Design support system, design support program, and storage medium

Abstract

To provide a design support system that assists even inexperienced design engineers in appropriately selecting new materials and components suitable for the purpose of product design, and that can also present the annual cost-effectiveness of design changes using new materials and components. [Solution] The design support system 1 is equipped with a database 120, a machine learning means 110, an input means 310, and a display means 320. The database stores design information for multiple materials and parts that make up a product in advance. The machine learning means generates a trained model 130 that extracts design information with a high degree of fit to the design target information, using the multiple design information as training data 112. The trained model then calculates the functional fit and cost fit for existing materials and parts and new design candidates individually, based on the design target values ​​input by the design engineer. Furthermore, the display means displays the respective fit scores side by side so that the design engineer can easily understand the comparison results.

Description

Technical Field

[0001] The present invention relates to a design support system that assists even non-expert design engineers in appropriately selecting new materials and components suitable for product design purposes. Furthermore, it relates to a design support system that can also present the annual cost effect when changing to the selected materials and components. It also relates to a design support program that enables a computer to function as a design support system, and a storage medium storing the design support program.

[0002] More specifically, it relates to a design support system that compares design target information obtained by input means with function information and cost information related to individual materials and components registered in a database. Specifically, it uses a learned model (AI) trained to individually derive function compatibility and cost compatibility to extract candidates for materials and components with high compatibility with the design target information. Furthermore, it relates to a design support system that contrastively displays the extracted candidates for materials and components on a display means so that they are easily intuitively understood by design engineers.

Background Art

[0003] Conventionally, as a method for organizing the relationships between design items of a product, methods such as the DSM method (Design Structure Matrix) are well-known. The DSM method visualizes the flow of information in a matrix representation and is widely used in semiconductor development, automobile development, etc. as a method for designing an efficient process from the perspective of overall optimization. In the DSM method, a matrix with design items as elements is created, and by inputting checks or numerical values into each cell, the presence or absence and strength of interdependencies between the elements can be readily discerned.

[0004] However, even if all the materials and components constituting the product are registered as elements of the matrix, from the interdependency relationships of the elements, one can only grasp the scope of influence when redesigning the materials and components, and it was not possible to easily select candidates for materials that meet the new design criteria or evaluate the cost-effectiveness.

[0005] Therefore, in conventional redesign processes, design engineers had no choice but to search for materials and components that met the design objectives based on their own experience using keywords from a database containing specification information. However, it is time-consuming and laborious for a person to select new materials that are highly suitable in terms of both function and cost. Moreover, there are no unified standards for evaluating the suitability of materials and components, and it is left to the experience of individual design engineers, so the materials and components that each design engineer deems optimal may differ.

[0006] Therefore, it was necessary to share information within the design group and double-check the selection of materials by the assigned engineers to ensure there were no errors or omissions, which hindered rapid product development and design. Furthermore, even if the selected materials had appropriate functionality, the lack of cost information made it difficult to quickly evaluate costs, requiring the time-consuming process of referring to past purchase orders or contacting suppliers.

[0007] Consequently, even if there were no problems with functionality, if poor cost-effectiveness was discovered during the cost evaluation stage, all the previous functional considerations and evaluations could potentially be wasted. Therefore, there was a need for a technology that would allow even inexperienced design engineers to quickly select appropriate new materials and components from both functional and cost perspectives, and that would also reduce the effort required for double-checking.

[0008] Patent Document 1 discloses a system technology for determining the optimal material for a product using a trained model. According to the technology described in this document, material suitability is determined in two stages, at the component level and the overall product level, based on constraints related to individual components and constraints related to the entire product. In addition, the system takes into account the specifications provided by the user and determines the optimal material from a set of registered material information.

[0009] The criteria include not only the material compatibility of individual components but also the material compatibility of the entire product, allowing for the selection of the optimal material, taking into account its interaction with other components. Specifically, this might involve limiting the use of aluminum if replacing steel with aluminum could lead to galvanic corrosion between adjacent steel components. Economic constraints may also be included.

[0010] However, the technology described in Patent Document 1 assigns a multi-level score to the degree of suitability for each constraint (structural resistance, corrosion resistance, tensile strength, etc.). Specifically, an example of assigning a score on a 10-point scale from 0.1 to 1.0 is illustrated. Even when this scoring is applied to cost constraints, it only provides an indication of whether or not the new material meets the cost constraints, and it is not possible to specifically estimate the annual cost-effectiveness. Therefore, accurately understanding the cost significance of design changes is time-consuming and requires considerable effort.

[0011] Patent Document 2 discloses a design support device technology that assists in optimizing product design using a trained model. According to the technology described in this document, multiple product performances (dependent variables) are predicted from a set of design factors including multiple design factors (explanatory variables) of the product, and candidate product design information is generated based on a comprehensive evaluation of the predicted performances. Furthermore, the degree of fit of the comprehensive evaluation may be calculated, and cost may be included as an indicator of performance.

[0012] However, in the technology described in Patent Document 2, the design factors (explanatory variables) in the example of a storage battery are "material composition ratio, electrode shape," etc., and the performance (dependent variable) is "capacity, output, cost," etc. In other words, "cost information" is not distinguished from functional information such as capacity, and the degree of performance suitability is calculated with cost and function mixed together. Since functional information has more items than cost information, if the degree of suitability is calculated with these mixed together, high-performance materials and components tend to be ranked as top candidates.

[0013] In other words, materials and components that are inexpensive but do not offer a high degree of functional suitability become lower-ranking candidates and are easily overlooked. Consequently, this approach is not suitable for redesign work that seeks materials and components that meet the best functional requirements within a limited budget, and its practical application presents challenges. Furthermore, because it is not possible to concretely estimate the annual cost-effectiveness, similar to Patent Document 1, accurately understanding the cost benefits of design changes is time-consuming and requires considerable effort. [Prior art documents] [Patent Documents]

[0014] Patent Document 1: Japanese Unexamined Patent Publication No. 2024-149396 Patent Document 2: Japanese Unexamined Patent Publication No. 2024-134116 [Overview of the Initiative] [Problems that the invention aims to solve]

[0015] The problem that this invention aims to solve is to provide a design support system and design support program that enables the rapid development of new materials and components, and that allows for the rapid and accurate estimation of annual cost-effectiveness, by standardizing the selection accuracy, which varies among design engineers, when selecting new materials and components that have a high degree of suitability to the design target values ​​of new materials and components. [Means for solving the problem]

[0016] The first invention of the present invention is a design support system for assisting the design of a product, comprising a storage means, a processing means, an input means, and a display means constituting a computer, wherein the storage means stores a plurality of design information as a database, relating functional information representing the functional characteristics of a plurality of materials and parts constituting the product, cost information of the plurality of materials and parts, and weight coefficients indicating the priority between items of functional characteristics, the processing means functions as a machine learning means for executing machine learning processing, a specific design information extraction means, a design target information creation means, and a suitability guidance means, wherein the processing means as a machine learning means generates a trained model that uses the plurality of design information as training data and infers design information with a high degree of suitability to the acquired design target information, the specific design information extraction means extracts first design information from the database that matches the information of existing materials and existing parts input from the input means, the design target information creation means creates the design target information based on the input functional target information and actual cost data, and the processing means executes inference processing based on the trained model, thereby enabling suitability The system functions as a target extraction means and a condition fit derivation means. The target extraction means acquires the design target information and causes the trained model to infer second design information with a high degree of fit to the acquired design target information. The condition fit derivation means compares the design target information with the first design information to individually derive the functional fit and cost fit, and also compares the design target information with the second design information to individually derive the functional fit and cost fit. For the functional fit, the system calculates the functional fit for each item of the functional characteristics from the error rate with the design target information. The system calculates the functional fit and obtains the weight coefficient from the database. The system then multiplies the functional fit of each item by the weight coefficient and adds the weighted functional fit of each item to derive a sum of functional fit values. The system is characterized in that the conformance target guidance means displays the sum of functional fit values ​​and cost fit values ​​derived by comparing the design target information and the first design information, and the sum of functional fit values ​​and cost fit values ​​derived by comparing the design target information and the second design information, as evaluation indicators used for design decisions, in a manner that can be compared with the display means.

