Titanium alloy alpha / beta phase transition temperature prediction method and related device

Abstract

The invention belongs to the technical field of calculation material science and alloy design optimization, and particularly relates to a titanium alloy alpha / beta phase transition temperature prediction method and related device.A linear regression algorithm is used for establishing a basic prediction model of the alpha / beta phase transition temperature; generating a correction term population of the basic prediction model through symbolic regression calculation based on a genetic algorithm; on the basis of a non-dominated sorting algorithm, according to the length of each individual in the correction term population and the prediction precision of the basic prediction model after the correction term is added, a correction term set on the Pareto front is screened out, and an optimal correction term is selected; fusing the basic prediction model and the optimal correction term to obtain a titanium alloy alpha / beta phase transition temperature prediction model; and calculating the alpha / beta phase transition temperature of the titanium alloy to be predicted by using the titanium alloy alpha / beta phase transition temperature prediction model. The problem that an existing machine learning method is insufficient in physical interpretation can be solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computational materials science and alloy design optimization, and particularly relates to a titanium alloy alpha / beta phase transition temperature prediction method and a related device. BACKGROUND

[0002] The mechanical properties of titanium alloys are strongly dependent on their microstructure characteristics, and the alpha / beta phase transition temperature is a key thermodynamic parameter that determines the phase composition and organizational form of titanium alloys, directly affecting the optimization effect of the final mechanical properties. Studies have shown that the alpha / beta phase transition temperature value is closely related to the content ratio of alpha-stabilizing elements (including aluminum) and beta-stabilizing elements (including vanadium and molybdenum). By changing the chemical composition of these elements, the alpha / beta phase transition temperature can be directly controlled, and the target microstructure and mechanical properties can be indirectly obtained.

[0003] The determination of the alpha / beta phase transition temperature mainly relies on experimental methods, empirical formula methods, and machine learning methods, but these methods have certain limitations. Traditional experimental methods are tedious and costly, and it is difficult to establish an explicit quantitative relationship between composition and temperature. Existing empirical formulas (such as the piecewise linear model based on the alpha / beta phase transition temperature of pure titanium) are simple in form, but they generally ignore the complex nonlinear interactions between elements, which are still applicable in binary systems, but the prediction reliability significantly decreases in multi-element titanium alloys. Machine learning methods can effectively solve the defects of relying on experimental methods and empirical formula methods, but the "black box" nature of machine learning methods limits the output of analytical expressions with clear physical meaning, restricting their practical application in composition design.

[0004] In summary, existing machine learning methods have insufficient interpretability, which restricts the further development of titanium alloy design. SUMMARY

[0005] To solve the problems in the prior art, the purpose of the present application is to provide a titanium alloy alpha / beta phase transition temperature prediction method and a related device, which can solve the problem of insufficient physical interpretability of existing machine learning methods.

[0006] To achieve the above-mentioned purpose, the following technical solutions are adopted in the present application: The titanium alloy alpha / beta phase transition temperature prediction method comprises the following processes: The content of alloying elements of a titanium alloy with known composition is input into a titanium alloy alpha / beta phase transition temperature prediction model that has been constructed, and the alpha / beta phase transition temperature of the titanium alloy is calculated using the titanium alloy alpha / beta phase transition temperature prediction model; The construction process of the titanium alloy alpha / beta phase transition temperature prediction model comprises: The basic prediction model of the alpha / beta phase transition temperature is established by using a linear regression algorithm based on a constructed titanium alloy data set, the titanium alloy data set including chemical composition information of the titanium alloy and measured values of the alpha / beta phase transition temperature corresponding to the chemical composition information, the chemical composition information including types of alloying elements contained in the titanium alloy and contents of the alloying elements; A correction term population of the basic prediction model is generated by symbolic regression calculation based on a genetic algorithm. The length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding the correction term are calculated. Based on a non-dominated sorting algorithm, all correction terms in the correction term population are traversed according to the length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding the correction term, and a Pareto front is formed. An optimal correction term is selected from the correction terms on the Pareto front. The titanium alloy alpha / beta phase transition temperature prediction model is obtained by fusing the basic prediction model and the optimal correction term.

[0007] Preferably, the prediction accuracy of the basic prediction model after adding the correction term is the average absolute error between the measured value of the alpha / beta phase transition temperature and the alpha / beta phase transition temperature predicted by the basic prediction model after adding the correction term.

