology projects, my country's traditional metallurgical industry is rapidly developing towards digitalization and intelligence. Technological innovation in continuous casting technology has begun to shift from primarily following to a new stage where following, keeping pace, and even leading coexist. However, due to weak forward-looking basic theoretical research on continuous casting and a lack of materials / process knowledge bases, continuous casting process formulation relies on experience accumulation and simple "trial and error," lacking process design theories and methods oriented towards the quality control needs of continuously cast billets. The unclear evolution law of process-microstructure-property in continuous casting has led to low yield and qualification rates, poor batch stability, high costs, and low self-sufficiency rates for key metal components used in high-end equipment in my country. High-end equipment and key components remain dependent on foreign suppliers.

[0004] CN202311463827.1 discloses an integrated optimization method for continuous casting of niobium-containing austenitic stainless steel based on the phase-field method. This method obtains the solidification microstructure characterization results of the continuously cast stainless steel billet through experiments and simulates and predicts the evolution of dendrite growth and solute distribution at the solid-liquid interface under different cooling rates. Solute concentration and grain boundary data are extracted as initial conditions for the model. Based on solute concentration, precipitate size, grain size, and grain boundary parameters, the grain growth, grain boundary diffusion, and precipitate evolution processes are described. Simultaneously, elastic strain energy is coupled to obtain the equivalent stress-strain distribution of the microstructure over time. Based on the equivalent stress-strain distribution, combined with a damage criterion model, damage factor parameters under different continuous casting process conditions are obtained, thereby optimizing the continuous casting process. This method establishes the relationship between the continuously cast billet and the microstructure, enabling rapid prediction and process optimization of the billet microstructure morphology and crack sensitivity in the actual production of niobium-containing austenitic stainless steel using different continuous casting processes.

[0005] CN202310610803.8 discloses a simulation and prediction method for continuous casting processes. This method integrates parameters such as storage device parameters, steel grade parameters, process parameters, and model parameters, and combines them with temperature simulation models, segregation prediction models, and solidification structure prediction models for simulation. Based on the simulation results, it performs billet quality monitoring, temperature control calibration, and process parameter optimization. This method, based on real production process parameters and combined with database methods, systematically and comprehensively analyzes the relationship between parameters in the continuous casting process and the solute distribution in the continuously cast billet, providing analytical tools and data support for many fields such as process parameter optimization, internal quality monitoring, casting machine capacity development, and new product development in continuous casting.

[0006] CN202310610798.0 discloses a method for temperature simulation and macrostructure prediction in continuous casting processes. This method, based on combinations of continuous casting process parameters and the physical properties of the steel grade, applies a temperature simulation sub-model and a macrostructure prediction sub-model to predict the macrostructure, and stores the temperature simulation and macrostructure prediction results in a MongoDB database. This method visualizes the relationship between continuous casting process temperature and macrostructure as a cloud map, providing effective data support for temperature control calibration, macrostructure control, and casting machine capability development in the continuous casting process, offering analytical tools and multi-sample data.

[0007] CN202311630918.X discloses a method for adjusting continuous casting process parameters. This method involves constructing a training set of continuous casting process parameters; training an initial network model based on this training set to obtain a continuous casting quality prediction model; and generating an adaptive optimization model for the continuous casting process parameters based on continuous casting process evaluation standards and the predicted model. This method, based on the analysis of large-scale data on the relationship between actual continuous casting production process data and the quality of continuously cast billets, derives an adaptive adjustment model for continuous casting process parameters. This makes the adjusted continuous casting process parameters closer to the continuous casting quality requirements, ensuring the quality of continuously cast products and effectively improving the yield of continuously cast products.

