JP2021154657A - Computer program, machine learning machine, information processing method, and molder - Google Patents

Abstract

To provide a molder capable of determining a setting display screen that should be displayed next so that screens transition in a favorable order for an operator to be able to easily use the molder.SOLUTION: A molder of this invention which displays a setting display screen on a display device, the setting display screen selected among plural setting display screens, comprises: an acquisition section that acquires a transition history of the setting display screens according to an operator's operation for operating the molder; and a learning section that learns a setting display screen that should be displayed next based on the transition history of the plural setting display screens acquired.SELECTED DRAWING: Figure 1

Description

The present invention relates to computer programs, machine learning machines, information processing methods and molding machines.

The display screen of the injection molding machine includes a large number of screens such as an operation setting value display screen, a machine operation monitoring information display screen, a production monitoring information display screen, and a warning display screen. Normally, these screens are layered with the menu screen as the top screen. The operator needs to switch screens by at least the number of layers in order to reach the desired screen, which causes a problem of poor operability.

Patent Document 1 discloses a technique of using AI technology to change the arrangement order of operation screen transition icons according to the frequency of use of icons operated by an operator.

Patent Document 2 discloses a technique for highlighting a setting item of molding conditions when the setting item of a molding condition is changed a plurality of times by an operator.

Patent Document 3 discloses a technique for immediately reading and displaying a required screen according to the state of an injection molding machine.

Japanese Unexamined Patent Publication No. 2017-138881 Japanese Unexamined Patent Publication No. 2010-52289 Japanese Unexamined Patent Publication No. 11-254497

An object of the present disclosure is to determine the setting display screen to be displayed next so that the screens transition in a convenient order by the operator by learning the transition history of the setting display screen of the molding machine operated by the operator. Provide computer programs, machine learning machines, display methods and molding machines that can be used.

The computer program according to the present disclosure is a computer program for causing a computer to display a setting display screen selected from a plurality of setting display screens for operating a molding machine on a display device, and is a setting display operated by an operator. The computer is made to execute a process of determining the setting display screen to be displayed next based on the screen transition history.

The machine learning machine according to the present disclosure is displayed next based on the acquisition unit that acquires the transition history of the setting display screen by the operation of the operator for operating the molding machine and the transition history of the acquired plurality of setting display screens. It is equipped with a learning unit that learns the setting display screen.

The information processing method according to the present disclosure is an information processing method for displaying a setting display screen selected from a plurality of setting display screens for operating a molding machine on a display device, and is a transition of the setting display screen by an operator's operation. Based on the history, the setting display screen to be displayed next is determined, and the determined setting display screen is displayed on the display device.

The molding machine according to the present disclosure is a molding machine that displays a setting display screen selected from a plurality of setting display screens on a display device, and displays a transition history of the setting display screens operated by an operator for operating the molding machine. It includes an acquisition unit to be acquired and a learning unit to learn the setting display screen to be displayed next based on the transition history of a plurality of acquired setting display screens.

According to the above, by learning the transition history of the setting display screen of the molding machine operated by the operator, it is possible to determine the setting display screen to be displayed next so that the screens transition in a convenient order by the operator. can.

It is a schematic diagram explaining the structural example of the injection molding machine which concerns on Embodiment 1. FIG. It is a block diagram which shows the hardware composition of the control device which concerns on Embodiment 1. FIG. It is a block diagram which shows the structural example of the machine learning machine which concerns on Embodiment 1. It is a block diagram which shows the structural example of the learning model which concerns on Embodiment 1. FIG. It is a flowchart which shows the processing procedure of reinforcement learning which concerns on Embodiment 1. It is a flowchart which shows the processing procedure of the screen display which concerns on Embodiment 1. It is a schematic diagram which shows the screen display method which concerns on Embodiment 1. FIG. It is a flowchart which shows the AI screen transition processing. It is a schematic diagram which shows an example of the transition history of a setting display screen. It is a schematic diagram which shows an example of the transition history of a setting display screen. It is a schematic diagram which shows the 1st example of the operation part for AI screen transition. It is a schematic diagram which shows the 2nd example of the operation part for AI screen transition. It is a schematic diagram which shows the 3rd example of the operation part for AI screen transition. It is a schematic diagram which shows the structural example of an injection molding system.

The computer program, machine learning machine, display method, and molding machine according to the embodiment of the present disclosure will be described below with reference to the drawings. It should be noted that the present disclosure is not limited to these examples, and is indicated by the scope of claims, and is intended to include all modifications within the meaning and scope equivalent to the scope of claims. In addition, at least a part of the embodiments described below may be arbitrarily combined.

(Embodiment 1)
FIG. 1 is a partial cross-sectional front view showing a configuration example of the injection molding machine 100 according to the first embodiment. The injection molding machine 100 according to the first embodiment includes an injection device 1, a mold clamping device 2 arranged in front of the injection device 1, a control device 3, a display device 4, and an operator information reading device 5. Be prepared.

