Estimation method of physical quantity relating to molding of molded body

Abstract

To provide a technique capable of estimating molding conditions or a discharge result with high accuracy.SOLUTION: Disclosed is an estimation method of a physical quantity relating to molding of a molded body by using a molding apparatus for molding a molded body by discharging a molding material from a nozzle, wherein the physical quantity includes molding conditions of a molded body or a discharge result of a molding material discharged from the nozzle, and according to a relational expression obtained based on a parameter relating to a viscosity of the molding material, molding conditions, and a discharge result, molding conditions corresponding to the discharge result or a discharge result corresponding to molding conditions are estimated.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present disclosure relates to a method for estimating a physical quantity related to the formation of a modeled object. [Background technology]

[0002] For example, Patent Document 1 discloses an injection molding condition generation system that generates injection molding conditions based on a relational equation that is generated based on data that links the material property values ​​of a resin material, the injection molding conditions to be input to an injection molding machine, and quality parameters related to the quality of the molded product. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2022-141495 A Summary of the Invention [Problem to be solved by the invention]

[0004] When estimating the modeling conditions that correspond to the discharge results in the modeling of a modeled object, or the discharge results that correspond to the modeling conditions, the viscosity of the modeling material is an important parameter. However, the viscosity of the modeling material is intricately related to the temperature, shear rate, and pressure of the modeling material, and it is difficult to obtain an accurate value. Therefore, there was a problem in that it was difficult to estimate the modeling conditions or the discharge results with high accuracy. [Means for solving the problem]

[0005] According to a first aspect of the present disclosure, there is provided a method for estimating a physical quantity related to the formation of a modeled object, the method being a method for estimating a physical quantity related to the formation of a modeled object by a modeling device that forms a modeled object by discharging a modeling material from a nozzle, the physical quantity including modeling conditions of the modeled object or a discharge result of the modeling material discharged from the nozzle, and estimating the modeling conditions corresponding to the discharge result or the discharge result corresponding to the modeling conditions according to a relational expression obtained based on a parameter related to a viscosity of the modeling material, the modeling conditions, and the discharge result. [Brief description of the drawings]

[0006] [Figure 1] FIG. 1 is an explanatory diagram showing a schematic configuration of a physical quantity estimation system. [Diagram 2] FIG. 2 is an explanatory diagram showing a schematic configuration of a molding apparatus. [Diagram 3] FIG. [Figure 4] Schematic plan view of a barrel. [Diagram 5] 4 is a process diagram of a physical quantity estimation process executed by the physical quantity estimation system. [Figure 6] FIG. 2 is a diagram for explaining machine learning in the first embodiment. [Figure 7] A process diagram for acquiring learning data. [Figure 8] FIG. 13 shows the evaluation results of a trained model using training data. [Figure 9] A diagram showing the evaluation results of a trained model based on evaluation data. [Figure 10] A plot of the true and estimated flow velocities at each data point in the training data. [Figure 11] A plot of the true and estimated flow velocities at each data point of the evaluation data. [Figure 12] FIG. 11 is a diagram for explaining machine learning in the second embodiment. [Figure 13] FIG. 11 is a process diagram of a learning data acquisition process in the second embodiment. [Figure 14]FIG. 13 shows the evaluation results of a trained model using training data. [Figure 15] A diagram showing the evaluation results of a trained model based on evaluation data. [Figure 16] A plot of the true and estimated flow velocities at each data point in the training data. [Figure 17] A plot of the true and estimated flow velocities at each data point of the evaluation data. [Figure 18] FIG. 13 is a diagram for explaining machine learning in the third embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0007] A. First embodiment: FIG. 1 is an explanatory diagram showing a schematic configuration of a physical quantity estimation system 10. The physical quantity estimation system 10 includes a modeling apparatus 100 and a physical quantity estimation apparatus 200. The modeling apparatus 100 models a modeled object by discharging a modeling material from a nozzle. In this embodiment, the modeling apparatus 100 is a three-dimensional modeling apparatus that models a three-dimensional object. The modeling apparatus 100 is configured to be able to communicate with the physical quantity estimation apparatus 200 via a network NT. The network NT may be, for example, a LAN, a WAN, or the Internet. Note that the modeling apparatus 100 may be connected to the physical quantity estimation apparatus 200 without going through the network NT.

[0008] In this embodiment, the physical quantity estimation device 200 estimates the discharge result of the modeling material discharged from the nozzle, which corresponds to the modeling conditions of the modeled object, by using machine learning. Here, the modeling conditions of the modeled object and the discharge result of the modeling material discharged from the nozzle are collectively referred to as physical quantities related to the modeling of the modeled object. In other words, the physical quantity estimation device 200 estimates physical quantities related to the modeling of the modeled object by using machine learning. Details of the physical quantity estimation device 200 will be described later.

[0009] FIG. 2 is an explanatory diagram showing a schematic configuration of the modeling apparatus 100. FIG. 2 shows arrows representing mutually orthogonal X, Y, and Z directions. The X and Y directions are parallel to the horizontal plane. The Z direction is parallel to the vertical direction. The X, Y, and Z directions in FIG. 2 and the X, Y, and Z directions in other figures indicate the same directions. When specifying a direction, positive and negative signs are used in combination to indicate the direction, with the positive direction indicated by the arrow being "+" and the negative direction opposite to the direction indicated by the arrow being "-".

[0010] The modeling apparatus 100 includes a control unit 110, a modeling unit 120, a stage 130, and a moving mechanism 140.

[0011] The control unit 110 controls the operation of the entire modeling apparatus. The control unit 110 is configured by a computer equipped with one or more processors, a memory, and an input / output interface for inputting and outputting signals from and to the outside. The control unit 110 performs various functions, such as a function of executing a modeling process for modeling a three-dimensional object, by the processor executing programs and instructions loaded onto the main storage device. Note that the control unit 110 may be realized by a configuration combining multiple circuits for realizing at least a part of each function, instead of being configured by a computer.

