Retrieval system, retrieval method, and retrieval program

Abstract

A retrieval system disclosed herein acquires a search condition related to a candidate downstream substance characteristic that is a characteristic of a candidate downstream substance that is a substance obtained in a downstream process of manufacturing, extracts one or more data records corresponding to the search condition from a database storing a plurality of data records indicating a combination of the candidate downstream substance characteristic and a related physical property that is a physical property related to the candidate downstream substance characteristic, sets a related physical property indicated by one data record selected from the extracted one or more data records as a retrieval target, acquires, for each of two or more candidate upstream substances that are substances used in an upstream process of manufacturing, a candidate upstream substance physical property that is a physical property of the candidate upstream substance, and executes an optimization method using the retrieval target and the candidate upstream substance physical property of each of the two or more candidate upstream substances to determine a blending ratio between the two or more candidate upstream substances.

Description

Search System, Search Method, and Search Program

[0001] One aspect of the present disclosure relates to a search system, a search method, and a search program.

[0002] Materials informatics (MI), an effort to improve the efficiency of material development using information science, has been introduced in various material development scenarios. For example, the following techniques related to or potentially related to MI are known.

[0003] Patent Document 1 describes an analysis device for combinatorial optimization problems. The computer used in this analysis device starts from an initial combinatorial state, determines the state to transition to using an evaluation function from among combinatorial states defined as adjacent states, and sequentially performs a search that repeats the transition to the determined transition destination. The computer attempts to discover an optimal combinatorial state that minimizes or maximizes the function value of an evaluation function composed of the sum of a function to be minimized or maximized and a penalty function representing the amount of constraint violation through this search.

[0004] Patent Document 2 describes a combinatorial optimization device having a storage unit that stores the values of a plurality of state variables included in a first energy function to which terms representing constraint conditions for the plurality of state variables are given, and a processing unit that performs a search for the values of the plurality of state variables that minimize the value of the first energy function. The search by the processing unit includes a first search performed using the first energy function, a second search performed using a second energy function obtained by removing the terms representing the constraint conditions from the first energy function after the first search, and a third search performed using the first energy function after the second search.

[0005] Patent Document 3 describes a parameter generation device including an input means that receives an input of a first objective function and constraint conditions defining a combination of elements related to the manufacture of a material, an objective function generation means that generates a second objective function in which probabilistic fluctuations are set for the parameters of the first objective function, an optimization processing means that optimizes a model including the second objective function and the constraint conditions, and an output means that outputs the values of the variables of the second objective function obtained by the optimization as a parameter set.

[0006] Patent Document 4 describes a composition proposal device that proposes the type and mass ratio of a certain polymer and other components to be included as constituent components in a polymer composite material. This device has a setting unit that sets target values ​​for the physical properties of a virtual polymer composite material, and an optimization unit that modifies the parameters including the descriptor of the virtual polymer composite material so that the physical properties of the virtual polymer composite material, predicted by inputting parameters including the descriptor of the virtual polymer composite material into a trained model, approach the target values.

[0007] Patent Document 5 describes a learning model generation program that causes a computer to execute a learning model generation step, in which machine learning is performed using learning data in which the blending ratio of raw materials constituting a known thermoplastic aromatic polyester resin composition is used as an explanatory variable and the physical property values ​​of the thermoplastic aromatic polyester resin composition are used as the objective variable.

[0008] Japanese Patent Publication No. 2003-223322, Japanese Patent Publication No. 2021-33657, International Publication No. 2023 / 209983, Japanese Patent No. 7407893, Japanese Patent Publication No. 2023-177308

[0009] There is a need for a mechanism to more quickly search for the optimal formulation of materials used in the upstream processes of a manufacturing process to achieve desired properties in materials obtained in the downstream processes of that manufacturing process.

[0010] A search system relating to one aspect of this disclosure comprises at least one processor. The at least one processor acquires search conditions related to candidate downstream material properties, which are characteristics of candidate downstream materials, which are materials obtained in the downstream processes of manufacturing. From a database that stores multiple data records showing combinations of candidate downstream material properties and related physical properties, which are physical properties related to the candidate downstream material properties, it extracts one or more data records corresponding to the search conditions. It sets the related physical properties indicated by one data record selected from the one or more extracted data records as a search target. For each of the two or more candidate upstream materials, which are materials used in the upstream processes of manufacturing, it acquires candidate upstream material properties, which are physical properties of the candidate upstream material. It executes an optimization method using the search target and the respective candidate upstream material properties of the two or more candidate upstream materials to determine the blending ratio between the two or more candidate upstream materials.

[0011] In this aspect, first, one or more data records showing combinations of candidate downstream material properties and related physical properties are extracted based on search conditions related to the properties of candidate downstream materials obtained in the downstream manufacturing process (candidate downstream material properties). Then, an optimization method is executed that uses the related physical properties indicated by one selected data record as the search target to determine the blending ratio between two or more candidate upstream materials used in the upstream manufacturing process. A two-stage process is executed sequentially: searching for related physical properties corresponding to downstream materials with desired properties, and searching for the blending ratio between two or more upstream materials to obtain those related physical properties. This mechanism allows for the search of a blend of materials used in the upstream manufacturing process that realizes the desired properties for materials obtained in the downstream manufacturing process in a shorter amount of time.

[0012] According to one aspect of this disclosure, it is possible to search in a shorter time for formulations of materials used in upstream processes of manufacturing that achieve desired properties for materials obtained in downstream processes of manufacturing.

[0013] This is a diagram showing an example of the functional configuration of a search system. This is a flowchart showing an example of the process for generating contained component / composition data. This is a flowchart showing an example of the process for determining the blending ratio. This is a flowchart showing an example of a search using the simulated annealing method. This is a diagram showing an example of the transition rule in the simulated annealing method.

[0014] The following describes various examples in this disclosure in detail with reference to the attached drawings. In the description of the drawings, identical or equivalent elements are denoted by the same reference numeral, and redundant descriptions are omitted.

