Operation amount determining device, molding device system, molding machine, computer program, operation amount determining method, and state display device

Abstract

The present invention provides an operation quantity determining device for determining the operation quantity related to a molding machine. The operation quantity determining device includes: an observation unit that acquires observation data obtained by observing physical quantities related to the molding process when the molding machine is performing molding; a state display unit that generates a state representation mapping diagram representing the state of the molding machine based on the observation data acquired by the observation unit; and an operation quantity output unit that outputs the operation quantity based on the state representation mapping diagram generated by the state display unit.

Description

Technical Field

[0001] This invention relates to an operation quantity determination device, a molding device system, a molding machine, a computer program, an operation quantity determination method, and a status display device. Background Technology

[0002] Patent document 1 discloses an injection molding apparatus system and a machine learner that uses a reinforcement learner to determine the optimal operating conditions with low power consumption and adjust the operating conditions.

[0003] Patent document 2 discloses an anomaly diagnosis device that diagnoses anomalies in an injection molding machine through machine learning.

[0004] Existing technical documents

[0005] Patent documents

[0006] Patent Document 1: Japanese Invention Patent No. 6346128

[0007] Patent Document 2: Japanese Invention Patent No. 6294268 Summary of the Invention

[0008] The problem the invention aims to solve

[0009] However, the injection molding apparatus system in Patent Document 1 is a system that determines the amount of operation of the molding machine through model-free reinforcement learning. When the environment changes, the strategy for determining the amount of operation may no longer be universal. In such cases, the strategy needs to be relearned from scratch. Learning the strategy requires a large amount of training data or training tasks, resulting in a lack of universality.

[0010] Although the device in Patent Document 2 can perform abnormality diagnosis in injection molding, it cannot determine the optimal operating amount of the molding machine.

[0011] The purpose of this invention is to provide an operation amount determination device, molding device system, molding machine, computer program, operation amount determination method, and status display device that are highly robust to environmental changes and capable of performing additional learning for adaptation to new environments in a short time.

[0012] Solution for solving the problem

[0013] The present invention provides an operation quantity determining device for determining operation quantities relating to a molding machine. The operation quantity determining device includes: an observation unit that acquires observation data obtained by observing physical quantities relating to the molding process during molding; a state representation unit that generates a state representation mapping diagram representing the state of the molding machine based on the observation data acquired by the observation unit; and an operation quantity output unit that outputs the operation quantity based on the state representation mapping diagram generated by the state representation unit.

[0014] The molding apparatus system of the present invention includes the above-described quantity determination device and molding machine.

[0015] The molding machine of the present invention includes the above-described operation amount determining device and operates according to the operation amount determined by the operation amount determining device.

[0016] The computer program of the present invention causes a computer to determine an operational quantity relating to a molding machine, the computer program causing the computer to perform the following processes: acquiring observation data obtained by observing physical quantities relating to the molding process while the molding machine is molding; generating a state representation map representing the state of the molding machine based on the acquired observation data; and outputting the operational quantity based on the generated state representation map.

[0017] The operation quantity determination method of the present invention determines the operation quantity related to a molding machine by obtaining observation data obtained from observing physical quantities related to the molding process when the molding machine is performing molding; generating a state representation mapping diagram representing the state of the molding machine based on the obtained observation data; and outputting the operation quantity based on the generated state representation mapping diagram.

[0018] The status display device of the present invention includes: an observation unit that acquires observation data obtained by observing physical quantities related to the molding process during molding in a molding machine; and a display processing unit that displays a mapping image and the actual molding result of the molding machine in a comparative manner, the mapping image relating to a status representation mapping image that represents the status of the molding machine based on the observation data acquired by the observation unit.

[0019] Invention Effects

[0020] As mentioned above, it exhibits excellent robustness to environmental changes and is capable of rapid additional learning to adapt to new environments. Attached Figure Description

[0021] Figure 1 This is a block diagram illustrating a structural example of the molding apparatus system according to Embodiment 1.

[0022] Figure 2 This is a functional block diagram of the molding apparatus system of Embodiment 1.

[0023] Figure 3 It is a block diagram representing the functions and input / output data of the status display unit.

[0024] Figure 4 It is a block diagram representing the function of the operational output section and the input / output data.

[0025] Figure 5 This is a schematic diagram representing an example of a mapping image.

[0026] Figure 6 These are schematic diagrams representing other examples of mapping graph images.

[0027] Figure 7 These are schematic diagrams representing other examples of mapping graph images.

[0028] Figure 8 These are schematic diagrams representing other examples of mapping graph images.

[0029] Figure 9 These are schematic diagrams representing other examples of mapping graph images.

[0030] Figure 10 These are schematic diagrams representing other examples of mapping graph images.

[0031] Figure 11 It is a timing diagram that shows the operation of the device determined by the amount of operation during the learning phase.

[0032] Figure 12 It is a timing diagram that shows how the operation of the device is determined by the amount of operation in the application stage.

[0033] Figure 13 This is a block diagram showing the molding machine of Embodiment 2. Detailed Implementation

[0034] Hereinafter, specific examples of the operation amount determination device, molding apparatus system, molding machine, computer program, operation amount determination method, and status display device according to embodiments of the present invention will be described with reference to the accompanying drawings. At least some of the embodiments described below can be combined arbitrarily. Furthermore, the present invention is not limited to these examples, but rather includes all modifications within the meaning and scope equivalent to the claims, as shown in the claims.

[0035] (Implementation Method 1)

[0036] Figure 1 This is a block diagram illustrating a structural example of the molding apparatus system according to Embodiment 1. Figure 2 This is a functional block diagram of the molding apparatus system of Embodiment 1.

