Proposed method, information processing device, and program

Abstract

We provide formulation proposal technology that enables material design that takes into account robustness to processing conditions. [Solution] The information processing device 10 comprises a candidate generation unit 110 and a calculation unit 130. The candidate generation unit 110 generates candidate information. The candidate information includes blending information indicating the blending ratio of each of several raw materials. The calculation unit 130 uses relational information to calculate a robust index. The relational information is information showing the relationship between processing conditions and the characteristic values ​​of the material manufactured according to the blending ratio and processing conditions. The robust index is an index that shows the degree of sensitivity of the characteristic values ​​of the material to the processing conditions.

Description

Technical Field

[0001] The present invention relates to a proposal method, an information processing apparatus, and a program.

Background Art

[0002] In the formulation design for obtaining a desired material, attempts have been made to utilize a computer.

[0003] Patent Document 1 describes incorporating interference factors that affect carbonaceous parameters into modeling. It also describes performing solution finding by robust optimization.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in Patent Document 1, the interference factors include elements of the carbonaceous material itself, sampling elements, and chemical inspection elements, and the robustness against processing conditions has not been considered.

[0006] One aspect of the present invention provides a formulation proposal technique that enables material design considering robustness against processing conditions.

Means for Solving the Problems

[0007] According to one form of the present invention, the following proposal method, information processing apparatus, and program are provided.

[0008] (1) A proposal method executed by one or more computers, A candidate generation step that generates candidate information including formulation information showing the blending ratio of each of several raw materials, The calculation step includes using relational information showing the relationship between processing conditions, the mixing ratio, and the characteristic values ​​of the material produced according to the processing conditions, to calculate a robust index indicating the degree of sensitivity of the characteristic values ​​to the processing conditions. Proposal method. (2) In the proposed method described in (1), The candidate generation step and the calculation step are repeated until a predetermined termination condition is met. Proposal method. (3) In the proposed method described in (2), In the candidate generation step, the next candidate information is generated based on the robust index calculated for one candidate piece of information. Proposal method. (4) In the proposed method described in (3), In the candidate generation step, the following candidate information is generated using an optimization algorithm. Proposal method. (5) In the proposed method described in (3) or (4), The method further includes a prediction step of predicting the characteristic values ​​of the material using the candidate information, In the candidate generation step, the next candidate information is generated based on the prediction result in the prediction step for the one candidate information. Proposal method. (6) In the proposed method described in any one of (1) to (5), The candidate information further includes processing information indicating the processing conditions. Proposal method. (7) In any of the proposed methods described in (1) to (6), The aforementioned relational information is a mathematical formula or a trained model. Proposal method. (8) In the proposed method described in any one of (1) to (6), The aforementioned related information is experimental data. Proposal method. (9) A candidate generation unit that generates candidate information including formulation information showing the blending ratio of each of the multiple raw materials, The system includes a calculation unit that uses relational information showing the relationship between processing conditions and the characteristic values ​​of the material produced according to the mixing ratio and processing conditions to calculate a robust index indicating the degree of sensitivity of the characteristic values ​​to the processing conditions. Information processing device. (10) Computers, Candidate generation means for generating candidate information including formulation information showing the blending ratio of each of multiple raw materials, and Using relational information showing the relationship between processing conditions, the mixing ratio, and the characteristic values ​​of the material manufactured according to the processing conditions, this system functions as a calculation means for calculating a robust index indicating the degree of sensitivity of the characteristic values ​​to the processing conditions. program. [Effects of the Invention]

[0009] According to one aspect of the present invention, a compounding proposal technology is provided that enables material design that takes into account robustness to processing conditions. [Brief explanation of the drawing]

[0010] [Figure 1] This is a diagram showing an overview of the information processing device according to the first embodiment. [Figure 2] This figure shows an overview of the proposed method according to the first embodiment. [Figure 3] This graph illustrates the relationship between processing conditions for obtaining a material and the characteristic values ​​of that material. [Figure 4] This diagram illustrates the functional configuration of the information processing device according to the first embodiment. [Figure 5] This is a flowchart illustrating the processing flow performed by the information processing device according to the first embodiment. [Figure 6] This is a table illustrating the output data produced by the output unit. [Figure 7] This diagram illustrates a computer used to implement an information processing device. [Figure 8]This figure illustrates the functional configuration of an information processing device according to the second embodiment. [Figure 9] This is a flowchart illustrating the processing flow performed by the information processing device according to the second embodiment. [Figure 10] This table illustrates the relationship information acquired by the candidate generation unit. [Figure 11] This figure illustrates a graph that approximates relational information using a quadratic function. [Figure 12] This figure illustrates the functional configuration of an information processing device according to the third embodiment. [Figure 13] This is a flowchart illustrating the processing flow performed by the information processing device according to the third embodiment. [Figure 14] This figure illustrates the functional configuration of an information processing device according to the fifth embodiment. [Figure 15] This is a flowchart illustrating the processing flow performed by the information processing device according to the fifth embodiment. [Modes for carrying out the invention]

[0011] Embodiments of the present invention will be described below with reference to the drawings. In all drawings, similar components are denoted by the same reference numerals, and their descriptions are omitted as appropriate.

[0012] (First embodiment) Figure 1 is a diagram showing an overview of an information processing device 10 according to the first embodiment. The information processing device 10 comprises a candidate generation unit 110 and a calculation unit 130. The candidate generation unit 110 generates candidate information. The candidate information includes blending information that shows the blending ratio of each of a plurality of raw materials. The calculation unit 130 calculates a robust index using relational information. The relational information is information that shows the relationship between processing conditions and the characteristic values ​​of the material manufactured according to the blending ratio and processing conditions. The robust index is an index that shows the degree of sensitivity of the characteristic values ​​of the material to the processing conditions.

