A data-driven multi-objective performance inverse design optimization method for wrought nickel-based superalloys

Abstract

The present application relates to a kind of data-driven deformation nickel-based superalloy multi-objective performance reverse design optimization method, belong to metal material design development technical field, including the following steps: based on first principle calculation stable high-strength deformation nickel-based superalloy system, and based on this, based on nickel-based superalloy field knowledge, construct deformation nickel-based superalloy composition design space;Based on deformation nickel-based superalloy composition design space, using empirical formula is combined with thermodynamics high-throughput calculation to reduce deformation nickel-based superalloy composition design space, obtain the reduced deformation nickel-based superalloy composition design space, and based on this, based on machine learning and genetic algorithm, establish deformation nickel-based superalloy reverse design model, filter target performance deformation nickel-based superalloy.Compared with prior art, the present application proposes a kind of fusion cross-scale calculation, field knowledge constraint and machine learning reverse design, realize the directional development method of high-strength and high-toughness nickel-based superalloy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of metal material design and development, and particularly relates to a data-driven multi-objective performance reverse design optimization method for deformed nickel-based high-temperature alloy. BACKGROUND

[0002] Nickel-based high-temperature alloy has become the preferred material for key components such as turbine disks and turbine blades of an aero-engine, due to its high strength and creep rupture life, as well as excellent resistance to corrosion, oxidation and fatigue at high temperatures.

[0003] In the process of data-driven design of nickel-based high-temperature alloy, it is difficult to reversely design the alloy composition according to the target performance, because the dimension of the composition vector is much larger than that of the predicted performance vector, and one performance parameter corresponds to multiple alloy compositions.

[0004] The current design method of nickel-based high-temperature alloy mainly includes: first principle calculation (such as density functional theory DFT) to predict the electronic structure, phase stability and basic physical properties of materials from the atomic scale by solving quantum mechanics equation. In the design of high-temperature alloy, this method can be used to evaluate the influence of element doping on thermodynamic stability, lattice distortion, interface energy and defect formation energy; the high-throughput thermodynamic technology based on CALPHAD (phase diagram calculation) can realize rapid phase equilibrium prediction and composition-technology window screening by constructing the Gibbs free energy database of multi-component alloy system. Typical applications include batch evaluation of the change of γ' phase content, phase composition and content with alloy composition and temperature in the thermodynamic calculation module, TCP phase precipitation tendency; the precipitation phase calculation module can also calculate the average particle size, volume fraction and TTT / CCT curve of the precipitated phase; machine learning (Machine Learning, ML) can realize the rapid prediction of material performance by establishing a nonlinear mapping model of “composition-technology-performance” (such as support vector machine, random forest), and the reverse design framework (such as Bayesian optimization, genetic algorithm) can reversely generate candidate alloy composition for target performance (yield strength / tensile strength / elongation).

[0005] However, the first principle calculation consumes huge resources and is only suitable for small-scale models (usually <1000 atoms), and it is difficult to directly correlate macroscopic performance (such as high-temperature yield strength and room-temperature elongation). The thermodynamic phase diagram calculation depends on the accuracy of experimental data, cannot handle non-equilibrium dynamic processes, and has limited prediction ability for complex performance (such as high-temperature yield strength and room-temperature elongation). Machine learning requires high quality of data set, and mainly relies on single performance optimization, lacks systematic coordination ability for multi-objective conflict of high-temperature alloy (such as the trade-off between strength and plasticity), and has insufficient model interpretability.

[0006] CN202310956599.5 discloses a high-temperature alloy composition and process design method, comprising the following steps: collecting data of a certain brand of high-temperature alloy manual, experiment, literature, and constructing a database; using machine learning algorithm to construct corresponding performance prediction model, using Bayesian optimization technology, taking the required optimized alloy composition range as the search range and taking the comprehensive performance index as the optimization target, realizing alloy composition optimization to obtain excellent comprehensive performance, realizing process optimization of small sample data comprehensive performance through machine learning algorithm, using optimized alloy composition and process to cast alloy, verifying the optimized comprehensive performance, training the model by collecting data as the training set, analyzing the influence law of alloy composition and processing technology on the structure performance, designing the composition and process of the optimized alloy, reducing the cost of manpower and material resources, and reducing the research and development cycle of cast nickel-based high-temperature alloy. However, the method in CN202310956599.5 mainly relies on domain knowledge, and the physical interpretability and engineering feasibility of composition design cannot be ensured.

[0007] CN202410138800.3 discloses a method and device for reverse design of target performance nickel-based high-temperature alloy composition, relating to the technical field of nickel-based high-temperature alloy composition design, which comprises: obtaining formed nickel-based high-temperature alloy composition data and performance prediction model; using K-means algorithm to cluster the formed nickel-based high-temperature alloy composition data to obtain several nickel-based high-temperature alloy data sets; using each nickel-based high-temperature alloy data set to train variational autoencoder to obtain nickel-based high-temperature alloy composition generation model; and using the nickel-based high-temperature alloy composition generation model and the performance prediction model to reverse design target performance nickel-based high-temperature alloy composition to obtain nickel-based high-temperature alloy composition design value, which can realize reverse design of nickel-based high-temperature alloy composition with target performance. However, the method in CN202410138800.3 only takes alloy chemical composition as input feature, does not incorporate domain knowledge and microstructure parameters as physical constraints, and does not prove the effectiveness of the method through iterative optimization and experimental verification.

