Prediction system, prediction method, prediction program, learned model generation system, learned model generation method, and learned model generation program

Abstract

To reduce the processing time and processing load required to analyze molded objects to be produced by a 3D printer.SOLUTION: A prediction device (10) includes: an acquisition unit configured to acquire information representing the shape, material, and molding conditions of a molded object to be produced by a 3D printer; and an output unit configured to output information representing mechanical properties of the molded object to be produced by inputting the information acquired by the acquisition unit into a learned model generated by machine learning, which uses information representing the shape, material, and molding conditions of a molded object to be produced by the 3D printer as input information and information representing mechanical properties of the molded object as output information.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a prediction system, a prediction method, a prediction program, a trained model generation system, a trained model generation method, and a trained model generation program for supporting the design of molded bodies to be formed by a three-dimensional printer. [Background technology]

[0002] A technology for supporting the design of molded objects to be printed using a three-dimensional printer has been proposed. For example, Patent Document 1 describes a design support device that acquires specification data indicating required specifications for at least one of the performance and function of a three-dimensional structure to be designed, and identifies candidate structures based on the required specifications from a database in which parameters describing the physical properties of multiple candidate structures that differ from each other in at least one of material and structure are registered. The design support device described in Patent Document 1 evaluates the performance and function of the candidate structures identified based on the required specifications by performing a simulation when each part of the candidate structures identified based on the required specifications is connected. If the evaluation results do not conform to the required specifications, it determines that the connection of the parts is not good, and repeats the process of estimating candidate structures for each part of the three-dimensional structure or for each region defined by a mesh set for each part. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-63853 Summary of the Invention [Problem to be solved by the invention]

[0004] The conventional techniques have the problem that the processing time and processing load required for analyzing molded bodies formed by three-dimensional printers is large. The technique described in Patent Document 1 also has the same problem.

[0005] One aspect of the present invention aims to realize a prediction system, a prediction method, a prediction program, a trained model generation system, a trained model generation method, and a trained model generation program that can reduce the processing time and processing load required to analyze molded bodies formed by a three-dimensional printer. [Means for solving the problem]

[0006] In order to solve the above problems, a prediction system according to one embodiment of the present invention comprises an acquisition unit that acquires information representing the shape, material, and modeling conditions of a molded body formed by a three-dimensional printer, and an output unit that outputs information representing the mechanical properties of the molded body to be formed, obtained by inputting the information acquired by the acquisition unit into a trained model generated by machine learning, the trained model having as input information the information representing the shape, material, and modeling conditions of the molded body to be formed by the three-dimensional printer, and as output information information representing the mechanical properties of the molded body.

[0007] In addition, a trained model generation system according to one embodiment of the present invention includes an acquisition unit that acquires training data including information representing the shape, material, and modeling conditions of a molded body to be formed by a three-dimensional printer, and information representing the mechanical properties of the molded body, and a generation unit that generates a trained model through supervised machine learning using the training data, with information representing the shape, material, and modeling conditions of the molded body as input and information representing the mechanical properties of the molded body as output.

[0008] Furthermore, a prediction method according to one embodiment of the present invention includes an acquisition step of acquiring information representing the shape, material, and molding conditions of a molded body to be molded, and an output step of outputting information representing the mechanical properties of the molded body to be molded, obtained by inputting the information acquired in the acquisition step into a trained model generated by machine learning, which receives as input information representing the shape, material, and molding conditions of a molded body to be molded by a three-dimensional printer and outputs information representing the mechanical properties of the molded body.

[0009] Furthermore, a prediction method according to one embodiment of the present invention includes an acquisition step of acquiring training data including information representing the shape, material, and modeling conditions of a molded body to be formed by a three-dimensional printer, and information representing the mechanical properties of the molded body, and a generation step of generating a trained model using the training data through supervised machine learning, in which the information representing the shape, material, and modeling conditions of the molded body is input and information representing the mechanical properties of the molded body is output.

[0010] The prediction system according to each aspect of the present invention may be realized by a computer. In this case, the prediction program of the prediction system that causes the computer to operate as each part (software element) of the prediction system to realize the prediction system on a computer, and the computer-readable recording medium on which the program is recorded, also fall within the scope of the present invention.

[0011] In addition, the trained model generation system according to each aspect of the present invention may be realized by a computer. In this case, the trained model generation program of the trained model generation system, which realizes the trained model generation system on a computer by causing the computer to operate as each part (software element) of the trained model generation system, and the computer-readable recording medium on which it is recorded, also fall within the scope of the present invention. [Effects of the Invention]

[0012] According to one aspect of the present invention, it is possible to reduce the processing time and processing load required for analyzing a molded body produced by a three-dimensional printer. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a block diagram showing an example of the configuration of a design support system according to a first embodiment of the present invention. [Figure 2] 1 is a block diagram showing an example of the configuration of a prediction device according to a first embodiment of the present invention. [Figure 3] FIG. 2 is a block diagram showing an example of the configuration of a user terminal according to the first embodiment of the present invention. [Figure 4] FIG. 3 is a sequence diagram showing an example of the flow of a prediction method according to the first embodiment of the present invention. [Figure 5] FIG. 2 is a diagram showing an example of an input screen according to the first embodiment of the present invention. [Figure 6] FIG. 3 is a diagram showing an example of a display screen of an analysis result according to the first embodiment of the present invention. [Figure 7] FIG. 4 is a diagram showing an example of a display screen for details of an analysis result according to the first embodiment of the present invention. [Figure 8] FIG. 2 is a sequence diagram showing an example of the flow of a modeling method according to the first embodiment of the present invention. [Figure 9] FIG. 10 is a block diagram showing an example of the configuration of a prediction device according to a second embodiment of the present invention. [Figure 10] FIG. 10 is a flow diagram showing an example of the flow of a trained model generation method according to a second embodiment of the present invention. [Figure 11] FIG. 10 is a flow diagram showing an example of the flow of a trained model generation method according to a second embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0014] [Embodiment 1] [Configuration of the design support system] An embodiment of the present invention will be described in detail below. FIG. 1 is a block diagram showing an example of the configuration of a design support system 1 according to this embodiment. The design support system 1 is a system for supporting the design of a molded body to be formed by a three-dimensional printer 30. The design support system 1 includes a prediction device 10, a user terminal 20, and the three-dimensional printer 30. The prediction device 10, the user terminal 20, and the three-dimensional printer 30 are communicably connected via a communication line N. The specific configuration of the communication line N does not limit this embodiment, but examples of the communication line include a wireless LAN (Local Area Network), a wired LAN, a WAN (Wide Area Network), a public line network, a mobile data communication network, or a combination thereof.

[0015] The three-dimensional printer 30 is a three-dimensional printer that forms a molded body. Examples of 3D printing methods include material extrusion and fused deposition modeling, but the method is not limited to these, and molded bodies may be formed using other methods. The prediction device 10 is a device that has a function to support the design of a molded body to be formed by the three-dimensional printer 30, and is an example of a prediction system according to the present disclosure. The prediction device 10 is, for example, a general-purpose computer. The user terminal 20 is a terminal used by a user of the design support system 1, and is, for example, a personal computer such as a laptop computer, a tablet terminal, or a smartphone.

[0016] [material] The material used by the three-dimensional printer 30 to form the molded body is, for example, a matrix resin or a composite material containing short fibers and a matrix resin. For example, the composite material is composed of short fibers and a matrix resin impregnated into the short fibers.

[0017] <Matrix resin> The matrix resin can be appropriately selected from resins that can be used for molding using the three-dimensional printer 30, and may be a thermoplastic resin or a photocurable resin. The matrix resin may be a single type or a combination of two or more types. Examples of thermoplastic resins include polycarbonate, acrylonitrile-butadiene-styrene (ABS), acrylate-styrene-acrylonitrile (ASA), polylactic acid, PETG (polyethylene terephthalate copolymerized with cyclohexanedimethanol), polyetherimide, polyurethane (TPU), polystyrene, nylon, acrylic, polyether ether ketone (PEEK), polyolefin, polyamide, and mixtures of two or more of these. Examples of photocurable resins include epoxy resin, acrylic resin, and mixtures of these. The matrix resin is preferably a thermoplastic resin from the viewpoint of being suitable for use in molding using the three-dimensional printer 30, and is preferably a photocurable resin from the viewpoint of exhibiting high strength after molding. The resin can be selected depending on the device.

[0018] <Short fibers> The short fibers in the composite material can be appropriately selected from fibers that have the property of reinforcing the molded body. The fibers may be one or more types. Examples of the short fibers according to the present disclosure include at least one of carbon fibers, aramid fibers, acrylic fibers, metal fibers, and glass fibers.

[0019] The fiber length of the short fibers can be determined appropriately, for example, in consideration of the type of fiber and the strength required for the molded article. From the viewpoint of improving the strength of the resulting molded article, the fiber length is preferably 10 μm or more, more preferably 50 μm or more, and even more preferably 100 μm or more. On the other hand, from the viewpoint of facilitating molding using a three-dimensional printer and better utilizing the fiber's characteristic of improving the rigidity of the molded article, the fiber length is preferably 100 mm or less, more preferably 50 mm or less, and even more preferably 10 mm or less.

