WO2021106225A1 - Abnormality monitoring device, abnormality monitoring method, and abnormality monitoring program - Google Patents

Abnormality monitoring device, abnormality monitoring method, and abnormality monitoring program Download PDF

Info

Publication number
WO2021106225A1
WO2021106225A1 PCT/JP2019/046921 JP2019046921W WO2021106225A1 WO 2021106225 A1 WO2021106225 A1 WO 2021106225A1 JP 2019046921 W JP2019046921 W JP 2019046921W WO 2021106225 A1 WO2021106225 A1 WO 2021106225A1
Authority
WO
WIPO (PCT)
Prior art keywords
abnormality monitoring
mold
captured image
divided portion
learning
Prior art date
Application number
PCT/JP2019/046921
Other languages
French (fr)
Japanese (ja)
Inventor
藤井　浩
泰彦 原
慎一 槇原
Original Assignee
菱洋エレクトロ株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 菱洋エレクトロ株式会社 filed Critical 菱洋エレクトロ株式会社
Priority to PCT/JP2019/046921 priority Critical patent/WO2021106225A1/en
Priority to JP2020515276A priority patent/JP6722836B1/en
Publication of WO2021106225A1 publication Critical patent/WO2021106225A1/en

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B29—WORKING OF PLASTICS; WORKING OF SUBSTANCES IN A PLASTIC STATE IN GENERAL
    • B29C—SHAPING OR JOINING OF PLASTICS; SHAPING OF MATERIAL IN A PLASTIC STATE, NOT OTHERWISE PROVIDED FOR; AFTER-TREATMENT OF THE SHAPED PRODUCTS, e.g. REPAIRING
    • B29C45/00—Injection moulding, i.e. forcing the required volume of moulding material through a nozzle into a closed mould; Apparatus therefor
    • B29C45/17—Component parts, details or accessories; Auxiliary operations
    • B29C45/76—Measuring, controlling or regulating
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B29—WORKING OF PLASTICS; WORKING OF SUBSTANCES IN A PLASTIC STATE IN GENERAL
    • B29C—SHAPING OR JOINING OF PLASTICS; SHAPING OF MATERIAL IN A PLASTIC STATE, NOT OTHERWISE PROVIDED FOR; AFTER-TREATMENT OF THE SHAPED PRODUCTS, e.g. REPAIRING
    • B29C45/00—Injection moulding, i.e. forcing the required volume of moulding material through a nozzle into a closed mould; Apparatus therefor
    • B29C45/17—Component parts, details or accessories; Auxiliary operations
    • B29C45/84—Safety devices