[0017] This invention may be operated by a single company or collaboratively by a group of companies. The database may be built on a general-purpose computer, a local server computer, or storage provided by a cloud computing service. The machine learning means may consist of a central processing unit such as a general-purpose computer or a server computer and storage means. The input means may be a general-purpose input device such as a mouse, keyboard, or touch panel. The display means may be a monitor, touch panel, etc.

[0018] The trained model generated by the processing means that functions as a machine learning tool can be either a linear regression model or a nonlinear regression model, and is not limited to either. For example, if it is a nonlinear regression model, it can be any known support vector machine, neural network, etc. If it is a linear regression model, it can be any known regression analysis method such as multiple regression analysis, ridge regression, lasso regression, etc.

[0019] During the training phase of the pre-trained model, functional information and cost information related to materials and parts are used as input training data, and corresponding output training data is used to train the pre-trained model using the aforementioned publicly known model, which provides specific material and part design information with a high degree of fit. The functional information constituting the design information is functional numerical data based on design specifications such as heat resistance, corrosion resistance, strength, and dimensions, linked to the material name and part name. The cost information constituting the design information is numerical data related to commercial transactions, such as the weight unit price of materials and the unit price of parts. Note that annual production quantities differ for each product, so they are not registered in the database but can be entered as one of the design target information. As a result, the pre-trained model of the present invention is generated as a model capable of inferring optimal material and part design information for the design target based on the input information.

[0020] The training data for functional information consists of a group of functional characteristic values. Functional characteristic values ​​should consist of a type of characteristic value and a unit, such as "heat resistance (°C)" or "strength (MPa)". In addition, functional characteristic values ​​may include evaluation values ​​without units, such as UV resistance, chemical resistance, ease of processing, and supplyability (ease of procurement). The evaluation values ​​may be multi-level numerical values, such as "2" in a 5-point scale.

[0021] The training data for the predetermined functional fit may be, for example, an acceptable value obtained by multiplying the design target information by a predetermined coefficient value, or it may be a deviation score relative to the design target value, but it is not limited to these. For example, if an acceptable coefficient value is set, it may be sufficient to determine that the predetermined functional fit is exceeded when at least 90% of the design target value is met. The training data also includes weight coefficients that indicate the priority between items of functional characteristics. The training data for condition fit may be modified as needed depending on the progress of machine learning, etc.

[0022] The training data for cost information includes at least product quoted prices such as "material weight unit price" and "part unit price." For a more accurate understanding of annual cost-effectiveness, the training data may also include "annual production quantity," "processing costs," and "procurement costs." The training data for the degree of suitability of cost information should be specific costs based on past development practices. Processing costs are not limited to those related to the processing of materials and parts, but "surface treatments" such as plating and "heat treatments" that alter material strength through heating are preferred.

[0023] Here, each component of the processing means will be described in detail. Note that the processing means performs inference processing based on the learned model only when executing the processing of the conformity target extraction means and the condition conformity degree derivation means described later. The specific design information extraction means searches the database based on the information of the existing materials / parts input by the design engineer using the input means, and extracts the first design information that matches the input information. The information of the existing materials / parts is information that can identify the materials / parts. For example, when the existing material is SUS304, which is stainless steel, only the name / model number information that can identify "SUS304" is sufficient. The first design information (existing materials / parts) is the function information such as heat resistance / strength and other function information stored in association with "SUS304" in the database, and the cost information such as the unit price of the material.

[0024] The design target information creation means creates design target information from the function target information input by the user and the performance data related to cost (hereinafter simply referred to as performance data). The function target information is, similar to the teacher data related to function information, a combination of a function name and a numerical value, etc. The performance data may be at least "material weight unit price (yen / kg)" in the case of materials, and at least "component unit price (yen / unit)" in the case of components. In either case, "processing cost" and "annual production quantity" may be included.

[0025] The suitability target extraction means identifies second design information (candidates for new materials and parts) with high suitability for the design target information from the database based on the learned content of the learned model. The second design information may be two or more candidates. The condition suitability derivation means derives the condition suitability of the existing materials and existing parts for the design target information as a specific numerical value based on the learned content of the learned model. Further, the condition suitability of the second design information is also calculated as a specific numerical value so as to be a comparison target with the existing materials, etc. At this time, for the suitability regarding functional characteristics, a total value that is a score obtained by multiplying the error rate of each function information by a weight coefficient (function suitability = Σ (error rate of each function × weight coefficient)) is calculated, and a numerical value converted to a percentage may be displayed. In this case, the error rate may be a numerical value calculated as a percentage or the like of the difference from the function target value, and the weight coefficient may be set to reflect the importance of the function.

[0026] The cost suitability may be in a form that allows the design engineer to grasp the cost variation associated with the review of materials and parts, and is not limited to ratios, error rates, etc. For example, it may be the annual cost itself such as "10 million yen per year", or the annual difference amount such as "(1 million yen reduction per year compared to the conventional part)". Since the cost suitability is calculated separately from the function suitability, it is easy to compare and examine only the cost superiority when there are multiple suitable materials and parts that meet the function target value.

[0027] Then, the suitability target guidance means causes the display means to display the function suitability and cost suitability derived individually, together with the respective design information, for the existing material / part product and the new design candidate in a comparable state. The comparable state may be, for example, a state where the existing materials / parts and the new design candidate are arranged side by side or above and below on the display means for comparison. Note that the condition suitability may be converted into a radar chart, bar graph, etc. and displayed so as to be easily visually judged.

[0028] According to the first invention, by using inference processing based on a trained model, candidates for new designs with a high degree of suitability can be extracted from a database, and the significance of design changes from existing materials and parts can be examined by individually comparing functional and cost aspects. This enables objective evaluation based on accumulated past data, without depending on the skill level of the design engineer, and has advantageous effects not found in conventional technology, such as homogenization of design quality, improvement of reliability, and acceleration.

[0029] The second invention of the present invention is a design support system of the first invention, wherein the input means comprises an image information input means capable of inputting image information, and the image information includes at least digital data of a design drawing containing the design information, or a paper medium; the specific design information extraction means performs optical character recognition processing or feature extraction processing on the character information contained in the input image information to identify character information relating to existing materials and existing parts, and extracts first design information from the database; the design target information creation means creates provisional design target information based on the identified first design information, and creates the design target information when further difference information is input from the input means.

[0030] The image information input means may read design drawings printed on paper using an optical reading device (scanner, OCR, etc.), or it may read digital data such as PDFs or JPEGs stored on a storage medium such as an HDD or USB memory. According to the second invention, the first design information is identified from the image information. Furthermore, since the specific design information extraction means automatically inputs the first design information as provisional design target information, the design engineer only needs to input the difference information.

[0031] Difference information may include not only functional and cost information related to design changes, but also information not registered in the database, such as the planned annual production quantity. This reduces the effort required for design engineers to input data, even when reviewing a large number of materials and parts simultaneously, and offers advantages not found in conventional technologies, such as reducing human error and data mix-ups.

[0032] A third aspect of the present invention is a design support system of the first invention, characterized in that the database registers at least the product price and processing costs as information constituting the cost information, the processing costs include at least one of surface treatment costs or heat treatment costs, the design target information creation means creates the design target information when the product price, processing costs and annual production quantity are input as actual data from the input means, and the condition suitability derivation means derives the cost suitability based on the actual data.

[0033] According to the third invention, the database includes not only the product price, which is the unit price of materials by weight and the unit price of parts, but also at least one of the processing costs, namely surface treatment costs and heat treatment costs. Here, surface treatment refers to processes such as plating and painting. Furthermore, in addition to the product price and processing costs, the annual production quantity is a required input item as actual cost data, and the annual improvement cost is created from these figures. This allows for quick and reliable cost analysis based on accurate annual cost-effectiveness, and provides advantageous effects not found in conventional technology, such as being able to grasp the cost reduction effect of design changes in concrete amounts.