The injection device 1 drives a heating cylinder, a screw provided so as to be driveable in the rotation direction and the axial direction in the heating cylinder, a rotary motor for driving the screw in the rotation direction, and a screw in the axial direction. It is composed of a motor and the like.

The mold clamping device 2 drives a toggle mechanism that opens and closes the mold and tightens the mold so that the mold is not opened when the molten resin injected from the injection device 1 is filled in the mold, and the toggle mechanism. It is equipped with a motor. The configuration of the injection molding machine 100 is not limited to the above. Further, although the present embodiment describes an injection molding machine 100 which is a kind of molding machine as an example, the molding machine of the present invention is not limited to the injection molding machine, and other than a press molding machine, a laminated molding machine and the like. It does not exclude the molding machine of.

FIG. 2 is a block diagram showing a hardware configuration of the control device 3 according to the first embodiment.
The control device 3 is a computer and includes a control unit 3a, a main storage unit 3b, an input / output I / F3c, and an auxiliary storage unit 3d.

The control unit 3a includes one or more CPUs (Central Processing Units), MPUs (Micro-Processing Units), GPUs (Graphics Processing Units), GPGPUs (General-purpose computing on graphics processing units), and TPUs (Tensor Processing Units).
It is configured by using an arithmetic processing unit such as.

The main storage unit 3b is a temporary storage area for SRAM (Static Random Access Memory), DRAM (Dynamic Random Access Memory), flash memory, etc., and temporarily stores data necessary for the control unit 3a to execute arithmetic processing. Remember.

As shown in FIGS. 1 and 3, the input / output I / F 3c is an interface to which the injection device 1, the display device 4, the operator information reading device 5, and the like are connected. The control unit 3a acquires state information indicating the state of the injection device 1 from the injection device 1 via the input / output I / F3c. The state information is information indicating, for example, the presence or absence of an abnormality in the injection device 1. In the present invention, the input / output I / F3c corresponds to an acquisition unit that acquires the transition history of the setting display screen.

Further, the control unit 3a outputs the screen information to the display device 4 via the input / output I / F 3c, so that the display device 4 displays the screen related to the screen information. As shown in FIG. 1, the display device 4 includes a touch panel 41 and a hardware key 42. For example, the setting display screen 41a and the menu screen 41b are displayed on the touch panel 41. The setting display screen 41a has, for example, a tab 41c for changing the setting display screen 41a. The operator can also switch the setting display screen 41a by touching the tab 41c. The menu screen 41b includes a hierarchical screen transition operation unit 41d and an AI screen transition operation unit 41e. The operator can switch the setting display screen 41a forming the hierarchical structure by operating the hierarchical screen transition operation unit 41d. That is, by operating the operation unit 41d for transition of the hierarchical screen, the operator can switch the setting display screen 41a of the upper layer to the setting display screen 41a of the lower layer, or from the setting display screen 41a of the lower layer to the setting display screen of the upper layer. It is possible to switch to 41a or switch to the top-level setting display screen 41a. Depending on the size of the touch panel 41, the setting display screen 41a can be arranged vertically on two screens. Further, the menu screen 41b may not be displayed at all times.

Further, the control unit 3a acquires the information read by the operator information reading device 5 via the input / output I / F3c. The operator information reading device 5 is, for example, a card reader that reads an IC card in which an operator ID is recorded, a keyboard for inputting an operator ID, a touch panel 41, and the like. get. The operator information reading device 5 is not limited to a card reader, a keyboard, or the like, and may be any device that can acquire information that can identify the operator, such as face authentication, fingerprint authentication, and vein authentication. Further, the operator information reading device 5 is not essential to the present invention, and may simply detect that the operator is in the operating position in front of the display device 4.

The auxiliary storage unit 3d is a storage device such as a hard disk, EEPROM (Electrically Erasable Programmable ROM), and a flash memory. The auxiliary storage unit 3d stores various data necessary for processing of the computer program P executed by the control unit 3a and the control unit 3a. Further, the auxiliary storage unit 3d stores the learning model 6, the standard hierarchical structure 7, and the operator DB 8.

The learning model 6 includes an operator ID, an operator attribute, a molded product information indicating a molded product, a mold information indicating a mold, or a molding machine state information indicating the state of the molding machine, and a setting display screen 41a operated by the operator. This is a model in which the setting display screen 41a to be displayed next is learned according to the transition history. Details will be described later.

The standard hierarchical structure 7 is information for hierarchically associating a plurality of setting display screens 41a necessary for operating the injection molding machine 100 and the like. The setting display screen 41a includes, for example, a screen for inputting a set value to the injection molding machine 100, a screen for monitoring the operation of the injection molding machine 100, a screen for performing production control, a screen for notifying an abnormality, and the like. , Functions or uses are layered (see FIG. 7).

