Prediction method, training method and system

Abstract

The embodiment of the invention provides a prediction method, a training method and a system. In the prediction method, a prediction system generates a plurality of core descriptors and a plurality of non-core descriptors for describing a target material based on proportion information corresponding to each element in the target material and attribute values of a plurality of attributes, and inputting at least part of the descriptors into the target network to obtain a predicted value of the amorphous critical cooling rate of the target material. The target network is a network obtained by training a plurality of sample materials by a training system, and comprises a plurality of sub-networks which have the same structure and are arranged according to levels, each sub-network comprises N input nodes and one output node, the output node of the i-th level sub-network is used as the input node of the (i + 1)-th level sub-network, and the output node of the (i + 1)-th level sub-network is used as the input node of the (i + 1)-th level sub-network; n core descriptors input into each sub-network in the first hierarchy are the same, N-N1 non-core descriptors are not completely the same, both N and N1 are greater than 1, and i is greater than or equal to 1.

Description

Technical Field

[0001] This specification relates to the field of chemical materials, and in particular to a prediction method, training method, and system. Background Art

[0002] Materials typically exist in two states: crystalline and amorphous. Crystalline materials have a regular atomic arrangement, and researchers have developed a relatively comprehensive theoretical framework for studying aspects of their equilibrium state. However, despite their enormous potential in scientific research and industrial applications, amorphous materials remain understudied due to the complexity of their preparation.

[0003] Metallic glass, as an amorphous material, has attracted extensive attention from researchers in recent years due to its unique mechanical and physical properties. In 1960, Duwez introduced the rapid cooling technology, which achieved a cooling rate higher than the critical cooling rate of amorphization (hereinafter referred to as the critical cooling rate of amorphization, R c ) cooling, that is, R c >10 7 K / s, ushering in a new era in the research of metallic glass. In the preparation process of metallic glass, the critical cooling rate of amorphization is a crucial parameter that determines whether the liquid metal can successfully transform into an amorphous structure.

[0004] At present, the research on the critical cooling rate of amorphization of materials mainly relies on experimental methods. Due to the complexity of experimental conditions, experimental measurement of the critical cooling rate of amorphization of materials requires a lot of time and cost.

[0005] The content of the background technology section is merely information known to the inventor personally, and does not mean that the above information has entered the public domain before the application date of this disclosure, nor does it mean that it can become the prior art of the present disclosure. Summary of the Invention

[0006] This specification provides a prediction method, training method and system that can accurately predict the critical cooling rate of amorphization of materials through a pre-trained target network, reducing the time and cost of research on amorphous materials.

[0007] In a first aspect, the present specification provides a method for predicting an amorphization critical cooling rate, comprising: generating K descriptors for describing the target material based on information about the proportion of at least one element included in the target material to be predicted, and attribute values of X attributes corresponding to each of the at least one element; determining N1 core descriptors and K-N1 non-core descriptors from the K descriptors, wherein the core descriptors have a higher degree of influence on the amorphization critical cooling rate than the non-core descriptors; and inputting at least some of the K descriptors into a target network for prediction to obtain a predicted value corresponding to the amorphization critical cooling rate of the target material, wherein , the target network is a network trained using Y sample materials, the target network includes multiple sub-networks with the same structure and arranged in layers, each sub-network includes N input nodes and 1 output node, the output node of the sub-network in the i-th layer serves as the input node of the sub-network in the i+1-th layer, and each sub-network in the 1st layer is input with the N1 core descriptors and N-N1 non-core descriptors, the N-N1 non-core descriptors are part of the K-N1 non-core descriptors, and the non-core descriptors input to different sub-networks are not exactly the same; wherein, the X, the K, the N1, the Y, and the N are all integers greater than 1, and the i is an integer greater than or equal to 1.

[0008] In a second aspect, the present specification provides a target network training method for predicting the critical amorphization cooling rate of a material, the method comprising: obtaining experimental values of the critical amorphization cooling rate for each of Y sample materials, each sample material comprising at least one element; for each sample material, generating K descriptors for describing the sample material based on information about the proportion of the at least one element in the sample material and attribute values of X attributes corresponding to each of the at least one element, and determining N1 core descriptors and K-N1 non-core descriptors from the K descriptors, wherein the core descriptors have a greater impact on the critical amorphization cooling rate than the non-core descriptors; and inputting some of the K descriptors corresponding to each sample material into the target network for prediction. A predicted value corresponding to the critical cooling rate of amorphization of the sample material is obtained, wherein: the target network includes multiple sub-networks with the same structure and arranged in a hierarchy, each sub-network includes N input nodes and 1 output node, the output node of the sub-network in the i-th level serves as the input node of the sub-network in the i+1-th level, each sub-network in the 1st level is input with the N1 core descriptors and N-N1 non-core descriptors, the N-N1 non-core descriptors are part of the K-N1 non-core descriptors, and the non-core descriptors input to different sub-networks are not exactly the same; and the parameters of the target network are updated with the training goal of minimizing the difference between the predicted value and the experimental value; wherein, X, K, N1, Y, and N are all integers greater than 1, and i is an integer greater than or equal to 1.

[0009] In a third aspect, the present specification provides a method for predicting a eutectic point, comprising: based on the method described in the first aspect, predicting the critical cooling rate of amorphization of multiple materials formed by multiple elements in a target multi-element system, each of the multiple materials including the multiple elements, and the proportion information of the multiple elements in different materials is different; and determining the material with the smallest corresponding critical cooling rate of amorphization among the multiple materials as the target material, and determining the eutectic point of the target multi-element system based on the target material.

[0010] In a fourth aspect, the present specification also provides a prediction system configured to predict the critical cooling rate of amorphization of a target material, comprising: at least one storage medium storing at least one instruction set; and at least one processor communicatively connected to the at least one storage medium, wherein when the prediction system is running, the at least one processor reads the at least one instruction set and executes the prediction method described in the first aspect according to the instructions of the at least one instruction set.

[0011] In a fifth aspect, the present specification also provides a training system configured to train a target network for predicting the critical cooling rate of amorphization of a material, comprising: at least one storage medium storing at least one instruction set; and at least one processor communicatively connected to the at least one storage medium, wherein when the training system is running, the at least one processor reads the at least one instruction set and executes the training method described in the second aspect according to the instructions of the at least one instruction set.

[0012] In a sixth aspect, this specification also provides a prediction system configured to predict the eutectic point of a multi-element system, comprising: at least one storage medium storing at least one instruction set; and at least one processor communicatively connected to the at least one storage medium, wherein when the prediction system is running, the at least one processor reads the at least one instruction set and executes the prediction method described in the third aspect according to the instructions of the at least one instruction set.

[0013] Other functions of the prediction method, training method and system provided in this specification will be partially listed in the following description. The creative aspects of the prediction method, training method and system provided in this specification can be fully explained by practicing or using the methods, systems and combinations described in the following detailed examples. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of this specification, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0015] Figure 1 Schematic diagram of a TTT curve provided according to an embodiment of this specification is shown;

[0016] Figure 2 A schematic diagram of a hierarchical structure of a target network provided according to an embodiment of this specification is shown;

[0017] Figure 3 A schematic diagram of the structure of a sub-network provided according to an embodiment of this specification is shown;

[0018] Figure 4 shows a hardware structure diagram of a system provided according to an embodiment of this specification;

[0019] Figure 5 A flowchart of a target network training method provided according to an embodiment of this specification is shown;

[0020] Figure 6FIG2 shows an iterative schematic diagram of a genetic algorithm provided according to an embodiment of this specification;

[0021] Figure 7 A schematic diagram illustrating an inference process of a target network provided according to an embodiment of this specification is shown;

[0022] Figure 8 A schematic diagram showing prediction results of different network models provided according to an embodiment of this specification is shown;

[0023] Figure 9 A schematic diagram showing prediction results of different element systems using a target network according to an embodiment of this specification is shown;

[0024] Figure 10 A flowchart showing a method for predicting a critical cooling rate of amorphization provided in accordance with an embodiment of this specification; and

[0025] Figure 11 A flowchart of a method for predicting a eutectic point according to an embodiment of this specification is shown. DETAILED DESCRIPTION

[0026] The following description provides specific application scenarios and requirements for this specification, with the goal of enabling those skilled in the art to make and use the contents of this specification. Various modifications to the disclosed embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of this specification. Therefore, this specification is not limited to the embodiments shown, but is intended to be accorded the broadest scope consistent with the claims.

[0027] The terms used herein are for the purpose of describing specific example embodiments only and are not intended to be limiting. For example, as used herein, the singular forms "a," "an," and "the" may also include the plural forms unless the context clearly indicates otherwise. When used in this specification, the terms "comprise," "include," and / or "contain" are intended to refer to the presence of the associated integers, steps, operations, elements, and / or components, but do not preclude the presence of one or more other features, integers, steps, operations, elements, components, and / or groups or the addition of other features, integers, steps, operations, elements, components, and / or groups in the system / method.

[0028] These and other features of this specification, as well as the operation and function of the associated elements of the structure, and the economical assembly and manufacture of the components, can be significantly improved with consideration of the following description. Reference is made to the accompanying drawings, all of which form a part of this specification. However, it should be expressly understood that the drawings are for illustration and description purposes only and are not intended to limit the scope of this specification. It should also be understood that the drawings are not drawn to scale.