Definitions

  • the present invention relates to an abnormality monitoring device, an abnormality monitoring method, and an abnormality monitoring program.
  • Japanese Patent No. 5220948 stores a plurality of reference image data to be used as a reference for normal operation in a predetermined step of an injection molding cycle based on a video signal, and a video signal in a sequentially repeated injection molding cycle.
  • the detected image data is acquired based on the above, the detected image data is compared with the reference image data of 1, and it is determined that the detected image data is dissimilar to the reference image data, the detected image data is stored.
  • an injection molding machine monitoring device that monitors an abnormal operation of the injection molding machine main body by comparing with the reference image data that has not yet been compared with the detected reference image data among the plurality of reference image data.
  • the present invention provides an abnormality monitoring device, an abnormality monitoring method, and an abnormality monitoring method capable of monitoring the abnormal state of the monitored target without causing complicated initial setting work by the user even when the monitored target is new.
  • the purpose is to provide an anomaly monitoring program.
  • the abnormality monitoring device of one aspect of the present invention includes a learned model generated by machine learning a learning image representing a divided portion of a mold, and an captured image acquired by imaging the divided portion with a camera. Based on the above, it has a control unit that determines whether or not the divided portion is in an abnormal state, and when it is determined that the divided portion is in an abnormal state, outputs a stop instruction for stopping the mold clamping. ..
  • a learned model generated by machine learning a learning image representing a divided portion of a mold is prepared, and the divided portion is imaged by a camera to obtain an captured image.
  • the divided unit is in an abnormal state based on the captured image and the learned model.
  • a stop instruction for stopping the mold clamping is output.
  • the abnormality monitoring program of one aspect of the present invention uses a machine learning image showing the divided portion of the mold in response to input of control information indicating the end of the extraction process of the molded product formed by mold clamping.
  • a program for determining whether or not the divided portion is in an abnormal state based on the trained model generated by learning and the captured image acquired by imaging the divided portion with a camera, and the divided portion. It consists of a program that outputs a stop instruction for stopping the mold clamping when it is determined that the unit is in an abnormal state.
  • an abnormality monitoring device capable of monitoring an abnormal state of a monitored object without causing complicated initial setting work by the user even when the monitored object is new.
  • Anomaly monitoring programs can be provided.
  • FIG. 1 is a block diagram showing an example of the configuration of the injection molding machine 1.
  • the injection molding machine 1 has a molding machine main body 2 and an abnormality monitoring device 3.
  • the molding machine main body 2 and the abnormality monitoring device 3 are connected to each other by a signal line.
  • the molding machine main body 2 has an injection mechanism 10, a mold clamping mechanism 20, an instruction input unit 30, a control unit 40, and a storage unit 50.
  • the injection mechanism 10 injects the molten resin material into the mold clamping mechanism 20.
  • the injection mechanism 10 includes a hopper 11, an injection cylinder 12, and an injection drive unit 13.
  • the hopper 11 is connected to one end of the injection cylinder 12.
  • the hopper 11 can charge the resin material, accommodates the charged resin material, and supplies the resin material to the injection cylinder 12.
  • the injection cylinder 12 has a nozzle 12a provided at the other end opposite to one, a heater provided at the outer peripheral portion, and a screw rotatably provided around the axis at the inner peripheral portion.
  • the injection drive unit 13 has a drive device such as a motor, for example. Under the control of the control unit 40, the injection drive unit 13 rotates the screw to convey the resin material to the other, heats the heater to melt the resin material, and ejects the melted resin material from the nozzle 12a.
  • a drive device such as a motor
  • the mold clamping mechanism 20 molds the resin material injected from the injection mechanism 10 by the mold clamping.
  • the mold clamping mechanism 20 includes a fixed plate 21, a movable plate 22, a fixed side mold 23, a movable side mold 24, a spool 25, a mold clamping drive unit 26, an ejector pin 27, and a take-out drive unit 28.
  • the fixing plate 21 is provided on the other side of the injection cylinder 12.
  • a nozzle connecting portion 21a for connecting the nozzle 12a is provided on one side of the fixing plate 21.
  • the movable board 22 is provided on the other side of the fixed board 21.
  • the movable platen 22 is guided by a tie bar 22a provided between the movable platen 22 and the fixed platen 21, and can move to one or the other.
  • the fixed side mold 23 and the movable side mold 24 are provided between the fixed plate 21 and the movable plate 22.
  • the fixed-side mold 23 has a fixed-side mounting plate 23a for attaching to the fixed plate 21 on one side, and is attached to a portion on the other side of the fixed plate 21. Further, the fixed-side mold 23 has a guide bush 23b, a fixed-side split portion 23c, and a cavity 23d on the other side.
  • the movable side mold 24 is arranged so as to face the fixed side mold 23.
  • the movable side mold 24 has a movable side mounting plate 24a for attaching to the movable board 22 on the other side, and is attached to a portion on one side of the movable board 22.
  • a resin material is poured on one side between the guide pin 24b guided by the guide bush 23b, the movable side dividing portion 24c that presses against the fixed side dividing portion 23c, and the cavity 23d. It has a core 24d that forms a space for it.
  • the spool 25 communicates the nozzle connecting portion 21a with the cavity 23d so that the resin material can flow.
  • the mold clamping drive unit 26 has, for example, a drive device such as a flood control, and opens and closes the movable side mold 24 with respect to the fixed side mold 23. More specifically, the mold clamping drive unit 26 moves the movable platen 22 to one side under the control of the control unit 40, presses the movable side mold 24 against the fixed side mold 23, and molds. After mold clamping, the mold clamping drive unit 26 moves the movable platen 22 to the other side and pulls the movable side mold 24 away from the fixed side mold 23.
  • a drive device such as a flood control
  • the ejector pin 27 is provided so as to penetrate the movable platen 22, the movable mold 24, and the inner peripheral portion of the core 24d.
  • the take-out drive unit 28 has, for example, a drive device such as a motor.
  • the take-out drive unit 28 projects the ejector pin 27 from the core 24d so that the molded product Mo can be taken out under the control of the control unit 40. It is pulled back and buried in the core 24d.
  • the instruction input unit 30 has various buttons, display panels, and other operating tools, and can input various instructions to the molding machine main body 2.
  • the instruction input unit 30 outputs various input instructions to the control unit 40.
  • the control unit 40 has a processor such as a CPU that executes various processes.
  • the function of the control unit 40 is realized by the processor executing various programs stored in the storage unit 50.
  • the control unit 40 controls various operations of the molding machine main body 2. More specifically, the control unit 40 performs an injection process for driving the injection mechanism 10 based on an instruction input from the instruction input unit 30, performs a mold clamping process for driving the mold clamping mechanism 20, and drives the product to be taken out. The take-out process for driving the unit 28 is performed. When the take-out process is completed, the control unit 40 outputs the control information Ci indicating the end of the take-out process. When the stop instruction St is input, the control unit 40 performs a stop process of stopping the drive of the mold clamping mechanism 20 until the drive restart instruction is input.
  • the storage unit 50 has storage elements such as ROM, RAM, HDD, and SSD, and stores various information and programs for controlling various operations of the molding machine main body 2.
  • the abnormality monitoring device 3 monitors whether or not the molding machine main body 2 is in an abnormal state.
  • the abnormality monitoring device 3 has a camera 60 and a monitoring device main body 70.
  • the camera 60 images the movable side split portion 24c and outputs the captured image Im to the monitoring device main body 70.
  • the camera 60 is held by the arm 61.
  • FIG. 2 is an explanatory diagram for explaining an example of the mounting state of the camera 60.
  • the arm 61 has a mounting tool 62 having a camera 60 attached to one end and a magnet at the other end opposite to one end.
  • the arm 61 is attached to the fixing plate 21 or the like by the attractive force of the magnet of the attachment 62, and holds the camera 60.
  • the monitoring device main body 70 has a control unit 71 and a storage unit 72.
  • the control unit 71 has a processor such as a CPU and a GPU that executes various processes.
  • the function of the control unit 71 is realized by the processor reading and executing various information stored in the storage unit 72.
  • the control unit 71 controls various operations of the abnormality monitoring device 3, and also executes a program of the abnormality determination unit Pj in response to the input of the control information Ci.
  • the storage unit 72 has storage elements such as ROM, RAM, HDD, and SSD, and stores various programs and various data, as well as the program of the abnormality determination unit Pj and the trained model Md.
  • Abnormality determination unit Pj performs abnormality determination processing.
  • the abnormality determination process determines whether or not the mold clamping mechanism 20 is in an abnormal state based on the learned model Md and the captured image Im, and when it is determined that the mold clamping mechanism 20 is in an abnormal state, the mold clamping mechanism 20 is in an abnormal state.
  • a stop instruction St for stopping 20 is output.
  • the abnormality determination unit Pj calculates an evaluation value by inputting the captured image Im into the trained model Md, and when the evaluation value is within a predetermined range, the mold clamping mechanism 20 is in an abnormal state. Is determined.
  • FIG. 3 is an explanatory diagram for explaining the movable side divided portion 24c and the molded product Mo.
  • FIG. 4 is an explanatory diagram for explaining the extraction process.
  • FIG. 5 is an explanatory diagram for explaining an example of an abnormal state of the mold clamping mechanism 20.
  • the trained model Md machine-learns teacher data based on techniques such as deep learning and neural networks so that the calculation result of the evaluation value falls within a predetermined range when the mold clamping mechanism 20 is in an abnormal state. Is generated by.
  • the learning image showing the movable side divided portion 24c having no molded product Mo has a correlation with the normal state of the mold clamping mechanism 20 after the extraction process.
  • a learning image showing the movable side divided portion 24c having no molded product Mo is used as an indicator of the normal state of the mold clamping mechanism 20.
  • the molded product Mo is left on the movable side split portion 24c due to catching or the like.
  • the learning image of the movable side split portion 24c having the molded product Mo has a correlation with the abnormal state of the mold clamping mechanism 20 after the extraction process.
  • a learning image showing the movable side split portion 24c having the molded product Mo is used as an indicator of the abnormal state of the mold clamping mechanism 20.
  • the learning image is created based on the captured image Im obtained by performing an injection molding process at the development stage and capturing the movable side split portion 24c with the camera 60. More specifically, the captured image Im representing the movable side divided portion 24c having the molded product Mo is acquired by imaging the movable side divided portion 24c after molding and before taking out the molded product Mo. The captured image Im showing the movable side divided portion 24c having no molded product Mo is acquired by imaging the movable side divided portion 24c after taking out the molded product Mo and before the next mold clamping.
  • the captured image Im is created by imaging the movable side divided portion 24c from various angles such as upward, lateral, and downward. Further, the captured image Im may be inflated by horizontal inversion, vertical inversion, rotation, enlargement, reduction, brightness change, or the like.
  • a part of the captured image Im is used for evaluation of the trained model Md.
  • the abnormality monitoring device 3 monitors whether or not the mold clamping mechanism 20 of the molding machine main body 2 to be monitored is in an abnormal state.
  • the monitoring target is a mold provided in the mold clamping mechanism 20 and a split portion of the mold. More specifically, the monitoring targets are the movable side mold 24 and the movable side dividing portion 24c.
  • the control unit 71 is based on the learned model Md generated by machine learning the learning image representing the divided portion of the mold and the captured image Im acquired by imaging the divided portion with the camera 60. , It is determined whether or not the divided portion is in an abnormal state, and when it is determined that the divided portion is in an abnormal state, a stop instruction St for stopping the mold clamping is output.
  • the control unit 71 calculates the evaluation value by inputting the captured image Im into the trained model Md, and determines that the division unit is in an abnormal state when the evaluation value is within a predetermined range.
  • the trained model Md is obtained by machine learning a learning image showing a divided portion having no molded product Mo showing a normal state and a learning image showing a divided portion having a molded product Mo showing an abnormal state. Will be generated.
  • a learned model Md generated by machine learning a learning image showing a divided portion of a mold is prepared, and the divided portion is imaged by a camera 60 to acquire an captured image Im.
  • the divided unit is in an abnormal state based on the captured image Im and the learned model Md.
  • a stop instruction St for stopping the mold clamping is output.
  • the abnormality monitoring program machine-learns a learning image showing the divided portion of the mold in response to the input of the control information Ci indicating the end of the extraction process of the molded product Mo formed by the mold clamping.
  • a program for determining whether or not the divided portion is in an abnormal state and a program for determining whether the divided portion is abnormal It consists of a program that outputs a stop instruction St for stopping the mold clamping when it is determined that the state is in the state.
  • FIG. 6 is a flowchart showing an example of the flow of the injection molding process.
  • the control unit 40 outputs an instruction for closing the movable mold 24.
  • the mold clamping drive unit 26 moves the movable mold 24 to one side in response to an instruction, presses the movable mold 24 against the fixed mold 23, and clamps the mold.
  • the control unit 40 outputs an instruction for injecting the resin material.
  • the injection drive unit 13 injects the resin material from the nozzle 12a into the mold clamping mechanism 20 in response to an instruction.
  • the resin material flows through the spool 25, flows into the space formed between the cavity 23d and the core 24d, and is molded to form a molded product Mo.
  • the control unit 40 Open the movable mold 24 (S3).
  • the control unit 40 outputs an instruction for opening the movable mold 24.
  • the mold clamping drive unit 26 moves the movable side mold 24 to the other side and pulls the movable side mold 24 away from the fixed side mold 23 in response to an instruction.
  • the molded product Mo appears on the movable side divided portion 24c.
  • the control unit 40 takes out the molded product Mo (S4).
  • the control unit 40 outputs an instruction for taking out the molded product Mo.
  • the take-out drive unit 28 projects the ejector pin 27 from the core 24d in response to an instruction, and then pulls back the ejector pin 27 and buries it in the core 24d.
  • the control unit 40 outputs the control information Ci indicating the end of the retrieval process.
  • the control unit 71 inputs the captured image Im input from the camera 60 into the trained model Md according to the control information Ci, and calculates an evaluation value.
  • the control unit 71 determines that the evaluation value is within the predetermined range, the process proceeds to S6 (S6: YES).
  • the evaluation value is not within the predetermined range, the process returns to S1 (S6: NO).
  • the control unit 71 outputs a stop instruction St for stopping the mold clamping mechanism 20.
  • the control unit 40 stops the mold clamping mechanism 20 in response to the stop instruction St.
  • the treatments S1 to S8 are injection molding treatments.
  • the processing of S1 and S3 is a mold clamping process.
  • the process of S2 is an injection process.
  • the process of S4 is a take-out process.
  • the processing of S5 and S6 is an abnormality determination processing.
  • the processing of S7 and S8 is a stop processing.
  • the abnormality monitoring device 3 is complicated by the user even when the movable side mold 24 to be monitored is new, such as when the molding machine main body 2 is newly installed or when the movable side mold 24 is replaced.
  • the evaluation value is calculated by inputting the captured image Im into the trained model Md without any initial setting work, and it is determined whether or not the mold clamping mechanism 20 is in an abnormal state according to the evaluation value. Can be done.
  • the abnormality monitoring device 3 can monitor the abnormal state of the monitoring target without causing complicated initial setting work by the user even when the monitoring target is new.
  • the trained model Md is not additionally trained, but the trained model Md may be configured to be additionally trained.
  • FIG. 7 is a flowchart showing an example of the flow of the additional learning process.
  • the additional learning process starts according to a predetermined schedule.
  • the captured image Im is stored in the storage unit 72 (S11).
  • the control unit 71 stores the imaged image Im representing each of the movable side divided portion 24c having the molded product Mo and the movable side divided portion 24c having no molded product Mo in the storage unit 72.
  • the captured image Im showing the movable side divided portion 24c having the molded product Mo is acquired by imaging the movable side divided portion 24c after molding and before taking out the molded product Mo.
  • the captured image Im showing the movable side divided portion 24c having no molded product Mo is acquired by imaging the movable side divided portion 24c after taking out the molded product Mo and before the next mold clamping.
  • the control unit 71 determines whether or not to start machine learning according to the schedule read from the storage unit 72. Machine learning is preferably performed at a time when the drive of the injection molding machine 1 is stopped, such as at night. When starting machine learning, the process proceeds to S13 (S12: YES). When machine learning is not started, the process returns to S11 (S12: NO).
  • the captured image Im is read from the storage unit 72 (S13).
  • the control unit 71 performs machine learning based on the read captured image Im, and updates the trained model Md.
  • the captured image Im of the movable side split portion 24c having no molded product Mo is used as an indicator of the normal state of the mold clamping mechanism 20.
  • the captured image Im of the movable side divided portion 24c having the molded product Mo is used as an indicator of the abnormal state of the mold clamping mechanism 20.
  • the storage unit 72 stores the captured image Im
  • the control unit 71 updates the learned model Md by machine learning the captured image Im.
  • the abnormality monitoring device 3 can perform additional learning processing of the trained model Md based on the captured image Im captured by the camera 60, and improve the accuracy of determining the abnormal state of the mold clamping mechanism 20. Can be done.
  • the abnormality monitoring device 3 is not connected to the server 4, but the abnormality monitoring device 3 may be configured to be connected to the server 4 (two-dot chain line in FIG. 1).
  • the control unit 71 is connected to the server 4 by wired communication or wireless communication via a network such as the Internet or LAN.
  • the control unit 71 can output the captured image Im to the server 4.
  • the server 4 performs additional learning processing based on the captured image Im input from the monitoring device main body 70, and outputs the result of the additional learning processing to the monitoring device main body 70.
  • control unit 71 updates the trained model Md based on the result of the additional learning process input from the server 4.
  • the server 4 can receive the input of the captured image Im from the plurality of abnormality monitoring devices 3 and perform the additional learning process, and can perform the additional learning process based on more captured images Im. is there.
  • the server 4 outputs a keep-alive signal to the control unit 71 at a predetermined timing.
  • the control unit 71 receives the keep-alive signal
  • the control unit 71 outputs a response signal to the server 4.
  • the server 4 warns the user that the abnormality monitoring device 3 is in the stopped state by a notification means such as an e-mail.
  • the server 4 outputs a signal to the control unit 71 at a predetermined timing, and outputs a warning when the response signal cannot be received from the control unit 71 within the predetermined time.
  • the abnormality monitoring device 3 can more reliably determine whether or not the mold clamping mechanism 20 is in an abnormal state.
  • the learning image showing the movable mold 24 having the molded product Mo and the learning image showing the movable mold 24 having the molded product Mo are machine-learned. Generated by, but not limited to.
  • the movable side split portion 24c made of metal appears, while in the abnormal state, the molded product Mo made of resin appears on the movable side split portion 24c.
  • the learning image of the movable side split portion 24c showing the metal has a correlation with the normal state of the mold clamping mechanism 20. Further, the learning image showing both the metal and the resin or the resin has a correlation with the abnormal state of the mold clamping mechanism 20.
  • the trained model Md is a machine learning image showing a metal as showing the normal state of the mold clamping mechanism 20 and a learning image showing both metal and resin or resin as showing an abnormal state. It is generated by learning.
  • the abnormality monitoring device 3 can determine that the mold clamping mechanism 20 is in an abnormal state when the resin appears in the captured image Im indicating after the extraction process.
  • the trained model Md has a first trained model, a second trained model, and a third trained model.
  • the first trained model is generated based on a learning image created by changing the type of lighting that illuminates the movable side dividing portion 24c and the type of the camera 60 and taking an image.
  • the second learned model is generated based on the learning image created by changing the type of the movable side dividing portion 24c and taking an image.
  • the third trained model is generated based on the learning image created by changing the color of the molded product Mo and taking an image.
  • the control unit 71 has a first evaluation value calculated based on the first trained model and the captured image Im, a second evaluation value calculated based on the second trained model and the captured image Im, and a third learning. It is determined whether or not the divided portion of the mold clamping mechanism 20 is in an abnormal state according to the completed model and the third evaluation value calculated based on the captured image Im.
  • control unit 71 weights each of the first evaluation value, the second evaluation value, and the third evaluation value by a predetermined weighting coefficient, and totals the weighting results to calculate the evaluation value. , When the evaluation value is within the predetermined range, it is determined that the state is abnormal.
  • the abnormality monitoring device 3 can determine whether or not the mold clamping mechanism 20 of more types is in an abnormal state with higher accuracy.
  • the camera 60 monitors the movable side split portion 24c, but the present invention is not limited to this.
  • the molded product Mo appears in the fixed-side divided portion 23c instead of the movable-side divided portion 24c after molding and before taking out the molded product Mo, the fixed-side divided portion 23c may be monitored.
  • the abnormality monitoring device 3 captures the movable side split portion 24c with the camera 60, but when the molded product Mo is formed on the fixed side mold 23, the fixed side split portion 23c is captured by the camera. An image may be taken by 60 to determine whether or not the mold clamping mechanism 20 is in an abnormal state. Further, the abnormality monitoring device 3 may image both the movable side dividing portion 24c and the fixed side dividing portion 23c to determine whether or not the mold clamping mechanism 20 is in an abnormal state.
  • the camera 60 is located above the fixed plate 21 by the arm 61, but the position of the camera 60 is not limited to this.
  • the camera 60 may be located on the side or lower part of the fixing plate 21 by the arm 61. Further, the camera 60 may be located on the side of the movable mold 24 as long as the core 24d can be imaged. Further, the camera 60 may be attached to other than the fixed plate 21.
  • the storage unit 72 stores the abnormality determination unit Pj and the learned model Md, but these may be stored in the storage unit 50 or the server 4. In that case, the function of the control unit 71 may be realized by the control unit 40 or the server 4.
  • the determination of the start of the additional learning process (S12) was performed by reading the predetermined schedule from the storage unit 72, but the determination of the start of the additional learning process was performed by the instruction input unit 30. Alternatively, it may be performed by a user's instruction input via an instruction unit (not shown) connected to the control unit 71.
  • Each step of each procedure in the present embodiment may be executed at the same time by changing the execution order, or may be executed in a different order for each execution, as long as the property is not contrary to the property. Further, all or a part of each step of each procedure in the present embodiment may be realized by hardware.
  • the present invention is not limited to the above-described embodiment, and various modifications, modifications, and the like can be made without changing the gist of the present invention.