The operator DB 8 is a database that stores operator IDs, operator attributes, operator names, and the like in association with each other. The operator attributes include information indicating the authority level, process in charge, possession qualification, skill level, etc. assigned to each operator, and information that can identify the operator such as face authentication, fingerprint authentication, and vein authentication.

The auxiliary storage unit 3d may be an external storage device connected to the control device 3. The computer program P may be written in the auxiliary storage unit 3d at the manufacturing stage of the control device 3, or the control device 3 acquires what is distributed by the remote server device by communication and stores it in the auxiliary storage unit 3d. You may. The computer program P may be readable and recorded on a recording medium 30 such as a magnetic disk, an optical disk, or a semiconductor memory.
Further, in the first embodiment, an example in which the control device 3 performs the learning process of the learning model 6 will be described, but as described in the second embodiment, the learning process may be performed by a device other than the control device 3. good. In this case, the data related to the learned learning model 6 may be provided in the form of distribution via the network, or may be provided in the form recorded on the recording medium 30, as in the computer program P. ..

In the control device 3 according to the first embodiment configured in this way, the control unit 3a reads out and executes the computer program P stored in the auxiliary storage unit 3d, so that the input / output unit 31 and the molding machine control processing unit 32, the machine learning machine 33, the image display processing unit 34, and the like are realized as software-like functional units.

The input / output unit 31 communicates with the injection molding machine 100 and acquires molding machine state information indicating the state of the molding machine. Further, the input / output unit 31 communicates with the operator information reading device 5 to acquire the operator ID. By referring to the operator DB8 using the operator ID as a key, the attributes of the operator can be obtained. Further, the input / output unit 31 communicates with the display device 4 to acquire the molded product information indicating the molded product, the mold information indicating the mold, and the switching operation information of the setting display screen 41a by the operator's operation. By acquiring the switching operation information of the setting display screen 41a by the operator over time, the transition history of the setting display screen 41a can be obtained.

The molding machine control processing unit 32 controls the operation of the injection molding machine 100.

The machine learning machine 33 is a functional unit that learns the setting display screen 41a to be displayed next from the operator's attributes and the transition history of the setting display screen 41a so far. The machine learning machine 33 has a function of a learning unit that learns the setting display screen to be displayed next based on the acquired transition history of the plurality of setting display screens.

The image display processing unit 34 displays the setting display screen 41a and the menu screen 41b on the display device 4, and executes a process of switching the setting display screen 41a according to the operation of the operator.

FIG. 3 is a block diagram showing a configuration example of the machine learning machine 33 according to the first embodiment. FIG. 4 is a block diagram showing a configuration example of the learning model 6 according to the first embodiment. The machine learning machine 33 learns the display order of the setting display screen 41a by, for example, reinforcement learning, and includes an observation unit 33a, a learning model 6, a setting display screen selection unit 33b, a reward calculation unit 33c, and the like. It includes an update processing unit 33d.

The observation unit 33a observes a state quantity indicating the current state. The state quantity is, for example, an operator ID, an operator attribute, a molded product information indicating a molded product, a mold information indicating a mold, or a molding condition information stored in association with the mold information, or a state of a molding machine. It includes the molding machine state information shown and the transition history of the setting display screen 41a operated by the operator. The transition history is, for example, a vector amount indicating the setting display screen 41a displayed from a certain point in time to the present time by the operation of the operator. For example, when the operation of the injection molding machine 100 is started and the screen A, the screen B, and the screen C are operated in order, the transition history can be expressed as (P1, P2, P3,0,0, ...). .. Note that "0" indicates that the operation has not been performed yet. P1, P2, P3 ... Are identifiers indicating the setting display screen 41a. The number of dimensions of the vector is not particularly limited.

The learning model 6 is, for example, DQN (Deep Q Network). The learning model 6 is a model that outputs an action value for evaluating the setting display screen 41a to be displayed next on the display device 4 when a state quantity indicating the current state is input.
As shown in FIG. 4, the learning model 6 has the value of displaying the input layer 61 into which the state quantity is input, the intermediate layer 62 for recognizing the state quantity, and then the specific setting display screen 41a (behavior). It has an output layer 63 that outputs an action value indicating.
The input layer 61 includes at least one of an operator ID, an operator attribute, a molded product information indicating a molded product, a mold information indicating a mold, a molding machine state information indicating the state of the molding machine, and a setting display screen operated by the operator. It has a plurality of neurons that accept the input of the transition history of 41a, and passes the input information to the intermediate layer 62. The intermediate layer 62 compresses each information input to the input layer 61 and extracts the characteristics of the current state. The output layer 63 has a plurality of neurons corresponding to the plurality of setting display screens 41a, and each neuron outputs the action value of transitioning to each setting display screen 41a. The neural network including the neurons is not limited to the above. For example, a neural network having two or more intermediate layers 62 and performing deep learning may be provided.

