Design support device, design support method, and program

Abstract

The present invention provides a design support device, method, and program that reduce the workload of workers in finding the desired formulation. [Solution] In the design support system 1, the design support device comprises an input receiving unit that receives input of first mixing conditions from an operator, a search condition acquisition unit that generates or acquires search conditions for searching a database that stores past experimental data based on the first mixing conditions, a search unit that searches the database using the search conditions, and a display control unit that displays the search results of the search unit on a display device. If no experimental data matching the first mixing conditions exists in the database, the search unit generates or acquires a second mixing condition that relaxes the first mixing conditions and searches the database using search conditions based on the second mixing conditions. If no experimental data matching the first mixing conditions exists, the display control unit displays experimental data matching the second mixing conditions as reference point data close to the first mixing conditions on the display device.

Description

Technical Field

[0001] The present disclosure relates to a design support device, a design support method, and a program.

Background Art

[0002] In recent years, in the field of material development, the efficiency of material development has been improved by utilizing artificial intelligence (AI). Workers who perform material development using AI have been having AI predict physical properties by quantitatively specifying a formulation from a large number of options.

[0003] For example, in a research and development support system aimed at improving the efficiency of research and development, a technique of using a research and development chatbot using AI as a personal research assistant has been conventionally known (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, when it is necessary to quantitatively specify a formulation from a large number of options in order to find a desired formulation, the labor of the worker's input increases, and the worker's workload may increase. Note that Patent Document 1 does not describe such content.

[0006] An object of the present disclosure is to provide a design support device, a design support method, and a program that reduce the workload of a worker who finds a desired formulation.

Means for Solving the Problems

[0007] The present disclosure includes the following configurations.

[0008] [1] An input receiving unit that receives input of the first mixing conditions from the worker in a conversational format, A prompt generation unit that generates a prompt incorporating the first formulation conditions, A first search unit searches a database of past experimental data for experimental data that matches the first blending conditions, using search conditions generated by the language model based on the prompt, and if no experimental data matches the first blending conditions, searches for experimental data that matches a second blending condition which is a relaxed version of the first blending condition. A prediction unit predicts the physical properties of multiple candidate formulations generated from experimental data formulations that match the second formulation conditions, A second search unit searches for candidate formulations that match the first formulation conditions from the plurality of candidate formulations based on the search conditions, A formulation display unit that displays candidate formulations that meet the first formulation conditions, A design support device equipped with the following features.

[0009] [2] If there are no candidate formulations that match the first formulation conditions, the second search unit searches for candidate formulations that match the second formulation conditions obtained by relaxing the first formulation conditions. The aforementioned blending display unit displays the candidate blending blends that meet the second blending conditions. [1] Design support device as described above.

[0010] [3] The system further comprises a search condition generation unit which generates the search conditions based on the prompts, with the language model generating the search conditions based on the prompts. The design support device described in [1] or [2].

[0011] [4] The input receiving unit receives the non-quantitative input of the first blending conditions from the operator in a conversational format. The search condition generation unit generates the search conditions based on the prompt, which incorporates the non-quantitative first blending conditions received from the operator, and the language model generates the search conditions. [3] Design support device as described in [3].

[0012] [5] The prediction unit predicts the physical properties of the candidate formulation using a machine learning model that has learned the correspondence between the formulation and the physical properties. A design support device as described in any one of items [1] through [4].

[0013] [6] The formulation display unit displays experimental data that matches the first formulation conditions if experimental data that matches the first formulation conditions is available. A design support device as described in any one of items [1] through [5].

