Information processing system, injection molding machine, and program

Abstract

The present invention addresses the problem of reducing the workload required to change a simulation model in accordance with the configuration, usage mode, or the like of a device, for a model used in a simulation relating to the control of the device. This information processing system is provided with: a first mechanical operation (506) that, on the basis of a command for controlling the operation of a device to be controlled, estimates relating to the operation of a first movable part among movable parts of the device to be controlled, using a simulation model of the first movable part; the first movable part is a part which is not influenced by a use mode during action; and a second mechanical operation (507) that receives the estimation result of the first mechanical operation (506) and performs an estimation relating to the operation of a second movable part, which is a part affected by the use mode during the operation, using a simulation model of the second movable part among the movable parts of the device to be controlled.

Description

Technical Field

[0001] This application claims priority based on Japanese Patent Application No. 2024-121279, filed on July 26, 2024. The entire contents of that Japanese application are incorporated herein by reference.

[0002] This invention relates to an information processing system, an injection molding machine, and a program. Background Technology

[0003] In the control of the equipment, the following process is performed: simulation is conducted, and the control information is adjusted based on the simulation results.

[0004] Patent document 1 discloses the following: In an injection molding system, a PID (Proportional-Integral-Differential) controller is adjusted based on the error between the injection pressure detected by a sensor and the pressure setpoint represented by a control model in order to control the injection pressure of the injection molding machine.

[0005] Patent document 2 discloses the following: For the apparatus constituting an injection molding machine, a mathematical model obtained by modeling the actual equipment is linearized and simulated to design the control system of the apparatus.

[0006] Patent Document 1: Japanese Patent Publication No. 2020-533197

[0007] Patent Document 2: Japanese Patent Application Publication No. 6-8296

[0008] In simulations related to equipment control, it is sometimes desirable to use different simulation models depending on the equipment's structure or usage. However, changing the simulation model based on the structure or usage conditions of various equipment is labor-intensive. Summary of the Invention

[0009] The purpose of this invention is to reduce the workload required to modify the simulation model based on the structure or usage of the equipment, for simulations related to equipment control.

[0010] One aspect of the present invention is an information processing system, characterized by comprising: a first inference unit, which, based on an instruction for controlling the operation of a controlled object device, uses a simulation model of a first movable part of a movable part of the controlled object device to make inferences related to the operation of the first movable part, wherein the first movable part is a part that is not affected by the mode of use during operation; and a second inference unit, which receives the inference result of the first inference unit and uses a simulation model of a second movable part of a movable part of the controlled object device to make inferences related to the operation of the second movable part, wherein the second movable part is a part that is affected by the mode of use during operation.

[0011] Invention Effects

[0012] According to one aspect of the present invention, for the model used in simulations related to the control of the device, the amount of work required to change the simulation model according to the structure or usage of the device can be reduced. Attached Figure Description

[0013] Figure 1 This is a diagram showing the structure of the injection molding machine according to this embodiment.

[0014] Figure 2 This is a diagram showing the structure of the control device.

[0015] Figure 3 It is a diagram showing the structure of a data processing device.

[0016] Figure 4 It is a diagram showing the structure of an information processing device.

[0017] Figure 5 This is a diagram illustrating an example of the hardware structure of a computer that implements an information processing device.

[0018] Figure 6 This is a diagram illustrating the structure of a simulation model used in the analysis performed by the analysis unit of the processing unit of an information processing device.

[0019] Figure 7 It is a graph comparing the output waveforms of the simulation model.

[0020] Figure 8 This is a diagram of another structural example of a simulation model used for analysis performed by the analysis unit of the processing unit of an information processing device.

[0021] Figure 9 This is a diagram representing a structural example of a linear multiple regression model.

[0022] Figure 10 This is a diagram illustrating the methods for deriving machine learning models.

[0023] In the diagram: 10-Injection molding machine, 20-Injection device, 30-Mold closing device, 100-Control device, 200-Data processing device, 400-Information processing device, 410-Communication unit, 420-Processing unit, 421-Analysis unit, 430-Storage unit, 431-Simulation model, 501-Instruction generation, 502-Pattern generation, 503-Pressure control, 504-Speed ​​control, 505-Current control, 506-First mechanical action, 507-Second mechanical action. Detailed Implementation

[0024] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0025] <Device Structure>

[0026] Figure 1 This diagram illustrates the structure of the injection molding machine according to this embodiment. The injection molding machine 10 includes an injection unit 20, a mold clamping unit 30, a control unit 100, and a data processing unit 200. The injection molding machine 10 is an example of a controllable device. Furthermore, an information processing unit 400 is connected to the control unit 100 and the data processing unit 200.