[0012] The modeling unit 120, under the control of the control unit 110, discharges the modeling material, which is a pasty modeling material obtained by plasticizing a solid material, onto a stage 130 that serves as a base for a three-dimensional object. The modeling unit 120 includes a material supply unit 301, a plasticizing unit 302, and a discharging unit 303.

[0013] The material supply unit 301 supplies the material for generating the modeling material to the plasticization unit 302. The material supply unit 301 is, for example, configured by a hopper. The material supply unit 301 stores pellet-shaped or powder-shaped materials. For example, thermoplastic resins such as polypropylene resin (PP), polyethylene resin (PE), and polyacetal resin (POM) are used as the material. The material supply unit 301 supplies the material to the plasticization unit 302 via a communication passage 315.

[0014] The plasticizing unit 302 plasticizes at least a part of the material supplied from the material supply unit 301, generates a pasty modeling material having fluidity, and guides it to the discharge unit 303. Here, "plasticization" is a concept that includes melting, and means changing from a solid to a fluid state. Specifically, in the case of a material that undergoes glass transition, plasticization means raising the temperature of the material to the glass transition point or higher. In the case of a material that does not undergo glass transition, plasticization means raising the temperature of the material to the melting point or higher. The plasticizing unit 302 includes a flat screw 320, a screw case 330, a drive motor 340, and a barrel 350.

[0015] The flat screw 320 is housed in a screw case 330. The upper surface side of the flat screw 320 is connected to a drive motor 340. The flat screw 320 rotates in the screw case 330 by a rotational driving force generated by the drive motor 340. The axial direction of the rotation axis RX of the flat screw 320 is along the Z direction. The rotation speed of the flat screw 320 is controlled by the control unit 110 controlling the rotation speed of the drive motor 340. The control unit 110 controls the rotation speed of the drive motor 340 so that the rotation speed of the flat screw 320 becomes a value set by the user. The flat screw 320 may be driven by the drive motor 340 via a reducer. The flat screw 320 has a groove forming surface 321 on which a groove 323 is formed on the opposite side to the upper surface in the direction along the rotation axis RX. A specific configuration of the flat screw 320 will be described later. The flat screw 320 is also called a rotor or a screw. Hereinafter, the flat screw 320 will also be referred to simply as the screw 320.

[0016] The barrel 350 is installed on the -Z direction side of the screw 320. An opposing surface 351, which is the upper surface of the barrel 350, faces the groove forming surface 321 of the screw 320. A communication hole 352 is formed in the center of the barrel 350, which penetrates the barrel 350 along the Z direction and forms a part of a flow path 362 through which the modeling material flows. A heater is embedded in the barrel 350 to heat the material supplied to the groove 323 of the screw 320. In this embodiment, four rod-shaped heaters are arranged along the Y direction. Each heater is arranged below the opposing surface 351. Of the four heaters, two heaters arranged at a position away from the communication hole 352 are called third heaters 353, and two heaters arranged closer to the communication hole 352 than the third heater 353 are called fourth heaters 354. The third heater 353 and the fourth heater 354 are also collectively called second heaters. The temperature of each heater is controlled by the control unit 110. In other words, the temperatures of the third heater 353 and the fourth heater 354 are individually controlled by the control unit 110. The control unit 110 controls the temperatures of the heaters so that the temperatures of the heaters become the values ​​set by the user.

[0017] FIG. 3 is a perspective view showing a schematic configuration of the screw 320. The screw 320 has a generally cylindrical shape in which the length in the direction along the rotation axis RX is smaller than the length in the direction perpendicular to the rotation axis RX. A spiral groove 323 is formed on the groove forming surface 321, centered on the central portion 322. The groove 323 is connected to a material inlet 324 formed on the side surface of the screw 320. The groove 323 forms a part of the flow path 362. The material supplied from the material supply unit 301 is supplied to the groove 323 through the material inlet 324. The groove 323 is formed by being separated by a convex portion 325. FIG. 3 shows an example in which three grooves 323 are formed, but the number of grooves 323 may be one or more. The groove 323 is not limited to a spiral shape, but may be a spiral shape or an involute curve shape, or may be a shape extending from the central portion 322 to the outer periphery in an arc.

[0018] 4 is a schematic plan view of barrel 350. A plurality of guide grooves 355 are formed around communicating hole 352 in opposing surface 351. One end of each guide groove 355 is connected to communicating hole 352, and extends in a spiral shape from communicating hole 352 toward the outer periphery of opposing surface 351. Note that one end of guide groove 355 does not have to be connected to communicating hole 352. Also, barrel 350 does not have to have guide groove 355 formed therein.

[0019] The material supplied to the groove 323 of the screw 320 flows along the groove 323 while being plasticized in the groove 323 by the rotation of the screw 320 and the heat of the heater, and is guided to the center part 322 of the screw 320 as a modeling material. The pasty modeling material that has flowed into the center part 322 and exhibits fluidity is supplied to the discharge part 303 through the communication hole 352. Note that it is not necessary for all types of substances constituting the modeling material to be plasticized in the plasticizing part 302. It is sufficient that the modeling material is converted into a state having fluidity as a whole by plasticizing at least some types of substances constituting the modeling material.

[0020] The discharge unit 303 discharges the modeling material. The discharge unit 303 includes a nozzle 361, a flow path 362, a discharge control unit 363, a pressure sensor 364, a flow rate sensor 365, and a first heater 356.

[0021] The nozzle 361 is connected to the communication hole 352 of the barrel 350 through a flow path 362. The nozzle 361 discharges the modeling material generated in the plasticizing section 302 from a nozzle opening 366 at the tip of the nozzle 361 toward the stage 130.

[0022] The discharge control unit 363 includes a discharge adjustment unit 367 that opens and closes the flow path 362, and a suction unit 368 that sucks in the modeling material and temporarily stores it.

