Design support devices, methods, methods for manufacturing materials, programs, and organic materials

Abstract

To provide a design support device, method, material manufacturing method, program, and organic material that accurately predict the properties of a material. [Solution] The learning process performed by the control unit of the design support device involves acquiring first learning data including structural information indicating the structure of the material and characteristic information indicating the properties of the material, generating structural information, calculating a sensitivity analysis of the first learning data to expand the first learning data, generating second learned data, and storing the second learned data to perform the learning process.

Description

[Technical Field]

[0001] The present disclosure relates to a design support apparatus, a method, a method for producing a material, a program, and an organic material. [Background Art]

[0002] A technique for predicting properties of materials based on machine learning is known. For example, Patent Document 1 discloses a prediction apparatus that inputs a material composition of a prediction target material into a trained model to predict a phase fraction at each temperature of the prediction target material within a predetermined temperature interval, arranges each component of the prediction target material on a first axis, and displays a heat map in which each temperature within the predetermined temperature interval is arranged on a second axis. [Prior Art Documents] [Patent Documents]

[0003] [Patent Document 1] International Publication No. 2024 / 143033 [Summary of the Invention] [Problem to be Solved by the Invention]

[0004] However, in the prior art, there is room for improving prediction accuracy. For example, in machine learning, it is known that the greater the amount of training data, the higher the prediction accuracy, but collecting a large amount of training data requires enormous cost.

[0005] The present disclosure provides a technique for accurately predicting properties of a material. [Means for Solving the Problem]

[0006] A design support device according to a first aspect of the present disclosure is a design support device having a control unit, the control unit acquires first learning data including structural information indicating the structure of a material and characteristic information indicating the properties of the material, generates second learning data which is an extension of the first learning data by performing a sensitivity analysis on the first learning data, and predicts the properties of the material based on the second learning data.

[0007] According to the first aspect of this disclosure, the properties of the material can be predicted with high accuracy.

[0008] A second aspect of this disclosure is a design support device according to the first aspect, wherein the control unit generates a plurality of structural information that are candidates for the structure, and searches for structural information among the plurality of structural information whose predicted values ​​of the characteristics satisfy the required characteristics.

[0009] According to a second aspect of this disclosure, it is possible to accurately search for a material structure that satisfies the required characteristics.

[0010] A third aspect of the present disclosure is a design support device relating to the first or second aspect, wherein the control unit generates second structural information obtained by changing the structural information included in the first learning data, calculates second characteristic information based on the results of sensitivity analysis of the second structural information, and generates second learning data obtained by adding the second structural information and the second characteristic information to the first learning data.

[0011] According to a third aspect of this disclosure, the training data used to predict characteristics can be extended.

[0012] A fourth aspect of this disclosure is a design support device according to any of the first to third aspects, wherein the control unit generates a trained model that outputs characteristic information when structural information is input, based on the second training data, and predicts the characteristics of the material by inputting the structural information to be predicted into the trained model.

[0013] According to a fourth aspect of this disclosure, material properties can be predicted with high accuracy based on augmented training data.

[0014] A fifth aspect of this disclosure is a design support device relating to any of the first to fourth aspects, wherein the material includes an organic material.

[0015] According to a fifth aspect of this disclosure, the properties of organic materials can be predicted with high accuracy.

[0016] A sixth aspect of this disclosure is a design support device relating to any of the first to fifth aspects, wherein the characteristics include at least one of mechanical properties, thermal properties, electrical properties, heat resistance, wear resistance, or dimensional stability.

[0017] According to a sixth aspect of this disclosure, mechanical properties, thermal properties, electrical properties, heat resistance, wear resistance, or dimensional stability can be predicted with high accuracy.

[0018] A method according to a seventh aspect of the present disclosure involves a control unit of a design support device acquiring first learning data including structural information indicating the structure of a material and characteristic information indicating the properties of the material, performing sensitivity analysis on the first learning data to generate second learning data which is an extension of the first learning data, and predicting the properties of the material based on the second learning data.

