High-performance novel alloy design method and system

Abstract

The invention discloses a high-performance novel alloy design method and system. The high-performance novel alloy design method comprises the steps that a first historical alloy data set is obtained; calculating element characteristics of historical alloy; screening the element features to obtain key element features; for each historical alloy, calculating a phase diagram, and extracting key phase diagram parameters; obtaining a second historical alloy data set based on the first historical alloy data set, the key element features and the key phase diagram parameters; constructing an alloy prediction model based on a machine learning model, and determining an optimal model by taking the second historical alloy data set as a sample data set; on the basis of the optimal model, key element characteristic values and process data corresponding to the optimal performance data are obtained; and the optimal key element characteristic value is converted into component data, so that the component data and the process data of the high-performance novel alloy are obtained. According to the method, automation and intelligentization of alloy design can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of material design, and in particular to a method and system for designing a high-performance new alloy. Background Art

[0002] In the field of materials science, alloy materials play a pivotal role in industry and high-tech fields due to their unique physical and chemical properties. Traditional alloy design mainly relies on empirical trial and error to improve material properties by adding relatively small amounts of various elements to a main element, such as steel, aluminum alloy, titanium alloy, magnesium alloy and copper alloy.

[0003] With the advancement of science and technology, the demand for the development of new materials is growing, and the limitations of traditional methods in the development of new materials are gradually becoming apparent. Due to the large number of elements involved and the huge space for component process design, the cost of traditional empirical trial and error methods is gradually increasing.

[0004] In recent years, machine learning technology, as a multi-disciplinary subject involving probability theory and statistics, has provided a new perspective for alloy material design. Machine learning uses algorithms to allow computers to efficiently construct complex nonlinear relationships between input data and output targets from existing data, and use this relationship to build models to predict unknown data.

[0005] Although the application of machine learning techniques in alloy material design has shown great potential, there are still some significant shortcomings. First, machine learning models are often criticized as "black box" problems because they generally lack transparency and interpretability, which is particularly problematic in materials science because it is crucial to understand and explain the relationship between the microstructure and macroscopic properties of materials. In addition, existing machine learning methods often ignore the basic principles of metallurgy for alloy design, such as phase diagrams, crystal structures, and interactions between elements, which are key factors in determining alloy properties.

[0006] Further analysis of the application of machine learning methods in alloy design shows that these methods can only capture the simple linear relationship between composition and process, but cannot deeply analyze the nonlinear coupling between them, which has a decisive influence on the final performance of the alloy. When sample data is scarce, the predictive ability of the model will be significantly reduced due to the mutual interference between samples, resulting in reduced prediction accuracy. This situation is particularly common in the early stages of research and development of new materials, because the available data is usually very limited.

[0007] In practical applications, due to the scarcity of sample data and the strong coupling relationship between alloy composition and process, machine learning models often find it difficult to capture these complex interactions. When there is less sample data, the model's predictive ability will be significantly reduced due to the mutual interference between samples, resulting in reduced prediction accuracy. In this case, due to the lack of in-depth understanding of the complex coupling relationship between alloy composition and process, the model is often unable to accurately learn and simulate the true relationship between alloy composition and process parameters, making it difficult to achieve integrated and precise design and optimization of alloy composition and process, thereby limiting its application potential in the development of high-performance alloy materials. Summary of the invention

[0008] The purpose of the present invention is to provide a high-performance new alloy design method and system, which can realize the automation and intelligence of alloy design.

[0009] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0010] In a first aspect, a method for designing a high-performance new alloy is provided, comprising the following steps:

[0011] S1: Obtain a first historical alloy data set; the first historical alloy data set includes a plurality of first historical alloy data, each of which corresponds to composition data, process data and performance data of a historical alloy, wherein the composition data of the historical alloy includes the content and proportion of each element included in the historical alloy, and the sum of the proportions of all elements of each historical alloy is equal to 100%, so as to ensure the integrity and accuracy of the data;

[0012] S2: For the historical alloy, obtain the characteristic factor of each element therein, and calculate the statistical properties of the characteristic factors of all elements therein to obtain the element characteristics of the historical alloy;

[0013] S3: For historical alloys, element characteristics are screened to obtain key element characteristics;

[0014] This step is the process of extracting element characteristics from historical alloy data, providing key input for subsequent performance data prediction of new alloys;

[0015] S4: For each historical alloy, calculate the phase diagram and extract the key phase diagram parameters;

[0016] Among them, key phase diagram parameters include but are not limited to liquidus temperature and solid solubility;

[0017] These parameters are then used as input features of the model to guide the alloy design process and provide direct guidance from metallurgical principles.

