Method and device for determining alloy preparation strategy, and bidirectional prediction system

Abstract

The application provides a method and device for determining an alloy preparation strategy and a bidirectional prediction system. The method comprises: obtaining a target alloy performance corresponding to an alloy to be prepared; performing the following training and prediction operations for the target alloy performance: determining a current training set in a target data sample set; training a preset preparation strategy prediction model using the current training set, predicting the target alloy performance by using the trained preparation strategy prediction model, and outputting a current preparation strategy corresponding to the alloy to be prepared; training a preset alloy performance prediction model using the current training set, predicting the current preparation strategy by using the trained alloy performance prediction model, and outputting a predicted alloy performance corresponding to the current preparation strategy; calculating a deviation value according to the target alloy performance and the predicted alloy performance; and determining a target preparation strategy based on the deviation value and a current training and prediction number. The application can provide a more accurate alloy preparation strategy for the target alloy performance.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of software, in particular to an alloy preparation strategy determination method and device and a bidirectional prediction system. BACKGROUND

[0002] Al-Zn-Mg-Cu alloy, as a multi-component aging alloy, has excellent strength-to-weight ratio, corrosion resistance and low density, and has been widely used in the aviation industry. At present, the development of the aviation industry puts forward higher requirements for the performance of Al-Zn-Mg-Cu alloy, especially for strength and toughness.

[0003] Fundamentally, the main methods to improve the performance of Al-Zn-Mg-Cu alloy include element design, control of heat treatment strategy and development of advanced processing technology. The emergence of some advanced processing technologies can undoubtedly improve the performance of Al-Zn-Mg-Cu alloy, such as severe plastic deformation (SPD), thereby realizing the synchronous improvement of strength and plasticity. However, these technologies always have problems of high cost, complex operation and limited product size. On the other hand, optimizing material elements and formulating a reasonable heat treatment strategy are two feasible methods to improve the performance of Al-Zn-Mg-Cu alloy, but the experimental range of unexplored element content and heat treatment parameters for Al-Zn-Mg-Cu alloy is extremely extensive, resulting in that the conventional experimental-trial-and-error method is time-consuming and laborious. On this basis, it is of great significance to propose an intelligent design system for Al-Zn-Mg-Cu alloy. Fortunately, the machine learning (ML) method can achieve this goal.

[0004] Based on rich algorithms and powerful multidimensional modeling capabilities, the machine learning method can effectively promote material discovery, which has attracted great interest of researchers in recent materials research. However, the current machine learning applications for Al-Zn-Mg-Cu alloy mainly focus on the optimization of heat treatment process and the prediction of performance under specific alloy composition or elements, and cannot provide an efficient alloy preparation strategy determination method including the determination process of alloy element content and heat treatment process for the target performance requirements. SUMMARY

[0005] The application aims to provide an alloy preparation strategy determination method and device and a bidirectional prediction system, through a bidirectional real-time training and prediction process of preparation strategy and alloy performance, a more accurate alloy preparation strategy can be given for the target alloy performance.

[0006] In a first aspect, the embodiments of the present application provide a method for determining an alloy preparation strategy, the method comprising: obtaining a target alloy performance corresponding to an alloy to be prepared; performing the following training and prediction operations for the target alloy performance: determining a current training set in a target data sample set; training a preset preparation strategy prediction model using the current training set, predicting the target alloy performance using the trained preparation strategy prediction model, and outputting a current preparation strategy corresponding to the alloy to be prepared; the current preparation strategy comprising: alloy element content and heat treatment parameter; training a preset alloy performance prediction model using the current training set, predicting the current preparation strategy using the trained alloy performance prediction model, and outputting a predicted alloy performance corresponding to the current preparation strategy; calculating a deviation value based on the target alloy performance and the predicted alloy performance; if the deviation value is less than a preset deviation threshold, determining the current preparation strategy as a target preparation strategy corresponding to the alloy to be prepared; if the deviation value is greater than the preset deviation threshold, determining whether a number of training and prediction times exceeds a number threshold; if the number of training and prediction times does not exceed the number threshold, continuing to perform the training and prediction operations; and if the number of training and prediction times exceeds the number threshold, determining a current preparation strategy corresponding to a minimum deviation value in prediction results within the number threshold as the target preparation strategy corresponding to the alloy to be prepared.