[0027] The injection unit 20 is configured to include a cylinder for heating the molding material, a screw that can rotate within the cylinder and move forward and backward axially, a rotary motor that drives the screw in the direction of rotation, and a motor that drives the screw axially. The molding material is, for example, resin. While rotating the screw, the injection unit 20 extends forward from the injection unit 20 toward the mold clamping device 30, thereby injecting the molding material, heated to a liquid state within the cylinder, into the mold of the mold clamping device 30 located in front of the injection unit 20. The injection unit 20 performs, for example, metering, filling, and holding pressure processes in the manufacturing process of the molded article. The filling and holding pressure processes are also collectively referred to as the injection process.

[0028] The mold closing device 30 is configured to include a mold, a clamping mechanism for securing the mold, and a motor for driving the clamping mechanism. The mold closing device 30 closes the mold, receiving molding material injected from the injection unit 20 into the mold. At this time, the mold closing device 30 uses the clamping mechanism to secure the mold (mold closing) so that the mold does not open due to the filling of molding material. The molding material filled into the mold solidifies, thereby producing a molded article. Then, the mold closing device 30 opens the mold, allowing the produced molded article to be removed. The mold closing device 30 is used in the molding process, for example, to perform mold closing, pressurization, mold closing, depressurization, and mold opening processes.

[0029] The control device 100 is a device for controlling the operation of the injection device 20 and the mold closing device 30. The data processing device 200 is a device for processing data obtained accompanying the operation of the injection device 20 and the mold closing device 30.

[0030] The information processing unit 400 acquires waveform data obtained from the control unit 100 and the data processing unit 200 using the operation of the injection molding machine 10, and calculates the characteristics of the injection molding machine 10 based on the acquired waveform data. Based on the calculated characteristics of the injection molding machine 10, the information processing unit 400 outputs information for controlling each injection molding machine 10 and provides it to the control unit 100. Details regarding the processing by the information processing unit 400 and the output information will be described later.

[0031] <Structure of Control Device 100>

[0032] Figure 2 This diagram illustrates the structure of the control device 100. The control device 100 controls the operation of the injection unit 20 and the mold closing unit 30. The control device 100 is implemented, for example, by a computer. The control device 100 includes a control unit 110 and a storage unit 120. The control device 100 controls the injection unit 20 and the mold closing unit 30 to repeatedly perform processes related to the manufacture of the molded article, thereby repeatedly manufacturing the molded article. Processes related to the manufacture of the molded article include metering processes, mold closing processes, pressure boosting processes, mold closing processes, filling processes, pressure holding processes, cooling processes, pressure release processes, mold opening processes, ejection processes, etc. Hereinafter, these manufacturing-related processes will sometimes be collectively referred to as "manufacturing processes." Furthermore, the series of actions used to obtain the molded article (e.g., actions from the metering process in the above-described manufacturing processes to the start of the next metering process) will be referred to as "injection," "molding cycle," etc. Additionally, the above-described processes for manufacturing the molded article are merely examples. For example, other processes not included above may be included as processes performed during a single injection.

[0033] The control unit 110 controls the injection unit 20 and the mold clamping unit 30 based on control information. The control information consists of conditions set by the user, generated for example, based on information input by the user using an input device (not shown). The control information includes, for example, molding conditions such as resin temperature, mold temperature (cylinder temperature), injection holding time, metering value, VP switching position, holding pressure, injection speed (filling speed), screw speed, screw back pressure, and clamping force. These molding conditions are determined by multiple combinations depending on the molded product or mold. Hereinafter, the combination data of these molding conditions will also be referred to as a molding condition dataset. Furthermore, the control information includes control data such as voltage, current, pressure, speed, and acceleration for drive units such as mechanisms or motors. The molding condition dataset is prepared according to the type of molded product or mold and stored in the storage unit 120. Molding conditions are an example of operating conditions.

[0034] The control unit 110 uses the aforementioned molding condition dataset to control the injection unit 20 and the mold clamping unit 30, implementing processes related to the manufacture (injection) of the molded article, including the aforementioned steps. When manufacturing the molded article begins, the control unit 110 reads the molding condition dataset corresponding to the molded article to be manufactured from the storage unit 120. Then, the control unit 110 controls the operation of the injection unit 20 and the mold clamping unit 30 based on control information containing the read molding condition dataset. Specifically, the control unit 110 controls the injection unit 20 and the mold clamping unit 30 to ensure that the data obtained from the injection unit 20 and the mold clamping unit 30 during the manufacturing process matches the set values ​​of the molding condition dataset.

[0035] The storage unit 120 stores the control information used by the control unit 110 to control the injection unit 20 and the mold clamping unit 30. The molding condition datasets included in the control information are prepared in a manner that establishes a corresponding association with the molded product or mold of the manufactured object. The storage unit 120 stores the molding condition datasets for each molded product of the manufactured object or each mold. Furthermore, although not shown, the storage unit 120 stores the program for the control unit 110 to control the injection unit 20 and the mold clamping unit 30.