[0018] S5: Based on the first historical alloy data set, key element characteristics and key phase diagram parameters, a second historical alloy data set is obtained; the second historical alloy data set includes a plurality of second historical alloy data, each of which corresponds to a key element characteristic value, process data, performance data and key phase diagram parameter;

[0019] S6: Build an alloy prediction model based on the machine learning model, use the second historical alloy data set as a sample data set, apply the intelligent optimization algorithm, automatically search for the optimal parameter combination of the alloy prediction model, and determine the best model;

[0020] This process performs hyperparameter self-optimization and model self-screening based on the model's generalization ability and test accuracy to enhance the model's understanding and application of metallurgical principles in alloy design, determine the optimal model, and thus achieve integrated and precise design and optimization of alloy composition and process.

[0021] S7: Based on the optimal model, the key element characteristic values ​​and process data corresponding to the optimal performance data are obtained; the key element characteristic values ​​corresponding to the optimal performance data are converted into composition data (elements and their contents), thereby obtaining the composition data and process data of the high-performance new alloy.

[0022] Furthermore, the proportion is a mole fraction or a mass fraction.

[0023] Furthermore, the extraction accuracy of the element content is selected as 0.01 atomic percent (at.%) to ensure the accuracy of the data and the reliability of the model prediction.

[0024] Furthermore, the characteristic factors include various physical property characteristics; the physical property characteristics include atomic size information, electronic information and energy information.

[0025] Further, the S2 includes:

[0026] For each historical alloy, the weighted average value, weighted mean square error and weighted average deviation of the same characteristic factor of all elements are calculated as the average characteristic factor, mean square error characteristic factor and average deviation characteristic factor of the historical alloy; wherein the weighted value of each element is the proportion of the corresponding element in the historical alloy; thereby, the statistical properties of the elements in the historical alloy can be quantified;

[0027] For each historical alloy, the element characteristics include the average characteristic factor, the mean square deviation characteristic factor, and the average deviation characteristic factor;

[0028] Further, the S3 includes:

[0029] S31: Set the screening threshold r th ;

[0030] S32: Pearson correlation coefficient is used to calculate the correlation coefficient between element characteristics and element characteristics of historical alloys, and between element characteristics and performance data;

[0031] S32: For element features whose absolute value of correlation coefficient is greater than the screening threshold, only one element feature with a large correlation coefficient with the performance data is retained;

[0032] S33: Select the basic model, use the wrapper method to select the element features retained by S32, use different element feature combinations as the input of the basic model, use the performance data as the output of the basic model, establish different models, judge the accuracy of all models, and use the element feature combination used when the model has the highest accuracy as the final selected key element feature combination.

[0033] Furthermore, in S4, the extraction accuracy of key phase diagram parameters is 0.1°C and 0.01at.%, so as to ensure the accuracy of phase diagram information, thereby providing accurate metallurgical guidance for alloy design.

[0034] Furthermore, in S4, the phase diagram parameters are obtained by theoretical calculation, experimental testing and literature research; the theoretical calculation includes first-principles calculation and Thermo-Cal software calculation; the phase diagram types include binary equilibrium phase diagram, ternary equilibrium phase diagram and multivariate equilibrium or non-equilibrium phase diagram.

[0035] Furthermore, the different process parameters of solution treatment-first cold rolling-first aging-second cold rolling-second aging are: solution treatment temperature is 600°C to 950°C; solution treatment time is 5 to 480 min; first cold rolling reduction and second cold rolling reduction are 0 to 95%; aging temperature is 300°C to 500°C; first aging time and second aging time are 1 to 1440 min.

[0036] Furthermore, the parameter combination includes the number of neural network layers, the number of neuron nodes, the learning rate and the loss rate. The number of neural network layers is 2 to 30 layers, the number of neuron nodes is 8 to 1024, and the learning rate is 10 -6 Up to 10 -1 , the dropout rate is 0.01 to 0.6.

[0037] Optionally, the number of iterations of the intelligent optimization algorithm is set to 500 to 1500, preferably 1000, to obtain the optimal parameter combination within a reasonable computing time.

[0038] In the second aspect, a prediction system for high-performance new alloys based on machine learning and phase diagram information is provided. The system realizes the prediction and design of high-performance new alloys through integrated module operations, and specifically includes the following modules:

[0039] A first data acquisition module is used to acquire a first historical alloy data set; the first historical alloy data set includes a plurality of first historical alloy data, each of which corresponds to composition data, process data and performance data of a historical alloy, wherein the composition data of the historical alloy includes the content and proportion of each element included in the historical alloy, and the sum of the proportions of all elements of each historical alloy is equal to 100%, so as to ensure the integrity and accuracy of the data;

[0040] A data processing module is used to obtain the characteristic factor of each element in the historical alloy and calculate the statistical characteristics of the characteristic factors of all elements therein to obtain the element characteristics of the historical alloy;

[0041] The data processing module can pre-process the collected historical alloy data sets, including data cleaning, normalization and feature extraction, so as to form the element characteristics of each historical alloy;

[0042] Element feature screening module, used to screen element features for historical alloys to obtain key element features;

[0043] Phase diagram information acquisition module, used to calculate the phase diagram for each historical alloy and extract key phase diagram parameters;

[0044] Among them, the key phase diagram parameters include liquidus temperature and solid solubility;

[0045] These parameters are then used as input features of the model to guide the alloy design process and provide direct guidance from metallurgical principles.