[0020] In the embodiment of the present invention, the fiber length of the short fibers in the molded body is determined as a measured value of the length of the fibers at any location including the surface and cross section of the resin molded body, for example.

[0021] The thickness of the fibers can be appropriately determined within the range that allows the molded article to exhibit the desired strength, and may be, for example, 3 to 150 μm.

[0022] Examples of the fibers include glass fibers, aramid fibers, carbon fibers, polyethylene fibers, and mixtures thereof. From the viewpoint of developing the desired mechanical properties of the molded article, it is preferable that the fibers be one or more types of fibers selected from the group consisting of glass fibers, aramid fibers, and carbon fibers.

[0023] The fiber content in the molded product is substantially the same as the fiber content in the composite material. From the viewpoint of fully expressing the fiber properties in the molded product, the fiber content in the composite material is preferably 10% by mass or more, more preferably 20% by mass or more, even more preferably 30% by mass or more, and even more preferably 40% by mass or more. Furthermore, from the viewpoint of fully expressing the properties of materials other than the fiber in the molded product, the fiber content in the composite material is preferably 80% by mass or less, more preferably 70% by mass or less, and even more preferably 60% by mass or less. If the content of materials other than the fiber in the composite material is above the lower limit, the resin properties are fully expressed, and if it is below the upper limit, the fiber properties are fully expressed, which is preferable.

[0024] The fiber may be in the form of a single fiber or a fiber bundle.

[0025] <Other ingredients> In an embodiment of the present invention, the molded article and composite material may further contain other components in addition to the matrix resin and short fibers, as long as the effects of the present invention are obtained. The other components may be one or more. Examples of the other components include a filler, a polymerization initiator, and a lubricant.

[0026] [Configuration of prediction device] 2 is a block diagram showing an example of the configuration of the prediction device 10. The prediction device 10 includes a control unit 110, a storage unit 120, a communication unit 130, an input unit 140, and an output unit 150.

[0027] <Communications Department> The communication unit 130 communicates with devices external to the prediction device 10 via a communication line N. The communication unit 130 transmits data supplied from the control unit 110 to other devices, and supplies data received from other devices to the control unit 110.

[0028] <Input section> The input unit 140 is configured to receive input to the prediction device 10, and includes, for example, input devices such as a keyboard, a mouse, a touch panel, a camera, and a microphone. The input unit 140 may also be configured to receive data from the input devices via an interface such as a USB (Universal Serial Bus).

[0029] <Output section> The output unit 150 is a component for performing output from the prediction device 10, and includes, for example, output devices such as a display, a printer, a touch panel, a speaker, etc. The output unit 150 may be configured to include, for example, an interface such as a USB, and to output data to the output device via the interface.

[0030] <Storage section> The memory unit 120 stores various types of information referenced by the control unit 110. Examples of such information include input information D1, output information D2, a first trained model M1, a second trained model M2, and a database DB1. The input information D1 is information that the simulation execution unit 12, which will be described later, inputs to the first trained model M1 or the second trained model M2. The output information D2 is information that is output from the first trained model M1 or the second trained model M2.

[0031] Database DB1 is a database that stores information indicating the type of material and mechanical property information indicating the mechanical properties of the material in association with each other. Database DB1 is a database referenced by a simulation execution unit 12, which will be described later. Examples of the information indicating the type of material include information indicating the type of composite material, the type of matrix resin, or the type of fiber. Examples of types of matrix resin include polypropylene, polycarbonate, ABS, ASA, polylactic acid, PETG, PEI, TPU, polystyrene, nylon, and acrylic. Examples of types of fiber include carbon fiber, aramid fiber, acrylic fiber, metal fiber, and glass fiber.

[0032] (Mechanical property information) The mechanical property information indicating the mechanical properties of the material includes, for example, at least one of the mechanical properties of the matrix resin, the mechanical properties of the short fibers, the mechanical properties of the composite material, and the mechanical properties of the molded body. Examples of the mechanical properties of the matrix resin include the tensile modulus of the resin, the shear modulus of the resin, Poisson's ratio, linear expansion coefficient, thermal conductivity, density, specific heat, and volume content of the resin. Examples of the mechanical properties of the short fibers include the volume content, tensile modulus of the short fibers, fiber length, fiber diameter, fiber radius, and average inter-fiber distance in the molded body. Examples of the mechanical properties of the composite material include the Poisson's ratio in one or more directions of the composite material, the linear expansion coefficient in one or more directions of the composite material, thermal conductivity, density, and specific heat. The mechanical properties of the composite material may also include the tensile modulus of the composite material. Here, the tensile modulus of the composite material may, for example, be an experimentally measured value. Examples of the mechanical properties of the molded body include the orientation coefficient indicating the fiber orientation of the molded body and the shear modulus of the composite material. Note that the mechanical property information is not limited to the above examples and may include other information. Furthermore, the mechanical property information does not need to include all of the above-mentioned information, but may include only a part of the above-mentioned information.

[0033] (First trained model) The first trained model M1 is a model generated by machine learning, and is a model for estimating the mechanical properties of a molded body to be formed by the three-dimensional printer 30. Note that storing the first trained model M1 in the storage unit 120 means that parameters defining the first trained model M1 are stored in the storage unit 120. As an example, the first trained model M1 can be realized by a model having a neural network structure such as a CNN (Convolutional Neural Network). However, the first trained model M1 is not limited to a neural network model, and may be another model generated by machine learning.

[0034] (Input information for the first trained model) The input information of the first trained model M1 includes information representing (i) the shape, (ii) the material, and (iii) the modeling conditions of the molded body to be modeled by the three-dimensional printer 30. The input information may also include information other than that described above.

[0035] (shape information, modeling conditions) Examples of shape information representing the shape of the molded body include, but are not limited to, G-code or CAD data. Furthermore, as an example, the shape information may include information indicating at least one of the long dimension, short dimension, height dimension, thickness dimension, diameter, roundness, cylindricity, coaxiality, concentricity, and profile of the molded body to be modeled. As an example, the modeling conditions include at least one of the modeling temperature (nozzle temperature of the 3D printer 30), ambient temperature, table temperature, and modeling width. As an example, the shape information and modeling conditions are received from the user terminal 20.

[0036] (Materials information) The material information representing the material includes, for example, information indicating at least one of the type of material (such as the type of resin or the type of fiber), information indicating the mechanical properties of the material, and the fiber content. For example, the material information is received from the user terminal 20. Furthermore, for example, the material information may include information read by the simulation execution unit 12 from the database DB1.

[0037] (Output information of the first trained model) The output information of the first trained model M1 includes information representing the mechanical properties of the molded body. For example, the output information includes at least one of the temperature distribution, deformation amount, principal stress, shear stress, von Mises stress, normal stress, shear strain, and normal strain of the molded body. The output information may also include information representing the deformation amount of the molded body. The output information may also include information representing a change in the temperature distribution of the molded body over time, or may also include information representing a change in the principal stress of the molded body over time. For example, the output information is image data representing a still image or a moving image, but the output information is not limited to image data and may also be numerical data representing the mechanical properties of the molded body.

[0038] (Second trained model) The second trained model M2 is a trained model generated by machine learning, and is a trained model for estimating the amount of deformation of a molded body due to an external force. Note that storing the second trained model M2 in the memory unit 120 means that parameters defining the second trained model M2 are stored in the memory unit 120. As an example, the second trained model M2 can be realized by a model having a neural network structure such as CNN. However, the second trained model M2 is not limited to a neural network model, and may be another model generated by machine learning.

[0039] (Input and output information of the second trained model) The input information of the second trained model M2 includes information representing an external force applied to the molded body. As an example, the information representing the external force is received from the user terminal 20. Furthermore, the output information of the second trained model M2 includes information representing a change in the molded body.

[0040] <Control unit> The control unit 110 includes an acquisition unit 11, a simulation execution unit 12, an orientation coefficient calculation unit 13, and a molded body analysis parameter calculation unit 14. The simulation execution unit 12 is an example of an output unit according to the present disclosure. Each unit of the control unit 110 is realized by the control unit 110 reading and executing instructions of a computer program stored in the memory unit 120.

[0041] (Acquisition Department) The acquisition unit 11 acquires information representing the shape, material, and molding conditions of the molded body. As an example, the acquisition unit 11 receives the information by receiving the information from the user terminal 20 via the communication unit 130. As another example, the acquisition unit 11 may acquire the information by reading the information from a storage location specified by the user of the prediction device 10 (which may be a storage device within the prediction device 10 or a storage device external to the prediction device 10). The acquisition unit 11 may also acquire the information input to the input unit 140.