[0037] The molding apparatus system of Embodiment 1 includes an operation amount determination device 1, a molding machine 2, and a measuring unit 3. The operation amount determination device 1 functions as a status display device.

[0038] The molding machine 2 is, for example, an injection molding machine, a blow molding machine, a film molding machine, an extruder, a two-screw extruder, a spinning extruder, a granulator, a magnesium injection molding machine, etc. Hereinafter, the molding machine 2 described in Embodiment 1 is an injection molding machine. The injection molding machine includes an injection unit and a mold clamping device disposed in front of the injection unit. The injection unit consists of a heated barrel, a screw disposed within the heated barrel that can be driven in both the rotational and axial directions, a rotary motor that drives the screw in the rotational direction, and a motor that drives the screw in the axial direction. The mold clamping device includes a toggle mechanism and a motor that drives the toggle mechanism. The toggle mechanism opens and closes the mold, clamping it when molten resin injected from the injection unit is filled into the mold, preventing the mold from opening.

[0039] Molding machine 2 is set with specified operating parameters for molding conditions, including: resin temperature inside the mold, nozzle temperature, barrel temperature, hopper temperature, clamping force, injection speed, injection acceleration, peak injection pressure, injection stroke, resin pressure at the front end of the barrel, check ring seat status, holding pressure switching pressure, holding pressure switching speed, holding pressure switching position, holding pressure end position, buffer position, metering back pressure, metering torque, metering completion position, screw retraction speed, cycle time, mold closing time, injection time, holding time, metering time, and mold opening time. Molding machine 2 operates according to these operating parameters. The optimal operating parameters vary depending on the environment of molding machine 2 and the molded product.

[0040] The measuring unit 3 is a device that measures physical quantities related to the molding process during molding in the molding machine 2. The measuring unit 3 outputs the physical quantity data obtained through measurement processing to the operation quantity determination device 1. The physical quantities include temperature, position, velocity, acceleration, current, voltage, pressure, time, image data, torque, force, strain, power consumption, etc.

[0041] The information measured by the measuring unit 3 includes, for example, information about the molded product, molding conditions (measured values), peripheral equipment settings (measured values), and atmosphere information. This peripheral equipment is equipment that constitutes a system linked to the molding machine 2, including a mold clamping device or mold. Examples of peripheral equipment include a molded product removal device (robot), an insert insertion device, a film feeding device for in-mold transfer, a tape conveyor for tape molding, a gas injection device for gas-assisted molding, a gas injection device and long fiber injection device for foam molding using supercritical fluids, a material mixing device for LIM molding, a deflashing device for the molded product, a runner cutting device, a molded product weighing meter, a molded product strength testing machine, an optical inspection device for the molded product, a camera and image processing device for the molded product, and a robot for handling the molded product.

[0042] Molded product information includes, for example, camera images of the molded product, deformation data obtained from a laser displacement sensor, optical measurements such as color and brightness obtained from an optical measuring instrument, weight measured by a weighing meter, and strength measured by a strength measuring instrument. This information indicates whether the molded product is normal, the type of defect, and the severity of the defect, and is also used in reward calculations.

[0043] Molding conditions include information measured by thermometers, pressure gauges, speed measuring devices, acceleration measuring devices, position sensors, timers, and weighing meters, such as the resin temperature inside the mold, nozzle temperature, barrel temperature, hopper temperature, clamping force, injection speed, injection acceleration, peak injection pressure, injection stroke, resin pressure at the front end of the barrel, check ring seat status, holding pressure switching pressure, holding pressure switching speed, holding pressure switching position, holding pressure end position, buffer position, metering back pressure, metering torque, metering completion position, screw retraction speed, cycle time, mold closing time, injection time, holding pressure time, metering time, and mold opening time.

[0044] The peripheral equipment settings include information such as mold temperature (set to a fixed value) and mold temperature (set to a variable value), and pellet supply, which are measured by thermometers, measuring instruments, etc.

[0045] Atmospheric information includes atmospheric temperature, atmospheric humidity, and information related to convection (Reynolds number, etc.) obtained from thermometers, hygrometers, flow meters, etc.

[0046] In addition, the measuring unit 3 can also measure the mold opening amount, return flow rate, tie rod deformation amount, and heater heating rate.

[0047] The operation quantity determination device 1 is a computer, and its hardware structure includes a processor 10, a storage unit 11, and input / output interfaces (not shown). The processor 10 has arithmetic circuits such as a CPU (Central Processing Unit), multi-core CPU, GPU (Graphics Processing Unit), GPGPU (General-purpose computing on graphics processing units), TPU (Tensor Processing Unit), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), and NPU (Neural Processing Unit), internal storage devices such as ROM (Read Only Memory) and RAM (Random Access Memory), and I / O terminals. The processor 10 performs the functions of the initial information acquisition unit 12, the physical quantity acquisition unit 13, the control unit 14, the learner 15, and the mapping output unit 16 by running the computer program 11a stored in the storage unit 11 (described later). Furthermore, each functional unit of the operation quantity determination device 1 can be implemented in software, or some or all of them can be implemented in hardware.

[0048] In addition, the operation quantity determination device 1 may also be a server device connected to a network not shown.

[0049] Storage unit 11 is a non-volatile memory such as a hard disk, EEPROM (Electrically Erasable Programmable ROM), or flash memory. Storage unit 11 stores a computer program 11a for causing the computer to implement the operation quantity determination method of Embodiment 1. Additionally, storage unit 11 stores a state representation mapping generated by learner 15 (described later). Figure 11 b.