[0013] Figure 2 is a diagram illustrating an overview of the proposed method according to this embodiment. The proposed method according to this embodiment is executed by one or more computers. The proposed method according to this embodiment includes a candidate generation step S10 and a calculation step S20. In the candidate generation step S10, one or more computers generate candidate information. The candidate information includes blending information that shows the blending ratio of each of a plurality of raw materials. In the calculation step S20, one or more computers calculate a robust index using relational information. The relational information is information that shows the relationship between processing conditions and the characteristic values ​​of the material manufactured according to the blending ratio and processing conditions. The robust index is an index that shows the degree of sensitivity of the characteristic values ​​of the material to the processing conditions.

[0014] In some cases, a composition containing multiple raw materials is processed to obtain a material exhibiting desired properties. In this case, the properties of the obtained material may depend on the processing conditions (e.g., heating temperature and heating time) used to obtain the material. While processing conditions for obtaining a material with desired properties are usually defined, these conditions cannot always be adhered to. If even slight variations in processing conditions cause large variations in the properties of the obtained material, the material may no longer exhibit the desired properties. Therefore, it is preferable that, at least near the desired properties, the variation in the material's properties is small in response to variations in processing conditions. In other words, it is preferable that the material's properties have low sensitivity to processing conditions, i.e., that the material's properties have high robustness to processing conditions.

[0015] Figure 3 is a graph illustrating the relationship between processing conditions and material properties for obtaining a material. Curve 91 indicates a more robust situation than curve 92. In the situation shown by curve 91, the desired material can be obtained over a relatively wide range, such as range 910, with values ​​exceeding the lower limit. On the other hand, in the situation shown by curve 92, the desired material can only be obtained within a limited range, such as range 920. In other words, processing conditions must be strictly controlled, which increases manufacturing costs.

[0016] As explained above, it is preferable to design the formulation to ensure robustness to processing conditions. In contrast, according to this embodiment, the calculation unit 130 calculates a robustness index. Therefore, it becomes possible to design materials that take robustness to processing conditions into consideration.

[0017] Figure 4 is a diagram illustrating the functional configuration of the information processing device 10 according to this embodiment. In the example in Figure 4, the information processing device 10 further comprises a prediction unit 150, an output unit 190, and a relational information storage unit 102. However, the relational information storage unit 102 may be a storage device provided outside the information processing device 10. Figure 5 is a flowchart illustrating the flow of processing performed by the information processing device 10 according to this embodiment.

[0018] Candidate information includes at least formulation information. Candidate information can also be described as information that shows a recipe for obtaining the material. Formulation information shows the mixing ratio of each of several raw materials. Formulation information can be, for example, a vector. Formulation information may also be information that shows the mixing ratio for each raw material included in a given group of raw materials. The mixing ratio may be, for example, a value that shows the volume ratio or mass ratio of the raw materials in the material. In the formulation information, the mixing ratios may be normalized so that the sum of the mixing ratios of all the raw materials is 1. Formulation information may also be a vector whose elements are the mixing ratios of each raw material. Here, in the formulation information, zero may be shown as the mixing ratio of raw materials that are not included in the material from the given group of raw materials.

[0019] Candidate information may further include processing information indicating processing conditions. Multiple processing steps may be required to obtain the material. Also, there may be multiple processing condition items for obtaining the material. A processing step is a step performed at at least one of the following points in the process from multiple raw materials to obtaining the material. A processing step is, for example, one or more of the following: heat treatment, cooling treatment, mixing treatment, pressurization treatment, vacuum treatment, oxidation treatment, alkalizing treatment, and neutralization treatment. However, the processing steps are not limited to these examples. The processing information may include information on multiple processing steps that differ from each other in at least one of their content and timing.

[0020] Examples of processing conditions include processing time, processing temperature, heating rate, cooling rate, rotation speed, pressure, and pH. In other words, processing conditions are values ​​that represent one or more of the following: processing time, processing temperature, heating rate, cooling rate, rotation speed, pressure, and pH. However, some or all of the processing conditions for obtaining the material may be predetermined. That is, the values ​​of some of the multiple processing conditions necessary to obtain the material may be predetermined, and the values ​​of the remaining items may be shown in the processing information. The user of the information processing device 10 may be able to input some or all of the processing conditions for obtaining the material to the information processing device 10.

[0021] The material is obtained by mixing multiple raw materials in the proportions indicated in the formulation information, and by performing a process according to the processing conditions at at least one of the steps in obtaining the material from the multiple raw materials. For example, a composition is prepared by mixing multiple raw materials in the proportions indicated in the formulation information, and the material is obtained by processing the resulting composition according to the processing conditions described above. In this case, the composition may be distributed and processed by the purchaser.

[0022] The material may be an organic material, an inorganic material, or an organic-inorganic composite material. The application of the material is not particularly limited, but examples include electronic equipment or semiconductor devices, energy-related applications, medical applications, bio-applications, automotive applications, and aerospace applications. Examples of materials for electronic equipment or semiconductor devices include semiconductor encapsulants. The material may be a liquid material or a solid material. The material may be in paste form. The material may be in the form of a molded product. In that case, the processing conditions may be molding conditions.

[0023] Examples of raw materials include resins, fillers, and additives. Multiple raw materials may contain multiple different fillers. Multiple raw materials may contain multiple different resins. Multiple raw materials may be identified by part numbers or the like.

