WO2023171774A1 - 材料創出を支援するシステム及び方法、プログラム - Google Patents

材料創出を支援するシステム及び方法、プログラム Download PDF

Info

Publication number
WO2023171774A1
WO2023171774A1 PCT/JP2023/009174 JP2023009174W WO2023171774A1 WO 2023171774 A1 WO2023171774 A1 WO 2023171774A1 JP 2023009174 W JP2023009174 W JP 2023009174W WO 2023171774 A1 WO2023171774 A1 WO 2023171774A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
data
inference
data set
manufacturing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/JP2023/009174
Other languages
English (en)
French (fr)
Japanese (ja)
Inventor
貴文 木俣
崇弘 冨田
恭平 阿閉
健司 森本
真吾 惣川
泰輔 加藤
孝文 西脇
篤則 稲垣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
NGK Insulators Ltd
Original Assignee
NGK Insulators Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by NGK Insulators Ltd filed Critical NGK Insulators Ltd
Priority to EP23766950.2A priority Critical patent/EP4492389A4/en
Priority to JP2023543393A priority patent/JP7392209B1/ja
Priority to CA3245482A priority patent/CA3245482A1/en
Priority to KR1020247029475A priority patent/KR20240144988A/ko
Publication of WO2023171774A1 publication Critical patent/WO2023171774A1/ja
Priority to JP2023198164A priority patent/JP2024015482A/ja
Priority to US18/823,832 priority patent/US20240428900A1/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/23—Updating
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/70—Machine learning, data mining or chemometrics
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures

Definitions

  • the present invention generally relates to techniques that support material creation.
  • MI Materials informatics
  • MI inorganic materials
  • properties are affected by a complex higher-order structure (for example, particle size, particle size distribution, grain boundary phase, etc.) in which a large number of crystal grains are gathered, and therefore it is difficult to simulate the properties. For this reason, it is difficult to obtain highly accurate simulation data regarding the characteristics, and therefore, it is difficult to apply MI.
  • the material creation support system includes first to third models and a processing section.
  • the first model is a model in which a manufacturing method recipe data set representing a manufacturing method recipe is input and a material property data set representing material characteristics is output.
  • the second model is a model whose input is a material feature data set representing material features and whose output is a material property data set.
  • the third model is a model in which the manufacturing method recipe data set is input and the material characteristic data set is output.
  • the processing unit performs either inference using the first model or inference using the second model and the third model, depending on the accuracy of at least one of the first to third models. Inference is made, and in the inference, recipe characteristic data, which is data representing the association between the manufacturing recipe of the material and the material characteristics, is generated or updated.
  • FIG. 1 shows an example of a system configuration in an embodiment of the present invention.
  • An example of the configuration of the MI platform system 100 is shown.
  • An example of functional blocks of the AI section 108 is shown.
  • An example of the flow of processing performed by the learning unit 351 is shown.
  • An example of the flow of processing performed by the inference unit 352 is shown.
  • Recipe provision is schematically shown.
  • a modified example of learning of the first model 301 is shown.
  • an "interface device” may be one or more interface devices.
  • the one or more interface devices may be at least one of the following: - One or more I/O (Input/Output) interface devices.
  • the I/O (Input/Output) interface device is an interface device for at least one of an I/O device and a remote display computer.
  • the I/O interface device for the display computer may be a communication interface device.
  • the at least one I/O device may be a user interface device, eg, an input device such as a keyboard and pointing device, or an output device such as a display device. - One or more communication interface devices.
  • the one or more communication interface devices may be one or more of the same type of communication interface device (e.g., one or more NICs (Network Interface Cards)) or two or more different types of communication interface devices (e.g., one or more NICs). It may also be an HBA (Host Bus Adapter).
  • HBA Hypervisor Adapter
  • memory refers to one or more memory devices that are an example of one or more storage devices, and may typically be a main storage device. At least one memory device in the memory may be a volatile memory device or a non-volatile memory device.
  • persistent storage may be one or more persistent storage devices, which is an example of one or more storage devices.
  • the persistent storage device may typically be a non-volatile storage device (for example, an auxiliary storage device), and specifically, for example, an HDD (Hard Disk Drive), an SSD (Solid State Drive), or an NVME (Non-Volatile). It may be a Memory Express drive or an SCM (Storage Class Memory).
  • a “storage device” may be at least a memory and a persistent storage device.
  • a "processor” may refer to one or more processor devices.
  • the at least one processor device may typically be a microprocessor device such as a CPU (Central Processing Unit), but may also be another type of processor device such as a GPU (Graphics Processing Unit).
  • At least one processor device may be single-core or multi-core.
  • the at least one processor device may be a processor core.
  • At least one processor device is a circuit that is a collection of gate arrays (for example, an FPGA (Field-Programmable Gate Array), a CPLD (Complex Programmable Logic Device), or an ASIC (Application Interface)) using a hardware description language that performs part or all of the processing.
  • plication It may be a processor device in a broad sense such as a specific integrated circuit).
  • functions may be explained using the expression "yyy part", but functions may be realized by one or more computer programs being executed by a processor, or by one or more computer programs being executed by a processor. It may be realized by the above hardware circuit (for example, FPGA or ASIC), or a combination thereof.
  • a function is realized by a program being executed by a processor, the specified processing is performed using a storage device and/or an interface device as appropriate, so the function may be implemented as at least a part of the processor. good.
  • a process described using a function as a subject may be a process performed by a processor or a device having the processor. Programs may be installed from program source.
  • the program source may be, for example, a program distribution computer or a computer-readable recording medium (for example, a non-temporary recording medium).
  • the description of each function is an example, and a plurality of functions may be combined into one function, or one function may be divided into a plurality of functions.
  • MI is an abbreviation for materials informatics
  • DB is an abbreviation for database
  • FIG. 1 shows an example of a system configuration in an embodiment of the present invention.
  • a material creation support system 10 that supports material creation is constructed.
  • the material creation support system 10 may be a physical computer system, a logical computer system based on a physical computer system, or a combination of at least part of the physical computer system and at least part of the logical computer system. It may also be used in combination with the section.
  • a physical computer system may be composed of one or more physical computers, and may include an interface device, a storage device, and a processor connected thereto.
  • the logical computer system may include a virtual machine or a system as a cloud computing service.
  • the material creation support system 10 includes a data presentation section 50, a data acquisition section 60, and a data conversion section 70. These functions 50, 60, and 70 may be realized by a computer program being executed by a processor.
  • the data presentation unit 50 acquires data from the data mart 107 periodically or each time an inquiry is received, and presents the target data.
  • the data in the data mart 107 will be referred to as "organized data.”
  • the target data is at least one of the acquired organized data and data based on an inference result obtained by inputting the organized data into a machine learning model.
  • Data mart 107 is an example of a first data source.
  • the data preferably includes data regarding ceramic materials.
  • the data acquisition unit 60 stores experimental data, which is data obtained by an experiment based on the target data, in the electronic experiment notebook 131, the production DB 121, and the evaluation DB 122.
  • the electronic experiment notebook 131, the production DB 121, and the evaluation DB 122 are examples of the second data source.
  • "experimental data” may be data related to the experiment, such as a summary, details, results, or evaluation of the experiment.
  • the data conversion unit 70 acquires experimental data from the electronic experiment notebook 131, the production DB 121, and the evaluation DB 122, and converts the experimental data into data included in the existing or new data mart 107.
  • the target data is updated or It is based on the newly created data mart 107, that is, the organized data of experimental data of experiments conducted based on target data presented in the past. Therefore, regardless of whether or not the type of material to be created is one for which highly accurate simulation data regarding material properties can be obtained, it is possible to increase the possibility that data useful for material creation will be presented. . As a result, the types of materials that can be supported in material creation using informatics methods will be expanded. Specifically, for example, even if the type of material to be created is an inorganic material (ceramics), data useful for material creation can be presented.
  • the interface device may be communicably connected to at least one of the experiment system 110, the data acquisition unit 60, and the researcher terminal 11A.
  • the storage device may store at least some of the data of the electronic experiment notebook 131, the production DB 121, the evaluation DB 122, the various data 150, the data lake 102, and the data mart 107.
  • the processor may implement the functions 50, 60, and 70 described above by executing a computer program.
  • a data presentation unit 50 and a data conversion unit 70 are provided in the MI platform system 100.
  • the data acquisition unit 60 is provided outside (or inside) the MI platform system 100.
  • the MI platform system 100 may be a physical computer system, a logical computer system, or a combination of at least part of a physical computer system and at least part of a logical computer system.
  • the MI platform system 100 is a physical computer system, as shown in FIG. 2, and includes an interface device 201, a storage device 202, and a processor 203 connected thereto. Communication may be made with the researcher terminal 11 (at least the trainee terminal 11A) through the interface device 201.
  • the storage device 202 may be provided with a data lake 102 or a data mart 107 to store data. Additionally, programs may be stored in the storage device 202. At least some of the functions of the data presentation unit 50 and data conversion unit 70 in the MI platform system 100 may be realized by the processor 203 reading and executing a program from the storage device 202.
  • the data presentation section 50 includes an AI (Artificial Intelligence) section 108 and an IF (Interface) section 109.
  • AI Artificial Intelligence
  • IF Interface
  • the AI unit 108 performs machine learning model learning and inference using the machine learning model. For example, in response to an instruction from the IF unit 109, the AI unit 108 outputs to the IF unit 109 an inference result obtained by inputting the organized data acquired from the data mart 107 into a machine learning model.
  • the IF unit 109 receives an inquiry from the researcher terminal 11A, and presents the target data to the researcher terminal 11A in response to the inquiry.
  • the researcher terminal 11A is an information processing terminal (for example, a personal computer or a smartphone) of the materials researcher 5A (an example of a user).
  • the researcher terminal 11A is an example of a sender of an inquiry about the target data, and an example of a destination to which the target data is presented.
  • the IF unit 109 collects target data (organized data acquired from the data mart 107 and/or information obtained by instructing the AI unit 108).
  • the object represented by the data (based on the inference results obtained) is presented.
  • the object presented may be, for example, a manufacturing recipe or material properties.
  • a “manufacturing recipe” refers to a method of creating a material, and may typically include a material composition and/or a synthesis process.
  • the "material composition” may be, for example, a formulation composition, and the “synthesis process” may be the type of raw materials (for example, different particle sizes), process conditions, etc.
  • the manufacturing method recipe may include a material development strategy.
  • the material researcher 5A creates new materials based on the presented target data and conducts experiments based on the target data.
  • the presentation destination (transmission destination) of the target data may be a system such as the experiment system 110 instead of or in addition to the researcher terminal 11A.
  • the materials researcher 5A conducts an experiment using the experimental system 110 based on the presented target.
  • the experiment using the experiment system 110 may be a combinatorial experiment.