[0042] The acquisition unit 11 also acquires information indicating the deformation tolerance of the molded body set by the user. As an example, the acquisition unit 11 receives the information by receiving the information from the user terminal 20 via the communication unit 130. As another example, the acquisition unit 11 may acquire the information by reading the information from a storage location specified by the user of the prediction device 10 (which may be a storage device within the prediction device 10 or a storage device external to the prediction device 10). The acquisition unit 11 may also acquire the information input to the input unit 140.

[0043] The acquisition unit 11 also acquires information representing an external force to be applied to a molded body that is a modeling target. As an example, the acquisition unit 11 receives the information by receiving the information from the user terminal 20 via the communication unit 130. As another example, the acquisition unit 11 may acquire the information by reading the information from a storage location specified by the user of the prediction device 10 (which may be a storage device within the prediction device 10 or a storage device external to the prediction device 10). The acquisition unit 11 may also acquire the information input to the input unit 140.

[0044] In addition, when the material of the molded body is a composite material, the acquisition unit 11 may acquire, as information representing the material of the molded body, information indicating the mechanical properties of the short fibers contained in the composite material containing short fibers and a matrix resin, information indicating the mechanical properties of the matrix resin, and the tensile modulus of the composite material.

[0045] (Simulation execution unit) The simulation execution unit 12 outputs information representing the mechanical properties of the molded body to be modeled, which is obtained by inputting input information including the information acquired by the acquisition unit 11 into the first trained model M1. As an example, the simulation execution unit 12 outputs the information by transmitting the information to the user terminal 20 via the communication unit 130. The simulation execution unit 12 may also output the information by writing the information to a storage destination specified by the user of the prediction device 10 (which may be a storage device within the prediction device 10 or a storage device external to the prediction device 10). The simulation execution unit 12 may also output the information to an output device such as a display via the output unit 150.

[0046] (Orientation coefficient calculation section) The orientation coefficient calculation unit 13 calculates the orientation coefficient η0, which represents the fiber orientation of the molded body, using the information acquired by the acquisition unit 11 and a model that represents the relationship between the information and the orientation coefficient of the composite material. Details of the process by which the orientation coefficient calculation unit 13 calculates the orientation coefficient η0 will be described later.

[0047] (Parameter calculation section for compact analysis) The molded body analysis parameter calculation unit 14 calculates molded body analysis parameters based on the orientation coefficient η0, which represents the fiber orientation of the molded body. The molded body analysis parameters are parameters used in analyzing molded bodies formed by a three-dimensional printer. The molded body analysis parameters include the tensile modulus and shear modulus of the composite material in one or more directions. The molded body analysis parameters may further include at least one of the Poisson's ratio in one or more directions of the composite material, the linear expansion coefficient in one or more directions of the composite material, the thermal conductivity, the density, and the specific heat. However, the molded body analysis parameters are not limited to the above examples and may include other parameters.

[0048] [User device configuration] 3 is a block diagram showing the configuration of the user terminal 20. The user terminal 20 includes a control unit 210, a storage unit 220, a communication unit 230, an input unit 240, and an output unit 250. The user terminal 20 is, for example, a general-purpose computer. The communication unit 230 communicates with devices external to the user terminal 20 via a communication line N.

[0049] <Input section> The input unit 240 is configured to receive input to the user terminal 20, and includes, for example, input devices such as a keyboard, a mouse, a touch panel, a camera, and a microphone. The input unit 240 may also be configured to receive data from the input devices via an interface such as a USB (Universal Serial Bus).

[0050] <Output section> The output unit 250 is configured to perform output from the user terminal 20, and includes, for example, output devices such as a display, a printer, a touch panel, a speaker, etc. The output unit 250 may also be configured to include an interface such as a USB, and to output data to the output device via the interface.

[0051] <Storage section> The storage unit 220 stores various types of information referenced by the control unit 210. Examples of such information include user input data D21 and display data D22. The user input data D21 is data that a user inputs to the user terminal 20 using an input device. The display data D22 is data received from the prediction device 10 and represents content to be displayed on a display.

[0052] <Control unit> The control unit 210 includes an application execution unit 21. The application execution unit 21 is realized by the control unit 210 reading and executing instructions of an application program stored in the storage unit 220. The application execution unit 21 executes the application program stored in the storage unit 220, and performs processing to transmit information representing the shape, material, and molding conditions of a molded body that is the object to be molded to the prediction device 10, and processing to display simulation results of the molded body. An application implemented by the application execution unit 21 is, for example, a general-purpose web browser, but is not limited to this. The application execution unit 21 may also be a dedicated application that communicates with the prediction device 10 to assist in the design of a molded body.

[0053] [Prediction method flow] 4 is a sequence diagram showing an example of the flow of the prediction method according to this embodiment. As an example, the process in FIG. 4 is started when the user of the user terminal 20 performs an operation to start an application using an input device.

[0054] <Step S11> First, in step S11, the application execution unit 21 displays on the display an input screen for the user to input the shape, material, and modeling conditions of the molded body that is the modeling target.

[0055] FIG. 5 is a diagram showing an example of an input screen. In the example of FIG. 5, screen SC11 includes text boxes A11 to A16, a pull-down list A17, and buttons B11 to B13. The text boxes A11 to A15 are text boxes for inputting the printing conditions. The text boxes A11 to A15 are used to input the case title, printing temperature (C), ambient temperature (C), table temperature (C), and printing width (mm), respectively. The text box A16 is used to input the allowable range of deformation of the molded body. The pull-down list A17 is used to select a material. The button B11 is used to return to the main screen. The button B12 is used to upload the G-code. The button B13 is used to start analysis.

[0056] The user uses the input device to input the case title, modeling temperature, ambient temperature, table temperature, modeling width, and allowable deformation range into text boxes A11 to A16. The user also uses the input device to select a material from a pull-down list A17. The selected material may be, for example, a matrix resin or a composite material containing a matrix resin and short fibers.

[0057] The user also uses the input device to select button B12 to upload the G-code. When button B12 is selected, the application execution unit 21 executes a process for uploading the G-code stored in, for example, the storage unit 220 to the prediction device 10. Note that other data (for example, CAD data) may be input instead of the G-code as information representing the shape of the molded body.

[0058] <Step S12> When the user inputs various modeling conditions in the text boxes A11 to A15, selects a material in the pull-down list A17, and selects the button B13 to start the analysis, in step S12, the application execution unit 21 transmits input information including the various information input in the text boxes A11 to A16, information representing the material selected in the pull-down list A17, and shape information such as G-code to the prediction device 10.

[0059] <Steps S13 to S16> In step S13, the acquisition unit 11 receives input information including shape information, modeling conditions, and material information of a molded body to be modeled. In step S14, if the shape information included in the acquired input information is not a G-code (NO in step S13), the acquisition unit 11 executes the processes of steps S15 to S16 to generate a G-code. First, in step S15, the acquisition unit 11 sets a modeling direction and start position of the molded body based on the acquired CAD data, and generates a G-code based on the set information and CAD data. At this time, the modeling direction and start position of the molded body may be set, for example, by an administrator of the prediction device 10. To generate the G-code, for example, general-purpose software that generates G-code from CAD data can be used.

[0060] <Step S17> In step S17, the simulation execution unit 12 executes a process for identifying the molded body analysis parameters to be used for simulating the molded body. Examples of the process for identifying the molded body analysis parameters include (i) a method of identifying the parameters by referring to database DB1, and (ii) a method of calculating the molded body analysis parameters based on the orientation coefficient. As an example, if the molded body analysis parameters are stored in database DB1, the above method (i) is used. On the other hand, if the molded body analysis parameters are not stored in database DB1, the above method (ii) is used. These methods will be explained in order.

[0061] ((i) Example of identifying by referencing database DB1) In this example, the acquisition unit 11 refers to the database DB1 to identify the molded body analysis parameters stored in association with the material type information received from the user terminal 20. The molded body analysis parameters include, for example, the tensile modulus and shear modulus of the composite material in one or more directions. The molded body analysis parameters may further include at least one of the Poisson's ratio in one or more directions of the composite material, the linear expansion coefficient in one or more directions of the composite material, the thermal conductivity, the density, and the specific heat.

[0062] ((ii) Example of calculating parameters for compact analysis based on orientation coefficient) In this example, the acquisition unit 11 acquires, from the database DB1, material information corresponding to the material type information received from the user terminal 20. For example, the information acquired by the acquisition unit 11 from the database DB1 is the volume content V f , tensile modulus of short fiber E f , tensile modulus of the matrix resin E m , shear modulus of matrix resin G m , fiber length L, fiber diameter D, fiber radius r, average fiber spacing R, and tensile modulus E of the composite material c Includes:

[0063] The orientation coefficient calculation unit 13 calculates the orientation coefficient η0, which represents the fiber orientation of the molded body, using the information acquired by the acquisition unit 11 and a model that represents the relationship between the information and the orientation coefficient of the composite material. Here, for example, the Cox-Krenchel model is used as the model. The Cox-Krenchel model calculates the tensile modulus E of a composite material having fibers that are ideally arranged in space. c The Cox-Krenchel model is expressed by the following equation (1): The Cox-Krenchel model is described by adding coefficients to a model that simply follows the linear law of mixtures, taking into account the influence of orientation and fiber length.

number

[0064] In the above formula (1), E c is the tensile modulus of the composite, and V f is the volume fraction of short fibers. E f is the tensile modulus of the short fiber. E m is the tensile modulus of the matrix resin, and G m is the shear modulus of the matrix resin.