[0024] The resin is not particularly limited, but for example, the raw materials may include one or more resins selected from the group consisting of epoxy resins, phenolic resins, melamine resins, unsaturated polyester resins, divinylbenzene polymers, divinylbenzene-styrene copolymers, divinylbenzene-acrylic acid ester copolymers, diallyl phthalate polymers, triallyl isocyanurate polymers, and benzoguanamine polymers.

[0025] The fillers are not particularly limited, but for example, multiple raw materials may include one or more fillers selected from the group consisting of conductive particles and non-conductive particles. Examples of conductive particles include gold, silver, copper, carbon black, tin oxide, metal plating particles, etc. Examples of non-conductive particles include silica, alumina, diamond, boron nitride, organic fillers, etc. Multiple raw materials may include one or more fillers selected from the group consisting of silver, gold, copper, carbon black, tin oxide, metal plating particles, silica, alumina, diamond, boron nitride, and organic fillers.

[0026] Furthermore, the multiple raw materials may include one or more selected from the group consisting of curing agents, coupling agents, curing accelerators, diluents, radical initiators, and stress reducers. The multiple raw materials may or may not contain solvents. In addition, the multiple raw materials may include compositions as raw materials.

[0027] However, the raw materials are not limited to these examples.

[0028] The candidate generation unit 110 may, for example, generate candidate information randomly. Specifically, the candidate generation unit 110 may randomly determine the blending ratio for each raw material included in a predetermined group of raw materials and generate blending information indicating the blending ratio of each raw material. The candidate generation unit 110 includes the generated blending information in the candidate information.

[0029] Furthermore, the candidate generation unit 110 may use randomly identified processing conditions as processing information, or it may use processing conditions randomly selected from a predetermined set of options as processing information.

[0030] Constraints may be set for the candidate information. Constraints may be input to the information processing device 10 by the user. The candidate generation unit 110 may generate candidate information in a manner that satisfies the constraints. That is, the candidate generation unit 110 may generate candidate information randomly within a range that satisfies the constraints. In this way, candidate information that satisfies the desired conditions can be obtained.

[0031] Relational information is information that shows the relationship between processing conditions and material property values. In this embodiment, relational information is, for example, a regression model. The explanatory variables of this regression model include the blending ratio of each of several raw materials and one or more processing conditions. The dependent variable of this regression model includes material property values. Relational information may also be a mathematical formula that shows the relationship between processing conditions and material property values. Relational information may also be a trained model obtained by machine learning. Even when relational information is a trained model, it is possible to express the relationship between processing conditions and material property values ​​mathematically based on the trained model. In this embodiment, relational information is generated by an existing method based on previously acquired data. Alternatively, relational information may be a theoretical formula. In this embodiment, relational information is stored in advance in the relational information storage unit 102.

[0032] Furthermore, the explanatory variables of the regression model may include the characteristic values ​​of one or more raw materials included in the multiple raw materials. In that case, the candidate information generated by the candidate generation unit 110 may include the characteristic values ​​of the raw materials.

[0033] In this embodiment, the prediction unit 150 reads and acquires relational information from the relational information storage unit 102. The prediction unit 150 can then predict the material's characteristic values ​​using the candidate information generated by the candidate generation unit 110 and the relational information. Specifically, the prediction unit 150 inputs the blending ratio and processing conditions included in the candidate information into the model, which serves as relational information. The prediction unit 150 may also set predetermined values ​​as part of the multiple explanatory variables input into the model. The prediction unit 150 can obtain the predicted characteristic values ​​as the output of the model.

[0034] The property values ​​of a material are values ​​for specific properties. Examples of properties include mechanical properties, thermal properties, optical properties, chemical properties, and physical properties. Examples of mechanical properties include tensile strength, tensile modulus, and compressive modulus. Examples of thermal properties include melting point, glass transition temperature (Tg), thermal expansion coefficient, and thermal conductivity. Examples of optical properties include refractive index and absorbance. Examples of chemical properties include acid resistance and alkali resistance. Examples of physical properties include density and specific gravity.

[0035] The robustness index is an index that indicates the degree of sensitivity of a material's characteristic values ​​to processing conditions. As described above, a single material may have multiple characteristics, and there may be multiple items for processing conditions, but one robustness index is calculated for each combination of a single characteristic and a single processing condition item. The method for calculating the robustness index will be described below, but in this embodiment, the calculation unit 130 may similarly calculate the robustness index for each of the multiple processing condition items. Alternatively, the calculation unit 130 may calculate the robustness index for each of the multiple characteristics. Relational information may be prepared for each material characteristic and stored in the relational information storage unit 102.

[0036] The degree of sensitivity of a material's property values ​​to processing conditions refers to the extent to which those property values ​​change when the processing conditions are altered. By calculating a robustness index, it becomes possible to quantitatively address robustness.

[0037] The method for calculating robustness metrics is described below. However, the calculation method for robustness metrics is not limited to the examples below.

[0038] For the purpose of explanation, the value of a certain property of the material is given by variable y. i This is shown as follows, and a certain processing condition is expressed as variable x j This is shown as follows. And the robustness index for its characteristics and processing conditions is given by R ij This is shown as follows.

[0039] Also, the reference point (x j,0 ,yi,0 ) is defined. The reference value x of the processing conditions j,0 In this case, the characteristic value of the material predicted by the regression model is y i,0 is. The robust index R ij is calculated as the degree of sensitivity of the characteristic value of the material to the processing conditions in the vicinity of the reference point. The reference value x of the processing conditions j,0 may be input to the information processing apparatus 10 by the user of the information processing apparatus 10, or may be the value indicated in the processing information of the candidate information. Further, for the regression model, together with other explanatory variables, the reference value x is used as the processing condition j,0 By inputting, the characteristic value y i,0 can be derived. At this time, as other explanatory variables, a value based on the candidate information or a predetermined value may be set.