  • the experimental system 110 may include one or more devices, such as a ceramic fabrication device 111 (eg, an air firing furnace) and a ceramic evaluation device 112 (eg, a device that evaluates the coefficient of thermal expansion).
  • Experimental data is output from experimental devices such as the manufacturing device 111 and the evaluation device 112.
  • the output experimental data is transmitted to the data server 120 and stored.
  • a combinatorial experiment is an example of an experiment, and instead of or in addition to a combinatorial experiment, at least one of a high-throughput experiment, an automated experiment using a robot, and an experiment that mainly involves manual work is adopted. It's okay.
  • the data acquisition section 60 includes a data server 120 and an experiment notebook section 130.
  • the data server 120 manages a DB in which experiment data from the experiment system 110 is stored.
  • a DB managed by the data server 120 may exist for each type of experimental device.
  • a creation DB 121 that stores experimental data from the manufacturing device 111
  • an evaluation DB 122 that stores experimental data from the evaluation device 112 may be managed.
  • Experiment data from the experiment system 110 is converted into structured data by the data server 120 and stored in a DB.
  • the experiment notebook section 130 manages the experiment data acquired from the DB managed by the data server 120 as an electronic experiment notebook 131.
  • the electronic experiment notebook 131 is structured data, such as a DB.
  • the data conversion unit 70 includes a collection unit 101, a feature calculation unit 103, an image analysis unit 104, a natural language analysis unit 105, and an arrangement unit 106.
  • the collection unit 101 collects experimental data from the electronic experiment notebook 131 and the DB of the data server 120, formats the collected experimental data, and stores the formatted experimental data in the data lake 102.
  • "Formatting” as used herein means arranging the structure of experimental data into a predetermined structure.
  • the collection unit 101 may collect various data 150.
  • the various data 150 may include experimental data.
  • the various data 150 may include numerical data of past experiments, material databases available for a fee or free of charge, linguistic data representing past experiments (for example, experiment contents and experiment results), calculations such as simulations, etc. It may include at least one of the obtained material data and the language data of patent documents and papers.
  • the data source for numerical data from past experiments may be different from the data source for linguistic data from patent documents and papers.
  • the various data 150 may include sensor data, which is data including measured values by sensors. At least some of the various data 150 may be stored in the data lake 102.
  • the feature amount calculation unit 103 calculates feature amounts of one or more predetermined types of data in the formatted experimental data. Thereby, each of the one or more types of data is digitized, and as a result, processing by the AI unit 108 becomes possible.
  • the one or more types of data may be image data and text data.
  • the image data is analyzed by the image analysis unit 104, and based on the result of the analysis, the feature amount calculation unit 103 calculates the feature amount of the image data.
  • the text data is subjected to text mining by the natural language analysis unit 105, and the feature amount calculation unit 103 calculates the feature amount of the text data based on the result of the text mining.
  • the above-mentioned neural network may be used to calculate the feature amount.
  • a copy of the experimental data may be generated in the data lake 102, and each of one or more predetermined types of data in the copy may be converted into a calculated feature amount (numeric value).
  • a organizing unit 106 organizes the data lake 102 into one or more data marts 107 as a data set that meets predetermined conditions.
  • the data mart 107 includes organized data that can be presented as at least part of the target data not via the AI unit 108, and the data mart 107 that includes organized data that can be presented as at least part of the target data via the AI unit 108.
  • a data mart 107 may be generated.
  • the data may be presented to the materials researcher 5A without going through the MI platform system 100.
  • another materials researcher 5B may use the trainee terminal 11B to acquire experimental data from the electronic experiment notebook 131, organize the experimental data, and input the organized data to the AI section 140.
  • the AI section 140 may be implemented outside (or inside) the trainee terminal 11B.
  • the AI unit 140 may also perform inference by inputting the input organized data into a machine learning model. Data based on the results of this inference may be presented to the material researcher 5A.
  • the material researcher 5A may conduct an experiment based on this presented data. Data for this experiment may also be stored in the data server 120 from the experiment system 110 and collected in the MI platform system 100.
  • Experiments based on the presented target data may be facilitated by a predetermined function implemented within or outside the researcher terminal 11A.
  • the function may decide what kind of experiment to recommend to the materials researcher 5A, and present the determined items to the materials researcher 5A.
  • the target data may include data representing what kind of experiment is recommended.
  • the cycle of presenting target data ⁇ experiment ⁇ collecting experimental data ⁇ converting experimental data to organized data ⁇ presenting target data based on organized data can be performed with high efficiency and speed. You can expect it. In this way, MI based on experimental data is realized. Therefore, it is possible to support the creation of materials for which it is difficult to obtain highly accurate simulation data regarding material properties, such as materials with complex higher-order structures such as ceramics.
  • various data internal or external to an organization (for example, a company) that provides the MI platform system 100 can be aggregated in the MI platform system 100, structured and managed, and reflected in the data mart 107. Therefore, the possibility that the presented target data is useful for material creation is increased.
  • ⁇ Expression 1> Periodically or each time a query is received, data is obtained from a first data source, and at least the data is based on the inference result obtained by inputting the data into a machine learning model.
  • a data presentation unit (50) that presents certain target data
  • a data acquisition unit (60) that stores experimental data, which is data obtained by an experiment based on the target data, in a second data source
  • a material creation support system comprising: a data conversion unit (70) that acquires experimental data from the second data source and converts the experimental data into data included in an existing or new first data source.
  • ⁇ Expression 3> (A) periodically or each time a query is received, obtain data from a first data source and combine the data with data based on inference results obtained by inputting the data into a machine learning model; present target data that is at least one of the (B) storing experimental data, which is data obtained by an experiment based on the target data, in a second data source; (C) obtaining experimental data from the second data source and converting the experimental data into data contained in an existing or new first data source; Material creation support method.
  • the machine learning model acquires updated data included in an existing or new first data source by acquiring the experimental data, and learns parameters.
  • the material creation support method described in Expression 3. Specifically, for example, it is as follows. Based on the experimental data acquired by (C) (for example, experimental data acquired by the data conversion unit 70), data in the existing data mart 107 is updated, or data based on the acquired experimental data A new data mart 107 containing the following is generated. Therefore, the existing or new data mart 107 will include updated data that is data based on the acquired experimental data.
  • the AI unit 108 may acquire update data from the data mart 107 and learn (for example, update) parameters of a machine learning model used by the AI unit 108.
  • ⁇ Expression 5> collecting existing or new experimental data and converting it into data contained in a first data source; Periodically or each time a query is received, data is obtained from a first data source, and at least the data is based on the inference result obtained by inputting the data into a machine learning model.
  • presenting certain target data Material creation support method.
  • this material creation support method may be performed by the MI platform system 100.
  • the material creation support system 10 does not need to have the data presentation section 50 and the data conversion section 70 but the data acquisition section 60.
  • ⁇ Expression 6> Among the data acquired from the first data source, the presented target data, the experimental data, the data included in the existing or new first data source, and the data based on the inference result.
  • At least one of the data includes data regarding the ceramic material; A material creation support method described in any one of Expressions 3 to 5.
  • ⁇ Expression 8> The data acquired from the first data source, the presented target data, the experimental data, the data included in the existing or new first data source, the learning data, and the recommended inference result. at least one of the data representing data includes data regarding the ceramic material; A method for generating a material creation support model described in Expression 7.
  • ⁇ Expression 9> A program for causing a computer to execute the material creation support method described in any one of Expressions 3 to 6.
  • ⁇ Expression 10> A program for causing a computer to execute the material creation support model generation method described in Expression 7 or 8.
  • a "data set” is a logical mass of electronic data seen from a program such as an application program, and may be any one of a record, a file, a key-value pair, a tuple, etc.
  • Researcher input information is information input by researcher 5A. Researcher input information is input to the IF section 109. In this embodiment, the researcher input information is typically information representing a manufacturing method recipe or material characteristics.
  • Researcher output information is information output (provided) to the researcher 5A.
  • the researcher output information is output from the IF section 109.
  • the researcher output information is typically information representing material properties or a manufacturing recipe.
  • Manufacturing method recipe data set is a data set representing a manufacturing method recipe for materials.
  • the manufacturing method recipe data set may be a data set based on the researcher input information.
  • Medical property data set is a data set representing material properties.
  • the researcher output information is information representing material properties, the researcher output information may be information based on a material property data set.
  • FIG. 3 shows an example of functional blocks of the AI section 108.
  • the AI unit 108 may receive a dataset from the IF unit 109 (that is, a dataset based on researcher input information may be generated by the IF unit 109), or may receive researcher input information from the IF unit 109. A dataset may be generated based on the information input by the researcher concerned. Further, the AI unit 108 may output the dataset to the IF unit 109 (that is, researcher output information based on the dataset may be generated by the IF unit 109), or the researcher output information based on the dataset may be output to the IF unit 109. The output information may be output to the IF section 109. Further, the IF section 109 may be included in the AI section 108.
