CN110010210B - Multi-component alloy component design method based on machine learning and oriented to performance requirements - Google Patents

Multi-component alloy component design method based on machine learning and oriented to performance requirements Download PDF

Info

Publication number
CN110010210B
CN110010210B CN201910252935.1A CN201910252935A CN110010210B CN 110010210 B CN110010210 B CN 110010210B CN 201910252935 A CN201910252935 A CN 201910252935A CN 110010210 B CN110010210 B CN 110010210B
Authority
CN
China
Prior art keywords
performance
alloy
data
neural network
component
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910252935.1A
Other languages
Chinese (zh)
Other versions
CN110010210A (en
Inventor
付华栋
谢建新
王长胜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Science and Technology Beijing USTB
Original Assignee
University of Science and Technology Beijing USTB
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Science and Technology Beijing USTB filed Critical University of Science and Technology Beijing USTB
Priority to CN201910252935.1A priority Critical patent/CN110010210B/en
Publication of CN110010210A publication Critical patent/CN110010210A/en
Application granted granted Critical
Publication of CN110010210B publication Critical patent/CN110010210B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a multi-component alloy component design method based on machine learning and oriented to performance requirements, which relates to the technical field of metal material design, and can realize the purpose of quickly and accurately designing alloy components according to the performance requirements by mining a large amount of existing data about the alloy components and the performance and unlocking an implicit complex relation between the components and the performance by adopting a machine learning technology; the method comprises the following steps: s1, establishing a data set according to the historical data; s2, establishing and training a C2P model and a P2C model; s3, inputting the target performance as input data into P2C to obtain an initial design component; s4, inputting the initial design components as input data to C2P to obtain the predicted performance; and S5, judging whether the error of the predicted performance relative to the target performance is in an acceptable range, if not, reestablishing the model, and if so, completing the design. The technical scheme provided by the invention is suitable for the process of designing the alloy components.

