Design assistance device and design assistance method

Abstract

A design assistance device (1) according to the present invention is characterized by assisting a user by comprising: an input processing unit (21) that receives a design document relating to a design of a product, a creation instruction relating to the creation of a system model of the product, an ontology model, and an instance; a prompt creation unit (22) that creates a prompt for a generative AI to which the design document, the creation instruction, the ontology model, and the instance have been added; a model information extraction unit (23) that inputs the created prompt to the generative AI and receives model information relating to the creation of the system model extracted from the design document by the generative AI; and a system model creation unit (24) that creates the system model using the extracted model information.

Description

Design Support Device and Design Support Method

[0001] The present invention relates to a design support device and a design support method, and particularly relates to a technique for assisting in creating a system model related to product design.

[0002] Due to the complexity of products, systems, etc. (hereinafter referred to as products), in product design, it is often necessary to consider multiple areas such as electricity, vibration, and strength. Also, not only the performance of the product but also multiple KPIs (Key Performance Indexes) such as environmental friendliness, reliability, and safety often need to be satisfied simultaneously. Thus, as one of the techniques for complex product design, the utilization of system models has been proposed.

[0003] A system model captures various areas related to a product as an across-system, models information related to the system, and describes it structurally. By expressing product design in a system model, it is possible to eliminate the ambiguity of expression due to natural language and the ambiguity of relationships between elements, and more efficient product design can be expected. By using a system model, for example, when a design change is made, the impact of the design change on existing functions can be quickly identified.

[0004] As an advanced form of system model, MBSE (Model-Based Systems Engineering) has been proposed. MBSE models the associations between components such as the component structure, required functions, use cases, design parameters, and internal processing flow of a product, and expresses them in a graph that is visually easy to understand through the description of a modeling language, aiming to improve design efficiency and prevent design errors. An example of a modeling language is SysML (OMG (Object Management Group) Systems Modeling Language). However, to create a system model or an MBSE model from design documents, advanced expertise in system modeling is required and it often requires a huge amount of man-hours, so it is desired to create system models efficiently.

[0005] As a technology to solve these problems, Patent Document 1 describes a specification determination support device comprising: a design parameter input receiving unit configured to receive input of product design parameters; a feasibility determination unit configured to determine whether the design parameters are feasible; a system specification model generation unit configured to generate a system specification model by reflecting the design parameters in the acquired system metamodel of the product if the design parameters are feasible; and a document generation unit configured to generate a product specification based on the system specification model (see Claim 1 of Patent Document 1).

[0006] Furthermore, in order to create a system model or MBSE model from design documents, it is necessary to extract the information (knowledge) required for model creation from the design documents. As a technique for extracting information from a vast database, the use of knowledge graphs or ontology models has been proposed. An ontology model describes the concepts (classes) and their relationships (relationships) that appear in knowledge in a tree structure, and defines the inclusion relationships and constituent elements of concepts as a hierarchical structure.

[0007] For example, Patent Document 2 describes a knowledge metadata generation device that generates knowledge metadata related to content provided to a user, characterized in that it comprises a knowledge metadata generation means for generating the knowledge metadata based on an information resource related to the content and an ontology of a field related to the information resource that has been stored in advance, and a storage means for storing the knowledge metadata obtained by the knowledge metadata generation means (see Claim 1 of Patent Document 2).

[0008] Japanese Patent Publication No. 2023-83915 Japanese Patent Publication No. 2007-249907

[0009] Patent Document 1 discloses a technique for generating a system specification model from product design parameters, but does not mention the creation of system models other than specification models. Patent Document 2 discloses a technique for extracting information based on ontologs, but does not mention means for utilizing ontologs in generating AI. Furthermore, Patent Document 2 does not mention a method for creating a system model from the extracted information. The present invention has been made in view of these problems, and its object is to automatically create a system model that conforms to the user's intentions from design documents.