[0007] In a preferred embodiment of the present application, the construction process of the preset preparation strategy prediction model and the preset alloy performance prediction model is as follows: the preset preparation strategy prediction model and the preset alloy performance prediction model are respectively taken as a current to-be-constructed model, and the following steps are performed: obtaining a target data sample set; the samples in the sample set comprise a corresponding relationship between a preparation strategy and an alloy performance; training a plurality of preset machine learning models using the samples in the target data sample set; the plurality of preset machine learning models at least comprise a linear regression model, a support vector regression model, and an artificial neural network model; and determining a model with the highest prediction accuracy among the plurality of trained models as a target model corresponding to the current to-be-constructed model.

[0008] In a preferred embodiment of the present application, the step of obtaining the target data sample set comprises: obtaining alloy element features, heat treatment parameter features, and alloy performance features from literature; wherein the alloy element features comprise a plurality of alloy elements and a corresponding occurrence number and content of each element; extracting a target element with an occurrence number rate reaching a number threshold and a content reaching a content threshold from the alloy element features; determining a target heat treatment parameter from the heat treatment parameter features; determining a target alloy performance from the alloy performance features; wherein the target heat treatment parameter comprises single-step solid solution and aging; the target alloy performance comprises tensile strength, yield strength, and elongation; and generating the target data sample set based on the target element content, the target heat treatment parameter, and the target alloy performance.

[0009] In the preferred embodiments of the present application, the target data sample set corresponds to a plurality of training set proportions; each training set proportion corresponds to an extraction number interval; the step of determining the current training set in the target data sample set comprises: obtaining the current training prediction number; determining the target training set proportion corresponding to the target extraction number interval according to the current training prediction number; extracting a sample number corresponding to the target training set proportion from the target data sample set to obtain the current training set.

[0010] In the preferred embodiments of the present application, the step of training the preset preparation strategy prediction model using the current training set comprises: inputting the samples in the current training set into the preset preparation strategy prediction model for training, taking the preparation strategy as the output and taking the alloy performance as the input, to obtain the trained preparation strategy prediction model.

[0011] In the preferred embodiments of the present application, the step of training the preset alloy performance prediction model using the current training set comprises: inputting the samples in the current training set into the preset alloy performance prediction model for training, taking the preparation strategy as the input and taking the alloy performance as the output, to obtain the trained alloy performance prediction model.

[0012] In the preferred embodiments of the present application, the step of calculating the deviation value according to the target alloy performance and the predicted alloy performance comprises: subtracting the value corresponding to the target alloy performance from the value corresponding to the predicted alloy performance to obtain a difference value; and dividing the difference value by the value corresponding to the target alloy performance to obtain the deviation value.

[0013] In a second aspect, the embodiments of the present application further provide a device for determining an alloy preparation strategy, the device comprising: a performance obtaining module configured to obtain a target alloy performance corresponding to an alloy to be prepared; a bidirectional prediction module configured to perform the following training and prediction operations for the target alloy performance: determining a current training set in a target data sample set, training a preset preparation strategy prediction model using the current training set, predicting the target alloy performance by using the trained preparation strategy prediction model, and outputting a current preparation strategy corresponding to the alloy to be prepared; the current preparation strategy comprising an alloy element content and a heat treatment parameter; training a preset alloy performance prediction model using the current training set, predicting the current preparation strategy by using the trained alloy performance prediction model, and outputting a predicted alloy performance corresponding to the current preparation strategy; a deviation calculating module configured to calculate a deviation value according to the target alloy performance and the predicted alloy performance; and a strategy determining module configured to determine the current preparation strategy as a target preparation strategy corresponding to the alloy to be prepared if the deviation value is less than a preset deviation threshold, and determine, if the deviation value is greater than the preset deviation threshold, whether a number of times of training and prediction is greater than a number threshold, continue to perform the training and prediction operations if the number of times of training and prediction is not greater than the number threshold, and determine, if the number of times of training and prediction is greater than the number threshold, a current preparation strategy corresponding to a minimum deviation value in prediction results within the number threshold as the target preparation strategy corresponding to the alloy to be prepared.

[0014] In a third aspect, the embodiments of the present application further provide a bidirectional prediction system for alloy performance and preparation strategy, the system storing a target data sample set, a preset preparation strategy prediction model and a preset alloy performance prediction model; the system being configured to predict a target preparation strategy corresponding to a target alloy performance by using the method for determining an alloy preparation strategy according to the first aspect based on the target data sample set, the preset preparation strategy prediction model and the preset alloy performance prediction model; and the system being further configured to predict the target alloy performance corresponding to the target preparation strategy by using the preset alloy performance prediction model.

[0015] In a fourth aspect, the embodiments of the present application further provide a computer readable storage medium storing computer executable instructions, the computer executable instructions causing a processor to implement the method according to the first aspect when the computer executable instructions are called and executed by the processor.