The setting display screen selection unit 33b is a functional unit that selects one or more setting display screens 41a to be displayed next based on the action value output from the learning model 6. Normally, the setting display screen 41a selects the display candidate setting display screen 41a in descending order of action value.

The reward calculation unit 33c calculates the reward based on the transition history of the setting display screen 41a switched by the operation of the operator, and outputs the calculated reward to the update processing unit 33d.

The update processing unit 33d machine-learns the learning model 6 based on the action value calculated before the action and the TD error obtained by the evaluation after the action based on the reward output from the reward calculation unit 33c. The action value update formula is expressed by the following formula (1).

Q (st, at) ← Q (st, at) + α (R + γmaxQ (st + 1, at + 1) -Q (st, at) ... (1)
However,
st: State quantity at time t: Action at time t (displays a specific setting display screen 41a)
Q (st, at): Action value α: Learning coefficient R: Immediate reward γ: Time discount rate max Q (st + 1, at + 1): Value of the next step Maximum action value at the time of action selection

The update processing unit 33d optimizes the weighting coefficient of the learning model 6 by an error backpropagation method, an error gradient descent method, or the like so that the TD error, that is, the second term on the right side of the above equation (1) approaches zero. The learning model 6 is machine-learned.

Although Q-learning has been described as a machine learning method, other value iterative methods such as sarsa may be used. Further, the policy gradient method or the like may be used as the machine learning method. Further, although DQN has been described as the learning model 6, a table in which the state quantity and the action value are associated with each other may be used.

FIG. 5 is a flowchart showing a processing procedure of reinforcement learning according to the first embodiment. FIG. 5 shows a method of machine learning the learning model 6 by batch processing. By batch processing, the learning model 6 can be pre-learned to some extent so that the setting display screen 41a is displayed in a standard order.

First, the control unit 3a collects the state quantities of a plurality of episodes (step S11). Here, the transition history of the setting display screen 41a performed by using the hierarchical screen transition operation unit 41d may be collected.

The control unit 3a calculates the reward for the transition history of the collected setting display screen 41a (step S12). The calculation method of the reward will be described.
When the control unit 3a transitions from a certain state or one setting display screen 41a to another setting display screen 41a, the control unit 3a gives a positive reward according to the number of times. Alternatively, the reward for transitioning from the setting display screen 41a to another setting display screen 41a is not performed by associating only between the two screens, but by grouping the display order of the setting display screens 41a of three or more screens. Positive rewards may be given to groups that are frequently displayed by the operator.
When the control unit 3a transitions to one setting display screen 41a in a certain state and the display time of the setting display screen 41a is less than a predetermined time, the setting display screen 41a may not need to be displayed. Assuming that the screen does not need to be displayed, a negative reward is given to such a screen transition. The predetermined time is, for example, within 3 seconds. On the contrary, when the display time of the setting display screen 41a is longer than a predetermined time, a positive reward may be applied to the screen transition. Alternatively, the setting display screen that is displayed but the setting is not input may be given a negative reward as a screen that may not be displayed or a screen that does not need to be displayed. However, the setting display screen in which the setting is not input does not include a screen for displaying only.
Further, when the control unit 3a transitions to one setting display screen 41a in a certain state, or when the operation is switched to the operation using the hierarchical screen transition operation unit 41d, the setting display screen 41a may not need to be displayed. Assuming that it was a certain screen, a negative reward is given to such a screen transition.
Further, the reward may be calculated according to a series of operations of the injection molding machine 100, for example, a screen operation time until the setup operation is completed, and the number of transition screens. For example, as the number of transition screens is smaller, a positive reward should be given assuming that there are no extra screen transitions. Further, for example, a positive reward may be given to the screen transition when the time until the setting input such as the setup operation is completed is short.

Further, the control unit 3a manually receives the reward with the touch panel 41 or the hardware key 42 (step S13). For example, when the injection device 1 is in a state quantity in which the injection device 1 is in a predetermined state, the control unit 3a gives a large reward to the screen transition to the predetermined setting display screen 41a. The predetermined state is, for example, a state at the time of starting up the injection molding machine 100, a state at the end of continuous molding, a state at the time of occurrence of an abnormality, and the like. By giving the reward in this way, when the injection device 1 is in a certain predetermined state, the predetermined setting display screen 41a is displayed.

Further, for example, the control unit 3a displays the reduced images of a series of a plurality of setting operation screens displayed by the operator's operation side by side on the display device 4, and accepts the rearrangement, deletion, etc. of the setting operation screens on the touch panel 41. , The reception result may be accepted as a reward.
For example, when the specific setting display screen 41a is deleted, a negative reward is given to the screen transition to the specific setting display screen 41a while the previous setting display screen 41a is displayed.
For example, when the order of the first setting display screen 41a displayed next to the specific setting display screen 41a and the second setting display screen 41a displayed next to the specific setting display screen 41a is changed, the specific setting is received. A negative reward is set for the screen transition from the display screen 41a to the first setting display screen 41a, and a positive reward is set for the screen transition from the specific setting display screen 41a to the second setting display screen 41a.