[0008] Therefore, there is a need for a directional development method for high-strength and high-toughness nickel-based high-temperature alloy. SUMMARY

[0009] The present application aims to overcome the defects of the prior art and provides a data-driven deformation nickel-based superalloy multi-objective performance reverse design optimization method, which is a directional development method of high strength and toughness nickel-based superalloy by combining cross-scale calculation, field knowledge constraint and machine learning reverse design. The composition design space is constructed by screening the thermodynamic stable system through the first principle, combining the organizational stability criterion, the strengthening effect regulation and the processing performance hard constraint; the microstructure parameters are calculated by the high-throughput thermodynamic software, the feature dimension of machine learning is expanded, and the performance-component reverse mapping is established based on machine learning and genetic algorithm (XGBoost model and NSGA-II algorithm). Through multiple iterations (EI index driving) optimization and experimental verification, a new deformation nickel-based superalloy with strong plasticity synergy is finally obtained.

[0010] The object of the present application can be achieved by the following technical solutions:

[0011] The object of the present application is to develop a data-driven deformation nickel-based superalloy multi-scale reverse optimization design method. By combining the first principle and the high-throughput calculation of thermodynamics, the computational materials science method is used to provide a multi-source data basis for the alloy reverse design facing the performance requirements by using the machine learning method. New rules and new knowledge are mined from traceable material data, and based on the machine learning model enhanced by the field professional knowledge, the alloy chemical composition design window is reversely established. Through the cross-scale fusion design and machine learning reverse optimization, a new deformation nickel-based superalloy with strong plasticity synergy and processability is developed.

[0012] The present application provides a data-driven deformation nickel-based superalloy multi-objective performance reverse design optimization method, comprising the following steps:

[0013] S1, based on the first principle, a stable high-strength deformation nickel-based superalloy system is calculated;

[0014] S2, based on the deformation nickel-based superalloy system obtained in step S1, the deformation nickel-based superalloy composition design space is constructed based on the field knowledge of nickel-based superalloy;

[0015] S3, based on the deformation nickel-based superalloy composition design space constructed in step S2, the deformation nickel-based superalloy composition design space is reduced by using the empirical formula combined with the high-throughput calculation of thermodynamics, and the reduced deformation nickel-based superalloy composition design space is obtained;

[0016] S4, based on the reduced deformation nickel-based superalloy composition design space obtained in step S3, a deformation nickel-based superalloy reverse design model is established based on machine learning and genetic algorithm, and a target performance deformation nickel-based superalloy is screened.

[0017] Further, step S1 includes the following process:

[0018] Thermodynamic stability, lattice mismatch, and stacking fault energy of alloys are calculated based on first-principles calculations to screen stable and high-strength wrought nickel-based superalloy systems.

[0019] Furthermore, the deformed nickel-based superalloy system is a Ni-Fe-Cr-Co-Nb-Ta-Al-Ti-Mo-W alloy system.

[0020] Further, step S2 includes the following process:

[0021] Based on the wrought nickel-based superalloy system obtained in step S1, a composition design space for wrought nickel-based superalloys is constructed based on the composition calculation space of hot workability, microstructure stability and oxidation corrosion resistance.

[0022] Furthermore, step S2 specifically includes the following process:

[0023] Based on the wrought nickel-based superalloy system obtained in step S1, Ni is set to >40wt% and is a balancing element to ensure Ni serves as the matrix, Cr >15wt% to ensure resistance to oxidation and corrosion, and Al+Ti <6wt% with Ti / Al ≥1.5wt%. The required γ′ volume fraction (Vg) of the designed wrought nickel-based superalloy is... γ′ The following condition must be met: 20% < (V) γ′ <50%, constructing the composition design space for deformed nickel-based superalloys.

[0024] Furthermore, step S3 includes the following process:

[0025] S31. Using empirical formulas to quickly screen alloy components with high-temperature stability of γ and γ′ phases, thus reducing the compositional traversal space of deformed nickel-based high-temperature alloys.

[0026] S32. Thermodynamic high-throughput calculations traverse the composition space of the previous step, accurately screen the compositional stability of wrought nickel-based superalloys, narrow down the compositional design space of wrought nickel-based superalloys, and obtain the narrowed compositional design space of wrought nickel-based superalloys.