[0008] Preferably, the process of selecting the optimal correction term from the correction terms on the Pareto front includes: The crowding distance of all correction terms on the Pareto front is calculated, and the correction term with the largest crowding distance is selected as the optimal correction term.

[0009] Preferably, the fusion method of fusing the basic prediction model and the optimal correction term is linear addition of the basic prediction model and the optimal correction term.

[0010] Preferably, the alloying elements contained in the titanium alloy include at least one of Al, Mo, V, Cr, Fe, Nb, Sn, Zr, Cu, Si, C, H, O and N.

[0011] Preferably, the construction process of the titanium alloy data set includes: At least one of historical experimental data and literature data is used to construct an initial data set of the titanium alloy. The initial data set of the titanium alloy is subjected to missing value filling and abnormal data removal to obtain the titanium alloy data set.

[0012] Preferably, the specific process of filling the missing values in the initial data set of the titanium alloy and removing the abnormal data includes: The missing values in the initial data set of the titanium alloy are filled with 0, and the abnormal data is identified and removed by using the Rajda criterion.

[0013] The application further provides a titanium alloy alpha / beta phase transition temperature prediction system for implementing the titanium alloy alpha / beta phase transition temperature prediction method of the application. The input module is configured to input the content of alloy elements of a titanium alloy with known composition into the titanium alloy alpha / beta phase transition temperature prediction model that has been constructed. The calculation module is configured to calculate the alpha / beta phase transition temperature of the titanium alloy by using the titanium alloy alpha / beta phase transition temperature prediction model and the content of the alloy elements. The output module is configured to output the alpha / beta phase transition temperature calculated by the titanium alloy alpha / beta phase transition temperature prediction model. The construction process of the titanium alloy alpha / beta phase transition temperature prediction model comprises the following steps. A basic prediction model of the alpha / beta phase transition temperature is established by using a linear regression algorithm based on a constructed titanium alloy data set; the titanium alloy data set comprises chemical composition information of a titanium alloy and measured values of the alpha / beta phase transition temperature corresponding to the chemical composition information; the chemical composition information comprises the types of alloy elements contained in the titanium alloy and the content of each alloy element. A correction term population of the basic prediction model is generated by using a symbolic regression calculation based on a genetic algorithm. The length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding the correction term are calculated. Based on a non-dominated sorting algorithm, all correction terms in the correction term population are traversed according to the length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding the correction term, and a Pareto front is formed. An optimal correction term is selected from the correction terms on the Pareto front. The basic prediction model and the optimal correction term are fused to obtain the titanium alloy alpha / beta phase transition temperature prediction model.

[0014] The application further provides an electronic device comprising: one or more processors; a memory having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the titanium alloy alpha / beta phase transition temperature prediction method of the application.

[0015] The application further provides a storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the titanium alloy alpha / beta phase transition temperature prediction method of the application.

[0016] The application has the following beneficial effects: The titanium alloy alpha / beta phase transition temperature prediction method constructs a basic prediction model based on a linear regression algorithm and a correction term based on genetic algorithm symbolic regression calculation of a titanium alloy alpha / beta phase transition temperature prediction model; wherein the basic prediction model is obtained based on the titanium alloy data set and by using the linear regression algorithm, so the basic prediction model is a mathematical analytical formula, the titanium alloy alpha / beta phase transition temperature prediction model is obtained by further optimizing the basic prediction model, and the titanium alloy alpha / beta phase transition temperature prediction model is also an alpha / beta phase transition temperature analytical formula with clear physical meaning, thereby effectively solving the limitation of insufficient physical interpretation of the existing machine learning model in the titanium alloy alpha / beta phase transition temperature prediction. In the technical scheme of the present application, the basic prediction model obtained by using the linear regression algorithm has a clear mathematical analytical formula, and the mathematical analytical formula has element information and content information of each alloying element of the titanium alloy, thereby avoiding the black box characteristics of the existing machine learning method, the basic prediction model gives the final titanium alloy alpha / beta phase transition temperature prediction model of the present application basic interpretability, the correction term generated by the symbolic regression calculation based on the genetic algorithm can accurately capture the nonlinear interaction between alloying elements, the optimal correction term selected by the Pareto frontier construction and the non-dominated sorting algorithm can not only compensate for the fitting defects of the basic prediction model and improve the prediction accuracy, but also ensure the practicality and simplicity of the basic prediction model by controlling the length of the correction term, and finally form a titanium alloy alpha / beta phase transition temperature prediction model with clear physical meaning and high prediction accuracy, which can clearly reflect the quantitative relationship between the chemical composition information of the titanium alloy and the alpha / beta phase transition temperature. At the same time, the titanium alloy alpha / beta phase transition temperature prediction model only needs to input the content of the alloying elements of the titanium alloy to quickly output the titanium alloy alpha / beta phase transition temperature prediction result, thereby greatly improving the acquisition efficiency of the alpha / beta phase transition temperature in the titanium alloy design and reducing the cost waste caused by blind experiments, and the selection of the optimal correction term also ensures the stability and reliability of the titanium alloy alpha / beta phase transition temperature prediction model, thereby improving the application value of the titanium alloy alpha / beta phase transition temperature prediction model in the titanium alloy design. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The figure is a flow chart for constructing the titanium alloy alpha / beta phase transition temperature prediction model in the embodiment of the present application.