[0050] The computer program 11a in this embodiment can also be in the form of being computer-readablely recorded on the recording medium 4. The storage unit 11 stores the computer program 11a read from the recording medium 4 by a readout device (not shown). The recording medium 4 is a semiconductor memory such as flash memory. Alternatively, the recording medium 4 can also be an optical disc such as a CD (Compact Disc)-ROM, DVD (Digital Versatile Disc)-ROM, or BD (Blu-ray Disc). The recording medium 4 can also be a floppy disk, a hard disk, a magneto-optical disk, or the like. In addition, the computer program 11a of this embodiment can be downloaded from an external server (not shown) connected to a communication network (not shown) and stored in the storage unit 11.

[0051] The initial information acquisition unit 12 acquires initial information about the molding machine 2 or the molded product, which is a prerequisite for determining the operating quantity of the molding machine 2. The initial information includes, for example, molding machine information, mold information, resin information, etc. The initial information acquisition unit 12 outputs the acquired initial information to the learner 15. The molding machine information includes information indicating the type or characteristics of the molding machine 2, the clamping unit, the injection unit, the barrel, the screw, and the nozzle.

[0052] Mold information includes the shape of the molded product, the weight of the molded product, the number of ejections, the flow length, the thickness of the molded product, the resin pressure, and the resin temperature.

[0053] Resin information includes type, grade, manufacturer, viscosity, crystallinity, glass transition temperature, melting point, and composite materials.

[0054] Furthermore, it is difficult to generate a state representation mapping by strictly quantifying all initial information. Figure 11 b. Or, a mapping representation of the initial information construction state. Figure 11 b. Therefore, it is preferable to define the required initial information using a unique heat representation.

[0055] The physical quantity acquisition unit 13 acquires the physical quantity data measured and output by the measurement unit 3 during molding in the molding machine 2. The physical quantity acquisition unit 13 outputs the acquired physical quantity data to the control unit 14.

[0056] like Figure 2 As shown, the control unit 14 includes an observation unit 14a, a reward calculation unit 14b, and an operation quantity correction unit 14c. The observation unit 14a, the reward calculation unit 14b, and the operation quantity correction unit 14c are input with physical quantity data output from the measurement unit 3.

[0057] The observation unit 14a observes the state of the molding machine 2 by analyzing physical quantity data and outputs the observed data to the state display unit 15a of the learner 15. Because the physical quantity data contains a large amount of information, the observation unit 14a can generate observation data that compresses the information of the physical quantity data. The observation data contains information such as the operating state of the molding machine 2 and the state of the molded product.

[0058] For example, the observation unit 14a calculates observation data such as feature quantities representing the appearance characteristics of the molded article, the dimensions, area, volume of the molded article, and the optical axis offset of optical components (molded article) based on camera images and measurements from a laser displacement sensor. Furthermore, the observation unit 14a can preprocess time-series waveform data such as injection speed, injection pressure, and holding pressure to extract feature quantities from the time-series waveform data as observation data. Additionally, time-series waveform data and image data representing time-series waveforms can also be used as observation data.

[0059] The reward calculation unit 14b calculates the reward data based on the physical quantity data and outputs the calculated reward data to the status display unit 15a of the learner 15. This reward data is the benchmark for the quality of the operation in the current state of the molding machine 2.

[0060] For example, the reward calculation unit 14b calculates molding results indicating whether the molded product is normal, the degree of defect, and the type of defect by analyzing physical quantity data such as camera images, laser displacement sensor measurements, and weighing scale measurements. Examples of defect types include warpage / deformation, weld lines, flash, short shots, shrinkage marks, voids, gloss defects, uneven color, black spots / contamination, charring, flow marks, silver streaks, jetting marks, and discoloration. The degree of defect includes the extent of the defect and its occurrence rate. Furthermore, the reward calculation unit 14b calculates reward data indicating whether the environment of the molding machine 2 and the operating parameters set for the molding machine 2 are appropriate based on whether the molded product is normal and the degree of defect. The reward data may be, for example, numerical data where a normal state of the molded product is set to 1, and a defective state of the molded product is set to less than 1.

[0061] Reward data can also be a vector representing the state of a finished product according to defect type. For example, reward data can be a vector where the state of no defect of that type is 1, and the state of having a defect of that type is less than 1.

[0062] In addition, although an example of calculating the molding result by analyzing physical quantity data was given, the molding result can also be received by the operator through an operation panel with buttons, touch panel, etc.

[0063] The operation amount correction unit 14c corrects the operation amount output from the learner 15 as needed, and outputs the corrected operation amount to the molding machine 2. For example, if an upper limit value, a lower limit value, etc. are set for the operation amount, the operation amount can be corrected so that the value related to the molding conditions does not exceed the upper limit value or the lower limit value.

[0064] If no correction is required, the operation amount correction unit 14c will output the operation amount output from the learner 15 to the molding machine 2 as is.

[0065] Learner 15 learns the state representation mapping of the state of molding machine 2. Figure 11 b (environment model) is used to map the state representation. Figure 11 b. The amount of operation is determined by the model reinforcement learning. For example... Figure 2 As shown, the learner 15 has a state representation unit 15a, a state representation learning unit 15b, and an operation output unit 15c. The learned state representation mapping... Figure 11 b is stored in storage unit 11.

[0066] The molding apparatus system of this embodiment 1 has a learning state representation mapping. Figure 11 b's learning phase and usage state representation mapping Figure 11 b. Optimize the operation volume for the molding application stage. The molding device system can switch between the learning stage and the application stage via an operation panel (not shown).