[0023] The discharge adjustment unit 367 is provided in the flow path 362 and changes the opening degree of the flow path 362 by rotating in the flow path 362. In this embodiment, the discharge adjustment unit 367 is configured by a butterfly valve. The discharge adjustment unit 367 is driven by a first driving unit 369 under the control of the control unit 110. The first driving unit 369 is configured by, for example, a stepping motor. The control unit 110 can adjust the flow rate of the modeling material flowing from the plasticizing unit 302 to the nozzle 361, that is, the discharge amount of the modeling material discharged from the nozzle 361, by controlling the rotation angle of the butterfly valve using the first driving unit 369. The discharge adjustment unit 367 can adjust the discharge amount of the modeling material and can control the on / off of the outflow of the modeling material.

[0024] The suction unit 368 is connected between the discharge adjustment unit 367 and the nozzle opening 366 in the flow path 362. When the discharge of the modeling material from the nozzle 361 stops, the suction unit 368 temporarily sucks the modeling material in the flow path 362, thereby suppressing the tailing phenomenon in which the modeling material hangs like a string from the nozzle opening 366. In this embodiment, the suction unit 368 is configured with a plunger. The suction unit 368 is driven by a second drive unit 370 under the control of the control unit 110. The second drive unit 370 is configured with, for example, a stepping motor or a rack-and-pinion mechanism that converts the rotational force of the stepping motor into the translational motion of the plunger.

[0025] The pressure sensor 364 is provided in the flow path 362 upstream of the discharge adjustment unit 367. The pressure sensor 364 measures the pressure of the modeling material in the flow path 362, i.e., the internal pressure of the flow path 362. The value of the internal pressure of the flow path 362 measured by the pressure sensor 364 is transmitted to the control unit 110.

[0026] The flow rate sensor 365 is provided on the side surface of the nozzle 361. The flow rate sensor 365 measures the flow rate of the modeling material discharged from the nozzle opening 366. The value of the flow rate of the modeling material measured by the flow rate sensor 365 is transmitted to the control unit 110. The flow rate of the modeling material is also called the discharge amount of the modeling material.

[0027] The first heater 356 is provided near the nozzle 361. The first heater 356 heats the nozzle 361. Note that in this embodiment, the discharge unit 303 does not necessarily have to include the first heater 356.

[0028] The stage 130 is disposed at a position facing the nozzle opening 366 of the nozzle 361. The modeling apparatus 100 ejects a modeling material from the nozzle 361 onto a modeling surface 131, which is the upper surface of the stage 130, to stack modeling layers, thereby forming a three-dimensional object.

[0029] The moving mechanism 140 changes the relative position between the nozzle 361 and the stage 130. In this embodiment, the moving mechanism 140 moves the stage 130 with respect to the nozzle 361, whose position is fixed. The change in the relative position of the nozzle 361 with respect to the stage 130 is also simply referred to as the movement of the nozzle 361. The moving mechanism 140 is configured with a three-axis positioner that moves the stage 130 in three axial directions, that is, the X, Y, and Z directions, by the driving forces of three motors. Each motor of the moving mechanism 140 is driven under the control of the control unit 110. Note that the moving mechanism 140 may not be configured to move the stage 130, but may be configured to move the nozzle 361 with the position of the stage 130 fixed. Also, the moving mechanism 140 may be configured to move both the stage 130 and the nozzle 361.

[0030] The physical quantity estimation device 200 will be described below with reference to FIG. 1. The physical quantity estimation device 200 is configured as a computer including a processing unit 201, a storage unit 202, a communication unit 203, a display unit 204, and an input unit 205. The processing unit 201 includes one or more processors. The processing unit 201 controls the operation of the physical quantity estimation device 200 by executing a program stored in the storage unit 202. The storage unit 202 includes a main storage unit such as a RAM and an auxiliary storage unit such as a hard disk drive. The communication unit 203 includes a communication circuit for controlling communication with the molding apparatus 100. The display unit 204 displays various screens. The display unit 204 includes, for example, a liquid crystal display. The input unit 205 accepts input from a user. The input unit 205 includes, for example, a mouse and a keyboard. The display unit 204 and the input unit 205 may be integrated as a touch panel.

[0031] The processing unit 201 includes a learning data acquisition unit 210, a model generation unit 220, and an estimation unit 230. The learning data acquisition unit 210, the model generation unit 220, and the estimation unit 230 are realized by the processing unit 201 executing a program stored in the storage unit 202. These may also be realized by circuits.

[0032] The learning data acquisition unit 210 acquires data from the molding apparatus 100 via the communication unit 203, and generates learning data using the acquired data. Here, the learning data is data used as learning data when generating a trained model by machine learning.

[0033] The model generation unit 220 generates a trained model by performing machine learning using the training data.

[0034] The estimation unit 230 estimates the discharge result by inputting the modeling conditions to the learned model. The estimation unit 230 causes the display unit 204 to display the estimated discharge result.

[0035] The storage unit 202 includes a model storage unit 260. The model storage unit 260 stores the trained model generated by the model generation unit 220.

[0036] 5 is a process diagram of the physical quantity estimation process executed by the physical quantity estimation system 10. The physical quantity estimation process includes a learning process and an estimation process executed after the learning process. In the learning process, a trained model is generated by machine learning using learning data. In the estimation process, a discharge result corresponding to a modeling condition is estimated using the generated trained model.

[0037] FIG. 6 is a diagram for explaining machine learning in the first embodiment. In this embodiment, the learning data includes the viscosity of the modeling material at two different specific positions on the flow channel 362, the set value of the rotation speed of the screw 320, parameters related to the shape of the flow channel 362, the flow rate of the modeling material discharged from the nozzle 361, and the internal pressure of the flow channel 362. Hereinafter, the parameters related to the shape of the flow channel 362 are also referred to as flow channel parameters. The flow channel parameters are, for example, the inner diameter of the flow channel 362. The viscosity of the modeling material at a specific position on the flow channel 362 is calculated by the Cross-WLF model using the set temperature of the heater and the shear rate at a specific position on the flow channel 362. A method for calculating the viscosity of the modeling material will be described later.