[0040] Processing condition x j The change of affects the characteristic value y i As an index for evaluating the influence of, there is the normalized sensitivity derivative coefficient S shown in Equation (1) i,j is.

[0041]

Equation

[0042] Furthermore, the normalized sensitivity derivative coefficient S i,j can also be expressed as in Equation (2) using an approximation formula by finite difference. Here, Δx j = ω, and the analysis interval near the reference value is set to [x j,min = x j,0 - ω / 2, x j,max = x j,0 + ω / 2]. The value of ω may be predetermined or the user may be able to input the value of ω to the information processing apparatus 10. Δy j is the change amount of y j corresponding to Δx j in the regression model.

[0043]

Equation

[0044] Robust index R ij This is the normalized sensitivity derivative S i,j It can be calculated based on the following. The smaller the sensitivity, that is, the normalized sensitivity derivative S i,j The smaller the value, the more robust the system is considered to be. Therefore, the robustness index R ij For example, the normalized sensitivity derivative S i,j Based on this, it can be calculated using formula (3). Alternatively, the robust index R ij This may be calculated using equation (4). ε is a small positive constant to avoid numerical instability, for example, ε = 10 -6 It can be done this way.

[0045]

number

number

[0046] Normalized sensitivity derivative S i,j Several examples of methods for calculating this are described below. Here, relational information is given by N explanatory variables x1, x2, ..., x N Regression model y i =f i (x1,x2,...,x N It is expressed as ).

[0047] <Example 1> Regression model y i =f i (x1,x2,...,x N ) is the target explanatory variable x i If it is partially differentiable, the normalized sensitivity derivative S i,j This is calculated by equation (5). In equation (5), the bold x represents the vector (x1, x2, ..., x N This means ). The same applies to equations (6), (7), and (8) below.

[0048]

number

[0049] <Example 2> Regression model y i =f i (x1,x2,...,x N ) is the explanatory variable x of the subject. i If the regression model y is not partially differentiable, i =f i (x1,x2,...,x N ) can be expressed as a form g that is differentiable about a reference point, as shown in equation (6) below. i、x0 (x1,x2,...,x N After converting to (), partial differentiation is performed as in equation (7). The conversion to a form differentiable about a reference point can be done using existing methods such as Local interpretable model-agnostic explanations (LIME).

[0050]

number

number

[0051] <Example 3> Regardless of whether the regression model is partially differentiable, the normalized sensitivity derivative S can be obtained by using finite differences as shown in equation (8). i,j It is also possible to calculate this.

[0052]

number

[0053] The processing flow executed by the information processing device 10 according to this embodiment will be explained using Figure 5. In step S101, the candidate generation unit 110 generates candidate information as described above. In step S102, the prediction unit 150 predicts the characteristic values ​​of the material obtained based on the candidate information using a regression model as relational information, as described above.

[0054] In step S103, the calculation unit 130 reads and obtains the regression model as relational information from the relational information storage unit 102. The calculation unit 130 also identifies the reference value for the processing conditions. Then, the calculation unit 130 calculates the robust index using the regression model and the reference value by one of the methods described above.

[0055] In step S104, the output unit 190 outputs output data that includes candidate information generated by the candidate generation unit 110 and robust indicators calculated by the calculation unit 130. In addition, information indicating reference values ​​for processing conditions may be associated with the robust indicators in the output data. In this embodiment, the output unit 190 may further output output data that includes characteristic values ​​predicted by the prediction unit 150.

[0056] The output destination of the output data from the output unit 190 may be, for example, a display connected to the information processing device 10, a device other than the information processing device 10, or a storage device accessible by the information processing device 10.

[0057] Figure 6 is a table illustrating the output data output by the output unit 190. This table contains multiple candidate information. For each candidate information, the table includes the identification number (No), the blending ratio of each of the multiple raw materials (raw material A, raw material B, and raw material C), processing conditions (heating time and heating temperature), predicted characteristic values ​​of the material (elastic modulus, Tg, and thermal conductivity), and the robustness index (R). Tg,time and R Tg,temp ) is shown. Robust index R Tg,time This indicates the degree of sensitivity of the material's Tg to heating time. Furthermore, this robustness index RTg,time The value is shown as "T0=60" because it was derived with a baseline of 60 min. Robust index R Tg,temp This indicates the degree of sensitivity of the material's Tg to heating temperature. Furthermore, this robustness index R Tg,temp The value "Temp0=150" indicates that it was derived using a baseline value of 150K.

[0058] According to the information processing device 10 of this embodiment, candidate information and a robustness index for the material obtained based on that candidate information are output together. Therefore, the user of the information processing device 10 can quantitatively grasp the level of robustness of the material obtained based on the candidate information.

[0059] Furthermore, in addition to calculating a robust index indicating the degree of sensitivity to processing conditions, the calculation unit 130 may also calculate a robust index indicating the degree of sensitivity to the blending ratio of any of the raw materials. Alternatively, instead of calculating a robust index indicating the degree of sensitivity to processing conditions, the calculation unit 130 may calculate a robust index indicating the degree of sensitivity to the blending ratio of any of the raw materials. Doing so makes it possible to design materials that take into account the tolerance for blending errors of the raw materials. The robust index indicating the degree of sensitivity to the blending ratio is calculated based on the processing conditions x described above. i This is then replaced with a mixing ratio and calculated similarly.