[0014] [7] Computers, An input reception step in which the operator inputs the first blending conditions in a conversational format, A prompt generation step that generates a prompt incorporating the first formulation conditions described above, A first search step involves using search conditions generated by the language model based on the prompt to search for experimental data matching the first formulation conditions from a database storing past experimental data, and if no experimental data matching the first formulation conditions is found, searching for experimental data matching the second formulation conditions, which are relaxed versions of the first formulation conditions. A prediction step of predicting the physical properties of multiple candidate formulations generated from experimental data formulations that match the second formulation conditions, A second search step involves searching for candidate formulations from the aforementioned plurality of candidate formulations that match the first formulation conditions based on the search conditions, A formulation display step that displays candidate formulations that meet the first formulation conditions, A design support method that includes the following features.

[0015] [8] Computers, An input reception procedure that receives input of the first blending condition from the worker in a conversational format. A prompt generation procedure for generating a prompt incorporating the first formulation conditions described above, Using the search conditions generated by the language model based on the prompt, search for experimental data that matches the first formulation condition from a database that stores past experimental data. If there is no experimental data that matches the first formulation condition, search for experimental data that matches the second formulation condition obtained by relaxing the first formulation condition. This is the first search procedure. A prediction procedure for predicting the physical properties of a plurality of candidate formulations generated from the formulations of the experimental data that match the second formulation condition. A second search procedure for searching for the candidate formulation that matches the first formulation condition from the plurality of candidate formulations based on the search conditions. A formulation display procedure for displaying the candidate formulation that matches the first formulation condition. A program for executing the above.

[0016] [9] A design support system comprising a plurality of computers, An input reception unit that receives an input of the first formulation condition from an operator in a conversation format, A prompt generation unit that generates a prompt incorporating the first formulation condition, Using the search conditions generated by the language model based on the prompt, search for experimental data that matches the first formulation condition from a database that stores past experimental data. If there is no experimental data that matches the first formulation condition, search for experimental data that matches the second formulation condition obtained by relaxing the first formulation condition. This is the first search unit. A prediction unit that predicts the physical properties of a plurality of candidate formulations generated from the formulations of the experimental data that match the second formulation condition, A second search unit that searches for the candidate formulation that matches the first formulation condition from the plurality of candidate formulations based on the search conditions, A formulation display unit that displays the candidate formulation that matches the first formulation condition, A design support system comprising the above.

Advantages of the Invention

[0017] According to this disclosure, a design support device, a design support method, and a program can be provided that reduce the workload of an operator in finding a desired formulation. [Brief explanation of the drawing]

[0018] [Figure 1] This is a diagram illustrating an example of a design support system according to this embodiment. [Figure 2] This is a hardware configuration diagram of an example of a computer according to this embodiment. [Figure 3] This is a functional configuration diagram of an example of a design support system according to this embodiment. [Figure 4] This is a flowchart illustrating an example of the processing of the design support system according to this embodiment. [Figure 5] This is an illustrative image of an example screen that accepts input for the first blending condition in a conversational format. [Figure 6] This is an illustrative image of an example screen that accepts input for the first blending condition in a conversational format. [Figure 7] This is an explanatory diagram of an example of a prompt incorporating the first formulation condition. [Figure 8] This is an explanatory diagram illustrating an example of search criteria for finding experimental data from past experimental data that matches the first formulation condition. [Figure 9] This is an illustrative image of a screen displaying experimental data that matches the second formulation condition, which is a relaxed version of the first formulation condition. [Figure 10] This is an explanatory diagram illustrating an example of the conditions for formula production. [Figure 11] This is an illustrative image of an example screen that displays candidate combinations that match the first combination condition. [Figure 12] This is an example image of a screen for entering the blending conditions. [Figure 13] This is an explanatory diagram illustrating an example of a large-scale language model used in the design support system according to this embodiment. [Modes for carrying out the invention]

[0019] Next, embodiments of the present invention will be described in detail. However, the present invention is not limited to the following embodiments.

[0020] [First Embodiment] <System Configuration> Figure 1 is a configuration diagram of an example of a design support system according to this embodiment. The design support system 1 in Figure 1 includes a design support device 10 and a user terminal 12. The design support device 10 and the user terminal 12 are connected via a communication network 18 such as a local area network (LAN) or the internet, enabling data communication.