Description

Multi-component alloy component design method based on machine learning and oriented to performance requirements
[ technical field ] A method for producing a semiconductor device
The invention relates to the technical field of metal material design, in particular to a method for designing components of a multi-component alloy based on machine learning and oriented to performance requirements.
[ background of the invention ]
With the rapid development of economy and society, the high-performance requirements of key metal materials such as copper alloy, aluminum alloy, magnesium alloy, titanium alloy, high-temperature alloy and the like widely applied to the high-end manufacturing fields such as aerospace, transportation, energy equipment, electronic information and the like are more and more prominent, and a new alloy component design method which can aim at actual requirements and set performance requirements is particularly expected to be developed, so that the research and development efficiency and the engineering application efficiency of the new material are remarkably improved.
The design of metal alloy components so far mainly depends on experimental trial and error methods, a model capable of quantitatively describing the relationship between the alloy components and the performance is lacked, and the alloy components are difficult to design quickly and accurately according to given performance requirements, so that the research and development efficiency of new materials is low, the period is long, and the cost is high. The development of a method for accurately and rapidly designing alloy components according to given performance requirements becomes a technical problem which needs to be broken through urgently in the field of metal material design.
Machine learning, as a typical data mining analysis method, is an effective means for discovering rules and establishing models from existing data. Raccuglia and other application machine learning algorithms of Harvard university in 2016 analyze waste experimental data, so that the formation condition of the inorganic matter synthesized by the organic template can be efficiently predicted, the success rate reaches 89%, and the efficiency is remarkably improved compared with that of the traditional synthetic scheme. Lookman et al in the Alamoss national laboratory thinks that even if a high-throughput calculation and high-throughput experiment method is adopted, the optimization can be performed only in a narrow range (local optimization), and a widely-applied machine learning method can realize global optimization on the basis of a certain sample amount, search for a new material with given performance requirements, establish a relation model between alloy components and thermal hysteresis temperature by using a support vector machine, and search for the alloy components with the optimal thermal hysteresis temperature. However, both the studies conducted to date and the existing methods have made it difficult to achieve the design of alloy compositions to a given target property (such design is often referred to as reverse design).
Accordingly, there is a need for a new data-driven, performance-oriented alloy design approach that addresses or alleviates one or more of the problems set forth above.
[ summary of the invention ]
In view of the above, the invention provides a multi-component alloy composition design method based on machine learning and oriented to performance requirements, which achieves the purpose of quickly and accurately designing alloy compositions according to the performance requirements by mining a large amount of existing data about alloy compositions and performance and unlocking an implicit complex relationship between composition and performance by adopting a machine learning technology.
In one aspect, the invention provides a method for designing a multi-component alloy composition based on machine learning and oriented to performance requirements, which is characterized by comprising the following steps:
s1, establishing a data set according to the historical data;
s2, establishing an alloy component → performance BP neural network model according to the data set, and training the model;
s3, establishing an alloy performance → component BP neural network model according to the data set, and training the model;
s4, inputting the target performance as input data into the trained alloy performance → component BP neural network model to obtain an initial design component;
s5, inputting the initial design components as input data into the trained alloy component → performance BP neural network model to obtain the predicted performance;
s6, calculating the error of the predicted performance relative to the target performance;
s7, judging whether the error is within a preset error range; if yes, finishing the alloy design; if not, the alloy properties → component BP neural network model is re-established and re-trained, and S4-S7 are repeated.
As for the above-mentioned aspect and any possible implementation manner, there is further provided an implementation manner, where the specific step of S1 includes:
s11, collecting historical data of the alloy to establish a basic data set, wherein the historical data comprises component data and performance data;
s12, classifying the data in the basic data set, and selecting the data of the type corresponding to the data as an initial data set according to the type of the alloy to be designed;
and S13, carrying out normalization processing on the initial data set to obtain the data set.
The above aspects and any possible implementations further provide an implementation, and the performance data in S11 includes tensile strength, elongation after tensile break, electrical conductivity, compressive yield strength, and hardness, and one or more combinations thereof.
The above aspect and any possible implementation further provide an implementation, and the data classification in S12 is performed by one or more of machining process, strengthening method, and alloy type.
The above aspects and any possible implementations further provide an implementation, where the alloy includes an aluminum alloy, a copper alloy, a magnesium alloy, a titanium alloy, a superalloy, and a high entropy alloy.
As for the above-mentioned aspect and any possible implementation manner, there is further provided an implementation manner, where the specific step of S2 includes:
s21, taking the component data in the data set as input and the performance data as output, setting model parameters of a BP neural network, and establishing an alloy component → performance BP neural network model;
s22, selecting a transfer function and a training method to train the alloy composition → performance BP neural network model, so that the precision of the alloy composition → performance BP neural network model reaches the preset precision.
The above-described aspects and any possible implementations further provide an implementation where the alloy composition → performance BP neural network model has a pre-set precision of no less than 0.90.
As for the above-mentioned aspect and any possible implementation manner, there is further provided an implementation manner, where the specific step of S3 includes:
s31, taking the performance data and component data in the data set as input and component data as output, setting model parameters of a BP neural network, and establishing an alloy performance → component BP neural network model;
s32, selecting a transfer function and a training method to train the alloy performance → component BP neural network model, so that the precision of the alloy performance → component BP neural network model reaches the preset precision.