[0046] A second data acquisition module, for obtaining a second historical alloy data set based on the first historical alloy data set, key element characteristics and key phase diagram parameters; the second historical alloy data set includes a plurality of second historical alloy data, each second historical alloy data corresponding to a key element characteristic value, process data, performance data and key phase diagram parameter;

[0047] The model optimization module builds an alloy prediction model based on the machine learning model, uses the second historical alloy data set as a sample data set, applies an intelligent optimization algorithm, automatically searches for the optimal parameter combination of the alloy prediction model, and determines the best model;

[0048] This process is based on the model's generalization ability and test accuracy to enhance the model's understanding and application of metallurgical principles in alloy design, determine the best model, and thus achieve integrated and precise design and optimization of alloy composition and process.

[0049] The alloy design module obtains the key element characteristic values ​​and process data corresponding to the optimal performance data based on the optimal model; the key element characteristic values ​​corresponding to the optimal performance data are converted into composition data (elements and their contents), thereby obtaining the composition data and process data of the high-performance new alloy.

[0050] In a third aspect, the present application provides an electronic device, including: a memory and a processor;

[0051] The memory is used to store computer programs;

[0052] The processor is used to call the computer program to execute the method as described above.

[0053] In a fourth aspect, the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed on an electronic device, the electronic device implements the method described above.

[0054] In a fifth aspect, the present application provides a computer program product, including a computer program, and when the computer program is run on an electronic device, the electronic device implements the method as described above.

[0055] The specific implementation methods of the second to fifth aspects of the present application can refer to the implementation methods of the first aspect, which will not be repeated here.

[0056] Compared with the prior art, the present invention has the advantages of:

[0057] (1) Significant improvement in the theoretical basis of alloy design: By integrating phase diagram information, the present invention combines the thermodynamic basis of alloy design with machine learning technology, providing a solid theoretical support for the prediction of performance data. The introduction of phase diagram information makes alloy design more scientific and accurate, and also enhances the generalization ability of the model, enabling it to adapt to different alloy systems and process conditions.

[0058] (2) Integrated and precise design of alloy composition and process: The present invention uses advanced machine learning technology, combined with historical alloy data sets and phase diagram information, to achieve integrated and precise prediction of composition and process. By screening key element characteristics through correlation analysis, gradient descent method and SHAP method, the present invention can identify the factors that have the most significant impact on performance data, thereby optimizing alloy design and improving alloy performance.

[0059] (3) Automated optimization of model parameters: By applying intelligent optimization algorithms, such as genetic algorithms or particle swarm optimization (PSO), the present invention realizes the automation of model parameter optimization, reduces the interference of human factors, and improves the efficiency of model training and the accuracy of prediction. This automated hyperparameter optimization process ensures that the model can maintain stable performance on a wide range of data sets.

[0060] (4) Accelerating the development of high-performance, low-cost alloys: The present invention reduces the number of experiments and material waste through precise design methods, which not only reduces R&D costs but also accelerates the development of high-performance, low-cost alloys. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a method flow chart of an embodiment of the present application.

[0062] Figure 2 A system block diagram of an embodiment of the present application. DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] This application aims to address the limitations of existing machine learning methods in alloy design, especially the lack of guidance from metallurgical principles, which leads to insufficient analysis of the complex coupling between alloy composition and preparation process, as well as the decrease in model prediction accuracy when there is less sample data. To this end, this application proposes an efficient integrated design method for alloy composition and process based on machine learning and phase diagram information, which can make full use of material science principles and advanced data analysis technology to realize the automation and intelligence of alloy design.

[0065] The new alloy composition and process design process of this application can be divided into two stages, A and B; stage A is the screening of the optimal alloy characteristic factor types; stage B is the optimization of alloy characteristic factors and processing technology; stage C is the process of converting the optimal alloy characteristic factors into the optimal alloy composition. For stage A, all the candidate alloy characteristic factors are converted from the elements in the first historical alloy data set, all characteristic factors are independent of each other, and there is no order between the characteristic factors, so the model uses a parallel model. The alloy factor conversion process and screening process are shown in S2~S3, which will not be repeated here; for stage B, the alloy characteristic factor types are obtained by screening in stage A, and the specific values ​​are calculated from the elements and their contents in the second historical alloy data set. In addition, there is a sequence of processing technology, so the model uses a time series model; for stage C, the alloy composition and its content are reversed from the optimal alloy characteristic factor, and an optimal alloy characteristic factor combination may correspond to a variety of different alloy compositions and their contents, and the specific alloy composition can be selected according to specific requirements (such as cost, environmental friendliness, etc.). Therefore, the method described in the present invention can develop a variety of new alloys and their processing and preparation processes that meet different service requirements and are low-cost, which greatly enriches the types of materials and accelerates the research and development of metal materials.