[0018] Figure 2 The figure is a Pareto frontier distribution diagram of the expression in the symbolic regression calculation in the embodiment of the present application.

[0019] Figure 3A tree structure diagram of an expression of a titanium alloy α / β phase transition temperature prediction model in an embodiment of the present application.

[0020] Figure 4 A structure block diagram of a titanium alloy α / β phase transition temperature prediction system in an embodiment of the present application.

[0021] Figure 5 A structure schematic diagram of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0022] The present application will be further described in detail below with reference to specific embodiments and the accompanying drawings. A person of ordinary skill in the art can implement the present application based on these descriptions. In addition, the embodiments of the present application involved in the following descriptions are generally only embodiments of a part of the present application, rather than all embodiments. Therefore, all other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application without making creative efforts should belong to the protection scope of the present application.

[0023] Embodiment 1 Referring to Figure 1 The titanium alloy α / β phase transition temperature prediction method in the embodiment includes the following steps: Step one, at least one of historical experimental data and literature data is collected to construct a titanium alloy initial data set of titanium alloy component α / β phase transition temperature; the titanium alloy initial data set includes chemical component information of the titanium alloy and α / β phase transition temperature measured values corresponding to the chemical component information, wherein the elements of the titanium alloy include Ti and alloy elements, the alloy elements include at least one of Al, Mo, V, Cr, Fe, Nb, Sn, Zr, Cu, Si, C, H, O and N, the chemical component information in the present application includes the types of alloy elements contained in the titanium alloy and the contents of each alloy element, and the content of each alloy element can be the atomic percentage or the mass percentage of each alloy element in the titanium alloy. The data scale in the titanium alloy initial data set is not less than 50 groups (wherein, for a titanium alloy with a known composition, the chemical component information of the titanium alloy and the α / β phase transition temperature measured values corresponding to the chemical component information are regarded as one group of data), so as to ensure the prediction reliability of the basic prediction model established in step three.

[0024] Step two, missing value filling and abnormal data elimination are performed on the titanium alloy initial data set constructed in step one to obtain a titanium alloy data set; the specific process of performing missing value filling and abnormal data elimination on the titanium alloy initial data set comprises: using 0 to fill the missing values in the titanium alloy initial data set, and using the Raje criterion to identify and eliminate abnormal data. The titanium alloy data set is similar to the titanium alloy initial data set in step one, and includes chemical composition information of the titanium alloy and the measured value of the α / β phase transition temperature corresponding to the chemical composition information.

[0025] Step three, a basic prediction model of the α / β phase transition temperature is established based on the titanium alloy data set constructed in step two by using a linear regression algorithm. The prediction residual of the basic prediction model is calculated, and the prediction residual of the basic prediction model refers to the difference between the predicted value of the α / β phase transition temperature and the measured value of the α / β phase transition temperature, wherein the predicted value of the α / β phase transition temperature is the α / β phase transition temperature calculated by the basic prediction model, and the calculation equation of the prediction residual is as follows:

[0026] wherein, is the prediction residual, is the measured value of the α / β phase transition temperature, and the unit is ℃; is the predicted value of the α / β phase transition temperature, and the unit is ℃.