Abstract

An abnormality monitoring device 3 has a control unit 71 that, on the basis of a captured image Im that is obtained by imaging a division part of a mold by a camera 60 and a learned model Md that is generated through performing machine learning of learning images showing the division part, determines whether or not the division part is in an abnormal state, and that outputs a stop instruction St for stopping mold closing when determining that the division part is in an abnormal state.

Description

異常監視装置、異常監視方法、及び、異常監視プログラムAbnormality monitoring device, abnormality monitoring method, and abnormality monitoring program
　本発明は、異常監視装置、異常監視方法、及び、異常監視プログラムに関する。 The present invention relates to an abnormality monitoring device, an abnormality monitoring method, and an abnormality monitoring program.
　従来、監視対象の異常状態を監視する異常監視装置がある。例えば、特許第５２２０９４８号公報には、ビデオ信号を基に、射出成形サイクルの所定の工程における正常動作の基準とすべき複数の基準画像データを記憶し、順次繰り返される射出成形サイクルにおいて、ビデオ信号を基に検出画像データを取得し、検出画像データと１の基準画像データとを比較し、当該検出画像データが当該基準画像データと非類似であると判定したとき、当該検出画像データを、記憶された複数の基準画像データのうち未だ当該検出画像データとの比較を行っていない基準画像データと比較することにより射出成形機本体の異常動作を監視する、射出成形機監視装置が開示される。 Conventionally, there is an abnormality monitoring device that monitors the abnormal state of the monitoring target. For example, Japanese Patent No. 5220948 stores a plurality of reference image data to be used as a reference for normal operation in a predetermined step of an injection molding cycle based on a video signal, and a video signal in a sequentially repeated injection molding cycle. When the detected image data is acquired based on the above, the detected image data is compared with the reference image data of 1, and it is determined that the detected image data is dissimilar to the reference image data, the detected image data is stored. Disclosed is an injection molding machine monitoring device that monitors an abnormal operation of the injection molding machine main body by comparing with the reference image data that has not yet been compared with the detected reference image data among the plurality of reference image data.
特許第５２２０９４８号公報Japanese Patent No. 5220948
　しかし、従来の異常監視装置は、基準画像データを有しない新規の監視対象を監視するとき、正常状態の基準とすべき基準画像データを登録するためのユーザによる煩雑な初期設定作業が発生する。 However, in the conventional abnormality monitoring device, when monitoring a new monitoring target that does not have the reference image data, a complicated initial setting work by the user for registering the reference image data that should be the reference of the normal state occurs.
　そこで、本発明は、監視対象が新規であるときにおいても、ユーザによる煩雑な初期設定作業が発生することなく、監視対象の異常状態を監視することができる、異常監視装置、異常監視方法、及び、異常監視プログラムを提供することを目的とする。 Therefore, the present invention provides an abnormality monitoring device, an abnormality monitoring method, and an abnormality monitoring method capable of monitoring the abnormal state of the monitored target without causing complicated initial setting work by the user even when the monitored target is new. , The purpose is to provide an anomaly monitoring program.
　本発明の一態様の異常監視装置は、金型の分割部を表した学習用画像を機械学習することによって生成された学習済みモデルと、カメラによって前記分割部を撮像して取得された撮像画像とに基づいて、前記分割部が異常状態であるか否かを判定し、前記分割部が異常状態であると判定したとき、型締めを停止させるための停止指示を出力する、制御部を有する。 The abnormality monitoring device of one aspect of the present invention includes a learned model generated by machine learning a learning image representing a divided portion of a mold, and an captured image acquired by imaging the divided portion with a camera. Based on the above, it has a control unit that determines whether or not the divided portion is in an abnormal state, and when it is determined that the divided portion is in an abnormal state, outputs a stop instruction for stopping the mold clamping. ..
　本発明の一態様の異常監視方法は、金型の分割部を表した学習用画像を機械学習することによって生成された学習済みモデルを用意し、カメラによって前記分割部を撮像して撮像画像を取得し、制御部により、金型の型締めによって成形された成形物の取出処理の終了を示す制御情報の入力に応じ、前記撮像画像と前記学習済みモデルに基づいて、前記分割部が異常状態であるか否かを判定し、前記分割部が異常状態であると判定したとき、前記型締めを停止させるための停止指示を出力する。 In the abnormality monitoring method of one aspect of the present invention, a learned model generated by machine learning a learning image representing a divided portion of a mold is prepared, and the divided portion is imaged by a camera to obtain an captured image. In response to the input of control information that is acquired and indicates the end of the take-out process of the molded product formed by the mold clamping by the control unit, the divided unit is in an abnormal state based on the captured image and the learned model. When it is determined that the split portion is in an abnormal state, a stop instruction for stopping the mold clamping is output.
　本発明の一態様の異常監視プログラムは、金型の型締めによって成形された成形物の取出処理の終了を示す制御情報の入力に応じ、前記金型の分割部を表した学習用画像を機械学習することによって生成された学習済みモデルと、カメラによって前記分割部を撮像して取得された撮像画像とに基づいて、前記分割部が異常状態であるか否かを判定するプログラムと、前記分割部が異常状態であると判定したとき、前記型締めを停止させるための停止指示を出力するプログラムからなる。 The abnormality monitoring program of one aspect of the present invention uses a machine learning image showing the divided portion of the mold in response to input of control information indicating the end of the extraction process of the molded product formed by mold clamping. A program for determining whether or not the divided portion is in an abnormal state based on the trained model generated by learning and the captured image acquired by imaging the divided portion with a camera, and the divided portion. It consists of a program that outputs a stop instruction for stopping the mold clamping when it is determined that the unit is in an abnormal state.
　本発明によれば、監視対象が新規であるときにおいても、ユーザによる煩雑な初期設定作業が発生することなく、監視対象の異常状態を監視することができる、異常監視装置、異常監視方法、及び、異常監視プログラムを提供することができる。 According to the present invention, an abnormality monitoring device, an abnormality monitoring method, and an abnormality monitoring method capable of monitoring an abnormal state of a monitored object without causing complicated initial setting work by the user even when the monitored object is new. , Anomaly monitoring programs can be provided.
実施形態に係る、射出成形機の構成の一例を示すブロック図である。It is a block diagram which shows an example of the structure of the injection molding machine which concerns on embodiment. 実施形態に係る、射出成形機におけるカメラの取付状態の一例を説明するための説明図である。It is explanatory drawing for demonstrating an example of the mounting state of the camera in the injection molding machine which concerns on embodiment. 実施形態に係る、射出成形機における可動側分割部及び成形物を説明するための説明図である。It is explanatory drawing for demonstrating the movable side division part and the molded article in the injection molding machine which concerns on embodiment. 実施形態に係る、射出成形機における取出処理を説明するための説明図である。It is explanatory drawing for demonstrating the take-out process in an injection molding machine which concerns on embodiment. 実施形態に係る、射出成形機の型締機構の異常状態の一例を説明するための説明図である。It is explanatory drawing for demonstrating an example of the abnormal state of the mold clamping mechanism of the injection molding machine which concerns on embodiment. 実施形態に係る、射出成形機における射出成形処理の流れの一例を示すフローチャートである。It is a flowchart which shows an example of the flow of the injection molding process in the injection molding machine which concerns on embodiment. 実施形態に係る、射出成形機における追加学習処理の流れの一例を示すフローチャートである。It is a flowchart which shows an example of the flow of the additional learning process in an injection molding machine which concerns on embodiment.
　以下、図面を参照しながら、実施形態を説明する。 Hereinafter, embodiments will be described with reference to the drawings.
（構成）　
　図１は、射出成形機１の構成の一例を示すブロック図である。 (Constitution)
FIG. 1 is a block diagram showing an example of the configuration of the injection molding machine 1.
　図１に示すように、射出成形機１は、成形機本体２、及び、異常監視装置３を有する。成形機本体２と異常監視装置３は、信号線によって互いに接続される。 As shown in FIG. 1, the injection molding machine 1 has a molding machine main body 2 and an abnormality monitoring device 3. The molding machine main body 2 and the abnormality monitoring device 3 are connected to each other by a signal line.
　成形機本体２は、射出機構１０、型締機構２０、指示入力部３０、制御部４０、及び、記憶部５０を有する。 The molding machine main body 2 has an injection mechanism 10, a mold clamping mechanism 20, an instruction input unit 30, a control unit 40, and a storage unit 50.
　射出機構１０は、型締機構２０内に溶融した樹脂材料を射出する。射出機構１０は、ホッパー１１、射出筒１２、及び、射出駆動部１３を有する。 The injection mechanism 10 injects the molten resin material into the mold clamping mechanism 20. The injection mechanism 10 includes a hopper 11, an injection cylinder 12, and an injection drive unit 13.
　ホッパー１１は、射出筒１２の一方の端部と接続される。ホッパー１１は、樹脂材料を投入可能であり、投入された樹脂材料を収容し、射出筒１２に樹脂材料を供給する。 The hopper 11 is connected to one end of the injection cylinder 12. The hopper 11 can charge the resin material, accommodates the charged resin material, and supplies the resin material to the injection cylinder 12.
　射出筒１２は、一方とは逆の他方の端部に設けられたノズル１２ａと、外周部に設けられたヒータと、内周部に軸周りに回転可能に設けられたスクリューとを有する。 The injection cylinder 12 has a nozzle 12a provided at the other end opposite to one, a heater provided at the outer peripheral portion, and a screw rotatably provided around the axis at the inner peripheral portion.
　射出駆動部１３は、例えば、モータ等の駆動装置を有する。射出駆動部１３は、制御部４０の制御の下、スクリューを回転させて樹脂材料を他方へ搬送し、ヒータを加熱させて樹脂材料を溶融し、溶融した樹脂材料をノズル１２ａから射出させる。 The injection drive unit 13 has a drive device such as a motor, for example. Under the control of the control unit 40, the injection drive unit 13 rotates the screw to convey the resin material to the other, heats the heater to melt the resin material, and ejects the melted resin material from the nozzle 12a.
　型締機構２０は、型締めによって射出機構１０から射出された樹脂材料を成形する。型締機構２０は、固定盤２１、可動盤２２、固定側金型２３、可動側金型２４、スプール２５、型締駆動部２６、エジェクタピン２７、及び、取出駆動部２８を有する。 The mold clamping mechanism 20 molds the resin material injected from the injection mechanism 10 by the mold clamping. The mold clamping mechanism 20 includes a fixed plate 21, a movable plate 22, a fixed side mold 23, a movable side mold 24, a spool 25, a mold clamping drive unit 26, an ejector pin 27, and a take-out drive unit 28.
　固定盤２１は、射出筒１２よりも他方に設けられる。固定盤２１の一方側には、ノズル１２ａを接続するためのノズル接続部２１ａが設けられる。 The fixing plate 21 is provided on the other side of the injection cylinder 12. A nozzle connecting portion 21a for connecting the nozzle 12a is provided on one side of the fixing plate 21.
　可動盤２２は、固定盤２１よりも他方に設けられる。可動盤２２は、固定盤２１との間に設けられたタイバー２２ａにガイドされ、一方又は他方へ移動可能である。 The movable board 22 is provided on the other side of the fixed board 21. The movable platen 22 is guided by a tie bar 22a provided between the movable platen 22 and the fixed platen 21, and can move to one or the other.
　固定側金型２３及び可動側金型２４は、固定盤２１と可動盤２２の間に設けられる。 The fixed side mold 23 and the movable side mold 24 are provided between the fixed plate 21 and the movable plate 22.
　固定側金型２３は、一方側に固定盤２１に取り付けるための固定側取付板２３ａを有し、固定盤２１の他方側の部位に取り付けられる。また、固定側金型２３は、他方側に、ガイドブッシュ２３ｂ、固定側分割部２３ｃ、及び、キャビティ２３ｄを有する。 The fixed-side mold 23 has a fixed-side mounting plate 23a for attaching to the fixed plate 21 on one side, and is attached to a portion on the other side of the fixed plate 21. Further, the fixed-side mold 23 has a guide bush 23b, a fixed-side split portion 23c, and a cavity 23d on the other side.
　可動側金型２４は、固定側金型２３と対向するように配置される。可動側金型２４は、他方側に可動盤２２に取り付けるための可動側取付板２４ａを有し、可動盤２２の一方側の部位に取り付けられる。また、可動側金型２４は、一方側に、ガイドブッシュ２３ｂによってガイドされるガイドピン２４ｂと、固定側分割部２３ｃに押し当たる可動側分割部２４ｃと、キャビティ２３ｄとの間に樹脂材料を流し込むための空間を形成するコア２４ｄとを有する。 The movable side mold 24 is arranged so as to face the fixed side mold 23. The movable side mold 24 has a movable side mounting plate 24a for attaching to the movable board 22 on the other side, and is attached to a portion on one side of the movable board 22. Further, in the movable mold 24, a resin material is poured on one side between the guide pin 24b guided by the guide bush 23b, the movable side dividing portion 24c that presses against the fixed side dividing portion 23c, and the cavity 23d. It has a core 24d that forms a space for it.
　スプール２５は、樹脂材料を流すことができるように、ノズル接続部２１ａとキャビティ２３ｄを連通させる。 The spool 25 communicates the nozzle connecting portion 21a with the cavity 23d so that the resin material can flow.