The above aspect and any possible implementation further provide an implementation that the alloy property → predetermined precision of the component BP neural network model is not less than 0.85.
The above-mentioned aspect and any possible implementation manner further provide an implementation manner, and the method for calculating the error in S6 is: (target performance-predicted performance)/target performance; the preset error range in the S7 is 5% -15%.
Compared with the prior art, the invention can obtain the following technical effects: the rapid and accurate design of the novel high-performance multi-element alloy based on data driving and according to given performance requirements can be realized, the research and development efficiency of new materials is obviously improved, and the requirements of economic and social development on the new materials are met; the method can realize the efficient utilization of data such as scientific research, production and the like, further excavate and utilize the value of the accumulated data, and promote the innovation capability and level of the material science and technology field.
Of course, it is not necessary for any one product in which the invention is practiced to achieve all of the above-described technical effects simultaneously.
[ description of the drawings ]
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a flow chart of a method for designing a multi-component alloy composition based on machine learning and oriented to performance requirements according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an alloy composition design system based on machine learning and oriented to performance requirements according to an embodiment of the present invention.
[ detailed description ] embodiments
For better understanding of the technical solutions of the present invention, the following detailed descriptions of the embodiments of the present invention are provided with reference to the accompanying drawings.
It should be understood that the described embodiments are only some embodiments of the invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The terminology used in the embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the examples of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
A multi-component alloy component design method based on machine learning and oriented to performance requirements is characterized in that a large amount of existing data about alloy components and performance are mined, an implicit complex relation between the components and the performance is unlocked by adopting a machine learning technology, the alloy components are quickly and accurately designed according to the performance requirements, and the research, development and application of new materials are accelerated. The method comprises the following specific steps:
(1) data collection: acquiring composition data and performance data of the alloy by collecting and sorting publicly published literature data, experimental tests or enterprise production data, and establishing a basic data set;
alloys include, but are not limited to, the following alloys: aluminum alloys, copper alloys, magnesium alloys, titanium alloys, high temperature alloys, and high entropy alloys;
alloy property data include, but are not limited to: tensile strength, elongation after tensile failure, electrical conductivity, compressive yield strength and hardness, and one or a combination of more of the above properties;
alloy composition data: the amount of other microalloying elements contained in addition to the main alloying element is one or more;
(2) data cleaning: classifying the collected alloy data according to a certain principle, and selecting data of a type corresponding to the collected alloy data as an initial data set according to the type of the designed alloy;
the alloy classification principle is at least one of a processing technology, a strengthening mode and an alloy type;
(3) normalization treatment: respectively carrying out normalization processing (0-1) on the alloy composition data and the performance data in the initial data set;
(4) establishing an alloy component → performance BP neural network model: taking the component data subjected to normalization processing in the step (3) as input, taking the performance data as output, setting the number of initial hidden layers and the number of nodes of the initial hidden layers of the neural network, and establishing a neural network model (namely an alloy component → performance BP neural network model, which is abbreviated as a C2P model) for predicting the performance of the alloy by components;
before the BP neural network model is trained, the normalized data set can be randomly divided into a training set (proportion is 40% -90%), a verification set (proportion is 5% -30%) and a test set (proportion is 5% -30%), and the sum of the proportions of the training set, the verification set and the test set is always 1; the training set is used for training the neural network model, the verification set is used for preventing overfitting of the neural network model, and the training set and the verification set are used during training of the BP neural network model; the test set is used for testing the precision of the neural network model and is used after the model training is finished;
the training set is used for training the neural network model; the verification set is used for controlling the training process, namely an early stopping strategy is adopted, and the training is stopped after the error of the verification set continuously decreases for a set number of times, mainly used for preventing overfitting; the early stopping strategy is suitable for the condition that the data in the field of the current monitoring materials are few, not only can overfitting be effectively prevented, but also the training time can be saved.
(5) Training the C2P model: selecting a transfer function and a training method and training a C2P model; the comprehensive accuracy (training, verifying and testing) characteristic of the C2P model is not less than 0.90;
the precision is that the data set is used as input, a predicted value is obtained through a trained model, and the linear regression correlation between the predicted value and an actual value is the precision of the model;
the training method mainly includes a steepest descent method, an additional momentum method, a self-adaptive learning rate method, a momentum-self-adaptive learning rate adjustment method, a quantization conjugate gradient algorithm and the like, and the steepest descent algorithm is selected in the following embodiment, and other algorithms can be selected.
(6) Establishing an alloy performance → component BP neural network model: taking the performance data subjected to normalization processing in the step (3) as input, taking component data as output, setting the number of initial hidden layers and the number of nodes of the initial hidden layers of the neural network, and establishing a neural network model (abbreviated as P2C model) with alloy performance design components;
(7) training the P2C model: selecting a transfer function and a training method and training a P2C model;
the comprehensive accuracy (training, verification and testing) characteristic of the P2C model is not less than 0.85;
the precision and training method are the same as in (5).
(8) Inputting the given target performance into the P2C model trained in the step (7) to obtain an initial design component;
(9) taking the initial design components in the step (8) as the input of the C2P model, and outputting the predicted performance;
(10) comparing the predicted performance obtained in the step (9) with a given target performance, and calculating an error e which is (target performance-predicted performance)/target performance, wherein generally, e is selected to be 5-15%;