[0021] The user terminal 12 is an information processing terminal operated by the worker, such as a PC, tablet, or smartphone. The user terminal 12 displays a screen on its display device that accepts information input from the worker and accepts information input from the worker. The user terminal 12 also transmits the information received from the worker to the design support device 10, which then executes a process to reduce the workload of the worker in finding the desired mixture.

[0022] The user terminal 12 receives information on the execution results of the design support device 10 and displays it on the display device for the operator to confirm. For example, the user terminal 12 receives information on the mixture requested by the operator and displays it on the display device for the operator to confirm.

[0023] The design support device 10 is an information processing device such as a PC that assists the worker in finding the desired mixture. The design support device 10 performs processing to reduce the workload of the worker in finding the desired mixture. The design support device 10 transmits information on the results of the processing and displays the information on the user terminal 12.

[0024] For example, the design support device 10 uses a language model to understand the mixing conditions (hereinafter referred to as the first mixing conditions) input by the operator in a conversational format, and searches a database for past experimental data that matches the first mixing conditions.

[0025] If there is past experimental data that matches the first blending condition, the design support device 10 displays it on the user terminal 12. If there is no past experimental data that matches the first blending condition, the design support device 10 uses a language model to search for past experimental data that matches a blending condition that relaxes the first blending condition (hereinafter referred to as the second blending condition) as a reference point.

[0026] The design support device 10 generates candidate formulations similar to those in past experimental data from formulations that match the second formulation conditions, and predicts the physical properties of the candidate formulations using a machine learning model that has already learned the correspondence between formulations and physical properties.

[0027] If the design support device 10 finds a candidate formulation that matches the first formulation conditions, it displays the candidate formulation that matches the first formulation conditions on the user terminal 12. If the design support device 10 finds no candidate formulation that matches the first formulation conditions, it displays a candidate formulation that matches the second formulation conditions on the user terminal 12.

[0028] The design support device 10 utilizes a database that stores past experimental data. The design support device 10 may use its own internal database, or it may use a database such as a database server connected via the communication network 18.

[0029] Furthermore, the machine learning methods used for machine learning models that have learned the correspondence between formulations and physical properties (pre-trained machine learning models) include linear, generalized linear, partial least squares, kernel ridge, Gaussian process, k-nearest neighbors, decision trees, random forests, AdaBoost, bagging, gradient boosting, support vector machines, or neural networks. When predicting the physical properties of candidate formulations using a pre-trained machine learning model, quantitative candidate formulations are input into the pre-trained machine learning model, and the physical properties of the candidate formulations are output.

[0030] Furthermore, the design support device 10 may use a large-scale language model such as ChatGPT (registered trademark) as its language model. The design support device 10 may use a language model stored in its own memory, or it may use a language model of a server (including cloud services) connected via the communication network 18.

[0031] The design support system 1 in Figure 1 may be implemented using a design support device 10 with web server functionality and a user terminal 12 that runs a web application using a web browser. The design support system 1 in Figure 1 may also be implemented by having an application installed on the user terminal 12 cooperate with a program installed on the design support device 10 to perform processing.

[0032] It should be noted that the design support system 1 in Figure 1 is merely an example, and there are various system configurations depending on the application and purpose. For example, the design support device 10 may be implemented using multiple computers, implemented as a cloud computing service, or implemented in conjunction with a cloud computing service. Furthermore, the design support system 1 in Figure 1 may be implemented using a standalone computer.

[0033] <Hardware Configuration> The design support device 10 and user terminal 12 in Figure 1 are implemented, for example, by a computer 500 with the hardware configuration shown in Figure 2.

[0034] Figure 2 is a hardware configuration diagram of an example of a computer according to this embodiment. The computer 500 in Figure 2 is equipped with an input device 501, a display device 502, an external interface 503, RAM 504, ROM 505, a CPU 506, a communication interface 507, and an HDD 508, and each is interconnected via bus B. Note that the input device 501 and the display device 502 may be used in a connected configuration.