[0067] Figure 3 This is a block diagram representing the functions and input / output data of the status display unit 15a.

[0068] The mapping is represented in the learning state. Figure 11 In the case of learning stage b, such as Figure 3 As shown, the state representation unit 15a is input with initial information output from the initial information acquisition unit 12, observation data output from the observation unit 14a, reward data output from the reward calculation unit 14b, and operation quantity output from the operation quantity output unit 15c. The state representation unit 15a includes a state representation learning unit 15b, which learns a state representation mapping based on the input initial information, observation data, operation quantity, and reward data. Figure 11 b.

[0069] State representation mapping Figure 11 For example, model b outputs the reward g for setting the action a in state s and the state transition probability (confidence) Pt for the next state s′, given arbitrary initial information, observation data (state s), and an action a. The reward g can be considered as information about whether the finished product obtained in state s is normal when a certain action a is set. State representation mapping... Figure 11b. Preferably, the state of the molding machine 2 is represented by data with a lower dimension than the observed data.

[0070] In addition, state representation mappings can also be prepared according to the type of initial information. Figure 11 b.

[0071] The state representation learning unit 15b generates or updates the state representation mapping based on empirical data (state s, action a, next state s′, reward g) or historical data used as learning data. Figure 11 b. For example, the state representation learning unit 15b can use maximum likelihood estimation, Bayesian inference, etc., to calculate the state transition probability Pt. The state transition probability Pt is equivalent to the value obtained by dividing the number of visits n to (state s, action a, next state s′) by the number of visits Σn to (state s, action a, any next state s′∈S). In addition, the state representation unit 15a can use maximum likelihood estimation, Bayesian inference, etc., to calculate the reward g (information representing the quality of the finished product). The reward g is equivalent to the value obtained by dividing the reward G of (state s, action a) by the number of visits Σn to (state s, action a, any next state s′).

[0072] Additionally, state representation mapping Figure 11 b can also be constructed using a learned model that utilizes a neural network. A neural network is a known structure with an input layer, one or more hidden layers, and an output layer. The state representation learning unit 15b enables the neural network to learn such that, given the input of learning data (state s, action a) to the neural network, the output from the neural network is (next state s′, reward g), which is (state transition probability Pt for transitioning to the next state s′, reward g).

[0073] When the number of states represented by the initial state and observation data is large, the state and state representation map can be approximated using parameters smaller than the degrees of freedom of the state.

[0074] In the state representation mapping used Figure 11 In the application phase where molding machine 2 is running, the status display unit 15a is input with initial information, observation data, and the operation quantity output from the operation quantity output unit 15c. The status display unit 15a inputs the initial information, observation data, and operation quantity representing the current status into the status display mapping. Figure 11 b. Calculate the state representation data and output the state representation data to the operation output unit 15c. The state representation data represents the state transition probability Pt and reward g of transitioning to the next state s′ from the current state.

[0075] Furthermore, during the learning and application phases, the state representation unit 15a maps the state representations. Figure 11b, along with the initial information and observation data, is output to the mapping output unit 16.

[0076] The operation quantity output unit 15c determines the maximum operation quantity based on the state representation data output from the state representation unit 15a, and outputs the determined operation quantity to the operation quantity correction unit 14c and the state representation unit 15a. For example, the operation quantity output unit 15c determines the operation quantity using known methods such as dynamic programming or linear programming, which are based on the value iteration method.

[0077] Figure 4 This is a block diagram illustrating the function and input / output data of the operation quantity output unit 15c. The operation quantity output unit 15c includes a switching unit 15d, a first evaluation unit 15e, a second evaluation unit 15f, and an operation quantity determination unit 15g.

[0078] When the switching unit 15d is in the application phase, it outputs the status representation data to the first evaluation unit 15e; when it is in the learning phase, it outputs the status representation data to the second evaluation unit 15f.

[0079] The first evaluation unit 15e has a first objective function for adjusting the operating amount to achieve a state in which a normal molded product can be obtained. The first evaluation unit 15e calculates an evaluation value as the expected return (discount cumulative reward) by inputting state representation data and operating amount into the first objective function. The expected return is the expected value of the reward that can be obtained in the future.

[0080] The second evaluation unit 15f has a second objective function, which is used to adjust the operation amount to change the state of the molded product, in order to explore the state representation mapping. Figure 11 b. The second evaluation unit 15f calculates the evaluation value by inputting state representation data and operation quantity into the second objective function. For example, the more unknown the molding result of the state and operation quantity of the molding machine 2, that is, the fewer the number of trials, the larger the evaluation value. In addition, the second evaluation unit 15f can also use exploratory methods such as the so-called ε-greedy method or UCB1 to calculate the evaluation value.

[0081] The operation quantity determination unit 15g determines the operation quantity with the maximum evaluation value calculated by the first evaluation unit 15e when the application phase is in progress, and determines the operation quantity with the maximum evaluation value calculated by the second evaluation unit 15f when the learning phase is in progress. The operation quantity output unit 15c outputs the operation quantity determined by the operation quantity determination unit 15g to the status display unit 15a and the operation quantity correction unit 14c.

[0082] The operation amount determination unit 15g can determine the operation amount so that the change in the operation amount for each step in the learning phase is greater than the change in the operation amount for each step in the application phase. Alternatively, the operation amount determination device 1 can also be configured to accept the operator's setting of the change in the operation amount for each step via an operation panel (not shown). The operation amount determination unit 15g indicates mapping in the update state. Figure 11 In case b, use the accepted change amount to change the operation amount, and explore the state representation mapping. Figure 11 b. Update. If there are significant changes in the mold, molding machine 2, peripheral equipment type, or the physical properties of the resin, the change amount of the operation quantity during the learning phase can be set to be large.

