Molding condition correction device, molding machine, molding condition correction method, and computer program

Abstract

A molding condition correction device that corrects molding conditions of a molding machine includes: an acquisition unit that acquires measurement data obtained by measuring a state of the molding machine and inspection result data obtained by inspecting a state of a molded article molded by the molding machine; and a learner that learns a relationship between the measurement data and the inspection result data and a correction amount of the molding condition, and determines the correction amount from the acquired measurement data and inspection result data. At least one of the measurement data, the inspection result data, and the correction amount processed by the learner is normalized or normalized.

Description

Technical Field

[0001] The present invention relates to a molding condition correction device, a molding machine, a molding condition correction method and a computer program. Background Art

[0002] When performing injection molding using an injection molding machine, it is necessary to first perform condition determination work, correct the set values ​​of various molding condition items, and find the molding conditions. The correction of molding conditions is based on the operator's experience, and in order to obtain appropriate molding conditions, it is necessary to try and explore repeatedly. The same is true when performing extrusion molding using an extruder.

[0003] Patent Document 1 discloses an injection molding machine system that uses reinforcement learning to correct molding conditions of an injection molding machine.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2019-166702 Summary of the invention

[0007] Incidentally, in order to build a learner (AI model) to correct molding conditions, learning data from thousands of injections is required. In addition, a learner is required for each mold or product.

[0008] Since the data required for learning also includes data of defective products, defective products are deliberately produced. There is a technical problem that producing a large number of defective products in a mass production factory will have a negative impact on the factory's production plan.

[0009] The purpose of the present disclosure is to provide a molding condition correction device, a molding machine, a molding condition correction method and a computer program having a learner that serves as a standard model for correcting molding conditions and can be applied to the production of any molded product.

[0010] Means used to solve problems

[0011] According to one aspect of the present disclosure, a molding condition correction device is a molding condition correction device for correcting the molding conditions of a molding machine, comprising: an acquisition unit, which acquires measurement data obtained by measuring the state of the molding machine and inspection result data obtained by inspecting the state of a molded product molded by the molding machine; and a learner, which learns the relationship between the measurement data and the inspection result data and the correction amount of the molding condition, and determines the correction amount based on the acquired measurement data and the inspection result data, wherein at least one of the measurement data, the inspection result data and the correction amount processed by the learner is standardized or normalized.

[0012] The molding machine according to one aspect of the present disclosure has the above-described molding condition correction device.

[0013] The molding condition correction method according to one aspect of the present disclosure is a molding condition correction method for correcting the molding conditions of a molding machine, including: a step of acquiring measurement data obtained by measuring the state of the molding machine and inspection result data obtained by inspecting the state of a molded product molded by the molding machine; and a step of determining the correction amount based on the acquired measurement data and inspection result data by using a learner that has learned the relationship between the measurement data, the inspection result data, and the correction amount of the molding conditions, wherein at least one of the measurement data, the inspection result data, and the correction amount processed by the learner is standardized or normalized.

[0014] The computer program according to one aspect of the present disclosure is a computer program for causing a computer to execute a process of correcting the molding conditions of a molding machine, which causes the computer to execute: a step of acquiring measurement data obtained by measuring the state of the molding machine and inspection result data obtained by inspecting the state of a molded product molded by the molding machine; and a determination step of determining the correction amount based on the acquired measurement data and inspection result data by using a learner that has learned the relationship between the measurement data, the inspection result data, and the correction amount of the molding conditions, wherein the determination step uses the learner in which at least one of the measurement data, the inspection result data, and the correction amount is standardized or normalized to determine the correction amount.

[0015] Advantageous Effects of the Invention

[0016] According to the present disclosure, it is possible to provide a molding condition correction device, a molding machine, a molding condition correction method, and a computer program having a learner that is a standard model for correcting molding conditions and can be applied to the manufacture of any molded product. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram illustrating a configuration example of an injection molding device according to the present embodiment.

[0018] Figure 2 It is a block diagram showing a configuration example of the molding condition correction device according to the present embodiment.

[0019] Figure 3 It is a functional block diagram of a processor according to the present embodiment.

[0020] Figure 4 It is a flowchart showing a method for generating and reusing a learner (standard model) according to the present embodiment.

[0021] Figure 5 This is a conceptual diagram showing a method of repurposing a learner (standard model) according to the present embodiment.

[0022] Figure 6 This is a conceptual diagram showing a method of additional learning of a learner (standard model) according to the present embodiment.

[0023] Figure 7 This is a flowchart showing a process of generating a learner (standard model).

[0024] Figure 8 This is a flowchart showing a process of importing a learner (standard model).

[0025] Figure 9 This is a flowchart showing a molding process.

[0026] Figure 10 This is a flowchart showing a process of additional learning of a learner (standard model). Detailed Embodiments

[0027] Hereinafter, specific examples of a molding condition correction device, a molding machine, a molding condition correction method, and a computer program according to an embodiment of the present invention will be described with reference to the accompanying drawings. At least a part of the embodiments described below can be arbitrarily combined. It should be noted that the present invention is not limited to these examples, but is shown by the scope of the claims and is intended to include all modifications within the meaning and scope equivalent to the claims.

[0028] <Summary of the Present Embodiment>

[0029] In the present embodiment, an example will be introduced in which a learner (AI model) for correcting molding conditions specifically for specific defects (such as burrs and insufficient material) is pre-constructed and repurposed for other molds. By standardizing and normalizing three parameters closely related to the operation of the learner for correcting molding conditions, a learner that absorbs mold differences can be constructed. The three parameters for which standardization or normalization is performed are: measurement data obtained by measuring the state of an injection molding machine, inspection result data obtained by inspecting the state of a molded product molded by the injection molding machine, and the amount of correction of molding conditions (the amount of molding conditions changed by one correction).