[0036] <Structure of Data Processing Device 200>

[0037] Figure 3 This diagram illustrates the structure of the data processing device 200. The data processing device 200 acquires and processes data obtained during the actions performed by the injection unit 20 and the mold clamping unit 30 in the aforementioned processes related to the manufacture of the molded article. The data processing device 200 is implemented, for example, by a computer. The data processing device 200 includes a data acquisition unit 210, a processing unit 220, and a storage unit 230.

[0038] The data acquisition unit 210 acquires data about the object being processed from the injection unit 20 and the mold closing unit 30. Various sensors, detectors, etc., are installed on the injection unit 20 and the mold closing unit 30. Furthermore, various measuring devices are sometimes connected to the injection unit 20 or the mold closing unit 30. The data acquired using these sensors, detectors, measuring devices, etc. (hereinafter referred to as "acquisition data") serves as information representing the molding results of the injection unit 20 and the mold closing unit 30, and is used for quality management of the molded product. The data acquisition unit 210 receives the acquisition data sent from the sensors, detectors, measuring devices, etc., and stores it in the storage unit 230.

[0039] The processing unit 220 processes the acquired data stored in the storage unit 230. Specifically, the processing unit 220 performs processes such as extracting representative values ​​of the acquired data from each process and generating time-series data by temporally sequencing the acquired data from each process. When extracting representative values, the processing unit 220 performs statistical processing on the acquired data, such as calculating the average value, determining the acceptable range of values, and determining the maximum or minimum value.

[0040] Storage unit 230 stores the acquired data obtained by data acquisition unit 210. The data format for the acquired data stored in storage unit 230 can be, for example, binary, text, CSV (Comma Separated Values), INI, YAML (YAML Ain't Markup Language), JSON (JavaScript Object Notation), etc. By setting data files based on these common data formats, the data files stored in storage unit 230 can be exchanged with other information processing devices or edited with data files obtained from external devices. Furthermore, although not shown, storage unit 230 stores a program for processing data performed by processing unit 220.

[0041] <Structure of Information Processing Device 400>

[0042] Figure 4 This diagram illustrates the structure of the information processing device 400. The information processing device 400 uses a model of the injection molding machine 10 to simulate the operation of the injection molding machine 10. Then, the information processing device 400 generates and outputs control information for the injection molding machine 10 based on the simulation results. The information processing device 400 is implemented, for example, by a computer. The information processing device 400 includes a communication unit 410, a processing unit 420, and a storage unit 430. The processing unit 420 has an analysis unit 421 that performs simulations using machine learning models and the like to analyze the characteristics of the injection molding machine 10. The storage unit 430 stores the simulation model 431 used in the analysis performed by the analysis unit 421.

[0043] The communication unit 410 performs data communication with the control device 100 and the data processing device 200. Specifically, the communication unit 410 sends control information of the injection molding machine 10 generated by the processing unit 420 to the control device 100. Furthermore, the communication unit 410 is also used to acquire information used for simulation or analysis from the control device 100 or the data processing device 200. For example, it can acquire control information of the injection device 20 and the clamping device 30 used in the processing of the processing unit 420 from the control device 100. The data received can be various data obtained in the form of waveform data when the injection molding machine 10 operates. Specifically, examples include voltage, current supplied to various mechanisms, speed, acceleration, pressure, and other data during the operation of various mechanisms.

[0044] The processing unit 420 processes the control information from the injection unit 20 and the mold clamping unit 30 to determine the characteristics of the injection molding machine 10. The control information from the injection unit 20 and the mold clamping unit 30 can be obtained by receiving it from the injection unit 20 and the mold clamping unit 30 through the communication unit 410, or control information prepared separately for simulation can be used.

[0045] The analysis unit 421 processes the control information of the injection unit 20 and the mold clamping unit 30 to generate input variables for input to the simulation model 431. Then, the analysis unit 421 inputs the generated input variables into the simulation model 431 and outputs variables equivalent to the actual values ​​(hereinafter referred to as "output values"). The processing performed by the analysis unit 421 and the details of the simulation model 431 will be described later.

[0046] The storage unit 430 stores control information of the injection device 20 and mold clamping device 30 used in the simulation, information obtained from processing by the processing unit 420, and the simulation model 431 used by the analysis unit 421 of the processing unit 420. The data format of the data file stored in the storage unit 430 can be, for example, CSV, XML (Extensible Markup Language), JSON, etc.

[0047] <Hardware Structure of Information Processing Device 400>

[0048] Figure 5 This is a diagram illustrating an example of the hardware structure of a computer that implements the information processing device 400. Figure 5The computer shown includes a processor 401 as a processing unit, a main storage device (main memory) 402 as a storage unit, and an auxiliary storage device 403. The processor 401 can be, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), a FPGA (Field-Programmable Gate Array), or various other processing circuits. The processor 401 reads programs stored in the auxiliary storage device 403 into the main storage device 402 for execution. The main storage device 402 can be, for example, RAM (Random Access Memory). The auxiliary storage device 403 can be, for example, a disk drive or an SSD (Solid State Drive).