[0083] The mapping output unit 16 includes a mapping image generation unit 16a, a drawing unit 16b, and a display processing unit 16c.

[0084] The mapping image generation unit 16a generates a mapping of state representations. Figure 11 b. A visualized map image (see reference) Figure 5 , Figure 6 For example, the mapping image generation unit 16a can map according to the state representation. Figure 11 b generates the function f(X, Y). (X, Y) is, for example, a representative value obtained by approximating any state (initial information, observed data) and operation quantity of molding machine 2 as a two-dimensional representation. That is, the representative value (X, Y) represents information such as the molding conditions and environment set for molding machine 2. Function f returns a value indicating whether the molded product obtained under the operation quantity of a certain state set for molding machine 2 is normal or defective. The value indicating whether the molded product is normal is, for example, mapped to the state representation. Figure 11 Input b is the value equivalent to the output reward data for that state and operation. The function f(X, Y) can be visualized, for example, as a contour plot.

[0085] The drawing unit 16b performs the process of drawing the actual state (X1, Y1) during molding onto the mapping image (see reference). Figure 5 , Figure 6 In addition, the drawing unit 16b performs processing that displays the actual molding result, i.e., whether the molded product is normal, in a recognizable manner using different drawing images (see reference). Figure 7 Furthermore, the drawing unit 16b performs a process of drawing multiple states (X1, Y1), (X2, Y2), ... on a mapping image during multiple actual molding processes, and displaying an image representing the direction of change of that state on the mapping image (see reference). Figure 8 For example, the change in the center of gravity of a drawing group can be displayed as an arrow image.

[0086] The display processing unit 16c performs the process of displaying a mapping image, a mapping image showing the actual state when it has been formed, etc., on a display unit (not shown).

[0087] Figure 5 This is a schematic diagram representing an example of a mapping image. In the diagram, the horizontal and vertical axes represent the aforementioned representative values ​​(X, Y). Figure 5 In the diagram, the thick lines represent the boundary between molding conditions that yield normal molded products and molding conditions that yield defective molded products. Hereinafter, the states (initial information, observation data) and operational quantities will be simply referred to as molding conditions. Furthermore, the line segments on the mapping image that represent the boundary between molding conditions that yield normal molded products and molding conditions that yield defective molded products will be simply referred to as boundary lines.

[0088] For example, the mapping image generation unit 16a can distinguish between molding conditions where the value of function f is above a predetermined threshold and a normal molded product can be obtained, and molding conditions where the value of function f is below a predetermined threshold and a defective product can be obtained, using two different colors. Furthermore, using two colors is one example; as long as the molding conditions for obtaining a normal molded product and the molding conditions for obtaining a defective molded product can be displayed in different forms, any other known method can be used for display.

[0089] Figure 5 In the diagram, the black dots represent the actual molding conditions that were achieved.

[0090] Figure 6 These are schematic diagrams illustrating other examples of mapped images. The horizontal and vertical axes are... Figure 5 Similarly, it represents the value (X, Y). Figure 6 In the model, the normal and defective parts of the molded product, and the degree of defect, are displayed using a contour map. Specifically, the mapping image generation unit 16a generates a contour map based on the reward g.

[0091] Figure 7 These are schematic diagrams illustrating other examples of mapped images. The horizontal and vertical axes are... Figure 5 Similarly, it represents the value (X, Y). Figure 7 In the middle, the upper right part (normal area) represents the molding conditions under which a normal molded product can be obtained, and the lower left part (defect area) represents the molding conditions under which a normal molded product cannot be obtained.

[0092] The circles and X marks on the mapping image indicate the molding conditions during multiple molding processes in the past, and show whether the molded product was normal. A circle indicates a normal molded product, and an X mark indicates a non-normal molded product.

[0093] like Figure 7 As shown, even under molding conditions that would yield a normal molded product, molding defects still occur during actual molding.

[0094] exist Figure 7 In this diagram, circle and X marks are examples of images used to indicate whether a molded product is normal. Other graphic images or text images can also be used, as long as they can indicate whether a molded product is normal or the degree of its defects. Additionally, the state of the molded product can be represented by color differentiation.

[0095] Figure 8 These are schematic diagrams illustrating other examples of mapped images. The horizontal and vertical axes are... Figure 7 Similarly, the representative values ​​(X, Y) are represented. The drawing of the circle markers indicates the molding conditions when molding was actually performed multiple times in the past. In addition, the drawing unit 16b displays the trajectory of the molding condition changes, or in other words, the history of the center of gravity changes of the molding condition group, as an arrow image overlaid on the mapping image. More specifically, the time-averaged velocity of the center of gravity on the mapping image drawn in the actual molding process can be calculated, and an arrow image representing the average velocity is overlaid on the mapping image.

[0096] Figure 9 This is a schematic diagram illustrating other examples of mapped images. The mapped image generation unit 16a is capable of displaying a status representation of the mapped image. Figure 11 The history of changes in the forming conditions during the learning process of b, and the changes in the boundary lines in the mapping image.

[0097] Mapping in the learning state representation Figure 11 At step b, in order to explore the environmental model, learner 15 explores and changes the forming conditions. Drawing unit 16b maps the state representation. Figure 11 The overlapping molding conditions during the learning process of b are displayed on the mapping image. Additionally, arrows representing changes in molding conditions are shown.