[0027] Step four, a correction term population of the basic prediction model obtained in step three is generated by sign regression calculation based on a genetic algorithm, and the correction term in the correction term population is used to offset the prediction residual of the basic prediction model; wherein the process of sign regression calculation based on a genetic algorithm comprises: initializing calculation parameters and iterative evolution; wherein the calculation parameters include population size, evolution generation number, crossover probability, mutation probability and initial population generation method, and the initial population generation method can adopt a random generation method.

[0028] Step five, the length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding the correction term are calculated; wherein the prediction accuracy of the basic prediction model after adding the correction term is the average absolute error between the measured value of the α / β phase transition temperature and the α / β phase transition temperature predicted by the basic prediction model after adding the correction term.

[0029] Step six, based on the non-dominated sorting algorithm, according to the length of each correction item in the correction item population and the prediction accuracy of the base prediction model after adding the correction item, all correction items in the correction item population are traversed to form a Pareto front; the optimal correction item is selected from the correction item on the Pareto front; specifically, the process of selecting the optimal correction item from the correction item on the Pareto front includes: calculating the crowding distance of all correction items on the Pareto front, and selecting the correction item with the largest crowding distance as the optimal correction item.

[0030] Step seven, fusing the base prediction model obtained in step three and the optimal correction item obtained in step six to obtain a final titanium alloy α / β phase transition temperature prediction model. Specifically, the fusion method of fusing the base prediction model and the optimal correction item is linear addition of the base prediction model and the optimal correction item, and the mathematical expression of the titanium alloy α / β phase transition temperature prediction model is as follows:

[0031] In the formula, The final α / β phase transition temperature prediction value is represented by T, and the unit is ℃. The optimal correction item is represented by ΔT, and the unit is ℃.

[0032] Step eight, for the titanium alloy to be predicted α / β phase transition temperature, input the content of the alloying element of the titanium alloy to be predicted into the titanium alloy α / β phase transition temperature prediction model constructed in step seven, and calculate the α / β phase transition temperature of the titanium alloy by using the titanium alloy α / β phase transition temperature prediction model.

[0033] In the above scheme of the present application, by constructing a prediction framework of fusing a linear reference model (i.e. the base prediction model obtained in step three) and a symbolic regression correction item (i.e. the optimal correction item), a multi-objective optimization selection mechanism of the Pareto front based on the crowding distance is adopted, an α / β phase transition temperature analytical formula (i.e. the final titanium alloy α / β phase transition temperature prediction model obtained in step seven) with high prediction accuracy and clear physical meaning is generated, the problem of insufficient physical interpretation of traditional machine learning models is effectively solved, a reliable theoretical basis and practical calculation tool are provided for titanium alloy design, and the efficiency and engineering applicability of titanium alloy design are significantly improved.

[0034] Example 2 The titanium alloy α / β phase transition temperature prediction method of the present embodiment includes the following steps: Step one, collect the chemical composition information of titanium alloy containing 14 alloying elements including Al, Mo, V, Cr, Fe, Nb, Sn, Zr, Cu, Si, C, H, O and N from public literature and experimental data, and the measured values of the alpha / beta phase transition temperature corresponding to the chemical composition information, and construct a titanium alloy initial data set containing 200 groups of data based on the measured values of the alpha / beta phase transition temperature corresponding to the collected chemical composition information.

[0035] Step two, fill in the missing values and remove the abnormal data of the titanium alloy initial data set constructed in step one to obtain a titanium alloy data set.

[0036] Step three, using a linear regression algorithm, based on the titanium alloy data set constructed in step two, a basic prediction model of the alpha / beta phase transition temperature is established. The prediction residual of the basic prediction model is calculated.

[0037] Step four, by symbol regression calculation based on genetic algorithm, a correction term population of the basic prediction model obtained in step three is generated, and the correction terms in the correction term population are used to offset the prediction residual of the basic prediction model; wherein the process of symbol regression calculation based on genetic algorithm includes: initializing calculation parameters and iterative evolution; wherein the population size of the calculation parameters is set to 1000, the evolution generation number is set to 20, the crossover probability is set to 0.2, the mutation probability is set to 0.2, and the initial population generation method adopts a random generation method.

[0038] Step five, the length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding the correction term are calculated.

[0039] Step six, based on the non-dominated sorting algorithm, according to the length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding the correction term, all correction terms in the correction term population are traversed to form a Pareto front; the optimal correction term is selected from the correction terms on the Pareto front; specifically, the process of selecting the optimal correction term from the correction terms on the Pareto front includes: calculating the crowding distance of all correction terms on the Pareto front, and selecting the correction term with the largest crowding distance as the optimal correction term.

