A design method of CrMo high-strength and high-toughness steel based on coupling of atomic physical characteristics and machine learning and a preparation method thereof

Abstract

The application belongs to the technical field of CrMo high strength and toughness steel design, and discloses a design method of CrMo high strength and toughness steel based on atomic physical feature coupling machine learning and a preparation method thereof. By introducing atomic scale physical and chemical features (especially "weighted variance" features) as key inputs of machine learning, the application accurately designs a CrMo alloy steel which still maintains excellent plasticity under ultrahigh strength (>1400MPa) or maintains high strength under high toughness. The application first introduces atomic physical and chemical feature parameters such as "ionization energy variance" into steel design, reveals the strengthening mechanism at the electron level, avoids blind component trial and error, and significantly reduces the research and development cost.

Description

Technical Field

[0001] This invention relates to the field of CrMo high-strength and high-toughness steel design technology, and in particular to a design method and preparation method of CrMo high-strength and high-toughness steel based on atomic physical characteristics coupled with machine learning. Background Technology

[0002] Medium-carbon chromium-molybdenum steel (such as the classic 42CrMo / AISI 4140) is widely used in key load-bearing components of high-end equipment such as automotive axles, oil drilling tools, and wind turbine gears due to its excellent comprehensive mechanical properties, hardenability, and machinability. As modern industry increasingly demands lightweight equipment and adaptability to extreme working conditions, higher expectations are placed on structural steel materials: they must possess ultra-high strength (e.g., tensile strength > 1200 MPa) to withstand enormous loads, and also possess excellent ductility and toughness to prevent sudden brittle fracture. However, existing technologies face insurmountable bottlenecks in improving the performance of CrMo steel, mainly in the following three aspects: (1) Existing technologies mainly rely on adjusting the content of macroscopic alloying elements such as C, Cr, and Mo to control performance. Although increasing the carbon content or the total amount of alloying elements can improve strength, this often comes at the cost of sacrificing plasticity and impact toughness (i.e., the "strong plasticity inversion" phenomenon). For example, when pursuing a strength of over 1100 MPa, the elongation after fracture of traditional 42CrMo steel usually drops significantly to below 10%, and its low-temperature impact toughness deteriorates sharply, making it difficult to meet the requirements of complex service environments such as deep sea and extreme cold.

[0003] (2) Traditional materials research and development follows an Edison-style trial-and-error process of "experience-design-smelting-testing-correction". This method is not only time-consuming and costly, but also often gets stuck in local optima. In recent years, although machine learning has been introduced into materials research and development, most existing solutions only use "macroscopic composition percentage" as input variable. This "black box" data fitting ignores the physical and chemical interactions between elements at the atomic scale, resulting in poor extrapolation ability of the model, failing to truly reveal the physical origin of material performance, and making it difficult to discover ultra-high performance formulations outside the conventional composition range.

[0004] (3) At the level of microscopic strengthening mechanism, existing technologies overemphasize macroscopic solid solution strengthening and second-phase precipitation strengthening, while neglecting the profound influence of the statistical distribution of atomic-scale physical characteristics (such as atomic radius, electronegativity, ionization energy, etc.) on the lattice distortion field. In fact, the interaction between alloying elements at the electronic level (such as the change in electron cloud overlap density caused by the difference in ionization energy, and the local stiffness mismatch caused by the difference in bulk modulus) is the key internal factor determining the ultimate strength of the material, but this key dimension has long been blank in the existing CrMo steel design system.

[0005] In summary, how to break through the limitations of traditional trial-and-error methods and start from the underlying logic of atomic physical characteristics to accurately design a new type of CrMo steel that can break through the bottleneck of strong plasticity inversion and still maintain excellent plasticity under ultra-high strength is a technical problem that urgently needs to be solved. Summary of the Invention

[0006] The purpose of this invention is to provide a design and preparation method for CrMo high-strength and high-toughness steel based on atomic physical features coupled with machine learning, overcoming the low efficiency of traditional trial-and-error methods. By introducing atomic-scale physicochemical features (especially "weighted variance" features) as key inputs for machine learning, a CrMo alloy steel that maintains excellent plasticity at ultra-high strength (>1400MPa) or high strength at high toughness can be precisely designed.