[0016] This application provides a method, apparatus, and bidirectional prediction system for determining an alloy preparation strategy. First, the target alloy properties corresponding to the alloy to be prepared are obtained. For the target alloy properties, the following training and prediction operations are performed: The current training set in the target data sample set is determined; a preset preparation strategy prediction model is trained using the current training set; the target alloy properties are predicted using the trained preparation strategy prediction model, and the current preparation strategy corresponding to the alloy to be prepared is output; the current preparation strategy includes: alloy element content and heat treatment parameters; a preset alloy property prediction model is trained using the current training set; the current preparation strategy is predicted using the trained alloy property prediction model, and the predicted alloy properties corresponding to the current preparation strategy are output; a deviation value is calculated based on the target alloy properties and the predicted alloy properties; if the deviation value is less than a preset deviation threshold, the current preparation strategy is determined as the target preparation strategy corresponding to the alloy to be prepared; if the deviation value is greater than the preset deviation threshold, it is determined whether the current training prediction count exceeds a threshold; if it does not exceed the threshold, the training and prediction operation continues; if it exceeds the threshold, the current preparation strategy corresponding to the smallest deviation value among the prediction results within the threshold is determined as the target preparation strategy corresponding to the alloy to be prepared. In the embodiments of this application, a more accurate alloy preparation strategy can be provided for the target alloy properties through a two-way real-time training and prediction process of preparation strategy and alloy properties. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a method for determining an alloy preparation strategy provided in this application embodiment;

[0019] Figure 2 A schematic diagram illustrating the process of determining an alloy preparation strategy as provided in an embodiment of this application;

[0020] Figure 3 A flowchart illustrating the sample set determination method in an alloy preparation strategy determination method provided in this application embodiment;

[0021] Figure 4 A comparative chart showing the predictive capabilities of three machine learning models for three key mechanical properties of aluminum alloys, provided in an embodiment of this application.

[0022] Figure 5A flowchart of a current sample set determination method in a method for determining an alloy preparation strategy provided by an embodiment of the present application is shown in FIG. 1.

[0023] Figure 6 A structural block diagram of a device for determining an alloy preparation strategy provided by an embodiment of the present application is shown in FIG. 2.

[0024] Figure 7 A structural block diagram of a bidirectional prediction system provided by an embodiment of the present application is shown in FIG. 3. DETAILED DESCRIPTION

[0025] The technical solutions of the present application will be described in detail below with reference to the embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0026] Currently, the application of machine learning to Al-Zn-Mg-Cu alloy mainly focuses on the optimization of heat treatment process and the prediction of performance under specific alloy composition or elements, and cannot provide an efficient method for determining an alloy preparation strategy, including the determination process of alloy element content and heat treatment process, for a target performance requirement. Based on this, the embodiments of the present application provide a method and device for determining an alloy preparation strategy and a bidirectional prediction system. Through the bidirectional real-time training and prediction process of the preparation strategy and the alloy performance, a more accurate alloy preparation strategy can be given for a target alloy performance.

[0027] To facilitate the understanding of the embodiments, first, a method for determining an alloy preparation strategy disclosed by the embodiments of the present application is described in detail.

[0028] Figure 1 A flowchart of a method for determining an alloy preparation strategy provided by an embodiment of the present application is shown in FIG. 1. Figure 2 A schematic diagram of a determination process of an alloy preparation strategy provided by an embodiment of the present application is shown in FIG. 2. Figure 1 As shown in the figure, the method specifically includes the following steps:

[0029] Step S101, obtaining a target alloy performance corresponding to an alloy to be prepared.

[0030] Step S102, performing the following training and prediction operations for the target alloy performance:

[0031] Step S1021, determining a current training set in a target data sample set. The determination process of the target data sample set and the current training set can be referred to the following description. Each time the training and prediction operation is performed, the current training set needs to be determined from the target data sample set, that is, the samples in the training set are different each time.

[0032] Step S1022, the current training set is applied to train the preset preparation strategy prediction model, the target alloy performance is predicted through the trained preparation strategy prediction model, and the current preparation strategy corresponding to the alloy to be prepared is output. The current preparation strategy includes: alloy element content, heat treatment parameter.

[0033] In specific implementation, the sample in the current training set is input into the preset preparation strategy prediction model for training by taking the preparation strategy as the output and taking the alloy performance as the input, so as to obtain the trained preparation strategy prediction model.

[0034] Step S1023, the current training set is applied to train the preset alloy performance prediction model, the current preparation strategy is predicted through the trained alloy performance prediction model, and the predicted alloy performance corresponding to the current preparation strategy is output.