[0060] Furthermore, in addition to calculating a robust index indicating the degree of sensitivity to processing conditions, the calculation unit 130 may also calculate a robust index indicating the degree of sensitivity to the characteristic value of any of the raw materials. Alternatively, instead of calculating a robust index indicating the degree of sensitivity to processing conditions, the calculation unit 130 may calculate a robust index indicating the degree of sensitivity to the characteristic value of any of the raw materials. Doing so makes it possible to design materials that take into account robustness to the characteristic value of the raw materials. The robust index indicating the degree of sensitivity to the characteristic value of the raw materials is calculated based on the processing conditions x described above. i This is then replaced with the characteristic values ​​of the raw materials and calculated similarly.

[0061] As illustrated in Figure 6, the candidate generation unit 110 may generate multiple candidate information. The candidate generation unit 110 may generate the multiple candidate information randomly, or it may generate it using a quasi-Monte Carlo method or the like. When the candidate generation unit 110 generates multiple candidate information, the calculation unit 130 calculates a robust index for each candidate information. The output unit 190 then associates the robust index with each of the multiple candidate information and includes it in the output data.

[0062] When the output unit 190 outputs multiple candidate information, the output unit 190 may include in the output data the results of at least one of the following: classification, filtering, or ranking of the multiple candidate information. At least one of classification, filtering, or ranking can be performed using a robustness index in accordance with predetermined rules.

[0063] The hardware configuration of the information processing device 10 is described below. Each functional component of the information processing device 10 (candidate generation unit 110, calculation unit 130, prediction unit 150, and output unit 190) can be realized by a combination of hardware and software (e.g., a combination of an electronic circuit and a program to control it).

[0064] Figure 7 illustrates a computer 1000 for implementing the information processing device 10. Computer 1000 is any computer. For example, computer 1000 could be an SoC (System on Chip), a Personal Computer (PC), a server machine, a tablet terminal, or a smartphone. Computer 1000 may be a dedicated computer designed to implement the information processing device 10, or it may be a general-purpose computer. Furthermore, the information processing device 10 may be implemented by a single computer 1000, or by a combination of multiple computers 1000.

[0065] Computer 1000 includes a bus 1020, a processor 1040, memory 1060, a storage device 1080, an input / output interface 1100, and a network interface 1120. Bus 1020 is a data transmission path for the processor 1040, memory 1060, storage device 1080, input / output interface 1100, and network interface 1120 to send and receive data to and from each other. However, the method of connecting the processor 1040 and the other components is not limited to bus connection. Examples of the processor 1040 include various processors such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and an FPGA (Field-Programmable Gate Array). Memory 1060 is a main memory device implemented using RAM (Random Access Memory), etc. Storage device 1080 is an auxiliary storage device implemented using a hard disk, SSD (Solid State Drive), memory card, or ROM (Read Only Memory), etc.

[0066] The input / output interface 1100 is an interface for connecting the computer 1000 with input / output devices. For example, input devices such as a keyboard and output devices such as a display are connected to the input / output interface 1100. The method by which the input / output interface 1100 connects to the input and output devices may be wireless or wired.

[0067] The network interface 1120 is an interface for connecting the computer 1000 to a network. Examples of such networks include LANs (Local Area Networks) and WANs (Wide Area Networks). The network interface 1120 may connect to the network via a wireless connection or a wired connection.

[0068] The storage device 1080 stores program modules that realize each functional component of the information processing device 10. The processor 1040 reads these program modules into the memory 1060 and executes them to realize the functions corresponding to each program module.

[0069] Furthermore, if the relational information storage unit 102 is located inside the information processing device 10, for example, the relational information storage unit 102 may be implemented using a storage device 1080.

[0070] According to this embodiment, the calculation unit 130 calculates a robustness index using the relevant information. Therefore, it becomes possible to design materials that take into account robustness to processing conditions.

[0071] (Second embodiment) Figure 8 is a diagram illustrating the functional configuration of the information processing device 10 according to the second embodiment. Figure 9 is a flowchart illustrating the processing flow performed by the information processing device 10 according to this embodiment. The information processing device 10 according to this embodiment is the same as the information processing device 10 according to the first embodiment, except for the points described below.

[0072] The information processing device 10 according to this embodiment includes an acquisition unit 140. The acquisition unit 140 acquires relational information. In this embodiment, the relational information is a table showing the relationship between processing conditions and material characteristic values. More specifically, the relational information is a table showing multiple processing conditions and characteristic values ​​for each of those multiple processing conditions. Here, the multiple processing conditions include multiple condition values ​​for the same item. The relational information may be experimental data. The relational information can be said to be multiple data points showing the relationship between processing conditions and characteristic values. In this embodiment, it is not necessary for a regression model like the one described in the first embodiment to exist in advance.

[0073] In the example shown in Figure 8, the information processing device 10 does not have a relational information storage unit 102. Below, an example in which relational information is input to the information processing device 10 and the acquisition unit 140 acquires the input relational information will be described in detail. However, even in this embodiment, the acquisition unit 140 may acquire experimental data as relational information from the relational information storage unit 102. Alternatively, the acquisition unit 140 may acquire relational information from another device.