Next, the control unit 3a machine-learns the learning model 6 based on the calculated or received reward (step S14), and finishes the process. As described above, machine learning is performed by optimizing the weighting coefficient of the learning model 6 so that the TD error approaches zero.

FIG. 6 is a flowchart showing a screen display processing procedure according to the first embodiment, and FIG. 7 is a schematic view showing a screen display method according to the first embodiment. The control unit 3a determines whether or not the hierarchical screen transition operation unit 41d has been operated (step S31). When it is determined that the hierarchical screen transition operation unit 41d has been operated (step S31: YES), the control unit 3a selects and selects the setting display screen 41a based on the standard hierarchical structure 7 stored in the auxiliary storage unit 3d. The set setting display screen 41a is displayed on the display device 4 (step S32).

For example, as shown in the upper hierarchical structure (standard) of FIG. 7, a case where the screen A, the screen C, the screen D, the screen B, and the screen E are sequentially displayed will be described. The method of determining the setting display screen to be displayed next based on the hierarchical structure and the operator's operation is the first method. First, the operator operates the setting input screen transition icon to transition to the setting input selection screen. .. Next, by the operation of the operator, the screen A is transitioned to the lower layer screen A, and the screen C is further transitioned to the lower layer screen C. Next, in order to display the screen D, the screen A is once returned to the upper layer, and the screen D is changed to the lower layer. Next, in order to display the screen B and the screen E, the screen once returns to the setting input screen of the upper layer, transitions to the screen B of the lower layer, and further transitions to the screen E of the lower layer.

When it is determined that the hierarchical screen transition operation unit 41d is not operated, or when the operator operates the transition operation button in the setting display screen 41a to transition to another screen (step S31: NO), or step S32. When the process of is completed, the control unit 3a determines whether or not the AI screen transition operation unit 41e has been operated (step S33). When it is determined that the AI screen transition operation unit 41e has been operated (step S33: YES), the AI screen transition process for selecting and displaying the setting display screen 41a to be displayed next is executed using the learning model 6 (step S33: YES). Step S34). The method of determining the setting display screen to be displayed next based on the transition history of the setting display screen operated by the operator is the second method, and either the first method or the second method is selected by the operator. Can be accepted. Further, the setting input by the second method may be switched from the middle of the setting by the first method, or the setting input by the first method may be switched from the middle of the setting by the second method.

For example, as shown in the continuous screen (AI) at the bottom of FIG. 7, if the learning of the learning model 6 is completed, the operation of going up and down the hierarchical structure is unnecessary, and the control unit 3a The screen A, the screen C, the screen D, the screen B, and the screen E can be sequentially transitioned and displayed on the display device 4. However, even when the setting display screen 41a is displayed based on the learning result of the learning model 6, the operator can interrupt display another setting display screen 41a (for example, screen F after screen D) by input work. Then, when the screen F is interrupted after the screen D, it is reflected in the learning model 6. That is, reinforcement learning is performed according to the number of times the screen F is interrupted after the screen D, and when the predetermined number of times is exceeded, the screen F is displayed after the screen D. Further, even if an abnormality occurs while the operator is displaying the setting display screen 41a by the learning model 6 and the screen G for confirming the abnormality is opened, if the number of times the screen G is opened is extremely small, the screen by the initial learning model 6 is used. Does not affect the display order.

When it is determined that the AI screen transition operation unit 41e is not operated (step S33: NO), or when the processing of step S34 is completed, the control unit 3a determines whether or not to end the processing (step). S35). If it is determined that the process is not completed (step S35: NO), the process is returned to step S31. When it is determined that the process is finished (step S35: YES), the process is finished.

8 is a flowchart showing the AI screen transition process, FIGS. 9 and 10 are schematic views showing an example of the transition history of the setting display screen 41a, and FIG. 11 is a first example of the AI screen transition operation unit 41e. It is a schematic diagram. The process of learning the transition of the setting display screen 41a desired by the operator while executing the process of estimating and displaying the setting display screen 41a to be displayed next will be described using the learning model 6.

The control unit 3a acquires the current state quantity (step S51). Then, by inputting the acquired state quantity into the learning model 6, the action value of the plurality of setting display screens 41a is calculated (step S52). Then, the control unit 3a determines a candidate for the setting display screen 41a based on the calculated action value (step S53). Here, the setting display screen 41a having the highest action value is determined. Then, as shown in FIG. 11, the control unit 3a instructs the icon 41f indicating the setting display screen 41a candidate to be displayed next, the icon 41g indicating the previously displayed setting table screen candidate, and the transition to the next screen. The forward icon 41h and the back icon 41i for instructing the transition to the previous screen are displayed on the display device 4 (step S54).