  • the AI unit 108 inputs a manufacturing method recipe data set 311 and outputs (predicts) a material property data set 313. Furthermore, in the present embodiment, when the researcher input information is information representing material characteristics, the AI unit 108 identifies the manufacturing method recipe data set 311 of the manufacturing method recipe that is expected to create a material having the material characteristics, The manufacturing method recipe data set 311 (or researcher output information based on the manufacturing method recipe data set 311) is output to the IF section 109. That is, if the researcher 5A inputs information representing a manufacturing method recipe into the MI platform system 100, the MI platform system 100 can provide information representing the predicted material properties of the material assumed to have been created using the manufacturing method recipe. Can receive. Furthermore, by inputting information representing material properties into the MI platform system 100, the researcher 5A receives from the MI platform system 100 information representing a manufacturing recipe for a material expected to create a material having the material properties. be able to.
  • the AI unit 108 includes a learning unit 351 that performs learning of a machine learning model, and an inference unit 352 that performs inference using the learned machine learning model (that is, the above-mentioned inference). Furthermore, there are first to third models 301 to 303 as machine learning models. The first to third models 301 to 303 are stored in the storage device 202.
  • the first model 301 is a model that has a manufacturing method recipe data set 311 as input and a material property data set 313 as output.
  • Ceramics have a complex higher-order structure (e.g., grain size, grain size distribution, grain boundary phase, etc.) made up of a large number of crystal grains, and material characteristics including this higher-order structure affect the material properties of ceramics. Therefore, it is generally difficult to predict material properties with high accuracy from a manufacturing recipe.
  • the material feature data set 312 (a data set representing material features) is input based on the fact that material features influence material properties, in other words, there is a correlation between material features and material properties.
  • a second model 302 is prepared that outputs a material property data set 313.
  • the manufacturing method recipe data set 311 is input and the material feature data set 312 is predicted.
  • a third model 303 to be output is prepared.
  • the "manufacturing method recipe" represented by the manufacturing method recipe data set 311 may include a material composition and/or a synthesis process, and specifically, for example, one or more recipe items (for example, synthesis temperature). and a value (typically a numerical value) for each of the one or more recipe items.
  • Machine properties represented by the material property data set 313 means properties exhibited by a substance as a material (for example, a structure), and specifically, for example, one or more property items (for example, strength, thermal expansion, etc.). rate, etc.) and a value (typically a numerical value) for each of the one or more characteristic items.
  • the “material characteristics” represented by the material characteristics data set 312 means the physical and/or chemical state exhibited by a substance as a material, and specifically, for example, one or more characteristic items (for example, crystal particle size, particle size distribution, etc.) and values (typically numerical values) for each of the one or more characteristic items.
  • a feature amount may be included in the material feature data set 312 for each feature item.
  • the second model 302 can be used to predict material properties that include values for one or more property items from material features that include values for multiple feature items, or to predict values for one feature item. It is also possible to predict material properties including values for one or more property items from material features including.
  • the material feature data set 312 may be a data set representing the feature amount calculated by the feature amount calculation unit 103, and may be stored in the data mart 107 and may be acquired from the data mart 107.
  • One or more feature quantities represented by the material feature data set 312 are extracted by the feature quantity calculation unit 103 from image data (for example, SEM image) and spectral data using data processing (for example, processing using CNN). Ru.
  • image data for example, SEM image
  • spectral data using data processing (for example, processing using CNN).
  • Ru for example, processing using CNN.
  • the model is a linear regression (for example, Ridge regression, Lasso regression, Elastic Net regression), logistics regression, SVM (Support Vector Machine), decision tree model (for example, Random Forest, XGBoost (Xtreme Gradient Boosting), LightGBM (Light Gradient Boosting Machine)), neural networks (e.g. CNN, RNN (Recurrent Neural Network), Recurrent Neural Network) sNet (Residual Network)), Bayesian optimization, k-NN (k- Nearest Neighbor), or a combination of two or more of these models.
  • linear regression for example, Ridge regression, Lasso regression, Elastic Net regression
  • logistics regression for example, SVM (Support Vector Machine), decision tree model (for example, Random Forest, XGBoost (Xtreme Gradient Boosting), LightGBM (Light Gradient Boosting Machine)
  • neural networks e.g. CNN, RNN (Recurrent Neural Network), Recurrent Neural Network) sNet (Residual Network)
  • the model is a random forest, XGBoost, LightGBM, CNN, Bayesian optimization, or any two or more models thereof.
  • a combination is fine.
  • the following is expected: -Although the number of manufacturing method recipe data sets 311 as learning data is relatively small, Bayesian optimization is capable of making predictions including prediction deviations, is excellent at finding data addition candidate points, and is Find the optimal value while increasing it.
  • - Decision tree models eg, Random Forest, XGBoost, LightGBM
  • XGBoost or LightGBM prevents overfitting and increases robustness by creating multiple decision trees and applying weights to targets with prediction errors.
  • ⁇ CNN has a high degree of freedom, can handle complex systems, is useful when there are many features, and can adjust the degree of overfitting by appropriately controlling the number of learning times (typically the number of epochs). .
  • each of the first to third models 301 to 303 is a regression equation.
  • FIG. 4 shows an example of the flow of processing performed by the learning section 351.
  • the learning unit 351 stores learning data for the first model 301, specifically, one or more manufacturing method recipe data sets 311 and one or more material characteristics corresponding to the one or more manufacturing method recipe data sets 311.
  • the data set 313 is acquired (S401). These data sets 311 and/or 313 may be input by the researcher 5A in the learning process, may be prepared in advance as learning data, or may be data such as experimental data and/or various data 150, etc. or manufacturing method recipe data representing a manufacturing method recipe with narrowed recipe items or value ranges (or manufacturing method recipes with expanded recipe items or value ranges) in a manufacturing method recipe represented by a certain manufacturing method recipe data set 311.
  • the set 311 (manufacturing recipe data set 311 newly generated by the learning unit 351) may be used.
  • the learning unit 351 learns the first model 301 using the acquired data sets 311 and 313 (S402). Specifically, for example, the learning unit 351 stores a manufacturing method recipe data set 311 and a material property data set 313 (a manufacturing method recipe and the material characteristics of the material created by the manufacturing method recipe) corresponding to the manufacturing method recipe data set 311. The first model 301 is learned by using the first model 301.
  • the learning unit 351 determines whether to finish learning the first model 301 (S403). For example, when the accuracy (prediction accuracy) of the first model 301 is greater than or equal to the first accuracy threshold (or when the accuracy of the first model 301 is higher than the accuracy of the second model 302 and the third model 303) ), the determination result in S403 is true (S403: Yes). In this case, the process of the learning unit 351 ends.
  • the learning unit 351 uses the learning data of the second model 302, specifically, the one or more material feature data sets 312, and the one or more material feature data sets 312.
  • One or more material property data sets 313 corresponding to the material property data set 312 are acquired (S404). These data sets 312 and/or 313 may be input by the researcher 5A in the learning process, may be prepared in advance as learning data, or may be data such as experimental data and/or various data 150, etc.
  • Material feature data representing a feature item or a material feature with a narrowed value range (or a material feature with an expanded feature item or value range) in the material feature represented by a certain material feature data set 312 The set 312 (material feature data set 312 newly generated by the learning unit 351) may be used.
  • the learning unit 351 learns the second model 302 using the acquired datasets 312 and 313 (S405). Specifically, for example, the learning unit 351 uses a material feature data set 312 and a material property data set 313 (material features and material properties of a material having the material features) corresponding to the material feature data set 312. A second model 302 is trained. The purpose of this learning is to find material features that have a strong correlation with material properties, or in other words, to identify material features that correspond to material properties.
  • the learning unit 351 determines whether to finish learning the second model 302 (S406). For example, when the accuracy (prediction accuracy) of the second model 302 is equal to or greater than the second accuracy threshold, the determination result in S406 is true (S406: Yes).
  • the fact that the accuracy of the second model 302 is equal to or greater than the threshold value means that important feature items and/or values (feature amounts) have been identified. In other words, if important feature items and/or values (feature amounts) are not identified, the accuracy of the second model 302 typically does not exceed the threshold value.
  • the learning unit 351 uses the learning data of the third model 303, specifically, the one or more material feature data sets 312 specified in the learning of the second model 302. , one or more manufacturing method recipe data sets 311 corresponding to the one or more material characteristic data sets 312 are acquired (S407).
  • the manufacturing method recipe data set 311 corresponding to the material characteristic data set 312 may be, for example, the manufacturing method recipe data set 311 corresponding to the material characteristic data set 313 corresponding to the material characteristic data set 312.
  • These data sets 312 and/or 311 may be data sets acquired from the data sets used in this learning process, or may be data sets obtained by narrowing down recipe items or value ranges in a manufacturing method recipe represented by a certain manufacturing method recipe data set 311.
  • the manufacturing method recipe data set 311 (a manufacturing method recipe data set 311 newly generated by the learning unit 351) representing a manufacturing method recipe (or a manufacturing method recipe with expanded recipe items and value ranges) may be used.
  • the learning unit 351 learns the third model 303 using the acquired data sets 312 and 311 (S408). Specifically, for example, the learning unit 351 acquires a material feature data set 312 and a manufacturing method recipe data set 311 (a material feature and a manufacturing method recipe for creating a material having the material feature) corresponding to the material feature data set 312.
  • the third model 303 is learned using the following.
  • the learning of the third model 303 may be performed in parallel with the learning of the second model 302 (S404 and S405).
  • the material feature data set 312 acquired in S407 may be a data set representing the features specified (narrowed down) during learning of the second model 302, or may be material feature data before such identification. It may also be set 312.
  • the learning unit 351 determines whether to finish learning the third model 303 (S409). If the accuracy (prediction accuracy) of the third model 303 is greater than or equal to the third accuracy threshold, the determination result in S409 is true (S409: Yes). Note that the first to third accuracy thresholds described above may be the same value or may be different values.
  • the learning unit 351 performs S401 again.