[0008] The main drawbacks of the aforementioned existing technologies are that they fail to establish the relationship between continuous casting process, microstructure, and properties. Secondly, they are highly dependent on and require a large amount of continuous casting process parameters and performance data from actual production processes. These existing technologies differ significantly from our invention and fail to solve the technical problem we aim to address. Therefore, we have invented a novel interpretable machine learning method for establishing the relationship between continuous casting process, microstructure, and properties. Summary of the Invention

[0009] The purpose of this invention is to provide a method for establishing the relationship between continuous casting process, microstructure, and performance by optimizing the combination of continuous casting process parameters based on the performance of the target product, thereby significantly improving the consistency of product performance, and establishing an interpretable machine learning method for establishing the relationship between continuous casting process, microstructure, and performance.

[0010] The objective of this invention can be achieved through the following technical measures: a method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning, comprising:

[0011] Step 1: Conduct laboratory scale-down experiments to produce continuous casting billet samples under different process conditions;

[0012] Step 2: Take a sample from the center of the continuously cast billet and prepare an EBSD sample to observe its microstructure.

[0013] Step 3: Extract different tissue characteristics and measure the Vickers hardness value of samples under different process conditions;

[0014] Step 4: Analyze the correlation between tissue characteristics and hardness;

[0015] Step 5: Apply the grid method to divide the microstructure feature cloud map and select the four items with the highest correlation to make contribution superposition to establish the relationship between continuous casting process, microstructure and properties.

[0016] The objective of this invention can also be achieved through the following technical measures:

[0017] In step 1, when producing continuous casting billet samples under different process conditions, the billet pulling speed is 0.5 to 2 m / min, the step size interval is 0.3 mm / min, and the cooling water flow rate in the secondary cooling zone is 30 to 50 L / (t·min).

[0018] In step 1, the steel liquid composition range is: C 350~550×10 -6 wt%, Si accounts for 45-70×10 -6 wt%, Mn accounts for 1550~2350×10 -6 wt%, P accounts for 130~190×10 -6 wt%, TS accounts for 70-110×10 -6 wt%, Al accounts for 260-400×10 -6 wt%, Cr content 140~210×10 -6 wt%, Ni accounts for 75–110 × 10 -6 wt%, Cu accounts for 75-110×10 -6 wt%, Mo accounts for 30~50×10 -6 wt%.

[0019] In step 2, small squares were taken from the central region of the continuously cast billet under different process conditions, and EBSD samples were prepared by electrolytic polishing and their microstructure was observed.

[0020] In step 2, the electrolytic polishing solution used is a perchloric acid-alcohol mixture with a volume ratio of 1:3. The electrolytic polishing voltage is 20-21V, the electrolysis time is 10-12s, and the electrolytic polishing temperature is 0-1℃.

[0021] In step 2, a small square of 10 is taken from the center area of ​​the continuously cast billet. 10 A small square, 10mm in diameter.

[0022] In step 3, the extracted microstructure features include eight characteristic indicators: average grain area size, average grain boundary angle, proportion of large-angle grain boundaries, proportion of dynamically recrystallized grain size, average geometrically required dislocation density, proportion of dynamically recovered grains, aspect ratio, and proportion of deformed grain size.

[0023] In step 4, the SHAP analysis method is applied to analyze the correlation between tissue characteristics and hardness.

[0024] The interpretable machine learning method for establishing the relationship between continuous casting process, microstructure, and properties also includes, after step 4, using interpolation to draw cloud maps of single microstructure features of samples under different process conditions.

[0025] In step 5, the cloud map of a single microstructure feature is divided into regions using the grid method, and the grid is superimposed based on the contribution of the single microstructure feature to the performance. Finally, the correctness of the SHAP analysis method is verified, and the optimal combination of continuous casting process parameters for the corresponding steel grade under the target performance in the continuous casting process is obtained.