[0038] As shown in FIG. 5, in the learning process, a learning data acquisition process is executed in step S11.

[0039] FIG. 7 is a process diagram of the learning data acquisition process. In step S110, the user sets the temperature of the third heater 353, the temperature of the fourth heater 354, the rotation speed of the screw 320, and the flow path parameters, and operates the modeling apparatus 100 to discharge the modeling material from the nozzle 361. The learning data acquisition unit 210 acquires the above-mentioned four set values, the flow rate of the modeling material measured by the flow sensor 365, and the internal pressure of the flow path 362 measured by the pressure sensor 364, and stores these as one data set in the storage unit 202. In this embodiment, the set value of the rotation speed of the screw 320 and the flow path parameters correspond to the modeling conditions. In addition, the flow rate of the modeling material measured by the flow sensor 365 and the internal pressure of the flow path 362 measured by the pressure sensor 364 correspond to the discharge result. The learning data acquisition unit 210 acquires multiple data sets with different types of materials and the above-mentioned four set values, and stores them in the storage unit 202.

[0040] In step S120, the learning data acquisition unit 210 calculates the average flow rate for each rotation speed of the screw 320 and flow path parameter using a data set having the same rotation speed of the screw 320 and flow path parameters from among multiple data sets stored in the memory unit 202.

[0041] In step S130, the learning data acquisition unit 210 calculates the shear rate of the modeling material at a specific position on the flow channel 362 using the formula (1). · is the shear rate, Q is the average value of the flow rate calculated in step S120, and d is the radius of the flow channel 362 at a specific position on the flow channel 362. Hereinafter, the specific position on the flow channel 362 located near the third heater 353 is also referred to as the first position, and the specific position on the flow channel 362 located near the fourth heater 354 is also referred to as the second position. In step S130, the learning data acquisition unit 210 calculates the shear rate at the first position and the shear rate at the second position. The learning data acquisition unit 210 executes the process of step S130 for all data sets stored in the storage unit 202.

[0042]

number

[0043] In step S140, the learning data acquisition unit 210 calculates the viscosity of the modeling material at a specific position on the flow path 362 using the Cross-WLF model, which is an equation for calculating the viscosity of the modeling material. The Cross-WLF model is a model for calculating the viscosity of a thermoplastic resin. In this embodiment, the viscosity of the modeling material calculated using the Cross-WLF model is also referred to as a parameter related to the viscosity of the modeling material. The Cross-WLF model is expressed by the formula (2). Here, η is the viscosity of the modeling material, γ · is the shear rate, T is the temperature, P is the pressure, η0 is the zero shear viscosity, τ * is the inflection point with respect to shear rate, and n is the gradient with respect to shear rate. η0 is calculated using equation (3). T * and A2 are calculated by equations (4) and (5), respectively. Here, D1 is the average viscosity, A1 is the gradient with respect to temperature, and A ~ 2 and D2 are the inflection points with respect to temperature, and D3 is the pressure dependence.

[0044]

number

number

number

number

[0045] D1, A1, and A included in formulas (2) to (5) ~ 2, D2, n, τ *, D3 are coefficients input to the Cross-WLF model. Hereinafter, the above seven coefficients input to the Cross-WLF model will be referred to as Cross-WLF coefficients. The Cross-WLF coefficients are calculated by fitting the results of measuring the viscosity of the material at various temperatures, shear rates, and pressures using a rheometer with the Cross-WLF model. The value of the Cross-WLF coefficients varies depending on the type of material.

[0046] The learning data acquisition unit 210 adds γ · The viscosity of the modeling material at a specific position on the flow channel 362 is calculated by substituting the shear rate of the modeling material at the specific position on the flow channel 362 calculated in step S130 as γ, the set temperature of the heater located in the vicinity of the specific position on the flow channel 362 as T, the general pressure of the flow channel 362 when the modeling apparatus 100 models an object as P, and the seven Cross-WLF coefficient values. The general pressure of the flow channel 362 when the modeling apparatus 100 models an object is, for example, 1 MPa or 2 MPa. Specifically, the learning data acquisition unit 210 adds γ · The viscosity of the modeling material at the first position is calculated by substituting the shear rate at the first position as V and the set temperature of the third heater 353 as T. The viscosity of the modeling material at the second position is calculated in a similar manner. The learning data acquisition unit 210 executes the process of step S140 for all data sets stored in the storage unit 202. As described above, the learning data acquisition unit 210 acquires a plurality of learning data. The acquired learning data is stored in the storage unit 202.

[0047] In step S12 of FIG. 5, the model generation unit 220 generates a trained model by performing machine learning using the learning data. Specifically, the model generation unit 220 performs supervised learning using the viscosity at the first position, the viscosity at the second position, the rotation speed of the screw 320, and the flow path parameters as explanatory variables, and the flow rate of the modeling material discharged from the nozzle 361 and the internal pressure of the flow path 362 as objective variables to generate a trained model. The model generation unit 220 generates the trained model using a machine learning algorithm such as a neural network, a support vector machine, or a random forest. That is, the model generation unit 220 generates a relational expression for estimating the discharge result corresponding to the modeling condition based on the parameters related to the viscosity of the modeling material, the modeling condition, and the discharge result. In this embodiment, the discharge result corresponds to the teacher data of the machine learning.

[0048] In step S13, the model generation unit 220 stores the trained model generated in step S12 in the storage unit 202. The learning process is performed as described above.

[0049] Next, the estimation process will be described. In step S21, the estimation unit 230 acquires input data to be input to the trained model. The input data is parameters related to the viscosity of the modeling material and modeling conditions, specifically, the viscosity at the first position, the viscosity at the second position, the rotation speed of the screw 320, and the flow path parameters. The viscosity at the first position and the viscosity at the second position are calculated in the same manner as the training data acquisition process described in FIG. 7.

[0050] In step S22, the estimation unit 230 inputs the input data acquired in step S21 into the trained model.