  • the manufacturing method recipe data set 311 specified (or newly generated) during learning of the third model 303 and the material property data set 313 corresponding to the manufacturing method recipe data set 311 may be acquired.
  • the process of the learning unit 351 ends. Note that even if the determination in S403 is performed N times (N is an integer of 2 or more), if the accuracy of the first model 301 is less than the first accuracy threshold (if the determination result in S403 is not true), the learning unit The process of 351 may be completed.
  • re-learning means that the process of the learning unit 351 is once terminated and then the process of the learning unit 351 is performed again.
  • Relearning may be started at a predetermined timing.
  • the predetermined trigger may be, for example, at least one of the following.
  • the learning data has increased sufficiently since the previous learning (for example, the learning data has increased by more than a certain amount).
  • the accuracy of at least one of the models 301 to 303 has deteriorated to less than the first accuracy threshold.
  • the inference by the inference unit 352 may be performed in parallel with the processing by the learning unit 351. Further, a data set obtained through inference by the inference unit 352 may be used for learning any of the models 301 to 303. Specifically, for example, at least one of the following may be adopted. - If the determination result in S403 is true (learning of the first model 301 is completed), inference is performed using the first model 301, and a manufacturing recipe may be provided as appropriate. - If the determination result in S406 is false (learning of the second model 302 has not been completed), inference is performed using the second model 302, and material characteristics may be provided as appropriate.
  • Inference using the second model 302 may be possible, for example, when the difference between the accuracy of the second model 302 and the second accuracy threshold is less than or equal to a predetermined difference. - If the determination result in S409 is false (learning of the third model 303 has not been completed), inference is performed using the third model 303, and a manufacturing recipe may be provided as appropriate. Inference using the third model 303 may be possible, for example, when the difference between the accuracy of the third model 303 and the third accuracy threshold is less than or equal to a predetermined difference.
  • the second Inference is made using the model 302 and the third model 303, and a manufacturing recipe may be provided as appropriate.
  • a manufacturing recipe may be provided as appropriate.
  • the manufacturing method recipe data set 311, material feature data set 312, and material property data set 313 obtained from the experimental data are used in learning or relearning to make any of the determination results in S403, S406, and S409 true. May be used.
  • the new conditions may include conditions based on the inference results, or may include conditions determined independently of the inference results based on the knowledge of the researcher 5A.
  • FIG. 5 shows an example of the flow of processing performed by the inference unit 352.
  • the inference unit 352 determines whether relearning has been performed (S501). If the determination result in S501 is true (S501: Yes), the inference unit 352 invalidates the list (S502).
  • the "list” referred to here is data generated or updated in inference, and is data in which the relationship between material characteristics and manufacturing recipe and/or material characteristics is recorded. The list is stored in the storage device 202 (see FIG. 2) by the reasoning unit 352, and can be referenced by the IF unit 109. Furthermore, "invalidating" a list means making it impossible for the IF unit 109 to refer to the list.
  • the storage device 202 may mean deleting the list from the storage device 202, or It may be to associate data that means invalidity, or it may be to save a list from one storage area to another storage area. Further, S502 may be omitted. For example, each time a list is newly generated or updated, the latest list may be made valid (the list that is referenced for providing manufacturing recipe or material properties).
  • the inference unit 352 determines whether learning of the first model 301 has been completed (S503). If the determination result in S503 is true (S503: Yes), the inference unit 352 performs inference using the first model 301 (S504). Specifically, for example, the inference unit 352 acquires a plurality of manufacturing method recipe data sets 311 and inputs the manufacturing method recipe data set 311 into the first model 301 for each acquired manufacturing method recipe data set 311. A material property data set 313 may be obtained. Each of the plurality of manufacturing method recipe data sets 311 may be a data set acquired from newly added experimental data or various data 150, or may be data generated by different combinations of recipe items or values for each recipe item.
  • the inference unit 352 may increase the manufacturing method recipe data set 311 by repeatedly changing the manufacturing method recipe, such as increasing or decreasing the value of the synthesis temperature by a fixed value).
  • the inference unit 352 lists the manufacturing recipe represented by the manufacturing recipe data set 311 and the material characteristics represented by the material property data set 313 for each pair of the input manufacturing method recipe data set 311 and the acquired material property data set 313. to be recorded. Further, a material property may be recorded in the list when the material property satisfies a predetermined condition (for example, when the value of a certain property item falls within a predetermined value range).
  • the inference unit 352 determines whether learning of the third model 303 has been completed (S505). If the determination result in S505 is true (S505: Yes), the inference unit 352 performs inference using the second and third models 302 and 303 (S506). Specifically, for example, the inference unit 352 obtains the material property data set 313 by inputting the material feature data set 312 into the second model 302.
  • the inference unit 352 may obtain the material feature data set 312 by inputting the manufacturing method recipe data 311 to the third model 303 .
  • the inference unit 352 extracts the manufacturing method recipe and material characteristic data represented by the manufacturing method recipe data set 311 for each set of the input manufacturing method recipe data set 311, the acquired material characteristic data set 312, and the acquired material characteristic data set 313.
  • the material characteristics represented by set 312 and the material characteristics represented by material property data set 313 are recorded in a list. The list may also record material characteristics corresponding to the manufacturing recipe and material properties.
  • a material property may be recorded in the list when the material property satisfies a predetermined condition (for example, when the value of a certain property item falls within a predetermined value range).
  • each of the plurality of manufacturing method recipe data sets 311 or material characteristic data sets 312 may be a data set acquired from newly added experimental data or various data 150, or may be a data set acquired from newly added experimental data or various data 150, or a combination of each item or a value for each item.
  • the inference unit 352 may increase the manufacturing method recipe data set 311 by repeatedly changing the manufacturing method recipe, such as increasing or decreasing the value of the synthesis temperature by a fixed value.
  • the inference unit 352 determines whether the accuracy of inference using the second and third models 302 and 303 is greater than or equal to a threshold value. You may decide whether or not. If the result of this determination is also true, the inference unit 352 may perform S506 (inference using the second and third models 302 and 303).
  • the inference unit 352 determines whether learning of the second model 302 has been completed (S507). If the determination result in S507 is true (S507: Yes), the inference unit 352 performs inference using the second model 302 and the third model 303 (S508). Specifically, for example, the inference unit 352 first uses the second model 302 to determine the material characteristics necessary to satisfy the target material properties (the material property data set 313 representing the target material properties is the output). A material feature data set 312) as an input is identified.
  • the inference unit 352 uses the third model 303 to infer a manufacturing recipe for obtaining the specified material characteristics (specifically, the material characteristic data set 312 representing the specified material characteristics (Identifies the manufacturing method recipe data set 311 as an input that is the output).
  • the inference unit 352 records the manufacturing method recipe, material characteristics, and material properties in a list for each set of the manufacturing method recipe data set 311 as input, the acquired material characteristic data set 312, and the acquired material characteristic data set 313. do.
  • each of the plurality of manufacturing method recipe data sets 311 or material characteristic data sets 312 may be a data set acquired from newly added experimental data or various data 150, or may be a combination of recipe items or a value for each recipe item.
  • the inference unit 352 may increase the manufacturing method recipe data set 311 by repeatedly changing the manufacturing method recipe, such as increasing or decreasing the value of the synthesis temperature by a fixed value.
  • the inference unit 352 performs inference using two models, the second and third models 302 and 303, or the first model 301 (S509). . Specifically, for example, the inference unit 352 performs the following (S509-1) or (S509-2). More specifically, for example, if the prediction accuracy of the second model 302 and the third model 303 is higher than the prediction accuracy of the first model 301, the inference unit 352 performs (S509-1). , if the prediction accuracy of the second model 302 and the third model 303 is the same as or lower than the prediction accuracy of the first model 301, (S509-2) may be performed.
  • each of the plurality of manufacturing method recipe data sets 311 or material characteristic data sets 312 may be a data set acquired from newly added experimental data or various data 150, or a combination of each item or a data set for each item.
  • the inference unit 352 may increase the manufacturing method recipe data set 311 by repeatedly changing the manufacturing method recipe, such as increasing or decreasing the value of the synthesis temperature by a fixed value.) .
  • the inference unit 352 obtains the material property data set 313 by inputting the material property data set 312 to the second model 302.
  • the inference unit 352 Toward creating a material having the material properties represented by the acquired material property data set 313 (that is, the manufacturing method recipe data set 311 whose output is the material feature data set 312 input for obtaining the material property data set 313) ), the inference unit 352 obtains a material feature data set 312 by inputting the manufacturing method recipe data 311 to the third model 303 .
  • the inference unit 352 extracts the manufacturing method recipe and material characteristic data set 312 represented by the manufacturing method recipe data 311 for each set of the input manufacturing method recipe data 311, the acquired material characteristic data set 312, and the acquired material characteristic data set 313.
  • the material characteristics represented by and the material characteristics represented by the material property data set 313 are recorded in a list.
  • the inference unit 352 acquires the material property data 313 by inputting the manufacturing method recipe data 311 into the first model 301 (in an actual experiment, the material property data set 313 is used to evaluate the material properties). may also be collected). The inference unit 352 lists the manufacturing recipe represented by the manufacturing recipe data set 311 and the material characteristics represented by the material property data set 313 for each pair of the input manufacturing method recipe data set 311 and the acquired material property data set 313. to be recorded.
  • the inference unit 352 determines whether to end the inference process (S510). If the determination result in S510 is true (S510: Yes), the process ends. If the determination result in S510 is false (S510: No), the process returns to S501.