[0030] A learner in which the above parameters have been standardized and normalized will be appropriately referred to as a standard model. According to the standard model, molding conditions can be appropriately corrected even in a mold different from the mold used to construct the standard model. In addition, the performance can be improved by making the standard model perform additional learning using data collected from different molds.

[0031] <Configuration of Injection Molding Machine>

[0032] Figure 1 This is a schematic diagram showing a configuration example of an injection molding machine 101 according to the present embodiment. The injection molding machine 101 of the present embodiment includes: a mold clamping device 2 for clamping the mold 21; an injection device 3 for plasticizing and injecting a molding material; a control device 4; and an inspection device 5. The mold clamping device 2 and the injection device 3 constitute a molding machine main body 1. The control device 4 functions as a molding condition correction device according to the present embodiment.

[0033] The mold clamping device 2 includes a fixed platen 22 fixed to a base 20, a mold clamping housing 23 slidably provided on the base 20, and a movable platen 24 slidably provided on the base 20 in the same manner. The fixed platen 22 and the mold clamping housing 23 are connected by a plurality of (for example, four) tie bars 25, 25,.... The movable platen 24 is configured to be slidable between the fixed platen 22 and the mold clamping housing 23. A mold clamping mechanism 26 is provided between the mold clamping housing 23 and the movable platen 24.

[0034] The mold clamping mechanism 26 is constituted by, for example, a toggle mechanism. It should be noted that the mold clamping mechanism 26 may be constituted by a direct pressure type mold clamping mechanism, that is, a mold clamping cylinder. Fixed molds 21a and movable molds 21b are respectively provided on the fixed platen 22 and the movable platen 24. After the mold clamping mechanism 26 is driven, the mold 21 is opened and closed.

[0035] The injection device 3 is provided on a base 30. The injection device 3 includes: a heating cylinder 31 having a nozzle 31a at its front end; and a screw 32 disposed in the heating cylinder 31 so as to be rotatable in the circumferential and axial directions. The screw 32 is driven by a drive mechanism 33 in the rotational direction and the axial direction. The drive mechanism 33 is constituted by a rotation motor for driving the screw 32 in the rotational direction and a motor for driving the screw 32 in the axial direction, etc. It should be noted that Figure 1 the drive mechanism 33 shown is covered by a cover, so the internal structure is not shown.

[0036] A hopper 34 into which the molding material is charged is provided near the rear end portion of the heating cylinder 31. In addition, the injection molding machine 101 includes a nozzle contact device 35 for moving the injection device 3 in the front-rear direction ( Figure 1 the left-right direction in the figure). After the nozzle contact device 35 is driven, the injection device 3 advances, so that the nozzle 31a of the heating cylinder 31 contacts the close contact portion of the fixed platen 22.

[0037] Figure 2FIG. 0 is a block diagram showing a configuration example of the control device 4 according to the present embodiment. The control device 4 is a computer that controls the operations of the mold clamping device 2 and the injection device 3, and includes a processor 41, a storage unit 42, a control signal output unit 43, a first acquisition unit 44, a second acquisition unit 45, and an operation panel 46 as a hardware structure. The processor 41 has a learner 40 as a functional unit. Note that a part of the learner 40 may be implemented in hardware form.

[0038] The control device 4 is a device that corrects the molding conditions of the injection molding machine 101 (refer to Figure 1 ). Note that the control device 4 may also be a server device connected to a network. In addition, the control device 4 may be configured to perform distributed processing by a plurality of computers, may be implemented by a plurality of virtual machines provided in one server, or may be implemented using a cloud server.

[0039] The processor 41 has an arithmetic circuit such as a CPU (Central Processing Unit), a multi-core CPU, a GPU (Graphics Processing Unit), a GPGPU (General-purpose computing on graphics processing units), a TPU (Tensor Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), an NPU (Neural Processing Unit), etc., an internal storage device such as a ROM (Read Only Memory) and a RAM (Random Access Memory), an I / O terminal, a timing unit, etc. The processor 41 implements the molding condition correction method according to the present embodiment by executing a computer program (program product) 42a stored in the storage unit 42 (described later). Note that each functional unit of the control device 4 may be implemented in software form, or may be partially or entirely implemented in hardware form.

[0040] The storage unit 42 is a non-volatile memory such as a hard disk, an EEPROM (Electrically Erasable Programmable ROM), or a flash memory. The storage unit 42 stores a computer program 42a for strengthening the learning of a method for correcting molding conditions corresponding to the states of the injection molding machine 101 and the molded product, and causes the computer to execute molding condition correction processing. The storage unit 42 stores various coefficients representing the learner 40 as a reinforcement learning model.

[0041] The computer program 42a according to the present embodiment may be in a form that can be read by a computer and stored in the storage medium 49. The storage unit 42 stores the computer program 42a read from the storage medium 49 by a reading device. The storage medium 49 is a semiconductor memory such as a flash memory. In addition, the storage medium 49 may also be an optical disc such as a CD (Compact Disc) -ROM, a DVD (Digital Versatile Disc) -ROM, or a BD (Blu-ray (registered trademark) disc). Further, the storage medium 49 may also be a magnetic disk (such as a floppy disk or a hard disk), a magneto-optical disc, or the like.

[0042] In addition, the computer program 42a according to the present embodiment can be downloaded from an external server connected to a communication network and stored in the storage unit 42.