　型締駆動部２６は、例えば、油圧等の駆動装置を有し、固定側金型２３に対して可動側金型２４を開閉させる。より具体的に、型締駆動部２６は、制御部４０の制御の下、可動盤２２を一方へ移動させ、可動側金型２４を固定側金型２３に押し当て、型締めする。型締め後、型締駆動部２６は、可動盤２２を他方へ移動させ、可動側金型２４を固定側金型２３から引き離す。 The mold clamping drive unit 26 has, for example, a drive device such as a flood control, and opens and closes the movable side mold 24 with respect to the fixed side mold 23. More specifically, the mold clamping drive unit 26 moves the movable platen 22 to one side under the control of the control unit 40, presses the movable side mold 24 against the fixed side mold 23, and molds. After mold clamping, the mold clamping drive unit 26 moves the movable platen 22 to the other side and pulls the movable side mold 24 away from the fixed side mold 23.
　エジェクタピン２７は、可動盤２２、可動側金型２４、及び、コア２４ｄの内周部に貫通するように設けられる。 The ejector pin 27 is provided so as to penetrate the movable platen 22, the movable mold 24, and the inner peripheral portion of the core 24d.
　取出駆動部２８は、例えば、モータ等の駆動装置を有する。取出駆動部２８は、制御部４０の制御の下、成形物Ｍｏを取り出すことができるように、エジェクタピン２７をコア２４ｄから突出させ、突出後、次回の型締処理に備えてエジェクタピン２７を引き戻してコア２４ｄに埋没させる。 The take-out drive unit 28 has, for example, a drive device such as a motor. The take-out drive unit 28 projects the ejector pin 27 from the core 24d so that the molded product Mo can be taken out under the control of the control unit 40. It is pulled back and buried in the core 24d.
　指示入力部３０は、各種ボタン、表示パネル等の操作具を有し、成形機本体２に対する各種指示の入力が可能である。指示入力部３０は、入力された各種指示を制御部４０に出力する。 The instruction input unit 30 has various buttons, display panels, and other operating tools, and can input various instructions to the molding machine main body 2. The instruction input unit 30 outputs various input instructions to the control unit 40.
　制御部４０は、各種処理を実行するＣＰＵ等のプロセッサを有する。制御部４０の機能は、プロセッサが記憶部５０に記憶された各種プログラムを実行することによって実現される。 The control unit 40 has a processor such as a CPU that executes various processes. The function of the control unit 40 is realized by the processor executing various programs stored in the storage unit 50.
　制御部４０は、成形機本体２の各種動作を制御する。より具体的には、制御部４０は、指示入力部３０から入力された指示に基づいて、射出機構１０を駆動する射出処理を行い、型締機構２０を駆動する型締処理を行い、取出駆動部２８を駆動する取出処理を行う。取出処理が終了すると、制御部４０は、取出処理の終了を示す制御情報Ｃｉを出力する。停止指示Ｓｔの入力があると、制御部４０は、駆動再開指示の入力があるまで型締機構２０の駆動を停止する停止処理を行う。 The control unit 40 controls various operations of the molding machine main body 2. More specifically, the control unit 40 performs an injection process for driving the injection mechanism 10 based on an instruction input from the instruction input unit 30, performs a mold clamping process for driving the mold clamping mechanism 20, and drives the product to be taken out. The take-out process for driving the unit 28 is performed. When the take-out process is completed, the control unit 40 outputs the control information Ci indicating the end of the take-out process. When the stop instruction St is input, the control unit 40 performs a stop process of stopping the drive of the mold clamping mechanism 20 until the drive restart instruction is input.
　記憶部５０は、ＲＯＭ、ＲＡＭ、ＨＤＤ、ＳＳＤ等の記憶素子を有し、成形機本体２の各種動作を制御するための各種情報及びプログラムが記憶される。 The storage unit 50 has storage elements such as ROM, RAM, HDD, and SSD, and stores various information and programs for controlling various operations of the molding machine main body 2.
　異常監視装置３は、成形機本体２が異常状態であるか否かを監視する。異常監視装置３は、カメラ６０及び監視装置本体７０を有する。 The abnormality monitoring device 3 monitors whether or not the molding machine main body 2 is in an abnormal state. The abnormality monitoring device 3 has a camera 60 and a monitoring device main body 70.
　カメラ６０は、可動側分割部２４ｃを撮像し、撮像画像Ｉｍを監視装置本体７０に出力する。カメラ６０は、アーム６１によって保持される。 The camera 60 images the movable side split portion 24c and outputs the captured image Im to the monitoring device main body 70. The camera 60 is held by the arm 61.
　図２は、カメラ６０の取付状態の一例を説明するための説明図である。 FIG. 2 is an explanatory diagram for explaining an example of the mounting state of the camera 60.
　図２に示すように、アーム６１は、一端部に、カメラ６０を取り付け、一端部とは逆の他端部に、磁石を有する取付け具６２を有する。アーム６１は、取付け具６２の磁石の吸引力によって固定盤２１等に取り付けられ、カメラ６０を保持する。 As shown in FIG. 2, the arm 61 has a mounting tool 62 having a camera 60 attached to one end and a magnet at the other end opposite to one end. The arm 61 is attached to the fixing plate 21 or the like by the attractive force of the magnet of the attachment 62, and holds the camera 60.
　図１に戻り、監視装置本体７０は、制御部７１及び記憶部７２を有する。 Returning to FIG. 1, the monitoring device main body 70 has a control unit 71 and a storage unit 72.
　制御部７１は、各種処理を実行するＣＰＵ及びＧＰＵ等のプロセッサを有する。制御部７１の機能は、プロセッサが、記憶部７２に記憶された各種情報を読み込み、実行することによって実現される。制御部７１は、異常監視装置３の各種動作を制御する他、制御情報Ｃｉの入力に応じて異常判定部Ｐｊのプログラムを実行する。 The control unit 71 has a processor such as a CPU and a GPU that executes various processes. The function of the control unit 71 is realized by the processor reading and executing various information stored in the storage unit 72. The control unit 71 controls various operations of the abnormality monitoring device 3, and also executes a program of the abnormality determination unit Pj in response to the input of the control information Ci.
　記憶部７２は、ＲＯＭ、ＲＡＭ、ＨＤＤ、ＳＳＤ等の記憶素子を有し、各種プログラム及び各種データの他、異常判定部Ｐｊのプログラム、及び、学習済みモデルＭｄも記憶される。 The storage unit 72 has storage elements such as ROM, RAM, HDD, and SSD, and stores various programs and various data, as well as the program of the abnormality determination unit Pj and the trained model Md.
　異常判定部Ｐｊは、異常判定処理を行う。異常判定処理は、学習済みモデルＭｄと撮像画像Ｉｍに基づいて、型締機構２０が異常状態であるか否かを判定し、型締機構２０が異常状態であると判定したとき、型締機構２０を停止させるための停止指示Ｓｔを出力する。より具体的には、異常判定部Ｐｊは、撮像画像Ｉｍを学習済みモデルＭｄに入力することによって評価値を算出し、評価値が所定範囲内にあるとき、型締機構２０が異常状態であると判定する。 Abnormality determination unit Pj performs abnormality determination processing. The abnormality determination process determines whether or not the mold clamping mechanism 20 is in an abnormal state based on the learned model Md and the captured image Im, and when it is determined that the mold clamping mechanism 20 is in an abnormal state, the mold clamping mechanism 20 is in an abnormal state. A stop instruction St for stopping 20 is output. More specifically, the abnormality determination unit Pj calculates an evaluation value by inputting the captured image Im into the trained model Md, and when the evaluation value is within a predetermined range, the mold clamping mechanism 20 is in an abnormal state. Is determined.
（学習済みモデルＭｄ）　
　続いて、学習済みモデルＭｄについて説明をする。 (Trained model Md)
Next, the trained model Md will be described.
　図３は、可動側分割部２４ｃ及び成形物Ｍｏを説明するための説明図である。図４は、取出処理を説明するための説明図である。図５は、型締機構２０の異常状態の一例を説明するための説明図である。 FIG. 3 is an explanatory diagram for explaining the movable side divided portion 24c and the molded product Mo. FIG. 4 is an explanatory diagram for explaining the extraction process. FIG. 5 is an explanatory diagram for explaining an example of an abnormal state of the mold clamping mechanism 20.
　学習済みモデルＭｄは、型締機構２０が異常状態であるとき、評価値の算出結果が所定範囲内となるように、ディープラーニングや、ニューラルネットワーク等の技術に基づいて、教師データを機械学習させることによって生成される。 The trained model Md machine-learns teacher data based on techniques such as deep learning and neural networks so that the calculation result of the evaluation value falls within a predetermined range when the mold clamping mechanism 20 is in an abnormal state. Is generated by.
　図３に示すように、型締処理において、型締め後、可動側金型２４を引き離すと、可動側分割部２４ｃ上には、成形物Ｍｏが表れる。図４に示すように、取出処理において、エジェクタピン２７がコア２４ｄから突出すると、成形物Ｍｏは、押し出され、型締機構２０から取り出される。 As shown in FIG. 3, in the mold clamping process, when the movable mold 24 is pulled apart after the mold is compacted, the molded product Mo appears on the movable side divided portion 24c. As shown in FIG. 4, when the ejector pin 27 protrudes from the core 24d in the ejection process, the molded product Mo is extruded and ejected from the mold clamping mechanism 20.
　取出処理の後、正常状態では、可動側分割部２４ｃ上には成形物Ｍｏが残されない。成形物Ｍｏを有しない可動側分割部２４ｃを表した学習用画像は、取出処理後における型締機構２０の正常状態と相関関係を有する。教師データは、型締機構２０の正常状態を示すものとして、成形物Ｍｏを有しない可動側分割部２４ｃを表した学習用画像が用いられる。 After the take-out process, in the normal state, no molded product Mo is left on the movable side dividing portion 24c. The learning image showing the movable side divided portion 24c having no molded product Mo has a correlation with the normal state of the mold clamping mechanism 20 after the extraction process. As the teacher data, a learning image showing the movable side divided portion 24c having no molded product Mo is used as an indicator of the normal state of the mold clamping mechanism 20.
　一方、図５に示すように、取出処理の後、異常状態では、成形物Ｍｏが、引っ掛かり等によって可動側分割部２４ｃ上に残される。成形物Ｍｏを有する可動側分割部２４ｃの学習用画像は、取出処理の後における型締機構２０の異常状態と相関関係を有する。教師データは、型締機構２０の異常状態を示すものとして、成形物Ｍｏを有する可動側分割部２４ｃを表した学習用画像が用いられる。 On the other hand, as shown in FIG. 5, after the take-out process, in an abnormal state, the molded product Mo is left on the movable side split portion 24c due to catching or the like. The learning image of the movable side split portion 24c having the molded product Mo has a correlation with the abnormal state of the mold clamping mechanism 20 after the extraction process. As the teacher data, a learning image showing the movable side split portion 24c having the molded product Mo is used as an indicator of the abnormal state of the mold clamping mechanism 20.
　学習用画像は、開発段階において、射出成形処理を行い、カメラ６０によって可動側分割部２４ｃを撮像して取得した撮像画像Ｉｍに基づいて作成される。より具体的に、成形物Ｍｏを有する可動側分割部２４ｃを表した撮像画像Ｉｍは、型締め後から成形物Ｍｏの取出し前における可動側分割部２４ｃを撮像することによって取得される。成形物Ｍｏを有しない可動側分割部２４ｃを表した撮像画像Ｉｍは、成形物Ｍｏの取出し後から次回の型締め前の可動側分割部２４ｃを撮像することによって取得される。 The learning image is created based on the captured image Im obtained by performing an injection molding process at the development stage and capturing the movable side split portion 24c with the camera 60. More specifically, the captured image Im representing the movable side divided portion 24c having the molded product Mo is acquired by imaging the movable side divided portion 24c after molding and before taking out the molded product Mo. The captured image Im showing the movable side divided portion 24c having no molded product Mo is acquired by imaging the movable side divided portion 24c after taking out the molded product Mo and before the next mold clamping.
　撮像画像Ｉｍは、可動側分割部２４ｃを上方、側方、下方等、様々な角度から撮像して作成することが好ましい。また、撮像画像Ｉｍは、水平反転、垂直反転、回転、拡大、縮小、明度変更等によって水増しさせてもよい。 It is preferable that the captured image Im is created by imaging the movable side divided portion 24c from various angles such as upward, lateral, and downward. Further, the captured image Im may be inflated by horizontal inversion, vertical inversion, rotation, enlargement, reduction, brightness change, or the like.
　撮像画像Ｉｍの一部は、学習済みモデルＭｄの評価に使用される。 A part of the captured image Im is used for evaluation of the trained model Md.
　すなわち、異常監視装置３は、監視対象である成形機本体２の型締機構２０が異常状態であるか否かを監視する。監視対象は、より具体的には、型締機構２０に設けられた金型及び金型の分割部である。さらに具体的には、監視対象は、可動側金型２４及び可動側分割部２４ｃである。 That is, the abnormality monitoring device 3 monitors whether or not the mold clamping mechanism 20 of the molding machine main body 2 to be monitored is in an abnormal state. More specifically, the monitoring target is a mold provided in the mold clamping mechanism 20 and a split portion of the mold. More specifically, the monitoring targets are the movable side mold 24 and the movable side dividing portion 24c.
　制御部７１は、金型の分割部を表した学習用画像を機械学習することによって生成された学習済みモデルＭｄと、カメラ６０によって分割部を撮像して取得された撮像画像Ｉｍとに基づいて、分割部が異常状態であるか否かを判定し、分割部が異常状態であると判定したとき、型締めを停止させるための停止指示Ｓｔを出力する。 The control unit 71 is based on the learned model Md generated by machine learning the learning image representing the divided portion of the mold and the captured image Im acquired by imaging the divided portion with the camera 60. , It is determined whether or not the divided portion is in an abnormal state, and when it is determined that the divided portion is in an abnormal state, a stop instruction St for stopping the mold clamping is output.
　制御部７１は、撮像画像Ｉｍを学習済みモデルＭｄに入力することによって評価値を算出し、評価値が所定範囲内にあるとき、分割部が異常状態であると判定する。 The control unit 71 calculates the evaluation value by inputting the captured image Im into the trained model Md, and determines that the division unit is in an abnormal state when the evaluation value is within a predetermined range.