[0066] Embodiment 1:

[0067] As attached Figure 1 As shown, the embodiment of the present application provides a method for designing a high-performance new alloy. Taking the Cu-Be-XY (X, Y are both unknown elements) alloy as an example, this embodiment elaborates on the selection of elements and optimization of the preparation process to the prediction of performance.

[0068] S1: Obtain a first historical alloy data set; the first historical alloy data set includes a plurality of first historical alloy data, each of which corresponds to composition data, process data and performance data of a historical alloy, wherein the composition data of the historical alloy includes the content and proportion of each element included in the historical alloy, and the sum of the proportions of all elements of each historical alloy is equal to 100%, so as to ensure the integrity and accuracy of the data;

[0069] Among them, the acquisition methods include literature collection and experimental supplement.

[0070] S2: For the historical alloy, obtain the characteristic factor of each element therein, and calculate the statistical properties of the characteristic factors of all elements therein to obtain the element characteristics of the historical alloy;

[0071] In some embodiments, the characteristic factor of an element is a physical property characteristic of the element. The physical property characteristics of the element are known, including atomic size information, electronic information and energy information of each element. Among them, atomic size information includes lattice constant, atomic radius, Cordero covalent radius and Pyykko covalent radius; electronic information includes: valence electron concentration, electronegativity, electronic energy band, valence electron, electron affinity, conductivity, resistivity; energy information includes: melting point, boiling point, enthalpy cohesive energy, heat of fusion, heat of vaporization, specific heat, thermal conductivity and thermal expansion coefficient; physical information includes: atomic mass, nuclear charge number, density, proportion, atomic hardness, shear modulus, Young's modulus, Poisson's ratio, family number, period number, space group number. The above information is integrated and calculated according to the element content in the first historical alloy data set to obtain the average value, variance, and deviation of each characteristic factor in each historical alloy sample.

[0072] S3: For historical alloys, the element characteristics are screened to obtain key element characteristics.

[0073] In some embodiments, the element features can be screened by using correlation analysis, gradient descent method and SHAP method, wherein the correlation analysis method includes Pearson correlation analysis method.

[0074] For any historical alloy, the element characteristics are screened by Pearson correlation analysis to obtain the key element characteristics, including:

[0075] S31: Set the screening threshold r th ;

[0076] In some embodiments, r can be set th It is 0.7 to 0.9, preferably 0.8.

[0077] S32: Pearson correlation coefficient is used to calculate the correlation coefficient between element characteristics and element characteristics of historical alloys, and between element characteristics and performance data;

[0078] Exemplarily, for the element feature x and the element feature y of the historical alloy, the calculation process of the correlation coefficient r is as follows:

[0079]

[0080] where x i and i are the values ​​of element characteristics x and element characteristics y of the i-th historical alloy sample, respectively, x m and x m are the means of element characteristics x and element characteristics y of N historical alloy samples, respectively; N is the number of historical alloy samples;

[0081] S32: For element features whose absolute value of correlation coefficient is greater than the screening threshold, only one element feature with a large correlation coefficient with the performance data is retained;

[0082] Through the above steps, the correlation coefficients between element features and between element features and targets (performance) are compared successively, and the element feature combination that can better describe the target is screened out. When the absolute value of the correlation coefficient between element features is greater than the threshold, it means that there is a strong correlation between the two element features X and Y, and the information they include and the degree of contribution to model construction are similar. Only the one with a higher degree of correlation with performance data needs to be retained to avoid redundancy.

[0083] S33: Select the basic model, use the packaging method to select the element features retained by S32, use different element feature combinations as the input of the basic model, use the performance data as the output of the basic model, establish different models, judge the accuracy of all models, and use the element feature combination used when the model has the highest accuracy as the final selected key element feature combination.

[0084] In some embodiments, the accuracy of all models is judged based on the calculation results of MAPE (Mean Absolute Percentage Error, referred to as MAPE).

[0085] The calculation method of MAPE is:

[0086]

[0087] Among them, V Predictive value.i and V Measured value.i are the model predicted performance and true performance of the i-th historical alloy sample, respectively, and N is the number of samples.

[0088] This step further screens the features retained after S32 screening.

[0089] Among them, the basic model can select a random forest (RF) model or a support vector regression (SVR) model.

[0090] Among them, the wrapper method can adopt the recursive feature elimination method (RFE for short).

[0091] During the feature screening process, RFE will use different element feature combinations as the input of the basic model and the performance data as the output of the basic model to establish different models. The accuracy of all models will be judged based on the calculation results of MAPE, and the element feature combination used when MAPE is the largest will be used as the key element feature combination finally screened out.