[0043] The control signal output unit 43 outputs a control signal for controlling the operation of the injection molding machine 101 to the molding machine main body 1 in accordance with the control of the processor 41 based on the molding conditions.

[0044] The operation panel 46 is an interface for setting molding conditions of the injection molding machine 101 and operating the operation of the injection molding machine 101. The operation panel 46 has a display panel and an operation device. The display panel is a display device such as a liquid crystal display panel or an organic EL display panel, and displays a reception screen for receiving the setting of the molding conditions of the injection molding machine 101, or displays the state of the injection molding machine 101, the implementation status of the molding condition correction method according to the present embodiment, etc. according to the control of the processor 41. The operation device is an input device for inputting and correcting the molding conditions of the injection molding machine 101, and has operation buttons, a touch screen, etc. The operation device delivers data representing the received molding conditions to the processor 41.

[0045] The injection molding machine 101 is set with set values related to various molding conditions. The molding conditions include the injection start position, the resin temperature in the mold, the nozzle temperature, the cylinder temperature (heater temperature), the hopper temperature, the clamping force, the injection speed, the injection acceleration, the injection peak pressure (injection pressure), and the injection stroke. The molding conditions also include the resin pressure at the front end of the cylinder, the seating state of the check ring, the holding pressure, the holding pressure switching speed, the holding pressure switching position, the holding pressure completion position, the cushion position, the metering back pressure, and the metering torque. In addition, the molding conditions also include the metering completion position, the screw retraction speed, the cycle time, the mold closing time, the injection time, the holding pressure time, the metering time, and the mold opening time. In addition, the molding conditions also include the cooling time, the screw rotation speed, the mold opening and closing speed, the ejection speed, and the number of ejections.

[0046] The injection molding machine 101 set with these set values operates according to the set parameters. Among the above molding conditions, in particular, the holding pressure, the holding pressure switching position, and the injection speed are related to molding defects such as burrs and material shortages in the molded product. In the present embodiment, an example of using the learner 40 to correct the holding pressure [MPa], the holding pressure switching position [mm], and the injection speed [mm / s] will be described.

[0047] The first acquisition unit 44 is an input circuit that acquires measurement data obtained by measuring the state of the injection molding machine 101.

[0048] One or more sensors 1a are provided in the clamping device 2 and the injection device 3 (see Figure 1 ), and the one or more sensors detect physical quantities necessary for controlling the operation of the molding machine main body 1, which represent the state of the injection molding machine 101. The sensor 1a is connected to the first acquisition unit 44. The physical quantities include, for example, forces such as voltage, current, temperature, humidity, torque, and pressure, the speed, acceleration, rotation angle, position of the movable part, and the flow rate and speed of the fluid. The sensor 1a is, for example, a current sensor, a voltage sensor, a temperature sensor, a humidity sensor, a torque sensor, a pressure sensor, a speed sensor, an acceleration sensor, a rotation angle sensor, a position measurement sensor, a flow sensor, a flow velocity meter, etc. The sensor 1a outputs a measurement signal representing the physical quantity to the control device 4. The measurement signal output from the sensor 1a is input to the first acquisition unit 44, and the first acquisition unit 44 acquires the measurement signal as measurement data.

[0049] The measurement data is data representing the operating condition of the injection molding machine 101. For example, it includes the cycle time [s], the injection time [s], the holding pressure time [s], the holding pressure switching position [mm], the holding pressure switching speed [mm / s], the holding pressure switching pressure [Pa], the cushion position [mm], the holding pressure completion position [mm], the metering time [s], the back pressure [Pa], the metering completion position [mm], etc.

[0050] The second acquisition unit 45 is an input circuit that acquires inspection result data obtained by inspecting the state of a molded product molded by the injection molding machine 101. The inspection device 5 is connected to the second output unit. The inspection device 5 is a camera, a distance measuring sensor, a weighing scale, etc. that detect the state of the molded product, measure physical quantities related to the state of the molded product, and obtain inspection result data indicating the state of the molded product based on the measured physical quantity data. For example, the inspection result data is data indicating the burr area [mm 2 and the short shot area [mm 2 of the molded product.

[0051] Figure 3 is a functional block diagram of the processor 41 according to the present embodiment. The processor 41 of the control device 4 serves as the learner 40. The learner 40 includes an observation unit 41a, a reward calculation unit 41b, an agent 41c, and a correction unit 41d. It should be noted that the learner 40 and each functional unit of the control device 4 can be implemented in software form, or part or all of them can be implemented in hardware form.

[0052] The observation unit 41a acquires measurement data and inspection result data from the first acquisition unit 44 and the second acquisition unit 45, respectively. The observation unit 41a includes: a normalization processing unit 41e that normalizes the measurement data acquired by the first acquisition unit 44; and a normalization processing unit 41f that normalizes the inspection result data acquired by the second acquisition unit 45.

[0053] The normalization processing unit 41e normalizes (Ave-Std scale transformation) the measurement data acquired by the first acquisition unit 44 by using the average value and standard deviation of the measurement data pre-calculated for data normalization before the start of mass production molding. For example, the normalization processing unit 41e normalizes the measurement data obtained by the first acquisition unit 44 as shown in the following formula (1) so that the average value of the measurement data is 0 and the standard deviation is 1. For the value of the normalized measurement data, a predetermined numerical range such as ±3σ can be set between 0 and 255, etc. The scale of the normalized measurement data can be arbitrarily set by confirming the distribution of the measurement data. The scale can be different for each measurement data. In addition, when the normalized measurement data falls outside a predetermined numerical range such as ±3σ, the maximum or minimum value of this numerical range can be used as the normalized measurement data. It should be noted that when the normalized measurement data falls outside a predetermined numerical range such as ±3σ many times, the median or n-th quantile (n is an integer of 3 or more) can be used instead of the average value for normalization.