[0027] Furthermore, the empirical formula includes:

[0028] The formula for determining the stability of the γ phase used to judge tissue stability is shown in equation (1) below:

[0029] (1)

[0030] In the formula C i Let i be the mole fraction of element i in the alloy. Md id orbital energy of element i, eV, is calculated according to formula (1) of different deformed nickel-based superalloy compositions at 650℃ Md The value meets Md The deformed nickel-based superalloy with <0.94eV is considered to have γ stability;

[0031] The γ' phase stability empirical judgment formula for judging the stability of the organization is shown in the following formula (2):

[0032] (2)

[0033] In the formula, M γ′ is the ratio of (Nb+Ta+Ti) and Al content, C Ti , C Nb , C Ta , C Al are the mole fractions of the corresponding metal elements Ti, Nb, Ta and Al respectively.

[0034] Further, in the process of narrowing the deformed nickel-based superalloy composition design space, the γ volume fraction V γ and the γ' volume fraction V γ′ of the deformed nickel-based superalloy meet the following conditions: V γ+γ′ >99.9vol%, 20%<V γ′ <50%, the solidus temperature T s and the γ' solubility temperature T γ′s of the deformed nickel-based superalloy meet the following conditions: T γ′s ≤1150℃, T s -T γ′s >50℃.

[0035] Further, formula (1) and formula (2) are used to screen the high temperature stability of γ and γ' phases, that is, the stability of the organization. The strengthening effect is the content relationship of each element set, and the stability of the organization and the strengthening effect affect each other according to the proportion of the alloy composition. The hot workability is the solidus temperature T s and the γ' solubility temperature T γ′s ≤1150℃, T s -T γ′s >50℃, also known as the hot working window.

[0036] Further, step S4 includes the following process:

[0037] S41, based on the narrowed deformed nickel-based superalloy composition design space obtained in step S3, a deformed nickel-based superalloy composition and organizational performance quantization relationship database is established;

[0038] S42, based on the established database, data preprocessing and feature engineering are performed, the best machine learning model is screened, and a deformation nickel-based superalloy reverse design model is established;

[0039] S42, based on the deformation nickel-based superalloy reverse design model, a genetic algorithm multi-objective optimization is used to search for a target performance deformation nickel-based superalloy composition, the model is evaluated and iteratively optimized.

[0040] Further, the best machine learning model is selected from one of Gaussian kernel support vector machine (SVR), Gaussian process regression (GP), random forest (RF) and extreme gradient boosting tree (XGBoost).

[0041] Further preferably, the best machine learning model is extreme gradient boosting tree (XGBoost).

[0042] Further, step S42 includes the following process:

[0043] Based on the prediction results of the best machine learning model for high-temperature yield strength and room temperature elongation, the EI values of each target performance of the alloy are calculated as the fitness function in the execution of the NSGA-II algorithm, and then the Pareto front of the EI values of the alloy in the composition space is determined, and candidate experimental alloys are selected from the Pareto front for experimental preparation and performance test verification. The experimental results of the high-temperature yield strength, tensile strength and room temperature elongation of the candidate experimental alloys are fed back to the initial data set (machine learning training data set), and iterative optimization is performed to obtain the optimized composition of the deformation nickel-based superalloy.

[0044] Further, in the experimental preparation and performance test verification, the specific experimental preparation includes vacuum arc melting alloy ingot + hot rolling, homogenization + solid solution + double-stage aging heat treatment, and the specific performance test verification includes microstructure and phase composition characterization analysis: chemical composition quantitative analysis, grain size and grain size statistics, phase composition and phase content analysis, etc., as well as room temperature tensile test, 650℃ tensile test.

[0045] Further, the domain knowledge includes microstructure stability, strengthening effect and hot workability.

[0046] Further, the microstructure stability: the excellent performance of nickel-based superalloy is derived from the microstructure composed of L12 ordered structure γ' precipitated strengthening phase (L12-Ni3Al) and FCC structure γ matrix phase (FCC-Ni), so the primary condition for designing a superalloy is to ensure that the alloy has a γ+γ' dual-phase structure and avoid the formation of topologically close-packed phases. Therefore, referring to the γ stability judgment method based on d-orbital energy level (Md) proposed by Morinage et al., the formula is , where Ci is the mole fraction of element i in the alloy, Md iThe d-orbital energy of element i, eV. According to the formula, the Md value of different alloy components at 650 DEG C is calculated, and the nickel-based high-temperature alloy with Md < 0.94 eV is regarded as having γ stability.

[0047] Further, the strengthening effect: each element in the nickel-based high-temperature alloy can be used based on its respective characteristics, for example, increasing the Fe content to improve the forgeability, adding Cr to improve the oxidation corrosion resistance, adding Al, Ti and Ta to achieve precipitation strengthening, adding Mo and W to improve the creep resistance, and adding C, B and Zr to achieve grain boundary strengthening. However, the interaction between different alloy elements should also be considered when designing the nickel-based high-temperature alloy, for example, the geometrically close-packed phases such as η-Ni3Ti and δ-Ni3Nb increase the high-temperature instability of the γ' phase, but in the present method, a small amount of η phase is allowed to exist, which precipitates at the grain boundary during heat treatment to strengthen the grain boundary, thereby improving the endurance performance, which is related to the content of Al, Ti, Nb and Ta, and the relationship can be described as Therefore, in order to maximize the solid solution and precipitation strengthening effect, the proportion of Nb+Mo+Ta solid solution strengthening elements and Al+Ti precipitation strengthening elements is mainly controlled, and the grain boundary strengthening elements such as C, B and Zr are not considered for the time being.