(11) if the error e is not larger than the set error, the predicted composition output in the step (8) is a reasonable value, otherwise, the P2C neural network parameters in the step (6) are reset, the P2C model in the step (7) is retrained, and then the steps (8) to (11) are repeated until alloy compositions which meet the error requirement and contain or do not contain a certain specific element are screened out;
the resetting is mainly to reset the structural parameters (such as the number of hidden layers, the number of nodes, a transfer function and the like) and the training algorithm of the BP neural network model. The data set used in the retraining is re-randomly distributed by the system, and the overall data content is unchanged and is consistent with the previous data.
(12) Performing inverse normalization treatment on the alloy components determined in the step (11) by using the same parameters as those in the step (3) to obtain the final design components of the alloy;
the purpose of normalization and reverse normalization is to perform scale transformation on the original data without changing the original data distribution rule so as to improve the generalization capability of the model; the calculation method is a commonly used algorithm in mathematics, and is not described herein again.
Example 1: by applying the method provided by the invention, five high-performance copper alloy components are designed under the condition that the performance error is not more than 10% by taking tensile strengths of 600MPa, 625MPa, 650MPa, 700MPa and 750MPa and conductivity of 50% IACS as performance targets, and the design method is specifically described as follows:
(1) data collection: establishing a basic data set by collecting and arranging publicly published components, strengthening modes, tensile strength and corresponding electric conductivity of Cu-Fe-P series, Cu-Ni-Si series and Cu-Cr-Zr series alloys;
(2) data cleaning: dividing the collected three copper alloy data into two types according to a strengthening mode, namely a deformation-precipitation strengthening type and a non-deformation-precipitation strengthening type, and discarding non-deformation-precipitation strengthening type data, such as an alloy initial data set obtained by processing methods such as pack rolling, equal-diameter angular extrusion, high-pressure knob and the like;
(3) normalization treatment: respectively carrying out normalization processing (0-1) on the alloy composition data and the performance data in the initial data set;
(4) establishing an alloy component → performance BP neural network model: taking the component data subjected to normalization processing in the step (3) as input, taking tensile strength and conductivity data as output, setting the number of initial hidden layers of the neural network to be 2, and the number of nodes of the initial hidden layers to be 10 and 16 respectively, and establishing a neural network model (abbreviated as C2P model) for predicting the alloy performance by components;
(5) training the C2P model: selecting Simoid as a hidden layer transfer function and a linear transfer function as an output layer transfer function, and training a C2P model by adopting a steepest descent method;
(6) establishing an alloy performance → component BP neural network model: taking the data of the tensile strength and the electric conductivity after the normalization processing in the step (3) as input, taking the component data as output, setting the number of initial hidden layers of the neural network to be 2, and the number of nodes of the initial hidden layers to be 12 and 16 respectively, and establishing a neural network model (abbreviated as P2C model) with the components designed by the alloy performance;
(7) training the P2C model: selecting Simoid as a hidden layer transfer function and a linear transfer function as an output layer transfer function, and training a P2C model by adopting a steepest descent method;
(8) inputting the tensile strengths of 600MPa, 625MPa, 650MPa, 700MPa and 750MPa, and the electric conductivities of 5 tensile strengths of 50% IACS as target performances (given performances) into the P2C model trained in the step (7) to obtain initial design components;
(9) taking the initial design components in the step (8) as the input of the C2P model, and outputting the predicted performance;
(10) comparing the predicted performance obtained in the step (9) with a given target performance, and calculating an error e ═ target performance-predicted performance)/target performance;
(11) if the error e is less than or equal to 10 percent (10 percent is the error set value of the embodiment), the predicted components output in the step (8) are reasonable values, otherwise, the P2C neural network parameters in the step (6) are reset, the P2C model in the step (7) is retrained, and then the steps (8) to (11) are repeated until the alloy components meeting the error requirement are screened out;
(12) and (3) performing reverse normalization treatment on the alloy components determined in the step (11) by using the same parameters as those in the step (3) to obtain the final design components of 5 alloys, as shown in table 1:
Figure BDA0002012843170000091
Figure BDA0002012843170000101
TABLE 1
Example 2: by applying the method provided by the invention, the tensile strength of 600MPa, the tensile strength of 625MPa and the tensile strength of 650MPa, and the electric conductivity of 60% IACS are respectively taken as performance targets, and the components of three alloys are designed under the condition that the performance error is not more than 10%, wherein the design method is specifically described as follows:
(1) data collection: establishing a basic data set by collecting and arranging publicly published components, strengthening modes, tensile strength and corresponding electric conductivity of Cu-Fe-P series, Cu-Ni-Si series and Cu-Cr-Zr series alloys;
(2) data cleaning: dividing the collected three copper alloy data into two types according to a strengthening mode, namely a deformation-precipitation strengthening type and a non-deformation-precipitation strengthening type, and discarding non-deformation-precipitation strengthening type data, such as an alloy initial data set obtained by processing methods such as pack rolling, equal-diameter angular extrusion, high-pressure knob and the like;
(3) normalization treatment: respectively carrying out normalization processing (0-1) on the alloy composition data and the performance data in the initial data set;
(4) establishing an alloy component → performance BP neural network model: taking the component data subjected to normalization processing in the step (3) as input, taking tensile strength and conductivity data as output, setting the number of initial hidden layers of the neural network to be 2, and the number of nodes of the initial hidden layers to be 12 and 8 respectively, and establishing a neural network model (abbreviated as C2P model) for predicting the alloy performance by components;
(5) training the C2P model: selecting Simoid as a hidden layer transfer function and a linear transfer function as an output layer transfer function, and training a C2P model by adopting a steepest descent method;
(6) establishing an alloy performance → component BP neural network model: taking the tensile strength and conductivity data subjected to normalization processing in the step (3) as input, taking the component data as output, setting the number of initial hidden layers of the neural network to be 2, and the number of nodes of the initial hidden layers to be 9 and 12 respectively, and establishing a neural network model (abbreviated as P2C model) with alloy performance design components;