[0034] The fourth invention of the present invention is a design support system of the first to third inventions, wherein the design support system comprises a server computer and a plurality of operating terminals, the server computer functions as the processing means and the database, the operating terminals are provided with the input means and the display means, the storage area of ​​the database is partitioned according to the divisions of a plurality of business units to which each of the operating terminals belongs, the design information is stored independently for each division, the design target information creation means creates the design target information from the functional target information and the actual data, as well as the division designation information input from the input means, and the suitability target extraction means infers second design information corresponding to the design information in the storage area assigned to the designated division.

[0035] The server computer may be an on-premises server or a cloud server. The application of the trained model may be stored in the server computer's storage means, similar to a database, or it may be stored on individual user terminals. When the application is stored on the server computer, it is preferable because if the high-performance processing means constituting the server computer functions as the trained model, the user terminals only need to utilize the calculation results of the trained model, and a general-purpose computer can be used.

[0036] In companies that separate development departments by group company, production base, and product type, even for the same materials and parts, the availability (ease of procurement) and cost differ from one business unit to another due to factors such as inventory status, transportation costs, and customs duties. According to the fourth invention, by making the database storage area independent for each business unit, cost conditions can be managed separately from those of other business units.

[0037] Furthermore, by specifying the business unit to which no design engineers belong, it is possible to consider optimal materials and components across business unit boundaries. This has the advantageous effect of allowing for the deriving of a more accurate degree of suitability, taking into account circumstances where it is not appropriate to manage products using unified indicators such as supplyability and cost, which differ from business unit to business unit.

[0038] The fifth invention of the present invention is a design support system of the fourth invention, wherein the design support system is applied to a group of companies consisting of manufacturers and users of the product, the server computer comprises communication means, user information management means, and design information registration management means, the processing means functions as the user information management means and the registration management means, the user information management means manages access rights for each manufacturer and each user, the database stores the classification further subdivided by access rights, and when an operating terminal belonging to either the manufacturer or the user accesses the server computer via the communication means, the registration management means allows editing of the design information stored in the database within the scope corresponding to the access rights, and also allows selection of whether to make each piece of design information public or private.

[0039] According to the fifth invention, the system is applicable to a group of companies consisting of product manufacturers and users. Specifically, a database built on the storage area of ​​a server computer is jointly used by the manufacturers and users in sections that correspond to their respective access rights. These sections are further subdivided by business unit.

[0040] Therefore, manufacturers can register their products in the database within the scope of their access rights, and can choose to make registered products public or private. This makes it easier to promote the sale of new products that are not yet widely known. In addition, price changes for existing products can be quickly reflected in the database and made public, which is expected to reduce the effort required for users to request quotes.

[0041] Furthermore, since specific products can be kept private, information on discontinued products, for example, can be made accessible only to the user as part of the history, thereby reducing unnecessary inquiries. In addition, because the registration management system allows editing of design information only within the scope of access rights, data tampering can also be prevented.

[0042] Users can access specifications and cost information for new products from other companies without having to register them in their own internal database. As a result, materials and components that were previously unknown to the user can be considered as candidates for new designs. Furthermore, for materials with large price fluctuations, such as precious metals, it becomes easier to grasp accurate annual cost-effectiveness based on the latest transaction prices without having to request quotes, thus shortening the evaluation time.

[0043] This offers several advantages not found in conventional technologies: it allows manufacturers to boost sales of new products, enables users to consider products from other companies that their design engineers may not be familiar with, and promotes collaboration among multiple companies. Furthermore, it reduces the effort required for design engineers to keep the design information registered in the database up-to-date.

[0044] The sixth invention of the present invention is a design support program that is executed on a computer and assists in the design of a product, wherein the computer is configured to include a processing means, a storage means, an input means, and a display means, and the design support program causes the processing means to function as a machine learning means for executing machine learning processing, a specific design information extraction means, a design target information creation means, and a suitability guidance means, and causes the storage means to function as a database, the database storing a plurality of design information relating functional information representing the functional characteristics of a plurality of materials and parts constituting the product, cost information of the plurality of materials and parts, and weight coefficients indicating the priority between items of functional characteristics, as a first step, the machine learning means generates a trained model that uses the plurality of design information as training data and infers the design information with a high degree of fit to the acquired design target information, as a second step, the specific design information extraction means extracts first design information from the database that corresponds to the design information of existing materials and parts entered by the user, as a third step, the design target information creation means extracts functional target information and cost information entered by the user From actual usage data, design target information is created. In the fourth step, the processing means functions as a means for extracting suitable targets by performing inference processing based on the trained model, and acquires the design target information. The trained model then infers second design information with a high degree of suitability to the acquired design target information. In the fifth step, the processing means functions as a means for deriving conditional suitability by performing inference processing based on the trained model, and compares the design target information with the first design information to individually derive functional suitability and cost suitability. The functional fit and cost fit are individually derived by comparing the target information and the second design information. For the functional fit, the functional fit for each item is calculated from the error rate with the target information for each functional characteristic item. Furthermore, the weight coefficient is obtained from the database, and the functional fit for each item is multiplied by the weight coefficient. The weighted functional fit for each item is then added to derive the sum of the functional fits. In the sixth step, the conformance target guidance means compares the target information and the first design information to determine the sum of the functional fits and the cost fit.The system is characterized by having the computer process data so that the sum of the functional fit scores and the cost fit scores, derived by comparing the aforementioned design target information with the second design information, are displayed on the display means in a comparable manner as evaluation indicators used for design decisions.

[0045] According to the sixth invention, the same advantageous effects as the first invention can be obtained. Furthermore, because it can be provided as a program, it is easy to implement on general-purpose computers and the like that design engineers (users) use on a daily basis, making it highly versatile.

[0046] The seventh aspect of the present invention is a computer-readable recording medium that records the design support program of the sixth aspect for execution by a computer. According to the seventh aspect, the design support program can be provided as a computer-readable recording medium. This makes it possible to use the design support program of the sixth aspect via a physically isolated storage medium, even on a computer isolated from the internet to prevent unauthorized access.

[0047] The eighth invention of the present invention is a design support system for assisting the design of a product, comprising a database, machine learning means, input means, and display means, wherein the database stores a plurality of design information relating functional information representing the functional characteristics of a plurality of materials and parts constituting the product and cost information of the plurality of materials and parts, the functional information includes ease of processing as essential information, the machine learning means generates a trained model that uses the plurality of design information as training data and extracts the design information with a high degree of fit to the design target information, and the trained model comprises a design target information creation means, a fitting target extraction means, a condition fit degree derivation means, and The system functions as a target selection guide means, and is characterized in that the design target information creation means creates design target information based on input functional target information and actual cost data, the target selection means extracts at least one second design information from the database in which the ease of processing exceeds a predetermined functional fit for the design target information, and the condition fit derivation means individually derives the functional fit and cost fit from the design target information and the second design information, and displays the second design information for multiple candidates, along with their corresponding functional fit and cost fit, in descending order of fit, in a manner that can be compared with the display means.

[0048] The eighth invention is configured to extract second design information using only design target information, without inputting first design information related to existing materials and parts. Furthermore, it requires the functional characteristic of ease of processing as functional information. Ease of processing can be any numerical representation of the ease of bending, cutting, etc., based on mechanical properties such as strength and softening temperature for materials, in a multi-stage manner.

[0049] For example, using a 5-point scale, resin materials could be assigned a "5," general structural rolled steel a "3," and stainless steel a "1," with the evaluation values ​​set so that materials and parts that are easier to process receive higher ratings for suitability. Then, when extracting the second design information from the database, only those whose processability exceeds a predetermined suitability score are filtered out. The predetermined suitability score can be set based on past design results, and there are no specific numerical limitations. For example, a predetermined suitability score (acceptable range) of 90% or more of the design target value could be set.