[0051] In step S23, the estimation unit 230 acquires an estimated value of the discharge result output from the trained model. Specifically, the estimation unit 230 acquires an estimated value of the flow rate of the modeling material discharged from the nozzle 361 and an estimated value of the internal pressure of the flow path 362. The estimation unit 230 causes the display unit 204 to display the acquired estimated value of the discharge result. The estimation unit 230 may store the acquired estimated value of the discharge result in the storage unit 202. The physical quantity estimation process is performed as described above.

[0052] The following describes cross-validation of a trained model. The trained model described below is a model generated by executing the learning process of the physical quantity estimation process using training data of three types of materials. In the cross-validation of the trained model, evaluation data of one type of material different from the above-mentioned three types of materials is used, which is acquired by executing the same process as the training data acquisition process.

[0053] FIG. 8 is a diagram showing the evaluation results of the trained model using the learning data of three types of materials. The horizontal axis of FIG. 8 indicates the true value of the flow rate of the modeling material discharged from the nozzle 361, and the vertical axis indicates the estimated value of the flow rate by the trained model. Here, the flow rate of the modeling material discharged from the nozzle 361 is a value converted from the flow rate of the modeling material discharged from the nozzle 361. The true value of the flow rate is a flow rate converted from the flow rate measured by the flow rate sensor 365, and the estimated value of the flow rate is a flow rate converted from the estimated value of the flow rate output from the trained model. In FIG. 8, the true value and the estimated value of the flow rate of each learning data are plotted.

[0054] Figure 9 shows the evaluation results of the trained model using evaluation data for one type of material different from the three types of materials used in the training data. The horizontal and vertical axes in Figure 9 are the same as those in Figure 8. Figure 9 plots the true and estimated values ​​of the flow velocity for each evaluation data.

[0055] Fig. 10 is a plot of the true and estimated values ​​of the flow velocity at each data point of the learning data shown in Fig. 8. Fig. 11 is a plot of the true and estimated values ​​of the flow velocity at each data point of the evaluation data shown in Fig. 9. The horizontal axis of Fig. 10 and Fig. 11 indicates the data number of the learning data and the evaluation data, respectively. The vertical axis of Fig. 10 and Fig. 11 indicates the flow velocity of the modeling material discharged from the nozzle 361.

[0056] In the above-mentioned cross-validation, the coefficient of determination of the evaluation data was 0.6 or higher, confirming that the trained model can accurately estimate the discharge results.

[0057] According to the first embodiment described above, the learning data acquisition unit 210 calculates the viscosity of the modeling material at a specific position on the flow path 362 by using the Cross-WLF model, which is an equation for calculating the viscosity of the modeling material. The estimation unit 230 estimates the discharge result corresponding to the modeling conditions by using a trained model generated by machine learning using the viscosity calculated by the above-mentioned method. Therefore, even if it is difficult to calculate an accurate value of the viscosity of the modeling material, the discharge result corresponding to the modeling conditions can be estimated with high accuracy.

[0058] In the present embodiment, the viscosity of the modeling material at two different specific positions on the flow channel 362 is used as learning data. Therefore, the estimation accuracy of the discharge result can be improved compared to the case where the viscosity of the modeling material at one specific position on the flow channel is used as learning data.

[0059] In addition, in this embodiment, the viscosity of the modeling material is calculated using the Cross-WLF model, which is a model for calculating the viscosity of thermoplastic resin, and therefore, it is possible to improve the estimation accuracy of the discharge result when a model is formed using thermoplastic resin.

[0060] In addition, in the present embodiment, a trained model is generated by machine learning, so that a relational expression for estimating a discharge result corresponding to a modeling condition can be generated in a short time.

[0061] B. Second embodiment: In the second embodiment, the learning data, the modeling conditions, and the parameters related to the viscosity of the modeling material are different from those in the first embodiment. The configuration of each part of the physical quantity estimation system 10 in the second embodiment is the same as that in the first embodiment.

[0062] 12 is a diagram illustrating machine learning in the second embodiment. In the second embodiment, the learning data are the set temperature of the third heater 353, the set temperature of the fourth heater 354, the set value of the rotation speed of the screw 320, the flow path parameters, the Cross-WLF coefficient, the flow rate of the modeling material discharged from the nozzle 361, and the internal pressure of the flow path 362. In the second embodiment, the modeling conditions are the set temperature of the third heater 353, the set temperature of the fourth heater 354, the set value of the rotation speed of the screw 320, and the flow path parameters. In the second embodiment, the parameter related to the viscosity of the modeling material is the Cross-WLF coefficient.

[0063] The physical quantity estimation process in the second embodiment will be described below with reference to Fig. 5. Note that a description of the part in which the same process as the physical quantity estimation process in the first embodiment is executed will be omitted. First, in step S11, a learning data acquisition process is executed.

[0064] FIG. 13 is a process diagram of the learning data acquisition process in the second embodiment. In step S111, the user sets the temperature of the third heater 353, the temperature of the fourth heater 354, the rotation speed of the screw 320, and the flow path parameters, and operates the modeling apparatus 100 to discharge the modeling material from the nozzle 361. The learning data acquisition unit 210 acquires the above-mentioned four set values, the flow rate of the modeling material measured by the flow rate sensor 365, and the internal pressure of the flow path 362 measured by the pressure sensor 364. The learning data acquisition unit 210 stores the above-mentioned six acquired values ​​and the Cross-WLF coefficient corresponding to the type of material used as one data set of learning data in the storage unit 202. The learning data acquisition unit 210 changes the above-mentioned four set values ​​and the type of material to acquire multiple data sets of learning data, and stores them in the storage unit 202.

[0065] 5, the model generation unit 220 generates a trained model by performing machine learning using the training data. Specifically, the model generation unit 220 performs supervised learning using the set temperature of the third heater 353, the set temperature of the fourth heater 354, the rotation speed of the screw 320, the flow path parameters, and the Cross-WLF coefficient as explanatory variables, and the flow rate of the modeling material discharged from the nozzle 361 and the internal pressure of the flow path 362 as objective variables, to generate a trained model.