[0015] [System Overview] The search system described herein is a computer system that searches for a blend of raw materials that realizes the constituent components of a composition having desired properties. To perform the search, the search system accesses constituent component / composition data that shows combinations of the properties of candidate compositions and the physical properties of candidate constituent components, and raw material data that shows the physical properties of candidate raw materials for producing the candidate constituent components. The search system extracts constituent component / composition data that corresponds to the search conditions for the properties of the candidate composition. Then, based on the physical properties of the candidate constituent component selected from the extraction results and the physical properties of two or more candidate raw materials selected for the candidate constituent component, the search system determines the blending ratio between the two or more candidate raw materials to realize the physical properties of the candidate constituent component. In this way, the search system continuously performs a two-stage process: searching for constituent components corresponding to a composition having desired properties, and searching for the blending ratio between two or more raw materials that constitute those constituent components. By using the search system, users can obtain a blend of raw materials that realizes the constituent components of a composition having desired properties in a shorter amount of time.

[0016] Candidate compositions refer to compositions that may be searched by users. Candidate composition properties refer to the physical properties unique to the candidate composition, and are also referred to in this disclosure as "candidate composition properties." Candidate components refer to the substances that make up the candidate composition. Candidate component properties refer to the physical properties of the candidate component, and are also referred to in this disclosure as "candidate component properties." Candidate raw materials refer to raw materials that may be selected to produce the candidate component. Candidate raw material properties refer to the physical properties of the candidate raw material, and are also referred to in this disclosure as "candidate raw material properties." The properties of the candidate composition may be expressed by one or more parameters. The properties of the candidate component and the properties of the candidate raw materials may be expressed by one or more common parameters.

[0017] In one example, each candidate composition is a photosensitive resin composition, each candidate component is a polymer, and each candidate raw material is a monomer.

[0018] When the candidate composition is a photosensitive resin composition, the characteristics of the candidate composition may be expressed by at least one of the following parameters: the minimum time required for the photosensitive resin composition to be developed, sensitivity to light, light transmittance, resolution indicating the density of developable images, minimum resist line width, adhesion to the substrate, foaming or cohesiveness when developing the photosensitive resin composition with a developer, edge fuse characteristics (the amount of photosensitive resin composition that spills out from the end face of the dry film roll due to winding pressure during storage of the dry film roll), flexibility of the developed photocurable photosensitive resin composition, adhesion between the dry film and the base film or cover film, color stability of the photosensitive resin composition, peelability when peeling the photosensitive resin composition from the substrate, size of the photosensitive resin composition (peeled piece) peeled from the substrate, and tear rate when tenting the tent holes of the substrate with dry film.

[0019] The polymer as a candidate component may be an alkali-soluble polymer, an ethylenically unsaturated bond-containing compound, a photopolymerization initiator, a resin having repeating units containing an acid-degradable group (for example, a group deprotected by an acid), a phenolic resin, a photoacid generator, a dissolution inhibitor, a sensitizer, a polymerization inhibitor, an adhesive, or a plasticizer. When the candidate component is a polymer, the properties of the candidate component may be expressed by at least one of the following parameters: solubility parameter, number of a given functional group, total number of atoms per repeating unit, number of atoms of each element per repeating unit, average molecular weight, acid dissociation constant (pKa), polar surface area, partition coefficient (LogP), glass transition temperature (Tg), melting point, boiling point, refractive index, density, viscosity, and specific gravity.

[0020] When the candidate raw material is a monomer, the properties of the candidate raw material may be expressed by at least one of the following parameters: solubility parameter, number of a given functional group, total number of atoms per repeating unit, number of atoms of each element per repeating unit, average molecular weight, pKa, polar surface area, LogP, glass transition temperature (Tg), melting point, boiling point, refractive index, density, viscosity, and specific gravity.

[0021] The mixing ratio between two or more candidate raw materials is expressed by the mixing ratio of each candidate raw material. For example, the mixing ratio of six candidate raw materials Ma, Mb, Mc, Md, Me, and Mf is expressed as Ra:Rb:Rc:Rd:Re:Rf, where Ra is the mixing ratio of candidate raw material Ma, Rb is the mixing ratio of candidate raw material Mb, Rc is the mixing ratio of candidate raw material Mc, Rd is the mixing ratio of candidate raw material Md, Re is the mixing ratio of candidate raw material Me, and Rf is the mixing ratio of candidate raw material Mf. The mixing ratio and mixing ratio may be expressed by copolymerization ratio (%) or by molar concentration (mol / L or mol%).

[0022] As an alternative to the method described herein, it is conceivable to determine the blending ratio of candidate raw materials in a single step using a predictive model that does not involve separate optimization processing, based on the characteristics of the target composition. However, this alternative method requires considering not only the combinations of physical properties of candidate components but also the combinations of blending ratios of candidate raw materials. Therefore, the number of numerical combinations input to the predictive model may become enormous, potentially increasing computational costs. Typically, the number of combinations of physical properties of candidate components is around 10 to the power of 9, and the number of numerical combinations input to the predictive model is around 10 to the power of 20, thus increasing computational costs. In contrast, the search system described herein makes it possible to obtain the blending ratio of raw materials in a short time without having to input a potentially enormous number of combinations into the predictive model. Furthermore, the search system allows the user to select the physical properties of the components to achieve the desired characteristics of the composition, thereby increasing the transparency of the search system and the reliability of the search results.

[0023] A candidate composition is an example of a candidate downstream substance, which is a substance obtained in the downstream process of manufacturing, and the candidate composition properties are an example of candidate downstream substance properties, which are properties of the candidate downstream substance. A candidate component is an example of a candidate midstream substance, which is a substance obtained in the midstream process of manufacturing. The candidate component properties are an example of related properties, which are properties related to the candidate downstream substance properties, and are also an example of candidate midstream substance properties, which are properties of the candidate midstream substance. A candidate raw material is an example of a candidate upstream substance, which is a substance used in the upstream process of manufacturing, and the candidate raw material properties are an example of candidate upstream substance properties, which are properties of the candidate upstream substance. Therefore, the search system according to this disclosure can also be said to be a computer system that searches for formulations of substances used in the upstream process of manufacturing that realize desired properties for a substance obtained in the downstream process of manufacturing in a shorter amount of time. In order to perform the search, the search system accesses data (e.g., component / composition data) that shows combinations of candidate downstream substance properties and related properties. The search system extracts one or more related properties that correspond to search conditions for candidate downstream substance properties. The search system then determines the mixing ratio between the two or more candidate upstream materials to achieve the desired properties of the candidate downstream material, based on the relevant physical properties selected from the extraction results and the physical properties of the two or more candidate upstream materials. In this way, the search system continuously performs two stages of processing: searching for relevant physical properties corresponding to a downstream material having the desired properties, and searching for the mixing ratio between the two or more upstream materials to achieve those relevant physical properties. By using the search system, users can obtain the formulation of materials used in the upstream manufacturing process that achieves the desired properties of a material obtained in the downstream manufacturing process in a shorter amount of time.