[0019] An eighth aspect of this disclosure is a method relating to a seventh aspect, wherein the control unit generates a plurality of structural information that are candidates for the structure, and searches for structural information among the plurality of structural information in which the predicted value of the characteristics satisfies the required characteristics.

[0020] A method for manufacturing a material according to a ninth aspect of the present disclosure includes the step of manufacturing the material based on the structural information retrieved by the method according to the eighth aspect.

[0021] According to the ninth aspect of this disclosure, a material that satisfies the required characteristics can be manufactured.

[0022] A program according to a tenth aspect of the present disclosure causes a control unit included in a design support apparatus to execute processes of: acquiring first learning data including structure information indicating a structure of a material and characteristic information indicating characteristics of said material; performing sensitivity analysis on said first learning data to generate second learning data obtained by expanding said first learning data; and predicting characteristics of said material based on said second learning data.

[0023] An organic material according to an eleventh aspect of the present disclosure is found using the design support apparatus according to any one of the first to sixth aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] [Figure 1] It is a block diagram showing an example of the hardware configuration of a design support apparatus. [Figure 2] It is a flowchart showing an example of a learning process. [Figure 3] It is a flowchart showing an example of a proposal process. MODES FOR CARRYING OUT THE INVENTION

[0025] Each embodiment will be described below with reference to the accompanying drawings. In the present specification and the drawings, constituent elements having substantially the same functional configuration are denoted by the same reference numerals, and redundant explanations are omitted.

[0026] [Embodiment] One embodiment of the present disclosure is an example of an information processing apparatus that supports material design. Hereinafter, the information processing apparatus according to the present embodiment will be referred to as a "design support apparatus".

[0027] The design support apparatus has a function of predicting characteristics of a material. The design support apparatus also has a function of proposing a material structure that satisfies required characteristics. In order to propose a material structure, the design support apparatus predicts characteristics of the material based on material structure information, and searches for structure information for which a predicted characteristic value satisfies the required characteristics.

[0028] In this embodiment, the objective is to accurately predict the properties of a material. To this end, in this embodiment, the training data, which includes structural information indicating the structure of the material and property information indicating the properties of the material, is expanded by performing sensitivity analysis on the training data, and the properties of the material are predicted based on the expanded training data.

[0029] Sensitivity analysis is a method for evaluating the impact of input changes on output. For example, it is used to derive the direction of updating design variables based on gradient information (derivative) of the target variable when seeking the optimal solution (i.e., the optimal structure) for a topology optimization problem. According to sensitivity analysis, for example, the amount of change in properties (e.g., stiffness) when the structure of a material (e.g., void location) is slightly changed can be estimated.

[0030] In one respect, this embodiment allows for accurate prediction of material properties even with a small amount of collectible training data, because the training data is expanded through sensitivity analysis. In another respect, this embodiment allows for accurate prediction of material properties, enabling the accurate proposal of material structures that satisfy the required properties. This shortens the lead time in the development of materials or products manufactured using those materials.

[0031] <Hardware configuration of the design support system> Figure 1 is a block diagram showing an example of the hardware configuration of a design support device. As shown in Figure 1, the design support device 100 includes a processor 101, memory 102, auxiliary storage device 103, operating device 104, display device 105, communication device 106, and drive device 107. Each piece of hardware in the design support device 100 is interconnected via a bus 108.

[0032] The processor 101 has various computing devices such as a CPU (Central Processing Unit). The processor 101 reads various programs installed in the auxiliary storage device 103 into the memory 102 and executes them.

[0033] Memory 102 has main memory devices such as ROM (Read Only Memory) and RAM (Random Access Memory). The processor 101 and memory 102 form a so-called computer (hereinafter also referred to as the "control unit"), and the computer realizes various functions by the processor 101 executing various programs read into memory 102.

[0034] The auxiliary storage device 103 (hereinafter also referred to as the "storage unit") stores various programs and various data used when these programs are executed by the processor 101.