[0074] In step S201, the candidate generation unit 110 generates candidate information in the same manner as in the first embodiment. In step S202, the output unit 190 outputs the generated candidate information. Then, in step S203, the acquisition unit 140 acquires the relational information input to the information processing device 10. The user of the information processing device 10 can prepare materials according to the candidate information output in step S202 and measure the characteristic values ​​of the prepared materials. The output unit 190 may display an input screen on a display connected to the information processing device 10 for inputting characteristic values ​​corresponding to the candidate information. The user can then input the characteristic values ​​obtained as measurement results into the information processing device 10. In step S203, it is preferable that the relational information acquired by the acquisition unit 140 includes data points for at least the reference value of the processing conditions. Furthermore, it is preferable that the relational information acquired by the acquisition unit 140 includes data points for two processing conditions that sandwich the reference value.

[0075] In step S204, the calculation unit 130 generates a regression model using relational information. Then, in step S205, the calculation unit 130 calculates a robust index using the regression model.

[0076] The calculation unit 130 can generate a regression model using relational information. Here, the explanatory variables of the regression model generated by the calculation unit 130 do not need to include items other than the target processing condition items, nor do they need to include the blending ratio. Furthermore, the regression model generated by the calculation unit 130 only needs to be a model that is effective at least near the reference point. The calculation unit 130 can then use the generated regression model to calculate a robust index in the manner described in the first embodiment.

[0077] The method by which the calculation unit 130 according to this embodiment calculates the robust index will be explained below with specific examples.

[0078] In this specific example, the calculation unit 130 uses a finite difference to determine the characteristic value y with respect to the modulus of elasticity. i And, the processing condition x regarding the heating temperature j The robust index R related to this ij Calculate the standard value x for heating temperature. j,0 Let ω be 450K and ω be 10K. That is, the lower limit P of the analysis interval. min This becomes 445K, and the upper limit P max This will be 455K.

[0079] Figure 10 is a table illustrating the relational information acquired by the candidate generation unit 110. This relational information includes data points corresponding to the reference value, the upper limit of the analysis interval, and the lower limit of the analysis interval. From the viewpoint of model accuracy, it is preferable that the relational information further includes other data points within the analysis interval.

[0080] Figure 11 illustrates a graph in which relational information is approximated by a quadratic function. The data points represented by the relational information are shown as black circles in Figure 11. The calculation unit 130 fits multiple data points included in the relational information with a predetermined type of function. In this way, an approximation formula can be obtained as a regression model. The type of function used for fitting may be predetermined for each combination of the target processing conditions and characteristic values. Examples of usable functions include quadratic functions, linear functions, and constant functions. However, the types of functions that can be used are not limited to these examples.

[0081] Next, the calculation unit 130 applies the reference value x to the obtained regression model. j,0 By substituting this, the reference value x j,0 Characteristic value y for i,0 The calculation unit 130 also uses a regression model to identify the maximum and minimum values ​​of the characteristic values ​​in the analysis interval. Then, using these values ​​and the above-mentioned equations (4) and (8), the calculation unit 130 calculates the robust index R. ij Calculate.

[0082] Step S206 is the same as step S104 in Figure 5.

[0083] If the information processing device 10 according to this embodiment further includes a prediction unit 150, the prediction unit 150 may predict characteristic values ​​using the regression model generated in step S204.

[0084] The hardware configuration of the computer implementing the information processing device 10 according to this embodiment is shown, for example, in Figure 7, similar to the first embodiment. However, the storage device 1080 of the computer 1000 implementing the information processing device 10 according to this embodiment further stores program modules that realize the functions of the acquisition unit 140.

[0085] According to this embodiment, the same operation and effect as in the first embodiment can be obtained. In addition, according to this embodiment, the relational information is a table showing the relationship between processing conditions and material characteristic values, and the calculation unit 130 generates a regression model using the relational information. Therefore, a robust index can be obtained even if the regression model is not prepared in advance.

[0086] (Third embodiment) Figure 12 is a diagram illustrating the functional configuration of the information processing device 10 according to the third embodiment. Figure 13 is a flowchart illustrating the processing flow performed by the information processing device 10 according to this embodiment. The information processing device 10 according to this embodiment is the same as the information processing device 10 according to at least one of the first and second embodiments, except for the points described below.

[0087] In the proposed method according to this embodiment, the information processing device 10 repeats the step S301 of generating candidate information and the step S303 of calculating a robustness index until a predetermined termination condition is met. In this way, the user of the information processing device 10 can compare and examine multiple candidate information while taking robustness into consideration.

[0088] In this embodiment, the information processing device 10 includes a termination determination unit 170. The following description will refer to Figures 12 and 13 to explain an example of when the information processing device 10 calculates the robust index using the method according to the first embodiment. However, the information processing device 10 in this embodiment may also calculate the robust index using the method according to the second embodiment.

[0089] Steps S301 to S303 are the same as steps S101 to S103 described above. Following step S303, the termination determination unit 170 determines whether or not the termination conditions have been met (step S304).

[0090] The termination condition may be, for example, that a predetermined stop operation has been performed on the information processing device 10. This allows the user to stop the process at any time. Alternatively, the termination condition may be that the number of repetitions has reached a predetermined number, or that the elapsed time since the start of the first step S301 has reached a predetermined time. The termination condition can be at least one of these examples. The predetermined number of repetitions and the predetermined time can be determined by the user inputting them to the information processing device 10.

[0091] If the termination condition is not met (No. in step S304), the process returns to step S301, and the candidate generation unit 110 generates candidate information again. Multiple candidate information is generated through this repeated process. The candidate generation unit 110 may generate multiple candidate information randomly, or it may generate multiple candidate information using, for example, a quasi-Monte Carlo method.

[0092] If the termination condition is met (Yes in step S304), the output unit 190 then outputs the output data (step S305). The output unit 190 outputs each of the multiple candidate information generated by the candidate generation unit 110, associating it with the robust index calculated by the calculation unit 130 and the characteristic value predicted by the prediction unit 150.