[0049] Furthermore, the computer can be configured to include a display device 404 for displaying images and an input device 405 for users to perform input operations. The input device 405 can be, for example, a keyboard, mouse, or touch panel. When using a touch panel integrated with the display device 404 as the input device 405, the user performs input operations by touching the operation screen displayed on the display device 404 with their finger or a pen-like device. Additionally, Figure 5 The computer structure shown is only one example, and the computer used in this embodiment is not limited to that type. Figure 5 For example, it can also be configured to have a non-volatile memory such as flash memory or ROM (Read Only Memory) as a storage device.

[0050] In the information processing device 400 Figure 5 In the computer implementation shown, the communication unit 410 is implemented, for example, by a processor 401 that loads and executes programs, and a communication interface (not shown). The functions of the processing unit 420 are implemented, for example, by loading and executing programs through the processor 401. The storage unit 430 is implemented, for example, by an auxiliary storage device 403.

[0051] <Analysis conducted by Analysis Department 421>

[0052] Figure 6This diagram illustrates a structural example of the simulation model 431 used for analysis performed by the analysis unit 421 of the processing unit 420 of the information processing device 400. The analysis unit 421 receives molding conditions contained in the control information and performs various simulations using different models. The analysis unit 421 performs simulations targeting the operations of the injection unit 20 and the clamping unit 30 constituting the injection molding machine 10. Furthermore, it can also perform simulations targeting the controls performed by the control unit 100 of the injection molding machine 10. For example, as... Figure 6 As shown, the simulation performed by the analysis unit 421 can be composed of simulation models 431 such as instruction generation 501, pattern generation 502, speed control 504, current control 505, first mechanical action 506, and second mechanical action 507.

[0053] Instruction generation 501 and pattern generation 502 are simulation models that generate instructions for operating the injection molding machine 10 based on the molding conditions used to control the injection molding machine 10. Here, in the actual injection molding machine 10, the control device 100 generates instructions for controlling the operation of the injection unit 20 and the clamping unit 30 according to the molding conditions. The calculation model used in instruction generation 501 and pattern generation 502 is the same as the calculation model used to generate the instructions in the control device 100. The analysis unit 421, which uses instruction generation 501 and pattern generation 502 to perform the simulation, is an example of an instruction generation unit.

[0054] First, the analysis unit 421 generates instructions for operating the injection molding machine 10 based on the simulation model generated by the instruction generation 501. In the instruction generation 501, the molding conditions for injection molding are converted into instructions for operating the injection molding machine 10. These instructions are commands that specify the speed or pressure, etc., during the operation of the injection molding machine 10. The instructions generated by the instruction generation 501 correspond to the user's settings for the operation of the injection molding machine 10.

[0055] Next, the analysis unit 421 converts the instructions generated by the instruction generation 501 into instructions that the actual device (injection molding machine 10) can follow, based on the simulation model of the pattern generation 502. Since the actual device is subject to physical constraints, it is difficult to operate according to instructions corresponding to the molding conditions. Therefore, in the pattern generation 502, the instructions generated by the instruction generation 501 are converted into instructions that the actual device can follow. Specifically, for example, if the speed of the action changes in the instructions generated by the instruction generation 501, the instruction is converted as follows: the action becomes an action where the speed changes after a certain delay due to acceleration, rather than immediately switching to a different speed at the moment the speed changes.

[0056] Speed ​​control 504 and current control 505 are simulation models that generate information representing the control content of the drive source, i.e., the motor, of the injection molding machine 10. In speed control 504 and current control 505, pattern information is output representing the control pattern of the motor when the motor of the injection molding machine 10 operates according to the instructions generated by pattern generation 502. For example, the pattern information is generated in the form of waveform data. Furthermore, the format of the waveform data is not particularly limited. For example, it can be data obtained by listing values ​​based on CSV, or it can be a waveform graph. The analysis unit 421, which uses speed control 504 and current control 505 to perform the simulation, is an example of a pattern information generation unit.

[0057] First, the analysis unit 421 simulates the motor's operating speed based on the simulation model of the speed control 504, generating pattern information (current command) of current changes. This pattern information is generated, for example, in the form of waveform data representing the relationship between the elapsed time and the current. Next, the analysis unit 421 simulates the current supplied to the motor based on the simulation model of the current control 505, generating pattern information (voltage command) of voltage changes. In the current control 505, the voltage supplied to the motor is determined to achieve control over the motor's operating speed generated by the speed control 504. This pattern information is output, for example, in the form of waveform data representing the relationship between the elapsed time and the voltage supplied to the motor.

[0058] The first mechanical action 506 and the second mechanical action 507 are simulation models of the operation of the injection molding machine 10. The first mechanical action 506 is a simulation model of the operation of the drive source of the injection molding machine 10, namely the motor. The second mechanical action 507 is a simulation model of the operation of the movable part in the injection molding machine 10 that receives the driving force of the motor and performs the operation. The object simulated using the first mechanical action 506, namely the motor, is an example of the first movable part. Similarly, the object simulated using the second mechanical action 507, namely the movable part, is an example of the second movable part.