Then, the control unit 3a accepts the selection operation operated by the operator (step S55), transitions to the selected setting display screen 41a, and displays the display on the display device 4 (step S56). For example, in the example shown in FIG. 11, when the advance icon 41h is operated by the operator, the mold opening / closing setting screen is displayed on the display device 4.

Next, the control unit 3a calculates the reward based on the operation result of the operator (step S57). For example, if the display time of the setting display screen 41a operated by the operation of the forward icon 41h is less than a predetermined time, it is presumed that the operator did not need to see the setting display screen 41a, so a negative reward is set. The method of granting the reward is the same as that of batch processing.

Next, the control unit 3a machine-learns the learning model 6 by the reward calculated in step S57 (step S58), and finishes the process.

By repeating the above process, the setting display screen 41a is displayed in the order desired by the operator. For example, according to the learning model 6 that learned the work time from the mold setup to the molding, as shown in FIGS. 9 and 10, the condition reading screen, the temperature setting screen, the setup screen, the material purge screen, the mold opening / closing setting screen, The weighing setting screen, the injection setting screen, the production quantity setting screen, and the like can be displayed by transitioning in this order. The operator can transition the setting display screen 41a in the above order simply by operating the forward icon 41h.

The candidate screen prepared by the machine-learned learning model 6 and displayed next may have a mode in which candidates (for example, icons 41f and 41g) are not shown as shown in FIG. Specifically, the transition to the next screen may be performed by simply touching an icon or key such as "to the next screen" by the operator. Further, the transition to the next screen may be performed by the operator swiping or flicking the currently displayed setting display screen 41a.

These machine learning learning models 6 may be constructed by grasping the operators by the operator information reading device 5 and customizing them for each operator. Alternatively, the standard learning model 6 may be constructed by adding up all the screen display orders output by a plurality of operators.

FIG. 12 is a schematic view showing a second example of the AI screen transition operation unit 41e. The AI screen transition operation unit 41e according to the second example is configured so that a plurality of setting display screens 41a can be selected as candidates for the setting display screen 41a to be displayed next. Specifically, the control unit 3a selects a plurality of setting display screens 41a in descending order of action value, and displays an AI screen transition operation unit 41e including characters or images indicating the selected plurality of setting display screens 41a. do. More specifically, as shown in FIG. 12, the AI screen transition operation unit 41e includes the names and icons of the plurality of selected setting display screens 41a.

The operator can display the setting display screen 41a on the display device 4 by selecting an icon indicating the desired setting display screen 41a. The control unit 3a gives a positive reward to the action of transitioning to the setting display screen 41a to reinforce the learning model 6.

FIG. 13 is a schematic view showing a third example of the AI screen transition operation unit 41e. Similar to the second example, the AI screen transition operation unit 41e according to the third example displays a reduced image of a plurality of setting display screens 41a as candidates for the setting display screen 41a to be displayed next, and the operator desires the operation unit 41e. It is configured so that the setting display screen 41a can be selected. Specifically, the control unit 3a selects a plurality of setting display screens 41a in descending order of action value, and displays an AI screen transition operation unit 41e including a reduced image of the selected plurality of setting display screens 41a.

The operator can display the setting display screen 41a on the display device 4 by selecting a reduced image showing the desired setting display screen 41a. Further, the operator can exclude the setting display screen 41a from the display candidates by swiping or flicking the unnecessary setting display screen 41a with respect to the reduced image in a predetermined direction, for example, in the horizontal direction.
The control unit 3a redisplays the AI screen transition operation unit 41e including the plurality of setting display screens 41a having high action value, except for the unnecessary setting display screen 41a. For example, in the example shown in FIG. 13, the "weighing" screen is removed as an unnecessary setting display screen 41a, and a new "production monitoring" screen is displayed.
The control unit 3a gives a negative reward to the action of transitioning to the excluded setting display screen 41a to train the learning model 6. Further, as in the second example, the learning model 6 is trained by giving a positive reward to the action of transitioning to the selected setting display screen 41a.

According to the computer program P, the machine learning machine 33, the information processing method, and the injection molding machine 100 configured as described above, the operator is strengthened and learned by learning the transition history of the setting display screen 41a of the molding machine by the operation of the operator. It is possible to determine the setting display screen 41a to be displayed next in the order of convenience.

Although reinforcement learning has been described in the first embodiment, the learning model 6 may be learned by supervised learning. For example, training data to which teacher data indicating a transition method of the setting display screen 41a based on the operation of an expert is added to the state information is prepared, and the learning model 6 is trained based on the training data. In this case, when state information is input to the learning model 6, the learning model 6 outputs information indicating the probability of transition to the plurality of setting display screens 41a. The control unit 3a transitions the setting display screen 41a based on the information output from the learning model 6. Further, RNN (Recurrent Neural Network) may be used as a model for supervised learning.