[0065] Also, η l is a correction coefficient, which is used to calculate the effective elastic modulus of the fiber from the difference in displacement between the short fiber and the matrix resin when a displacement is applied to the composite material. Here, L is the fiber length of the short fiber, D is the fiber diameter of the short fiber, r is the fiber radius of the short fiber, and R is the average value of the inter-fiber distance of the short fiber. Also, X i is the packing factor, which is 4.0 for square-packing.

[0066] In addition, in formula (1), η0 is the orientation coefficient. For example, the orientation coefficient η0 takes a value of 1 when the fibers are uniaxially oriented in the molded object, 3 / 8 when they are 2D randomly oriented, and 1 / 5 when they are 3D randomly oriented.

[0067] Thus, in the Cox-Krenchel model, the tensile modulus of the composite material, E c is the tensile modulus of short fiber E f , tensile modulus of matrix resin E m , short fiber volume content V f It can be seen that the fiber diameter D and fiber length L (i.e., aspect ratio) affect the fiber strength.

[0068] When the Cox-Krenchel model is used, the orientation coefficient calculation unit 13 calculates the orientation coefficient η by substituting the information and the tensile modulus acquired by the acquisition unit 11 into the Cox-Krenchel model. Specifically, the orientation coefficient calculation unit 13 calculates the orientation coefficient η by substituting the volume content V acquired by the acquisition unit 11 into the Cox-Krenchel model. f , tensile modulus E f , tensile modulus E m , shear modulus G m , fiber length L, fiber diameter D, fiber radius r, average fiber spacing R, and tensile modulus E cThe orientation coefficient η0 is calculated by substituting into the above equation (1).

[0069] The molded body analysis parameter calculation unit 14 generates a homogenized model by a homogenization method using the orientation coefficient η calculated by the orientation coefficient calculation unit 13. Here, the homogenization method involves replacing the heterogeneous composite material with an equivalent homogenized material model (homogenized model) and analyzing it using equivalent physical property values. The homogenized model is a model in which a molded body made of a composite material is replaced with a single material using the homogenization method, and is a model that simulates the fiber orientation of the composite material. The molded body analysis parameter calculation unit 14 calculates molded body analysis parameters using the generated homogenized model. For example, general-purpose structural analysis software is used to calculate the molded body analysis parameters. When general-purpose structural analysis software is used, the molded body analysis parameter calculation unit 14 can obtain the molded body analysis parameters by, for example, inputting the generated homogenized model and the mechanical property information of the matrix resin and short fibers acquired by the acquisition unit 11 into the general-purpose structural analysis software.

[0070] <Step S18> In step S18, the simulation execution unit 12 executes a modeling simulation by inputting input information to the first trained model M1. The input information input by the simulation execution unit 12 to the first trained model M1 includes, for example, modeling conditions input by the user, information representing the type of material, and information representing the shape of the molded body (such as a G-code).

[0071] The input information may also include information representing the mechanical properties of a material identified by the acquisition unit 11 with reference to the database DB1. In this case, the simulation execution unit 12 may refer to a database that stores material types and mechanical property information indicating the mechanical properties of the material in association with each other, acquire the mechanical property information corresponding to the material represented by the information acquired by the acquisition unit 11, and input the input information including the acquired mechanical property information to the first trained model M1.

[0072] The input information may also include an orientation coefficient calculated by the orientation coefficient calculation unit 13. In this case, in other words, the simulation execution unit 12 can input the input information including the orientation coefficient calculated by the calculation unit to the first trained model M1.

[0073] The input information may also include molded body analysis parameters calculated by the molded body analysis parameter calculation unit 14. In this case, in other words, the simulation execution unit 12 can input input information including the molded body analysis parameters calculated by the molded body analysis parameter calculation unit 14 to the first trained model M1. Note that if the input information of the first trained model M1 does not include an orientation coefficient and molded body analysis parameters, the processing of step S17 in FIG. 4 does not need to be executed.

[0074] <Step S19> In step S19, the simulation execution unit 12 transmits information indicating the analysis results to the user terminal 20 based on the output information obtained by inputting the above input information into the first trained model M1. The information indicating the analysis results is, for example, image data indicating at least one of the temperature distribution, deformation amount, principal stress, shear stress, von Mises stress, normal stress, shear strain, and normal strain of the molded body, but is not limited to this.

[0075] In addition, if the output information of the first trained model M1 includes information indicating the deformation amount of the formed body, the simulation execution unit 12 may transmit the comparison result between the deformation amount of the formed body and the deformation allowance acquired by the acquisition unit 11.

[0076] <Step S20> In step S20, the application execution unit 21 receives information indicating the analysis result from the prediction device 10, and displays the received information indicating the analysis result on the display.

[0077] Fig. 6 is a diagram showing an example of a display screen of analysis results that the application execution unit 21 displays on the display. In the example of Fig. 6, screen SC12 includes three types of analysis results A21 to A23 and buttons B21 to B22. The analysis results A21 to A23 are information indicating simulation results when a molded body is molded at a different ambient temperature. Button B21 is a button for returning to the main screen of the application. Button B22 is a button for exiting the application.

[0078] 6 is displayed, the user can select one of the displayed analysis results. When the user performs an operation to select one of the analysis results using the input device, the application execution unit 21 displays details of the selected analysis result on the display.

[0079] FIG. 7 is a diagram showing an example of a display screen for details of the analysis results. In the example of FIG. 7, screen S13 includes the molding conditions specified by the user and an image showing details of the analysis results of the molded body. Screen S13 also includes a button B31 for returning to the screen of FIG. 6 and a button B32 for exiting the application. Screen S13 also includes link information A31 for displaying a dynamic image showing the change over time in the principal stress of the molded body, and link information A32 for displaying a dynamic image showing the change over time in the temperature distribution of the molded body.

[0080] When the user selects link information A31 or link information A32 on the screen of Figure 7, the application execution unit 21 sends an acquisition request for the selected link information to the prediction device 10 and displays the moving image received from the prediction device 10 on the display.

[0081] <Step S21> In step S21, the simulation execution unit 12 displays a screen for inputting external force conditions on the display, and transmits the external force conditions input by the user to the prediction device 10. The input screen for external force conditions may be a screen different from the screens exemplified in Figures 5 to 7, and any of the screens in Figures 5 to 7 may include a text box for inputting external force conditions.

[0082] <Step S22> In step S22, the simulation execution unit 12 executes an external force simulation by inputting the external force conditions into the second trained model M2.

[0083] <Step S23> In step S23, the simulation executing unit 12 transmits the external force simulation result to the user terminal 20. In other words, the simulation executing unit 12 transmits to the user terminal 20 information obtained by inputting the information representing the external force acquired by the acquiring unit 11 into the second learned model M2.

[0084] <Step S24> In step S24, the application execution unit 21 receives the external force simulation result and displays the received external force simulation result on the display.

[0085] [Modeling method flow] 8 is a sequence diagram showing an example of the flow of the modeling method according to this embodiment. In step S21, the user of the user terminal 20 inputs modeling conditions using the input device, and the application execution unit 21 accepts the input modeling conditions. In step S22, the application execution unit 21 transmits a modeling instruction including the input modeling conditions to the prediction device 10.

[0086] In step S23, the prediction device 10 instructs the three-dimensional printer 30 to model a molded body based on the modeling instruction received from the user terminal 20. In step S24, the three-dimensional printer 30 receives the modeling instruction from the prediction device 10 and acquires the modeling conditions, and in step S25 starts modeling of the molded body.

[0087] Furthermore, based on the analysis results obtained in step S20, the three-dimensional printer 30 may be instructed to form a molded body. In this case, the three-dimensional printer 30 receives information indicating the analysis results obtained in step S20 and a modeling instruction, acquires modeling conditions, and starts forming a molded body in step S25.

[0088] As described above, according to this embodiment, the prediction device 10 acquires information representing the shape, material, and modeling conditions of a molded body produced by the three-dimensional printer 30, and outputs information representing the mechanical properties of the molded body to be modeled, which information is obtained by inputting the acquired information into the first trained model M1 generated by machine learning. By using the first trained model M1 generated by machine learning, it is possible to reduce the processing time and processing load required to analyze a molded body produced by the three-dimensional printer 30, compared to when general-purpose structural analysis software is used for analysis.

[0089] [Embodiment 2] Other embodiments of the present invention will be described below. For ease of explanation, the same reference numerals will be used to designate components having the same functions as those described in the above embodiment, and the description thereof will not be repeated.