　学習済みモデルＭｄは、正常状態を示す成形物Ｍｏを有しない分割部を表した学習用画像と、異常状態を示す成形物Ｍｏを有する分割部を表した学習用画像とを機械学習することによって生成される。 The trained model Md is obtained by machine learning a learning image showing a divided portion having no molded product Mo showing a normal state and a learning image showing a divided portion having a molded product Mo showing an abnormal state. Will be generated.
　また、異常監視方法は、金型の分割部を表した学習用画像を機械学習することによって生成された学習済みモデルＭｄを用意し、カメラ６０によって分割部を撮像して撮像画像Ｉｍを取得し、制御部７１により、金型の型締めによって成形された成形物Ｍｏの取出処理の終了を示す制御情報Ｃｉの入力に応じ、撮像画像Ｉｍと学習済みモデルＭｄに基づいて、分割部が異常状態であるか否かを判定し、分割部が異常状態であると判定したとき、型締めを停止させるための停止指示Ｓｔを出力する。 Further, as an abnormality monitoring method, a learned model Md generated by machine learning a learning image showing a divided portion of a mold is prepared, and the divided portion is imaged by a camera 60 to acquire an captured image Im. In response to the input of the control information Ci indicating the end of the extraction process of the molded product Mo formed by the mold clamping by the control unit 71, the divided unit is in an abnormal state based on the captured image Im and the learned model Md. When it is determined that the split portion is in an abnormal state, a stop instruction St for stopping the mold clamping is output.
　また、異常監視プログラムは、金型の型締めによって成形された成形物Ｍｏの取出処理の終了を示す制御情報Ｃｉの入力に応じ、金型の分割部を表した学習用画像を機械学習することによって生成された学習済みモデルＭｄと、カメラ６０によって分割部を撮像して取得された撮像画像Ｉｍとに基づいて、分割部が異常状態であるか否かを判定するプログラムと、分割部が異常状態であると判定したとき、型締めを停止させるための停止指示Ｓｔを出力するプログラムからなる。 In addition, the abnormality monitoring program machine-learns a learning image showing the divided portion of the mold in response to the input of the control information Ci indicating the end of the extraction process of the molded product Mo formed by the mold clamping. Based on the trained model Md generated by the camera 60 and the captured image Im acquired by imaging the divided portion with the camera 60, a program for determining whether or not the divided portion is in an abnormal state, and a program for determining whether the divided portion is abnormal It consists of a program that outputs a stop instruction St for stopping the mold clamping when it is determined that the state is in the state.
（動作）　
　続いて、射出成形機１の動作について説明をする。図６は、射出成形処理の流れの一例を示すフローチャートである。 (motion)
Subsequently, the operation of the injection molding machine 1 will be described. FIG. 6 is a flowchart showing an example of the flow of the injection molding process.
　可動側金型２４を閉じる（Ｓ１）。制御部４０は、可動側金型２４を閉じるための指示を出力する。型締駆動部２６は、指示に応じ、可動側金型２４を一方へ移動させ、可動側金型２４を固定側金型２３に押し当て、型締めする。 Close the movable mold 24 (S1). The control unit 40 outputs an instruction for closing the movable mold 24. The mold clamping drive unit 26 moves the movable mold 24 to one side in response to an instruction, presses the movable mold 24 against the fixed mold 23, and clamps the mold.
　樹脂材料を射出する（Ｓ２）。制御部４０は、樹脂材料を射出するための指示を出力する。射出駆動部１３は、指示に応じ、樹脂材料をノズル１２ａから型締機構２０内に射出する。樹脂材料は、スプール２５を流れ、キャビティ２３ｄとコア２４ｄの間に形成された空間に流れ込み、成形され、成形物Ｍｏを形成する。 Inject the resin material (S2). The control unit 40 outputs an instruction for injecting the resin material. The injection drive unit 13 injects the resin material from the nozzle 12a into the mold clamping mechanism 20 in response to an instruction. The resin material flows through the spool 25, flows into the space formed between the cavity 23d and the core 24d, and is molded to form a molded product Mo.
　可動側金型２４を開く（Ｓ３）。制御部４０は、可動側金型２４を開くための指示を出力する。型締駆動部２６は、指示に応じ、可動側金型２４を他方へ移動させ、可動側金型２４を固定側金型２３から引き離す。可動側分割部２４ｃ上に成形物Ｍｏが表れる。 Open the movable mold 24 (S3). The control unit 40 outputs an instruction for opening the movable mold 24. The mold clamping drive unit 26 moves the movable side mold 24 to the other side and pulls the movable side mold 24 away from the fixed side mold 23 in response to an instruction. The molded product Mo appears on the movable side divided portion 24c.
　成形物Ｍｏを取り出す（Ｓ４）。制御部４０は、成形物Ｍｏを取り出すための指示を出力する。取出駆動部２８は、指示に応じ、エジェクタピン２７をコア２４ｄから突出させ、続いて、エジェクタピン２７を引き戻してコア２４ｄに埋没させる。制御部４０は、取出処理の終了を示す制御情報Ｃｉを出力する。 Take out the molded product Mo (S4). The control unit 40 outputs an instruction for taking out the molded product Mo. The take-out drive unit 28 projects the ejector pin 27 from the core 24d in response to an instruction, and then pulls back the ejector pin 27 and buries it in the core 24d. The control unit 40 outputs the control information Ci indicating the end of the retrieval process.
　異常状態であるか否かの判定を行う（Ｓ５）。制御部７１は、制御情報Ｃｉに応じ、カメラ６０から入力された撮像画像Ｉｍを学習済みモデルＭｄに入力し、評価値を算出する。評価値が所定範囲内であると制御部７１が判定すると、処理はＳ６に進む（Ｓ６：ＹＥＳ）。一方、評価値が所定範囲内ではないとき、処理はＳ１に戻る（Ｓ６：ＮＯ）。 Determine whether or not it is in an abnormal state (S5). The control unit 71 inputs the captured image Im input from the camera 60 into the trained model Md according to the control information Ci, and calculates an evaluation value. When the control unit 71 determines that the evaluation value is within the predetermined range, the process proceeds to S6 (S6: YES). On the other hand, when the evaluation value is not within the predetermined range, the process returns to S1 (S6: NO).
　停止指示Ｓｔを出力する（Ｓ６）。制御部７１は、型締機構２０を停止させるための停止指示Ｓｔを出力する。 Output the stop instruction St (S6). The control unit 71 outputs a stop instruction St for stopping the mold clamping mechanism 20.
　型締機構２０を停止させる（Ｓ７）。制御部４０は、停止指示Ｓｔに応じ、型締機構２０を停止させる。 Stop the mold clamping mechanism 20 (S7). The control unit 40 stops the mold clamping mechanism 20 in response to the stop instruction St.
　駆動再開指示があるまで待機する（Ｓ８）。ユーザは、型締機構２０内を点検し、成形物Ｍｏが残留している場合には取り除き、型締機構２０を正常状態に復帰させる。正常状態に復帰した後、ユーザは、指示入力部３０によって駆動再開指示を入力する。指示入力部３０から駆動再開指示があると、処理はＳ１に戻る（Ｓ８：ＹＥＳ）。指示入力部３０から駆動再開指示があるまで、制御部４０は、待機する（Ｓ８：ＮＯ）。 Wait until there is a drive restart instruction (S8). The user inspects the inside of the mold clamping mechanism 20, removes any residual molded product Mo, and restores the mold clamping mechanism 20 to the normal state. After returning to the normal state, the user inputs a drive restart instruction by the instruction input unit 30. When there is a drive restart instruction from the instruction input unit 30, the process returns to S1 (S8: YES). The control unit 40 stands by (S8: NO) until there is a drive restart instruction from the instruction input unit 30.
　Ｓ１～Ｓ８の処理は、射出成形処理である。Ｓ１とＳ３の処理は、型締処理である。Ｓ２の処理は、射出処理である。Ｓ４の処理は、取出処理である。Ｓ５とＳ６の処理は、異常判定処理である。Ｓ７とＳ８の処理は、停止処理である。 The treatments S1 to S8 are injection molding treatments. The processing of S1 and S3 is a mold clamping process. The process of S2 is an injection process. The process of S4 is a take-out process. The processing of S5 and S6 is an abnormality determination processing. The processing of S7 and S8 is a stop processing.
　これにより、異常監視装置３は、成形機本体２を新設したときや、可動側金型２４を交換したとき等、監視対象となる可動側金型２４が新規であるときにおいても、ユーザによる煩雑な初期設定作業が発生することなく、撮像画像Ｉｍを学習済みモデルＭｄに入力することによって評価値を算出し、評価値に応じて型締機構２０が異常状態であるか否かを判定することができる。 As a result, the abnormality monitoring device 3 is complicated by the user even when the movable side mold 24 to be monitored is new, such as when the molding machine main body 2 is newly installed or when the movable side mold 24 is replaced. The evaluation value is calculated by inputting the captured image Im into the trained model Md without any initial setting work, and it is determined whether or not the mold clamping mechanism 20 is in an abnormal state according to the evaluation value. Can be done.
　実施形態によれば、異常監視装置３は、監視対象が新規であるときにおいても、ユーザによる煩雑な初期設定作業が発生することなく、監視対象の異常状態を監視することができる。 According to the embodiment, the abnormality monitoring device 3 can monitor the abnormal state of the monitoring target without causing complicated initial setting work by the user even when the monitoring target is new.
（実施形態の変形例１）　
　実施形態及び他の変形例では、学習済みモデルＭｄが追加学習されないが、学習済みモデルＭｄは、追加学習されるように構成してもよい。 (Modified Example 1 of the Embodiment)
In the embodiment and other modifications, the trained model Md is not additionally trained, but the trained model Md may be configured to be additionally trained.
　図７は、追加学習処理の流れの一例を示すフローチャートである。 FIG. 7 is a flowchart showing an example of the flow of the additional learning process.
　追加学習処理は、予め定められたスケジュールに応じて開始する。 The additional learning process starts according to a predetermined schedule.
　撮像画像Ｉｍを記憶部７２に記憶させる（Ｓ１１）。射出成形処理において、制御部７１は、成形物Ｍｏを有する可動側分割部２４ｃと、成形物Ｍｏを有しない可動側分割部２４ｃとの各々を表した撮像画像Ｉｍを記憶部７２に記憶させる。成形物Ｍｏを有する可動側分割部２４ｃを表した撮像画像Ｉｍは、型締め後から成形物Ｍｏの取出し前における可動側分割部２４ｃを撮像することによって取得される。成形物Ｍｏを有しない可動側分割部２４ｃを表した撮像画像Ｉｍは、成形物Ｍｏの取出し後から次回の型締め前の可動側分割部２４ｃを撮像することによって取得される。 The captured image Im is stored in the storage unit 72 (S11). In the injection molding process, the control unit 71 stores the imaged image Im representing each of the movable side divided portion 24c having the molded product Mo and the movable side divided portion 24c having no molded product Mo in the storage unit 72. The captured image Im showing the movable side divided portion 24c having the molded product Mo is acquired by imaging the movable side divided portion 24c after molding and before taking out the molded product Mo. The captured image Im showing the movable side divided portion 24c having no molded product Mo is acquired by imaging the movable side divided portion 24c after taking out the molded product Mo and before the next mold clamping.
　機械学習を開始するか否かを判定する（Ｓ１２）。制御部７１は、記憶部７２から読み込みをしたスケジュールに応じ、機械学習を開始するか否かを判定する。機械学習は、例えば夜間等、射出成形機１の駆動が停止される時間に行うのが好ましい。機械学習を開始するとき、処理はＳ１３に進む（Ｓ１２：ＹＥＳ）。機械学習を開始しないとき、処理は、Ｓ１１に戻る（Ｓ１２：ＮＯ）。 Determine whether to start machine learning (S12). The control unit 71 determines whether or not to start machine learning according to the schedule read from the storage unit 72. Machine learning is preferably performed at a time when the drive of the injection molding machine 1 is stopped, such as at night. When starting machine learning, the process proceeds to S13 (S12: YES). When machine learning is not started, the process returns to S11 (S12: NO).
　記憶部７２から撮像画像Ｉｍを読み込む（Ｓ１３）。 The captured image Im is read from the storage unit 72 (S13).
　機械学習を行う（Ｓ１４）。制御部７１は、読み込みをした撮像画像Ｉｍに基づいて機械学習を行い、学習済みモデルＭｄを更新する。教師データは、型締機構２０の正常状態を示すものとして、成形物Ｍｏを有しない可動側分割部２４ｃの撮像画像Ｉｍが用いられる。また、教師データは、型締機構２０の異常状態を示すものとして、成形物Ｍｏを有する可動側分割部２４ｃの撮像画像Ｉｍが用いられる。 Perform machine learning (S14). The control unit 71 performs machine learning based on the read captured image Im, and updates the trained model Md. As the teacher data, the captured image Im of the movable side split portion 24c having no molded product Mo is used as an indicator of the normal state of the mold clamping mechanism 20. Further, as the teacher data, the captured image Im of the movable side divided portion 24c having the molded product Mo is used as an indicator of the abnormal state of the mold clamping mechanism 20.
　すなわち、記憶部７２は、撮像画像Ｉｍを記憶し、制御部７１は、撮像画像Ｉｍを機械学習することによって学習済みモデルＭｄを更新する。 That is, the storage unit 72 stores the captured image Im, and the control unit 71 updates the learned model Md by machine learning the captured image Im.
　これにより、異常監視装置３は、カメラ６０によって撮像した撮像画像Ｉｍに基づいて、学習済みモデルＭｄの追加学習処理をすることができ、型締機構２０の異常状態の判定の精度を向上させることができる。 As a result, the abnormality monitoring device 3 can perform additional learning processing of the trained model Md based on the captured image Im captured by the camera 60, and improve the accuracy of determining the abnormal state of the mold clamping mechanism 20. Can be done.
（実施形態の変形例２）　
　実施形態及び他の変形例では、異常監視装置３がサーバ４と接続されないが、異常監視装置３は、サーバ４と接続するように構成してもよい（図１の２点鎖線）。 (Modification 2 of the embodiment)
In the embodiment and other modifications, the abnormality monitoring device 3 is not connected to the server 4, but the abnormality monitoring device 3 may be configured to be connected to the server 4 (two-dot chain line in FIG. 1).
　制御部７１は、インターネット又はＬＡＮ等のネットワークを介し、有線通信又は無線通信によってサーバ４と接続される。 The control unit 71 is connected to the server 4 by wired communication or wireless communication via a network such as the Internet or LAN.
　制御部７１は、撮像画像Ｉｍをサーバ４に出力可能である。 The control unit 71 can output the captured image Im to the server 4.