[0035] The input device 501 includes a touch panel, operation keys and buttons, a keyboard and mouse, etc., used by the operator to input various signals. The display device 502 consists of a display such as a liquid crystal or organic EL that displays the screen, and a speaker that outputs sound data such as voice and sound. The communication I / F 507 is an interface for the computer 500 to perform data communication.

[0036] Furthermore, HDD508 is an example of a non-volatile storage device that stores programs and data. The programs and data stored include the OS, which is the basic software that controls the entire computer 500, and applications that provide various functions on the OS. Note that computer 500 may use a drive device that uses flash memory as a storage medium (for example, a solid-state drive: SSD) instead of HDD508.

[0037] External I / F 503 is an interface to external devices. External devices include recording media 503a, etc. This allows computer 500 to read and / or write to recording media 503a via external I / F 503. Recording media 503a include flexible disks, CDs, DVDs, SD memory cards, USB memory, etc.

[0038] ROM505 is an example of non-volatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. ROM505 stores programs and data such as the BIOS, OS settings, and network settings that are executed when the computer 500 starts up. RAM504 is an example of volatile semiconductor memory (storage device) that temporarily holds programs and data.

[0039] The CPU 506 is an arithmetic unit that controls and implements the functions of the entire computer 500 by reading programs and data from storage devices such as ROM 505 and HDD 508 onto RAM 504 and executing processing. In this embodiment, the computer 500 can implement various functions of the design support device 10 and user terminal 12, as described later, by executing programs.

[0040] <Functional Configuration> The functional configuration of the design support system 1 according to this embodiment will now be described. Figure 3 is a functional configuration diagram of an example of the design support system according to this embodiment. Note that parts of the configuration diagram in Figure 3 that are not necessary for the explanation of this embodiment have been omitted as appropriate.

[0041] The design support device 10 includes a request receiving unit 20, a response transmission unit 22, an input receiving unit 24, a prompt generation unit 26, a search condition generation unit 28, a first search unit 30, a control unit 32, a prediction unit 34, a second search unit 36, a formulation display unit 38, a database storage unit 50, a machine learning model storage unit 52, and a language model storage unit 54. The user terminal 12 includes an information display unit 60, an operation receiving unit 62, a request transmission unit 64, and a response receiving unit 66.

[0042] The information display unit 60 displays information on the display device 502, including a screen for receiving information input from the operator and the execution results of the design support device 10's processing. The operation reception unit 62 receives operations from the operator, such as information input. The request transmission unit 64 transmits processing requests to the design support device 10 in response to the information input from the operator. The response reception unit 66 receives responses from the design support device 10 to the processing requests transmitted by the request transmission unit 64.

[0043] The request receiving unit 20 receives a processing request from the user terminal 12. The response transmission unit 22 responds with the result of the processing performed in accordance with the processing request. The input receiving unit 24 works in conjunction with the user terminal 12 to receive input of the first blending conditions desired by the worker in a conversational format.

[0044] The prompt generation unit 26 generates a prompt by incorporating the first blending conditions received in conversational format into the variable portion of the template prompt. Details of the prompts generated by the prompt generation unit 26 will be described later.

[0045] The search condition generation unit 28 obtains the search conditions generated (output from the language model) by inputting the prompt generated by the prompt generation unit 26 into the language model. The language model understands the first blending condition incorporated into the prompt and generates a search condition (search condition expression) for searching for experimental data that matches the first blending condition from past experimental data.

[0046] Furthermore, the search condition generation unit 28 understands the first blending condition incorporated into the prompt and generates search conditions to search for experimental data from past experimental data that match the second blending condition, which is a relaxed version of the first blending condition.

[0047] The first search unit 30 searches for experimental data that matches the first blending conditions from past experimental data using search conditions. If experimental data matching the first blending conditions is found in the database of past experimental data, the design support device 10 terminates its search for the blending formula desired by the operator.