[0050] This allows for objective evaluation based on accumulated historical data, even when designing from scratch, without relying on the skill level of the design engineer. This enables homogenization of design quality, improvement of reliability, and acceleration of the design process. Furthermore, because ease of processing is a crucial condition for selection, processing costs that cannot be estimated solely from the procurement costs of materials and parts can be anticipated at the design stage, resulting in a design support system that is tailored to the realities of design and development. [Effects of the Invention]

[0051] According to the first aspect of the present invention, objective evaluation based on accumulated past data becomes possible without depending on the skill level of the design engineer, resulting in advantageous effects not found in conventional technology, such as homogenization of design quality, improvement of reliability, and acceleration of the process. According to the second invention of this invention, even when reviewing a large number of materials and parts simultaneously, it is possible to reduce the effort required for design engineers to input data, thereby reducing human error and data misinterpretation, which is an advantageous effect not found in conventional technology. According to the third invention of this invention, it is possible to quickly and reliably conduct cost studies based on accurate annual cost-effectiveness, and to grasp the cost reduction effect of design changes in concrete monetary terms, which is an advantageous effect not found in conventional technology.

[0052] According to the fourth invention of this invention, it is possible to derive a more accurate degree of suitability for conditions, taking into account circumstances where it is not suitable to manage each business unit using unified indicators such as ease of procurement and cost. According to the fifth invention of this invention, manufacturers can promote the sale of new products, and users can consider products from other companies that design engineers are not familiar with, thereby promoting collaboration among multiple companies—advantages not found in conventional technology. Furthermore, design engineers can reduce the effort required to keep the design information registered in the database up to date. According to the sixth invention of this invention, the same effects as the first invention can be obtained, and it is also easy to introduce into general-purpose computers and the like that design engineers (users) use on a daily basis, making it highly versatile. According to the seventh invention of this invention, in order to prevent unauthorized access, the design support program of the sixth invention can be used even on a computer isolated from the internet, via a physically isolated storage medium. According to the eighth aspect of the present invention, even when designing from scratch, objective evaluation based on accumulated past data becomes possible without depending on the skill level of the design engineer, thereby enabling homogenization of design quality, improvement of reliability, and acceleration of the process. [Brief explanation of the drawing]

[0053] [Figure 1] Block diagram of the design support system (Example 1). [Figure 2] A concrete example of a user interface for a design support system (Example 1). [Figure 3] Configuration diagram of the table related to the first design information (Example 1). [Figure 4] Configuration diagram of the table related to the second design information (Example 1). [Figure 5] A concrete example of a design information table (Example 1). [Figure 6] A specific example of an existing materials / components input table (Example 1). [Figure 7] A concrete example of a design goal table (Example 1). [Figure 8] A specific example of an tolerance coefficient table (Example 1). [Figure 9] A concrete example of a weight coefficient table (Example 1). [Figure 10] A concrete example of the first design information management table (Example 1). [Figure 11] A concrete example of the second design information management table (Example 1). [Figure 12] Overall operation flow diagram of the design support system (Example 1). [Figure 13] A flowchart illustrating the operation involved in obtaining design conditions and creating a management table (Example 1). [Figure 14] Operational flowchart for information processing based on a trained model (Example 1). [Figure 15] Diagram illustrating a design support system involving multiple companies (Example 2). [Figure 16] Operational flowchart for design information registration and login authentication (Example 2). [Modes for carrying out the invention]

[0054] The design support system, which assists in product design, is equipped with a processing means, a storage means that functions as a database, an input means, and a display means. The database stores design information for multiple materials and parts that make up the product in advance. The processing means, which functions as a machine learning means, generates a trained model that extracts design information with a high degree of fit to the design target information using the multiple design information as training data. Then, based on the trained model, the processing means calculates the functional fit and cost fit for existing materials and parts and new design candidates individually, using the design target values ​​input by the design engineer as a reference. Furthermore, to make it easier for the design engineer to understand the comparison results, the display means displays the respective fit scores side by side. [Examples]

[0055] In Example 1, a design support system 1 applied to a single company with multiple business divisions will be described with reference to Figures 1 to 14. Figure 1 shows a block diagram of the design support system. In Figure 1, any configuration is enclosed in a dashed line. Figure 2 shows a specific example of the user interface; Figure 2(A) shows an example where design target information is manually entered, and Figure 2(B) shows an example where provisional design target information is entered as an image and difference information is manually entered.

[0056] Figures 3 and 4 show the configuration diagrams of tables that manage the design information used as input data by the trained model. Figure 3 shows the linkage of data tables related to the first design information (design information of existing materials and parts), and Figure 4 shows the linkage of data tables related to the second design information (design information of new materials and parts). Figure 5 shows a list of design information stored in the database. Figures 6 to 11 show examples of design information to be stored in each table shown in Figures 3 and 4. Figures 12 to 14 show the operation flow from the stage of registering design information in the database to the comparative display of design information.

[0057] The design support system 1 consists of a server computer 100, communication means 200, and operating terminals 300 belonging to each business unit. The server computer 100 is equipped with machine learning means 110 and a database 120, and the operating terminals 300 are equipped with input means 310 and display means 320 (see Figure 1). The server computer 100 is equipped with processing means 10 that functions as machine learning means 110, and storage means 20 for storing the machine learning means application 111 and training data 112. The processing means can be any known central processing unit, and the storage means can be any known HDD, SSD, etc.

[0058] The machine learning method application 111 can be created using any programming language available for publicly known AI development purposes, but is not limited to that. For example, Python (trademark registered) is preferred because it is a programming language suitable for data collection and data analysis. The machine learning method may perform machine learning using either a publicly known linear regression model or a nonlinear regression model. For example, a nonlinear regression model could be a publicly known support vector machine, neural network, etc. Alternatively, a linear regression model could be a publicly known regression analysis method such as multiple regression analysis, ridge regression, or lasso regression.

[0059] The machine learning means 110 generates a trained model 130 that, using the machine learning described above, uses multiple design information as training data 112. When design target information is input from the input means, the model is trained to extract design information with a high degree of fit from a group of design information registered in the database 120 (see Figure 1). The design information used as training data 112 includes functional information 121 and cost information 122 of materials and parts that constitute the product (see Figures 1 and 5), as well as tolerance coefficient information 123 used to narrow down the design information (see Figures 1 and 8), and weight coefficient information 124 used to calculate the degree of fit (see Figures 1 and 9). The tolerance coefficient information includes "greater than" and "less than" information indicating the tolerance direction along with the tolerance coefficient.

[0060] Furthermore, the processing means 10 functions as a specific design information extraction means 131, a design target information creation means 132, a suitable target extraction means 133, a conditional suitability derivation means 134, and a suitable target guidance means 135. Of these, the processing means 10 performs inference processing based on the trained model 130 in the processing of the suitable target extraction means 133 and the conditional suitability derivation means 134. The specific design information extraction means 131 identifies existing material and component identification information from at least one of the character information or image information entered by the user from the input means of the operating terminal. Furthermore, it extracts materials and components with matching identification information from the group of design information registered in the database 120, and identifies their functional information and cost information (first design information). In the case of zero-based design, the input of existing material and component identification information and the extraction of first design information are omitted.

[0061] The aforementioned identification information can be any information that uniquely identifies a material or component, such as the name of the material or component or an internal management number (see Figure 6). The method of inputting the identification information can be text input or selection from a pull-down menu, if the input means 310 is a keyboard, mouse, etc. Alternatively, image information may be read from the storage means of the operation terminal 300, or input may be made by reading image information printed on paper using an external device such as a scanner. The input image information is acquired by the server computer via the communication means 200. The image information can be material or component specifications, safety data sheets, drawing data, etc. The data format of the image information is preferably PDF with text information, CAD data, etc.

[0062] If the image data format is JPEG, BMP, PNG, etc., which do not contain text information, the feature extraction processing means 131a, such as a known character recognition application, can be used to extract the features of the character image and then extract the text information. In the case of drawings, specifications, two-dimensional codes, etc., printed on paper, the image information can be acquired using a known scanner or two-dimensional code reader connected to the operation terminal 300, and the identification information can be identified as text information using the optical character recognition processing means 131b.

[0063] The design target information creation means 132 creates design target information (see Figure 7) from the functional characteristics and actual values ​​entered by the user from the input means 310, and from the cost information items and actual values. The conformance target extraction means 133 narrows down the materials and parts that exceed a predetermined degree of conformance to the design target information from the design information registered in the database 120, and extracts second design information. Here, the narrowing down is performed by whether or not the tolerance value obtained by multiplying the actual value of the design target information by a predetermined tolerance coefficient is met.