[0007] To achieve the above-mentioned objectives, the present invention provides the following technical solution: This invention provides a design method for CrMo high-strength and high-toughness steel based on atomic physical features coupled with machine learning, comprising the following steps: (1) Collect data on CrMo steel, including the composition of CrMo steel, the heat treatment process of CrMo steel and the properties of CrMo steel; (2) Perform feature engineering on the data collected in step (1) to convert the composition of CrMo steel into atomic physical and chemical characteristics, and then calculate the mean and variance to obtain the variance of atomic physical and chemical characteristics; use the variance of atomic physical and chemical characteristics and the heat treatment process of CrMo steel as input values, and use XGBoost and RCDS methods to screen features to obtain the screened features. (3) Use the features selected in step (2) to train a machine learning model to establish a machine learning model of composition-process-performance; (4) Using SHAP analysis to explain the machine learning model of composition-process-performance; (5) The machine learning model of composition-process-performance is optimized by Bayesian optimization and genetic algorithm to obtain candidate compositions of CrMo high strength and toughness steel.

[0008] Preferably, the CrMo steel in step (1) comprises 17 elements, specifically: C, N, Si, Al, Mn, Cr, Ti, Co, Cu, B, Mo, Nb, S, V, Ni, P and Fe; the heat treatment process of the CrMo steel includes austenitization, tempering, normalizing and quenching; the properties of the CrMo steel include tensile strength and elongation.

[0009] Preferably, after collecting the data of CrMo steel in step (1), the collected data is further cleaned to remove outliers and null values, and then the data distribution analysis is performed.

[0010] Preferably, the machine learning model in step (3) includes the BP model, the ANN model, and the MLP model.

[0011] As a preferred option, the specific process of step (3) includes establishing a machine learning model of composition-process-performance using the screened features as input values ​​and tensile strength and plasticity as output values.

[0012] This invention also provides a CrMo high-strength and high-toughness steel obtained by the above-mentioned design method of CrMo high-strength and high-toughness steel based on atomic physical feature coupling machine learning.

[0013] This invention also provides a method for preparing CrMo high-strength and high-toughness steel, comprising the following steps: The raw materials were weighed according to the composition of CrMo high strength and toughness steel, and the raw materials were melted into ingots by vacuum induction melting, and then forged to obtain forgings; the forgings were quenched to obtain a martensitic matrix; the martensitic matrix was tempered to obtain CrMo high strength and toughness steel.

[0014] Preferably, the forging temperature is 1100~1200℃; the forging time is 1~3h.

[0015] Preferably, the quenching conditions are: temperature of 820~870℃, holding time of 30~60min, and cooling method of water cooling.

[0016] Preferably, the tempering process is performed using either method one or method two. The conditions for Method 1 are: temperature of 550~580℃, holding time of 40~60min, and cooling method of air cooling; The conditions for Method 2 are: temperature of 430~460℃, holding time of 40~60min, and cooling method of air cooling.

[0017] In this invention, a composition design strategy driven by "atomic physical and chemical characteristic variance" is employed. Unlike traditional methods that only input component proportions, this invention constructs a feature engineering system containing physical descriptors such as the weighted variance of the first ionization energy and the weighted variance of the bulk modulus. This system is then combined with machine learning methods to optimize the composition of CrMo high-strength and high-toughness steel. The design principle lies in: utilizing the "electron friction" effect generated by the ionization energy variance to enhance solid solution strengthening, and utilizing the microscopic "stiffness mismatch" effect generated by the bulk modulus variance to improve plasticity, thereby achieving synergistic control of strength and plasticity.

