Titanium alloy component optimization design method and titanium alloy preparation method

Abstract

The invention relates to the technical field of materials, and provides a titanium alloy component optimization design method and a titanium alloy preparation method. The method comprises the following steps: step a, acquiring component related data and performance related data of a titanium alloy, and constructing a data set based on the two related data; b, training a relation model of a corresponding relation between components and performance of the titanium alloy through the data set and a random forest algorithm to obtain a trained relation model; step c, determining each performance value corresponding to each component in the preset component population of the titanium alloy; and d, setting a multi-performance target of the titanium alloy, determining the priority of each component based on the multi-performance target and each performance value corresponding to each component, carrying out genetic manipulation on the component with the highest priority to generate a next-generation component, updating the component population, and returning to the step c until a first condition is met, so as to obtain an optimal component combination meeting the multi-performance target. According to the scheme, the complicated test process is avoided, and the complexity of determining titanium alloy component design is reduced.

Description

Technical Field

[0001] The present invention relates to the field of material technology, and particularly to a method for optimizing the design of titanium alloy components and a method for preparing titanium alloy. Background Art

[0002] In the traditional method for designing titanium alloy components, the corresponding component combinations that meet different performance requirements are determined through repeated experiments. This method requires a large number of experiments and comprehensive analysis of the experimental results to roughly obtain the optimal combination that meets the corresponding performance requirements. The accuracy of obtaining the optimal combination is not high, the entire process is time-consuming and costly, and the efficiency is low, making it difficult to meet the demands of modern industry for titanium alloy.

[0003] The existing patent document with the application number 202411199233.9 discloses a method for designing titanium alloy components with high comprehensive performance, including: constructing an initial data set; gradually screening the data through Pearson correlation screening, recursive elimination, and feature weight ranking to determine the key material descriptor combinations that affect the performance of titanium alloy; then constructing a data set containing alloy components, key material descriptors, and alloy performance according to the above screening results; constructing a prediction model, establishing an initial XGBoost model for key features using the training set, and using the Grid Search and K-Fold Cross Validation algorithms to double-optimize the hyperparameters of the XGBoost initial model to obtain an improved XGBoost model; using the test set to evaluate the prediction accuracy of the improved XGBoost model and perform optimization design to obtain the best XGBoost prediction model; and also using the multi-objective optimization strategy of the non-dominated sorting genetic algorithm to co-optimize multiple alloy performances to obtain titanium alloy components with optimal comprehensive performance, and conducting experimental verification and iterative optimization. It can be seen from the above that although this method avoids the steps of conducting a large number of experiments and comprehensive analysis of the experimental results, the entire process involves multiple data preparations and a large amount of data calculation processes, increasing the complexity of titanium alloy component design from a computational perspective.

[0004] In summary, there is an urgent need for a method for designing titanium alloy components that can avoid both the steps of conducting a large number of experiments and comprehensive analysis of the experimental results and the processes of multiple data preparations and data calculations. Summary of the Invention

[0005] Aiming at the problem that although the prior art avoids the steps of conducting a large number of experiments and comprehensive analysis of the experimental results, the entire process involves multiple data preparations and a large amount of data calculation processes, increasing the complexity of titanium alloy component design from a computational perspective, the present disclosure provides a method for optimizing the design of titanium alloy components and a method for preparing titanium alloy.

[0006] According to the first aspect of the present invention, a method for optimizing the design of a titanium alloy composition is provided, including: Step a, obtaining composition-related data and performance-related data of the titanium alloy, and constructing a data set based on the two related ones; Step b, training a relationship model of the correspondence between the composition and performance of the titanium alloy through the data set and the random forest algorithm until relevant parameters of the relationship model are determined, obtaining a trained relationship model; Step c, determining performance values corresponding to each composition in the preset composition population of the titanium alloy through the trained relationship model; Step d, setting multi-performance targets for the titanium alloy, determining the priority of each composition based on the multi-performance targets and the performance values corresponding to each composition, performing genetic operations on the composition with the highest priority to generate the next-generation composition and updating the composition population, and returning to step c until a first condition is reached, obtaining an optimal composition combination that meets the multi-performance targets.