[0054] X = (x - xave) / σ...(1)

[0055] Where,

[0056] X: Normalized measurement data

[0057] x: Measurement data before normalization

[0058] xave: Average value of measurement data

[0059] σ: Standard deviation of measurement data

[0060] The normalization processing unit 41f normalizes (Min - Max scaling transformation) the measurement data obtained by the second acquisition unit 45 using the maximum and minimum values of the inspection result data determined in advance for data normalization before the start of mass production molding. For example, the normalization processing unit 41f normalizes the inspection result data obtained by the second acquisition unit 45 as shown in the following formula (2) so that the minimum value of the inspection result data is 0 and the maximum value is 1.

[0061] Y = (y - ymin) / (ymax - ymin)...(2)

[0062] Here,

[0063] Y: Normalized inspection result data

[0064] y: Inspection result data before normalization

[0065] ymin: Minimum value of inspection result data

[0066] ymax: Maximum value of inspection result data

[0067] The observation unit 41a outputs the measurement data standardized by the standardization processing unit 41e and the inspection result data normalized by the normalization processing unit 41f as observation data to the agent 41c. And the observation unit 41a outputs the normalized inspection result data to the reward calculation unit 41b. In the case of being configured to calculate the reward data in consideration of the operating conditions of the injection molding machine 101, the observation unit 41a can also be configured to output the standardized measurement data and the normalized inspection result data to the reward calculation unit 41b together.

[0068] The reward calculation unit 41b calculates the reward data obtained by evaluating the currently set molding conditions based on the data observed by the observation unit 41a, especially the inspection result data, and outputs the calculated reward data to the agent 41c. Specifically, the reward calculation unit 41b determines the value of the reward data according to the inspection result data in such a way that the larger the burr area and the underfill area, the smaller the reward value. When the burr area and the underfill area are zero, the reward value is the largest.

[0069] The agent 41c is, for example, a reinforcement learning model with a deep neural network (such as DQN (Deep Q-Network), A3C, D4PG, etc.), a model-based reinforcement learning model such as PlaNet, SLAC, etc. An example in which the agent 41c has DQN will be described below.

[0070] The agent 41c uses DQN to determine an action a corresponding to the state s of the injection molding machine 101 represented by the observation data. The state s related to the observation data includes measurement data and inspection result data.

[0071] DQN is a neural network model that, when input with the state s represented by the observation data, outputs the value of each of multiple actions a. The multiple actions a include items to be corrected and correction amounts. The numerical range of the correction amount is normalized, for example, a value between 0 and 1. An action a with a high value indicates an appropriate correction item and correction amount determined for the condition. The agent 41c selects an action a with a high value, and the selected action a causes the injection molding machine 101 to transition to another state. After the state transition, the agent 41c receives the reward calculated by the reward calculation unit 41b and causes the agent 41c to learn so that the profit, that is, the cumulative reward, is maximized.

[0072] More specifically, DQN has an input layer, an intermediate layer, and an output layer. The input layer includes multiple nodes, and the state s, that is, the observation data, is input to these nodes. The output layer includes multiple nodes, which respectively correspond to the multiple actions a and output the value Q(s, a) of the action a in the input state s.

[0073] Based on the state s, action a, and reward r obtained from this action, taking the value Q represented by the following formula (3) as the correct solution as teaching data, the DQN of the agent 41c can perform reinforcement learning by correcting various weighting coefficients characterizing DQN.

[0074] Q(s,a)←Q(s,a)+α(r+γmaxQ(snext,anext)-Q(s,a))...(3)

[0075] Among them,

[0076] s: State

[0077] a. Action

[0078] α: Learning coefficient

[0079] r: Reward

[0080] γ: Discount rate

[0081] maxQ(snext,next): The maximum value among the Q-values for the next possible actions

[0082] The agent 41c has an inverse transformation unit 41g that inverse-transforms the correction amount of the normalized molding conditions into the actual correction amount. The agent 41c outputs the correction amount inverse-transformed by the inverse transformation unit 41g to the correction unit 41d.

[0083] The inverse transformation unit 41g inverse-transforms the correction amount (Min-Max scaling transformation) using the maximum and minimum values of the correction amount determined in advance for data normalization before the start of mass production molding. For example, as shown in the following formula (4), the inverse transformation unit 41g inverse-transforms the normalized correction amount calculated by the agent 41c into the actual correction amount.

[0084] z = Z × (zmax – zmin) + zmin...(4)

[0085] Here,

[0086] z: Inverse-transformed correction amount

[0087] Z: Normalized correction amount

[0088] zmin: Minimum value of the correction amount

[0089] zmax: Maximum value of the correction amount

[0090] The correction unit 41d corrects the molding conditions based on the correction amount calculated by the agent 41c and provides the corrected molding conditions to the control signal output unit 43. The control signal output unit 43 outputs a control signal corresponding to the corrected molding conditions to the molding machine main body 1.

[0091] <Method for generating and transferring a standard model>

[0092] Figure 4 A flowchart showing the method for generating and transferring the learner 40 (standard model) according to the present embodiment, Figure 5 A conceptual diagram showing the method for transferring the learner 40 (standard model) according to the present embodiment, Figure 6 A conceptual diagram showing the additional learning method of the learner 40 (standard model) according to the present embodiment.