  • the list 600 shown in FIG. 6 is generated or updated.
  • the IF unit 109 receives an inquiry specifying material properties from the researcher 5A, it identifies a manufacturing recipe for creating a material with the material properties based on the list 600, and researches the identified manufacturing recipe. Provided to person 5A.
  • the AI unit 140 shown in FIG. 1 may have the same function as the AI unit 108 described above.
  • the "experimental data” collected by the collection unit 101 includes data from combinatorial experiments, data obtained from the electronic experiment notebook 131 in which past experimental data is accumulated, data obtained from external sites, etc. may include at least some data of .
  • “experimental data” may include calculation data (for example, data predicted by first-principles calculation or separate machine learning), or may include linguistic data.
  • at least one of the data sets 311 to 313 may also include calculation data or linguistic data. -Also, images may be generated during learning, but in the embodiment, image generation is not required during learning. As a result, it is expected that the calculation load will be small and the model will be simple.
  • the material creation support system 10 includes, for example, an interface device 201 connected to a user device (for example, a researcher terminal 11A), which is a device having an input device and a display device, and a storage device 202 in which models 301 to 303 are stored. It may include an interface device 201 and a processor 203 connected to a storage device 202. The processor 203 may realize the AI section 108 and the IF section (interface section) 109 by executing a computer program, for example.
  • the AI unit 108 performs inference using the first model 301, and inference using the second model 302 and the third model, depending on the accuracy (for example, prediction accuracy) of at least one of the first to third models 301 to 303.
  • the list 600 (an example of recipe characteristic data, which is data representing the association between a manufacturing method recipe and material characteristics) is generated or updated.
  • the IF unit 109 may receive an inquiry specifying a material characteristic, specify a manufacturing method recipe corresponding to the characteristic specified in the inquiry based on the list 600, and provide the specified manufacturing method recipe.
  • a second model 302 is prepared based on the correlation between material features and material properties, and it is possible to predict material property data set 313 from manufacturing method recipe data set 311 via an intermediate data set called material feature data set 312.
  • a third model 303 is prepared for this purpose. The presence of the second and third models 302 and 303 allows researchers to create materials with desired material properties even if the accuracy of the first model 301 is insufficient due to insufficient amount of experimental data or other reasons. You can contribute to the proposal of manufacturing method recipes for.
  • the AI unit 108 may be generated or updated in inference using the model 301 of 1. Thereby, inferences for generating or updating the list 600 used to provide the manufacturing method recipe can be performed with a smaller computational load (without inputting/outputting an intermediate data set as the material characteristic data set 312).
  • the accuracy of the second model 302 and the third model 303 does not mean the individual accuracy of the second model 302 and the third model 303, but the accuracy of the second model 302 and the third model 303. This is the accuracy when both models 303 are used.
  • the AI unit 108 may learn the first model 301 using the manufacturing method recipe data set 311 and material property data set 313 specified or generated during learning or inference of the second and third models 302 and 303. .
  • the learning or inference of the second and third models 302 and 303 it may be determined that it is preferable for the recipe item of synthesis time to be present or absent in the manufacturing recipe.
  • recipe items are expanded or reduced, and values for recipe items are expanded or reduced, and as a result, the learning of the first model 301 It is expected that the number of preferred manufacturing method recipe data sets 311 will increase. Therefore, it can be expected that the accuracy of the first model 301 will be improved.
  • the AI unit 108 may perform inference using the first model 301 and/or inference using the second and third models 302 and 303 in parallel with the learning of the models 301 to 303. .
  • the AI unit 108 may switch (select) a model to be used in inference as learning is switched (depending on which model learning has been completed).
  • the AI unit 108 selects one or more material feature data sets from the plurality of material feature data sets 312 as factors that cause the accuracy of the second model 302 to be equal to or higher than the second accuracy threshold. 312 may be identified or generated.
  • the AI unit 108 extracts data corresponding to one or more material feature data sets 312 specified or generated in the learning or inference of the second model 302 from the plurality of manufacturing recipe data sets 311.
  • One or more manufacturing recipe data sets 311 may be identified or generated.
  • the AI unit 108 stores one or more manufacturing method recipe data sets 311 specified or generated in the learning or inference of the third model 303, and one or more material characteristics corresponding to the one or more manufacturing method recipe data sets 311.
  • the first model may be trained using the data set 313.
  • the material feature data set 312 is narrowed down during learning or inference of the third model 303, so the accuracy of the third model 303 can be increased with a small calculation load.
  • the manufacturing method recipe data set 311 corresponding to such narrowed-down material feature data set 312 is used for learning the first model 301, the accuracy of the first model 301 can be increased with a small calculation load. .
  • the material feature data set 312 may be a data set composed of numerical feature amounts for each feature item belonging to the material feature. Thereby, the computational load required for learning and inference can be reduced, and the second and third models 302 and 303 can be simplified.
  • the AI unit 108 When performing relearning, the AI unit 108 discards the list 600 and performs inference using the first model 301 after relearning and/or the second model 302 after relearning and the second model 302 after relearning.
  • the list 600 In the inference using the model 303 of No. 3, the list 600 may be newly generated or updated. This makes it possible to avoid the IF unit 109 from providing a recipe for a list generated or updated in inference using an old model whose accuracy has not yet been further improved.
  • the accuracy of the first model 301 is less than the first accuracy threshold (for example, even if the learning of the first model 301 is performed a predetermined number of times, the AI unit 108 (if the accuracy of the model 301 does not exceed the first accuracy threshold), in addition to the manufacturing method recipe data set 311 specified or generated during learning of the second and third models 302 and 303, the manufacturing method recipe data set 311 Learning of the first model 301 (process recipe, material feature set, and material property You may learn the correspondence relationship with This can be expected to increase the possibility that the accuracy of the first model 301 can exceed the first accuracy threshold.
  • the manufacturing method recipe data set 311 Learning of the first model 301 process recipe, material feature set, and material property You may learn the correspondence relationship with This can be expected to increase the possibility that the accuracy of the first model 301 can exceed the first accuracy threshold.
  • the AI unit 108 creates a list 600 in which material features are associated with material properties in addition to the manufacturing method recipe (that is, the relationship between the manufacturing method recipe, material characteristics, and material properties).
  • a recorded list 600) may be created or updated.
  • the IF unit 109 receives an inquiry in which material characteristics are specified in addition to the material properties, specifies a manufacturing method recipe corresponding to the material characteristics and material characteristics specified in the inquiry based on the list 600, and You may provide a recipe for the manufacturing method.
  • the material characteristics may include (1) density and porosity quantified by the Archimedes method, and (2) SEM image characteristics (for example, composition area ratio and composition area deviation).
  • Characteristic items (characteristic factors) that can affect the target properties may include porosity, crystal circumference, compositional area ratio, and compositional area deviation ("compositional area deviation” refers to (This is a feature amount representing the variation in the composition ratio of a plurality of divided regions constituting one image (here, a SEM image)).
  • the first model 301 may be Ridge regression (input may be manufacturing conditions such as composition ratio, material relay, and firing temperature, and output may be bending strength and coefficient of thermal expansion).
  • the second model 302 may be Ridge regression or LightGBM (inputs may be the above-mentioned (1) density and porosity, and (2) SEM image features, and outputs may be bending strength and coefficient of thermal expansion) .
  • the third model 303 may be Ridge regression (inputs may be manufacturing conditions such as composition ratio, material particle size, and firing temperature, and outputs may be the above-mentioned (1) density and porosity, and (2) SEM image (Features may be used.)
  • a manufacturing method recipe may be provided (may be displayed on the researcher terminal 11A).
  • any of the models 301 to 303 may include a regression equation for each element (for example, objective variable) as an output. That is, any of the models 301 to 303 may include one or more regression equations.
  • the data set as input may be common for all of the models 301 to 303, explanatory variables and weighting are typically different for each regression equation in the model.
  • the purposes of inference are (1) to find a manufacturing recipe that produces a material with desired material properties, and (2) to find manufacturing recipe candidates for adding experimental data to improve model accuracy.
  • the inference in S504 in FIG. 5 corresponds to the purpose (1) (this is because model learning has been completed).
  • the inferences in S506, S508, and S509 in FIG. 5 correspond to the purpose (2) (this is because model learning is not completed).
  • a dataset e.g., a manufacturing method recipe
  • combinations e.g., combinations of a manufacturing method recipe and material properties
  • a data set obtained through actual experiments based on the data set is added as a learning data set, and model learning is performed using the added data set. Since the dataset is augmented in this way, the model can be trained with greater accuracy.
  • the AI unit 108 Inference may be made using the second model 302 and the third model 302.
  • the AI unit 108 may learn the second model 302 before learning the third model 303.
  • the AI unit 108 specifies that the data set input or output in the inference using the second model 302 and the third model 303 is used for learning at least one of the first to third models.
  • the second case may be given priority over the second case.
  • First case A case where the inference using the second model 302 and the third model 303 is an inference performed when the accuracy of the third model 303 is less than a third threshold.
  • Second case A case in which the inference using the second model 302 and the third model 303 is performed when the accuracy of the third model 303 is equal to or greater than the third threshold.
  • the AI unit 108 Inference may be performed using the model 302 and the third model 303, and if the accuracy of the second model 302 and the third model 303 is the same as or lower than the accuracy of the first model 301, the AI The unit 108 may perform inference using the first model 301.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • Health & Medical Sciences (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
PCT/JP2023/009174 2022-03-10 2023-03-09 材料創出を支援するシステム及び方法、プログラム Ceased WO2023171774A1 (ja)