[0035] In specific implementation, the sample in the current training set is input into the preset alloy performance prediction model for training by taking the preparation strategy as the input and taking the alloy performance as the output, so as to obtain the trained alloy performance prediction model.

[0036] The process of determining the current training set in the target data sample set, applying the current training set to train the preset preparation strategy prediction model, and applying the current training set to train the preset alloy performance prediction model is not shown in Figure 2 . It can be understood that, Figure 2 the P-S model in may be regarded as a model trained by applying the current training set to the preset preparation strategy prediction model; and the S-P model may be regarded as a model trained by applying the current training set to the preset alloy performance prediction model.

[0037] The input of the P-S model is the target alloy performance, and the output of the model is the current preparation strategy; the input of the S-P model is the current preparation strategy, and the output is the predicted alloy performance.

[0038] Step S103, the deviation value is calculated according to the target alloy performance and the predicted alloy performance.

[0039] In specific implementation, the target alloy performance and the predicted alloy performance each correspond to a numerical value, and the performance is represented by the numerical value. Therefore, when the deviation value is calculated, the value corresponding to the predicted alloy performance can be subtracted from the value corresponding to the target alloy performance to obtain a difference value; and then the difference value is divided by the value corresponding to the target alloy performance to obtain the deviation value.

[0040] Step S104, whether the deviation value is less than a preset deviation threshold is judged; for example, Figure 2 whether the deviation value is less than the PT preset deviation threshold, in the embodiment of the present application, the PT value is finally set to 5% after repeated verification.

[0041] If the deviation value is less than the preset deviation threshold, step S105 is performed to determine the current preparation strategy as the target preparation strategy corresponding to the alloy to be prepared; that is Figure 2 the output result 1 shown in

[0042] If the deviation value is greater than the preset deviation threshold, step S106 is performed to determine whether the current training prediction times exceeds the times threshold; for example Figure 2 whether the training prediction times shown in

[0043] If the times threshold is not exceeded, the training prediction operation of step S102 is continued to be performed;

[0044] If the times threshold is exceeded, step S107 is performed to determine the current preparation strategy corresponding to the minimum deviation value in the prediction results within the times threshold as the target preparation strategy corresponding to the alloy to be prepared, for example Figure 2 the output result 2 shown in

[0045] The determination method of the alloy preparation strategy provided in the embodiments of the present application can give a more accurate alloy preparation strategy for the target alloy performance through the bidirectional real-time training and prediction process of the preparation strategy and the alloy performance.

[0046] The embodiments of the present application also provide another determination method of an alloy preparation strategy, which is implemented on the basis of the above-mentioned embodiments; the model training process and the bidirectional prediction process are mainly described in the present embodiment.

[0047] Referring to Figure 3 The construction process of the above-mentioned preset preparation strategy prediction model and the preset alloy performance prediction model is as follows:

[0048] The preset preparation strategy prediction model and the preset alloy performance prediction model are respectively taken as the current to-be-constructed model, and the following steps are performed:

[0049] Step S302: Obtain a target data sample set; the samples in the sample set include the corresponding relationship between the preparation strategy and the alloy performance.

[0050] In specific implementation, the target data sample set can be obtained through the following steps:

[0051] (1) Obtain the alloy element features, heat treatment parameter features and alloy performance features from the literature; wherein the alloy element features include a plurality of alloy elements and the occurrence times and contents of each element;

[0052] (2) from the alloy element characteristics, extract the target elements whose occurrence frequency rate reaches the frequency threshold and whose content reaches the content threshold; determine the target heat treatment parameters from the heat treatment parameter characteristics; determine the target alloy performance from the alloy performance characteristics; wherein the target heat treatment parameters include: single-step solid solution and aging; and the target alloy performance includes: tensile strength, yield strength and elongation;

[0053] (3) based on the target element corresponding content, the target heat treatment parameters and the target alloy performance, generate a target data sample set.

[0054] In specific implementation, through analysis of the alloy composition in the retrieved literature, 24 elements, i.e. Al, Zn, Mg, Cu, Zr, Si, Fe, Mn, Ti, Cr, Sc, La, Er, Y, Be, Pr, Yb, Ag, Ni, Ce, Li, Sr, Gd and Sn, are selected as initial composition characteristics. The solid solution temperature and time, and the aging temperature and time are defined as initial characteristics of the heat treatment parameters. In terms of performance characteristics, in addition to the tensile strength and elongation, other characteristics such as yield strength, hardness, fracture toughness and grain size are also collected. Next, the original data set is further cleaned. The elements (Cu, Mg, Cr, Zn, Zr, Sc, Al) with an occurrence of more than 100 times and an average content of more than 0.01wt% are selected as the final composition characteristics. Since the multi-stage heat treatment data is relatively small, single-step solid solution and aging are used as the heat treatment characteristics. The tensile strength, yield strength and elongation, which are the most concerned in this work, are selected as the performance characteristics.