[0024] [System Configuration] The search system consists of one or more computers. When multiple computers are used, these computers are connected via a communication network such as the Internet or an intranet to logically construct a single search system.

[0025] A computer constituting a search system generally comprises a processor, storage device (memory), and communication interface as hardware components. The processor is, for example, a CPU or GPU. Storage devices consist of flash memory, hard disks, etc. Communication interfaces consist of network cards, wireless communication modules, etc. Each functional module of the search system is realized when the processor executes a program stored in the storage device.

[0026] A search program for enabling a computer to function as a search system includes program code for implementing each functional module of the search system. This search program may be provided on a non-temporary recording medium such as a CD-ROM, DVD-ROM, or semiconductor memory. Alternatively, the search program may be provided via a communication network as a data signal superimposed on a carrier wave. The provided search program is then recorded, for example, on a storage device.

[0027] Figure 1 shows the functional configuration of a search system 10 in one example. In this example, the search system 10 connects to a database 30 and a user terminal 40 via a communication network. The communication network is typically constructed by the internet, an intranet, or a combination thereof. The communication network can be constructed by a wired network, a wireless network, or a combination thereof.

[0028] The database 30 is a storage device that stores various types of data used by the search system 10. The database 30 may be a component of the search system 10 or it may be located outside the search system 10. In one example, the database 30 stores contained component / composition data and raw material data.

[0029] The component / composition data represents multiple combinations of candidate component properties and candidate composition characteristics. For example, each data record in the component / composition data includes an identifier for a candidate component, the candidate component's properties, an identifier for a candidate composition, and the candidate composition's characteristics. For both the candidate component and the candidate composition, the identifier may be expressed in various forms, such as a number, name, chemical formula, or structural formula.

[0030] Raw material data is data that shows the physical properties of each of several candidate raw materials. In one example, each data record in the raw material data includes an identifier for the candidate raw material and the physical properties of the candidate raw material. The identifier for the candidate raw material may be expressed in various forms such as a number, name, chemical formula, or structural formula.

[0031] The user terminal 40 is a computer used by the user of the search system 10. The user terminal 40 can be any type of computer, such as a personal computer, workstation, tablet, smartphone, or wearable device. The user terminal 40 is equipped with a display device such as a liquid crystal display.

[0032] The search system 10 includes a processor 101 that functions as a data generation unit 11, a search unit 12, and a search unit 13. The data generation unit 11 is a functional module that generates data records of contained components / composition data using a prediction model 20. The prediction model 20 is a trained model generated by machine learning to calculate the properties of a composition composed of contained components from the physical properties of the contained components. The physical properties of the contained components can be represented by a combination of one or more physical property values ​​(parameter values) of the contained components. The prediction model 20 is implemented, for example, by a neural network. The prediction model 20 is an example of a prediction model generated by machine learning to calculate the properties from physical properties related to the properties of downstream materials (i.e., related physical properties or candidate midstream material physical properties). The search unit 12 is a functional module that extracts contained component / composition data corresponding to search conditions related to candidate composition properties. The search unit 13 is a functional module that determines the blending ratio between two or more candidate raw materials constituting a selected candidate contained component from the extracted contained component / composition data using an optimization method.

[0033] [System Operation] The operation of the search system 10 will be described below, along with the search method related to this disclosure.

[0034] (Generation of Ingredient / Composition Data) The process for generating ingredient / composition data will be explained with reference to Figure 2. Figure 2 is a flowchart showing an example of this process as process flow S1.

[0035] In step S11, the data generation unit 11 inputs the physical properties of one candidate component into the prediction model 20 to calculate the candidate composition characteristics. As described above, in one example, the data generation unit 11 inputs a combination of one or more physical property values ​​(parameter values) of one candidate component as the candidate component physical properties into the prediction model 20 to calculate the candidate composition characteristics. For example, for the first parameter of the candidate component physical properties, the data generation unit 11 randomly selects a parameter value within a specified range, or selects multiple parameter values ​​at predetermined intervals. The data generation unit 11 similarly selects a parameter value for the second parameter of the candidate component physical properties. The data generation unit 11 inputs the combination of parameter values ​​for the first and second parameters into the prediction model 20 to calculate the candidate composition characteristics. The number of such combinations of physical property values ​​is, for example, about 10 to the power of 9. The data generation unit 11 may read the candidate component physical properties stored in a predetermined storage device (memory), receive the candidate component physical properties from another computer such as a user terminal 40, or accept the candidate component physical properties input by the administrator of the search system 10. The prediction model 20 calculates candidate composition characteristics from the physical properties of candidate components, and the data generation unit 11 acquires these candidate composition characteristics.

[0036] In step S12, the data generation unit 11 generates a data record showing the combination of the candidate component properties input to the prediction model 20 and the calculated candidate composition characteristics. If the data record includes identifiers for the candidate component and the candidate composition, the data generation unit 11 may automatically generate these identifiers according to predetermined rules, or it may set them according to input by the user or administrator.

[0037] As shown in step S13, the data generation unit 11 may repeatedly generate data records. If the generation of data records is to be continued (NO in step S13), the process returns to step S11. In the repeated step S11, the data generation unit 11 inputs the physical properties of another candidate component into the prediction model 20 to calculate the candidate composition characteristics. In the repeated step S12, the data generation unit 11 generates another data record showing the combination of the candidate component properties and the candidate composition characteristics. On the other hand, if the generation of data records is to be terminated (YES in step S13), the process proceeds to step S14. In step S14, the data generation unit 11 stores the set of one or more generated data records as component / composition data in the database 30.