[0093] First, a standard model for correcting molding conditions specifically for specific defects (such as burrs and insufficient material), that is, the normalized and standardized learner 40, is generated (step S11). The learner 40 as the standard model is, for example, as Figure 5 and Figure 6As shown, it is generated in a model generation factory. The model generation factory is a factory different from the mass production factory. For example, the model generation factory is a factory of a company that manufactures and sells injection molding machines 101.

[0094] Figure 7 FIG. 4 is a flowchart showing the processing procedure of the generation learner 40 (standard model). In the present embodiment, it is assumed that the injection molding machine 101 used in the model generation factory is the same as the injection molding machine 101 used in the mass production factory, and is an injection molding machine that generates the learner 40 through the processing of the processor 41.

[0095] By operating the injection molding machine 101 in the model generation factory, measurement data, inspection result data, and correction amounts for standard model learning are collected (step S31).

[0096] Next, the processor 41 calculates the average value and standard deviation of the collected measurement data (step S32). The processor 41 also specifies the maximum value and minimum value of the collected inspection result data (step S33). Similarly, the processor 41 specifies the maximum value and minimum value of the collected correction amounts (step S34).

[0097] Next, the processor 41 normalizes the collected measurement data based on the average value and standard deviation calculated in step S32 (step S35). In addition, the processor 41 normalizes the collected inspection result data based on the maximum value and minimum value calculated in step S33 (step S36). Similarly, the processor 41 normalizes the collected correction amounts based on the maximum value and minimum value calculated in step S34 (step S37).

[0098] The processor 41 performs reinforcement learning on the relationship between the measurement data, inspection result data, and correction amounts for the molding conditions based on the normalized measurement data, inspection result data, and correction amounts, thereby generating the learner 40 as the standard model (step S38). The processor 41 stores various coefficients representing the generated learner 40 as the standard model in the storage unit 42.

[0099] Return to Figures 4 to 6 , the conversion of the standard model will be described. The staff who imports the learner 40 imports the learner 40 generated through the processing of steps S11 and S31 to S38 into the injection molding machine 101 provided in the mass production factory (step S12), and produces molded products (step S13). Step S13 is a molding process for evaluating the operation result of the learner 40 of the standard model. It should be noted that mass production of molded products can be started in step S13.

[0100] Figure 8It is a flowchart showing the import process of the learner 40 (standard model). The staff member who imports the learner 40 transfers the injection molding machine 101 of the mass production factory, which is the learner 40 vector of the standard model, to the injection molding machine 101 (step S51). Specifically, the staff member sends various coefficients characterizing the learner 40 as the standard model to the control device 4 of the injection molding machine 101 and stores them in the storage unit 42. The transfer of the learner 40 as the standard model can be carried out through a wired or wireless communication network, or can be transferred to the control device 4 through a removable recording medium. The mold 21 used by the injection molding machine 101 in the mass production factory is different from the mold 21 used when generating the standard model in the model generation factory.

[0101] Next, before the start of mass production in the mass production factory, the staff member operates the injection molding machine 101 to collect measurement data, inspection result data, and correction amounts for standardization and normalization (step S52). Since it is data for standardizing and normalizing the measurement data, inspection result data, and correction amounts, the injection molding machine 101 can be operated by changing the molding conditions so that the values of each data fluctuate relatively greatly.

[0102] It should be noted that in this embodiment, examples of calculating the average value, standard deviation, minimum value, and maximum value for standardization and normalization are described. However, if the average value, standard deviation, minimum value, and maximum value for standardization and normalization calculated and specified when generating the standard model are appropriate, these values can be directly used.

[0103] In addition, it can also be configured not to produce molded products, but to appropriately set the average value, standard deviation, minimum value, and maximum value for standardization and normalization.

[0104] The processor 41 of the control device 4 calculates the average value and standard deviation of the collected measurement data for standardization (step S53). In addition, the processor 41 designates the maximum value and minimum value of the collected inspection result data for normalization (step S54). Similarly, the processor 41 designates the maximum value and minimum value of the collected correction amount for normalization (step S55). The processor 41 stores the average value, standard deviation, maximum value, and minimum value for standardization and normalization in the storage unit 42 (step S56) and ends the process.

[0105] Figure 9The flowchart shows the molding process. The processor 41 controls the operation of the injection molding machine 101 and performs molding cycle control (step S71). That is, the injection molding machine 101 performs molding to produce molded products. For example, the molding cycle includes well-known processes such as the mold closing process, the mold clamping process, the injection unit advancing process, the injection process, the cooling process, the metering process, the injection unit retracting process, the mold opening process, and the ejection process. The processor 41 controls the operation of the molding machine main body 1 so that these processes are executed in sequence. It should be noted that the process from the mold closing process to the ejection process for manufacturing one molded product is regarded as one cycle, and the time required for one cycle is called the cycle time.

[0106] Next, the processor 41 obtains measurement data and inspection result data by the first acquisition unit 44 and the second acquisition unit 45 (step S72). The processor 41 standardizes the obtained measurement data according to the average value and standard deviation of the measurement data stored in the storage unit 42 (step S73). The processor 41 normalizes the obtained inspection result data according to the maximum value and minimum value of the inspection result data stored in the storage unit 42 (step S74).

[0107] The processor 41 calculates the normalized correction amount of the molding conditions by inputting the standardized and normalized measurement data and inspection result data into the agent 41c of the learner 40 (step S75). Then, the processor 41 performs inverse transformation on the normalized correction amount according to the maximum value and minimum value of the correction amount stored in the storage unit 42 (step S76), and corrects the correction amount according to the inverse-transformed correction amount (step S77). The corrected molding conditions are set, and in the next molding, molding is performed according to the corrected molding conditions.