[0048] Further, the processing performance: according to the amount (mass fraction) of Al+Ti, the alloy can be divided into three types, namely high plastic deformation alloy (Al+Ti < 3%), which is in the single-phase γ region within the forging temperature range, that is, the single-phase austenite region without γ' phase, which is beneficial to forging; medium plastic deformation alloy (3% < Al+Ti < 6%), which has grain boundary carbides within the hot working temperature range, which is not conducive to hot working; and low plastic deformation alloy (Al+Ti > 6%), which has grain boundary carbides and unsolved γ' phase within the forging temperature range, and the alloy is in a multi-phase state, and the process plasticity is low. Therefore, in order to ensure that the alloy has plastic deformation and hot workability during the smelting preparation process and does not sacrifice the high-temperature strength, the content of Al+Ti elements is controlled to be less than 6% and the volume fraction of γ' phase is less than 50%, the solidus temperature (T s ) and the γ' phase complete dissolution temperature (T γ′s ) difference is greater than 100 DEG C, and the γ' phase complete dissolution temperature (T γ′s ) is less than 1150 DEG C.

[0049] Compared with the prior art, the present application has the following technical advantages:

[0050] 1) The present application provides a data-driven deformation nickel-based superalloy multi-objective performance reverse design optimization method, which is a multiscale integrated computing method for reverse design and optimization of deformation nickel-based superalloy by combining first principles, domain knowledge, thermodynamic high-throughput calculation and machine learning. Compared with the traditional "trial and error method" for developing new deformation nickel-based superalloy, the efficiency is greatly improved, and the mechanical properties of the alloy obtained by reverse design are verified by experiments to meet the target requirements and theoretical design expectations, which can reduce the development cycle of new deformation nickel-based superalloy.

[0051] 2) The present application provides a data-driven deformation nickel-based superalloy multi-objective performance reverse design optimization method, which proposes a multiscale integrated computing method for reverse design and optimization of deformation nickel-based superalloy by combining first principles, domain knowledge, thermodynamic high-throughput calculation and machine learning. The microstructure parameters are included in the machine learning features, so that the model can associate the complex mapping of composition-microstructure-mechanical properties; the model bias is corrected through 5 rounds of "prediction-experiment-data feedback" cycle. Traditional methods usually rely on single verification, while the present application dynamically includes experimental data into the training set, which significantly improves the optimization efficiency.

[0052] 3) The present application provides a data-driven deformation nickel-based superalloy multi-objective performance reverse design optimization method, which utilizes the structured collaboration of multiscale computing method to construct a series of work flow of "electronic scale screening→ thermodynamic scale parameter supplement→ machine learning modeling→ experimental verification feedback iteration", and through the embedding of microstructure parameters, the blindness of machine learning "pure data black box" is avoided, which not only significantly improves the design efficiency (5 rounds of iteration to obtain the optimal alloy and the real effectiveness is verified by experiment), but also establishes the internal relationship between chemical composition-microstructure-mechanical properties, providing a generalizable design idea and method for complex alloy system.

[0053] 4) The present application provides a data-driven deformation nickel-based superalloy multi-objective performance reverse design optimization method, a new type of deformation nickel-based superalloy obtained by integrated computing reverse design can be prepared by conventional vacuum arc melting and hot rolling process, and the room temperature and high temperature tensile mechanical properties are better than those of GH4706 high temperature alloy for turbine disc, and subsequent iteration optimization can be carried out to develop new deformation nickel-based superalloy with better mechanical properties. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The flow chart of the data-driven deformation nickel-based superalloy multi-objective performance reverse design optimization method.

[0055] Figure 2 The alloy ingot after hot rolling is prepared by experimental melting.

[0056] Figure 3Figure 1 is a uniaxial equilibrium phase diagram of alloy #1 calculated based on Thermo-Calc software.

[0057] Figure 4 Figure 2 is a distribution diagram of precipitated phase size of alloy #1 calculated based on Pandat software.

[0058] Figure 5 Figure 3 is a comparison curve of tensile engineering stress-strain of preferred alloy #1 and GH4706 alloy. DETAILED DESCRIPTION

[0059] The present application will be described in detail below with reference to the accompanying drawings and specific embodiments. In the technical solution, if the component model, material name, connection structure, control method, algorithm and other features are not explicitly stated, they are considered as common technical features disclosed in the prior art.

[0060] The technical concept of the present application is as follows:

[0061] 1) First, through first-principle pre-screening, unstable component systems are quickly excluded by calculating thermodynamic stability, lattice mismatch and stacking fault energy, thereby narrowing the component design space (Ni-Fe-Cr-Co-Nb-Ta-Al-Ti-Mo-W system).