Priority Applications (6)

Application Number Priority Date Filing Date Title
EP23766950.2A EP4492389A4 (en) 2022-03-10 2023-03-09 SYSTEM, METHOD AND PROGRAM FOR AID IN THE CREATION OF A MATERIAL
JP2023543393A JP7392209B1 (ja) 2022-03-10 2023-03-09 材料創出を支援するシステム及び方法、プログラム
CA3245482A CA3245482A1 (en) 2022-03-10 2023-03-09 SYSTEM, METHOD AND PROGRAM FOR AID IN THE CREATION OF A MATERIAL
KR1020247029475A KR20240144988A (ko) 2022-03-10 2023-03-09 재료 창출을 지원하는 시스템 및 방법, 프로그램
JP2023198164A JP2024015482A (ja) 2022-03-10 2023-11-22 材料創出を支援するシステム及び方法、プログラム
US18/823,832 US20240428900A1 (en) 2022-03-10 2024-09-04 Material creation support system, method, and program

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2022037335 2022-03-10
JP2022-037335 2022-03-10

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/823,832 Continuation US20240428900A1 (en) 2022-03-10 2024-09-04 Material creation support system, method, and program

Publications (1)

Publication Number Publication Date
WO2023171774A1 true WO2023171774A1 (ja) 2023-09-14

Family

ID=87935292

Family Applications (2)

Application Number Title Priority Date Filing Date
PCT/JP2023/009174 Ceased WO2023171774A1 (ja) 2022-03-10 2023-03-09 材料創出を支援するシステム及び方法、プログラム
PCT/JP2023/009175 Ceased WO2023171775A1 (ja) 2022-03-10 2023-03-09 材料創出を支援するシステム及び方法、プログラム

Family Applications After (1)

Application Number Title Priority Date Filing Date
PCT/JP2023/009175 Ceased WO2023171775A1 (ja) 2022-03-10 2023-03-09 材料創出を支援するシステム及び方法、プログラム