[0059] The first mechanical action 506 and the second mechanical action 507 are implemented, for example, in the form of a machine learning model. As a machine learning model, for example, a linear multiple regression model is used. In the first mechanical action 506 and the second mechanical action 507, motion information representing the result of the mechanical actions constituting the actual injection molding machine 10 is output. Figure 6 In the example shown, motion information for the motion speed of the injection molding machine 10 is generated. The analysis unit 421 that simulates the first mechanical motion 506 is an example of the first inference unit. And, the analysis unit 421 that simulates the second mechanical motion 507 is an example of the second inference unit.

[0060] First, the analysis unit 421 performs a simulation of the motor's movement based on the simulation model of the first mechanical action 506. The analysis unit 421 extracts input variables from the waveform data output from the self-current control 505 and inputs them into the first mechanical action 506. In the first mechanical action 506, the simulation of the motor's movement is performed, and the output variables (inferred values) corresponding to the input variables are calculated. The input variables are variables related to the motor's torque. For example, d-axis current, q-axis current, and speed can be examples of input variables. The output variables are selected from information that can be used to calculate the motor's torque in the second mechanical action 507. For example, the inferred value of the motor's torque itself or the inferred value of the current supplied to the motor can be examples of output variables.

[0061] Next, the analysis unit 421 simulates the movement of the movable part of the injection molding machine 10 based on the simulation model of the second mechanical action 507. The analysis unit 421 inputs the inferred values ​​output from the first mechanical action 506 to the second mechanical action 507 as input variables. In the second mechanical action 507, the simulation of the movement of the movable part of the injection molding machine 10 is performed, and the output variables (inferred values) corresponding to the input inferred values ​​are calculated. The output variables are selected from various information related to the movement of the movable part that is the object of simulation. For example, inferred values ​​of the movement speed, acceleration, and position of the movable part, and inferred values ​​of the current supplied to the injection molding machine 10 can be given. Furthermore, information (parameters) that can be used to calculate inferred values ​​such as movement speed or current can also be output instead of inferred values ​​of movement speed or current. In addition, the simulation model of the second mechanical action 507 includes information such as the weight of the mold, the viscosity of the molding material (resin), and the difference between belt drive and DD drive (drive based on a direct drive motor).

[0062] Figure 7 This is a graph comparing the output waveforms of the simulation model. Here, for each output of instruction generation 501, pattern generation 502, and the second mechanical action 507, waveform data representing the relationship between the motor's operating speed and time are compared. Figure 7 In the diagram, each waveform data is represented by a waveform graph showing the relationship between velocity and time. Figure 7 In the diagram, waveform diagram 510, with dashed lines, shows the waveform data as the output of instruction generation 501. Waveform diagram 520, with thin solid lines, shows the waveform data as the output of pattern generation 502. Waveform diagram 530, with thick solid lines, shows the waveform data as the output of the second mechanical action 507.

[0063] As described above, the waveform data (waveform 510) output by instruction generation 501 abruptly changes speed at a certain moment. This is an action based on the molding conditions set by the user, but in actual devices, the motor may be difficult to operate as shown in waveform 510 due to inertia or output control. The waveform data (waveform 520) output by pattern generation 502 follows the speed change shown in waveform 510 after a certain delay caused by acceleration at the moment of speed switching in waveform 510.

[0064] The waveform data (waveform diagram 530) output as the second mechanical action 507 shows the state of speed change, which also reflects the inertia or motion characteristics of the movable part of the motor or injection molding machine 10. Figure 7 In waveform 530, a state is shown where the velocity curve changes while being further delayed than in waveform 520.

[0065] <Adjustment of the simulation model>

[0066] like Figure 6 As shown, the output value of the first mechanical action 506 (an inferred value of the current supplied to the motor) is fed back to the current control 505. The analysis unit 421 compares the output value related to the current of the first mechanical action 506 with the output value (current command) of the speed control 504 input to the current control 505, and calculates the error. Then, the analysis unit 421 adjusts the physical quantities (parameters) included in the simulation model of the first mechanical action 506 to minimize the error of the output value of the current control 505 relative to the value calculated by the linear multiple regression model for the output value of the first mechanical action 506.

[0067] Furthermore, the output value of the second mechanical action 507 (the inferred value of the movement of the movable part of the injection molding machine 10) is fed back to the speed control 504. The analysis unit 421 compares the output value related to the speed of the second mechanical action 507 with the output value (speed command) of the pattern generation 502 input to the speed control 504, and calculates the error. Then, the analysis unit 421 adjusts the physical quantities (parameters) included in the simulation model of the second mechanical action 507 so that the error of the output value of the first mechanical action 506 relative to the value calculated by the linear multiple regression model for the output value of the second mechanical action 507 is minimized.