[0064] Specifically, if the actual value for the functional characteristic "strength (MPa)" is entered as "400", and the tolerance coefficient is set to "0.9" and the tolerance direction to "greater than or equal to", the tolerance value is specified as "360 MPa or greater". The conformance target extraction means 133 extracts second design information from the database that satisfies the tolerance value for all functional characteristics. Alternatively, the tolerance direction may be set to "±", allowing tolerance values ​​to be specified on both the upper and lower sides of the actual value. If there are no design information entries within the tolerance range, this may be displayed on the display means to prompt the design engineer to review the design target value and check for incorrect input.

[0065] The condition suitability derivation means 134 derives the functional suitability from the functional characteristic values ​​of the design target information and the functional characteristic values ​​of the second design information. For example, if the design target information is "strength 400 (MPa)" and the second design information is "380 MPa", the error rate with respect to the design target value is "-5%", so the individual suitability for strength should be set to "95%". Setting the upper limit of the functional suitability to "100%" is preferable to prevent overestimating the functional characteristic values.

[0066] Then, the functional fit for each functional characteristic value is multiplied by a predetermined weighting coefficient, and the sum of the multiplied values ​​is calculated (sum value = Σ functional fit for each functional characteristic value × weighting coefficient). This sum value may be used as the functional fit, or it may be averaged to make it easier for design engineers to understand intuitively. Alternatively, the functional fit may be converted to a percentage (for example, 95.0%) by dividing the sum value by the number of functional characteristic items. Even if the sum value is displayed as is, the level of functional fit can be judged by the magnitude of the value.

[0067] Furthermore, when narrowing down the second design information, a standard deviation can be used instead of the error rate from the design target value. In this case, a standard deviation related to the individual functional characteristic value is calculated for all materials and parts registered in the database. For example, when comparing the strength of all materials, the standard deviation is stored linked to the functional characteristic value, such as "50" for general structural rolled steel and "60" for stainless steel. Then, the standard deviation of the input functional target value is calculated, and the second design information in which that standard deviation falls within the acceptable range is extracted.

[0068] For example, if the standard deviation of the design target value for strength is "40" and the acceptable range for the standard deviation is set to "±5", then materials and parts with a strength standard deviation in the range of "35 to 45" should be extracted. Functional suitability can also be derived from the error rate of the standard deviation of the functional target value. In this case, an objective evaluation that considers variability (magnitude of standard deviation) rather than the actual numerical value of the functional characteristic becomes possible, so that candidates with high functional suitability are extracted without being missed and can be presented to the design engineer as a comparison target.

[0069] The condition suitability derivation means 134 derives the cost suitability from the ratio of actual cost data to cost information based on the first or second design information, or from the error rate. For materials, the cost is calculated as "material reference price (yen / kg)" plus "unit volume (cm³)". 3 " / piece)" and "Specific gravity (g / cm³)" 3You can calculate the product cost per part (material cost) by multiplying by ").. The ratio is calculated as follows: if the actual data shows a unit price of 50 yen and the unit price in the first design information is 100 yen, then "50 yen / 100 yen = 50%", so the cost fit is 50%.

[0070] Furthermore, when processing materials and parts, the sum of the material cost (yen / kg) and processing cost (yen / kg) should be multiplied by the unit volume and specific gravity. Processing costs should consist of at least one of the following: surface treatment costs such as plating and coating, and heat treatment costs that change the surface properties through heating. Surface treatment costs and heat treatment costs are relatively easy to estimate from design drawings and specifications, and are also likely to increase costs. Therefore, estimating these processing costs at the design stage is effective for accurately understanding the costs.

[0071] Furthermore, the condition suitability derivation means 134 may use "annual improvement cost" as the cost suitability. The annual improvement cost is derived when "annual production quantity" is input as design target information. Specifically, it can be derived using "annual improvement cost = product price per part × annual production quantity". In this case, if material A has an annual improvement cost of 2.5 million yen and material B has an annual improvement cost of 3 million yen, material A, which has a lower cost after cost improvement, should be judged to have a higher cost suitability and given a higher priority.

[0072] Alternatively, the "annual effect difference" may be used as the cost-fitness indicator. The annual effect difference is the difference between the annual improvement cost in the first design information and the annual improvement cost in the second design information. In this case, materials and parts with a higher annual effect difference should be judged as having a higher cost-fitness indicator and given higher priority. Even when determining cost-fitness as a ratio, it is preferable to display both the annual improvement cost and the annual effect difference in the display method to make it easier to grasp the cost improvements resulting from design changes (see Figure 2).

[0073] The conformance target guidance means 135 displays the first design information (existing materials and parts) and the second design information side-by-side on the display means (see Figures 1 and 2). In this case, if multiple second design information sources are extracted, the candidates are sorted in descending order of conformance to the conditions. Furthermore, when a design engineer is conducting a review, it is desirable to allow them to sort the display priority of the extracted candidates for the second design information by functional conformance, cost conformance, or overall conformance (see the Sort column in Figure 2).

[0074] Prioritizing cost-effectiveness is suitable for redesigning mass-produced products; prioritizing functional effectiveness is suitable for redesigning custom-made products; and prioritizing overall evaluation is suitable for redesigning mid-range products where functionality is slightly prioritized. When prioritizing overall effectiveness, it is sufficient to sort the products in descending order of the average values ​​of functional effectiveness and cost-effectiveness. In all cases, it is possible to compensate for the lack of experience of design engineers according to the design objectives, and to shorten the review time while ensuring high reliability.

[0075] The storage means 20 of the server computer stores an application 111 for operating the machine learning means 110 and training data 112 (see Figure 1). Furthermore, the design information of materials and parts that the processing means 10 extracts through inference processing based on the trained model 130 is stored as a database 120. The training data 112 may be the same as the group of design information registered in the database 120. For large companies with separate business divisions, the database 120 stores functional information, cost information, tolerance coefficient information, and weight coefficient information related to materials and parts, linked to the divisional information 125 of each individual business division. Note that the divisional information has an arbitrary configuration, and even companies with separate business divisions may use a unified database.

[0076] Functional information and other data are stored independently in databases for each business unit, allowing for different values ​​to be stored to reflect the specific circumstances of each unit. For example, in a particular business unit, procurement of certain materials and parts may be easier compared to other units, allowing for settings that reflect the actual working environment, such as setting lower cost information. In addition, settings that reflect differences in factory equipment levels can be adjusted, such as by adjusting the ease of processing value to be higher than in other business units. Furthermore, some materials and parts may be registered only in the memory categories of specific business units.

[0077] The operating terminal 300 operated by the design engineer can be any known general-purpose computer (see Figure 1). The operating terminal is equipped with an input means 310 and a display means 320. The input means 310 can be a keyboard, mouse, scanner that optically recognizes paper documents, a two-dimensional code reader, etc. Identification information of existing materials and parts (such as names), design target information, classification information, etc. are input from the input means and acquired by the server computer 100 via the communication means 200.

[0078] The display means 320 only needs to be able to display a user interface (hereinafter referred to as UI), and can be a well-known monitor, touch panel, etc. The display means 320 displays, in addition to input fields for design target information, first design information and second design information extracted from the database by the processing means by performing inference processing based on a trained model, in comparison. Here, an example of the UI will be briefly explained with reference to Figure 2.

[0079] In UI400, a top bar is displayed at the top of the screen, an input field 420 is displayed on the left side of the center of the screen, and a comparison field 430 of extracted design information is displayed on the right side of the center of the screen. The top bar 410 displays an operation mode selection field 411 and a category selection field 412 for selecting the business unit category. Although the operation mode selection field is displayed in the UI here, the operation mode selection field is an optional configuration and may be omitted.

[0080] In the operation mode selection field 411, you can select either "manual input" or "image input" using a checkbox. If "manual input" is checked, the words "Select existing material / part" and the input field 421a for existing material / part information will be displayed above the input field 420. The design engineer will enter the material name / part name into the input field by keyboard input or by selecting from a pull-down menu (see Figure 2(A)). If there are no existing materials or parts, such as in zero-based design, the input field 421a should be left blank.