Country Status (6)

Country Link
US (1) US20240428900A1 (https=)
EP (1) EP4492389A4 (https=)
JP (3) JP7392209B1 (https=)
KR (1) KR20240144988A (https=)
CA (1) CA3245482A1 (https=)
WO (2) WO2023171774A1 (https=)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2025069920A1 (ja) * 2023-09-29 2025-04-03 日本電気株式会社 機械学習装置、機械学習方法、及びプログラム
WO2025115306A1 (ja) * 2023-11-28 2025-06-05 株式会社日立ハイテク データ収集装置、データ収集方法、及びデータ収集システム
WO2025248833A1 (ja) * 2024-05-27 2025-12-04 日本碍子株式会社 セラミックス粉体生成方法、乾燥方法、及び、セラミックス粉体生成システム

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2025238536A1 (en) * 2024-05-14 2025-11-20 Sacmi Cooperativa Meccanici Imola Societa' Cooperativa Method for creating a database for a continuous cycle production line for decorated ceramic products and computer server containing the database

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2020060955A (ja) * 2018-10-10 2020-04-16 国立研究開発法人物質・材料研究機構 予測管理システム、予測管理方法、データ構造、予測管理装置及び予測実行装置
JP2021043959A (ja) 2019-09-05 2021-03-18 国立大学法人東京工業大学 作製評価システム、作製評価方法及びプログラム
JP2021043325A (ja) * 2019-09-11 2021-03-18 東京応化工業株式会社 情報処理システム、情報処理装置、情報処理方法及びプログラム

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP7201981B2 (ja) * 2017-06-30 2023-01-11 学校法人 明治薬科大学 予測装置、予測方法および予測プログラム
GB201805296D0 (en) * 2018-03-29 2018-05-16 Benevolentai Tech Limited Shortlist Selection Model For Active Learning
JP2020198003A (ja) * 2019-06-04 2020-12-10 ジャパンモード株式会社 生成物推定プログラム及びシステム
WO2021015134A1 (ja) * 2019-07-23 2021-01-28 昭和電工株式会社 材料設計システム、材料設計方法、及び材料設計プログラム
JP2020057364A (ja) * 2019-07-30 2020-04-09 学校法人東北工業大学 対象化合物の特性の予測を行うための方法、コンピュータシステム、プログラム
US20210098084A1 (en) * 2019-09-30 2021-04-01 Nissan North America, Inc. Method and System for Material Screening
WO2021124392A1 (ja) * 2019-12-16 2021-06-24 日本電信電話株式会社 材料開発支援装置、材料開発支援方法、および材料開発支援プログラム
BR112022011151A2 (pt) * 2019-12-18 2022-08-23 Dow Global Technologies Llc Visão de máquina para caracterização com base em dados analíticos
JP7398334B2 (ja) * 2020-04-30 2023-12-14 株式会社日立製作所 ユーザに提案する材料を決定するシステム

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2020060955A (ja) * 2018-10-10 2020-04-16 国立研究開発法人物質・材料研究機構 予測管理システム、予測管理方法、データ構造、予測管理装置及び予測実行装置
JP2021043959A (ja) 2019-09-05 2021-03-18 国立大学法人東京工業大学 作製評価システム、作製評価方法及びプログラム
JP2021043325A (ja) * 2019-09-11 2021-03-18 東京応化工業株式会社 情報処理システム、情報処理装置、情報処理方法及びプログラム

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4492389A4

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2025069920A1 (ja) * 2023-09-29 2025-04-03 日本電気株式会社 機械学習装置、機械学習方法、及びプログラム
WO2025115306A1 (ja) * 2023-11-28 2025-06-05 株式会社日立ハイテク データ収集装置、データ収集方法、及びデータ収集システム
WO2025248833A1 (ja) * 2024-05-27 2025-12-04 日本碍子株式会社 セラミックス粉体生成方法、乾燥方法、及び、セラミックス粉体生成システム

Also Published As

Publication number Publication date
CA3245482A1 (en) 2025-01-23
EP4492389A1 (en) 2025-01-15
WO2023171775A1 (ja) 2023-09-14
JP7392208B1 (ja) 2023-12-05
EP4492389A4 (en) 2026-03-18
KR20240144988A (ko) 2024-10-04
US20240428900A1 (en) 2024-12-26
JPWO2023171775A1 (https=) 2023-09-14
JP2024015482A (ja) 2024-02-02
JPWO2023171774A1 (https=) 2023-09-14
JP7392209B1 (ja) 2023-12-05

Similar Documents

Publication Publication Date Title
JP7392209B1 (ja) 材料創出を支援するシステム及び方法、プログラム
Abdollahzadeh et al. A multi-objective optimization algorithm for feature selection problems
Arróyave et al. A perspective on Bayesian methods applied to materials discovery and design
Liang et al. A survey of surrogate-assisted evolutionary algorithms for expensive optimization: J. Liang et al.
Martínez-Frutos et al. Kriging-based infill sampling criterion for constraint handling in multi-objective optimization
Gao et al. A novel hybrid PSO based on levy flight and wavelet mutation for global optimization
JP7411977B2 (ja) 機械学習支援方法及び機械学習支援装置
Mehdizadeh Khorrami et al. Forecasting heating and cooling loads in residential buildings using machine learning: a comparative study of techniques and influential indicators
Khodaparasti et al. Prediction of the concrete compressive strength using improved random forest algorithm
Bellotti Optimized conformal classification using gradient descent approximation
Liu et al. Improved GWO and its application in parameter optimization of Elman neural network
WO2018194565A1 (en) Monitoring the thermal health of an electronic device
Huang et al. MBCT: tree-based feature-aware binning for individual uncertainty calibration
Han et al. A feature selection method based on feature-label correlation information and self-adaptive MOPSO
Liu et al. A sequential multi-fidelity surrogate model-assisted contour prediction method for engineering problems with expensive simulations
Lee et al. K2vTune: A workload-aware configuration tuning for RocksDB
Parhi et al. Application of metaheuristic spotted hyena optimization in strength prediction of concrete
Cai et al. A Novel Power-sum Time-varying Grey Prediction Model and Its Applications.
Mainini et al. Analytical benchmark problems and methodological framework for the assessment and comparison of multifidelity optimization methods
Dubey et al. A review: Most important and frequently used algorithms in supervised learning
CN116861256B (zh) 一种固废焚烧过程的炉温预测方法、系统、设备及介质
Mandli et al. Selection of most relevant features from high dimensional data using ig-ga hybrid approach
Tchaye-Kondi et al. A new hashing based nearest neighbors selection technique for big datasets
Abdalla et al. Evaluation and Benchmarking of Performance of Machine Learning and Symbolic Regression: Datasets, Software Tools and Prediction Models
Ansari Shiri et al. A New Hybrid Filter-Wrapper Feature Selection using Equilibrium Optimizer and Simulated Annealing

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 2023543393

Country of ref document: JP

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23766950

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 20247029475

Country of ref document: KR

Kind code of ref document: A

WWE Wipo information: entry into national phase

Ref document number: 1020247029475

Country of ref document: KR

WWE Wipo information: entry into national phase

Ref document number: 2023766950

Country of ref document: EP

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2023766950

Country of ref document: EP

Effective date: 20241010