[0062] 2) Based on domain knowledge (organizational stability, strengthening effect, processability), the component boundary is constructed, and the component design space is further narrowed to a controllable range, thereby improving the search efficiency.

[0063] 3) Traditional design methods only use chemical composition and heat treatment to predict performance, ignoring the key influence of microstructure parameters (such as gamma' volume fraction, precipitated phase size radius, antiphase domain boundary energy) and the like. Microstructure parameters are calculated by commercial thermodynamic calculation software as machine learning input features, thereby enhancing the characterization ability of machine learning on the strengthening mechanism of the deformed nickel-based high-temperature alloy, significantly improving the physical interpretability of the model, and significantly improving the prediction accuracy.

[0064] 4) There is a trade-off between high-temperature strength and room-temperature plasticity, and it is difficult to simultaneously optimize the traditional method. The expected improvement (EI) is used as the fitness function to search for the strength-plasticity balance point on the Pareto frontier, and the experimental data is fed back to the initial data set for iterative optimization, thereby dynamically correcting the model bias and solving the reverse design problem of multi-objective performance constraints, thereby significantly improving the efficiency compared with the traditional “trial and error method”.

[0065] The present application combines first-principles calculation, physical metallurgy rules specific to the field of nickel-based superalloys to construct the composition space, and uses microstructure parameters from thermodynamic calculations as physical constraints as input features of the machine learning model to significantly improve the material identification ability of the machine learning model. This dual-drive mode of "computational physics + field knowledge" combines a design framework that integrates physical mechanisms and field knowledge, a machine learning model enhanced by microstructure parameters, and a multi-objective iterative optimization oriented to engineering application requirements, ensuring the physical interpretability and engineering feasibility of the composition design. The present application is mainly aimed at the composition design and development of wrought nickel-based superalloys, not only using alloy chemical composition as input features, but also incorporating field knowledge and microstructure parameters as physical constraints, and through iterative optimization and experimental verification, the effectiveness of the multi-scale reverse design method proposed in the present patent is proved.

[0066] Embodiment

[0067] As shown in Figure 1 , in the present embodiment, a data-driven multi-objective performance reverse design optimization method for wrought nickel-based superalloys is provided, comprising the following steps:

[0068] I. Based on first-principles calculation, a stable high-strength wrought nickel-based superalloy system is obtained, specifically, the thermodynamic stability, lattice mismatch and dislocation energy of the alloy are calculated, and a stable high-strength (Ni-Fe-Cr-Co-Nb-Ta-Al-Ti-Mo-W) alloy system is selected.

[0069] II. Based on the wrought nickel-based superalloy system obtained in step I, a nickel-based superalloy design strategy is constructed based on the field knowledge of nickel-based superalloys for organizational stability, strengthening effect and hot workability, a wrought nickel-based superalloy composition design space is constructed, and based on the constructed wrought nickel-based superalloy composition design space, an empirical formula is used in combination with high-throughput thermodynamic calculations to narrow the wrought nickel-based superalloy composition design space, and a narrowed wrought nickel-based superalloy composition design space is obtained. Specifically, it includes:

[0070] Based on the wrought nickel-based superalloy system obtained in step S1, set Ni>40wt% and balance element to ensure Ni as matrix, Cr>15wt% to ensure oxidation and corrosion resistance, Al+Ti<6wt% and Ti / Al (wt%)≥1.5, and require the designed wrought nickel-based superalloy to have a γ' volume fraction V γ′ satisfying the following conditions: 20%<V γ′ <50%, a wrought nickel-based superalloy composition design space is constructed;

[0071] An empirical formula is used to preliminarily screen alloy compositions with high-temperature stability of γ and γ', narrowing the wrought nickel-based superalloy composition traversal space;

[0072] Thermodynamic high-throughput calculation traverses the composition space of the previous step to screen the composition-structure stability of the designed nickel-based superalloy, and to narrow down the composition design space of the wrought nickel-based superalloy, thereby obtaining the narrowed-down composition design space of the wrought nickel-based superalloy.

[0073] The specific construction of the nickel-based superalloy design strategy in step II is as follows:

[0074] ① Structure stability: The excellent performance of the nickel-based superalloy is derived from the microstructure composed of the L12 ordered structure γ' precipitate strengthening phase (L12-Ni3Al) and the FCC structure γ matrix phase (FCC-Ni), so the primary condition for designing the superalloy is to ensure that the alloy has a γ+γ' dual-phase structure and avoid the generation of topologically close-packed phases. Therefore, referring to the γ stability judgment method based on the d-orbital energy level (Md) proposed by Morinage et al., the formula is , where Ci is the mole fraction of element i in the alloy, and Md i is the d-orbital energy of element i, eV. According to the formula, the Md value of different alloy compositions at 650℃ is calculated, and the nickel-based superalloy satisfying Md < 0.94eV is considered to have γ stability.