[0081] Furthermore, if "Image Input" is checked (see Figure 2(B)), multiple existing part information can be entered into the input field 421b by dragging and dropping digital data such as JPEG, PDF, or CAD files of part drawings. Alternatively, part drawings printed on paper can be acquired as digital data using optical scanning or the like, and entered into the "Existing Part Selection" input field 421b.

[0082] When existing part information is entered as an image, the words "Select Existing Part" appear, and a list of part names corresponding to the image information is displayed in the input field 421b for existing parts. The design engineer can identify the existing part to be modified by selecting a part name from the list (see the thick border for selecting part α in Figure 2(B)). When a part name is selected, the main diagram of the selected part is displayed in the input field 421b for "Select Existing Part". Since the design engineer can see the main diagram along with the part name (product number), it helps prevent parts mix-ups and input errors. It would be desirable for the main diagram to be enlargeable by selecting it with a mouse or similar method.

[0083] Furthermore, in the input field 422 for design target information, the functional information and cost information (first design information) of the selected parts are read from the database and automatically entered as "provisional design target information (initial value)". The design engineer only needs to input actual data regarding the functional information and costs related to the design changes (hereinafter referred to as difference information). When the user changes a numerical value such as "strength (MPa)" as difference information, the design target information creation means, as described above, replaces part of the provisional design information with the difference information to create the official design target information.

[0084] In the category selection field 412, if the database storage area is divided by business unit and the numerical values ​​of the design information are fine-tuned for each category, the user specifies the category of design information that the matching target extraction means should extract. If the operating terminal belongs to the first business unit, normally only the category of the business unit to which it belongs is selected. On the other hand, for example, when considering which business unit should be responsible for the design and development of a new product, selecting multiple business units allows for the consideration of the optimal materials and parts while comparing the procurement and processing capabilities of each business unit.

[0085] Once the design target information and classification information have been entered, clicking the execute button 424 will cause the matching target extraction means to extract the second design information with a high degree of matching to the design target information, and the functional and cost matching degrees of the first design information (existing) 431 and the second design information (new) 432 will be displayed side by side in the comparison column 430 on the right side of the center of the screen. If multiple second design information 432s have been extracted, the comparison candidates can be changed using the scroll bar 433, etc.

[0086] Here, the "annual production quantity" is also entered in the input field on the left side of the screen (see Figure 2(A)). Therefore, in the comparison column on the right side of the screen, the "Second Design Information" column displays not only the "Cost Suitability (100%)" as a percentage, but also the "Annual Improvement Cost (1 million yen)" and the "Annual Effect Difference (2 million yen)". The "Annual Improvement Cost" shows the annual cost itself when using that material / part, and the "Annual Effect Difference" shows the difference between the cost of the First Design Information and the cost when changing to the material related to the Second Design Information. Similarly, the "First Design Information" column also displays the cost suitability, but since there is no annual effect difference, "0 yen" is displayed.

[0087] Next, the linkage structure of the data tables that make up the database and the registration information of each data table will be explained with reference to Figures 3 to 11. Figure 3 shows the linkage structure of the data tables related to the first design information (existing materials and parts), and Figure 4 shows the linkage structure of the data tables related to the second design information (new design). Figure 5 shows a specific example of a design information table related to materials. Figure 6 shows the existing materials and parts input table. Figure 7 shows the design target input table. Figure 8 shows the tolerance coefficient table. Figure 9 shows the weight coefficient table. Figure 10 shows the first design information management table, and Figure 11 shows the second design information management table.

[0088] The first design information management table T1 is linked to five tables: the design information table T10, the existing materials / parts input table T20, the design target input table T30, the tolerance coefficient table T40, and the weight coefficient table T50 (see Figure 3). The second design information management table T2 is linked to four tables: the design information table T10, the design target input table T30, the tolerance coefficient table T40, and the weight coefficient table T50 (see Figure 4). Each table is assigned unique table identification information (such as T1). In addition, each table manages individual data using an address ID consisting of an alphabetical row number and a numerical column number so that it can be uniquely identified (see Figures 5 to 11).

[0089] This section explains the information related to tables T10 to T50. Design information table T10 contains multiple functional information and cost information (see Figure 5). Note that only materials are shown to avoid data complexity. Functional information includes functional characteristic values ​​(see Figures 5(A) and 5(B)) which contain numerical values ​​and units such as material name, specific gravity, strength, rigidity, heat resistance, abrasion resistance, corrosion rate, water absorption, and thermal expansion coefficient, and functional characteristic values ​​(see Figures 5(B) and 5(C)) which contain only multi-stage evaluation values ​​such as chemical resistance, UV resistance, electrical insulation, supplyability (ease of supply), and ease of processing (see Figures 5(B) and 5(C)). Items with multi-stage evaluation are marked with "(*)".

[0090] The cost information includes at least the material cost (raw material cost) and surface treatment costs (processing costs). Although not shown in the diagram, heat treatment costs (processing costs), etc., are also registered (see Figure 5(C)). Note that the material cost shown in the diagram is merely an example and does not correspond to actual market prices. For ease of understanding, processing costs are shown here as "one-tenth of the material cost" for convenience, but in reality, they can be set flexibly according to past design experience, the extent of processing, etc.

[0091] The existing materials and parts input table T20 stores identification information of existing materials and parts entered by the design engineer (see Figure 6). Here, an example is shown where "SUS304," which is stainless steel, is entered as the "name" of the identification information. The design target table T30 stores functional target information and actual data entered by the design engineer (see Figure 7). Functional target information does not need to be entered for all functional characteristic values ​​registered in the design information table. If there are items that have not been entered, those items should be excluded and the functional conformance should be calculated. Also, if the annual production quantity item is not entered in the actual data, only the cost conformance as a percentage will be derived, and the annual improvement cost and annual effect difference amount will be omitted.

[0092] The tolerance coefficient table T40 stores the tolerance direction and tolerance coefficient set for each item of functional information (see Figure 8). For example, for strength, the tolerance direction is set to "greater than or equal to" and the tolerance coefficient to "0.9". If the design target value for strength is "400 MPa", then multiplying this by "0.9" gives "360 MPa ≤" (360 MPa or greater), which becomes the tolerance value when extracting the second design information.

[0093] The weight coefficient table T50 stores the weight coefficients set for each functional information item, as well as notes on changing the weight coefficients (see Figure 9). The weight coefficient is a coefficient value that indicates which function should be given priority in the design when deriving the degree of functional suitability. Here, the weight coefficients are set in the order of increasing importance for items that are more directly related to product performance: Strength (0.25) > Heat resistance (0.2) > Rigidity = Wear resistance = Corrosion rate = Machinability (0.15) > Others (0.1). The weight coefficients may be adjusted by the database administrator, taking into account the notes column and past design results.

[0094] Now, the first design information management table T1 (see Figure 3) retrieves the "name," which is the identification information of existing materials and parts, from the address ID that stores the "name" in the existing materials and parts input table T20. It retrieves the "classification information," "functional objective information," and "actual data" from the address ID in the design objective input table T30. Once the "identification information" and "classification information" of existing materials and parts are available, the materials and parts with matching "names" are extracted from the design information table T10 that matches the classification information (for example, T10-1 with a sub-number for the first business unit) to form the first design information.

[0095] Then, the address IDs for reading the design information of materials and parts related to the first design information are stored in columns 3 through 20 of row B of the first design information management table T1 (see Figure 3). As a result, specific functional characteristic values ​​corresponding to each address ID are stored in row B of the first design information management table (see Figure 10). Furthermore, "tolerance coefficient information" is obtained from the address IDs of the tolerance coefficient table T40, and "weighting coefficient information" is obtained from the address IDs of the weighting coefficient information T50, and stored in the first design information table. The condition suitability derivation means derives the functional suitability and cost suitability individually based on the first design information, design target information, tolerance coefficient information, and weighting coefficient information stored in the first design information management table T1.