[0092] This step comprehensively evaluates and screens the element characteristics to identify and extract key element characteristics that have a significant impact on the performance data.

[0093] S4: For each historical alloy, the phase diagram is accurately calculated based on the thermodynamic data and the key phase diagram parameters are extracted;

[0094] Based on thermodynamic principles and computational materials science methods, binary or multivariate phase diagrams of alloys can be accurately drawn.

[0095] Among them, key phase diagram parameters include but are not limited to liquidus temperature and solid solubility;

[0096] Taking Cu-Be (copper-beryllium) alloy as an example, Therm-Cal software is used to calculate the Cu-Be binary phase diagram. The liquidus temperature A (℃) corresponding to the Be content in the historical alloy sample in the first historical alloy data set and the maximum solid solubility B (wt.%) at the corresponding aging temperature are obtained. These two values ​​will have an important impact on the alloy solid solution process and aging process respectively. In order to ensure that all features are at the same order of magnitude during the modeling process, the following formulas A and B are used for processing:

[0097]

[0098] SS=B×10000 (2)

[0099] Where A is the liquidus temperature corresponding to the Be content in the historical alloy sample, B is the maximum solid solubility (mass ratio) at the corresponding aging temperature, ST is the solid solution temperature corresponding to the historical alloy sample, SM is 100 times the ratio of the solid solution temperature to the liquidus temperature, and SS is 10,000 times the maximum solid solubility B. When modeling, SM and SS are added to the training as supplementary features in addition to composition and process.

[0100] As a precipitation-strengthened copper alloy, the choice of solution treatment system is particularly important for the preparation of high-performance Cu-Be alloy. If the solution temperature is too high or the holding time is too long, the alloy grain size is likely to increase significantly or even overburn, resulting in poor uniformity of the material structure. At the same time, the genetic characteristics of the coarse grains and uneven structure of the copper matrix will have an adverse effect on improving the structure and performance through the comprehensive regulation of subsequent processing and aging processes. If the solution temperature is too low or the time is too short, it is difficult for the solute atoms to dissolve into the copper matrix. These undissolved alloy atoms will exist in the form of large primary phases or segregated forms, which will undoubtedly have an adverse effect on the uniformity of the material structure. Subsequent processing and aging processes will also find it difficult to transform these alloy atoms into uniform and fine precipitation phases, which will seriously deteriorate the mechanical properties and electrical conductivity of the copper alloy. SM represents the ratio of the solution temperature to the liquidus temperature. This value is of great significance for the choice of solution process: when SM is large, it means that the solution temperature is high for the Cu-Be alloy of this composition, and vice versa. Therefore, SM is an important parameter to measure whether the solution process system is reasonable. The experimental results show that it is reasonable and effective to use it as an important parameter of the solution process time step of the timing model to participate in model training.

[0101] The selection of aging temperature and time has an important influence on the final properties of Cu-Be alloy. If the aging temperature is too high or the time is too long, the size of the precipitated phase will increase and the alloy will be recrystallized and softened, thereby reducing the mechanical properties of the alloy; if the aging temperature is too low or the aging time is too short, the solute atoms cannot be fully precipitated, and the precipitation strengthening effect is difficult to reflect; in addition, the atoms dissolved in the copper matrix will also cause serious lattice distortion in the matrix, which will have an adverse effect on the electrical conductivity of the alloy. SS represents the limiting solid solubility of the alloy at the aging temperature. The limiting solid solubility is the maximum amount of Be atoms that will remain in the copper matrix when the Cu-Be alloy is aged for a long time at a certain temperature. When the concentration of Be atoms in the solid solution is reduced to the limiting solid solubility, it is difficult to precipitate the remaining Be atoms even if the aging time is further extended. Therefore, SS has an important reference significance for the selection of aging time. At the same time, the limiting solid solubility of Be atoms in Cu increases with increasing temperature, which means that the higher the aging temperature, the more solute atoms will remain in the copper matrix in theory. However, the increase in temperature will increase the precipitation motive force of Be atoms, which is conducive to the full precipitation of Be atoms. Therefore, the SS parameter is also of great reference value for the selection of alloy aging temperature.

[0102] S5: Based on the first historical alloy data set, key element characteristics and key phase diagram parameters, a second historical alloy data set is obtained; the second historical alloy data set includes a plurality of second historical alloy data, each of which corresponds to a key element characteristic, process data, performance data and key phase diagram parameter of a historical alloy;

[0103] This step obtains the second historical alloy data set, which integrates the key element characteristics of the first historical alloy data set obtained in S1 and the key element characteristics screened in S3, and incorporates the key phase diagram parameters extracted in step 4. Specifically, each data includes: (1) key element characteristics obtained in S3. (2) process data in the first historical alloy data set obtained in S1, and the processing technology includes solid solution, primary cold rolling, primary aging, secondary cold rolling, and secondary aging. Specific parameters include: solid solution temperature, solid solution time, primary cold rolling reduction, primary aging temperature, primary aging time, secondary cold rolling reduction, secondary aging temperature, and secondary aging time. (3) phase diagram information obtained in S4. (4) performance data of different stages in the first historical alloy data set. The stages include solid solution state and aging state. The performance data of the solid solution state include hardness, conductivity, and elongation; the performance data of the cold rolled state include hardness, conductivity, and elongation; the performance data of the aging state include hardness, conductivity, and elongation.