[0098] Furthermore, when the position of the boundary line changes during the learning process, the mapping image generation unit 16a overlaps the boundary line before and after learning on the mapping image. Additionally, the mapping image generation unit 16a displays arrows indicating the direction of the boundary line change on the mapping image.

[0099] Figure 10 These are schematic diagrams illustrating other examples of mapping imagery. Thick solid lines represent boundary lines in the mapping imagery. Discontinuous lines represent the confidence intervals of these boundary lines.

[0100] It goes without saying that the confidence interval can also be shown in the above. Figures 5 to 9 The state shown represents a mapping. Figure 11 b on.

[0101] Figures 5 to 10The text illustrates examples of showing the state of obtaining a normal molded product and the state of obtaining a defective product, but it can also identifiably display the state of obtaining a defective product of what type of defect.

[0102] For example, the reward data can be set as a vector representing the state of the molded product according to the defect type, and the state representation can be mapped. Figure 11 b is configured to output reward data as a vector when given the input state and operation amount. Furthermore, the function f(X, Y) can also be a function that returns the values ​​(f1, f2, f3, ..., fN) representing the state of the molded product obtained when an operation amount indicating a certain state is set on the molding machine 2, according to the defect type. fi (i is a natural number from 1 to N representing the defect type) is, for example, a value indicating whether a defect of defect type i can appear on the molded product. fi = 1 represents a state where a defect of that type does not appear, and f1 = 0 represents a state where a defect of that type appears.

[0103] The mapping image generation unit 16a generates a mapping image based on the values ​​of f1, f2, ... fN. The mapping image represents the molding conditions represented by the representative values ​​(X, Y) in a way that can identify the state in which a normal molded product can be obtained and the multiple states in which defects of various defect types occur.

[0104] With this configuration, the operator can visually identify the state in which a defect of a certain type can occur from the mapping image.

[0105] Figure 11 This is a timing diagram showing how the amount of operation during the learning phase determines the action of device 1. Figure 11 In the diagram, the area enclosed by the dotted line on the left represents the real-world environment, namely the actions of the molding machine 2 and the measurement actions of the measuring unit 3. The area enclosed by the discontinuous line in the center represents the processing actions of the control unit 14, and the area enclosed by the discontinuous line on the right represents the processing actions of the learner 15.

[0106] When molding is performed in molding machine 2, measuring unit 3 measures physical quantities related to molding machine 2 and molded product, and outputs the measured physical quantity data to observation unit 14a (step S11).

[0107] The observation unit 14a acquires physical quantity data output from the measurement unit 3, generates observation data based on the acquired physical quantity data, and outputs the generated observation data to the status display unit 15a (step S12).

[0108] State representation unit 15a acquires observation data output from observation unit 14a, and applies the observation data and initial state, etc., to the state representation mapping. Figure 11b. Generate state representation data and output the generated state representation data to the operation quantity output unit 15c (step S13).

[0109] The operation quantity output unit 15c determines the operation quantity of the molding machine 2 based on the status display data output from the status display unit 15a, and outputs the determined operation quantity to the status display unit 15a and the operation quantity correction unit 14c (steps S14 and S15). For example, the operation quantity output unit 15c determines the operation quantity with the maximum evaluation value obtained from the second objective function, as described above.

[0110] Similar to step S11, the measuring unit 3 outputs the measured physical quantity data to the operation quantity correction unit 14c (step S16). The operation quantity correction unit 14c corrects the operation quantity as needed based on the physical quantity data and outputs the corrected operation quantity to the molding machine 2 (step S17). The molding machine 2 sets the operation quantity and performs molding processing according to that operation quantity. The physical quantities related to the operation of the molding machine 2 and the molded product are input to the measuring unit 3 (step S18). The molding process can also be repeated multiple times.

[0111] When molding is performed in molding machine 2, measuring unit 3 measures physical quantities related to molding machine 2 and molded product, and outputs the measured physical quantity data to observation unit 14a and reward calculation unit 14b (steps S19 and S20).

[0112] The observation unit 14a acquires physical quantity data output from the measurement unit 3, generates observation data based on the acquired physical quantity data, and outputs the generated observation data to the status display unit 15a (step S21).

[0113] The reward calculation unit 14b calculates the reward data based on the physical quantity data measured by the measurement unit 3, according to whether the molded product is normal and the degree of defect, and outputs the calculated reward data to the status display unit 15a (step S22).

[0114] The state representation learning unit 15b of the state representation unit 15a updates the state representation model based on the observation data output from the observation unit 14a, the reward data output from the reward calculation unit 14b, the operation quantity output from the operation quantity output unit 15c, and the initial information (step S23). The state representation learning unit 15b can update the state representation model using, for example, maximum likelihood estimation or Bayesian inference.

[0115] Figure 12 This is a timing diagram showing the operation of the device 1, which determines the amount of operation during the application stage. When the molding machine 2 is molding, the measuring unit 3 measures the physical quantities related to the molding machine 2 and the molded product, and outputs the measured physical quantity data to the observation unit 14a (step S31).

[0116] The observation unit 14a acquires physical quantity data output from the measurement unit 3, generates observation data based on the acquired physical quantity data, and outputs the generated observation data to the status display unit 15a (step S32).

[0117] State representation unit 15a acquires observation data output from observation unit 14a, and applies the observation data and initial state, etc., to the state representation mapping. Figure 11 b. Generate status representation data and output the generated status representation data to the operation quantity output unit 15c (step S33).