[0096] The second design information management table T2 (see Figures 4 and 11) first retrieves "classification information," "functional target information," and "actual data" from the address ID of the design target input table T30. Next, it retrieves "tolerance coefficient information" from the address ID of the tolerance coefficient table T40. Here, the tolerance value is calculated by multiplying the target value by the tolerance coefficient and stored in row C of the second design information table. In the extraction process of the second design information, materials and parts corresponding to this tolerance value are extracted from the design information table T10. Therefore, the processing method differs from that of the first design information, where identification information (name, etc.) is used for identification.

[0097] Once the second design information is identified, the address ID from which to read that design information is stored in the second design information column (rows A and B / column 24) of the second design information management table T2 (see Figure 11). Note that the second design information column may span multiple columns. If multiple second design information entries are extracted, the respective candidate design information is stored in columns 25 and beyond. After storing the address ID of the second design information, the "weighting coefficient information" is obtained from the address ID of the weighting coefficient information T50. The condition fit derivation means individually derives the functional fit and cost fit based on the second design information, tolerance coefficient information, and weighting coefficient information stored in the second design information management table T2.

[0098] The following briefly describes the operation flow of the design support system with reference to Figures 12 to 14. Figure 12(A) shows the overall flow diagram, and Figure 12(B) shows the detailed steps for acquiring design conditions entered from the input means. Figure 13 shows the detailed steps for creating the management table by the specific design information extraction means and the design target information creation means. Figure 14 shows the steps from the inference processing by the suitability target extraction means and the condition suitability derivation means to the comparative display by the suitability target guidance means. Steps enclosed by dashed lines are optional and may be omitted. The table numbers (T1, etc.) correspond to the tables shown in Figures 3 to 11. The information to be displayed on the display means corresponds to the UI in Figure 2.

[0099] In step 100, design information for materials and parts is registered in the database (S100). If the storage area is to be divided by business division, multiple design information tables (T10-1, ...) are created with sub-numbers attached to the table identification numbers, and the design information is registered in them. In step 200, tolerance coefficients and weight coefficients are set and registered in each table (T40, T50) (S200). In step 300, the server computer acquires design conditions such as design target information and tolerance coefficients entered from the input means (S300). These are the initial setup steps.

[0100] In step 400, the specific design information extraction means and the design target information creation means create design information management tables T1 and T2 based on the design conditions (S400). In step 500, based on the information stored in the design information management tables T1 and T2, the matching target extraction means and the condition fit derivation means derive the condition fit through inference processing based on the trained model (S500). In step 600, the matching target guidance means displays the first design information and its condition fit, and the second design information and its condition fit, side by side on the display means (S600). If the initial setup is complete, processing can start from step 400.

[0101] Here, we will explain the detailed flow of step 300 (see Figure 12(B)). In step 310, the operation mode conditions entered by the design engineer are acquired (S310). Specifically, these are operation conditions such as selecting either a material or a component, or whether the input is by direct keyboard input or image input (see Figure 2). In step 320, the classification information is acquired and the branch number of the design information table from which the design information is read (e.g., T10 "-1") is identified (S320).

[0102] In step 330, the identification information (such as names) of existing materials and parts entered by the design engineer is acquired (S330). In step 340, the design target information entered by the design engineer is acquired (S340). Steps S310 to S340 can be performed in any order. Furthermore, since the acquisition of operation mode conditions and classification specification information is an arbitrary configuration, this step may be omitted depending on the configuration of the design support system.

[0103] Next, we will explain the detailed flow of step 400, which creates design information management tables T1 and T2 (see Figure 13). In step 410, the first design information management table T1 is created (S410). In step 420, the identification information of existing materials and parts entered by the design engineer is stored in table T1 (S420). In step 430, the first design information read from the database based on the identification information is stored in table T1 (S430).

[0104] In step 440, the tolerance coefficient and weight coefficient are stored in table T1 (S440). In step 450, the design target information is stored in table T1 (S450). In step 460, the second design information management table T2 is created (S460). In step 470, the tolerance coefficient and weight coefficient are stored in table T2. In step 480, the design target information is stored in table T2. The processing order of each step is just an example, and the order may be changed.

[0105] Next, we will explain the detailed flow of step 500, which derives the degree of suitability (see Figure 14). In step 510, the degree of functional suitability related to the first design information stored in table T1 is derived (S510). In step 520, the degree of cost suitability related to the first design information is derived (S520). In step 530, the tolerance coefficient is multiplied by the design target information stored in table T2 to calculate the tolerance value for the functional information (S530). In step 540, the second design information that satisfies the tolerance value is extracted from the database (design information table T10) (S540).

[0106] Step 550 derives the functional fit related to the extracted second design information (S550). Step 560 derives the cost fit (percentage) related to the extracted second design information (S560). Step 570 calculates the annual improvement cost and the annual cost difference between the first and second design information (S570). Step 570 is not executed if the annual production quantity is omitted from the design target information, in which case the cost fit is derived only as a percentage. [Examples]

[0107] In Example 2, a design support system 2 applicable to a group of multiple companies will be described with reference to Figures 15 and 16. Figure 15(A) shows a conceptual diagram of collaboration by a group of multiple companies, and Figure 15(B) shows a block diagram of the server computer that constitutes the design support system 2. Figure 16 shows the operation flow added in the design support system 2. In Example 2, components identical to those in Example 1 are denoted by the same reference numerals and their explanation is omitted.

[0108] In design support system 2, a single server computer 100 is shared by multiple users 500 and multiple manufacturers 600, each within the scope of their assigned access rights (see Figure 15(A)). Here, manufacturers 600 refer to those who provide and sell materials and parts, mainly material manufacturers and parts manufacturers. On the other hand, users 500 refer to those who use materials and parts, mainly companies to which design engineers belong.

[0109] Each manufacturer 600 can access the server computer 100 via a communication means 200 such as the internet, register the design information of their products in their assigned design information table, and expand their sales channels by encouraging users to use their products. On the other hand, users can consider new products that are not yet registered in their assigned design information table as second design information. Furthermore, since the effort of registering new products can be transferred to the manufacturers, the database 120 can be expanded without any effort on the part of the users.

[0110] The processing unit 10, which constitutes the server computer, functions not only as described in Example 1, but also as a user information management unit 136 and a registration management unit 137 (see Figure 15(B)). The user information management unit 136 registers users for each user and manufacturer, grants access rights to each user, and authenticates users who access the server computer by logging them in.

[0111] The registration management means 137 manages registration information to allow the registration and editing of design information in the database within the scope of access rights, and also manages the public / private status of individual design information. For example, manufacturer α can register and edit design information in the memory area allocated to manufacturer α, and can also choose whether to make the registered design information public or private.

[0112] Therefore, no one other than manufacturer α can edit the design information registered by manufacturer α, thus preventing tampering. Furthermore, if the option to make the design information public is selected, even those other than manufacturer A are permitted to extract that design information. In cases of sales suspension or other reasons, making the information private allows it to remain in the database as a history while preventing others from extracting it.

[0113] Database 120 has memory areas divided for each user (A, B, etc.) and each manufacturer (α, β, etc.), and each user classification information 126 and manufacturer classification information 127 are stored together with access permission information 128, public / private information 129, design information, etc. (see Figure 15(B)). If the user consists of multiple business units, this classification may be further subdivided by business unit. Access permissions may be granted by an administrator representing a group of multiple companies, or by a management company different from the group of multiple companies.

[0114] Next, the operation flow added from Example 1 will be explained with reference to Figure 16. The added operation flow is included in steps 100 and 300. In step 110, prior to the registration of design information, user registration processing is performed by the user information management means (S110). For example, unique information that identifies the user, such as name, address, and technical field, is registered. In step 120, access rights linked to the user information are granted, storage categories corresponding to the access rights are created in the database, and a login ID, password, etc., are issued (S120).

[0115] In step 130, the user enters their login ID and other information from the operating terminal, and login authentication is performed on the server (S130). In step 140, design information is registered and edited (S140). For users, this design information should consist of uploading known material and component information to the server. For manufacturers, this design information should consist of uploading design information for new products aimed at expanding sales channels. The data format to be uploaded is not limited; any data from a publicly known spreadsheet software is acceptable. In step 150, public and private information is registered and edited for each design (S150). In step 200, each user can set and register tolerance coefficients and weight coefficients.