[0038] As shown in processing flow S1, the data generation unit 11 inputs the physical properties of each of the one or more candidate constituent components into the prediction model 20 to calculate the candidate composition characteristics, and stores a single data record showing the combination of the candidate constituent component properties and the candidate composition characteristics in the database 30 as one of multiple data records of constituent component / composition data. Processing flow S1 is an example of a process in which, for each of the one or more related physical properties, the related physical properties are input into the prediction model to calculate the candidate downstream material characteristics, and a single data record showing the combination of the related physical properties and the candidate downstream material characteristics is stored in the database as one of multiple data records. The search system 10 may execute processing flow S1 when constituent component / composition data has not been stored in the database 30. Alternatively, the search system 10 may execute processing flow S1 when one or more data records of constituent component / composition data have already been stored in the database 30 to increase the amount of constituent component / composition data.

[0039] (Determination of Mixing Ratio) The process for determining the mixing ratio will be explained with reference to Figure 3. Figure 3 is a flowchart showing an example of this process as process flow S2.

[0040] In step S21, the search unit 12 receives search conditions related to candidate composition characteristics from the user terminal 40. This reception process is an example of a process for acquiring search conditions. In one example, the search unit 12 provides the user terminal 40 with a user interface for executing a two-stage process of searching for contained components and searching for the mixing ratio between two or more raw materials constituting the contained components. The user terminal 40 displays the user interface on a display device and accepts the search conditions input by the user via the user interface. The search conditions are expressed, for example, by the numerical range of each of one or more parameters representing candidate composition characteristics. The user terminal 40 transmits the input search conditions to the search system 10, and the search unit 12 receives the search conditions.

[0041] In step S22, the search unit 12 extracts contained component / composition data corresponding to the search conditions from the database 30. The search unit 12 accesses the database 30 and extracts one or more data records of the contained component / composition data corresponding to the search conditions, that is, one or more data records that match the search conditions, from the database 30.

[0042] In step S23, the search unit 12 transmits the extracted contained component / composition data to the user terminal 40. The user terminal 40 receives and displays the data. That is, the search unit 12 displays one or more data records of the extracted contained component / composition data on the display device of the user terminal 40.

[0043] In step S24, the search unit 13 sets the physical properties of the candidate contained component selected by the user as the search target. On the user terminal 40, the user interface receives a user operation for selecting one data record from one or more displayed data records, and transmits the physical properties of the candidate contained component indicated by the selected one data record to the search system 10. The search unit 13 sets the physical properties of the candidate contained component as the search target. That is, the search unit 13, in response to a user operation for selecting one data record from one or more data records displayed on the display device of the user terminal 40, sets the physical properties of the candidate contained component indicated by the one data record as the search target. For each of one or more parameters expressing the physical properties of the candidate contained component, it can be said that the parameter value is the target value for the parameter.

[0044] In step S25, the search unit 13 transmits a plurality of candidate raw materials to the user terminal 40. For example, the search unit 13 reads out the raw material data pre-stored in the database 30 and transmits the identifier of each of the plurality of candidate raw materials to the user terminal 40. The user terminal 40 receives and displays the data. That is, the search unit 13 displays a plurality of candidate raw materials on the display device of the user terminal 40.

[0045] In step S26, the search unit 13 acquires the physical properties of each of two or more candidate raw materials selected by the user. On the user terminal 40, the user interface receives a user operation for selecting two or more candidate raw materials from the plurality of displayed candidate raw materials, and transmits the identifiers of the selected two or more candidate raw materials to the search system 10. The selected two or more candidate raw materials are information selected for the candidate contained component having the physical properties of the candidate contained component set as the search target in step S24. For each of the identifiers of the two or more selected candidate raw materials, the search unit 13 reads out the physical properties of the candidate raw material corresponding to the identifier from the raw material data in the database 30. That is, the search unit 13, in response to a user operation for selecting two or more candidate raw materials from the plurality of candidate raw materials displayed on the display device of the user terminal 40, acquires the physical properties of each of the two or more candidate raw materials.

[0046] In step S27, the search unit 13 executes an optimization method using the search target and the physical properties of each of the two or more selected candidate raw materials to determine the blending ratio between the two or more candidate raw materials. In one example, the search unit 13 uses Simulated Annealing (SA) as the optimization method. The search using Simulated Annealing will be explained with reference to Figure 4. Figure 4 is a flowchart showing an example of such a search.

[0047] In step S271, the search unit 13 sets an initial value for the temperature. Temperature is a parameter that controls the rate of change in the state (explanatory variables in the search space). The higher the temperature, the higher the probability that the state will change in a way that worsens the solution. As the search progresses, the temperature gradually decreases and the rate of change in the state gradually decreases.

[0048] In step S272, the search unit 13 sets an initial value for the blending ratio between two or more candidate raw materials, calculates an initial provisional solution for the physical properties of the candidate components based on that blending ratio, and calculates an initial evaluation value that shows the difference between that provisional solution and the search target. An example of this series of processes will be explained below.

[0049] First, the search unit 13 randomly sets the initial value of the blending ratio. The blending ratio may be set to 0 for at least one candidate raw material.

[0050] Next, the search unit 13 calculates the parameter value of the candidate component for each of the one or more parameters based on the parameter values ​​and blending ratios of two or more candidate raw materials. Then, the search unit 13 expresses an initial provisional solution for the physical properties of the candidate component using the one or more parameter values ​​of the candidate component. The search unit 13 calculates the parameter value of the candidate component for each of the one or more parameters using one of the following two calculation methods.

[0051] In the first calculation method, the search unit 13 calculates the product of the parameter value and the blending ratio for each of the two or more candidate raw materials. Then, the search unit 13 calculates the sum of the two or more products corresponding to the two or more candidate raw materials as the parameter value of the candidate component. Let N be the number of types of candidate raw materials, and let p be the parameter value and blending ratio of the i-th candidate raw material.ｉ ,w ｉ Therefore, the parameter value v of the candidate component ｙ This is obtained by equation (1).