[0055] After verifying the reliability of the data set, the StandardScalar method can also be applied to standardize the data set and normalize the characteristics to [-1, 1]. Finally, a data set containing 1967 labeled samples is established, including element content, heat treatment parameters and material performance. After preliminary screening, an aluminum alloy data set is successfully created, including element content, processing technology and material performance.

[0056] Step S304, applying samples in the target data sample set to train a plurality of preset machine learning models; the plurality of preset machine learning models at least include: a linear regression model, a support vector regression model and an artificial neural network model;

[0057] Step S306, determining the model with the highest prediction accuracy from the plurality of trained models as the target model corresponding to the current to-be-constructed model.

[0058] In the embodiments of the present application, three ML algorithms, linear regression (LR), support vector regression (SVR) and artificial neural network (ANN), are used to train the data set to construct a multi-performance prediction model, which is named S-P in the embodiments of the present application, i.e., the alloy performance prediction model. The preparation strategy composed of element content and heat treatment parameters is the input feature of the model, and the material performance is the output feature of the model. In this work, a grid search is performed to determine the optimal parameters of the SVR model; in the LR model, a regularization method is applied to avoid overfitting; in the construction of the ANN model, the network structure, the number of iterations and the learning rate and other hyperparameters play a crucial role in the accuracy of the model. In the present application, after analysis, a three-layer neural network structure of 12-16-3 and a learning rate of 0.0035 are finally determined. The predicted values of the three models are compared with the actual values, as shown in Figure 4 The comparison chart of the prediction ability of the three machine learning models for the three key mechanical properties of the aluminum alloy (tensile strength, yield strength and elongation) can be found that the relative error and bias of SVR are the smallest. Therefore, the S-P model is finally constructed using the SVR algorithm. Similarly, the P-S model in the system is also constructed based on the SVR algorithm, and the function of the model is to predict the material performance according to the preparation strategy.

[0059] The target data sample set corresponds to a plurality of training set proportions; each training set proportion corresponds to an extraction number interval; for example, the proportions of the training set and the validation set are 9:1, 8:2, 7:3, 6:4, 5:5 and 4:6; and the training set proportions are 0.9, 0.8, 0.7, 0.6, 0.5 and 0.4, respectively. Each training set proportion corresponds to an extraction number interval; for example, when the training set proportion is 0.9, the corresponding extraction number interval is [1-50], when the training set proportion is 0.8, the corresponding extraction number interval is [50-100], and so on. That is, each training set proportion corresponds to 50 extraction numbers. The steps for determining the current training set in the target data sample set include the following steps, as shown in Figure 5

[0060] Step S502, obtaining the current training prediction number; assuming that the current training prediction number is 9.

[0061] Step S504, determining the target training set proportion corresponding to the target extraction number interval according to the current training prediction number; for example, the target extraction number interval corresponding to the current training prediction number 9 is [1-50], so the training set proportion 0.9 corresponding to [1-50] is taken as the target training set proportion.

[0062] Step S506, extracting samples of a sample number corresponding to the target training set proportion from the target data sample set to obtain the current training set.​

[0063] Assuming that there are a total of 100 samples in the target data sample set, the target training set proportion corresponds to the number of samples extracted from the target data sample set, that is, 90 samples are randomly extracted from 100 samples as the current training set.

[0064] It should be noted that in model construction, in addition to the three machine learning algorithms that have been compared, other machine learning regression algorithms can also be used for modeling. In addition to being applicable to the design of aluminum alloys, this method can also be applied to the design of other materials, including copper alloys, titanium alloys, high-entropy alloys, steels, semiconductor materials, batteries, and the like.

[0065] The alloy preparation strategy determination method provided by the embodiments of the present application can quickly predict the approximate performance of the alloy according to the aluminum alloy composition and heat treatment system, reducing the waste of experiments and manual work; on the contrary, the suitable alloy composition and corresponding heat treatment parameters can also be quickly predicted according to the input alloy performance, so as to realize arbitrary design of performance and meet the actual industrial needs, such as extremely high strength requirement and low elongation (ductility) requirement in some cases; such as high elongation requirement and low strength requirement, and the like.