(7) training the P2C model: selecting Simoid as a hidden layer transfer function and a linear transfer function as an output layer transfer function, and training a P2C model by adopting a steepest descent method;
(8) inputting the conductivity of 60% IACS under the three strengths of 600MPa, 625MPa and 650MPa of tensile strength as target performance (given performance) into the P2C model trained in the step (7) to obtain initial design components;
(9) taking the initial design components in the step (8) as the input of the C2P model, and outputting the predicted performance;
(10) comparing the predicted performance obtained in the step (9) with a given target performance, and calculating an error e ═ target performance-predicted performance)/target performance;
(11) if the error e is less than or equal to 10%, the predicted components output in the step (8) are reasonable values, otherwise, the parameters of the P2C neural network in the step (6) are reset, the model of the P2C in the step (7) is retrained, and then the steps (8) to (11) are repeated until the alloy components meeting the error requirement are screened out;
(12) and (3) performing inverse normalization treatment on the alloy components determined in the step (11) by using the same parameters as those in the step (3) to obtain the final design components of 3 alloys, as shown in Table 2:
Figure BDA0002012843170000111
TABLE 2
Example 3: by applying the method provided by the invention, five refractory high-entropy alloy components are designed under the condition that the performance error is not more than 15% by taking three strengths of compressive yield strength 1700MPa, 1750MPa and 1800MPa as performance targets, and the design method is specifically described as follows:
(1) data collection: establishing a basic data set by collecting and sorting publicly published components of the refractory high-entropy alloy, a preparation process and compressive yield strength;
(2) data cleaning: dividing the collected refractory high-entropy alloy data into two types according to preparation modes, namely preparing by adopting an electric arc melting furnace and other methods, and abandoning the initial data set of alloy preparation by other process methods, such as vacuum casting, powder metallurgy and other processes;
(3) normalization treatment: respectively carrying out normalization processing (0-1) on the alloy composition data and the performance data in the initial data set;
(4) establishing an alloy component → performance BP neural network model: taking the component data subjected to normalization processing in the step (3) as input, taking the compressive yield strength data as output, setting the number of initial hidden layers of the neural network to be 2, and the number of nodes of the initial hidden layers to be 6 and 8 respectively, and establishing a neural network model (abbreviated as C2P model) for predicting the alloy performance by components;
(5) training the C2P model: selecting Simoid as a hidden layer transfer function and a linear transfer function as an output layer transfer function, and training a C2P model by adopting a steepest descent method;
(6) establishing an alloy performance → component BP neural network model: taking the compression yield strength data after normalization processing in the step (3) as input, taking the component data as output, setting the number of initial hidden layers of the neural network to be 2, and the number of nodes of the initial hidden layers to be 5 and 8 respectively, and establishing a neural network model (abbreviated as P2C model) with alloy performance design components;
(7) training the P2C model: selecting Simoid as a hidden layer transfer function and a linear transfer function as an output layer transfer function, and training a P2C model by adopting a steepest descent method;
(8) inputting compressive strengths of 1700MPa, 1750MPa and 1800MPa as target performances into the P2C model trained in the step (7) to obtain initial design components;
(9) taking the initial design components in the step (8) as the input of the C2P model, and outputting the predicted performance;
(10) comparing the predicted performance obtained in the step (9) with a given target performance, and calculating an error e ═ target performance-predicted performance)/target performance;
(11) if the error e is less than or equal to 15%, the predicted components output in the step (8) are reasonable values, otherwise, the parameters of the P2C neural network in the step (6) are reset, the P2C model in the step (7) is retrained, and then the steps (8) to (11) are repeated until the alloy components meeting the error requirement are screened out;
(12) and (3) performing inverse normalization treatment on the alloy components determined in the step (11) by using the same parameters as those in the step (3) to obtain the final design components of 3 alloys, as shown in Table 3:
Figure BDA0002012843170000131
TABLE 3
The method for designing the components of the multi-component alloy based on machine learning and oriented to the performance requirements, provided by the embodiment of the application, is described in detail above. The above description of the embodiments is only for the purpose of helping to understand the method of the present application and its core ideas; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
As used in the specification and claims, certain terms are used to refer to particular components. As one skilled in the art will appreciate, manufacturers may refer to a component by different names. This specification and claims do not intend to distinguish between components that differ in name but not function. In the following description and in the claims, the terms "include" and "comprise" are used in an open-ended fashion, and thus should be interpreted to mean "include, but not limited to. "substantially" means within an acceptable error range, and a person skilled in the art can solve the technical problem within a certain error range to substantially achieve the technical effect. The description which follows is a preferred embodiment of the present application, but is made for the purpose of illustrating the general principles of the application and not for the purpose of limiting the scope of the application. The protection scope of the present application shall be subject to the definitions of the appended claims.
It is also noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a good or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such good or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a commodity or system that includes the element.
It should be understood that the term "and/or" as used herein is merely one type of association that describes an associated object, meaning that three relationships may exist, e.g., a and/or B may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship.
The foregoing description shows and describes several preferred embodiments of the present application, but as aforementioned, it is to be understood that the application is not limited to the forms disclosed herein, but is not to be construed as excluding other embodiments and is capable of use in various other combinations, modifications, and environments and is capable of changes within the scope of the application as described herein, commensurate with the above teachings, or the skill or knowledge of the relevant art. And that modifications and variations may be effected by those skilled in the art without departing from the spirit and scope of the application, which is to be protected by the claims appended hereto.