[0052] In the second calculation method, the search unit 13 calculates a quotient for each of the two or more candidate raw materials by dividing the blending ratio by the parameter value. Then, the search unit 13 calculates the parameter value of the candidate component as the reciprocal of the sum of the two or more quotients corresponding to the two or more candidate raw materials. Let N be the number of types of candidate raw materials, and let p be the parameter value and blending ratio of the i-th candidate raw material. ｉ ,w ｉ Therefore, the parameter value v of the candidate component ｙ This is obtained by equation (2).

[0053] In one example, the search unit 13 calculates parameter values ​​for parameters other than the glass transition temperature using a first calculation method, and calculates the parameter value for the glass transition temperature using a second calculation method.

[0054] Next, the search unit 13 calculates an initial evaluation value that shows the difference between the initial provisional solution (current provisional solution) and the search target. In one example, the search unit 13 calculates the difference between the calculated parameter value and the target value of the search target for each of the one or more parameters, and calculates the error rate, which is the ratio of this difference to the target value. Then, the search unit 13 calculates the sum of the error rates of one or more corresponding to one or more parameters as the evaluation value. The smaller this evaluation value, the closer the provisional solution is to the search target. Let M be the number of types of parameters, and let v be the parameter value and target value for the j-th parameter. ｊ , g ｊ Therefore, the evaluation value E, which represents the difference between the provisional solution and the search target, can be obtained by equation (3).

[0055] In step S273, the search unit 13 changes the blending ratio between two or more candidate raw materials, calculates a new provisional solution for the physical properties of the candidate components based on that blending ratio, and calculates a new evaluation value that shows the difference between that provisional solution and the search target.

[0056] First, the search unit 13 changes the blending ratio based on a predetermined transition rule in the annealing process. In one example, the search unit 13 selects two candidate raw materials from two or more candidate raw materials and changes the blending ratio between the two candidate raw materials so that the sum of the blending ratios between the two candidate raw materials does not change. Figure 5 shows an example of such a transition rule. The search unit 13 may exchange the blending ratios between the two candidate raw materials (Example 201), or it may transfer a portion of the blending ratio of one candidate raw material to the blending ratio of the other candidate raw material (Example 202). Alternatively, the search unit 13 may average the blending ratios between the two candidate raw materials, in which case the blending ratios of both become the same (Example 203). Alternatively, the search unit 13 may exchange the blending ratio between one candidate raw material with the minimum blending ratio and one candidate raw material with a blending ratio of 0, and this process can be said to be an exchange of raw materials (Example 204).

[0057] Next, the search unit 13 calculates the parameter value of the candidate component for each of the one or more parameters based on the parameter values ​​and blending ratios of two or more candidate raw materials. Subsequently, the search unit 13 expresses a new provisional solution for the physical properties of the candidate component using the one or more parameter values ​​of the candidate component. Then, the search unit 13 calculates a new evaluation value that shows the difference between the new provisional solution and the search target. The search unit 13 performs these series of processes using the same method as in step S272.

[0058] In step S274, the search unit 13 decides whether or not to adopt a new provisional solution, using a probability based on temperature as needed. The search unit 13 adopts the new provisional solution if a better evaluation value is obtained. When the evaluation value E obtained by equation (3) above is used, a better evaluation value is an evaluation value lower than all the evaluation values ​​calculated so far in step S27. If a better evaluation value is not obtained, the search unit 13 probabilistically decides whether or not to adopt a new provisional solution based on the probability p = exp(-ΔE / T). ΔE represents the difference between the current provisional solution and the new provisional solution, and T represents the temperature. If the new provisional solution is adopted, the search unit 13 updates the current provisional solution with the new provisional solution. If the new provisional solution is not adopted, the search unit 13 discards the new provisional solution and maintains the current provisional solution.

[0059] As shown in step S275, the search unit 13 repeats steps S273 and S274 until it reaches an equilibrium state at the current temperature. An equilibrium state refers to a situation where the current provisional solution does not change even after the process is repeated.

[0060] If equilibrium is reached (YES in step S275), the process proceeds to step S276. In step S276, the search unit 13 updates the temperature. This update is a process that lowers the temperature, and is achieved, for example, by exponential annealing.

[0061] As shown in step S277, the search unit 13 repeats the processing in steps S273 to S276 until a predetermined termination condition is met. The termination condition may be that a better evaluation value can no longer be obtained, that is, that the evaluation value is no longer updated. Alternatively, the termination condition may be that a predetermined processing time has elapsed.

[0062] If the termination condition is met (YES in step S277), the process proceeds to step S278. In step S278, the search unit 13 determines the blending ratio. The search unit 13 determines the blending ratio corresponding to the provisional solution that was finally adopted as the solution of the optimization method (simulated annealing). That is, the search unit 13 determines the blending ratio that minimizes the evaluation value, which shows the difference between the provisional solution of the candidate component properties calculated based on the physical properties of each of the two or more candidate raw materials and the search target. This process is an example of determining the blending ratio based on that evaluation value.

[0063] As explained with reference to Figure 4, in the annealing process, the search unit 13 repeatedly searches for a blending ratio by changing the blending ratio between two candidate raw materials selected from two or more candidate raw materials so that the sum of the blending ratios between the two candidate raw materials does not change, and determines the blending ratio.

[0064] Returning to Figure 3, in step S28, the search unit 13 transmits the determined blending ratio to the user terminal 40. This transmission is an example of a process that outputs the blending ratio. The user terminal 40 receives the blending ratio. The user terminal 40 may display the blending ratio on a display device or store the blending ratio in a predetermined storage device (memory).

[0065] [Variations] The technology of this disclosure has been described in detail above based on various examples. However, the technology of this disclosure is not limited to the examples above. Various modifications are possible without departing from the gist of this disclosure.

[0066] In the example above, the optimization method is simulated annealing, but the search system (search unit) may perform other types of optimization methods to determine the blending ratio between two or more candidate raw materials. For example, the search system (search unit) may use random search as the optimization method.