[0040] Step seven, linearly adding the basic prediction model and the optimal correction term to obtain the final titanium alloy alpha / beta phase transition temperature prediction model.

[0041] In the above step one, four samples are randomly selected from the titanium alloy initial data set, and the mass percentages of alloying elements in the titanium alloy corresponding to the samples are as follows: Sample 1: Al: 3.2%, V: 10.3%, Fe: 2.2%, H: 0.016%, N: 0.009%, alpha / beta phase transition temperature: 805℃; Sample 2: Al: 6.495%, V: 3.3%, Fe: 0.1%, O: 0.2%, α / β phase transition temperature: 1021℃; Sample 3: Al: 2.5%, Mo: 1%, Sn: 11%, Si: 0.2%, α / β phase transition temperature: 925℃; Sample 4: Al: 2%, V: 11.5%, Fe: 0.2%, Zr: 11%; C: 0.08%; H: 0.015%, O: 0.015%, N: 0.05%, α / β phase transition temperature: 738℃; The measured and original values ​​(i.e., the α / β phase transition temperatures of each sample given in published literature and experimental data) of these four samples are shown in Table 1: Table 1

[0042] In Table 1, the measured α / β phase transition temperature refers to the measured value of the α / β phase transition temperature of the sample, and the original α / β phase transition temperature refers to the original value of the α / β phase transition temperature of the sample. As shown in Table 1, the absolute value of the relative error between the measured and original values ​​of the α / β phase transformation temperature of the four samples does not exceed 1%. Therefore, the α / β phase transformation temperature values ​​recorded in the initial dataset of titanium alloys constructed in step one are used.

[0043] In step two above, the initial dataset of titanium alloy is checked to ensure there are no missing values. Then, the Laida criterion is used to identify and delete outlier data in the initial dataset of titanium alloy.

[0044] In step three above, using all alloying elements as independent variables and the α / β phase transformation temperature as the dependent variable, a basic prediction model for the α / β phase transformation temperature is established using a multiple linear regression algorithm. In this scheme, the process of establishing the basic prediction model for the α / β phase transformation temperature using a multiple linear regression algorithm with all alloying elements as independent variables and the α / β phase transformation temperature as the dependent variable is a conventional process. The specific construction process will not be elaborated upon in this invention. The mathematical expression of the obtained basic prediction model is as follows:

[0045] in, for The mass percentage of the element for The mass percentage of the element for The mass percentage of the element for The mass percentage of the element for The mass percentage of the element For the mass percentage of the element, For the mass percentage of the element, For the mass percentage of the element, For the mass percentage of the element, For the mass percentage of the element, For the mass percentage of the element, For the mass percentage of the element, For the mass percentage of the element, For the mass percentage of the element.

[0046] Based on the initial data set of the titanium alloy, the prediction values of all samples were calculated by the base prediction model, and the prediction residual was calculated. The average absolute error of the base prediction model was 12.1℃.

[0047] In the above step four, when generating the correction term population of the base prediction model obtained in step three by symbolic regression calculation based on genetic algorithm, the function set used includes addition, subtraction, multiplication, division and square root operation. After iterative evolution, a total of 12450 mathematical expression individuals were generated in the final correction term population.

[0048] In the above step five, the prediction accuracy (i.e. average absolute error) and individual length of the base prediction model after adding the correction term of each individual obtained in the above step four were calculated. The calculation results show that the average absolute error of all individuals is distributed in the range of 9.24℃ to 18.5℃, and the individual length is distributed between 5 and 30.

[0049] In the above step six, based on the average absolute error and individual length calculated in step five, all individuals were non-dominantly sorted. After screening, 8 Pareto optimal solutions were finally identified to constitute the Pareto front, and the distribution of the Pareto front is as shown in Figure 2 The crowding distance of each individual in the Pareto front was calculated, and an optimal correction term was selected, and the analytical expression of the optimal correction term is = The expression tree structure corresponding to the optimal correction term is as shown in Figure 3

[0050] ​​In step seven, the base prediction model and the optimal correction term are fused by linear addition to obtain the mathematical expression of the final titanium alloy α / β phase transition temperature prediction model as follows:

[0051] wherein, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element, is the mass percentage of the element,

[0052] The average absolute error of the titanium alloy α / β phase transition temperature prediction model is 9.36 ℃, and the prediction accuracy is improved by 10% compared with the base prediction model.