Claims (8)

1. A multi-component alloy component design method based on machine learning and oriented to performance requirements is characterized by comprising the following steps:
s1, establishing a data set according to the historical data;
s2, establishing a BP neural network model from alloy components to performance according to the data set, and training the BP neural network model;
s3, establishing a BP neural network model of alloy properties to components according to the data set, and training the BP neural network model;
s4, inputting the target performance as input data into the trained BP neural network model of the alloy performance-to-component to obtain an initial design component;
s5, inputting the initial design components as input data into the trained BP neural network model of the alloy components to obtain the predicted performance;
s6, calculating the error of the predicted performance relative to the target performance;
s7, judging whether the error is within a preset error range; if yes, finishing the alloy design; if not, re-establishing and re-training the BP neural network model of the alloy performance to composition, and repeating S4-S7;
the specific steps of S1 include:
s11, collecting historical data of the alloy to establish a basic data set, wherein the historical data comprises component data and performance data;
s12, classifying the data in the basic data set, and selecting the data of the type corresponding to the data as an initial data set according to the type of the alloy to be designed;
s13, carrying out normalization processing on the initial data set to obtain the data set;
the performance data in S11 is one or more of tensile strength, elongation after tensile failure, electric conductivity, compressive yield strength and hardness.
2. The method of claim 1, wherein the data in S12 is classified into one or more of processing, strengthening and alloy type.
3. The method of machine learning-based and performance-requirement-oriented multi-component alloy composition design according to any one of claims 1-2, wherein the alloy comprises an aluminum alloy, a copper alloy, a magnesium alloy, a titanium alloy, a superalloy, and a high-entropy alloy.
4. The method of claim 1, wherein the step S2 comprises the steps of:
s21, taking the component data in the data set as input and the performance data as output, setting model parameters of a BP neural network, and establishing a BP neural network model from the alloy components to the performance;
s22, selecting a transfer function and a training method to train the BP neural network model from the alloy components to the performance, so that the precision of the BP neural network model from the alloy components to the performance reaches the preset precision.
5. The machine-learning-based performance-requirement-oriented multi-component alloy composition design method of claim 4, wherein the predetermined precision is not less than 0.90.
6. The method of claim 1, wherein the step S3 comprises the steps of:
s31, taking the performance data and component data in the data set as input and the component data as output, setting model parameters of a BP neural network, and establishing a BP neural network model from the alloy performance to the components;
s32, selecting a transfer function and a training method to train the alloy performance → component BP neural network model, so that the precision of the alloy performance to component BP neural network model reaches the preset precision.
7. The machine-learning-based performance-requirement-oriented multi-component alloy composition design method of claim 6, wherein the predetermined precision is not less than 0.85.
8. The method for designing the components of the multi-component alloy based on machine learning and oriented to the performance requirements of claim 1, wherein the error in S6 is calculated by: (target performance-predicted performance)/target performance; the preset error range in the S7 is 5% -15%.
CN201910252935.1A 2019-03-29 2019-03-29 Multi-component alloy component design method based on machine learning and oriented to performance requirements Active CN110010210B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910252935.1A CN110010210B (en) 2019-03-29 2019-03-29 Multi-component alloy component design method based on machine learning and oriented to performance requirements