[0018] As can be seen from the above technical solution, compared with the prior art, the present invention has the following beneficial effects: (1) Breaking through performance limits: The steel prepared by this invention achieves an ultra-high tensile strength of 1420MPa under tempering at 430~460℃, while still retaining an engineering elongation of nearly 20.37%; under tempering at 550~580℃, it achieves an excellent match of 1050MPa strength and 16.73% elongation, which is significantly better than traditional 42CrMo steel (usually the elongation is difficult to exceed 12% at the 1000MPa level).

[0019] (2) Scientific design and shortened cycle: This invention introduces atomic physical and chemical characteristic parameters such as "ionization energy variance" into steel design for the first time, revealing the strengthening mechanism at the electronic level, avoiding blind component trial and error, and significantly reducing R&D costs.

[0020] (3) Strong process adaptability: The composition system of the present invention has good response characteristics to heat treatment temperature. It can be flexibly switched between "high strength and toughness" and "ultra-high strength" by adjusting the tempering temperature to meet the needs of different working conditions. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0022] Figure 1 This is a flowchart illustrating the design method of CrMo high-strength and high-toughness steel based on atomic physical features coupled with machine learning, as described in this invention. Figure 2 The mechanical property test results of the CrMo high-strength and high-toughness steel in Example 2 are shown. Figure 3 The image shows the metallographic structure of the CrMo high-strength and high-toughness steel from Example 2. Figure 4 The mechanical property test results are for the CrMo high-strength and high-toughness steel in Example 3; Figure 5 The image shows the metallographic structure of the CrMo high-strength and high-toughness steel in Example 3. Figure 6 The mechanical property test results are for the CrMo high-strength and high-toughness steel in Example 4. Detailed Implementation

[0023] This invention provides a design method for CrMo high-strength and high-toughness steel based on atomic physical features coupled with machine learning, comprising the following steps: (1) Collect data on CrMo steel, including the composition of CrMo steel, the heat treatment process of CrMo steel and the properties of CrMo steel; (2) Perform feature engineering on the data collected in step (1) to convert the composition of CrMo steel into atomic physical and chemical characteristics, and then calculate the mean and variance to obtain the variance of atomic physical and chemical characteristics; use the variance of atomic physical and chemical characteristics and the heat treatment process of CrMo steel as input values, and use XGBoost and RCDS methods to screen features to obtain the screened features. (3) Use the features selected in step (2) to train a machine learning model to establish a machine learning model of composition-process-performance; (4) Using SHAP analysis to explain the machine learning model of composition-process-performance; (5) The machine learning model of composition-process-performance is optimized by Bayesian optimization and genetic algorithm to obtain candidate compositions of CrMo high strength and toughness steel.

[0024] In this invention, the composition of the CrMo steel in step (1) includes 17 elements, specifically: C, N, Si, Al, Mn, Cr, Ti, Co, Cu, B, Mo, Nb, S, V, Ni, P and Fe; the heat treatment process of the CrMo steel includes austenitization, tempering, normalizing and quenching; the properties of the CrMo steel include tensile strength and elongation.

[0025] In this invention, after collecting the data of CrMo steel in step (1), the collected data is further cleaned to remove outliers and null values, and then the data distribution analysis is performed.

[0026] In this invention, the atomic physical and chemical characteristics described in step (2) include bulk modulus, ionization energy, relative atomic mass, molar volume, thermal conductivity, melting point, boiling point, enthalpy of fusion, enthalpy of vaporization, enthalpy of atomization, atomic radius, van der Waals radius, covalent radius, number of valence electrons, lattice constant, Pauling electronegativity, Sanderson electronegativity, Allred Rochow electronegativity, and effective nuclear charge number.

[0027] In this invention, the machine learning model in step (3) includes the BP model, the ANN model, and the MLP model.

[0028] In this invention, the specific process of step (3) includes establishing a machine learning model of composition-process-performance using the screened features as input values ​​and tensile strength and plasticity as output values.