[0010] The design support device of the present invention is characterized by supporting the user by comprising: an input processing unit that accepts design documents relating to product design, creation instructions relating to the creation of a system model of a product, an ontology model, and an instance; a prompt creation unit that creates a prompt for a generating AI to which the design documents, creation instructions, ontology model, and instance have been added; a model information extraction unit that inputs the created prompt to the generating AI and receives model information relating to the creation of the system model extracted by the generating AI from the design documents; and a system model creation unit that creates the system model from the extracted model information. Other means will be described in the embodiments for carrying out the invention.

[0011] According to the present invention, a system model that aligns with the user's intentions can be automatically created from design documents. As a result, the man-hours required to create the system model can be significantly reduced, and various effects such as improved design efficiency and a shorter design process can be expected.

[0012] This is a diagram illustrating the configuration of the design support device. This is a diagram showing an example of a prompt template. This is a diagram showing an example of a creation know-how database. This is a diagram showing an example of an ontology model. This is a diagram showing an example of a text description of an ontology model. This is a diagram showing an example of an instance. This is an example of a flowchart for the prompt creation unit. This is a diagram showing an example of model information. This is an example of a flowchart for the system model creation unit. This is a diagram showing an example of a system model created by the generation AI. This is a diagram showing an example of visualizing the script of the system model in Figure 10. This is an example of a flowchart for Example 2.

[0013] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. While the drawings show specific embodiments in accordance with the principles of the present invention, they are for the purpose of understanding the present invention and are not intended to be used to restrict its interpretation.

[0014] <Example 1> Figure 1 is a diagram illustrating the configuration of the design support device 1. The design support device 1 is a general-purpose computer and includes a central control unit 11, input devices 12 such as a camera, microphone, mouse, and keyboard, output devices 13 such as a speaker and display, main memory 14, and auxiliary memory 15.

[0015] The input processing unit 21, prompt creation unit 22, model information extraction unit 23, and system model creation unit 24 in the main memory 14 are programs. In the following description, when the subject is indicated as "the ○○ unit," it means that the central control unit 11 reads the program from the auxiliary storage device 15 into the main memory 14 and realizes the function of the program. The auxiliary storage device 15 stores the generated AIDB 31, design document DB 32, prompt template DB 33, and creation know-how DB 34 (details below). These have the form of electronic files, relational databases, etc. "AI" stands for "Artificial Intelligence," and "DB" stands for "database."

[0016] The auxiliary storage device 15 may be configured independently as a separate enclosure. That is, the design support device 1 performs information processing using the generated AIDB 31, etc., but this information may be stored in a cloud server or the like located outside the design support device 1. In this case, the design support device 1 and the separate enclosure constitute a design support system.

[0017] Generative AIDB 31 is a collection of generative AIs that use machine learning algorithms, such as neural networks. Each generative AI outputs a prediction result by performing a given task based on input information. Generative AIs are particularly large language models (LLMs) used in the field of natural language processing, and are capable of automatically generating text or other answers to user questions. By using such generative AIs, it is possible to extract specific information from text or input files. The generative AIDB as a collection is assigned the code 31, and each individual generative AI is assigned the codes 31a, 31b, ...

[0018] The design document DB32 is a collection of design documents related to product design, such as requirements specifications, functional definitions, component information, design study documents, product defect information, CAD (Computer-Aided Design), and BOM (Bill of Materials). These design documents are used as training data or input data for the generating AI. As is clear from the specific examples of design documents, design documents are "documents" in a broad sense and include images. More generally, the generating AI can generate specific information from the design documents, which include images or text.

[0019] The prompt template DB33 is a collection of prompt templates pre-written with various instructions for generating AI to create a system model from design documents. A prompt generally means "an instruction or question that a user inputs to an interactive system." The prompts in Example 1 are also instructions or questions input to the generating AI.

[0020] As is well known, whether a generating AI can output text that aligns with the user's intent depends on the quality of the prompt. Therefore, the design support device 1 prepares multiple prompt templates according to their intended use. A template is a "template," and the user inputs specific text into the "blanks" of the prompt template. A prompt template with information added to the blanks becomes a prompt. The prompt template DB as a set is assigned the code 33, and individual prompt templates are assigned the codes 33a, 33b, ... Although we are still in the middle of Figure 1, the explanation will now move on to Figure 2.