[0026] The interpretable machine learning-based method for establishing the continuous casting process-microstructure-property relationship in this invention can provide the optimal range of continuous casting process parameters for specific types of steel products and give specific combinations of continuous casting process parameters based on performance requirements. This invention employs scaled-down experiments under laboratory conditions to conduct controlled variable experiments on two key indicators in the continuous casting process: billet drawing speed and cooling water flow rate in the secondary cooling zone. Continuous casting billets under different process parameter combinations are produced, and the microstructure and hardness of the billets are observed and tested. Combining the machine learning SHAP (SHapley Additive exPlanations) analysis method, the types of microstructure parameters that have a significant impact on performance are identified, and a grid method is applied to analyze and verify the corresponding microstructure parameters under different process parameters. This method requires less data and, combined with interpretable machine learning, clearly elucidates the continuous casting process-microstructure-property relationship. Compared with existing technologies, this invention has the following advantages:

[0027] (1) The method described in this invention can significantly reduce the amount of experimental data required.

[0028] (2) The method described in this invention can establish the relationship between continuous casting process and performance by means of microstructure morphology.

[0029] (3) The method described in this invention significantly reduces the cost compared to industrial experiments. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the correlation analysis of the screening results of key tissue characteristic factors in a specific embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of the SHAP values ​​of key microstructure characteristics that affect the properties of continuously cast billets in a specific embodiment of the present invention.

[0032] Figure 3 This is a tissue feature cloud map obtained by the mesh division difference method in a specific embodiment of the present invention;

[0033] Figure 4 This is a schematic diagram illustrating the contribution analysis of a single tissue feature to performance in a specific embodiment of the present invention;

[0034] Figure 5 This is a schematic diagram of the contribution of a single organizational feature to performance by mesh overlay in a specific embodiment of the present invention;

[0035] Figure 6 This is a comparison diagram of the region with excellent theoretical performance and the region with excellent actual performance under the synergistic influence of multiple tissue features in a specific embodiment of the present invention.

[0036] Figure 7 This is a flowchart of a specific embodiment of the method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning according to the present invention. Detailed Implementation

[0037] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0038] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, and / or combinations thereof.

[0039] like Figure 7 As shown, Figure 7This is a flowchart illustrating the method for establishing the continuous casting process-microstructure-property relationship using interpretable machine learning according to the present invention. The method includes:

[0040] Step 1: Conduct laboratory scaled-down continuous casting experiments to obtain continuous casting billet samples under different process conditions;

[0041] The billet drawing speed is 0.5–2 m / min, with a step size interval of 0.3 mm / min. The cooling water flow rate in the secondary cooling zone is 30–50 L / (t·min). The steel composition range is: C 350–550 × 10⁻⁶. -6 wt%, Si accounts for 45-70×10 -6 wt%, Mn accounts for 1550~2350×10 -6 wt%, P accounts for 130~190×10 -6 wt%, TS accounts for 70-110×10 -6 wt%, Al accounts for 260-400×10 -6 wt%, Cr content 140~210×10 -6 wt%, Ni accounts for 75–110 × 10 -6 wt%, Cu accounts for 75-110×10 -6 wt%, Mo accounts for 30~50×10 -6 wt%.

[0042] Step 2: Take 10 samples from the central region of the continuously cast billet under different process conditions. 10 10mm cubes were used to prepare EBSD (Electron Back Scatter Diffraction) samples using electropolishing and their tissue morphology was observed.

[0043] The electrolytic solution used for electropolishing is a perchloric acid-alcohol mixture with a volume ratio of 1:3. The electropolishing voltage is 20-21V, the electrolysis time is 10-12s, and the electropolishing temperature is 0-1℃.

[0044] Step 3: Use Aztec software to extract different tissue characteristics and measure the Vickers hardness values ​​of samples under different process conditions.

[0045] Tissue feature extraction includes eight feature indicators: average grain area size, average grain boundary angle, proportion of large-angle grain boundaries, proportion of dynamically recrystallized grain size, average geometrically necessary dislocation density, proportion of dynamically recovered grains, aspect ratio, and proportion of deformed grain size.

[0046] Step four: Apply the SHAP analysis method to analyze the correlation between tissue characteristics and hardness;

[0047] Step 5: Use interpolation to draw cloud maps of single tissue features of samples under different process conditions;

[0048] Step 6: Apply the grid method to divide the microstructure feature cloud map and select the four items with the highest correlation to make contribution superposition to establish the relationship between continuous casting process, microstructure and properties.