[0029] This invention also provides a CrMo high-strength and high-toughness steel obtained by the above-mentioned design method of CrMo high-strength and high-toughness steel based on atomic physical feature coupling machine learning.

[0030] This invention also provides a method for preparing CrMo high-strength and high-toughness steel, comprising the following steps: The raw materials were weighed according to the composition of CrMo high strength and toughness steel, and the raw materials were melted into ingots by vacuum induction melting, and then forged to obtain forgings; the forgings were quenched to obtain a martensitic matrix; the martensitic matrix was tempered to obtain CrMo high strength and toughness steel.

[0031] In this invention, the forging temperature is preferably 1100~1200℃, more preferably 1150~1200℃, and even more preferably 1200℃; the forging time is preferably 1~3h, more preferably 2~3h, and even more preferably 2h.

[0032] In this invention, the quenching conditions are as follows: the temperature is preferably 820~870℃, more preferably 840~860℃, and even more preferably 850℃; the holding time is preferably 30~60min, more preferably 40~50min, and even more preferably 45min; and the cooling method is preferably water cooling.

[0033] In this invention, the tempering process is performed using either method one or method two; The conditions for Method 1 are as follows: the temperature is preferably 550~580℃, more preferably 555~570℃, and even more preferably 560℃; the heat preservation time is preferably 40~60min, more preferably 45~55min, and even more preferably 50min; and the cooling method is preferably air cooling. The conditions for Method 2 are as follows: the temperature is preferably 430~460℃, more preferably 440~455℃, and even more preferably 450℃; the heat preservation time is preferably 40~60min, more preferably 45~55min, and even more preferably 50min; and the cooling method is preferably air cooling.

[0034] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] Example 1

[0036] This embodiment provides a design method for CrMo high-strength and high-toughness steel based on atomic physical features coupled with machine learning. The flowchart is shown below. Figure 1 As shown, it includes the following steps: (1) Data on 211 CrMo steels were collected through literature review, including the composition of CrMo steel, the heat treatment process of CrMo steel and the properties of CrMo steel; the composition of CrMo steel includes 17 elements, namely: C, N, Si, Al, Mn, Cr, Ti, Co, Cu, B, Mo, Nb, S, V, Ni, P and Fe; the heat treatment process of CrMo steel includes austenitization, tempering, normalizing and quenching; the properties of CrMo steel include tensile strength and elongation. The collected data was cleaned to remove outliers and null values, and then a data distribution analysis was performed. The data distribution analysis showed that the composition, heat treatment process and performance all approximately followed a normal distribution. (2) Perform feature engineering on the data collected in step (1) to convert the composition of CrMo steel into atomic physical and chemical characteristics. The atomic physical and chemical characteristics include bulk modulus, ionization energy, relative atomic mass, molar volume, thermal conductivity, melting point, boiling point, enthalpy of fusion, enthalpy of vaporization, enthalpy of atomization, atomic radius, van der Waals radius, covalent radius, number of valence electrons, lattice constant, Pauling electronegativity, Sanderson electronegativity, Allred Rochow electronegativity, and effective nuclear charge. Then, calculate the mean and variance to obtain the variance of the atomic physical and chemical characteristics. Using the variance of atomic physical and chemical characteristics and the heat treatment process of CrMo steel as input values, two methods were used for feature selection: XGBoost (multi-round XGBoost algorithm to obtain the feature importance ranking of strength and plasticity) and RCDS (clustering to construct a global feature map, and constructing a large number of sub-models based on linear (Lasso) and nonlinear (XGBoost) strict nested cross-validation evaluation) to obtain the selected features. (3) Use the features selected in step (2) to train a machine learning model to establish a machine learning model of composition-process-performance; Specifically, the models include BP model, ANN model and MLP model, which use the selected features as input values ​​and tensile strength and plasticity as output values ​​to establish a machine learning model of composition-process-performance. (4) The SHAP analysis method was used to interpret the machine learning model of composition-process-performance; the results showed that the model predictions were in high agreement with the actual values, and the determination coefficient R of the tensile strength prediction was high. 2 The R value for plasticity prediction reached 0.93. 2 The value was 0.89, which verifies that the model has good stability and reliability; (5) The composition-process-performance machine learning model was optimized using Bayesian optimization and genetic algorithm to obtain the Pareto fronts of tensile strength and plasticity after 1000 iterations. The prediction results show that the optimized CrMo steel material has better tensile strength and plasticity than existing materials, and exhibits good comprehensive mechanical properties, providing a theoretical basis for the synergistic optimization of material composition design and heat treatment process.