[0066] 5, the estimation unit 230 acquires input data to be input to the trained model. The input data are parameters related to the modeling conditions and the viscosity of the modeling material, specifically, the set temperature of the third heater 353, the set temperature of the fourth heater 354, the rotation speed of the screw 320, the flow path parameters, and the Cross-WLF coefficient.

[0067] The cross-validation of the trained model in the second embodiment will be described below. The trained model described below is a model generated by executing the learning process of the physical quantity estimation process using training data of three types of materials. In the cross-validation of the trained model, evaluation data of one type of material different from the above-mentioned three types of materials is used, which is acquired by executing the same process as the training data acquisition process.

[0068] Fig. 14 is a diagram showing the evaluation results of the trained model using the training data. The horizontal and vertical axes in Fig. 14 are the same as those in Fig. 8. In Fig. 14, the true and estimated values ​​of the flow velocity of each training data are plotted.

[0069] Fig. 15 is a diagram showing the evaluation results of the trained model using the evaluation data. The horizontal and vertical axes in Fig. 15 are the same as those in Fig. 14. In Fig. 15, the true and estimated values ​​of the flow velocity of each evaluation data are plotted.

[0070] Fig. 16 is a plot of the true and estimated values ​​of the flow velocity at each data point of the learning data shown in Fig. 14. Fig. 17 is a plot of the true and estimated values ​​of the flow velocity at each data point of the evaluation data shown in Fig. 15. The horizontal axes of Fig. 16 and Fig. 17 indicate the data numbers of the learning data and the evaluation data, respectively. The vertical axes of Fig. 16 and Fig. 17 indicate the flow velocity of the modeling material discharged from the nozzle 361.

[0071] In the above-mentioned cross-validation, the coefficient of determination of the evaluation data was 0.6 or higher, confirming that the trained model can accurately estimate the discharge results.

[0072] According to the second embodiment described above, a trained model is generated by machine learning using the modeling conditions, the Cross-WLF coefficients, and the discharge results as learning data, and the discharge results corresponding to the modeling conditions are estimated using the trained model. Therefore, the discharge results corresponding to the modeling conditions can be estimated without calculating the viscosity of the modeling material. Even if it is difficult to calculate the viscosity of the modeling material using the Cross-WLF model, the discharge results corresponding to the modeling conditions can be estimated. Here, a case where it is difficult to calculate the viscosity of the modeling material using the Cross-WLF model is, for example, a case where the average value of the flow rate of the modeling material is significantly different from the actual flow rate.

[0073] C. Third embodiment: In the third embodiment, the physical quantity estimation device 200 estimates, by using machine learning, the modeling conditions of a modeling object, which correspond to the result of the ejection of a modeling material ejected from a nozzle 361. The configurations of the respective parts of the physical quantity estimation system 10 in the third embodiment are the same as those in the second embodiment.

[0074] In the third embodiment, the estimation unit 230 estimates the modeling conditions by inputting the discharge result to a trained model. The estimation unit 230 causes the display unit 204 to display the estimated modeling conditions.

[0075] 18 is a diagram for explaining machine learning in the third embodiment. The learning data, the modeling conditions, and the parameters related to the viscosity of the modeling material in the third embodiment are the same as those in the second embodiment.

[0076] The physical quantity estimation process in the third embodiment will be described below with reference to Fig. 5. Note that a description of the part in which the same process as the physical quantity estimation process in the second embodiment is executed will be omitted.

[0077] In step S12, the model generation unit 220 generates a trained model by performing machine learning using the learning data. Specifically, the model generation unit 220 performs supervised learning using the flow rate of the modeling material discharged from the nozzle 361, the internal pressure of the flow path 362, and the Cross-WLF coefficient as explanatory variables, and the set temperature of the third heater 353, the set temperature of the fourth heater 354, the rotation speed of the screw 320, and the flow path parameters as objective variables to generate a trained model. That is, the model generation unit 220 generates a relational expression for estimating the modeling conditions corresponding to the discharge result based on the parameters related to the viscosity of the modeling material, the modeling conditions, and the discharge result. In the third embodiment, the modeling conditions correspond to the teacher data of the machine learning.

[0078] In step S21, the estimation unit 230 acquires input data to be input to the trained model. The input data are parameters related to the discharge result and the viscosity of the modeling material, specifically, the flow rate of the modeling material discharged from the nozzle 361, the internal pressure of the flow path 362, and the Cross-WLF coefficient.

[0079] In step S23, the estimation unit 230 acquires estimated values ​​of the modeling conditions output from the trained model. Specifically, the estimation unit 230 acquires an estimated value of the set temperature of the third heater 353, an estimated value of the set temperature of the fourth heater 354, an estimated value of the rotation speed of the screw 320, and an estimated value of the flow path parameters. The estimation unit 230 displays the acquired estimated values ​​of the modeling conditions on the display unit 204. The estimation unit 230 may store the acquired estimated values ​​of the modeling conditions in the memory unit 202.

[0080] According to the third embodiment described above, a trained model is generated by machine learning using the modeling conditions, the Cross-WLF coefficients, and the discharge results as learning data, and the modeling conditions corresponding to the discharge results are estimated using the trained model. Therefore, it is possible to estimate the modeling conditions corresponding to the discharge results without calculating the viscosity of the modeling material.

[0081] D. Other embodiments: (D-1) In the above embodiment, the material is a thermoplastic resin. However, the material may be a metal. In this case, the equation for calculating the viscosity of the molding material is, for example, equation (6). Here, Cu is D 60 / D 10 D 10 , D 50 , D 60 are the particle sizes at 10 vol%, 50 vol%, and 60 vol%, respectively. Also, η is the viscosity of the molding material, E a is the activation energy, R is the gas constant, T is the temperature, η0 is the zero shear viscosity, λ is the time constant, and a and n are parameters determined from experiments.