[0104] S6: Build an alloy prediction model based on the machine learning model, and use the second historical alloy data set as a sample data set. Apply the intelligent optimization algorithm to automatically search for the optimal parameter combination of the alloy prediction model and determine the best model.

[0105] The model combines the first historical alloy data set obtained in step S1 and the key element characteristics screened in step S3, and incorporates the phase diagram parameters extracted in step S14 to enhance the model's understanding and application of metallurgical principles in alloy design.

[0106] In the second historical alloy data set, the composition data of the historical alloys are converted into key element characteristic values ​​through calculation, and the types of converted key element characteristics are consistent with the types screened in S3.

[0107] Among them, through intelligent optimization algorithms, including sparrow optimization algorithm, genetic algorithm or particle swarm optimization (PSO).

[0108] At the same time, this step integrates key element characteristics and key phase diagram parameters into the neural network model, and then performs parameter self-optimization and model self-screening, and determines the best performance model based on model generalization ability and test accuracy. The alloy prediction model uses key element characteristics, phase diagram information, and process data as input and performance data as output, which can achieve accurate prediction of alloy properties.

[0109] Among them, the machine learning model can adopt a time series neural network model, and the optional models include LSTM, Transformer, and LSTM-Transformer.

[0110] In this step, the intelligent optimization algorithm is used to try different parameter combinations to construct the LSTM-Transformer model. In this model, for example: the key element characteristics are used as the input of the first time step, and the cast properties are predicted; the solution temperature and solution time are used as the input of the second step, and the performance data of the solution state is predicted; the relevant parameters of the first cold rolling and the first aging are used as the input of the third time step, and the first aging performance is predicted; the relevant parameters of the second cold rolling and the second aging are used as the input of the fourth time step, and the second aging performance is predicted. Then the intelligent optimization algorithm performs parameter combination and optimization on the LSTM-Transformer model.

[0111] Furthermore, different key phase diagram parameters obtained in S4 are used as inputs at different time steps in the LSTM-Transformer model as needed to guide the alloy design process and provide direct guidance on metallurgical principles.

[0112] Specifically, the intelligent optimization algorithm selected in this example is the sparrow optimization algorithm, which is a prior art and will not be described in detail here.

[0113] Among them, the solution temperature, solution time, primary cold rolling reduction, primary aging temperature, and primary aging time can be set manually; the SM phase diagram information and SS phase diagram information can be calculated based on the alloy composition.

[0114] Furthermore, in this embodiment, the key element characteristics screened out are lattice constant, electron affinity, thermal expansion coefficient and elastic modulus, and these key element characteristics are used as inputs of the first time step to participate in modeling.

[0115] Furthermore, the different process parameters of solution treatment-first cold rolling-first aging-second cold rolling-second aging are: the solution treatment temperature can be set to 600°C to 950°C; the solution treatment time can be set to 5 to 480 min; the first cold rolling reduction and the second cold rolling reduction can be set to 0 to 95%; the first aging temperature and the second aging temperature can be set to 300 to 500°C, and the first aging time and the second aging time can be set to 1 min to 1440 min.

[0116] Furthermore, the parameter combination includes the number of neural network layers, the number of neuron nodes, the learning rate and the loss rate. The number of neural network layers is 2 to 30 layers, the number of neuron nodes is 8 to 1024, and the learning rate is 10 -6 Up to 10 -1 , the dropout rate is 0.01 to 0.6.

[0117] Optionally, the number of iterations of the intelligent optimization algorithm is set to 500 to 1500, preferably 1000, to obtain the optimal parameter combination within a reasonable computing time.

[0118] S7: Based on the optimal model, the key element characteristic values ​​and process data corresponding to the optimal performance data are obtained; the key element characteristic values ​​corresponding to the optimal performance data are converted into composition data (elements and their contents), thereby obtaining the composition data and process data of the high-performance new alloy.

[0119] Specifically, the key element characteristic values ​​of different alloy compositions are calculated and compared with the key element characteristic values ​​corresponding to the optimal performance data, so as to confirm the optimal alloy composition.

[0120] The elements contained in the new alloy are not limited to those contained in the historical alloy, but can also be other elements in the periodic table.

[0121] In this way, the automation and intelligence of alloy design can be achieved.