[0066] Based on the above method embodiments, the embodiments of the present application also provide an alloy preparation strategy determination device, as shown in Figure 6 The device includes:

[0067] The performance acquisition module 62 is configured to acquire a target alloy performance corresponding to the alloy to be prepared; the bidirectional prediction module 64 is configured to perform the following training and prediction operations for the target alloy performance: determining a current training set in a target data sample set, training a preset preparation strategy prediction model by using the current training set, predicting the target alloy performance by using the trained preparation strategy prediction model, and outputting a current preparation strategy corresponding to the alloy to be prepared; the current preparation strategy includes an alloy element content and a heat treatment parameter; training a preset alloy performance prediction model by using the current training set, predicting the current preparation strategy by using the trained alloy performance prediction model, and outputting a predicted alloy performance corresponding to the current preparation strategy; the deviation calculation module 66 is configured to calculate a deviation value according to the target alloy performance and the predicted alloy performance; the strategy determination module 68 is configured to determine the current preparation strategy as a target preparation strategy corresponding to the alloy to be prepared if the deviation value is less than a preset deviation threshold; and if the deviation value is greater than the preset deviation threshold, determining whether a training and prediction frequency exceeds a frequency threshold; if the training and prediction frequency does not exceed the frequency threshold, continuing to perform the training and prediction operation; and if the training and prediction frequency exceeds the frequency threshold, determining a current preparation strategy corresponding to a minimum deviation value in the prediction results within the frequency threshold as the target preparation strategy corresponding to the alloy to be prepared.

[0068] In the preferable embodiments of the present application, the device further comprises a model construction module, configured to perform the following construction process of the preset preparation strategy prediction model and the preset alloy performance prediction model: taking the preset preparation strategy prediction model and the preset alloy performance prediction model as the current to-be-constructed model respectively, and performing the following steps: obtaining a target data sample set; the samples in the sample set comprise a corresponding relationship between the preparation strategy and the alloy performance; applying the samples in the target data sample set to train a plurality of preset machine learning models; the plurality of preset machine learning models at least comprise a linear regression model, a support vector regression model and an artificial neural network model; and determining the model with the highest prediction accuracy among the plurality of trained models as the target model corresponding to the current to-be-constructed model.

[0069] In the preferable embodiments of the present application, the model construction module is configured to obtain alloy element features, heat treatment parameter features and alloy performance features from literature; wherein the alloy element features comprise a plurality of alloy elements and the corresponding occurrence frequency and content of each element; the target element with the occurrence frequency reaching a frequency threshold and the content reaching a content threshold is extracted from the alloy element features; the target heat treatment parameter is determined from the heat treatment parameter features; the target alloy performance is determined from the alloy performance features; wherein the target heat treatment parameter comprises single-step solid solution and aging; the target alloy performance comprises tensile strength, yield strength and elongation; and the target data sample set is generated based on the target element content, the target heat treatment parameter and the target alloy performance.

[0070] In the preferable embodiments of the present application, the target data sample set corresponds to a plurality of training set proportions; each of the training set proportions corresponds to an extraction frequency interval; and the bidirectional prediction module is further configured to obtain a current training prediction frequency; determine the target training set proportion corresponding to the target extraction frequency interval according to the current training prediction frequency; extract a sample number corresponding to the target training set proportion from the target data sample set to obtain a current training set.

[0071] In the preferable embodiments of the present application, the bidirectional prediction module is further configured to input the samples in the current training set to the preset preparation strategy prediction model as the output and the alloy performance as the input for training, to obtain the trained preparation strategy prediction model.

[0072] In the preferable embodiments of the present application, the bidirectional prediction module is further configured to input the samples in the current training set to the preset preparation strategy prediction model as the output and the alloy performance as the input for training, to obtain the trained preparation strategy prediction model.

[0073] In the preferred embodiment of the present application, the deviation calculation module is configured to: subtract the value corresponding to the target alloy performance from the value corresponding to the predicted alloy performance to obtain a difference value; and divide the difference value by the value corresponding to the target alloy performance to obtain the deviation value.

[0074] The device provided in the embodiments of the present application has the same implementation principle and technical effects as the foregoing method embodiments. For brief description, the part of the device embodiments not mentioned can be referred to the corresponding content in the foregoing method embodiments.

[0075] Based on the foregoing method embodiments, the embodiments of the present application further provide an alloy performance and preparation strategy bidirectional prediction system, as shown in Figure 7 The system stores a target data sample set 71, a preset preparation strategy prediction model 72 and a preset alloy performance prediction model 73. The system is configured to predict a target preparation strategy corresponding to a target alloy performance based on the target data sample set, the preset preparation strategy prediction model and the preset alloy performance prediction model by using the alloy preparation strategy determination method described in the foregoing method embodiments. The system is further configured to predict the target alloy performance corresponding to the target preparation strategy by using the preset alloy performance prediction model.