[0029] The flowcharts used in this specification illustrate operations implemented by systems according to some embodiments of the present specification. It should be clearly understood that the operations of the flowcharts may not be implemented in sequence. Rather, the operations may be implemented in reverse order or simultaneously. Furthermore, one or more additional operations may be added to the flowcharts. One or more operations may be removed from the flowcharts.

[0030] In the past, researchers have proposed various methods to determine the critical cooling rate (R c For example, Johnson's theory proposed that the critical cooling rate of amorphization is related to the critical size of the material.

[0031] Figure 1 Figure (a) shows the relationship between the theoretical and experimental values of the critical cooling rate for amorphization, derived from Johnson's theory. The blue dots represent the theoretical values of the critical cooling rate before correction, while the red dots represent the theoretical values after correction using the coefficient. Overall, there is still a significant gap between the theoretical values before and after correction and the experimental values of the critical cooling rate for amorphization. Therefore, this theory cannot accurately determine the critical cooling rate for a material.

[0032] In addition, researchers can also draw the time-temperature-transformation (TTT) curve of the material, such as Figure 1 b), c) and d) in the figure to determine the critical cooling rate of amorphization of the material. For example, the critical cooling rate of amorphization can be calculated by the following formula:

[0033] R c =(T m -T n ) / t n

[0034] Where: T m is the melting point temperature, T n is the nose tip temperature, t n is the nucleation time on the TTT curve. The parameters in the formula can be obtained based on the TTT curve.

[0035] Although TTT curves are important for understanding the formation process of amorphous materials, their construction requires precise isothermal crystallization behavior measurements over a wide temperature range. This requires precise control of experimental conditions such as temperature and time, and the measurements must be performed over a wide temperature range. This requires a large number of experiments, with independent isothermal crystallization measurements performed at each temperature point. This makes the experiments complex and difficult, hindering the rapid study of amorphous materials.

[0036] In recent years, machine learning (ML) technology has rapidly developed in the field of materials science. These models can learn patterns from large amounts of experimental data, constructing mappings between the material properties under investigation and material attributes (e.g., descriptors), thereby predicting relevant material properties. Descriptors are information that describes various properties or characteristics of a material.

[0037] Typically, the predictive performance of machine learning models is highly dependent on the quantity and quality of training samples. For a given sample material, the more descriptors describing the sample material, the more likely it is to train a high-performance prediction network. However, a greater number of descriptors requires more experimental data on the critical cooling rate of amorphization of the sample material. This data is difficult to obtain and requires significant time and effort. Therefore, how to overcome the difficulty of training a prediction network to meet expectations while ensuring the accuracy of the prediction results due to insufficient samples and a large number of material descriptors remains a challenge in current research.

[0038] The following combination Figure 2 The application scenario of predicting the critical cooling rate of amorphization provided in this manual is introduced.

[0039] Figure 2 FIG. 1 shows a schematic diagram of an application scenario for predicting the critical cooling rate of amorphization according to an embodiment of this specification. Figure 2 As shown, scenario 001 may include a training system 100 and a prediction system 200 .

[0040] Scenario 001 can be divided into two stages: training stage and prediction stage.

[0041] Training phase

[0042] The training system 100 can train a target network capable of predicting the critical cooling rate of amorphization of a material based on a plurality of sample materials. After the training of the training system 100 is completed, the target network can be deployed in the prediction system 200.

[0043] Prediction stage

[0044] The prediction system 200 can call the target network to predict the critical cooling rate of amorphization of the target material. In addition, the prediction system 200 can also be used to predict the eutectic point of a multi-element system.

[0045] Eutectic Point: The material with a specific element ratio in a multi-element system undergoes a eutectic reaction at a constant temperature (the material changes from liquid to solid state of each element at the same time). The specific element ratio is the eutectic point of the multi-element system. For example, the multi-element system is Zr-Cu-Al. If the ratio of Zr, Cu and Al in the material is 50:40:10 respectively, then a eutectic reaction can occur at a constant temperature. Then Zr 50 Cu 40 Al 10 It is the eutectic point in the Zr-Cu-Al system.

[0046] In some embodiments, the training system 100 may store data and instructions for implementing a training method for a target network, and may execute or be used to execute the data and instructions. In some embodiments, the training system 100 may include a hardware device with data information processing capabilities and the necessary programs required to drive the hardware device to operate.

[0047] In some embodiments, prediction system 200 may store data and instructions for implementing a method for predicting the critical cooling rate of amorphization or the eutectic point, and may execute or be used to execute such data and instructions. In some embodiments, prediction system 200 may include hardware capable of data information processing and the necessary programs to operate the hardware.

[0048] It is understandable that the training system 100 and the prediction system 200 may correspond to the same system or to different systems, and this specification does not impose any limitation on this.

[0049] It should be noted that the training system 100 can correspond to a single device or a device cluster, and this specification does not impose any restrictions on this. When the training system 100 corresponds to a single device, the target network training method can be executed entirely on the device. When the training system 100 corresponds to a device cluster, the target network training method can be executed in coordination on multiple devices corresponding to the device cluster, or the target network training method can be executed in other ways, and this specification does not impose any restrictions on this.

[0050] The prediction system 200 may correspond to a single device or a cluster of devices, and this specification does not impose any restrictions on this. When the prediction system 200 corresponds to a single device, the method for predicting the critical cooling rate of amorphization or the eutectic point may be executed entirely on the device. When the prediction system 200 corresponds to a cluster of devices, the method for predicting the critical cooling rate of amorphization or the eutectic point may be executed in coordination on multiple devices corresponding to the cluster, or the method for predicting the critical cooling rate of amorphization or the eutectic point may be executed in other ways, and this specification does not impose any restrictions on this.

[0051] The target network is a hierarchical neural network with a network structure such as Figure 3 As shown, there are F levels, namely the 1st to the Fth level networks. Each level network can include one or more sub-networks, and the number of sub-networks included in each level network gradually decreases from the 1st to the Fth level. For example, the 1st level network includes m1 sub-networks, the 2nd level network includes m2 sub-networks, m2 < m1, and so on. represents the first subnetwork in the first level, The first subnetwork in the second layer is represented by , and so on. The network in the Fth layer includes one subnetwork. The output of each subnetwork in the first layer serves as the input of each subnetwork in the second layer, and so on. The network structure of each subnetwork can be the same, and this specification does not limit this.

[0052] This specification does not limit the manner in which the number of subnetworks at each level decreases. For example, the number of subnetworks at each level may decrease in a regular manner, such as by a fixed amount, where the difference between the number of subnetworks at the first level and the number of subnetworks at the second level is the same as the difference between the number of subnetworks at the second level and the number of subnetworks at the third level. Alternatively, the number of subnetworks at each level may decrease in an irregular manner, such as by a non-fixed amount.

[0053] It should be noted that this specification does not limit the number of network layers of the target network and the number of sub-networks included in each layer.

[0054] Exemplarily, the sub-network is an artificial neural network (ANN) or a convolutional neural network (CNN). This specification does not limit the network type of the sub-network.

[0055] Figure 4 4 shows a hardware structure diagram of a system 400 provided according to an embodiment of this specification.

[0056] like Figure 4 As shown, the system 400 can be Figure 1 The training system 100 or prediction system 200 in .

[0057] The system 400 includes at least one storage medium 430 and at least one processor 420. In some embodiments, the system 400 may further include a communication port 450 and an internal communication bus 410. In addition, the system 400 may further include an I / O component 460.

[0058] The internal communication bus 410 can connect various system components. For example, the internal communication bus 410 can connect the storage medium 430 , the processor 420 , the communication port 450 , and the I / O component 460 .

[0059] I / O components 460 support input / output between the system 400 and other components.

[0060] The communication port 450 is used for data communication between the system 400 and the outside world. For example, the communication port 450 can be used for data communication between the system 400 and a network. The communication port 450 can be a wired communication port or a wireless communication port.

[0061] In some embodiments, the network can be any type of wired or wireless network, or a combination thereof. For example, the network can include a cable network, a wired network, a fiber optic network, a telecommunications network, an intranet, the Internet, a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), a public switched telephone network (PSTN), a Bluetooth™ network, a short-range wireless network (ZigBee™), a near field communication (NFC) network, or the like.

[0062] In some embodiments, the network may include one or more network access points. For example, the network may include a wired or wireless network access point, such as a base station or an Internet exchange point. Through the access point, one or more components of each device corresponding to the system 400 can connect to the network to exchange data or information.

[0063] The storage medium 430 may include a data storage device. The data storage device may be a non-transitory storage medium or a temporary storage medium. For example, the data storage device may include one or more of a disk 432, a read-only storage medium (ROM) 434, or a random access storage medium (RAM) 436. The storage medium 430 also includes at least one instruction set stored in the data storage device. The instruction set may include a computer program code, which may include a program, routine, object, component, data structure, process, module, etc. for executing the method for predicting the critical cooling rate of amorphization or the eutectic point or the training method of the target network provided in this specification.

[0064] The processor 420 can be in communication with the storage medium 430. The processor 420 is configured to execute the at least one instruction set. When the system 400 is running, the processor 420 reads the at least one instruction set and, in accordance with the instructions of the at least one instruction set, executes the method for predicting the critical cooling rate of amorphization or the eutectic point, or the method for training the target network, as provided in this specification.