[0075] ② Strengthening effect: Each element in the nickel-based superalloy can be used based on its own characteristics, for example, increasing the Fe content to improve forgeability, adding Cr to improve oxidation and corrosion resistance, adding Al, Ti and Ta to achieve precipitation strengthening, adding Mo and W to improve creep resistance, and adding C, B and Zr to achieve grain boundary strengthening. However, when designing the nickel-based superalloy, the interaction between different alloy elements also needs to be considered, for example, the geometrically close-packed phases such as η-Ni3Ti and δ-Ni3Nb can increase the high-temperature instability of the γ' phase, but in this method, a small amount of η phase is allowed to precipitate at the grain boundary during heat treatment to strengthen the grain boundary, thereby improving the endurance performance, which is related to the content of Al, Ti, Nb and Ta, and the relationship can be described as . Therefore, to maximize the solid solution and precipitation strengthening effect, the proportion of Nb+Mo+Ta solid solution strengthening elements and Al+Ti precipitation strengthening elements is focused on, and the grain boundary strengthening elements such as C, B and Zr are not considered.

[0076] ③ Hot workability: According to the amount of Al+Ti (mass fraction, the same below), the alloys can be divided into three categories, i.e. high plastic deformation alloy (Al+Ti<3%), in the range of forging temperature, the alloy is in single-phase γ region, that is, single-phase austenite region without γ' phase which is beneficial to forging; medium plastic deformation alloy (3%<Al+Ti<6%), in the range of hot working temperature, there are grain boundary carbides, which are not conducive to hot working; low plastic deformation alloy (Al+Ti>6%), in the range of forging temperature, there are grain boundary carbides and unsolved γ' phase, the alloy is in a multi-phase state, and the process plasticity is very low. Therefore, in order to ensure the plastic deformation and hot workability of the alloy in the smelting preparation process and not to sacrifice the high temperature strength, the content of Al+Ti is controlled to be less than 6%, the volume fraction of γ' phase is less than 50%, the difference between the solidus temperature (T s ) and the complete dissolution temperature of γ' phase (T γ′s ) is greater than 100℃, and the complete dissolution temperature of γ' phase (T γ′s ) is less than 1150℃.

[0077] In summary, the composition design space of nickel-based superalloy is constructed, Ni>40wt% and is a balanced element to ensure that Ni is the matrix, Cr>15wt% to ensure oxidation and corrosion resistance, Al+Ti<6wt% and Ti / Al (wt%)≥1.5 (only when Ti / Al ratio>1.5, η phase can be precipitated at the grain boundary during heat treatment), and the volume fraction of γ phase V γ and the volume fraction of γ' phase V γ′ of the designed alloy meet the following conditions: V γ+γ′ >99.9vol%, 20%<V γ′ <50%. The composition design space of the deformed nickel-based superalloy is shown in Table 1.

[0078] Table 1 Composition design space of deformed nickel-based superalloy (Ni Bal., wt%)

[0079]

[0080] III. Machine learning dataset construction and preprocessing: To eliminate the performance differences caused by forming processes and improve the learning ability of machine learning on strengthening mechanisms, the machine learning training dataset only collects the chemical composition, heat treatment process, room temperature and 650°C tensile property data of precipitation strengthening and solid solution strengthening of deformed nickel-based superalloys prepared by general processes such as rolling and forging, and does not include performance data of other special forming processes such as powder metallurgy, directional solidification, additive manufacturing, etc. The data comes from “China High Temperature Alloy Handbook”, “China Aviation Materials Manual Volume 2. Deformed High Temperature Alloy. Cast High Temperature Alloy”, existing alloy grades at home and abroad, publicly available databases on the network, literature data, and a total of 726 samples are collected after screening. The alloy composition is all in mass percent (wt%), and the interval data (such as [a, b] or a ~ b) is converted to the average value. Heat treatment includes solution treatment temperature, solution treatment time, aging treatment temperature, aging treatment time, test temperature is 650°C, and the design target performance is room temperature yield strength, room temperature elongation and 650°C yield strength.

[0081] In fact, the performance of nickel-based superalloy materials is closely related to their microstructure, and only considering composition and heat treatment parameters as factors affecting alloy strength will result in the loss of some information. The integration of more strength-related knowledge or information may further improve the search efficiency of the target alloy. Therefore, it is necessary to add material parameters related to microstructure to improve the material recognition ability of the machine learning model, including the volume fraction of γ' phase (V γ′ ), the antiphase boundary energy E APB , the γ / γ' lattice mismatch (δ), the complete dissolution temperature of γ' phase (γ' dissolution temperature, T γ′s ), the secondary γ' phase precipitation size radius (r s ), and the tertiary γ' phase precipitation size radius (r t ). In addition, some missing values of microstructure material parameters in the collected deformed nickel-based superalloy dataset were calculated by Python scripts through the API interface of the commercial thermodynamic calculation software Thermal-Calc and Pandat.