[0090] [Configuration of prediction device] 9 is a block diagram showing an example of the configuration of a prediction device 10B according to this embodiment. The prediction device 10B includes a control unit 110B, a storage unit 120B, a communication unit 130, an input unit 140, and an output unit 150.

[0091] The control unit 110B includes an estimation phase execution unit 111 and a learning phase execution unit 112. The estimation phase execution unit 111 includes an acquisition unit 11, a simulation execution unit 12, an orientation coefficient calculation unit 13, and a molded body analysis parameter calculation unit .

[0092] The learning phase execution unit 112 trains the first trained model M1 and the second trained model M2. The learning phase execution unit 112 includes a teacher data acquisition unit 16, a trained model generation unit 17, and a teacher data generation unit 18.

[0093] The teacher data acquisition unit 16 acquires teacher data used for training the first trained model M1. The teacher data used for training the first trained model M1 includes multiple sets of information representing the shape, material, and modeling conditions of the molded body to be modeled by the 3D printer 30, and information representing the mechanical properties of the molded body. The information representing the mechanical properties of the molded body is, for example, information obtained by general-purpose structural analysis software. However, the information representing the mechanical properties of the molded body is not limited to the above-mentioned examples, and may be, for example, actual measured values ​​obtained by experiment.

[0094] Furthermore, the teacher data acquisition unit 16 acquires teacher data used for training the second trained model M2. The teacher data used for training the second trained model M2 includes multiple sets of information representing the external force applied to the molded body and information representing the change in the molded body. The information representing the change in the molded body is, for example, information obtained by inputting the information representing the external force applied to the molded body into general-purpose structural analysis software. However, the information representing the change in the molded body is not limited to the above-mentioned examples and may be, for example, actual measured values ​​obtained by experiment.

[0095] The trained model generation unit 17 generates a first trained model M1 through supervised machine learning using training data. The trained model generation unit 17 also generates a second trained model M2 through supervised machine learning using training data.

[0096] The teacher data generation unit 18 acquires the results of a simulation of the molding of a molded body by the three-dimensional printer 30 using information representing the shape, material, and molding conditions of the molded body to be molded by the three-dimensional printer 30, and generates the teacher data including the above information and the acquired results.

[0097] The storage unit 120B stores the input information D1, the output information D2, the first trained model M1, the second trained model M2, the database DB1, and the training data TD1. The training data TD1 is training data used to train at least one of the first trained model M1 and the second trained model M2.

[0098] [Process for generating trained models] 10 and 11 are diagrams showing an example of the flow of a trained model generation method according to this embodiment. Fig. 10 is a diagram showing an example of the flow of a method for generating a first trained model M1, and Fig. 11 is a diagram showing an example of the flow of a method for generating a second trained model M2.

[0099] <Step S31> The teacher data generation unit 18 generates teacher data by executing the processes of steps S31 to S36. Here, a case will be described in which the output of general-purpose structural analysis software is used as the teacher data. First, in step S31, the teacher data generation unit 18 acquires mechanical property information (information indicating the mechanical properties of the short fibers and information indicating the mechanical properties of the matrix resin), shape information indicating the shape of the molded body, and the tensile modulus of elasticity of the composite material. As an example, the teacher data generation unit 18 acquires the above information from the memory unit 120B. As an example, the information indicating the mechanical properties of the short fibers and the information indicating the mechanical properties of the matrix resin acquired by the teacher data generation unit 18 is actual values ​​measured in advance for each material. Furthermore, as an example, the information indicating the shape of the molded body acquired by the teacher data generation unit 18 is information indicating the shape specified by the customer who ordered the molding of the molded body and CAD data.

[0100] In addition, the tensile modulus of elasticity of the composite material acquired by the teacher data generation unit 18 is, as an example, an actual measured value obtained by conducting a tensile test on a molded body that has been previously produced using the composite material using a three-dimensional printer.

[0101] <Step S32> In step S32, the teacher data generating unit 18 calculates the orientation coefficient η0 by substituting the information acquired in step S31 into the Cox-Krenchel model. At this time, the teacher data generating unit 18 calculates the tensile modulus E when "1" is substituted for the orientation coefficient η0. c Calculate the tensile modulus E c and the tensile modulus E obtained in step S31 c The result of the comparison may be used as a guide for the calculated orientation coefficient η0.

[0102] <Steps S33 and S34> In step S33, the teacher data generation unit 18 generates a homogenized model by a homogenization method using the orientation coefficient η calculated in step S32. In step S34, the teacher data generation unit 18 calculates parameters for analyzing the molded body using the homogenized model generated in step S33. For example, general-purpose structural analysis software is used for the calculation process of the parameters for analyzing the molded body. When general-purpose structural analysis software is used, the teacher data generation unit 18 can, for example, obtain the parameters for analyzing the molded body by inputting the homogenized model generated in step S33 and the mechanical property information of the matrix resin and short fibers acquired by the teacher data generation unit 18 into the structural analysis software.

[0103] <Step S35> In step S35, the teacher data generation unit 18 executes a simulation using the calculated parameters for analyzing the molded body, and analyzes the molded body to be manufactured by the three-dimensional printer 30. For example, general-purpose additive manufacturing analysis software can be used for analyzing the molded body. Parameters to be input to the additive manufacturing analysis software include, for example, "Young's modulus [MPa]," "rigidity modulus [MPa]," "Poisson's ratio," "linear expansion coefficient [ / °C]," and "density [kg / m3 ]," "Specific heat [J / kg / m 3 ] and "thermal conductivity [W / m / ℃]". "Young's modulus [MPa]" is the tensile modulus of elasticity of the molded body in each of the three directions (e.g., MD direction, TD direction, ND direction). "Modulus of rigidity [MPa]" is the shear modulus of elasticity of the molded body in each of the three directions. "Poisson's ratio" is the Poisson's ratio in each of the three directions. "Linear expansion coefficient [ / ℃]" is the linear expansion coefficient in each of the three directions. "Density [kg / m 3 ]," "Specific heat [J / kg / m 3 ]" and "thermal conductivity [W / m / °C]" are the density, specific heat, and thermal conductivity of the homogenized model, respectively. However, the parameters used in the additive manufacturing analysis are not limited to the above examples, and parameters such as the manufacturing temperature, the ambient temperature during manufacturing, and the ambient temperature during cooling may also be used.

[0104] The analysis results obtained by the analysis processing of the molded body include at least one of the temperature distribution, deformation amount, principal stress, shear stress, von Mises stress, normal stress, shear strain, and normal strain of the molded body. The analysis results may also include information indicating the deformation amount of the molded body. The analysis results may also include information representing changes in the temperature distribution of the molded body over time, and may also include information representing changes in the principal stress of the molded body over time. One example of the analysis results is image data representing a still image or a moving image, but the analysis results are not limited to image data and may also be numerical data indicating the mechanical properties of the molded body.

[0105] <Steps S36 and S37> In step S36, the teacher data generation unit 18 generates teacher data including the information acquired in step S31 and the analysis result of step S35. In step S37, the teacher data acquisition unit 16 acquires the teacher data generated by the teacher data generation unit 18, and the trained model generation unit 17 generates a first trained model M1 by supervised machine learning using the teacher data.

[0106] <Steps S41 and S42> 11 is a flow diagram showing an example of the flow of a method for generating a second trained model. In step S41, the training data acquisition unit 16 acquires training data to be used for generating the second trained model. In step S42, the trained model generation unit 17 generates the second trained model M2 by supervised machine learning using the training data acquired in step S41.

[0107] As described above, according to this embodiment, the prediction device 10B generates a first trained model M1 through supervised machine learning using training data including information representing the shape, material, and modeling conditions of a molded body to be molded by a three-dimensional printer and information representing the mechanical properties of the molded body, and generates a second trained model M2 through supervised machine learning using training data including information representing external forces applied to the molded body and information representing changes in the molded body. It is possible to generate a first trained model M1 and a second trained model M2 that can reduce the processing time and processing load required for analyzing a molded body to be molded by a three-dimensional printer.

[0108] [Variations] The functions of the prediction devices 10 and 10B described above may be shared and implemented by multiple devices. For example, the prediction devices 10 and 10B described above may be realized as a prediction system in which two or more devices are connected via a communication network. In this case, the prediction system may be, for example, a system including a first device including an acquisition unit 11 and a simulation execution unit 12, and a second device including an orientation coefficient calculation unit 13 and a molded body analysis parameter calculation unit 14. In this case, the functions of the prediction devices 10 and 10B described above are realized by the cooperation of the first device and the second device.

[0109] Furthermore, in the above-described first embodiment, as shown in FIG. 1, the prediction device 10 and the three-dimensional printer 30 are configured as separate devices, but the prediction device 10 and the three-dimensional printer 30 may also be configured as an integrated device.

[0110] Furthermore, at least some of the functions of the prediction devices 10 and 10B described above may be implemented in the user terminal 20. For example, the functions of the simulation execution unit 12 of the prediction device 10 may be implemented as an application of the user terminal 20.