[0049] The cloud map was divided into regions using the grid method, and the grid was superimposed based on the contribution of a single microstructure to the performance. Finally, the correctness of the SHAP analysis method was verified, and the optimal combination of continuous casting process parameters for the corresponding steel grade under the target performance in the continuous casting process was obtained.

[0050] In a specific embodiment of the present invention, the method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning includes the following steps:

[0051] Step 1: Conduct laboratory scaled-down continuous casting experiments to obtain continuous casting billet samples under different process conditions.

[0052] The molten steel used in the continuous casting process was prepared in the laboratory, and its raw materials came from a steel plant in Jiangsu Province. The elemental composition (unit: 10) is as follows: -6 (×wt%) is shown in the table below:

[0053] Table 1. Elemental Composition of Molten Steel in Tundish

[0054]

[0055] Experiments were conducted with a billet pulling speed of 0.5–2 m / min and a step size interval of 0.3 mm / min, and a cooling water flow rate of 30–50 L / (t·min) and a step size interval of 5 L / (t·min) in the secondary cooling zone. A total of 30 samples were obtained.

[0056] Step 2: Take 10 samples from the central region of the continuously cast billet under different process conditions. 10 EBSD samples were prepared by electropolishing using 10 mm cubes. The electropolishing solution was a 1:3 volume ratio of perchloric acid to alcohol mixture. The electropolishing voltage was 20–21 V, the electropolishing time was 10–12 s, and the electropolishing temperature was 0–1 ℃.

[0057] Step 3: Use Aztec software to extract eight characteristic indicators, including average grain area, average grain boundary angle, proportion of large-angle grain boundaries, proportion of dynamically recrystallized grain size, average geometrically required dislocation density, proportion of dynamically recovered grains, aspect ratio, and proportion of deformed grain size.

[0058] Step 4: Measure the Vickers hardness value of the samples under different process conditions.

[0059] The SHAP analysis method was used to analyze the correlation between tissue characteristics and stiffness. The correlation analysis of the screening results of key tissue characteristic factors is shown in the attached figure. Figure 1 As shown, the accuracy of the experimental set is 95%, and the accuracy of the test set is 85%, which is relatively high.

[0060] The key microstructure characteristics affecting the properties of continuously cast billets were obtained using the SHAP analysis method, as shown in the attached figure. Figure 2 As shown, among the eight microstructural characteristic factors analyzed, the top four with the highest correlation were selected. Among them, the proportion of large-angle grain boundaries is positively correlated with the performance of continuously cast billets, while the average grain size, the proportion of dynamically recrystallized grains, and the average grain boundary angle are negatively correlated with the performance of continuously cast billets.

[0061] Step 5: Apply the interpolation method to draw cloud maps of four key tissue characteristic factors, and then apply the grid method to divide them, as shown in the attached figure. Figure 3 As shown.

[0062] Step 6: Based on the correlation results between single microstructure characteristics and performance obtained in Step 4, screen out the grid numbers that are beneficial to performance improvement. If all three microstructure characteristics under a certain grid are beneficial to performance, then it is determined that the continuous casting billet performance is better under the process conditions corresponding to this grid.

[0063] The grid numbers where all three organizational characteristics are beneficial to performance are selected by the red box. The contribution analysis results are attached. Figure 4 As shown.

[0064] The grid numbers were organized into a 10×10 grid for comparison with the hardness field distribution. The contributing grids are superimposed as shown in the attached figure. Figure 5 As shown.

[0065] A schematic diagram of the hardness field distribution under all process conditions was drawn using interpolation, and compared with the attached diagram. Figure 5 The grid areas in the image are overlaid for comparison, and the comparison results are shown in the attached figure. Figure 6 As shown.

[0066] The comparison shows that the regions with excellent theoretical performance under the synergistic influence of multiple organizational features are almost identical to the regions with excellent actual performance, indicating that the relationship between continuous casting process, microstructure, and performance is established based on interpretable machine learning.