[0035] The operating device 104 is an operating device for the user of the design support device 100 to perform various operations. The display device 105 is a display device that displays the processing results of various processes performed by the design support device 100.

[0036] The communication device 106 is a communication device for communicating with external devices via a communication network.

[0037] The drive device 107 is a device for setting the storage medium 109. The storage medium 109 here includes media that store information optically, electrically, or magnetically, such as CD-ROMs, flexible disks, and magneto-optical disks. The storage medium 109 may also include semiconductor memory that stores information electrically, such as ROMs and flash memory.

[0038] The various programs to be installed in the auxiliary storage device 103 are installed, for example, when the distributed storage medium 109 is set in the drive device 107 and the various programs stored in the storage medium 109 are read by the drive device 107. Alternatively, the various programs to be installed in the auxiliary storage device 103 may be installed by downloading them from the network via the communication device 106.

[0039] <Flowchart of Design Support Methods> The design support method performed by the design support device 100 will be explained with reference to Figures 2 and 3. The design support method includes a learning process (see Figure 2) and a proposal process (see Figure 3).

[0040] ≪Learning Process≫ Figure 2 is a flowchart illustrating an example of the learning process. The learning process generates a trained model for predicting material properties.

[0041] In step S1, the control unit of the design support device 100 acquires learning data. The learning data may include structural information indicating the structure of the material and characteristic information indicating the properties of the material. The control unit may read learning data that has been pre-stored in the memory unit. The control unit may also accept input of learning data from the user of the design support device 100.

[0042] Hereafter, the training data obtained in step S1 will also be referred to as "initial training data." The initial training data may be small in quantity. The initial training data may not be sufficient to generate a trained model. The initial training data is an example of the first training data.

[0043] The material may include organic materials. Examples of organic materials include polymer materials or rubber materials. The material may also include heterogeneous materials. Heterogeneous materials may include composite materials obtained by kneading polymer materials or rubber materials with fillers or fibers. For example, a composite material may be obtained by kneading polyimide and a thermally conductive filler.

[0044] The material structure may include phase-separated structures, crystalline and amorphous structures, and may also include the dispersion structure of composite materials. Structural information may include, for example, morphological information such as the internal structure and microstructure of the material, including the shape and orientation of particles and pores. Structural information may also include information showing the particle dispersion structure calculated by particle dispersion simulation. Structural information may also include information showing the structure measured experimentally.

[0045] The material properties may include at least one of the following: mechanical properties, thermal properties, electrical properties, heat resistance, wear resistance, or dimensional stability. Examples of material properties may include thermal conductivity, thermal diffusivity, electrical conductivity, dielectric properties, coefficient of linear expansion, tensile strength, modulus of elasticity, temperature resistance, compressibility, surface tension, surface free energy, Young's modulus, Poisson's ratio, or hygroscopicity. The property information may include information showing property values ​​calculated by CAE (Computer-Aided Engineering) analysis. The structural information may include information showing measured property values ​​obtained through experiments.

[0046] In step S2, the control unit of the design support device 100 generates structural information by sensitivity analysis. The control unit may generate new structural information based on the structural information contained in the training data acquired in step S1. The control unit may generate new structural information by changing the structural information contained in the training data. The control unit may make minute changes to the structural information contained in the training data. The control unit may make local changes to the structural information contained in the training data. A minute change means that the amount of change before and after the change is small. Local changes mean that only a part of the structure is changed.

[0047] The control unit may generate one or more structural information. The number of structural information generated in step S2 may be determined according to the number of initial training data. The control unit may generate a sufficient amount of structural information to generate a trained model. Note that new structural information only needs to differ slightly from existing structural information, so there is no upper limit to the number of structural information that can be generated. New structural information is an example of second structural information.

[0048] In step S3, the control unit of the design support device 100 generates characteristic information. The control unit may generate new characteristic information based on the structural information generated in step S2. The control unit may generate new characteristic information by performing a sensitivity analysis on the structural information. The control unit may generate new characteristic information based on the results of the sensitivity analysis on the structural information. The results of the sensitivity analysis on the structural information may include the amount of change in characteristics caused by changes in the structural information. The control unit may generate new characteristic information by changing the characteristic information contained in the training data by the amount of change calculated by the sensitivity analysis. The new characteristic information is an example of second characteristic information.