[0116] In step 301, the design engineer accesses the server and enters a login ID, etc., and login authentication is performed by the user information management means (S301). In step 302, access rights are read from the category associated with the login ID (S302). In obtaining category specification information in step 320, first in step 321 the scope of the category relating to the business unit is identified (S321). Here, it is determined whether design information classified under other business units should be included as extraction target as second design information.

[0117] Furthermore, step 322 specifies the scope of classifications relating to other companies (S322). Here, it is determined whether or not to include the design information within the manufacturer's public scope as a target for extraction of the second design information. The specification of the public scope may be for all items, or it may be for a technical field registered as unique information. The technical field may be an industrial classification such as automobile manufacturing, production machinery manufacturing, or pharmaceutical manufacturing, or it may be a detailed classification based on past usage history such as for 3D printers, daily necessities, amusement machines, or computers.

[0118] (others) In this embodiment, a specific example of inputting information on existing materials and parts (material name, etc.) was described. However, when designing from scratch, it is sufficient to omit inputting existing information and instead input functional target information related to the new design. • In this embodiment, we have described a specific example in which the design information table T10 is a separate table for each business division, but a unified design information table T10 may be used for multiple business divisions. The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The technical scope of the present invention is indicated by the claims, not limited to the foregoing description, and all modifications within the meaning and scope equivalent to the claims are intended. [Explanation of Symbols]

[0119] 1,2…Design support system, 100...Server computer, 10...Processing means, 20...Storage means, 110...Machine learning methods, 111...Applications, 112...Training data, 120...Database, 121...Functional information, 122...Cost information, 123...Tolerance coefficient information, 124...Weight coefficient information, 125...Classification information, 126...User classification information, 127...Manufacturer classification information, 128... Access permission information, 129... Public / private information, 130...Trained model, 131...Specific design information extraction means, 131a... Optical character recognition means, 131b... Feature extraction means, 132...Means for creating design target information, 133...Means for extracting suitable targets, 134...Means for deriving the degree of suitability, 135...Means for guiding the suitable target, 136...User information management means, 137...Registration management means, 200... means of communication, 300...Operating terminal, 310...Input means, 320...Display means, 400...User Interface (UI), 410...Top bar, 411...Operation Mode selection field, 412...Category selection field, 420... Input field, 421a... Input field for existing material / part information, 421b... Input field for existing parts, 422... Input field for design target information, 423...Difference information, 424...Execute button, 430...Comparison column, 431...First design information, 432...Second design information, 433... scroll bar, 500…user, 600…manufacturer

Claims

1. A design support system that assists in the design of a product, A computer includes storage means, processing means, input means, and display means, The storage means stores, as a database, a plurality of design information relating functional information representing the functional characteristics of a plurality of materials and parts constituting the product, cost information of the plurality of materials and parts, and weight coefficients indicating the priority between the items of functional characteristics. The processing means functions as a machine learning means for executing machine learning processing, a means for extracting specific design information, a means for creating design target information, and a means for guiding suitable targets. The processing means of the machine learning means generates a trained model that uses multiple pieces of the design information as training data and infers the design information with a high degree of fit to the acquired design target information. The specified design information extraction means extracts first design information from the database that matches the information on existing materials and existing parts input from the input means. The design target information creation means creates the design target information based on the input functional target information and actual cost data. The processing means functions as a fitting target extraction means and a condition fit deriving means by performing inference processing based on the trained model. The matching target extraction means acquires the design target information and causes the trained model to infer second design information with a high degree of fit to the acquired design target information. The condition suitability derivation means derives the functional suitability and cost suitability individually by comparing the design target information and the first design information, and also derives the functional suitability and cost suitability individually by comparing the design target information and the second design information. Regarding the functional fit, for each functional characteristic item, the functional fit for that item is calculated from the error rate with the design target information, and further, the weight coefficient is obtained from the database, and the functional fit for that item is multiplied by the weight coefficient, and the weighted functional fit for that item is added to derive the sum of the functional fits. The conformance target guidance means displays the sum of functional conformance and cost conformance derived by comparing the design target information and the first design information, and the sum of functional conformance and cost conformance derived by comparing the design target information and the second design information, as evaluation indicators used for design judgment, in a manner that can be compared on the display means. A design support system characterized by the following features.

2. The input means includes an image information input means capable of inputting image information, The aforementioned image information includes at least digital data or paper media of a design drawing containing the aforementioned design information. The specified design information extraction means performs optical character recognition processing or feature extraction processing on the character information contained in the input image information to identify character information relating to existing materials and existing parts, and extracts the first design information from the database. The design target information creation means creates provisional design target information based on the specified first design information, and when further difference information is input from the input means, it creates the design target information. The design support system according to feature 1.

3. The aforementioned database registers at least the product price and processing costs as information constituting the cost information. The aforementioned processing costs include at least one of surface treatment costs or heat treatment costs. The design target information creation means creates the design target information when the product price, processing costs, and annual production quantity are input as the actual data from the input means. The condition suitability derivation means derives cost suitability based on the actual data. The design support system according to feature 1.

4. The design support system comprises a server computer and multiple operating terminals. The server computer functions as the processing means and the database, The aforementioned operating terminal is equipped with the input means and the display means, The storage area of ​​the database is partitioned according to the divisions of the multiple business units to which each of the operating terminals belongs, and the design information is stored independently for each division. The design target information creation means creates the design target information from the functional target information and the actual data, as well as the classification information input from the input means. The matching target extraction means causes the second design information corresponding to the design information of the storage area assigned to the specified category to be inferred. A design support system according to any one of claims 1 to 3.

5. The aforementioned design support system is applied to a group of companies consisting of manufacturers and users of the aforementioned product. The server computer comprises communication means, user information management means, and design information registration management means. The processing means functions as the user information management means and the registration management means, The user information management means manages access rights for each manufacturer and each user. The database further subdivides and stores the categories according to the access rights, When the operating terminal belonging to either the manufacturer or the user accesses the server computer via the communication means, The registration management means allows editing of the design information stored in the database within the scope of the access rights, and also allows the user to select whether to make each piece of design information public or private. The design support system according to feature 4.

6. A design support program that runs on a computer and assists in the design of a product, The computer is configured to include processing means, storage means, input means, and display means. The design support program makes the processing means function as a machine learning means for executing machine learning processing, a means for extracting specific design information, a means for creating design target information, and a means for guiding suitable targets, and also makes the storage means function as a database. The database stores multiple design information items, each relating functional information representing the functional characteristics of multiple materials and components constituting the product, cost information of the multiple materials and components, and weighting coefficients indicating the priority between the items of functional characteristics. As the first step, the machine learning means generates a trained model that uses multiple pieces of design information as training data and infers design information with a high degree of fit to the acquired design target information. As the second step, the specific design information extraction means extracts first design information from the database that corresponds to the design information of existing materials and existing parts entered by the user. In the third step, the design target information creation means creates design target information from the functional target information and cost data entered by the user. In the fourth step, the processing means functions as a fitting target extraction means by performing inference processing based on the trained model, acquires the design target information, and causes the trained model to infer second design information with a high degree of fit to the acquired design target information. As the fifth step, the processing means functions as a condition fit derivation means by performing inference processing based on the trained model, and individually derives the functional fit and cost fit by comparing the design target information and the first design information, and also individually derives the functional fit and cost fit by comparing the design target information and the second design information. Regarding the functional fit, for each functional characteristic item, the functional fit for that item is calculated from the error rate with the design target information, and further, the weight coefficient is obtained from the database, and the functional fit for that item is multiplied by the weight coefficient, and the weighted functional fit for that item is added to derive the sum of the functional fits. As the sixth step, the conformance target guidance means causes the computer to perform processing so that the sum of the functional conformance and cost conformance derived by comparing the design target information and the first design information, and the sum of the functional conformance and cost conformance derived by comparing the design target information and the second design information, are displayed on the display means in a comparable manner as evaluation indicators to be used for design judgment. A design support program characterized by the following features.

7. A computer-readable recording medium, Recorded in order to have a computer execute the design support program described in claim 6, A recording medium characterized by the following features.

Images