[0007] In some embodiments, the multi-performance targets of the titanium alloy include: the tensile strength of the titanium alloy is greater than or equal to 900 MPa and the elongation is greater than or equal to 20%.

[0008] In some embodiments, step d includes: Comparing the contribution sizes of the performance values corresponding to each composition under the multi-performance targets through the non-dominated sorting algorithm, calculating the crowding degree of each composition, and determining the priority of each composition according to the principle that the lower the crowding degree under the same domination condition, the higher the priority.

[0009] In some embodiments, step d further includes: Performing crossover operations on the composition with the highest priority, and its formula is as follows,

[0010] wherein, the x p and the x q are two parent components, the β is a crossover factor generated based on a random number, and the X new (k) is the pre-mutation next-generation component generated; Performing mutation operations on the pre-mutation next-generation component, and its formula is as follows,

[0011] wherein, the δ is a small perturbation for polynomial mutation, and the is the next-generation component generated.

[0012] In some embodiments, step d further includes: Until the maximum number of iterations is reached, update the component population with the next generation of components obtained in the final round of iteration and determine the optimal component combination that meets the multi-performance objectives based on it.

[0013] In some embodiments, the preset component population is generated by a random algorithm, and the component population includes a number of component vectors X i , and its expression is as follows:

[0014] Wherein, the component vector x in represents the nth element in the component vector X i , and the sum of the corresponding contents from the x i1 to x in is 1.

[0015] In some embodiments, the step b includes: Divide the data set into a training set and a test set in a ratio of 2:1, train the relationship model between the components and properties of the titanium alloy through the training set and the random forest algorithm, optimize the relationship model through cross-validation and grid search methods, determine the relevant parameters of the relationship model, and verify the accuracy of the relationship model through the test set. If the verification passes, obtain the trained relationship model.

[0016] In some embodiments, the step a includes: obtaining the component-related data and performance-related data of the titanium alloy, performing missing value processing, data format unification, and vectorization processing on the associated component-related data and performance-related data, and constructing a data set based on the processed data.

[0017] In some embodiments, the component-related data of the titanium alloy includes the types and contents of alloying elements, the weights and contents of impurity elements. The alloying elements include aluminum, vanadium, molybdenum, and tin, and the impurity elements include iron, oxygen, nitrogen, and carbon.

[0018] In some embodiments, the relevant parameters of the relationship model include: the number of decision trees, the maximum depth, the minimum number of samples for splitting, the minimum number of samples for leaf nodes, and the maximum number of features.

[0019] According to the second aspect of the present invention, there is also provided a method for preparing a titanium alloy, including: preparing various raw materials according to the method for optimizing the design of the titanium alloy components described in any one of the above; preprocessing the various raw materials and then putting them into a melting furnace, controlling the relevant parameters of the melting process, and then casting and forming to obtain the titanium alloy.

[0020] The above method for optimizing the design of a titanium alloy composition and the method for preparing a titanium alloy obtain data related to the composition and data related to the performance of the titanium alloy, construct a data set based on the two related ones, and train a relationship model between the composition and performance of the titanium alloy through the data set and the random forest algorithm until the relevant parameters of the relationship model are determined, obtaining a trained relationship model. Then, through the trained relationship model, the performance values corresponding to each composition in the preset composition population of the titanium alloy are determined, the multi-performance objectives of the titanium alloy are set, and the priority of each composition is determined based on this. The composition with the highest priority is subjected to genetic operations to generate the next-generation composition and update the composition population. Then, continue to determine the performance values corresponding to each composition in the preset composition population of the titanium alloy through the trained relationship model until the first condition is met, obtaining the optimal composition combination that meets the multi-performance objectives. The method of the present invention does not require a large number of experiments and comprehensive analysis of experimental data, avoiding the complicated experimental process. At the same time, it also does not require multiple data preparations and a large amount of data calculation processes, reducing the complexity of determining the titanium alloy composition design by algorithms and improving the efficiency and accuracy of determining the titanium alloy composition design.

[0021] Meanwhile, the method for preparing a titanium alloy of the present application can also achieve the above technical effects, which will not be elaborated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other embodiments can be obtained based on these drawings.