[0122] The optimal alloy composition in this example is Be: 1.5-2.0wt.%, Ni: 0.3-0.8wt.%, Co 0.3-0.8wt.%, Cr 0.1-0.4wt.%, Mg: 0.05-0.2wt.%, and the rest is Cu.

[0123] Obviously, those skilled in the art should understand that the above-mentioned units or steps of the present application can be implemented by a general-purpose computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the present application is not limited to any specific combination of hardware and software.

[0124] Embodiment 2:

[0125] The present application embodiment provides a prediction system for high-performance new alloys based on machine learning and phase diagram information, as shown in the attached Figure 2 shown.

[0126] A first data acquisition module is used to acquire a first historical alloy data set; the first historical alloy data set includes a plurality of first historical alloy data, each of which corresponds to composition data, process data and performance data of a historical alloy, wherein the composition data of the historical alloy includes the content and proportion of each element included in the historical alloy, and the sum of the proportions of all elements of each historical alloy is equal to 100%, so as to ensure the integrity and accuracy of the data;

[0127] A data processing module is used to obtain the characteristic factor of each element in the historical alloy and calculate the statistical characteristics of the characteristic factors of all elements therein to obtain the element characteristics of the historical alloy;

[0128] The data processing module can pre-process the collected historical alloy data sets, including data cleaning, normalization and feature extraction, so as to form the element characteristics of each historical alloy;

[0129] Element feature screening module, used to screen element features for historical alloys to obtain key element features;

[0130] Phase diagram information acquisition module, used to calculate the phase diagram for each historical alloy and extract key phase diagram parameters;

[0131] Among them, the key phase diagram parameters include liquidus temperature and solid solubility;

[0132] These parameters are then used as input features of the model to guide the alloy design process and provide direct guidance from metallurgical principles.

[0133] A second data acquisition module is used to obtain a second historical alloy data set based on the first historical alloy data set, key element characteristics and key phase diagram parameters; the second historical alloy data set includes a plurality of second historical alloy data, each second historical alloy data corresponding to a key element characteristic, process data, performance data and key phase diagram parameter of a historical alloy;

[0134] The model optimization module builds an alloy prediction model based on the machine learning model, uses the second historical alloy data set as a sample data set, applies an intelligent optimization algorithm, automatically searches for the optimal parameter combination of the alloy prediction model, and determines the best model;

[0135] This process is based on the model's generalization ability and test accuracy to enhance the model's understanding and application of metallurgical principles in alloy design, determine the best model, and thus achieve integrated and precise design and optimization of alloy composition and process.

[0136] The alloy design module obtains the key element characteristic values ​​and process data corresponding to the optimal performance data based on the optimal model; the key element characteristic values ​​corresponding to the optimal performance data are converted into composition data (elements and their contents), thereby obtaining the composition data and process data of the high-performance new alloy.

[0137] The specific implementation of this embodiment refers to Embodiment 1 and will not be described in detail here.

[0138] Embodiment 3:

[0139] An embodiment of the present application provides an electronic device, including: a memory and a processor;

[0140] The memory is used to store computer programs;

[0141] The processor is used to call the computer program to execute the method as described above.

[0142] Embodiment 4:

[0143] An embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed on an electronic device, the electronic device implements the method as described above.

[0144] Embodiment 5:

[0145] An embodiment of the present application further provides a computer program product, including a computer program. When the computer program is executed on an electronic device, the electronic device implements the method as described above.

[0146] The embodiments of the present application also provide a system, an electronic device, a computer-readable storage medium, and a computer program product. The specific implementation methods can refer to the specific embodiments of the above methods and will not be repeated here.

[0147] It should be understood that in the embodiments of the present invention, the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the type of device.

[0148] The readable storage medium is a computer-readable storage medium, which may be an internal storage unit of the controller described in any of the foregoing embodiments, such as a hard disk or memory of the controller. The readable storage medium may also be an external storage device of the controller, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the controller. Furthermore, the readable storage medium may also include both an internal storage unit of the controller and an external storage device. The readable storage medium is used to store the computer program and other programs and data required by the controller. The readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0149] Based on this understanding, the technical solution of the present invention, in essence or in other words, the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned readable storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., various media that can store program codes.