[0065] The processor 420 may be in the form of one or more processors. In some embodiments, the processor 420 may include one or more hardware processors, such as a microcontroller, a microprocessor, a reduced instruction set computer (RISC), an application-specific integrated circuit (ASIC), an application-specific instruction set processor (ASIP), a central processing unit (CPU), a graphics processing unit (GPU), a physical processing unit (PPU), a microcontroller unit, a digital signal processor (DSP), a field programmable gate array (FPGA), an advanced RISC machine (ARM), a programmable logic device (PLD), any circuit or processor capable of performing one or more functions, or any combination thereof.

[0066] For illustrative purposes only, only one processor 420 is shown in the system 400 in the accompanying drawings. However, it should be noted that the system 400 described in this specification may also include multiple processors. Therefore, the operations and / or method steps disclosed in this specification may be performed by one processor as described in this specification, or may be performed jointly by multiple processors. For example, if the processor 420 of the system 400 described in this specification performs step A and step B, it should be understood that step A and step B may also be performed jointly or separately by two different processors 420 (e.g., the first processor performs step A, the second processor performs step B, or the first and second processors perform steps A and B together).

[0067] Figure 5 FIG1 shows a flow chart of a target network training method P500 provided according to an embodiment of the present specification. The training system 100 can execute the target network training method P500, wherein the target network is used to predict the critical cooling rate of amorphization of a material. Figure 5 As shown, the training method P500 of the target network includes the following steps.

[0068] S510: Obtaining experimental values of amorphization critical cooling rates of Y sample materials, each sample material including at least one element.

[0069] Wherein, Y is an integer greater than 1. Different sample materials may include different numbers of elements and different proportions of elements. For example, the sample material is an alloy material. For example, the sample material Ti 63 Be 37For example, 63 is the proportion information corresponding to the element Ti, and 37 is the proportion information corresponding to the element Be.

[0070] This manual collects experimental data corresponding to 119 metallic glasses, and after cleaning them, obtains the experimental values of the critical cooling rate of amorphization for each of Y sample materials.

[0071] S520: For each sample material, based on information about a proportion of at least one element in the sample material and attribute values of X attributes corresponding to each of the at least one element, generate K descriptors for describing the sample material, and determine N1 core descriptors and K1 non-core descriptors from the K descriptors. The core descriptors have a greater impact on the critical cooling rate of amorphization than the non-core descriptors, where K1 = K - N1.

[0072] In some embodiments, the X attributes include at least two of the following dimensions: the periodic table dimension; the thermodynamic dimension; the physics dimension; and the crystallographic dimension. In this embodiment, by considering multiple dimensions of elemental properties, the resulting K descriptors can more comprehensively characterize the sample material, covering a wide range of material properties and avoiding omission of important descriptors, thus providing a good foundation for improving the prediction accuracy and generalization ability of the target network.

[0073] Illustratively, the X attributes refer to at least part of Table 1.

[0074] Table 1

[0075]

[0076]

[0077] In Table 1, BCC refers to body-centered cubic lattice, FCC refers to face-centered cubic lattice, ICSD refers to the Inorganic Crystal Structure Database, and GS refers to the ground state of the object.

[0078] For example, the sample material is Zr 55 Al 19 Co 19 Cu7, describing Zr 55 Al 19 Co 19 The K descriptors of Cu7 can be generated based on the following: the proportion of Zr element The corresponding proportion of Al element The proportion of Co element The proportion of Cu element And the attribute values of at least part of the attributes listed in Table 1 corresponding to the Zr element, the attribute values of at least part of the attributes corresponding to the Al element, the attribute values of at least part of the attributes corresponding to the Co element, and the attribute values of at least part of the attributes corresponding to the Cu element.

[0079] The following takes sample material #1 as an example to introduce the generation process of K descriptors describing sample material #1. The K descriptors describing other sample materials can be generated in the same way.

[0080] In some embodiments, the K descriptors describing sample material #1 include at least one of the following:

[0081] The first set of descriptors is used to describe the statistical characteristics of sample material #1 on the X attributes;

[0082] The second set of descriptors is used to describe the valence electron occupation state of sample material #1;

[0083] A third set of descriptors, used to describe the ionicity of sample material #1; and

[0084] The fourth set of descriptors is used to describe the configurational entropy (CE) of sample material #1.

[0085] In this embodiment, K descriptors can describe different aspects of the sample material's characteristics, providing a comprehensive description of the sample material, supporting the prediction of complex multi-element materials, and improving the prediction accuracy and generalization capability of the target network.

[0086] Exemplarily, the first set of descriptors is generated as follows: for each of the X attributes, the training system 100 performs statistical analysis of the characteristics of sample material #1 for the attribute from multiple statistical dimensions based on the proportion of each element included in sample material #1 and the attribute value of each element for the attribute, obtaining P statistical values. The first set of descriptors is then generated based on the P statistical values corresponding to each of the X attributes of sample material #1. The number of descriptors in the first set is equal to X*P. In this example, the first set of descriptors involves multiple statistical dimensions, and both the number of descriptors and the statistical dimensions involved are relatively large. This exploits the basic statistical characteristics of the sample material, accurately and comprehensively characterizing the sample material, and facilitating the training of the target network.

[0087] Taking attribute #1 as an example, training system 100 can generate P statistical values for the characteristics of sample material #1 with respect to attribute #1 from multiple statistical dimensions. These P statistical values represent all descriptors corresponding to attribute #1 in the first set of descriptors. The following describes the generation process for all descriptors corresponding to attribute #1 in the first set of descriptors. The same process applies to descriptors corresponding to other attributes in the first set of descriptors and is not further elaborated.

[0088] For example, the multiple statistical dimensions include statistical dimension #1, statistical dimension #2, statistical dimension #3, statistical dimension #4, and statistical dimension #5. Statistical dimension #1 is used to calculate the weighted average of the attribute values of each element included in sample material #1 on attribute #1. The formula corresponding to statistical dimension #1 is as follows:

[0089]

[0090] Among them, x i represents the proportion of the i-th element in the chemical formula of sample material #1, f i Indicates the attribute value of the i-th element on attribute #1, Indicates the descriptor corresponding to attribute #1 obtained from statistical dimension #1. Assume that sample material #1 is Zr 55 Al 19 Co 19 Cu7, the training system 100 can determine based on formula (1)

[0091] Statistical dimension #2 is used to calculate the weighted deviation of the attribute values of each element in sample material #1 on attribute #1. The formula corresponding to statistical dimension #2 is as follows:

[0092]

[0093] in, Indicates the descriptor corresponding to attribute #1 obtained from statistical dimension #2.

[0094] Statistical dimension #3 is used to calculate the standard deviation of the attribute values of each element in sample material #1 on attribute #1. The formula corresponding to statistical dimension #3 is as follows:

[0095]

[0096] Here, f represents the descriptor corresponding to attribute #1 obtained from statistical dimension #3.

[0097] Statistical dimension #4 is used to calculate the maximum, minimum, and value range of the attribute values or weighted attribute values of each element in attribute #1 for sample material #1. The formula corresponding to statistical dimension #4 is as follows:

[0098] (f i )_min,max,range (4)

[0099] (x i *f i )_min,max,range (5)

[0100] The training system 100 can obtain six descriptors corresponding to attribute #1 from statistical dimension #4, which are as follows:

[0101] (f i )_min represents the minimum value among the attribute values of each element included in the sample material #1 on attribute #1; (f i )_max represents the maximum value of the attribute values of each element included in the sample material #1 on the attribute #1; (f i )_range represents the value range of each element included in the sample material #1 in the attribute value of attribute #1, which can be (f i )_max and (f i )_min difference. (x i *f i )_min represents the minimum value among the weighted attribute values of each element included in sample material #1 on attribute #1; (x i *f i )_max represents the maximum value of the weighted attribute values of each element included in sample material #1 on attribute #1; (x i *f i )_range represents the value range of the weighted attribute value of each element included in sample material #1 on attribute #1, which can be (x i *f i )_max and (x i *f i )_min difference.

[0102] Statistical dimension #5 is used to calculate the degree of deviation of different elements in attribute #1 among all elements included in sample material #1. The formula corresponding to statistical dimension #5 is as follows:

[0103]

[0104] Among them, x i1 represents the proportion of the i1th element in the chemical formula of sample material #1, x i2 represents the proportion of the i2th element in the chemical formula of sample material #1, f i1 Indicates the attribute value of the i1th element on attribute #1, f i2The AP represents the attribute value of the i2th element on attribute #1, and the AP represents the degree of deviation of different elements on attribute #1. 加权 Indicates the degree of deviation of the weighted attribute values of different elements on attribute #1.

[0105] The training system 100 can obtain eight descriptors corresponding to attribute #1 from statistical dimension #5, which are as follows:

[0106] AP_ave represents the average value of the deviations of all elements in sample material #1 on attribute #1; AP_min represents the minimum value of the deviations of all elements in sample material #1 on attribute #1; AP_max represents the maximum value of the deviations of all elements in sample material #1 on attribute #1; AP_range represents the range of the deviations of all elements in sample material #1 on attribute #1, which can be the difference between AP_max and AP_min; AP 加权 _ave represents the average value of the weighted attribute value deviations of different elements on attribute #1 among all elements included in sample material #1; AP 加权 _min represents the minimum value of the weighted attribute value deviation of different elements on attribute #1 among all elements included in sample material #1; AP 加权 _max represents the maximum value of the weighted attribute value deviation of different elements on attribute #1 among all elements included in sample material #1; AP 加权 _range represents the value range of the weighted attribute value deviation of different elements in attribute #1 among all elements included in sample material #1, which can be AP 加权 _max and AP 加权 _min difference. AP, AP 加权 The training system 100 considers AP, AP and AP when generating the first set of descriptors. 加权 It is very important to describe the material comprehensively. In addition, if sample material #1 consists of a single element, the above 8 descriptors are all zero.