Further, in the first embodiment, although the reinforcement learning has been described, the learning model 6 may be learned by unsupervised learning.
Furthermore, although model-free reinforcement learning has been described, screen transitions may be learned by model-based reinforcement learning.

Furthermore, although the example of learning and displaying the setting display screen 41a to be displayed next by machine learning has been described in the embodiment, the setting display screen 41a to be displayed may be determined based on the rule.
For example, the control unit 3a may collect, for example, a plurality of screen transition histories, create a frequency distribution of the screen transition history, and use the image transition history having the highest frequency to transition and display the setting display screen 41a. ..

Furthermore, the learning model 6 may be configured to learn the setting value to be input to the setting display screen 41a. In this case, the combination of the transition setting display screen 41a and the set value may be set as an action, and the action value of the action may be output when the state quantity is input to the learning model 6.

Furthermore, in the case where the molding conditions and the like are input in a predetermined order, whether or not the set value should be input is stored for each setting display screen 41a, and the screen transition operation is performed without inputting. If so, the injection molding machine 100 may be configured to output a warning.

Furthermore, the injection molding machine 100 may be configured to store the normal range of the set value for each setting item and output a warning when the set value outside the normal range is set.

Furthermore, in the first embodiment, an example of reinforcement learning of the screen transition of the setting display screen 41a in a specific work such as molding condition setting has been described, but setup work, molding condition setting work, tabulation work, and temperature confirmation work have been described. , Injection side operation confirmation work, mold clamping side operation confirmation work, etc., may be configured to reinforce and learn screen transitions for each type of work. In this case, it is preferable to store the learning model 6 which is different for each type of work in the auxiliary storage unit 3d. Of course, one learning model 6 may be used by including a variable indicating the type of work in the state quantity. Further, it may be a reinforcement learning of screen transitions that span two or more tasks.
The control unit 3a displays the AI screen transition operation unit 41e including icons indicating a plurality of types of work. When one icon is operated by the operator, the control unit 3a selects the learning model 6 corresponding to the selected icon, and uses the selected learning model 66a to display the setting display screen 41a to be displayed next. Decide and display. When the state quantity includes information indicating the type of work, the control unit 3a can display the setting display screen 41a in order according to the type of work by the same process as in the first embodiment.

The learning model 6 learned by one injection molding machine 100 can be diverted to another injection molding machine 100. For example, when a new injection molding machine 100 is shipped from an injection molding machine manufacturer, the control device 3 can be equipped with a learning model 6 that has already been trained by the computer program P and another injection molding machine 100. Further, the learning model 6 mounted on the new injection molding machine 100 can be further evolved by operating the screen of the operator.

(Embodiment 2)
FIG. 14 is a schematic view showing a configuration example of an injection molding system. The injection molding system according to the second embodiment is different in that the information processing device 9 which is a server executes the transition process of the setting display screen 41a. Therefore, the above differences will be mainly described below. Since other configurations and actions and effects are the same as those in the first embodiment, the corresponding parts are designated by the same reference numerals and detailed description thereof will be omitted.

The injection molding system according to the second embodiment includes an information processing device 9 and one or more injection molding machines 100. The information processing device 9 and the injection molding machine 100 are communicated and connected via a network N such as a LAN (Local Area Network) or the Internet.

The information processing device 9 is a computer and includes a control unit 9a, a main storage unit 9b, a communication unit 9c, and an auxiliary storage unit 9d. The communication unit 9c is a communication circuit for transmitting and receiving data to and from the injection molding machine 100 via the network N. The hardware configurations of the control unit 9a, the main storage unit 9b, and the auxiliary storage unit 9d are the same as those of the injection molding machine 100 described in the first embodiment. The computer program P, the learning model 6, the standard hierarchical structure 7, the operator DB 8, and the recording medium 90 stored in the auxiliary storage unit 9d are also the same as the various programs and models of the first embodiment.
The information processing device 9 may be a multi-computer composed of a plurality of computers, or may be a virtual machine virtually constructed by software. The information processing device 9 may be a local server installed in the same facility or factory as the injection molding machine 100, or may be a cloud server communicatively connected to the injection molding machine 100 via the Internet or the like.

The information processing device 9 configured in this way acquires a state quantity from the injection molding machine 100 via the network N, and based on the acquired state quantity, displays one or a plurality of setting display screens 41a to be displayed next. Information indicating the determined and determined one or more setting display screens 41a is transmitted to the image apparatus. The injection molding machine 100 acquires the information transmitted from the information processing device 9, displays the AI screen transition operation unit 41e in the same manner as in FIG. 1, and displays the setting display screen 41a according to the operator's operation.
Further, the information processing apparatus 9 collects a state quantity from one or a plurality of injection molding machines 100, and learns a learning model 6 based on the collected state quantity.