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910252935.1A CN110010210B (en) 2019-03-29 2019-03-29 Multi-component alloy component design method based on machine learning and oriented to performance requirements

Publications (2)

Publication Number Publication Date
CN110010210A CN110010210A (en) 2019-07-12
CN110010210B true CN110010210B (en) 2021-03-19

Family

ID=67169044

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910252935.1A Active CN110010210B (en) 2019-03-29 2019-03-29 Multi-component alloy component design method based on machine learning and oriented to performance requirements

Country Status (1)

Country Link
CN (1) CN110010210B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110387485B (en) * 2019-07-17 2021-02-02 西北工业大学 Component design method of metastable beta titanium alloy
CN112214925A (en) * 2020-09-17 2021-01-12 上海微亿智造科技有限公司 Sintering process product quality characteristic processing method and system suitable for deep model training
CN112582032B (en) * 2020-11-27 2024-05-17 北京科技大学 High-thermal-stability iron-based soft magnetic amorphous alloy designed based on machine learning model
CN112927764B (en) * 2021-03-03 2023-03-17 上海大学 Design method and system for target flame-retardant polymer composite material preparation model
CN113477929A (en) * 2021-04-15 2021-10-08 中国工程物理研究院材料研究所 High-flux preparation and component process optimization method of high-strength and high-toughness ODS steel
CN113862436A (en) * 2021-10-09 2021-12-31 河北工程大学 High-entropy alloy heat treatment method based on artificial intelligence prediction system
CN113870957B (en) * 2021-10-22 2024-09-03 中国航发北京航空材料研究院 Eutectic high-entropy alloy composition design method and device based on machine learning
CN115017816A (en) * 2022-06-17 2022-09-06 重庆大学 Proxy model-based copper rotor formula optimization method for induction motor
CN116682518B (en) * 2023-06-20 2024-09-10 小米汽车科技有限公司 Alloy formula determination method, device, medium and electronic equipment
CN117894387A (en) * 2023-11-22 2024-04-16 哈尔滨理工大学 Method for predicting Ni-Mn-based shape memory alloy thermal hysteresis ring Thys based on machine learning

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1718802A (en) * 2005-08-05 2006-01-11 武汉大学 Optimization designing method of nickel base high temperature alloy composition
CN102632082A (en) * 2011-02-11 2012-08-15 宝山钢铁股份有限公司 Performance prediction model based dynamic control method for mechanical property of hot strip
CN105631160A (en) * 2016-01-30 2016-06-01 西北工业大学 Method for forecasting room-temperature mechanical property of TC4 alloy forge piece
CN107577850A (en) * 2017-08-11 2018-01-12 中国航发北京航空材料研究院 Using the method for BP neural network prediction TC4 titanium alloy casting shrinkage cavity defects

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050232350A1 (en) * 2004-03-30 2005-10-20 Mana Hamada Inverse prediction apparatus and decoding apparatus using the same
CN1793931A (en) * 2005-11-30 2006-06-28 天津市产品质量监督检测技术研究院 Prediction method for mechanical performance of hot milling seamless steel tube
EP2457378A4 (en) * 2009-07-23 2016-08-10 Ericsson Telefon Ab L M Method and apparatus for encoding and decoding of images
GB201302743D0 (en) * 2013-02-18 2013-04-03 Rolls Royce Plc Method and system for designing a material
CN103886371A (en) * 2014-03-28 2014-06-25 郑州大学 Method for controlling component and thermal treatment technological process of pre-hardening plastic die steel
CN104899416B (en) * 2015-04-23 2018-08-10 江苏大学 A kind of prediction technique and its system of the anti-lamellar tearing performance of Steels for High Rise Buildings
CN105095962B (en) * 2015-07-27 2017-07-28 中国汽车工程研究院股份有限公司 A kind of material dynamic mechanical performance prediction method based on BP artificial neural networks
CN105740467B (en) * 2016-03-07 2019-02-05 东北大学 A kind of method for digging of C-Mn steel industry big data
CN107506565A (en) * 2017-10-16 2017-12-22 安徽工业大学 Method for building up based on neutral net alloy cast steel roll Expert System for Materials Design
CN107609647A (en) * 2017-10-16 2018-01-19 安徽工业大学 One kind is based on BP neural network roll alloy mechanical property Forecasting Methodology
CN107817521B (en) * 2017-10-27 2019-07-12 河海大学 It is a kind of based on improved at lithology distributions probability recognition methods golden in decigram
CN107801026B (en) * 2017-11-09 2019-12-03 京东方科技集团股份有限公司 Method for compressing image and device, compression of images and decompression systems
CN108197419A (en) * 2017-11-23 2018-06-22 合肥国轩高科动力能源有限公司 Method for analyzing abnormal performance of lithium battery material based on fault tree
CN108256689B (en) * 2018-02-06 2020-02-21 华中科技大学 Artificial neural network prediction method for amorphous alloy thermoplastic forming performance