[0023] Figure 1 It is a flowchart of a method for optimizing the design of a titanium alloy composition provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] The following will further describe in detail the embodiments of the present disclosure in conjunction with the drawings. The detailed description and drawings of the following embodiments are used to exemplarily illustrate the principles of the present disclosure, but cannot be used to limit the scope of the present disclosure. The present disclosure can be implemented in many different forms, not limited to the specific embodiments disclosed herein, but including all technical solutions falling within the scope of the claims.

[0025] The present disclosure provides these embodiments to make the present disclosure thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that: unless otherwise specifically stated, the relative arrangements of components and steps, the compositions of materials, numerical expressions, and numerical values set forth in these embodiments should be construed as merely exemplary, rather than as limitations.

[0026] In addition, the terms "first", "second", and similar terms used in the present disclosure do not denote any order, quantity, or importance, but are merely used to distinguish different parts. Terms such as "comprising" or "including" mean that the elements before the term cover the elements listed after the term, and do not exclude the possibility of also covering other elements.

[0027] All terms used in the present disclosure have the same meaning as understood by those of ordinary skill in the art to which the present disclosure pertains, unless otherwise specifically defined. It should also be understood that terms defined in a general dictionary, such as those, should be construed to have a meaning consistent with their meaning in the context of the relevant art, and should not be interpreted in an idealized or overly formal sense, unless specifically defined as such herein.

[0028] Techniques, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the techniques, methods, and devices should be regarded as part of the specification.

[0029] It should be understood that the embodiments of the present invention shown in the exemplary embodiments are merely illustrative. Although only a few embodiments of the present invention have been described in detail, those skilled in the art can easily appreciate that various modifications are feasible without substantially departing from the teachings of the subject matter of the present invention. Accordingly, all such modifications should be included within the scope of the present invention. Without departing from the gist of the present invention, other substitutions, modifications, variations, and deletions can be made to the design, operating conditions, parameters, etc. of the following exemplary embodiments.

[0030] Please refer to Figure 1 , Figure 1 which shows a flowchart of a method for optimizing the design of a titanium alloy composition and a method for preparing a titanium alloy provided by an embodiment of the present invention. A method for optimizing the design of a titanium alloy composition and a method for preparing a titanium alloy 100 includes: Step a, obtaining composition-related data and performance-related data of the titanium alloy, and constructing a data set based on the two that are associated with each other; Specifically, the titanium alloys include TC4, TC11, etc. The composition-related data of the titanium alloys include the types and contents of alloying elements, the weights and contents of impurity elements. The alloying elements include aluminum, vanadium, molybdenum, tin, etc., and the impurity elements include iron, oxygen, nitrogen, carbon, etc. The performance-related data of the titanium alloys include mechanical property-related data, chemical property-related data, physical property-related data, etc. Among them, the mechanical properties include strength, plasticity, toughness, etc. For titanium alloys, they are composed of multiple components, and at the same time, they have multiple properties themselves. Therefore, there is a corresponding relationship between the composition-related data and the performance-related data of different titanium alloys, and multiple composition-related data correspond to multiple performance-related data. Before constructing the data set, it is necessary to preprocess the composition-related data and the performance-related data, and sequentially complete the missing value processing, data format unification, and vectorization processing of the data. For example, the data is standardized through Python, and then the data set is constructed based on the processed data.

[0031] Step b: Train the relationship model of the corresponding relationship between the composition and performance of the titanium alloy through the data set and the random forest algorithm until the relevant parameters of the relationship model are determined, and the trained relationship model is obtained; Specifically, the random forest algorithm is an ensemble learning algorithm based on decision trees, which can accurately classify samples. The representation of the random forest algorithm is the content of the prior art and will not be elaborated here. The data set is divided into a training set and a test set. Preferably, the data set is divided into a training set and a test set according to a ratio of 2:1. The relationship model of the corresponding relationship between the composition and performance of the titanium alloy is trained through the training set and the random forest algorithm, and the relevant parameters of the relationship model are optimized through the optimization algorithm until the optimal effect is achieved, and the final relevant parameters of the relationship model are determined. Among them, the optimization algorithms include cross-validation algorithms and grid search algorithms, etc. The final relevant parameters of the determined relationship model include the number of decision trees (n_estimators), the maximum depth (max_depth), the minimum number of samples for splitting (min_samples_split), the minimum number of samples for leaf nodes (min_samples_leaf), the maximum number of features (max_features), etc. In a specific embodiment, the relationship model of the corresponding relationship between the composition and performance of the titanium alloy is trained through the Sklearn package in the Python environment, and the trained relationship model is obtained. For example, the composition-performance (tensile strength and elongation of TC4 titanium alloy) relationship model is obtained.