[0053] ​​​​​​​​​​​​​​Step eight, the chemical formula of a titanium alloy with an α / β phase transition temperature to be predicted is Ti-5Al-2Mo-5Sn-2Zr-0.25Si, and the actual measured phase transition temperature is 980℃. The content of the alloying elements of the titanium alloy is input into the titanium alloy α / β phase transition temperature prediction model for calculation, and the α / β phase transition temperature of the titanium alloy to be predicted is 984.0675℃. It can be seen that the predicted value and the actual value of the titanium alloy α / β phase transition temperature in this embodiment are close, and the deviation is 0.42%, so the prediction result of the α / β phase transition temperature is relatively accurate. When the content of the alloying elements of the titanium alloy is input into the titanium alloy α / β phase transition temperature prediction model, the mass percentage of each alloying element in the titanium alloy is used to replace the symbol of each alloying element in the titanium alloy α / β phase transition temperature prediction model, and then calculation is performed. The calculation result is the predicted value of the titanium alloy α / β phase transition temperature.

[0054] In addition, the embodiment of the present application also provides a titanium alloy α / β phase transition temperature prediction system for realizing the titanium alloy α / β phase transition temperature prediction method of the present application as described above, as shown in the figure, the system comprises: Figure 4 An input module is configured to input the content of the alloying elements of a titanium alloy with known composition into a titanium alloy α / β phase transition temperature prediction model that has been constructed; A calculation module is configured to calculate the α / β phase transition temperature of the titanium alloy by using the titanium alloy α / β phase transition temperature prediction model and the content of the alloying elements; An output module is configured to output the α / β phase transition temperature calculated by the titanium alloy α / β phase transition temperature prediction model; The construction process of the titanium alloy α / β phase transition temperature prediction model comprises: A basic prediction model of the α / β phase transition temperature is established by using a linear regression algorithm based on a constructed titanium alloy data set. The titanium alloy data set comprises chemical composition information of a titanium alloy and actual measured values of the α / β phase transition temperature corresponding to the chemical composition information. The chemical composition information comprises the types of alloying elements contained in the titanium alloy and the content of each alloying element; A correction term population of the basic prediction model is generated by symbol regression calculation based on a genetic algorithm; The length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding the correction term are calculated; Based on a non-dominated sorting algorithm, all correction terms in the correction term population are traversed according to the length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding the correction term, and a Pareto front is formed; An optimal correction term is selected from the correction terms on the Pareto front; ​By integrating the basic prediction model and the optimal correction term, the prediction model for the α / β phase transformation temperature of the titanium alloy is obtained.

[0055] The embodiments of the present invention also provide corresponding electronic devices and computer-readable storage media for implementing the solutions provided in the embodiments of the present invention.

[0056] Among them, such as Figure 5 As shown, the electronic device includes a memory and one or more processors. The memory is used to store one or more programs, and the one or more processors are used to execute the one or more programs to cause the electronic device to perform the titanium alloy α / β phase transformation temperature prediction method according to any embodiment of this application.

[0057] The storage medium stores a computer program, which, when executed by a processor, implements the method for predicting the α / β phase transformation temperature of titanium alloys according to any embodiment of this application.

[0058] In summary, the technical solution of this invention breaks through the black box limitation of existing machine learning methods in predicting the α / β phase transformation temperature of titanium alloys, providing a reliable theoretical tool for titanium alloy design and thus significantly improving R&D efficiency.

[0059] Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention.