[0111] In the above-described embodiment, the orientation coefficient calculation unit 13 and the teacher data generation unit 18 calculated the orientation coefficient η0 using the mechanical property information and the tensile modulus of elasticity, but the orientation coefficient calculation unit 13 and the teacher data generation unit 18 may calculate the orientation coefficient η0 by other methods. For example, the orientation coefficient calculation unit 13 and the teacher data generation unit 18 may calculate the orientation coefficient η0 by performing image analysis processing on the composite material.

[0112] In the above-described embodiments, the prediction devices 10 and 10B execute the prediction method shown in Fig. 4 and the trained model generation method shown in Fig. 10 and Fig. 11. However, at least some of the steps included in the prediction method and trained model generation method according to the present disclosure may be performed by a user of the prediction devices 10 and 10B, etc. For example, the process of calculating the orientation coefficient in step S17 in Fig. 4 may be performed not by the prediction device 10 but by, for example, the user of the prediction device 10.

[0113] [Software implementation example] The functions of the prediction device 10, 10B (hereinafter referred to as the "device") can be realized by a program that causes a computer to function as the device, and a program that causes a computer to function as each control block of the device (particularly each part included in the control unit 110).

[0114] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program, thereby realizing the functions described in each of the above embodiments.

[0115] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0116] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0117] Furthermore, each process described in each of the above embodiments may be executed by AI (Artificial Intelligence). In this case, the AI ​​may run on the control device or on another device (for example, an edge computer or a cloud server).

[0118] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0119] [summary] (Summary A) The prediction system according to aspect A1 of the present invention comprises an acquisition unit that acquires information representing the shape, material, and modeling conditions of a molded body to be modeled by a three-dimensional printer, and an output unit that outputs information representing the mechanical properties of the molded body to be modeled, obtained by inputting the information acquired by the acquisition unit into a trained model generated by machine learning, the trained model having as input information the information representing the shape, material, and modeling conditions of the molded body to be modeled by the three-dimensional printer and as output information information representing the mechanical properties of the molded body.

[0120] According to the above aspect, it is possible to reduce the processing time and processing load required for analyzing a molded body formed by a three-dimensional printer.

[0121] A prediction system according to aspect A2 of the present invention is the prediction system according to aspect A1 above, wherein the information output by the output unit includes at least one of the temperature distribution, deformation amount, principal stress, shear stress, von Mises stress, normal stress, shear strain, and normal strain of the molded body.

[0122] According to the above aspect, the time and processing load required to analyze at least one of the temperature distribution, deformation amount, principal stress, shear stress, von Mises stress, normal stress, shear strain, and normal strain of the molded body can be reduced.

[0123] A prediction system according to aspect A3 of the present invention is the prediction system described in aspect A2 above, wherein the information output by the output unit includes information indicating the deformation amount of the molded body, the acquisition unit acquires information indicating the deformation allowance of the molded body set by the user, and the output unit outputs the comparison result between the deformation amount of the molded body and the deformation allowance.

[0124] According to the above aspect, the user can know whether the amount of deformation of the molded body to be shaped is within the allowable range.

[0125] A prediction system according to aspect A4 of the present invention is a prediction system according to any one of aspects A1 to A3 above, wherein the acquisition unit acquires information representing the external forces to be applied to the molded body that is the object to be molded, and the output unit outputs information obtained by inputting the information representing the external forces acquired by the acquisition unit into a trained model that takes as input the information representing the external forces to be applied to the molded body and outputs information representing changes in the molded body.

[0126] According to the above aspect, it is possible to reduce the processing time and processing load required to predict a change in a molded body due to an external force applied to the molded body.

[0127] A prediction system according to Aspect A5 of the present invention is the prediction system according to any one of Aspects A1 to A4, wherein the shaping conditions include at least one of shaping temperature, ambient temperature, table temperature, and shaping width.

[0128] According to the above aspect, it is possible to reduce the processing time and processing load required to analyze a molded body formed under a forming condition including at least one of the forming temperature, the ambient temperature, the table temperature, and the forming width.

[0129] A prediction system according to aspect A6 of the present invention is a prediction system according to any one of aspects A1 to A5 above, wherein the information representing the material includes information indicating at least one of the type of material, information indicating the mechanical properties of the material, and information indicating the fiber content.

[0130] According to the above aspect, it is possible to reduce the processing time and processing load required to analyze a molded body made from a material represented by information indicating at least one of the type of material, information indicating the mechanical properties of the material, and fiber content.

[0131] A prediction system according to aspect A7 of the present invention is a prediction system according to any one of aspects A1 to A6 above, wherein the acquisition unit refers to a database that stores material types in association with mechanical property information indicating the mechanical properties of the materials, acquires mechanical property information corresponding to the material represented by the information acquired by the acquisition unit, and inputs input information including the acquired mechanical property information into the trained model.

[0132] According to the above aspect, by using the mechanical property information of the material obtained by referring to the database as input to the trained model, the analysis results of the molded body to be formed can be predicted with greater accuracy.

[0133] The prediction system according to aspect A8 of the present invention is a prediction system according to any one of aspects A1 to A7 above, wherein the acquisition unit acquires, as information representing the material of the molded body, information indicating the mechanical properties of the short fibers contained in a composite material containing short fibers and a matrix resin, information indicating the mechanical properties of the matrix resin, and the tensile modulus of the composite material, and further includes an orientation coefficient calculation unit that calculates an orientation coefficient representing the fiber orientation of the molded body using the tensile modulus acquired by the acquisition unit, and the output unit inputs input information including the orientation coefficient calculated by the orientation coefficient calculation unit to the trained model.

[0134] According to the above aspect, by including the orientation coefficient calculated using the tensile modulus of the composite material in the input of the trained model, the analysis results of the molded body to be formed can be predicted more accurately.

[0135] The prediction system of aspect A9 of the present invention is a prediction system described in any one of aspects A1 to A8 above, further comprising a molded body analysis parameter calculation unit that calculates molded body analysis parameters including a tensile modulus and a shear modulus in one or more directions of a composite material containing short fibers and a matrix resin based on an orientation coefficient representing the fiber orientation of the molded body that is the object to be molded, and the output unit inputs input information including the molded body analysis parameters calculated by the molded body analysis parameter calculation unit to the trained model.

[0136] According to the above aspect, by including parameters for analyzing the molded body calculated based on the orientation coefficient in the input of the trained model, the analysis results of the molded body to be formed can be predicted more accurately.

[0137] The prediction system according to aspect A10 of the present invention is a prediction system according to any one of aspects A1 to A9 above, further comprising a teacher data acquisition unit that acquires teacher data including information representing the shape, material, and molding conditions of a molded body to be molded by a 3D printer, and information representing the mechanical properties of the molded body, and a trained model generation unit that generates the trained model by supervised machine learning using the teacher data.

[0138] According to the above aspect, it is possible to generate a trained model that can reduce the processing time and processing load required to analyze a molded body formed by a three-dimensional printer.

[0139] The trained model generation system according to aspect A11 of the present invention comprises an acquisition unit that acquires training data including information representing the shape, material, and modeling conditions of a molded body to be formed by a three-dimensional printer, and information representing the mechanical properties of the molded body, and a generation unit that uses supervised machine learning using the training data to generate a trained model that takes as input the information representing the shape, material, and modeling conditions of the molded body and outputs information representing the mechanical properties of the molded body.

[0140] According to the above aspect, it is possible to generate a trained model that can reduce the processing time and processing load required to analyze a molded body formed by a three-dimensional printer.

[0141] The trained model generation system of aspect A12 of the present invention is the trained generation system described in aspect A11 above, further comprising a training data generation unit that acquires the results of a simulation of the formation of a molded body by a three-dimensional printer using information representing the shape, material, and formation conditions of the molded body to be formed by the three-dimensional printer, and generates the training data including the information and the acquired results.

[0142] According to the above aspect, by using the results of a modeling simulation as training data for machine learning, the predictive accuracy of the trained model can be improved.

[0143] (Summary B) The prediction method according to aspect B1 of the present invention includes an acquisition step of acquiring information representing the shape, material, and molding conditions of a molded body to be molded, and an output step of outputting information representing the mechanical properties of the molded body to be molded, obtained by inputting the information acquired in the acquisition step into a trained model generated by machine learning, which receives as input information representing the shape, material, and molding conditions of a molded body to be molded by a three-dimensional printer and outputs information representing the mechanical properties of the molded body.

[0144] According to the above aspect, it is possible to reduce the processing time and processing load required for analyzing a molded body formed by a three-dimensional printer.

[0145] A prediction method according to aspect B2 of the present invention is the prediction method according to aspect B1 above, wherein the information output in the output step includes at least one of the temperature distribution, deformation amount, principal stress, shear stress, von Mises stress, normal stress, shear strain, and normal strain of the molded body.

[0146] According to the above aspect, the time and processing load required to analyze at least one of the temperature distribution, deformation amount, principal stress, shear stress, von Mises stress, normal stress, shear strain, and normal strain of the molded body can be reduced.