[0067] In the example above, the data generation unit 11 generates component / composition data (one or more data records showing combinations of candidate component properties and candidate composition characteristics) using the prediction model 20. However, the component / composition data may be generated by a computer system separate from the search system, or it may be generated manually. Therefore, the search system does not need to have functions equivalent to the data generation unit 11 and the prediction model 20, and does not need to execute the processing flow S1.

[0068] In the above example, the search unit 13 displays a plurality of candidate raw materials on the display device of the user terminal 40, and in response to a user operation to select two or more candidate raw materials from the plurality of candidate raw materials, it acquires the physical properties of each of the two or more candidate raw materials. As a variation of this, the search system (search unit) may acquire the physical properties of each of two or more predetermined candidate raw materials that are prepared independently of user operation. Alternatively, the search system (search unit) may acquire the physical properties of each of the two or more candidate raw materials in response to a user operation to select one or more candidate raw materials from a plurality of candidate raw materials. In this variation, if only one candidate raw material is selected by user operation, the search system (search unit) automatically selects one or more other candidate raw materials and acquires the physical properties of the single candidate raw material selected by user operation and the one or more other candidate raw materials that were automatically selected.

[0069] In the example above, the search system 10 acts as a server in a client-server system. In another example, the functions of the search system 10 and the database 30 may be implemented on a standalone computer. Alternatively, the search system may be implemented on a user terminal that can access the database 30 via a communication network.

[0070] In the above example, the related properties are the properties of the candidate constituent components, but the related properties may also be the properties of the candidate composition. In another example, the candidate downstream substance may be the candidate constituent component, the properties of the candidate downstream substance may be the properties of the candidate constituent component, and the related properties may be the properties of the candidate constituent component. Each substance considered in the exploration system may be appropriately selected depending on the manufacturing process. For example, the exploration system may determine the blending ratio between candidate upstream substances based on the substances in the upstream and downstream processes, or based on the substances in the upstream, midstream, and downstream processes, respectively.

[0071] The processing steps for a method executed by at least one processor are not limited to the examples above. For example, some of the steps described above may be omitted, or each step may be performed in a different order. Also, any two or more of the steps described above may be combined, or some of the steps may be modified or deleted. Alternatively, other steps may be performed in addition to each of the steps described above.

[0072] In comparing the relative magnitudes of two numerical values ​​in this disclosure, either of the two criteria, "greater than or equal to" and "greater than," may be used, or either of the two criteria, "less than or equal to" and "less than," may be used.

[0073] In this disclosure, the expression "at least one processor executes a first process, a second process, ... and the nth process," or a corresponding expression, refers to a concept that includes cases where the entity executing the n processes from the first process to the nth process, i.e., the processor, changes along the way. In other words, this expression refers to a concept that includes both cases where all n processes are executed by the same processor and cases where the processor changes at an arbitrary rate for the n processes.

[0074] [Note] As can be seen from the various examples above, this disclosure includes the following embodiments: (Note 1) A search system comprising at least one processor, the at least one processor acquires search conditions related to candidate downstream material properties, which are the properties of a candidate downstream material, which is a substance obtained in a downstream process of manufacturing; extracts one or more data records corresponding to the search conditions from a database that stores a plurality of data records showing combinations of the candidate downstream material properties and related physical properties, which are physical properties related to the candidate downstream material properties; sets the related physical properties indicated by one data record selected from the one or more extracted data records as a search target; acquires candidate upstream material properties, which are the physical properties of each of the candidate upstream materials, for each of the two or more candidate upstream materials used in an upstream process of manufacturing; and executes an optimization method using the search target and the candidate upstream material properties of each of the two or more candidate upstream materials to determine the blending ratio between the two or more candidate upstream materials. (Note 2) The search system according to Note 1, wherein the candidate downstream substance is a candidate composition, the candidate downstream substance properties are candidate composition properties which are properties of the candidate composition, the related physical properties are candidate component properties which are properties of candidate component constituting the candidate composition, the candidate upstream substance is a candidate raw material selected for the candidate component, and the candidate upstream substance properties are candidate raw material properties which are properties of the candidate raw material. (Note 3) The search system according to Note 1 or 2, wherein at least one processor inputs the related physical properties to a prediction model generated by machine learning to calculate the characteristics of a downstream substance from physical properties related to the characteristics of the downstream substance for each of the plurality of related physical properties, calculates the candidate downstream substance properties, and stores the data record showing the combination of the related physical properties input to the prediction model and the calculated candidate downstream substance properties in the database as one of the plurality of data records.(Note 4) The search system according to any one of Notes 1 to 3, wherein at least one processor displays the extracted one or more data records on a display device, and in response to a user operation to select one of the displayed one or more data records, sets the related physical properties indicated by the one data record as the search target, and displays the determined mixing ratio on the display device. (Note 5) The search system according to any one of Notes 1 to 4, wherein at least one processor displays a plurality of candidate upstream substances on a display device, and in response to a user operation to select one or more candidate upstream substances from the plurality of displayed candidate upstream substances, acquires the physical properties of each of the two or more candidate upstream substances. (Note 6) The search system according to any one of Notes 1 to 5, wherein the related physical properties are the physical properties of a candidate midstream material, which is a material obtained in the midstream process of the manufacturing process, and at least one processor determines the blending ratio based on an evaluation value that shows the difference between the provisional solution of the candidate midstream material properties calculated based on the respective candidate upstream material properties of the two or more candidate upstream materials and the search target during the execution of the optimization method. (Note 7) The search system according to Note 6, wherein each of the candidate upstream material properties and the candidate midstream material properties is expressed by one or more parameters, and at least one processor calculates the parameter value of the candidate midstream material for each of the one or more parameters based on the respective parameter values ​​and blending ratios of the two or more candidate upstream materials, and expresses the provisional solution of the candidate midstream material properties using one or more parameter values ​​of the candidate midstream material. (Note 8) The search system according to Note 7, wherein at least one processor calculates the product of the parameter value and the mixing ratio for each of the two or more candidate upstream materials for each of the two or more candidate upstream materials, and calculates the sum of the two or more products corresponding to the two or more candidate upstream materials as the parameter value of the candidate midstream material.(Note 9) The search system according to Note 7, wherein at least one processor calculates a quotient obtained by dividing the blending ratio by the parameter value for each of the two or more candidate upstream materials for each of the two or more candidate upstream materials, for each of the two or more parameter selected from the one or more parameter, and calculates the parameter value of the candidate midstream material as the reciprocal of the sum of the two or more quotients corresponding to the two or more candidate upstream materials. (Note 10) The search system according to any one of Notes 1 to 9, wherein the optimization method is simulated annealing, and at least one processor, in the simulated annealing method, repeatedly searches for the blending ratio while changing the blending ratio between the two candidate upstream materials selected from the two or more candidate upstream materials so that the sum of the blending ratios between the two candidate upstream materials does not change, and determines the blending ratio. (Note 11) A search method performed by a search system comprising at least one processor, comprising: acquiring search conditions related to candidate downstream material properties, which are the properties of a candidate downstream material, which is a substance obtained in a downstream process of manufacturing; extracting one or more data records corresponding to the search conditions from a database storing a plurality of data records indicating combinations of the candidate downstream material properties and related physical properties, which are physical properties related to the candidate downstream material properties; setting the related physical properties indicated by one data record selected from the one or more extracted data records as a search target; acquiring candidate upstream material properties, which are the physical properties of each of the two or more candidate upstream materials, which are substances used in an upstream process of manufacturing; and determining the blending ratio between the two or more candidate upstream materials by executing an optimization method using the search target and the respective candidate upstream material properties of the two or more candidate upstream materials.(Note 12) A search program that causes a computer to perform the following steps: acquiring search conditions related to candidate downstream material properties, which are the characteristics of a candidate downstream material, which is a substance obtained in a downstream process of manufacturing; extracting one or more data records corresponding to the search conditions from a database that stores a plurality of data records showing combinations of the candidate downstream material properties and related physical properties, which are physical properties related to the candidate downstream material properties; setting the related physical properties indicated by one data record selected from the one or more extracted data records as a search target; acquiring candidate upstream material properties, which are the physical properties of each of the two or more candidate upstream materials, which are substances used in the upstream process of manufacturing; and executing an optimization method using the search target and the candidate upstream material properties of each of the two or more candidate upstream materials to determine the blending ratio between the two or more candidate upstream materials.