Claims

1. A method for predicting the α / β phase transformation temperature of titanium alloys, characterized in that, The process includes the following: The content of alloying elements of a titanium alloy with known composition is input into a pre-constructed titanium alloy α / β phase transformation temperature prediction model, and the α / β phase transformation temperature of the titanium alloy is calculated using the titanium alloy α / β phase transformation temperature prediction model. The construction process of the titanium alloy α / β phase transformation temperature prediction model includes: Using a linear regression algorithm, a basic prediction model for the α / β phase transformation temperature is established based on a pre-constructed titanium alloy dataset. The titanium alloy dataset includes the chemical composition information of the titanium alloy and the measured values ​​of the α / β phase transformation temperature corresponding to the chemical composition information. The chemical composition information includes the types of alloying elements contained in the titanium alloy and the content of each alloying element. A population of correction terms for the basic prediction model is generated by symbolic regression calculation based on a genetic algorithm. Calculate the length of each correction term in the correction term population and the prediction accuracy of the base prediction model after adding correction terms; Based on the non-dominated sorting algorithm, according to the length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding correction terms, all correction terms in the correction term population are traversed to form a Pareto front. Select the optimal correction term from the correction terms on the Pareto front; By integrating the basic prediction model and the optimal correction term, the prediction model for the α / β phase transformation temperature of the titanium alloy is obtained.

2. The method for predicting the α / β phase transformation temperature of titanium alloys according to claim 1, characterized in that, The prediction accuracy of the basic prediction model after adding the correction term is the average absolute error between the measured value of the α / β phase transition temperature and the α / β phase transition temperature predicted by the basic prediction model after adding the correction term.

3. The method for predicting the α / β phase transformation temperature of titanium alloys according to claim 1, characterized in that, The process of selecting the optimal correction term from the correction terms on the Pareto front includes: Calculate the crowding distance of all correction terms on the Pareto front, and select the correction term with the largest crowding distance as the optimal correction term.

4. The method for predicting the α / β phase transformation temperature of titanium alloys according to claim 1, characterized in that, The fusion method of the basic prediction model and the optimal correction term is to linearly sum the basic prediction model and the optimal correction term.

5. The method for predicting the α / β phase transformation temperature of titanium alloys according to claim 1, characterized in that, Titanium alloys contain at least one of the following alloying elements: Al, Mo, V, Cr, Fe, Nb, Sn, Zr, Cu, Si, C, H, O, and N.

6. The method for predicting the α / β phase transformation temperature of titanium alloys according to claim 1, characterized in that, The process of constructing the titanium alloy dataset includes: An initial dataset for titanium alloys was constructed using at least one of historical experimental data and literature data. The initial titanium alloy dataset is filled with missing values ​​and outliers are removed to obtain the titanium alloy dataset.

7. The method for predicting the α / β phase transformation temperature of titanium alloys according to claim 6, characterized in that, The specific process of imputing missing values ​​and removing outliers in the initial titanium alloy dataset includes: The missing values ​​in the initial dataset of the titanium alloy were filled with 0, and outlier data were identified and removed using the Laida criterion.

8. A system for predicting the α / β phase transformation temperature of titanium alloys, characterized in that, include: Input module: Used to input the content of alloying elements of titanium alloys with known compositions into the pre-built titanium alloy α / β phase transformation temperature prediction model; Calculation module: used to calculate the α / β phase transformation temperature of the titanium alloy using a titanium alloy α / β phase transformation temperature prediction model and the content of the alloying elements; Output module: Used to output the α / β phase transformation temperature calculated by the titanium alloy α / β phase transformation temperature prediction model; The construction process of the titanium alloy α / β phase transformation temperature prediction model includes: Using a linear regression algorithm, a basic prediction model for the α / β phase transformation temperature is established based on a pre-constructed titanium alloy dataset. The titanium alloy dataset includes the chemical composition information of the titanium alloy and the measured values ​​of the α / β phase transformation temperature corresponding to the chemical composition information. The chemical composition information includes the types of alloying elements contained in the titanium alloy and the content of each alloying element. A population of correction terms for the basic prediction model is generated by symbolic regression calculation based on a genetic algorithm. Calculate the length of each correction term in the correction term population and the prediction accuracy of the base prediction model after adding correction terms; Based on the non-dominated sorting algorithm, according to the length of each correction term in the correction term population and the prediction accuracy of the basic prediction model after adding correction terms, all correction terms in the correction term population are traversed to form a Pareto front. Select the optimal correction term from the correction terms on the Pareto front; By integrating the basic prediction model and the optimal correction term, the prediction model for the α / β phase transformation temperature of the titanium alloy is obtained.

9. An electronic device, characterized in that, include: One or more processors; A memory on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for predicting the α / β phase transformation temperature of titanium alloys as described in any one of claims 1-7.

10. A storage medium, characterized in that, It stores a computer program, wherein the computer program, when executed by a processor, implements the method for predicting the α / β phase transformation temperature of titanium alloys as described in any one of claims 1-7.

Images