[0107] Thus, training system 100 can obtain 17 descriptors corresponding to attribute #1 using the above multiple statistical dimensions. If the X attributes are 53 as shown in Table 1, the first set of descriptors obtained by training system 100 for sample material #1 can include 17 * 53 = 901 descriptors, fully covering the basic statistical attributes of sample material #1.

[0108] In some embodiments, the second set of descriptors includes: occupation information of Q valence electrons in sample material #1, wherein the occupation information of the zth valence electron is obtained as follows:

[0109] For each element in sample material #1, the training system 100 determines the number of each type of valence electron and the total number of valence electrons in the element based on the attribute values of at least some of the X attributes corresponding to the element; for the zth type of valence electron, the training system 100 determines the occupancy information of the zth type of valence electron based on the corresponding proportion information of each element in sample material #1, the number of the zth type of valence electron, and the total number of valence electrons corresponding to each element.

[0110] In other words, each descriptor in the second set of descriptors is used to describe the valence electron occupation state. The second set of descriptors obtained by training system 100 includes Q descriptors, one for each valence electron. Valence electrons play a significant role in the preparation of amorphous materials. Therefore, considering the occupation states of multiple valence electrons in the descriptors can characterize key properties of the sample material and improve the target network's prediction accuracy of the critical cooling rate of amorphization.

[0111] Exemplarily, the training system 100 obtains the descriptor corresponding to the z-th valence electron based on the following formula:

[0112]

[0113] Among them, F z Indicates the descriptor corresponding to the zth valence electron, E z represents the number of z-th valence electrons in the i-th element in the chemical formula of sample material #1, E n represents the total number of Q valence electrons in the i-th element, such as valence electrons p, s, d, and f. Specifically, the descriptor corresponding to the valence electron p is The descriptors corresponding to other valence electrons are deduced in the same way, so that the training system 100 can obtain 4 descriptors.

[0114] In some embodiments, the third set of descriptors is derived based on the percentage information of each element in sample material #1, the electronegativity of each element, and any of the following: the maximum electronegativity of each element in sample material #1; or the average electronegativity of each element in sample material #1. Electronegativity is a measure of the ability of an element's atoms to attract electrons in a compound. Therefore, incorporating electronegativity into descriptors can characterize a key property of the sample material and improve the accuracy of the target network's prediction of the critical cooling rate for amorphization.

[0115] Exemplarily, the training system 100 obtains the third set of descriptors based on the following formula:

[0116]

[0117] Among them, I represents the third group of descriptors, represents the maximum value of the electronegativity of each element included in the sample material #1, or the average value of the electronegativity of each element included in the sample material #1; When represents the maximum value of electronegativity, the training system 100 can obtain the first descriptor in the third group of descriptors based on formula (9); when When represents the average value of electronegativity, the training system 100 can obtain the second descriptor in the third group of descriptors based on formula (9); when When representing the average value of electronegativity, the training system 100 can also obtain the third descriptor in the third group of descriptors. Specifically, when the I calculated by the training system 100 based on formula (9) exceeds 1.7, the third descriptor is 1. If the calculated I does not exceed 1.7, the third descriptor is 0 (this is also the standard for determining the formation of ionic bonds).

[0118] In some embodiments, the fourth set of descriptors is derived based on the elemental composition information, the Boltzmann constant, and the room temperature in sample material #1. Configuration entropy is an important parameter that measures the degree of atomic disorder and is closely related to the critical cooling rate of amorphous materials. Considering configuration entropy when generating descriptors can characterize key properties of the sample material and improve the target network's prediction accuracy for the critical cooling rate of amorphization.

[0119] Exemplarily, the training system 100 obtains the fourth set of descriptors based on the following formula:

[0120] ΔS con =-k B T∑x i *lnx i (10)

[0121] Where, ΔS con represents configuration entropy; k B represents the Boltzmann constant; T represents room temperature, such as 298 K. The training system 100 can obtain a descriptor based on formula (10).

[0122] In summary, when there are 53 types of element attributes, for each sample material, the training system 100 generates 901 first group descriptors, 4 second group descriptors, 3 third group descriptors, and 1 fourth group descriptor based on formulas (1) to (10), that is, the total number of K descriptors is 909.

[0123] In the field of materials science, experimentally determining the critical cooling rate of amorphization requires significant time and effort. Therefore, when a small number of sample materials are available, training a neural network using 909 descriptors as features of a sample material can easily lead to overfitting. This is especially true when the sample material contains a large number of elements, making it even more difficult to train a target network with relatively accurate predictions. For example, in an ANN with two hidden layers, inputting 909 descriptors would require training to determine over a billion hyperparameters, making it difficult to achieve the desired training results when the number of sample materials is small. Therefore, this specification utilizes a genetic algorithm (GA) to reduce the number of input descriptors in a single subnetwork within the target network.

[0124] Exemplarily, the training system 100 groups the K descriptors based on a competitive swarm optimizer (CSO) algorithm, with the descriptors in each group competing in pairs. After each competition, the winning descriptor enters the next iteration, and the losing descriptor can learn from the winning descriptor to update its position and velocity. The basic principle of the CSO algorithm can be seen in the following formula:

[0125]

[0126]

[0127] Where t represents the number of iterations; and is in [0,1] n Three vectors randomly generated in the range; for each descriptor, '1' indicates that the descriptor is selected after competition, and '0' indicates that the descriptor is not selected after competition; 'n' indicates the number of descriptors, such as 'n' is 909; the variable and Represent the winning descriptor and the losing descriptor respectively, the variable represents the average position of the descriptor group in the tth iteration, and φ determines influence, represents the velocity of the i-th descriptor in the t-th iteration.

[0128] In this example, the training system 100 uses the 1-R 2As the fitness function, the fitness function is minimized by strategically selecting descriptors. Specifically, at the beginning of the iteration, the training system 100 randomly selects multiple descriptor combinations and their speeds, and then randomly selects two descriptor combinations from among them for fitness function evaluation. The descriptor combination with the lower fitness function value is regarded as the winner, and the descriptor combination with the higher fitness function value is regarded as the loser. During the iteration process, updates and iterations are performed according to formulas (11) and (12). When the fitness function value no longer decreases, the iteration process terminates.

[0129] Figure 6 The GA iteration process is illustrated. The x-axis represents the number of GA iterations, the left y-axis represents the number of descriptors (d), and the right y-axis represents the test R 2 During the GA iteration process, multiple preferred descriptor combinations may appear, and the difference between the predicted value and the experimental value of the amorphization critical cooling rate of the sample material predicted by the preferred descriptor combination is less than or equal to a first preset threshold.

[0130] For example, see Figure 6 After iterative selection, the training system 100 determines that when N is 25, the prediction accuracy reaches a turning point, and the test R 2 That is to say, when the number of descriptors input to each sub-network is 25, the critical cooling rate of amorphization of the material can be predicted relatively accurately.

[0131] It should be noted that the determination of N may be related to the number of sample materials participating in GA, so this specification does not limit the specific value of N.

[0132] In addition, during the iteration process, descriptors whose appearance frequency in multiple preferred descriptor combinations is greater than or equal to the second preset threshold are used as core descriptors required for predicting the critical cooling rate of amorphization, and descriptors whose appearance frequency is less than the second preset threshold are used as non-core descriptors required for predicting the critical cooling rate of amorphization.

[0133] Table 2 gives the six core descriptors required to predict the critical cooling rate of amorphization.

[0134] Table 2

[0135]

[0136]

[0137] For example, AP_ave_GSvolume_pa represents the average value of the deviation of the ground state atomic volume of different elements in the sample material, which can be obtained based on formula (6). AP_max_weighted_Atomic_Number represents the maximum value of the weighted deviation of the atomic number of different elements in the sample material, which can be obtained based on formula (7). std_phi represents the standard deviation of the property values of each element in the sample material on the work function, which can be obtained based on formula (3). std_Melting T represents the standard deviation of the property values of each element in the sample material on the melting point, which can be obtained based on formula (3). range_weight_HeatCapacityMolar represents the range of values of the weighted property values of different elements in the sample material on the molar heat capacity, which can be obtained based on formula (5). AP_ave_phi represents the average value of the deviation of the work function of different elements in the sample material, which can be obtained based on formula (6).

[0138] This specification uses a genetic algorithm to screen 25 key descriptors from 909 descriptors and further refines 6 core descriptors, significantly reducing the training complexity of the target network while retaining descriptors that are crucial for predicting the critical cooling rate of amorphization. Furthermore, existing research indicates that the formation process of amorphous materials is closely related to properties such as atomic volume. The nonlinear relationship between these six core descriptors and the critical cooling rate of amorphization reveals the complex physical mechanisms underlying the formation of amorphous materials and is consistent with the formation mechanisms of amorphous materials in existing research. This provides a theoretical basis for further research on amorphous materials and plays a key role in the prediction accuracy of the target network.

[0139] S530: Inputting some of the K descriptors corresponding to each sample material into a target network for prediction, thereby obtaining a predicted value corresponding to the critical cooling rate of amorphization of the sample material. The target network includes multiple subnetworks with the same structure and arranged in a hierarchical manner, each subnetwork including N input nodes and 1 output node. The output node of the subnetwork in the i-th level serves as the input node of the subnetwork in the i+1-th level. Each subnetwork in the first level receives the N1 core descriptors and N2 non-core descriptors as input. The N2 non-core descriptors are part of the K1 non-core descriptors, and the non-core descriptors input to different subnetworks are not completely the same, where N2 = N-N1.

[0140] Where X, K, K1, N1, N2, Y, and N are all integers greater than 1, and i is an integer greater than or equal to 1. Using the target network can avoid overfitting when there are many descriptors but insufficient sample material. It should be noted that each subnetwork can have different weighting factors during training.

[0141] In some embodiments, see Figure 7 , the training system 100 generates m1 groups of descriptors based on the K descriptors, each group of descriptors includes the N1 core descriptors and N-N1 non-core descriptors; the m1 groups of descriptors are respectively input into the m1 sub-networks of the first layer of the target network to obtain m1 intermediate prediction values of the critical cooling rate of amorphization; according to the order of i from 2 to F, the m1 sub-networks of the i-1 layer of the target network are respectively input into the m1 sub-networks of the first layer of the target network to obtain m1 intermediate prediction values of the critical cooling rate of amorphization; i-1 The critical cooling rate m of amorphization obtained by the sub-network i-1 The intermediate prediction values are divided into m i After grouping, input the target network layer i m i The critical cooling rate of amorphization is m i The intermediate prediction value of the amorphization critical cooling rate output by the sub-network of the Fth layer is then used as the final prediction value of the amorphization critical cooling rate. i-1 are all integers greater than 1, m i is an integer greater than or equal to 1.

[0142] In this embodiment, K descriptors are divided into multiple groups and input into multiple sub-networks, which can avoid the problem of overfitting of the sub-network due to the excessive number of hyperparameters required to train the sub-network due to insufficient sample materials. Figure 7 The hierarchical structure of the target network shown here allows descriptors to be distributed across multiple subnetworks, preventing overfitting. This allows the number of descriptors to be expanded to 909, enriching the material characterization dimensionality. Each subnetwork requires only a small number of input descriptors, resolving the machine learning paradox of a large number of descriptors and a small number of sample materials. Furthermore, testing (see below for details) has shown that the target network can fully utilize information from a large number of descriptors, achieving higher prediction accuracy than other neural networks, facilitating the study of amorphous materials.

[0143] For example, when K is 909, the training system 100 can obtain 909-6=903 non-core descriptors by random combination. Non-core descriptor groups, each non-core descriptor group plus 6 core descriptors can be obtained descriptor groups. That is, the number of subnetworks in the first level can reach However, this specification does not limit the number of first-level neutron networks.

[0144] Different descriptor groups play different roles in predicting the critical cooling rate of amorphization. This role can be determined during the training of the target network. Different input descriptor groups for each sub-network in the first layer can improve the generalization ability of the target network and reduce the possibility of overfitting. Compared with the traditional method of training a network with a single descriptor combination, the training method of this specification takes into account various descriptor combinations and can capture or explore the potential impact of different descriptor combinations on predicting the critical cooling rate of amorphization, thereby training a target network with higher prediction accuracy. Traditional algorithms, such as XGBoost and CNN, do not take into account the differences in the contributions of different descriptor groups to the prediction of the critical cooling rate of amorphization, nor do they consider the differences in the performance of the target network caused by different descriptor groups.

[0145] For the first level of the target network, the input of each sub-network is a descriptor group, and each descriptor group includes N (for example, 25) descriptors; that is, an input node of each sub-network in the first level corresponds to a descriptor. The input of is descriptor group 1, and so on. Each descriptor group includes N1 (for example, 6) core descriptors and N2 (for example, 19) non-core descriptors. Different descriptor groups include at least one different non-core descriptor. For the second level, the input of each sub-network is the intermediate prediction value output by the N sub-networks in the first level, such as The input of the third layer is the output of N second layer sub-networks. to The N intermediate prediction values output are to The intermediate prediction value of , and so on. For the Fth level, the sub-network F The input is the intermediate prediction value of the N F-1th layer neutron network output.

[0146] During training, the target network learns from the iterative knowledge of different subnetworks. This knowledge can include both overlapping and unique knowledge between subnetworks, thereby improving prediction accuracy. Given the limited number of sample materials, training a single neural network using 909 descriptors avoids the excessive number of hyperparameters that can affect training accuracy. As the number of target network layers increases, the predicted value for the critical cooling rate of amorphization gradually converges.

[0147] For example, in the process of training the target network, the core descriptors input by each sub-network in the first level are, for example, MD1 to MD6, and the outputs of each sub-network after the first level are predicted values of the critical cooling rate of amorphization.

[0148] S540: updating the parameters of the target network with minimizing the difference between the predicted value and the experimental value of the critical cooling rate of amorphization of the sample material as a training goal.

[0149] For example, the present specification may use the coefficient of determination (R 2 ) and mean absolute percentage error (MAPE) as the evaluation indicators of the target network. MAPE and R 2 The calculation formula is as follows:

[0150]

[0151]

[0152] Among them, n represents the number of sample materials in the current training round, y i represents the experimental value of the critical cooling rate of amorphization for the i-th sample material, represents the average experimental value of the critical cooling rate of amorphization of the sample material, F(x i ) represents the predicted value of the critical cooling rate of amorphization of the i-th sample material predicted by the target network. R 2 A higher value and a lower MAPE value indicate a better performance of the target network, that is, a more accurate prediction result.

[0153] Exemplarily, the training system 100 divides the Y sample materials into a training set and a test set, and trains the following network models, including the target network: 145-XGBoost, 909-XGBoost, 145-CNN, 909-CNN, 145-HNN, and 909-HNN. The numbers represent the number of descriptors input to the network model, and HNN represents the target network.

[0154] For example, "145-XGBoost" means that the network to be trained is XGBoost and the input of XGBoost is the 145 descriptors currently used by other machine learning models; "909-XGBoost" means that the network to be trained is XGBoost and the input of XGBoost is the 909 descriptors provided in this specification; "145-CNN" means that the network to be trained is CNN and the input of CNN is the 145 descriptors; "909-CNN" means that the network to be trained is CNN and the input of CNN is the 909 descriptors; "145-HNN" means that the network to be trained is the target network and the input of the target network is the 145 descriptors; "909-HNN" means that the network to be trained is the target network and the input of the target network is the 909 descriptors.

[0155] Figure 8 The learning results of each of the above network models are shown.

[0156] For example, the horizontal axis in a) represents different network models, and the vertical axis on the left represents the test R 2 b) shows the training and test results of “909-CNN”. c) shows the training and test results of “909-HNN”. 2 The R of the network model for the sample materials in the training set c The relationship between the predicted value and the experimental value; test R 2 The R of the network model for the sample material in the test set c The relationship between the predicted and experimental values.

[0157] See also Figure 8 It can be seen that when the traditional CNN model uses 909 descriptors, the test set R 2 It is only 0.58, while the HNN model has a test set R 2 It reaches 0.942, and the prediction effect is significantly better than other models. In addition, the CNN model is prone to overfitting under high-dimensional descriptors (such as 909 descriptors), while the target network provided in this specification introduces a hierarchical structure and decomposes the complex neural network into multiple sub-networks, which effectively avoids the problem of overfitting and can be optimized layer by layer during training, significantly improving the prediction accuracy. And while ensuring the prediction accuracy, since the number of descriptors input to each sub-network is greatly reduced, compared with training the traditional network model with 909 descriptors, the parallel training of the hierarchical structure fully utilizes the computing resources of each sub-model, and the training time can be reduced by 30%-40%, thereby improving the prediction efficiency.

[0158] The target network trained in this specification can be applied to interpolated predictions of multi-element systems included in sample materials. For example, using the multi-element Zr-Cu-Al system, the target network can predict the critical cooling rate of amorphization for different materials with varying percentages of Zr, Cu, and Al.

[0159] Furthermore, the target network trained in this specification can also be used for extrapolating predictions for multi-element systems beyond those found in the experimental data. For example, if the Ge-Sb-Te system is absent from the sample material, the target network can predict the critical cooling rate for amorphization of different materials with varying proportions of Ge, Sb, and Te.

[0160] See also Figure 9, where a), c) and d) show the prediction results of the critical cooling rate of amorphization for different element proportions in the Zr-Cu-Al system. b) shows the prediction results of the critical cooling rate of amorphization for different element proportions in the Ge-Sb-Te system. From the prediction results shown in a), we can derive the area with the lowest predicted value of the critical cooling rate of amorphization (the circled area in the figure), as well as the proportion information of each element included in the material corresponding to this area. The critical cooling rate of amorphization can quantitatively characterize the glass forming ability (GFA), and the GFA of the material is inversely correlated with the critical cooling rate of amorphization. The proportion information of the elements in the material corresponding to the area in a) is highly consistent with the proportion information of each element in the materials with higher GFA in the Zr-Cu-Al system in existing studies, verifying the prediction accuracy and wide applicability of the target network.

[0161] In addition, this specification also uses multi-element systems (such as Mg-Zn-Ca and Ca-Cu-Mg) that do not appear in the training and test sets for extrapolated prediction verification. Table 3 shows the comparison results of the experimental values and the predicted values of the target network. It can be seen that the predicted values of the target network deviate very little from the experimental values, with prediction errors less than 5%. Therefore, the accuracy of the prediction results of the target network in the extrapolated prediction is also relatively high, proving the reliability of the target network.

[0162] Table 3

[0163] Material <![CDATA[Experimental value R c (K / s)]]> <![CDATA[Predicted value R c (K / s)]]> error <![CDATA[Mg 68 Zn 25 Ca7]]> 3708.0 3480.3 4.9％ <![CDATA[Ca 50 With 25 Mg 25 ]]> 2509.8 2480.1 1.2％ <![CDATA[Ge2Sb2Te5]]> <![CDATA[1.9×10 9 ]]> <![CDATA[2.5×10 9 ]]> 3.2％ <![CDATA[Ca 50 With 27.5 Mg 22.5 ]]> 2923.1 3028.8 3.6％

[0164] In summary, the training method and training system of the target network provided in this specification expand the number of descriptors describing sample materials, enrich the description dimension of sample materials, and through multiple levels of target networks, K descriptors can be grouped for parallel training of sub-networks, avoiding the contradiction caused by too many descriptors and insufficient training samples. First, the significant increase in the number of descriptors combined with the target network can improve the prediction accuracy of the amorphous critical cooling rate. Even a small amount of training samples can significantly enhance the prediction ability of the target network, which can reduce the experimental cost in the research process of amorphous materials. Furthermore, when grouping descriptors, each group includes the same core descriptors and non-core descriptors that are not exactly the same. This takes into account the importance of the core descriptors and explores the potential impact of different non-core descriptor combinations on the amorphous critical cooling rate. Combined with the prediction performance data given above, the prediction accuracy of the target network is higher than that of other networks, which is conducive to accelerating the process of innovative research on amorphous materials.

[0165] Figure 10 1 shows a flowchart of a method P1000 for predicting amorphization critical cooling rate according to an embodiment of the present specification. The prediction system 200 can execute the method P1000 for predicting amorphization critical cooling rate. Figure 10 As shown, the prediction method P1000 for the critical cooling rate of amorphization includes the following steps.

[0166] S1010: Generate K descriptors for describing the target material based on information about the proportion of at least one element included in the target material to be predicted and attribute values of X attributes corresponding to each of the at least one element.

[0167] In some embodiments, the X properties include at least two of the following dimensions: a periodic table dimension; a thermodynamic dimension; a physics dimension; and a crystallographic dimension.

[0168] Specifically, the X attributes can be found in Table 1 and will not be described in detail.

[0169] In some embodiments, the K descriptors include at least one of the following:

[0170] The first set of descriptors is used to describe the statistical characteristics of the target material on the X properties;

[0171] The second set of descriptors is used to describe the valence electron occupation state of the target material;

[0172] A third set of descriptors, used to describe the ionicity of the target material; and

[0173] The fourth group of descriptors is used to describe the configuration entropy of the target material.

[0174] In some embodiments, the first set of descriptors is generated as follows:

[0175] For each of the X attributes, the prediction system 200 performs statistics on the target material's characteristics for the attribute from multiple statistical dimensions based on the proportion information of the at least one element in the target material and the attribute values of the at least one element for the attribute, obtaining P statistical values. The prediction system 200 also generates a first set of descriptors based on the P statistical values corresponding to each of the X attributes of the target material. The number of descriptors in the first set is equal to X*P.

[0176] Specifically, the method for generating the first group of descriptors corresponding to the target material is similar to the method for generating the first group of descriptors corresponding to the sample material #1. The specific generation method can be referred to above and will not be described in detail.

[0177] In some embodiments, the second set of descriptors includes: occupation information of Q valence electrons in the target material, wherein the occupation information of the zth valence electron is obtained as follows:

[0178] For each of the at least one element, the prediction system 200 determines the number of each type of valence electron and the total number of valence electrons in the element based on the attribute values of at least some of the X attributes corresponding to the element; for the zth type of valence electron, the prediction system 200 determines the occupancy information of the zth type of valence electron based on the proportion information of the at least one element in the target material, the number of the zth type of valence electron corresponding to each of the at least one element, and the total number of valence electrons corresponding to each of the at least one element.

[0179] Specifically, the method for generating the second set of descriptors corresponding to the target material is similar to the method for generating the second set of descriptors corresponding to the sample material #1. The specific generation method can be referred to above and will not be described in detail.

[0180] In some embodiments, the third set of descriptors is obtained based on information about the proportion of the at least one element in the target material, the electronegativity of each of the at least one element, and any one of the following:

[0181] The maximum value among the electronegativity corresponding to each of the at least one element;

[0182] The average value of the electronegativity corresponding to each of the at least one element.

[0183] Specifically, the method for generating the third group of descriptors corresponding to the target material is similar to the method for generating the third group of descriptors corresponding to the sample material #1. The specific generation method can be referred to above and will not be described in detail.

[0184] In some embodiments, the fourth set of descriptors is obtained based on information about the proportion of the at least one element in the target material, the Boltzmann constant, and room temperature.

[0185] Specifically, the generation method of the fourth group of descriptors corresponding to the target material is similar to the generation method of the fourth group of descriptors corresponding to the sample material #1. The specific generation method can be referred to above and will not be repeated here.

[0186] S1020: Determine N1 core descriptors and K1 non-core descriptors from the K descriptors, wherein the core descriptors have a higher impact on the critical cooling rate of amorphization than the non-core descriptors.

[0187] In some embodiments, the N1 core descriptors include multiple of the following:

[0188] MD1, characterizing the average value of the deviations of the ground state atomic volumes of different elements in the at least one element;

[0189] MD2, representing the maximum value of weighted deviations in atomic numbers of different elements in the at least one element;

[0190] MD3, representing the standard deviation of the property value of the at least one element on the work function;

[0191] MD4, characterizing the standard deviation of the property value of the at least one element at the melting point;

[0192] MD5, representing a value range of weighted attribute values of different elements in the at least one element in terms of molar heat capacity; and

[0193] MD6 represents an average value of deviations in work function between different elements in the at least one element.

[0194] Exemplarily, after the prediction system 200 determines the N1 core descriptors in MD1 to MD6 , it uses the remaining descriptors in the K descriptors as non-core descriptors.

[0195] S1030: Input at least some of the K descriptors into a target network for prediction to obtain a predicted value corresponding to the critical cooling rate of amorphization of the target material. The target network is a network trained using Y sample materials. The target network includes multiple sub-networks with the same structure and arranged in a hierarchical manner. Each sub-network includes N input nodes and 1 output node. The output node of the sub-network in the i-th level serves as the input node of the sub-network in the i+1-th level. Each sub-network in the first level is input with the N1 core descriptors and N2 non-core descriptors. The N2 non-core descriptors are part of the K1 non-core descriptors, and the non-core descriptors input to different sub-networks are not exactly the same.

[0196] In some embodiments, the prediction system 200 generates m1 groups of descriptors based on the K descriptors, each group of descriptors includes the N1 core descriptors and N-N1 non-core descriptors; the m1 groups of descriptors are respectively input into the m1 sub-networks of the first layer of the target network to obtain m1 intermediate prediction values of the critical cooling rate of amorphization; and the m1 sub-networks of the i-1th layer of the target network are input into the m1 sub-networks of the first layer of the target network in the order of i from 2 to F. i-1 The critical cooling rate m of amorphization obtained by the sub-network i-1 The intermediate prediction values are divided into m i After grouping, input the mth layer of the target network i The critical cooling rate of amorphization is m i and taking the intermediate prediction value of the amorphization critical cooling rate output by the sub-network of the Fth layer as the final prediction value of the amorphization critical cooling rate.

[0197] For example, the prediction system 200 may input the m1 group descriptors into the following Figure 7The m1 subnetworks of the first layer of the target network shown in the figure obtain m1 intermediate prediction values of the critical cooling rate of amorphization; according to the order of i from 2 to F, the m i-1 The critical cooling rate m of amorphization obtained by the sub-network i-1 The intermediate prediction values are divided into m i After grouping, input the target network T i-th layer m i The critical cooling rate of amorphization is m i The intermediate prediction value of the amorphization critical cooling rate output by the sub-network of the Fth layer is used as the final prediction value of the amorphization critical cooling rate.

[0198] It is understood that some technical details involved in the prediction method P1000 for the critical cooling rate of amorphization can be referred to the description of the target network training method P500, and will not be repeated here. The beneficial effects of each embodiment of the above method P1000 can also be referred to the beneficial effects of the embodiment of method P500, and will not be repeated here.

[0199] In summary, the prediction method and prediction system for the critical cooling rate of amorphization provided in this specification expand the number of descriptors of the target material, enrich the description dimension of the target material, and through multiple levels of target networks, K descriptors can be grouped and input into the sub-network for parallel prediction. First, the significant increase in the number of descriptors combined with the target network can improve the prediction accuracy of the critical cooling rate of amorphization of the target material, and can reduce the experimental cost in the research process of amorphous materials. Furthermore, when grouping the descriptors, each group includes the same core descriptors and non-core descriptors that are not completely the same. This takes into account the importance of the core descriptors and explores the potential impact of different non-core descriptor combinations on the critical cooling rate of amorphization of the target material. Combined with the prediction data given in the previous article, the prediction accuracy of the target network is higher than that of other networks, which is conducive to accelerating the process of innovative research on amorphous materials.

[0200] Furthermore, this specification uses the P1000 method for predicting the critical amorphization cooling rate for multiple materials in different multi-element systems. It was found that, for a given multi-element system, the material with the minimum predicted critical amorphization cooling rate is close to the eutectic point of that system. This provides important guidance for the design of amorphized materials (e.g., novel glass materials). Therefore, this specification also provides a method for predicting the eutectic point.

[0201] Figure 11 FIG2 shows a flowchart of a method P1100 for predicting a eutectic point according to an embodiment of the present specification. The prediction system 200 can execute the method P1100 for predicting a eutectic point. Figure 11 As shown, the eutectic point prediction method P1100 includes the following steps.

[0202] S1110: Predicting the critical cooling rate of amorphization of multiple materials formed by multiple elements in a target multi-element system through the prediction method P1000 for the critical cooling rate of amorphization, wherein each of the multiple materials includes the multiple elements, and the proportion information of the multiple elements in different materials is different.

[0203] The multiple materials may cover a variety of proportion information of the multiple elements in the target multi-element system. Theoretically, the more proportion information the multiple materials cover, the more accurate the prediction result of the eutectic point will be.

[0204] For example, the target multi-element system is Zr-Cu-Al, Zr i Cu j Al z is the chemical formula of multiple materials in the target multi-element system, wherein at least one of the proportion information i, j, and z corresponding to the elements Zr, Cu, and Al in the multiple materials is different.

[0205] The prediction system uses the amorphization critical cooling rate prediction method P1000 to obtain the amorphization critical cooling rate of each material in the target multi-element system. The specific content can be referred to the above description and will not be repeated here.

[0206] S1120: Determine the material with the smallest amorphization critical cooling rate among the multiple materials as the target material, and determine the eutectic point of the target multi-element system based on the target material.

[0207] For example, the prediction system predicts the critical cooling rate of amorphization of all materials formed by elements Zr, Cu and Al in the Zr-Cu-Al system by method P1000, and determines the critical cooling rate of Zr 50 Cu 40 Al 10 The critical cooling rate of amorphization of Zr is the minimum value among the critical cooling rates of amorphization of all materials in the Zr-Cu-Al system. 50 Cu 40 Al 10 As the target material, the eutectic point of the Zr-Cu-Al system is determined to be Zr 50 Cu 40 Al 10 .

[0208] Table 4 shows the comparison results between the predicted values and actual values of the eutectic points of multiple multi-element systems predicted by the prediction system using the prediction method P1000.

[0209] Table 4

[0210]

[0211]

[0212] As shown in Table 4, for different multi-element systems, the error between the predicted value and the actual value of the eutectic point obtained by the prediction system is basically less than 5%, and the accuracy is high.

[0213] In summary, the prediction method and system for eutectic point provided in this specification improve the traditional method of determining the eutectic point through a large number of experiments and thermodynamic simulations, greatly improve the prediction efficiency and accuracy of the eutectic point, reduce the time and cost of studying the eutectic point, and promote the progress of material research.

[0214] On the other hand, this specification provides a computer-readable non-transitory storage medium storing at least one set of executable instructions for predicting the critical cooling rate of amorphization or amorphous point or training a target network. When the executable instructions are executed by a processor, the executable instructions instruct the processor to implement the steps of the method P1000 or P1100 for predicting the critical cooling rate of amorphization, or the method P500 for training the target network described in this specification. In some possible implementations, various aspects of this specification can also be implemented in the form of a program product, which includes program code. When the program product is run on the system 400, the program code is used to cause the system 400 to execute the method P1000 or P1100 for predicting the critical cooling rate of amorphization or amorphous point, or the steps of the method P500 for training the target network described in this specification. The program product for implementing the above method can include program code in a portable compact disk read-only memory (CD-ROM) and can be run on the system 400. However, the program product of this specification is not limited to this. In this specification, a readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system. The program product can use any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. The computer-readable storage medium can include a data signal propagated in baseband or as part of a carrier wave, which carries the readable program code. This propagated data signal can take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium may also be any readable medium other than a readable storage medium that can send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing. The program code for performing the operations of this specification may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and conventional procedural programming languages such as "C" or similar programming languages.The program code may execute entirely on the system 400, partly on the system 400, as a stand-alone software package, partly on the system 400 and partly on a remote computing device, or entirely on the remote computing device.

[0215] In the embodiments of this specification, the term "and / or" describes the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0216] The terms "first", "second" and the like in this specification are used to distinguish similar or similar objects or entities, and are not necessarily meant to limit a specific order or sequence.

[0217] Unless otherwise stated, the term “plurality” mentioned in this specification should be understood to mean two or more.

[0218] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0219] In summary, after reading this detailed disclosure, those skilled in the art will appreciate that the foregoing detailed disclosure may be presented by way of example only and may not be limiting. Although not expressly stated herein, those skilled in the art will understand that this specification encompasses various reasonable changes, improvements, and modifications to the embodiments. Such changes, improvements, and modifications are intended to be suggested by this specification and are within the spirit and scope of the exemplary embodiments of this specification.

[0220] Furthermore, certain terms in this specification have been used to describe embodiments of this specification. For example, “one embodiment,” “an embodiment,” and / or “some embodiments” mean that a particular feature, structure, or characteristic described in connection with that embodiment may be included in at least one embodiment of this specification. Therefore, it is emphasized and should be understood that two or more references to “an embodiment,” “one embodiment,” or “an alternative embodiment” in various parts of this specification do not necessarily refer to the same embodiment. Furthermore, particular features, structures, or characteristics may be appropriately combined in one or more embodiments of this specification.

[0221] It should be understood that in the foregoing descriptions of the embodiments of this specification, to facilitate understanding of a feature and to simplify this specification, various features are combined in a single embodiment, figure, or description thereof. However, this does not necessarily mean that these features are combined. When reading this specification, a person skilled in the art may label some of the devices as separate embodiments. In other words, the embodiments of this specification can also be understood as the integration of multiple sub-embodiments. This also applies when each sub-embodiment contains fewer than all the features of a single previously disclosed embodiment.

[0222] Each patent, patent application, publication of a patent application, and other materials, such as articles, books, specifications, publications, documents, and the like, cited in this disclosure (excluding any historical review documents related thereto) is hereby incorporated by reference for all purposes relevant to this disclosure, such as within the specification and claims of this disclosure. However, if there is any inconsistency or conflict between the descriptions, definitions, and / or terminology of such materials and the descriptions, definitions, and / or terminology used in this disclosure, the descriptions, definitions, and / or terminology used in this disclosure shall control.

[0223] Finally, it should be understood that the embodiments of the application disclosed herein are illustrative of the principles of the embodiments of this specification. Other modified embodiments are also within the scope of this specification. Therefore, the embodiments disclosed in this specification are merely examples and not limitations. Those skilled in the art can adopt alternative configurations based on the embodiments in this specification to implement the application in this specification. Therefore, the embodiments of this specification are not limited to the embodiments precisely described in the application.

Claims

1. A method for predicting the critical cooling rate of amorphization, comprising: Generate K descriptors for describing the target material based on information about a proportion of at least one element included in the target material to be predicted and attribute values of X attributes corresponding to each of the at least one element; Determining N1 core descriptors and K-N1 non-core descriptors from the K descriptors, wherein the core descriptors have a greater influence on the critical cooling rate of amorphization than the non-core descriptors. as well as At least part of the K descriptors is input into the target network for prediction to obtain a predicted value corresponding to the critical cooling rate of amorphization of the target material, wherein: The target network is a network trained using Y sample materials. The target network includes multiple sub-networks with the same structure and arranged in layers. Each sub-network includes N input nodes and 1 output node. The output node of the sub-network in the i-th layer serves as the input node of the sub-network in the i+1-th layer. Each sub-network in the first layer is input with the N1 core descriptors and N-N1 non-core descriptors. The N-N1 non-core descriptors are part of the K-N1 non-core descriptors. Moreover, the non-core descriptors input to different sub-networks are not exactly the same; Wherein, X, K, N1, Y, and N are all integers greater than 1, and i is an integer greater than or equal to 1.

2. The method according to claim 1, wherein The X attributes include at least two of the following dimensions: Periodic table dimensions; thermodynamic dimension; physical dimension; or Crystallographic dimension.

3. The method according to claim 1, wherein The K descriptors include at least one of the following: A first set of descriptors is used to describe the statistical characteristics of the target material on the X properties; a second set of descriptors for describing the valence electron occupation state of the target material; A third set of descriptors is used to describe the ionicity of the target material; as well as The fourth group of descriptors is used to describe the configuration entropy of the target material.

4. The method according to claim 3, wherein: The first set of descriptors is generated as follows: For each of the X attributes, based on information about the proportion of the at least one element in the target material and the attribute values of the at least one element for the attribute, statistically analyze the characteristics of the target material for the attribute from multiple statistical dimensions to obtain P statistical values; as well as A first group of descriptors is generated based on P statistical values corresponding to the X properties of the target material, where the number of the first group of descriptors is equal to X*P.

5. The method according to claim 3, wherein The second set of descriptors includes: occupation information of Q valence electrons in the target material, wherein the occupation information of the zth valence electron is obtained by: For each element of the at least one element, determining the number of each type of valence electrons and the total number of valence electrons in the element based on the attribute values of at least some of the X attributes corresponding to the element; and For the zth valence electron, the occupancy information of the zth valence electron is determined based on the proportion information of the at least one element in the target material, the number of the zth valence electrons corresponding to each of the at least one element, and the total number of valence electrons corresponding to each of the at least one element.

6. The method according to claim 3, wherein: The third set of descriptors is obtained based on the proportion information of the at least one element in the target material, the electronegativity corresponding to each of the at least one element, and any one of the following: The maximum value among the electronegativity corresponding to each of the at least one element; or The average value of the electronegativity corresponding to each of the at least one element.

7. The method according to claim 3, wherein: The fourth set of descriptors is obtained based on the proportion information of the at least one element in the target material, the Boltzmann constant and the room temperature.

8. The method according to claim 1, wherein The N1 core descriptors include multiple items of the following: a descriptor characterizing an average of deviations in ground state atomic volume between different elements of the at least one element; a descriptor characterizing a maximum value of weighted deviations in atomic numbers of different elements in the at least one element; a descriptor characterizing a standard deviation of property values of the at least one element on the work function; a descriptor characterizing a standard deviation of the property values of the at least one element at the melting point; a descriptor characterizing a range of weighted property values of different elements in the at least one element on molar heat capacity; as well as A descriptor characterizing an average value of deviations in work function of different elements in the at least one element.

9. The method according to claim 1, wherein The target network includes F levels, and inputting at least part of the K descriptors into the target network for prediction to obtain a predicted value corresponding to the amorphization critical cooling rate includes: Generate m1 groups of descriptors based on the K descriptors, each group of descriptors including the N1 core descriptors and N-N1 non-core descriptors; Inputting the m1 group descriptors into the m1 sub-networks of the first layer of the target network respectively to obtain m1 intermediate prediction values of the critical cooling rate of amorphization; In the order of i values from 2 to F, the m of the i-1th layer of the target network is i-1 The critical cooling rate m of amorphization obtained by the sub-network i-1 The intermediate prediction values are divided into m i After grouping, input the mth layer of the target network i subnetworks, and the critical cooling rate m of amorphization is obtained. i intermediate predictions; and Using the intermediate prediction value of the amorphization critical cooling rate output by the sub-network of the Fth layer as the final prediction value of the amorphization critical cooling rate; Among them, the m1, the F, the m i-1 are all integers greater than 1, and the m i is an integer greater than or equal to 1.

10. A method for training a target network for predicting a critical cooling rate of amorphization of a material, the method comprising: Obtaining experimental values of the amorphization critical cooling rate for each of Y sample materials, each sample material including at least one element; For each sample material, based on information about the proportion of the at least one element in the sample material and attribute values of the X attributes corresponding to each of the at least one element, K descriptors are generated to describe the sample material, and N1 core descriptors and K-N1 non-core descriptors are determined from the K descriptors, wherein the core descriptors have a greater influence on the amorphization critical cooling rate than the non-core descriptors. Part of the K descriptors corresponding to each sample material is input into the target network for prediction to obtain a predicted value corresponding to the critical cooling rate of amorphization of the sample material, where: The target network includes multiple sub-networks with the same structure and arranged in layers. Each sub-network includes N input nodes and 1 output node. The output node of the sub-network in the i-th layer serves as the input node of the sub-network in the i+1-th layer. Each sub-network in the first level is input with the N1 core descriptors and N-N1 non-core descriptors, where the N-N1 non-core descriptors are part of the K-N1 non-core descriptors. The non-core descriptors input to different sub-networks are not exactly the same; and Updating the parameters of the target network with minimizing the difference between the predicted value and the experimental value as a training goal; Wherein, X, K, N1, Y, and N are all integers greater than 1, and i is an integer greater than or equal to 1.

11. The method according to claim 10, wherein: The X attributes include at least two of the following dimensions: Periodic table dimensions; thermodynamic dimension; physical dimension; or Crystallographic dimension.

12. The method according to claim 10, wherein: The K descriptors include at least one of the following: The first group of descriptors is used to describe the statistical characteristics of the sample material on the X properties; a second set of descriptors for describing the valence electron occupation state of the sample material; A third set of descriptors is used to describe the ionicity of the sample material; as well as The fourth group of descriptors is used to describe the configuration entropy of the sample material.

13. The method according to claim 12, wherein: The first set of descriptors is generated as follows: For each of the X attributes, based on information about the proportion of the at least one element in the sample material and the attribute values of the at least one element for the attribute, statistically analyze the characteristics of the sample material for the attribute from multiple statistical dimensions to obtain P statistical values; as well as A first group of descriptors is generated based on P statistical values corresponding to the X attributes of the sample material. The number of the first group of descriptors is equal to X*P.

14. The method according to claim 12, wherein: The second set of descriptors includes: occupation information of Q types of valence electrons in the sample material, wherein the occupation information of the zth type of valence electrons is obtained by: For each element of the at least one element, determining the number of each type of valence electrons and the total number of valence electrons in the element based on the attribute values of at least some of the X attributes corresponding to the element; and For the zth valence electron, the occupancy information of the zth valence electron is determined based on the proportion information of the at least one element in the sample material, the number of the zth valence electrons corresponding to each of the at least one element, and the total number of valence electrons corresponding to each of the at least one element.

15. The method according to claim 12, wherein: The third set of descriptors is obtained based on the proportion information of the at least one element in the sample material, the electronegativity corresponding to each of the at least one element, and any one of the following: The maximum value among the electronegativity corresponding to the at least one element; or The average electronegativity of the at least one element.

16. The method according to claim 12, wherein: The fourth set of descriptors is obtained based on the proportion information of the at least one element in the sample material, the Boltzmann constant and the room temperature.

17. The method according to claim 10, wherein The N1 core descriptors include multiple items of the following: a descriptor characterizing an average of deviations in ground state atomic volume between different elements of the at least one element; a descriptor characterizing a maximum value of weighted deviations in atomic numbers of different elements in the at least one element; a descriptor characterizing a standard deviation of property values of the at least one element on the work function; a descriptor characterizing a standard deviation of the property values of the at least one element at the melting point; a descriptor characterizing a range of weighted property values of different elements in the at least one element on molar heat capacity; as well as A descriptor characterizing an average value of deviations in work function of different elements in the at least one element.

18. The method according to claim 10, wherein The target network includes F levels, and inputting some of the K descriptors corresponding to each sample material into the target network for prediction to obtain a predicted value corresponding to the amorphization critical cooling rate of the sample material includes: Generate m1 groups of descriptors based on the K descriptors, each group of descriptors including the N1 core descriptors and N-N1 non-core descriptors; Inputting the m1 group descriptors into the m1 sub-networks of the first layer of the target network respectively to obtain m1 intermediate prediction values of the critical cooling rate of amorphization; According to the order of i from 2 to F, the m of the i-1 layer of the target network is i-1 The critical cooling rate m of amorphization obtained by the sub-network i-1 The intermediate prediction values are divided into m i After grouping, input the target network layer m i subnetworks, and the critical cooling rate m of amorphization is obtained. i intermediate predictions; and Using the intermediate prediction value of the critical cooling rate of amorphization outputted by the sub-network of the Fth layer as the final prediction value of the critical cooling rate of amorphization of the sample material; Among them, the m1, the F, the m i-1 are all integers greater than 1, and the m i is an integer greater than or equal to 1.

19. A method for predicting a eutectic point, comprising: Predicting, based on the method of any one of claims 1 to 9, the critical cooling rate of amorphization of a plurality of materials formed from a plurality of elements in a target multi-element system, wherein each of the plurality of materials includes the plurality of elements, and the proportions of the plurality of elements in different materials are different; as well as The material having the smallest amorphization critical cooling rate among the multiple materials is determined as a target material, and the eutectic point of the target multi-element system is determined based on the target material.

20. A prediction system, characterized in that Configured to predict the critical cooling rate for amorphization of target materials, including: at least one storage medium storing at least one instruction set; and at least one processor, in communication with the at least one storage medium; Wherein, when the prediction system is running, the at least one processor reads the at least one instruction set and executes the method according to any one of claims 1 to 9 according to the instructions of the at least one instruction set.

21. A training system, characterized in that: The target network configured to train for predicting the critical cooling rate of amorphization of materials includes: at least one storage medium storing at least one instruction set; and at least one processor, in communication with the at least one storage medium; Wherein, when the training system is running, the at least one processor reads the at least one instruction set and executes the method according to any one of claims 10-18 according to the instructions of the at least one instruction set.

22. A prediction system, characterized in that: Configured to predict eutectic points for multi-element systems, including: at least one storage medium storing at least one instruction set; and at least one processor, in communication with the at least one storage medium; Wherein, when the prediction system is running, the at least one processor reads the at least one instruction set and executes the method of claim 19 according to the instructions of the at least one instruction set.

Images