　サーバ４は、監視装置本体７０から入力された撮像画像Ｉｍに基づいて追加学習処理を行い、追加学習処理の結果を監視装置本体７０に出力する。 The server 4 performs additional learning processing based on the captured image Im input from the monitoring device main body 70, and outputs the result of the additional learning processing to the monitoring device main body 70.
　また、制御部７１は、サーバ４から入力された追加学習処理の結果に基づいて、学習済みモデルＭｄを更新する。 Further, the control unit 71 updates the trained model Md based on the result of the additional learning process input from the server 4.
　これにより、サーバ４は、複数の異常監視装置３から撮像画像Ｉｍの入力を受け、追加学習処理を行うことができ、より多くの撮像画像Ｉｍに基づいて、追加学習処理を行うことが可能である。 As a result, the server 4 can receive the input of the captured image Im from the plurality of abnormality monitoring devices 3 and perform the additional learning process, and can perform the additional learning process based on more captured images Im. is there.
（実施形態の変形例３）　
　実施形態及び他の変形例では、異常監視装置３の駆動状況を監視することができないが、サーバ４が異常監視装置３の駆動状況を監視するように構成してもよい。 (Modification 3 of the embodiment)
In the embodiment and other modifications, the drive status of the abnormality monitoring device 3 cannot be monitored, but the server 4 may be configured to monitor the drive status of the abnormality monitoring device 3.
　サーバ４は、所定のタイミングによって制御部７１にキープアライブ信号を出力する。制御部７１は、キープアライブ信号を受信すると、サーバ４に応答信号を出力する。所定時間内に応答信号を受信できないとき、サーバ４は、メール等の通知手段によってユーザに異常監視装置３が停止状態にあることを警告する。 The server 4 outputs a keep-alive signal to the control unit 71 at a predetermined timing. When the control unit 71 receives the keep-alive signal, the control unit 71 outputs a response signal to the server 4. When the response signal cannot be received within the predetermined time, the server 4 warns the user that the abnormality monitoring device 3 is in the stopped state by a notification means such as an e-mail.
　すなわち、サーバ４は、所定のタイミングによって制御部７１に信号を出力し、所定時間内に制御部７１から応答信号を受信できないとき、警告を出力する。 That is, the server 4 outputs a signal to the control unit 71 at a predetermined timing, and outputs a warning when the response signal cannot be received from the control unit 71 within the predetermined time.
　これにより、異常監視装置３は、より確実に、型締機構２０が異常状態であるか否かを判定することができる。 As a result, the abnormality monitoring device 3 can more reliably determine whether or not the mold clamping mechanism 20 is in an abnormal state.
（実施形態の変形例４）　
　実施形態及び他の変形例では、成形物Ｍｏを有しない可動側金型２４を表した学習用画像と、成形物Ｍｏを有する可動側金型２４を表した学習用画像とを機械学習させることによって生成されたが、これに限定されない。 (Modified Example 4 of the Embodiment)
In the embodiment and other modifications, the learning image showing the movable mold 24 having the molded product Mo and the learning image showing the movable mold 24 having the molded product Mo are machine-learned. Generated by, but not limited to.
　取出処理の後、正常状態では、金属を材質とした可動側分割部２４ｃが表れ、一方、異常状態では、可動側分割部２４ｃ上に樹脂を材質とした成形物Ｍｏが表れる。金属を表した可動側分割部２４ｃの学習用画像は、型締機構２０の正常状態と相関関係を有する。また、金属及び樹脂の両方、又は、樹脂を表した学習用画像は、型締機構２０の異常状態と相関関係を有する。 After the take-out process, in the normal state, the movable side split portion 24c made of metal appears, while in the abnormal state, the molded product Mo made of resin appears on the movable side split portion 24c. The learning image of the movable side split portion 24c showing the metal has a correlation with the normal state of the mold clamping mechanism 20. Further, the learning image showing both the metal and the resin or the resin has a correlation with the abnormal state of the mold clamping mechanism 20.
　学習済みモデルＭｄは、型締機構２０の正常状態を示すものとして金属を表した学習用画像と、異常状態を示すものとして金属と樹脂の両方、又は、樹脂を表した学習用画像とを機械学習させることによって生成される。 The trained model Md is a machine learning image showing a metal as showing the normal state of the mold clamping mechanism 20 and a learning image showing both metal and resin or resin as showing an abnormal state. It is generated by learning.
　これにより、異常監視装置３は、取出処理後を示す撮像画像Ｉｍに樹脂が表れているとき、型締機構２０が異常状態であると判定することができる。 As a result, the abnormality monitoring device 3 can determine that the mold clamping mechanism 20 is in an abnormal state when the resin appears in the captured image Im indicating after the extraction process.
（実施形態の変形例５）　
　実施形態及び他の変形例では、１つの学習済みモデルＭｄによって型締機構２０が異常状態であるか否かを判定したが、これに限定されない。学習済みモデルＭｄは、複数であってもよい。 (Modified Example 5 of the Embodiment)
In the embodiment and other modifications, it is determined by one trained model Md whether or not the mold clamping mechanism 20 is in an abnormal state, but the present invention is not limited to this. The trained model Md may be plural.
　例えば、学習済みモデルＭｄは、第１学習済みモデル、第２学習済みモデル、及び、第３学習済みモデルを有する。 For example, the trained model Md has a first trained model, a second trained model, and a third trained model.
　第１学習済みモデルは、可動側分割部２４ｃを照明する照明の種類及びカメラ６０の種類を変化させて撮像することによって作成した学習用画像に基づいて生成される。 The first trained model is generated based on a learning image created by changing the type of lighting that illuminates the movable side dividing portion 24c and the type of the camera 60 and taking an image.
　第２学習済みモデルは、可動側分割部２４ｃの種類を変化させて撮像することによって作成した学習用画像に基づいて生成される。 The second learned model is generated based on the learning image created by changing the type of the movable side dividing portion 24c and taking an image.
　第３学習済みモデルは、成形物Ｍｏの色を変化させて撮像することによって作成した学習用画像に基づいて生成される。 The third trained model is generated based on the learning image created by changing the color of the molded product Mo and taking an image.
　制御部７１は、第１学習済みモデルと撮像画像Ｉｍに基づいて算出された第１評価値と、第２学習済みモデルと撮像画像Ｉｍに基づいて算出された第２評価値と、第３学習済みモデルと撮像画像Ｉｍに基づいて算出された第３評価値とに応じ、型締機構２０の分割部が異常状態であるか否かを判定する。 The control unit 71 has a first evaluation value calculated based on the first trained model and the captured image Im, a second evaluation value calculated based on the second trained model and the captured image Im, and a third learning. It is determined whether or not the divided portion of the mold clamping mechanism 20 is in an abnormal state according to the completed model and the third evaluation value calculated based on the captured image Im.
　より具体的には、制御部７１は、第１評価値、第２評価値、及び、第３評価値の各々を所定の重み付け係数によって重み付けし、重み付けの結果を合計して評価値を算出し、評価値が所定範囲内にあるとき、異常状態であると判定する。 More specifically, the control unit 71 weights each of the first evaluation value, the second evaluation value, and the third evaluation value by a predetermined weighting coefficient, and totals the weighting results to calculate the evaluation value. , When the evaluation value is within the predetermined range, it is determined that the state is abnormal.
　これにより、異常監視装置３は、より多くの種類の型締機構２０について、より高い精度によって異常状態であるか否かを判定可能である。 Thereby, the abnormality monitoring device 3 can determine whether or not the mold clamping mechanism 20 of more types is in an abnormal state with higher accuracy.
　なお、実施形態及び変形例では、カメラ６０は、可動側分割部２４ｃを監視するが、これに限定されない。型締め後であって成形物Ｍｏの取出し前に、可動側分割部２４ｃではなく、固定側分割部２３ｃに成形物Ｍｏが表れるとき、固定側分割部２３ｃを監視してもよい。 In the embodiment and the modified example, the camera 60 monitors the movable side split portion 24c, but the present invention is not limited to this. When the molded product Mo appears in the fixed-side divided portion 23c instead of the movable-side divided portion 24c after molding and before taking out the molded product Mo, the fixed-side divided portion 23c may be monitored.
　なお、実施形態及び変形例では、異常監視装置３は、可動側分割部２４ｃをカメラ６０によって撮像するが、固定側金型２３に成形物Ｍｏが形成されるときには、固定側分割部２３ｃをカメラ６０によって撮像し、型締機構２０が異常状態であるか否かを判定してもよい。また、異常監視装置３は、可動側分割部２４ｃおよび固定側分割部２３ｃの両方を撮像し、型締機構２０が異常状態であるか否かを判定してもよい。 In the embodiment and the modified example, the abnormality monitoring device 3 captures the movable side split portion 24c with the camera 60, but when the molded product Mo is formed on the fixed side mold 23, the fixed side split portion 23c is captured by the camera. An image may be taken by 60 to determine whether or not the mold clamping mechanism 20 is in an abnormal state. Further, the abnormality monitoring device 3 may image both the movable side dividing portion 24c and the fixed side dividing portion 23c to determine whether or not the mold clamping mechanism 20 is in an abnormal state.
　なお、実施形態及び変形例では、キャビティ２３ｄ及びコア２４ｄが各１つ設けられた例を説明したが、キャビティ２３ｄ及びコア２４ｄは、各２つ以上設けられてもよい。また、実施形態及び変形例では、ランナーを説明していないがランナーを有してもよい。 In the embodiment and the modified example, an example in which one cavity 23d and one core 24d are provided has been described, but two or more cavities 23d and two or more cores 24d may be provided. Moreover, although the runner is not described in the embodiment and the modified example, the runner may be included.
　なお、実施形態及び変形例では、カメラ６０がアーム６１によって固定盤２１の上部に位置するが、カメラ６０の位置は、これに限定されない。カメラ６０は、アーム６１によって固定盤２１の側部又は下部に位置してもよい。また、カメラ６０は、コア２４ｄを撮像できる位置であれば、可動側金型２４の側方に位置してもよい。また、カメラ６０は、固定盤２１以外に取り付けられてもよい。 In the embodiment and the modified example, the camera 60 is located above the fixed plate 21 by the arm 61, but the position of the camera 60 is not limited to this. The camera 60 may be located on the side or lower part of the fixing plate 21 by the arm 61. Further, the camera 60 may be located on the side of the movable mold 24 as long as the core 24d can be imaged. Further, the camera 60 may be attached to other than the fixed plate 21.
　なお、実施形態及び変形例では、記憶部７２は、異常判定部Ｐｊ及び学習済みモデルＭｄを記憶するが、これらは記憶部５０又はサーバ４に記憶されてもよい。その場合、制御部７１の有する機能は、制御部４０又はサーバ４によって実現されてもよい。 In the embodiment and the modified example, the storage unit 72 stores the abnormality determination unit Pj and the learned model Md, but these may be stored in the storage unit 50 or the server 4. In that case, the function of the control unit 71 may be realized by the control unit 40 or the server 4.
　なお、変形例１では、追加学習処理の開始の判定（Ｓ１２）が、予め定められたスケジュールを記憶部７２から読み込むことによって行われたが、追加学習処理の開始の判定は、指示入力部３０又は制御部７１に接続される指示部（不図示）を介したユーザの指示入力によって行われてもよい。 In the first modification, the determination of the start of the additional learning process (S12) was performed by reading the predetermined schedule from the storage unit 72, but the determination of the start of the additional learning process was performed by the instruction input unit 30. Alternatively, it may be performed by a user's instruction input via an instruction unit (not shown) connected to the control unit 71.
　本実施形態における各手順の各ステップは、その性質に反しない限り、実行順序を変更し、複数同時に実行し、あるいは実行毎に異なった順序で実行してもよい。さらに、本実施形態における各手順の各ステップの全てあるいは一部をハードウェアにより実現してもよい。 Each step of each procedure in the present embodiment may be executed at the same time by changing the execution order, or may be executed in a different order for each execution, as long as the property is not contrary to the property. Further, all or a part of each step of each procedure in the present embodiment may be realized by hardware.
　本発明は、上述した実施の形態に限定されるものではなく、本発明の要旨を変えない範囲において、種々の変更、改変等が可能である。 The present invention is not limited to the above-described embodiment, and various modifications, modifications, and the like can be made without changing the gist of the present invention.
１　　　　　　射出成形機
２　　　　　　成形機本体
３　　　　　　異常監視装置
４　　　　　　サーバ
１０　　　　　　射出機構
１１　　　　　　ホッパー
１２　　　　　　射出筒
１２ａ　　　　　　ノズル
１３　　　　　　射出駆動部
２０　　　　　　型締機構
２１　　　　　　固定盤
２１ａ　　　　　　ノズル接続部
２２　　　　　　可動盤
２２ａ　　　　　　タイバー
２３　　　　　　固定側金型
２３ａ　　　　　　固定側取付板
２３ｂ　　　　　　ガイドブッシュ
２３ｃ　　　　　　固定側分割部
２３ｄ　　　　　　キャビティ
２４　　　　　　可動側金型
２４ａ　　　　　　可動側取付板
２４ｂ　　　　　　ガイドピン
２４ｃ　　　　　　可動側分割部
２４ｄ　　　　　　コア
２５　　　　　　スプール
２６　　　　　　型締駆動部
２７　　　　　　エジェクタピン
２８　　　　　　取出駆動部
３０　　　　　　指示入力部
４０　　　　　　制御部
５０　　　　　　記憶部
６０　　　　　　カメラ
６１　　　　　　アーム
６２　　　　　　取付け具
７０　　　　　　監視装置本体
７１　　　　　　制御部
７２　　　　　　記憶部
Ｃｉ　　　　　　制御情報
Ｉｍ　　　　　　撮像画像
Ｍｄ　　　　　　学習済みモデル
Ｍｏ　　　　　　成形物
Ｐｊ　　　　　　異常判定部
Ｓｔ　　　　　　停止指示 1 Injection molding machine 2 Molding machine body 3 Abnormality monitoring device 4 Server 10 Injection mechanism 11 Hopper 12 Injection cylinder 12a Nozzle 13 Injection drive unit 20 Mold clamping mechanism 21 Fixed plate 21a Nozzle connection part 22 Movable plate 22a Tie bar 23 Fixed side mold 23a Fixed side mounting plate 23b Guide bush 23c Fixed side split part 23d Cavity 24 Movable side mold 24a Movable side mounting plate 24b Guide pin 24c Movable side split part 24d Core 25 Spool 26 Mold tightening drive part 27 Ejector pin 28 Ejection drive part 30 Instruction Input unit 40 Control unit 50 Storage unit 60 Camera 61 Arm 62 Fixture 70 Monitoring device body 71 Control unit 72 Storage unit Ci Control information Im Captured image Md Learned model Mo Molded product Pj Abnormality determination unit St Stop instruction