[0048] If there is no experimental data in the database that stores past experimental data that matches the first blending condition, the first search unit 30 searches the database that stores past experimental data for experimental data that matches the second blending condition using search conditions for past experimental data.

[0049] The prediction unit 34 generates multiple candidate formulations that are close to the experimental data formulations that match the second formulation conditions, and predicts the physical properties of the multiple candidate formulations using, for example, a trained machine learning model stored in the machine learning model storage unit 52.

[0050] The second search unit 36 ​​refers to the physical properties of the multiple predicted candidate formulations and searches for a candidate formulation that matches the first formulation condition from among the multiple candidate formulations. If there is no candidate formulation that matches the first formulation condition, the second search unit 36 ​​searches for a candidate formulation that matches the second formulation condition from among the multiple candidate formulations.

[0051] The formulation display unit 38, if it finds experimental data in the database that stores past experimental data that matches the first formulation conditions, presents the experimental data that matches the first formulation conditions to the operator by displaying it on the user terminal 12.

[0052] Furthermore, if the formulation display unit 38 does not have experimental data in the database that stores past experimental data that matches the first formulation condition, it will present candidate formulations that match the first formulation condition or candidate formulations that match the second formulation condition to the operator by displaying them on the user terminal 12.

[0053] The database storage unit 50 stores past experimental data. The machine learning model storage unit 52 stores machine learning models that have learned the correspondence between formulations and physical properties. The language model storage unit 54 stores language models that generate search conditions based on input prompts.

[0054] The control unit 32 controls the request receiving unit 20, response transmission unit 22, input receiving unit 24, prompt generation unit 26, search condition generation unit 28, first search unit 30, prediction unit 34, second search unit 36, combination display unit 38, database storage unit 50, machine learning model storage unit 52, and language model storage unit 54 shown in Figure 3.

[0055] Note that the configuration of the design support system 1 in Figure 3 is just one example. The design support system 1 according to this embodiment can be realized with various configurations. For example, the database storage unit 50, the machine learning model storage unit 52, and the language model storage unit 54 may be a storage device, computer, or cloud storage that can communicate data with the design support device 10 via a communication network 18.

[0056] <Processing> The design support system 1 according to this embodiment assists the worker in finding the desired mix, for example, by following the procedure shown in Figure 4. Figure 4 is a flowchart showing an example of the processing of the design support system according to this embodiment.

[0057] In step S10, the input receiving unit 24 of the design support device 10 displays, for example, the screen 1000 shown in Figures 5 and 6 on the user terminal 12. Figures 5 and 6 are illustrative images of an example of a screen that receives input of the first mixing conditions in a conversational format. Figure 5 is an example of screen 1000 that has received a conversational specification from the operator as the first mixing condition: "What is a mixing formula where material C is used in an amount of 0.4 or more and physical property A is between 30,000 and 50,000?" Figure 6 is an example of screen 1000 that has received a conversational specification from the operator as the first mixing condition: "What is a mixing formula that increases physical property A?" As shown in Figure 6, the operator can specify non-quantitative first mixing conditions in a conversational format.

[0058] In step S12, when the user terminal 12 receives an operation from the worker by pressing the send button on the screen 1000, it transmits the first blending conditions in conversational format entered on the screen 1000 to the design support device 10. The input receiving unit 24 of the design support device 10 receives the first blending conditions in conversational format.

[0059] In step S14, the prompt generation unit 26 generates a prompt, for example, as shown in Figure 7, by incorporating the first conversational formatting conditions received from the user terminal 12 into the variable portion of the template prompt.

[0060] Figure 7 is an explanatory diagram of an example of a prompt incorporating the first blending condition. The prompt shown in Figure 7 incorporates the first blending condition, "What blend uses material C at a level of 0.4 or higher and has physical property A between 30,000 and 50,000?", which was specified by the operator in conversational format on screen 1000 of Figure 5. Note that the prompt in Figure 7 is written in English to save tokens, but it may also be written in a language other than English, such as Japanese.