[0118] The operation quantity output unit 15c determines the operation quantity of the molding machine 2 based on the status display data output from the status display unit 15a, and outputs the determined operation quantity to the status display unit 15a and the operation quantity correction unit 14c (steps S34 and S35). For example, the operation quantity output unit 15c determines the operation quantity that maximizes the expected return (discount accumulation reward) obtained from the first objective function, as described above.

[0119] Similar to step S31, the measuring unit 3 outputs the measured physical quantity data to the operation quantity correction unit 14c (step S36). The operation quantity correction unit 14c corrects the operation quantity as needed based on the physical quantity data and outputs the corrected operation quantity to the molding machine 2 (step S37). The molding machine 2 sets the operation quantity and performs molding processing according to the operation quantity. The physical quantities related to the operation of the molding machine 2 and the molded product are input to the measuring unit 3 (step S38).

[0120] When the environment in which the molding device system is used changes, the operator can appropriately switch the action mode from the application phase to the learning phase to update the state representation mapping. Figure 11 b. The operation quantity determination device 1, which switches from the application phase to the learning phase, can execute... Figure 11 The process shown is used to update the state representation mapping. Figure 11 b.

[0121] Even when the environment changes, the state representation mapping Figure 11 The content of b will not change drastically. In most cases, it is sufficient to correct the boundary between the molding conditions that yield a normal molded product and the molding conditions that yield a defective molded product. Therefore, the operation quantity output unit 15c performs the state representation mapping to be updated by outputting the operation quantity near this boundary. Figure 11 The exploration of region b. Furthermore, the operational output unit 15c can use the operational start state, equivalent to the boundary line on the mapping image, to represent the mapping. Figure 11 The exploration of b. Then, the state representation learning unit 15b updates the state representation mapping based on the empirical or historical data obtained during the exploration process. Figure 11 b.

[0122] The molding apparatus system, operation quantity calculation device, operation quantity calculation method, and computer program 11a according to Embodiment 1 have excellent robustness to environmental changes and can perform additional learning to adapt to new environments in a short time.

[0123] According to Embodiment 1, the operation quantity determination device 1 is able to generate a state representation mapping that takes into account the initial information involving the molding machine 2 or the molded product. Figure 11 b.

[0124] According to this embodiment 1, the state representation mapping can be updated even when the environment changes. Figure 11 b. Adapt to environmental changes.

[0125] According to this embodiment 1, during the learning phase, the state representation mapping can be updated efficiently by outputting an operation quantity such as a change in the shaping result. Figure 11 b.

[0126] According to this implementation method 1, the state representation mapping is explored during the learning phase. Figure 11 b increases the magnitude of changes in the operational quantity, enabling efficient updates to the state representation mapping. Figure 11 b.

[0127] According to this embodiment 1, the strategy related to the determination of operational quantities can be changed during the learning phase and the application phase. State representation mapping can be performed efficiently during the learning phase. Figure 11 The update of b, and the efficient adjustment of the operation amount to the state that can obtain the molded product during the application stage.

[0128] According to Embodiment 1, the mapping output unit 16 is capable of mapping state representations. Figure 11 b is displayed visually as a mapping image. The mapping image can identify and display the molding conditions that result in normal molded products and the molding conditions that result in defective molded products.

[0129] Furthermore, the mapping output unit 16 can draw the actual current forming conditions onto the state representation mapping. Figure 11 Displayed on b.

[0130] Furthermore, the mapping output unit 16 can display different drawn images to show whether the molded product obtained under the actual molding conditions is normal or defective. The operator can intuitively identify whether the current molding conditions are in a state that can produce a normal molded product, and to what extent the current molding conditions have deviated from the state that can produce a normal molded product.

[0131] Furthermore, the mapping output unit 16 can display the boundary lines of the state of a normal molded product and the state of a molded product with defects, and map the state representation. Figure 11 The changes in the boundary lines formed by the learning of b are displayed on the mapping image. The operator can intuitively identify the state representation mapping. Figure 11 b's learning situation.

[0132] (Implementation Method 2)

[0133] The molding machine of Embodiment 2 differs from that of Embodiment 1 in that the operation amount determining device is located within the molding machine; therefore, the following mainly describes the aforementioned differences. Other structures and effects are the same as in Embodiment 1; therefore, the same reference numerals are used in the corresponding parts, and detailed descriptions are omitted.

[0134] Figure 13 This is a block diagram showing the molding machine 202 of Embodiment 2. The molding machine 202 of Embodiment 2 includes an injection unit 221, a mold clamping unit 222 disposed in front of the injection unit 221, and a control device 220 for controlling the operation of the molding machine 202. The control device 220 includes the operation amount determination device 1 described in Embodiment 1.

[0135] According to embodiment 2, the molding machine 202 is capable of learning state representation mapping. Figure 11 b. Determine the amount of operation to reduce the degree of defects, and operate according to the determined amount of operation.