[0108] After the processing of step S77 is completed, the processor 41 determines whether to end the production of molded products (step S78). In the case where it is determined not to end the production of molded products (step S78: No), the process returns to step S71, and the molding cycle process is repeated. It should be noted that the correction of the molding conditions can be performed each time a molded product is molded, or can be performed every predetermined number of moldings. In the case where it is determined to end the production of molded products (step S78: Yes), the processor 41 ends the control process related to the molding cycle process.

[0109] Back to Figures 4 to 6, processing such as additional learning based on the operation results using the standard model will be described. After the staff who imports the learner 40 imports the standard model and performs shaping in steps S12 and S13, it is judged whether the operation result of the learner 40 is good (step S14). For example, by intentionally setting molding conditions that are likely to cause molding defects and judging whether the defects of the molded product are improved to within a predetermined number of times by correcting the molding conditions, it is judged whether the operation result of the learner 40 is good. In addition, it is also possible to judge whether the operation result of the learner 40 is good based on the defect rate of the molded product produced under the intentionally set molding conditions that are likely to cause molding defects. The method for judging whether the operation result of the learner 40 is good is an example and is not limited to the above method.

[0110] It should be noted that when the performance of the learner 40 as the standard model is high enough, the conversion process of the learner 40 can be ended without performing the process of judging whether the operation result is good and performing additional learning, etc. (steps S14 to S17).

[0111] When it is judged that the operation result is good (step S14: Yes), the process related to the conversion of the learner 40 is ended. After that, as Figure 5 shown, the standard model imported into the injection molding machine 101 in the mass production factory can be directly used to operate the injection molding machine 101.

[0112] When it is judged that the operation result is poor (step S14: No), the staff who imports the learner 40 judges whether the reason for the poor operation result lies in the problem of standardization or normalization (step S15). For example, when many data in the standardized measurement data are values outside 2σ or 3σ, it is judged that there is a problem with standardization. In addition, whether qualified products or defective products are produced, when there is an abnormality in the inspection result data or the correction amount value is concentrated near the minimum or maximum value, it is judged that there is a problem with normalization.

[0113] When it is judged that there is a problem with standardization or normalization (step S15: Yes), as Figure 6 shown, the staff performs the preprocessing of standardization and normalization performed in step S12 again (step S16), and returns the process to step S13.

[0114] When it is judged that there is no problem with standardization or normalization (step S15: No), additional learning of the learner 40 is performed (step S17), and the process is returned to step S13.

[0115] Figure 10It is a flowchart showing the additional learning process of the learner 40 (standard model). By operating the injection molding machine 101 in the mass production factory, measurement data, inspection result data, and correction amounts for additional learning are collected (step S91).

[0116] Next, the processor 41 calculates the average value and standard deviation of the collected measurement data (step S92). In addition, the processor 41 also specifies the maximum value and minimum value of the collected inspection result data (step S93). Similarly, the processor 41 specifies the maximum value and minimum value of the collected correction amounts (step S94).

[0117] Next, the processor 41 normalizes the collected measurement data based on the average value and standard deviation calculated in step S92 (step S95). In addition, the processor 41 also normalizes the collected inspection result data based on the maximum value and minimum value calculated in step S93 (step S96). Similarly, the processor 41 normalizes the collected correction amounts based on the maximum value and minimum value calculated in step S94 (step S97).

[0118] The processor 41 performs additional learning of the learner 40 by performing reinforcement learning on the relationship between the measurement data and inspection result data and the correction amounts of the molding conditions based on the normalized and standardized measurement data, inspection result data, and correction amounts (step S98). The processor 41 stores various coefficients representing the learner 40 that has undergone additional learning in the storage unit 42 (step S99).

[0119] When the learner 40 is used in the molding control process after additional learning, here, the stored average value and standard deviation, maximum value and minimum value can be used to perform the correction process of the molding conditions.

[0120] According to the injection molding machine 101 and the control device 4 (molding condition correction device) of the present embodiment configured as above, etc., the learner 40 as a standard model can be reused to correct the molding conditions of injection molding.

[0121] Basically, the control device 4 of the injection molding machine 101 can directly use the learner 40 of the standard model to correct the molding conditions. In addition, even when the operation result of the learner 40 is not good, the molding conditions can be corrected more effectively by performing additional learning of the learner 40. Since the learner 40 is normalized and standardized, additional learning can be performed with less data compared to the case where it is not normalized and standardized.

[0122] In this way, it is possible to reduce the amount of data collection for learning required when introducing the learning machine 40 in the mass production factory. Since the data for learning can be reduced, the number of defective products produced before the mass production process can be reduced, and the impact on the production plan of the molded products in the mass production factory can be lowered.

[0123] In particular, by standardizing or normalizing the measurement data, inspection result data, and correction amount that are closely related to the correction of the molding conditions of the injection molding machine 101, the learning machine 40 of the standard model can be made into a more general-purpose model. Therefore, the control device 4 of the injection molding machine 101 can more effectively absorb the differences in the mold 21 and correct the molding conditions.

[0124] In addition, by standardizing the measurement data and normalizing the inspection result data and correction amount, the learning machine 40 of the standard model can be made into a more general-purpose model. Since the average value and standard deviation of the measurement data vary depending on the mold 21, the standardization of the measurement data can generalize its characteristics to a greater extent than normalization. In addition, since the inspection result data is the burr area and the underfill area, normalization is more suitable than standardization. Since the correction amount is the correction amount of the molding conditions, normalization is more suitable than standardization. By differentiating the use of standardization and normalization according to the characteristics of the data input to the learning machine 40 and output from it, the generality of the learning machine 40 can be improved.