[0093] The hardware configuration of the computer implementing the information processing device 10 according to this embodiment is shown, for example, in Figure 7, similar to the first embodiment. However, the storage device 1080 of the computer 1000 implementing the information processing device 10 according to this embodiment further stores a program module that implements the functions of the termination determination unit 170.

[0094] According to this embodiment, the same functions and effects as in the first embodiment can be obtained. In addition, according to this embodiment, the information processing device 10 repeats the step S301 of generating candidate information and the step S303 of calculating a robustness index until a predetermined termination condition is met. Therefore, the user of the information processing device 10 can compare and examine multiple candidate information while taking robustness into consideration.

[0095] (Fourth embodiment) The functional configuration of the information processing device 10 according to the fourth embodiment is illustrated by Figure 12, similar to the third embodiment. The processing flow executed by the information processing device 10 according to this embodiment is illustrated by Figure 13, similar to the third embodiment. The information processing device 10 according to this embodiment is the same as the information processing device 10 according to the third embodiment, except for the points described below.

[0096] The candidate generation unit 110 according to this embodiment generates the next candidate information based on at least one of the calculation results of robust indicators for the candidate information and the prediction results of characteristic values. More specifically, in step S301, the candidate generation unit 110 can generate the next candidate information using an optimization algorithm. By doing so, candidate information that is closer to the desired conditions can be obtained. In optimization, the information processing device 10 can use at least one of the robust indicators and characteristic values ​​as the target variable (target) for optimization. The target of optimization may include multiple robust indicators or multiple characteristic values. The user can set a target value for each target of optimization and input it to the information processing device 10. In optimization, the candidate information can be said to be a candidate solution.

[0097] To obtain candidate information with high robustness while acquiring the desired characteristics, it is preferable to perform multi-objective optimization that targets both robustness indicators and characteristic values ​​for optimization.

[0098] Optimization algorithms that can be used include genetic algorithms, the Nelder-Mead method, or particle swarm optimization.

[0099] In step S301, which generates candidate information, the candidate generation unit 110 may generate the next candidate information based on a robust index calculated for one candidate information. That is, in the second and subsequent steps S301, the candidate generation unit 110 generates candidate information based on a robust index calculated using the candidate information generated in the previous step S301. In this way, new candidate information is generated to further improve the obtained robust index. By quantifying robustness, it can be treated as the target variable for optimization. Note that the candidate generation unit 110 may generate candidate information not only based on the robust index calculated using the previous candidate information, but also based on multiple robust indexes calculated up to that point.

[0100] If the proposed method executed by the information processing device 10 includes a step S302 in which the characteristic values ​​of a material are predicted using candidate information, the candidate generation unit 110 may generate the next candidate information in step S301 based on the prediction result in step S302 for one candidate information. That is, in the second and subsequent steps S301, the candidate generation unit 110 generates candidate information based on the characteristic values ​​predicted using the candidate information generated in the previous step S301. In this way, new candidate information is generated to obtain characteristic values ​​that are closer to the desired values. The candidate generation unit 110 may generate candidate information not only based on the characteristic values ​​predicted using the previous candidate information, but also based on multiple characteristic values ​​predicted up to that point. Furthermore, the candidate generation unit 110 may generate the next candidate information based on the prediction results of characteristic values ​​for multiple characteristics.

[0101] Steps S302 to S305 are as described in the third embodiment. However, examples of termination conditions in this embodiment further include conditions related to the amount of improvement and conditions related to the convergence of the search space. The condition related to the amount of improvement is, for example, that the amount of improvement of the optimization target variable has fallen below a predetermined threshold. The amount of improvement of the optimization target variable is the value obtained by |ΔP1-ΔP2|, where ΔP1 is the difference between the target variable and the target value for the candidate information newly generated in step S301, and ΔP2 is the difference between the target variable and the target value for the candidate information generated in the previous step S301. The condition related to the convergence of the search space is, for example, whether the most recent sample point falls within the convex hull of the search space so far.

[0102] In step S305, the output unit 190 outputs at least the candidate information last generated by the candidate generation unit 110, the robust index calculated using that candidate information, and the characteristic value predicted using that candidate information in the output data. The output unit 190 may also output multiple candidate information generated by the candidate generation unit 110, the robust index calculated using each candidate information, and the characteristic value predicted using each candidate information, not limited to the candidate information last generated by the candidate generation unit 110.

[0103] Furthermore, if multi-objective optimization is performed, the output unit 190 may also include graphs or other data to visualize the set of Pareto solutions in the output data.

[0104] The information processing device 10 according to this embodiment may perform optimization by imposing constraints on at least one of the formulation information, processing information, characteristic values, and robust indicators using existing methods.

[0105] According to this embodiment, the same functions and effects as in the first embodiment can be obtained. In addition, according to this embodiment, the candidate generation unit 110 generates the next candidate information based on at least one of the calculation result of a robust index for a given candidate information and the prediction result of a characteristic value. Therefore, candidate information that is closer to the desired conditions can be obtained.

[0106] (Fifth embodiment) Figure 14 is a diagram illustrating the functional configuration of the information processing device 10 according to the fifth embodiment. Figure 15 is a flowchart illustrating the processing flow performed by the information processing device 10 according to this embodiment. The information processing device 10 according to this embodiment is the same as the information processing device 10 according to the fourth embodiment, except for the points described below.