[0067] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0068] Except for the technical features described in the specification, all other technologies are known to those skilled in the art.

Claims

1. A method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning, characterized in that, The interpretable machine The methods for establishing the microstructure-property relationship in continuous casting processes include: Step 1: Conduct laboratory scale-down experiments to produce continuous casting billet samples under different process conditions; Step 2: Take a sample from the center of the continuously cast billet and prepare an EBSD sample to observe its microstructure. Step 3: Extract different tissue characteristics and measure the Vickers hardness value of samples under different process conditions; Step 4: Analyze the correlation between tissue characteristics and hardness; Step 5: Apply the grid method to divide the microstructure feature cloud map and select the four key microstructure feature factors with the highest correlation affecting the performance of the continuously cast billet. Perform contribution superposition to obtain the optimal combination of continuous casting process parameters for the corresponding steel grade under the target performance in the continuous casting process. Based on the optimal combination of continuous casting process parameters and the key microstructure feature factors affecting the performance of the continuously cast billet, establish the relationship between continuous casting process, microstructure and performance.

2. The method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning according to claim 1, characterized in that, In step 1, when producing continuous casting billet samples under different process conditions, the billet pulling speed is 0.5 to 2 m / min, the step size interval is 0.3 mm / min, and the cooling water flow rate in the secondary cooling zone is 30 to 50 L / (t·min).

3. The method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning according to claim 2, characterized in that, In step 1, the steel liquid composition range is: C 350~550×10 -6 wt%, Si accounts for 45-70×10 -6 wt%, Mn accounts for 1550~2350×10 -6 wt%, P accounts for 130~190×10 -6 wt%, TS accounts for 70-110×10 -6 wt%, Al accounts for 260-400×10 -6 wt%, Cr content 140~210×10 -6 wt%, Ni accounts for 75–110 × 10 -6 wt%, Cu accounts for 75-110×10 -6 wt%, Mo accounts for 30~50×10 -6 wt%.

4. The method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning according to claim 1, characterized in that, In step 2, small squares were taken from the central region of the continuously cast billet under different process conditions, and EBSD samples were prepared by electrolytic polishing and their microstructure was observed.

5. The method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning according to claim 4, characterized in that, In step 2, the electrolytic polishing solution used is a perchloric acid-alcohol mixture with a volume ratio of 1:

3. The electrolytic polishing voltage is 20-21V, the electrolysis time is 10-12s, and the electrolytic polishing temperature is 0-1℃.

6. The method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning according to claim 1, characterized in that, In step 3, the extracted microstructure features include eight characteristic indicators: average grain area size, average grain boundary angle, proportion of large-angle grain boundaries, proportion of dynamically recrystallized grain size, average geometrically required dislocation density, proportion of dynamically recovered grains, aspect ratio, and proportion of deformed grain size.

7. The method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning according to claim 1, characterized in that, In step 4, the SHAP analysis method is applied to analyze the correlation between tissue characteristics and hardness.

8. The method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning according to claim 7, characterized in that, The interpretable machine learning method for establishing the relationship between continuous casting process, microstructure, and properties also includes, after step 4, using interpolation to draw cloud maps of single microstructure features of samples under different process conditions.

9. The method for establishing the relationship between continuous casting process, microstructure, and properties using interpretable machine learning according to claim 8, characterized in that, In step 5, the cloud map of a single microstructure feature is divided into regions using the grid method, and the grid is superimposed based on the contribution of the single microstructure feature to the performance. The hardness field distribution under the process conditions is plotted using the interpolation method, and compared with the grid region of the contributing superimposed grid. By comparing whether the theoretically excellent performance region under the synergistic influence of multiple microstructure features is close to the actual performance cloud map region, the correctness of the SHAP analysis method is finally verified, and the optimal combination of continuous casting process parameters for the corresponding steel grade under the target performance in the continuous casting process is obtained.

Images