[0032] Step c: Determine the performance values corresponding to each component in the preset component population of the titanium alloy through the trained relationship model; Specifically, the preset component population of the titanium alloy is obtained through the existing component generation model. For example, several component vectors X are randomly generated through a random algorithm. i, and its expression is as follows: (Formula 1) Among them, the component vector x in represents the nth element in the component vector X i , and the sum of the corresponding contents from x i1 to x in is 1. For example, the content of the jth element in the component vector X i is represented as x ij , which satisfies .

[0033] Through the trained relationship model and the above-mentioned preset component population, the performance values corresponding to each component in the preset component population are determined.

[0034] Step d: Set multiple performance targets for the titanium alloy. Based on these and the performance values corresponding to each component, determine the priority of each component. Perform genetic operations on the component with the highest priority to generate the next generation of components and update the component population, and then return to step c until the first condition is met to obtain the optimal component combination that meets the multiple performance targets.

[0035] Specifically, the multiple performance targets of the titanium alloy refer to the targets of multiple performances. For example, the tensile strength of the titanium alloy is greater than or equal to 900 MPa and the elongation is greater than or equal to 20%, or the tensile strength of the titanium alloy is greater than or equal to 800 MPa and the elongation is greater than or equal to 20%, etc. Specifically, the multiple performance targets of the titanium alloy can be flexibly adjusted according to actual needs and are not limited to the multiple performance targets listed above.

[0036] Specifically, according to the non-dominated sorting algorithm (NSGA-II), sort the component vectors. By comparing the performance values corresponding to each component vector, determine which components perform better in the process of optimizing the multiple performance targets. For any two component vectors x i and x j , if the following formula 2 is satisfied, (Formula 2) where k, m ∈ {1, 2, 3}k, then x i dominates x j, under the condition of the same domination degree, the components are further sorted by calculating the crowding degree. The crowding degree reflects the density of the component points in the target space. The smaller the density of the component points, the lower the crowding degree, and the higher the priority. According to the principle of determining the priority level, the priority corresponding to each component is obtained. The component with the highest priority is subjected to genetic operations through the genetic algorithm, and the crossover operation (for example, generating a new offspring component xnew(k) through the simulated binary crossover (SBX) operator) and the mutation operation (for example, introducing a small perturbation δ for polynomial mutation on the basis of the offspring component) are carried out in sequence to generate the next generation of components and update the component population. Then, based on the updated component population, the respective performance values corresponding to each component in the updated component population are determined through the trained relationship model. The iteration update is carried out according to the above steps until the first condition is reached, such as reaching the maximum number of iterations, and the component population is updated with the next generation of components obtained in the final round of iteration, and the optimal component combination meeting the multi-performance objectives is determined based on it.

[0037] The above method for optimizing the design of titanium alloy components does not require a large number of experiments and comprehensive analysis of experimental data, avoiding the complicated experimental process. At the same time, it also does not require multiple data preparations and a large amount of data calculation processes, reducing the complexity of determining the titanium alloy component design through algorithms and improving the efficiency and accuracy of determining the titanium alloy component design.

[0038] According to several embodiments of the present invention, step d includes: comparing the contribution sizes of the respective performance values corresponding to the respective components under the multi-performance objectives through the non-dominated sorting algorithm, calculating the crowding degree of the respective components, and determining the priority of each component according to the principle that the lower the crowding degree under the same domination condition, the higher the priority.