[0082] IV、In order to realize the inverse design of the multi-objective performance of the wrought nickel-based superalloy, data preprocessing and feature engineering are performed, the best machine learning model is selected, and the inverse design model of the wrought nickel-based superalloy is established. Specifically, first, the performance of four machine learning models, including Gaussian kernel support vector machine (SVR), Gaussian process regression (GP), random forest (RF), and extreme gradient boosting tree (XGBoost), is evaluated using the leave-one-out method. The training set and the test set are divided in a ratio of 8:2 for training and modeling. Ten-fold cross-validation method is used to optimize the model parameters. The root mean square error (RMSE) is selected as the evaluation index of the model performance. The smaller the RMSE, the higher the prediction accuracy of the model. Based on the inverse design model of the wrought nickel-based superalloy, the genetic algorithm is used to search for the composition of the wrought nickel-based superalloy with the target performance. The model is evaluated and iteratively optimized. Specifically, based on the prediction results of the high-temperature yield strength and the room-temperature elongation of the best machine learning model (XGBoost), the EI values of the target performance of the alloy are calculated, which are used as the fitness function in the execution of the NSGA-Ⅱ algorithm. Then, the Pareto frontier of the EI values of the alloy in the composition space is determined. Finally, candidate experimental alloys are selected from the Pareto frontier for experimental preparation and performance testing and verification. Figure 2 To prepare the hot-rolled alloy plate for experimental melting, all candidate alloys use metal element pellets with a purity of >99.95% to match the alloy composition. The ingots with a size of about 70mmx45mmx12mm are prepared by vacuum arc melting, and then homogenized at 1150℃ for 24h. After homogenization, the total deformation of hot rolling is 50%, followed by solid solution + double-stage aging heat treatment (980℃ for 1h air cooling to room temperature, 730℃ for 8h, cooling rate of 55℃ / h to 620℃ for 8h, air cooling). Specific performance testing and verification includes microstructure and phase composition characterization analysis: mass spectrometer (IPC-MS) is used for quantitative analysis of alloy chemical composition, optical microscope is used for grain size and grain size statistics by intercept method, X-ray diffraction (XRD) is used for phase composition analysis by comparing XRD standard cards, field emission scanning electron microscope (SEM) is used to observe the volume fraction and size of γ and γ', and room temperature tensile test, 650℃ tensile test. The experimental results of the high-temperature yield strength, tensile strength and room-temperature elongation of the new alloy are fed back to the initial data set for iterative optimization. After 5 iterations, a new wrought nickel-based superalloy is obtained, named "#1". As shown in Figure 3 The uniaxial equilibrium phase diagram of #1 alloy calculated based on Thermo-Calc software is shown in FIG. 1. It is shown that #1 alloy only exists in γ and γ' phases (FCC_L12 and FCC_L12#2, respectively) at 650℃, without other phases. Figure 4The size distribution of the multi-peak precipitates of alloy #1 under the heat treatment regime calculated based on the Pandat software is shown in FIG. 1. The deformed nickel-based superalloy “#1” obtained satisfies the target performance requirements of a room temperature yield strength > 900 MPa, a high temperature 650 °C yield strength > 820 MPa, and a room temperature elongation > 12%, and has a significant improvement in the tensile yield strength at room temperature and 650 °C without a significant reduction in elongation compared to the existing GH4706 grade deformed superalloy (e.g. Figure 5 as shown in FIG. 1).

[0083] The above describes the embodiments of the method of the present application in conjunction with the accompanying drawings, but the present application is not limited to the above-described embodiments, and can be changed in various ways according to the purpose of the inventive idea of the present application, and any parameter change or calculation simplification made according to the principle of the technical solution of the present application, as long as it meets the inventive purpose of the present application, as long as it does not deviate from the principle and concept of the data-driven deformed nickel-based superalloy multi-objective performance reverse design optimization method of the present application, all belong to the protection scope of the present application.

[0084] The above description of the embodiments is for the convenience of the ordinary skilled person in the art to understand and use the application. Those skilled in the art can obviously make various modifications to these embodiments, and apply the general principles described herein to other embodiments without having to go through creative labor. Therefore, the present application is not limited to the above-described embodiments, and any improvement and modification made by those skilled in the art according to the disclosure of the present application without departing from the scope of the present application should be within the protection scope of the present application.