[0107] In the fourth embodiment, the optimization performed by the information processing device 10 can be described as static optimization that does not involve updating the regression model. In contrast, in the fifth embodiment, the information processing device 10 can be described as performing dynamic optimization that involves updating the regression model. In this embodiment, the information processing device 10 can, for example, generate candidate information using a Bayesian optimization method or update the regression model.

[0108] The information processing device 10 according to this embodiment includes an acquisition unit 140, an update unit 180, and a model storage unit 103. Steps S402 and S403 are the same as steps S202 and S203 according to the second embodiment, respectively. Step S405 is the same as at least one of steps S103 and S205. Step S407 is the same as step S305 according to the fourth embodiment.

[0109] The model storage unit 103 holds a regression model. The explanatory variables of this regression model include the blending ratio of each of several raw materials and one or more processing conditions. The dependent variable of this regression model includes the characteristic values ​​of the material. The accuracy of the model in its initial state is not considered. In step S404, the update unit 180 updates the model held in the model storage unit 103 based on the relational information acquired by the candidate generation unit 110.

[0110] In step S401, the candidate generation unit 110 generates an acquisition function based on the model stored in the model storage unit 103. Then, the candidate generation unit 110 generates new candidate information using the acquisition function.

[0111] Step S406 is the same as step S304. However, the example of the termination condition in this embodiment further includes a condition relating to the convergence of the acquisition function. The condition relating to the convergence of the acquisition function is, for example, that the output result for candidate information of the acquisition function becomes smaller than a predetermined criterion.

[0112] Furthermore, the information processing device 10 may perform a combination of the static optimization described in the fourth embodiment and the dynamic optimization described in the fifth embodiment. For example, the information processing device 10 may switch between static optimization and dynamic optimization in the first and second halves of a plurality of iterative processes. In addition, the information processing device 10 may use a multifidelity optimization method.

[0113] The information processing device 10 according to this embodiment may perform optimization by imposing constraints on at least one of the formulation information, processing information, characteristic values, and robust indicators using existing methods.

[0114] The hardware configuration of the computer implementing the information processing device 10 according to this embodiment is shown, for example, in Figure 7, similar to the first embodiment. However, the storage device 1080 of the computer 1000 implementing the information processing device 10 according to this embodiment further stores program modules that implement the functions of the acquisition unit 140 and the update unit 180.

[0115] Furthermore, if the model storage unit 103 is located inside the information processing device 10, for example, the model storage unit 103 may be implemented using a storage device 1080. However, the model storage unit 103 may be a storage device located outside the information processing device 10.

[0116] According to this embodiment, the same actions and effects as those of the first embodiment can be obtained. In addition, the same actions and effects as those of the fourth embodiment can be obtained.

[0117] The embodiments of the present invention have been described above with reference to the drawings, but these are merely examples of the present invention, and various other configurations can also be adopted.

[0118] Furthermore, while the flowcharts used in the above description show multiple steps (processes) in sequence, the execution order of the steps performed in each embodiment is not limited to the order in which they are described. In each embodiment, the order of the illustrated steps can be changed to the extent that it does not impede the content. Also, the above embodiments can be combined to the extent that their contents do not conflict. [Explanation of Symbols]

[0119] 10 Information Processing Devices 102 Related Information Storage Unit 103 Model Memory Unit 110 Candidate generation section 130 Calculation Unit 140 Acquisition Department 150 Prediction Section 170 Termination Determination Unit 180 Update Department 190 Output section 1000 calculator 1020 Bus 1040 processor 1060 memory 1080 Storage Devices 1100 Input / Output Interface 1120 Network Interface

Claims

1. A proposed method that is executed by one or more computers, A candidate generation step that generates candidate information including formulation information showing the blending ratio of each of several raw materials, The calculation step includes using relational information showing the relationship between processing conditions, the mixing ratio, and the characteristic values ​​of the material produced according to the processing conditions, to calculate a robust index indicating the degree of sensitivity of the characteristic values ​​to the processing conditions. Proposal method.

2. In the proposed method described in claim 1, The candidate generation step and the calculation step are repeated until a predetermined termination condition is met. Proposal method.

3. In the proposed method described in claim 2, In the candidate generation step, the next candidate information is generated based on the robust index calculated for one candidate piece of information. Proposal method.

4. In the proposed method described in claim 3, In the candidate generation step, the following candidate information is generated using an optimization algorithm. Proposal method.

5. In the proposed method described in claim 3 or 4, The method further includes a prediction step of predicting the characteristic values ​​of the material using the candidate information, In the candidate generation step, the next candidate information is generated based on the prediction result in the prediction step for the one candidate information. Proposal method.

6. In the proposed method according to any one of claims 1 to 4, The candidate information further includes processing information indicating the processing conditions. Proposal method.

7. In the proposed method according to any one of claims 1 to 4, The aforementioned relational information is a mathematical formula or a trained model. Proposal method.

8. In the proposed method according to any one of claims 1 to 4, The aforementioned related information is experimental data. Proposal method.

9. A candidate generation unit generates candidate information that includes formulation information showing the blending ratio of each of several raw materials, The system includes a calculation unit that uses relational information showing the relationship between processing conditions and the characteristic values ​​of the material produced according to the mixing ratio and processing conditions to calculate a robust index indicating the degree of sensitivity of the characteristic values ​​to the processing conditions. Information processing device.

10. Computers, Candidate generation means for generating candidate information including formulation information showing the blending ratio of each of multiple raw materials, and Using relational information showing the relationship between processing conditions, the mixing ratio, and the characteristic values ​​of the material manufactured according to the processing conditions, this system functions as a calculation means for calculating a robust index indicating the degree of sensitivity of the characteristic values ​​to the processing conditions. program.

Images