[0039] According to several embodiments of the present invention, step d further includes: performing a crossover operation on the component with the highest priority, and its formula is as follows, (Formula 3) where x p and x q are two parent components, β is a crossover factor generated based on a random number, and X new (k) is the next generation of components before mutation; Perform a mutation operation on the next generation of components before mutation, and its formula is as follows, (Formula 4) where δ is a small perturbation for polynomial mutation, is the generated next generation of components.

[0040] For a further understanding of the method for optimizing the design of titanium alloy components of the present invention, the following is further elaborated in detail in a specific embodiment.

[0041] Step 1: Collect data related to the composition and properties of titanium alloys through laboratory, factory, and paper data, etc.

[0042] Step 2: Use Python to standardize the above-mentioned data related to composition and properties, and successively complete missing value processing, data format unification, and vectorization processing. Use the pandas library in Python to first comprehensively check the data, identify the fields or records where missing values are located, and according to the characteristics of the data, through appropriate missing value processing strategies, delete missing values, fill missing values, and mark missing values. Use Python to perform data format unification processing, such as data type conversion, date and time format processing, and text data processing. Finally, perform vectorization processing on all features, and construct a data set based on the processed data.

[0043] Step 3: Divide the data set into a training set and a test set according to a ratio of 2:1, and train a high-precision random forest machine learning model through the training set to establish a relationship model between composition and properties (tensile strength, yield strength, and elongation of TC4 titanium alloy). To ensure the rationality of the model, use the methods of cross-validation and grid search to determine the relevant parameters of the random forest model. The relevant parameters include the number of decision trees (n_estimators), maximum depth (max_depth), minimum number of samples for splitting (min_samples_split), minimum number of samples for leaf nodes (min_samples_leaf), maximum number of features (max_features), etc. The above-mentioned training process is completed by the Sklearn package based on the Python environment to obtain a relationship model between composition and properties (tensile strength and elongation of TC4 titanium alloy) for subsequent calls by the non-dominated sorting genetic algorithm.

[0044] Step 4: Set multiple performance objectives, such as setting the tensile strength of the titanium alloy to be greater than or equal to 900 MPa and the elongation to be greater than or equal to 20%. Use a random algorithm to generate an initial composition population, such as randomly generating 400 composition vectors X i , the expression of which is shown in Formula 1 and will not be elaborated here.

[0045] Step 5: Sort the composition vectors according to the non-dominated sorting algorithm (NSGA-II). Non-dominated sorting determines which components perform superiorly in the multi-objective optimization process by comparing the respective performance values corresponding to each composition vector. For any two composition vectors x i and x j, if formula 2 is satisfied, under the condition of the same dominance, the components are further sorted by calculating the crowding degree. The crowding degree reflects the density of the component points in the target space. The smaller the density of the component points, the lower the crowding degree and the higher the priority. Specifically, the component points are first projected into two-dimensional space, and each component point is triangulated using the Delaunay triangulation method to calculate the area of ​​the triangle that wraps the point. The larger the area, the lower the density around the point, and thus the smaller the crowding degree and the higher the priority. Delaunay triangulation is an algorithm in computational geometry that is used to divide a set of points on a plane into a series of non-overlapping triangles so that the set of these triangles can completely cover the convex hull of the original point set.

[0046] Step 6, determine the priority of each component according to the sorting results, select the component with the best performance, that is, the component with the highest ranking in the non-dominated sorting and the crowded sorting, and perform genetic operations on them, including crossover and mutation operations. The crossover operation formula is as shown in Formula 3, and the mutation operation is as shown in Formula 4, which will not be repeated here.

[0047] Step 7, repeat steps 6 and 7, after generating each new generation of components, re-predict and sort them by relational model and non-dominated sorting algorithm, and then generate the next generation of components by genetic operation until the preset iteration step number g is reached. max .

[0048] Step 8, in the last generation, select the components that meet the performance requirements. These components perform well in the multi-objective optimization and meet the pre-set multi-performance objectives.

[0049] According to a second aspect of the present invention, there is also provided a method for preparing a titanium alloy, comprising: preparing various raw materials according to any of the above-mentioned methods for optimizing the design of titanium alloy composition; placing the various raw materials into a smelting furnace after pretreatment, controlling relevant parameters of the smelting process, and then casting to obtain the titanium alloy.