[0082]

number

[0083] (D-2) In the second and third embodiments, the modeling conditions are the set temperature of the third heater 353, the set temperature of the fourth heater 354, the set value of the rotation speed of the screw 320, and the flow path parameters. In contrast to this, the modeling conditions may further include the set temperature of the first heater 356. In addition, the temperature of the first heater 356 may be used as learning data.

[0084] (D-3) In the first embodiment, the viscosity of the modeling material at the first position and the viscosity of the modeling material at the second position are used as learning data. In contrast to this, the viscosity of the modeling material at three or more different specific positions on the flow path 362 may be used as learning data.

[0085] (D-4) In the second and third embodiments, the set temperature of the third heater 353 and the set temperature of the fourth heater 354 are used as learning data. In contrast, the molding apparatus 100 may be provided with three or more heaters, and the set temperature of each heater may be used as learning data.

[0086] (D-5) In the above embodiment, the discharge result is the flow rate of the modeling material discharged from the nozzle 361 and the internal pressure of the flow path 362. In contrast, the discharge result may include the discharge speed of the modeling material from the nozzle 361 and the amount of fluctuation in the discharge speed of the modeling material from the nozzle 361. The amount of fluctuation in the discharge speed of the modeling material from the nozzle 361 is the difference between the maximum value and the average value of the discharge speed, the difference between the minimum value and the average value, etc.

[0087] (D-6) In the above embodiment, a plurality of heaters are provided along opposing surface 351 in a direction perpendicular to communication hole 352. In contrast, a plurality of heaters may be provided along communication hole 352 and flow path 362 in discharge portion 303 in a direction perpendicular to flow path 362.

[0088] (D-7) In the above embodiment, the ejection result corresponding to the modeling condition, or the relational expression for estimating the modeling condition corresponding to the ejection result, is generated by machine learning using the parameters related to the viscosity of the modeling material, the modeling condition, and the ejection result as learning data. In contrast, the above-mentioned relational expression may be generated by a method other than machine learning, for example, by using a program for deriving the above-mentioned relational expression.

[0089] (D-8) In the above embodiment, the modeling apparatus 100 is a three-dimensional modeling apparatus. However, the modeling apparatus 100 may be an injection molding apparatus that ejects a modeling material from the nozzle 361 into a mold to form a model.

[0090] E. Other forms: The present disclosure is not limited to the above-mentioned embodiment, and can be realized in various forms without departing from the spirit of the present disclosure. For example, the present disclosure can be realized in the following forms. The technical features in the above-mentioned embodiments corresponding to the technical features in each form described below can be appropriately replaced or combined in order to solve some or all of the problems of the present disclosure, or to achieve some or all of the effects of the present disclosure. Furthermore, if the technical feature is not described as essential in this specification, it can be appropriately deleted.

[0091] (1) According to one aspect of the present disclosure, there is provided a method for estimating a physical quantity related to the formation of an object, the method being a method for estimating a physical quantity related to the formation of an object by a modeling device that forms an object by discharging a modeling material from a nozzle, the physical quantity including modeling conditions of the object or a discharge result of the modeling material discharged from the nozzle, and estimating the modeling conditions corresponding to the discharge result or the discharge result corresponding to the modeling conditions in accordance with a relational expression obtained based on a parameter related to a viscosity of the modeling material, the modeling conditions, and the discharge result. According to this embodiment, by using parameters related to the viscosity of the modeling material, even when it is difficult to obtain an accurate value of the viscosity of the modeling material, the modeling conditions corresponding to the ejection result, or the ejection result corresponding to the modeling conditions, can be estimated with high accuracy.

[0092] (2) In the above aspect, the relational expression may be generated by machine learning using the parameters, the modeling conditions, and the ejection results as learning data. According to this embodiment, the time required to generate the relational expression can be reduced.

[0093] (3) In the above-mentioned form, the relational equation is generated by the machine learning using the modeling conditions or the discharge result as training data, and the discharge result includes at least one of the pressure of the flow path in the modeling device through which the modeling material flows, the discharge amount of the modeling material from the nozzle, the discharge speed of the modeling material from the nozzle, and a fluctuation amount of the discharge speed, and the training data may include any of the measured values ​​of the pressure of the flow path, the discharge amount, the discharge speed, or the fluctuation amount of the discharge speed included in the discharge result. According to this embodiment, it is possible to estimate any one of the pressure of the flow path, the amount of the modeling material discharged from the nozzle, the discharge speed of the modeling material from the nozzle, or the fluctuation amount of the discharge speed, which corresponds to the modeling condition. It is also possible to estimate the modeling condition corresponding to any one of the pressure of the flow path, the amount of the modeling material discharged from the nozzle, the discharge speed of the modeling material from the nozzle, or the fluctuation amount of the discharge speed.

[0094] (4) In the above embodiment, the modeling device includes a plasticization unit that plasticizes a material to generate the modeling material, a flow path through which the modeling material flows, the nozzle that ejects the modeling material, and a first heater that heats the nozzle, and the plasticization unit includes a flat screw that has a groove forming surface on which a groove is formed and rotates around a rotation axis, a barrel that has an opposing surface that faces the groove forming surface in a direction along the rotation axis and is provided with a communicating hole that forms part of the flow path, and a second heater that heats the modeling material supplied to the groove, and the modeling conditions may include at least one of a set temperature of the first heater, a set temperature of the second heater, a rotation speed of the flat screw, and a shape of the flow path. According to this embodiment, it is possible to estimate the discharge result corresponding to any one of the set temperature of the first heater, the set temperature of the second heater, the rotation speed of the flat screw, or the shape of the flow path. Also, it is possible to estimate any one of the set temperature of the first heater, the set temperature of the second heater, the rotation speed of the flat screw, or the shape of the flow path that corresponds to the discharge result.

[0095] (5) In the above embodiment, the second heater may include a third heater and a fourth heater arranged closer to the communication hole than the third heater, and the molding conditions may include a set temperature of the third heater and a set temperature of the fourth heater. According to this embodiment, it is possible to estimate the discharge result corresponding to the set temperature of the third heater and the set temperature of the fourth heater. Also, it is possible to estimate the set temperature of the third heater and the set temperature of the fourth heater corresponding to the discharge result.