[0147] A prediction method according to aspect B3 of the present invention is the prediction method described in aspect B2 above, in which the information output in the output process includes information indicating the deformation amount of the molded body, and in the acquisition process, information indicating the deformation allowance of the molded body set by the user is acquired, and in the output process, the comparison result between the deformation amount of the molded body and the deformation allowance is output.

[0148] According to the above aspect, the user can know whether the amount of deformation of the molded body to be shaped is within the allowable range.

[0149] A prediction method according to aspect B4 of the present invention is a prediction method according to any one of aspects B1 to B3 above, in which, in the acquisition step, information representing the external forces to be applied to the molded body that is the object to be molded is acquired, and in the output step, information obtained by inputting the information representing the external forces acquired in the acquisition step into a trained model that takes the information representing the external forces to be applied to the molded body as input and outputs information representing changes in the molded body is output.

[0150] According to the above aspect, it is possible to reduce the processing time and processing load required to predict a change in a molded body due to an external force applied to the molded body.

[0151] A prediction method according to aspect B5 of the present invention is the prediction method according to any one of the above aspects B1 to B4, wherein the shaping conditions include at least one of shaping temperature, ambient temperature, table temperature, and shaping width.

[0152] According to the above aspect, it is possible to reduce the processing time and processing load required to analyze a molded body formed under a forming condition including at least one of the forming temperature, the ambient temperature, the table temperature, and the forming width.

[0153] A prediction method according to aspect B6 of the present invention is a prediction method according to any one of aspects B1 to B5 above, wherein the information representing the material includes information indicating at least one of the type of material, information indicating the mechanical properties of the material, and information indicating the fiber content.

[0154] According to the above aspect, it is possible to reduce the processing time and processing load required to analyze a molded body made from a material represented by information indicating at least one of the type of material, information indicating the mechanical properties of the material, and fiber content.

[0155] A prediction method according to aspect B7 of the present invention is a prediction method according to any one of aspects B1 to B6 above, in which, in the acquisition step, a database that stores material types in correspondence with mechanical property information indicating the mechanical properties of the materials is referenced to acquire mechanical property information corresponding to the material represented by the information acquired in the acquisition step, and input information including the acquired mechanical property information is input to the trained model.

[0156] According to the above aspect, by using the mechanical property information of the material obtained by referring to the database as input to the trained model, the analysis results of the molded body to be formed can be predicted with greater accuracy.

[0157] The prediction method according to aspect B8 of the present invention is the prediction method according to any one of aspects B1 to B7 above, further comprising an orientation coefficient calculation step of acquiring, in the acquisition step, information indicating the mechanical properties of the short fibers contained in a composite material comprising short fibers and a matrix resin, information indicating the mechanical properties of the matrix resin, and the tensile modulus of the composite material as information representing the material of the molded body, and calculating an orientation coefficient representing the fiber orientation of the molded body using the tensile modulus acquired in the acquisition step, and in the output step, input information including the orientation coefficient calculated in the orientation coefficient calculation step is input to the trained model.

[0158] According to the above aspect, by including the orientation coefficient calculated using the tensile modulus of the composite material in the input of the trained model, the analysis results of the molded body to be formed can be predicted more accurately.

[0159] The prediction method according to aspect B9 of the present invention is the prediction method according to any one of aspects B1 to B8 above, further comprising a molded body analysis parameter calculation step of calculating molded body analysis parameters including a tensile modulus and a shear modulus in one or more directions of a composite material including short fibers and a matrix resin based on an orientation coefficient representing the fiber orientation of the molded body to be molded, and in the output step, input information including the molded body analysis parameters calculated in the molded body analysis parameter calculation step is input to the trained model.

[0160] According to the above aspect, by including parameters for analyzing the molded body calculated based on the orientation coefficient in the input of the trained model, the analysis results of the molded body to be formed can be predicted more accurately.

[0161] The prediction method according to aspect B10 of the present invention is a prediction method according to any one of aspects B1 to B9 above, further comprising a training data acquisition step of acquiring training data including information representing the shape, material, and modeling conditions of a molded body to be molded by a three-dimensional printer, and information representing the mechanical properties of the molded body, and a trained model generation step of generating the trained model by supervised machine learning using the training data.

[0162] According to the above aspect, it is possible to generate a trained model that can reduce the processing time and processing load required to analyze a molded body formed by a three-dimensional printer.

[0163] The trained model generation method according to aspect B11 of the present invention includes an acquisition step of acquiring training data including information representing the shape, material, and modeling conditions of a molded body to be formed by a three-dimensional printer, and information representing the mechanical properties of the molded body, and a generation step of generating a trained model by supervised machine learning using the training data, in which the information representing the shape, material, and modeling conditions of the molded body is used as input and information representing the mechanical properties of the molded body is used as output.

[0164] According to the above aspect, it is possible to generate a trained model that can reduce the processing time and processing load required to analyze a molded body formed by a three-dimensional printer.

[0165] The trained model generation method according to aspect B12 of the present invention is the trained model generation method described in aspect B11 above, and further includes a training data generation process for obtaining results of a simulation of the molding of a molded body by a three-dimensional printer using information representing the shape, material, and molding conditions of the molded body to be molded by the three-dimensional printer, and generating the training data including the information and the obtained results.

[0166] According to the above aspect, by using the results of a modeling simulation as training data for machine learning, the predictive accuracy of the trained model can be improved.

[0167] (Summary C) The prediction program according to aspect C1 of the present invention is a prediction program for causing a computer to function as a prediction device, and causes the computer to function as an acquisition unit that acquires information representing the shape, material, and molding conditions of a molded body to be molded, and an output unit that outputs information representing the mechanical properties of the molded body to be molded, obtained by inputting the information acquired by the acquisition unit into a trained model generated by machine learning, which takes as input information representing the shape, material, and molding conditions of a molded body to be molded by a 3D printer and outputs information representing the mechanical properties of the molded body.

[0168] According to the above aspect, it is possible to reduce the processing time and processing load required for analyzing a molded body formed by a three-dimensional printer.

[0169] The prediction program according to aspect C2 of the present invention is the prediction program described in aspect C1 above, wherein the information output by the output unit includes at least one of the temperature distribution, deformation amount, principal stress, shear stress, von Mises stress, normal stress, shear strain, and normal strain of the molded body.

[0170] According to the above aspect, the time and processing load required to analyze at least one of the temperature distribution, deformation amount, principal stress, shear stress, von Mises stress, normal stress, shear strain, and normal strain of the molded body can be reduced.

[0171] The prediction program of aspect C3 of the present invention is the prediction program described in aspect C2 above, in which the information output by the output unit includes information indicating the deformation amount of the molded body, the acquisition unit acquires information indicating the deformation allowance of the molded body set by the user, and the output unit outputs the comparison result between the deformation amount of the molded body and the deformation allowance.

[0172] According to the above aspect, the user can know whether the amount of deformation of the molded body to be shaped is within the allowable range.

[0173] The prediction program according to aspect C4 of the present invention is a prediction program described in any one of aspects C1 to C3 above, wherein the acquisition unit acquires information representing the external forces to be applied to the molded body that is the object to be molded, and the output unit outputs information obtained by inputting the information representing the external forces acquired by the acquisition unit into a trained model that takes as input the information representing the external forces to be applied to the molded body and outputs information representing changes in the molded body.

[0174] According to the above aspect, it is possible to reduce the processing time and processing load required to predict a change in a molded body due to an external force applied to the molded body.

[0175] A prediction program according to aspect C5 of the present invention is the prediction program according to any one of aspects C1 to C4, wherein the printing conditions include at least one of a printing temperature, an atmosphere temperature, a table temperature, and a printing width.

[0176] According to the above aspect, it is possible to reduce the processing time and processing load required to analyze a molded body formed under a forming condition including at least one of the forming temperature, the ambient temperature, the table temperature, and the forming width.

[0177] A prediction program according to aspect C6 of the present invention is a prediction program according to any one of aspects C1 to C5 above, wherein the information representing the material includes information indicating at least one of the type of material, information indicating the mechanical properties of the material, and information indicating the fiber content.

[0178] According to the above aspect, it is possible to reduce the processing time and processing load required to analyze a molded body made from a material represented by information indicating at least one of the type of material, information indicating the mechanical properties of the material, and fiber content.

[0179] A prediction program according to aspect C7 of the present invention is a prediction program according to any one of aspects C1 to C6 above, wherein the acquisition unit refers to a database that stores material types in correspondence with mechanical property information indicating the mechanical properties of the materials, acquires mechanical property information corresponding to the material represented by the information acquired by the acquisition unit, and inputs input information including the acquired mechanical property information into the trained model.

[0180] According to the above aspect, by using the mechanical property information of the material obtained by referring to the database as input to the trained model, the analysis results of the molded body to be formed can be predicted with greater accuracy.