[0068] <Structure of the Simulation Model Group>

[0069] The structure of the simulation model group will be further explained. The simulation model used in the analysis performed by the analysis unit 421 is... Figure 6In the example shown, instruction generation 501, pattern generation 502, speed control 504, current control 505, first mechanical action 506, and second mechanical action 507 are illustrated. That is, in the above structural example, as a simulation model of the action of the injection molding machine 10, a simulation model of the action of the motor, namely the first mechanical action 506, and a simulation model of the action of the movable part of the injection molding machine 10, namely the second mechanical action 507, are provided.

[0070] Here, the object of the simulation of the first mechanical action 506 is the motor, whose operating characteristics do not change when using the injection molding machine 10. In other words, the motor is a movable part of the injection molding machine 10 that is not affected by the mode of use. Therefore, the operating characteristics of the motor can be determined at the time of manufacture of the injection molding machine 10. On the other hand, the operating characteristics of various movable parts in the injection molding machine 10 other than the motor may sometimes change significantly due to the mode of use of the injection molding machine 10. Specifically, examples include situations where inertia changes due to mold changes or where the applied pressure differs due to different molding conditions or molding materials.

[0071] Therefore, this embodiment separately constructs a simulation model for the motor's operation, namely the first mechanical action 506, and a simulation model for the operation of the movable part of the injection molding machine 10, namely the second mechanical action 507. By making the first mechanical action 506 independent, a simulation model of the first mechanical action 506 can be exported when the injection molding machine 10 leaves the factory. Furthermore, for injection molding machines 10 using the same motor, the first mechanical action 506 based on a common simulation model can be applied.

[0072] On the other hand, by separating the simulation model for the motor's movement from the second mechanical action 507, it is sufficient to export only the simulation model for the movement of movable parts other than the motor, based on how the injection molding machine 10 is used. Furthermore, exporting the simulation model requires actually operating the injection molding machine 10 to obtain control information or motion characteristic information, and adjusting the simulation model based on the acquired information. However, typically, the injection molding machine 10 does not have sensors installed to obtain information such as the motor's torque and other motion characteristics. Therefore, if the simulation model corresponding to the usage is exported after the injection molding machine 10 is manufactured, exporting it is easy if the simulation model for the motor is separate.

[0073] Furthermore, the first mechanical action 506 is not essential in the simulation model used for the analysis performed by the analysis unit 421. By omitting the simulation based on the first mechanical action 506, the current control 505, which outputs the voltage command input to the first mechanical action 506, is also omitted. That is, it can be configured such that the input variable extracted from the waveform data output from the speed control 504 is directly input to the second mechanical action 507. Even with this configuration, the operating speed of the movable part of the injection molding machine 10 can be inferred based on the output of the second mechanical action 507. However, in this case, since the inference of the motor's operating characteristics performed by the first mechanical action 506 is not reflected, the accuracy of the output inferred value decreases compared to the case where the first mechanical action 506 is included.

[0074] Furthermore, pattern generation 502 and current control 505 are also not essential. Without pattern generation 502, the input variables extracted from the output of instruction generation 501 are input to the simulation model following pattern generation 502. Without current control 505, the input variables extracted from the output of the simulation model preceding current control 505 are input to the simulation model following the first mechanical action 506. In either case, the inference based on the omitted simulation models is not reflected; therefore, the accuracy of the inferred value output from the second mechanical action 507 decreases compared to the case where these simulation models are present.

[0075] In addition, Figure 6 In the illustrated structural example, instructions for operating the injection molding machine 10 are generated (simulated) using instruction generation 501 and pattern generation 502. Alternatively, the control device 100 of the injection molding machine 10 can be used to generate instructions for actually controlling the operation of the injection unit 20 and the clamping unit 30. The generated instructions are sent from the control device 100 to the information processing device 400. Furthermore, the analysis unit 421 of the information processing device 400 generates pattern information based on speed control 504 and current control 505 according to the acquired instructions. (The following will be discussed further.) Figure 8 The same applies to the variant examples shown.

[0076] <Application in Control Device 100>

[0077] After adjusting each simulation model as described above, the user performs a simulation using that model in the information processing device 400. Then, the user generates control information based on the inference values ​​obtained from the simulation to achieve the desired action according to the characteristics of the injection molding machine 10, and sends it to the control device 100. The control device 100 controls the operation of the injection unit 20 and the mold clamping unit 30 according to the control information obtained from the information processing device 400, and performs the processes related to the manufacturing of the molded article.