[0061] In step S16, the search condition generation unit 28 generates a first search condition by inputting the prompt generated by the prompt generation unit 26 into the language model. The language model understands the first blending condition incorporated into the prompt and generates a search condition, for example, as shown in Figure 8, to search for experimental data from past experimental data that matches the first blending condition.

[0062] Figure 8 is an explanatory diagram of an example of search criteria for searching for experimental data that matches the first formulation condition from past experimental data. The search criteria in Figure 8 are an example of search criteria for searching for experimental data that matches the first formulation condition, "What formulations use material C at a ratio of 0.4 or higher and have physical property A between 30,000 and 50,000?"

[0063] In step S18, the first search unit 30 searches past experimental data for experimental data that matches the first formulation conditions using the search conditions generated by the search condition generation unit 28 in step S16.

[0064] In step S20, if the first search unit 30 finds past experimental data that matches the first blending conditions, it proceeds to the process in step S22. In step S22, the first search unit 30 notifies the blending display unit 38 of the experimental data that matches the first blending conditions. The blending display unit 38 presents the experimental data that matches the first blending conditions to the operator by displaying it on the user terminal 12.

[0065] In step S20, if there is no past experimental data that matches the first blending condition, the search condition generation unit 28 proceeds to the process in step S24. In step S24, the search condition generation unit 28 understands the first blending condition incorporated in the prompt and generates search conditions for searching past experimental data that match the second blending condition, which is a relaxed version of the first blending condition. The search condition generation unit 28 may also generate the second search conditions by inputting a prompt to the language model that causes the language model to generate search conditions for searching past experimental data that match the second blending condition, which is a relaxed version of the first blending condition.

[0066] In step S26, the first search unit 30 searches past experimental data for experimental data that matches the second formulation conditions using the search conditions generated by the search condition generation unit 28 in step S24.

[0067] In step S28, the first search unit 30 notifies the formulation display unit 38 of experimental data that matches the second formulation conditions. The formulation display unit 38 presents the experimental data that matches the second formulation conditions to the operator by displaying it on the user terminal 12, for example, as shown in screen 1100 in Figure 9.

[0068] Figure 9 is an illustrative image of an example screen displaying experimental data that matches the second formulation condition, which is a relaxed version of the first formulation condition. Screen 1100 in Figure 9 does not match the first formulation condition, "a formulation in which material C is used at a level of 0.4 or higher and physical property A is between 30,000 and 50,000," but displays past experimental data that is close to the first formulation condition. Subsequently, the formulation display unit 38 displays the message 1200 in Figure 9.

[0069] In step S30, the prediction unit 34 generates several candidate formulations that are close to the formulations in experimental data that match the second formulation conditions, for example, using the formulation generation conditions shown in Figure 10. Figure 10 is an explanatory diagram of an example of formulation generation conditions. The formulation generation conditions shown in Figure 10 can be created using past experimental data that matches the second formulation conditions. For example, if the amount of "material A" in past experimental data that matches the second formulation conditions is in the range of "0.1 to 0.5", then the amount of "material A" can be set using a random number in the range of "0.1 to 0.5".

[0070] In step S32, the prediction unit 34 predicts the physical properties of multiple candidate formulations using, for example, a trained machine learning model stored in the machine learning model storage unit 52.

[0071] In step S34, the second search unit 36 ​​uses the search conditions generated by the search condition generation unit 28 in step S16 to search for a candidate blend from a plurality of candidate blends that matches the first blending conditions.

[0072] Furthermore, if there are no candidate formulations that match the first formulation conditions, the second search unit 36 ​​uses the search conditions generated by the search condition generation unit 28 in step S24 to search for candidate formulations that match the second formulation conditions from among multiple candidate formulations.