[0181] The prediction program according to aspect C8 of the present invention is a prediction program according to any one of aspects C1 to C7 above, wherein the acquisition unit acquires, as information representing the material of the molded body, information indicating the mechanical properties of the short fibers contained in a composite material comprising short fibers and a matrix resin, information indicating the mechanical properties of the matrix resin, and the tensile modulus of the composite material, and further includes an orientation coefficient calculation unit that calculates an orientation coefficient representing the fiber orientation of the molded body using the tensile modulus acquired by the acquisition unit, and the output unit inputs input information including the orientation coefficient calculated by the orientation coefficient calculation unit to the trained model.

[0182] According to the above aspect, by including the orientation coefficient calculated using the tensile modulus of the composite material in the input of the trained model, the analysis results of the molded body to be formed can be predicted more accurately.

[0183] The prediction program of aspect C9 of the present invention is a prediction program described in any one of aspects C1 to C8 above, further comprising a molded body analysis parameter calculation unit that calculates molded body analysis parameters including a tensile modulus and a shear modulus in one or more directions of a composite material containing short fibers and a matrix resin based on an orientation coefficient representing the fiber orientation of the molded body that is the object to be molded, and the output unit inputs input information including the molded body analysis parameters calculated by the molded body analysis parameter calculation unit to the trained model.

[0184] According to the above aspect, by including parameters for analyzing the molded body calculated based on the orientation coefficient in the input of the trained model, the analysis results of the molded body to be formed can be predicted more accurately.

[0185] The prediction program according to aspect C10 of the present invention is a prediction program according to any one of aspects C1 to C9 above, further comprising a teacher data acquisition unit that acquires teacher data including information representing the shape, material, and molding conditions of a molded body to be molded by a 3D printer, and information representing the mechanical properties of the molded body, and a trained model generation unit that generates the trained model by supervised machine learning using the teacher data.

[0186] According to the above aspect, it is possible to generate a trained model that can reduce the processing time and processing load required to analyze a molded body formed by a three-dimensional printer.

[0187] The trained model generation program of aspect C11 of the present invention is a trained model generation program for causing a computer to function as a trained model generation device, and causes the computer to function as: an acquisition unit that acquires training data including information representing the shape, material, and modeling conditions of a molded body to be formed by a 3D printer, and information representing the mechanical properties of the molded body; and a generation unit that generates a trained model through supervised machine learning using the training data, which takes as input information representing the shape, material, and modeling conditions of the molded body and outputs information representing the mechanical properties of the molded body.

[0188] According to the above aspect, it is possible to generate a trained model that can reduce the processing time and processing load required to analyze a molded body formed by a three-dimensional printer.

[0189] The trained model generation program of aspect C12 of the present invention is the trained model generation program described in aspect C11 above, and further includes a training data generation unit that acquires the results of a simulation of the formation of a molded body by a three-dimensional printer using information representing the shape, material, and formation conditions of the molded body to be formed by the three-dimensional printer, and generates the training data including the information and the acquired results.

[0190] According to the above aspect, by using the results of a modeling simulation as training data for machine learning, the predictive accuracy of the trained model can be improved. [Explanation of symbols]

[0191] 1 Design support system 10, 10B Prediction Device 11 Acquisition Department 12 Simulation execution unit 13 Orientation coefficient calculation section 14. Parameter calculation section for compact analysis 16 Teacher data acquisition section 17 Trained model generation unit 18 Teacher data generation unit 110, 110B, 210 control unit 111 Estimation Phase Execution Unit 112 Learning Phase Execution Unit 120, 120B, 220 storage section 130, 230 Communications Department 140, 240 input section 150, 250 output section 20 User terminal 21 Application execution unit 30 3D printer

Claims

1. an acquisition unit that acquires information representing the shape, material, and modeling conditions of a molded body by a three-dimensional printer; an output unit that inputs the information acquired by the acquisition unit into a trained model generated by machine learning, the trained model having input information representing the shape, material, and modeling conditions of a molded body to be modeled by a three-dimensional printer and output information representing the mechanical properties of the molded body; and A prediction system comprising:

2. The information output by the output unit includes at least one of a temperature distribution, a deformation amount, a principal stress, a shear stress, a von Mises stress, a normal stress, a shear strain, and a normal strain of the molded body. The prediction system of claim 1 .

3. The information output by the output unit includes information indicating a deformation amount of the molded body, the acquisition unit acquires information indicating a deformation tolerance of the molded body set by a user, the output unit outputs a comparison result between the deformation amount of the molded body and the deformation allowance. The prediction system of claim 2 .

4. the acquisition unit acquires information representing an external force to be applied to the molded body that is the modeling target, The output unit information representing an external force applied to a molded body is used as input information, and information representing a change in the molded body is used as output information. Information obtained by inputting the information representing the external force acquired by the acquisition unit into a trained model generated by machine learning is output. The prediction system according to claim 1 or 2.

5. The modeling conditions include at least one of a modeling temperature, an ambient temperature, a table temperature, and a modeling width. The prediction system according to claim 1 or 2.

6. The information representing the material includes information representing at least one of a type of the material, information representing mechanical properties of the material, and information representing a fiber content of the material. The prediction system according to claim 1 or 2.

7. The acquisition unit refers to a database that stores material types and mechanical property information indicating the mechanical properties of the materials in association with each other, acquires mechanical property information corresponding to the material represented by the information acquired by the acquisition unit, and inputs input information including the acquired mechanical property information into the trained model. The prediction system according to claim 1 or 2.

8. the acquisition unit acquires, as information representing the material of the molded body, information indicating mechanical properties of short fibers contained in a composite material including short fibers and a matrix resin, information indicating mechanical properties of the matrix resin, and a tensile modulus of elasticity of the composite material; An orientation coefficient calculation unit that calculates an orientation coefficient representing the fiber orientation of the molded body using the tensile modulus acquired by the acquisition unit, The output unit inputs input information including the orientation coefficient calculated by the orientation coefficient calculation unit to the trained model. The prediction system according to claim 1 or 2.

9. a molded body analysis parameter calculation unit that calculates molded body analysis parameters, including a tensile modulus and a shear modulus in one or more directions of a composite material including short fibers and a matrix resin, based on an orientation coefficient that represents the fiber orientation of the molded body that is the object to be shaped; The output unit inputs input information including the molded body analysis parameters calculated by the molded body analysis parameter calculation unit to the trained model. The prediction system according to claim 1 or 2.

10. a training data acquisition unit that acquires training data including information representing the shape, material, and modeling conditions of a molded body to be modeled by a three-dimensional printer, and information representing the mechanical properties of the molded body; and a trained model generation unit that generates the trained model by supervised machine learning using the training data. The prediction system according to claim 1 or 2.

11. an acquisition unit that acquires training data including information representing the shape, material, and modeling conditions of a molded body to be modeled by a three-dimensional printer, and information representing mechanical properties of the molded body; A generation unit that generates a trained model by supervised machine learning using the training data, with information representing the shape, material, and molding conditions of the molded body as input and information representing the mechanical properties of the molded body as output; A trained model generation system comprising:

12. and a training data generation unit that acquires the results of a simulation of molding of a molded body by a three-dimensional printer using information representing the shape, material, and molding conditions of the molded body to be molded by the three-dimensional printer, and generates the training data including the information and the acquired results. The trained model generation system according to claim 11.

13. an acquisition step of acquiring information representing the shape, material, and modeling conditions of a molded body to be modeled; an output process in which information representing the mechanical properties of the molded body to be modeled is input to a trained model generated by machine learning, the trained model receiving the information acquired in the acquisition process and outputting information representing the mechanical properties of the molded body; and A forecasting method including:

14. an acquisition process for acquiring training data including information representing the shape, material, and modeling conditions of a molded body to be modeled by a three-dimensional printer, and information representing the mechanical properties of the molded body; A generation process of generating a trained model by supervised machine learning using the training data, in which information representing the shape, material, and molding conditions of a molded body is input and information representing the mechanical properties of the molded body is output; A method for generating a trained model including:

15. A prediction program for causing a computer to function as a prediction device, the program comprising: an acquisition unit that acquires information representing the shape, material, and modeling conditions of a molded body that is a modeling target; an output unit that inputs information representing the shape, material, and modeling conditions of a molded body to be modeled by a three-dimensional printer and outputs information representing the mechanical properties of the molded body, the information being obtained by inputting the information acquired by the acquisition unit into a trained model generated by machine learning, the trained model having input information representing the shape, material, and modeling conditions of the molded body to be modeled by a three-dimensional printer and outputting information representing the mechanical properties of the molded body; A prediction program to function as a

16. A trained model generation program for causing a computer to function as a trained model generation device, the computer comprising: an acquisition unit that acquires training data including information representing the shape, material, and modeling conditions of a molded body to be modeled by a three-dimensional printer, and information representing mechanical properties of the molded body; A generation unit that generates a trained model by supervised machine learning using the training data, with information representing the shape, material, and molding conditions of the molded body as input and information representing the mechanical properties of the molded body as output; A trained model generation program to function as a

Images