[0050] So far, various embodiments of the present disclosure have been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed here.

[0051] Although some specific embodiments of the present disclosure have been described in detail by way of examples, those skilled in the art should understand that the above examples are only for illustration and not for limiting the scope of the present disclosure. Those skilled in the art should understand that the above embodiments can be modified or partial technical features can be equivalently replaced without departing from the scope and spirit of the present disclosure. In particular, as long as there is no structural conflict, the various technical features mentioned in each embodiment can be combined in any way.

Claims

1. A method for optimizing the design of titanium alloy composition, characterized in that: include: Step a, obtaining composition-related data and performance-related data of titanium alloy, and constructing a data set based on the associated data; Step b, training the relationship model of the corresponding relationship between the composition and performance of the titanium alloy by using the data set and the random forest algorithm until the relevant parameters of the relationship model are determined to obtain a trained relationship model; Step c, determining the performance values ​​corresponding to each component in the preset component population of the titanium alloy through the trained relationship model; Step d: set multiple performance targets for the titanium alloy, determine the priority of each component based on the performance values ​​corresponding to the components, perform genetic operations on the components with the highest priority to generate the next generation of components and update the component population, and return to step c until the first condition is met to obtain the optimal component combination that meets the multiple performance targets.

2. The method for optimizing the composition of titanium alloy according to claim 1, characterized in that: The multiple performance targets of the titanium alloy include: the tensile strength of the titanium alloy is greater than or equal to 900 MPa and the elongation is greater than or equal to 20%.

3. The method for optimizing the composition of titanium alloy according to claim 2, characterized in that: The step d comprises: The contribution of each performance value corresponding to each component under the multiple performance objectives is compared through a non-dominated sorting algorithm, and the congestion of each component is calculated. The priority of each component is determined according to the principle that the lower the congestion, the higher the priority under the same dominating condition.

4. The method for optimizing the composition of titanium alloy according to claim 3, characterized in that: The step d also includes: The highest priority components are cross-operated, and the formula is as follows: Among them, the x p and the x q are two parent components, β is a crossover factor generated based on random numbers, and X new (k) The next generation component before the generated mutation; The mutation operation is performed on the next generation components before the mutation, and the formula is as follows: Wherein, δ is a small perturbation for polynomial mutation, To generate the next generation components.

5. The method for optimizing the composition of titanium alloy according to claim 4, characterized in that: The step d also includes: Until the maximum number of iterations is reached, the component population is updated with the next generation components obtained in the final round of iterations and based on the next generation components, the optimal component combination that meets the multiple performance objectives is determined.

6. The method for optimizing the composition of titanium alloy according to claim 1, characterized in that: The preset component population is generated by a random algorithm, and the component population includes several component vectors X i , which is expressed as follows: Among them, the component vector x in Represents the component vector X i The nth element in the i1 to x in The sum of the corresponding contents is 1.

7. The method for optimizing the composition of titanium alloy according to claim 1, characterized in that: The step b comprises: The data set is divided into a training set and a test set in a ratio of 2:

1. The relationship model of the correspondence between the composition and performance of the titanium alloy is trained by the training set and the random forest algorithm, and the relationship model is optimized by cross-validation and grid search methods to determine the relevant parameters of the relationship model. The accuracy of the relationship model is verified by the test set. If the verification passes, a trained relationship model is obtained.

8. The method for optimizing the composition of titanium alloy according to claim 1, characterized in that: The step a comprises: Obtain composition-related data and performance-related data of titanium alloy, process missing values, unify data formats, and vectorize the associated composition-related data and performance-related data, and build a data set based on the processed data.

9. The method for optimizing the composition of titanium alloy according to claim 1, characterized in that: The relevant parameters of the relationship model include: the number of decision trees, the maximum depth, the minimum number of sample splits, the minimum number of sample leaf nodes, and the maximum number of features.

10. A method for preparing a titanium alloy, characterized in that: include: Prepare various raw materials according to the method for optimizing the composition of titanium alloy according to any one of claims 1 to 9; The various raw materials are pretreated and placed in a smelting furnace, and the relevant parameters of the smelting process are controlled, and then cast to obtain a titanium alloy.

Images