[0075] According to appendices 1, 11, and 12, first, one or more data records showing combinations of candidate downstream material properties and related physical properties are extracted based on search conditions related to the properties of candidate downstream materials obtained in the downstream manufacturing process (candidate downstream material properties). Then, an optimization method is executed that uses the related physical properties indicated by one selected data record as the search target to determine the blending ratio between two or more candidate upstream materials used in the upstream manufacturing process. A two-stage process is executed sequentially: searching for related physical properties corresponding to a downstream material having the desired properties, and searching for the blending ratio between two or more upstream materials to obtain those related physical properties. This mechanism allows for the search of a blend of materials used in the upstream manufacturing process that realizes the desired properties for a material obtained in the downstream manufacturing process in a shorter amount of time. The more types of candidate upstream materials there are, the more enormous the number of possible blending ratio patterns becomes. However, even in this case, it becomes possible to search for a blend of upstream materials that realizes a downstream material having the desired properties in a shorter amount of time.

[0076] According to Appendix 2, first, one or more data records showing combinations of candidate composition characteristics and candidate component properties are extracted based on search conditions related to candidate composition characteristics. Then, an optimization method is executed that uses the candidate component property indicated by one of the selected data records as the search target, and the blending ratio between two or more candidate raw materials to produce the selected candidate component is determined. Since the two-stage process of searching for a component corresponding to a composition with desired characteristics and searching for the blending ratio between two or more raw materials constituting that component is executed in succession, the blending ratio of raw materials that realize the component constituting a composition with desired characteristics can be searched in a shorter time.

[0077] According to Appendix 3, candidate downstream material properties are calculated from related physical properties using a prediction model, and data records showing combinations of related physical properties and candidate downstream material properties are automatically stored in the database. Therefore, the database used by the search system can be prepared in a short time. For example, a dataset showing multiple combinations of related physical properties and candidate downstream material properties can be prepared in a short time without having to repeat experiments to identify candidate downstream material properties or collect information on candidate downstream material properties.

[0078] According to Appendix 4, the results of a two-stage processing process—the search results for related physical properties corresponding to downstream materials with desired characteristics, and the search results for the mixing ratio between two or more upstream materials to obtain those related physical properties—are displayed sequentially on the display device. This mechanism for the user interface allows the user to receive the results of these two stages in an easily understandable format.

[0079] According to Appendix 5, the user is given the opportunity to select two or more candidate upstream substances to determine the mixing ratio, allowing the user's requirements for determining the mixing ratio to be reflected more flexibly in the search. This contributes to improving the usability of the search system.

[0080] According to Appendix 6, the mixing ratio is determined based on an evaluation value that shows the difference between the provisional solutions for the properties of the candidate midstream material and the search target. By adopting this mixing ratio, it becomes possible to increase the probability of realizing a downstream material with the desired properties.

[0081] According to Appendix 7, for each of the one or more parameters that represent the properties of the candidate midstream material, the parameter value of the candidate midstream material is calculated based on the parameter values ​​and mixing ratios of two or more candidate upstream materials. Then, a provisional solution for the properties of the candidate midstream material is expressed by the one or more parameter values ​​of the candidate midstream material. By representing the properties with one or more parameters while calculating each parameter independently of the others, the calculations required to determine the mixing ratio can be made relatively simpler even when the properties of the candidate midstream material are complex. This leads to a simplification of the search process and, consequently, can contribute to a reduction in computation time.

[0082] According to Appendix 8, it is expected that specific parameters can be calculated more accurately by using the so-called weighted sum calculation method. Consequently, the properties of candidate midstream materials expressed by one or more parameters can be determined more accurately.