[0078] <Variation Example>

[0079] Figure 8 This is a diagram illustrating another structural example of the simulation model used for analysis performed by the analysis unit 421 of the processing unit 420 of the information processing device 400. Figure 8 The simulation model in the example shown can be composed of simulation models such as instruction generation 501, pattern generation 502, pressure control 503, speed control 504, current control 505, first mechanical action 506, and second mechanical action 507. Compared to Figure 6 The structural example shown, Figure 8 The illustrated structural example includes an additional pressure control 503. Pressure control 503 is positioned between pattern generation 502 and speed control 504. That is, the output of pattern generation 502 is input to pressure control 503, and the output of pressure control 503 is input to speed control 504. The simulation models for instruction generation 501, pattern generation 502, speed control 504, current control 505, the first mechanical action 506, and the second mechanical action 507 are... Figure 6 The simulation models shown are identical.

[0080] exist Figure 8 In the structure shown, pressure control 503 is a simulation model that generates information representing the control content of the movable part of injection molding machine 10. Pressure control 503 outputs pattern information representing the control content of the pressure when injection molding machine 10 operates according to the instructions generated by pattern generation 502. The pattern information is generated, for example, in the form of waveform data. The format of the waveform data is not particularly limited; it can be data obtained by listing values ​​based on CSV, etc., or it can be a waveform graph. The analysis unit 421, which uses pressure control 503 for simulation, is an example of a pattern information generation unit.

[0081] exist Figure 8 In the structure shown, the analysis unit 421 generates pattern information of the simulated object, i.e., the movable part, in the injection molding machine 10 based on the simulation model of the pressure control 503. This pattern information is generated, for example, in the form of waveform data representing the relationship between the elapsed time and the change in pressure. Then, the analysis unit 421 generates speed pattern information through the speed control 504.

[0082] like Figure 8As shown, when pressure control 503 is provided, the output value of the second mechanical action 507 is fed back to pressure control 503. Analysis unit 421 compares the pressure-related output value of the second mechanical action 507 with the output value of pressure control 503 and calculates the error. Furthermore, analysis unit 421 adjusts the simulation model of speed control 504 to minimize the error between the output value of pressure control 503 and the output value of the second mechanical action 507.

[0083] The above is for reference. Figure 6 The example provided is used to infer the operating speed and current of the movable part of the injection molding machine 10, and in reference... Figure 8 The simulation of pressure inference was further explained in the modified example described. In addition, by selecting waveform data that can be obtained from the injection molding machine 10 based on the output of sensors, etc., inferred values ​​such as clamping force acting on the movable part, acceleration of the movable part, position of the movable part (displacement or rotation angle, etc.), and position of the pressure plate can be obtained.

[0084] <Specific examples of methods for deriving machine learning models>

[0085] Next, a specific example of the method for deriving the machine learning model used as the simulation model for the first mechanical action 506 and the second mechanical action 507 will be explained. Here, as an example, the example of deriving the simulation model of the motor's motion used in the first mechanical action 506 will be explained. First, a linear multiple regression model as an example of a machine learning model will be explained, and then the method for deriving the machine learning model will be explained.

[0086] Figure 9 This is a diagram illustrating the structure of a linear multiple regression model. The illustrated model accepts multiple inputs and utilizes multiple nodes representing multiple processing units to perform an analysis that considers the correlations between the inputs, thereby obtaining a single output. Figure 9 In the model shown, x1 to x3 are input variables, and y1 is the output. w1 to w4, recorded in the intermediate layers, represent the weight values ​​set for each node. That is, in Figure 9 In the model shown, if input variables x1 to x3 are provided, the output y1 is obtained after processing using the weight values ​​w1 to w4 of each node. Furthermore, by changing the weight values ​​w1 to w4 of each node, different outputs y1 can be obtained for the same input variables x1 to x3. Additionally, Figure 9 The number of input variables and nodes shown is for illustrative purposes only and is not limited to the number shown in the figure.

[0087] In Figure 9The linear multiple regression model shown is applied to the relationship between current and voltage. For example, the actual current is input to input variables x1 to x3, and a voltage command is obtained as the output y1. Furthermore, physical constants are set as weight values ​​w1 to w4 for each node. The types of physical constants set as weight values ​​w1 to w4 are selected based on the physical quantities of the object being analyzed. For example, when the actual current of the motor is used as input and the voltage command is used as output, physical constants such as the motor's resistance (R), d-axis inductance (Ld), q-axis inductance (Lq), back electromotive force constant (Ke), torque constant (Kt), viscous friction (Dm), and inertia (J) can be assigned to each node, and weight values ​​can be set.

[0088] Figure 10 This is a diagram illustrating the derivation method of the machine learning model. The simulation model of the first mechanical action 506, for example, is obtained by using, as shown in the reference... Figure 9 The training and learning of a linear multiple regression model is used to analyze and derive waveform data acquired from the controlled device, injection molding machine 10. Here, as an example, it is assumed that the simulation model obtains voltage and acceleration from the actual current and actual speed of the motor. Furthermore, in Figure 10 In this paper, the intermediate layers in the model are grouped together into a single node.