[0049] In step S4, the control unit of the design support device 100 expands the training data. The control unit may expand the initial training data acquired in step S1. The control unit may also expand the initial training data by adding the structural information generated in step S2 and the characteristic information generated in step S3 to the initial training data.

[0050] Hereinafter, the training data acquired in step S2 will also be referred to as "enhanced training data." The enhanced training data is sufficient in quantity to generate a trained model. The control unit may store the enhanced training data in the memory unit. The enhanced training data is an example of second training data.

[0051] In step S5, the control unit of the design support device 100 generates a trained model. The control unit may generate a trained model based on the expanded training data generated in step S4. The control unit may also generate a trained model by learning the relationship between structural information and characteristic information contained in the expanded training data. The trained model may be a machine learning model that takes structural information as input and outputs predicted values ​​of characteristic information.

[0052] The trained model may be any type of machine learning model. Examples of machine learning models include linear regression, logistic regression, LASSO (Least Absolute Shrinkage and Selection Operator) regression, Gaussian process regression (GPR), partial least squares regression (PLS), neural networks, random forests, or gradient boosting decision trees. Examples of gradient boosting decision trees include LightGBM (Light Gradient Boosting Machine) or XGBoost (eXtreme Gradient Boosting).

[0053] In step S6, the control unit of the design support device 100 stores the trained model. The control unit may store the trained model generated in step S5. The control unit may store the trained model in the storage unit.

[0054] ≪Proposal Processing≫ Figure 3 is a flowchart showing an example of the proposal process. The proposal process is a process of proposing a material structure that satisfies the required characteristics.

[0055] In step S11, the control unit of the design support device 100 acquires the required characteristics. The required characteristics may include target values ​​for the material properties. The required characteristics may include target values ​​for each of a plurality of characteristics. The target values ​​may be a numerical range that includes at least one of an upper limit or a lower limit. The control unit may read out the required characteristics that have been previously stored in the memory unit. The control unit may accept input of required characteristics from the user of the design support device 100.

[0056] In step S12, the control unit of the design support device 100 generates candidate structures. Candidate structures are structural information that can be candidates for the proposed structure. The control unit may generate candidate structures randomly. The control unit may generate candidate structures based on structural information contained in the expanded training data. The control unit may generate new candidate structures based on candidate structures generated in step S12 previously executed.

[0057] The control unit may generate one or more candidate structures. The number of candidate structures generated in step S12 may be predetermined. The control unit may generate multiple candidate structures each time step S12 is executed. The control unit may generate one candidate structure each time step S12 is executed and generate multiple candidate structures by repeating step S12 multiple times.

[0058] In step S13, the control unit of the design support device 100 predicts the material properties. The control unit may predict the material properties based on the expanded training data. The control unit may predict the material properties based on a trained model generated based on the expanded training data. The control unit may predict the material properties based on a trained model read from the storage unit. The control unit may predict the material properties based on the candidate structures generated in step S12. The control unit may predict the material properties by inputting the candidate structures to be predicted into the trained model. The control unit may predict the material properties for each of the multiple candidate structures.

[0059] In step S14, the control unit of the design support device 100 determines whether the candidate structure satisfies the required characteristics. The control unit may also determine whether the predicted values ​​of the characteristics predicted based on the candidate structure satisfy the required characteristics. The control unit may also determine whether the predicted values ​​of the characteristics predicted in step S13 satisfy the required characteristics obtained in step S11. The control unit may determine that the required characteristics are satisfied if the predicted values ​​of the characteristics satisfy the target values ​​indicated in the required characteristics. If the required characteristics include target values ​​for multiple characteristics, the control unit may determine that the required characteristics are satisfied if the predicted values ​​of the characteristics satisfy the target values ​​for all characteristics.