[0037] Example 2

[0038] Using the method of Example 1, preferred components (numbered 1#) were screened out. The specific chemical composition (mass fraction) is as follows: C: 0.45%, Si: 0.17%, Mn: 0.80%, Cr: 0.90%, Mo: 0.25%, Ni: 0.71%, Al: 0.07%, Ti: 0.03%, N: 0.01%, Co: 0.17%, balance Fe.

[0039] The preparation method includes the following steps: The raw materials were weighed according to the composition of CrMo high strength and toughness steel. The raw materials were vacuum induction melted into ingots, and then forged at 1200℃ for 2 hours to obtain forgings. The forgings were heated to 850℃ for quenching, held for 45 minutes, and then water-cooled to obtain a martensitic matrix. The martensitic matrix was tempered at 560℃ for 50 minutes and then air-cooled to obtain CrMo high strength and toughness steel.

[0040] The mechanical property test results of the CrMo high-strength and high-toughness steel in this embodiment are as follows: Figure 2 As shown. By Figure 2 It can be seen that under tempering at 560℃, the tensile strength of the material reaches 1050MPa and the elongation after fracture is as high as 16.73%.

[0041] The metallographic structure of the CrMo high-strength and high-toughness steel in this embodiment is shown below. Figure 3 As shown. By Figure 3 It can be seen that the metallographic structure is uniform and fine tempered sorbite. Carbides are dispersed in the ferrite matrix, and no network carbides are observed. This microstructure ensures excellent plasticity.

[0042] Example 3

[0043] This embodiment is specifically referred to in Embodiment 2, except that the tempering conditions for the martensitic matrix are: holding at 450℃ for 50 minutes, then air cooling after removal from the furnace to obtain CrMo high-strength and tough steel.

[0044] The mechanical property test results of the CrMo high-strength and high-toughness steel in this embodiment are as follows: Figure 4 As shown. By Figure 4It can be seen that when the tempering temperature is reduced to 450℃, the material strength increases explosively, with the tensile strength reaching 1473MPa, while still retaining an elongation of 20.37%.

[0045] The metallographic structure of the CrMo high-strength and high-toughness steel in this embodiment is shown below. Figure 5 As shown. By Figure 5 It can be seen that the metallographic structure is mainly tempered troostite and a small amount of residual martensite.

[0046] Example 4

[0047] Using the method of Example 1, a second group of preferred compositions (numbered 2#) was selected. Microalloying elements V and Nb were introduced, with the following specific chemical composition (mass fraction): C: 0.42%, Si: 0.35%, Mn: 0.52%, Cr: 1.20%, Mo: 0.25%, Ni: 0.41%, V: 0.30%, Nb: 0.05%, Cu: 0.05%, N: 0.01%, Al: 0.04%, Ti: 0.03%, Co: 0.11%, balance Fe.

[0048] The preparation method includes the following steps: The raw materials were weighed according to the composition of CrMo high strength and toughness steel. The raw materials were vacuum induction melted into ingots, and then forged at 1200℃ for 2 hours to obtain forgings. The forgings were heated to 850℃ for quenching, held for 45 minutes, and then water-cooled to obtain a martensitic matrix. The martensitic matrix was tempered at 550℃ for 50 minutes and then air-cooled to obtain CrMo high strength and toughness steel.