[0021] Figure 2 shows an example of a prompt template. Prompt template 33a is a prompt template for the generating AI 31a that generates model information, and consists of four parts: "#Creation Instruction", "#Design Document", "#Ontology Model", and "#Instance". "#Creation Instruction" describes the task that the generating AI 31a will perform. "#Design Document" describes the contents of the design document as input data. The section marked "~~Insert design document information here" is the aforementioned blank space in the prompt template. Specific information will be added to this section later (the same applies below). "#Ontology Model" describes the definition of the ontology model as a conceptual connection of the information to be extracted. "#Instance" describes the instance entered by the user as a concrete example of the ontology model.

[0022] The prompt template 33b is a prompt template for the generating AI 31b that generates a system model, and consists of three parts: "#Creation Instructions", "#Model Information", and "#Creation Know-how". "#Creation Instructions" describes the tasks that the generating AI 31b will perform. "#Model Information" describes the "Model Information", which is any information necessary to create the system model. "Creation Know-how" describes the grammar related to the system model. The explanation returns to Figure 1.

[0023] As described above, the creation know-how DB34 is a collection of creation know-how that describes the grammar related to system models. Each piece of creation know-how is information at the record (row) level in the creation know-how DB. By including these grammars in the prompts for the generating AI, the generating AI can generate a system model that conforms to the grammar. In some cases, the system model generated by the generating AI31b may not be what the user desires, due to the generating AI31b not having sufficiently learned the grammar related to system models. The creation know-how grammatically describes the cause of the problem and how to fix it in such cases.

[0024] Figure 3 shows an example of the creation know-how DB34. The creation know-how DB34 is a compilation of creation know-how, which is a grammar for describing each model element, based on the language specifications of SysML, a modeling language.

[0025] The input processing unit 21 accepts user input as "creation instructions" in natural language, such as instructions, requests, and notes regarding the creation of a system model. For example, the user might input, "I want to extract safety requirements from the design documents of a digital camera." The input creation instructions are reflected when selecting a prompt template for the generation AI. To accurately execute the instructed task, the user inputs an ontology model and instances as auxiliary data. An ontology model is a tree-like description of concepts (classes) and their relationships (relationships) that appear in knowledge, defining the relationships between concepts (inclusion relationships, hierarchical relationships, causal relationships, etc.) and constituent elements as a hierarchical structure.

[0026] Figure 4 shows an example of an ontology model. This ontology model relates to the requirements diagram of the system to be designed, and graphs the connections between concepts such as "system requirements," "use cases," "blocks," "activities," "test cases," and "rationales" included in the requirements diagram. Each concept (ellipse) is associated with an instance that represents the specific content of that concept. The letters "has," etc., attached to the links between two concepts indicate the relationship between those concepts.

[0027] Figure 5 shows an example of a text description of an ontology model. When the ontology model in Figure 4 is converted into text using RDFS (RDF Schema), the schema language of the Resource Description Framework (RDF), it becomes as shown in Figure 5. RDFS provides a basic structure for defining vocabulary and its meanings for RDF data, and expresses the relationships between resources and the hierarchy of class properties. By using such a framework, the ontology model can be saved as a text file, and the user can input the ontology model text file into the system.

[0028] Figure 6 shows an example of an instance. In the context of an ontology model, an instance represents a concrete example and entity of a class (concept) in the ontology model. Instances play a role in linking abstract concepts and categories within the ontology model to concrete data and events. In a broader sense, the instances in Example 1 also include interpretations of classes (concepts), descriptions of concrete examples of classes (concepts) and their locations in design documents. In Example 1, instances are described for each of the system model concepts such as "system requirements," "use cases," "blocks," "activities," and "test cases." Instances can be entered in natural language or described as text files. One or more instances are described for a single concept.

[0029] The prompt creation unit 22 creates prompts for the generation AI to extract information related to the creation of a system model from the data received by the input processing unit 21.