[0060] If the control unit determines that the candidate structure satisfies the required characteristics (YES), it proceeds to step S15. On the other hand, if the control unit determines that the candidate structure does not satisfy the required characteristics (NO), it skips step S15 and proceeds to step S16.

[0061] In step S15, the control unit of the design support device 100 stores candidate structures in the storage unit. The control unit may also store candidate structures that it determined in step S14 to satisfy the required characteristics in the storage unit. The control unit may control the number of candidate structures stored in the storage unit to be less than or equal to a predetermined number. For example, if the control unit has stored candidate structures with lower predicted characteristic values ​​than the candidate structures that it determined to satisfy the required characteristics, it may store the candidate structures that it determined to satisfy the required characteristics in the storage unit and delete the candidate structures stored in the storage unit.

[0062] In step S16, the control unit of the design support device 100 determines whether or not to terminate the search. Whether or not to terminate the search can be determined by whether or not a predetermined number of candidate structures have been stored in the memory device, whether or not a predetermined number of iterations has been exceeded, whether or not a candidate structure that satisfies the required characteristics has been obtained, etc. If it is determined to terminate the search (YES), the control unit proceeds to step S17. On the other hand, if it is determined not to terminate the search (NO), the control unit returns to step S12.

[0063] Returning to step S12, the control unit of the design support device 100 generates a new candidate structure. At this time, the control unit may generate a new candidate structure based on the already generated candidate structure. The control unit may also generate a new candidate structure based on the candidate structure stored in the memory unit (in other words, a candidate structure that satisfies the required characteristics). For example, the control unit may generate a new candidate structure by changing the candidate structure stored in the memory unit.

[0064] Subsequently, the control unit of the design support device 100 repeats the processes from step S13 to step S16 based on the new candidate structure. In this way, the control unit repeatedly executes the processes from step S12 to step S16 until it determines in step S16 that the search is complete. In other words, the control unit searches for a candidate structure among the multiple candidate structures generated in step S12 in which the predicted values ​​of the characteristics predicted in step S13 satisfy the required characteristics obtained in step S11.

[0065] In step S17, the control unit of the design support device 100 outputs proposed information. The proposed information includes candidate structures that satisfy the required characteristics obtained in step S11. The control unit may present the proposed information to the user of the design support device 100. For example, the control unit may display the proposed information on the display device 105. The control unit may transmit the proposed information to an external information processing device or information processing system.

[0066] The user of the design support device 100 may refer to the presented proposal information and use it for the design, development, or manufacture of materials. The user may determine new required properties based on the characteristic information contained in the proposal information and run the proposal process again. This allows the user to explore materials whose properties may be further improved. The user may also experiment with materials based on the structural information shown in the proposal information. This allows the user to verify whether a material with the proposed structure meets the required properties.

[0067] Furthermore, the user of the design support device 100 may manufacture materials based on the structural information shown in the proposed information. This allows the user to manufacture materials that meet the required characteristics. The user may then manufacture products using the manufactured materials. This allows the user to manufacture products with superior performance.

[0068] <Summary> As described above, according to one embodiment of the present disclosure, the properties of a material can be predicted with high accuracy. For example, the design support device 100 generates second learning data by extending the first learning data, which includes structural information indicating the structure of the material and characteristic information indicating the properties of the material, by performing sensitivity analysis on first learning data, and predicts the properties of the material based on the second learning data.

[0069] In one respect, this embodiment allows for accurate prediction of material properties even with a small amount of collectible training data, because the training data is expanded through sensitivity analysis. In another respect, this embodiment allows for accurate prediction of material properties, enabling the accurate proposal of material structures that satisfy the required properties. This shortens the lead time in the development of materials or products manufactured using those materials.

[0070] The design support device 100 may generate multiple structural information pieces that serve as candidates for the material's structure. The design support device 100 may search among the multiple structural information pieces for structural information pieces whose predicted properties satisfy the required properties. In one aspect, according to this embodiment, a material structure that satisfies the required properties can be searched with high accuracy.