[0089] First, the d-axis current, q-axis current, and velocity are extracted from the waveform data acquired by the injection molding machine 10 as input variables X, and then input into the simulation model. Furthermore, the actual values ​​(Y(actual)) of the d-axis voltage, q-axis voltage, and acceleration are extracted from the waveform data. These actual values ​​are equivalent to the output calculated based on the input variables of the simulation model, namely the d-axis current, q-axis current, and velocity.

[0090] On the other hand, the simulation model outputs (Y) based on the input d-axis current, q-axis current, and velocity to obtain the d-axis voltage, q-axis voltage, and acceleration, respectively. Here, the d-axis voltage, q-axis voltage, and acceleration are each calculated based on the d-axis current, q-axis current, and velocity, respectively.

[0091] Next, the actual values ​​(Y(actual)) of d-axis voltage, q-axis voltage, and acceleration extracted from the waveform data are compared with the output (Y) of d-axis voltage, q-axis voltage, and acceleration obtained using the simulation model, and the error is calculated. Then, the weight values ​​W of each node in the intermediate layer are adjusted to minimize this error; in other words, to make the output (Y) of d-axis voltage, q-axis voltage, and acceleration consistent with the actual values ​​(Y(actual)). If the weight value W that minimizes the error between the output (Y) of d-axis voltage, q-axis voltage, and acceleration and the actual values ​​(Y(actual)) of d-axis voltage, q-axis voltage, and acceleration is obtained, then this weight value W represents a characteristic related to the source of the input waveform data, namely the motor of the injection molding machine 10.

[0092] In the above example, the derivation of the simulation model for the motor's motion used in the first mechanical action 506 was described, but the derivation of the simulation model for the motion of the movable part used in the second mechanical action 507 is performed in the same way. In the derivation of the simulation model for the second mechanical action 507, input and output variables are set according to the motion of the movable part, which is the object of the simulation. As described above, the object of the simulation for the second mechanical action 507, i.e., the movable part, sometimes requires a change in the simulation model when the usage of the injection molding machine 10 is changed. Therefore, when the usage of the injection molding machine 10 is changed, the following process is performed: the injection molding machine 10 is operated to obtain physical quantities related to the motion of the movable part based on the new usage, input and output variables are set, the simulation is executed, and the simulation model is adjusted.

[0093] The embodiments of the present invention have been described above, but the technical scope of the present invention is not limited to the above embodiments. For example, in the above embodiments, waveform data is analyzed and control information is generated in the information processing device 400 connected to the injection molding machine 10. Alternatively, the structure could be configured such that waveform data is analyzed and control information is generated in the control device 100 or the data processing device 200 of the injection molding machine 10. In other words, the function of the information processing device 400 could also be integrated into the injection molding machine 10. Furthermore, various modifications or structural substitutions that do not depart from the scope of the technical concept of the present invention are also included in the present invention.

Claims

1. An information processing system, characterized in that, have: The first inference unit, based on the instruction for controlling the action of the controlled object device, uses a simulation model of the first movable part of the movable part of the controlled object device to make inferences related to the action of the first movable part, wherein the first movable part is a part that is not affected by the mode of use during the action; and The second inference unit accepts the inference result of the first inference unit and uses a simulation model of the second movable part in the movable part of the control object device to make inferences related to the action of the second movable part, which is the part affected by the mode of use during the action.

2. The information processing system according to claim 1, characterized in that, As the first movable part, the first inference unit makes inferences related to the operation of the drive source of the controlled object device, namely the motor.

3. The information processing system according to claim 1, characterized in that, It also has an instruction generation unit. The instruction generation unit receives information related to the operating conditions of the controlled device and generates the instruction based on the received information. The first inference unit receives instructions generated by the instruction generation unit to perform inferences related to the action of the first movable part.

4. The information processing system according to claim 1, characterized in that, It also has a pattern information generation unit. The pattern information generation unit generates pattern information related to the operation of the controlled object device according to the instruction. The first inference unit receives pattern information generated by the pattern information generation unit to make inferences related to the movement of the first movable part. The pattern information generation unit modifies the computational model used to generate the pattern information based on the inference results of at least one of the first inference unit and the second inference unit.

5. An injection molding machine, characterized in that, have: Injection molding equipment and mold clamping equipment perform the processes in injection molding; and The control device controls the operation of the injection device and the mold closing device. The control device includes: The information processing system according to any one of claims 1 to 4; and The control unit controls the operation of the injection device and the mold closing device based on control information reflecting the inference results of the information processing system.

6. A program, characterized in that, To enable the computer to function as a unit in the following ways: The first inference unit, based on the instruction for controlling the action of the controlled object device, uses a simulation model of the first movable part of the movable part of the controlled object device to make inferences related to the action of the first movable part, wherein the first movable part is the part that is not affected by the mode of use during the action; and The second inference unit accepts the inference result of the first inference unit and uses a simulation model of the second movable part in the movable part of the control object device to make inferences related to the action of the second movable part, which is the part affected by the mode of use during the action.

Images