[0076] The alloy performance and preparation strategy bidirectional prediction system provided in the embodiments of the present application is an Al-Zn-Mg-Cu alloy rapid design system (ARDS). The system can be used to customize a preparation strategy under a required performance or inversely predict an alloy performance based on an input preparation strategy. It is worth noting that the prediction reliability of the system designed by experimental data is verified. The results show that the upper limit of the system for accurately predicting the tensile strength, yield strength and elongation of the aluminum alloy is about 800 MPa, 730 MPa and 30%, respectively. The aluminum alloy rapid design system (ARDS) developed in the embodiments of the present application can not only be used to guide the discovery of Al-Zn-Mg-Cu alloy, but also can be extended to the design of other materials.

[0077] The system provided in the embodiments of the present application has the same implementation principle and technical effects as the foregoing method embodiments. For brief description, the part of the system embodiments not mentioned can be referred to the corresponding content in the foregoing method embodiments.

[0078] The embodiments of the present application further provide a computer readable storage medium storing computer executable instructions. When the computer executable instructions are called and executed by a processor, the computer executable instructions cause the processor to implement the foregoing method. The specific implementation can be referred to the foregoing method embodiments, which will not be described herein again.

[0079] The computer program product of the method, the device and the electronic equipment provided in the embodiment of the application includes a computer readable storage medium storing program codes, the program codes include instructions for executing the method described in the foregoing method embodiments, and specific implementation can be referred to the method embodiments, and details are not described herein.

[0080] Unless specifically stated, the relative steps, numerical expressions and values of the components and steps set forth in these embodiments do not limit the scope of the application.

[0081] If the functions are implemented in the form of software function units and sold or used as independent products, the functions can be stored in a nonvolatile computer readable storage medium executable by a processor. Based on such understanding, the technical solutions of the application essentially or the parts that make contributions to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0082] In the description of the application, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like indicate the orientation or positional relationship shown in the drawings, and are only used to facilitate the description of the application and simplify the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the application. In addition, the terms "first", "second", "third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0083] Finally, it should be noted that the above-described embodiments are merely specific embodiments of the present application, which are used to illustrate the technical solutions of the present application, but not to limit the same. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that any skilled person in the art can still modify or easily think of changes to the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some of the technical features, within the technical scope disclosed by the present application. The modifications, changes or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method of determining an alloying strategy, characterized in that, The method includes: Obtain the target alloy properties corresponding to the alloy to be prepared; For the target alloy properties, the following training and prediction operations are performed: Get the current number of training predictions; Based on the current training prediction count, determine the target training set percentage corresponding to the target extraction count interval; each training set percentage corresponds to an extraction count interval; Extract the number of samples corresponding to the proportion of the target training set from the target data sample set to obtain the current training set; Determine the current training set in the target data sample set; The current training set is used to train a preset preparation strategy prediction model. The trained preparation strategy prediction model is used to predict the target alloy properties and outputs the current preparation strategy corresponding to the alloy to be prepared. The current preparation strategy includes: alloy element content and heat treatment parameters. The current training set is used to train the preset alloy property prediction model. The trained alloy property prediction model is used to predict the current preparation strategy, and the predicted alloy properties corresponding to the current preparation strategy are output. Calculate the deviation value based on the target alloy properties and the predicted alloy properties; If the deviation value is less than a preset deviation threshold, the current preparation strategy is determined as the target preparation strategy corresponding to the alloy to be prepared; If the deviation value is greater than the preset deviation threshold, determine whether the current number of training predictions exceeds the threshold; if it does not exceed the threshold, continue to perform the training prediction operation; if it exceeds the threshold, determine the current preparation strategy corresponding to the smallest deviation value among the prediction results within the threshold as the target preparation strategy for the alloy to be prepared. The construction process of the preset preparation strategy prediction model and the preset alloy performance prediction model is as follows: Using the preset preparation strategy prediction model and the preset alloy property prediction model as the current models to be constructed, respectively, the following steps are performed: Obtain the target data sample set; the samples in the sample set include: the correspondence between preparation strategies and alloy properties; The samples in the target data sample set are used to train various preset machine learning models; the various preset machine learning models include at least: linear regression model, support vector regression model and artificial neural network model; The model with the highest prediction accuracy among the various trained models is determined as the target model corresponding to the current model to be constructed.