[0071] The design support device 100 may generate second structural information by changing the structural information contained in the first training data. The design support device 100 may calculate second characteristic information based on the results of sensitivity analysis of the second structural information. The design support device 100 may generate second training data by adding the second structural information and the second characteristic information to the first training data. In one aspect, according to this embodiment, the training data used to predict characteristics can be expanded.

[0072] The design support device 100 may generate a trained model based on second training data that outputs characteristic information when structural information is input. The design support device 100 may predict the material properties by inputting the structural information to be predicted into the trained model. In one aspect, according to this embodiment, the material properties can be predicted with high accuracy based on the extended training data.

[0073] The material may include organic materials. In one respect, according to this embodiment, the properties of organic materials can be predicted with high accuracy.

[0074] The characteristics may include at least one of mechanical, thermal, or electrical characteristics. In one aspect, according to this embodiment, mechanical, thermal, or electrical characteristics can be predicted with high accuracy.

[0075] The user of the design support device 100 may manufacture materials based on the retrieved structural information. In one aspect, according to this embodiment, materials that satisfy the required characteristics can be manufactured.

[0076] [supplement] Each of the embodiments described above can be implemented by one or more processing circuits. Hereinafter, "processing circuit" as used herein includes processors programmed to execute each function by software, such as CPUs (Central Processing Units) or GPUs (Graphics Processing Units) implemented by electronic circuits, as well as devices such as ASICs (Application Specific Integrated Circuits), DSPs (Digital Signal Processors), FPGAs (Field Programmable Gate Arrays), and conventional circuit modules designed to execute each of the functions described above.

[0077] Although embodiments have been described above, it should be understood that various modifications to the form and details are possible without departing from the spirit and scope of the claims. [Explanation of symbols]

[0078] 100 Design Support Devices

Claims

1. A design support device having a control unit, The control unit, First learning data is obtained, which includes structural information indicating the structure of the material and characteristic information indicating the properties of the material. By performing sensitivity analysis on the first training data, a second training data is generated that is an extension of the first training data. Based on the second learning data, the properties of the material are predicted. Design support equipment.

2. The control unit, Multiple structural information candidates for the aforementioned structure are generated, The system searches for structural information among the aforementioned multiple structural information pieces in which the predicted value of the characteristic satisfies the required characteristic. The design support device according to claim 1.

3. The control unit, A second structural information is generated by changing the structural information contained in the first training data. Based on the results of sensitivity analysis of the second structural information, the second characteristic information is calculated. The second training data is generated by adding the second structural information and the second characteristic information to the first training data. The design support device according to claim 1.

4. The control unit, Based on the second training data, a trained model is generated that outputs characteristic information when the structural information is input. By inputting the structural information to be predicted into the trained model, the properties of the material are predicted. The design support device according to claim 1.

5. The aforementioned material includes an organic material. A design support device according to any one of claims 1 to 4.

6. The aforementioned properties include at least one of mechanical properties, thermal properties, electrical properties, heat resistance, wear resistance, or dimensional stability. A design support device according to any one of claims 1 to 4.

7. The control unit of the design support device, First learning data is obtained, which includes structural information indicating the structure of the material and characteristic information indicating the properties of the material. By performing sensitivity analysis on the first training data, a second training data is generated that is an extension of the first training data. Based on the second learning data, the properties of the material are predicted. method.

8. The control unit, Multiple structural information candidates for the aforementioned structure are generated, The system searches for structural information among the aforementioned multiple structural information pieces in which the predicted value of the characteristic satisfies the required characteristic. The method according to claim 7.

9. The process includes manufacturing the material based on the structural information obtained by the method of claim 8, Method for manufacturing materials.

10. The control unit of the design support device has First learning data is obtained, which includes structural information indicating the structure of the material and characteristic information indicating the properties of the material. By performing sensitivity analysis on the first training data, a second training data is generated that is an extension of the first training data. Based on the second learning data, the properties of the material are predicted. A program to execute a process.

11. An organic material discovered using the design support apparatus described in any one of claims 1 to 4.

Images