[0030] Figure 7 is an example of a flowchart of the prompt creation unit 22. In Figure 7, solid arrows indicate the flow of processing, and dashed lines indicate the flow of information (the same applies to Figures 9 and 12). As a prerequisite for the flowchart in Figure 7 to start, the input processing unit 21 is assumed to be accepting the following inputs from the user via the input device 12. Note that the user here may be one person or multiple people. - Instructions created by the user regarding the creation of a system model of the product or information extraction - Design documents created by the user regarding the design of the product - Ontology models created by the user - Instances created by the user

[0031] The prompt creation unit 22 performs the following steps. In step S201, the prompt creation unit 22 determines an appropriate prompt template based on the creation instruction and reads the appropriate prompt template from the prompt template DB 33. For the sake of explanation, let's assume here that the prompt creation unit 22 determines and reads prompt template 33a in Figure 2 as the appropriate prompt template. Then, the prompt creation unit 22 adds the creation instruction that the user had previously entered to the blank "#Creation Instruction" field of prompt template 33a.

[0032] In step S202, the prompt creation unit 22 adds the contents of the design document, which is the information source for creating the system model, to the prompt. Specifically, the prompt creation unit 102 extracts design information such as text from the design document previously entered by the user and adds it to the blank space in “#Design Document” of the prompt template 33a, or adds a link to the design document file to the said blank space.

[0033] In step S203, the prompt creation unit 22 adds the ontology model to the prompt. Specifically, the prompt creation unit 102 adds the text description of the ontology model, as shown in Figure 5, to the blank space "#Ontology Model" in the prompt template 33a.

[0034] In step S204, the prompt creation unit 22 adds the instance to the prompt. Specifically, the prompt creation unit 22 adds the instance text description, as shown in Figure 6, to the blank "#instance" field of the prompt template 33a.

[0035] Steps S201 to S204 create a prompt for the generated AI 31a. As mentioned above, the prompt is input information (search term) for the generated AI, and for example, it takes the form of a question.

[0036] The model information extraction unit 23 inputs the prompt created by the prompt creation unit 22 to the generation AI 31a and executes it, thereby extracting information related to the creation of the system model (hereinafter also called "model information") as the output of the generation AI 31a. The extracted model information is saved in the auxiliary storage device 15 in the form of a file or database. After that, the flowchart in Figure 7 ends.

[0037] Figure 8 shows an example of model information. The model information extraction unit 23 receives the results of the generation AI extracting each model element, including use cases, blocks, test cases, activities, requirements, etc., from the design document. The model information extraction unit 23 assigns an ID (identifier) ​​to each model element, extracts the name (title), text (specific content), dependents, and dependencies, organizes them in a table format (Figure 8), and outputs them as model information.

[0038] The system model creation unit 24 creates a system model from the model information extracted by the model information extraction unit 23.

[0039] Figure 9 is an example of a flowchart of the system model creation unit 24. Following the flowchart in Figure 7, the system model creation unit 104 performs the following steps.

[0040] In step S401, based on the creation instruction, the system model creation unit 24 determines an appropriate prompt template and reads the appropriate prompt template from the prompt template DB 33. For the convenience of explanation, here, it is assumed that the system model creation unit 24 determines and reads the prompt template 33b in FIG. 2 as the appropriate prompt template. Then, the system model creation unit 24 adds the creation instruction that the user has input in advance to the blank of "# creation instruction" in the prompt template 33b.

[0041] In step S402, the system model creation unit 24 adds the model information extracted by the model information extraction unit 23 to the blank of "# model information" in the prompt template 33b.

[0042] In step S403, the system model creation unit 24 extracts the creation know-how indicated by the user from the creation know-how DB 34 and adds it to the blank of "# creation know-how" in the prompt template 33b.

[0043] From step S401 to S403, a prompt for the generation AI 31b for creating a system model is generated, and the system model creation unit 24 inputs this to the generation AI 31b, so that the system model instructed by the user is output from the generation AI 31b. Although it is in the middle of FIG. 9, the explanation will once move to FIG. 10.

[0044] FIG. 10 is a diagram showing an example of a system model created by the generation AI 31b. FIG. 10 is a requirement diagram of the target system described in a script conforming to the SysML specification. The explanation returns to FIG. 9.