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1718802A (en) * 2005-08-05 2006-01-11 武汉大学 Optimization designing method of nickel base high temperature alloy composition
CN102632082A (en) * 2011-02-11 2012-08-15 宝山钢铁股份有限公司 Performance prediction model based dynamic control method for mechanical property of hot strip
CN105631160A (en) * 2016-01-30 2016-06-01 西北工业大学 Method for forecasting room-temperature mechanical property of TC4 alloy forge piece
CN107577850A (en) * 2017-08-11 2018-01-12 中国航发北京航空材料研究院 Using the method for BP neural network prediction TC4 titanium alloy casting shrinkage cavity defects

Also Published As

Publication number Publication date
CN110010210A (en) 2019-07-12

Similar Documents

Publication Publication Date Title
CN110010210B (en) Multi-component alloy component design method based on machine learning and oriented to performance requirements
Zhao et al. Machine learning-assisted discovery of strong and conductive Cu alloys: Data mining from discarded experiments and physical features
El-Awady Unravelling the physics of size-dependent dislocation-mediated plasticity
Tofigh et al. Efficient optimum solution for high strength Al alloys matrix composites
Pickett et al. Failure prediction for advanced crashworthiness of transportation vehicles
Chaudry et al. Machine learning-aided design of aluminum alloys with high performance
Kumar et al. High temperature investigation on EDM process of Al 2618 alloy reinforced with Si3N4, ALN and ZrB2 in-situ composites
Inamdar et al. Studies on the prediction of springback in air vee bending of metallic sheets using an artificial neural network
Dewangan et al. Microstructure exploration and an artificial neural network approach for hardness prediction in AlCrFeMnNiWx High-Entropy Alloys
Hannula et al. Simulation-based exergy and LCA analysis of aluminum recycling: linking predictive physical separation and re-melting process models with specific alloy production
Belayadi et al. Neural network model for 7000 (Al-Z) alloys: Classification and prediction of mechanical properties
Li et al. Performance-oriented multistage design for multi-principal element alloys with low cost yet high efficiency
Vafaeenezhad et al. Application of constitutive description and integrated ANFIS–ICA analysis to predict hot deformation behavior of Sn-5Sb lead-free solder alloy
Pettersson et al. Experimental and computational study of nitride precipitation in a CrMnN austenitic stainless steel
Singh et al. Modeling and optimization of multi-performance characteristics of powder-mixed EDM of tungsten carbide alloy using intelligent decision-making tools
Kappatos et al. Assessment of the effect of existing corrosion on the tensile behaviour of magnesium alloy AZ31 using neural networks
Shirai et al. Mechanical properties prediction of gray cast iron considering trace elements based on deep learning
Dong et al. Just-in-time learning-based soft sensor for mechanical properties of strip steel via multi-block weighted semisupervised models
Krupiński et al. Prediction of mechanical properties of cast Mg-Al-Zn alloys
CN109977488A (en) Cast aluminium alloy gold mobility prediction technique based on support vector regression
Das et al. Development of multivariate process monitoring strategy for a typical process industry
Bhadrakali et al. A hybrid approach consisting of multi-objective and multivariate analyses for WAAM specimens
Kumar et al. Optimization of process parameter for stir casted AA 6063 metal matrix composite on hardness, tensile strength and impact energy
Mocellin et al. Numerical prediction of distortions during machining of large aluminium aeronautical parts: Numerische Vorhersage der während der Zerspanung von großen luftfahrttechnischen Aluminiumbauteilen auftretenden Verzüge
Poudel et al. AlloyManufacturingNet for discovery and design of hardness-elongation synergy in multi-principal element alloys

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant