System, method, and program for predicting characteristics of multiphase material using artificial intelligence

Abstract

The present application relates to a system, a method, and a program for predicting characteristics of a multiphase material using artificial intelligence. A system, method, and program predicts characteristics of a material having multiple phases. A system for implementing an AI model that predicts characteristics of a material having multiple phases includes: a memory configured to store executable instructions; and one or more processors configured to execute the instructions to perform operations including: inputting first material information including information about a material into a first AI model to output first feature data; inputting first phase information including information on a first phase of the material into the second AI model to output first phase characteristic data; and inputting second phase information including information on a second phase of the material into the second AI model to output second phase characteristic data; and wherein the first characteristic data comprises information about a characteristic of the material according to a multiphase of the material.

Description

Technical Field

[0001] The present disclosure generally relates to a system, method, and program for predicting properties of materials having multiple phases, and more particularly, to a system, method, and program for predicting properties of materials having multiple phases using artificial intelligence. Background Art

[0002] Recently, with the diversification of consumer demands and trends and the variety of product development methods, the development of materials that can be used to manufacture products has been actively promoted. These materials may affect the properties of the products, and certain properties of the materials (e.g., composition) may become key factors in determining the properties of the manufactured products. Therefore, in order to efficiently develop and mass-produce materials for manufacturing products, it may be necessary to predict and analyze the properties of the materials.

[0003] Traditionally, in order to verify the properties of manufactured products based on the properties of materials, a method of developing various materials, verifying the properties of each material, and then applying them to the final product for testing to verify its properties has been used. These traditional methods require the process of actually manufacturing various materials, verifying their properties, then manufacturing products using these materials, and then verifying the properties of the products. However, this method requires a large amount of cost and time for material and property development, and it is also difficult to find materials with optimal properties.

[0004] As a method for improving this traditional approach, Korean Patent Application Publication No. 10-2022-0060346 discloses a method for predicting material properties using artificial intelligence.

[0005] Existing methods for predicting material properties using artificial intelligence assume that the material is single-phase, extract representative features of the material, and use them to predict the properties of the material.

[0006] However, when actually using materials, there are cases where they change into various phases depending on the environment, usage method, etc. In these cases, if a prediction method assuming a conventional single-phase is used, the prediction accuracy will be very low, making it difficult to apply the prediction results to the development of actual materials and products.

[0007] For example, the positive electrode of a lithium-ion secondary battery includes a positive electrode material containing lithium (Li). When the lithium-ion secondary battery is charged and discharged, lithium in the positive electrode material is absorbed and released in the form of lithium ions (Li + ). Since lithium atoms in the positive electrode material escape during charging under the operating principle of these lithium-ion secondary batteries, the composition ratio of lithium contained in the positive electrode material in the charged state is different from that in the discharged state. Therefore, the positive electrode material has multiple phases in actual use.

[0008] However, existing methods for predicting material properties using artificial intelligence assume that they are single-phase. Therefore, properties are predicted based on the lithium composition during the manufacture of the cathode material and these are used to predict the properties of the product. However, when the cathode material is actually applied to a lithium-ion secondary battery, due to changes in the lithium composition, it has multiple phases. Therefore, according to the conventional method of using single-phase prediction as described above, the predicted results of the material properties and product properties are different from the actual results, thereby reducing the practicality of the artificial intelligence prediction method. Summary of the Invention

[0009] Some embodiments of the present disclosure may provide a system, method, and program that use artificial intelligence to ensure representative features of the multiple phases that a material may have and use these to more accurately predict the properties of a material having multiple phases.

[0010] The problems to be solved by the present invention are not limited to the problems mentioned above, and those skilled in the art will clearly understand other problems not mentioned from the following description.

[0011] According to one aspect of the present disclosure, a system for implementing an AI model for predicting the properties of a material having multiple phases includes: at least one processor; and at least one memory that stores commands or information that allow the at least one processor to perform operations, wherein the operations performed by the commands or information include: inputting first material information including information about a material into a first AI model to output first feature data; inputting first phase information including information about a first phase of the material into a second AI model to output first phase feature data; and inputting second phase information including information about a second phase of the material different from the first phase into the second AI model to output second phase feature data; and wherein the first feature data includes information about the properties of the material according to the phase state of the material.

[0012] The operations performed by the commands or information may further include inputting the first feature data, the first phase feature data, and the second phase feature data into a third AI model.

[0013] The operations performed by the commands or information may further include: adding the first phase feature data to the first feature data and inputting them into the third AI model to output first prediction data; and adding the second phase feature data to the first feature data and inputting them into the third AI model to output second prediction data.

[0014] The system may further include a calculation unit that receives the first prediction data and the second prediction data and outputs a target value.

[0015] The operations performed by the commands or information further include adding the first feature data, the first phase feature data, and the second phase feature data and inputting them into the third AI model.

[0016] In an embodiment, the first feature data may have a class label form.

[0017] In other embodiments, the system may further include a transformation unit that receives the spliced embedding data and outputs transformation data, where the spliced embedding data may include: first feature data; second material information including information about the material; and at least one of information about the substance or device containing the material, where the second material information may be different from the first material information.

[0018] The transformation data may include second feature data, the second feature data may include information about the first feature data, and the second feature data may further include at least one of the second material information and information about the substance or device containing the material.

[0019] The operation performed by the command or information may further include inputting the transformation data into a fourth AI model to output a predicted value.

[0020] The predicted value may include characteristic values of the substance or device containing the material.

[0021] In an embodiment, at least one of the first feature data and the second feature data may be a feature vector.

[0022] According to another aspect of the present disclosure, a method for predicting the characteristics of a material having multiple phases by an AI model is executed by at least one processor. The method may include the following steps: inputting first material information including information about the material into a first AI model to output first feature data; inputting first phase information including information about the first phase of the material into a second AI model to output first phase feature data; and inputting second phase information including information about the second phase of the material different from the first phase into the second AI model to output second phase feature data; where the first AI model and the second AI model may be executed or learned by the processor, and where the first feature data may include information about the material characteristics according to the phase state of the material.

[0023] The method may further include inputting the first feature data, the first phase feature data, and the second phase feature data into a third AI model.

[0024] The method may further include the following steps: inputting the spliced embedding data into the transformation unit and outputting transformation data by the processor, where the spliced embedding data may include: first feature data; second material information including information about the material; and at least one of information about the substance or device containing the material, where the second material information may be different from the first material information.

[0025] A program according to another aspect of the present disclosure is connected to a computer, and the program may be a program stored on a computer-readable medium for executing a method of predicting properties of a material having multiple phases according to an embodiment of the present disclosure.

[0026] According to another aspect of the present disclosure, a system for implementing an AI model for predicting properties of a material having multiple phases may include: at least one processor; and at least one memory that stores commands or information that allows the at least one processor to perform operations, wherein the operations performed by the commands or information may include: inputting embedding data including at least one of first feature data about a material, first material information of the material, and information about a substance or device containing the material into a transformation unit to output transformed data; and inputting the transformed data into a fourth AI model to output a predicted value, wherein the first feature data may include information about material properties according to the phase state of the material.

[0027] According to a specific embodiment of the present disclosure, since an AI model that learns by inputting material information and multi-phase information is utilized, the learned feature data already includes information about the phase state of the material. Therefore, when predicting the properties of the material and / or the properties of a substance or device containing the material, the feature data can be used as a representative vector of the material as input data without obtaining or inputting information about the multi-phase.

[0028] In addition, a method for predicting the properties of a material and / or the properties of a substance or device containing the material according to a specific embodiment of the present disclosure may use the feature data derived from a pre-learned AI model as an input representative vector, and since the feature data already includes information about the properties according to the phase state of the material, even when predicting the properties of a material having multiple phases, property prediction can be accurately performed without separately inputting phase information.

[0029] In addition, according to some embodiments of the present disclosure, the prediction of material properties can be accurately performed only by using the information input into the feature data, so accurate property prediction can be achieved by using only a small amount of information.

[0030] The effects of the present invention are not limited to the effects mentioned above, and those skilled in the art will clearly understand other effects not mentioned from the following description. Brief Description of the Drawings

[0031] Figure 1 is a schematic diagram of a system for implementing a method of obtaining feature data of material composition information based on artificial intelligence according to an embodiment of the present disclosure.

[0032] Figure 2It is a block diagram for explaining the configuration of a device for a method of obtaining characteristic data of material composition information based on artificial intelligence according to an embodiment of the present disclosure.

[0033] Figure 3 It is a flowchart for explaining a learning method of an AI model according to an embodiment of the present disclosure.

[0034] Figure 4 It is a block diagram for explaining a learning method of an AI model according to an embodiment of the present disclosure.

[0035] Figure 5 It is a block diagram for explaining a method of predicting the characteristics of a material and / or a substance or device containing the material according to an embodiment of the present disclosure.

[0036] Figure 6 It is a block diagram for explaining a learning method of an AI model according to another embodiment of the present disclosure. Detailed Embodiments

[0037] The following embodiments are provided as examples so that the gist of the present invention can be fully conveyed to those skilled in the art to which the present invention pertains. Therefore, the present invention is not limited to the embodiments described below and can be embodied in other forms.

[0038] Throughout the present disclosure, the same reference numerals refer to the same components. The present disclosure does not describe all elements of the embodiments and omits general content within the technical field to which the present invention pertains and overlapping details between the embodiments. Terms such as "unit", "module", "component", and "block" used in the specification can be implemented either in software or in hardware, and according to different embodiments, multiple "units", "modules", "components", and "blocks" may be implemented as a single component, or one "unit", "module", "component", and "block" may also include multiple components.

[0039] Throughout this specification, when a component is described as being "connected" to another component, this includes not only the case where they are directly connected but also the case where they are indirectly connected, and indirect connection includes connection via a wireless communication network.

[0040] In addition, when a component is said to "include" a specific component, this means that it may also include other components without excluding other components, unless otherwise explicitly stated to the contrary.

[0041] Throughout this specification, when a component is said to be "on" another component, this includes not only the case where the component is in contact with the other component but also the case where there is another component between the two components.

[0042] Terms such as first and second are used to distinguish one component from another, and these components are not limited by the above terms.

[0043] The singular expressions include plural meanings unless the context clearly indicates otherwise.

[0044] The identification symbols in each step are used for ease of explanation and do not indicate the order of each step, and each step can be executed in a different order from that specified unless the context clearly indicates a specific order.

[0045] The system for predicting material properties according to an embodiment of the present disclosure may include a device, and the device may include various devices capable of performing computational processing and providing results to a user. For example, the system for predicting material properties according to the present disclosure may include one or more of a computer, a processor, a server device, and / or a portable terminal, or may be in any form having the same or similar functions as these devices.

[0046] In addition, the system for predicting material properties according to an embodiment of the present disclosure may include an implementation manner of providing a service from a server to a user terminal, such as in the form of a network service or in the form of linking a server and a user terminal.

[0047] However, the method performed by the system for predicting material properties according to an embodiment of the present disclosure is not limited thereto, and the present disclosure may include all types of realizable systems.

[0048] Here, the computer may include, for example but not limited to, a laptop computer, a desktop computer, a notebook computer, a tablet computer, a slate computer, etc. equipped with a web browser.

[0049] The server device may be a server configured to process information or data by communicating with external devices, and may include an application server, a computing server, a database server, a file server, a game server, a mail server, a proxy server, and a network server, etc., but is not limited thereto.

[0050] The portable terminal is, for example, a wireless communication device providing portability and mobility, and may include all types of handheld wireless communication devices, such as PCS (Personal Communication System), GSM (Global System for Mobile Communications), PDC (Personal Digital Cellular), PHS (Personal Handy-phone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunications)-2000, CDMA (Code Division Multiple Access)-2000, WCDMA (Wideband Code Division Multiple Access), WiBro (Wireless Broadband Internet) terminal, a smart phone, and wearable devices, such as a watch, a ring, a bracelet, an anklet, a necklace, glasses, contact lenses, or a head-mounted device (HMD).

[0051] In some embodiments of the present disclosure, the AI model can be controlled, executed, learned, or operated by at least one processor configured to perform at least one task. The AI model can be stored in a memory, and the feature data according to the specific embodiments of the present disclosure can also be stored in the memory.

[0052] In addition, according to some embodiments of the present disclosure, the commands for causing at least one processor to perform operations can be included in at least one memory. The at least one processor can cause the AI model to run, and in addition to the AI model, can also cause other components of the system (such as a transformation unit, a computing unit, etc.) to run.

[0053] In addition, in specific embodiments of the present disclosure, the AI model can include a neural network, a machine learning model, etc. However, the present invention is not limited thereto.

[0054] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.

[0055] The present disclosure generally relates to systems, methods, and programs for predicting the properties of materials having multiple phases, and more particularly, to systems, methods, and programs for predicting the properties of materials having multiple phases using artificial intelligence.

[0056] Figure 1 is a schematic diagram of a system for implementing a method for predicting material properties according to an embodiment of the present disclosure.

[0057] Referring to Figure 1 , the system 1000 may include a device 100, a database 200, and an artificial intelligence (AI) model 300.

[0058] The device 100, the database 200, and the AI model 300 included in the system 1000 may communicate via a network W. For example, the network W may include a wired network and a wireless network. The network may include various networks such as a local area network (LAN), a metropolitan area network (MAN), and a wide area network (WAN).

[0059] In addition, the network W may include the World Wide Web (WWW). However, according to the embodiments of the present disclosure, the network W is not limited to the networks listed above and may include any type of network, for example, a known wireless data network, a known telephone network, or a known wired and / or wireless television network.

[0060] The device 100 may obtain or output information about the material based on the AI model 300, for example, feature data about the material information. For example, the feature data may include a feature vector as data representing the material information, and the feature value of the material information may be negative or positive and may include a numerical value after the decimal point.

[0061] The material can be a material having multiple phases. Here, a phase (phase state) refers to a state in which the material has uniform physical properties from a macroscopic perspective under specific conditions. The material may change its phase state continuously or discontinuously during use. For example, the material may include the positive electrode material of a secondary battery, and the positive electrode material may contain lithium (Li). In addition, the material may also include a thermally activated delayed fluorescence (TADF) material that can be used in an organic light-emitting diode (OLED) device. However, according to an embodiment of the present disclosure, the material is not limited to the above, and the material may include any type of material that may have multiple phases.

[0062] The material information may include chemical formula, elemental or atomic composition ratio, elemental composition ratio, physical and / or chemical information, molecular structure information, active ion information, and information about molecules that change during the phase change process, and the chemical information may include various types of molecular information (e.g., Simplified Molecular-Input Line-Entry System (SMILES), International Chemical Identifier (INCHI), or Self-Referencing Embedded Strings (SELFIES), etc.). However, the material information is not limited to these, and may include any type of information that can distinguish the differences between different materials.

[0063] In addition, the device 100 may obtain or output characteristic data (i.e., phase information) of each phase of the material having multiple phases based on the AI model 300. The characteristic data is data representing the phase information of the material, may include a characteristic vector, which can be a negative value or a positive value, and may include a numerical value after the decimal point.

[0064] The phase information of the material may include information that distinguishes the differences between each phase, information representing each phase, etc. For example, the phase information may include the composition ratio of elements or atoms, elemental composition ratio, physical and / or chemical information, molecular structure information, etc. For example, if the material contains a lithium positive electrode material, during the charging / discharging of the secondary battery, lithium in the lithium positive electrode material will move out of the positive electrode material in the form of active ions (Li + )), thereby changing the composition ratio of lithium, and then transforming into a different phase. Therefore, the phase information may include the lithium composition ratio in the positive electrode material, the change amount of the lithium composition ratio, the ratio of lithium to other elements except lithium (molar ratio, mass ratio, etc.), etc.

[0065] The device 100 may calculate or predict the properties of the material, the properties of the device or substance containing the material based on the obtained characteristic data. In one embodiment, the device 100 may be used to calculate or predict the properties of the material containing the lithium positive electrode material, the capacity, average voltage, etc. of the lithium-ion secondary battery containing the material.

[0066] The database 200 can store various types of learning data for learning the AI model 300. In addition, the database 200 can store material information, phase information, simulation result information, etc., and in various embodiments, the database 200 can also store the output data output by the AI model 300. However, after the learning of the AI model 300 is completed, the system 1000 may not include the database 200.

[0067] Figure 1 An exemplary embodiment showing the database 200 implemented as a device separate from the device 100 is shown. In this embodiment, the database 200 can be connected to the device 100 through a network W (e.g., a wired and / or wireless network). However, this is just one embodiment, and the database 200 can also be included in the device 100. For example, the database 200 can be implemented as a component of the device 100.

[0068] Figure 1 An exemplary embodiment showing the AI model 300 implemented or provided outside the device 100 (e.g., a cloud-based AI model) is shown, but is not limited thereto, and the AI model 300 can also be included in the device 100 as a component in the device 100.

[0069] Figure 2 It is a block diagram illustrating the configuration of a device for performing a method of obtaining feature data according to an embodiment of the present disclosure or a device for performing a method of predicting the properties of a material having multiple phases using artificial intelligence.

[0070] Referring to Figure 2 , the device 100 may include a memory 110, a communication module or communicator 120, a display 130, an input module 140, and a processor 150. However, the present invention is not limited to these, and the software and hardware configurations of the device 100 can be modified, added, and / or omitted within the scope obvious to those skilled in the art to perform one or more appropriate operations according to the embodiments of the present disclosure. In addition, the device 100 can be replaced by a system, and the device 100 can include multiple devices. In this case, each component included in the device 100 can be included in at least one of the multiple devices.

[0071] The memory 110 can store data supporting or executing various functions of the device 100, programs for the operation of the processor 150 and / or executed in the processor 150, and input and / or output data, and can store multiple applications or apps driven or executed in the device 100, data, commands, and AI models for the operation of the device 100. One or more of these applications can be downloaded from an external server through wireless communication. The memory 110 can store commands or information that cause the processor 150 to perform operations.

[0072] This memory 110 may include any type of storage medium, such as but not limited to flash memory type, hard disk type, solid state drive type (SSD type), silicon disk drive type (SDD type), multimedia card micro, card type memory (e.g., SD, XD memory, etc.), RAM (random access memory), SRAM (static random access memory), ROM (read only memory), EEPROM (electrically erasable programmable read only memory), PROM (programmable read only memory), magnetic memory, magnetic disks, and optical disks.

[0073] In addition, the memory 110 may be a device separate from the device 100 and may include a database connected by wire or wirelessly. Figure 1 The database 200 shown in may be implemented as a component of the memory 110.

[0074] The communication module or communicator 120 may include one or more components configured to perform communication with external devices and may include, for example, one or more of a broadcast receiving module, a wired communication module, a wireless communication module, a short - range communication module, or a location information module.

[0075] The wired communication module may include various wired communication modules, such as a local area network (LAN) module, a wide area network (WAN) module, or a value - added network (VAN) module, and various cable communication modules, such as a universal serial bus (USB), a high - definition multimedia interface (HDMI), a digital visual interface (DVI), a recommended standard 232 (RS - 232), power line communication, or a plain old telephone service (POTS).

[0076] In addition to a WiFi module and a wireless broadband module, the wireless communication module may also include a wireless communication module capable of performing or supporting various wireless communication methods such as a global system for mobile communications (GSM), code division multiple access (CDMA), wideband code division multiple access (WCDMA), universal mobile telecommunications system (UMTS), time division multiple access (TDMA), long - term evolution (LTE), 4G, 5G, and 6G.

[0077] The display 130 displays or outputs the information or data processed in the device 100, the data input or output through the AI model 300, etc. In addition, the display 130 may also display the execution screen information of an application program (e.g., an app) driven or executed in the device 100 or the user interface (UI) or graphical user interface (GUI) information according to such execution screen information.

[0078] The input module 140 may be configured to receive information from a user. When information is input through the user input unit, the processor 150 may control the operation of the device 100 in response to the input information.

[0079] For example, the input module 140 may include physical buttons of a hardware type (e.g., buttons, keycap switches, rollers, toggle switches, etc., located at one or more positions on the front, back, or side of the device 100) and touch buttons of a software type. For example, the touch buttons may include virtual buttons, soft buttons, or visual buttons displayed on a display of a touch screen type through software processing, or may include touch buttons placed on parts of the device 100 other than the touch screen. In addition, the virtual buttons or visual buttons may be displayed in various forms on the touch screen and may be formed as, for example, graphics, text, icons, videos, or combinations thereof.

[0080] The processor 150 may also include a memory and at least one processor. The memory is configured to store data of an algorithm for controlling the operation of components in the device 100 (e.g., learning or performing one or more operations of an AI model) or a program configured to reproduce the algorithm, and the at least one processor is configured to perform the above operations using the data stored in the memory. In this exemplary embodiment, the memory and the processor may be implemented as separate multiple chips or may be implemented as a single chip.

[0081] In one embodiment, the system 1000 or the device 100 according to an embodiment of the present disclosure may include at least one processor. And, in the case where the system 1000 or the device 100 includes multiple processors, the multiple processors may be respectively included in multiple different devices 100.

[0082] In addition, the processor 150 may combinatorially control the above one or more components to implement various embodiments of the present disclosure described below in the device 100.

[0083] Figure 3 is a flowchart for explaining a learning method of an AI model according to an embodiment of the present disclosure, and Figure 4 is a block diagram for explaining a learning method of an AI model according to an embodiment of the present disclosure.

[0084] Refer to Figure 3 and Figure 4, The learning method of the AI model according to an embodiment includes: inputting material information 410 into the first AI model 431 to output first feature data 440 (S411); inputting each of a plurality of first phase information 421 and second phase information 422 into the second AI model 432 to output a plurality of phase feature data 451 and 452 corresponding to each of the first phase information 421 and the second phase information 422 (S412); performing a summation operation on the first feature data 440 and each of the first phase feature data 451 and the second phase feature data 452, and inputting the summation result into the third AI model 460 to output a plurality of first prediction data 471 and second prediction data 472 corresponding to each of the first phase information 421 and the second phase information 422 (S420); inputting the plurality of first prediction data 471 and second prediction data 472 into a calculation unit or calculator 480 to output a target value 490 (S430); and repeating the operations of S411 to S430 above (S440).

[0085] The learning method of the AI model according to some embodiments of the present disclosure can be executed by at least one processor. In addition, commands or information for causing one or more processors to execute or perform operations may be included in at least one memory. One or more processors or one or more memories may be included in multiple devices. Additionally, the learning method of the AI model according to a specific embodiment of the present disclosure can be executed through one or more operations performed via commands or information.

[0086] Each step of learning the AI model will be described in detail below.

[0087] First, the learning method of the AI model according to an embodiment includes: inputting material information 410 into the first AI model 431 to output first feature data 440 (S411), and inputting each of a plurality of first image information 421 and second image information 422 into the second AI model 432 to output a plurality of first image feature data 451 and second image feature data 452 corresponding to each of the first image information 421 and the second image information 422 (S412).

[0088] For example, in S411, the material information 410 may include information or factors representing the material whose properties are to be predicted. The material information 410 may include chemical formula, composition ratio of elements or atoms, composition ratio of elements, chemical information, molecular structure information, active ion information, information about molecules that change during phase change, etc. In the case of a lithium cathode material, the material information 410 may include chemical components of the material constituting the lithium cathode material and information about representative active ions (Li + )

[0089] The first AI model 431 configured to perform learning according to an embodiment of the present disclosure may be a preset algorithm for outputting feature data of material information 410, or may be a model learned based on various information of the material (e.g., formation enthalpy, refractive index, bandgap, phonon frequency, bulk modulus, Debye temperature, thermal conductivity, coefficient of thermal expansion, decomposition enthalpy, etc.). The algorithm of the first AI model 431 may be a multi-layer perceptron (MLP), a graph neural network (GNN), a transformer encoder, etc., but is not limited thereto.

[0090] In one embodiment, the first feature data 440 output by inputting the material information 410 into the first AI model 431 may have N values of N dimensions (C 1 、C 2 、C 3 ……C N ). The first feature data 440 may include a feature vector. The first AI model 431 may vectorize the input material information 410 through element embedding, fractional embedding, etc., and then output the vectorized input material information as feature data through a regressor including multi-head attention and a multi-layer perceptron (MLP) or a residual network (e.g., a model configured to predict feature data from the vectorized information). The output feature data may be in matrix format or in multiple dimensions, and only some feature data may be selected and output from all the feature data. When only some feature data are selected, the selected feature data may be values representing the compositional information of the material. However, the method by which the first AI model 431 outputs the first feature data 440 through vectorization is not limited thereto.

[0091] The material information 410 may include factors that can be used as variables for predicting materials and the properties of devices or materials containing these materials through the learning of the AI model. For example, when learning an AI model for predicting the properties of a lithium cathode material and a secondary battery containing the lithium cathode material, if the variables for predicting the properties are the chemical compositions of the materials constituting the lithium cathode material, the material information 410 input into the first AI model 431 may include information or factors regarding the chemical compositions of the materials constituting the cathode material. Therefore, according to one embodiment, the properties according to the chemical composition of the cathode material can be predicted through the learned AI model.

[0092] In S412, the plurality of first phase information 421 and second phase information 422 may include information that can distinguish the differences between the phases, information representing each phase, etc. For example, the first phase information 421 and the second phase information 422 may include the composition ratios of elements or atoms, the composition ratios of elements, physical / chemical information, molecular structure information, etc. For example, if the material contains a lithium cathode material, during charging or discharging of the secondary battery, lithium in the lithium cathode material will be in the form of active ions (Li +) is removed from the positive electrode material in the form of, thereby changing the composition ratio of lithium, and then transformed into another phase. Therefore, the phase information may include, for example, the lithium composition ratio in the positive electrode material, the change amount of the lithium composition ratio, and the ratio (molar ratio, mass ratio, etc.) of lithium to other elements except lithium as representatives.

[0093] In one embodiment, the plurality of phase information may include first phase information 421 and second phase information 422. The first phase information 421 and the second phase information 422 respectively include information about the first phase and information about the second phase. The first phase and the second phase are different phases from each other. When the material has different phases, the plurality of phase information may include information about each phase as representative factors. For example, when the material is a lithium positive electrode material, the change of active ions (Li + ) will cause a phase change. During discharge, the active ions (Li + ) are almost completely contained in the positive electrode material in the form of lithium, while during charging, at least a part of the lithium contained in the positive electrode material will be released in the form of active ions (Li + ). In this case, the first phase information 421 and the second phase information 422 may each respectively include information about the lithium ratio during discharge and information about the lithium ratio during charging. In this case, the ratio of lithium may include, for example, the ratio of lithium to the entire positive electrode material (including but not limited to molar ratio, mass ratio, etc.).

[0094] In various embodiments, the plurality of first phase information 421 and second phase information 422 may be obtained from actual experimental data, simulation data, known data, etc. For example, if the material is a lithium positive electrode material and the first phase information 421 and the second phase information 422 are to be used as information about the lithium ratio during charging or discharging, the first phase information 421 and the second phase information 422 may be obtained using data obtained through simulation. However, the present invention is not limited thereto.

[0095] In addition, Figure 4 An AI learning method using two pieces of phase information 421 and 422 is shown, but the present invention is not limited thereto. The AI learning method according to the present invention may include inputting three or more pieces of phase information.

[0096] The second AI model 432 configured to perform learning according to an embodiment of the present disclosure may be a preset algorithm for outputting feature data of a plurality of first phase information 421 and second phase information 422, or may be a model learned based on a plurality of information about substances. The algorithm of the second AI model 432 may be (for example, but not limited to) a multi-layer perceptron (MLP), a graph neural network (GNN), a transformer encoder, etc.

[0097] In one embodiment, the phase feature data 451 and 452 output by inputting a plurality of first-phase information 421 and second-phase information 422 into the second AI model 432 may have N values of N dimensions (C 1 、C 2 、C 3 ……C N ). The phase feature data 451 and 452 may include feature vectors. The plurality of first-phase feature data 451 and second-phase feature data 452 correspond to the input plurality of first-phase information 421 and second-phase information 422, and with reference to Figure 4 , the feature data corresponding to the first-phase information 421 may be the first-phase feature data 451, and the feature data corresponding to the second-phase information 422 may be the second-phase feature data 452.

[0098] In various embodiments, the operation of using the first AI model 431 to output the first feature data 440 (S411) and the operation of using the second AI model 432 to output the phase feature data 451 and 452 (S412) may be executed simultaneously or sequentially. When these operations are executed sequentially, the order of execution is not limited.

[0099] Since the plurality of first-phase information 421 includes information about each phase of a material having multiple phases, when the material is used in the form of having multiple phases as a device, material, etc., according to some embodiments of the present disclosure, the AI model can be used to predict the properties of the material and the properties of the device or substance containing the material. For example, in the case of learning an AI model for predicting the properties of a lithium cathode material and a secondary battery containing the lithium cathode material, since the lithium cathode material changes phases according to charging or discharging, the AI model learned according to one embodiment can be used to predict the properties of the cathode material and the secondary battery containing the cathode material.

[0100] In an embodiment, the first feature data 440 may have the form of a class label. The first feature data 440 having the form of a class label may be a representative vector that represents factors related to material information and multiple-phase information. Therefore, the first feature data 440 that has undergone repeated learning can be used as a representative vector (representation form) during the learning of the AI model when the material involved in the input material information has multiple phases, and this representative vector includes information about what properties the material has.

[0101] Next, the first feature data 440 is added to each of the phase feature data 451 and 452 and input into the third AI model 460, and a plurality of first prediction data 471 and second prediction data 472 corresponding to each of the first phase information 421 and the second phase information 422 are output (S420). The first feature data 440 can be added and input into the third AI model 460 and output as the first prediction data 471. The first feature data 440 can be added and input into the third AI model 460 and output as the second prediction data 472.

[0102] The learning method of the AI model according to some embodiments of the present disclosure may include one or more summation operations. The summation operation may include summing the first feature data 440 with each of the phase feature data 451 and 452. For example, if the first feature data 440 is an 8-dimensional vector and each of the phase feature data 451 and 452 is an 8-dimensional vector, the vector input into the third AI model 460 may be a 16-dimensional vector.

[0103] In one embodiment, the third AI model 460 that performs learning according to the embodiments of the present disclosure may be a preset algorithm or a pre-learned model. The algorithm of the third AI model 460 may be (for example, but not limited to) a multi-layer perceptron (MLP), a graph neural network (GNN), a transformer encoder, etc.

[0104] Next, the first prediction data 471 and the second prediction data 472 are input into a calculation unit or a calculator 480 to output a target value 490 (S430).

[0105] The calculation unit or calculator 480 may be configured to perform various operations or may include an AI model. The AI model may be a preset algorithm or a pre-learned model. The algorithm of the third AI model 460 may be (for example, but not limited to) a multi-layer perceptron (MLP), a residual network, etc.

[0106] The target value 490 may include information about the properties of the material and / or the properties of the substance or device containing the material. For example, if the material is a lithium cathode material and the model predicts the properties of a secondary battery containing the lithium-based cathode material, the target value 490 may include the capacity, average voltage, etc. of the secondary battery.

[0107] In one embodiment, the error rates of the information included in the target value 490 may be different or the same. The information included in the target value 490 can be obtained from, for example, actual experimental data, simulation data, known data, etc. In this case, the error rates of the information output to the target value 490 may be different. For example, the target value 490 may include a first output value and a second output value, and the error rates of the first output value and the second output value can be set to be different or the same. In another embodiment, the first output value may be a capacity and the second output value may be an average voltage. In this case, the error rate between the first output value and the second output value can be adjusted within the range of 1:9 to 9:1. In various embodiments, the error rate between the first output value and the second output value can be any one of 9:1, 5:5, or 1:9.

[0108] Subsequently, the learning process of S411 to S430 can be repeated (S440). In this case, the material information 410 and the multiple phase information 421, 422 input during the repeated learning process can be input in various modified forms.

[0109] In a specific embodiment of the present disclosure, the first AI model 431, the second AI model 432, and the third AI model 460 can be learned using the output value or output data. During the repeated learning process, the parameters of the algorithm can be adjusted in various ways.

[0110] In an embodiment of the present disclosure, the first AI model 431, the second AI model 432, and the third AI model 460 can perform at least one of the tasks executed, learned, or driven by at least one processor. In addition, one or more of the first AI model 431, the second AI model 432, or the third AI model 460 can perform at least one of the tasks executed, learned, or driven by another processor.

[0111] According to the AI model learning method of an embodiment of the present disclosure, the first feature data 440 includes data obtained by repeatedly learning by inputting the material information 410 and the first phase information 421 and the second phase information 422 of multiple phases of the material into the AI model and using the output value. Furthermore, the first feature data 440 can include information about the material properties according to the phase state of the material. Therefore, when attempting to predict the properties of a material, even when it is not fully clear how the multiple phases of the material change, the properties of the material with multiple phases and / or the properties of the substance or device containing the material can be predicted more accurately by using the first feature data 440.

[0112] In particular, compared with the conventional method of predicting the properties of materials using artificial intelligence, the conventional method assumes the existence of a single phase. Therefore, it predicts the properties based on the lithium composition during the manufacture of the cathode material and uses these to predict the product properties. Thus, when used in an actual lithium-ion secondary battery (such as the cathode material), if the lithium composition changes according to the usage state (e.g., charging or discharging), and if the information of each phase is not input, the prediction accuracy decreases. Therefore, when using the conventional prediction method, the information of each phase needs to be input for each prediction.

[0113] However, according to some embodiments of the present disclosure, since the AI model configured to learn by inputting material information and multiple phase information is utilized, the first feature data 440 already includes information according to the phase state of the material. Therefore, at each prediction, without re-acquiring and inputting the information of each phase, the first feature data 440 can be used as the representative vector of the material as the input data, thereby improving the accuracy of property prediction.

[0114] Figure 5 It is a conceptual diagram illustrating a method of predicting the properties of a material and / or the properties of a substance or device including the material according to an embodiment of the present disclosure.

[0115] Referring to Figure 5 According to Figure 3 and Figure 4 The first feature data 440 generated according to the exemplary embodiments described with reference to

[0116] and the embedding data 510 including at least one of other material information 511 of the material or information 512 about the substance or device including the material are input to the transformation unit 520. The embedding data 510 may include (e.g., but not limited to) concatenated embedding data.

[0117] Information 512 about a substance or device containing the material may include the applied current, applied voltage, usage conditions, etc. of the substance or device. For example, when the material is a lithium cathode material, the substance or device containing the material may be a lithium-ion secondary battery, and in this case, information 512 about the substance or device containing the material may be the applied current and applied voltage. However, information 512 about the substance or device containing the material is not limited thereto, and various information may be included according to the type of the material, substance, or device containing the material.

[0118] The first feature data 440, other material information 511 of the material, and information 512 about the substance or device containing the material may be included in the embedded data 510 and may be spliced, for example, in the spliced embedded data.

[0119] The transformation unit 520 may perform various computing functions and may also include an AI model. The transformation unit 520 may include a transformer encoder.

[0120] The embedded data 510 input to the transformation unit 520 may be output as transformation data 510a, and the transformation data 510a may include second feature data 515.

[0121] The second feature data 515 includes information about the first feature data 440 and includes information about at least one of the other material information 511 of the material and information 512 about the substance or device containing the material. The second feature data 515 may include a feature vector.

[0122] Next, when the second feature data 515 is input to the fourth AI model 530, a predicted value 540 is output.

[0123] In one embodiment, the fourth AI model 530 may be a preset algorithm or a pre-learned model. The algorithm of the third AI model 460 may be (for example, but not limited to) a multi-layer perceptron (MLP), a graph neural network (GNN), a transformer encoder, etc.

[0124] The predicted value 540 includes data or characteristic values that need to be predicted among the characteristics of the material having multiple phases and / or the characteristics of the substance or device containing the material. For example, the predicted value 540 may include capacity, average voltage, etc.

[0125] The method for predicting the properties of a material and / or the properties of a substance or device including the material according to an embodiment of the present invention may use the first feature data 440 obtained from a pre-learned AI model as input feature data, and since the first feature data 440 already contains information about the properties of the multiphase material through the AI model learning method according to an embodiment of the present disclosure, even when predicting the properties of a material having multiple phases, the property prediction can be accurately performed without inputting phase information.

[0126] For example, one of the research and development goals of the positive electrode material of a lithium-ion secondary battery is to improve the battery capacity, lifespan, etc., and the lithium-ion secondary battery has multiphase properties due to the change in the amount of lithium in the positive electrode material during charging or discharging. According to an embodiment of the present disclosure, by inputting the chemical composition of the positive electrode material as material information (information on a representative state) and the phase information of the positive electrode material in the charged or discharged state, feature data (used as a representation) containing information about the above material information and phase information can be ensured through a learned AI model, and by using this feature data, the properties (e.g., capacity, lifespan) of a lithium-ion secondary battery with a specific chemical composition can be predicted more accurately without additional phase information input.

[0127] In addition, the method for predicting the properties of a material and / or the properties of a substance or device including the material according to an embodiment of the present disclosure may be executed in a device different from the device Figure 3 and Figure 4 described for performing the AI model learning method. In this case, the first AI model 431, the second AI model 432, and the third AI model 460 may execute at least one of the tasks executed, learned, or driven by at least one first processor, Figure 5 and the fourth AI model 530 may execute at least one of the tasks executed, learned, or driven by at least one second processor different from the first processor.

[0128] However, the present invention is not limited thereto. The AI model learning method and the method for predicting the properties of a material and / or the properties of a substance or device including the material according to some embodiments of the present disclosure may be executed in one or more devices and may also be executed by one or more processors.

[0129] In addition, in a specific embodiment of the present disclosure, the positive electrode material of a lithium-ion secondary battery is mainly used as an example for illustration, but the present invention is not limited thereto. The present invention is applicable to various materials having multiple phases including OLED materials.

[0130] Figure 6 is a block diagram illustrating the AI model learning method according to other embodiments of the present disclosure.

[0131] Figure 6 The implementation manner of Figure 4 differs from the implementation manner of

[0132] in that the first prediction data 471 and the second prediction data 472 are output, the operation of inputting the first prediction data 471 and the second prediction data 472 into the calculation unit 480 is omitted, and the first feature data 440 and the multiple phase feature data 451, 452 are all added together and input into the third AI model 460.

[0132] Referring to Figure 6 , the first feature data 440 and the multiple phase feature data 451, 452 are all added together, Figure 4 the calculation unit 480 of Figure 6 is omitted in the implementation manner of

[0133] Thereby simplifying the AI learning.

[0133] The following detailed descriptions that are repetitive with the content described with reference to Figure 3 and Figure 4 are omitted here.

[0134] The following experimental examples explain the prediction accuracy of a system and method for predicting the properties of a multi-phase material using artificial intelligence according to an embodiment of the present disclosure.

[0135] The following table is a table comparing the prediction accuracy performance of a model according to an embodiment of the present disclosure and a model according to a comparative example. The following Example 1 and the comparative example are the results of an experiment on the prediction accuracy of the properties of a lithium cathode material.

[0136] Example 1 uses the model according to the Figure 6 implementation manner of

[0137] [Table 1]

[0138]

[0139] As shown in Table 1, compared with the existing model, when using the prediction model according to the embodiment of the present disclosure, the accuracy of predicting the material properties is improved.

[0140] Example 2 uses the model according to Figure 5For the model of the embodiment, the capacity is predicted from battery experiment data. The input information types and the models used in the comparative examples are as follows. As a comparative example, a model based on CrabNet (npj Computational Materials, 7(1):77, 2021) was used, which can obtain expression information from the composition of materials. The lower the mean absolute error (MAE) value, the better the prediction performance, and the higher the R2 score value, the higher the prediction accuracy.

[0141] [Table 2]

[0142]

[0143] As shown in Table 2, compared with the existing models, when using the prediction model according to the embodiment of the present disclosure, the accuracy of predicting material properties is improved.

[0144] In addition, according to some embodiments of the present disclosure, a method for training an AI model and a method for predicting the properties of a material and / or a substance or device containing the material can be implemented by referring to Figure 1 and Figure 2 the system described.

[0145] In addition, specific disclosed embodiments can be implemented in the form of a recording medium storing commands executable by a computer. These commands can be stored in the form of program code, and when executed by a processor, can generate program modules to perform the operations of the disclosed embodiments. The recording medium can be implemented as a computer-readable recording medium.

[0146] Computer-readable recording media include all types of recording media storing instructions interpretable by a computer. For example, it may include ROM (Read Only Memory), RAM (Random Access Memory), magnetic tapes, magnetic disks, flash memories, optical data storage devices, etc.

[0147] As described above, the disclosed embodiments have been described with reference to the accompanying drawings. Those skilled in the art to which the present disclosure pertains should understand that the present disclosure can be implemented in other forms different from the disclosed embodiments without changing the technical gist or basic features of the present disclosure. The disclosed embodiments are merely examples and should not be construed as restrictive.

[0148] Cross-reference to related applications

[0149] This application claims the priority of Korean Patent Application No. 10-2023-0168608, filed on November 28, 2023, the entire content of which is incorporated herein by reference.

Claims

1. A system for implementing an artificial intelligence (AI) model to predict properties of a material having multiple phases, the system comprising: a memory configured to store executable instructions; as well as One or more processors configured to execute the instructions to perform operations including: inputting first material information into a first AI model to output first feature data, the first material information including information about the material having the multiple phases including a first phase and a second phase; inputting first phase information into a second AI model to output first phase characteristic data, the first phase information comprising information about the first phase of the material; and inputting second phase information into the second AI model to output second phase characteristic data, the second phase information including information about a second phase of the material different from the first phase; and Therein, the first characteristic data includes information on properties of the material according to the multiple phases of the material.

2. The system according to claim 1, wherein: The operations performed by the one or more processors also include inputting the first feature data, the first phase feature data, and the second phase feature data into a third AI model.

3. The system according to claim 2, wherein: The operations performed by the one or more processors also include: summing the first phase characteristic data and the first characteristic data, and inputting the sum of the first phase characteristic data and the first characteristic data into the third AI model to output first prediction data; and The second phase characteristic data is summed with the first characteristic data, and the sum of the second phase characteristic data and the first characteristic data is input into the third AI model to output second prediction data.

4. The system according to claim 3, wherein: The operations performed by the one or more processors also include receiving the first prediction data and the second prediction data, and outputting a target value in response to the first prediction data and the second prediction data.

5. The system according to claim 2, wherein: The operations performed by the one or more processors also include: summing the first feature data, the first phase feature data, and the second phase feature data, and inputting the sum of the first feature data, the first phase feature data, and the second phase feature data into the third AI model.

6. The system according to claim 1, wherein: The first feature data is in the form of a category label.

7. The system according to claim 1, in, The operations performed by the one or more processors further include: receiving concatenated embedded data and outputting transformed data, Wherein, the spliced ​​embedded data includes: the first characteristic data; second material information including information about the material; and at least one of information about a substance or a device containing the material, and The second material information is different from the first material information.

8. The system according to claim 7, wherein: The transformation data includes second feature data, The second characteristic data includes at least one of the second material information and information about a substance or a device including the material, and information about the first characteristic data.

9. The system according to claim 8, wherein: The operations performed by the one or more processors also include inputting the transformed data into a fourth AI model to output a predicted value.

10. The system according to claim 9, wherein: The predicted value includes a characteristic value of a substance or a device including the material.

11. The system according to claim 1, wherein: At least one of the first feature data or the second feature data is a feature vector.

12. A computerized method comprising the steps of: inputting first material information into a first AI model to output first feature data, the first material information including information about a material having multiple phases including a first phase and a second phase; inputting first phase information into a second AI model to output first phase characteristic data, the first phase information comprising information about the first phase of the material; as well as inputting second phase information into the second AI model to output second phase characteristic data, the second phase information including information about a second phase of the material different from the first phase, wherein the first AI model and the second AI model are executed or learned by one or more processors, and Therein, the first characteristic data includes information on properties of the material according to the multiple phases of the material.

13. The computerized method of claim 12, further comprising the steps of: The first characteristic data, the first phase characteristic data, and the second phase characteristic data are input into a third AI model.

14. The computerized method of claim 12, further comprising the steps of: outputting transformed data in response to the input concatenated embedded data, Wherein, the spliced ​​embedded data includes: the first characteristic data; second material information including information about the material; and at least one of information about a substance or a device containing the material, and The second material information is different from the first material information.

15. A program stored on a non-transitory computer-readable recording medium, the program storing instructions executable by one or more processors to perform the operations included in the computerized method according to claim 12.

16. A system for implementing an AI model to predict properties of a material having multiple phases, the system comprising: a memory configured to store executable instructions; as well as One or more processors configured to execute the instructions to perform operations including: outputting transformed data by inputting embedded data, the embedded data including at least one of characteristic data about the material, material information about the material, and information about a substance or a device including the material; and Input the transformed data into an AI model to output a predicted value, Therein, the characteristic data include information about properties of the material depending on the multiphase nature of the material.

Images