[0049] The mechanical property test results of the CrMo high-strength and high-toughness steel in this embodiment are as follows: Figure 6 As shown. By Figure 6 It can be seen that while maintaining high strength, the composition can further improve the resistance to tempering softening by strengthening the precipitation of V and Nb carbonitrides, and the overall score is expected to be better than that of Example 2.

[0050] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A design method for CrMo high-strength and high-toughness steel based on atomic physical features coupled with machine learning, characterized in that, Includes the following steps: (1) Collect data on CrMo steel, including the composition of CrMo steel, the heat treatment process of CrMo steel and the properties of CrMo steel; (2) Perform feature engineering on the data collected in step (1) to convert the composition of CrMo steel into atomic physical and chemical characteristics, and then calculate the mean and variance to obtain the variance of atomic physical and chemical characteristics; Using the variance of atomic physical and chemical characteristics and the heat treatment process of CrMo steel as input values, feature selection was performed using two methods, XGBoost and RCDS, to obtain the selected features. (3) Use the features selected in step (2) to train a machine learning model to establish a machine learning model of composition-process-performance; (4) Using SHAP analysis to explain the machine learning model of composition-process-performance; (5) The machine learning model of composition-process-performance is optimized by Bayesian optimization and genetic algorithm to obtain candidate compositions of CrMo high strength and toughness steel.

2. The design method for CrMo high-strength and high-toughness steel based on atomic physical features coupled with machine learning according to claim 1, characterized in that, The composition of the CrMo steel in step (1) includes 17 elements, specifically: C, N, Si, Al, Mn, Cr, Ti, Co, Cu, B, Mo, Nb, S, V, Ni, P and Fe; the heat treatment process of the CrMo steel includes austenitization, tempering, normalizing and quenching; the properties of the CrMo steel include tensile strength and elongation.

3. The design method for CrMo high-strength and high-toughness steel based on atomic physical features coupled with machine learning according to claim 1, characterized in that, Step (1) involves collecting data on CrMo steel, followed by cleaning the collected data to remove outliers and null values, and then performing data distribution analysis.

4. The design method for CrMo high-strength and high-toughness steel based on atomic physical features coupled with machine learning according to claim 1, characterized in that, The machine learning models mentioned in step (3) include BP model, ANN model and MLP model.

5. The design method for CrMo high-strength and high-toughness steel based on atomic physical features coupled with machine learning according to claim 1, characterized in that, The specific process of step (3) includes using the screened features as input values ​​and tensile strength and plasticity as output values ​​to establish a machine learning model of composition-process-performance.

6. The CrMo high-strength and high-toughness steel obtained by the design method of CrMo high-strength and high-toughness steel based on atomic physical feature coupling machine learning as described in any one of claims 1 to 5.

7. The method for preparing CrMo high-strength and high-toughness steel according to claim 6, characterized in that, Includes the following steps: The raw materials were weighed according to the composition of CrMo high strength and toughness steel, and the raw materials were melted into ingots by vacuum induction melting, and then forged to obtain forgings; the forgings were quenched to obtain a martensitic matrix; the martensitic matrix was tempered to obtain CrMo high strength and toughness steel.

8. The preparation method according to claim 7, characterized in that, The forging temperature is 1100~1200℃; the forging time is 1~3h.

9. The preparation method according to claim 7, characterized in that, The quenching conditions are: temperature of 820~870℃, holding time of 30~60min, and water cooling.

10. The preparation method according to claim 7, characterized in that, The tempering process is performed using either Method 1 or Method 2; The conditions for Method 1 are: temperature of 550~580℃, holding time of 40~60min, and cooling method of air cooling; The conditions for Method 2 are: temperature of 430~460℃, holding time of 40~60min, and cooling method of air cooling.

Images