Claims

1. A data-driven inverse design optimization method for the multi-objective performance of a wrought nickel-based superalloy, characterized in that, Comprise the following steps: S1, based on first principles calculation of stable high strength deformation nickel-based superalloy system; S2, based on the deformation nickel-based superalloy system obtained in step S1, the composition design space of the deformation nickel-based superalloy is constructed based on the knowledge of nickel-based superalloy field; S3, based on the composition design space of the deformation nickel-based superalloy constructed in step S2, the composition design space of the deformation nickel-based superalloy is reduced by using empirical formula combined with thermodynamic high-throughput calculation, and the reduced composition design space of the deformation nickel-based superalloy is obtained; S4, based on the reduced composition design space of the deformation nickel-based superalloy obtained in step S3, a reverse design model of the deformation nickel-based superalloy is established based on machine learning and genetic algorithm, and a target performance deformation nickel-based superalloy is screened; Step S2 specifically includes the following process: Based on the deformed nickel-based superalloy system obtained in step S1, set Ni>40wt% and be the balance element to ensure Ni as the matrix, Cr>15wt% to ensure the oxidation and corrosion resistance, Al+Ti<6wt% and Ti / Al≥1.5wt%, require the designed deformed nickel-based superalloy γ' volume fraction V γ′ Meet the following conditions: 20%<V γ′ <50%, build a deformed nickel-based superalloy component design space; Step S3 includes the following process: S31, the alloy composition with high temperature stability of γ and γ' is screened by using empirical formula to reduce the composition traversal space of the deformation nickel-based superalloy; S32, the composition of the deformation nickel-based superalloy is screened by thermodynamic high-throughput calculation to reduce the composition design space of the deformation nickel-based superalloy, and the reduced composition design space of the deformation nickel-based superalloy is obtained; The empirical formula includes: The empirical judgment formula for judging the stability of γ phase is shown in the following formula (1): （1） wherein C i is the mole fraction of element i in the alloy, Md i is the d-orbital energy of element i in eV, calculated according to equation (1) for the composition of the different wrought nickel-based superalloys at 650 °C Md values, satisfying Md wrought nickel-based superalloys with values of < 0.94 eV are considered to have γ-stability; The empirical judgment formula for judging the stability of γ' phase is shown in the following formula (2): （2） wherein M γ′ is the ratio of (Nb+Ta+Ti) to Al content, C Ti , C Nb , C Ta , C Al are the mole fractions of the respective metal elements.

2. The data-driven inverse design and optimization method of the property of the deformed nickel-based superalloy according to claim 1, characterized in that, Step S1 includes the following process: The thermodynamic stability, lattice mismatch and dislocation energy of the alloy are calculated based on first principles to screen the stable high-strength deformation nickel-based superalloy system; The deformation nickel-based superalloy system is Ni-Fe-Cr-Co-Nb-Ta-Al-Ti-Mo-W alloy system.

3. The data-driven multi-objective performance inverse design optimization method of a wrought nickel-based superalloy according to claim 2, wherein, Step S2 includes the following process: Based on the deformation nickel-based superalloy system obtained in step S1, the composition design space of the deformation nickel-based superalloy is constructed based on hot workability, microstructure stability and oxidation corrosion resistance.

4. The data-driven inverse design and optimization method of property variations of a wrought nickel-based superalloy according to claim 1, wherein, In the process of narrowing the composition design space of wrought nickel-base superalloys, the γ volume fraction V γ and the γ' volume fraction V γ′ are set to satisfy the following conditions: V γ+γ′ > 99.9 vol%, 20% < V γ′ < 50%; the solidus temperature T s and the γ' solvus temperature T γ′s of the wrought nickel-base superalloy are set to satisfy the following conditions: T γ′s ≤ 1150℃, T s - T γ′s > 100℃, to ensure that the alloy has a sufficient hot working window, thereby ensuring the hot workability of the alloy.

5. The data-driven inverse design and optimization method of property variations of a wrought nickel-based superalloy according to claim 1, wherein Step S4 includes the following process: S41, based on the reduced composition design space of the deformation nickel-based superalloy obtained in step S3, a composition and microstructure performance quantitative relationship database of the deformation nickel-based superalloy is established; S42, based on the established database, data preprocessing and feature engineering are carried out, the best machine learning model is selected, and a reverse design model of the deformation nickel-based superalloy is established; S42, based on the reverse design model of the deformation nickel-based superalloy, the genetic algorithm is used to search the target performance deformation nickel-based superalloy composition, the model is evaluated and iteratively optimized.

6. The data-driven multi-objective performance inverse design optimization method of a wrought nickel-based superalloy according to claim 5, wherein, The best machine learning model is selected from one of Gaussian kernel support vector machine, Gaussian process regression, random forest and extreme gradient boosting tree.

7. The data-driven multi-objective performance inverse design optimization method of a wrought nickel-based superalloy according to claim 5, wherein, Step S42 includes the following process: Based on the prediction results of the target performance of the optimal machine learning model, the EI values of each target performance of the alloy are calculated as the fitness function in the execution of the NSGA-II algorithm, and the Pareto front of the EI values of the alloy in the composition space is determined. The candidate experimental alloys are screened from the Pareto front for experimental preparation and performance test verification. The experimental results of the high-temperature yield strength, tensile strength and room-temperature elongation of the candidate experimental alloys are fed back to the machine learning training data set, and iterative optimization is performed to obtain the optimized deformed nickel-based high-temperature alloy composition. The target performance is selected from one or more of room-temperature yield strength, 650 DEG C yield strength and room-temperature elongation.

Images