[0150] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0151] It should be noted that the terms such as "first", "second" and the like in the specification and claims of the present application and the above-mentioned drawings are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is a specific relationship or order between these entities or operations. It should be understood that the words "first", "second" and the like do not limit the quantity and execution order, and the words "first" and "second" and the like do not limit them to be necessarily different. It should be understood that the data used in this way can be interchanged under appropriate circumstances. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, article or device including a series of elements includes not only those elements clearly listed, but also other elements that are not clearly listed, or may also include elements inherent to such process, method, product or device. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0152] The above description of the embodiments of the present application is only a partial embodiment of the present application, which is used to enable professionals in the field to implement or use the content of the present application, and is not used to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for designing a high-performance new alloy, characterized in that: include: S1: Obtain a first historical alloy data set; the first historical alloy data set includes a plurality of first historical alloy data, each of which corresponds to composition data, process data and performance data of a historical alloy, wherein the composition data of the historical alloy includes the content and proportion of each element included in the historical alloy, and the sum of the proportions of all elements of each historical alloy is equal to 100%; S2: For the historical alloy, obtain the characteristic factor of each element therein, and calculate the statistical properties of the characteristic factors of all elements therein to obtain the element characteristics of the historical alloy; S3: For historical alloys, element characteristics are screened to obtain key element characteristics; S4: For each historical alloy, calculate the phase diagram and extract the key phase diagram parameters; S5: Based on the first historical alloy data set, key element characteristics and key phase diagram parameters, a second historical alloy data set is obtained; the second historical alloy data set includes a plurality of second historical alloy data, each of which corresponds to a key element characteristic, process data, performance data and key phase diagram parameter of a historical alloy; S6: Build an alloy prediction model based on the machine learning model, use the second historical alloy data set as a sample data set, apply the intelligent optimization algorithm, automatically search for the optimal parameter combination of the alloy prediction model, and determine the best model; S7: Based on the optimal model, the key element characteristic values ​​and process data corresponding to the optimal performance data are obtained; the key element characteristic values ​​corresponding to the optimal performance data are converted into composition data, thereby obtaining the composition data and process data of the high-performance new alloy.

2. The method according to claim 1, characterized in that The extraction accuracy of the content of the elements was selected to be 0.01 atomic percent.

3. A method according to claim 1, characterized in that The characteristic factors include atomic size information, electronic information and energy information.

4. A method according to claim 1, characterized in that The S2 includes: For each historical alloy, the weighted average value, weighted mean square error and weighted average deviation of the same characteristic factor of all elements are calculated as the average characteristic factor, mean square error characteristic factor and average deviation characteristic factor of the historical alloy; wherein the weighted value of each element is the proportion of the corresponding element in the historical alloy; For each historical alloy, the element characteristics include the average characteristic factor, the mean square deviation characteristic factor and the mean deviation characteristic factor.

5. A method according to claim 1, characterized in that: The S3 includes: S31: Set the screening threshold r th ; S32: Pearson correlation coefficient is used to calculate the correlation coefficient between element characteristics and element characteristics of historical alloys, and between element characteristics and performance data; S32: For element features whose absolute value of correlation coefficient is greater than the screening threshold, only one element feature with a large correlation coefficient with the performance data is retained; S33: Select the basic model, use the packaging method to select the element features retained by S32, use different element feature combinations as the input of the basic model, use the performance data as the output of the basic model, establish different models, judge the accuracy of all models, and use the element feature combination used when the model has the highest accuracy as the final selected key element feature combination.

6. The method according to claim 1, characterized in that The extraction accuracy of the key phase diagram parameters is 0.1°C and 0.01 at.%.

7. A high-performance new alloy design system, comprising: A first data acquisition module is used to acquire a first historical alloy data set; the first historical alloy data set includes a plurality of first historical alloy data, each of which corresponds to composition data, process data and performance data of a historical alloy, wherein the composition data of the historical alloy includes the content and proportion of each element included in the historical alloy, and the sum of the proportions of all elements of each historical alloy is equal to 100%; A data processing module is used to obtain the characteristic factor of each element in the historical alloy and calculate the statistical characteristics of the characteristic factors of all elements therein to obtain the element characteristics of the historical alloy; The data processing module can pre-process the collected historical alloy data sets, including data cleaning, normalization and feature extraction, so as to form the element characteristics of each historical alloy; Element feature screening module, used to screen element features for historical alloys to obtain key element features; Phase diagram information acquisition module, used to calculate the phase diagram for each historical alloy and extract key phase diagram parameters; A second data acquisition module, for obtaining a second historical alloy data set based on the first historical alloy data set, key element characteristics and key phase diagram parameters; the second historical alloy data set includes a plurality of second historical alloy data, each second historical alloy data corresponding to a key element characteristic value, process data, performance data and key phase diagram parameter; A model optimization module is used to build an alloy prediction model based on a machine learning model, and use the second historical alloy data set as a sample data set, apply an intelligent optimization algorithm, automatically search for the optimal parameter combination of the alloy prediction model, and determine the best model; The alloy design module obtains the key element characteristic values ​​and process data corresponding to the optimal performance data based on the optimal model; the key element characteristic values ​​corresponding to the optimal performance data are converted into composition data, thereby obtaining the composition data and process data of the high-performance new alloy.

8. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer programs; The processor is configured to call the computer program to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed on an electronic device, the electronic device implements the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed on an electronic device, the electronic device implements the method according to any one of claims 1 to 6.

Images