[0096] (6) In the above embodiment, the parameter is the viscosity calculated using a Cross-WLF model, which is an equation for calculating the viscosity, and the viscosity may be calculated using a shear rate calculated from the flow rate of the modeling material. According to this aspect, when the modeling material is a thermoplastic resin, it is possible to improve the estimation accuracy of the modeling conditions or the discharge results.

[0097] (7) In the above aspect, the viscosity may be the viscosity at a plurality of positions in a flow path in the modeling device through which the modeling material flows. According to this aspect, it is possible to improve the estimation accuracy of the modeling conditions or the discharge results, compared to the case where the viscosity at one position in the flow path is a parameter.

[0098] (8) In the above aspect, the parameters may be a plurality of coefficients input to a Cross-WLF model, which is an equation for calculating the viscosity. According to this aspect, it is possible to estimate the modeling conditions or the ejection results without calculating the viscosity of the modeling material.

[0099] (9) In the above embodiment, the modeling device may include a three-dimensional modeling device that models the object by stacking the modeling material ejected from the nozzle, or an injection molding device that models the object by ejecting the modeling material from the nozzle into a molding die. [Explanation of symbols]

[0100] 10...physical quantity estimation system, 100...modeling device, 110...control unit, 120...modeling unit, 130...stage, 131...modeling surface, 140...movement mechanism, 200...physical quantity estimation device, 201...processing unit, 202...storage unit, 203...communication unit, 204...display unit, 205...input unit, 210...learning data acquisition unit, 220...model generation unit, 230...estimation unit, 260...model memory unit, 301...material supply unit, 302...plasticization unit, 303...discharge unit, 315...communication passage, 320...flat screw, 321...groove formation surface, 322...center unit , 323...groove, 324...material inlet, 325...ridge portion, 330...screw case, 340...driving motor, 350...barrel, 351...opposing surface, 352...communicating hole, 353...third heater, 354...fourth heater, 355...guide groove, 356...first heater, 361...nozzle, 362...flow path, 363...discharge control section, 364...pressure sensor, 365...flow rate sensor, 366...nozzle opening, 367...discharge adjustment section, 368...suction section, 369...first driving section, 370...second driving section, NT...network, RX...rotating shaft

Claims

1. A method for estimating a physical quantity related to the formation of a modeling object by a modeling apparatus that forms a modeling object by discharging a modeling material from a nozzle, comprising the steps of: the physical quantity includes a modeling condition of the object or a discharge result of the modeling material discharged from the nozzle, estimating the modeling conditions corresponding to the discharge result, or the discharge result corresponding to the modeling conditions, according to a relational expression obtained based on a parameter related to the viscosity of the modeling material, the modeling conditions, and the discharge result; A method for estimating physical quantities related to the formation of a model.

2. A method for estimating a physical quantity related to the formation of a structure according to claim 1, comprising: The relational expression is generated by machine learning using the parameters, the modeling conditions, and the discharge results as learning data. A method for estimating physical quantities related to the formation of a model.

3. A method for estimating a physical quantity related to the formation of a structure according to claim 2, comprising the steps of: the relational expression is generated by the machine learning using the modeling conditions or the discharge results as training data, The discharge result includes at least one of a pressure of a flow path in the modeling device through which the modeling material flows, a discharge amount of the modeling material from the nozzle, a discharge speed of the modeling material from the nozzle, and a fluctuation amount of the discharge speed, The teacher data includes any one of the pressure of the flow path, the discharge amount, the discharge speed, or a measured value of the fluctuation amount of the discharge speed, which is included in the discharge result. A method for estimating physical quantities related to the formation of a model.

4. A method for estimating a physical quantity related to the formation of a structure according to claim 2, comprising the steps of: The modeling apparatus includes a plasticizing unit that plasticizes a material to generate the modeling material; A flow path through which the building material flows; The nozzle that ejects the modeling material; A first heater that heats the nozzle, The plasticizing section includes: a flat screw having a groove forming surface on which grooves are formed and rotating around a rotation axis; a barrel having an opposing surface facing the groove forming surface in a direction along the rotation axis and having a communication hole that forms a part of the flow path; A second heater that heats the modeling material supplied to the groove, The molding conditions include at least one of a set temperature of the first heater, a set temperature of the second heater, a rotation speed of the flat screw, and a shape of the flow path. A method for estimating physical quantities related to the formation of a model.

5. A method for estimating a physical quantity related to the formation of a shaped object according to claim 4, comprising the steps of: The second heater is A third heater; a fourth heater disposed closer to the communication hole than the third heater, The modeling conditions include a set temperature of the third heater and a set temperature of the fourth heater. A method for estimating physical quantities related to the formation of a model.

6. A method for estimating a physical quantity related to the formation of a structure according to claim 1, comprising: The parameter is the viscosity calculated using the Cross-WLF model, which is an equation for calculating the viscosity, The viscosity is calculated using a shear rate calculated from the flow rate of the molding material. A method for estimating physical quantities related to the formation of a model.

7. A method for estimating a physical quantity related to the formation of a shaped object according to claim 6, comprising the steps of: The viscosity is the viscosity at a plurality of positions in a flow path of the building material in the building device. A method for estimating physical quantities related to the formation of a model.

8. A method for estimating a physical quantity related to the formation of a structure according to claim 1, comprising: The parameters are a plurality of coefficients input into the Cross-WLF model, which is an equation for calculating the viscosity. A method for estimating physical quantities related to the formation of a model.

9. A method for estimating a physical quantity related to the formation of a structure according to claim 1, comprising: The modeling device includes a three-dimensional modeling device that models the model by stacking the modeling material discharged from the nozzle, or an injection molding device that models the model by discharging the modeling material from the nozzle into a molding die. A method for estimating physical quantities related to the formation of a model.

Images