2. The method of claim 1, wherein, The steps to obtain the target data sample set include: The alloy element characteristics, heat treatment parameter characteristics, and alloy property characteristics are obtained from the literature; wherein, the alloy element characteristics include: multiple alloy elements and the frequency and content of each element; From the alloy element characteristics, target elements with occurrence rates reaching a threshold and content reaching a threshold are extracted; target heat treatment parameters are determined from the heat treatment parameter characteristics; target alloy properties are determined from the alloy property characteristics; wherein, the target heat treatment parameters include: single-step solution treatment and aging; the target alloy properties include: tensile strength, yield strength, and elongation. A target data sample set is generated based on the content of the target element, the target heat treatment parameters, and the target alloy properties.

3. The method of claim 1, wherein, The target data sample set has multiple training set proportions; each training set proportion has a range of extraction times. The steps to determine the current training set in the target data sample set include: Get the current number of training predictions; Based on the current number of training predictions, determine the proportion of the target training set corresponding to the target extraction frequency interval; The current training set is obtained by extracting samples from the target data sample set that correspond to the proportion of the target training set.

4. The method of claim 1, wherein, The steps of training a prediction model for a preset preparation strategy using the current training set include: The samples in the current training set are used as the output of the preparation strategy and the input of the alloy properties. These are then input into the preset preparation strategy prediction model for training, resulting in the trained preparation strategy prediction model.

5. The method of claim 1, wherein, The steps of training a preset alloy property prediction model using the current training set include: The samples in the current training set are used as input for the preparation strategy and as output for the alloy properties. These samples are then input into the preset preparation strategy prediction model for training, resulting in a trained alloy property prediction model.

6. The method of claim 1, wherein, The step of calculating the deviation value based on the target alloy properties and the predicted alloy properties includes: The difference is obtained by subtracting the value corresponding to the target alloy property from the value corresponding to the predicted alloy property. The deviation value is obtained by dividing the difference by the value corresponding to the target alloy property.

7. An apparatus for determining an alloying strategy, characterized in that The device includes: The performance acquisition module is used to acquire the target alloy properties corresponding to the alloy to be prepared. A bidirectional prediction module is used to perform the following training and prediction operations for the target alloy properties: obtaining the current training prediction count; determining the target training set percentage corresponding to the target extraction count interval based on the current training prediction count; each training set percentage corresponds to an extraction count interval; extracting samples corresponding to the target training set percentage from the target data sample set to obtain the current training set; determining the current training set in the target data sample set; applying the current training set to train a preset preparation strategy prediction model, predicting the target alloy properties through the trained preparation strategy prediction model, and outputting the current preparation strategy corresponding to the alloy to be prepared; the current preparation strategy includes: alloy element content and heat treatment parameters; applying the current training set to train a preset alloy property prediction model, predicting the current preparation strategy through the trained alloy property prediction model, and outputting the predicted alloy properties corresponding to the current preparation strategy; The deviation calculation module is used to calculate the deviation value based on the target alloy properties and the predicted alloy properties; The strategy determination module is used to determine the current preparation strategy as the target preparation strategy corresponding to the alloy to be prepared if the deviation value is less than a preset deviation threshold; if the deviation value is greater than the preset deviation threshold, determine whether the current training prediction count exceeds the count threshold; if it does not exceed the count threshold, continue to execute the training prediction operation; if it exceeds the count threshold, determine the current preparation strategy corresponding to the smallest deviation value among the prediction results within the count threshold as the target preparation strategy corresponding to the alloy to be prepared. The construction process of the preset preparation strategy prediction model and the preset alloy performance prediction model is as follows: Using the preset preparation strategy prediction model and the preset alloy property prediction model as the current models to be constructed, respectively, the following steps are performed: Obtain the target data sample set; the samples in the sample set include: the correspondence between preparation strategies and alloy properties; The samples in the target data sample set are used to train various preset machine learning models; the various preset machine learning models include at least: linear regression model, support vector regression model and artificial neural network model; The model with the highest prediction accuracy among the various trained models is determined as the target model corresponding to the current model to be constructed.

8. A two-way prediction system for alloy properties and preparation strategies, characterized in that, The system stores a target data sample set, a preset preparation strategy prediction model, and a preset alloy performance prediction model; the system is used to predict the target preparation strategy corresponding to the target alloy performance based on the target data sample set, the preset preparation strategy prediction model, and the preset alloy performance prediction model, using the alloy preparation strategy determination method as described in any one of claims 1-6. The system is also used to predict the target alloy properties corresponding to the target preparation strategy using the preset alloy property prediction model.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method according to any one of claims 1 to 6.

Images