[0073] If the second search unit 36 ​​finds a candidate blend that matches the first blending conditions, it proceeds to step S36. In step S36, the second search unit 36 ​​notifies the blending display unit 38 of the candidate blend that matches the first blending conditions.

[0074] The formulation display unit 38 presents candidate formulations that meet the first formulation conditions to the operator by displaying them on the user terminal 12, for example, as shown in the screen 1300 in Figure 11. Figure 11 is an illustrative image of an example screen that displays candidate formulations that meet the first formulation conditions. The screen 1300 in Figure 11 displays multiple candidate formulations that meet the first formulation condition, "a formulation in which material C is used in an amount of 0.4 or more and physical property A is between 30,000 and 50,000".

[0075] If the second search unit 36 ​​finds no candidate formulations that match the first formulation conditions, it proceeds to step S38. In step S38, the second search unit 36 ​​notifies the formulation display unit 38 of candidate formulations that match the second formulation conditions. The formulation display unit 38 presents the candidate formulations that match the second formulation conditions to the operator by displaying them on the user terminal 12.

[0076] According to the flowchart shown in Figure 4, the operator inputs the first blending condition in a conversational format. Even if the first blending condition is not necessarily quantitative, the language model generates appropriate search conditions, allowing the operator to find the desired blend from past experimental data. Furthermore, the flowchart in Figure 4 reduces the workload and improves usability compared to inputting the blending condition as shown in screen 1400 of Figure 12.

[0077] Figure 12 is an illustrative image of an example screen for inputting mixing conditions. The screen 1400 shown in Figure 12 requires users to select materials from a large number of options and quantitatively specify the minimum and maximum amounts of each material, which places a heavy workload on the operator. In particular, experiments using many materials require a large amount of input, increasing the workload on the operator.

[0078] Furthermore, according to the flowchart shown in Figure 4, by having the operator input the first blending conditions in a conversational format, the likelihood of finding a candidate blend that matches the first blending conditions increases, even if the desired blend is not found in past experimental data. Therefore, the likelihood of the operator finding an unexpectedly favorable candidate blend increases.

[0079] Furthermore, according to the flowchart shown in Figure 4, experimental data matching the second blending condition, which is a relaxed version of the first blending condition, is searched from past experimental data. Multiple candidate blends close to the blends in the experimental data that match the second blending condition are then randomly generated. This reduces the possibility of unrealistic candidate blends being proposed, even if the blending conditions are not necessarily quantitative. For example, according to the flowchart shown in Figure 4, by randomly generating multiple candidate blends close to the blends in the experimental data that match the second blending condition, it becomes easier to satisfy the conditions required for blending (such as the sum of the main chains being 1), and the possibility of proposing realistic candidate blends is increased.

[0080] The design support system 1 according to this embodiment can utilize a large-scale language model 100 whose functionality has been extended by a library, as shown in Figure 13, for example. Figure 13 is an explanatory diagram of an example of a large-scale language model used in the design support system according to this embodiment. The large-scale language model 100 can utilize, for example, GPT4. A library that can extend the functionality of the large-scale language model 100 can utilize, for example, LangChain.

[0081] The design support system 1 according to this embodiment uses a large-scale language model 100 with extended functionality through a library to execute programs such as Python, enabling the retrieval of past experimental data 110 and coordination with the prediction process of a trained machine learning model 120, as shown in Figure 13.

[0082] [Other embodiments] The formulation proposed by the design support device 10 according to this embodiment may be used, for example, to supply formulation information to a manufacturing device that synthesizes multiple materials, thereby allowing the materials to be synthesized.

[0083] As described above, the design support system 1 according to this embodiment provides a design support device, design support method, program, and design support system that reduce the workload of an operator searching for a desired mixture.