[0045] In step S404, the system model creation unit 24 determines whether the output system model is what the user desires as a result of the user verifying the output system model. Specifically, the user directly decodes the script of the output system model to verify the error. Or, the system model creation unit 24 visualizes the system model, and the user verifies the visualized model. Although it is in the middle of FIG. 9, the explanation will once move to FIG. 11.

[0046] FIG. 11 is a diagram showing an example in which the script of the system model of FIG. 10 is visualized. Visualization has the merit of improving the visibility of the system model and making it easier to verify. In the upper center of FIG. 11, there is a model element (rectangle) having "ID = 1". The model element corresponds to the record in the 5th row of the model information in FIG. 8. And "Text = ******" of the model element in FIG. 11 is the same as "******" stored in the text column of the record in the 5th row of the model information in FIG. 8. Further, as shown by the record in the 1st row of the instance in FIG. 6, the "Text = ******" is described in "1.1 Overview" and "1.2 Term Definition" of the design document. The explanation returns to FIG. 9.

[0047] As a result of the above determination, if there is no problem, the system model creation unit 24 ends the operation. If there is a problem, that is, if the system model generated by the generation AI 31b is not what the user desires, the system model creation unit 24 proceeds to step S405. In step S405, the system model creation unit 24 identifies the cause of the problem and modifies a part of the creation know-how. Then, the system model creation unit 24 reflects the modified creation know-how in the creation know-how DB 34 and updates the creation know-how DB 34. After that, the flowchart in FIG. 9 ends.

[0048] <Example 2> In Example 2, when the user inputs a creation instruction, the user directly instructs the creation of a system model. As a result, the design support device 1 directly creates a system model from the design document without going through the step of extracting model information. That is, in Example 2, the prompt creation unit 22 and the system model creation unit 24 generate a prompt for creating a system model, and the system model creation unit 24 inputs it to the generation AI to create a system model. In Example 2, the model information extraction unit 23 becomes unnecessary.

[0049] Figure 12 is an example of a flowchart for Embodiment 2. Embodiment 2 consists of the following steps. In step S501, the prompt generation unit 22 determines an appropriate prompt template based on the user's instructions for creating a system model, and reads the appropriate prompt template from the prompt template DB 33. For the sake of explanation, it is assumed here that the prompt creation unit 22 determines and reads prompt template 33c (not shown) as the appropriate prompt template.

[0050] Prompt template 33c is a combination of prompt templates 33a and 33b in Figure 2, and has blank spaces for "#Creation Instructions", "#Design Document", "#Ontology Model", "#Instance", and "#Creation Know-how". The prompt creation unit 22 then adds the creation instructions previously entered by the user to the blank space for "#Creation Instructions" in prompt template 33c.

[0051] In step S502, the prompt creation unit 22 adds the contents of the design document, which is the information source for creating the system model, to the prompt. Specifically, the prompt creation unit 22 extracts design information such as text from the design document previously entered by the user and adds it to the blank space in “#Design Document” of the prompt template 33c, or adds a link to the design document file to the said blank space.

[0052] In step S503, the prompt creation unit 22 adds the ontology model to the prompt. Specifically, the prompt creation unit 22 adds the text description of the ontology model, as shown in Figure 5, to the blank space "#Ontology Model" in the prompt template 33c.

[0053] In step S504, the prompt creation unit 22 adds the instance to the prompt. Specifically, the prompt creation unit 22 adds the instance text description, as shown in Figure 6, to the blank "#instance" field of the prompt template 33c.

[0054] In step S505, the system model creation unit 24 extracts the creation know-how instructed by the user from the creation know-how DB 34 and adds it to the blank space "#creation know-how" in the prompt template 33c.

[0055] Steps S501 to S505 create a prompt for the generating AI 31c to create the system model. The system model creation unit 24 inputs this prompt into the generating AI 31c, and the user-instructed system model is output. The generating AI 31c is a combination of generating AIs 31a and 31b, and generates the system model from the information input (added) to the prompt template 33c.