Claims (11)

  1. 　金型の分割部を表した学習用画像を機械学習することによって生成された学習済みモデルと、カメラによって前記分割部を撮像して取得された撮像画像とに基づいて、前記分割部が異常状態であるか否かを判定し、前記分割部が異常状態であると判定したとき、型締めを停止させるための停止指示を出力する、制御部を有する、異常監視装置。 The divided portion is in an abnormal state based on a learned model generated by machine learning a learning image representing the divided portion of the mold and an captured image obtained by imaging the divided portion with a camera. An abnormality monitoring device having a control unit that outputs a stop instruction for stopping mold clamping when it is determined that the split portion is in an abnormal state.
  2. 　前記制御部は、前記撮像画像を前記学習済みモデルに入力することによって評価値を算出し、前記評価値が所定範囲内にあるとき、前記分割部が異常状態であると判定する、請求項１に記載の異常監視装置。 The control unit calculates an evaluation value by inputting the captured image into the trained model, and determines that the divided unit is in an abnormal state when the evaluation value is within a predetermined range. Abnormality monitoring device described in.
  3. 　前記学習済みモデルは、
    　　正常状態を示す成形物を有しない前記分割部を表した前記学習用画像と、
    　　前記異常状態を示す前記成形物を有する前記分割部を表した前記学習用画像と、
    　を機械学習することによって生成された、
    　請求項１に記載の異常監視装置。     The trained model is
    The learning image showing the divided portion having no molded product showing a normal state, and
    The learning image showing the divided portion having the molded product showing the abnormal state, and the learning image.
    Generated by machine learning,
    The abnormality monitoring device according to claim 1.
  4. 　前記学習済みモデルは、正常状態を示すものとして金属を表した前記学習用画像と、前記異常状態を示すものとして金属と樹脂の両方、又は、樹脂を表した前記学習用画像とを機械学習させることによって生成された、請求項１に記載の異常監視装置。 The trained model is machine-learned from the learning image showing the metal as showing the normal state and the learning image showing both the metal and the resin or the resin as showing the abnormal state. The abnormality monitoring device according to claim 1, which is generated by the above.
  5. 　前記学習済みモデルは、
    　　前記分割部を照明する照明の種類及び前記カメラの種類を変化させて撮像することによって作成した前記学習用画像に基づいて生成された第１学習済みモデルと、
    　　前記分割部の種類を変化させて撮像することによって作成した前記学習用画像に基づいて生成された第２学習済みモデルと、
    　　成形物の色を変化させて撮像することによって作成した前記学習用画像に基づいて生成された第３学習済みモデルと、を有し、
    　前記制御部は、前記第１学習済みモデルと前記撮像画像に基づいて算出された第１評価値と、前記第２学習済みモデルと前記撮像画像に基づいて算出された第２評価値と、前記第３学習済みモデルと前記撮像画像に基づいて算出された第３評価値とに応じ、前記分割部が異常状態であるか否かを判定する、
    　請求項１に記載の異常監視装置。     The trained model is
    A first trained model generated based on the learning image created by changing the type of lighting that illuminates the divided portion and the type of the camera and taking an image.
    A second trained model generated based on the learning image created by changing the type of the divided portion and taking an image, and
    It has a third trained model generated based on the training image created by changing the color of the molded product and taking an image.
    The control unit includes a first evaluation value calculated based on the first trained model and the captured image, a second evaluation value calculated based on the second trained model and the captured image, and the above. It is determined whether or not the divided portion is in an abnormal state according to the third trained model and the third evaluation value calculated based on the captured image.
    The abnormality monitoring device according to claim 1.
  6. 　記憶部を有し、
    　前記記憶部は、前記撮像画像を記憶し、
    　前記制御部は、前記撮像画像を機械学習することによって前記学習済みモデルを更新する、
    　請求項１に記載の異常監視装置。     Has a memory
    The storage unit stores the captured image and stores the captured image.
    The control unit updates the trained model by machine learning the captured image.
    The abnormality monitoring device according to claim 1.
  7. 　サーバを有し、
    　前記サーバは、前記制御部から入力された前記撮像画像に基づいて追加学習処理を行い、追加学習処理の結果を前記制御部に出力する、請求項１に記載の異常監視装置。     Have a server
    The abnormality monitoring device according to claim 1, wherein the server performs additional learning processing based on the captured image input from the control unit, and outputs the result of the additional learning processing to the control unit.
  8. 　前記サーバは、所定のタイミングによって前記制御部に信号を出力し、所定時間内に前記制御部から応答信号を受信できないとき、警告を出力する、請求項７に記載の異常監視装置。 The abnormality monitoring device according to claim 7, wherein the server outputs a signal to the control unit at a predetermined timing, and outputs a warning when a response signal cannot be received from the control unit within a predetermined time.
  9. 　前記金型は、射出成形機に設けられた、可動側金型および固定型金型の少なくとも一方である、請求項１に記載の異常監視装置。 The abnormality monitoring device according to claim 1, wherein the mold is at least one of a movable mold and a fixed mold provided in an injection molding machine.
  10. 　金型の分割部を表した学習用画像を機械学習することによって生成された学習済みモデルを用意し、
    　カメラによって前記分割部を撮像して撮像画像を取得し、
    　制御部により、前記金型の型締めによって成形された成形物の取出処理の終了を示す制御情報の入力に応じ、前記撮像画像と前記学習済みモデルに基づいて、前記分割部が異常状態であるか否かを判定し、前記分割部が異常状態であると判定したとき、前記型締めを停止させるための停止指示を出力する、
    　異常監視方法。     Prepare a trained model generated by machine learning a training image showing the split part of the mold.
    The divided portion is imaged by a camera to acquire an captured image, and the captured image is acquired.
    The division unit is in an abnormal state based on the captured image and the learned model in response to input of control information indicating the end of the extraction processing of the molded product formed by the mold clamping of the mold by the control unit. When it is determined whether or not the part is in an abnormal state, a stop instruction for stopping the mold clamping is output.
    Abnormality monitoring method.
  11. 　金型の型締めによって成形された成形物の取出処理の終了を示す制御情報の入力に応じ、前記金型の分割部を表した学習用画像を機械学習することによって生成された学習済みモデルと、カメラによって前記分割部を撮像して取得された撮像画像とに基づいて、前記分割部が異常状態であるか否かを判定するプログラムと、
    　前記分割部が異常状態であると判定したとき、前記型締めを停止させるための停止指示を出力するプログラムからなる異常監視プログラム。     A trained model generated by machine learning a learning image showing the divided portion of the mold in response to input of control information indicating the end of the extraction process of the molded product formed by mold clamping. , A program for determining whether or not the divided portion is in an abnormal state based on the captured image obtained by imaging the divided portion with a camera.
    An abnormality monitoring program including a program that outputs a stop instruction for stopping the mold clamping when it is determined that the divided portion is in an abnormal state.
PCT/JP2019/046921 2019-11-29 2019-11-29 Abnormality monitoring device, abnormality monitoring method, and abnormality monitoring program WO2021106225A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/JP2019/046921 WO2021106225A1 (en) 2019-11-29 2019-11-29 Abnormality monitoring device, abnormality monitoring method, and abnormality monitoring program
JP2020515276A JP6722836B1 (en) 2019-11-29 2019-11-29 Abnormality monitoring device, abnormality monitoring method, and abnormality monitoring program