[0125] In addition, the learning machine 40 of the present embodiment is a general-purpose model capable of correcting molding defects such as burrs and underfills. Even when the mold 21 is replaced, it is possible to correct the molding conditions so that burrs and underfills do not occur only by standardizing and normalizing the data.

[0126] It should be noted that in the present embodiment, the conversion of the learning machine 40 for correcting the molding conditions of the injection molding machine 101 has been described, but the present technology can also be applied to the learning machine 40 for correcting the molding conditions of an extruder or other molding machines.

[0127] In addition, although examples of standardizing and normalizing the measurement data, inspection result data, and correction amount have been described, it can also be configured to standardize or normalize any one or both of these.

[0128] In addition, although examples of performing Ave - Std scale transformation on the measurement data and Min - Max scale transformation on the inspection result data and correction amount have been described, this is an example of the methods of standardization and normalization, and it can also be configured to perform standardization and normalization by other known methods.

[0129] In addition, in the present embodiment, an example is described in which the generation of the standard model and the import of the standard model are performed in different factories, but the generation and import of the standard model may also be performed in the same factory.

[0130] In addition, in the present embodiment, an example is described in which a reinforcement learning model is used to correct the molding conditions, but the learning device 40 may also be configured as a supervised learning model. A learning model related to supervised learning is a model that outputs the most appropriate molding conditions or molding amount when measurement data and inspection result data are input. This learning model can be generated by using learning data in which the most appropriate molding conditions or molding amount are added as teacher data in the measurement data and the inspection result data.

[0131] Means for solving the problems of the present disclosure are noted.

[0132] (Note 1)

[0133] A molding condition correction device for correcting the molding conditions of a molding machine, comprising:

[0134] An acquisition unit that acquires measurement data obtained by measuring the state of the molding machine and inspection result data obtained by inspecting the state of a molded product molded by the molding machine; and

[0135] A learning device that learns the relationship between the measurement data, the inspection result data, and the correction amount of the molding conditions, and determines the correction amount based on the acquired measurement data and inspection result data,

[0136] At least one of the measurement data, the inspection result data, and the correction amount processed by the learning device is standardized or normalized.

[0137] (Note 2)

[0138] In the molding condition correction device according to Note 1,

[0139] The measurement data, the inspection result data, and the correction amount are data that have been standardized or normalized.

[0140] (Note 3)

[0141] In the molding condition correction device according to Note 1 or Note 2,

[0142] The measurement data is data in which the distribution of the data has been standardized,

[0143] The inspection result data and the correction amount are data in which the minimum value and the maximum value have been normalized.

[0144] (Note 4)

[0145] In the molding condition correction device according to any one of Supplementary Notes 1 to 3,

[0146] the acquisition unit acquires the measurement data, the inspection result data, and the correction amount for normalization and standardization before mass production molding;

[0147] the learner:

[0148] calculates the average value and standard deviation of the measurement data according to the measurement data for standardization acquired by the acquisition unit;

[0149] determines the minimum value and maximum value of the inspection result data and the correction amount according to the inspection result data and the correction amount for normalization acquired by the acquisition unit;

[0150] standardizes the measurement data acquired by the acquisition unit during mass production molding according to the average value and standard deviation calculated using the inspection result data for standardization;

[0151] normalizes the inspection result data acquired by the acquisition unit during mass production molding using the maximum value and minimum value determined using the inspection result data for normalization;

[0152] performs an inverse transformation on the normalized correction amount determined by the learner using the maximum value and minimum value determined using the correction amount for normalization.

[0153] (Supplementary Note 5)

[0154] In the molding condition correction device according to any one of Supplementary Notes 1 to 4,

[0155] the acquisition unit acquires the measurement data, the inspection result data, and the correction amount for additional learning,

[0156] the learner:

[0157] calculates the average value and standard deviation of the measurement data according to the measurement data for additional learning;

[0158] standardizes the measurement data for additional learning according to the calculated average value and standard deviation;

[0159] determines the minimum value and maximum value of the inspection result data and the correction amount according to the inspection result data and the correction amount for additional learning;

[0160] normalizes the inspection result data and the correction amount for additional learning according to the determined minimum value and maximum value;

[0161] Perform additional learning using the measured data, the inspection result data, and the correction amount that have been standardized and normalized for additional learning.

[0162] (Supplementary Note 6)

[0163] In the molding condition correction device according to any one of Supplementary Notes 1 to 5,

[0164] The learner:

[0165] Standardize the measured data obtained by the acquisition unit during mass production molding using the average value and standard deviation calculated using the measured data for the additional learning;

[0166] Normalize the inspection result data obtained by the acquisition unit during mass production molding using the maximum value and minimum value determined using the inspection result data for the additional learning,

[0167] Perform inverse transformation on the normalized correction amount determined by the learner using the maximum value and minimum value determined using the correction amount for the additional learning.

[0168] (Supplementary Note 7)

[0169] In the molding condition correction device according to any one of Supplementary Notes 1 to 6,

[0170] The molding machine is an injection molding machine,

[0171] The measured data includes cycle time, injection time, holding pressure time, holding pressure switching position, holding pressure switching speed, holding pressure switching pressure, cushion position, holding pressure completion position, metering time, back pressure, or metering completion position,

[0172] The inspection result data includes the burr area or the short-shot area of the molded product,

[0173] The correction amount includes the correction amount of the holding pressure, the correction amount of the holding pressure switching position, or the correction amount of the injection speed.