[0056] In step S506, the system model creation unit 24 determines whether the system model is what the user wants based on the user's verification of the output system model. Specifically, the system model creation unit 24 directly deciphers the script of the output system model to verify for errors. Alternatively, the system model creation unit 24 visualizes the system model, and the user verifies the visualized model. If there are no problems as a result of the above determination, the system model creation unit 24 terminates its operation. If there are problems, that is, if the system model generated by the generating AI 31c is not what the user wants, the system model creation unit 24 proceeds to step S507.

[0057] In step S507, the system model creation unit 24 identifies the cause of the problem and modifies part of the creation know-how. The system model creation unit 24 reflects the modified creation know-how in the creation know-how DB 34 and updates the creation know-how DB 34. After that, the flowchart in Figure 12 is completed.

[0058] Example 2 is also an example in which the generated AI 31c also plays the roles of generated AIs 31a and 31b in Example 1. Thus, the generated AIDB 31 may store only a single general-purpose generated AI.

[0059] While several embodiments of the present invention have been described, these embodiments are presented as examples only and do not limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents.

[0060] 1 Design support device 11 Central control unit 12 Input device 13 Output device 14 Main memory 15 Auxiliary memory 21 Input processing unit 22 Prompt creation unit 23 Model information extraction unit 24 System model creation unit 31 Generated AI database 32 Design document database 33 Prompt template database 34 Creation know-how database

Claims

1. A design support device characterized by supporting a user by comprising: an input processing unit that accepts design documents relating to product design, creation instructions relating to the creation of a system model of a product, an ontology model, and an instance; a prompt creation unit that creates a prompt for a generating AI to which the design documents, creation instructions, ontology model, and instance have been added; a model information extraction unit that inputs the created prompt to the generating AI and receives model information relating to the creation of the system model extracted by the generating AI from the design documents; and a system model creation unit that creates the system model from the extracted model information.

2. The design support device according to claim 1, comprising: a generation AI database storing a plurality of generation AIs that generate specific information from a design document containing images or text; a design document database storing information relating to the design of a product, including requirements specifications, function definitions, component information, design study documents, product defect information, CAD, or BOM, as the design document; a prompt template database storing a plurality of prompt templates, which are templates of prompts input to the generation AI in order to create a system model using the generation AI; and a creation know-how database storing creation know-how for creating the system model.

3. The design support device according to claim 1, characterized in that the ontology model is described by the user in graph or text as elements to be extracted from the design document and the connections between those elements, and added to prompts for the generating AI.

4. The design support device according to claim 1, wherein the instance describes a specific example of an element defined by the ontology model or a location in the design document, and is added to a prompt for the generating AI.

5. The design support device according to any one of claims 2 to 4, characterized in that the prompt creation unit selects an appropriate prompt template based on the creation instructions from the user and adds the ontology model and instance entered by the user to the prompt template.

6. The design support device according to claim 2, characterized in that the system model creation unit creates a prompt for creating a system model by adding the extracted model information and the creation know-how extracted from the creation know-how database to the prompt template, and creates the system model instructed by the user by inputting the prompt to the generating AI.

7. The design support device according to claim 6, characterized in that the system model creation unit verifies whether the created system model is as desired by the user, and if the verification shows that the created system model is not as desired by the user, it modifies the creation know-how and updates the creation know-how database.

8. A design support device that supports a user by comprising: means for receiving design documents relating to product design, creation instructions relating to the creation of a system model of a product, an ontology model and an instance; means for creating a prompt for a generating AI to which the design documents, creation instructions, ontology model and instance have been added; and means for inputting the created prompt into a generating AI model and receiving the system model extracted by the generating AI from the design documents.

9. A design support method characterized by: an input processing unit of a design support device receiving design documents relating to the design of a product, creation instructions relating to the creation of a system model of a product, an ontology model, and an instance; a prompt creation unit of the design support device creating a prompt for a generating AI that adds the design documents, the creation instructions, the ontology model, and the instance; a model information extraction unit of the design support device inputting the created prompt to the generating AI and receiving model information relating to the creation of the system model extracted by the generating AI from the design documents; and a system model creation unit of the design support device creating the system model from the extracted model information.

Images