[0136] Explanation of reference numerals in the attached figures:

[0137] 1: Operating quantity determination device

[0138] 2: Molding machine

[0139] 3: Measurement Section

[0140] 4: Recording media

[0141] 10: Processor

[0142] 11: Storage Department

[0143] 11a: Computer program

[0144] 11b: State Representation Map

[0145] 12: Initial Information Acquisition Department

[0146] 13: Physical Quantity Acquisition Section

[0147] 14: Control Department

[0148] 14a: Observation Department

[0149] 14b: Reward Calculation Department

[0150] 14c: Operational quantity correction unit

[0151] 15: Learning device

[0152] 15a: Status Display Section

[0153] 15b: State Representation Learning Department

[0154] 15c: Operational output unit

[0155] 15d: Switching Unit

[0156] 15e: First Evaluation Department

[0157] 15f: Second Evaluation Department

[0158] 15g: Operational dosage determination department

[0159] 16: Mapping diagram output section

[0160] 16a: Mapping image generation unit

[0161] 16b: Drawing Department

[0162] 16c: Display Processing Unit

Claims

1. An operation amount determining device for determining the operation amount related to a molding machine, the operation amount determining device comprising: The observation unit acquires observation data obtained by observing physical quantities related to the molding process during the molding process in the molding machine; A state representation unit generates a state representation mapping diagram representing the state of the molding machine based on the observation data obtained by the observation unit; An operation quantity output unit outputs the operation quantity based on a state representation mapping generated by the state representation unit. The reward calculation unit calculates reward data that reflects the molding result of the molding machine based on the observation data obtained by the observation unit. and A state representation learning unit updates the state representation mapping based on the observation data obtained when the operation amount is set and the reward data calculated by the reward calculation unit. The state representation mapping is a model that, when given arbitrary initial information, the observed data, and the operation quantity, outputs a reward for the operation quantity and a state transition probability for the state corresponding to the observed data.

2. The operation quantity determination device according to claim 1, This includes an initial information acquisition unit that acquires initial information about the molding machine or molded article that serves as a basis for determining the operation amount. The state representation unit generates a state representation mapping diagram representing the state based on the observation data obtained by the observation unit and the initial information obtained by the initial information acquisition unit.

3. The operation quantity determining device according to claim 1, comprising: The state representation storage unit stores the state representation mapping diagram.

4. The operation quantity determination device according to claim 3, When the state representation mapping is updated, the operation quantity output unit outputs the operation quantity that changes the molding result of the molding machine.

5. The operation quantity determining device according to claim 4, The operation output unit includes: The first evaluation unit, when input with a state representation mapping diagram output from the state representation unit, outputs an evaluation value for the operation quantity. The higher the probability of transitioning to a state in which a normal molded product can be obtained, the higher the evaluation value. The second evaluation unit outputs an evaluation value for the operation quantity when a state representation mapping diagram output from the state representation unit is input. The more unknown the state of the molding machine and the molding result for the operation quantity are, the higher the evaluation value will be. A switching unit that switches between a first evaluation unit and a second evaluation unit for determining the operation amount; The operation amount determination unit determines the operation amount based on the evaluation value output by the first evaluation unit or the second evaluation unit switched from the switching unit.

6. The operation quantity determination device according to any one of claims 1 to 5, The system includes a display processing unit that displays the state representation map and the actual molding result of the molding machine in a comparable manner.

7. A molding apparatus system, comprising: The operation quantity determination device according to any one of claims 1 to 6, and Molding machine.

8. A molding machine, Includes the operation quantity determination device according to any one of claims 1 to 6, The molding machine operates according to the operating amount determined by the operating amount determining device.

9. A computer program product comprising a computer program that causes a computer to determine operating amounts relating to a molding machine, the computer program causing the computer to perform the following processes: Obtain observational data by observing physical quantities involved in the molding process during the molding process in the molding machine; A state representation mapping diagram representing the state of the molding machine is generated based on the obtained observation data; The operation quantity is output based on the generated state representation mapping; Based on the obtained observation data, reward data representing the molding result of the molding machine is calculated; and The state representation mapping is updated based on the observed data obtained when the operation amount is set and the calculated reward data. in, The state representation mapping is a model that, when given arbitrary initial information, the observed data, and the operation quantity, outputs a reward for the operation quantity and a state transition probability to the next state in the state corresponding to the observed data.

10. A method for determining the amount of operation involved in a molding machine. Obtain observational data by observing physical quantities involved in the molding process during the molding process in the molding machine; A state representation mapping diagram representing the state of the molding machine is generated based on the obtained observation data; The operation quantity is output based on the generated state representation mapping; Based on the obtained observation data, reward data representing the molding result of the molding machine is calculated; and The state representation mapping is updated based on the observed data obtained when the operation amount is set and the calculated reward data. in, The state representation mapping is a model that, when given arbitrary initial information, the observed data, and the operation quantity, outputs a reward for the operation quantity and a state transition probability to the next state in the state corresponding to the observed data.

11. A status display device, comprising: The observation unit acquires observation data obtained by observing physical quantities related to the molding process during molding in the molding machine; and The display processing unit displays a mapping image and the actual molding result of the molding machine in a comparative manner. The mapping image refers to a state representation mapping image that represents the state of the molding machine based on the observation data obtained by the observation unit. The state representation mapping is a model that, when given arbitrary initial information, the observed data, and the operation quantity, outputs a reward for the operation quantity and a state transition probability to the next state in the state corresponding to the observed data.

12. The status display device according to claim 11, The mapping image contains images that can identifiablely represent states where normal molded products can be obtained and states where defective molded products can be obtained.

13. The status display device according to claim 11, The system includes a drawing unit that displays a drawing image representing the actual molding state of the molding machine on the mapping image.

14. The status display device according to claim 13, The drawing unit displays the drawn image on the mapping image, and the drawn image varies depending on whether the actual molded product is normal.

15. The status display device according to claim 14, When the drawing unit performs multiple actual molding processes, it displays multiple drawn images representing the states of each molding process and an image representing the changes in the drawn images displayed on the mapping image on the mapping image.

16. The status display device according to any one of claims 11 to 15, The display processing unit, The boundary lines are displayed on the mapping image, representing the boundaries between the state where a normal molded product can be obtained and the state where a defective molded product can be obtained. If the state representation map has been updated, the boundary lines before and after the update are displayed.

Images