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2019/046921 WO2021106225A1 (en) 2019-11-29 2019-11-29 Abnormality monitoring device, abnormality monitoring method, and abnormality monitoring program

Publications (1)

Publication Number Publication Date
WO2021106225A1 true WO2021106225A1 (en) 2021-06-03

Family

ID=71523907

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2019/046921 WO2021106225A1 (en) 2019-11-29 2019-11-29 Abnormality monitoring device, abnormality monitoring method, and abnormality monitoring program

Country Status (2)

Country Link
JP (1) JP6722836B1 (en)
WO (1) WO2021106225A1 (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPWO2022162806A1 (en) * 2021-01-27 2022-08-04
JP2022167220A (en) * 2021-04-22 2022-11-04 ウシオライティング株式会社 Mold monitoring device and mold monitoring method

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005125561A (en) * 2003-10-22 2005-05-19 Sumitomo Heavy Ind Ltd Mold monitoring device, mold monitoring method and program
JP2019168973A (en) * 2018-03-23 2019-10-03 ファナック株式会社 Drive device and machine learning device

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005125561A (en) * 2003-10-22 2005-05-19 Sumitomo Heavy Ind Ltd Mold monitoring device, mold monitoring method and program
JP2019168973A (en) * 2018-03-23 2019-10-03 ファナック株式会社 Drive device and machine learning device

Also Published As

Publication number Publication date
JP6722836B1 (en) 2020-07-15
JPWO2021106225A1 (en) 2021-12-02

Similar Documents

Publication Publication Date Title
JP6557272B2 (en) State determination device
WO2021106225A1 (en) Abnormality monitoring device, abnormality monitoring method, and abnormality monitoring program
US9682506B2 (en) Control method and control device for injection molding machine
JP5421980B2 (en) Control device for injection molding machine
CN105228808B (en) For processing the machine operated method of plastics
CN109693352A (en) Mold monitoring device and mold monitor method
JP2007160642A (en) Molding machine control system, molding machine, control apparatus, and molding machine control method
CN108688107A (en) Injection (mo(u)lding) machine and injection moulding information processing unit
US20180311878A1 (en) Plastic injection-moulding machine having a data memory for storing text information and/or image information
CA3025398A1 (en) Method for monitoring and controlling an injection molding process using a strain gauge
CN107303718A (en) Metal mold supervising device
JP2008006785A (en) Ejector control device of injection molding machine
JP3562582B2 (en) Control method and control device for injection molding machine
US9701054B2 (en) Injection molding system
JPWO2014076752A1 (en) Injection molding machine
JP5661820B2 (en) Control device for injection molding machine having mold clamping force control function
JP2545465B2 (en) Method for automatically setting upper and lower limits of molding conditions of molding machine
JP6660443B2 (en) Injection molding apparatus and injection molding method
WO2022162806A1 (en) Abnormality monitoring device, abnormality monitoring method, and abnormality monitoring program
CN111940700B (en) Die casting machine control method and device, electronic equipment and storage medium
JP2019126907A (en) Control device in injection molding machine
JPH06210691A (en) Molding machine with molding condition determination support function
JPH06210692A (en) Molding machine with product schedule monitoring function
JP2005125709A (en) Mold monitoring apparatus, mold monitoring method and program
US20200147849A1 (en) Molding system, molding apparatus, inspection apparatus, inspection method, and program

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2020515276

Country of ref document: JP

Kind code of ref document: A

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19954478

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19954478

Country of ref document: EP

Kind code of ref document: A1