[0084] Although this embodiment has been described above, it will be understood that various modifications to the form and details are possible without departing from the spirit and scope of the claims. Although the present invention has been described above based on examples, the present invention is not limited to the above examples, and various modifications are possible within the scope described in the claims. This application claims priority to Basic Application No. 2023-155551 filed with the Japan Patent Office on September 21, 2023, the entire contents of which are incorporated herein by reference. [Explanation of Symbols]

[0085] 1. Design support system 10 Design support equipment 12 User terminals 18. Communication Networks 24 Input Reception Section 26 Prompt generation unit 28 Search Condition Generation Unit 30 First search section 34 Prediction Section 36 Second Search Section 38 Mixture display section

Claims

1. A design support device that presents information regarding the formulation desired by the operator, An input receiving unit that receives input of the first mixing conditions from the worker, A search condition acquisition unit generates or acquires search conditions for searching a database that stores past experimental data based on the first formulation conditions described above, A search unit that searches the database according to the aforementioned search conditions, A display control unit that displays the search results of the search unit on a display device, Equipped with, The search unit, if no experimental data matching the first blending conditions exists in the database, generates or obtains second blending conditions that relax the first blending conditions, and searches the database using search conditions based on the second blending conditions. The design support device is characterized in that, when no experimental data matching the first blending conditions exists, the display control unit displays experimental data matching the second blending conditions as reference point data close to the first blending conditions on the display device.

2. The design support apparatus according to claim 1, characterized in that the search condition acquisition unit generates a prompt incorporating the first formulation conditions and acquires the search conditions generated by the language model based on the prompt.

3. The design support apparatus according to claim 1 or 2, characterized in that the search unit generates the second blending conditions by relaxing at least one constraint among the first blending conditions when no experimental data matching the first blending conditions exist.

4. The design support apparatus according to claim 1 or 2, characterized in that the display control unit causes the display device to display messages indicating that no experimental data matching the first blending conditions exists and that the reference point data is close to the first blending conditions when displaying the reference point data.

5. The design support apparatus according to claim 1 or 2, characterized in that the display control unit displays the difference between the first blending condition and the second blending condition, or the content of the first blending condition that has been relaxed, on the display device along with the reference point data.

6. The design support device according to claim 1 or 2, characterized in that the input receiving unit receives input of the first blending conditions in a conversational format.

7. The design support apparatus according to claim 1 or 2, characterized in that, if the first blending conditions include non-quantitative conditions, the search condition acquisition unit generates or acquires search conditions that convert the non-quantitative conditions into searchable conditions for the database.

8. The design support apparatus according to claim 1 or 2, characterized in that the display control unit sorts the experimental data to be displayed as reference point data based on its proximity to the first formulation conditions and displays it on the display device.

9. A design assistance method performed by a computer, A process for receiving input of the first blending conditions from the worker, A step of generating or obtaining search conditions for searching a database that stores past experimental data based on the first formulation conditions, The steps include: searching the database using the aforementioned search conditions; The steps include displaying the results of the search on a display device, Includes, The search step involves, if no experimental data matching the first blending conditions exists, generating or obtaining second blending conditions that relax the first blending conditions, and searching the database using search conditions based on the second blending conditions. The aforementioned display step is a design support method characterized in that, when no experimental data matching the first blending conditions exists, experimental data matching the second blending conditions is displayed on the display device as reference point data close to the first blending conditions.

10. On the computer, A process for receiving input of the first blending conditions from the worker, A step of generating or obtaining search conditions for searching a database that stores past experimental data based on the first formulation conditions, A step of searching the database using the aforementioned search conditions, A step of displaying the results of the search on a display device, Make it run, In the search step, if no experimental data matching the first blending conditions exists, a second blending condition obtained by relaxing the first blending conditions is generated or acquired, and the database is searched using the search conditions based on the second blending condition. A program characterized in that, in the process of displaying the data, if no experimental data matching the first blending conditions exists, experimental data matching the second blending conditions is displayed on the display device as reference point data close to the first blending conditions.

Images