[0174] (Supplementary Note 8)

[0175] A molding machine including the molding condition correction device according to any one of Supplementary Notes 1 to 7.

[0176] Explanation of Reference Numerals

[0177] 1: Molding machine main body

[0178] 1a: Sensor

[0179] 2: Mold clamping device

[0180] 3: Injection device

[0181] 4: Control device

[0182] 5: Inspection device

[0183] 40: Learner

[0184] 41: Processor

[0185] 41a: Observation unit

[0186] 41b: Reward calculation unit

[0187] 41c: Agent

[0188] 41d: Correction unit

[0189] 41e: Standardization processing unit

[0190] 41f: Normalization processing unit

[0191] 41g: Inverse transformation unit

[0192] 42: Storage unit

[0193] 42a: Computer program

[0194] 43: Control signal output unit

[0195] 44: First acquisition unit

[0196] 45: Second acquisition unit

[0197] 46: Operation panel

[0198] 49: Recording medium

[0199] 101: Injection molding machine

Claims

1. A forming condition correction device for correcting the forming conditions of a forming machine, comprising: An acquisition unit that acquires measurement data obtained by measuring the state of the molding machine and inspection result data obtained by inspecting the state of a molded product molded by the molding machine; And A learner that learns the relationship between the measurement data, the inspection result data, and the correction amount of the molding conditions, and determines the correction amount based on the acquired measurement data and inspection result data. At least one of the measurement data, the inspection result data, and the correction amount processed by the learner is standardized or normalized.

2. The forming condition correction device according to claim 1, wherein, The measurement data, the inspection result data, and the correction amount are data that have been standardized or normalized.

3. The forming condition correction device according to claim 1 or claim 2, wherein, The measurement data is data whose distribution has been standardized. The inspection result data and the correction amount are data whose minimum and maximum values have been normalized.

4. The forming condition correction device according to any one of claims 1 to 3, wherein, The acquisition unit acquires the measurement data, the inspection result data, and the correction amount for normalization and standardization before mass production molding. The learner: Calculates the average value and standard deviation of the measurement data based on the measurement data for standardization acquired by the acquisition unit. Determines the minimum and maximum values of the inspection result data and the correction amount based on the inspection result data and the correction amount for normalization acquired by the acquisition unit. Standardizes the measurement data acquired by the acquisition unit during mass production molding based on the average value and standard deviation calculated using the inspection result data for standardization. Normalizes the inspection result data acquired by the acquisition unit during mass production molding using the maximum and minimum values determined using the inspection result data for normalization. Performs an inverse transformation on the normalized correction amount determined by the learner using the maximum and minimum values determined using the correction amount for normalization.

5. The forming condition correction device according to any one of claims 1 to 4, wherein, The acquisition unit acquires the measurement data, the inspection result data, and the correction amount for additional learning. The learner: Calculates the average value and standard deviation of the measurement data based on the measurement data for additional learning. Standardizes the measurement data for additional learning based on the calculated average value and standard deviation. Determines the minimum and maximum values of the inspection result data and the correction amount based on the inspection result data and the correction amount for additional learning. Normalizes the inspection result data and the correction amount for additional learning based on the determined minimum and maximum values. Performs additional learning using the measurement data, the inspection result data, and the correction amount for additional learning that have been standardized and normalized.

6. The forming condition correction device according to claim 5, wherein, The learner: Standardizes the measurement data acquired by the acquisition unit during mass production molding using the average value and standard deviation calculated using the measurement data for the additional learning. Normalizes the inspection result data acquired by the acquisition unit during mass production molding using the maximum and minimum values determined using the inspection result data for the additional learning. Inverse-transform the normalized correction amount determined by the learner using the maximum and minimum values determined for the correction amount for the additional learning.

7. The forming condition correction device according to any one of claims 1 to 6, wherein, The molding machine is an injection molding machine. The measurement data includes cycle time, injection time, holding time, holding pressure switching position, holding pressure switching speed, holding pressure switching pressure, cushion position, holding pressure completion position, metering time, back pressure, or metering completion position. The inspection result data includes the burr area or the short-shot area of the molded product. The correction amount includes a correction amount for holding pressure, a correction amount for holding pressure switching position, or a correction amount for injection speed.

8. A molding machine, comprising the molding condition correction device according to any one of claims 1 to 7.

9. A method for correcting molding conditions, for correcting the molding conditions of a molding machine, comprising: A step of obtaining measurement data obtained by measuring the state of the molding machine and inspection result data obtained by inspecting the state of the molded product molded by the molding machine. And A step of determining the correction amount according to the obtained measurement data and inspection result data using a learner that has learned the relationship between the measurement data, the inspection result data, and the correction amount of the molding conditions. At least one of the measurement data, the inspection result data, and the correction amount processed by the learner is standardized or normalized.

10. A computer program for causing a computer to execute a process of correcting the molding conditions of a molding machine, wherein, Cause the computer to execute: A step of obtaining measurement data obtained by measuring the state of the molding machine and inspection result data obtained by inspecting the state of the molded product molded by the molding machine. And A determination step of determining the correction amount according to the obtained measurement data and inspection result data using a learner that has learned the relationship between the measurement data, the inspection result data, and the correction amount of the molding conditions. In the determination step, use the learner in which at least one of the measurement data, the inspection result data, and the correction amount is standardized or normalized to determine the correction amount.

Images