[0083] According to Appendix 9, by calculating the reciprocal of the sum of quotients obtained by dividing the mixing ratio by the parameter value as the parameter value for the candidate midstream material properties, it is expected that specific parameters can be calculated more accurately. Consequently, the candidate midstream material properties expressed by one or more parameters can be determined more accurately.

[0084] According to Appendix 10, in the annealing process, the state (mixing ratio) can be transitioned so that the sum of the mixing ratios between the two candidate upstream substances remains unchanged, thereby simplifying the process related to the state transition.

[0085] 10... Search system, 11... Data generation unit, 12... Search unit, 13... Search unit, 20... Prediction model, 30... Database, 40... User terminal.

Claims

1. A search system comprising at least one processor, wherein the at least one processor acquires search conditions related to candidate downstream material properties, which are characteristics of candidate downstream materials, which are materials obtained in the downstream processes of manufacturing; extracts one or more data records corresponding to the search conditions from a database that stores a plurality of data records showing combinations of candidate downstream material properties and related physical properties, which are physical properties related to the candidate downstream material properties; sets the related physical properties indicated by one data record selected from the one or more extracted data records as a search target; acquires candidate upstream material properties, which are physical properties of each of the two or more candidate upstream materials, which are materials used in the upstream processes of manufacturing; and executes an optimization method using the search target and the candidate upstream material properties of each of the two or more candidate upstream materials to determine the blending ratio between the two or more candidate upstream materials.

2. The search system according to claim 1, wherein the candidate downstream substance is a candidate composition, the characteristics of the candidate downstream substance are candidate composition characteristics, the related physical properties are candidate component properties, the physical properties of candidate components constituting the candidate composition, the candidate upstream substance is a candidate raw material selected for the candidate component, and the physical properties of the candidate upstream substance are candidate raw material properties, the physical properties of the candidate raw material.

3. The search system according to claim 1 or 2, wherein at least one processor inputs the related physical properties to a predictive model generated by machine learning to calculate the characteristics of a downstream material from physical properties related to the characteristics of the downstream material for each of the plurality of related physical properties, calculates the candidate downstream material characteristics, and stores the data record showing the combination of the related physical properties input to the predictive model and the calculated candidate downstream material characteristics in the database as one of the plurality of data records.

4. The search system according to any one of claims 1 to 3, wherein at least one processor displays the extracted one or more data records on a display device, and in response to a user operation to select one data record from the displayed one or more data records, sets the related physical property indicated by the one data record as the search target, and displays the determined blending ratio on the display device.

5. The search system according to any one of claims 1 to 4, wherein at least one processor displays a plurality of candidate upstream materials on a display device, and in response to a user operation to select one or more candidate upstream materials from the displayed plurality of candidate upstream materials, the physical properties of the candidate upstream materials are obtained for each of the two or more candidate upstream materials.

6. The search system according to any one of claims 1 to 5, wherein the related physical properties are candidate midstream material properties, which are properties of a candidate midstream material obtained in the midstream process of the manufacturing, and at least one processor determines the blending ratio in the execution of the optimization method based on an evaluation value that shows the difference between the provisional solution of the candidate midstream material properties calculated based on the candidate upstream material properties of each of the two or more candidate upstream materials and the search target.

7. The search system according to claim 6, wherein each of the candidate upstream material properties and the candidate midstream material properties is represented by one or more parameters, and at least one processor calculates the parameter value of the candidate midstream material for each of the one or more parameters based on the parameter values ​​and blending ratios of the two or more candidate upstream materials, and expresses the provisional solution for the candidate midstream material properties using one or more parameter values ​​of the candidate midstream material.

8. The search system according to claim 7, wherein the at least one processor calculates the product of the parameter value and the blending ratio for each of the two or more candidate upstream materials for each of the two or more candidate upstream materials, and calculates the sum of the two or more products corresponding to the two or more candidate upstream materials as the parameter value of the candidate midstream material.

9. The search system according to claim 7, wherein the at least one processor calculates a quotient obtained by dividing the blending ratio by the parameter value for each of the two or more candidate upstream materials for each of the two or more candidate upstream materials, for each of the at least one parameter selected from the one or more parameters, and calculates the reciprocal of the sum of the two or more quotients corresponding to the two or more candidate upstream materials as the parameter value for the candidate midstream material.

10. The search system according to any one of claims 1 to 9, wherein the optimization method is an simulated annealing method, and at least one processor repeatedly searches for the blending ratio while changing the blending ratio between two candidate upstream materials selected from the two or more candidate upstream materials such that the sum of the blending ratios between the two candidate upstream materials remains unchanged, thereby determining the blending ratio.

11. A search method performed by a search system comprising at least one processor, comprising: obtaining search conditions related to candidate downstream material properties, which are characteristics of a candidate downstream material, which is a substance obtained in a downstream process of manufacturing; extracting one or more data records corresponding to the search conditions from a database storing a plurality of data records indicating combinations of the candidate downstream material properties and related physical properties, which are physical properties related to the candidate downstream material properties; setting the related physical properties indicated by one data record selected from the one or more extracted data records as a search target; obtaining candidate upstream material properties, which are physical properties of the candidate upstream material, for each of the two or more candidate upstream materials, which are substances used in an upstream process of manufacturing; and determining the blending ratio between the two or more candidate upstream materials by executing an optimization method using the search target and the respective candidate upstream material properties of the two or more candidate upstream materials.

12. A search program that causes a computer to perform the following steps:

11. Obtain search conditions related to candidate downstream material properties, which are the characteristics of candidate downstream materials, which are materials obtained in the downstream processes of manufacturing; 2. Extract one or more data records corresponding to the search conditions from a database that stores a plurality of data records showing combinations of candidate downstream material properties and related physical properties, which are physical properties related to the candidate downstream material properties; 3. Set the related physical property indicated by one data record selected from the one or more extracted data records as a search target; 4. Obtain candidate upstream material properties, which are the physical properties of each of the two or more candidate upstream materials, which are materials used in the upstream processes of manufacturing; and 5. Execute an optimization method using the search target and the candidate upstream material properties of each of the two or more candidate upstream materials to determine the blending ratio between the two or more candidate upstream materials.

Images