The information processing device 9 may be configured to deliver information defining the learned learning model 6 to the injection molding machine 100. The injection molding machine 100 may store the learning model 6 based on the distributed information, and use the learning model 6 to execute the transition process of the setting display screen 41a.

Similarly to the first embodiment, the information processing apparatus 9, the computer program P, and the information processing method according to the second embodiment also learn the transition history of the setting display screen 41a of the molding machine operated by the operator, so that the order is convenient for the operator. The setting display screen 41a to be displayed next can be determined by.

(Appendix 1)
A computer program that causes a computer to generate a neural network for selecting a setting display screen to be displayed next from a plurality of setting display screens for operating a molding machine.
Acquire at least one of the operator's identifier, the operator's attribute, the molded product information indicating the molded product, the mold information indicating the mold, or the molding machine state information indicating the state of the molding machine.
Acquire the transition history of the setting display screen by the operation of the operator,
Calculate the reward based on the acquired transition history,
At least one of the acquired operator identifier, operator attributes, molded product information indicating a molded product, mold information indicating a mold, or molding machine state information indicating the state of a molding machine, and a setting display screen operated by the operator. When a state quantity including a transition history is input to the neural network, the neural network is output so that the difference between the value of the setting display screen to be displayed next and the value based on the reward is small. A computer program for causing the computer to execute a process of strengthening learning.

100 Injection molding machine 1 Injection device 2 Mold clamping device 3 Control device 3a Control unit 3b Main storage unit 3c Input / output I / F
3d Auxiliary storage 4 Display device 5 Operator information reader 6 Learning model 7 Standard hierarchical structure 8 Operator DB
31 Input / output unit 32 Molding machine control processing unit 33 Machine learning machine 33a Observation unit 33b Setting display screen selection unit 33c Reward calculation unit 33d Update processing unit 34 Image display processing unit 41 Touch panel 41a Setting display screen 41b Menu screen 41c Tab 41d Hierarchical screen Operation unit for transition 41e AI Operation unit for screen transition 42 Hardware key 61 Input layer 62 Intermediate layer 63 Output layer 100 Injection molding machine

Claims (9)

A computer program for causing a computer to display a setting display screen selected from a plurality of setting display screens for operating a molding machine on a display device.
A computer program for causing the computer to execute a process of determining a setting display screen to be displayed next based on a transition history of the setting display screen operated by an operator. At least one of the operator's identifier, the operator's attributes, the molded product information indicating the molded product, the mold information indicating the mold, or the molding machine state information indicating the state of the molding machine, and the transition history of the setting display screen by the operator's operation. The computer program according to claim 1, wherein the computer executes a process of learning the setting display screen to be displayed next based on the above. In the transition history of the setting display screen, let the computer execute the process of learning that the setting display screen whose display time is less than the predetermined time or the setting display screen that is displayed but the setting is not input is considered to be a setting screen with low need to be displayed. The computer program according to claim 2. The computer program according to claim 2 or 3, for causing a computer to execute a process of learning a predetermined setting display screen to be displayed on the display device when the molding machine is in a predetermined state. At least one of the operator's identifier, the operator's attributes, the molded product information indicating the molded product, the mold information indicating the mold, or the molding machine state information indicating the state of the molding machine, and the transition history of the setting display screen by the operator's operation. When a state quantity including and is input, any one of claims 2 to 4 for causing a computer to perform a process of machine learning a neural network so as to output the value of the setting display screen to be displayed next. The computer program described in the section. Based on the hierarchical structure for switching and displaying multiple setting display screens and the operation of the operator, the first method of determining the setting display screen to be displayed next, and the transition history of the setting display screen by the operation of the operator. , Accepts the selection with the second method to determine the setting display screen to be displayed next,
When the first method is accepted, the setting display screen to be displayed next is determined based on the hierarchical structure and the operation of the operator.
When the second method is accepted, any one of claims 1 to 4 for causing the computer to execute a process of determining the setting display screen to be displayed next based on the transition history of the setting display screen operated by the operator. The computer program according to item 1. An acquisition unit that acquires the transition history of the setting display screen operated by the operator to operate the molding machine,
A machine learning machine equipped with a learning unit that learns the setting display screen to be displayed next based on the acquired transition history of a plurality of setting display screens. It is an information processing method that displays a setting display screen selected from a plurality of setting display screens for operating a molding machine on a display device.
The setting display screen to be displayed next is determined based on the transition history of the setting display screen operated by the operator.
An information processing method for displaying the determined setting display screen on the display device. A molding machine that displays a setting display screen selected from a plurality of setting display screens on a display device.
An acquisition unit that acquires the transition history of the setting display screen by the operator's operation to operate the molding machine,
A molding machine equipped with a learning unit that learns the setting display screen to be displayed next based on the acquired transition history of a plurality of setting display screens.

Images